article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
in refs . @xcite we have discussed the cosmological properties of light neutralinos ( _ i.e. _ neutralinos with a mass in the range 6 gev @xmath0 50 gev ) , which originate in supersymmetric schemes where gaugino - mass unification is not assumed . actually , the most remarkable features occur for neutralinos in the mass range : 7 gev @xmath0 25 gev .
namely , for relic neutralinos with these masses , direct and indirect detection rates are considerably high , and at the level of present experimental sensitivities .
furthermore , the range of the predicted values for the rates is quite narrow , at variance with what happens for neutralinos of higher masses , where the expected rates are spread over decades .
the properties of light neutralinos with respect to wimp direct measurements were analyzed in refs .
since then , some new results and/or analyses of previous data from experiments of wimp direct searches have appeared @xcite . in the present paper
, we examine whether these new data put some constraints on the relic neutralinos of light masses .
let us recall that the differential event rate @xmath1 ( @xmath2 being the nuclear recoil energy ) measured in wimp direct searches is a convolution of the wimp - nucleus cross section with the wimp phase - space distribution function of wimps , evaluated at the earth location . by assuming that : i ) in this phase - space distribution function , spatial and velocity dependence factorize , and 2 ) coherent interactions dominate over incoherent ones in the wimp - nucleus scattering ( which is usually the case for relic neutralinos ) , one recovers the expression : @xmath3 where : @xmath4 in the previous formulae , notations are : @xmath5 is the number of target nuclei per unit mass
, @xmath6 is the wimp mass , @xmath7 is the nucleus mass , @xmath8 is the wimp_nucleon _ reduced mass , @xmath9 the nuclear mass number , @xmath10 is the wimp nucleon coherent cross section , @xmath11 is the nuclear form factor , @xmath12 is the fraction of the mass density of the wimp in terms of the total local density for non - baryonic dark matter @xmath13 ( _ i.e. _ : @xmath14 ) , @xmath15 and @xmath16 denote the velocity distribution function ( df ) and wimp velocity in the earth frame , respectively ( @xmath17 ) .
it is natural to define the velocity distribution function in the galactic rest frame @xmath18 , where @xmath19 , @xmath20 being the earth velocity in the galactic rest frame .
the earth frame velocity df is then obtained by means of the transformation : @xmath21 .
it is implicitly understood that the velocity df @xmath22 is truncated at a maximal escape velocity @xmath23 , since the gravitational field of the galaxy can not bound arbitrarily fast wimps .
the value we adopt here is : @xmath24 km sec@xmath25 @xcite , although we will comment on the effect of a lower value , which we will set at @xmath26
km sec@xmath25 @xcite . finally , the quantity @xmath27 appearing in eq .
( [ eq : ivmin ] ) defines the minimal earth
frame wimp velocity which contributes to a given recoil energy @xmath2 : @xmath28^{1/2}\ , , \label{eq : vmin}\ ] ] where @xmath29 is the wimp_nucleus _ reduced mass .
( [ eq : rate ] ) and ( [ eq : ivmin ] ) are the basis for deriving information on the quantity @xmath30 from the measurements on the differential rate @xmath1 .
however , this procedure implies the use of a specific wimp distribution function , which determines both the value of the local dark matter density @xmath13 and the shape of the velocity df @xmath22 . in the present paper
we first discuss how upper bounds on @xmath31 , derived from experimental upper limits on @xmath1 , depend on the large uncertainties affecting the wimp distribution functions .
we then discuss what is the relevance of these upper bounds on @xmath30 for light relic neutralinos .
a0 & isothermal sphere + a1 & evans logarithmic @xcite + a2 & evans power law @xcite
+ a5 & nfw @xcite + & + + b1 & evans logarithmic @xcite + b2 & evans power law @xcite
+ b5 & nfw @xcite + & + + c2 & evans logarithmic + c3 & evans power law + & + + d1 & earth on major axis , radial anisotropy + [ cols="^,^,^,^",options="header " , ] function @xmath32 for an isothermal sphere ( model a0 of table [ tab : models ] ) and @xmath33 km sec@xmath25 .
the solid curves ( from top to bottom , as seen on the extreme left of the plot ) refer to the three values of @xmath34 ( and ensuing @xmath13 ) given in table [ tab : intervals ] : @xmath35 km sec@xmath25 ( top ) , @xmath36 km sec@xmath25 ( medium ) , @xmath37 km sec@xmath25 ( bottom ) .
the dashed line shows the modification of the median isothermal case when @xmath38 km sec@xmath25 . ]
values of @xmath39 as a function of the nuclear recoil energy @xmath2 , for a ge detector .
the different curves refer to wimp masses of : 10 , 20 , 50 , 100 , 200 gev and 1 tev , from top to bottom . ] in our analysis we consider a subset of the large sample of galactic halo models which were studied in detail in ref .
following ref .
@xcite , we classify the dfs into four categories , depending on the symmetry properties of the matter density ( or the corresponding gravitational potential ) and of the velocity dependence : a ) spherically symmetric matter density @xmath40 with isotropic velocity dispersion , b ) spherically symmetric matter density with non isotropic velocity dispersion , c ) axisymmetric models , d ) triaxial models @xcite for each category , different specific models are identified .
the models considered in the present analysis are listed in table [ tab : models ] . for a thorough definition of the different models and of the values of their intrinsic parameters , and for a detailed description of theoretical technicalities ,
we refer to ref .
here we just remind that for each model we calculate , either analytically ( when possible ) or numerically , the velocity df which accompanies a given matter density distribution . for the spherically symmetric and isotropic models of class a
, the velocity df is obtained by solving the eddington equation @xcite . for the spherically symmetric and non isotropic models of class b we assume the anisotropy to be described in terms of the osipkov
merrit parameter @xmath41 @xcite which defines the degree of anisotropy ( we fix @xmath42 ) : in this case , the velocity df can be obtained by a generalization of the eddington method @xcite .
axisymmetric models of class c allow the presence of a definite angular momentum . we choose them as a direct generalization of some of the models of class a :
for these models , analytical solutions to the relevant generalized eddington equation may be found . in this case , we also allow for a maximal co rotation or counter rotation of the galactic halo .
finally , class d presents a specific triaxial model .
each halo model is constrained by a number of observational inputs @xcite : i ) properties of the galactic rotational curve , namely the range of the allowed values for the local rotational velocity , @xmath43 @xcite , and the amount of flatness of the rotational curve at large distances from the galactic center , and ii ) the maximal amount of non halo components in the galaxy , @xmath44 ( _ i.e. _ the disk , the bulge , etc . ) .
these constraints determine first of all the value of the local dark matter density @xmath13 , which is a relevant parameter in the direct detection rate . depending on whether one allows for a maximal halo ( _ i.e. _ the contribution of the non - halo components is minimized ) or , the other way around ,
the contribution of the halo to the rotational curve is minimized , the value of @xmath13 is either increased or reduced , respectively .
this is discussed in detail in ref .
@xcite and is manifest also in table [ tab : intervals ] , where values of @xmath13 for each halo model and for the three representative values of @xmath34 are shown .
the difference in the values of @xmath13 depending on the assumption of a minimal or a maximal halo contribution is given in ref .
for instance , for the isothermal sphere @xmath13 falls in the range @xmath45 for @xmath46 , @xmath47 for @xmath48 , and @xmath49 for @xmath50 .
we therefore expect a significant variability in the extracted upper limits in direct detection experiments also for the simple and generally used isothermal sphere model .
we notice that the standard reference choice of @xmath51 for @xmath52 refers to the case of a minimal halo . the choice we make here for the values of @xmath13 associated to each representative value of @xmath34 is the following : for @xmath53 ( which correspond to its 95% c.l .
lower bound ) we adopt the case of a minimal halo , in order to determine the set of less constraining upper limits on @xmath54 , as is clear from eq .
( [ eq : rate ] ) . for @xmath55 ( which correspond to its 95% c.l .
upper bound ) we instead adopt a maximal halo : in this case we will determine the most constraining upper
limits on @xmath54 . in the case of the central ( and reference )
value @xmath56 , we adopt a minimal halo , which reproduces the standard choice @xmath57 for the isothermal sphere .
these considerations apply to all models of class a and b. for the models of class c and d , for which we can rely only on analytical solutions for the velocity df , we are forced to use always the case of a maximal halo : in fact , analytical solutions of class c and d are actually obtained only for a maximal halo contribution @xcite .
let us turn now to the direct effect of the velocity df on the detection rate , which is studied here in terms of the relevant function @xmath32 of eq .
( [ eq : ivmin ] ) .
[ fig : ivmin_isothermal ] shows @xmath32 for the isothermal halo and for the three values of @xmath34 listed in table [ tab : intervals ] .
we see that for low values of @xmath39 the larger contribution to the detection rate occurs when @xmath34 is smaller , since for smaller rotational velocities the velocity dispersion of the isothermal maxwellian distribution is also smaller , and in turn this enhances the average inverse velocity , which is related to the definition of @xmath32 : @xmath58 .
this can be analytically understood by remembering that for a pure isothermal sphere and a maximal halo the velocity distribution function is just an isotropic maxwellian with velocity dispersion given by @xmath34 : @xmath59^{-3/2 } \exp(-v^2/v_0 ^ 2)\ , ; \label{eq : fa}\ ] ] in this case the function @xmath32 reads ( in the limit @xmath60 ) @xcite : @xmath61\ , , \label{eq : ivmina}\ ] ] where @xmath62 and @xmath63 . from eq.([eq : ivmina ] ) we see that for small values of @xmath39 the larger @xmath32 occurs for smaller @xmath34 because of the inverse law dependence . on the contrary , for large values of @xmath39 , the almost exponential tail in @xmath32
is more severe when @xmath34 is small , and therefore the behaviour of @xmath32 with respect to @xmath34 is the opposite .
this again is understood from the simple expression of eq.([eq : ivmina ] ) .
the regime we are considering ( @xmath64 ) asymptotically can be studied as the limit @xmath65 in eq .
( [ eq : ivmina ] ) , which gives : @xmath66^{-1/2 } \exp(-{v_{\rm min}}^2/v_0 ^ 2)\ , .
\label{eq : ivminalim}\ ] ] this shows the discussed behaviour as a function of @xmath34 . the tail , due to the presence of a non vanishing @xmath67 in eq .
( [ eq : ivmina ] ) , is less severe than the one in eq .
( [ eq : ivminalim ] ) but nevertheless it follows the same behaviour . also the value of the escape velocity is relevant in the large @xmath39 tail of the function @xmath32 .
the results presented so far in fig . [ fig : ivmin_isothermal ] are obtained for a value of the escape velocity ( in the galactic frame ) @xmath68 .
however , values as low as @xmath69 have also been considered @xcite .
a lower escape velocity implies a cut in the high @xmath39 tail of @xmath32 @xcite .
this effect is shown in fig .
[ fig : ivmin_isothermal ] for the central case @xmath70 .
the discussion on the behaviour of @xmath32 has direct impact on the direct detection rate , since @xmath39 is directly related to the recoil energy .
( [ eq : vmin ] ) implies that very light wimps can produce recoil energies in the tens of kev range only if they possess large velocities . in this case
, the detection rate for such light wimps will be mostly determined by the almost exponential tail in the function @xmath32 discussed above .
on the contrary , heavy wimps can produce recoil in the same tens of kev range by possessing much lower velocities : they will be therefore more sensitive also to the low @xmath39 part of the function @xmath32 . the quantitative connection between @xmath27 and @xmath2 for a ge nucleus and different wimp masses is given in fig .
[ fig : ge_vmin ] . finally , fig .
[ fig : ivmin_distribs ] shows the function @xmath32 for all the other halo models listed in table [ tab : models ] . as for the symmetric and isotropic models
, we see that in the case of a power
law behaviour of the gravitational potential ( model a2 ) or for the nfw density profile ( model a5 ) the large @xmath39 tail is less suppressed , mainly for low @xmath34 : this comes along with a larger predicted detection rate and it will translate into a more constraining upper limit for wimps lighter than a few tens of gev . in the case of anisotropic models , we notice that the most direct anisotropic generalization of the isothermal sphere , which is a cored spherical distribution with anisotropic velocity dispersion ( model b1 ) is the one which is more suppressed at large @xmath39 : this has the effect of reducing the sensitivity of the detector to light wimps . on the contrary , the axisymmetric model with a power
law gravitational potential c3 is the one with the highest tail in the function @xmath32 , and therefore more sensitive in constraining light wimps .
we also notice that for this type of models , which possess an enhaced @xmath39 tail , the effect of a lower escape velocity is more dramatic : panel ( h ) of fig .
[ fig : ivmin_distribs ] shows the sizeable reduction of @xmath32 at large @xmath39 for an escape velocity @xmath38 km sec@xmath25 in the case of model c3 and @xmath37 km sec@xmath25 .
the local matter density values @xmath13 of table [ tab : intervals ] and the results of fig .
[ fig : ivmin_isothermal ] and fig .
[ fig : ivmin_distribs ] are the key elements which will be used in the next section to determine upper limits on the wimp
nucleon scattering cross sections .
the solid lines show the upper limit on the quantity @xmath30 as a function of the wimp mass @xmath6 for the cdms detector and for an isothermal sphere ( model a0 of table [ tab : models ] ) .
the curves refer to the three values of @xmath34 ( and corresponding @xmath13 ) given in table [ tab : intervals ] : @xmath35 km sec@xmath25 ( top ) , @xmath36 km sec@xmath25 ( medium ) , @xmath37 km sec@xmath25 ( bottom ) .
the colored region shows the values of @xmath30 for neutralino dark matter , obtained in a scan of the minimal supersymmetric model defined in refs .
the funnel for low neutralino masses ( below 50 gev ) corresponds to supersymmetric models without gaugino mass unification . ]
ratio of the upper limits of fig .
[ fig : limits_isothermal ] , obtained for an isothermal sphere ( model a0 ) .
the upper curve is the ratio between the upper and the central curves in fig .
[ fig : limits_isothermal ] .
the lower curve is the ratio between the lower and central curves ] the solid lines show the summary of our analysis on the upper limit on the quantity @xmath31 as a function of the wimp mass @xmath6 for the cdms detector and for @xmath33 km sec@xmath25 .
the median line refers to the standard isothermal sphere with @xmath71 km sec@xmath25 and @xmath72 ( model a0 ) .
the upper and lower curves show the two extremes obtained in the analysis and refer to model b1 with @xmath73 km sec@xmath25 ( upper solid line ) and model c3 with @xmath74 km sec@xmath25 ( lower solid line ) .
the dashed line refers to model c3 with maximal counter
rotation of the galactic halo .
the dotted lines show the zeplin i limits obtained for the same galactic models .
the long dashed lines show the upper limits for cdms in the case of a lower escape velocity @xmath38 km sec@xmath25 : the upper line refers to model a0 , the lower one to model c3 . for model b1 ,
the limit coincides with the corresponding solid line .
the colored region shows the values of @xmath30 for neutralino dark matter , obtained in a scan of the minimal supersymmetric model defined in refs .
the funnel for low neutralino masses ( below 50 gev ) corresponds to supersymmetric models without gaugino mass unification . ] in refs .
@xcite upper limits on @xmath31 are obtained by using a standard isothermal distribution function with density parameter @xmath75 gev @xmath76 and @xmath36 km sec@xmath25 . here , we analyze this class of experimental data employing the sample of distribution functions discussed in the previous section .
our analysis and discussion is performed in terms of the experimental data of ref .
@xcite ( cdms ) , since these turn out to be the most constraining ones .
for very light wimps , with masses below 10 gev , also the results of ref .
@xcite ( zeplin ) may play a role , and we will add them in our final discussion . in order to obtain the exclusion
plot from the cdms data , we extract the neutralino
nucleon cross section that yields a number of events compatible with zero between 10 kev and 64 kev with an effective exposure @xmath77 kg day , which corresponds to the effective exposure for @xmath78 gev quoted in ref .
@xcite . by considering a poissonian fluctuation of the expected rate , we assume the upper bound of 2.3 events at 90% c.l . in the calculation of the expected rate
, we use helm nuclear form factors and a bolometric quenching factor equal to 1 , as quoted by the experimental collaboration @xcite . our procedure for extracting the exclusion
plot is less refined than the one adopted in ref .
@xcite , since it neglects the dependence of @xmath79 on the wimp mass and does not take into account the part of the cdms spectrum between 64 and 100 kev ( which can be included following for instance the statistical procedure of ref .
@xcite ) . nevertheless our procedure allows us to reproduce to a good degree of precision the cdms limit for the standard isothermal distribution , as shown by the central curve in fig .
[ fig : limits_isothermal ] , when this is compared with the upper bound displayed in fig .
39 of ref .
therefore , in our analysis we adopt our simpler procedure ; we have checked that adding a proper treatment of the efficiency and adopting the statistical procedure of ref .
@xcite yields quite similar results .
the upper limits for the isothermal sphere ( model a0 ) are shown in fig .
[ fig : limits_isothermal ] , for the three representative values of @xmath34 and the corresponding choices for the local dark matter density @xmath13 , as quoted in table [ tab : intervals ] .
as already mentioned , the central curve corresponds to the reference case of @xmath70 with @xmath80 .
the upper and lower curves are instead obtained for @xmath81 with @xmath82 and @xmath55 with @xmath83 , respectively .
an important effect is obviously due to the different values of @xmath13 which are associated to the different values of @xmath34 , as discussed in the previous section and in ref .
however , the difference in the function @xmath32 is quite relevant in the determination of the upper limits , especially at low wimp masses . in order to appreciate the difference in the exclusion plots , we show in fig .
[ fig : ratio ] the ratios of the upper limits obtained with @xmath81 ( lower curve ) and @xmath55 ( upper curve ) with respect to the central @xmath70 case .
the dashed horizontal lines show the ratios of the corresponding values of @xmath13 .
we can notice that at low wimp masses the difference in the exclusion plots is very large , much larger than the naive ratio of the corresponding @xmath13 s .
this is a consequence of the sizable difference in @xmath32 for large @xmath39 , which is the regime relevant for light wimps , as discussed before .
the steep behaviour of the ratios in fig .
[ fig : ratio ] is a consequence of the fact that the sensitivity of direct detection to very low wimp masses ( below about 10 gev ) rapidly vanishes .
on the contrary , at large wimp masses , the difference in the exclusion plots is much close to what one would expect on the basis of the difference in the @xmath13 values .
this is clear from our previous analysis , since for large wimp masses the relevant range of @xmath39 is in the 100300 @xmath84 range , which is where the difference in @xmath32 is small .
we can also notice that , for large wimp masses , it is even possible to revert the value of the ratio of the exclusion plots naively obtained by the ratio of the different @xmath13 s : this is a consequence of the behaviour of @xmath32 at small @xmath39 discussed in the previous section .
[ fig : limits_isothermal ] represents the maximal variability which occurs for the isothermal sphere : this quantifies the astrophysical uncertainty connected to this halo model . confronting the upper limits with the results obtained for light neutralinos in supersymmetric models without gaugino mass universality @xcite
, we can see that while all the configuration in the mass range @xmath85 are excluded for the central ( upper ) values of @xmath34 , only a small fraction are eliminated when @xmath34 assumes its lower bound value .
therefore , the conservative attitude which has to be taken when setting limits makes us to conclude that for the isothermal sphere , direct detection only mildly constrains the light neutralino sector of supersymmetric models without gaugino mass universality @xcite , in the 2040 gev mass range .
clearly the variation on the upper limits due to the difference in the halo properties has consequences also on the exploration of the supersymemtric parameter space for heavier neutralinos , as is shown in fig .
[ fig : limits_isothermal ] .
this is relevant also to gaugino mass universal models , for which the lower bound on the neutralino mass exceeds 50 gev @xcite .
the results for the other galactic halo models is shown in fig .
[ fig : limits_distribs ] .
the differences in the upper limits can be understood on the basis of the discussion on the isothermal sphere and on the properties of @xmath32 for the different models presented in the previous section .
we notice that some models , like c3 , d1 and b5 are more constraining , while in the case of models like a1 and b1 the limits imposed by direct detection are relatively less severe . finally , we report in fig . [
fig : limits_extreme ] the summary of our analysis : together with the standard central isothermal sphere , we show the more ( c3 ) and less ( b1 ) constraining models we obtain . from the analysis of this figure
, we conservatively conclude that from direct detection experiments there is currently no constraint on the light neutralino sector of supersymmetric models without gaugino universality @xcite
. should the local value of the rotational velocity be on its high range ( close to @xmath55 ) direct detection could be able to set stringent limits on these supersymmetric configuration : all the mass range above 78 gev ( depending on the actual halo model ) and below 25 gev would be excluded .
notice that , would this be the case , also the local density @xmath13 would be large ( above @xmath86 , as discussed in ref .
@xcite and shown in table [ tab : intervals ] ) : in this case the neutralino configurations below 78 gev , which are not constrained by direct detection , would be completely excluded by antiproton searches @xcite . however , due to astrophysical uncertainties which affect the different detection rates , currently it is not yet possible to set absolute limits , neither from indirect detection techniques @xcite or , as shown in the present analysis , by direct detection .
[ fig : limits_extreme ] also shows the effect of a lower escape velocity .
as discussed in the previous section , this implies a cut in the high @xmath39 tail of @xmath32 : this turns into a weaker sensitivity of direct detection to low mass neutralinos .
the effect is especially manifest for the most stringent models , like model c3 with @xmath87 km sec@xmath25 . for @xmath88 ,
all neutralino models below 9 gev are not constrained even for c3 model .
for the a0 model with @xmath71 km sec@xmath25 , there is also a sizeable difference for light wimps , although this is not relevant for the neutralino configurations .
finally , in the case of model b1 with @xmath73 km sec@xmath25 , the lower escape velocity does not produce differences , since in the high velocity tail @xmath32 was already depressed even for @xmath89 , as can be seen in fig .
[ fig : ivmin_distribs ] . for completeness
, fig.[fig : limits_extreme ] also shows the upper limits we obtain , for the isothermal sphere and for the two extreme cases , for the zeplin detector @xcite .
the limits we obtain for the isothermal sphere are slightly higher at low wimp masses than the ones quoted by the experimental collaboration @xcite .
we trace this effect to some differences in the analysis of the data ( we do not make use of the `` light response matrix '' discussed in ref .
@xcite , since we do not have it at our disposal ) .
[ fig : limits_extreme ] shows that for very low wimp masses zeplin could be slightly more sensitive than cdms .
nevertheless , even lowering by a factor of 2 the upper limits we obtain for zeplin , our conclusions on the limits imposed to light neutralinos remain unchanged .
finally , we wish to remind that an annual modulation effect in direct detection has been observed by the dama collaboration over seven years @xcite .
this result , when interpreted in terms of scalar wimp nucleus interactions , leads to an allowed region in the plane @xmath54 vs. @xmath6 , which extends also to light wimp masses .
the dama collaboration analysis of ref .
@xcite takes into account the same variability in galactic halo models of ref .
@xcite , which is also used here .
it is not possible to make direct comparison among the dama allowed region and the upper limits we obtain here for cdms and zeplin , since the dama region is the convolution obtained after varying all the galactic halo models , while the results presented here refer to single halo model .
a proper comparison between different experimental results can be made _ only _ at a fixed galactic halo model .
notice that a convolution of our results would be just the upper curve ( model b1 ) of fig.[fig : limits_extreme ] . as for the comparison between the light neutralinos of non universal gaugino models and the dama allowed region , we comment by reminding that these light neutralinos are totally compatible with the allowed dama region , as we showed in ref .
@xcite : they could in fact explain the annual modulation effect .
we gratefully acknowledge financial support provided by research grants of the italian ministero dellistruzione , delluniversit e della ricerca ( miur ) , of the universit di torino and of the istituto nazionale di fisica nucleare ( infn ) within the _ astroparticle physics project_. the class of models discussed in ref .
@xcite represents a quite exhaustive summary of dark matter distributions functions as possible departures from the isothermal sphere model .
notice , however , that it does not include the possibility of non thermal components in the @xmath22 , which could in principle modify the high velocity tail of the velocity df @xcite . for other analyses , see for instance : a.m. green
d * 68 * , 023004 ( 2003 ) ; erratum - ibid d * 69 * , 109902 ( 2004 ) ; a.m. green , phys.rev.d * 66 * , 083003 ( 2002 ) ; a.m. green , phys.rev.d * 63 * , 043005 ( 2001 ) ; m. kamionkowski , a. kinkhabwala , phys .
d * 57 * , 3256 ( 1998 ) ; j.d .
vergados , phys .
lett.*83 * , 3597 ( 1999 ) ; j.d .
vergados , d. owen , astrophys.j . * 589*,17 ( 2003 ) ; p. ullio , m. kamionkowski , jhep * 0103 * , 049 ( 2001 ) ; j. edsjo , m. schelke , p. ullio , jcap * 0409 * , 004 ( 2004 ) ; g. gelmini , p. gondolo , phys .
rev d * 63 * , 036006 ( 2001 ) .
see for instance : v. berezinsky , a. bottino , j. r. ellis , n. fornengo , g. mignola and s. scopel , astropart .
phys . , 1 ( 1996 ) ; a. bottino , f. donato , n. fornengo , s. scopel , phys . rev .
d * 63 * , 125003 ( 2001);j .
r. ellis , k. a. olive , y. santoso and v. c. spanos , phys .
d * 71 * , 095007 ( 2005 ) ; e. accomando , r. arnowitt , b. dutta and y. santoso , nucl .
b * 585 * , 124 ( 2000 ) ; u. chattopadhyay , a. corsetti and p. nath , phys .
d * 68 * , 035005 ( 2003 ) ; j. l. feng , k. t. matchev and f. wilczek , phys .
b * 482 * , 388 ( 2000 ) ; l. bergstrom and p. gondolo , astropart .
phys . * 5 * , 263 ( 1996 ) ; y. g. kim , t. nihei , l. roszkowski and r. ruiz de austri , jhep * 0212 * , 034 ( 2002 ) ; v. a. bednyakov and h. v. klapdor - kleingrothaus , phys .
rev.d * 62 * , 043524 ( 2000 ) .
|
new upper bounds on direct detection rates have recently been presented by a number of experimental collaborations working on searches for wimps . in this paper
we analyze how the constraints on relic neutralinos which can be derived from these results is affected by the uncertainties in the distribution function of wimps in the halo .
various different categories of velocity distribution functions are considered , and the ensuing implications for supersymmetric configurations derived .
we conservatively conclude that current experimental data do not constrain neutralinos of small mass ( below 50 gev ) .
|
an important goal in modern galaxy formation is an improved understanding of the physical processes that regulate the rate at which stars form in galaxies .
these processes include the cooling and accretion of gas within dark matter halos , the transformation of the accreted gas into molecular clouds and stars , and the effects of `` feedback '' from massive stars and accreting black holes on the gas in and around galaxies .
the lambda - cold dark matter model provides detailed predictions for how dark matter halos assemble over time . within these halos ,
gas will cool and settle into a rotationally - supported disk .
one hotly debated issue is the degree to which gas loses angular momentum to the surrounding dark matter during this process . in many `` semi - analytic '' models , the angular momentum of the gas
is assumed to be conserved .
this produces a size - circular velocity relation for galactic disks that is in good agreement with observations ( mo , mao & white 1998 ) .
more detailed numerical hydrodynamical simulations of disk galaxy formation in a @xmath1cdm universe demonstrate , however , that the amount of angular momentum that is lost during the collapse of the dark matter halo depends sensitively on how feedback processes are included in the models ( navarro & steinmetz 1997 ; governato et al 2004,2007 ) . in many simulations , the disks form early and are too compact .
the resulting star formation timescales are then too short to be consistent with observations ( e.g. oser et al 2010 ) .
another important unsolved problem is to understand why galaxies divide into two distinct `` families '' those with ongoing active star formation , and those where star formation has been quenched ( strateva et al 2001 ; kauffmann et al 2003b ; baldry et al 2004 ) .
it has now become clear that these distinctions existed already at redshifts as high as @xmath2 ( williams et al 2010 ; wuyts et al 2011 ) , when cold gas accretion rates in almost all dark matter halos are predicted to be very high .
many recent theoretical galaxy formation models that aim to reproduce the statistical properties of the massive galaxy population ( e.g. croton et al 2006 ; bower et al 2006 ; cattaneo et al 2006 de lucia & blaizot 2007 ; somerville et a 2008 ; guo et al 2011 ; lu et al 2011 ) invoke star - formation quenching mechanisms that set in at a characteristic dark matter halo _ mass_.
the characteristic mass is associated with the transition between the regime where gas cooling times are short compared to the dynamical time of the dark matter halo and gas accretes in the form cold , condensed clouds , and the regime where cooling times are long and gas accretes from a corona of gas that is in virial equilibrium with the surrounding halo .
one might ask _ why _ lowered rates of star formation in galaxies should be linked with this transition .
one argument that is often used is that relativistic jets , generated when gas accretes onto black holes , heat the surrounding hot gas and prevent it from forming stars ( see , for example , croton et al .
the conditions under which black holes produce jets are not well understood .
there is clear observational evidence that radio - emitting jets play a role in regulating the cooling of gas in nearby galaxy clusters ( see mcnamara & nulsen 2007 for a recent review ) .
theoreticians have thus made an `` ansatz '' that jets from radio - loud agn will prevent gas from cooling and forming stars in _ all _ dark matter halos that are predicted to contain a hot gas atmosphere .
the characteristic halo mass threshold that separates star - forming and quiescent galaxies is the main factor that determines how massive galaxies evolve in current models , so it is clearly important to test the existence of such a threshold using real observations .
we note that accretion and quenching processes affect the _ gas components _ of galaxies .
the most direct empirical constraints on how they operate thus come from observations of the gas in galaxies .
this has been the primary motivation for the galex arecibo sdss survey ( gass ) , as well as the co legacy database for gass ( cold gass ) , which are measuring the atomic and molecular gas contents of an unbiased sample of several hundred galaxies with redshifts between 0.025 and 0.05 , and with stellar masses in the range @xmath3 . in both surveys , the strategy is to observe each galaxy until the hi / co lines are detected , or until upper limits in the ratio of atomic and molecular gas mass to stellar mass of @xmath4 reached .
the aim of the program is to carry out a census of the condensed baryons in galaxies in the local universe , to study scaling relations between the gas and stellar properties of these galaxies , and to understand gas accretion and quenching in galaxies .
details of the gass and cold gass survey designs , as well as target selection and observing procedures are given in catinella et al ( 2010 ) and saintonge et al ( 2011a ) .
these two papers also presented relations between the hi and h@xmath5 mass fractions of galaxies ( defined as @xmath6/m@xmath7 and @xmath8/m@xmath7 ) and global galaxy parameters such as stellar mass m@xmath7 , stellar mass surface density @xmath9 , galaxy bulge - to - disc ratio ( as parameterized by the concentration index @xmath10 of the @xmath11-band light ) , and specific star formation rate sfr / m@xmath7 ( see also schiminovich et al 2010 ) .
the two surveys uncovered sharp thresholds in @xmath9 and @xmath10 below which all galaxies have a measurable cold gas component , but above which the detection rate of the co and hi lines drops suddenly , suggesting that `` quenching processes '' have occurred in these systems . in this paper
, we compare the observed gas fraction relations from the cold gass survey with predictions from the semi - analytic models of fu et al ( 2010 ; hereafter f10 ) , which track the formation of molecular gas in disk galaxies .
the f10 models are based on the l - galaxies semi - analytic code , described in detail in croton et al ( 2006 ) and updated in de lucia & blaizot ( 2007 ) .
the models are currently being configured to operate on the latest update of the l - galaxies code by guo et al .
( 2011 ) ; the comparison in this paper will be restricted to the version of the models in the published f10 paper .
the main new aspect of these models is that galactic discs are represented by a series of concentric rings in order to track the evolution in the gas and stellar surface density profiles of galaxies over cosmic time .
two simple prescriptions for molecular gas formation processes are included : one is based on the analytic calculations by krumholz , mckee & tumlinson ( 2009 ; hereafter kmt ) , and the other is a prescription where the h2 fraction is determined by the pressure of the interstellar medium ( blitz & rosolowsky 2006 , hereafter br ) , with an implementation similar to that in obreschkow et al .
the free parameters of the models that regulate the rate at which gas is turned into stars and the efficiency with which supernovae reheat gas as a function of halo mass have been tuned to reproduce a number of key observables , including the observed present - day galaxy luminosity function , the gas - phase mass - metallicity relation , and the mean hi gas fraction as a function of b - band luminosity ( see de lucia , kauffmann & white ( 2004 ) for a more detailed discussion ) .
there have been a number of other cosmological models that follow the formation of molecular gas in disks .
lagos et al ( 2011 ) have developed semi - analytic models that track molecular gas formation in cosmological simulations of galaxy formation , but these models do not predict the detailed radial profiles of galaxies as in f10 .
robertson & kravtsov ( 2008 ) and dutton et al ( 2010 ) model the radial profiles of the gas and stars in galaxies , but their models are not embedded within full cosmological n - body simulations .
gnedin et al ( 2009 ) have carried out very high resolution cosmological simulations of molecular gas formation that include radiative transfer and detailed treatment of the chemistry , but their box sizes are too small to make statistical predictions .
the f10 paper demonstrated that their models could fit the radial hi , h@xmath5 , stellar mass and sfr profiles of spiral galaxies from the things / heracles surveys . in this paper , we investigate whether the models also yield correct gas fraction scaling relations and distribution functions .
as will be seen , it is useful to break this analysis into two distinct parts : 1)comparison between models and observations of the distribution of gas fractions in the population of galaxies with detectable gas , 2 ) comparison of how _ absence of gas _ , which we refer to as `` quenching '' , is manifested as a function of observable quantities such as stellar mass , stellar surface density , and bulge - to - disk ratio . as we will show , such a comparison elucidates those aspects of the models which work well , and those that fail . throughout this paper ,
we have assumed a cosmology with @xmath12 , @xmath13 and @xmath14 km s@xmath15 mpc@xmath15 to derive our observational quantities .
the cosmology assumed for the f10 model is a @xmath1cdm models with @xmath16 , @xmath17 , @xmath18 km s@xmath15 mpc@xmath15 and @xmath19 .
the semi - analytic models used in this paper are nearly identical to those described in detail in f10 .
we have made only one small change .
the free parameters in the f10 models were chosen so that reasonable fits to a variety of observables could be obtained .
these observables included the h@xmath5 mass function of keres et al ( 2003 ) derived using galaxies drawn from the fcrao extragalactic co survey ( young et al 1995 ) .
these authors adopted a co - to - h@xmath5 conversion factor that was a factor of 1.5 times larger than the one adopted by saintonge et al ( 2011a ) .
( k km s@xmath15 pc@xmath20)@xmath15 , which does not include a correction for the presence of helium . ] to bring the simulations back into agreement with the h@xmath5 data using the conversion factor adopted by saintonge et al , we simply increase the supernova reheating rate ( given by the parameter @xmath21 ) in table 1 of f10 by a similar factor ( @xmath22 instead of 3.5 ) .
we note that changing @xmath21 simply shifts the amplitude of the gas mass fraction scaling relations up or down .
it does not change the predicted slope or scatter in these relations .
we also note that the @xmath23 mass function derived from the cold gass data for galaxies with @xmath24 agrees well with that of keres et al ( 2003 ) at the high mass end , once the same co - to - h@xmath5 conversion factor is adopted ( fu et al 2012 , in preparation ) .
the observational sample is an update to that described in saintonge et al 2011(a ) , consisting of 299 galaxies with co(1 - 0 ) line observations with the iram 30 m telescope .
hi line measurements from the gass survey ( catinella et al 2010 ) are available for 270 out of the 299 galaxies .
because the observational sample is selected only by stellar mass , it is easy to compare simulation results directly with the data .
there are only two observational selection issues that require careful treatment : 1 . in both the gass and cold gass surveys , targets
are selected so that the resulting stellar mass distribution is approximately flat .
this was done in order to be able to compute gas fractions in different stellar mass bins with roughly the same errors .
the sampling must be taken into account when we look at gas fraction distributions as a function of parameters other than stellar mass . in this paper , we adopt two approaches : a ) we either weight each galaxy in the sample by the inverse of @xmath25 , where @xmath25 is the `` sampling rate function '' required to transform the true stellar mass distribution to a flat one , b ) we apply the sampling rate function @xmath25 to the simulated galaxies to create `` mock catalogue '' with flat stellar mass distributions that can be compared directly with the data .
it is important to take account of the detection limits of the survey when comparing model galaxies with the data .
this is illustrated in the top panels of figure 1 , where we plot m(hi)/m@xmath7 ( left ) and m(h@xmath5)/m@xmath7 ( right ) as a function of stellar mass for 299 galaxies that have been observed in co as of july 2011 ( 270 of these have hi observations from the gass survey ) .
galaxies where the hi or co line was detected ( see catinella et al 2010 and saintonge et al 2011a for details about line detection procedures ) are plotted in blue .
galaxies without detections are plotted in red at the positions of their 5@xmath26 upper limits in m(hi)/@xmath27 or m(h@xmath5)/@xmath27 .
+ to a very close approximation , the hi mass fraction limit log [ m(hi)/m@xmath7]@xmath29 is -1.82 ( corresponding to a hi mass fraction limit 0.015 ) for galaxies with @xmath30 . for galaxies with @xmath31 , log [ m(hi)/m@xmath7]@xmath32 .
+ the co line detection limits are somewhat more complicated .
as discussed in saintonge et al ( 2011a ) , the integration times are set so that log [ m(h@xmath5)/m@xmath7]@xmath29=-1.82 for galaxies with @xmath33 . for lower mass galaxies , integration times
are nominally set to reach a fixed r.m.s .
of around 1.1 mk per 20 km / s wide channel , but the actual value fluctuated according to weather conditions at the time of observation .
therefore , when creating our mock catalogues from the simulations , we simply impose a distribution of upper limits similar to that seen in the top right panel of figure 1 .
we take [ m(h@xmath5)/m@xmath7]@xmath34 , the _ maximum _ possible value of the h@xmath5 mass fraction limit , as log [ m(h@xmath5)/m@xmath7]@xmath34 = -1.72 if @xmath33 , and log [ m(h@xmath5)/m@xmath7]@xmath34 = 8.78- log m@xmath7 if @xmath35 . the h@xmath5 mass fraction limit that we adopt
is randomly distributed between the maximum value and a minimum value that is 0.35 dex smaller . ) and specific star formation rate ( log sfr / m@xmath7 ) are plotted as a function of the @xmath36 .
blue points denote galaxies with hi line detections .
red points denote galaxies where hi was not detected . in the right panels , h@xmath5
mass fraction ( log m(h@xmath5)/m@xmath7 ) and specific star formation rate ( log sfr / m@xmath7 ) are plotted as a function of the log m@xmath7 .
blue points denote galaxies with co line detections .
red points denote galaxies where co was not detected .
[ sample],width=294 ] in this way , we are able to classify galaxies in the simulations as hi / co `` detections '' or `` non - detections '' and compare properties such as stellar masses , stellar surface densities , specific star formation rates and concentration indices with those of the real galaxies in our survey .
we note that stellar masses and star formation rates are standard outputs of semi - analytic models . because the f10 models are able to track radial profiles of the gas and the stars in galaxies , they are also able to predict stellar surface mass densities , which are defined as @xmath37 , where @xmath38 is the radius containing half the stellar mass of the galaxy . in semi - analytic models ,
bulges form either as a result of a merger between two galaxies , or as a consequence of instabilities that set in when disks reach a certain critical threshold density . in the models ,
the total mass of the bulge is more reliably predicted than its size . as shown in figure 1 of weinmann et al ( 2009 )
, there is a tight correlation between the concentration index @xmath10 ( defined as the ratio of the radii containing 90% and 50% of the @xmath11-band light ) and galaxy bulge - to - disk radio derived using 2-dimensional decomposition codes ( gadotti 2009 ) . we have therefore elected to transform the bulge - to - disk ratios predicted by the models to concentration indices using a fit to this relation : c=1.8 + 2.14@xmath39 .
in the analysis presented in this paper , we will attempt to answer two questions : 1 . can the simple disk formation models described in the f10 paper explain observed gas fraction scaling relations for galaxies with gas ? 2 .
does the transition between the population of galaxies with gas and the population without gas occur in the same way in the models and in the data ? we have chosen the gas fraction detection threshold of our two surveys as the nominal division between the population of galaxies we will henceforth refer to as `` active '' and the population that we will call quenched " .
note that it is possible that some fraction of galaxies classified as `` active '' are actually losing their gas and evolving into quenched systems .
conversely , some fraction of galaxies classified as `` quenched '' may have simply run out of gas just prior to the next accretion episode . such galaxies
may be thought of as `` transition '' systems .
in the bottom panels of figure 1 , we plot specific star formation rate versus stellar mass for the galaxies in our sample . in the left panel , galaxies with hi detections are colour - coded blue and those where the hi line was not detected are colour - coded red . in the right panel
, we do the same thing according to whether the co line was detected or not .
as can be seen , galaxies with log sfr / m@xmath40 are not usually detected in hi and almost never in co. this means that there is a rather clean distinction between `` active '' and `` quenched '' galaxies that applies to both gas fraction and specific star formation rate for the majority of galaxies in our sample . in section 3.2.2
, we return to the issue of whether we can identify a sub - population of galaxies that is in the _ process _ of being quenched . in figures 2 and 3
we compare the mean hi and h@xmath5 gas fractions as a function of @xmath27 , @xmath9 , @xmath10 and sfr/@xmath27 for data and for models .
the reader is referred to catinella et al ( 2010 ) and saintonge et al ( 2011a ) for details on how hi and co line fluxes are transformed into hi and h@xmath5 masses .
we note that we have adopted a constant galactic conversion factor @xmath41 ( k km s@xmath15 pc@xmath20)@xmath15 , which does not include a correction for the presence of helium .
a constant conversion factor should be a reasonable approximation for the galaxies in our survey , which all have stellar masses greater than @xmath42 and gas - phase metallicities near solar ( saintonge et al 2011b ) . in figures 2 and 3 ,
the blue points show the mean values of m(hi)/m@xmath7 and m(h@xmath5)/m@xmath7 calculated using the galaxies with hi and co line detections .
error bars are calculated using bootstrap resampling .
the red and black curves indicate the mean values calculated from the model galaxies that are classified as detections ( see section 2 ) .
the black curves show results from the f10 model that uses the kmt prescription for the conversion of atomic to molecular gas .
the red curves show results from the f10 model that uses the br pressure - based prescription .
we note that there are many more simulated galaxies than observed galaxies . in generating the curves ,
we have ordered the galaxies by increasing @xmath27,@xmath9 , @xmath10 and sfr/@xmath27 and computed means for bins of 70 galaxies .
figures 2 and 3 show that the models reproduce the observed scalings between mean hi / h@xmath5 mass fraction and galaxy stellar mass , surface mass density and concentration quite well .
the differences between the hi mass fractions predicted by the kmt and br prescriptions are small . given the systematic uncertainties in the calibration of the br prescription , the fact that the kmt prescription is based on simple analytic calculations , the fact that our semi - analytic disc formation models are extremely idealized , and the fact that the model free parameters were not tuned to fit the cold gass scaling relations , we find it remarkable how well these results agree with the data . the relation between h@xmath5 mass fraction and specific star formation is _ shallower _ than predicted by the models , which lie on a relation with slope unity .
the discrepancy in the relation between m(h@xmath5)/m@xmath7 and sfr/@xmath27 is easily understood in light of the results presented in saintonge et al ( 2011b ) .
the f10 models adopt the assumption that @xmath43 with @xmath44 constant ( corresponding to an effective molecular gas depletion timescale of 2 gyr ) in all galaxies .
this assumption was motivated by the results in leroy et al ( 2008 ) .
saintonge et al ( 2011 ) showed , however , that there was more than a factor of 5 variation in the _ global _ molecular gas depletion timescale in different galaxies , and that the least actively star - forming galaxies ( i.e lowest values of sfr/@xmath27 ) had the longest depletion times .
this explains the shallow observed relation in the bottom right panel of figure 3 . in figure 4
, we show scatter plots of log sfr / m@xmath7 versus log m(h@xmath5)/m@xmath7 and log m(hi)/m@xmath7 for our survey galaxies ( blue points ) and for galaxies from our mock catalogue ( black points ) .
results are only shown for the kmt atomic - to - molecular gas prescription , because results for the pressure prescription are virtually the same . as can be seen
the models reproduce the scatter in specific star formation rate versus atomic gas fraction quite well , but not in specific star formation rate versus molecular gas fraction .
clearly , there are additional processes at work that determine the rate at which molecular gas will form stars .
this is the subject of a future paper ( saintonge et al , in preparation ) . ) for cold gass galaxies with hi line detections ( blue points ) is plotted as a function of stellar mass , stellar mass surface density , concentration and specific star formation rate .
error bars are calculated using bootstrap resampling .
results for the f10 models are shown as black ( kmt atomic - to - molecular gas prescription ) and red ( br pressure prescription ) curves .
[ detecth1],width=294 ] mass fraction ( log m(h@xmath5)/m@xmath7 ) for cold gass galaxies with co line detections ( blue points ) is plotted as a function of stellar mass , stellar mass surface density , concentration and specific star formation rate .
error bars are calculated using bootstrap resampling .
results for the f10 models are shown as black ( kmt atomic - to - molecular gas prescription ) and red ( br pressure prescription ) curves .
[ detecth2],width=294 ] as well as looking at mean hi and h@xmath5 mass fraction relations for galaxies with detections , it is instructive to analyze _ full distribution functions _ of gas mass fractions .
these are illustrated in figures 5 and 6 .
the data are shown as blue histograms , with error bars computed from bootstrap resampling .
black and red curves are for the models , as before .
we plot distributions of log m(hi)/m@xmath7 and log m(h@xmath5)/m@xmath7 in three different intervals of stellar mass , stellar mass surface density and concentration .
these plots show that the models provide a good representation not only of the mean hi / h@xmath5 mass fractions , but also of the _
scatter around the mean_. as discussed in f10 and will be illustrated below , there is a tight relation between stellar mass and dark matter halo mass in the models .
the scatter around the mean gas fraction at fixed dark matter halo mass is set by a ) the spin parameter of the halo , which determines the contraction factor of the accreted gas and hence the rate at which it will be consumed into stars , and b ) the recent gas accretion history of the galaxy .
the fact that the scatter in the models and data agree so well lends support to this basic picture .
we note that in recent years , there have been a number of papers questioning the treatment of disk formation in semi - analytic models ( e.g. sales et al 2009 ) .
the primary objection is the assumption that the angular momentum of a galaxy expressed in units of that of its surrounding halo ( @xmath45 ) , correlates with mass of the galaxy divided by the mass of its dark matter halo ( @xmath46 ) in a manner that is insensitive to feedback , i.e. @xmath47 is assumed for all halos . in our scheme
this is _ not _ the case .
we only assume that @xmath47 for the gas that is cooling instantaneously at any given time in any given halo . because we track the growth of galaxies and dark matter halos with time , and because feedback effects will affect the mass of gas that cools in halos of different masses at different redshifts , the resulting relation between the angular momentum of the galaxy and its mass fraction will depend on feedback .
figure 5 also shows that the models miss a minority population of galaxies with high atomic gas fractions and with low stellar masses , surface mass densities and concentrations .
what are these missing galaxies ? in three recent papers , we have studied the properties of galaxies in our survey that contain significantly more atomic gas than would be predicted from their uv / optical colours and sizes ( moran et al 2010 ; wang et al 2011 ; moran et al 2011 ) .
the main conclusion is that such galaxies have very blue outer disks with young stellar populations and low gas - phase metallicities .
these results led us to propose that these unusually hi - rich systems have experienced a recent gas accretion event , resulting in growth of the outer disk . in the f10 models ,
the accreted gas is always added to the disk with an exponential surface density profile . in real galaxies ,
the accreted gas may initially be distributed in the outer regions of the galaxy .
dynamical perturbations in the form of spiral density waves , interaction with companions or with the non - axisymmetric gravitational potential of the surrounding dark matter halo , will eventually the cause the gas to flow inwards and reach high enough densities to form molecular clouds .
if we change the gas accretion prescriptions so that accreted gas is initially dumped in the outskirts of the disk , this may produce a tail of hi - rich galaxies .
observations of sfr , metallicity and gas _ profiles _ will be required to constrain models of this nature .
mass fraction ( log m(h@xmath5)/m@xmath7,left ) and hi mass fraction ( log m(hi)/m@xmath7 , right ) for cold gass galaxies are shown as blue points .
results from the f10 models with kmt atomic - to - molecular gas prescription are shown as black points .
[ sfgas],width=294 ] ) for cold gass galaxies with hi line detections in 3 ranges of stellar mass , stellar mass surface density and concentration ( blue histograms ) .
error bars are calculated using bootstrap resampling .
results for the f10 models are shown as black and red curves .
the colour - coding of the model curves has the same meaning as in figures 2 and 3 .
[ fracfunh1],width=317 ] mass fraction ( log m(h@xmath5)/m@xmath7 ) for cold gass galaxies with co line detections in 3 ranges of stellar mass , stellar mass surface density and concentration ( blue histograms ) .
error bars are calculated using bootstrap resampling .
results for the f10 models are shown as black and red curves .
the colour - coding of the model curves has the same meaning as in figures 2 and 3 .
[ fracfunh2],width=317 ] in this section , we investigate whether the `` transition '' between the population of galaxies with detectable atomic and molecular gas and the populations with gas fractions below @xmath48 occurs in the same way in models as in the data . in figures 7 and 8
, we plot the fraction of hi / co - detected galaxies as a function of stellar mass , stellar mass surface density , concentration index and specific star formation rate .
results from the survey are shown as solid blue lines .
the dashed blue lines indicate the @xmath49 uncertainty in the detected fraction as a function of these parameters .
models results are shown in red and black ( the colour - coding is the same as in figures 2 and 3 ) .
the main conclusion from these two figures is that the models and the data do not agree .
first , the dependence of the detected fraction on stellar mass is weaker in the data than in the models .
this is true for both the hi and the co - detected fractions .
second , the fraction of co - detected galaxies drops very strongly as a function of both stellar surface density and concentration in the real data .
this is not seen in the models .
errors can be read off directly .
results for the f10 models are shown as black ( kmt atomic - to - molecular gas prescription ) and red ( br pressure prescription ) curves .
[ nodetecth1],width=294 ] , width=294 ] it is also instructive to look at the distribution of detected and non - detected galaxies in two - dimensional planes of stellar mass and galaxy structural parameters .
this is illustrated in figures 9 and 10 .
we plot hi / co detected galaxies in blue and hi / co non - detected galaxies in red in the stellar surface density versus stellar mass , concentration versus stellar mass , and concentration versus stellar surface density planes .
results from the survey are shown in the top panels and results for `` mock catalogues '' of the same size generated from the simulations are shown in the bottom panels . in the data , there are clear _ thresholds _ in both @xmath9 ( @xmath50 kpc@xmath51 ) and @xmath10 ( @xmath52 ) , that demarcate the location of almost all galaxies without detectable gas .
no such threshold is seen in stellar mass @xmath27 .
this is true for _ both _ the hi and the co non - detections . in the models , the parameters that most clearly demarcates the location of the hi non - detections are stellar mass surface density and stellar mass .
it is the behaviour of the co non - detections , however , that is most discrepant with the models . _
galaxies without @xmath23 are not confined to a specific location in structural parameter space in the same way as in the data . _
we also note that the tight correlation between stellar surface density and concentration index seen in the top right panels of both figures , is not present in the models .
in addition , the distribution of concentration indices in the models extends to much lower values than in the real data .
the tight correlation between @xmath10 and @xmath53 tells us that in the real universe , galaxies with larger bulge - to - disk ratios have disks with higher stellar surface densities . in the f10 models ,
the effects of gas inflows on the disk are not taken into account , which may explain why there is no clear relation between bulge - to - disk ratio and stellar surface density . in a scenario where bulges form
when gas flows inwards as a result of bar - driven inflows or other dynamical instabilities , the gas flows will not only form the bulge , but also increase the stellar surface mass density in the inner disk .
we now elucidate the origin of the strong trend in the fraction of quenched galaxies as a function of stellar mass that is seen for the model galaxies . in current semi - analytic models ,
there are two physical processes that remove the supply of new gas to a galaxy and shut down star formation . 1 .
_ radio mode feedback .
_ black holes are able to grow by accreting hot gas from the surrounding halo .
the growth rate is given by @xmath54 ( croton et al 2006 ; guo et al 2011 ) , where @xmath55 is the ratio of hot gas mass to dark matter mass in the surrounding halo or sub - halo , @xmath56 is the virial velocity of the halo , @xmath57 is the black hole mass , and @xmath58 is an efficiency parameter .
some fraction of the rest mass energy of the accreted material is assumed to be transferred to the surrounding hot gas by radio jets .
the models assume an energy input rate @xmath59 , where @xmath60 is the speed of light .
this leads to a reduction in the cooling rate of hot gas of the form @xmath61 .
2 . _ gas stripping . _
when a galaxy is accreted by a more massive dark matter halo , it becomes a `` satellite '' . in the models of de lucia & blaizot ( 2007 ) , the hot gas surrounding the satellite
is stripped instantaneously , leading to a sharp reduction in the cooling rate onto the satellite . in the more recent models of guo
et al ( 2011 ) , the dark matter and hot gas surrounding the satellite are removed more gradually both by tidal forces and by ram - pressure stripping .
the parameter @xmath58 is tuned to reproduce the high mass end of the stellar mass function .
the net effect is illustrated in figure 11 where we plot galaxies from our model catalogue in the 2d planes of subhalo mass versus @xmath62 , subhalo mass versus @xmath63 , and subhalo mass versus @xmath10 .
galaxies that are predicted to be detected in co are colour - coded blue , while those predicted to be non - detections are colour - coded red .
figure 11 shows a clear transition between active and quenched galaxies at a subhalo mass of @xmath64 , which is a factor of @xmath65 larger than the mass where dark matter halos are predicted to transition to hosting a static halo of hot gas ( birnboim & dekel 2003 ; dekel & birnboim 2006 ; croton et al 2006 ) .
the reason why the transition is very sharp in halo mass is because the primary dependence of the mass that is accreted by the black hole is on halo virial velocity and not on black hole mass .
we note that there is an additional population of quenched galaxies in low mass subhalos .
these are the satellite galaxies , where the subhalo has been stripped by tidal forces .
as can be seen , they form a small minority of the quenched population in the stellar mass range of the galaxies in the gass survey .
figure 11 clearly illustrates _
why _ quenching in the model galaxies is most strongly dependent on stellar mass and surface density , and largely independent of concentration .
this is because in the model , stellar mass and surface density correlate with the virial mass of the host halo , but concentration does not .
there is direct observational evidence in support of these results .
mandelbaum et al ( 2006 ) used weak gravitational lensing to derive halo masses of galaxies as a function of stellar mass and morphology .
they found stellar mass to be a good proxy for halo mass .
for a given halo mass , the stellar mass was independent of concentration / morphology below @xmath66 . in the data ,
quenching thresholds are seen as a function of stellar surface density and concentration , but not as a function of stellar mass . even for galaxies with stellar masses as high as @xmath67 , which are observed to reside in halos with masses in the range @xmath68 , hi and co are generally still detected if concentrations and densities are low .
this strongly suggests that _ processes associated with bulge formation must be responsible for shutting off the gas supply in galaxies . _
we note that this hypothesis was already put forward by kauffmann et al ( 2006 ) , based on an analysis of the _ scatter _ in the colours and spectral properties of galaxies as a function of stellar mass , stellar surface density and concentration .
kauffmann et al ( 2006 ) made an ansatz that scatter in specific star formation rate reflected scatter in gas content .
the fact that specific star formation rates as well as their scatter decreased sharply above a characteristic density threshold of @xmath69 kpc@xmath51 and concentration index of @xmath70 and that this threshold was largely independent of the stellar mass of the galaxy , was taken as evidence that gas accretion was no longer occurring in bulge - dominated systems .
the cold gass and gass surveys have now demonstrated that the neutral gas content of galaxies is weakly dependent on stellar mass and decreases sharply at @xmath71 kpc@xmath51 and @xmath72 .
we have not yet ascertained _ why _ bulge - dominated galaxies no longer accrete gas and form stars efficiently .
figure 3 shows that galaxies with @xmath71 kpc@xmath51 have molecular gas mass fractions that are slightly depressed relative to model predictions .
could such galaxies be transitioning to the red sequence on short timescales as proposed by schawinski et al ( 2009 ) ?
we note that molecular gas is generally concentrated towards the inner regions of galaxies .
the sdss fiber spectra probe the central 1 - 2 kpc central regions of the gass galaxies and contain diagnostics of the recent star formation history in the bulge . in figure 12 , we plot h@xmath73 as a function of d@xmath74(4000 ) for the galaxies in the survey .
galaxies where co was not detected are colour coded red , those with detections and @xmath75 are colour coded - blue , and those with detections and @xmath76 are colour - coded cyan . as discussed in detail in kauffmann et al ( 2003a ) , by combining these two stellar absorption line indices , we can diagnose if the recent star formation histories of a population of galaxies have been smooth or `` bursty '' on average .
starbursts of duration less than 1 - 2 hundred million years , displace a significant fraction of galaxies to higher values of h@xmath73 at a given value of d@xmath74(4000 ) . likewise ,
if star formation is truncated over a timescale of less than a few hundred million years , galaxies will also be displaced to higher than average values of h@xmath73 for around a gyr following the truncation event ( see for example , kauffmann et al 2004 ) .
index is plotted as a function of 4000 break strength for cold gass with co line detections and @xmath75 ( blue ) .
co line detections and @xmath76 ( cyan ) and for cold gass galaxies with no co line detections ( red ) .
the errors on each measurement are also indicated for each galaxy .
[ nodetecth1],width=294 ] as can be seen , most gass galaxies lie on the same locus in the plane of h@xmath73 versus d@xmath74(4000 ) .
there is no evidence of any net displacement as a function of stellar surface density or as a function of @xmath23 mass fraction , indicating that the star formation histories of most high surface density galaxies with gas have been smooth .
we note , however , that galaxies with co line detections and high stellar surface densities _ do span the largest range in stellar population parameters_. the have d@xmath74(4000 ) values as low as those of the low - density h@xmath5-rich population , and as high as those of the `` quenched '' population .
we also note that the three post - starburst galaxies that are clearly displaced to higher valued of h@xmath73 , are all detected in @xmath23 have have high stellar surface densities .
all of them have unusually high molecular - to - atomic gas ratios , and one is clearly the end product of a recent merger ( tidal features are visible in the sdss image ) .
this region of parameter space is clearly quite complex , and may consist of several distinct sub - populations of galaxies on different evolutionary trajectories .
if so , additional data will be required to pull these apart . in particular
, studies of the _ distribution and the kinematics of the gas _ may shed significant light on its origin and its eventual fate .
we compare the semi - analytic models of galaxy formation of fu et al .
( 2010 ) , which track the evolution of atomic and molecular gas in galaxies , with gas fraction scaling relations derived from a stellar mass - limited sample of 299 galaxies from the cold gass survey .
these galaxies have measurements of the co(1 - 0 ) line from the iram 30-m telescope and the hi line from arecibo , as well as measurements of stellar masses , structural parameters and star formation rates derived from galex+sdss photometry .
our analysis addresses two key questions : 1 ) can the semi - analytic disk formation models explain the observed gas fraction scaling relations in galaxies with gas and ongoing star formation ? ,
2 ) does the transition between the population of galaxies with gas and the `` quenched '' population without gas occur in the same way in the models and in the data ? in answer to the first question , we conclude that the disk models provide a reasonable description of how condensed baryons are partitioned into stars , atomic gas and molecular gas as a function of galaxy mass and size .
trends as a function of bulge - to - disk ratio are not well reproduced .
in particular , the models do not account for the fact that at fixed stellar mass , there is a tight relation between the size of a galaxy and its bulge - to - disk ratio . in answer to the second question , we conclude that our data disagree with the current _ implementation _ of radio - mode feedback in the models . in the models ,
the observable parameter that best predicts whether a galaxy has been quenched is its stellar mass .
our data shows that the fraction of quenched galaxies is largely independent of stellar mass , but depends strongly on galaxy bulge - to - disk ratio and stellar surface density . in other words ,
even low mass galaxies with bulges have high probability of being quenched .
we conclude that processes associated with bulge formation are thus likely to be responsible both for depleting the neutral gas in galaxies and for shutting off the gas supply in these systems .
solving these problems will require substantive changes to the way gas transport and bulge and black hole formation is treated in the models , as well as to the way feedback from agn is implemented .
we now outline our best guess as to how this might work in practice . in the current models ,
newly accreted gas is assumed to have an exponential profile .
if the profile is initially shallower than exponential , more atomic gas would collect in an largely inert reservoir in the outer regions of the galaxy .
this gas would later be driven towards the center of the galaxy by dynamical perturbations , where it would form molecular gas and stars in the inner region of the disk .
the inflowing gas would also contribute to the growth of the bulge .
a prescription of this kind may result in a better correlation between stellar surface mass density and bulge - to - disk ratio .
there is now evidence from studies of complete samples of nearby galaxies that bars are very common in disk galaxies .
barazza , jogee & marinova ( 2008 ) find that 70% of disk - dominated galaxies host bars .
moreover , barred galaxies ( in particular galaxies with strong bars ) are associated with higher central molecular gas fractions and enhanced rates of central star formation ( sakamoto et al 1999 ; ellison et al 2011 ; wang et al 2012 , in preparation ) , suggesting that bulges are forming in these systems . finally , barred galaxies do not exhibit any axcess of nearby companions , suggesting that they are not triggered by interactions ( li et al .
2009 ) .
all of this suggests that internally - driven gas transport processes are important in the formation of bulges , but resolved maps of the atomic and molecular gas distributions in complete samples of galaxies and more detailed theoretical calculation of gas inflow rates in galaxies are required in order to quantify this in more detail . in the current models
, radio - mode feedback suppresses the cooling of gas in halos more massive than @xmath64 .
in addition , in massive , high density galaxies , atomic gas is transformed very efficiently into molecular gas and thence into stars .
these two mechanisms appear to be sufficient to place all `` hi - quenched '' galaxies in the high @xmath27 amd @xmath9 corner of parameter space , as seen in the bottom panel of figure 9 .
the same mechanisms are not , however , sufficient to place the `` h@xmath5-quenched '' galaxies in this same region of parameter space . nor
can they explain why galaxies deficient in both hi and h@xmath5 are almost always found in galaxies with substantial bulge components .
the simplest inference is that the neutral gas in galaxies is either consumed or removed when bulges form , and that subsequent gas accretion into the atomic phase is suppressed in a significant fraction of bulge - dominated galaxies .
more detailed studies of the spatially - resolved kinematics of the gas and how gas motions across galaxies correlate with the presence of accreting black holes , jets and the distribution of star formation , will be needed before we are able to pinpoint the actual physical mechanisms responsible for removing / depleting the gas in these systems ( see for example hopkins et al 2011 ) .
quenching may occur over short timescales during particular phases of bulge formation , so it may be necessary to survey large samples before a definitive conclusion can be reached .
in addition , the mechanisms that inhibit the late accretion of gas in bulge - dominated systems remain to be clarified .
this work is based on observations carried out with the iram 30 m telescope .
iram is supported by insu / cnrs ( france ) , mpg ( germany ) , and ign ( spain ) .
we sincerely thank the staff of the telescope for their help in conducting the cold gass observations and qi guo for helpful discussions .
|
we compare the semi - analytic models of galaxy formation of fu et al .
( 2010 ) , which track the evolution of the radial profiles of atomic and molecular gas in galaxies , with gas fraction scaling relations derived from a stellar mass - limited sample of 299 galaxies from the cold gass survey .
these galaxies have measurements of the co(1 - 0 ) line from the iram 30-m telescope and the hi line from arecibo , as well as measurements of star formation rates , stellar masses , galaxy sizes and concentration parameters from galex+sdss photometry .
the models provides a good description of how condensed baryons in star - forming galaxies are partitioned into atomic and molecular gas and stars as a function of galaxy stellar mass and stellar surface density .
the models do not reproduce the observed tight relation between stellar surface mass density and bulge - to - disk ratio in these galaxies .
the current implementation of `` radio - mode feedback '' in the models produces trends that disagree strongly with the data . in the models ,
gas cooling shuts down in nearly all galaxies in dark matter halos above a mass of @xmath0 . as a result ,
stellar mass is the observable that best predicts whether a galaxy has little or no neutral gas , i.e. whether a galaxy has been quenched .
in contrast , our data show that quenching is largely _ independent _ of stellar mass . instead , there are clear thresholds in bulge - to - disk ratio and in stellar surface density that demarcate the location of quenched galaxies in our chosen parameter space .
we speculate that the disagreement between the models and the observations may be resolved if radial transport of gas from the outer disk is included as an additional bulge - formation mechanism in the models .
in addition , we propose that processes associated with bulge formation play a key role in depleting the neutral gas in galaxies and that gas accretion is suppressed in a significant fraction of galaxies following the formation of the bulge , even in dark matter halos of low mass .
galaxies : fundamental parameters galaxies : evolution galaxies : ism radio lines : galaxies surveys
|
over the last decade the combination of hst and ground - based data has led to the measurement of the black - hole masses of some 40 inactive galaxies in the nearby universe via gas and stellar dynamics .
these black - hole mass measurements have led to the discovery of correlations between black - hole mass and the properties of the surrounding host galaxy , namely bulge luminosity and stellar velocity dispersion . in parallel to the discovery of these correlations in inactive galaxies , the last few years has also seen a new industry emerge in estimating the black - hole masses of active galaxies . in the first section of this review
i will discuss the recent advances in our understanding of the connection between black - hole mass and host - galaxy properties in both active and inactive galaxies at low redshift . in the second section
i will proceed to discuss how black - hole mass estimation in active galaxies has been used to investigate the connection between black - hole mass and radio power .
within the context of the study of agn black - hole masses , the most important aspect of the recent increase in the numbers of nearby quiescent galaxies with black - hole mass measurements has been the discovery of two correlations between black - hole mass and the properties of the surrounding host galaxy .
given the recent demonstration that at low redshift the host galaxies of all powerful agn are indistinguishable from nearby inactive ellipticals ( dunlop et al .
2003 ) , irrespective of radio luminosity , these correlations can now be applied to estimate the central black - hole masses of active galaxies .
the first of these correlations is between black - hole mass and the bulge luminosity of the host galaxy ( kormendy & richstone 1992 , magorrian et al .
this @xmath0 correlation has the advantage of being observationally inexpensive , particularly with respect to radio galaxies , but is traditionally thought to suffer from a large associated scatter @xmath1 dex ( although see below ) .
the second , more recently discovered , correlation is between black - hole mass and the velocity dispersion of the host galaxy stellar population ( gebhardt et al .
2000 , ferrarese & merritt 2000 ) .
this correlation has been hailed as being the more `` fundamental '' due to its very low associated scatter ( @xmath2 dex ) .
however , the @xmath3 correlation , although extremely tight , has several disadvantages with respect to studying the black - hole masses of active galaxies .
firstly , even at low redshift it is not possible to obtain central stellar velocity dispersions for powerful quasars due to the presence of the unresolved nuclear point source .
furthermore , although it is possible to obtain velocity dispersion measurements for radio galaxies at low redshift , this will not be practicable at @xmath4 even with 8m - class telescopes . consequently , before proceeding to discuss a more direct method of estimating black - hole mass in agn , i will digress slightly to discuss the possibility that the intrinsic scatter in the @xmath0 relation is significantly lower than has previously been appreciated .
it has occurred to a number of authors that the high level of scatter associated with the @xmath0 relation could well be an artifact of the difficulties in determining accurate bulge luminosities for late - type galaxies .
this issue is addressed in fig [ fig1 ] ( taken from mclure & dunlop 2002 ) which shows a comparison of the @xmath0 and @xmath3 relations for a sample of 18 quiescent galaxies drawn from the list of objects with reliable black - hole mass measurements published by kormendy & gebhardt ( 2001 ) .
these 18 objects comprise those objects from the kormendy & gebhardt compilation which have pure e - type morphology , and therefore allow an investigation of the influence of incorrectly determined bulge luminosity upon the scatter in the @xmath0 relation .
the best - fitting relations are shown as solid lines in both panels of fig [ fig1 ] and display an identical level of associated scatter ; @xmath2 dex .
this comparison demonstrates that when objects for which it is problematic to obtain accurate bulge luminosities are excluded , the scatter in the @xmath0 relation is comparable to that of the @xmath3 relation .
this fact has obvious consequences for the study of black - hole mass in radio - loud agn , particularly with respect to radio galaxies for which it is straightforward to obtain relatively accurate bulge luminosities ( eg . bettoni et al .
furthermore , graham et al . ( 2001 ) and erwin et al . (
2002 ) have recently demonstrated that , at least for quiescent galaxies , if sufficient care is taken over the modelling of the galaxy surface brightness profiles , then the scatter in both the @xmath0 relation and a correlation between black - hole mass and galaxy light concentration is only @xmath5 dex , irrespective of morphological type .
although the black - hole mass - host galaxy correlations discussed above are useful for estimating the black - hole masses of agn , there is a more direct method of black - hole mass estimation which can be applied to broad - line agn .
this method is the so - called virial black - hole mass estimator , and utilizes the width of broad emission lines in quasar spectra to directly trace the gravitational potential of the central black - hole on the scale of light - days .
the underlying assumption behind the method is that the motion of the line - emitting material is virialized . under this assumption
the width of the broad lines can be used to trace the keplerian velocity of the broad - line gas , and thereby allow an estimate of the central black - hole mass via the formula : @xmath6 , where @xmath7 is the broad - line region ( blr ) radius and @xmath8 is the keplerian velocity of the blr gas .
recent evidence to support the keplerian interpretation of agn broad - line widths has been presented by peterson & wandel ( 2000 ) and onken & peterson ( 2002 ) .
these authors demonstrate , at least for several well studied seyfert galaxies , that the fwhm of various emission lines follow the @xmath9 relation expected for keplerian motion .
it is also worth noting that peterson ( 2002 ) has recently demonstrated that the virial black - hole masses of reverberation mapped seyfert galaxies lie on the same @xmath3 relation as nearby inactive galaxies . by utilizing the virial black - hole mass estimator it is possible to investigate whether the host galaxies of agn do follow the same @xmath0 relation as inactive galaxies in the nearby universe ( laor 2001 , mclure & dunlop 2002 ) . in fig [ fig2 ] the results of a study to investigate the @xmath0 relation of 72 @xmath10 agn by mclure & dunlop ( 2002 )
are shown .
also shown is the sample of nearby inactive galaxies with reliable black - hole mass measurements previously discussed in fig [ fig1 ] .
the agn sample ( @xmath11 % radio - loud ) consists of objects for which the host - galaxy bulge luminosity was determined via modelling of either hst or high - resolution ground - based imaging .
it can clearly be seen that the agn host galaxies follow a tight ( 0.4 dex scatter ) correlation which is identical to that followed by nearby quiescent galaxies .
indeed , the best - fitting linear relation ( dotted line ) is equivalent to @xmath12 , identical to that determined by merritt & ferrarese ( 2001 ) for nearby quiescent galaxies .
in the second section of this review i will proceed to examine recent evidence regarding the connection between black - hole mass and radio power .
figure [ fig3 ] is taken from a recent paper by dunlop et al .
( 2003 ) and shows 5 ghz radio luminosity plotted against black - hole mass for two samples .
the first sample is comprised of 33 agn , radio galaxies , radio - loud quasars and radio - quiet quasars , who s bulge / black - hole masses have been determined from the analysis of hst imaging .
the second sample features the nearby quiescent galaxies investigated by franceschini et al .
( 1998 ) , who found a tight correlation between black - hole mass and 5 ghz radio power of the form : @xmath13 . in fig
[ fig3 ] there are three parallel relations plotted , each of the form @xmath14 , separated from each other by 2.5 decades .
the lowest of these three relations appears to represent a lower limit to the radio output of a black - hole of a given mass .
there are two points concerning fig [ fig3 ] which are worth highlighting .
firstly , it can be seen that many of the radio - quiet quasars from the dunlop et al .
sample lie up against the apparent radio power lower limit .
this is important because the radio - quiet and radio - loud quasars were selected by dunlop et al .
to have identical optical luminosities .
consequently , it can immediately be seen that the `` radio - quietness '' of these quasars is not due to their central engines being starved of fuel , because they are still producing large optical luminosities .
the second point refers to the hypothesized upper envelope plotted in fig [ fig3 ] . clearly ,
if the franceschini - type relation does represent a lower limit to the radio output of a given black - hole mass , then it is interesting to inquire about the form of the corresponding upper envelope . although the form of the upper envelope plotted in fig [ fig3 ] is speculative , at least based on the data from dunlop et al .
, further support is provided by the data from the study of lacy et al .
( 2001 ) .
figure [ fig4 ] is an adapted version of fig 2 from lacy et al .
( 2001 ) , which again shows the @xmath15 plane , but this time populated by objects from the first bright quasar survey ( fbqs ) , the pg quasar survey and several notable low redshift objects .
figure [ fig4 ] clearly demonstrates that radio power and black - hole mass are related , although a large dynamic range is required in both parameters for the relation to become apparent .
the dotted line in fig [ fig4 ] is the best - fitting relation determined by lacy et al .
( not including accretion rate as an additional parameter ) and has the form @xmath16 .
however , in order to offer an alternative interpretation of the lacy et al .
data i have annotated fig [ fig4 ] with the addition of the upper and lower limits on radio - power suggested by dunlop et al . (
the lower radio power limit describes the lacy et al .
data extremely well , and again shows radio - quiet quasars lying tight up against the relation defined by quiescent galaxies , despite being well supplied with fuel .
furthermore , the lacy et al .
data also obey the upper limit on radio - power over 4 orders of magnitude in black - hole mass and 11 orders of magnitude in radio luminosity . given that objects appear to exist between radio power boundaries spanning 5 orders of magnitude , it is therefore understandable that studies which adopt the radio - loudness parameter @xmath17 tend to find little or no correlation between radio - loudness and black - hole mass ( eg .
woo & urry 2002 ) .
one final interesting point is illustrated by fig [ fig4 ] .
although there is clearly a large scatter associated with the @xmath15 relation , it is still true that the most radio - loud quasars are those with the most massive black - holes .
for example , fig [ fig4 ] shows that quasars with @xmath18whzsr@xmath19 are virtually exclusively associated with black - holes of mass @xmath20 .
one consequence of this fact is that selecting samples based on extreme radio luminosity offers a method for selecting objects with the largest black - holes , and presumably bulge masses , at any given epoch . as a result
, it should therefore be possible to cleanly trace the evolution of the @xmath0 relation in a sub - set of the most massive galaxies in the universe .
the conclusions regarding the connection between black - hole mass and radio power can be summarizes as follows : * there does not exist a threshold in black - hole mass above which an agn must be radio - loud .
there is a large overlap in the black - hole mass distributions of radio - loud and radio - quiet quasars , although at a given optical luminosity radio - loud quasars do appear to be biased to higher black - hole masses than their radio - quiet counterparts . *
black - hole mass and radio luminosity are connected , although an extremely large dynamic range in both parameters is required for this connection to become apparent .
* the position of an object on the @xmath15 plane appears to be a function of an additional physical parameter .
although accretion rate may be playing a role , the lack of a tight correlation between radio and optical luminosity suggests that some other parameter , such as black - hole spin , may be involved . *
selecting the most radio - loud objects , as defined by their radio luminosity alone , is an effective method of isolating the objects with the most massive black - holes , and presumably host galaxies , at any epoch . * selecting the most radio - loud objects as a function of redshift provides an ideal method for cleanly studying the evolution of the @xmath0 relation within the sub - set of the most massive galaxies .
999 d. bettoni , et al.,_a&a _ , in press , astro - ph/0212162 , ( 2003 ) j. dunlop , et al . ,
_ mnras _ , in press , astro - ph/0108397 , ( 2003 ) p. erwin , et al . , _ carnegie observatories astrophysics series , vol . 1 : coevolution of black holes and galaxies , _ l. ho ed .
, 2002 l. ferrarese , d. merritt , _ apj _ * 539 * ( 2000 ) l9 a. franceschini , et al .
_ mnras _ * 297 * ( 1998 ) 817 k. gebhardt et al . ,
_ apj _ * 539 * ( 2000 ) l13 a. graham et al . , _ apj _ * 563 * ( 2001 ) l11 j. kormendy , d. richstone , _ apj _ * 393 * ( 1992 ) 559 j. kormendy , k. gebhardt , _
20th texas symposium on relativistic astrophysics , _ c. wheeler , h. martel ed .
2001 , p363 m. lacy , et al . , _ apj _ * 551 * ( 2001 ) 17 a. laor , _ apj _ * 553 * ( 2001 ) l01 j. magorrian , et al . _
aj _ * 115 * ( 1998 ) 2285 r. mclure , j. dunlop , _ mnras _ * 331 * ( 2002 ) 795 d. merritt , l. ferrarese , _ mnras _ * 320 * ( 2001 ) l30 c. onken , b. peterson , _ apj _ * 572 * ( 2002 ) 746 b. peterson , _ active galactic nuclei : from central engine to host galaxy _ , s. collin , f. combes , i. shlosman ed . , 2002 b. peterson , a. wandel , _ apj _ * 540 * ( 2000 ) l13 j. woo , c. urry , astro - ph/0211118 , 2002
|
in this review the correlations between host galaxy properties and black - hole mass determined from nearby quiescent galaxies are briefly discussed , before proceeding to investigate their application to active galactic nuclei ( agn ) .
the recent advances in estimating the black - hole masses of agn are then reviewed , followed by an examination of the connection between black - hole mass and radio luminosity .
galaxies : fundamental parameters - galaxies : active - quasars : general - quasars : emission lines
|
recent deep radio surveys ( @xmath9 mjy ) have shown that normalized radio source counts flatten below a few mjy .
this has been interpreted as being due to a new population which does not show up at higher fluxes where counts are dominated by active galactic nuclei ( agn ) . to clarify its nature
it is necessary to get detailed information on the radio properties of normal galaxies in the nearby universe ( z @xmath10 0.15 ) , down to faint flux limits and to have at hand large samples of mjy and sub - mjy sources , for subsequent optical identification and spectroscopic work . to this end
we have surveyed a large area ( @xmath11 26 sq .
degr ) with the atca at 1.4 ghz with a bandwidth of @xmath12 mhz .
the properties of normal nearby galaxies can be easily derived , because this area coincides with the region in which the esp redshift survey was conducted ( vettolani et al .
samples of faint galaxies over large areas are necessary to avoid bias due to local variations in their properties .
present samples of faint radio sources are confined to small regions with insufficient source statistics .
+ the present paper contains the radio source catalogue derived from the atesp survey . the full outline of the radio survey , its motivation in comparison with other surveys , and the description of the mosaic observing technique which was used to obtain an optimal combination of uniform and high sensitivity over the whole area has been presented in prandoni et al .
( 2000 , paper i ) .
the source counts derived from the atesp survey will be presented and discussed separately in a following paper .
the paper is organized as follows : in sect .
[ sec - det ] we describe the source detection and parameterization ; the catalogue is described in sect .
[ sec - cat ] and the accuracy of the parameters ( flux densities , sizes and positions ) are discussed in sect .
[ sec - errors ] .
the atesp survey consists of 16 radio mosaics with spatial resolution @xmath0 .
the survey was designed so as to provide uniform sensitivity over the whole region ( @xmath4 sq .
degrees ) of the esp redshift survey . to achieve this goal
a larger area was observed , but we have excluded from the analysis the external regions ( where the noise is not uniform and increases radially ) . in the region with uniform sensitivity the noise level varies from 69 @xmath3jy to 88 @xmath3jy , depending on the radio mosaic , with an average of 79 @xmath3jy ( see @xmath13 values reported in table 3 of paper i and repeated also in table [ tab - mospar ] of appendix [ sec - appb ] , at the end of this paper ) . for consistency with paper
i , in the following such sensitivity average values are denoted by the symbol @xmath13 .
this means that sensitivity values have been defined as the _ full width at half maximum _ ( fwhm ) of the gaussian that fits the pixel flux density distribution in each mosaic ( see paper i for more details ) .
+ a number of source detection and parameterization algorithms are available , which were developed for deriving catalogues of components from radio surveys .
we decided to use the algorithm _ image search and destroy _ ( imsad ) available as part of the _ multichannel image reconstruction , image analysis and display _ ( miriad ) package ( sault & killeen @xcite ) , as it is particularly suited to images obtained with the atca .
+ imsad selects all the connected regions of pixels ( _ islands _ ) above a given flux density threshold .
the _ islands _ are the sources ( or the source components ) present in the image above a certain flux limit .
then imsad performs a two - dimensional gaussian fit of the _ island _ flux distribution and provides the following parameters : position of the centroid ( right ascension , @xmath14 , and declination , @xmath15 ) , peak flux density ( @xmath16 ) , integrated flux density ( @xmath17 ) , fitted angular size ( major , @xmath18 , and minor , @xmath19 , fwhm axes , not deconvolved for the beam ) and position angle ( p.a . ) .
+ imsad proved to have an average success rate of @xmath20 down to very faint flux levels ( see below ) . since imsad attempts to fit a single gaussian to each _ island _ , it obviously tends to fail ( or to provide very poor source parameters ) when fitting complex ( _ i.e. _ non - gaussian ) shapes .
we used imsad to extract and parameterize all the sources and/or components in the uniform sensitivity region of each mosaic .
as a first step , a preliminary list containing all detections with @xmath21 ( where @xmath22 is the average mosaic rms flux density ) was extracted .
detection thresholds vary from 0.3 mjy to 0.4 mjy , depending on the radio mosaic .
we visually inspected all _ islands _ ( @xmath23 ) detected , in order to check for obvious failures and/or possibly poor parameterization , that need further analysis .
problematic cases were classified as follows : * _ islands _ for which the automated algorithm provides poor parameters and therefore are to be re - fitted with a single gaussian , constraining in a different way the initial conditions ( @xmath24 of the total ) . * _ islands _ that could be better described by two or more split gaussian components ( @xmath25 of the total ) ; * _ islands _ which can not be described by a single or multiple gaussian fit ( @xmath26 of the total ) ; * _ islands _ corresponding to obviously unreal sources , typically noise peaks and/or image artifacts in noisy regions of the images ( @xmath27 of the total ) ; the goodness of gaussian fit parameters was checked by comparing them with reference values , defined as follows . positions and peak flux densities were compared to the corresponding values derived by a second - degree interpolation of the _ island_. such interpolation usually provides very accurate positions and peak fluxes .
gaussian integrated fluxes were compared to the ones derived directly by summing pixel per pixel the flux density in the source area , defined as the region enclosed by the @xmath28 flux density contour .
flux densities were considered good whenever the difference between the gaussian and the reference value @xmath29 was @xmath30 .
positions were considered good whenever they did fall within the @xmath31 flux contour .
for the first three groups listed above _ ad hoc _ procedures were attempted aimed at improving the fit .
+ single - component fits were considered satisfactory whenever positions and flux densities satisfy the tolerance criteria defined above .
+ in a few cases gaussian fits were able to provide good values for positions and peak flux densities , but did fail in determining the integrated flux densities .
this happens typically at faint fluxes ( @xmath32 ) .
gaussian sources with a poor @xmath17 value are flagged in the catalogue ( see sect . [ sec - cformat ] ) .
+ the _ islands _ successfully split in two or three components are 67 in total ( 64 with two components and 3 with three components ) .
+ for non - gaussian sources we adopted as parameters the reference positions and flux densities defined above .
the source position angle was determined by the direction in which the source is most extended and the source axes were defined as _ largest angular sizes _ ( _ las _ ) , i.e. the maximum distance between two opposite points belonging to the @xmath33 flux density contour along ( major axis ) and perpendicular to ( minor axis ) the same direction .
all the non - gaussian sources are flagged in the catalogue ( see sect . [ sec - cformat ] ) .
the procedures described in the previous section yielded a preliminary list of sources ( or source components ) for further investigation . in order to minimize the incompleteness effects present at fluxes approaching the source extraction threshold ( see fig .
[ fig - fluxhist ] ) we decided to insert in the final catalogue only the sources with @xmath34 , where @xmath22 is the mosaic rms flux density .
this threshold has been chosen after inspection of the local noise distribution .
the local noise ( @xmath35 ) has been defined as the average noise value in a box of about @xmath36 size around a source .
usually the local noise does not show significant systematic departures from the mosaic average rms value : the @xmath37 distribution can be described fairly well by a gaussian with @xmath38 and peak value equal to 1.01 ( see fig .
[ fig - lrmshist ] ) . +
this can be seen also in fig .
[ fig - lrms ] , where we show , for each source , the signal - to - noise ratio defined using both @xmath35 and @xmath13 .
the two signal - to - noise ratios mostly agree with each other , although a number of significant departures are evident for the faintest sources .
this is due to the presence of some residual areas where the noise is not random due to systematic effects ( noise peaks and stripes ) .
this is caused by the limited dynamical range in presence of very strong sources ( stronger than @xmath39 mjy , see also paper i ) .
it is worth noting that also the systematic departures from the expected behavior at the brightest end of the plot ( @xmath40 ) are a consequence of the same problem .
+ in mosaic regions where local noise is significantly larger , we applied a @xmath41 cut - off if @xmath42 ( assuming a normal distribution for the noise , the probability to get a local to average noise value @xmath43 is @xmath44 ) .
this resulted in the rejection of 32 sources ( see the region in fig .
[ fig - lrms ] defined by @xmath45 , @xmath46 and @xmath47 ) . + the criteria discussed above proved to be very effective in selecting out noise artifacts from the catalogue .
nevertheless a few ( 6 ) sources , which satisfy both the average and the local noise constraints are , however , evident noise artifacts at visual inspection .
such objects have been rejected from the final catalogue .
+ the adopted criteria for the final catalogue definition also allowed us to significantly reduce the number of poor gaussian fits ( see sect .
[ sec - refit ] ) since a large fraction of them ( @xmath48 ) have fainter @xmath16 : we are left with 50 poor gaussian fits ( flagged _ ` s * ' _ , see sect .
[ sec - cformat ] ) in the final catalogue . + a total number of 3172 source components entered the final catalogue .
some of them have to be considered different components of a unique source and , as discussed later in sect .
[ sec - mult ] , the number of distinct sources in the atesp catalogue is 2960 . the ratio of the integrated flux to the peak flux is a direct measure of the extension of a radio source : @xmath49 where @xmath19 and @xmath18 are the source fwhm axes and @xmath50 and @xmath51 are the synthesized beam fwhm axes .
the flux ratio can therefore be used to discriminate between extended ( larger than the beam ) and point - like sources .
+ in fig .
[ fig - stspratio ] we have plotted the flux ratio as a function of the signal - to - noise for all the sources ( or source components ) in the atesp catalogue . the flux density ratio has a skew distribution , with a tail towards high flux ratios due to extended sources .
values for @xmath52 are due to the influence of the image noise on the measure of source sizes ( see sect . [ sec - errors ] ) . to establish a criterion for extension , such errors have to be taken into account .
we have determined the lower envelope of the flux ratio distribution ( the curve containing 90% of the @xmath53 sources ) and we have mirrored it on the @xmath54 side ( upper envelope in fig .
[ fig - stspratio ] ) . we have considered as unresolved all sources laying below the upper envelope .
the upper envelope can be characterized by the equation : @xmath55 \ ; .\ ] ] from this analysis we found that 1864 of the 3172 sources ( or source components ) in the catalogue ( i.e. @xmath56 ) have to be considered unresolved . + it is worth noting that the envelope does not converge to 1 going to large signal - to - noise values .
this is due to the radial smearing effect .
it systematically reduces the source peak fluxes , yielding larger @xmath57 ratios ( see discussion in sect .
[ sec - corr ] ) . from fig .
[ fig - stspratio ] we can quantify the smearing effect in @xmath24 on average . + deconvolved angular sizes are given in the catalogue only for sources above the upper curve ( filled circles in fig .
[ fig - stspratio ] ) . for unresolved sources ( dots in fig .
[ fig - stspratio ] ) deconvolved angular sizes are set to zero .
+ note that no bandwidth correction to deconvolved sizes has been applied .
correcting for such effect would be somewhat complicated by the fact that each source in the radio mosaics is a sum of contributions from several single pointings . in fig .
[ fig - doubles ] the ( nearest neighbor ) pair density distribution is shown as a function of distance ( histogram )
. also indicated is the expected distribution if all the sample sources ( components ) were randomly distributed in the sky .
the expected distribution has been scaled so as to have the same area below the curve and the observed histogram .
the excess at small distances is clearly due to physical associations and , because of the normalization chosen , is compensated by a deficiency at larger distances ( between @xmath58 and @xmath59 ) . +
all the components closer than @xmath60 ( _ i.e. _ about three times the beam size ) have been considered as possibly belonging to a unique double source .
triple sources are defined whenever one additional component is closer than @xmath60 to ( at least ) one of the pair components . for multiple sources the same criterion
is applied iteratively . applying this distance constraint
we expect that @xmath61 of the pairs are random superpositions .
+ the flux ratio distribution between the pair components has a large spread at all distances ( see fig .
[ fig - fratio ] ) . to reduce the contamination
we have discarded all the pairs with flux ratio larger than a factor 10 . for triples and multiple sources the probable core
is not considered when computing the flux ratios .
+ a few departures from the adopted criteria are present in the catalogue .
for example the triple source atesp j005620 - 394145 and the double source atesp j011029 - 393253
have @xmath62 but do not satisfy the flux ratio constraint .
all exceptions are based on source geometry considerations and/or the analysis of the source field .
+ in order to increase the multiple sources sub - sample completeness , we added 31 sources with distances @xmath63 , which show clear signs of physical associations between their components ( see fig .
[ fig - examples ] for some examples ) .
no flux ratio constraints have been applied to such sources . in fig .
[ fig - fratio ] are shown the flux ratios for all the pairs in the final sample of multiple sources ( filled circles ) . + as a final result we have 189 multiple sources : 168 doubles , 19 triples and 2 sources with four components . as a consequence , the initial list of 3172 radio components results in a catalogue of 2960 distinct radio sources .
lllrrrrrl & & & & & & & + & & & & + atesp j223235 - 402642 & 22:32:35.62 & -40:26:42.1 & 1.26 & 6.07 & 27.13 & 13.55 & 28.9 & s + atesp j223237 - 393113 & 22:32:37.71 & -39:31:13.2 & 0.69 & 0.71 & 0.00 & 0.00 & 0.0 & s + atesp j223238 - 394102 & 22:32:38.54 & -39:41:02.9 & 0.48 & 0.58 & 0.00 & 0.00 & 0.0 & s + atesp j223242 - 393054 & 22:32:42.10 & -39:30:54.3 & 2.33 & 2.46 & 0.00 & 0.00 & 0.0 & s + atesp j223248 - 401345 & 22:32:48.24 & -40:13:45.9 & 0.95 & 0.83 & 0.00 & 0.00 & 0.0 & s + atesp j223250 - 394059 & 22:32:50.79 & -39:40:59.7 & 0.57 & 0.93 & 10.47 & 0.00 & -88.4 & s + atesp j223252 - 401925 & 22:32:52.45 & -40:19:25.1 & 0.66 & 0.46 & 0.00 & 0.00 & 0.0 & s + atesp j223254 - 393652 & 22:32:54.54 & -39:36:52.0 & 1.03 & 1.70 & 9.89 & 5.39 & 48.3 & s + atesp j223255 - 395717 & 22:32:55.51 & -39:57:17.3 & 0.51 & 0.45 & 0.00 & 0.00 & 0.0 & s + atesp j223256 - 402010 & 22:32:56.10 & -40:20:10.3 & 1.13 & 1.08 & 0.00 & 0.00 & 0.0 & s + atesp j223301 - 393017 & 22:33:01.55 & -39:30:17.1 & 7.83 & 8.67 & 3.39 & 2.66 & -57.0 & s + atesp j223302 - 402817 & 22:33:02.31 & -40:28:17.5 & 0.66 & 0.71 & 0.00 & 0.00 & 0.0 & s + atesp j223303 - 401629 & 22:33:03.07 & -40:16:29.2 & 0.99 & 1.27 & 5.30 & 4.74 & 62.9 & s + atesp j223304 - 395639 & 22:33:04.92 & -39:56:39.0 & 1.85 & 2.94 & 23.51 & & &
m + atesp j223304 - 395639a & 22:33:04.43 & -39:56:41.4 & 1.85 & 2.18 & 6.66 & 1.43 & -24.2 & s + atesp j223304 - 395639b & 22:33:06.31 & -39:56:32.2 & 0.80 & 0.76 & 0.00 & 0.00 & 0.0 & s + atesp j223313 - 400216 & 22:33:13.42 & -40:02:16.1 & 7.60 & 9.18 & 4.94 & 3.65 & -45.7 & s + atesp j223314 - 394942 & 22:33:14.42 & -39:49:42.8 & 1.28 & 1.33 & 0.00 & 0.00 & 0.0 & s + atesp j223316 - 393124 & 22:33:16.90 & -39:31:24.4 & 0.52 & 0.68 & 0.00 & 0.00 & 0.0 & s + atesp j223317 - 393235 & 22:33:17.19 & -39:32:35.0 & 2.40 & 3.37 & 9.29 & 4.22 & 9.4 & s + atesp j223320 - 394713 & 22:33:20.14 & -39:47:13.8 & 0.96 & 1.08 & 0.00 & 0.00 & 0.0 & s + atesp j223322 - 401710 & 22:33:22.93 & -40:17:10.7 & 4.92 & 6.26 & 7.56 & 3.34 & 14.7 & s + atesp j223327 - 395836 & 22:33:27.45 & -39:58:36.9 & 3.24 & 5.16 & 13.60 & 3.39 & -5.9 & s + atesp j223327 - 394541 & 22:33:27.73 & -39:45:41.7 & 2.74 & 5.70 & 21.21 & & & m + atesp j223327 - 394541a & 22:33:27.08 & -39:45:40.2 & 2.74 & 3.65 & 5.98 & 4.83 & 60.7 & s + atesp j223327 - 394541b & 22:33:28.89 & -39:45:44.4 & 1.29 & 2.05 & 10.04 & 0.00 & -73.8 & s + atesp j223329 - 402019 & 22:33:29.10 & -40:20:19.6 & 1.41 & 1.82 & 6.85 & 3.75 & -32.1 & s + atesp j223330 - 395233 & 22:33:30.97 & -39:52:33.4 & 0.50 & 0.53 & 0.00 & 0.00 & 0.0 & s + atesp j223335 - 401337 & 22:33:35.21 & -40:13:37.7 & 0.63 & 0.71 & 0.00 & 0.00 & 0.0 & s + atesp j223337 - 394253 & 22:33:37.01 & -39:42:53.8 & 2.58 & 2.79 & 0.00 & 0.00 & 0.0 & s + atesp j223338 - 392919 & 22:33:38.49 & -39:29:19.6 & 5.26 & 5.62 & 0.00 & 0.00 & 0.0 & s + atesp j223339 - 393131 & 22:33:39.86 & -39:31:31.0 & 1.37 & 1.32 & 0.00 & 0.00 & 0.0 & s + atesp j223343 - 393811 & 22:33:43.21 & -39:38:11.2 & 1.00 & 1.05 & 0.00 & 0.00 & 0.0 & s + atesp j223343 - 402307 & 22:33:43.93 & -40:23:07.4 & 0.90 & 1.10 & 0.00 & 0.00 & 0.0 & s + atesp j223345 - 402815 & 22:33:45.23 & -40:28:15.6 & 0.56 & 0.65 & 0.00 & 0.00 & 0.0 & s + atesp j223346 - 393322 & 22:33:46.35 & -39:33:22.9 & 1.26 & 1.23 & 0.00 & 0.00 & 0.0 & s +
atesp j223351 - 394040 & 22:33:51.19 & -39:40:40.9 & 1.12 & 1.17 & 0.00 & 0.00 & 0.0 & s + atesp j223356 - 401949 &
22:33:56.57 & -40:19:49.7 & 1.71 & 2.28 & 36.37 & & & m + atesp j223356 - 401949a & 22:33:55.33 & -40:20:11.8 & 0.52 & 0.63 & 0.00 & 0.00 & 0.0 & s + atesp j223356 - 401949b & 22:33:57.05 & -40:19:41.2 & 1.71 & 1.65 & 0.00 & 0.00 & 0.0 & s + atesp j223358 - 400642 & 22:33:58.77 & -40:06:42.4 & 1.29 & 1.20 & 0.00 & 0.00 & 0.0 & s + atesp j223401 - 402310 & 22:34:01.25 & -40:23:10.8 & 0.58 & 1.18 & 10.98 & 9.14 & 32.8 & s + atesp j223401 - 393448 & 22:34:01.87 & -39:34:48.9 & 1.69 & 2.09 & 5.28 & 3.10 & -72.8 & s + atesp j223402 - 402357 & 22:34:02.36 & -40:23:57.7 & 0.59 & 0.57 & 0.00 & 0.00 & 0.0 & s + atesp j223402 - 400017 & 22:34:02.64 & -40:00:17.3 & 1.70 & 1.99 & 5.37 & 2.86 & -28.2 & s + atesp j223404 - 395831 & 22:34:04.00 & -39:58:31.5 & 6.16 & 7.45 & 5.53 & 3.43 & 29.7 & s
+ atesp j223404 - 393358 & 22:34:04.08 & -39:33:58.9 & 0.85 & 0.90 & 0.00 & 0.00 & 0.0 & s + atesp j223407 - 393721 & 22:34:07.34 & -39:37:21.9 & 3.14 & 6.41 & 25.71 & & & m + atesp j223407 - 393721a & 22:34:06.35 & -39:37:27.4 & 2.03 & 3.23 & 8.90 & 6.08 & 38.9 & s + atesp j223407 - 393721b & 22:34:08.36 & -39:37:16.3 & 3.14 & 3.19 & 0.00 & 0.00 & 0.0 & s + atesp j223409 - 394258 & 22:34:09.58 & -39:42:58.2 & 0.71 & 1.11 & 10.16 & 5.23 & -15.4 & s + atesp j223410 - 394427 & 22:34:10.82 & -39:44:27.6 & 1.54 & 1.59 & 0.00 & 0.00 & 0.0 & s +
atesp j223412 - 400254 & 22:34:12.80 & -40:02:54.6 & 0.66 & 0.69 & 0.00 & 0.00 & 0.0 & s + atesp j223413 - 393242 & 22:34:13.28 & -39:32:42.4 & 1.58 & 1.88 & 6.21 & 0.00 & 59.9 & s + atesp j223413 - 393650 & 22:34:13.74 & -39:36:50.2 & 0.48 & 1.71 & 25.27 & 8.28 & 38.4 & s + atesp j223413 - 395651 & 22:34:13.80 & -39:56:51.7 & 0.52 & 0.56 & 0.00 & 0.00 & 0.0 & s + atesp j223420 - 393150 & 22:34:20.47 & -39:31:50.5 & 0.55 & 0.73 & 0.00 & 0.00 & 0.0 & s + in the final catalogue we have 23 non - gaussian sources whose parameters have been defined as discussed in sect .
[ sec - refit ] .
in particular we notice that positions refer to peak positions , which , for non - gaussian sources does not necessarily correspond to the position of the core .
we also notice that we can have non - gaussian components in multiple sources .
some examples of single and multiple non gaussian sources are shown in fig .
[ fig - examples ] .
the electronic version of the full radio catalog is available through the atesp page at http://www.ira.bo.cnr.it .
its first page is shown as an example in table [ tab - cat ] .
the source catalogue is sorted on right ascension .
the format is the following : source iau name .
different components of multiple sources are labeled ` a ' , ` b ' , etc .
+ source position : right ascension and declination ( j2000 ) .
+ source peak ( @xmath16 ) and integrated ( @xmath17 ) flux densities in mjy ( baars et al .
@xcite scale ) .
the flux densities are not corrected for the systematic effects discussed in sect .
[ sec - corr ] .
+ intrinsic ( deconvolved from the beam ) source angular size .
full width half maximum of the major ( @xmath64 ) and minor ( @xmath65 ) axes in arcsec .
zero values refer to unresolved sources ( see sect .
[ sec - deconv ] for more details ) .
+ source position angle ( p.a .
, measured n through e ) for the major axis in degrees .
+ flag indicating the fitting procedure and parameterization adopted for the source or source component ( see sects .
[ sec - refit ] and [ sec - mult ] ) .
_ ` s ' _ refers to gaussian fits . _ ` s * ' _ refers to poor gaussian fits . _ ` e ' _ refers to non - gaussian sources . _
` m ' _ refers to multiple sources ( see below ) .
the parameters listed for non - gaussian sources are defined as discussed in sect .
[ sec - refit ] .
+ for multiple sources we list all the components ( labeled ` a ' , ` b ' , etc . ) preceded by a line ( flagged _ ` m ' _ ) giving the position of the radio centroid , total flux density and overall angular size of the source .
source positions have been defined as the flux - weighted average position of all the components ( source centroid ) . for sources with more than two components
the centroid position has been replaced with the core position whenever the core is clearly recognizable .
+ integrated total source flux densities are computed by summing all the component integrated fluxes .
+ the total source angular size is defined as _ las _ ( see sect .
[ sec - refit ] ) and it is computed as the maximum distance between the source components .
parameter uncertainties are the quadratic sum of two independent terms : the calibration errors , which dominate at high signal - to - noise ratios , and the internal errors , due to the presence of noise in the maps .
the latter dominate at low signal - to - noise ratios .
+ in the following sections we discuss the parameter internal accuracy of our source catalogue .
master equations for total rms error derivation , with estimates of the calibration terms are reported in appendix [ sec - appa ] . in order to quantify the internal errors we produced a one square degree residual map by removing all the sources detected above @xmath66 in the radio mosaic fld20to25 .
we performed a set of montecarlo simulations by injecting gaussian sources in the residual map at random positions and re - extracting them using the same detection algorithm used for the survey ( imsad ) .
the montecarlo simulations were performed by injecting samples of 30 sources at fixed flux and intrinsic angular size .
we sampled peak fluxes between @xmath66 and @xmath67 and intrinsic angular sizes ( fwhm major axis ) between @xmath68 and @xmath69 .
intrinsic sizes were convolved with the synthesized beam ( @xmath70 for mosaic fld20to25 ) before injecting the source in the residual map . + the comparison between the input parameters and the ones provided by imsad permitted an estimate of the internal accuracy of the parameters as a function of source flux and intrinsic angular size .
in particular we could test the accuracy of flux densities , positions and angular sizes and estimate both the efficiency and the accuracy of the deconvolution algorithm .
the flux density and fitted angular size errors for point sources are shown in in figs .
[ fig - spsim ] and [ fig - majminsim ] where we plot the ratio of the parameter value found by imsad ( output ) over the injected one ( input ) , as a function of the signal - to - noise ratio .
+ we notice that mean values very far from 1 could indicate the presence of systematic effects in the parameter measure .
the presence of such systematic effects is clearly present for peak flux densities in the faintest bins ( see fig .
[ fig - spsim ] ) .
this is the expected effect of the noise on the catalogue completeness at the extraction threshold . due to its gaussian distribution whenever an injected source falls on a noise dip , either the source flux is underestimated or the source goes undetected .
this produces an incompleteness in the faintest bins . as a consequence , the measured fluxes are biased toward higher values in the incomplete bins , beacause only sources that fall on noise peaks have been detected and measured .
we notice that the mean values found for @xmath71 are in good agreement with the ones expected taking into account such an effect ( see dashed line ) .
it is worth pointing out that our catalogue is only slightly affected by this effect because the detection threshold ( @xmath72 ) is much lower than the @xmath8-threshold chosen for the catalogue ( indicated by the vertical solid line in fig .
[ fig - spsim ] ) : at @xmath73 we expect flux over - estimations @xmath74 .
+ some systematic effects appear to be present also for the source size at @xmath66 : the major and minor axes tend to be respectively under- and over - estimated ( see fig . [ fig - majminsim ] ) .
such effects disappear at @xmath34 ( atesp cut - off ) .
for both the flux densities and the source axes , the rms values measured are in very good agreement with the ones proposed by condon ( @xcite ) for elliptical gaussian fitting procedures ( for details see appendix [ sec - appa ] ) : @xmath75 where we have applied eqs .
( 21 ) and ( 41 ) in condon ( @xcite ) to the case of atesp point sources ( see dotted lines in figs .
[ fig - spsim ] and [ fig - majminsim ] ) .
+ the fact that a source is extended does not affect the internal accuracy of the fitting algorithm for both the peak flux density and the source axes . in other words the errors quoted for point sources apply to extended sources as well .
+ however , this is not true for the deconvolution algorithm .
the errors for the deconvolved source axes depend on both the source flux and intrinsic angular size . the higher the flux and the larger the source , the smaller the error .
in particular , at 1 mjy ( @xmath76 ) the errors are in the range @xmath77
@xmath78 for angular sizes in the range @xmath79 @xmath69 . for fluxes @xmath80 the errors are always @xmath81 .
deconvolved angular sizes are unreliable for very faint sources ( @xmath82 ) , where only a very small fraction of sources can be deconvolved .
the deconvolution efficiency increases with the source flux . in particular
, the fraction of deconvolved sources with intrinsic dimension @xmath83 never reaches @xmath84 : it goes from @xmath85 at the lowest fluxes , to @xmath86 at 1 mjy , to @xmath87 at the highest fluxes .
we therefore can assume that @xmath68 is a critical value for deconvolution at the atesp resolution , and that atesp sources with intrinsic sizes @xmath83 are to be considered unresolved .
the positional accuracy for point sources is shown in fig .
[ fig - possim ] , where we plot the difference ( @xmath88 and @xmath89 ) between the position found by imsad ( output ) and the injected one ( input ) , as a function of flux .
no systematic effects are present and the rms values are in agreement with the ones expected for point sources ( condon @xcite , for details see appendix [ sec - appa ] ) : @xmath90 where we have assumed @xmath91 and @xmath92 , the average synthesized beam values of the atesp survey ( see dotted lines in fig . [ fig - possim ] ) .
the positional accuracy of atesp sources is therefore @xmath93 at the limit of the survey ( @xmath8 ) , decreasing to @xmath94 at @xmath95 ( @xmath96 mjy ) and to @xmath97 at 50@xmath22 .
two systematic effects are to be taken into account when dealing with atesp flux densities , the clean bias and the bandwidth smearing effect .
clean bias has been extensively discussed in paper i of this series ( see also appendix [ sec - appb ] at the end of this paper ) .
it is responsible for flux density under - estimations of the order of @xmath98 at the lowest flux levels ( @xmath99 ) and gradually disappears going to higher fluxes ( no effect for @xmath100 ) .
+ the effect of bandwidth smearing is well - known .
it reduces the peak flux density of a source , correspondingly increasing the source size in radial direction .
integrated flux densities are therefore not affected .
+ the bandwidth smearing effect increases with the angular distance ( @xmath101 ) from the the pointing center of phase and depends on the passband width ( @xmath102 ) , the observing frequency ( @xmath103 ) and the synthesized beam fwhm width ( @xmath104 ) .
the particular functional form that describes the bandwidth smearing is determined by the beam and the passband shapes .
it can be demonstrated , though , that the results obtained are not critically dependent on the particular functional form adopted ( e.g. bridle & schwab @xcite ) . + in the simplest case of gaussian beam and passband shapes , the bandwidth smearing effect can be described by the equation ( see eq . 12 in condon et al .
@xcite ) : @xmath105 where the @xmath106 ratio represents the attenuation on peak flux densities by respect to the unsmeared ( @xmath107 ) source peak value . +
we have estimated the actual smearing radial attenuation on atesp single fields , by measuring @xmath16 for a strong source ( atesp j233758 - 401025 ) detected in eight contiguous atesp fields ( corrected for the primary beam attenuation ) at increasing distance from the field center ( full circles in fig .
[ fig - smear1 ] , top panel ) .
the data were then fitted using eq .
[ eq - sm ] by setting @xmath108 ghz and @xmath109 ( from ( @xmath110 ) , as for the atesp data . the best fit ( dashed line in fig .
[ fig - smear1 ] , top panel ) gives @xmath111 mjy and an effective bandwidth @xmath112 mhz ( in very good agreement with the nominal channel width @xmath113 mhz ( see sect .
5.2 in paper i ) .
as expected , the measured integrated flux density ( stars in the same plot ) remains constant .
+ the mosaicing technique consists in a weighted linear combination of all the single fields in a larger mosaiced image ( see eq . 1 in paper i ) .
this means that , given single fields of size @xmath114 pixels , source flux measures at distances as large as @xmath115 from field centers are still used to produce the final mosaic ( even if with small weights ) .
as a consequence , the radial dependence of bandwidth smearing tends to cancel out .
+ for instance , since atesp pointings are organized in a @xmath116 spacing rectangular grid , a source located at the center of phase of one field ( @xmath107 ) is measured also at @xmath117 in the 4 contiguous e , w , s and n fields and at @xmath118 in the other 4 _ diagonally _ contiguous fields . using eq . ( 1 ) of paper i , we can estimate a 4% smearing attenuation for the mosaic peak flux of that source . in the same way we can estimate indicative values for mosaic smearing attenuations as a function of @xmath119 , defined as the distance to the closest field center
( see dotted line in fig .
[ fig - smear2 ] ) .
we notice that actual attenuations vary from source to source depending on the actual position of the source in the mosaic .
+ from fig .
[ fig - smear2 ] we can see that at small @xmath119 mosaic smearing is much worse than single field s one ( indicated by the solid line ) . the discrepancy becomes smaller going to larger distances and disappears at @xmath120 , which represents the maximum distance to the closest field center for atesp sources .
this maximum @xmath119 value gives an upper limit of @xmath121 to mosaic smearing attenuations .
+ the expected mosaic attenuations have been compared to the ones obtained directly estimating the smearing from the source catalogue . as already noticed ( sect .
[ sec - deconv ] ) , a ratio @xmath122 , is purely determined , in case of point sources and in absence of flux measurement errors , by the bandwidth smearing effect , which systematically attenuates the source peak flux , leaving the integrated flux unchanged .
+ we have then considered all the unresolved ( @xmath123 ) atesp sources with @xmath124 mjy and we have plotted the average values of the @xmath125 ratio in different distance intervals ( full dots in fig .
[ fig - smear2 ] ) .
the 2 mjy threshold ( @xmath126 ) was chosen in order to find a compromise between statistics and flux measure accuracy .
the average flux ratios plotted are in very good agreement with the expected ones , especially when considering that the most reliable measures are the intermediate distance ones , where a larger number of sources can be summed . in general we can conclude that on average smearing attenuations are @xmath24 and do not depend on the actual position of the source in the mosaics .
this result also confirms the @xmath127 estimate drawn from fig .
[ fig - stspratio ] .
we finally point out that smearing will affect to some extent also source sizes and source coordinates . to determine the quality of atesp source parameters we have made comparisons with external data .
unfortunately in the region covered by the atesp , there are not many 1.4 ghz data available .
the only existing 1.4 ghz radio survey is the nvss ( condon et al .
@xcite ) , which covers about half of the region covered by the atesp survey ( @xmath128 ) with a point source detection limit @xmath129 mjy .
+ the nvss has a poor spatial resolution ( @xmath60 fwhm beam width ) compared to atesp and this introduces large uncertainties in the comparison , especially for astrometry .
to test the positional accuracy we have therefore used data at other frequencies as well . in particular we have used vlbi sources extracted from the list of the standard calibrators at the atca and the catalogue of pmn compact sources with measured atca positions ( wright et al .
@xcite ) . in order to estimate the quality of the atesp flux densities we have compared atesp with nvss . to minimize the uncertainties due to the much poorer nvss resolution we should in principle consider only point - like atesp sources
nevertheless , in order to increase the statistics at high fluxes ( @xmath130 mjy ) , we decided to include extended atesp sources as well .
another source of uncertainty in the comparison is due to the fact that in the nvss distinct sources closer together than @xmath131 will be only marginally separated . to avoid this problem
we have restricted the comparison to bright ( @xmath132 mjy ) _ isolated _ atesp sources : we have discarded all multi - component sources ( as defined in sect .
[ sec - mult ] ) and all single - component sources whose nearest neighbor is at a distance @xmath133 ( as in the comparison between the first and the nvss by white et al .
we have not used isolated atesp sources fainter than 1 mjy because we have noticed that there are several cases where nvss point sources are resolved in two distinct objects in the atesp images , only one being listed in the atesp catalogue ( i.e. the other one has @xmath134 ) .
+ in fig .
[ fig - nvssatflux ] we have plotted the nvss against the atesp flux density for the 443 @xmath135 mjy atesp sources identified ( sources within the @xmath33 confidence circle in fig . [ fig - nvssatpos ] ) .
we have used integrated fluxes for the sources which appear extended at the atesp resolution ( full circles ) and peak fluxes for the unresolved ones ( dots ) .
also indicated are the @xmath136 confidence limits ( dashed curves ) , drawn taking into account both the nvss and the atesp errors in the flux measure . in drawing the upper line we have also taken into account an average correction for the systematic under - estimation of atesp fluxes due to clean bias and bandwidth smearing ( see sect . [ sec - corr ] ) .
the provided nvss fluxes are already corrected for any systematic effect .
+ the scatter plot shows that at high fluxes the atesp flux scale agrees with the nvss one within a few percents ( @xmath137 ) .
this gives an upper limit to calibration errors and/or resolution effects at high fluxes .
going to fainter fluxes the discrepancies between the atesp and the nvss fluxes become larger , reaching deviations as high as a factor of 2 at the faintest levels .
atesp fluxes tend to be lower than nvss fluxes .
this could be , at least partly , due to resolution effects .
such effects can be estimated from the comparison between nvss and atesp sources itself and from theoretical considerations . assuming the source integral angular size distribution provided by windhorst et al .
( @xcite ) we have that at the nvss limit ( @xmath138 mjy ) about 40% of the sources can be resolved by the atesp synthesized beam ( intrinsic angular sizes @xmath139 ) .
this fraction goes up to @xmath87 at @xmath140 mjy .
+ on the other hand , we point out that close to the nvss catalogue cut - off , incompleteness could affect nvss fluxes . for instance we have noticed that below 5 mjy , there are several cases where the flux given in the nvss catalogue is overestimated with respect to the one measured in the nvss image ( even taking into account the applied corrections ) .
the region covered by the atesp survey contains two vlbi sources from the atca calibrator catalogue : 2227 - 399 and 0104 - 408 .
these sources were not used to calibrate our data and therefore provide an independent check of our atca positions .
the offset between atesp and vlbi positions ( atesp vlbi ) for the first and the second source respectively are : @xmath141 and @xmath142 ; @xmath143 and @xmath144 .
such offsets indicate that the uncertainty in the astrometry should be within a fraction of arcsec .
obviously , we can not exclude the presence of systematic effects , but an analysis of the atesp
nvss positional offsets gives @xmath145 and @xmath146 ( see fig .
[ fig - nvssatpos ] ) .
+ a more precise comparison could be obtained by using the pmn sources with atca position measurements available .
unfortunately the number of such pmn sources in the region covered by the atesp survey is very small : we found only 12 identifications . using the 4.8 ghz positions for the pmn sources
, we derived ( atesp atpmn ) @xmath145 and @xmath147 .
+ we can conclude that all comparisons give consistent results and that the astrometry is accurate within a small fraction of an arcsec . also systematic offsets ,
if present , should be very small .
we acknowledge r. fanti for reading and commenting on an earlier version of this manuscript .
+ the authors aknowledge the italian ministry for university and scientific research ( murst ) for partial financial support ( grant cofin 98 - 02 - 32 ) .
this project was undertaken under the csiro / cnr collaboration programme .
the australia telescope is funded by the commonwealth of australia for operation as a national facility managed by csiro .
as discussed in sect . [ sec - errors ] , internal errors for the atesp source parameters are well described by condon ( @xcite ) equations of error propagation derived for two - dimensional elliptical gaussian fits in presence of gaussian noise . in order to get the total rms error on each parameter , a calibration term should be quadratically added .
using eqs .
( 21 ) and ( 41 ) in condon ( @xcite ) , total percentage errors for flux densities ( @xmath148 ) and fitted axes ( @xmath149 , @xmath150 ) can be calculated from : @xmath151 where @xmath152 is the calibration term and the effective signal - to - noise ratio , @xmath153 , is given by : @xmath154^{\alpha_{\rm m } } \left [ 1 + \left(\frac { \theta_{\rm n}}{\theta_{\rm min}}\right)^2 \right]^{\alpha_{\rm m } } \frac{s_{\rm peak}^2}{\sigma^2}\ ] ] where @xmath22 is the image noise ( @xmath155 @xmath3jy on average for atesp images ) , @xmath156 is the fwhm of the gaussian correlation length of the image noise ( assumed @xmath157 fwhm of the synthesized beam ) and the exponents are : @xmath158 similar equations hold for position rms errors ( condon @xcite , condon et al .
@xcite ) : @xmath159 where @xmath160 and @xmath161 represent the rms lengths of the major and minor axes respectively .
+ calibration terms are in general estimated from comparison with consistent external data of better accuracy than the one tested .
unfortunately there are no such data available in the region of sky covered by the atesp survey .
nevertheless , from our typical flux and phase calibration errors , we estimate calibration terms of about @xmath162 for both flux densities and source sizes . + as a caveat we remind ( see discussion in paper i ) that the 500 m baseline cutoff applied to our data makes the atesp survey progressively insensitive to sources larger than @xmath163 : assuming a gaussian shape , only @xmath87 of the flux for a @xmath163 large source would appear in the atesp images .
it is important to have this in mind when dealing with flux densities and source sizes of the largest atesp sources .
+ right ascension and declination calibration terms have been estimated from the astrometry results reported in sect .
[ sec - astr ] .
as already discussed , the atesp astrometry can be considered accurate within a small fraction of an arcsec , even though the scarcity of ( accurate ) external data available in the atesp region makes it difficult to quantify this statement . nevertheless from the rms dispersion of
the median offsets found between atesp and the external comparison samples ( see sect . [ sec - astr ] ) we can tentatively estimate @xmath164 and @xmath165 .
+ it can be easily demonstrated that the master equations ( [ eq - condon ] ) , ( [ eq - condonposa ] ) and ( [ eq - condonposb ] ) reduce to eqs .
( [ eq - sperr ] ) @xmath166 ( [ eq - decerr ] ) in sect .
[ sec - errors ] ( where the calibration term is neglected ) in the case of atesp point sources ( @xmath167 , p.a.@xmath168 ) , assuming @xmath169 ( or @xmath170 ) .
+ for a complete and detailed discussion of the error master equations of source parameters obtained through elliptical gaussian fits we refer to condon ( @xcite ) and condon et al .
( @xcite ) .
as already discussed in sect . [ sec - corr ] , two systematic effects are to be taken into account when dealing with atesp flux densities , the clean bias and the bandwidth smearing effect .
+ the flux densities reported in the atesp source catalogue are not corrected for such systematic effects .
the corrected flux densities ( @xmath171 ) can be computed as follows : @xmath172}\ ] ] where @xmath173 is the flux actually measured in the atesp images ( the one reported in the source catalogue ) .
the parameter @xmath174 represents the smearing correction .
it is set equal to 1 ( i.e. no correction ) when the equation is applied to integrated flux densities and @xmath175 when dealing with peak flux densities . from the analysis
reported in sect .
[ sec - corr ] we suggest to set @xmath176 ( @xmath127 smearing effect ) . + the clean bias correction is taken into account by the term in the square brackets . as discussed in paper i , sect . 5.3 , the importance of the clean bias effect varies from mosaic to mosaic depending on the average number of clean components ( cc s ) . in particular we derived the values for the parameters @xmath177 and @xmath178 in three different mosaics representing the case of low ( fld34to40 , 1616 cc s ) , intermediate ( fld44to50 , 2377 cc s ) and high ( fld69to75 , 3119 cc s ) average number of cc s ( see table 4 of paper i ) . + in correcting the source fluxes for the clean bias , we suggest to set @xmath179 whenever the mosaic average number of cc s is @xmath180 ( low cc s case ) ;
@xmath181 whenever the mosaic cc s average number is between 2000 and 3000 ( intermediate cc s case ) ;
@xmath182 whenever the mosaic cc s average number exceeds 3000 ( high cc s case ) .
the average number of clean components for each mosaic is reported in table [ tab - mospar ] . in order to trace back the sources to the original mosaics , table [ tab - mospar ] lists also the right ascension range covered by each mosaic ( indicated by the r.a .
of the first and the last source in that mosaic ) .
+ the clean bias is a function of the source signal - to - noise ratio @xmath183 .
since the noise level is fairly uniform within each mosaic , it is possible to assume @xmath22 equal to the mosaic average noise value ( @xmath184 in table [ tab - mospar ] , we refer to paper i for details on mosaic noise analysis and average noise value definition ) .
lrcc & & & + & & & + fld01to06 & 2033 & 78.7 & @xmath185 - @xmath186 + fld05to11 & 1796 & 77.8 & @xmath187 - @xmath188 + fld10to15 & 3104 & 88.1 & @xmath189 - @xmath190 + fld20to25 & 2823 & 83.0 & @xmath191 - @xmath192 + fld24to30 & 2716 & 82.8 & @xmath193 - @xmath194 + fld29to35 & 2044 & 79.2 & @xmath195 - @xmath196 + fld34to40 & 1616 & 76.3 & @xmath197 - @xmath198 + fld39to45 & 2535 & 81.2 & @xmath199 - @xmath200 + fld44to50 & 2377 & 78.0 & @xmath201 - @xmath202 + fld49to55 & 2168 & 78.6 & @xmath203 - @xmath204 + fld54to60 & 2504 & 77.3 & @xmath205 - @xmath206 + fld59to65 & 2447 & 79.4 & @xmath207 - @xmath208 + fld64to70 & 1899 & 75.1 & @xmath209 - @xmath210 + fld69to75 & 3119 & 81.9 & @xmath211 - @xmath212 + fld74to80 & 2558 & 77.1 & @xmath213 - @xmath214 + fld79to84 & 1522 & 68.9 & @xmath215 - @xmath216 + & & & + baars j.w.m . , genzel r. , pauliny - toth i.i.k . ,
witzel a. , 1977 , a&as 61 , 99 bridle a.h . ,
schwab f.r . , 1989 .
in : perley et al .
synthesis imaging in radio astronomy , asp conf .
6 , 247 condon j.j . , 1997 ,
pasp 109 , 166 condon j.j .
, 1998 . in : b.j .
mclean et al .
179 , new horizons from multi - wavelength sky surveys , p. 19
condon j.j . , cotton w.d .
, greiser e.w . , et al . , 1998 , aj 115 , 1693 prandoni i. , gregorini l. , parma p. , et al . ,
2000 , a&as 146 , 31 ( paper i ) sault r.j .
, killeen n. , 1995 , miriad users guide vettolani g. , zucca e. , zamorani g. , et al . , 1997 , a&a 325 , 954 white r.l . , becker r.h .
, helfand d.j . ,
gregg m.d . , 1997 ,
apj 475 , 479 windhorst r.a .
, mathis d. , neuschaefer l. , 1990 . in : kron r.g .
evolution of the universe of galaxies , asp conf .
10 , 389 wright a.e . , tasker n. , mcconnell d. , et al . , 1997 , http://www.parkes.atnf.csiro.au/databases/surveys/
|
this paper is part of a series reporting the results of the _ australia telescope eso slice project _ ( atesp ) radio survey obtained at 1400 mhz with the _ australia telescope compact array _ ( atca ) over the region covered by the _
eso slice project _
( esp ) galaxy redshift survey .
the survey consists of 16 radio mosaics with @xmath0 resolution and uniform sensitivity ( @xmath1 noise level @xmath279 @xmath3jy ) over the whole area of the esp redshift survey ( @xmath4 sq .
degrees at @xmath5 ) .
+ here we present the catalogue derived from the atesp survey .
we detected @xmath6 distinct radio sources down to a flux density limit of @xmath7 mjy ( @xmath8 ) , 1402 being sub - mjy sources .
we describe in detail the procedure followed for the source extraction and parameterization .
the internal accuracy of the source parameters was tested with monte carlo simulations and possible systematic effects ( e.g. bandwidth smearing ) have been quantified .
|
the existence of active galactic nuclei has long been taken as evidence for the existence of massive black holes in the centres of some galaxies ( lynden - bell 1969 ) . however , it is only relatively recently that high spatial resolution studies of the kinematics of galactic nuclei have revealed that essentially all galaxies harbour large central masses [ see ho ( 1999 ) for a review of the evidence ] .
the existence of these observations also means that there are now enough data to study the demographics of massive black holes , in order to seek clues to their origins .
the first significant discovery in this regard is that there is a correlation between the mass of the black hole , @xmath0 , and the mass of the host galaxy s spheroidal component , @xmath1 .
although there is a variety of possible biasses in measuring this correlation , it seems broadly to be the case that there is a linear relationship , such that @xmath2 ( magorrian et al .
1998 ) .
although this correlation is reasonably strong , there is still considerable scatter in the relation , such that there is more than a factor of ten variation in the inferred value of @xmath0 for galaxies of given spheroid mass ( magorrian et al .
1998 ) . some of this scatter can probably be attributed to the uncertainties in calculating black hole masses from relatively poor kinematic data and simplified dynamical models ( van der marel 1997 ) .
however , there are also astrophysical reasons why one might expect significant dispersion in this relation . for example , consider the simplest possible scenario in which galaxies form and evolve in near isolation . if the central black holes in these galaxies accrete mass fairly steadily from their hosts , then the mass of a black hole simply reflects the age of its host . under the currently - favoured hierarchical paradigm for galaxy formation , in which larger galaxies are formed from the merging of smaller galaxies ( white & rees 1978 ) , the simple linear correlation between galaxy mass and black hole mass is readily explained .
each time two galaxies merge to form a larger system , their black holes rapidly spiral to the centre of the new galaxy due to dynamical friction .
the black holes then merge , creating a proportionately - larger black hole .
however , a galaxy formed by this process of repeated mergers can not be characterized by a single age , so the above explanation for the scatter in black hole masses must be modified somewhat .
one measure of such a galaxy s age is the time since it last underwent a major merger , and kauffmann & haehnelt ( 2000 ) have shown that this timescale is a key factor in explaining the scatter in black hole masses . if the last merger happened long ago , then it will have occurred between relatively unevolved galaxies in which there would have been a large amount of cold gas .
if the black hole accretes some fixed fraction of this gas , then galaxies in which the last merger occurred longer ago will contain more massive black holes .
this picture , in which black holes acquire much of their mass through accretion of material from their host galaxies , seems quite credible .
however , it is not the only possible scenario .
stiavelli ( 1998 ) has argued that galaxies with essentially identical properties could be formed around pre - existing massive black holes . in this case
, the spread in black hole masses would simply reflect the stochastic nature of whatever physics was responsible for the formation of the primordial black holes . in this letter
, we investigate whether we can attribute the observed scatter in black hole masses to an astrophysical cause , and hence whether we can discriminate between the above scenarios .
specifically , we investigate whether the masses of black holes correlate with the ages of their host galaxies as determined by stellar absorption line diagnostics .
there are now several compilations of central black hole mass estimates in nearby galaxies ( e.g. ho 1999 ) .
the difficulty in using such compilations for quantitative studies is that they contain data obtained using a variety of heterogeneous techniques .
thus , not only are there likely to be systematic errors in the derived masses , but the nature of these errors will vary within the compilation . to minimize the impact of such uncertainties , we have chosen to consider a sample containing only objects from a single study where the analysis has been performed in a consistent fashion . although it could not be argued that such a sample is necessarily free of systematic errors , one might reasonably hope that the consistent analysis should produce relatively consistent results . for example , if one galaxy is found to have a more massive black hole than another optically - identical galaxy within a single sample , then it is likely that the two systems are intrinsically different ; one can not say the same if one compares two galaxies from different samples that have been analyzed using different techniques .
the largest consistently - analyzed sample available is that published by magorrian et al .
their study of galaxies stellar kinematics presented estimates for the masses of the central black holes and spheroidal components of 32 galaxies .
the only exceptional galaxy in this dataset is ngc 1399 .
the luminosity distribution of this galaxy has a very large diffuse envelope , making it one of the most extreme known examples of a cd galaxy ( schombert 1986 ) .
it is therefore almost impossible to disentangle the mass of this extensive galaxy from the mass of the cluster that surrounds it .
in fact , it is interesting to note that this galaxy lies well above magorrian et al.s ( 1998 ) mean relation between @xmath0 and @xmath1 .
however , if one adds to @xmath1 an estimate for the total mass in the cd envelope ( which was excluded from the original mass estimate ) , it is straightforward to place ngc 1399 right on the mean relation .
unfortunately , it is difficult to justify this ad hoc correction when one is trying to carry out a consistent analysis . given the undesirability of such a posteriori manipulation , and the fact that such extreme cd systems are likely to have evolved by very different mechanisms from regular ellipticals , ngc 1399 has been excluded from the sample , leaving a dataset of 31 galaxies .
we should , however , note that the presence of this galaxy in the sample makes no difference to the statistical significance of the conclusions presented below . as with the black hole
mass estimates , it is important that the galaxy age estimates be made in as consistent a manner as possible .
we have therefore used values from the recent catalogue of terlevich & forbes ( 2000 ) , which is compiled from a relatively homogeneous dataset of high - quality absorption line measurements for galaxies ( e.g. h@xmath3 , h@xmath4 , [ mgfe ] ) . using the stellar population model of worthey ( 1994 ) ,
these line indices can be used to break the age / metallicity degeneracy , thus giving separate age and metallicity estimates . for a few galaxies not in the terlevich & forbes ( 2000 ) catalogue ,
the same line indices have been measured by trager et al .
( 1998 ) using data of comparable quality . combining these measurements with the terlevich & forbes dataset , one can obtain consistent age estimates for 23 of the galaxies in the current sample .
the line index measurements come from the galaxies central regions and are luminosity weighted .
they are therefore dominated by the last major burst of star formation .
thus , the age estimate probably reflects the time since the galaxy s last major merger event , which will have induced significant amounts of star formation [ see also forbes , ponman & brown ( 1998 ) ] .
figure 1 shows the fraction of each galaxy s spheroidal component mass that resides in its central black hole as a function of the age inferred for the galaxy s stellar population .
there is clearly a large amount of scatter in this plot ; indeed , since there are sizeable uncertainties in both the black hole mass determinations and the age estimates , one could not expect to see a tight correlation .
however , there is a definite trend in the sense that older galaxies of a given total mass contain more massive black holes : the four youngest galaxies all have black holes whose masses lie below the mean of @xmath5 , while three of the four oldest galaxies lie above this line .
more quantitatively , a spearman rank test rejects the possibility that @xmath6 and @xmath7 are uncorrelated at @xmath8 confidence .
the robust nature of a rank test means that the significance of this correlation does not hang on the outlying points the same confidence level is reached if , for example , ngc 7332 is excluded from the analysis .
although there does appear to be a significant correlation between measured black hole mass and galaxy age estimate , it is not necessarily astrophysical in origin .
we must first consider the possibility that it arises from some systematic error in the analysis .
however , the kinematic data from which the black hole masses were inferred are completely independent from the line index data that provide the age estimates .
since the line index data were not selected with this project in mind , and the black hole mass estimates played no role in the choice of sample , the selection process can not have induced the correlation that is seen in fig . 1 .
further , the independent nature of the data sets used to measure the two ordinates means that there can be nothing in this analysis that might preferentially over - estimate the black hole masses in old galaxies , or underestimate the masses in young systems .
it should also be borne in mind that the absolute calibrations of the black hole masses and galaxy ages are significantly uncertain . in the case of the absorption line indices , for example ,
the age estimates are derived from spectral synthesis modelling , which remains a somewhat uncertain process , so the absolute values of the ages of two galaxies may be quite ill - determined .
however , the fact that one is older than the other can be determined relatively reliably by this modelling process , so the approximate ordering of galaxy ages can be determined quite robustly .
since the spearman rank test described above depends only on this ordering , the statistical significance of the correlation is not dependent on the details of the adopted calibration .
it would thus appear that there is an underlying astrophysical correlation between the fraction of a galaxy s mass in its central black hole and the age of its most recently formed stellar component .
hence , in addition to the established correlation between black hole mass , @xmath0 and galaxy mass , @xmath1 , there seems to be a `` second parameter '' correlation with the age of the youngest stellar component . at
any given value of @xmath1 , different age galaxies will have different values of @xmath0 , so this secondary correlation must go some way toward explaining the scatter in the primary relation .
we have sought to quantify the contribution of this second parameter to the scatter in the relation between @xmath1 and @xmath0 by calculating @xmath9 this process corrects the mass ratio for the effects of age by subtracting the simplest possible linear fit to the correlation in fig . 1 . as one would expect , this correction reduces the scatter in the relation : for the data in this sample , the dispersion in @xmath10 is 0.42 dex while that in @xmath11 is only 0.31 dex .
clearly , even the corrected mass ratio still contains considerable scatter .
however , given the large uncertainties in the individual black hole mass and galaxy age determinations , it would be very surprising if the dispersion were reduced below a factor of two ( @xmath12 dex ) .
the simplest explanation for the existence of the second parameter correlation is that a single physical process couples the growth of the central black hole to the triggering of star formation in a galaxy .
as outlined in the introduction , the hierachical picture of galaxy and black hole evolution described by kauffmann & haehnelt ( 2000 ) suggests that galaxy mergers lie behind both processes .
where the last major merger occurred long ago , it will have taken place in a gas - rich environment that will provide ample fuel to augment the mass of the black hole .
since the last major episode of star formation will also be triggered in the merger , such galaxies will contain old stellar populations and massive black holes .
conversely , galaxies formed in more recent mergers will contain under - massive black holes and younger stellar populations .
although the correlation between black hole mass and galaxy age is predicted by the hierarchical merging models , it should be borne in mind that such a correlation is a fairly generic prediction of any model in which the black hole mass grows over time .
even if galaxies form monolithically , those that form first and hence contain the oldest stellar populations will have had time to grow the largest black holes .
the models that do not fit easily with this correlation are those in which the black holes and stellar components form at entirely different times it would be hard to explain the observed correlation if , for example , the central black holes were entirely primordial .
the study of black hole demographics is maturing rapidly , and , as we hope we have shown , it is already possible to detect phenomena beyond the basic relation between @xmath0 and @xmath1 . in the near future , larger sets of both the kinematic and line - strength data will become available , and more sophisticated modelling techniques will be developed to refine the estimates for black hole masses and galaxy ages . with these tools , it will become possible to address subtler questions , such as whether a galaxy s environment plays a significant role in its black hole growth rate .
such analyses will provide key tests for theories of black hole formation within the broader context of galaxy evolution .
it is a pleasure to thank the referee , bob mann , for a range of helpful suggestions .
99 forbes , d.a . , ponman , t.j . ,
brown , r.j.n .
, 1998 , apj , 508 , l43 ho , l. , 1999 , in `` observational evidence for black holes in the universe , '' ed .
s. k. chakrabarti ( dordrecht : kluwer ) kauffmann , g. , & haehnelt , 2000 , mnras , submitted white , s.d.m . , & rees , m.j . , 1978 ,
mnras , 183 , 341 lynden - bell , d , 1969 , nat . , 223 , 690 magorrian , j. , et al . , 1998 , aj , 115 , 2285 terlevich , a.i . ,
forbes , d.a . , 2000 , in preparation trager , s.c . ,
worthey , g. , faber , s.m . ,
burstein , d. , gonzalez , j.j . , 1998 ,
apjs , 116 , 1 van der marel , r.p . , 1997 , in galaxy interactions at low and high redshift , proc .
iau symposium 186 ( kluwer , dordrecht ) worthey , g. , 1994 , apjs , 95 , 107
|
we present evidence that there is a significant correlation between the fraction of a galaxy s mass that lies in its central black hole and the age of the galaxy s stellar population . since the absorption line indices that are used to estimate the age are luminosity weighted
, they essentially measure the time since the last significant episode of star formation in the galaxy .
the existence of this correlation is consistent with several theories of galaxy formation , including the currently - favoured hierarchical picture of galaxy evolution , which predicts just such a relation between black hole mass and the time since the last burst of merger - induced star formation .
it is not consistent with models in which the massive black hole is primordial , and hence uncoupled from the stellar properties of the galaxy .
[ firstpage ] galaxies : formation , nuclei
quasars : general black hole physics
|
let @xmath0 be a @xmath1-dimensional _ lattice polytope _ in @xmath2 ; that is , @xmath0 is the convex hull of finitely many points in @xmath3 . further , let @xmath4 be a triangulation of @xmath0 .
we call @xmath4 _ dense _ if the points @xmath5 are precisely the vertices of @xmath4 . moreover , we call @xmath4 _ foldable _ if its dual graph is bipartite ; equivalently , the vertices of the @xmath6-skeleton considered as an abstract graph is @xmath7-colorable @xcite . in this case we may distinguish between `` black '' and `` white '' @xmath1-dimensional cells of @xmath4 . the _ signature _ @xmath8 is defined as the absolute value of the difference of the numbers of black and of white @xmath1-dimensional cells of @xmath4 that have odd normalized volume . the _ normalized volume _ of @xmath0 equals @xmath9 times the euclidean volume of @xmath0 . as @xmath0 is a lattice polytope
the normalized volume is an integer .
soprunova and sottile @xcite prove that @xmath8 is a lower bound for the number of real roots of so - called `` wronski systems '' of @xmath1 polynomials over the reals in @xmath1 unknowns associated with @xmath0 and @xmath4 . for @xmath10 that result amounts to the basic fact that every univariate real polynomial of odd degree has at least one real zero . in general , the signature @xmath8 can be interpreted as the topological degree of the map which `` folds '' the foldable triangulation @xmath4 onto the standard simplex of the same dimension ; see @xcite and the references given there . in various cases the soprunova sottile bound is sharp for generic coefficients . here
we are concerned with the planar case @xmath11 , where we provide a new lower bound for the signature , and then an explicit example of a wronski system for which the bound is sharp . in the literature foldable abstract simplicial complexes
are also called `` balanced '' ; for instance , see @xcite .
let @xmath12 and @xmath13 be lattice points in @xmath14 .
we say that the line segment @xmath15=\operatorname{conv}\{p , q\}$ ] is _ of type @xmath16 _ if the first coordinate of @xmath17 is odd and the other one is even . similarly , the segment is _ of type @xmath18 _ if the second coordinate is odd and the other one is even . for a segment _ of type @xmath19 _
both coordinates are odd .
the key to our main result is the following observation .
[ lemma1 ] let @xmath20 be a lattice triangle of odd normalized area in the plane
. then @xmath20 has precisely one edge of type @xmath16 , one of type @xmath18 and one of type @xmath19 .
up to a translation we can assume that the vertices of @xmath20 are @xmath21 , @xmath22 and @xmath23 with @xmath24 .
then the normalized area of @xmath20 is given by the absolute value of the determinant @xmath25 and our precondition says that this is odd .
it follows that the parities of the two products @xmath26 and @xmath27 are distinct .
after possibly exchanging the vertices @xmath22 and @xmath23 we may assume that @xmath26 is odd , and thus both @xmath28 and @xmath1 are odd .
if both @xmath29 and @xmath30 are even then the edge @xmath31 $ ] is of type @xmath16 , the edge @xmath32 $ ] is of type @xmath18 , and the third edge @xmath33 $ ] is of type @xmath19 .
if , however , @xmath29 is odd and @xmath30 is even , then @xmath31 $ ] is of type @xmath19 , @xmath32 $ ] is of type @xmath18 and @xmath33 $ ] is of type @xmath16 .
the situation is similar if @xmath29 is even and @xmath30 is odd .
the preceding lemma admits the following generalization .
let @xmath20 be an arbitrary lattice triangle .
then the normalized area of @xmath20 is congruent modulo @xmath34 to the number of edges of type @xmath35 , where @xmath36 is any type .
we omit the proof , which is similar to the above .
now let @xmath11 and let @xmath0 be a lattice polygon with a dense and foldable triangulation @xmath4 . for @xmath37 foldable triangulations
can also be characterized by the fact that all interior vertices of its graph have even degree .
( indeed , the interesting case is @xmath11 , as for @xmath10 all triangulations are foldable . )
it is also special to dimensions @xmath37 that a triangulation is dense if and only if it is unimodular , that is , each triangle has normalized area one .
as @xmath4 is foldable we may distinguish between `` black '' and `` white '' triangles .
we can also assign colors , either black or white , to the boundary edges of a foldable triangulation , according to the color of the unique triangle they are contained in .
thus we are ready for our main result .
( v1 ) at ( 0,1 ) ; ( v2 ) at ( 0,2 ) ; ( v4 ) at ( 1,1 ) ; ( v5 ) at ( 1,2 ) ; ( v6 ) at ( 1,3 ) ; ( v8 ) at ( 2,1 ) ; ( v9 ) at ( 2,2 ) ; ( v10 ) at ( 2,3 ) ; ( v11 ) at ( 2,4 ) ; ( v12 ) at ( 3,0 ) ; ( v13 ) at ( 3,1 ) ; ( v14 ) at ( 3,2 ) ; ( v15 ) at ( 3,3 ) ; ( v16 ) at ( 3,4 ) ; ( v17 ) at ( 3,5 ) ; ( v18 ) at ( 4,0 ) ; ( v19 ) at ( 4,1 ) ; ( v20 ) at ( 4,2 ) ; ( v21 ) at ( 4,3 ) ; ( v22 ) at ( 4,4 ) ; ( v23 ) at ( 5,0 ) ; ( v24 ) at ( 5,1 ) ; ( v25 ) at ( 5,2 ) ; ( v26 ) at ( 5,3 ) ; ( v27 ) at ( 5,4 ) ; = [ draw , thick,-,black ] ( v1 )
( v12 ) cycle ; ( v8 ) ( v4 ) ( v12 ) cycle ; ( v1 ) ( v4 ) ( v5 ) cycle ; ( v2 ) ( v5 ) ( v6 ) cycle ; ( v1 ) ( v2 ) ( v5 ) cycle ; ( v4 ) ( v5 ) ( v8 ) cycle ; ( v5 ) ( v8 ) ( v9 ) cycle ; ( v5 )
( v6 ) ( v11 ) cycle ; ( v5 ) ( v10 ) ( v11 ) cycle ; ( v5 ) ( v9 ) ( v10 ) cycle ; ( v9 ) ( v10 ) ( v14 ) cycle ; ( v10 ) ( v11 ) ( v15 ) cycle ; ( v11 )
( v15 ) ( v16 ) cycle ; ( v8 ) ( v12 ) ( v18 )
cycle ; ( v8 ) ( v9 )
( v14 ) cycle ; ( v8 ) ( v13 ) ( v14 ) cycle ; ( v8 ) ( v18 ) ( v23 ) cycle ; ( v8 ) ( v13 ) ( v23 ) cycle ; ( v10 ) ( v14 ) ( v20 ) cycle ; ( v10 ) ( v15 )
( v20 ) cycle ; ( v16 ) ( v20 ) ( v21 ) cycle ; ( v15 ) ( v16 ) ( v20 ) cycle ; ( v13 ) ( v14 ) ( v19 ) cycle ; ( v14 ) ( v19 ) ( v20 ) cycle ; ( v13 ) ( v19 ) ( v23 ) cycle ; ( v19 )
( v20 ) ( v23 ) cycle ; ( v20 ) ( v23 ) ( v24 ) cycle ; ( v20 ) ( v21 ) ( v24 ) cycle ; ( v21 ) ( v24 )
( v25 ) cycle ; ( v16 ) ( v21 )
( v22 ) cycle ; ( v21 ) ( v22 ) ( v26 ) cycle ; ( v21 ) ( v25 ) ( v26 ) cycle ; ( v22 ) ( v26 ) ( v27 ) cycle ; in v1,v2,v4,v5,v6,v8,v9,v10,v11,v12,v13,v14,v15,v16,v18,v19,v20,v21,v22,v23,v24,v25,v26,v27 ( ) circle ( 2pt ) ; ( x0 ) at ( @xmath38 ) ; ( x3 ) at ( @xmath39 ) ; ( x4 ) at ( @xmath40 ) ; ( x5 ) at ( @xmath41 ) ; ( x6 ) at ( @xmath42 ) ; ( x7 ) at ( @xmath43 ) ; ( y1 ) at ( @xmath44 ) ; ( y2 ) at ( @xmath45 ) ; ( y3 ) at ( @xmath46 ) ; ( y4 ) at ( @xmath47 ) ; ( y5 ) at ( @xmath48 ) ; ( xy0 ) at ( @xmath49 ) ; ( xy1 ) at ( @xmath50 ) ; let @xmath4 be a dense and foldable triangulation of a planar lattice polygon @xmath0 . then the signature @xmath8 equals the absolute value of the difference between the numbers of black and of white edges of type @xmath35 , for any fixed @xmath36 . by lemma [ lemma1 ]
each triangle of @xmath4 has precisely one edge of type @xmath16 , one edge of type @xmath18 , and one edge of type @xmath19 .
fix @xmath35 .
each @xmath35 edge is either an interior edge , in which case it is contained in exactly one black and one white triangle , or it is on the boundary .
conversely , each triangle contains a @xmath35 edge , and so the interior @xmath35 edges induce a perfect matching on the interior triangles of @xmath4 .
their total contribution to the signature is zero .
the claim follows .
this result can be interpreted as a combinatorial version of green s theorem on integrating a potential over a simply connected planar domain by integrating over the boundary .
it also bears some similarity with pick s theorem , which states that a lattice polygon of euclidean area @xmath51 and @xmath52 interior lattice points has exactly @xmath53 lattice points on the boundary ; see beck and robbins @xcite as well as the `` green s theorem style '' proof of pick s theorem by blatter @xcite .
figure [ fig : polygon ] shows a lattice hexagon with a dense and foldable triangulation .
it has 17 black triangles and 16 white ones ; so the signature equals one . at the same time it has three black and two white boundary edges of type @xmath16 , three black and two white boundary edges of type @xmath18 , as well as two black and one white boundary edges of type @xmath19 . for each type
there is a surplus of exactly one black edge .
the following corollary solves a problem which the first author presented at the oberwolfach meeting on geometric and topological combinatorics in 2007 .
the authors gratefully acknowledge the repeated hospitality of the mathematisches forschungsinstitut ; this paper was written during another stay at the institut in 2012 .
let @xmath0 be an axis - parallel lattice rectangle in the plane .
then the signature of any dense and foldable triangulation of @xmath0 vanishes
. there are no @xmath19 edges in the boundary .
our main result entails a general upper bound on the signature for lattice polygons .
let @xmath4 be a dense and foldable triangulation of a planar lattice polygon @xmath0 .
then the signature @xmath8 is bounded from above by @xmath54 where @xmath51 is the euclidean area and @xmath52 is the number of interior lattice points of @xmath0 . by pick s theorem
the number @xmath55 of lattice points on the boundary of @xmath0 equals @xmath56 .
now @xmath55 is also the number of boundary edges of the triangulation @xmath4 , and so there is some type @xmath36 such that the number of @xmath35-edges in the boundary does not exceed @xmath57 . since the signature of @xmath4 can not exceed the number of boundary edges of type @xmath35 the claim follows
there are lattice polygons for which this bound is sharp .
consider the triangle @xmath58 . here
@xmath59 , @xmath60 , and @xmath61 , which equals the signature of the trivial triangulation of @xmath0 .
more generally , the dilates @xmath62 for arbitrary integral @xmath63 are lattice triangles with a triangulation @xmath64 induced by the integral translates of the coordinate axes and the diagonal line @xmath65 .
this triangulation is dense and foldable .
all the boundary edges share the same color , and we have exactly @xmath66 boundary edges of each type .
the signature of @xmath64 equals @xmath66 and this coincides with our bound .
however , the example of axis - parallel rectangles shows that our bound on the signature can be arbitrarily bad in general .
the triangle @xmath67 is shown in figure [ fig : triangle ] below .
in the remainder of this note we will sketch how our results are related to bivariate polynomial systems . omitting most of the technical details we aim at explaining one simple example .
let @xmath0 be a lattice polygon with a dense and foldable triangulation @xmath4 .
as @xmath4 is foldable , the lattice points in @xmath0 can be assigned numbers from the set @xmath68 such that two points receive distinct labels whenever they lie on a common edge of @xmath4 . that is to say , the primal graph of @xmath4 ( which is clearly planar ) is @xmath69-colorable .
a proof follows from heawood s criterion on the @xmath69-colorability of maximal planar graphs since @xmath4 being foldable forces that the dual graph of @xmath4 does not contain any odd cycles @xcite .
in fact , up to relabeling such an assignment , @xmath70 , is unique .
now for any choice of non - zero real numbers @xmath71 the polynomial @xmath72 is a bivariate _
wronski polynomial _ with respect to @xmath4 . since the coefficients @xmath73 are non - zero the newton polygon of such a polynomial is @xmath0 .
a bivariate _ wronski system _ consists of two bivariate wronski polynomials with respect to the same triangulation @xmath4 .
provided that the coefficients are chosen sufficiently generic the two polynomials have just finitely many complex roots in common .
a classical result of kushnirenko says that in this case the number of common complex roots coincides with the normalized area of the polygon @xmath0 ; see @xcite and ( * ? ? ?
3.2 ) . under additional assumptions soprunova and sottile
show that the signature @xmath8 yields a lower bound for the number of common _ real _ roots ; see @xcite and ( * ? ? ?
* thm . 7.13 ) .
let @xmath0 be the lattice triangle @xmath74 with the dense and foldable triangulation @xmath4 shown in figure [ fig : triangle ] .
this figure also defines the labeling @xmath30 of the lattice points .
for instance , @xmath75 and @xmath76 .
choosing @xmath77 and @xmath78 as coefficient vectors we obtain the wronski system @xmath79 of bivariate polynomial equations .
the real conic defined by the first polynomial is a hyperbola , the second one defines an ellipse ; see figure [ fig : curves ] .
the normalized area of @xmath0 is four and so , by kushnirenko s theorem , the polynomial system has exactly four complex solutions ( the coefficients _ are _ sufficiently generic ) .
the signature of @xmath4 equals two and thus , by soprunova and sottile , at least two of the complex solutions must be real ( again all extra conditions are met in this case ) .
the results on wronski systems generalize to arbitrary dimension @xmath1 , in which case we have @xmath1 polynomials in @xmath1 indeterminates .
it is instructive to look at the case @xmath10 .
if @xmath80 is a univariate real polynomial with non - vanishing constant coefficient its @xmath6-dimensional newton polytope is the interval from @xmath81 to the degree of @xmath80 .
its normalized volume is the length of the interval , which equals the degree . if the coefficients of @xmath80 are sufficiently generic there are degree many complex roots . in this sense kushnirenko s theorem
generalizes the fundamental theorem of algebra . to obtain a lower bound for the number of _ real _ roots of @xmath80 , consider the unique dense triangulation of the interval @xmath82 $ ] into unit intervals
this is clearly foldable , and its signature is the parity of the degree .
this means that the result by soprunova and sottile generalizes the well known fact that a univariate real polynomial of odd degree has at least one real root .
|
we give a simple formula for the signature of a foldable triangulation of a lattice polygon in terms of its boundary .
this yields lower bounds on the number of real roots of certain systems of polynomial equations known as `` wronski systems '' .
|
over the last few years , the dissipative particle dynamics ( dpd ) model of complex fluids has received considerable attention .
it has matured from its somewhat arbitrary initial formulation into a model with a solid theoretical basis .
furthermore , it has been applied with considerable success to a large number of computer simulations of complex fluid systems such as colloidal suspensions , polymeric fluids , spinodal decomposition of binary immiscible fluids , and amphiphilic fluids .
dpd also looks promising for simulating multiphase flows and flow in porous media , and is now considered a useful technique alongside the other complex fluid algorithms : molecular dynamics , lattice - gas automata , and techniques based on the lattice - boltzmann equation . no single technique can yet be applied to all situations , and each has different strengths and weaknesses . while molecular dynamics is in principle the most accurate microscopic approach , in practice it is too slow in both its quantum ( car parinello ) and classical forms because of its excessive detail .
discrete mesoscopic methods developed from lattice - gas automata have had some success , but they too have problems , such as lacking galilean invariance .
the traditional approach of continuum fluid dynamics has met with limited success in modeling behavior on the length and time scales characteristic of complex fluids . in this paper we investigate phase separation from both symmetric ( critical ) and off - critical quenches in binary immiscible fluids of generally differing viscosity using the hydrodynamically - correct dpd model .
our motivation is to probe and extend knowledge of the behavior of differing - viscosity fluids , of application , for example , to the action of detergents and the extraction of oil from reservoir rocks .
phase separation in _ equal - viscosity _ binary immiscible fluid systems has been simulated using a variety of techniques , including dpd @xcite ; molecular dynamics @xcite ; monte carlo @xcite ; cell dynamical systems without hydrodynamics @xcite and with oseen tensor hydrodynamics @xcite ; time - dependent ginzburg landau models with @xcite and without hydrodynamics @xcite ; lattice - gas automata @xcite ; and lattice - boltzmann techniques @xcite
. spinodal decomposition of _ differing - viscosity _ immiscible fluids has previously been simulated in two dimensions by a time - dependent ginzburg landau model without hydrodynamics @xcite .
we discuss the effect of the proportion of each phase on the scaling behavior in both two and three spatial dimensions .
we also examine the behavior of equal - viscosity fluids , comparing our results to similar lattice - gas simulations in two dimensions @xcite . after describing our fluid model in the following section ,
we discuss the expected temporal development of the characteristic size of the separating domains in section [ temporal ] .
we then describe our method for calculating the characteristic domain size and its rate of growth in section [ estimating ] .
this is followed in sections [ results ] and [ discussion ] by information on the simulations performed and a discussion of the results , and by some conclusions in section [ conclusions ] . finally , as an appendix to this paper
, we make a few comments on the high - performance computing aspects of this work .
in 1992 , hoogerbrugge and koelman proposed dissipative particle dynamics @xcite as a novel particulate model for the simulation of complex fluid behavior .
dpd was developed in an attempt to capture the best aspects of molecular dynamics and lattice - gas automata .
it avoids the lattice - based problems of lattice - gas automata , yet maintains an elegant simplicity and larger scale that keeps the model much faster than molecular dynamics .
this simplicity also makes dpd highly extensible , such as for including the interactions of complex molecules or modeling flow in an arbitrary number of spatial dimensions .
the key features of the model are that the fluid is grouped into packets , termed `` particles '' , and that mass and momentum are conserved but energy is not .
particles are normally interpreted as representing a coarse - graining of the fluid , so that each particle contains many molecules @xcite . since the intrinsic time scale of dpd represents the correlated motion of mesoscopic packets of atoms or molecules , it is typically orders of magnitude larger than the time scale of molecular dynamics @xcite .
particle positions and momenta are real variables , and are not restricted to a grid .
espaol and warren s analysis @xcite showed that the original dpd model does not satisfy detailed balance , so the equilibrium states ( if they exist ) can not be simply characterized .
detailed balance is the condition equating the rates of forward and backward transition probabilities in a dynamical system , and is a sufficient ( but not necessary ) condition guaranteeing that the system has a ( gibbsian ) equilibrium state @xcite .
espaol and warren formulated a fokker
planck equation and equivalent set of stochastic differential equations which lead to an analogous continuous - time model , @xmath0 \smallskip\cr d{\bf x}_i = { \displaystyle { \bf p}_i \over\displaystyle m_i } dt . } \ ] ] in these equations , @xmath1 , @xmath2 , and @xmath3 denote the momentum , position , and mass of particle @xmath4 .
@xmath5 is a conservative force acting between particles @xmath4 and @xmath6 , while @xmath7 and @xmath8 are the dissipative and random forces .
@xmath9 are independent increments of a wiener process . by it calculus
@xmath10 so @xmath11 is an infinitesimal of order @xmath12 and @xmath11 can be written @xmath13 , where @xmath14 is a random variable with zero mean and unit variance @xcite .
detailed balance is satisfied by this continuous - time version of dpd with an appropriate choice for the form of the forces @xcite , and so equilibrium states are guaranteed to exist and be gibbsian . to ensure that the associated fluctuation dissipation theorem holds , espaol and warren suggested the forces assume the following forms @xcite : @xmath15 @xmath16 and @xmath17 where @xmath18 is the difference in velocities of particles @xmath6 and @xmath4 , @xmath19 is the unit vector pointing from particle @xmath6 to particle @xmath4 , and @xmath20 is a weighting function depending only on the distance separating particles @xmath4 and @xmath6 . the constants @xmath21 , @xmath22 , and @xmath23 are chosen to reflect the relative importance of the conservative , dissipative ( viscous ) , and random components in the fluid of interest . as a consequence of detailed balance and the fluctuation dissipation theorem , @xmath22 and @xmath23 are related to boltzmann s constant @xmath24 and the equilibrium temperature @xmath25 by @xmath26 in order to remain as close as possible to the original dpd model , espaol and warren chose the friction weight function to be @xmath27 within the constant cutoff length @xmath28 and zero otherwise , where @xmath29 is the distance between particles @xmath4 and @xmath6 .
adding eqs .
( [ fc])([fr ] ) together , the total force is @xmath30 \omega_{ij } { { \bf\hat{e}}_{ij}}.\ ] ] immiscible fluid mixtures exist because individual molecules attract similar and repel dissimilar molecules .
the most common example of such behavior arises in mixtures of oil and water below a critical temperature .
the nonpolar , hydrophobic molecules of oil attract one another through short - range van der waals forces , while the polar water molecules have complex , long - range hydrophilic attractions which are dominated by electrostatic interactions , including hydrogen bonds . at the atomistic level employed in molecular dynamics , such interactions demand a detailed treatment . however , to obtain mesoscopic and macroscopic level descriptions using dpd , the microscopic model can be drastically simplified . in order to model immiscible fluids ,
the simplest modification to the one - component dissipative particle dynamics algorithm is to introduce a new variable , called the `` color '' by analogy with rothman
keller @xcite .
for example , we could choose red to represent oil and blue to represent water .
when two particles of different color interact we increase the conservative force , thereby increasing the repulsion .
that is , @xmath31 where @xmath32 and @xmath33 are constants with @xmath34 . as a consequence of mass and momentum conservation , the navier stokes equation is obeyed within a single - phase dpd fluid and within regions of homogeneity of each of the two binary immiscible fluid phases @xcite .
likewise , detailed balance is also preserved , at least in the limit of continuous time @xcite .
the immiscible fluids described above are identical to each other .
however , it is often the case that the two fluids in a mixture will differ physically .
for example , oil and water typically have different viscosities . to model binary fluids with differing viscosity we again adopt the simplest approach : labeling one of the two phases ( colors ) as more viscous .
when two particles of the same viscous color interact , the dissipative ( viscous ) force is increased ; in order to keep the temperature constant , we must correspondingly decrease the random force according to eq .
( [ temperature ] ) , i.e. , @xmath35 @xmath36 where @xmath37 our previous study of finite difference methods applicable to simulations with non - conservative forces @xcite indicated that the finite - difference algorithm suggested by groot and warren @xcite is a good choice for dpd .
their method is a variation on the familiar velocity - verlet algorithm , adding a momentum estimate before the force evaluation : @xmath38
\cr { \bf p}_i(t + \delta t ) & = & { \bf p}_i(t ) + \frac{\delta t}{2 } { \bf f}_i(t ) \cr { \bf f}_i(t + \delta t ) & = & \displaystyle\sum_{j \ne i } { \bf f}_{ij}(t + \delta t ) \cr { \bf p}_i(t + \delta t ) & = & { \bf p}_i(t ) + \frac{\delta t}{2 } \left [ { \bf f}_i(t ) + { \bf f}_i(t + \delta t ) \right],\end{aligned}\ ] ] where @xmath39 is the time step size and @xmath40 is the force acting on particle @xmath4 at time @xmath41 .
the dpd units of length , mass , and time are specific to the particular set of model parameters , and the exact relationship between these parameters and a real fluid in a particular situation can be determined by considering the dimensionless groups relevant to the motion of that system , such as the mach , reynolds , and weber numbers .
many further modifications to the model have been suggested and implemented by others .
some of the most interesting include energy conservation , colloidal particles , and polymers .
in very recent work , it has been shown that it is possible to derive a modified version of dpd directly from the underlying molecular dynamics @xcite .
a central quantity in the study of growth kinetics is the characteristic domain size @xmath42 . for binary systems in the regime of sharp domain walls , this is usually thought to follow algebraic growth laws of the form @xmath43 for symmetric quenches without hydrodynamic interactions , dynamical scaling theory and experiment @xcite indicate that the scaling exponent @xmath44 . if flow effects are relevant , typically @xmath45 in two dimensions , and @xmath46 in three dimensions , where @xmath47 is the hydrodynamic length and @xmath48 is the diffusive length , expressed in terms of the absolute ( dynamic ) viscosity @xmath49 , density @xmath50 , surface tension coefficient @xmath51 , and diffusion coefficient @xmath52 .
these scaling laws follow directly from dimensional analysis of the macroscopic fluid dynamics equations ( so - called model h , or cahn hilliard coupled to navier stokes hydrodynamics ) in the appropriate regimes @xcite .
when @xmath53 in three dimensions , diffusive effects dominate and we expect the domains to form via the lifshitz slyozov wagner ( lsw ) evaporation condensation mechanism @xcite .
when @xmath54 in both two and three dimensions , we expect the growth mechanism to be surface tension driven by hydrodynamic flow , balanced by inertial effects @xcite .
if @xmath55 in three dimensions , we expect viscous hydrodynamic effects to dominate , as predicted by siggia @xcite ; in this regime the surface tension drives the transport of the fluid along the interface , which is possible only if both phases are continuous @xcite .
this third growth regime can not occur in two dimensions , and so we expect to see diffusive growth for @xmath56 @xcite . due to our choice of model parameters and the small size of our simulations , we do not expect to probe the viscous or inertial hydrodynamic regimes .
simulations in two spatial dimensions are especially useful to emphasize the importance of correct hydrodynamics : simulations which do not conserve momentum typically give @xmath57 for the diffusive regime ( @xmath58 ) @xcite .
simulation methods with correct hydrodynamic interactions typically give the expected result of @xmath59 ( see e.g. , refs .
it is worth noting that momentum conservation is not thought necessary to model spinodal decomposition in binary metallic alloys , since phase separation occurs by the migration of atoms to neighboring vacancies on the crystalline lattice @xcite .
simulations based on lattice - boltzmann techniques also typically display @xmath57 .
these observations are supported by a renormalization group approach which has shown that thermal fluctuations cause brownian motion - driven coalescence and play a crucial role in causing @xmath60 to assume the value of @xmath12 @xcite ; although some lattice - boltzmann techniques include these fluctuations @xcite , most do not .
there is some confusion in the literature over which scaling exponents should be observed for off - critical quenches , and whether or not the algebraic scaling law [ eq . ( [ scaling ] ) ] should in fact be obeyed for any off - critical binary immiscible fluid .
several authors have reported the coexistence of multiple length scales @xcite .
it is likewise uncertain as to what behavior should be observed from binary immiscible fluids of differing viscosity @xcite .
certainly , one growth mechanism we expect for off - critical quenches of both equal and differing viscosity is the lsw evaporation condensation mechanism , for which @xmath57 @xcite .
it should also be noted that there are still some experimental and theoretical challenges in unraveling the behavior of systems in which both the order parameter and the momentum are locally conserved @xcite .
experimentally , for example , it is difficult if not impossible to study two - dimensional fluid systems . as far as numerical studies are concerned , it is important to recognize that three - dimensional simulations are particularly demanding on all the aforementioned techniques , and so definitive results are harder to come by than in two dimensions .
indeed , cates s group in edinburgh has recently reported somewhat conflicting results from three - dimensional studies of binary immiscible fluid separation for equal viscosity fluids using dissipative particle dynamics and lattice - boltzmann methods @xcite .
while the former suggests the persistence of non - universal length scales , hypothesized to be due to the intrusion of a `` molecular '' or discretization length scale , this is not supported by the latter , from which finite - size effects were more rigorously excluded .
note , however , that whereas the lattice - boltzmann scheme was based on a landau free - energy approach , and is essentially no more than a finite - difference solution of the continuum model h equations , the macroscopic equations to which the spinodal dpd scheme corresponds have not yet been derived .
this makes it unclear whether the two systems being simulated are really one and the same .
both of these studies emphasize the importance of observing dynamical scaling over decades of time before drawing conclusions as to the true nature of the scaling regime .
one notable result of their work is the clarification of the relative time scales over which computer simulation techniques typically operate .
for the simulations they discuss , molecular dynamics time scales are on the order of @xmath61 in dimensionless units , lattice - gas automata and langevin dynamics probe time scales on the order of @xmath61 through @xmath62 , dpd typically @xmath63 to @xmath64 , and methods based on the lattice - boltzmann equation @xmath65 through @xmath66 .
grant and elder have recently argued @xcite that @xmath67 in any asymptotic scaling regime because the reynolds number @xmath68 ( where @xmath69 is a characteristic velocity ) can not diverge , and in fact must remain less than a critical value @xmath70 to avoid the onset of turbulence @xcite and possible turbulent remixing of the fluids .
the conclusion they draw is that the @xmath71 scaling regime must be transient @xcite .
however , their analysis neglects mention of the relative strength of the interface , quantified for example by the weber number @xmath72 .
if @xmath73 is small at the onset of turbulence we would expect turbulent remixing to be delayed , or perhaps even postponed indefinitely allowing @xmath74 to diverge . in any case , the separation dynamics would likely be affected by @xmath75 ; for example , the fluid viscosity in a turbulent region is roughly proportional to @xmath74 @xcite .
in order to characterize the phase separation kinetics within a binary immiscible fluid , we need a practical tool to measure the characteristic domain size corresponding to the state of the system at a given point in time .
the use of the static structure function for this purpose is widespread .
however , we have noticed bizarre `` early time '' behavior when using the static structure function to characterize simulations of highly off - critical quenches . specifically , the characteristic domain size would sometimes change suddenly by an order of magnitude .
this abrupt change in behavior did not correspond to anything observable in the time evolution of the positions of the dpd particles .
such anomalous behavior is likely due to the large fluctuations prevalent in the static structure function for small simulations of highly - mixed binary fluids , for which the characteristic domain size is very small .
the radial distribution function @xmath76 is a well - established tool for the analysis of single - phase fluids @xcite , and indicates the likelihood of finding two particles separated by a distance @xmath77 . for binary fluids we can also use the same - phase and differing - phase distribution functions @xcite .
the same - phase distribution function @xmath78 describes the likelihood of finding two particles of the _ same _ phase separated by a distance @xmath77 , and the differing - phase distribution function @xmath79 describes the likelihood of finding two particles of _ differing _ phase separated by a distance @xmath77 . from the peaks of the differing - phase radial distribution function , we can estimate the characteristic separation of particles of differing phase ( i.e. , the characteristic domain size ) .
consequently , we decided to calculate the distribution between particles of different phase ( color ) , @xmath80 v(r,\delta r)},\ ] ] where @xmath81 is the mass of each particle ( assumed all identical ) , @xmath82 is the number of particle pairs of differing phase with separation between @xmath83 and @xmath84 , @xmath85 is the total number of particles , @xmath86 is the fraction of particles of one phase ( the more viscous phase if the two phases are of different viscosity ) , and @xmath87 is the volume of the spherical shell of radius @xmath77 and thickness @xmath88 ( from @xmath83 to @xmath89 ) .
it is apparent from eq .
( [ gr ] ) that @xmath79 can only be calculated for @xmath90 in a simulation with a periodic box size of @xmath91 ; this should not be interpreted as an undesirable limitation as finite size effects are normally significant for @xmath92 @xcite .
a value of @xmath93 gives a reasonable compromise between noise and resolution for the size of simulation we discuss in this paper .
the principal difficulty lies in analyzing the results automatically , as we need to calculate @xmath79 at many time steps within each simulation in order to display the domain growth over time .
we chose to calculate the median of that portion of the smoothed @xmath79 curve extending above a threshold value @xmath94 , where @xmath95 is the standard deviation of the smoothed curve when it has effectively a constant unit value , estimated from the last tenth of the smoothed @xmath79 curve at @xmath96 .
we used a fourth order savitzky
golay smoothing filter @xcite to smooth @xmath79 over @xmath77 at each point in time with a symmetric window size chosen to reduce the noise while leaving significant features untouched ( 41 points in two dimensions and 101 points in three ) .
we chose the cutoff threshold of @xmath94 so as not to bias the median by the size of the periodic simulation box ; likewise , we chose the median in preference to the mean ( first moment ) because the median is less biased by outliers , such as those 2.5% of noisy points which effectively have unit value but extend above the threshold . both the median and mean give poor estimates of the characteristic domain size as it approaches @xmath97 , the limit of calculable @xmath79 .
when this situation is detected , we use the global maximum of @xmath79 to estimate the domain size instead ; this is a continuous transition for symmetrical peaks .
once we have calculated the characteristic domain size for a series of @xmath79 curves taken at different times from a single simulation , we begin a search for linear sections in the plot of @xmath98 vs. @xmath99 .
this has also been automated , using analysis of variance to decide whether a given section of the plot is linear or cubic .
we used the analytic expressions for the least - squares fits with moments taken about the means to minimize the effect of roundoff error .
we keep only the longest linear sections , subject to there not being any significant gaps in the coverage of the log log plot .
an ensemble average of a number of simulations yields a plot of the scaling exponent @xmath60 vs. @xmath99 , in which long horizontal ( zero gradient ) sections represent algebraic growth .
this procedure allows more accurate determination of the scaling exponent than a visual comparison of @xmath98 vs. @xmath99 to a straight line of a particular slope , and provides a statistically valid determination of whether or not the growth is truly algebraic over a particular time period .
finite size effects for @xmath100 normally result in non - algebraic growth or unusually low scaling exponents ; either case is easily detected by our method .
a further advantage is the automation we have described , permitting large ensemble averages with minimal effort .
we should note that estimating the characteristic domain size using the median can occasionally lead to discontinuities .
if these discontinuities are large enough they will cause a break between linear sections .
small discontinuities may be spanned by a single line , which would affect the slope ( @xmath60 ) and be detrimental to the results .
however , large discontinuities will not be spanned , and so will not affect the slope .
the results we obtained using these techniques are considerably better than those obtained with the static structure function , especially for highly - mixed fluids where the pair distribution function is not drowned by fluctuations to the extent that the static structure function is .
furthermore , the pair distribution function is more intuitive to analyze than the static structure function , which aids our interpretation of the results .
for the simulations of fluids with differing viscosity , we chose one phase to be an order of magnitude more viscous than the other , i.e. , @xmath101 . before simulating this new complex fluid
, it is advisable to verify that increasing the parameter @xmath22 , while keeping the temperature constant , does indeed increase the viscosity . for the lower - viscosity fluid we chose the model parameters shown in table [ lowerviscosityparameters ] .
the absolute ( dynamic ) viscosity of a homogeneous dpd fluid can be estimated theoretically from the continuous - time viscosity , ignoring the effect of the conservative forces ( i.e. , @xmath21=0 ) @xcite ; the continuous - time viscosity of this lower - viscosity fluid is @xmath102 . in order to verify this estimate
, we performed a series of simulations of steady shear using lees edwards periodic boundary conditions .
we performed a total of 63 simulations with a time step of @xmath103 ( in our dpd time units ) , each from a different random initial configuration and each allowed to settle to steady - state shear before measurement began .
we studied systems of both 1600 and 6400 particles , six simulations at each of nine different shear rates for the former and three simulations at each of three distinct shear rates for the latter .
as the results from the larger simulations gave a mean viscosity nearly identical to that of the smaller ones , we can conclude that finite size effects do not bias the viscosity of the smaller , faster simulations .
we calculated the velocity profile for each set of parameters , and found it to be statistically indistinguishable from linear in every case .
analyzing these simulations led to a conclusion of @xmath104 .
others have also found discrepancies between theory and simulation , particularly regarding the kinematic contribution to viscosity @xcite , so the difference between the simulated viscosity and the continuous - time viscosity is not entirely surprising . since molecular dynamics simulations containing only conservative forces give a finite viscosity , we would be surprised if the theoretical estimate did not differ from our calculations . in order to measure the viscosity of the more viscous fluid , we set up a series of 1600 particle simulations of a homogeneous dpd fluid as described above , using a total of 30 simulations .
we varied the shear rate in the range that gave the best results in the previous simulations .
the model parameters differed in that @xmath22 was a factor of ten larger while @xmath23 was a factor of @xmath105 larger to keep the temperature constant [ see eq .
( [ temperature2 ] ) ] ; these parameters are shown in table [ higherviscosityparameters ] .
we decreased the time step to @xmath106 due to the increased magnitude of the dissipative and stochastic forces . analyzing these results led to the conclusion that @xmath107 , which confirms the increase in viscosity .
this increase is close to a factor of ten , so that for similar model parameters it is reasonable to conclude that @xmath22 is approximately proportional to viscosity .
it is possible to measure the surface tension of a binary immiscible fluid by integrating the pressure tensor across a flat interface , or by verifying laplace s law for a series of equilibrium bubbles of varying radii @xcite .
laplace s law was verified for a dpd binary immiscible fluid in our previous simulations @xcite , and so surface tension measurements were omitted from the present study .
these calculations would , however , allow identification of the unit of time for comparison with other simulation techniques ( see e.g. , refs .
theoretical estimates for surface tension are also available , but are of similar accuracy to the viscosity estimate above . for the simulation of fluids of differing viscosity
, we must also choose the relatively small time step size of @xmath106 in order to ensure the stability of the algorithm as a result of the increased magnitude of the dissipative forces .
this has the consequence of making these simulations computationally much more expensive than equal - viscosity simulations . in this context
, it is worth commenting on the virtues of using dpd to perform simulations of differing - viscosity fluids compared with other models .
sappelt and jckle use an approach based on the cahn hilliard equation ( so - called model b without noise ) with a concentration - dependant mobility @xcite so do not include hydrodynamics , unlike our approach with dpd . the other mesoscale techniques which could be used to model these fluids include lattice - gas automata and methods based on the lattice - boltzmann equation .
as with dpd , there is a high computational price to the lattice - gas approach , which requires adjustment to the collisional outcomes of the look - up tables .
lattice - boltzmann ( more correctly , lattice - bgk ) models can be used , based for example on the swift
yeomans free energy functional approach @xcite , but there is a similar computational price , although with the additional complication of poorly understood numerical instabilities
. however , dpd offers the simplest algorithmic implementation which is thermodynamically consistent .
our differing - viscosity simulations used the model parameters shown in table [ differingmodelparameters ] .
each simulation had 6400 particles , and ran for 50,000 time steps .
we chose to use only 6400 particles in our simulations so that individual simulations would be quick enough to be run multiple times , allowing us to consider the effect of changing the phase fraction and viscosity , and to increase the confidence in our results and calculate accurate error estimates with ensemble averaging .
simulations with more particles would have given better resolution of small scale features ( relative to the system size ) and postponed the finite size effects , at the price of increased computational demands .
our resources would only have allowed a few simulations of the size and length used by jury _
_ @xcite ( 1,000,000 particles ) , making studies of the sort we describe in this paper impossible .
we calculated the pair distribution function at 64 logarithmically - spaced points in time during the course of each simulation , starting from time @xmath96 and finishing at @xmath108 ( in our dpd time units ) .
in both two and three dimensions , we performed ten simulations at each of nine different fractions of the viscous phase , ranging from @xmath109 ( 10% viscous phase ) to @xmath110 ( 90% viscous phase ) .
we show the time evolution of a single simulation at each value of the viscous fraction in figs .
[ 2d - glassy - time][3d - glassy - time2 ] for both two and three dimensions .
we represent the positions of the dpd particles in two dimensions by a gray scale map , in which the particle positions are weighted by eq .
( [ omega ] ) and assigned an intensity of gray based upon the proportion of each phase in this localized - average .
figures [ 3d - glassy - time ] and [ 3d - glassy - time2 ] display the three - dimensional surface of the interface between the two immiscible phases , as defined by there being equal proportions of each fluid in the localized - average . in these four figures the gradual development of domains can be seen .
we examined the results of the domain size analysis for each simulation in plots of @xmath98 vs. @xmath111 , and further examined each ensemble average in plots of @xmath60 vs. @xmath99 ; the latter yielded the mean values of @xmath60 and corresponding 68% confidence intervals for significant sections of algebraic growth .
we show these results for two and three dimensions in tables [ glassy ] and [ 3d - glassy ] , and figs .
[ glassy2 ] and [ 3d - glassy2 ] .
the range of @xmath99 we give in the tables should be taken as a rough guide , since there is often a high degree of correlation between the start and end times of a particular linear section and its growth exponent within the broad category of late time .
we should emphasize that we are using the term `` late time '' as a relative category in this paper , to distinguish these results from those obtained at very early times . due to the model parameters and small size of these simulations
, we do not expect to probe the viscous or inertial hydrodynamic regimes ( see eqs .
( [ scaling])([3d - exponents ] ) , and compare with refs .
@xcite ) . we also constructed a series of equal - viscosity simulations in both two and three dimensions .
these had @xmath112 , 6400 particles , and stopped at @xmath113 .
the model parameters were the same as in the differing - viscosity simulations , with the exception that both phases were identical to the less - viscous phase of the differing - viscosity simulations ( i.e. , @xmath114 and @xmath115 ) .
we calculated the pair distribution function at 64 different points from @xmath96 to @xmath113 , with ten simulations at each of eight different minority phase fractions from @xmath116 to @xmath117 in two dimensions , and ten simulations at each of five different fractions from @xmath109 to @xmath117 in three dimensions .
we show the time evolution of a single simulation at each value of the minority fraction in figs .
[ 2d - same - time ] and [ 3d - same - time ] for two and three dimensions respectively .
in these two figures the gradual development of domains can be seen , and is greatly slowed for small minority fraction .
we show the scaling exponents for two and three dimensional equal - viscosity fluids in tables [ same - viscosity ] and [ 3d - same ] , and figs .
[ same - viscosity2 ] and [ 3d - same2 ] , where we mirror the scaling exponents about @xmath118 to show the full range of minority phase fraction .
comparison with figs .
[ 2d - same - time ] and [ 3d - same - time ] qualitatively confirms the same behavior .
a feature common to all these simulations is the lack of scale - invariance at very early times , until approximately @xmath119 to @xmath120 ( in our dpd time units ) depending on the exact composition of the fluid .
this is apparent in the @xmath79 curves as multiple peaks of similar magnitude , but could not be seen in the frequency domain shown by the static structure function .
figure [ g00013.9 ] shows an example of the differing - phase radial distribution function for a three - dimensional equal - viscosity simulation ( @xmath118 ) at @xmath121 ; the crosses represent the actual data and the curve represents the smoothed data .
this is one of the simulations we show in fig .
[ 3d - same - time ] .
the multiple peaks typically evolve by changing their height and relative weight , but not their position .
this is noticeably different from our `` late time '' behavior of the distribution function , where a single peak gradually advances its position while broadening but retaining its height and general shape .
it is because of this `` early time '' behavior that it was decided not to use solely the global maximum of @xmath79 to estimate the characteristic domain size . at these `` early times '' in all of the simulations , we observed algebraic growth with a very small exponent , roughly @xmath122 in two dimensional equal - viscosity simulations , @xmath123 in two dimensional differing - viscosity simulations ( except @xmath124 ( 90% viscous phase ) , for which @xmath125 ) , and @xmath126 for simulations in three dimensions .
this corresponds to the region of breakdown of scale invariance we described at the beginning of this section .
these `` early time '' exponents are unaffected by viscous or minority phase fraction in all of the simulations , except the differing - viscosity simulations in two dimensions
. here @xmath60 decreases with increasing @xmath127 ( viscous phase fraction ) , probably due to the domain - growth arresting effect of increasing viscosity .
we observed a remarkable ( though short ) regime of @xmath125 for @xmath124 ( 90% viscous phase ) , for which we have no adequate explanation .
others normally discard similar early time regimes without comment @xcite or as an `` early stage '' or `` transient '' regime @xcite . however , there is growing evidence for the coexistence of multiple domain sizes and hence a breakdown in universality , at least in certain phase - ordering domains @xcite . for `` late time '' domain growth in the two - dimensional differing - viscosity simulations
( see fig .
[ glassy2 ] ) , we observed a fairly constant value of @xmath60 for @xmath128 ( 20% viscous phase ) through @xmath129 ( 60% viscous phase ) , decreasing both for @xmath130 and very slightly for @xmath131 and @xmath132 , then decreasing sharply at @xmath124 .
this asymmetry is consistent with the variation of the `` early time '' exponent , in that an increasingly viscous fluid is expected to develop domains more gradually . at increasingly rarefied fractions ( @xmath130 and @xmath124 ) ,
domain growth is retarded by the increased isolation of the droplets .
the `` late time '' growth exponent throughout is effectively @xmath133 , which suggests that the presence of fluids of differing viscosity interferes with the normal @xmath59 growth mechanism in two dimensions .
the growth exponent of @xmath133 is expected from the lsw evaporation condensation mechanism @xcite .
this is in some ways analogous to the effect obtained by deliberately breaking momentum conservation in symmetric quenches , as described previously @xcite .
our domains are considerably less circular at all viscous fractions than those observed by sappelt and jckle ( compare figs .
[ 2d - glassy - time ] and [ 2d - glassy - time2 ] with ref .
this is likely due to the lack of hydrodynamic interactions in their model and to the greater difference in viscosity between their two phases . as in their simulations
, our two - phase structure for fluids of differing viscosity is not very different from the structure for fluids of equal viscosity .
moreover , our simulations do not reveal any new insights regarding interfacial structure . in three dimensions ,
the `` late time '' domain growth of differing - viscosity fluids displays nearly the opposite behavior , with the scaling exponent increasing as the viscous fraction reaches its extremes .
this could be explained by the increased fluid mobility in simulations with an extra spatial dimension , as the majority phase is completely connected and so the domain growth could occur according to the @xmath71 mechanism , which is surface tension driven by hydrodynamic flow , balanced by inertial effects .
this is qualitatively substantiated by inspection of figs .
[ 3d - glassy - time ] and [ 3d - glassy - time2 ] , where a larger degree of connectivity of the majority phase can be observed at the extremes of viscous fraction than for @xmath134 .
however , a more obvious mechanism for the domain growth would be the @xmath57 lsw evaporation condensation mechanism @xcite ; it may be a combination of these two mechanisms that leads to our observed @xmath135 growth .
a slight asymmetry in the `` late time '' growth exponent is also evident , with domain growth proceeding more slowly with increasingly viscous fluids of non - extreme viscous fraction . for equal - viscosity fluids in two dimensions we observed the expected @xmath59 for symmetric quenches ( @xmath118 ) @xcite .
as we reduced the minority phase fraction , we observed a steady decrease in the scaling exponent until @xmath57 is reached at the extremes .
this confirms the results of other workers @xcite that increasingly off - critical quenches retard the domain growth , while providing support for the observed slowdown of growth at the extremes of viscous fraction in differing - viscosity fluids in two dimensions , which we commented on above . in three dimensions ,
the domain growth of equal - viscosity fluids appears largely unaffected by varying the minority phase fraction .
although there may be some increase in @xmath60 at the extremes of @xmath127 ( as seen in the differing - viscosity fluid in three dimensions ) , this is difficult to confirm definitely because of the large variation in rate of growth observed for the simulations with @xmath130 , and hence correspondingly large confidence interval .
the scaling exponent throughout is close to @xmath57 .
whereas jury _ et al . _
@xcite were intending to probe the viscous or inertial hydrodynamic regimes with their dpd simulations of equal - viscosity fluids in three dimensions , we aimed only to probe length scales below @xmath136 and @xmath137 [ see eqs .
( [ scaling])([3d - exponents ] ) ] . as such ,
our results are fully consistent with theirs .
our exclusion of finite size effects is more rigorous than theirs , and although not as extreme as that advocated by kendon _ et al . _
@xcite our method gives statistical confidence that these domains are scaling algebraically .
both jury _ et al . _
@xcite and kendon _ et al . _
@xcite were able to cover the time domain more fully in three - dimensional equal - viscosity symmetric quenches only at the cost of performing a large number of computationally very intensive and very expensive simulations .
in this paper , we have described simulations of the domain growth and phase separation of hydrodynamically - correct binary immiscible fluids of differing and equal viscosity as a function of minority phase concentration in both two and three spatial dimensions . due to our choice of model parameters and the small size of our simulations
, we did not expect to probe the viscous or inertial hydrodynamic regimes . in three dimensions
, we found that the characteristic domain size scales as @xmath138 for simulations of differing and equal - viscosity fluids developing from symmetric and slightly off - critical quenches .
for highly off - critical quenches we observe an increase in the scaling exponent . in two dimensions ,
we also observe @xmath138 in simulations of differing - viscosity fluids developing from symmetric and slightly off - critical quenches , although we observe a decrease in the scaling exponent for highly off - critical quenches . in equal - viscosity fluids in two dimensions , we observe @xmath139 for symmetric quenches and a roughly linear decrease to @xmath138 for highly off - critical quenches ; these results are in agreement with similar lattice - gas simulations in two dimensions @xcite . obtaining meaningful results for ensemble averages of highly off - critical binary immiscible fluids
was only made feasible by our automation of the calculation of the characteristic domain size by the pair correlation function .
it also made possible the identification of a regime of breakdown of scale invariance at very early times , which was not noticeable in our original analysis using the static structure function .
further simulations aimed to probe the viscous and inertial hydrodynamic regimes [ see eqs .
( [ scaling])([3d - exponents ] ) ] would be a useful addition to this work , as would simulations aimed to cover longer periods of time ; however , both would require substantially increased computational work .
many helpful discussions were had with peter bladon , bruce boghosian , alan bray , mike cates , pep espaol , simon jury , colin marsh , john melrose , and julia yeomans ; extra thanks is due to matthew probert and matt segall .
we would also like to thank nato for a grant which has supported this work in part , the epsrc ( u.k . ) e7 grand challenge in colloidal hydrodynamics for providing access to the cray t3d at the edinburgh parallel computing centre , the csar service at the university of manchester for access to their cray t3e and sgi origin2000 through the epsrc high performance computing for complex fluids grant , and the cambridge high performance computing facility for access to their hitachi sr2201 .
ken gratefully acknowledges financial support from nserc ( canada ) and the ors awards scheme ( u.k . ) .
in this appendix we provide a few comments regarding the running of dpd simulations on high performance computers .
we usually had easy access to single - processor workstations , with a large variety of types and speeds of processors .
multi - processor machines allowing parallel execution of simulations are much less common and are more difficult to obtain access to , although they have become more common during this research project .
we used both the cray t3d of the edinburgh parallel computing centre ( epcc ) and the hitachi sr2201 of the cambridge high performance computing facility ( hpcf ) for computing the results described in this paper ; the former consisted of 512 processor nodes and the latter consists of 256 nodes .
the implementation of the dissipative particle dynamics algorithm is very similar to that of conventional molecular dynamics algorithms @xcite .
for example , we divide the periodic spatial domain ( the simulation cell ) into a regular array of equally sized link - cells , such that each side of the rectangular domain has an integer number of cells and each cell is at least @xmath140 across .
each link - cell consists of a dynamically allocated array of particles , and pointers to the neighboring cells .
individual particles consist of the position momentum vector pair and a color index . for each time step
we iterate through the particles in each link - cell , calculating the force acting on each particle as it interacts with the particles in the same and neighboring link - cells .
since the dpd force acts between pairs of particles , we must ignore half of the neighboring cells to avoid duplication . when considering a different particle pair , we compare the square of the separation distance with @xmath141 , skipping to the next particles if the pair is out of range .
we then compute the new position and velocity as determined by the finite - difference algorithm ( see eq .
( [ groot - warren ] ) and ref .
. we may write the complete state of the system to file , and we can perform other calculations thereafter , for example to determine the temperature and pressure of the system .
we used the freely - available _ gnu - make _ utility to dictate the compilation process , since the decision structures it contains make it simple to write programs portable to a large range of architectures .
we created a comprehensive , automated test suite to make it easy to verify that optimizations of the calculations did not accidentally change the results of the computations .
given constant @xmath140 and number density @xmath142 , the dpd algorithm scales linearly ( in both computation time and memory size ) with increasing number of particles ( @xmath85 ) , and is limited by computation time on all but the smallest machines .
the main simulations we performed for this paper consisted of 6400 particles , and it is on the parallel and serial performance of this size of simulation that we will make most of the following comments .
details of the performance of this size of simulation in two dimensions are shown in tables [ hpc - small1 ] and [ hpc - small2 ] .
these tables give the elapsed time per node in seconds and relative parallel efficiency for the first 1000 time steps , including data for a variety of computers and partition sizes .
these data are for code compiled with the highest level of optimization , including some small reductions in floating - point accuracy .
table [ hpc - small1 ] describes the computers used to calculate the results in this paper , while table [ hpc - small2 ] describes the computers to which we have recently been allowed access , such as the computer services for academic research ( csar ) cray t3e and sgi origin2000 in manchester .
a typical simulation of 50,000 time steps takes 2.5 hours on a 350 mhz intel pentium ii pc , the fastest single - processor machine to which we had common access .
this same simulation would take 2.9 hours on a 16-node partition of the t3d at the epcc .
however , to minimize fragmentation of the machine , jobs using up to 32 nodes were limited to a total execution time of 30 minutes .
one possibility was to break up the run into 30 minute portions , but this introduces additional overhead and complications ; however , new jobs start instantly because they need not be queued . a better option was to run jobs on a 64-node partition , task farming four 16-node jobs to run simultaneously
there was a 12 hour limit to 64512 node jobs ( 6 hours during the week ) , but there was often a long wait in the queues .
if the efficient usage of billed time was a significant concern , sixteen 2-node jobs would complete in 8.0 hours .
however , during the week this meant restarting halfway through and waiting in the queue again .
similar comments apply to the hitachi sr2201 , although its queues were limited to 8 hours maximum . the extra administrative overheads involved and the billed usage means that we usually concentrated computation on the serial workstations
however , parallel execution becomes more attractive with larger simulations .
the parallel efficiency of dpd with 6400 particles is good only for a modest number of processor nodes .
this is particularly true of the more modern parallel machines such as the cray t3e and sgi origin2000 , which are proportionally faster in processing than communicating when compared with their older counterparts .
much better parallel efficiency has been observed with larger simulations .
the cray t3d shows an unusual increase in efficiency when going from a serial calculation to a 2-node parallel calculation with 6400 particles ; this could be explained by any number of hardware - specific arguments .
we should note that the results for the origin2000 include the effect of sharing the machine with other users , unlike all the other machines whose results appear in tables [ hpc - small1 ] and [ hpc - small2 ] , for which each node was dedicated to our calculations .
we decided to write the main simulation program in c / c++ as opposed to fortran .
this choice was made because c / c++ were believed to be the most appropriate languages for dealing with dpd simulations which consist of a large amount of book - keeping wrapped around fairly simple computations .
c / c++ and fortran are highly portable to different computer architectures , and although well - written fortran is more efficient on vector machines , for almost all other situations they are of similar speed , given equally good compilers .
the use of vector machines ( such as the hitachi sr2201 ) was not anticipated when this work on dpd began several years ago .
furthermore , it was not believed that the basic algorithm would vectorize well , due to the short vector length in typical computations .
large programs are easier to maintain in c / c++ than in fortran , although the increasingly well - supported fortran 90 and 95 make the difference less significant . finally , we comment on our findings in tuning the message passing interface ( mpi ) calls for the cray t3d . in our simulations , it was found that blocking calls ( sends and receives ) were faster than non - blocking calls and were easier to use correctly .
furthermore , better scaling was achieved by sending the size of a variable - size message in a separate message rather than probing incoming messages to determine their size . finally , using derived data types to remove unneeded data from messages was slower than sending everything .
ccc @xmath127 & approximate range of @xmath147 & @xmath60 + 0.2 & @xmath148 to @xmath149 & @xmath150 + 0.3 & @xmath151 to @xmath152 & @xmath153 + 0.4 & @xmath154 to @xmath155 & @xmath156 + 0.5 & @xmath154 to @xmath157 & @xmath158 + 0.6 & @xmath159 to @xmath160 & @xmath161 + 0.7 & @xmath162 to @xmath163 & @xmath164 + 0.8 & @xmath165 to @xmath166 & @xmath167 + 0.9 & @xmath168 to @xmath169 & @xmath170 + 0.1 & @xmath171 to @xmath172 & @xmath173 + 0.2 & @xmath174 to @xmath175 & @xmath176 + 0.3 & @xmath177 to @xmath178 & @xmath179 + 0.4 & @xmath180 to @xmath181 & @xmath182 + 0.5 & @xmath183 to @xmath184 & @xmath185 + 0.6 & @xmath186 to @xmath187 & @xmath188 + 0.7 & @xmath189 to @xmath190 & @xmath191 + 0.8 & @xmath192 to @xmath193 & @xmath194 + 0.9 & @xmath195 to @xmath196 & @xmath197 + ccc @xmath127 & approximate range of @xmath147 & @xmath60 + 0.1 & @xmath168 to @xmath198 & @xmath199 + 0.2 & @xmath200 to @xmath201 & @xmath202 + 0.3 & @xmath203 to @xmath204 & @xmath205 + 0.4 & @xmath206 to @xmath207 & @xmath208 + 0.5 & @xmath209 to @xmath210 & @xmath211 + 0.6 & @xmath212 to @xmath213 & @xmath214 + 0.7 & @xmath215 to @xmath216 & @xmath217 + 0.8 & @xmath218 to @xmath219 & @xmath220 + 0.9 & @xmath221 to @xmath222 & @xmath223 + 0.1 & @xmath224 to @xmath225 & @xmath226 + 0.2 & @xmath227 to @xmath228 & @xmath229 + 0.3 & @xmath230 to @xmath231 & @xmath232 + 0.4 & @xmath233 to @xmath234 & @xmath235 + 0.5 & @xmath236 to @xmath237 & @xmath238 + 0.6 & @xmath239 to @xmath240 & @xmath241 + 0.7 & @xmath242 to @xmath243 & @xmath244 + 0.8 & @xmath245 to @xmath246 & @xmath247 + 0.9 & @xmath248 to @xmath249 & @xmath250 + ccc @xmath127 & approximate range of @xmath147 & @xmath60 + 0.2 & @xmath203 to @xmath251 & @xmath252 + 0.25 & @xmath206 to @xmath253 & @xmath254 + 0.3 & @xmath255 to @xmath256 & @xmath257 + 0.4 & @xmath258 to @xmath259 & @xmath260 + 0.5 & @xmath261 to @xmath262 & @xmath263 + 0.05 & @xmath264 to @xmath265 & @xmath266 + 0.1 & @xmath267 to @xmath268 & @xmath269 + 0.15 & @xmath270 to @xmath271 & @xmath269 + 0.2 & @xmath272 to @xmath273 & @xmath274 + 0.25 & @xmath275 to @xmath276 & @xmath277 + 0.3 & @xmath278 to @xmath279 & @xmath280 + 0.4 & @xmath281 to @xmath282 & @xmath283 + 0.5 & @xmath284 to @xmath285 & @xmath286 + ccc @xmath127 & approximate range of @xmath147 & @xmath60 + 0.1 & @xmath203 to @xmath287 & @xmath288 + 0.2 & @xmath289 to @xmath290 & @xmath291 + 0.3 & @xmath292 to @xmath293 & @xmath291 + 0.4 & @xmath203 to @xmath294 & @xmath295 + 0.5 & @xmath203 to @xmath296 & @xmath297 + 0.1 & @xmath298 to @xmath299 & @xmath300 + 0.2 & @xmath301 to @xmath302 & @xmath303 + 0.3 & @xmath304 to @xmath305 & @xmath306 + 0.4 & @xmath307 to @xmath308 & @xmath309 + 0.5 & @xmath310 to @xmath311 & @xmath241 + lccc & number & elapsed time & parallel + machine & of nodes & per node & efficiency + dec 3000/400 axp & 1 & 624 & + ( 133 mhz alpha ev4 ) & & & + linux pc & 1 & 180 & + ( 350 mhz intel pentium ii ) & & & + sgi indigo2 & 1 & 200 & + ( 195 mhz mips r10000 ) & & & + epcc cray t3d & 1 & 1254 & 1.00 + ( 512 @xmath312 150 mhz alpha ev4 ) & 2 & 575 & 1.09 + & 4 & 354 & 0.89 + & 8 & 251 & 0.62 + & 16 & 206 & 0.38 + & 32 & 214 & 0.18 + hpcf hitachi sr2201 & 1 & 1202 & 1.00 + ( 256 @xmath313 150 mhz harp-1e ) & 2 & 634 & 0.95 + & 4 & 371 & 0.81 + & 8 & 255 & 0.59 + & 16 & 212 & 0.35 + & 32 & 243 & 0.15 + lccc & number & elapsed time & parallel + machine & of nodes & per node & efficiency + sgi octane & 1 & 152 & 1.00 + ( 250 mhz mips r10000 ) & 2 & 87 & 0.88 + csar cray t3e-1200e & 1 & 143 & 1.00 + ( 576 @xmath312 600 mhz alpha ev5 ) & 2 & 96 & 0.74 + & 4 & 75 & 0.48 + & 8 & 67 & 0.27 + & 16 & 70 & 0.13 + csar sgi origin2000 & 1 & 133 & 1.00 + ( 16 @xmath312 250 mhz mips r10000 ) & 2 & 81 & 0.83 + & 4 & 69 & 0.49 + & 8 & 60 & 0.28 +
|
we investigate the domain growth and phase separation of hydrodynamically - correct binary immiscible fluids of differing viscosity as a function of minority phase concentration in both two and three spatial dimensions using dissipative particle dynamics .
we also examine the behavior of equal - viscosity fluids and compare our results to similar lattice - gas simulations in two dimensions .
|
searching for companions among the nearby barred galaxies from the shapley - ames catalog , garca - barreto et al .
( 2003 ) identified ngc 6908 as a discoidal companion of ngc 6907 , 10 kpc away from it .
they estimated this distance based only on their angular separation . but according to these authors , the alignment of the objects could be a mere effect of geometrical superposition , since the narrow band h@xmath3 images of ngc 6907 did not reveal any emission in direction of ngc 6908 ( garca - barreto et al . 1996 ) .
surace et al . (
2004 ) recognized ngc 6907 and 6908 as a potential interacting system , but the low resolution of their far - infrared images did not allow any further conclusion . in a recent paper , madore et al .
( 2007 ) revealed that ngc 6908 , once classified as a pofg ( part of a galaxy ) , is in fact a lenticular s0(6/7 ) galaxy , as indicated by its exponential brightness profile observed in several filters .
based on long - slit observations , these authors found different radial velocities for ngc 6908 lines in absorption ( 3113 @xmath4 73 km s@xmath1 ) and in emission ( 3060 @xmath4 16 ) km.s@xmath1 . comparing with the radial velocity that they estimated for ngc 6907 ( 3190 @xmath4 5 km s@xmath1 )
, the velocity difference between the two systems would be in the interval of 77 km s@xmath1 to 130 km s@xmath1 . summing the evidences of morphological asymmetry in the arm structure of ngc 6907 , the behavior of its dust lanes and the tail of debris observed at low surface brightness
, they conclude that the two galaxies are certainly interacting . despite the convincing arguments in favor of the interacting nature of the two galaxies , one could still question whether the absorption and emission lines observed in the direction of ngc 6908 are indeed from that galaxy or originate from the rotating disk of ngc 6907 .
studies of rotation curves of spiral galaxies based on ionized gas ( emission lines ) and on the stellar component ( absorption lines ) have shown that , for most galaxies , the rotation of the two components is similar , but local differences of velocities up to 100 km s@xmath1 can occur ( saito et al .
1984 ; fillmore et al . 1986 ; kormendy & westpfahl 1989 ; vega beltrn et al . 2001 ; pizzella et al .
2004 and rhee et al .
2004 ) . on the other hand ,
only luminous s0 galaxies are expected to present detectable emission lines . since ngc 6908 is a low luminosity s0(6/7 ) galaxy ( m@xmath5=-17.4 , madore et al .
2007 ) superimposed on the spiral arm of ngc 6907 , the lines used to measure its velocity could well originate in the rotating material of the disk of ngc 6907 .
we present in this paper new observations , at radio and optical wavelengths , that allow us to definitely settle this question , and to propose a scenario that describes in more detail the recent collision between the two galaxies .
we performed high resolution radio observations at 21 cm with the giant meterwave radio telescope ( gmrt ) , and a detailed spectroscopic study over the field of the system ngc 6907 and 6908 , with the gemini north telescope . at 21 cm , using the aperture synthesis technique , the h i intensity maps and the velocity fields of both objects were obtained .
the emission of the two objects could be distinguished and their velocities precisely measured .
furthermore the excess in velocity in the velocity field was used to estimate the time passed from the start of the interaction .
the optical multi - object spectroscopy , employing the gmos instrument , allowed the study of the emission and absorption lines of selected h ii regions in the ngc 6907 disk and the nucleus of both galaxies . each observed spectrum
could be compared with its radio counterpart and the differences in absorption and emission velocities in direction of ngc 6908 could be explained by the idea that the emission lines were generated on the ngc 6907 disk and the absorption lines were formed on ngc 6908 .
arcsec@xmath6 . it is overlaid by the gmrt continuum observation .
the contours are associated with the levels 2 , 4 , 11 , 18 , 25 , 33 , 40 , 47 , 54 and 61 mjy / beam .
the external thicker contour corresponds to the limit of detection of 2 sigmas of the noise level for the combined continuum image ( @xmath7 0.6 mjy / beam ) .
optical isophotal contour observed at the gemini telescope ( gray ) and elliptical fit ( dashed line ) are presented too.,width=325 ] inside a field of @xmath8 centered on ngc 6907 ( figure [ fig1 ] ) , only two extragalactic objects dominate the optical emission : ngc 6907 itself , an sbbc spiral galaxy ( eskridge et al . , 2002 ) with an absolute b magnitude m@xmath5 = -21.34 and a surface brightness of 22.29 mag arcsec@xmath9 at half b total light ( lauberts & valentijn 1989 ) and ngc 6908 , a s0(6/7 ) galaxy with m@xmath5=-17.4 ( madore et al .
2007 ) , 43.6 arcsec north - west from ngc 6907 .
garcia ( 1993 ) mentioned ngc 6907 as the main member of a sparse group of 3 galaxies ( ic 4999 and ic 5005 , without considering ngc 6908 ) , also identified in the catalog of galaxy groups of maia et al ( 1989 ) , but these objects can not be seen in the field shown in the figure . in the radio continuum at 1.4 ghz ,
ngc 6907 is also the brightest object , as can be seen from the nrao vla sky survey ( condon et al .
the main properties of ngc 6907/8 galaxies are listed in table [ tbl-1 ] .
.main properties of the galaxies ngc 6907 and 6908 .
the index @xmath10 is for the optical observations ( combining the g@xmath11 and r@xmath11 filters ) and @xmath12 for the radio observations . @xmath13 and @xmath14 are the equatorial coordinates , @xmath15 is the systemic velocity , @xmath16 is the size , @xmath17 the inclination , @xmath18 the position angle , @xmath19 the surface brightness , @xmath20 the morphological type , @xmath21 the galactocentric hubble flow distance , @xmath22 the 21-cm flux density , @xmath23 the galactic extinction in b - band , @xmath24 the maximum velocity rotation , @xmath25-@xmath26 the color in these filters , @xmath27 the absolute b - band magnitude , @xmath28 the total b luminosity , @xmath29 the dynamical mass and @xmath30/l@xmath5 is the @xmath31 mass to b luminosity ratio .
the number for the references are : ( 1 ) eskridge et al .
( 2002 ) , ( 2 ) madore et al .
( 2007 ) and ( 3 ) amores & lpine ( 2005 ) . [
cols="<,<,<",options="header " , ] a more detailed view of the continuum emission of ngc 6907 shows a north - east elongated structure , with an asymmetric emission which does not provide good elliptical fitting ( figure [ fig1 ] ) .
the interaction between ngc 6907 and 6908 must be the main reason for such asymmetry .
almost all the continuum emission is confined to an isophote of 1.9 mjy / beam with approximately the same size than the optical emission centered at @xmath32=20@xmath3325@[email protected] and @xmath36=-24@xmath3748@[email protected] .
the total continuum flux measured for ngc 6907 is 112 mjy , in absolute agreement with condon et al .
( 1998 ) and vollmer et al .
( 2005 ) , if the same isophotal limit is considered .
the radial profile of this emission exhibits the expected behavior for an exponential disk inside a de - projected radius of 72 arcsec . subtracting the continuum image from the image cubes composed for uv - ranges up to 40 k@xmath39 it was possible to recover the spatial h i distribution for each channel , resolving structures larger than 10@xmath4012 arcsec .
that is enough to sample the integrated contribution of the components inside the galaxy ngc 6907 .
the bidimensional integration of the h i emission detected above @xmath41 of the noise level in all channels composes the 0th moment . in figure [ fig6 ]
this map was converted to the h i column density assuming the relation by spitzer ( 1978 ) .
it reveals a substantially more extended spiral structure than observed in optical images and the emission has two extended peaks coincident with the spiral arms .
image of ngc 6907 and 6908 .
the contour levels are at 0.3 , 8.8 , 17.3 , 25.9 , 34.4 , 42.9 , 51.4 , 60.0 , 68 .5 and 77.0 in units of @xmath42.,width=325 ] elliptical fitting for the isophote associated with a column density of 2.8@xmath43 provides the extension of the radio emission as ( 3.5 @xmath4 0.3 ) arcmin or almost 2.4 times larger than the optical emission ( table [ tbl-3 ] ) .
neither the optical center nor the radio center are coincident .
the integration of the column density on elliptical rings along ngc 6907 provides the radial column density profile presented in figure [ fig7 ] .
as expected , there is depletion of h i on the central regions of ngc 6907 , since the molecular hydrogen is formed in a denser environments .
after the peak emission at 10 kpc an exponential fall occurs up to 28 kpc , where the column density rises up to 34 kpc and falls back again .
such distribution is quite common as can be seen in broeils & van woerden ( 1994 ) , and the minimum inside the galactic disk can be related to the gas redistribution in the corotation radius . on the other hand ,
the integration of all the emission for each channel of the datacube after continuum subtraction provides the global h i emission profile for the galaxies ngc 6907 and 6908 as it can be seen in figure [ fig8 ] . . inside the distribution
two horizontal dashed lines are used to mark the h i line - widths estimated at 20% ( w@xmath44 ) and at 50% ( w@xmath45 ) of the peak intensities ; @xmath46 is the central weighted mean velocity distribution.,width=325 ] the double peaked emission is expected for rotating disks .
nevertheless , the different levels of the peaks indicates the presence of other components in the distribution of velocities measured in the field .
these effects can be attributed to the spatial velocity of ngc 6908 projected on the line - of - sight and to the intrinsic asymmetries on the velocity field of ngc 6907 .
richter & huchtmeier ( 1991 ) , chengalur et al ( 1993 ) and doyle et al ( 2005 ) observed these galaxies in lower resolution with different single dishes , what can possibly account for the main differences in the registered line profiles .
the mean differences are lower than @xmath47 , including the flux density ( 48.3 @xmath4 1.8 mjy.km s@xmath1 ) , suggesting that a beam smearing is the main responsible for the differences found at the flux peak .
the weighted mean velocity associated with the h i profile ( @xmath48 km s@xmath1 ) is absolutely compatible with the velocity measured by the robust mean of the entire 1st moment map ( 3183.2 @xmath4 6.7 km s@xmath1 ) .
taking into account the hubble flow distance ( supposing h = ( 73 @xmath4 5 ) km / sec / mpc ) , the galactocentric distance of this galaxy is estimated as ( 44.5 @xmath4 3.1 ) mpc . using this distance a h i mass of ( 8.3 @xmath4 0.4)@xmath49 @xmath50 was calculated .
considering the distribution of velocities associated with the bidimensional emission in each frequency of the 1st moment map , then the observed velocity field of the galaxy can be extracted ( figure [ fig9 ] ) . .
dashed lines represent the part of the velocity field going away from the observer while the opposite is represented by the continuous lines .
the thicker crossing lines represent the kinematical axis .
the velocity field was generated using an uv range of @xmath51 and a resolution of @xmath52 , to balance the sensitivity to the diffuse emission and the spatial resolution .
, width=325 ] one can note that the velocity field is strongly oriented by the spiral arm in the internal regions .
in fact , the high density of the h i emission in the spiral arms ( from 10 to 100 times higher than the interarm regions ) could generate a bias in the observed velocity field , if the velocity from the disk does not correspond to that measured in the spiral arms .
the kinematic major axis is not perpendicular to the kinematic minor axis ( figure [ fig9 ] ) . that can be easily verified by comparing the linear fit over the coordinates of the points with the systemic velocity ( minor axis ) and the coordinates of the nearest points to the minor axis of all isovelocity levels , which samples the major axis in an axisymetric rotation disk .
the angle between these axis is ( 78.2 @xmath4 3.1)@xmath37 and the point where they cross each other can be used to define the kinematic center ( @xmath53=20@xmath3325@[email protected] and @xmath54=-24@xmath3748@[email protected] ) .
since the the global h i profile suggest that a rotating disk must be the main component of the movement of the system , then the rotation curve of the galaxy ngc 6907 can be extracted supposing that the elliptical fitting made on the h i gas distribution represent the projection of the ngc 6907 disk on the sky and its gas follows circular orbits .
in this situation the observed line - of - sight velocity ignoring vertical and radial movements would be : @xmath55 taking into account that : @xmath56 @xmath57 where @xmath58 is the kinematic center , @xmath15 is the systemic velocity , @xmath59 is the position angle of the receding semi - major axis , @xmath17 is the inclination between the normal to the plane of the galaxy and the line - of - sight and @xmath60 and @xmath61 are the polar coordinates inside the plane of the galactic disk .
note that we introduce the term @xmath62 multiplying all distances measured in right ascension ( what it is not usually made in the literature ) , otherwise it would overestimate the galactocentric radius , depending on the declination of the source and the position angle of each target relative to the center of the galaxy . for the ngc 6907 these errors are predominantly of the order of 12% . using the combination of the velocity fields made with maps covering uv - ranges from 1k@xmath39 to 40k@xmath39 , taking into account that the most reliable information about the rotation curve is extracted near the kinematic major axis ( @xmath63 )
, then the previous equations can be used to determine @xmath64 , @xmath60 and @xmath61 for the receding and approaching regions of the galaxy ( see figure [ fig10 ] ) . the tilted ring model was applied to the ngc 6907 velocity field , considering the same limits used in the previous method .
variations of the inclinations and the position angles with the distance are lower than the uncertainties for the position angle and inclination of the whole galaxy so its results can be ignored . a raw estimate for the dynamical mass of ngc 6907 , considering the last point in the rotation curve
is @xmath65 @xmath50 . and r@xmath11 . for the stellar rotation curve
it was supposed a mass - to - light ratio of 1.5.,width=325 ] comparing the radial velocities from optical and h i spectra observed in the same directions ( figure [ fig11 ] and table [ tbl-4 ] it is possible to verify the overall correspondence between the optical and radio rotation curves @xmath66 and verify that the radio projection parameters provides the best correspondences .
the high dispersion of individual points is related to the higher uncertainties to extract radial velocities azimuthaly far away from the major axis . in spite of that ,
the mean dispersion is inferior to the limit of 50 km s@xmath1 , refereed by bosma ( 1981 ) as the limit related to large - scale asymmetries of velocity fields .
the exclusion of the objects in larger azimuthal angles relative to the major axis @xmath67 provides almost the same fit .
in particular , the point representing the slit which sampled ngc 6908 presents a higher velocity dispersion than the others , possibly due to the spatial motion of ngc 6908 . )
a first view on the ngc 6907 velocity field ( figure [ fig9 ] ) reveals that ngc 6908 is over the approaching part of it .
the extraction of the rotation curve in this part of the galaxy reveals contributions of radial components of the ngc 6908 movement , what does not occurs in the receding part of the galaxy .
this verification explains the different peak levels in the global h i emission profile shown in figure [ fig8 ] and in the rotation curves presented in figure [ fig10 ] .
taking into account that the h i global profile preserves the behavior expected for a rotating disk and different emission profiles agree with a exponential disk , then the effects of the interaction on the velocities can be considered as local . consequently the rotation curve extracted from the receding part of the velocity field is the real ( undisturbed ) ngc 6907 rotation curve . comparing the raw observed rotation curves from the receding and approaching parts of ngc 6907 in direction of ngc 6908 , it is possible to see that the relative emission velocity between these galaxies ( 130 @xmath4 17 km s@xmath1 , as measured by madore et al .
2007 ) , coincides exactly with the expected velocity for the undisturbed rotation curve , and not with the velocity measured in radio for the approaching velocity field ( where ngc 6908 can be found ) .
this apparent contradiction can be easily understood considering that the emission line velocity measured for ngc 6908 is in fact from the rotating material in the ngc 6907 disk .
ngc 6908 , as a lenticular galaxy , has less gas and new hot stars to heat it .
so the emission lines used by madore to measure the ngc 6908 velocity only could come from the material in the ngc 6907 disk .
the idea that the emission lines come from the ngc 6907 is corroborated by the spectrum extracted in direction of ngc 6908 ( figure [ fig4 ] ) .
it has a strong continuum superimposed by few emission lines , for which the radial velocity is ( 3092 @xmath4 52 ) km s@xmath1 , and by several absorption lines , associated to a radial velocity of ( 3147 @xmath4 21 ) km s@xmath1 ( table [ tbl-5 ] ) .
relative velocities spectroscopically measured in relation to the ngc 6907 center ( 3209 @xmath4 10 km s@xmath1 ) are 117 @xmath4 53 km s@xmath1 for emission lines ( compatible with the ngc 6907 rotation curve in that direction ) and ( 62 @xmath4 23 ) km s@xmath1 for absorption lines , in perfect agreement with madore et al .
( 2007 ) .
another confirmation about the origin of the emission lines is that the broaden hydrogen absorption lines ( typical of lenticular galaxies ) contain vestiges of narrow emission lines inside it ( figure [ fig5 ] ) .
these emission lines have the same radial velocity of the rotation curve of ngc 6907 at that radius .
that is the reason why different references and observations do not agree with the presence of certain hydrogen lines ( they can be hidden in lower resolutions observations ) and why @xmath68 is observed in absorption while @xmath69 in emission .
since @xmath69 line is more intense than @xmath68 the equivalent absorption line formed in a different source is not deep enough to hide it .
verifying the distribution of h i emission as a function of the velocity in the direction of ngc 6908 , it is possible to distinguish three tied gaussian components ( figure [ fig12 ] ) .
considering the relative velocities measured with respect the radio systemic velocity expected for the nucleus ( since the spectrum extracted in direction of the slit 19 does not sample the radio nucleus ) , the centroid of each component can be interpreted as associated with different structures in the interaction between ngc 6907 and 6908 .
the first one , at 3120 km s@xmath1 , is associated with ngc 6908 , since it has the same velocity of the absorption lines generated by its stellar content , the second one , at 3065 km s@xmath1 , is produced by the excited material from the disk of ngc 6907 during the interaction with ngc 6908 and the third one ( main component ) related to the higher velocity gas left behind ngc 6908 due the interaction with ngc 6907 . using the velocity contours above the approaching velocity of 220 km s@xmath1 , it is possible to trace the disturbed regions of the velocity field and see a converging distribution of isophotes over ngc 6908 , whose gradient is higher in direction of the rotation of ngc 6907 , following the spiral arms ( figure [ fig13 ] ) . in the approaching region of the ngc 6907 velocity field
overlaying the g@xmath11 image of this galaxy .
the interval between contours is 5 km.s@xmath1.,width=325 ] the striking triangular shape of the isophotes associated with high - velocity gas could be explained by the following scenario .
ngc 6908 crossed the disk of ngc 6907 in a recent past , and very probably the tidal forces created a bridge of gas between the two galaxies , similar to the magellanic stream in our galaxy .
the gas left behind ngc 6908 disk by the collision explains the excess of gas approaching us . in our galaxy , at the solar galactic radius , the time required for a gas cloud launched in the direction perpendicular to the plane to revert its motion and fall back onto it is about 35 myr ( see eg . lpine & duvert , 1994 ) .
this can be taken as an order of magnitude for the present case .
let us now consider the velocity component of ngc 6908 in the plane of the sky .
this component is directed approximately towards the center of ngc 6907 , as we can infer from the fact that presently ngc 6908 is situated closer to the center of ngc 6907 than the marks left in the gas of the disk by the crossing event . the gas stream produced by the collision
must also have a component of velocity in the same ( almost radial ) direction , which explains why the triangular isophotes point towards ngc 6908 .
finally , let us analyze the component of velocity of the stream in the plane of the sky , but in the direction of rotation of ngc 6907 .
the stream can , in a simplified way , be considered as a mixture of gas with velocities ranging from that of ngc 6908 ( with almost zero component in the direction of rotation of ngc 6907 ) to the gas which resulted to have about the same rotation velocity of the disk of ngc 6907 .
this explains the elongated shape of the isophotes in the direction of rotation , and at the same time , allows us to estimate in a new way the time elapsed since the collision .
the mean de - projected distance of the two extremes of the gas with perturbed velocity in figure [ fig13 ] is about @xmath70 arcsec , where the undisturbed rotation curve has the velocity of @xmath71 km @xmath72 . assuming @xmath73 km s@xmath1/mpc , this distance is @xmath74 kpc .
since the angle between the two extremes is @xmath75 , then the initial disturbed material would have traveled @xmath76 kpc azimuthally . from this
we derive that the collision occurred about @xmath2 years ago , which is consistent with order of magnitude previously estimated .
the combination of the gmrt radio observation and gmos / gemini optical observation provides a more complete view about the interaction between the galaxies ngc 6907 and 6908 .
photometric images observed at the gemini telescope and the h i maps observed at the gmrt reveals extended structures that can be understood as tidal debris in optical images , but resemble the continuation of the spiral arms in radio . by fitting ellipses to the last isophotes registered to each one of these observations for ngc 6907 we verified that the 21-cm emission is 2.4 times more extended than the optical emission .
the orientation of the fitted ellipses provide different parameters of projection for optical and radio observations , the differences being 15.3@xmath37 for the inclination and 24.1@xmath37 for the position angle .
geometric , photometric and kinematic centers are not coincident , reinforcing the evidences about the effect of an interaction in ngc 6907 .
the spectra observed using the gmos spectrograph are coherent with those observed in 21-cm in the same direction of the slits .
the parameters of projection in radio provides the best correspondences between the optical and radio velocities .
the details of the emission and absorption lines observed in direction of ngc 6908 show that they have different origins .
first , because the observed emission lines are not typical of lenticular galaxies like [ oii ] ( 3727 ) , [ oiii ] ( 4959 and 5007 ) and the hei ( 5876 ) , specially for those with low luminosity like ngc 6908 .
second , because the radial velocity observed in absorption is different from the velocity emission lines , which are coherent with the undisturbed rotation curve of ngc 6907 expected for that direction . finally , because inside the hydrogen absorption lines it is possible to see traces of narrower hydrogen recombination lines with a doppler effect suitable for the ngc 6907 rotating disk , confirming their different origin .
the h i column density distribution presents a discontinuity in its radial profile which can be can be related to the gas redistribution at the corotation radius .
the details of the global h i emission profile reveals that the rotation is the main component of the h i velocity distribution , and its asymmetry indicates the contribution of the ngc 6908 motion is local and in velocities near to those found in the ngc 6907 rotating disk .
the total h i mass obtained was @xmath77 @xmath50 . due to the dominant contribution to the h i emission in the spiral arms they can be the main contributors for the velocity field asymmetries , verified by the non perpendicularity of the kinematic axis ( difference of @xmath78 )
. the differences in the rotation curve extracted from the receding and approaching parts of the velocity field can be interpreted as consequence of line - of - sight components produced by the interaction with ngc 6908 .
assuming that the interaction only affected the side where ngc 6908 is observed in projection , and that the velocity field of the opposite side is the real rotation curve of ngc 6907 , we were able to get a raw estimative for the dynamical mass of @xmath79 @xmath50 .
looking at the h i spectrum in direction of ngc 6908 it is possible to distinguish three emission components : one coincident with the relative velocity for ngc 6908 , another coincident with the excited gas in ngc 6907 disk and the last one in agreement with the higher relative velocities gas left behind ngc 6908 by dynamic friction . using the difference in the rotation curves extracted from the disturbed region of the velocity field and that one away from this region , we found the limits for the disturbed material on the ngc 6907 .
supposing that this material has drifted inside the ngc 6907 disk , according the rotation curve , we estimated the time when the interaction started in @xmath80 years .
another work is in preparation to discuss the metallicity distribution in the ngc 6907 and its possible connection with the corotation radius .
we thank the staff of the gmrt that made these observations possible .
gmrt is run by the national center for radio astrophysics of the tata institute of fundamental research .
financial support for this work was provided by conselho nacional para o desenvolvimento cientfico e tecnolgico ( cnpq ) and coordenaco de aperfeioamento pessoal de nvel superior ( capes ) .
this research has been benefited by the nasa s astrophysics data system ( ads ) and extra - galactic database ( ned ) services .
their open software used in this research is greatly acknowledged .
the gemini programme i d for the data used in this paper is the gn-2005b - q-39 ( pi : s .
scarano jr . ) .
we thank c. mendes de oliveira and jayaram n. chengalur for the helpful discussions .
|
ngc 6908 , a s0 galaxy situated in direction of ngc 6907 , was only recently recognized as a distinct galaxy , instead of only a part of ngc 6907 .
we present 21 cm radio synthesis observations obtained with the gmrt and optical images and spectroscopy obtained with the gemini north telescope of this pair of interacting galaxies . from the radio observations we obtained the velocity field and the h i column density map of the whole region containing the ngc 6907/8 pair , and by means of the gemini multi - object spectroscopy we obtained high quality photometric images and @xmath0 resolution spectra sampling the two galaxies . by comparing the rotation curve of ngc 6907 obtained from the two opposite sides around the main kinematic axis
, we were able to distinguish the normal rotational velocity field from the velocity components produced by the interaction between the two galaxies .
taking into account the rotational velocity of ngc 6907 and the velocity derived from the absorption lines for ngc 6908 , we verified that the relative velocity between these systems is lower than 60 km s@xmath1 .
the emission lines observed in the direction of ngc 6908 , not typical of s0 galaxies , have the same velocity expected for the ngc 6907 rotation curve .
some emission lines are inside their equivalent absorption lines , reinforcing the idea they were not formed in ngc 6908 .
finally , the h i profile exhibits details of the interaction , showing three components : one for ngc 6908 , another for the excited gas in the ngc 6907 disk and a last one for the gas with higher relative velocities left behind ngc 6908 by dynamical friction , used to estimate the time when the interaction started in @xmath2 years ago .
galaxies : interactions galaxies : kinematics and dynamics galaxies : photometry radio lines : galaxies .
|
for a given complex manifold @xmath1 , many cohomological invariants can be defined , and many are known for compact complex surfaces . among these , one can consider _ bott - chern and aeppli cohomologies_. they are defined as follows : @xmath2 note that the identity induces natural maps @xmath3 \ar[d ] \ar[rd ] & \\
h^{\bullet,\bullet}_{{\overline{{\partial}}}}(x ) \ar[rd ] & h^{\bullet}_{dr}(x;{\mathbb{c } } ) \ar[d ] & h^{\bullet,\bullet}_{{\partial}}(x ) \ar[ld ] \\ & h^{\bullet,\bullet}_{a}(x ) & } \ ] ] where @xmath4 denotes the dolbeault cohomology and @xmath5 its conjugate , and the maps are morphisms of ( graded or bi - graded ) vector spaces . for compact khler manifolds ,
the natural map @xmath6 is an isomorphism .
assume that @xmath1 is compact .
the bott - chern and aeppli cohomologies are isomorphic to the kernel of suitable @xmath7th - order differential elliptic operators , see ( * ? ? ?
2.c ) . in particular , they are finite - dimensional vector spaces .
in fact , fixed a hermitian metric @xmath8 , its associated @xmath9-linear hodge-@xmath10-operator induces the isomorphism @xmath11 for any @xmath12 , where @xmath13 denotes the complex dimension of @xmath1 . in particular , for any @xmath12 , one has @xmath14 for the dolbeault cohomology , the frlicher inequality relates the hodge numbers and the betti numbers : for any @xmath15 , @xmath16 similarly , for bott - chern cohomology , the following inequality _ la _ frlicher has been proven in ( * ? ? ? * theorem a ) : for any @xmath17 , @xmath18 the equality in the frlicher inequality characterizes the degeneration of the frlicher spectral sequence at the first level .
this always happens for compact complex surfaces . on the other side , in (
* theorem b ) , it is proven that the equality in the inequality _ la _ frlicher for the bott - chern cohomology characterizes the validity of the _
@xmath19-lemma _ , namely , the property that every @xmath20-closed @xmath21-closed @xmath22-exact form is @xmath19-exact too , @xcite .
the validity of the @xmath19-lemma implies that the first betti number is even , which is equivalent to khlerness for compact complex surfaces .
therefore the positive integer numbers @xmath23 varying @xmath24 , measure the non - khlerness of compact complex surfaces @xmath1 .
compact complex surfaces are divided in seven classes , according to the kodaira and enriques classification , see , e.g. , @xcite . in this note , we compute bott - chern cohomology for some classes of compact complex ( non - khler ) surfaces . in particular , we are interested in studying the relations between bott - chern cohomology and de rham cohomology , looking at the injectivity of the natural map @xmath25 .
this can be intended as a weak version of the @xmath19-lemma , compare also @xcite .
more precisely , we start by proving that the non - khlerness for compact complex surfaces is encoded only in @xmath26 , namely , @xmath27 is always zero .
this gives a partial answer to a question by t. c. dinh to the third author .
* theorem [ thm : delta1=0 ] .
* _ let @xmath1 be a compact complex surface . then : _
a. the natural map @xmath28 induced by the identity is injective ; b. @xmath29 .
in particular , the non - khlerness of @xmath1 is measured by just @xmath30 . for compact complex surfaces in class @xmath0 ,
we show the following result , where we denote @xmath31 for @xmath32 .
* theorem [ thm : bc - vii ] . *
_ the bott - chern numbers of compact complex surfaces in class @xmath0 are : @xmath33 _ according to theorem [ thm : delta1=0 ] , the natural map @xmath28 is injective for any compact complex surface .
one is then interested in studying the injectivity of the natural map @xmath25 induced by the identity , at least for compact complex surfaces diffeomorphic to solvmanifolds .
in fact , by definition , the property of satisfying the @xmath19-lemma , @xcite , is equivalent to the natural map @xmath6 being injective .
note that , for a compact complex manifold of complex dimension @xmath13 , the injectivity of the map @xmath34 implies the @xmath35-th weak @xmath19-lemma in the sense of j. fu and s .- t .
yau , ( * ? ? ?
* definition 5 ) .
we then compute the bott - chern cohomology for compact complex surfaces diffeomorphic to solvmanifolds , according to the list given by k. hasegawa in @xcite , see theorem [ thm : bc - solvmfds ] .
more precisely , we prove that the cohomologies can be computed by using just left - invariant forms . furthermore , for such complex surfaces , we note that the natural map @xmath36 is injective , see theorem [ thm : inj - solvmfds ] .
we note that the above classes do not exhaust the set of compact complex non - khler surfaces , the cohomologies of elliptic surfaces being still unknown .
_ acknowledgments . _
the first and third authors would like to thank the aix - marseille university for its warm hospitality .
many thanks are due to the referee for her / his suggestions that improved the presentation .
the third author would like to dedicate this paper to the memory of his mother .
we recall that , for a compact complex manifold of complex dimension @xmath13 , for @xmath37 , we define the `` non - khlerness '' degrees , ( * ? ? ?
* theorem a ) , @xmath38 where we use the duality in @xcite giving @xmath39 . according to (
* ? ? ? * theorem b ) , @xmath40 for any @xmath15 if and only if @xmath1 satisfies the @xmath19-lemma , namely , every @xmath20-closed @xmath21-closed @xmath22-exact form is @xmath19-exact too .
in particular , for a compact complex surface @xmath1 , the condition @xmath41 is equivalent to @xmath1 being khler , the first betti number being even , @xcite , see also ( * ? ? ?
* corollaire 5.7 ) , and ( * ? ? ?
* theorem 11 ) .
we prove that @xmath27 is always zero for any compact complex surface .
in particular , a sufficient and necessary condition for compact complex surfaces to be khler is @xmath42 .
[ thm : delta1=0 ] let @xmath1 be a compact complex surface . then : a. the natural map @xmath28 induced by the identity is injective ; b. @xmath29 .
in particular , the non - khlerness of @xmath1 is measured by just @xmath30 .
a. let @xmath43 be such that @xmath44=0\in h^{2,1}_{{\overline{{\partial}}}}(x)$ ] .
let @xmath45 be such that @xmath46 . fix a hermitian metric @xmath8 on @xmath1 , and consider the hodge decomposition of @xmath47 with respect to the dolbeault laplacian @xmath48 :
let @xmath49 where @xmath50 , and @xmath51 .
therefore we have @xmath52 where we have used that any @xmath53-form is primitive and hence , by the weil identity , is self - dual . in particular , @xmath54 is @xmath19-exact , so it induces a zero class in @xmath55 .
b. on the one hand , note that @xmath56 & \subseteq & \ker{\overline{{\partial}}}\cap \wedge^{1,0}x \;=\ ; \frac{\ker{\overline{{\partial}}}\cap \wedge^{1,0}x}{\operatorname{im}{\overline{{\partial } } } } \;=\ ; h^{1,0}_{{\overline{{\partial}}}}(x ) \;. \end{aligned}\ ] ] it follows that @xmath57 where we use that the frlicher spectral sequence degenerates , hence in particular @xmath58 .
+ on the other hand , by part ( i ) , we have @xmath59 where we use the kodaira and serre duality @xmath60
. + by summing up , we get @xmath61 & \leq & 2\ , \left ( b_1 - \dim_{\mathbb{c}}h^{0,1}_{{\overline{{\partial}}}}(x ) + \dim_{\mathbb{c}}h^{0,1}_{{\overline{{\partial}}}}(x ) - b_1 \right ) \;=\ ; 0 \ ; , \end{aligned}\ ] ] concluding the proof .
in this section , we compute bott - chern cohomology for compact complex surfaces in class @xmath0 .
let @xmath1 be a compact complex surface . by theorem [ thm : delta1=0 ] ,
the natural map @xmath28 is always injective .
consider now the case when @xmath1 is in class @xmath0 . if @xmath1 is minimal , we prove that the same holds for cohomology with values in a line bundle .
we will also prove that the natural map @xmath62 is not injective .
let @xmath1 be a compact complex surface in class @xmath63 .
let @xmath64 .
the natural map @xmath65 induced by the identity is injective .
let @xmath66 be a @xmath67-exact @xmath68-form .
we need to prove that @xmath54 is @xmath69-exact too .
consider @xmath70 , where @xmath71 .
in particular , @xmath72 , hence @xmath73 defines a class in @xmath74 .
note that @xmath75 for surfaces of class @xmath63 , ( * ? ? ?
* remark 2.21 ) .
it follows that @xmath76 for some @xmath77 .
hence @xmath78 , that is , @xmath54 is @xmath69-exact .
we now compute the bott - chern cohomology of class @xmath0 surfaces .
[ thm : bc - vii ] the bott - chern numbers of compact complex surfaces in class @xmath0 are : @xmath33 it holds @xmath79 hence @xmath80 . on the other side , by theorem [ thm : delta1=0 ] , @xmath81 hence @xmath82 . similarly , it holds @xmath83 hence @xmath84 .
note that , from ( * ? ? ?
* theorem a ) , we have @xmath85 hence @xmath86 .
more precisely , from ( * ? ? ?
* theorem b ) and theorem [ thm : delta1=0 ] , we have that @xmath87 if and only if @xmath42 if and only if @xmath1 satisfies the @xmath19-lemma , in fact @xmath1 is khler , which is not the case . finally , we prove that @xmath88 .
consider the following exact sequences from ( * ? ? ?
* lemma 2.3 ) .
more precisely , the sequence @xmath89 is clearly exact .
furthermore , fix a gauduchon metric @xmath8 .
denote by @xmath90 the @xmath91-form associated to @xmath8 , where @xmath92 denotes the integrable almost - complex structure . by definition of @xmath8 being gauduchon , we have @xmath93 .
the sequence @xmath94 is exact .
indeed , firstly note that for @xmath95 , we have @xmath96 by applying twice the stokes theorem .
then , we recall the argument in ( * ? ? ?
* lemma 2.3(ii ) ) for proving that the map @xmath97 is injective .
take @xmath98 .
then @xmath99 where @xmath100 is the adjoint operator of @xmath101 with respect to @xmath102
. then @xmath103 , by extending ( * ? ? ?
* corollary 7.2.9 ) by @xmath9-linearity .
take @xmath104 such that @xmath105 .
then , by defining @xmath106 , we have @xmath107=\left[\alpha\right]\in \frac{\operatorname{im}\operatorname{d}\cap \wedge^{1,1}x}{\operatorname{im}{\partial}{\overline{{\partial}}}}$ ] , and @xmath108 , and @xmath109 where @xmath110 . in particular , @xmath111 is primitive .
since @xmath111 is primitive and of type @xmath91 , then it is anti - self - dual by the weil identity .
then @xmath112 and hence @xmath113 , and therefore @xmath114=0 $ ] .
since the space @xmath115 is finite - dimensional , being a sub - space of @xmath116 , and since the space @xmath117 is finite - dimensional , being a sub - space of @xmath118 , we get that @xmath119 and hence @xmath120 we get that @xmath121 . finally , we prove that the natural map @xmath62 is not injective .
let @xmath1 be a compact complex surface in class @xmath0 .
then the natural map @xmath62 induced by the identity is the zero map and not an isomorphism .
note that , for class @xmath0 surfaces , the pluri - genera are zero .
in particular , @xmath122 , by kodaira and serre duality . by theorem [ thm : bc - vii ] , one has @xmath123 . in this section , as an explicit example , we list the representatives of the cohomologies of a compact complex surface in class @xmath0 : namely , we consider the calabi - eckmann structure on the differentiable manifolds underlying the hopf surfaces .
consider the differentiable manifold @xmath124 .
as a lie group , @xmath125 has a global left - invariant co - frame @xmath126 such that @xmath127 and @xmath128 and @xmath129 .
hence , we consider a global left - invariant co - frame @xmath130 on @xmath1 with structure equations @xmath131 \operatorname{d}e^1 & = & - 2\ , e^2 \wedge e^3 \\[5pt ] \operatorname{d}e^2 & = & \phantom{+ } 2\ , e^1 \wedge e^3 \\[5pt ] \operatorname{d}e^3 & = & - 2\ , e^1 \wedge e^2 \end{array}\right . \;.\ ] ] consider the left - invariant almost - complex structure defined by the @xmath132-forms @xmath133 \varphi^2 & : = & e^3 + \operatorname{i}\ , f \end{array}\right . \;.\ ] ] by computing the complex structure equations , we get @xmath134 { \partial}\varphi^2 & = & 0 \end{array}\right . \qquad \text { and } \qquad \left\{\begin{array}{rcl } { \overline{{\partial}}}\varphi^1 & = & \operatorname{i}\ , \varphi^1 \wedge \bar\varphi^2 \\[5pt ] { \overline{{\partial}}}\varphi^2 & = & - \operatorname{i}\ , \varphi^1
\wedge \bar\varphi^1 \end{array}\right . \;.\ ] ] we note that the almost - complex structure is in fact integrable .
the manifold @xmath1 is a compact complex manifold not admitting khler metrics .
it is bi - holomorphic to the complex manifold @xmath135 considered by calabi and eckmann , @xcite , see ( * ? ? ?
* theorem 4.1 ) . consider the hermitian metric @xmath8 whose associated @xmath91-form is @xmath136 as for the de rham cohomology , from the knneth formula we get @xmath137 ( where , here and hereafter , we shorten , e.g. , @xmath138 ) . by ( * ? ? ?
* appendix ii , theorem 9.5 ) , one has that a model for the dolbeault cohomology is given by @xmath139 where @xmath140 is an element of bi - degree @xmath141 .
in particular , we recover that the hodge numbers @xmath142 are @xmath143 we note that the sub - complex @xmath144 is such that @xmath145 is an isomorphism .
more precisely , we get @xmath146 \right\rangle \oplus { \mathbb{c}}\left\langle \left [ \varphi^{12\bar1 } \right ] \right\rangle \oplus { \mathbb{c}}\left\langle \left [ \varphi^{12\bar1\bar2}\right ] \right\rangle \ ; , \end{aligned}\ ] ] where we have listed the harmonic representatives with respect to the dolbeault laplacian of @xmath8 . by (
* theorem 1.3 , proposition 2.2 ) , we have also @xmath147 isomorphism .
in particular , we get @xmath148 \right\rangle \oplus { \mathbb{c}}\left\langle \left[\varphi^{12\bar1}\right ] \right\rangle \oplus { \mathbb{c}}\left\langle \left[\varphi^{1\bar1\bar2}\right ] \right\rangle \oplus { \mathbb{c}}\left\langle \left[\varphi^{12\bar1\bar2}\right ] \right\rangle \ ; , \end{aligned}\ ] ] where we have listed the harmonic representatives with respect to the bott - chern laplacian of @xmath8 . by @xcite , we have @xmath149 \right\rangle \oplus { \mathbb{c}}\left\langle \left[\varphi^{\bar2}\right ]
\right\rangle \oplus { \mathbb{c}}\left\langle \left[\varphi^{2\bar2}\right ] \right\rangle \oplus { \mathbb{c}}\left\langle \left[\varphi^{12\bar1\bar2}\right ] \right\rangle \ ; , \end{aligned}\ ] ] where we have listed the harmonic representatives with respect to the aeppli laplacian of @xmath8 .
note in particular that the natural map @xmath28 induced by the identity is an isomorphism , and that the natural map @xmath150 induced by the identity is injective .
let @xmath1 be a compact complex surface diffeomorphic to a solvmanifold @xmath151 . by ( * ? ? ?
* theorem 1 ) , @xmath1 is either a complex torus , or a hyperelliptic surface , or a inoue surface of type @xmath152 , or a primary kodaira surface , or a secondary kodaira surface , or a inoue surface of type @xmath153 , and , as such , it is endowed with a left - invariant complex structure . in each case
, we recall the structure equations of the group @xmath154 , see @xcite . more precisely ,
take a basis @xmath155 of the lie algebra @xmath156 naturally associated to @xmath154 .
we have the following commutation relations , according to @xcite : a. differentiable structure underlying a _ complex torus _ : @xmath157=0 \qquad \text { for any } j , k\in\{1,2,3,4\ } \;;\ ] ] ( hereafter , we write only the non - trivial commutators ) ; b. differentiable structure underlying a _ hyperelliptic surface _ : @xmath158=e_2 \ ; , \quad \left[e_2,e_4\right]=-e_1 \;;\ ] ] c. differentiable structure underlying a _ inoue surface of type @xmath152 _ : @xmath158=-\alpha\ , e_1+\beta\ , e_2 \ ; , \quad \left[e_2,e_4\right]=-\beta\,e_1-\alpha\,e_2 \ ; , \quad \left[e_3,e_4\right]=2\alpha\,e_3 \;,\ ] ] where @xmath159 and @xmath160 ; d. differentiable structure underlying a _ primary kodaira surface _ : @xmath161=-e_3 \;;\ ] ] e. differentiable structure underlying a _ secondary kodaira surface _ : @xmath161=-e_3 \ ; , \quad \left[e_1,e_4\right]=e_2 \ ; , \quad \left[e_2,e_4\right]=-e_1 \;;\ ] ] f. differentiable structure underlying a _ inoue surface of type @xmath153 _ : @xmath162=-e_1 \ ; , \quad \left[e_2,e_4\right]=-e_2 \ ; , \quad \left[e_3,e_4\right]=e_3 \;.\ ] ] denote by @xmath163 the dual basis of @xmath164 .
we recall that , for any @xmath165 , for any @xmath166 , it holds @xmath167\right)$ ] .
hence we get the following structure equations : a. differentiable structure underlying a _ complex torus _ : @xmath168 \operatorname{d}e^2 & = & 0 \\[5pt ] \operatorname{d}e^3 & = & 0 \\[5pt ] \operatorname{d}e^4 & = & 0 \end{array } \right . \;;\ ] ] b. differentiable structure underlying a _ hyperelliptic surface _ : @xmath169 \operatorname{d}e^2 & = & - e^1 \wedge e^4 \\[5pt ] \operatorname{d}e^3 & = & 0 \\[5pt ] \operatorname{d}e^4 & = & 0 \end{array } \right . \;;\ ] ] c. differentiable structure underlying a _ inoue surface of type @xmath152 _ : @xmath170 \operatorname{d}e^2 & = & -\beta\ , e^1\wedge e^4 + \alpha\ , e^2\wedge e^4 \\[5pt ] \operatorname{d}e^3 & = & -2\alpha\ , e^3\wedge e^4 \\[5pt ] \operatorname{d}e^4 & = & 0 \end{array } \right . \;;\ ] ] d. differentiable structure underlying a _ primary kodaira surface _ : @xmath168 \operatorname{d}e^2 & = & 0 \\[5pt ] \operatorname{d}e^3 & = & e^1 \wedge e^2 \\[5pt ] \operatorname{d}e^4 & = & 0 \end{array } \right .
\;;\ ] ] e. differentiable structure underlying a _ secondary kodaira surface _ : @xmath169 \operatorname{d}e^2 & = & - e^1 \wedge e^4 \\[5pt ] \operatorname{d}e^3 & = & e^1 \wedge e^2 \\[5pt ] \operatorname{d}e^4 & = & 0 \end{array } \right .
\;;\ ] ] f. differentiable structure underlying a _ inoue surface of type @xmath153 _ : @xmath171 \operatorname{d}e^2 & = & e^2 \wedge e^4 \\[5pt ] \operatorname{d}e^3 & = & -e^3\wedge e^4 \\[5pt ] \operatorname{d}e^4 & = & 0
\end{array } \right .
\;.\ ] ] in cases ( a ) , ( b ) , ( c ) , ( d ) , ( e ) , consider the @xmath154-left - invariant almost - complex structure @xmath92 on @xmath1 defined by @xmath172 consider the @xmath154-left - invariant @xmath132-forms @xmath173 \varphi^2 \;:=\ ; e^3 + \operatorname{i}e^4 \end{array } \right . \;.\ ] ] in case ( f ) , consider the @xmath154-left - invariant almost - complex structure @xmath92 on @xmath1 defined by @xmath174 where @xmath175 .
consider the @xmath154-left - invariant @xmath132-forms @xmath176 \varphi^2 \;:=\ ; e^3 + \operatorname{i}e^4 \end{array } \right . \;.\ ] ] with respect to the @xmath154-left - invariant coframe @xmath177 for the holomorphic tangent bundle @xmath178 , we have the following structure equations .
( as for notation , we shorten , e.g. , @xmath179 . )
a. _ torus _ : @xmath180 \operatorname{d}\varphi^2 & = & 0 \end{array } \right.\ ] ] b. _ hyperelliptic surface _ : @xmath181 \operatorname{d}\varphi^2 & = & 0 \end{array } \right.\ ] ] c. _ inoue surface @xmath152 _ : @xmath182 \operatorname{d}\varphi^2 & = & -\operatorname{i}\alpha\ , \varphi^{2\bar2 } \end{array } \right.\ ] ] ( where @xmath159 and @xmath160 ) ; d. _ primary kodaira surface _ : @xmath180 \operatorname{d}\varphi^2 & = & \frac{\operatorname{i}}{2}\ , \varphi^{1\bar1 } \end{array } \right.\ ] ] e. _ secondary kodaira surface _ : @xmath181 \operatorname{d}\varphi^2 & = & \frac{\operatorname{i}}{2}\ , \varphi^{1\bar1 } \end{array } \right.\ ] ] f. _ inoue surface @xmath153 _ : @xmath183 \operatorname{d}\varphi^2 & = & \frac{1}{2\operatorname{i}}\ , \varphi^{2\bar2 } \end{array } \right . \;.\ ] ]
in this section , we compute the dolbeault and bott - chern cohomologies of the compact complex surfaces diffeomorphic to a solvmanifold .
[ thm : bc - solvmfds ] let @xmath1 be a compact complex surface diffeomorphic to a solvmanifold @xmath151 ; denote the lie algebra of @xmath154 by @xmath156 .
then the inclusion @xmath184 induces an isomorphism both in dolbeault and in bott - chern cohomologies .
in particular , the dimensions of the de rham , dolbeault , and bott - chern cohomologies and the degrees of non - khlerness are summarized in table [ table : cohom ] . in tables [ table : cohomologies-1 ] and [ table : cohomologies-2 ] and in tables [ table : cohomologies - dr-1 ] and [ table : cohomologies - dr-2 ] , we summarize the results of the computations .
the sub - complexes of left - invariant forms are depicted in figure [ fig : left - inv ] ( each dot represents a generator , vertical arrows depict the @xmath21-operator , horizontal arrows depict the @xmath20-operator , and trivial arrows are not shown . )
the dimensions are listed in table [ table : cohom ] . on the one side ,
recall that the inclusion of left - invariant forms into the space of forms induces an injective map in dolbeault and bott - chern cohomologies , see , e.g. , ( * ? ? ?
* lemma 9 ) , ( * ? ? ?
* lemma 3.6 ) . on the other side ,
recall that the frlicher spectral sequence of a compact complex surface @xmath1 degenerates at the first level , equivalently , the equalities @xmath185 and @xmath186 hold . by comparing the dimensions in table [ table : cohom ] with the betti numbers case by case , we find that the left - invariant forms suffice in computing the dolbeault cohomology for each case .
then , by ( * ? ? ? * theorem 3.7 ) , see also ( * ? ? ?
* theorem 1.3 , theorem 1.6 ) , it follows that also the bott - chern cohomology is computed using just left - invariant forms .
[ thm : inj - solvmfds ] let @xmath1 be a compact complex surface diffeomorphic to a solvmanifold . then the natural map @xmath187 induced by the identity is an isomorphism , and the natural map @xmath36 induced by the identity is injective . by the general result in theorem [ thm : delta1=0 ]
, the natural map @xmath187 is injective .
in fact , it is an isomorphism as follows from the computations summarized in tables [ table : cohomologies-1 ] and [ table : cohomologies-2 ] . as for the injectivity of the natural map @xmath36 ,
it is a straightforward computation from tables [ table : cohomologies-1 ] and [ table : cohomologies-2 ] and tables [ table : cohomologies - dr-1 ] and [ table : cohomologies - dr-2 ] . as an example
, we offer an explicit calculation of the injectivity of the map @xmath36 for the inoue surfaces of type @xmath188 , see @xcite , see also @xcite .
we will change a little bit the notation . recall the construction of inoue surfaces :
let @xmath189 be a unimodular matrix having a real eigenvalue @xmath190 and two complex eigenvalues @xmath191 .
take a real eigenvector @xmath192 and an eigenvector @xmath193 of @xmath194 .
let @xmath195 ; on the product @xmath196 consider the following transformations defined as @xmath197 denote by @xmath198 the group generated by @xmath199 ; then @xmath198 acts in a properly discontinuous way and without fixed points on @xmath196 , and @xmath200 is an inoue surface of type @xmath188 , as in case ( c ) in @xcite . denoting by @xmath201 and @xmath202 ,
consider the following differential forms on @xmath203 : @xmath204 ( note that @xmath205 and @xmath206 , and @xmath207 are @xmath198-invariant , and consequently they induce global differential forms on @xmath208 . )
we obtain @xmath209 consider the natural complex structure on @xmath208 induced by @xmath210 .
locally , we have @xmath211 considering the @xmath198-invariant @xmath68-bott - chern cohomology of @xmath208 , we obtain that @xmath212 \right\rangle \;.\ ] ] clearly @xmath213 and @xmath214 , therefore the de rham cohomology class @xmath215 = \left [ e^1 \wedge e^3 \wedge e^4 \right ] \in h^3_{dr}(\mathcal{s}_{m})$ ] is non - zero .
f. hirzebruch , _ topological methods in algebraic geometry _ , translated from the german and appendix one by r. l. e. schwarzenberger , with a preface to the third english edition by the author and schwarzenberger , appendix two by a. borel , reprint of the 1978 edition , classics in mathematics , springer - verlag , berlin , 1995 .
|
we study bott - chern cohomology on compact complex non - khler surfaces .
in particular , we compute such a cohomology for compact complex surfaces in class @xmath0 and for compact complex surfaces diffeomorphic to solvmanifolds .
|
in 1943 , luria and delbrck @xcite ( ld hereafter ) used a simple evolutionary model together with laboratory experiments to argue that bacterial resistance ( to bacteriophage , in their specific case ) arose via the selection of pre - existing mutations .
their model postulated a constant probability at which dividing normal cells would have a resistant daughter ; these mutations were assumed to not affect the growth rate until the bacteria were challenged . the ld model and various extensions thereof have been used extensively in recent years to study the emergence of antibiotic - resistant microbes @xcite as well as chemotherapy - resistant cancer cells @xcite
. similar models are used to study the generation of differentiated cells from stem - like precursors @xcite .
there has been considerable effort in the mathematics and statistical physics communities devoted to understanding the probability distribution that arises from the ld process @xcite . while early efforts focused on various quasi - deterministic versions of the model where both wild - type and mutant cells or alternatively just the wild - type are assumed to undergo continuous exponential growth , more recently interest has focussed on the fully stochastic model .
a major achievement has very recently been achieved by antal and krapivsky @xcite who solve for the generating function of the full joint distribution of the number of wild - types and mutants as a function of time , starting from a single wild - type individual .
also noteworthy is the solution by angerer @xcite for the probability distribution of the number of mutants given a fixed total population , @xmath0 , in the case of the symmetric birth - only version of the stochastic model , where both wild - type and mutant have identical birth rates and cell death is excluded .
however , it has proven difficult to obtain much physical insight from these solutions , and to relate these solutions to those of the simpler quasi - deterministic models . in particular , the remarkable fact that the distribution is these simpler models
becomes @xcite a universal @xmath2-stable ( lvy ) distribution in the limit of small mutation rate @xmath3 , and large @xmath0 such that @xmath4 has had no known analogue in the fully stochastic models . here
, we introduce a new strategy to solve this model , taking advantage of the fact that for most cases of interest the population size @xmath0 is very large and the mutation rate @xmath3 is very small .
for example , a tumor may have upwards of @xmath5 cells when it is first detected and the mutation rate , even if elevated above normal human mutation rates by genomic instability , would still be @xmath6 or smaller ( depending on exactly what range of mutations are assumed to be able to confer resistance ) .
note that @xmath7 is in general large , and hence a perturbative treatment would not be accurate .
we show that if one focuses attention on the fixed @xmath0 ensemble , one can derive an alternative generating function equation which leads immediately to a simple integral expression for the probability distribution .
one can recover the asymptotic properties of this distribution for the general case where the resistant population may have a different fitness than the wild - type .
the outline of this work is as follows . to illustrate our methodology and to lay out the general features of the ld process , in particular that it gives rise to a distribution @xmath8 for the number @xmath9 of mutants
which is heavy - tailed , we first present the special case of zero death rate .
we show the general features of the quasi - deterministic model , including the fact that at large @xmath7 one obtains a universal @xmath2-stable distribution is actually valid for the fully stochastic version as well .
next , we consider cell death in addition to division and mutation , and show that even for this case we recover for large @xmath7 the limiting landau distribution ( for equal birth rates ) and its generalization for the asymmetric case .
finally , we compare our results to those of antal and krapivsky , who formally solve the same problem with a fixed time ensemble .
some of our results have appeared previously in abbreviated form @xcite .
to illustrate our basic ideas , we start with the relatively simple case of no cell death ( i.e. a pure birth / mutation process ) and with equal birth rates for both the wild - type and mutant subpopulations . if we start from a fixed number @xmath10 of initial wild - type cells and wish to study the ensemble at fixed total population @xmath0 , the number of events is just @xmath11 . rather than computing the probability function at fixed time , we instead use the event number as the independent variable .
then , @xmath12 is just the probability that in the @xmath11 births since the beginning , there were no mutations , thus @xmath13 the master equation for @xmath14 , @xmath15 reads @xmath16 .\end{aligned}\ ] ] to derive what we will refer to as the scaling solution for the probability distribution , we will keep the leading order in @xmath0 term for each order in @xmath3 . later , we will see exactly when this is valid .
it is easy to check that this involves keeping the coefficient of @xmath17 for terms proportional to @xmath18 .
let us denote this coefficient in @xmath19 as @xmath20 , so that @xmath21 . substituting this term into the master equation
, we can check that the leading order in @xmath0 condition ( the @xmath22 term is automatically satisfied , but matching the next order in @xmath0 term ( proportional to @xmath23 ) leads to the non - trivial condition for @xmath24 : @xmath25 with @xmath26 the former condition comes from our exact solution of @xmath12 , eq .
( [ pn0 ] ) , above and the latter from the fact that @xmath27 for all @xmath15 vanishes if @xmath28 . to solve this equation , we define the generating function @xmath29 in terms of which the generating function , @xmath30 , of the probability distribution is given by @xmath31 from the recursion relation eq .
( [ scaling - master ] ) we easily obtain the differential equation @xmath32f(x , y ) = -y(1-x)f + xye^{-y } .\ ] ] as already discussed in @xcite , we can solve this equation by first converting it to a homogeneous equation via the transformation @xmath33 , so that @xmath32g(x , y ) = -y(1-x)g m , \ ] ] with @xmath34 .
the method of characteristics tells us to write @xmath35 , choosing @xmath36 , so that the characteristic is @xmath37 where @xmath2 labels the characteristic
. then the equation reads @xmath38 with the solution @xmath39 so that @xmath40 and thus @xmath41 this result was derived initially by lea and coulson @xcite , for a model which neglected the stochastic nature of the birth - death dynamics for the wild - type population . in our derivation , going to the scaling limit of @xmath0 large , @xmath42 , has resulted in a distribution which has no @xmath10 dependence .
since when @xmath10 is large , the wild - type stochasticity is negligible , the fully stochastic treatment has to reduce to the lea - coulson result in the scaling limit .
one can numerically evaluate the lea - coulson distribution from the generating function via a contour integration to pick out @xmath27 : @xmath43 this is more conveniently computed by moving the contour to hug the branch cut along the real axis @xmath44 : @xmath45 a computation of this for the case @xmath46 is shown in fig .
[ lcasym ] , ( labelled @xmath47 , together with results of a direct numerical solution of the master equation for the parameters @xmath48 , @xmath49 . the agreement is excellent even for this not so tiny value of @xmath3 . , large @xmath0 approximate distribution for @xmath46 for the pure birth - model , with the ratio of wild - type to mutant birth rate @xmath50 , @xmath51 , @xmath52 and @xmath53 ( solid lines ) .
this is shown together with the results ( filled circles ) of a direct numerical solution of the master equation for the parameters @xmath48 , @xmath49.,scaledwidth=70.0% ] it is clear by looking at the small @xmath54 behavior that eq .
( [ lc ] ) correctly reproduces @xmath55 . in the small @xmath7 limit , @xmath56 so that for @xmath57 @xmath58 which is monotonically decreasing and has a fat - tailed @xmath59 decay .
this @xmath59 behavior at large @xmath9 is in fact true for all @xmath7 , and is a direct result of the logarithmic singularity of @xmath30 at @xmath60 .
the large @xmath7 limit is also tractable , for @xmath9 s of order @xmath7 .
we write @xmath61 , and write @xmath62 , giving @xmath63 this is just a landau distribution , the one - sided @xmath2-stable ( lvy ) distribution with index @xmath64 , written in terms of the original variables as @xmath65 times the normalization factor @xmath66 .
the landau distribution rises very rapidly from zero , reaches it maximum and then falls with a @xmath59 fat tail .
the appearance of the @xmath2-stable distribution in the quasi - deterministic model @xcite is a consequence of the fact that the total number of mutants is the sum of the mutants descended from each mutation event , and from large @xmath0 these are independent and identically distributed with a power - law tail , and thus the conditions of the generalized central - limit theorem are satisfied .
the landau distribution arises in the fully stochastic model for the same reasons ; the only difference being an overall shift in the distribution .
thus , the lea - coulson distribution is seen to interpolate between the monotonic @xmath67 distribution at small @xmath7 to the fat - tailed one - sided @xmath2-stable distribution with its sharp rise for small @xmath9 and @xmath59 falloff at large @xmath9 .
a signpost value is @xmath46 , above which @xmath27 is no longer monotonic .
the position of the maximum moves rightward with increasing @xmath7 , asymptotically behaving as @xmath68 .
this analysis is summarized in fig .
[ figlc ] . as given by the lea - coulson formula for @xmath69 , @xmath70 and @xmath71 .
in addition , the small @xmath7 limiting form @xmath72 and the large @xmath7 landau distribution are also shown .
it is seen that the landau distribution correctly describes the distribution except for @xmath9 s of order unity.,scaledwidth=70.0% ] for @xmath9 of order @xmath0 , the lea - coulson result breaks down because we have neglected terms that become of comparable order in this new regime .
this outer regime @xcite describes how the distribution must drop to zero as we approach @xmath73 ; after all , the most extreme possibility is that every division resulted in adding one more mutant .
since @xmath3 is small , this part of the distribution arises when one of the first births lead to a mutant and hence the mutant sub - population has as much chance to expand exponentially as does the wild - type . to describe this scenario ,
one need only keep terms linear in @xmath3 .
this can be done explicitly in the symmetric birth - only model @xcite . however , even this problem is difficult to solve exactly in the general case , and is only of mathematical interest , since this outer part of the distribution describes extremely rare events and have essentially no physical relevance .
we now extend the previous results to account for a fitness ( aka birth rate ) difference between the wild - type and the resistant subpopulations .
normally it would be exacted that alleles conferring resistance are less fit in the absence of any drug , and hence we expect the division rate ratio @xmath74 of wild - type to resistant cell to be greater than one .
the master equation connecting the mutant number distribution at one overall population size to the next overall population size , must take into account the relative probability that it is a wild - type cell that divides versus that it is a mutant cell which divides .
if the current population has @xmath9 mutant cells and hence @xmath75 wild - type ones , this relative probability is @xmath76 .
then the equation clearly becomes @xmath77 .\end{aligned}\ ] ] again , we solve this equation in the limit of large @xmath0 , small @xmath3 by keeping only the leading term in @xmath0 for each poet of @xmath3 , again using the expansion eq .
( [ expan ] ) . substituting this into the master equation
, we obtain a generalization of eq .
( [ scaling - master ] ) : @xmath78 following the same procedure as before , we transform this into an equation for the generating function , @xmath79f(x , y ) = -y(1-x)f(x , y ) + xye^{-y } .\ ] ] again , the inhomogeneous solution is @xmath80 , and the homogeneous solution satisfies @xmath79g(x , y ) = -y(1-x)g(x , y ) , \ ] ] with @xmath81 .
again , we solve by the method of characteristics , with the characteristic labelled by @xmath2 given by @xmath82 so that @xmath83 with the solution @xmath84 where @xmath85 is a hypergeometric function . substituting @xmath86
, we get @xmath87 where we have employed a linear transformation identity , so that finally @xmath88 again , to evaluate @xmath89 we integrate around the branch cut , which again extends rightward from @xmath60 .
using standard identities involving hypergeometric functions , we have @xmath90 this allows us to evaluate the discontinuity along the branch cut , giving @xmath91 again , fig .
[ lcasym ] presents a comparison of this expression with results of a direct numerical solution of the master equation for the parameters @xmath48 , @xmath49 .
the agreement is again excellent .
the non - analyticity at @xmath60 as before governs the large @xmath9 asymptotics ( subject to @xmath92 , as usual ) .
to investigate the behavior near @xmath60 , we use eq .
( [ fr ] ) to obtain ( @xmath93 ) @xmath94 by standard tauberian theorems , this corresponds to a @xmath95 decay at large @xmath9 , independent at @xmath7 . in particular , for small @xmath7 , we have @xmath96 so that for @xmath57 : @xmath97{\displaystyle \to } { } \mu n \frac{r\gamma(1+r)}{m^{1+r } } .\ ] ] for large @xmath7 , as we saw in the case of @xmath50 , the contour integral giving @xmath27 is dominated by the @xmath98 singularity of the integrand . thus , @xmath99\right ) .
\label{pnmr}\ ] ] this is nothing but the one - sided @xmath2-stable distribution , now with index @xmath100 , consistent with the power - law decay . to see this , note that the characteristic function , @xmath101 ( the fourier transform of the probability density ) of the @xmath2-stable distribution with index @xmath74 has the form @xmath102 , \ ] ] where @xmath103 is called the shift parameter , which is just the mean of the distribution , and @xmath104 is the scale parameter . by rotating @xmath105 in eq .
( [ pnmr ] ) , we get for the shift parameter @xmath106 which is indeed the mean of the distribution eq .
( [ smallmur ] ) .
for the scale parameter , we get @xmath107^{1/r } = \left [ \mu n \frac{r \pi}{2\sin \pi r/2 } \right]^{1/r } .\ ] ]
we now study the general asymmetric ld model with death , so that each event is either a birth event with rates @xmath108 , @xmath109 or a death event with rates @xmath110 , @xmath111 .
we start from the probability distribution as a function of time . the master equations for @xmath112 , where @xmath0 denotes the total ( w.t .
+ mutated ) population and @xmath9 is the number of mutants , are : @xmath113p(n , m ) ; \qquad\qquad ( n>1),\end{aligned}\ ] ] with @xmath114 at @xmath115 . to simplify these expressions
, we make use of the fact that we are interested in a fixed value of @xmath0 , independent of how much time has elapsed since the beginning . from the biology perspective , what one knows ( ideally ) is the size of the tumor at discovery , with no idea how long the tumor has been growing .
thus , we are interested in the conditional probabilities @xmath116 integrating the master equation over time leads to the following equation for @xmath117 : @xmath118t_n(m ) - \delta_{n , n_0}\delta_{m,0 } \nonumber\\ & = b_w ( n-1-m)(1-\mu ) t_{n-1}(m)\nonumber\\ & \quad { } + b_w \mu ( n - m ) t_{n-1}(m-1 ) + b_m ( m-1)t_{n-1}(m-1)\nonumber\\ & \quad { } + d_w ( n+1-m)t_{n+1}(m ) + d_m ( m+1 ) t_{n+1}(m+1 ) .
\label{teq}\end{aligned}\ ] ] with the definition @xmath119 .
again we assume @xmath120 .
then , plugging this into the recursion relation , and noting that as always the @xmath17 term cancels , we find the generalized scaling - limit recursion @xmath121 & = b_w a_{m-1}^{k-1 } - b_w a_m^{k-1 } + b_m(m-1)a_{m-1}^k \nonumber\\ & \qquad{}+ d_m(m+1)a_{m+1}^k .\end{aligned}\ ] ] by now , the procedure for solving this equation should be obvious .
we rewrite this systems as an equation for the generating function @xmath122f(x , y ) = -b_wy(1-x)f + { \cal{i } } , \ ] ] where @xmath123 is the inhomogeneous term . again
@xmath60 is the singular point with the coefficient of @xmath124 vanishes .
defining @xmath125 , the ratio of _ net _ birth rates , the characteristic is given by @xmath126 the homogeneous equation now reads @xmath127 notice that except for factors , this is the exact same equation we encountered in the asymmetric pure - birth case , and so the solution is immediate : @xmath128 where @xmath129 is a normalization prefactor .
our desired conditional probability generating function , which we denote by @xmath30 is then @xmath130 we see that in the pure birth case , @xmath131 , this reproduces the results of the previous section
. the large @xmath9 behavior is , as we have seen , determined by the behavior in the vicinity of @xmath60 .
writing @xmath132 , we have @xmath133 the analytic structure is identical to the asymmetric pure - birth case , and so @xmath134 again decays as @xmath95 : @xmath135 though the exponent is only a function of the the ratio of the _ net _ birth rates , the coefficient is more complicated .
we can again evaluate @xmath134 from our generating function via an integration along the branch cut : @xmath136 we present in fig .
[ bdfig ] such a calculation for the case @xmath137 , @xmath138 , @xmath139 , @xmath140 , with @xmath141 .
also shown is a direct numerical solution of the master equation , eq .
( [ teq ] ) , for @xmath48 , @xmath49 , which confirms our asymptotic analysis . for comparison ,
the pure birth distribution with the same @xmath141 is shown .
for the asymmetric birth - death model , together with a direct solution of the master equation , eq .
( [ teq ] ) for @xmath48 , @xmath49 .
here @xmath137 , @xmath138 , @xmath139 , @xmath140 , with @xmath141 . the asymptotic power law , eq .
( [ bdbigm ] ) is also shown .
for comparison , the pure birth distribution with the same @xmath141 is shown . , scaledwidth=70.0% ] for small @xmath7 , we can write @xmath134 explicitly , since @xmath142\nonumber\\ & = 1 + \mu n \frac{b_w}{rb_m}\big[-{}_2f_1\left(1,r;1+r;\frac{d_m}{b_m}\right ) \nonumber\\ & \qquad\qquad\qquad\qquad{}+ \sum_{m=1}^\infty\frac{r^2\gamma(r)\gamma(m)}{\gamma(r+m+1)}\,{}_2f_1\left(r , m;r+m+1;\frac{d_m}{b_m}\right ) x^m\big ] , \end{aligned}\ ] ] so that , for @xmath15 , @xmath143 and @xmath144 it is straightforward , though tedious , to show that @xmath145 is an _
exact _ solution of the small @xmath3 version of eq .
( [ teq ] ) ( using the fact that to zeroth order in @xmath3 , @xmath146 : @xmath147t_n(m ) \nonumber\\ & = b_w ( n-1-m ) t_{n-1}(m ) + b_m ( m-1)t_{n-1}(m-1)\nonumber\\ & \quad { } + d_w ( n+1-m)t_{n+1}(m ) + d_m ( m+1 ) t_{n+1}(m+1 ) , \nonumber \\
[ ( b_w + d_w)(n-1 ) & + ( b_m + d_m)]t_n(1 ) \nonumber\\ & = b_w ( n-2 ) t_{n-1}(1 ) + d_w nt_{n+1}(1 ) + 2 d_m
t_{n+1}(2 ) + \mu , \nonumber\\ [ ( b_w + d_w)n]t_n(0 ) & = b_w ( n-1 ) t_{n-1}(0)+ d_w ( n+1)t_{n+1}(0 ) \nonumber\\ & \qquad{}+ d_m t_{n+1}(1 ) - \mu , \end{aligned}\ ] ] with @xmath148 .
however , the finite @xmath0 system is truncated , with the equation for @xmath149 being different , so that this solution does not represent an exact solution of this truncated system .
nevertheless , it turns out that @xmath150 and @xmath149 decouple from the other equations , and in fact @xmath151 _ is _ the exact solution of the @xmath152 small @xmath3 problem for @xmath153 . the perspicacious reader will have noticed that the problem we are solving in this section does not reduce exactly to the problem we posed when we considered the pure - birth asymmetric model .
the equations are the same with the identification that @xmath154 in our section maps to @xmath155 in that section .
however , the probabilities we construct from the @xmath156 do not correspond to the @xmath157 s in the previous section . here
, we weight each state by the time spent in the state , which we did not do there . to leading order in @xmath0
, however , the time spent in a state is independent of the state ( as long as @xmath42 ) , so the large-@xmath0 solution is identical in the two ensembles .
note that when @xmath158 , we get @xmath159 which is equivalent to a well - known result of ref .
the advantage of our approach is that it is clear what approximations have to be made to obtain this result , something that is not overly clear in this earlier reference .
it should also be noted in this context that the result obtained in ref .
@xcite for @xmath134 in the small @xmath7 limit is not correct .
we can use eq .
( [ gensmallt ] ) to find the large @xmath7 , @xmath160 limiting distribution , where the branch cut integral is as usual dominated by the singular point at @xmath161 .
given that the analytic structure here , as we have seen , is exactly the same as for the asymmetric pure - birth case , we immediately can conclude that the distribution for large @xmath7 reproduces the one - sided @xmath2-stable distribution with index @xmath74 .
the shift parameter is @xmath162 that this is correct can be seen from the calculation of @xmath163 in the original dl model : @xmath164 for the scale parameter , we get @xmath165^{1/b } \nonumber\\ & = \left(\frac{b_m}{b_m - d_m}\right)^{1 - 1/b } \left [ \mu n \left(\frac{b_w}{b_w - d_w}\right)\frac{b \pi}{2\sin\frac { \pi b}{2 } } \right]^{1/b } .\end{aligned}\ ] ] lastly , for completeness , we write down the results for the symmetric birth - death process . here
@xmath50 and the large @xmath7 limit is thus again a landau distribution .
for @xmath166 we have @xmath167 for the large @xmath7 limit , we have @xmath168 which is again a landau distribution , @xmath169 .
amazingly , introducing death to the ld problem hardly makes a change in the answer , just renormalizing the variable @xmath9 by the ratio @xmath170 .
the most general solution of ld problem is given in the recent paper by antal and krapivsky ( ak ) @xcite , who solve the problem of determining the probability distribution at a fixed time .
we first present their solution , modifying their notation slightly so as not to interfere destructively with the notation we have been using until now .
they write down the backward version of the master equation for the probabilities of @xmath171 wild - types and @xmath9 mutants in the form @xmath172 in this expression the division and death rates of the wild - type ( @xmath171 ) and mutant ( @xmath9 ) subpopulations are denoted by @xmath173 and @xmath174 with the proper subscript .
the superscript denotes a process which starts from a single wild - type individual ( superscript @xmath171 ) , a single mutant individual ( superscript @xmath9 ) , two wild - type ( @xmath175 ) and one of each ( @xmath176 ) .
a great simplification occurs in passing to the generating function version , in that the independent time evolution of any two individuals leads to factorization of the two - individual probabilities , so that the generating function for @xmath177 is the square of the generating function for @xmath178 , with similar results for @xmath179 and @xmath177 .
this then leads to a pair of coupled ode s for @xmath180 and @xmath181 , the generating functions defined as @xmath182 and the equations are @xmath183 to see how the solution of this system works and how it relates to our approach , we return for the moment to the simpler case of no death and symmetric division @xmath184 , and recall that the ak derivation so far is for @xmath185 .
we can rewrite these equations by picking the unit of time to set the wild - type growth rate @xmath186 equal to 1 .
this leads to the new system @xmath187 with @xmath188 , @xmath189 .
the initial condition for the @xmath190 equation is that @xmath191 , since by definition the state defining @xmath190 starts with one mutant cell .
the solution of this equation is @xmath192 substituting this into the @xmath193 equation , one obtains the solution @xmath194 the constant of integration @xmath195 is picked to satisfy the initial condition @xmath196 , yielding @xmath197 from this , one can obtain directly the distribution of mutant numbers , independent of the number of wild - type , given by @xmath198 ; setting @xmath161 gives @xmath199 and using the mean population size @xmath200 , we obtain @xmath201 } .\ ] ] to compare this result with our finding given above in eq .
[ lea - coulson ] , we go to the scaling limit , which is small @xmath3 , large @xmath202 , finite @xmath9 ( so that the number of wild - type is approximately equal to the total population ) .
we can then drop the @xmath203 term , and we have @xmath204 so that indeed @xmath205 is the relevant control parameter .
this is _ not _ the lea - coulson generating function , as has already been noted in our abbreviated initial paper @xcite .
in particular , @xmath206 , so that it does not decay exponentially with @xmath205 as we found above .
the difference lies in the fact that the ak calculation is performed for _ fixed time _
, not at _ fixed total population @xmath0 _ as before . for large time
, the distribution of the number of wild - type cells , independent of the number of mutants can be shown to be geometric : @xmath207 from this , given that the probability of zero mutants given @xmath171 wild - type cells is , for small @xmath3 , @xmath208 , it immediately follows that the probability of zero mutants is given by @xmath209 in agreement with the ak calculation . clearly , then
, the configurations at fixed time with small numbers of wild - types and thus a relatively large probability of never having spawned a resistant cell increase @xmath210 drastically . in fig .
[ figkrap ] we present the distribution resulting from the ak fixed time ensemble , for times corresponding to @xmath211 , 1 and @xmath71 .
we see that in all cases the distribution is monotonic , and looks nothing like the lea - coulson result , nor does it go over to the landau distribution at large @xmath205 .
, in the ak fixed time ensemble , for times corresponding to @xmath212 , 1 and @xmath71.,scaledwidth=70.0% ] looking at @xmath213 for small @xmath205 , we see that it does reproduce the small @xmath7 limit of the lea - coulson answer . more generally , the structure of the singularity at @xmath60 is the same of for lea - coulson , so that fixed - time ensemble does show the same @xmath59 behavior as our fixed-@xmath0 ensemble for the intermediate asymptotic regime @xmath214 .
the behavior in the cut - off regime @xmath215 is not surprisingly very different from the fixed-@xmath0 behavior , since the shift of the pole of @xmath213 to @xmath216 means that the decay of @xmath8 is exponential for @xmath215 of course , the full ak solution contains information about the wild - type distribution encoded in its dependence on @xmath217 which has been dropped in the derivation of @xmath218 . if we return to the full solution for @xmath193 , we can of course restrict ourself to a fixed value of @xmath219 by finding the coefficient of the term that goes like @xmath220 .
this is still not exactly the same problem we solved .
our initial ensemble allows the elapsed time to vary , whereas this new result fixes both the time and the total population .
however , the distribution of mutants as we have seen is controlled by the total population , and we do not expect the time taken to reach this population to matter significantly , at least as long as it is close to the average time . further , as we have already noted , for @xmath42 , the total population @xmath0 and the wild - type population @xmath171 are the same to leading order in @xmath0 .
thus , undaunted , we continue .
the projection of @xmath180 onto fixed @xmath219 , which we denote @xmath221 can be done exactly , since the @xmath217 dependence of @xmath193 is simple and takes the form @xmath222 with @xmath223^\mu}{1-x } , \qquad a_3=-\frac{x[(1-z)(1-x)]^\mu}{1-x},\ ] ] and where as before @xmath224 .
thus , @xmath225 where @xmath129 is a normalization factor , chosen so that @xmath226 .
the non - trivial normalization factor arises because we do not know a priori the overall probability of finding exactly the correct value of @xmath0 in the overall ensemble , and hence this must be divided out to obtain a meaningful probability distribution .
now , in the limit @xmath227 , @xmath228 thus , we can find the value of the normalization constant @xmath229 given this , we can now calculate the leading large-@xmath202 behavior of @xmath230 , which we wish to compare to @xmath213 . for fixed @xmath54 and large @xmath202 , @xmath231 and @xmath232^{1-\mu}}{x } - \frac{1}{x } \approx-1 + \frac{1}{{{{\mkern 1.5mu\overline{\mkern-1.5mun\mkern-1.5mu}\mkern 1.5mu } } } } - \mu\frac{(1-x)\ln(1-x)}{x } .\ ] ] also , @xmath233^\mu } \approx \frac{1}{n } .\ ] ] .
finally we obtain @xmath234 this is just the lea - coulson result that we had obtained above .
thus , our expectations that post - projecting the fixed - time ensemble for fixed wild - type abundance would reproduce the fixed population ensemble result in the intermediate asymptotic regime are indeed realized .
the above calculation was for @xmath152 . for general @xmath10 , the ak result for the generating function
@xmath235 is immediate : @xmath236^{n_0 } , \ ] ] due to the independence of the life histories of the initial @xmath10 wild - type .
clearly then , the ak result , i.e. conditioning on fixed time , is @xmath10 dependent , as opposed to the @xmath10 independence of the fixed total population result , for @xmath92 .
we can generalize our calculation of the fixed @xmath219 projection of the ak result to general @xmath237 .
now , we have @xmath238 . then , performing the taylor expansion of @xmath239 we get @xmath240 the normalization factor is now @xmath241^{n_0-{{{\mkern 1.5mu\overline{\mkern-1.5mun\mkern-1.5mu}\mkern 1.5mu}}}}\nonumber\\ & \approx { { { \mkern 1.5mu\overline{\mkern-1.5mun\mkern-1.5mu}\mkern 1.5mu}}}\left(\frac{e}{n_0}\right)^{n_0}(n_0 - 1 ) ! .\end{aligned}\ ] ]
then , @xmath242 recovering , as we expect , the @xmath10-independent lea - coulson result .
it is remarkable that the fixed population ensemble results are much more similar to the original semi - deterministic model of luria and delbrck , with its universal aspects such as @xmath10 independence and large @xmath7 @xmath2-stable distribution than is the fixed time ensemble .
we chose to first present our projection method in the simplest possible context of the pure - birth , symmetric model , as in this case the ak solution is expressible in terms of elementary functions and the structure of the calculation is most transparent . the procedure , however , works in essentially the same way for the most general asymmetric birth - death model for which ak presented their solution for the fixed time ensemble . the model is now parameterized in terms of wild - type and mutant birth rates , @xmath108 , @xmath109 as well as wild - type and mutant death rates , @xmath110 , @xmath111 .
ak choose to scale time so that the non - mutating wild - type birth process @xmath243 has rate unity .
ak then introduce the following rates : @xmath244 ak also introduce the auxiliary parameter @xmath245 note that in the small @xmath3 limit , @xmath246 is of order @xmath3 : @xmath247 ak define as well @xmath248 where the approximations are correct to leading order in @xmath3 .
then , the full solution derived in ak takes the extremely complicated form @xmath249 where @xmath250 \nonumber\\ & \qquad{}- z_0^c z(c-1)f_2(z)\left[f_1(z_0)(1+\omega ) + \lambda_m f_3(z_0)z_0\right ] \nonumber\\ & \ { } \qquad + z_0^c z^2 f_4(z ) \left [ f_1(z_0)(1+\omega ) + \lambda_m z_0 f_3(z_0)\right ] , \nonumber \\ t_3 & = z^c z_0 f_1(z ) f_2(z_0 ) - z_0^c z f_2(z ) f_1(z_0 ) , \nonumber\\ t_4 & = z^c z_0 f_1(z ) \left[f_2(z_0)((c-1)\lambda_m - 1 - \omega ) - z_0 \lambda_m f_4(z_0)\right]\nonumber\\ & \qquad { } + z z_0^c f_2(z ) \left[f_1(z_0)(1+\omega ) + \lambda_m z_0 f_3(z_0)\right ] , \end{aligned}\ ] ] with the function definitions @xmath251 and @xmath252 , \qquad z_0 = { { \mkern 1.5mu\overline{\mkern-1.5mum\mkern-1.5mu}\mkern 1.5mu}}z .\ ] ] here , @xmath253 and is related to @xmath202 , the average number of wild - type cells , by @xmath254 then , @xmath230 is given by @xmath255 for the calculation of the @xmath256 factor , it is sufficient to work to zeroth order in @xmath3 . to this order @xmath257 and @xmath258 .
then , we have @xmath259\left [ z^c z_0 ( b\lambda_m - 1 ) + z z_0^c\right ] - \left[-bz_0^c z \right]\left[z^cz_0-z_0^c z\right ] \nonumber\\ & = b^2\lambda_m z_0^{c+1 } z^{c+1 } .\end{aligned}\ ] ] for the @xmath260 term in the denominator , things are even simpler since we can use the fact that @xmath261 : @xmath262 so that @xmath263 for the calculation of @xmath264 we have to keep both the first @xmath1 and @xmath3 corrections since the ratio is to be raised to the @xmath202th power .
we have , since @xmath265 @xmath266 \equiv 1 + \mu f_1 ^ 1(x ) , \end{aligned}\ ] ] and @xmath267 and so @xmath268 .\ ] ] similarly , @xmath269 .\ ] ] thus @xmath270 .\ ] ] thus , we have our result @xmath271 since , when @xmath227 , @xmath272 and @xmath273 , the condition @xmath226 implies @xmath274 this recovers the result we obtained above , as expected .
it is amazing how our relatively simple scaling distribution is encoded in the immensely more complicated ak formalism .
being exact , their equation must contain all the details regarding the outer region ( when @xmath9 is the same order as @xmath0 ) , must allow for arbitrary values of @xmath3 and @xmath0 , and must reflect the intricate interplay of time and population size .
our approach is much less general , but has the distinct advantage of leading to answers that are simple enough for us to gain some insight into what is going on . in the last part of this section
, we will see that we can employ a trick to recover our results from the ak formalism in a much simpler fashion .
there is an alternate way to extract our desired fixed population distribution from the ak results .
we have seen that in the regime of interest , @xmath92 , the distribution is independent of @xmath10 .
if we take @xmath10 large however , the wild - type dynamics becomes deterministic and so the fixed time and fixed population ensembles become identical .
thus , all we need to do is to take the ak results for large @xmath10 , and this gives the desired generating function directly .
we now proceed to carry this out for the general non - symmetric case . following the notation in ak ,
we focus on @xmath275 with the definition @xmath276 where the @xmath277 are the same as given above , with the constants taking the values ( valid to leading order in @xmath3 ) : @xmath278 and where the constant @xmath195 is determined by the normalization condition @xmath279 : @xmath280 with @xmath281 the value of @xmath282 at @xmath115 .
note that in general @xmath282 is proportional to @xmath283 .
finally @xmath284 .
this seems to be an completely unwieldy mess .
but , we get an enormous simplification if we recall that we are interested in small @xmath3 and therefore can use the leading order results @xmath285 , @xmath286 , and @xmath287 is @xmath288 and hence can be dropped in the denominator , since @xmath195 is also @xmath288 . with the definition @xmath289
, we can derive the much simpler leading order expression @xmath290 finally , we use the fact that @xmath282 is small as it is related @xmath291 which equals @xmath292 raised to the positive power @xmath293 .
thus , @xmath294 is given by @xmath295 we need an explicit expression for @xmath296 .
we have @xmath297 thus , we have the final expression @xmath298 again reproducing what we have previously calculated .
we have analyzed here the pure birth as well as the birth - death stochastic luria - delbrck model in the limit of small mutation rate @xmath3 and large current population , @xmath0 , where the product @xmath7 is unrestricted .
the answer in all cases exhibits an intermediate asymptotic power - law decay @xmath95 for @xmath214 , where @xmath74 is the ratio of the wild - type net growth rate to the mutant net growth rate . for small @xmath7 ,
the distribution is monotonically decreasing , and approaches the pure power - law already for @xmath9 s of order unity .
as @xmath7 increases the probability of a small number of mutants decreases sharply , and above a critical @xmath7 of order 1 , the distribution develops a peak . the power - law decay sets in then only a distance past this peak . for large @xmath7 ,
the distribution becomes the one - sided lvy distribution with index @xmath74 .
the distribution for @xmath92 is in all cases independent of the initial wide type colony size @xmath10 .
it is also independent of the precise nature of the ensemble , as long as there are sufficient numbers of wild - type individuals in all members of the ensemble with non - vanishingly small weighting .
this can be guaranteed by fixing @xmath299 at either fixed time or in a time average , or alternatively by having a large number for the number @xmath10 of initial wild - type individuals .
the power - law tail , as well as the strong dependence of the distribution on @xmath10 for @xmath215 has significant implications for tasks such as using measured data to infer mutation rate @xcite .
this also has implications for the variation expected in patient responses to administered therapy , although it seems that the _ extrinsic _ variation due to different rates dominates over the _ intrinsic _ variation due to process stochasticity @xcite .
this work was supported by the nsf center for theoretical biological physics , ( grant no .
phy-1308264 ) .
hl was also supported by cprit scholar program of the state of texas , and dk was also supported by the israeli science foundation . ,
l.a . , willians , r.t . ,
wu , j. , kinde , i. , hecht , j.r . ,
berlin , j. , allen , b. , bozic , i. , reiter , j.g . , nowak , m.a .
, kinzler , k.w . , oliner , k.s . ,
vogelstein , b. : the molecular evolution of acquired resistance to targeted egfr blockade in colorectal cancers .
486*(7404 ) , 537540 ( 2012 )
|
one of the most popular models for quantitatively understanding the emergence of drug resistance both in bacterial colonies and in malignant tumors was introduced long ago by luria and delbrck . here
, individual resistant mutants emerge randomly during the birth events of an exponentially growing sensitive population . a most interesting limit of this process occurs when the population size @xmath0 is large and mutation rates are low , but not necessarily small compared to @xmath1 . here
we provide a scaling solution valid in this limit , making contact with the theory of levy @xmath2-stable distributions , in particular one discussed long ago by landau .
one consequence of this association is that moments of the distribution are highly misleading as far as characterizing typical behavior .
a key insight that enables our solution is that working in the fixed population size ensemble is not the same as working in a fixed time ensemble .
some of our results have been presented previously in shortened form @xcite
|
in the search for planetary companions outside of our solar system , a variety of techniques have been devised and implemented , of which radial velocity monitoring has proven the most successful to date .
such measurements have led to the discovery of planets with masses of @xmath8 @xmath9 and separations of a few au or less around @xmath10 nearby stars ( mayor & queloz 1995 ; marcy & butler 1996 ; cochran et al .
1997 ; noyes et al . 1997 ; udry , mayor , & queloz 2001 ; butler et al .
2000 ; fischer et al .
2001b ; tinney et al .
2001 ) . combining those results with data on spectroscopic binaries produces a distribution of companions from 1 - 1000 @xmath9 that is characterized by a minimum or brown dwarf desert " at 10 - 30 @xmath9 ( e.g. , marcy & butler 1998 ) . this feature is likely a reflection of two formation mechanisms , one that creates planets from disks and one that produces companion stars and brown dwarfs ( e.g. , core fragmentation ) .
however , to fully understand the processes of planet and binary star formation , it is essential that the frequency of companions is measured over the maximum possible range of separations .
for instance , the conspicuous paucity of brown dwarf companions at @xmath11 au may not exist at separations beyond 1000 au ( gizis et al .
2001 ) .
known extrasolar planetary systems are logical initial targets for a search for companions at modestly wide separations ( 10 - 100 au ) .
gravitational perturbations by outer substellar or stellar companions may be responsible for the eccentric orbits observed for many planetary companions . to properly evaluate the stability and evolution of these planets ( e.g. , rivera & lissauer 2000 ) ,
wide massive companions must be searched for and included in the analysis .
furthermore , if substellar companions can be resolved from primaries , spectroscopic analysis would provide constraints on masses , temperatures , surface gravities , and compositions , and thus contribute to our understanding of the physical characteristics of substellar objects ( marley et al .
1996 ; allard et al .
1996 ) . finally ,
as radial velocity measurements are accumulated over longer periods of time , they are becoming sensitive to companions at larger separations .
it already appears that at least 50% of stars with planets have additional distant companions ( marcy et al .
2001a , b ; fischer et al .
2001b ) , which suggests that a search for a substellar objects at still wider separations through high - resolution direct imaging could prove fruitful .
direct imaging has become increasingly effective in detecting faint companions to bright stars . in high - resolution imaging with wfpc2 aboard the _ hubble space telescope ( hst ) _ , schroeder et al .
( 2000 ) probed for massive and young brown dwarfs at 1 - 60 au from 23 nearby stars . from the ground ,
oppenheimer et al . (
2001 ) used optical coronographic data and near - infrared ( ir ) images to search for companions at 40 - 120 au and @xmath12 @xmath9 near most of the known stars within 8 pc , which resulted in the discovery of the brown dwarf companion gl 229b ( nakajima et al .
the tip - tilt techniques used for enhanced image quality by oppenheimer et al .
( 2001 ) have been followed by higher order adaptive optics ( ao ) systems that achieve comparable and sometimes superior spatial resolution and sensitivity to those of @xmath13 .
near - ir ao imaging already has been used to discover binary systems whose members are near the hydrogen burning mass limit ( martn et al .
2000 ; close et al .
in addition , ao has been performed in conjunction with radial velocity measurements of nearby stars ( delfosse et al .
1999 ) and has resulted in the discovery of a brown dwarf companion ( @xmath14 @xmath9 , @xmath15 ) to the planet - bearing star gl 86 ( els et al .
2001 ) . in this paper
, we present the results of a survey for companions to 25 planetary systems using near - ir ao imaging at keck observatory .
we report astrometry and photometry for several point sources found in these images and discuss their likely origin as companions or background field stars .
we characterize the detection limits of these data in both observational and physical units , which are then compared to the sensitivities of the radial velocity surveys , and we discuss the implications of these data for individual planet - bearing stars .
during the nights of 2000 may 10 , june 27 , and september 16 - 18 , we used the shack - hartmann ao system on the keck ii telescope ( wizinowich et al .
2000 ) in conjunction with kcam , a nicmos3 @xmath16 near - ir camera , to obtain images of 25 stars for which planetary companions have been discovered in previous radial velocity measurements .
properties of the planetary systems in our sample are given in table [ tab : prop ] .
the plate scale of kcam was @xmath17 pixel@xmath18 , corresponding to a total field of @xmath19 .
the lower right quadrant of the array was not functional .
a cold filter wheel contained @xmath20 , @xmath21 , and @xmath22 filters .
warm external filters included an open filter and neutral density filters that attenuated the signal by factors of 10 , 100 , and 1000 ( n1 , n2 , and n3 ) .
the h - band filter was selected for all observations because it produces the optimum combination of spatial resolution and sensitivity and because substellar objects are relatively bright in this band .
the targets were grouped into pairs of stars that were close together on the sky ( @xmath23 ) .
the stars in a pair were observed in succession so that the data for one star provided an estimate of the psf for the other star .
images for a given star were obtained in the following manner .
the star was first centered in one of the three operating quadrants of the array . the exposure time and neutral density filter
were selected such that the maximum number of counts in the image was between 50 and 80% of the saturation level , where 5 sec and n3 were typical choices for a star at @xmath24 .
multiple exposures were obtained at each of two positions that were separated by @xmath25 ( 5 pixels ) along the rows and columns of the array .
similar exposures were obtained through the less opaque neutral density filters and the open filter to reach successively fainter magnitudes .
the same steps were then followed for the star that was paired to this target .
these data are used to search for companions within the quadrant surrounding each star ( @xmath26 ) .
the following observations were designed to reach larger separations .
a target was centered in the upper left quadrant of the camera array . because the area of the array directly surrounding the target was already imaged and would be saturated in long exposures , the telescope was offset by @xmath27 ( 128 pixels ) in each direction on the array to place the star near the center of the lower right dead quadrant .
three images were obtained at each position in the same dither pattern used previously .
each image consisted of the sum of two exposures with integration times of 30 sec for most stars and 5 sec for the brightest targets , which were short enough to avoid saturation within the operating quadrants of the array .
as a result , the total integration time was typically 6 min .
these steps for the long exposures were repeated at two additional position angles at intervals of @xmath28 , providing full and partial coverage out to radii of @xmath6 and @xmath29 surrounding each star .
the position angle of the array was fixed during a given image .
we now comment on a few departures from the above observing strategy .
the observations of @xmath30 55 cnc and 47 uma differed from the standard procedure because the observations of these two objects were completed before the final strategy was adopted .
only the n3 and open external filters were used for these stars . in addition , only short exposure times were used for 47 uma .
similarly , only short exposures were obtained for hd 168443 on the first night of observations . as a result
, only the brightest two objects near hd 168443 in table [ tab : cand ] were detected in those data .
a complete data set with the final observing procedure was obtained for hd 168443 during the later observing runs .
the faint objects near the hd187123 , hd37124 , and hd168443 were noticed in the data that were reduced at the telescope . to facilitate the measurement of the positions of these sources relative to the primaries
, we obtained additional exposures in which both the faint sources and the primaries appeared in good quadrants of the array .
two of the stars with faint nearby sources in the keck data , hd 187123 and hd 37124 , were observed with the university of hawaii hokupaa ao system in conjunction with the @xmath31 near - ir camera quirc ( graves et al .
1998 ) on the gemini north telescope during the night of 2001 october 7 .
the plate scale of quirc was @xmath32 pixel@xmath18 , corresponding to a total field of @xmath33 .
we obtained twelve 30 sec exposures of hd 187123 and seven 1 sec exposures of hd 37124 through the @xmath21-band filter .
these exposure times were selected to minimize the saturation of the primary stars while providing sufficient sensitivity to detect the faint sources found in the keck data , thus optimizing the astrometry for those sources relative to the primaries .
the gemini ao data for hd 187123 and hd 37124 exhibited fwhm@xmath34 and @xmath35 , respectively .
standard data reduction procedures were followed for the ao images .
the keck images were divided by twilight sky flat frames , while the gemini data were divided by dome flat frames .
offsets between dithered frames were measured from the centroids of nonsaturated point sources or the isophotes around saturated stars .
dithered frames were then combined into one image .
the properties of the keck ao data are illustrated in figure [ fig : mosaic ] , which shows images of 51 peg at various signal levels .
the ao - corrected psf is characterized by a bright narrow core whose width ( fwhm@xmath36 pixels ) approaches the diffraction limit and a low - level broad halo that is comparable in size to the seeing disc ( fwhm@xmath37 pixels ) .
astronomical point sources must be distinguished from point - like speckles that reside within the halo .
this was accomplished by visually comparing the images of a given pair of targets that were observed consecutively and thus act as psf stars for each other .
in addition , because most of the psf features rotate on the array ( and thus the sky ) with time , we could distinguish speckles from astronomical sources by comparing the images at the individual dither positions for a star . at larger radii from a star ,
the features of the psf are radially elongated and therefore are not mistaken for companions .
the sources that were found in this manner are listed in table [ tab : cand ] .
images of the objects detected near hd 37124 , hd 168443 , hd 187123 , and hd 190228 are shown in in figs .
[ fig : hd37124]-[fig : hd190228 ] .
the array coordinates and photometry of the point sources near the target stars were measured with the iraf tasks imexamine and phot , respectively .
aperture photometry was extracted with a radius of 3 to 6 pixels in the keck data , where fainter sources were measured with smaller radii .
photometric calibration was derived from aperture photometry with radii of 40 pixels of the unsaturated images of the target stars .
when available , the @xmath21-band measurements from the two - micron all - sky survey were adopted .
otherwise , the @xmath21 magnitudes were estimated by combining the spectral types and visual magnitudes of the targets .
most of the keck images were obtained in photometric conditions and the calibrations implied by the various targets agree within @xmath38 mag . because the conditions were not photometric during the gemini observations , photometry is not measured from those data . after measuring the aperture corrections between 40 and 3 - 6 pixels from the unsaturated keck images of the targets
, we arrived at @xmath21-band magnitudes for the point sources , which are given in table [ tab : cand ] .
because the fraction of the star s light that falls within the psf core can change significantly between consecutive images , the absolute photometry has large uncertainties of @xmath39 mag .
multiple observations of the point source near hd 37124 suggest a smaller photometric error of @xmath40 mag . because the two sources near hd 187123 were measured in the same image , their relative photometry should have good precision ( @xmath41 mag ) . similarly , the relative positions of multiple sources near a star have small errors ( @xmath42 pixels ) . meanwhile , the offsets of the faint sources from the central stars are more uncertain because the latter were saturated or in the dead quadrant in the images where the former were detected .
as described in the previous section , extra images were obtained for hd 37124 , hd 187123 , and hd 168443 so that the faint sources and the primaries would both appear in good quadrants of the array .
the uncertainties in the offsets between saturated primaries and the surrounding objects are @xmath43 pixel in each direction on the array . in the gemini data , both the candidate companions and the primaries fell within the array as well .
for the faint sources near hd 190228 , we have only the original long exposures in which the primary fell in the dead quadrant . in these data , we used the isophotes of the primary s psf that extended into the good quadrants to estimate the position of the primary , and thus the offset between the faint sources and the primary .
the uncertainties in those offsets are @xmath44 pixels . by obtaining images of an artificial light source within the keck ao system at several positions across the array , we measured a plate scale of @xmath45 pixel@xmath18 for kcam
. a plate scale of @xmath46 pixel@xmath18 in the @xmath21-band has been measured for the quirc camera on gemini by f. rigaut .
these plate scales were combined with the position angles of the instruments and the pixel offsets to compute the offsets in right ascension and declination that are listed in table [ tab : cand ] .
we now characterize the sensitivity to companions of the keck ao data .
we began by measuring the @xmath21-band detection limits for point sources as a function of angular separation from the targets . to remove the large - scale shape of the psf of the target stars , each image was convolved with a two - dimensional gaussian function and divided into the original image . in the quotient image , we measured the standard deviation in five pixel square boxes at one pixel intervals along a line radially outward from star , which was then repeated for radial lines at several other angles around the star that avoided ghost images of the primary . at the image positions measured in this way , we inserted artificial stars that simulated the core of the psf , which were given the fwhm measured from the unsaturated image of the star and a peak intensity that equaled a multiple of the standard deviation measured at that point in the image .
after visually examining the artificial stars at several values of this multiple , we adopted six sigma as a reasonable representation of the detection limit .
the sources detected around hd 190228 in figure [ fig : hd190228 ] are near this detection limit .
in addition , eight artificial stars have been placed at the detection limit of @xmath1 at @xmath2 in one of the images of 51 peg in figure [ fig : mosaic ] . within the seeing disc ,
the presence of speckles complicates the characterization of the detection limit .
as discussed in the previous section , real point sources can be identified down to the amplitude of the speckles . as a result
, the speckles effectively define the detection limit for point sources .
because the six sigma deviations measured from 5 pixel boxes in the seeing disc were comparable to the speckles intensities ( i.e. , the speckles are sparsely distributed relative to this box size ) , we adopted the former measurement for characterizing the detection limit both inside and outside of the seeing disc .
for a given star , the @xmath21-band detection limits were measured for the various filters , exposure times , and position angles and the combined results were fit by one polynomial function .
the differences of the @xmath21-band detection limits for point sources and the magnitudes of the target stars ( @xmath47 ) are presented as a function of angular separation in figs .
[ fig : p1]-[fig : p7 ] .
the ghost images that appear in the long exposures result in an 2 - 3% incompleteness in areal coverage . because the observations of @xmath30 55 cnc and 47 uma were not optimized for dynamic range , the images were less sensitive to companions at intermediate separations , as indicated by the data for these stars in figure [ fig : p3 ] .
in addition , the detection limit at large separations for 47 uma was worse than for the other stars because only short exposures were obtained .
as an illustration of the sensitivity to companions of these data , in figs .
[ fig : p1]-[fig : p7 ] we have indicated the magnitudes of typical l and t dwarfs at the distances of the targets .
we considered all field dwarfs between spectral types of m9 v and t that have trigonometric parallaxes as compiled by kirkpatrick et al .
( 2000 ) . for these objects ,
we fit a polynomial function to spectral type versus @xmath48 for m9 v through l8 v , where the data were from kirkpatrick et al .
( 2000 ) and references therein .
the values of @xmath48 from this function for l0 v , l5 v , and l8 v and the measurements of @xmath48 for the t dwarfs gl 229b and gl 570d ( kirkpatrick et al .
2000 ; leggett et al .
1999 ) are plotted along the detection limits of separation versus @xmath47 in figs .
[ fig : p1]-[fig : p7 ] .
the @xmath21-band detection limits for companions have been transformed to units of mass and projected physical separation in the following manner .
conversions between spectral types and effective temperatures are from schmidt - kaler ( 1982 ) for @xmath49m0 v and from leggett et al .
( 1996 ) as presented by luhman ( 1999 ) for m1 v to m9 v. for l and t types , we adopted a conversion in which l0 v , l8 v , and the t dwarf gl 229b correspond to effective temperatures of 2000 , 1400 , and 900 k ( leggett et al . 1999 , 2001 ; kirkpatrick et al .
2000 ; reid et al . 2001b
; schweitzer et al . 2001 ) .
bolometric corrections are from kenyon & hartmann ( 1995 ) for @xmath50m6 and from bessell ( 1991 ) , monet et al .
( 1992 ) , tinney , mould , & reid ( 1993 ) , and leggett et al .
( 1996 ) for m6 to m9 . for l and t types ,
the @xmath21-band bolometric corrections were computed from the @xmath20-band corrections of reid et al .
( 2001a ) and the average @xmath51 colors of kirkpatrick et al .
( 2000 ) . by combining these conversions with the @xmath21-band magnitudes and angular separations of the detection limits , the distances of the targets , and theoretical mass - luminosity relations for 1 , 3 , and 10 gyr ( @xmath52 @xmath53 , baraffe at al . 1998 ; @xmath54 @xmath53 , burrows et al .
1997 ) , we have expressed the detection limits for companions in terms of mass and projected physical separation in figs .
[ fig : p1]-[fig : p7 ] . for comparison
, we also plot the positions of the companions to these stars that have been discovered through radial velocity measurements .
a typical detection limit of those studies is represented by a velocity modulation of 10 m s@xmath18 for 1 @xmath53 .
the actual detection limits for many of these stars are presented by cumming , marcy , & butler ( 1999 ) .
we have obtained keck near - ir ao images of 25 of the @xmath10 known extrasolar planetary systems . from these data ,
we have identified faint sources near four of the targets and have measured the detection limits for companions . using the keck and gemini ao systems , we have also obtained ao images at an additional epoch of three of the four primaries exhibiting nearby faint objects .
we now discuss the candidate companions individually , evaluate the achieved detection limits , and examine implications of these new constraints on substellar companions . for reference during this discussion , we list the 25 planetary systems in table [ tab : prop ] along with the latest values of separation and @xmath55 sin @xmath56 for their radial velocity companions , upper limits to the masses of these companions from _ hipparcos _ astrometry ( zucker & mazeh 2001 ; see also pourbaix & arenou 2001 ) , and recent age estimates .
faint point sources have been detected in the keck ao images of hd 37124 , hd 168443 , hd 187123 , and hd 190228 , which could be either companions or background field stars .
these four star are all located at low galactic latitude ( @xmath7 ) , which suggests that most of the faint sources are background field stars rather than companions . for the star hd 168443 in particular , which is only @xmath57 from the galactic center and
has @xmath58 , the presence of nine nearby objects at comparable magnitudes ( @xmath59 - 19 ) is suggestive of a background star population . because of the high proper motions of the targets in our sample and the high spatial resolution of ao imaging , multi - epoch observations spanning a year or less can easily distinguish companions from background stars .
such data were obtained for five of the 15 candidate companions , as listed in table [ tab : cand ] .
all of these five objects exhibit proper motions relative to the primary stars that are consistent with those expected of background stars . between 2000
september 18 and 2001 october 7 , the position of the object near hd 37124 changed by ( @xmath60 , @xmath61)=(@xmath62 , @xmath63)@xmath64 , which is close to the relative motion of ( @xmath65 , @xmath66 ) expected for a stationary background star given the _ hipparcos _ proper motion of hd 37124 .
the two objects near hd 187123 were observed at two epochs by keck . in data obtained at a third epoch by gemini , the fainter source fell below the detection limit .
the position of the brighter object changed by ( @xmath67 , @xmath68)@xmath64 relative to hd 187123 from 2000 june 27 to 2001 october 7 , which is consistent with the motion of ( @xmath69 , @xmath70 ) expected for a stationary background star . from 2000
june 27 to september 17 , the offsets between the two candidate companions to hd 187123 are expected to have changed by @xmath71 in right ascension and declination if one was a background star and the other was a companion .
since the sources remained fixed relative to each other at a level of @xmath72 , the fainter source also must be background star .
multi - epoch data were obtained for two of the sources near hd 168443 . between 2000 may 10 and september 17 , the positions of two of the objects relative to hd 168443 changed by ( @xmath73 , @xmath73)@xmath64 and ( @xmath74 , @xmath75)@xmath64 .
these measurements are consistent with the motion of background stars , which is expected to be ( @xmath60 , @xmath61)=(@xmath76 , @xmath77 ) given the proper motion of hd 168443 . because of the low galactic latitude of hd 168443 and hd 190228 ( @xmath78 ) , the remaining faint sources detected near these two stars are probably background stars as well . rather than simply the discovery of low - mass companions , the primary objective of this study has been to obtain new , highly sensitive constraints on the presence of companions near a sample of planetary systems .
these constraints are summarized and quantified by the point source detection limits in figs .
[ fig : p1]-[fig : p7 ] . from the plots of separation versus @xmath47 , we see that ao on the keck ii telescope produces high - contrast images in which point sources at @xmath1 can be detected at @xmath2 from a bright star ( see figure [ fig : mosaic ] ) . as a result , these data are sensitive to l and t dwarf companions to the targets at separations from @xmath3 ( l0 ) to @xmath4 ( gl 570d ) . for comparison , the companion search towards field stars by oppenheimer et al .
( 2001 ) probed 3-@xmath79 and discovered the brown dwarf gl 228b at a separation of @xmath80 . in recent coronographic
ao images of the planet - bearing star gl 86 , els et al .
( 2001 ) identified a probable brown dwarf companion at @xmath81 , @xmath15 , and @xmath1 .
gl 86 and @xmath30 55 cnc have similar distances and brightnesses , and our detection limit for the latter indicates that the companion to gl 86 would have been detected in our program as well .
when these detection limits are transformed to units of companion mass and physical separation , they can be compared to the sensitivities of other methods of searching for substellar companions .
the most fruitful of these techniques has been radial velocity monitoring , which currently reaches companions at @xmath11 au that induce modulations of @xmath82 m s@xmath18 . as demonstrated in figs .
[ fig : p1]-[fig : p7 ] , our ao survey nicely complements the radial velocity work by probing for companions at wider separations , typically stars at 3 - 10 au and brown dwarfs at 10 - 100 au .
the sensitivity to companions in direct imaging is improved with less massive and closer primary stars , which is illustrated by our data for the m dwarf gl 876 , where brown dwarfs are detectable at 1 - 10 au .
several stars in our sample deserve further comment given the new constraints on the presence of companions from this study .
as already noted , the ao detection limits approach but do not reach the separations and masses of the substellar objects that have been inferred from radial velocity measurements to date .
however , in a few cases , the ao data do place upper limits on the masses of those unseen companions . on the dates of our observations ,
radial velocity measurements predict minimum angular separations of @xmath83 and @xmath84 for the two companions to 47 uma and @xmath85 for the companion to 14 her ( g. marcy , private communication ) .
while we can not constrain the mass of the inner companion to 47 uma , our non - detections correspond to a mass upper limit of 0.3 @xmath53 for each of the other two companions .
the currently measured orbit for the putative companion to @xmath86 eri is too uncertain to reliably predict the minimum angular separation , but our data do suggest that the companion is unlikely to be a star .
as the time baselines of the radial velocity studies become longer , additional companions may be discovered at larger separations ( see table [ tab : prop ] ) , and these ao data could be useful in constraining their masses as well .
however , upper limits from _ hipparcos _
are often much lower .
finally , gonzales et al .
( 2001 ) found that hd 46375 and hd 37124 are overluminous on the hertzsprung - russell diagram by 0.5 and 0.25 mag at @xmath87 , respectively .
they suggest that these stars may have unresolved stellar companions that are beyond the maximum separations of 5 au probed by radial velocity measurements .
however , at projected separations of @xmath88 ( @xmath89 au for each star ) , we find no companions that are sufficiently bright to account for the discrepancies in the luminosities .
other constraints on the presence of companions from the literature are noted in table [ tab : prop ] .
we have conducted a survey for companions to 25 of the @xmath10 known extrasolar planetary systems using near - ir ao imaging at keck observatory .
the keck ii ao system has produced high - resolution images that approach the diffraction limit with a typical value of @xmath83 for the fwhm in the @xmath21 band . using a variety of neutral density filters , exposure times , and position angles , we arrived at a set of images for each target that provides optimum sensitivity for all separations from @xmath90 to @xmath6 . after searching the ao images for candidate companions , near hd 37124 , hd 168443 , hd 187123 , and hd 190228
we have identified 15 faint point sources ( @xmath5 - 20 ) , all of which would be at or below the hydrogen burning mass limit if they were companions . using ao imaging at keck and gemini observatories ,
we have obtained images at a second epoch for five of the brightest candidate companions .
these sources all exhibit proper motions relative to the central stars that are consistent with those expected of background stars .
the remaining faint sources are probably background stars as well given the low galactic latitude of the four targets .
we have measured the point source detection limits for the ao images , which are presented in terms of both angular separation versus @xmath47 and projected physical separation versus companion mass .
the sensitivities correspond to @xmath1 at @xmath2 and l and t dwarfs at @xmath3 to @xmath4 from the planet - bearing stars ( @xmath91 - 6 ) .
these data reach companion stars and brown dwarfs at 3 - 10 au and 10 - 100 au from the planet - bearing stars in this study , providing an essential complement to previous radial velocity measurements of these systems ( @xmath8 @xmath9 , @xmath92 au ) .
we wish to thank peter wizinowich , scott acton , david le mignant and the rest of the keck observatory staff for their support .
we also thank kathy roth , francois rigaut , mark chun , olivier guyon , and dan potter for their assistance in obtaining the observations at gemini observatory .
we are grateful to joan najita for help with an earlier keck proposal , and to geoff marcy for calculating the positions of the radial velocity companions to 47 uma and 14 her .
we thank geoff marcy , debra fischer , and jamie lloyd for useful discussions .
k. l. was supported by a postdoctoral fellowship at the harvard - smithsonian center for astrophysics .
r. j. was supported by a miller research fellowship .
this work was supported in part by the smithsonian institution , nasa , and nsf grants to geoff marcy , and a nasa grant to r.j .
administered by the aas .
this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center , funded by the national aeronautics and space administration and the national science foundation .
some of the data presented herein were obtained at the w. m. keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california , and the national aeronautics and space administration .
the observatory was made possible by the generous financial support of the w. m. keck foundation .
we wish to extend special thanks to those of hawaiian ancestry on whose sacred mountain we are privileged to be guests . without their generous hospitality , some of the observations presented herein would not have been possible .
some of these observations were obtained at the gemini observatory , which is operated by the association of universities for research in astronomy , inc .
, under a cooperative agreement with the nsf on behalf of the gemini partnership : the national science foundation ( united states ) , the particle physics and astronomy research council ( united kingdom ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council ( australia ) , cnpq ( brazil ) and conicet ( argentina ) .
this paper includes data obtained with the adaptive optics system hokupaa / quirc , developed and operated by the university of hawaii adaptive optics group , with support from the national science foundation .
finally , we thank the anonymous referee for useful comments that improved this manuscript .
allard , f. , hauschildt , p. h. , baraffe , i. , & chabrier , g. 1996 , , 465 , l123 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p. h. 1998 , , 337 , 403 bessell , m. s. 1991 , , 101 , 662 brown , t. m. , charbonneau , d. , gilliland , r. l. , noyes , r. w. , & burrows , a. 2001 , , 552 , 699 burrows , a. , et al .
1997 , , 491 , 856 butler , r. p. , et al .
1999 , , 526 , 916 butler , r. p. , marcy , g. w. , vogt , s. s. , & apps , k. 1998 , , 110 , 1389 butler , r. p. , marcy , g. w. , williams , e. , hauser , h. , & shirts , p. 1997
, , 474 , l115 butler , r. p. , vogt , s. s. , marcy , g. w. , fischer , d. a. , henry , g. w. , & apps , k. 2000 , , 545 , 504 charbonneau , d. , brown , t. m. , latham , d. w. , & mayor , m. 2000 , , 529 , l45 close , l. m. , potter , d. , brandner , w. , lloyd - hart , m. , liebert , j. , burrows , a. , & siegler , n. 2002 , , in press cochran , w. d. , hatzes , a. p. , butler , r. p. , & marcy , g. w. 1997 , , 483 , 457 cumming , a. , marcy , g. w. , & butler , r. p. 1999
, , 526 , 890 delfosse , x. , forveille , t. , beuzit , j .- l . ,
udry , s. , mayor , m. , & perrier , c. 1999 , , 344 , 897 delfosse , x. , forveille , t. , mayor , m. , perrier , c. , naef , d. , & queloz , d. 1998 , , 338 , 67 els , s.g .
, sterzik , m.f . , marchis , f. , pantin , e. , endl , m. , & krster , m. 2001 , , 370 , l1 fischer , d. a. , marcy , g. w. , butler , r. p. , laughlin , g. , & vogt , s. s. 2001a , , in press fischer , d. a. , marcy , g. w. , butler , r. p. , vogt , s. s. , & apps , k. 1999 , , 111 , 50 fischer , d. a. , marcy , g. w. , butler , r. p. , vogt , s. s. , frink , s. , apps , k. 2001b , , 551 , 1107 ford , e. b. , rasio , f. a. , & sills , a. 1999 , , 514 , 411 fuhrmann , k. , pfeiffer , m. j. , & bernkopf , j. 1997 , , 326 , 1081 fuhrmann , k. , pfeiffer , m. j. , & bernkopf , j. 1998 , , 336 , 942 gizis , j. e. , kirkpatrick , j. d. , burgasser , a. , reid , i. n. , monet , d. g. , liebert , j. , & wilson , j. c. 2001 , , 551 , l163 gonzalez , g. 1998 , , 334 , 221 gonzalez , g. 1999 , , 308 , , 447 gonzalez , g. , & laws , c. 2000 , , 119 , 390 gonzalez , g. , laws , c. , tyagi , s. , & reddy , b. 2001 , , 121 , 432 gonzalez , g. , & vanture , a. d. 1998 , , 339 , 29 graves , j. e. , northcott , m. j. , roddier , f. j. , roddier , c. a. , & close , l. m. 1998 , proc .
, 3353 , 34 greaves , j. s. , et al .
1998 , , 506 , 133 hatzes , a. p. , et al .
2000 , , 544 , l145 henry , g. w. , baliunas , s. l. , donahue , r. a. , fekel , f. c. , & soon , w. h. 2000a , , 531 , 415 henry , g. w. , marcy , g. w. , butler , r. p. , & vogt , s. s. 2000b , , 529 , l41 kenyon , s. j. , & hartmann , l. 1995 , , 101 , 117 kirkpatrick , j. d. , et al .
2000 , , 120 , 447 lachaume , r. , dominik , c. , lanz , t. , & habing , h. j. 1999 , , 348 , 897 leggett , s. k. , allard , f. , berriman , g. , dahn , c. c. , & hauschildt , p. h. 1996 , , 104 , 117 leggett , s. k. , allard , f. , geballe , t. r. , hauschildt , p. h. , & schweitzer , a. 2001 , , 548 , 908 leggett , s. k. , toomey , d. w. , geballe , t. r. , & brown , t. r. 1999 , , 517 , l139 luhman , k. l. 1999 , , 525 , 466 marcy , g. w. , & butler , r. p. 1996
, , 464 , l153 marcy , g. w. , & butler , r. p. 1998
, , 36 , 57 marcy , g. w. , butler , r. p. , vogt , s. s. , fischer , d. , & lissauer , j. j. 1998 , , 505 , l147 marcy , g. w. , butler , r. p. , vogt , s. s. , fischer , d. , & liu , m. 1999 , , 520 , 239 marcy , g. w. , butler , r. p. , & vogt , s. s. 2000a , , 536 , l43 marcy , g. w. , cochran , w. d. , & mayor , m. 2000b , in _ protostars and planets iv _
v. mannings , a.p .
boss , s.s .
russell ( tucson : university of arizona press ) , 1285 marcy , g. w. , butler , r. p. , fischer , d. a. , vogt , s. s. , lissauer , j. j. , & rivera , e. j. 2001a , , 556 , 296 marcy , g. w. , butler , r. p. , vogt , s. s. , liu , m. c. , laughlin , g. , apps , k. , graham , j. r. , lloyd , j. , luhman , k. l. , & jayawardhana , r. 2001b , , 555 , 418 marley , m. s. , saumon , d. , guillot , t. , freedman , r. s. , hubbard , w. b. , burrows , a. , & lunine , j. i. 1996 , science , 272 , 1919 martn , e. l. , koresko , c. d. , kulkarni , s. r. , lane , b. f. , & wizinowich , p. l. 2000 , , 529 , l37 mazeh , t. , et al .
2000 , , 532 , l55 mayor , m. , & queloz , d. 1995 , nature , 378 , 355 monet , d. g. , dahn , c. c. , vrba , f. j. , harris h. c. , pier , j. r. , luginbuhi , c. b. , & ables , h. d. 1992 , , 103 , 638 naef , d. , mayor , m. , pepe , f. , queloz , d. , santos , n. c. , udry , s. , & burnet , m. 2001 , , 375 , 205 nakajima , t. , oppenheimer , b. r. , kulkarni , s. r. , golimowski , d. a . , matthews , k. , & durrance , s. t. 1995 , nature , 378 , 463 ng , y. k. , & bertelli , g. 1998 , , 329 , 943 noyes , r. w. , et al .
1997 , , 483 , l111 oppenheimer , b. r. , golimowski , d. a. , kulkarni , s. r. , matthews , k. , nakajima , t. , creech - eakman , m. , & durrance , s. t. 2001 , , 121 , 2189 pourbaix , d. , & arenou , f. 2001 , , 372 , 935 queloz , d. , mayor , m. , naef , d. , pepe , f. , santos , n.c . ,
udry , s. , & burnet , m. 2001 , in planetary systems in the universe : observation , formation and evolution , iau symp .
202 , eds .
a. penny , p. artymowicz , a .- m . lagrange and s. russel asp conf .
, in press reid , i. n. , burgasser , a. j. , cruz , k. l. , kirkpatrick , j. d. , & gizis , j. e. 2001a , , 121 , 1710 reid , i. n. , gizis , j. e. , kirkpatrick , j. d. , & koerner , d. w. 2001b , , 121 , 1710 rivera , e. j. , & lissauer , j. j. 2000 , , 530 , 454 schmidt - kaler , t. 1982 , in landolt - bornstein , group vi , vol .
hellwege ( berlin : springer ) , 454 schroeder , d. j. , et al .
2000 , , 119 , 906 schweitzer , a. , gizis , j. e. , hauschildt , p. h. , allard , f. , howard , e. m. , & kirkpatrick , j. d. 2001 , , in press sivan , j .-
p . , mayor , m. , naef , d. , queloz , d. , udry , s. , perrier - bellet , c. , & beuzit , j. 2001 , in planetary systems in the universe : observation , formation and evolution , iau symp .
202 , eds .
a. penny , p. artymowicz , a .- m .
lagrange and s. russel asp conf .
, in press smith , v. v. , cunha , k. , & lazzaro , d. 2001 , , 121 , 3207 song , i. , caillault , j .-
, barrado y navasces , d. , stauffer , j. r. , & randich , s. 2000 , , 533 , 41 tinney , c. g. , et al .
2001 , , 551 , 507 tinney , c. g. , mould , j. r. , & reid , i. n. 1993 , , 105 , 1045 udry , s. , mayor , m. , & queloz , d. , 2001 , in planetary systems in the universe : observation , formation and evolution , iau symp .
202 , eds .
a. penny , p. artymowicz , a .-
lagrange and s. russel asp conf .
, in press vogt , s. s. , marcy , g. w. , butler , r. p. , & apps , k. 2000 , , 536 , 902 wizinowich , p. , et al .
2000 , , 112 , 315 zucker , s. , & mazeh , t. 2001 , , in press lllllll @xmath93 and & 0.059 & 0.58 & 1 & & @xmath94 , @xmath95 , @xmath96 , @xmath97 , @xmath98 , @xmath99 & 2 , 3 , 4 , 5 , 6 , 7 + & 0.828 & 2.05 & 8 , 9 & & & + & 2.56 & 4.29 & 8 , 9 & @xmath100 & & + hd 10697 & 2.12 & 6.08 & 10 & @xmath101 & @xmath102 , 6.0 & 11 + hd 12661 & 0.795 & 2.84 & 12 & @xmath103 & @xmath104 , 8.4 & 11 + hd 16141 & 0.351 & 0.22 & 13 & @xmath105 & @xmath106 , 6.7 & 11 + hd 19994 & 1.26 & 1.83 & 14 & @xmath107 & @xmath108 , 4.3 & 3 , 15 + @xmath86 eri & 3.36 & 0.88 & 16 & @xmath109 & 0.5 - 1 , @xmath110 & 17 , 18 + hd 37124 & 0.547 & 1.13 & 10 & @xmath111 & 3.9 & 11 + hd 38529 & 0.129 & 0.77 & 9 & & @xmath112 , 3.7 & 11 + hd 46375 & 0.041 & 0.25 & 13 & & 4.5 & 11 + hd 52265 & 0.493 & 1.14 & 12 , 19 & @xmath103 & @xmath113 , 4.0 , 3.5 & 8 , 8 , 19 + @xmath30 55 cnc & 0.118 & 0.93 & 1 & @xmath114 & @xmath92 , @xmath115 , 5 , @xmath116 , 4 - 5 & 2 , 20 , 21 , 4 , 7 + 47 uma & 2.09 & 2.54 & 22 & @xmath117 & @xmath118 , @xmath119 , @xmath120 , 6.3 , @xmath121 , 6 & 23 , 3 , 24 , 21 , 5 , 7 + & 3.73 & 0.76 & 25 & & & + @xmath122 crb & 0.224 & 0.99 & 26 & @xmath123 & @xmath124 , @xmath125 , @xmath126 , 12.3 , @xmath127 , 6 & 2 , 3 , 24 , 21 , 4 , 7 + 14 her & 3.17 & 4.05 & 27 & @xmath128 & 6 & 28 , 21 + hd 168443 & 0.295 & 7.64 & 29 & & @xmath129 , 7.4 , 8 & 11 , 11 , 30 + & 2.87 & 16.96 & 30 , 31 & @xmath130 & & + 16 cyg b & 1.69 & 1.68 & 32 & @xmath131 & @xmath132 , @xmath133 , 9 , 6 - 7 & 2 , 24 , 21 , 7 + hd 187123 & 0.042 & 0.54 & 33 & & @xmath134 , 5.5 & 28 , 21 + hd 190228 & 2.25 & 5.01 & 34 & @xmath135 & & + hd 195019 & 0.136 & 3.55 & 35 & @xmath136 & 3 & 35 + hd 209458 & 0.046 & 0.63 & 36 , 37 & & 4.5 , @xmath137 , 4.3 & 38 , 11 , 11 + hd 210277 & 1.12 & 1.29 & 29 , 19 & @xmath138 & @xmath139 , 8.5 & 28 , 21 + gl 876 & 0.207 & 1.89 & 39 , 40 & @xmath140 & 1 - 10 & 39 + & 0.130 & 0.56 & 41 & & & + 51 peg & 0.052 & 0.46 & 42 & @xmath141 & @xmath142 , @xmath143 , @xmath144 , @xmath145 , 7 , @xmath146 , 7.1 & 23 , 3 , 4 , 5 , 7 , 11 , 11 + hd 217107 & 0.072 & 1.29 & 35 & @xmath147 & 8 , @xmath148 , 5.6 & 35 , 11 , 11 + hd 222582 & 1.35 & 5.18 & 10 & @xmath149 & @xmath150 , 5.6 & 11 + llllll hd 37124 & @xmath1513.076 , @xmath152 & 0.314 , 0.702 & 14.9 & 8.9 & 9/18/00 , 10/07/01 + hd 168443 & 2.802 & @xmath1510.413 & 18.5 & 13.1 & 9/17/00 + & 3.516 , 3.594 & 0.129 , 0.207 & 16.4 & 11.0 & 5/10/00 , 9/17/00 + & 3.166 , 3.243 & 0.318 , 0.403 & 17.7 & 12.3 & 5/10/00 , 9/17/00 + & 2.691 & 1.585 & 18.8 & 13.4 & 9/17/00 + & 1.554 & 2.789 & 19.0 & 13.6 & 9/17/00 + & 0.402 & 2.677 & 17.6 & 12.2 & 9/17/00 + & @xmath1510.364 & 2.949 & 18.6 & 13.2 & 9/17/00 + & @xmath1513.215 & 2.380 & 19.0 & 13.6 & 9/17/00 + & 2.973 & 1.502 & @xmath153 & @xmath154 & 9/17/00 + hd 187123 & 3.364,3.361,3.171 & 0.973,0.994,1.130 & 14.2 & 7.8 & 6/27/00 , 9/16/00 , 10/07/01 + & 3.142,3.137 & 0.286,0.310 & 18.3 & 11.9 & 6/27/00 , 9/16/00 + hd 190228 & 0.999 & @xmath1511.796 & @xmath155 & @xmath156 & 9/18/00 + & @xmath1511.343 & @xmath1511.644 & @xmath157 & @xmath158 & 9/18/00 + & @xmath1511.664 & @xmath1512.312 & @xmath155 & @xmath156 & 9/18/00 +
|
we have performed a keck adaptive optics ( ao ) imaging survey of 25 extrasolar planetary systems discovered by the radial velocity programs . typically , the high - resolution ( fwhm@xmath0 ) near - infrared images are able to detect point sources at @xmath1 at @xmath2 , l and t dwarfs at @xmath3 to @xmath4 , and stars and brown dwarfs at 3 - 10 au and 10 - 100 au from the targets .
the ao images reveal 15 faint sources ( @xmath5 - 20 ) near the stars hd 37124 , hd 168443 , hd 187123 , and hd 190228 out to a search radius of @xmath6 .
we have obtained ao images at a second epoch for five of the candidate companions at keck and gemini observatories .
the resulting measurements of proper motions relative to the primary stars indicate that these five sources are background stars .
given that the four primaries are near the galactic plane ( @xmath7 ) , the remaining faint sources are also likely to be background stars rather than substellar companions .
we present the companion detection limits for each target as a function of separation and compare them to the sensitivities of radial velocity surveys , demonstrating the complementary nature of the two techniques .
|
cavity optomechanics @xcite , which explores the interaction between light and mechanical motion , provides a unique platform for various applications , such as the fundamental test of quantum theory @xcite , quantum information processing @xcite and high - precision measurements @xcite .
recent theoretical and experimental efforts have demonstrated optomechanically induced transparency @xcite , optomechanical storage @xcite , normal mode splitting @xcite , quantum - coherent coupling between optical modes and mechanical modes @xcite , state transfer at different optical ( electromagnetic ) wavelengths @xcite , quantum entanglement @xcite , squeezing @xcite and nonlinear quantum optomechanics @xcite . for most applications ,
it is a prerequisite to cool the mechanical resonators close to the quantum ground state so as to suppress the thermal noise . in the past few years
, numerous experiments have demonstrated cooling of mechanical resonators by employing pure cryogenic cooling @xcite , feedback cooling ( active cooling , or cold - damping ) @xcite and backaction cooling ( passive cooling , or self - cooling ) @xcite .
the backaction cooling is proved to be efficient @xcite , especially in the resolved sideband limit , where the mechanical resonance frequency is greater than the decay rate of the optical cavity .
recently many efforts have been taken to extend or improve the backaction cooling , for example , cooling with dissipative coupling @xcite , cooling with quadratic coupling @xcite , cooling with hybrid systems @xcite , cooling in the single photon strong coupling regime @xcite and pulsed laser cooling @xcite . besides ground state cooling , another crucial condition for quantum operation is strong coupling @xcite , where the light - enhanced optomechanical coupling strength exceeds the cavity decay rate .
in such a regime , the optical and mechanical modes hybridize , leading to normal mode splitting @xcite . in the time domain ,
the energy exchange between the optical and mechanical modes is reversible , which allows state swapping .
however , it also brings about swap heating , which leads to the saturation of cooling rates . to overcome this problem
, we have recently @xcite proposed a dynamic dissipative cooling approach , which significantly accelerates the cooling process and reduces the cooling limits . in this paper
we extend our previous results and present a detailed exploration of the cooling dynamics and cooling limits in the strong coupling regime .
we systematically investigate the cooling dynamics under the rotating - wave approximation ( rwa ) and without rwa .
rabi - like oscillations and frequency mixing phenomenon in the time evolution of mean phonon number are studied . to obtain the lowest cooling limits
, we analytically derive the frequency matching condition and a small cavity decay rate is preferred .
the rest of this paper is organized as follows . in sec .
ii we use the linearized quantum langevin equations and quantum master equation to describe the system . in sec .
iii we discuss the simplified model under rwa . in sec .
iv we study the full cooling dynamics without the rwa , where the zero - temperature and finite - temperature cases are considered in succession . in sec .
v we discuss the reduced instantaneous - state cooling limits with dynamic dissipative cooling approach .
a summary is presented in sec . vi .
( under rwa ) as functions of @xmath0 and @xmath1 for @xmath2 , @xmath3 , @xmath4 and @xmath5 .
the purple dashed curves correspond to the cases of @xmath6 , @xmath7 and @xmath8 .
the color bar is in log scale .
( c ) @xmath9 as a function of @xmath0 for @xmath10 ( red circles and red solid curve ) and @xmath11 ( blue stars and blue dashed curve ) .
the circles and stars are numerical results obtained from eq .
( [ drwa ] ) and the curves are analytical results calculated from eq .
( [ nbrwa ] ) .
the gray dotted curves denote the envelopes given by @xmath12 ( upper curve ) and @xmath13 ( lower curve ) . ]
we consider a typical optomechanical system involving an optical mode and a mechanical mode , with a coherent laser driving the optical cavity [ fig .
[ fig1](a ) ] .
the system hamiltonian reads @xmath14here the first ( second ) term describes the energy of the optical ( mechanical ) mode , with the angular resonance frequency @xmath15 ( @xmath16 ) , the annihilation operator @xmath17 ( @xmath18 ) and the creation operator @xmath19 ( @xmath20 ) .
the third term represents the optomechanical interaction @xcite , with the single - photon optomechanical coupling rate @xmath21 .
the last term describes the driving of the input laser , where @xmath22 denotes the driving strength with the input laser power @xmath23 , initial phase @xmath24 , frequency @xmath25 and the input - cavity coupling rate @xmath26 . in the interaction picture ,
the quantum langevin equations are given by @xmath27 where @xmath28 denotes the input - cavity detuning ; @xmath29 stands for the intrinsic cavity dissipation rate ; @xmath30 represents the total cavity dissipation rate ; @xmath31 is the dissipation rate of the mechanical mode ; @xmath32 , @xmath33 and @xmath34 are the noise operators associated with the intrinsic cavity dissipation @xmath29 , external cavity dissipation @xmath26 and mechanical dissipation @xmath31 , which have zero mean values and obey the correlation functions @xmath35 , @xmath36 , @xmath37 and @xmath38 . here
@xmath39^{-1}$ ] describes the equilibrium mean thermal phonon number , where @xmath40 is the temperature of the reservoir and @xmath41 is boltzmann constant . for strong coherent laser input , both
the optical and mechanical modes reach a new steady state , and thus we can rewrite the operators as @xmath42 and @xmath43 . here
@xmath44 and @xmath45 represent the @xmath46-number steady state values of the optical and mechanical modes , while @xmath47 and @xmath48 are the corresponding fluctuation operators . from eqs .
( [ a ] ) and ( [ b ] ) , the quantum langevin equations for the quantum fluctuations are given by @xmath49 where @xmath50 represents the optomechanical - coupling modified detuning .
for strong driving @xmath51 , the nonlinear terms in eqs .
( [ a1 ] ) and ( [ b1 ] ) can be neglected , yielding linearized quantum langevin equations .
the corresponding quadratic hamiltonian is given by @xmath52 where @xmath53 describes the light - enhanced optomechanical coupling strength .
in this hamiltonian , the rotating - wave terms @xmath54 correspond to the beam - splitter interaction , while the counter - rotating - wave terms @xmath55 describe the two - mode squeezing interaction .
starting from the hamiltonian eq .
( [ hl ] ) in the linear regime , the time evolution of the system density matrix @xmath56 is described by the quantum master equation @xmath57+\frac{\kappa}{2}\left ( 2{a}_{1}\rho{a_{1}^{\dag}}-{{a_{1}^{\dag}{a}_{1}\rho-\rho a_{1}^{\dag}{a}_{1}}}\right ) \nonumber\\ & + \frac{\gamma}{2}(n_{\mathrm{th}}+1)\left ( 2{b}_{1}\rho{b_{1}^{\dag } -{b_{1}^{\dag}b_{1}\rho-\rho b_{1}^{\dag}b_{1}}}\right ) \nonumber\\ & + \frac{\gamma}{2}n_{\mathrm{th}}\left ( 2{b_{1}^{\dag}}\rho{b}_{1}-{{b_{1}{b_{1}^{\dag}}\rho-\rho b_{1}b_{1}^{\dag}}}\right ) .
\label{master}\ ] ] we focus on the time evolution of the mean phonon number @xmath58 .
( [ master ] ) , @xmath59 is determined by a linear system of ordinary differential equations involving all the second - order moments @xmath60 , @xmath59 , @xmath61 ,
@xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 @xcite , with the equations given by @xmath69 where @xmath70 , and the elements of the matrices @xmath71 and @xmath72 are presented in the appendix .
we focus on the strong coupling regime , where the light - enhanced optomechanical coupling strength @xmath73 is far greater than the cavity decay rate @xmath74 .
when @xmath75 is also satisfied , the rotating - wave approximation can be made , so that we can concisely describe the main characteristic of the cooling dynamics . under the rwa ,
the counter - rotating - wave terms @xmath76 are neglected , and eq .
( [ d ] ) reduce to [ drwa]@xmath77f,\end{gathered}\ ] ] where @xmath78 represents the coherence between the optical and mechanical modes .
we consider the red sideband resonant case with @xmath2 , _
i. e. _ , the cooling process is on resonance .
the time evolution of the mean phonon number is given by @xmath79 } { { \kappa+\gamma}}. \label{nbrwa}\ ] ] it shows that the phonon number is proportional to the environmental thermal phonon number @xmath80 , which reveals that the rotating - wave optomechanical coupling only modifies the effective mechanical dissipation rate . in fig .
[ fig1](b ) we plot the exact numerical results of @xmath9 as functions of @xmath0 and @xmath81 for @xmath82 , @xmath4 and @xmath5 , with fig .
[ fig1](c ) plots both numerical and analytical results of @xmath9 for @xmath10 and @xmath11 .
it shows that the mean phonon number undergoes rabi - like oscillations with the period of @xmath83 , which implies the energy exchange between the optical mode and the mechanical mode , together with the normal mode splitting with the frequency difference of @xmath84 .
the dissipation is characterized by the envelopes , where the upper envelope is approximately described by @xmath12 , and the lower envelope reads @xmath85 , as shown in fig .
[ fig1](c ) . from figs .
[ fig1](b ) and ( c ) we find small discrepancy for small @xmath73 , which is because the strong coupling condition is weakly satisfied for @xmath86 .
taking the dissipations into account , the normal mode splitting is given by @xmath87 , which is slightly smaller than @xmath84 . thus the oscillation period is slightly larger than @xmath83 , which agrees with figs .
[ fig1](b ) and ( c ) . when the system reaches the steady state , the final phonon number is given by @xmath88 , as inferred from eq .
( [ nbrwa ] ) .
nevertheless , the minimum phonon number is obtained near the end of the first half rabi oscillation cycle , @xmath89 , with the instantaneous - state cooling limit @xmath90 it shows that the instantaneous - state cooling limit does not depend on the cavity decay rate @xmath91 .
this is a great improvement compared with the steady - state cooling limit , which is constrained by the cavity decay rate . in fig .
[ fig2 ] we plot @xmath92 as a function of @xmath1 .
it reveals that a large coupling strength leads to a low cooling limit .
this is because the time to reach the minimum is in inverse proportional to the coupling strength , and short evolution time suffers less dissipation , as compared in fig .
[ fig1](c ) for @xmath93 and @xmath11 . the comparison between the steady - state cooling limit and the instantaneous - state cooling limit is also shown in fig .
[ fig2 ] , which reveals the advantage of the instantaneous - state cooling limit with a factor of @xmath94 . as a function of @xmath1 .
the red circles are numerical results obtained from eq .
( [ drwa ] ) and the red solid curve is the analytical result calculated from eq .
( [ ninsrwa ] ) .
the steady - state cooling limit is plotted for comparison ( blue dashed curve ) .
other parameters are the same as fig . [
fig1](b ) .
when the coupling strength @xmath73 is comparable to the mechanical resonance frequency @xmath16 , the effect of the counter - rotating interactions become important .
in this case we need to solve eq .
( [ d ] ) , without the the rwa . for convenience ,
we first consider the zero - temperature case , where the environmental thermal phonon number @xmath80 is set to be zero . for @xmath95 ,
all the initial values of the second - order moments in eq .
( [ d ] ) are zero . with the time evolution , these moments become nonzero due to the the quantum backaction resulting from the vacuum fluctuations , i. e. , they are created from vacuum by the counter - rotating interactions . for the red sideband resonant case @xmath2 , by diagonalizing the hamiltonian eq . ( [ hl ] ) , we obtain @xmath96 , where @xmath97 are the eigenmodes with the corresponding eigenfrequencies @xmath98 .
note that the the eigenfrequencies reduce to @xmath99 with the rwa as shown in the previous section .
the rotating - wave interaction is characterized by the frequency @xmath100[@xmath101 with rwa as appeared in eq .
( [ nbrwa ] ) ] , while the counter - rotating - wave interaction is characterized by the frequency @xmath102 .
therefore , by taking both interactions into consideration , the system dynamics is described by two frequencies @xmath103 . to provide quantitative results ,
we derive the time evolution of the mean phonon number in this case as @xcite@xmath104 } { { 2(\omega_{\mathrm{m}}^{2}-4}\left\vert { g}\right\vert ^{2})}. \label{nb0}\ ] ] unlike @xmath9 as shown in eq .
( [ nbrwa ] ) , here @xmath105 does not depend on the environmental thermal phonon number @xmath80 , which reveals that it originates from the quantum backaction associated with the vacuum fluctuations . as functions of @xmath0 and @xmath1 .
the purple dashed curves correspond to the cases of @xmath106 , @xmath107 and @xmath108 .
the island in the top left corner is labeled as @xmath109 .
the color bar is in log scale .
the black contour curves denote @xmath110 , @xmath111 , @xmath112 , @xmath113 , @xmath114 and @xmath115 .
( b ) and ( c ) : @xmath116 as a function of @xmath0 for @xmath117 ( b ) and @xmath118 ( c ) .
the circles and stars are numerical results obtained from eq .
( [ d ] ) , and the curves are analytical results calculated from eq .
( [ nb0 ] ) .
other parameters are the same as fig .
[ fig1](a ) except that @xmath95 . ] in fig .
[ fig3](a ) we present the exact numerical results of @xmath105 as functions of @xmath0 and @xmath81 with contour plots .
we note that a number of islands regularly appear in the contour map .
this is a result of the carrier - envelope - type frequency mixing as described by the term @xmath119 in eq .
( [ nb0 ] ) , where the carrier frequency @xmath102 corresponds to the counter - rotating - wave interaction and the envelope frequency @xmath120 corresponds to the rotating - wave interaction .
the local minimum value of @xmath105 is obtained when @xmath121 , which yields the following frequency matching condition [ mat]@xmath122 where @xmath123 and @xmath124 are both odd integers or both even integers , and @xmath125 .
for example , @xmath126 , @xmath127 , @xmath128 , @xmath129 ... ; @xmath130 , @xmath131 , @xmath132 , @xmath133 ... we can label these islands in fig . [
fig3](a ) as @xmath134 , where we have marked @xmath135 as an example . from eq .
( [ mat ] ) we derive the corresponding @xmath136 and @xmath0 for each islands @xmath134 as @xmath137 because of the dissipation , the optimal minimum value is reached for the island with the shortest time , which corresponds to @xmath109 , the labeled island in the top left corner of fig .
[ fig3](a ) .
in this case we obtain @xmath138 and @xmath139 . some of the other minimum values for @xmath126 [ along the leftmost purple dashed curve in fig . [
fig3](a ) ] can be obtained as @xmath140 , @xmath141 , @xmath142 ... (@xmath143 , @xmath129 , @xmath144 ... ) in figs . [
fig3](b ) and ( c ) we plot @xmath105 for @xmath117 and @xmath118 , corresponding to the frequency matched and unmatched regions , respectively . for the former ,
the time evolution of phonon number exhibits periodic oscillations , with the minimum phonon number lower than @xmath145 ; for the latter , the phonon number does not show obvious periodicity , and the minimum phonon number is larger than @xmath146 .
although there exist some discrepancies between the analytical results expressed by eq .
( [ nb0 ] ) and the exact results numerically computed from eq .
( [ d ] ) in the form of some high frequency fluctuations , the analytical expression characterizes the minimum phonon number very well , especially for the frequency - matched case . for the islands with @xmath126 ,
the minimum phonon number is obtained near the end of the first half rabi oscillation cycle , @xmath147 . the instantaneous - state cooling limit in this case
is analytically given by @xmath148 where we have neglected the terms containing mechanical dissipation rate @xmath149 since in the experiments typically @xmath150 .
we plot @xmath151 as a function of @xmath1 in fig .
the overall trend is that large coupling strength leads to large cooling limit , which is a result of quantum backaction heating in the strong coupling regime .
the analytical expression eq .
( [ nins0 ] ) describes this trend quite well .
meanwhile , the oscillations are signatures of frequency matching , where @xmath152 , @xmath153 and @xmath141 correspond to the frequency - matched regions with low phonon number . the steady - state cooling limit reads @xmath154 $ ] , which is also plotted in fig .
[ fig4 ] for comparison . as a function of @xmath1 .
the red circles are numerical results obtained from eq .
( [ d ] ) and the red dotted curve is the analytical result calculated from eq .
( [ nins0 ] ) .
the steady - state cooling limit is plotted for comparison ( blue dashed curve ) .
the dotted vertical lines indicate @xmath155 , @xmath153 and @xmath156 from left to right .
other parameters are the same as fig . [
fig3](a ) . ] as functions of @xmath0 and @xmath1 .
the purple dashed curves correspond to the cases of @xmath106 , @xmath107 and @xmath108 .
the color bar is in log scale .
the black contour curves denote @xmath157 , @xmath158 , @xmath159 and @xmath160 .
( b ) @xmath161 as a function of @xmath0 for @xmath162 ( red dashed curve ) and @xmath118 ( blue dotted curve ) , which are numerical results obtained from eq .
( [ d ] ) ; @xmath9 as a function of @xmath0 for @xmath117 ( red solid curve ) and @xmath118 ( blue dash - dotted curve ) , which are numerical results obtained from eq .
( [ drwa ] ) .
other parameters are the same as fig .
[ fig1](a ) . ] for finite - temperature case @xmath163 , the phonon number originates from both the modified mechanical dissipation and quantum backaction . by solving eq .
( [ d ] ) , in fig .
[ fig5](a ) we plot the exact numerical results of the time evolution of the mean phonon number @xmath161 as functions of @xmath0 and @xmath1 .
the contour map shows that @xmath161 also exhibits _ _ rabi - like oscillations and a number of islands .
the main features of the rabi - like oscillation is similar to the rwa case as shown in fig .
[ fig1 ] , while the island appears only when both @xmath123 and @xmath124 are odd integers , corresponding to the rabi - like oscillation dips .
this is because the contribution of @xmath105 is hidden by @xmath9 except for the regions where @xmath9 reaches the minimum , i. e. , @xmath164 , corresponding to odd @xmath124 . in fig .
[ fig5](b ) we compare @xmath161 and @xmath9 for both @xmath117 and @xmath118 , which demonstrates the significance of frequency matching . for @xmath117 ( frequency - matched case ) ,
the difference between @xmath161 and @xmath9 is small and the minimum phonon number is less than @xmath165 .
however , for @xmath166 ( frequency - unmatched case ) , @xmath161 is different from @xmath9 near the rabi - oscillation dips , since the frequency matching condition is not satisfied . in this case
the quantum backaction plays an important role and the minimum phonon number is as large as @xmath167 . in fig .
[ fig6](a ) , we present the instantaneous - state cooling limit @xmath168 [ obtained near @xmath169 as a function of @xmath1 ( blue stars ) , which reveals obvious difference between the frequency matched and unmatched cases .
as a function of @xmath1 for @xmath170 .
( b ) @xmath168 as a function of @xmath91 for @xmath171 .
the red circles are numerical results obtained from eq .
( [ d ] ) with dynamic cavity dissipation modulation ; the red dashed and dotted curves are the analytical results calculated from eqs .
( [ nins ] ) and ( [ ninsopt ] ) ; the blue stars are the numerical results obtained from eq .
( [ d ] ) without dynamic cavity dissipation modulation .
other parameters are the same as fig .
[ fig1](a ) . ] by employing the dynamic dissipative cooling approach @xcite , the instantaneous - state cooling limit can be can be significantly reduced . in fig .
[ fig6](a ) we plot the reduced instantaneous - state cooling limits with dynamic cavity dissipation modulation as a function of @xmath1 for @xmath82 ( red circles ) .
analytical derivations lead to the cooling limits @xcite @xmath172 where @xmath173 denotes the instantaneous - state cooling limits for the optimal coupling strength @xmath1 ( frequency - matched case ) , and @xmath168 corresponds to the frequency - unmatched case .
it reveals that these two expressions set the upper and lower bounds of the instantaneous - state cooling limits , as shown in fig .
[ fig6](a ) .
it is worth noting that the lower bound is the sum of the two cooling limits given by eq .
( [ ninsrwa ] ) and eq .
( [ nins0 ] ) , i. e. , @xmath174 .
typically , the minimum cooling limit is obtained for @xmath175 , as shown in fig . [ fig6](a ) . in fig .
[ fig6](b ) the instantaneous - state cooling limits as functions of the cavity decay rate @xmath91 for the optimal coupling strength @xmath152 are plotted .
it shows that a small @xmath74 is preferred to obtain a low cooling limit .
therefore , to attain the best cooling performance in the strong coupling regime , the frequency matching condition should be satisfied and a small cavity decay rate is required .
this is in contrast to the steady - state cooling limit which requires an optimal value for @xmath91 to balance the quantum limit and cavity bandwidth limitation @xcite .
in summary , we have examined the backaction cooling of mesoscopic mechanical resonators in the strong coupling regime . when the rotating - wave approximation is satisfied , the mean phonon number undergoes simple damped rabi - like oscillations with the rabi frequency of @xmath84 and the exponential damping envelope scales as @xmath176 . at the first half oscillation cycle ,
the minimum mean phonon number is derived as @xmath177 . for large coupling strength where the rotating - wave approximation fails
, the mean phonon number oscillates as functions of both the evolution time and the coupling strength , which is a result of frequency mixing . under the frequency matching condition ,
the optimal coupling strengths are derived as @xmath178 , @xmath179 , @xmath180 ... by employing the dynamic dissipative approach @xcite , the reduced instantaneous - state cooling limits reach the lower bound @xmath181 $ ] .
this provides a guideline for achieving the lowest cooling limit , which is reached when the frequency matching condition is satisfied and when the cavity decay rate @xmath91 is small .
compared with the steady - state cooling limit , the unique advantage is that it does not require an optimal value for @xmath91 , allowing for long - coherence - time quantum operations deeply in the strong coupling regime .
the parameter ranges @xmath182 can be realized in various optomechanical systems , for example , in the microtoroid system studied in ref .
@xcite , with the parameters @xmath183 @xmath184 , @xmath185 @xmath184 , @xmath186 @xmath187 and @xmath188 @xmath184 , or in the superconducting aluminium membrane system studied in ref .
@xcite , with the parameters @xmath189 @xmath184 , @xmath190 @xmath187 , @xmath191 @xmath192 and @xmath193 .
this work is supported by the 973 program ( 2013cb328704 , 2013cb921904 ) , nsfc ( 11004003 , 11222440 , and 11121091 ) , and rfdph ( 20120001110068 ) .
y.c.l is supported by the scholarship award for excellent doctoral students granted by the ministry of education .
y.f.s was supported by the national fund for fostering talents of basic science ( grants no .
j1030310 and no .
j1103205 ) @xmath194{ccccc}-{\kappa } & 0 & -ig & ig^{\ast } & ig^{\ast}\\ 0 & -{\gamma } & ig & -ig^{\ast } & ig^{\ast}\\ -ig^{\ast } & ig^{\ast } & -i\left ( \delta^{\prime}+{\omega_{\mathrm{m}}}\right ) -\frac{{\kappa+\gamma}}{2 } & 0 & 0\\ ig & -ig & 0 & i\left ( \delta^{\prime}+{\omega_{\mathrm{m}}}\right ) -\frac{{\kappa+\gamma}}{2 } & 0\\ -ig & -ig & 0 & 0 & i\left ( \delta^{\prime}-{\omega_{\mathrm{m}}}\right ) -\frac{{\kappa+\gamma}}{2}\\ ig^{\ast } & ig^{\ast } & 0 & 0 & 0\\ 0 & 0 & 0 & -2ig & -2ig\\ 0 & 0 & 2ig^{\ast } & 0 & 0\\ 0 & 0 & -2ig & 0 & -2ig^{\ast}\\ 0 & 0 & 0 & 2ig^{\ast } & 0 \end{array } \right .
\nonumber\\ & \left .
\begin{array } [ c]{ccccc}-ig & 0 & 0 & 0 & 0\\ -ig & 0 & 0 & 0 & 0\\ 0 & 0 & -ig & ig^{\ast } & 0\\ 0 & ig^{\ast } & 0 & 0 & -ig\\ 0 & -ig^{\ast } & 0 & -ig & 0\\ -i\left ( \delta^{\prime}-{\omega_{\mathrm{m}}}\right ) -\frac{{\kappa+\gamma } } { 2 } & 0 & ig & 0 & ig^{\ast}\\ 0 & 2i\delta^{\prime}-{\kappa } & 0 & 0 & 0\\ 2ig^{\ast } & 0 & -2i\delta^{\prime}-{\kappa } & 0 & 0\\ 0 & 0 & 0 & -2i{\omega_{\mathrm{m}}}-{\gamma } & 0\\ 2ig & 0 & 0 & 0 & 2i{\omega_{\mathrm{m}}}-{\gamma}\end{array } \right),\end{aligned}\ ] ] t. j. kippenberg and k. j. vahala , * 321 * , 1172 ( 2008 ) ; f. marquardt and s. m. girvin , * 2 * , 40 ( 2009 ) ; m. aspelmeyer , t. j. kippenberg , f. marquardt , : 1303.0733 ( 2013 ) ; y .- c .
liu , y .- w .
hu , c. w. wong and y .- f .
xiao , * 22 * , 114213 ( 2013 ) . m. d. lahaye , o. buu , b. camarota , k. c. schwab , * 304 * , 74 ( 2004 ) ; g. anetsberger _
et al . _ , * 5 * , 909 ( 2009 ) ; j. d. teufel , r. donner , m. a. castellanos - beltran , j. w. harlow , and k. w. lehnert , * 4 * , 820 ( 2009 ) ; a. g. krause , m. winger , t. d. blasius , q. lin , o. painter , * 6 * , 768 ( 2012 ) ; t. p. purdy , r. w. peterson , and c. a. regal , * 339 * , 801 ( 2013 ) ; y .- d .
wang and a. a. clerk , * 108 * , 153603 ( 2012 ) ; l. tian , * 108 * , 153604 ( 2012 ) ; s. barzanjeh , m. abdi , g. j. milburn , p. tombesi , and d. vitali , * 109 * , 130503 ( 2012 ) ; j. t. hill , a. h. safavi - naeini , j. chan , and o. painter , * 3 , * 1196 ( 2012 ) ; c. dong , v. fiore , m. c. kuzyk , and h. wang , * 338 , * 1609 ( 2012 ) ; y. liu , m. davanco , v. aksyuk , and k. srinivasan , * 110 * , 223603 ( 2013 ) .
d. vitali _ et al .
_ , * 98 * , 030405 ( 2007 ) ; l. tian , * 110 * , 233602 ( 2013 ) ; y .- d .
wang and a. a. clerk , * 110 * , 253601 ( 2013 ) ; h .- k li , x .- x ren , y .- c .
liu , y .- f .
xiao , * 88 * , 053850 ( 2013 ) ; m. c. kuzyk , s. j. van enk , and h. wang : 1307.2844 ( 2013 ) .
p. rabl , * 107 * , 063601 ( 2011 ) ; a. nunnenkamp , k. brkje , and s. m. girvin , * 107 * , 063602 ( 2011 ) ; m. ludwig , a. h. safavi - naeini , o. painter , and f. marquardt , * 109 * , 063601 ( 2012 ) ; m .- a .
lemonde , n. didier , and a. a. clerk , * 111 * , 053602 ( 2013 ) ; k. brkje , a. nunnenkamp , j. d. teufel , and s. m. girvin , * 111 * , 053603 ( 2013 ) ; y .- c .
liu , y .- f .
xiao , y .- l .
chen , x .- c .
yu , and q. gong , * 111 * , 083601 ( 2013 ) ; a. kronwald and f. marquardt , * 111 * , 133601 ( 2013 ) . s. mancini , d. vitali , and p. tombesi , * 80 * , 688 ( 1998 ) ; p .- f .
cohadon , a. heidmann , and m. pinard , * 83 * , 3174 ( 1999 ) ; d. kleckner and d. bouwmeester , * 444 * , 75 ( 2006 ) ; t. corbitt _
et al . _ , * 99 * , 160801 ( 2007 ) ; m. poggio , c. l. degen , h. j. mamin , and d. rugar , * 99 * , 17201 ( 2007 ) . f. elste , s. m. girvin , and a. a. clerk , * 102 * , 207209 ( 2009 ) ; m. li , w. h. p. pernice , and h. x. tang , * 103 * , 223901 ( 2009 ) ; a. xuereb , r. schnabel , and k. hammerer , * 107 * , 213604 ( 2011 ) ; t. weiss and a. nunnenkamp , * 88 * , 023850 ( 2013 ) ; m .- y .
yan , h .- k .
li , y .- c .
liu , w .- l
. jin , and y .- f .
xiao , * 88 * , 023802 ( 2013 ) .
m. r. vanner _
_ , * 108 * , 16182 ( 2011 ) ; y. li , l .- a .
wu , and z. d. wang , * 83 * , 043804 ( 2011 ) ; j .- q .
liao and c. k. law , * 84 * , 053838 ( 2011 ) ; x. wang , s. vinjanampathy , f. w. strauch , and k. jacobs , * 107 * , 177204 ( 2011 ) ; s. machnes _ et al .
_ , * 108 * , 153601 ( 2012 ) .
|
laser cooling of mesoscopic mechanical resonators is of great interest for both fundamental studies and practical applications .
we provide a general framework to describe the cavity - assisted backaction cooling in the strong coupling regime . by studying the cooling dynamics
, we find that the temporal evolution of mean phonon number oscillates as a function of the optomechanical coupling strength depending on frequency mixing .
the further analytical result reveals that the optimal cooling limit is obtained when the system eigenmodes satisfy the frequency matching condition .
the reduced instantaneous - state cooling limits with dynamic dissipative cooling approach are also presented .
our study provides a guideline for optimizing the backaction cooling of mesoscopic mechanical resonators in the strong coupling regime .
|
multi - agent systems such as crowds of human @xcite , schools of fish @xcite , flocks of birds @xcite , colonies of molds @xcite and ants @xcite often exhibit discrete phase transitions due to variations of the driven force of interaction among members @xcite .
specially , detecting transitions in humans @xcite is an ongoing research problem @xcite .
abrupt changes upon variation of some parameters such as speed , coordination and structure @xcite shift the system from one state to another @xcite .
numerous types of swarm decisions which determine the group dynamics are not only influenced by the intrinsic social interaction among members @xcite , but also by some outside factors such as threats @xcite and presence of predators or food sources @xcite .
however , with a majority of observation done via videos , the classical approaches of detecting transitions , such as , tracking individuals and monitoring their dynamics , are constrained by the size of the group and the scale of the problem @xcite .
being inspired by manifold representation of collective motion , here , we develop a method of detecting phase transitions in a multi - agent system as it switches from one unique behavior to another . in manifold theory , we can reveal an invariant manifold , @xmath0 , in an abstract higher dimensional space which describes the collective behavior of a group , such that each frame , partitioned from the collective motion video , corresponds to a point @xmath1 on the manifold @xcite . in this setup
, the whole group evolves according to the underlying flow @xmath2 at time @xmath3 @xcite . due to behavioral changes of the group
, this mapping switches from one phase space to another and indicates a transition of the motion .
thus , in the presence of a phase transition , the system can be represented as two distinct sub - manifolds , @xmath4 for @xmath5 and @xmath6 , along with singularities where the transition physically exists @xcite . here
, the most salient scenario is that the sub - manifolds intersect and make a locus @xmath7 of singularities as @xmath8 as an example , we estimate the dimensionalities of two distinct phases of a human crowd given as a video in @xcite that switches phase from walking ( fig .
[ fig : crowd_walk_run](a ) ) to running ( fig .
[ fig : crowd_walk_run](b ) ) .
we utilize an established dimensionality reduction routine called isomap @xcite to obtain corresponding scaled residual variances of two embedding spaces ( fig . [
fig : crowd_walk_run](c ) ) of each phase .
figure [ fig : crowd_walk_run](c ) shows that the two phases are embed on manifolds with different dimensionalities .
this example acts as a proof - of - concept for developing an routine to detect phase transitions . detecting transitions
should be naively implemented on videos before utilize dimensionality reduction schemes such as isomap @xcite , local linear embedding @xcite , as those may otherwise contain transitions .
a transition in a multi - agent system is defined as a switching of the current smooth embedding manifold into a another smooth manifold with different dimensionality as trajectories of agents evolve in the phase space .
our approach of detecting transitions is based on revealing high curvature on the manifold which differentiate phases of the motion .
we hypothesize that a phase transition is manifested in the form of change in local curvature that can be detected by a ratio of singular values from sampled points .
we first formalize this concept in two dimensions by proving a relation between curvature and singular value ratio in a curve , and then extend it to higher - dimensions by using the shape operator .
then , we justify that the same transition can also be detected by analyzing the ratios of the smallest singular value to the largest singular value which are computed locally upon neighborhoods of points sampled on the manifold . based on the distribution of a moving sum of absolute moving difference of the singular value ratios ,
transitions are detected and their magnitudes are ordered .
this paper is organized as follows : section [ sec : method ] describes the method of detecting phase transitions . in section [ sec : algorithm ] , the method along with the detailed algorithm of detecting transitions in collective behavior is presented .
section [ sec : examples ] describes performance of the method using one synthetic dynamical simulation of the vicsek model and three natural experimental data sets of a human crowd , a birds flock , and a fish school .
we conclude the work in section [ sec : conclusion_discussion ] with a discussion of the method performance and future work .
we declare that , in presence of a phase transition , the curvature of the underlying manifold is abruptly changed .
therein , we first approximate the point - wise curvature of a curve in two dimensions and then extend it to higher dimensions using the shape operator . finally , we attest that the same transition can also be approximated locally by singular value ratios computed on the data sampled on the manifold
. point - wise curvature of a curve is approximated by using singular value ratio computed over the data sampled in a neighborhood at the point .
we assume that the data is evenly distributed along the curve . as in figure [ fig : trans_circle ] , for any given point @xmath1 on the curve , we can superimpose a neighborhood in it with an arc @xmath9 of a translating circle such that the arc subtend a small angle @xmath10 at the center , @xmath11 .
we approximate the curvature of the translating circle instead of that of the curve since they are similar in aforesaid setup . at point @xmath1 with an arc @xmath9 which subtend an small angle of @xmath10 at the origin of a translating circle . ]
[ fig : trans_circle ] [ thm : curvature ] given a circle centered at the origin with radius @xmath12 , let @xmath13 number of points are uniformly distributed with density @xmath14 on an arc which subtends an angle of @xmath10 at the center .
let @xmath15 and @xmath16 , such that @xmath17 , are two singular values computed upon the data on the curve , then , @xmath18 where @xmath19 , and @xmath20 .
a point @xmath21 on the arc is given by @xmath22 , @xmath23 for @xmath24 $ ] .
expected values of variables @xmath25 and @xmath26 are computed as @xmath27 and @xmath28 respectively .
we then compute pairwise covariance between variable as @xmath29 @xmath30 @xmath31 and @xmath32 the covariance matrix , @xmath33 , of the data is @xmath34 the covariance matrix which is approximated to @xmath35 order of @xmath36 by using the taylor s expansion is @xmath37 let , eigenvalues of @xmath38 are @xmath39 and @xmath40 , such that @xmath41 , then , @xmath42 as eigenvalues are computed upon the covariance matrix , they also relate to principal components , denoted by @xmath15 and @xmath16 where @xmath17 , such that , @xmath43 @xcite . then , @xmath44 further , @xmath45 since @xmath36 is small .
let , @xmath13 points are uniformly distributed with the density @xmath14 on the arc , then , @xmath46 by ( [ eqn : sigma_ratio ] ) and ( [ eqn : value_t ] ) , @xmath47 we extend the two dimensional assertion into higher dimensions by intersecting principal sections , made through the shape operator , with the manifold .
extrinsic curvatures of a manifold in orthogonal tangential directions are measured using eigenvalues and eigenvectors of the shape operator @xcite , such that , eigenvalues provide magnitudes and eigenvectors provide directions of principal curvatures along tangential directions @xcite .
let @xmath48 represents the parametric form of the manifold embedded in the euclidean space @xmath49 .
for @xmath50 , @xmath51 provide mutually orthogonal tangent and unit tangent , vectors of @xmath52 at @xmath1 respectively .
thus , @xmath53 is an orthonormal basis for the tangential space at @xmath1 .
the shape operator , @xmath54 , of the manifold @xmath55 at the point @xmath1 is defined as @xmath56 @xcite .
let ( @xmath57 ) for @xmath50 denote eigenpairs of @xmath54 , then , magnitudes and directions of principal curvatures are given by @xmath58 and @xmath59 respectively @xcite .
the two - dimensional plane in @xmath49 which is spanned by the principal direction @xmath60 and the unit normal at @xmath1 , @xmath61 , is defined as the @xmath62-th principal section , @xmath63 thus , for @xmath50 , @xmath64 are mutually orthogonal planes .
intersection of @xmath64 s with the manifold makes curves @xmath65 such that each passes through @xmath1 . while figure [ fig : curvature_local ] illustrates this scenario when @xmath66 , a computational example of producing principal sections is attached in appendix [ app : shape_operator ] . without loss of generality , for @xmath50
, we assume that the data is distributed uniformly with a sufficient density on @xmath67 to contain @xmath13 points in a neighborhood at the point @xmath1 . for all @xmath62 , singular values @xmath68 and
@xmath69 , such that @xmath70 , are computed upon the data sampled in the neighborhood of @xmath1 on @xmath67 . by theorem [ thm : curvature ] , curvature @xmath71 on the curve is approximated as @xmath72 for some @xmath73 .
a phase transition differentiates the manifold into two sub - manifolds such that each map with a different euclidean space @xcite .
thus , under a transition , geometry permits that the curvature of some curves , @xmath67 s , undergo abrupt changes which also result abrupt changes of the ratios @xmath74 .
@xmath75 as a simple geometric example to illustrate high curvature on a manifold at a transition , we use a three - dimensional joined - manifold that we call a ` sombrero - hat ' ( fig .
[ fig : example_sombrero](a ) ) of 2000 points produced by the equations @xmath76 for @xmath77 $ ] and @xmath78 $ ] .
we observe this sombrero - hat joins two sub - manifold , brim ( green ) and crown ( blue ) along with the locus ( red ) representing a transition . instead of constructing principal sections using the shape operator , for simplicity , we intuitively find a principal section in this example at the point @xmath79 .
since the curvature of the manifold is same in all tangential direction at this point , we choose one principal direction is @xmath80 , the unit vector along the @xmath25-axis .
the unit normal at this point is @xmath81 which is the unit vector along the @xmath82-axis .
thus , we define the principal section as the plane @xmath83 . intersection of this principal section with the sombrero - hat gives a curve given in [ fig : example_sombrero](b ) which has a high curvature at red points .
isomap residual plots indicate different embedding dimensionalities at the locus than those of two sub - manifolds . to produce the curve ( b ) .
isomap residual plots ( c ) those show embedding dimensionalities by elbows reveal that the dimensionalities of two sub - manifolds ( blue and green ) are two while the dimensionality of the locus ( red ) is three.,scaledwidth=100.0% ] as constructing the shape operator at a neighborhood of each point on the manifold is computationally expensive , here we present an alternative approach to compute the singular value ratios and detect transitions .
therein , we first make a neighborhood @xmath84 at each point , @xmath1 , such that it contains @xmath85 points by running nearest neighbor search algorithm given in @xcite .
then , we perform singular value decomposition for some @xmath85 , and two unitary matrices @xmath86 and @xmath87 , such that @xmath88 , those provide the decomposition @xmath89 for @xmath90 with @xmath91 @xcite . ] for data in each @xmath92 and denote the descending order by @xmath93 for some @xmath85 . without loss of generality ,
we assume that the data distribution in each @xmath67 is dense enough for @xmath92 to contain data from each curve , @xmath67 ; @xmath50 , made in section [ sub : curvature_mani ] .
however , @xmath92 contains at least one point , @xmath1 , from each @xmath67 as shown in figure [ fig : curvature_local ] which depicts this scenario for a two dimensional manifold , @xmath94 , in @xmath95 . according to this setting
, we can assert that the abrupt changes which are detected by the ratio @xmath74 in some curves created through @xmath1 are also detected by the ratio @xmath96 computed over the data in @xmath92 on the manifold , @xmath52 . for clarity , from here onward ,
we denote the point @xmath1 on the manifold as the @xmath97-th point .
let consider @xmath98 data points are embedded on the manifold , thus , the distribution , @xmath99 provides the magnitudes of the phase at each point .
a group of agents evolving with mutual interaction can be explained in terms of a hidden manifold structure such that each frame corresponds to a single data point on the manifold .
behavioral changes between consecutive frames are measured by moving differences of singular value ratios computed over neighborhoods of corresponding points as @xmath100 since we observe the distribution @xmath101 is highly volatile , we utilize @xmath13-points moving sum , @xmath102 \cap \mathbb{n } } t_n \hspace{10pt } ; \hspace{10pt } n\in \left[\frac{\alpha}{2 } , n-1-\frac{\alpha}{2}\right]\cap \mathbb{n},\ ] ] to smooth it . after a transition occurs ,
manifold curvature is abruptly changed and then so the singular values .
thus , a transition at a point can be quantified by the magnitude of @xmath103 . on a two dimensional manifold ( @xmath94 ) .
principal sections @xmath104 and @xmath105 are created by using the shape operator around @xmath1 , and curves @xmath106 and @xmath107 are produced by intersecting @xmath104 and @xmath105 with @xmath94 . ]
for this study , we use frames partitioned from a collective motion videos as data .
the total pixel dimension of a frame is the dimensionality of the space where the manifold is embedded in .
this algorithm requires two inputs , one is the parameter @xmath13 for number of nearest neighbors , and the other is data matrix @xmath108 constructed as bellow .
we assume unprocessed data given is grayscale images partitioned from a video of collective motion .
we reshape each frame into a row matrix and produce the data matrix @xmath109 by concatenating each row matrix vertically such that @xmath97-th row in @xmath109 represents the pixel intensities of the @xmath97-th frame , for some @xmath97 @xcite . since @xmath97-th row in @xmath109 is the @xmath97-th point on the manifold , euclidean distance , @xmath110 , between any two points , @xmath97 and @xmath111 , on the manifold is computed as @xmath112 based on this distance , @xmath13 nearest neighbors for the @xmath97-th point are extracted as @xmath113 ( same as @xmath92 in section [ sub : local_approximation ] ) .
then , we compute @xmath114 from @xmath113 as explained in the section [ sub : local_approximation ] .
algorithm outputs magnitudes of transitions , @xmath115 , so we can choose the largest among them as significant transitions .
the method of detecting transitions is given as algorithm [ alg : detecting_transitions ] .
perform nearest neighbor search in @xcite to obtain @xmath13 nearest neighbors for each frame in @xmath109 . here , we denote @xmath13 nearest neighbors of the @xmath97-th frame in @xmath109 by the set @xmath113 .
perform singular value decomposition in @xcite on @xmath113 and denote the descending order of singular values by @xmath116 for some @xmath85 . compute the ratio @xmath117 for all @xmath113 and denote by @xmath118 where @xmath119 .
compute absolute moving difference of ratios , @xmath120 , for @xmath121 . compute @xmath122points
moving sum , @xmath123 \cap \mathbb{n } } t_n$ ] , of @xmath101 for @xmath124\cap \mathbb{n}$ ] .
largest values of @xmath103 are extracted as significant transitions .
in this section , we evaluate the transition detection algorithm on four datasets : a synthetic dynamical simulation of the vicsek model , and three natural data sets ; a crowd of human , a flock of birds and a school of fish .
a swarm of self propelled particles with three imposed transitions simulated by the vicsek model @xcite is examined to investigate the sensitivity of the algorithm .
since we simulate positions of particles in each frame , to avoid ambiguity , notation for set of nearest neighbors is moderated as @xmath125 to denote all the nearest - neighbors of the @xmath126-th agent within a unit distance at the @xmath97-th time - step ( each time - step will be converted into a frame later ) .
the vicsek model @xcite updates the orientation @xmath127 of the @xmath126-th agent at the @xmath97-th time step as @xmath128 and @xmath129 is the noise parameter sampled from a gaussian distribution with mean zero and standard deviation @xmath130 . here , @xmath131 is the average direction of motion of all the agent in @xmath132 . the position @xmath133 of the @xmath126-th agent at the @xmath97-th time step is therefore updated as @xmath134 where @xmath135 is the time - step size and @xmath136 is the speed of the @xmath126-th agent at the time - step ( @xmath1371 ) .
we run a simulation of this model and generate a synthetic data set of a particle swarm of 50 agents in 200 time - steps with periodic boundary conditions .
we set @xmath138 , @xmath139 for all @xmath126 and @xmath97 , and impose three transitions by varying the noise , @xmath140 , using a variable standard deviation , @xmath141 for all @xmath126 in the gaussian distribution . for @xmath142 , we convert point - mass positions of particles at the @xmath97-th time - step , @xmath143 , into a graycolored frame such that each particle is represented by a black square of pixels @xmath144 .
therein , we first convert the positions of particles into a sparse matrix , @xmath145 , at each time step @xmath142 .
we make a rotationally symmetric gaussian low - pass filter @xmath146 of size @xmath144 pixels with standard deviation of 10 pixels and filter the matrix @xmath145 by @xmath146 to generate @xmath147 for @xmath142 @xcite .
then , each matrix , @xmath148 , is converted into a frame , and all the frames are finally converted into a one data matrix , @xmath109 .
we run transition detection algorithm upon frames with several @xmath13 values and observe that the distribution @xmath149 generated by @xmath150 ( fig .
[ fig : example_viksec](a ) ) shows clear transitions with respect to noise changes .
the plot of descending order magnitudes of 20 largest transitions ( fig .
[ fig : example_viksec](b ) ) reveals that three transitions at frame numbers @xmath151 and @xmath152 are significant than others .
thus , the manifold representing the whole motion is partitioned into transition free 4 sub - manifolds ranging @xmath153 and @xmath154 .
we run isomap on each range of frames with the neighborhood parameter four and obtain the residual plots given in figure [ fig : example_viksec](c ) .
the isomap residual plots indicating the embedding dimensionality by an elbow , reveal that the sub - manifolds are embedded in three , six , two and four dimensions respectively . versus frame numbers .
therein , the range of frames representing each sub - manifold is marked by a left - right arrow and the frame containing each transition is marked by a red circle along with the frame number associated .
( b ) descending order of 20 largest transitions including frame numbers of three largest transitions .
( c ) isomap residual variance versus dimensionality of each sub - manifold .
, scaledwidth=100.0% ] a video of a human crowd available on - line at @xcite containing a transition at the @xmath155-th frame from walk to run is considered next . as example
[ sec : example_vicsek ] , we first convert all the frames into a data matrix , @xmath109 , and then run transition detection algorithm on it .
when @xmath156 , we observe that the distribution ( @xmath157 given in figure [ fig : example_crowd](a ) shows clear trade - off about the transition .
figure [ fig : example_crowd](b ) shows only the first transition , having a magnitude of @xmath158 , is significant .
the manifold representing the crowd is partitioned such that frames @xmath159 represent one sub - manifold and frames @xmath160 represent the other sub - manifold .
the snapshots in figure [ fig : example_crowd](a ) show the phases , walking ( left ) and running ( right ) , of the crowd .
isomap running on frames @xmath159 and @xmath160 with @xmath156 reveals that embedding dimensionalities of sub - manifolds are three and four respectively . versus frame numbers .
therein , left - right arrows and the red circle represent ranges of frames in each sub - manifold and the frame at the transition respectively , while the snapshots show instances in each phase .
( b ) plot of descending order of 20 largest transitions representing the largest transition in red along with the frame number .
, scaledwidth=100.0% ] we now detect a transition , differentiating phases of sitting and flying , at the @xmath161-th frame in a video of a bird flock obtained on - line at @xcite .
we run transition detection algorithm on the data matrix with @xmath156 and obtain the distribution ( @xmath157 in figure [ fig : example_birds](a ) . magnitudes of the descending order of 20 largest transitions in figure [ fig : example_birds](b ) illustrates that one transition with magnitude of @xmath162 is significant at the @xmath161-th frame .
then , the manifold is partitioned into two sub - manifolds representing frames in the ranges @xmath163 and @xmath164 .
isomap running on ranges of frames @xmath163 and @xmath164 with @xmath156 reveals that the embedding dimensionalities of corresponding sub - manifolds are two and three .
shows ranges of frames representing two sub - manifolds by left - right arrows and instances of phases by snapshots .
( b ) descending order of 20 largest transitions .
the frame at the largest transition is represented by red in figures ( a ) and ( b).,scaledwidth=100.0% ] finally , we use a video of a fish school having few transitions to validate the method .
this school is stimulated with panics and the behavior is recorded .
we run the algorithm on the data matrix with @xmath165 and obtain the distribution @xmath166 ( fig .
[ fig : example_fish](a ) ) . the plot of transitions ( fig .
[ fig : example_fish](c ) ) reveals that the first four transitions located at frames @xmath167 and @xmath168 are significant .
now , the manifold is partitioned into transition free sub - manifolds representing frames in ranges @xmath169 and @xmath170 as shown by left - right arrows in figure [ fig : example_fish](a ) . according to figure [ fig : example_fish](b )
, the pair of snapshots at the 40-th frame exhibits a large global abrupt change since the school reacts to the panic together .
however , by the evidence observed through figure [ fig : example_fish](b ) and snapshots , the second and the third transitions are abrupt and local while the last is gradual and local .
isomap is run on all sub - manifolds with @xmath165 and embedding dimensionalities are obtained as two , four , three , two and six respectively . with left - right arrows showing ranges of frames in sub - manifolds , and red dots showing frames at transitions .
( b ) snapshots of the school before ( left ) and after ( right ) each transition .
( c ) descending order of 20 largest transitions consisting four largest transitions marked in red with their frame numbers.,scaledwidth=100.0% ]
in this study , we propose a robust method to detect phase transitions in collective motion using ratio of singular values encountering high curvature of a underlying manifold .
therein , we first introduce an assertion to approximate the curvature of a curve by means of singular value ratios computed on it and then we extend the assertion to higher dimensions using the shape operator . finally , we assert that the same transition can also be detected through singular value ratios computed over the local data distribution on the manifold .
we validate the method through four diverse examples ; one from a simulation of the vicsek model @xcite containing three transitions , and the other three from natural instances of a human crowd , a bird flock and a fish school .
we run the transition detection algorithm on each data set with a predetermined nearest neighbor parameter ( @xmath13 ) .
algorithm outputs the distribution @xmath171 and 20 largest transitions which are ranked according their magnitudes . based on these outputs , manifold representing the whole collective motion
is partitioned into transitions free sub - manifolds . in example
[ sec : example_vicsek ] , we simulate a training data set of a particle swam consisting known transitions to test the method s performance .
noteworthy shifts of the ratios @xmath172 from one transition to another justify the sensitivity of the method .
therein , our method is capable of detecting the exact transitions at frames 50 and 150 , however , it approximates the transition at @xmath173-th frame as at @xmath174-th frame with -1 frame of error .
since this approximation is accurate enough , we partition the manifold into four sub - manifolds about frames 50 , 99 and 150 . thus
, the method is adept at perceiving the nature of the collective motion and spatial distribution of agents to approximate transitions .
the algorithm running on a fish school consisting four transitions ( ex . [ sec : example_fish ] ) ranks their magnitudes such that it provides a good comparison between them .
embedding dimensionalities of sub - manifolds in each example revealed by isomap affirm that the data is embedded in distinct smooth sub - manifolds which are joined at singularities . examples also ensure the applicability of the method for variety of data sets ranging from simulations to natural , as it requires one input parameter for @xmath13 .
as we detect transitions based on the curvature which is a local property on the manifold , choosing the best value @xmath13 is always important for the method to extract correct traditions .
this is done by either using the prior - knowledge about the data or running the algorithm with several @xmath13 s to determine the best value which differentiates and highlights transitions .
we can always rely on a small @xmath13 value when the data on the manifold is sufficiently dense @xcite .
since the current techniques of making videos with high frame rates , we can always generate sufficiently large quantity of frames those would then be densely represented as data points on a manifold .
thus , in this study , we run the transition detection algorithm with small @xmath13 values taken to be all natural numbers less than 10 and then decide the best value by analyzing corresponding transition plots . in the future
, we will fabricate an automated method to determine the best @xmath13 value which is giving the best transition plot to extract transitions .
we will also establish a novel approach of detecting transitions under gradual changes of the behavior of a multi - agent system which is not addressed in this work .
here we presented a transitions detection method of multi - agent systems using the curvature of the manifold representing the system .
the method was validated on several instances ranging from simulation to natural to justify the broad applicability and the accuracy .
kelum gajamannage and erik m. bollt are supported by the national science foundation under grant no .
cmmi- 1129859 .
here , we provide an example of computing principal sections of a two dimensional manifold , called a saddle surface , in @xmath95 given in parametric form , @xmath175 shown by figure [ fig : principal_curvature_saddle ] .
for an arbitrary point , @xmath177 , on the surface , we have , @xmath178 ) \implies \hat{\bs{v}}_{\bs{p}}^{(1)}=\frac{\bs{v}_{\bs{p}}^{(1)}}{|\bs{v}_{\bs{p}}^{(1)}|}=\frac{(1 , 0 , 3[x_1 ^ 2-x_2 ^ 2])}{\sqrt{9(x_1 ^ 2-x_2 ^ 2)^2 + 1}},\ ] ] and @xmath179 then , while the unit normal at @xmath1 , @xmath61 , is @xmath180 , 6x_1x_2 , 1)}{\sqrt{9(x_1 ^ 2+x_2 ^ 2)^2 + 1}},\ ] ] tangential directions are @xmath181 , x_2[9\{x_1 ^ 4-x_2 ^ 4\}-1 ] , 3x_1[x_1 ^ 2+x_2 ^ 2]\right)}{\left[9(x_1 ^ 2+x_2 ^ 2)^2 + 1\right]^{3/2}},\ ] ] and @xmath182 , x_1[9\{x_1 ^ 4-x_2 ^ 4\}+1 ] , -3x_2[x_1 ^ 2+x_2 ^ 2]\right)}{\left[9(x_1 ^ 2+x_2 ^ 2)^2 + 1\right]^{3/2}}.\ ] ] by ( [ eqn : first_tangents ] ) , ( [ eqn : second_tangents ] ) , ( [ eqn : x_derivative_of_n ] ) and ( [ eqn : y_derivative_of_n ] ) , we compute @xmath183[9(x_1 ^ 2-x_2 ^ 2)^2 + 1]}},\ ] ] @xmath184[36x_1 ^ 2x_2 ^ 2 + 1]}},\ ] ] @xmath185[9(x_1 ^ 2-x_2 ^ 2)^2 + 1]}},\ ] ] @xmath186[36x_1 ^ 2x_2 ^ 2 + 1]}}.\ ] ] particularly , the shape operator , @xmath187 , at @xmath188 is @xmath189 computed eigenpairs , @xmath190 and @xmath191l , of @xmath192 describe two pairs of principal curvatures and directions in the tangential space at @xmath188 .
these two dimensional principal directions are made to be three dimensions as @xmath193 and @xmath194 by introducing the third dimension .
the unit normal at @xmath188 , @xmath195 , is @xmath196 , thus , the principal sections in @xmath95 at @xmath1 are given by @xmath197 data set of detection of unusual crowd activity available at robotics and vision laboratory , department of computer science and engineering , university of minnesota .
available from : http://mha.cs.umn.edu/proj_events.shtml .
e. l. andrade , s. blunsden , and r. b. fisher .
hidden markov models for optical flow analysis in crowds . in _
pattern recognition , 2006 .
icpr 2006 .
18th international conference on _ , ieee , * 1 * ( 2006 ) , 460463 .
m. ballerini , n. cabibbo , r. candelier , a. cavagna , e. cisbani , i. giardina , a. orlandi , g. parisi , a. procaccini , m. viale , et al . empirical investigation of starling flocks : a benchmark study in collective animal behaviour . , * 76 * ( 2008 ) , 201215 . p. n. yianilos . data structures and algorithms for nearest neighbor search in general metric spaces . in _ proceedings of the fourth annual acm - siam symposium on discrete algorithms _ , society for industrial and applied mathematics , ( 1993 ) , 311321
t. zhao and r. nevatia .
tracking multiple humans in crowded environment . in _
computer vision and pattern recognition , 2004 .
cvpr 2004 .
proceedings of the 2004 ieee computer society conference on _ , ieee , ( 2004 ) , ii406 .
|
if a given behavior of a multi - agent system corresponds to restrict the phase variable to a invariant manifold , then we define a phase transition as change of physical characteristics such as speed , coordination and structure .
we define such a phase transition as splitting an underlying manifold into two sub - manifolds with distinct dimensionalities around the singularity where the transition is physically represented on . here
, we propose a method of detecting phase transitions and splitting the manifold into transitions free sub - manifolds .
therein , we utilize a relationship between curvature and singular value ratio of points sampled in a curve , and then extend the assertion to higher dimensions using the shape operator .
then we attest that , the same transition can also be approximated by singular value ratios computed locally over the data in a neighborhood on the manifold .
we validate the transitions detection method using one particle simulation and three real world examples .
kelum gajamannage erik m. bollt ( communicated by the associate editor name )
|
in dimensions greater than one , the spectrum of the i.i.d . random conductance laplacian displays a sharp transition between complete localization and complete homogenization .
this is the result of the present paper in combination with recent papers from flegel , heida , and slowik @xcite and neukamm , schffner , and schlmerkemper @xcite .
while the other two papers cover spectral homogenization , we investigate the localization phase .
a simple moment condition distinguishes between the two phases . more precisely , we investigate the spectrum of the laplacian @xmath6 associated with the random conductance model on the euclidean lattice @xmath0 .
the laplacian acts on real - valued functions @xmath7 as @xmath8 we assume that the conductances @xmath2 are positive , independent and identically distributed ( i.i.d . )
random variables .
we describe the almost - sure behavior of the principal eigenvector with zero dirichlet conditions outside a growing centered ball @xmath9 .
it turns out that its behavior strongly depends on the lower tails of the conductances . to be more precise ,
let us denote the expectation with respect to the conductances by @xmath10 and define @xmath11 < \infty \}$ ] .
then we show that , under some further regularity assumptions , @xmath12 on the other hand , as a special case the results in @xcite and @xcite imply that @xmath13 we comment on this in section [ subsec : literature ] . together , and imply that in the i.i.d .
random conductance model there is a dichotomy between a completely homogenized and a completely localized phase .
moreover , it is remarkable that the critical exponent @xmath4 coincides with the critical exponent for the validity of a local central limit theorem ( lclt ) of the corresponding random walk ( see e.g. @xcite ) : @xmath14 the validity of a local clt is a very strong kind of heat - kernel homogenization .
but in contrast to the principal dirichlet eigenvector , the heat kernel does not display such a completely different behavior for @xmath15 .
although the heat kernel decays anomalously for @xmath16 small enough @xcite , a quenched functional clt ( qfclt ) still holds under minimal assumptions on the i.i.d .
environment @xcite .
this is indeed _ not _ a contradiction since the qfclt associates with macroscopic properties of the random walk , whereas the anomalous heat - kernel bounds as well as the local clt and the principal dirichlet eigenvector are all sensitive to microscopic trapping structures .
this paper is organized as follows : in section [ subsec : model ] we define the model and our main objects .
we present our main results in section [ sec : results ] . in section [ subsec : literature ]
we compare our results to former results .
section [ subsec : heatkernel ] contains a heuristic explanation why the critical exponent @xmath4 decides between spectral homogenization and localization . in the same section we comment on how a subdiffusive upper bound for the principal dirichlet eigenvalue contradicts diffusive heat - kernel upper bounds .
we survey the proofs concerning the eigenvalues in sections [ subsec : proofupper ] and [ subsec : proofseigval ] where we rely on technical results from the subsequent sections .
section [ sec : bc ] contains borel - cantelli arguments , which extend results from @xcite , @xcite and @xcite . in section [ sec : perc ] we adapt some standard results on percolation theory from @xcite , @xcite and @xcite to our needs .
section [ sec : path ] contains a path argument similar to the one in @xcite . finally , we prove the localization of the principal eigenvector in section [ sec : proofthmloc ] .
we consider the lattice with vertex set @xmath17 ( @xmath1 ) and edge set @xmath18 . if two sites @xmath19 are neighbors according to @xmath20 , we also write @xmath21 . to each edge
@xmath22 we assign a positive random variable @xmath23 . in analogy to a @xmath24-dimensional resistor network , we call the random weights @xmath23 _ conductances_. we take @xmath25 as the underlying measurable space and assume that an environment @xmath26 is a family of i.i.d . positive random variables with law @xmath27 .
we denote the expectation w.r.t . to @xmath27 by @xmath10 .
if @xmath28 is the edge between the sites @xmath19 , we will also write @xmath29 or @xmath30 instead of @xmath23 .
note that by definition of the edge set @xmath20 , the edges are undirected , whence @xmath31 .
if we want to refer to an arbitrary copy of the conductances in general , we simply write @xmath2 , i.e. , for a set @xmath32 , the expression @xmath33 $ ] equals @xmath34 $ ] for an arbitrary edge @xmath28 .
we call @xmath35\colon u\mapsto { \mathbb{p}}[w\leq u]\end{aligned}\ ] ] the distribution function of the conductances . given a realization @xmath36 of the environment , we consider the markov chain on @xmath0 with transitions rates given by the conductances @xmath29 .
its generator @xmath6 is defined as in .
this markov chain is known as the variable - speed random walk among random conductances . during the last decades
both physicists and mathematicians have analyzed the random conductance model extensively and many questions regarding central limit theorems and heat - kernel behavior have been answered ( for reviews see @xcite and @xcite , respectively ) .
the model became popular for the description of materials where the transition rates between different states are independent of the states energy levels .
this is the case e.g. for the spectral transport of optical excitations among impurity ions @xcite , or charge transport in the one - dimensional ionic conductor hollandite @xcite .
our goal is to study the behavior of the principal eigenvalue @xmath37 and eigenvector @xmath38 of the sign - inverted generator @xmath39 in the ball @xmath40^d \cap { \mathbb{z}}^d\end{aligned}\ ] ] with zero dirichlet conditions at the boundary . for
a real - valued function @xmath41 let us define the dirichlet energy @xmath42 w.r.t . to the operator @xmath39 by @xmath43 where @xmath44 denotes the standard scalar product . then , according to the courant - fischer theorem , the principal dirichlet eigenvalue is given by the variational formula @xmath45 the function @xmath46 that minimizes the rhs of is the principal dirichlet eigenfunction @xmath38 , whence @xmath47 .
[ rem : pf ] for a given box @xmath48 the operator @xmath6 together with the zero dirichlet boundary conditions can be written as a @xmath49-matrix with non - negative entries everywhere except on the diagonal . since the matrix is finite - dimensional
, we can add a multiple of the identity to obtain a non - negative primitive matrix without changing the matrix spectrum . by the perron - frobenius theorem ( see e.g. ( * ? ? ?
* ch . 1 ) )
it follows that its principal eigenvalue is unique and we can assume without loss of generality that its principal eigenvector is positive . in this paper
we are especially interested in the behavior of the principal dirichlet eigenvalue and eigenfunction for dimensions @xmath1 and for conductances with a very heavy tail near zero .
more precisely , we consider those cases where the conductances are distributed such that a local central limit theorem is not valid ( cf .
* remark 1.10 ) ) . under different circumstances ,
the principal dirichlet eigenvalue and eigenfunctions were studied before : boivin and depauw @xcite proved that the top of the spectrum homogenizes .
recent results from neukamm , schffner , and schlmerkemper @xcite and flegel , heida , and slowik @xcite imply that the uniform ellipticity condition can be weakened to suitable moment conditions .
the one - dimensional case was thoroughly covered by faggionato @xcite . in section [ subsec : literature ]
we comment on this background and how our results relate to this previous work .
first we give asymptotic lower and upper bounds for the principal dirichlet eigenvalue @xmath37 .
how can we determine whether a function @xmath50 that decreases monotonically to zero , is such an asymptotic lower or upper bound ? to settle this , we will see that it is crucial to determine whether the box @xmath48 contains a site such that all the @xmath51 incident conductances are less than or equal to @xmath52 .
we call such a site a @xmath52-trap .
a function @xmath53 that carries the information about how many @xmath52-traps we can expect in the box @xmath48 , is defined by @xmath54^{2d}\ , .
\label{equ : deflambda}\end{aligned}\ ] ] note that the factor @xmath55 scales like the number of sites in the box @xmath48 and the factor @xmath56^{2d}$ ] relates to the probability that for a given site all the @xmath51 incident links carry a conductance less than or equal to @xmath52 .
we will see in lemma [ lem : bc : ii ] that if @xmath57 diverges fast enough , then @xmath27-a.s .
for @xmath58 large enough the box @xmath48 contains at least one @xmath52-trap . on the other hand ,
if @xmath57 decreases fast enough to zero , then @xmath48 does @xmath27-a.s .
not contain a @xmath52-trap for @xmath58 large enough , see lemma [ lem : bc : i ] . in our results
we often require recurring conditions on the function @xmath59 . [
ass : g ] let @xmath59 .
[ ass : rvless ] the function @xmath60 varies regularly at infinity with index less than @xmath61 .
[ ass : mon ] the function @xmath62 is monotone and has a finite limit as @xmath63 tends to infinity .
the function @xmath62 converges monotonically to zero as @xmath63 tends to infinity .
our first theorem gives a sufficient and a necessary condition for when the function @xmath60 is an asymptotic upper bound for the principal dirichlet eigenvalue @xmath37 .
note that , given one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mondec ] is true , then the sufficient and necessary conditions coincide up to the case where @xmath57 scales exactly like @xmath64 .
we summarize all the conditions of the following two theorems in a graphical overview ( see fig . [
fig : tails ] ) .
[ thm : eigvalsupperbound ]
let @xmath65 be a function that converges monotonically to zero and let @xmath57 be as in .
then the following statements are true : 1 .
[ thm : eigvalsupperboundsufficient ] if there exists @xmath66 such that for all @xmath58 large enough @xmath67 then there exists a constant @xmath68 such that @xmath27-a.s .
for @xmath58 large enough @xmath69 .
[ thm : eigvalsupperboundnecessary ] on the other hand , if @xmath70 and one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mondec ] is true , then @xmath27-a.s .
@xmath71 .
we prove part ( i ) of the theorem in section [ subsec : proofupper ] and part ( ii ) in section [ subsec : proofseigval ] .
note that in ( ii ) the assumptions [ ass : g ] [ ass : rvless ] and [ ass : mondec ] correspond to the fact that we can only deduce that the limit superior diverges if we assume that @xmath60 is in @xmath72 .
this is because in the diffusive regime @xmath37 scales like @xmath73 . in the case where the distribution function @xmath74 scales like @xmath75 with @xmath76 , theorem [ thm : eigvalsupperboundsufficient ] implies the following corollary .
[ cor : upperboundpolylaw ] let @xmath77 . if @xmath78 varies regularly at zero with index @xmath79 , then @xmath27-a.s .
for @xmath58 large enough the function @xmath80 is an asymptotic upper bound for @xmath37 .
note that if @xmath78 varies regularly at zero with index @xmath79 , then @xmath81 < \infty \}$ ] , as defined in the introduction .
the second theorem gives conditions for when the function @xmath60 is an asymptotic lower bound of the principal dirichlet eigenvalue @xmath37 .
note that , given one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mon ] is true and @xmath57 is bounded , then the condition in is sharp .
we further comment on these conditions in section [ subsec : proofseigval ] . as with the conditions of theorem [ thm : eigvalsupperbound ] , we summarize them in the graphical overview fig .
[ fig : tails ] .
[ thm : eigvalslowerbound ] let @xmath65 be a decreasing function that fulfills one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mon ] .
let @xmath57 be as in .
then the following statements are true : if @xmath82 and @xmath57 is bounded from above , then there exists a constant @xmath83 such that @xmath27-a.s . for @xmath58 large enough @xmath84 .
if , on the other hand , condition does not hold , then @xmath27-a.s .
@xmath85 .
we prove the first part of this theorem in section [ subsec : proofseigval ] .
the second part , i.e. , where condition does not hold , is covered in section [ subsec : proofupper ] .
similarly as for theorem [ thm : eigvalsupperbound ] , we obtain the following corollary .
[ cor : lowerboundpolylaw ] let @xmath77 .
if @xmath78 varies regularly at zero with index @xmath86 $ ] , then @xmath27-a.s .
for @xmath58 large enough the function @xmath87 is an asymptotic lower bound for @xmath37 .
furthermore , if @xmath78 varies regularly at zero with index @xmath5 , then there exists @xmath83 such that @xmath88 is an asymptotic lower bound for @xmath37 .
then @xmath89 for all @xmath90 , which is a poincar inequality for functions with bounded support .
when we set @xmath91 , then theorems [ thm : eigvalsupperboundnecessary ] and [ thm : eigvalslowerbound ] directly imply the following corollary .
[ cor : liminfsup ] assume that there exists @xmath92 such that @xmath93 is invertible and that the function @xmath94 converges monotonically to zero .
then @xmath95 we comment on this behavior in remark [ rem : liminfsup ] in section [ subsec : proofupper ] .
note that in the special case where there exists @xmath76 such that the law @xmath27 of the conductances fulfills @xmath96 = a^\gamma$ ] for @xmath97 $ ] , corollary [ cor : liminfsup ] implies that @xmath98 finally , we show that if we define the local speed measure @xmath99 by @xmath100 then @xmath27-a.s .
as @xmath58 tends to infinity , the first dirichlet eigenvector @xmath38 localizes in the sequence of sites @xmath101 that minimize @xmath102 over @xmath48 .
[ thm : loc ] let @xmath78 vary regularly at zero with index @xmath103 .
further , assume that there exists @xmath92 such that @xmath93 is invertible .
in the case where @xmath104 , assume additionally that there exists @xmath105 such that the product @xmath106 converges monotonically to zero as @xmath58 grows to infinity . for @xmath107 let @xmath108 be the site that minimizes @xmath102 over @xmath48 .
then @xmath27-a.s .
the mass of the principal dirichlet eigenvector @xmath38 with zero dirichlet conditions outside the box @xmath48 increasingly concentrates in the site @xmath108 , i.e. , @xmath109 more precisely , @xmath27-a.s . for @xmath58 large enough @xmath110 where for @xmath76 the value of @xmath105
is chosen such that @xmath111 . as a consequence @xmath37 @xmath27-a.s
. behaves like @xmath112 for large @xmath58 , i.e. , @xmath113 = 1\ , .
\label{equ : principaleigvalminpi } \end{aligned}\ ] ] we prove this theorem in section [ sec : proofthmloc ] .
if @xmath78 varies regularly at zero with index @xmath103 , then there exists @xmath114 such that the expectation @xmath115 $ ] diverges .
note that we can not expect that a result like theorem [ thm : loc ] holds in dimension one .
this is because in dimension one , the probabilistic cost to generate a hardly reachable area is independent of the area s diameter . if the conductances are bounded from above , we conjecture that , qualitatively , the above results should also hold for the constant - speed random conductance model , i.e. , where the laplacian is given by @xmath116 in this case , the critical exponent @xmath117 ( cf .
* theorem 1.8 ( 1 ) ) ) .
further , the typical trapping structures are not single sites but pairs of sites ( cf .
* figure 1 ) ) . in a similar way
as we adapt the proof techniques of @xcite for the variable - speed case , this should be possible for the constant - speed model
. however , the proofs become much more technical . and [ thm : eigvalslowerbound ] for a fixed distribution function @xmath78 .
the figure shows the space of functions @xmath118 that decrease to zero .
the space is depicted such that if @xmath119 , then @xmath46 appears left of @xmath60 . for simplicity
we assume that @xmath60 fulfills one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mondec ] . if @xmath120 , then @xmath121 diverges .
if @xmath60 even decays slowly enough such that condition is fulfilled , then there exists @xmath68 such that @xmath27-a.s . for @xmath58 large enough @xmath122 . on the other hand ,
if @xmath123 , then @xmath121 converges to zero .
if @xmath60 even decays fast enough such that is fulfilled , then there exists @xmath83 such that @xmath27-a.s . for @xmath58 large enough @xmath124 .
the figure also shows that around @xmath125 there is an interval where @xmath60 is definitely neither an a.s .
asymptotic upper nor an a.s .
asymptotic lower bound . ]
our investigation on the spectral behavior of the random conductance generator supplements the results of former research .
boivin and depauw @xcite proved spectral homogenization for stationary and ergodic conductances that fulfill the uniform ellipticity condition , i.e. , where there exist positive and finite constants @xmath126 that uniformly bound the conductances from above and below .
as a special case they showed that for uniformly elliptic i.i.d .
conductances there exists a constant @xmath83 such that if @xmath127 are the first @xmath128 dirichlet eigenvalues of @xmath39 , then for almost every realization of the conductance landscape @xmath129 where @xmath130 is the @xmath128th eigenvalue of the operator @xmath131 in @xmath132 with zero dirichlet conditions .
additionally , the principal dirichlet eigenfunction of @xmath39 converges , properly rescaled , to the principal dirichlet eigenfunction of the operator @xmath131 in @xmath132 with zero dirichlet conditions ( * ? ? ?
* theorem 1 , corollary 1 ) . in the special case of dimension one ,
the uniform ellipticity condition was already weakened by faggionato @xcite : she showed that for @xmath133 a finite inverse moment of @xmath2 is sufficient for spectral homogenization ( * ? ? ?
* proposition 2.6 ) .
further , if the inverse conductances @xmath134 are i.i.d . and in the domain of attraction of an @xmath135-stable law with @xmath136
, then faggionato showed that the vector of the first @xmath128 dirichlet eigenvalues rescaled by @xmath137 times a slowly varying function converges in distribution to the vector of the first @xmath128 dirichlet eigenvalues of a random generalized differential operator ( * ? ? ?
* theorem 2.5 ) .
recent results from two teams of authors imply that the uniform ellipticity condition can also be weakened in higher dimensions : neukamm , schffner , and schlmerkemper ( * ? ? ?
* corollary 3.4 , proposition 3.18 ) proved amongst other results that for @xmath5 the dirichlet energy of @xmath39 @xmath138-converges to a deterministic , homogeneous integral .
this together with their compactness result ( * ? ? ?
* lemma 3.9 ) and ( * ? ? ?
* theorem 13.5 ) implies that conditions i iv of ( * ? ? ?
* chapter 11 ) are fulfilled and spectral convergence follows . on the other hand , flegel , heida , and slowik @xcite
use the method of stochastic two - scale convergence by zhikov and pyatniskii @xcite to show that the poisson equation homogenizes .
their approach is similar to the one of faggionato @xcite who already employed two - scale convergence in order to show homogenization for a laplacian with shifted spectrum and bounded conductances . from the homogenization of the poisson equation , the spectral
homogenization follows again by ( * ? ? ?
* chapter 11 ) .
furthermore , flegel , heida , and slowik identify the corresponding limit operator .
the basis for both @xcite and @xcite are poincar and sobolev inequalities that were already used by andres , deuschel , and slowik @xcite to prove a quenched local clt under suitable moment conditions .
let us consider the variational formula .
the equation implies that for any real - valued test function @xmath139 with @xmath140 and @xmath141 we can estimate @xmath142 suppose that @xmath108 is a random site that minimizes @xmath102 ( see ) in @xmath48 .
now we choose the function @xmath46 such that its whole mass is concentrated in the site @xmath143 , i.e. , @xmath144 .
when we insert this into the variational formula , then we obtain that @xmath145 it remains to find conditions under which the above rhs can be bounded from above by a decreasing function @xmath52 . as we have already mentioned before ,
a quantity which carries this information , is the function @xmath57 defined in , as we see in the two following proofs .
condition together with lemma [ lem : bc : ii ] implies that @xmath27-a.s .
for @xmath58 large enough there exists a site @xmath146 such that @xmath147 .
choose the test function @xmath148 and insert it into the variational formula .
the claim follows .
if condition fails , then @xmath149 let @xmath150 be the set of edges incident to the origin and note that @xmath151 . a substitution of variables and lemma [ lem : bc : i ] imply that for any @xmath83 the following event occurs @xmath27-a.s .
infinitely often as @xmath152 : there exists a site @xmath153 such that all edges in @xmath154 have conductance smaller than or equal to @xmath155 . here , @xmath156
( @xmath157 ) denotes the spatial shift operator .
every time this event occurs , we choose the test function @xmath148 ( as in the proof of theorem [ thm : eigvalsupperboundsufficient ] ) , insert it into the variational formula and immediately obtain that @xmath27-a.s .
@xmath158 we now show that this implies the claim .
suppose that @xmath159 .
we have assumed that one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mon ] is true . in any case
it follows that there exists @xmath160 independent of @xmath161 such that eventually @xmath162 .
it follows that @xmath27-a.s .
this holds for any @xmath159 , implying that @xmath27-a.s .
@xmath164 .
[ rem : liminfsup ] now we can intuitively understand the result of corollary [ cor : liminfsup ] : for the choice @xmath91 , the function @xmath57 is constant one . therefore for every @xmath83 @xmath27-a.s .
there exists an infinite subsequence @xmath165 where the box @xmath166 contains a @xmath167-trap .
however , as we will see in section [ subsec : proofseigval ] , @xmath27-a.s . there also exists an infinite subsequence @xmath168 where the box @xmath169 does not contain a sufficiently good trap .
it follows that the asymptotics of @xmath37 fluctuate around the asymptotics of @xmath170 .
our first aim in this section is to give a heuristic argument why the @xmath171-moment of the inverse conductance @xmath134 decides between spectral homogenization and localization .
our second goal is to explain why the subdiffusive scaling of the principal dirichlet eigenvalue contradicts the validity of a local central limit theorem .
let us first consider a realization of the environment in which all conductances are equal to one .
in this situation the operator @xmath6 generates a simple random walk on @xmath0 .
it is known that in this case the principal dirichlet eigenvalue @xmath37 of @xmath39 in @xmath48 with zero dirichlet conditions scales like @xmath73 ( see e.g. ( * ? ? ?
* eq . ( 7 ) ) ) .
we call this diffusive scaling .
in contrast , if @xmath172 , we say that @xmath37 scales subdiffusively .
we will now investigate under what circumstances we can expect subdiffusive behavior .
let us recall , i.e. , that @xmath173 . for simplicity
we assume that there exists @xmath76 such that the law @xmath27 of the conductances fulfills @xmath174 = a^\gamma \qquad\text{for } a\in[0,1]\ , , \label{equ : polylaw}\end{aligned}\ ] ] and we choose @xmath175 with @xmath176 .
note that @xmath81 < \infty \}$ ] as defined in the introduction .
we recall that a @xmath52-trap is a site where all the @xmath51 incident conductances are less than or equal to @xmath52 .
we observe that @xmath177 which diverges if @xmath77 and converges to zero if @xmath178 .
correspondingly , the borel - cantelli arguments of lemmas [ lem : bc : i ] and [ lem : bc : ii ] imply that if we fix an environment @xmath179 , then @xmath27-a.s .
for @xmath58 large enough there exists a @xmath52-trap in @xmath48 if @xmath77 . on the other hand , there does not exist a @xmath52-trap in @xmath48 if @xmath178 .
this was already pointed out in ( * ? ? ?
* remark 1.10 ) .
now we simply note that if and only if @xmath15 , then there exists @xmath77 such that @xmath180 is in @xmath72 .
hence the principal dirichlet eigenvalue scales subdiffusively if @xmath15 .
it is remarkable that this condition is also sufficient to prove a complete asymptotic localization of the principal dirichlet eigenvector , see theorem [ thm : loc ] . on the other hand ,
if @xmath181 , then these considerations do not provide us with a sub - diffusive upper bound for @xmath37 . to the contrary : for @xmath5 theorem [ thm : eigvalslowerbound ] implies that there exists a constant @xmath83 such that @xmath27-a.s .
for @xmath58 large enough @xmath182 .
note that this is equivalent to a poincar inequality for functions with bounded support , which states that there exists @xmath83 such that @xmath27-a.s .
for @xmath58 large enough @xmath183 which for i.i.d .
conductances with finite expectation of @xmath184 is a consequence of ( * ? ? ?
* proposition 2.4 ) ( with @xmath185 , @xmath186 a step function and @xmath187 replaced a @xmath188 which for each neighbor sums over the optimal detour from the @xmath51 independent paths in figure 2 of @xcite ) .
furthermore , @xmath182 is also a necessary consequence of a diffusive scaling of quenched heat - kernel upper bounds and thus the validity of a quenched local central limit theorem .
the local clt was established in 2015 by the two teams of authors andres , deuschel , slowik @xcite and boukhadra , kumagai , mathieu @xcite who require that for i.i.d .
conductances there exists @xmath66 such that @xmath189 $ ] and @xmath190 $ ] are finite .
let us briefly comment on this .
first we define the random walk among random conductances .
this is the markov chain @xmath191 that is generated by the operator @xmath6 .
its behavior is as follows : when the walker is at a site @xmath192 it waits for an exponential time with expectation @xmath193 ( see ) and then jumps to one of the neighboring sites .
this is why we call @xmath194 the local speed of the random walk at the site @xmath195 . to which neighbor the random walker jumps is random with probabilities proportional to the corresponding conductances : if @xmath196 is a specific neighbor of @xmath195 , the random walker jumps to @xmath196 with probability @xmath197 .
we call @xmath198 the probability w.r.t . to the random walk where the superscript @xmath199 refers to a fixed environment ( quenched probability ) and
the subscript @xmath195 refers to the starting point of the random walker : @xmath200 = 1 $ ] .
the corresponding expectation is called @xmath201 .
let @xmath202 be the escape time from a set @xmath203 , i.e. , @xmath204 .
there exists a natural relation between the principal dirichlet eigenvalue of the operator @xmath205 and the expected escape time @xmath206 $ ] from the box @xmath48 , see ( * ? ? ?
* section 8.4.1 ) : @xmath207 { \aftergroup\egroup\originalright})^{-1}\ , .\end{aligned}\ ] ] thus , an upper bound for the principal dirichlet eigenvalue @xmath37 implies a lower bound on the maximal expected escape time from the box @xmath48 .
now lemma 2.1(i ) of @xcite implies that if the heat kernel @xmath208\qquad ( x , y\in{\mathbb{z}}^d\,,\ ; t\geq 0)\ , , \end{aligned}\ ] ] has a diffusive on - diagonal upper bound , i.e. there exists @xmath209 and a random @xmath210 such that @xmath211 then @xmath212\sim n^{-2}$ ] .
diffusive heat - kernel upper bounds are a necessary condition for the validity of a local clt .
but if we assume that the principal dirichlet eigenvalue scales subdiffusively , i.e. , @xmath172 , then eventually @xmath212\gg n^{-2}$ ] and therefore a subdiffusively scaling principal dirichlet eigenvalue contradicts the validity of a local clt .
we can explain the exploding escape times by showing that a large box contains some sites where the expected time to even leave the initial position is anomalously long .
although this effect is related to the one responsible for the anomalous heat - kernel decay observed in @xcite , it is still a different one . in @xcite ,
the dominating effect is that a random walk finds a trap elsewhere and then returns to its initial position .
this behavior , however , has a more complex dependence on the laplacian s eigenvalues . for the lower bound of the principal dirichlet eigenvalue we have to put in significantly more work than for the upper bound .
the key idea , however , is linked to the considerations for the shape theorem of first - passage percolation , see e.g. cox and durrett @xcite for the sample case @xmath213 .
the philosophy is that we have to show that each site in the box @xmath48 is sufficiently well reachable by conductances that are significantly greater than the lower bound candidate @xmath60 .
note that this is similar to the idea of lemma 4.6 in @xcite where the authors proved this for a polynomial tail of the conductances with parameter @xmath16 and the candidate @xmath214 with @xmath215 .
it turns out that a crucial element of the proof is to give a condition that implies that @xmath27-a.s . for @xmath58 large enough all sites in the box @xmath48
have at least one link with conductance greater than @xmath52 , similar to @xcite and @xcite .
in general , if @xmath60 is monotonically decreasing and @xmath216 are @xmath51 independent copies of the conductance @xmath2 , and @xmath217 = d\int_0^\infty u^{-1 } \lambda_g ( u){\mathrm{d}}u < \infty\ , , \label{equ : homcondheur}\end{aligned}\ ] ] then @xmath27-a.s . for @xmath58 large enough , all sites in the box @xmath48 have at least one link with conductance greater than @xmath52 .
this together with a path argument , which we adapt from @xcite gives the @xmath27-a.s .
lower bound for the principal dirichlet eigenvalue @xmath37 ( given that @xmath52 is not asymptotically larger than @xmath73 ) .
on the other hand , if condition is violated , then the same arguments as in cox and durrett @xcite yield that @xmath27-a.s .
as @xmath58 tends to infinity , the box @xmath48 contains a @xmath52-trap infinitely often .
we have already dealt with this case at the end of section [ subsec : proofupper ] . in what follows
we give a survey on the proofs of theorem [ thm : eigvalsupperboundnecessary ] as well as theorem [ thm : eigvalslowerbound ] if condition ( or equivalently ) holds .
the arguments described above are made rigorous in several auxiliary lemmas , which we present in the subsequent sections . by virtue of lemma [ lem :
bc : i ] with @xmath218 it follows that @xmath27-a.s .
there exists @xmath219 such that for all @xmath220 all sites @xmath221 have an incident edge with conductance greater than @xmath52 .
further , let @xmath222 be such that @xmath223 .
since we assumed that @xmath57 is bounded from above , it follows by virtue of corollary [ cor : bc : i3 ] ( with @xmath224 and @xmath225 ) that there exists @xmath66 such that @xmath27-a.s .
there exists @xmath226 such that for all @xmath227 and for all @xmath228 the box @xmath229 contains at most @xmath230 links with conductance less than or equal to @xmath231 .
set @xmath232 .
the claim follows by virtue of proposition [ prop : path ] .
condition together with the borel - cantelli argument of lemma [ lem : bc : iii ] implies that for any @xmath233 , @xmath27-a.s . as the box size @xmath58 grows to infinity , there exists a random subsequence @xmath234 along which each site @xmath235 has at least one incident link @xmath28 such that @xmath236 .
further , by corollary [ cor : bc : i4 ] we know the following : for a fixed @xmath222 there exists @xmath66 such that @xmath27-a.s . for @xmath58 large enough , there are at most @xmath51 edges in any subbox @xmath237 with conductance smaller than or equal to @xmath231 .
it follows that we can apply proposition [ prop : path ] with @xmath238 instead of @xmath60 and obtain that there exists @xmath239 ( independent of @xmath240 ) such that @xmath27-a.s . along the subsequence @xmath241 and for @xmath128 large enough the following holds : for any @xmath242 with @xmath243 it is @xmath244 since this holds for any @xmath245 , this implies the claim .
in this section we always assume that the dimension @xmath1 and that the conductances are i.i.d . with law @xmath27 .
we further let @xmath246 be a function that decreases monotonically to zero .
moreover , we use the following abbreviations : for @xmath247 and an edge set @xmath248 we define the event @xmath249 for a set @xmath250 we define @xmath251 to be the set of edges that connect a site in @xmath252 with a neighbor in positive axes direction ( i.e. , right , above , in front , etc . ) , i.e. , @xmath253 where @xmath254 is the set of unit base vectors of @xmath17 . for @xmath248
we write @xmath255 for the translation of @xmath256 by @xmath257 .
[ lem : bc : i ] if @xmath258 and @xmath259 is an edge set with @xmath260 , then [ _ n _ zb_n+b j_g(n ) ( _ z ) ] = 1 , & if @xmath261^m { \mathrm{d}}u < \infty\ , , $ ] [ equ : bc : ia ] + 0 , & otherwise .
[ equ : bc : ib ] i.e. , if and only if the integral @xmath261^m { \mathrm{d}}u$ ] is finite , then @xmath27-a.s . for @xmath58
large enough for all sites @xmath228 the edge set @xmath262 contains a conductance greater than @xmath52 .
otherwise the counter event occurs infinitely often .
the result of lemma [ lem : bc : i ] as well as the proof are generalizations of the considerations of cox and durrett @xcite and kesten @xcite ( there , @xmath224 and @xmath263 ) . for the sake of completeness , we included the proofs here . for : we first show that @xmath264 = { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ \limsup_{|z|_\infty\rightarrow\infty } { \mathopen{}\mathclose\bgroup\originalleft } ( j_{g(|z|_\infty - b ) } ( \tau_z \circ\mathfrak{a } ) { \aftergroup\egroup\originalright})^{\rm c}{\aftergroup\egroup\originalright}]\ , .
\label{equ : proofbc : i1}\end{aligned}\ ] ] we achieve this by applying the first borel - cantelli lemma , i.e. , we have to estimate @xmath265 & = \sum_{k = b+1}^\infty \sum_{|z|_\infty = k } { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq g(|z|_\infty - b ) { \aftergroup\egroup\originalright}]^m\\ & \leq 2d \sum_{k = b+1}^\infty ( 2k+1)^{d-1 } { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g(k - b ) { \aftergroup\egroup\originalright}]^m\ , .\end{aligned}\ ] ] since @xmath266 is monotonically decreasing , @xmath267 $ ] is monotonically decreasing as well .
further , there exists an index @xmath268 such that @xmath269 for all @xmath270 .
it follows that there exists @xmath68 such that @xmath271 & \leq c + 4^d d \sum_{k = b+1}^\infty ( 2^{-1}(k+1))^{d-1 } { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g{\mathopen{}\mathclose\bgroup\originalleft } ( 2^{-1}(k+1 ) { \aftergroup\egroup\originalright } ) { \aftergroup\egroup\originalright}]^m\ , .\end{aligned}\ ] ] this implies that there exists @xmath272 such that the lhs is bounded from above by @xmath273^m { \mathrm{d}}u$ ] , which is finite by assumption .
the claim follows from the first borel - cantelli lemma .
to arrive at the claim of the lemma , we observe that implies that there @xmath27-a.s .
exists @xmath274 such that for all @xmath275 the set @xmath262 contains at least one conductance with @xmath276 .
if @xmath277 and @xmath278 , i.e. , @xmath279 $ ] , this means that @xmath262 contains at least one conductance with @xmath280 ( recall that @xmath60 is monotonically decreasing ) .
since @xmath281 is finite and @xmath60 decreases monotonically to zero , it also follows that there exists a finite @xmath282 such that for all edges @xmath283 we have @xmath284 .
thus , @xmath285 is true @xmath27-a.s . for @xmath58 large enough . for : let @xmath286^m { \mathrm{d}}u = \infty$ ] .
we want to show that this implies @xmath287 = 0\ , .
\label{lem : proofbc : i2}\end{aligned}\ ] ] let us define the set @xmath288 .
it suffices to prove the claim for the intersection over @xmath289 , which in turn follows by the second borel - cantelli lemma if @xmath290 = \infty\ , , \end{aligned}\ ] ] since the events @xmath291 are independent . to prove that the above sum diverges
, we observe that there exists a constant @xmath239 such that @xmath292 & \geq 2d ( 2b+1)^d \sum_{k=1}^\infty ( 2k-1)^{d-1 } { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g((2b+1)k+b ) { \aftergroup\egroup\originalright}]^m\\ & \geq c \int_0^\infty u^{d-1}{\mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq g(u ) { \aftergroup\egroup\originalright}]^m { \mathrm{d}}u\ , .\end{aligned}\ ] ] by the assumption that @xmath286^m { \mathrm{d}}u = \infty$ ] , the sum diverges .
[ cor : bc : i2 ] let @xmath293 and @xmath294 .
then the following equivalence holds : @xmath295^m { \mathrm{d}}u < \infty ~\leftrightarrow~ { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ \liminf_{n\rightarrow\infty } \bigcap_{\mathfrak{a}\subseteq { \mathfrak{e}}(b_b),\atop |\mathfrak{a}|\geq m } \bigcap_{z\in b_{n+b } } j_{g(n ) } ( \tau_z \circ \mathfrak{a}){\aftergroup\egroup\originalright } ] = 1\ , .
\label{equ : hom : m } \end{aligned}\ ] ] for `` @xmath296 '' , we apply lemma [ lem : bc : i ] for an arbitrary @xmath297 with @xmath298 . for `` @xmath299 '' , note that since @xmath300 is finite , the intersection over the edge sets @xmath256 on the rhs of runs over finitely many events . by virtue of lemma [ lem :
bc : i ] the claim holds for each of these events and therefore also for the finite intersection .
[ cor : bc : i3 ] let @xmath293 , @xmath301 with @xmath302 .
if @xmath303^m$ ] is bounded from above , then @xmath304 = 1\qquad\text{for all } { \epsilon}\in [ 0,k(m+k)^{-1})\ , .
\label{equ : hom : m+1}\end{aligned}\ ] ] we show that the integral @xmath305^{m+k}{\mathrm{d}}v$ ] is finite and then we apply corollary [ cor : bc : i2 ] .
the change of variable @xmath306 yields @xmath307^{m+k}{\mathrm{d}}v & = ( 1-{\epsilon})^{-1}\int_0^\infty u^{d(1-{\epsilon})^{-1}-1 } { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g{\mathopen{}\mathclose\bgroup\originalleft } ( u{\aftergroup\egroup\originalright } ) { \aftergroup\egroup\originalright}]^{m+k } \,{\mathrm{d}}u\ , .\end{aligned}\ ] ] now we consider that @xmath308^{m+k}= u^{d{\mathopen{}\mathclose\bgroup\originalleft}((1-{\epsilon})^{-1 } -1 -\frac{k}{m}{\aftergroup\egroup\originalright})-1}{\mathopen{}\mathclose\bgroup\originalleft } ( u^d { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g(u ) { \aftergroup\egroup\originalright}]^m { \aftergroup\egroup\originalright})^{1+\frac{k}{m}}\ , .\end{aligned}\ ] ] since both @xmath303^m$ ] and @xmath309^m$ ] are bounded from above , we obtain that @xmath310^{m+k } \,{\mathrm{d}}u \leq \int_0^{u_1 } u^{d(1-{\epsilon})^{-1 } -1 } \,{\mathrm{d}}u + c \int_{u_1}^\infty u^{d{\mathopen{}\mathclose\bgroup\originalleft}((1-{\epsilon})^{-1 } -1 -\frac{k}{m}{\aftergroup\egroup\originalright})-1}\ , { \mathrm{d}}u < \infty\end{aligned}\ ] ] for any @xmath311 and a suitable @xmath68 . since @xmath312 and @xmath1 , the first integral on the rhs is finite .
further , since @xmath313 , the second integral on the rhs is finite as well . for the next three results , we define @xmath57 as in .
[ cor : bc : i4 ] let @xmath314 and assume that @xmath315 is bounded from above
. then there exists @xmath316 such that @xmath317 = 1\ , .
\label{equ : bc : i4}\end{aligned}\ ] ] we show that the integral in is finite for @xmath318 and @xmath319 instead of @xmath320 .
the assumption on the function @xmath57 implies that there exists @xmath68 such that @xmath321^{2d+1}}{{\mathopen{}\mathclose\bgroup\originalleft } ( \log\log u^{1-\epsilon}{\aftergroup\egroup\originalright})^{1 + \frac{1}{2d } } } < c\quad \text{for all $ u>0 $ and $ \epsilon\in ( 0,1)$.}\end{aligned}\ ] ] it follows that @xmath322^{2d+1 } \leq c u^{-\frac{1}{2}+\epsilon{\mathopen{}\mathclose\bgroup\originalleft}(d+ \frac{1}{2}{\aftergroup\egroup\originalright } ) } { \mathopen{}\mathclose\bgroup\originalleft } ( \log\log u^{1-\epsilon}{\aftergroup\egroup\originalright})^{1 + \frac{1}{2d}}\quad\text{for all $ u>0$.}\end{aligned}\ ] ] for all @xmath323 , this implies that the integral @xmath324^{2d+1 } \,{\mathrm{d}}u$ ] is finite .
the claim follows by virtue of corollary [ cor : bc : i2 ] .
r.2 for the next lemma we need the following definition : the even lattice is the set @xmath325 accordingly , the odd lattice is @xmath326 , see fig .
[ fig : evenodd ] for a sketch .
now , the next lemma states that if a certain condition is fulfilled , then @xmath27-a.s .
for @xmath58 large enough there exist sites @xmath327 and @xmath328 such that all incident links to these sites have conductance less than or equal to @xmath52 .
this implies that there are at least two sites in @xmath48 whose incident links are all less than or equal to @xmath52 .
[ lem : bc : ii ] let @xmath150 and @xmath329 .
then the following implication is true : if there exists @xmath66 and @xmath330 such that @xmath331 = 0\ , , \end{aligned}\ ] ] i.e. , @xmath27-a.s .
for @xmath58 large enough there exists a site @xmath332 that is completely surrounded by edges with conductances less than or equal to @xmath52 .
we first prove the claim for the subsequence @xmath333 with @xmath334 and with @xmath335 instead of @xmath52 .
then we show how to infer the claim along the whole sequence @xmath107 . for the first part , note that the sets @xmath336 and @xmath337 have cardinality greater than @xmath338 .
further , for any @xmath247 the events @xmath339 are independent , as well as are the events @xmath340 .
let @xmath341 be @xmath51 independent copies of @xmath2 .
then for both @xmath342 we can estimate @xmath343 = { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ \max \ { w_1,\ldots , w_{2d } \ } >
g(2n_j ) { \aftergroup\egroup\originalright}]^{{\mathopen{}\mathclose\bgroup\originalleft}| b_{n_j}\cap a { \aftergroup\egroup\originalright}|}\\ & \qquad\qquad\leq { \mathopen{}\mathclose\bgroup\originalleft } ( 1 - { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g(2n_j ) { \aftergroup\egroup\originalright}]^{2d } { \aftergroup\egroup\originalright})^{2^{d-1 } n_j^d } \leq \exp{\mathopen{}\mathclose\bgroup\originalleft } ( - \frac{1}{2 } { \mathopen{}\mathclose\bgroup\originalleft } ( 2n_j{\aftergroup\egroup\originalright})^d { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w \leq g(2n_j ) { \aftergroup\egroup\originalright}]^{2d } { \aftergroup\egroup\originalright})\ , .\end{aligned}\ ] ] the assumption on @xmath57 implies that the rhs is summable along the sequence @xmath333 .
thus , it follows directly by the borel - cantelli lemma that the statement of this lemma holds along the subsequence @xmath344 and with @xmath345 instead of @xmath346 . to infer the claim of the lemma along the entire sequence , we define @xmath347
note that both @xmath348 are monotonically decreasing in @xmath58 . by the first part of the proof we know that @xmath349 = 1\ , .
\label{equ : liminfsubseq}\end{aligned}\ ] ] now for @xmath350 we choose @xmath351 such that @xmath352 since @xmath353 and @xmath354 are all monotonically decreasing , this implies that @xmath355 whence especially @xmath356 thus , the claim follows by .
[ lem : bc : iii ] let @xmath357 be as in lemma [ lem : bc : ii ] . if the function @xmath358 decreases monotonically to zero and @xmath359 = 1 \quad\forall c>0\ , .
\label{equ : bc : iiilocal } \end{aligned}\ ] ] for @xmath360 , a fixed @xmath83 , and a fixed function @xmath60 let us abbreviate @xmath361 let us briefly outline the idea of the proof : it is sufficient to show that the claim is true along the subsequence @xmath362 .
first we show that @xmath363 = \infty \label{equ : sumhinfty}\end{aligned}\ ] ] which , since @xmath364 , implies that @xmath365 = \infty$ ] .
note that since for @xmath366 with @xmath367 the intersection @xmath368 the events @xmath369 are independent .
thus , we can infer by the second borel - cantelli lemma that @xmath370 = 1\ , .
\label{equ : pbc : limsup}\end{aligned}\ ] ] then we show that @xmath371 = 1\ , .
\label{equ : pbc : liminf}\end{aligned}\ ] ] since by definition @xmath372 together with implies the claim of the lemma .
let us start with the proof of .
we note that for @xmath373 and @xmath374 as defined in the fkg - inequality implies that @xmath375 = { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ h^{n_j}_{a_{\rm e } \cap b_{n_j } } \cap h^{n_j}_{a_{\rm o } \cap b_{n_j}}{\aftergroup\egroup\originalright } ] \geq { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ h^{n_j}_{a_{\rm e } \cap b_{n_j } } { \aftergroup\egroup\originalright}]^2\end{aligned}\ ] ] then we recall that @xmath373 was constructed such that @xmath376 is the intersection of less than @xmath377 i.i.d .
subevents @xmath378 , each with probability @xmath379 = 1 - { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(n_j ) { \aftergroup\egroup\originalright}]^{2d}\ , .\end{aligned}\ ] ] thus for @xmath380 large enough , there exists @xmath68 such that @xmath375 & \geq { \mathopen{}\mathclose\bgroup\originalleft } ( 1 - { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(n_j ) { \aftergroup\egroup\originalright}]^{2d } { \aftergroup\egroup\originalright})^{2(2n_j+1)^d}\nonumber\\ & = { \mathopen{}\mathclose\bgroup\originalleft } ( { \mathopen{}\mathclose\bgroup\originalleft } ( 1 - { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(n_j ) { \aftergroup\egroup\originalright}]^{2d } { \aftergroup\egroup\originalright})^{{\mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(n_j ) { \aftergroup\egroup\originalright}]^{-2d } } { \aftergroup\egroup\originalright})^{2(2n_j+1)^d { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(n_j ) { \aftergroup\egroup\originalright}]^{2d}}\nonumber\\ & \geq \exp{\mathopen{}\mathclose\bgroup\originalleft } ( -c n_j^d { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(n_j ) { \aftergroup\egroup\originalright}]^{2d}{\aftergroup\egroup\originalright } ) = \exp{\mathopen{}\mathclose\bgroup\originalleft } ( -c \lambda_{cg } { \mathopen{}\mathclose\bgroup\originalleft } ( n_j { \aftergroup\egroup\originalright } ) { \aftergroup\egroup\originalright})\ , .
\label{equ : boundph}\end{aligned}\ ] ] now we observe that the assumptions on @xmath60 and @xmath57 imply that the rhs of is not summable for any @xmath83 .
this concludes the argument for .
let us proceed with the proof of .
note that for any @xmath316 we have @xmath381 with @xmath382 .
this is because for @xmath380 large enough and any @xmath316 we have @xmath383 thus , is a consequence of @xmath384 = 1 \quad \forall c>0\ , .\end{aligned}\ ] ] by virtue of lemma [ lem : bc : i ] we can thus verify by showing that for all @xmath83 the integral @xmath385^{2d } { \mathrm{d}}u <
\infty\ , .\end{aligned}\ ] ] to see that the integral is indeed finite , we consider the following : there exists a constant @xmath68 such that @xmath385^{2d } { \mathrm{d}}u & \leq c + \int_2^\infty u^{-1 } { \mathopen{}\mathclose\bgroup\originalleft } ( \frac{u}{h(u)}{\aftergroup\egroup\originalright})^d { \mathopen{}\mathclose\bgroup\originalleft } ( h(u)^d\ , { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ w\leq cg(h(u ) ) { \aftergroup\egroup\originalright}]^{2d } { \aftergroup\egroup\originalright } ) \,{\mathrm{d}}u\\ & = c + \int_2^\infty u^{-1 } { \mathopen{}\mathclose\bgroup\originalleft } ( \log u { \aftergroup\egroup\originalright})^{-d ( 1-\epsilon ) } \lambda_{cg } ( h(u ) ) \,{\mathrm{d}}u\ , .\end{aligned}\ ] ] by the condition on @xmath57 and the monotonicity condition on @xmath60 follows that the above rhs is finite .
in this section we adapt three standard percolation results that we need for the path arguments of the next section in order to establish the lower bound for the principal dirichlet eigenvalue .
we consider the standard bernoulli bond percolation on the graph @xmath386 , i.e. , we assume that the conductances are independent random variables with common law @xmath387 such that an individual conductance is @xmath388 with probability @xmath389 and @xmath390 otherwise . for an introduction to percolation
we refer the reader to @xcite . as in the previous section ,
we call @xmath391 an environment and we denote the law of the environment by @xmath392 .
if the conductance @xmath23 of an edge @xmath28 is equal to 1 , then we call @xmath28 an open edge
. otherwise we call the edge @xmath28 closed . given a realization @xmath199 of the environment , we denote the set of open edges by @xmath393
. consider the random graph @xmath394 .
following the terminology of grimmet @xcite , we call the connected subgraphs of this graph _ open clusters _ and , for @xmath395 , we write @xmath396 for the open cluster that contains the site @xmath195 . note that @xmath397 is a graph .
we define the clusters this way in order to make sense of dirichlet forms defined as in below
. however , when we write @xmath398 , we refer to the number of sites in @xmath396 .
furthermore , when @xmath399 is a cluster and @xmath400 is a site in the vertex set of @xmath399 , then we use the shorthand notation @xmath401 .
similarly , if @xmath28 is in the edge set of @xmath399 , then we write @xmath402 .
we say that a path @xmath403 is open if and only if @xmath404 for all @xmath405 .
given a realization @xmath199 of the environment , we write @xmath406 for the length of the shortest open path between the sites @xmath195 and @xmath400 .
we say that @xmath407 whenever @xmath195 and @xmath400 are not connected through an open path .
let @xmath408 be the critical probability such that @xmath27-a.s .
there exists an infinite open cluster @xmath409 .
this cluster is @xmath27-a.s .
we assume that @xmath410 .
note that @xmath409 contains all sites @xmath195 that are connected to infinity through an open path _ as well as _
all open edges that are incident to a site in @xmath409 .
we further define @xmath411 as the complement of @xmath409 in @xmath0 , i.e. , we regard @xmath411 as a set of sites .
the main object of this section is to collect results from the literature and adapt the details such they exactly fit our needs .
[ lem : propiicbox ] let @xmath412 .
then for @xmath389 sufficiently close to one , there exist constants @xmath413 and @xmath83 such that @xmath414 \leq c { \rm e}^{-c n}\quad \text { for all $ n\geq 1$. } \label{equ : propiicbox } \end{aligned}\ ] ] for the proof of this lemma we refer the reader to ( * ? ? ?
* lemma 4.2 ) .
the second lemma is an implication of lemma [ lem : propiicbox ] above .
[ lem : injectivemap ] for @xmath389 sufficiently close to one and @xmath392-a.s .
for @xmath58 large enough there exists an injective map @xmath415 such that for any site @xmath416 the distance @xmath417
. the proof of this lemma follows the lines of the first paragraph of the proof of ( * ? ? ?
* lemma 4.7 ) but we included the proof here for completeness .
for @xmath157 and @xmath418 , we denote @xmath419 .
choose the percolation parameter @xmath389 such that is fulfilled with @xmath420 .
let @xmath421 and consider the disjoint partition @xmath422 of @xmath0 .
then lemma [ lem : propiicbox ] implies that there exist @xmath423 such that @xmath424 & \leq \sum_{b\in\mathcal{p}_m,\atop b\cap b_n \neq \emptyset } { { \mathbb{p}}}{\mathopen{}\mathclose\bgroup\originalleft } [ |b \cap { { \mathscr{c}}_{\infty}}| \leq \eta |b|{\aftergroup\egroup\originalright}]\\ & \leq c ( 2n+1)^d \exp{\mathopen{}\mathclose\bgroup\originalleft } ( -c { \mathopen{}\mathclose\bgroup\originalleft}(\log n{\aftergroup\egroup\originalright})^{d+1 } { \aftergroup\egroup\originalright})\ , , \end{aligned}\ ] ] which is summable . by the borel - cantelli lemma it follows that @xmath392-a.s .
for @xmath58 large enough we have @xmath425 in any @xmath426 with @xmath427 .
now we construct @xmath428 as follows : for @xmath429 choose @xmath426 ( unique ) such that @xmath430 .
choose @xmath431 in an injective way - this is possible since @xmath432 .
the @xmath433-distance between @xmath195 and @xmath434 is thus smaller than or equal to @xmath435 . for @xmath436 with @xmath437
we define the dirichlet - form @xmath438 : @xmath439 as well as the norm @xmath440 . in the following lemma
we give a lower bound for the principal dirichlet eigenvalue on @xmath441 .
the lemma is similar to theorem 1.3 from @xcite with the difference that @xmath441 is in general not connected and we do not include the condition that @xmath442 . [
lem : spectralgapperc ] let @xmath1 and choose @xmath389 such that lemma [ lem : propiicbox ] holds with @xmath420 .
then there exists a deterministic constant @xmath83 such that @xmath392-a.s .
for @xmath58 large enough and all real - valued functions @xmath139 with @xmath140 we have @xmath443 we prove this lemma at the end of the section .
it is rather standard given the relative isoperimetric inequality in lemma [ lem : isoper ] below ( see e.g. @xcite ) but since the details are slightly different , we include the proof for the convenience of the reader . for the proof we need some additional definitions .
let @xmath107 . if @xmath444^d\cap{\mathbb{z}}^d$ ] , we write @xmath445 for the shortest open path between @xmath195 and @xmath400 that is contained in @xmath48 .
we further say that @xmath446 is an open cluster in @xmath48 if and only if @xmath446 is an open cluster and @xmath447 for all @xmath448 . the largest open cluster in @xmath48 is denoted by @xmath449 .
note that it is _ not _ necessarily @xmath450 .
let @xmath451 be a set of sites .
we define the relative edge boundary of @xmath252 with respect to @xmath449 as the edge set @xmath452 further we need the following auxiliary lemma about properties of @xmath449 and a relative isoperimetric inequality . in this lemma
we collect well - known results from @xcite and @xcite and adapt them to our needs .
[ lem : isoper ] let @xmath1 and choose @xmath389 such that lemma [ lem : propiicbox ] holds with @xmath420 .
then there exists @xmath453 such that @xmath27-a.s .
exists @xmath330 such that for all @xmath454 the following statements are true : i ) @xmath450 , ii ) @xmath455 , iii ) @xmath456 , and iv ) the relative isoperimetric inequality holds : @xmath457 since @xmath458 , item ( i ) follows from lemma 2.19 and remark 2 on page 3049 of barlow @xcite . item ( iii ) follows from lemma [ lem : propiicbox ] together with a borel - cantelli argument .
item ( iv ) follows by virtue of ( * ? ? ?
* proposition 2.11 , proposition 2.12(a ) ) and the borel - cantelli lemma , i.e. , there exists @xmath83 such that @xmath27-a.s .
there exists @xmath459 such that for all @xmath460 the following statement is true : for all connected open subsets @xmath461 with @xmath462 and such that @xmath463 is connected , the isoperimetric inequality @xmath464 holds . by the same arguments as in (
3.1 ) , this implies that also holds for arbitrary @xmath461 with @xmath462 .
finally we show ( ii ) , i.e. , that there exists @xmath465 such that @xmath392-a.s .
there exists @xmath466 such that @xmath455 for all @xmath467 .
the proof is similar to the one in appendix b of @xcite but we present it here for the convenience of the reader .
assume that there exists a site @xmath468 .
we estimate @xmath469 \leq { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ { \mathopen{}\mathclose\bgroup\originalleft}\{\exists x \in b_{\frac{n}{2\rho}}\cap{{\mathscr{c}}_{\infty}}\backslash{{\mathscr{c}}^\vee_n}{\aftergroup\egroup\originalright}\ } \cap { \mathopen{}\mathclose\bgroup\originalleft}\ { { { \mathscr{c}}^\vee_n}\subset { { \mathscr{c}}_{\infty}}{\aftergroup\egroup\originalright}\ } { \aftergroup\egroup\originalright } ] + { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ { { \mathscr{c}}^\vee_n}\not\subset { { \mathscr{c}}_{\infty}}{\aftergroup\egroup\originalright}]\end{aligned}\ ] ] from ( i ) it follows that the second term on the rhs is negligible and we therefore only consider the first term . if @xmath470 , then @xmath27-a.s . for @xmath58 large enough there
exists a site @xmath471 with norm @xmath472 , for a sketch see fig [ fig : percolation ] .
but this implies that the @xmath473-distance between @xmath195 and @xmath400 can be estimated from above and below by @xmath474 further , since @xmath475 but @xmath471 , it follows that the shortest open path between @xmath195 and @xmath400 must leave the box @xmath48 , see e.g. fig [ fig : percolation ] .
thus , the chemical distance between the sites @xmath195 and @xmath400 is bounded from below by the @xmath476-distances of @xmath195 and @xmath400 to the boundary @xmath477 , i.e. @xmath478 by virtue of ( * ? ? ?
* theorem 1.1 ) there exists @xmath479 and @xmath480 such that @xmath481 \leq { \rm e}^{-\beta |x - y|_1}\ , .\end{aligned}\ ] ] let us choose @xmath482 such that @xmath483 .
then it follows that @xmath484\\ & \qquad\qquad\qquad\leq \sum_{x\colon |x|_\infty\leq \frac{n}{2\rho } } \sum_{y\colon |y|_\infty = \frac{n}{\rho } } { \rm e}^{-\beta |x - y|_1 } \leq ( 2n + 1)^{2d } { \rm e}^{-\beta{\mathopen{}\mathclose\bgroup\originalleft}(\frac{n}{2\rho}-1{\aftergroup\egroup\originalright})}\ , , \end{aligned}\ ] ] which is summable .
it follows that @xmath27-a.s .
there exists @xmath466 such that @xmath455 for all @xmath467 .
let @xmath281 be as in lemma [ lem : isoper ] and let @xmath454 .
further let @xmath485 such that @xmath486 .
it is sufficient to show that holds if @xmath46 is the principal dirichlet eigenvector of the generator belonging to the dirichlet form @xmath487 . similarly to remark [ rem : pf ] , the perron - frobenius theorem implies that we can assume w.l.o.g .
that @xmath488 .
we apply hlder s inequality to obtain @xmath489 now we use a standard approach which is known as the co - area formula ( see e.g. @xcite ) : @xmath490 where we can interchange the double sum and the integral since @xmath46 has bounded support .
if we define the set of sites @xmath491 , then we see that @xmath492 we aim to apply the relative isoperimetric inequality .
let us choose @xmath482 as in lemma [ lem : isoper ] . since @xmath486 and @xmath454 , we have @xmath493 for all @xmath494 by lemma [ lem : isoper ] ( ii ) .
this implies @xmath495 as well .
we further have @xmath496 by lemma [ lem : isoper ] ( ii ) and ( iii ) . by the relative isoperimetric inequality
we therefore obtain @xmath497 together with this implies that @xmath498
in this section we give the two propositions [ prop : spectralgapd ] and [ prop : path ] , which transfer the knowledge we obtained by the borel - cantelli arguments in section [ sec : bc ] to lower bounds of dirichlet forms . in order to achieve this , lemma [ lem : pathvsrw ] generalizes and modifies the path argument in ( * ? ? ?
* lemma 4.7 ) .
before we start , we give a definition which is crucial for the remaining part of the paper .
[ def : ewg ] let @xmath499 be an undirected graph and @xmath500 . for @xmath501 , we define the dirichlet energy on @xmath502 : @xmath503 [ rem : envperc ] for @xmath504 let us define @xmath505 ( @xmath506 ) .
let us call an edge @xmath28 open if and only if @xmath507 and let @xmath399 be an open cluster in the environment @xmath508 .
then , with reference to , we obtain that @xmath509 for all real - valued functions @xmath139 . since we apply a similar argument for two slightly different situations ( i.e. , once for the proofs of theorems [ thm : eigvalsupperbound ] and [ thm : eigvalslowerbound ] , see proposition [ prop : path ] , and once for the proof of theorem [ thm : loc ] , see proposition [ prop : spectralgapd ] ) , we kept the conditions of the following lemma as general as necessary .
[ lem : pathvsrw ] let @xmath499 be an undirected graph and let @xmath510 be a subgraph of @xmath502 .
assume that @xmath511 and @xmath512 are such that the following conditions are fulfilled : 1 .
[ item : poincarecond ] there exists a constant @xmath513 such that for all @xmath501 with @xmath514 the following inequality holds : @xmath515 2 .
there exists an injective map @xmath516 such that the following holds : from any @xmath517 there exists a ( self - avoiding ) directed path @xmath518 to @xmath519 in @xmath502 such that 1
. [ item : allbinl ] all @xmath520 fulfill @xmath521 , 2 .
[ item : upperboundl ] @xmath522 .
then for all @xmath501 with @xmath514 the following holds : @xmath523 we generalize the proof of ( * ? ? ?
* lemma 4.7 ) .
let @xmath501 with @xmath514 .
for the following calculation we abbreviate @xmath524 where @xmath525 is the ( directed ) link from site @xmath400 to its neighbor @xmath196 . for @xmath517
we write @xmath526 as a telescopic sum @xmath527 we apply the cauchy - schwarz inequality and expand the terms on the rhs by the conductances : @xmath528 now we sum over all @xmath517 and use the upper bound for @xmath529 according to condition [ item : upperboundl ] : @xmath530 let us look at the last term on the rhs : by definition @xmath531 is injective and its image is in @xmath532 .
this means that @xmath533 due to the limited length of the path @xmath518 the sum over @xmath534 on the rhs in uses each edge at most @xmath535 times , whence @xmath536 completing the sum to all sites @xmath537 and using the comparability between @xmath538 and @xmath539 , we obtain by virtue of condition [ item : poincarecond ] : @xmath540 from the path argument in lemma [ lem : pathvsrw ] we can use our observations from section [ sec : bc ] to obtain lower bounds of the dirichlet forms .
we use similar arguments as in ( * ? ? ?
* lemma 5.1 ) .
we fix @xmath504 such that @xmath541>p_c(d)\ , .
\label{equ : condxi}\end{aligned}\ ] ] moreover , we fix an environment @xmath542 and define a new environment @xmath543 by setting @xmath544 similarly as in remark [ rem : envperc ] .
we denote the unique infinite cluster of the environment @xmath543 by @xmath545 and we use the same shorthand notations as explained at the beginning of section [ sec : perc ] .
further we define @xmath546 as the restriction of @xmath545 to the box @xmath48 and and similarly the holes @xmath547 .
additionally , we define a second percolation environment @xmath548 for @xmath118 by setting @xmath549 thus , links with conductance less than or equal to @xmath52 are considered to be closed and all others keep their original conductance . with this terminology we can now define the following clusters .
[ def : indn ] let @xmath550 be the unique infinite open cluster of @xmath548 .
regarding this cluster , we use the same shorthand notations as introduced at the beginning of section [ sec : perc ] .
furthermore , let @xmath551 be the set of holes . for a fixed function @xmath60 and a fixed @xmath66 we often abbreviate @xmath552 and @xmath553 .
[ def : sparse ] we say a set @xmath554 is * sparse * if the set @xmath555 does not contain any neighboring sites .
[ prop : spectralgapd ] let @xmath556 be a decreasing function .
let @xmath222 be such that @xmath223 .
further , assume that there exists @xmath316 such that along the subsequence @xmath557 for all @xmath228 the box @xmath229 contains at most @xmath230 links with conductance less than or equal to @xmath231 .
define @xmath558 .
then there exists @xmath559 ( independent of @xmath60 ) such that @xmath27-a.s . for @xmath128 large enough , and all real - valued functions @xmath560 with @xmath561
we obtain @xmath562 with @xmath563 as in definition [ def : ewg ] .
[ prop : path ] let the assumptions of proposition [ prop : spectralgapd ] be true as well as one of assumptions [ ass : g ] [ ass : rvless ] or [ ass : mon ] . further , assume that there exists an infinite subsequence @xmath557 such that for all @xmath564 there exists an incident edge with conductance greater than @xmath565
. then there exists @xmath83 such that @xmath27-a.s . for @xmath128 large enough , and all real - valued functions @xmath566 with @xmath561
we obtain @xmath567 if one of the assumptions [ ass : g ] [ ass : rvless ] or [ ass : mondec ] is fulfilled , then the constant @xmath161 can be chosen independently of @xmath60 .
we prove these propositions in the next section .
we apply lemma [ lem : pathvsrw ] .
let @xmath568 .
further , let @xmath569 .
in order to define @xmath570 as required by lemma [ lem : pathvsrw ] , we fix @xmath504 according to and let @xmath571 .
since @xmath60 is decreasing to zero , @xmath572 for @xmath58 large enough .
by lemma [ lem : spectralgapperc ] we know that there exists @xmath160 such that @xmath27-a.s . for @xmath58 large enough and all real - valued functions
@xmath139 with @xmath573 we have @xmath574 thus , if we choose @xmath575 in the place of @xmath576 in , then condition [ item : poincarecond ] of lemma [ lem : pathvsrw ] is fulfilled .
we are now going to construct the map @xmath577 and the path @xmath518 .
for the next paragraph we say that a conductance is `` bad '' if it is smaller than or equal to @xmath231 . by lemma
[ lem : injectivemap ] there exists an injective map @xmath578 such that there exists a directed path @xmath579 in @xmath386 from @xmath195 to @xmath434 of length @xmath580 .
let @xmath581 . by assumption , along the subsequence @xmath165
, each subbox @xmath582 with @xmath228 contains at most @xmath230 bad conductances .
in addition , by the construction of @xmath583 , each site @xmath584 is connected to infinity by a path that contains only good conductances .
thus , we construct the path @xmath518 by the following algorithm : the path @xmath518 follows @xmath585 until it hits an edge with a bad conductance . in this case
it takes the shortest detour around that edge to meet @xmath585 again without using any edge with a bad conductance .
since each @xmath582 with @xmath228 contains at most @xmath230 bad conductances , this detour contains less than @xmath586 edges . after the detour
, @xmath518 continues again on @xmath585 until it hits the next bad conductance and so on . by this construction
there exists a @xmath68 , such that for all @xmath587 and for @xmath58 large enough the length of the resulting path @xmath588
. we can now apply lemma [ lem : pathvsrw ] and obtain that along the subsequence @xmath165 @xmath27-a.s . for @xmath128 large enough for all @xmath589 with @xmath590
the following holds : @xmath591 since both @xmath592 and @xmath593 are independent of @xmath60 , the claim follows .
again , we apply lemma [ lem : pathvsrw ] .
let @xmath594 and @xmath595 .
further , let @xmath596 be as in proposition [ prop : spectralgapd ] and let @xmath597
. then condition [ item : poincarecond ] of lemma [ lem : pathvsrw ] is fulfilled with @xmath598 by assumption , along the subsequence @xmath165 , each site @xmath599 has only neighbors in @xmath600 and there exists a a neighbor @xmath519 such that the conductance @xmath601 .
since each box @xmath229 for @xmath228 contains at most @xmath230 links with conductance less than or equal to @xmath52 and @xmath602 , the map @xmath603 is injective . by the choice of @xmath531
, there exists a path @xmath518 of length one that fulfills the requirements of lemma [ lem : pathvsrw ] .
it follows that @xmath27-a.s .
for @xmath128 large enough the following holds for all @xmath604 with @xmath590 : @xmath605 we have assumed that one of assumptions [ ass : g ] [ ass : rvless ] or [ ass : mon ] is fulfilled .
let us first assume that assumption [ ass : g ] [ ass : mon ] is true and that the limit of @xmath606 is smaller than @xmath607 .
it follows that eventually @xmath608 and therefore @xmath27-a.s . for @xmath128 large enough @xmath609
if we assume that assumption [ ass : g ] [ ass : mondec ] is fulfilled , then eventually even @xmath610 and thus the lower bound becomes independent of @xmath611 , and @xmath593 .
let us now assume that condition [ ass : rvless ] is true .
then there exists @xmath612 such that we can write @xmath613 where @xmath614 varies slowly at infinity .
it follows that eventually @xmath615 it follows that in this case @xmath27-a.s . for @xmath128 large enough @xmath616
we start with a collection of auxiliary lemmas and give the proof of theorem [ thm : loc ] in section [ subsec : proofloc ] .
let @xmath617 be as in theorem [ thm : loc ] and let @xmath281 be the smallest integer greater than or equal to @xmath618 . for @xmath454 we abbreviate @xmath619 note that @xmath620 for @xmath454 . in lemma [ lem : uniquemax ]
we are going to see that the principal eigenvector @xmath38 tends to concentrate in the site @xmath143 that minimizes the measure @xmath621 .
in addition to @xmath108 we are also interested in properties of the site @xmath622 which is the location of the second - smallest value of @xmath621 for @xmath221 .
[ lem : loc : in ] let @xmath60 be as in and @xmath623 .
let @xmath293 such that @xmath223 . then @xmath27-a.s .
for @xmath58 large enough and for all @xmath228 the box @xmath229 contains at most @xmath230 links with conductance less than or equal to @xmath624 .
furthermore , if @xmath583 is as in definition [ def : indn ] with @xmath625 , then @xmath27-a.s . for @xmath58 large enough
the set @xmath626 is sparse and both @xmath627 . to show that @xmath628 is @xmath27-a.s .
sparse , we have to show that no two sites in @xmath628 are neighbors .
let us assume that for infinitely many @xmath58 there exists a pair of neighbors @xmath629 .
we note that at least @xmath630 closed links are necessary to separate two neighboring sites from an infinite open cluster in @xmath0 . but @xmath631 in any dimension @xmath1 .
this is a contradiction .
therefore the above considerations imply that @xmath27-a.s . for @xmath58 large enough the set @xmath628 is sparse .
for the last statement consider the following : since the quotient @xmath632 diverges for @xmath58 growing to infinity , lemma [ lem : bc : ii ] implies that @xmath27-a.s .
for @xmath58 large enough @xmath633 .
but this already implies that eventually @xmath627 .
[ lem : loc : massdn ] let the function @xmath60 be as in .
assume that there exists @xmath105 such that one of the two cases occurs : @xmath60 varies regularly at infinity with index @xmath634 or the product @xmath635 converges monotonically to zero as @xmath58 grows to infinity .
further , let @xmath636 and @xmath583 be as in definition [ def : indn ]
. then @xmath27-a.s . for @xmath58 large enough @xmath637 by lemma [ lem : loc : in ] , we can apply proposition [ prop : spectralgapd ] to the set @xmath583 , i.e. , there exists @xmath83 such that @xmath27-a.s . for @xmath58 large enough @xmath638 for any function @xmath639 with @xmath140 . in any case , the assumptions imply that the product @xmath635 converges to zero as @xmath58 grows to infinity .
therefore @xmath640 converges to zero as well .
therefore , if @xmath641 implies that @xmath27-a.s .
for @xmath58 large enough @xmath642 on the other hand , we know that for any @xmath643 the term @xmath644 diverges .
let us specifically choose @xmath645 .
now we use theorem [ thm : eigvalsupperboundsufficient ] and the fact that the dirichlet form @xmath646 majorizes @xmath647 to infer that there exists @xmath648 such that @xmath27-a.s . for @xmath58 large enough @xmath649
when we solve this inequality for @xmath650 , we obtain that @xmath651 to finish the proof we use one of the additional assumptions about @xmath60 : if @xmath60 varies regularly at infinity with index @xmath634 , then we can write @xmath652 where @xmath614 varies slowly at infinity . in this case
we observe that eventually @xmath653 which implies the claim . in the other case , i.e. , if the product @xmath635 converges monotonically to zero as @xmath58 tends to infinity , we observe that eventually @xmath654 which implies the claim as well .
we assume the contrary , i.e. , we assume that @xmath661 then we define a new function @xmath662 by setting @xmath663 note that since implies that @xmath664 , we observe that @xmath665 .
obviously , @xmath666 and @xmath667 .
therefore , by the variational formula and remark [ rem : pf ] , the dirichlet energy @xmath668 is larger than the principal dirichlet eigenvalue @xmath37 .
however , the dirichlet energy of @xmath669 is given by @xmath670\nonumber\\ & \qquad + { \mathopen{}\mathclose\bgroup\originalleft}[\sum_{x : x\sim z } w_{xz } { \mathopen{}\mathclose\bgroup\originalleft } ( \psi_1^{(n)}(x ) - \phi(z ) { \aftergroup\egroup\originalright})^2- \sum_{x : x\sim z } w_{xz } { \mathopen{}\mathclose\bgroup\originalleft } ( \psi_1^{(n)}(x ) - \psi_1^{(n)}(z ) { \aftergroup\egroup\originalright})^2{\aftergroup\egroup\originalright}]\ , .
\label{equ : direnvarphi}\end{aligned}\ ] ] evaluation of the first bracketed summand on the rhs gives : @xmath671 where the last inequality follows by the definition of @xmath672 and since the assumption implies that @xmath664 .
further , we evaluate the second bracketed summand on the rhs of as @xmath673 since we assumed that @xmath38 is nonnegative and since the assumption implies that @xmath665 , we conclude that @xmath674 where the last equality follows by the definition of @xmath675 .
when we insert and into , then we obtain @xmath676\ , .\end{aligned}\ ] ] under assumption and the assumption that @xmath659 is nonnegative it follows that the dirichlet energy of @xmath669 is not larger than @xmath37 .
this is a contradiction to the perron - frobenius theorem , see remark [ rem : pf ] .
let @xmath679 $ ] be the laplace transform of @xmath78 .
then the laplace transform of @xmath677 fulfills @xmath680 = { \mathopen{}\mathclose\bgroup\originalleft } ( \mathscr{l } [ f]{\aftergroup\egroup\originalright})^{2d}\ , .
\end{aligned}\ ] ] by virtue of the tauberian theorems , more precisely by virtue of theorem 3 in ( * ? ? ?
* xiii.5 ) ( or , equivalently theorem 1.7.1 of @xcite ) , @xmath679 $ ] varies regularly at infinity with index @xmath681 .
it follows that @xmath682 $ ] varies regularly at infinity with index @xmath683 .
hence , by another application of theorem 3 in ( * ? ? ?
* xiii.5 ) we obtain that @xmath677 varies regularly at zero with index @xmath678 .
recall the definition of the local speed measure @xmath102 , i.e. , @xmath684 for @xmath395 .
if @xmath621 is very small , then @xmath196 is a potential trap .
the smaller @xmath621 , the better the trap @xmath196 .
we now show that the random sequence @xmath101 of sites defined by @xmath685 is the sequence that we are looking for .
note that @xmath27-a.s .
the site , where the minimum is attained , is unique .
let @xmath622 be the site in @xmath48 with second - smallest measure @xmath102 .
let @xmath686 and abbreviate @xmath687 for @xmath454 .
let us recall that we assumed that one of the two following cases occurs : @xmath688 or there exists @xmath105 such that the product @xmath635 converges monotonically to zero as @xmath58 grows to infinity . in the case where @xmath76
the inverse @xmath689 varies regularly at zero with index @xmath690 ( see e.g. @xcite ) .
it follows that @xmath60 varies regularly at infinity with index @xmath691 . since in addition @xmath15
, there exists @xmath105 such that @xmath692 . in both cases
we define @xmath583 as in definition [ def : indn ] with @xmath693 .
let @xmath626 .
by virtue of lemma [ lem : loc : in ] we know that @xmath27-a.s . for @xmath58 large enough
the set @xmath628 is sparse in the sense of definition [ def : sparse ] and @xmath627 .
now , let @xmath694 and note that @xmath695 however , by virtue of lemma [ lem : loc : massdn ] we know that @xmath27-a.s . for @xmath58 large enough @xmath696 and thus @xmath27-a.s .
the limit superior of the first event on the rhs vanishes . in order to estimate the probability of the second event on the rhs of
, we now estimate @xmath697 in terms of @xmath698 . by virtue of remark [ rem : pf ] , we can assume without loss of generality that @xmath38 nonnegative .
let @xmath699 and define @xmath700 . on the event where @xmath628 is sparse
, it is further @xmath701 .
therefore we know by virtue of lemma [ lem : uniquemax ] that @xmath702 . by definition @xmath703 and thus it follows that @xmath704 moreover , on the event where @xmath628 is sparse , any neighbor of @xmath705 is in @xmath583 and therefore @xmath706 on the event where is true and @xmath628 is sparse , we hence infer that @xmath707 it remains to show that @xmath27-a.s . for @xmath58 large enough @xmath708 .
we achieve this by an investigation of the extreme value statistics of the local speed measure @xmath709 , which follows below .
for @xmath107 let @xmath710 be an ordering of the set @xmath711 .
note that this means that @xmath712 and @xmath713 .
we abbreviate @xmath714 and @xmath715 . from the previous considerations we already know that @xmath716 \leq { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ \limsup_{n\to \infty } { \mathopen{}\mathclose\bgroup\originalleft } ( e_n \cap g_n { \aftergroup\egroup\originalright } ) { \aftergroup\egroup\originalright}]\\ & \qquad\qquad\qquad\leq \lim_{n\to\infty}{\mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ e_n \cap g_n { \aftergroup\egroup\originalright } ] + \lim_{n\to\infty}{\mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ \bigcup_{k= n+1}^\infty { \mathopen{}\mathclose\bgroup\originalleft } ( e_k \cap g_k \cap e_{k-1}^{\rm c}{\aftergroup\egroup\originalright } ) { \aftergroup\egroup\originalright}]\end{aligned}\ ] ] we now split the event @xmath717 and observe that @xmath718 and thus @xmath716\\ & \quad\leq \lim_{n\to\infty}{\mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ e_n \cap g_n { \aftergroup\egroup\originalright } ] + \lim_{n\to\infty}\sum_{k = n+1}^\infty{\mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ e_k \cap g_k { \mathopen{}\mathclose\bgroup\originalleft}| \pi_{2,|b_n| } < \pi_{2,|b_{n-1}| } { \aftergroup\egroup\originalright}.{\aftergroup\egroup\originalright}]\ , { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft}[\pi_{2,|b_n| } < \pi_{2,|b_{n-1}| } { \aftergroup\egroup\originalright}]\ , .\end{aligned}\ ] ] first , we estimate the factor @xmath719 $ ] . to this end
we recall that the set @xmath720 consists of dependent random variables .
however @xmath194 is independent from @xmath721 if the sites @xmath195 and @xmath400 are not neighbors .
let us recall the definition of the even lattice @xmath722 and the odd lattice @xmath723 as in and below . with these definitions
the set @xmath724 is a set of i.i.d .
random variables , as well as is @xmath725 .
let @xmath726 be the ordering of @xmath724 and @xmath727 be the ordering of @xmath725 .
now we observe that for @xmath728 the statement @xmath729 implies that @xmath730 or @xmath731 .
thus , by virtue of ( * ? ? ?
* proposition 4.3 ) , there exists a constant @xmath68 such that for @xmath728 we have : @xmath732 & \leq { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ { \mathopen{}\mathclose\bgroup\originalleft}\ { \pi^{\rm e}_{2,|b_n| } < \pi^{\rm e}_{2,|b_{n-1}|}{\aftergroup\egroup\originalright}\ } \cup { \mathopen{}\mathclose\bgroup\originalleft}\ { \pi^{\rm o}_{2,|b_n| } < \pi^{\rm o}_{2,|b_{n-1}|}{\aftergroup\egroup\originalright}\ } { \aftergroup\egroup\originalright } ] \leq cn^{-1}\ , .\end{aligned}\ ] ] we now define @xmath677 as the distribution function of @xmath102 . by definition
, @xmath733 is an increasing function and thus @xmath734 our next aim is to extract the factor @xmath735 from inside the function argument . to this end
we observe that by virtue of lemma [ lem : distrpi ] and the assumption that @xmath78 varies regularly at zero with index @xmath103 , we know that @xmath677 varies regularly at zero with index @xmath678 . for any @xmath736
it follows that @xmath737 thus , there exists @xmath738 such that for all @xmath739 we have @xmath740 .
since both @xmath741 and @xmath742 are converging to zero as a function of @xmath58 , we obtain that for @xmath58 large enough @xmath743 we observe that by definition the random variable @xmath744 ( @xmath395 ) is an inverse pareto variable with parameter @xmath745 .
it follows that the random variable @xmath746 is exponentially distributed with parameter @xmath51 . in analogy to the definitions above , we define @xmath747 as an ordering of the set @xmath748 .
finally , we notice that the event @xmath749 implies that the sites @xmath108 and @xmath750 are not neighbors .
thus , when we apply bayes theorem to express the probability @xmath751 $ ] as a conditional probability , we notice that @xmath752 \leq { \mathbb{p}}{\mathopen{}\mathclose\bgroup\originalleft } [ { \mathopen{}\mathclose\bgroup\originalleft}.\sigma_{1,|b_n| } - \sigma_{2,|b_n| } \leq -\log { \mathopen{}\mathclose\bgroup\originalleft } ( 1 - 4\sqrt{d}\ , \alpha_n { \aftergroup\egroup\originalright } ) \,{\aftergroup\egroup\originalright}|\ , z_n \not\sim z_{(n,2 ) } { \aftergroup\egroup\originalright}]\ , .\end{aligned}\ ] ] by the condition @xmath753 , the random variables @xmath754 and @xmath755 become independent except for the condition that @xmath756 .
but since the exponential distribution is memoryless , this condition simply implies that @xmath757 is again a nonnegative exponential random variable with parameter @xmath51 .
it follows that @xmath752 \leq 1-{\rm e}^{2d \log ( 1 - 4\sqrt{d}\ , \alpha_n ) } = 1 - { \mathopen{}\mathclose\bgroup\originalleft}(1 - 4\sqrt{d}\ , \alpha_n { \aftergroup\egroup\originalright})^{2d } \leq 8 d^{\frac{3}{2 } } \alpha_n\ , .\end{aligned}\ ] ] with the same argument ( i.e. , that the exponential distribution is memoryless ) we infer that @xmath758 \leq 8 d^{\frac{3}{2 } } \alpha_k\ , , \end{aligned}\ ] ] and thus @xmath716 \leq 8 d^{\frac{3}{2 } } \lim_{n\to\infty}{\mathopen{}\mathclose\bgroup\originalleft } ( \alpha_n + c\sum_{k = n+1}^\infty \alpha_k k^{-1}{\aftergroup\egroup\originalright})\ , , \end{aligned}\ ] ] which is zero since @xmath759 decreases as @xmath760 . in remains to prove .
we already know that there exists @xmath761 such that @xmath27-a.s .
for @xmath58 large enough @xmath762 further , by virtue of the variational formula the event @xmath763 implies that @xmath764\ , .
\end{aligned}\ ] ] the claim follows .
i express my gratitude to wolfgang knig , martin slowik , renato soares dos santos , christian hirsch , and jan nagel for fruitful discussions and valuable hints , as well as michiel renger for useful suggestions .
i thank wolfgang knig for making me acquainted with the question of this paper .
further , i acknowledge the financial support of dfg within the rtg 1845 `` stochastic analysis with applications in biology , finance and physics '' .
h. kesten .
aspects of first passage percolation . in p.l .
hennequin , editor , _
cole dt de probabilits de saint flour xiv - 1984 _ , volume 1180 of _ lecture notes in mathematics _ ,
pages 125264 .
springer berlin heidelberg , 1986 .
h. kesten .
first - passage percolation . in pierre picco and jaime san martin , editors ,
_ from classical to modern probability _ , volume 54 of _ progress in probability _
, pages 93143 .
birkhuser basel , 2003 .
l. saloff - coste .
lectures on finite markov chains . in pierre bernard ,
editor , _ lectures on probability theory and statistics _ , volume 1665 of _ lecture notes in mathematics _ ,
pages 301413 .
springer berlin heidelberg , 1997 .
|
we study the asymptotic behavior of the principal eigenvector and eigenvalue of the random conductance laplacian in a large domain of @xmath0 ( @xmath1 ) with zero dirichlet condition .
we assume that the conductances @xmath2 are positive i.i.d .
random variables , which fulfill certain regularity assumptions near zero .
if @xmath3<\infty \}<1/4 $ ] , then we show that for almost every environment the principal dirichlet eigenvector asymptotically concentrates in a single site and the corresponding eigenvalue scales subdiffusively .
the threshold @xmath4 is sharp . indeed , other recent results imply that for @xmath5 the top of the dirichlet spectrum homogenizes . our proofs are based on a spatial extreme value analysis of the local speed measure , borel - cantelli arguments , the rayleigh - ritz formula , results from percolation theory , and path arguments .
|
radio emission from a pulsar is believed to originate in the open field line region of the polar cap of the neutron star .
the size , shape and location of regions of radio emission in the average profiles of pulsars is thus expected to reflect the arrangement of emission regions in the pulsar magnetosphere .
pulsar average profiles exhibit a great diversity in shape , and their classification based on the number of emission components is a useful starting point to study the emission characteristics of pulsars .
rankin ( 1993 , 1990 ; and references therein ) has carried out such a detailed classification , and has come up with the conclusion that there are two kinds of emission components core and conal in pulsar profiles , which result from two distinct types of emission mechanisms .
further , rankin proposes that the conal components arise from two nested hollow cones of emission , which along with a central core emission region , make up the complete pulsar emission beam .
the actual profile observed for a given pulsar depends on the cut that the observer s line - of - sight makes through this emission beam . from the above work
, rankin also concludes that core radiation originates from very close to the neutron star surface whereas the conal radiation comes from regions higher up in the magnetosphere .
the outer cone is postulated to originate higher up in the magnetosphere than the inner cone , but along the same set of field lines .
recent work by @xcite which compares frequency dependence of pulsar radiation features for average profiles with those for single pulses , comes up with evidence in support of such a conal beam model . a somewhat different model of the pulsar emission beam
is described by @xcite who propose that the emission within the beam is patchy , i.e. the distribution of component locations within the beam is random , rather than organized in one or more hollow cones .
one of the problems in resolving this conflict is related to a correct determination of the total number of emission components for a pulsar and the significance of their arrangement within the pulse window . in a recent paper ( gangadhara & gupta , 2001 ;
hereafter paper i ) , we have addressed these issues with the help of a novel method of studying the emission geometry of pulsars . using the sensitive `` window - thresholding '' technique ( w - t technique hereafter ) on single pulse data for a given pulsar , we first identify all possible conal emission components in the profile that can be detected given the sensitivity of the data , and determine accurate locations in pulse longitude for these . from this
we determine the total number and arrangement of emission cones in the observed profile .
then , from detection of retardation and aberration effects in the locations of the conal components , we estimate the emission height for each cone . coupling this with a dipolar field geometry and the available information about the relative orientations of the spin and rotation axes of the pulsar and the line of sight , we are able to identify the field lines from which each cone originates , thereby giving a complete solution for the emission geometry for each cone . using this technique on single pulse data for psr b0329 + 54 at two different radio frequencies
, we showed in paper i that there are as many as 9 identifiable components in this pulsar , arranged in the form of 4 concentric rings around the central core component .
the emission heights for the cones were found to increase systematically from inner to outer cones for a given frequency , ranging from @xmath0 160 to @xmath0 1150 km .
also , for a given cone , the height was found to reduce with increasing frequency , providing direct proof of the commonly accepted `` radius to frequency mapping '' paradigm . in this paper , we extend our study of emission geometry to another six pulsars . section 2 describes the observations and data analysis steps . in section 3
, we present our results , where we are able to solve for the emission geometry for 5 of the pulsars studied , and have tentative results for the sixth . in section 4
we discuss the implications of our findings for the current understanding of pulsar emission physics .
.properties of the pulsars used for emission geometry studies . [ tbl1 ] [ cols="^,^,<,^,^,^,^,^ " , ] for our study of emission geometry , we selected a set of pulsars with clearly defined properties that make it easy to apply the techniques used for psr b0329 + 54 in paper i. these included the following criteria : ( i ) the presence of multiple emission components ( or indications thereof ) in the existing profiles ; ( ii ) the presence of a reasonably well identified core component at our frequency of observation , which was the 325 mhz band of the gmrt ; ( iii ) strong enough mean flux such that reasonable signal to noise ratio ( snr ) could be expected for single pulse observations with the incoherent array mode of the gmrt .
the first two constraints above confine the selected pulsars to the category of `` triple '' or `` multiple '' pulsars as defined by @xcite , or equivalently , to the category of `` cones with cores '' pulsars , as defined by @xcite . of the possible candidates
, we avoided pulsars with very wide profiles ( e.g. psr b0826@xmath134 , psr b1541 + 09 ) as it is not easy to unambiguously identify discrete emission components in these profiles due to the extended emission over a large fraction of the pulse period .
the final list of pulsars selected for our study is given in table [ tbl1 ] , along with relevant details .
here , @xmath2 and @xmath3 refer to the standard definitions of the angle between the rotation and magnetic axes and the impact angle of the line of sight with respect to the magnetic axis , respectively .
the values for these angles are taken from @xcite .
these are not very different from the corresponding values given by @xcite the typical difference for @xmath2 is @xmath0 a few degrees and that for @xmath3 is less than @xmath0 0.5 to 1.0 degree .
the observations were carried out at the gmrt , in the 325 mhz band , using the incoherent array mode ( see @xcite for more details about the pulsar modes of operation of the gmrt ) .
the data were obtained by incoherent addition of the dual polarization signals from 12 antennas .
the bandwidth used was 16 mhz , divided into 256 spectral channels by the digital back - ends .
the raw data were integrated to a time resolution of 0.516 milliseconds before being recorded for off - line analysis , where the data were dedispersed and gated to obtain the single pulse sequence for each pulsar . during this analysis , care was taken to check and flag the data for radio frequency interference signals . for a majority of the pulsars ,
the centre frequency of the observations was 318 mhz ; for one case it was 328 mhz , and 333 mhz for another ( see table [ tbl1 ] ) .
all of the 318 mhz observations were carried out on 10 - 11 dec 2000 . for most cases ,
the total number of pulses were obtained from a single observing run .
for some pulsars ( e.g. psr b2045@xmath116 ) , data from different observing sessions ( on different days ) was combined ( after proper alignment ) to obtain the final single pulse sequence with enough pulses for subsequent analysis .
the total number of pulses for each pulsar is given in column 7 of table [ tbl1 ] .
the average profile obtained for each pulsar is shown in figure 1 .
the time resolution here is the basic sample time interval of 0.516 millisec , except for the case of psr b2111 + 46 , where the data has been integrated by a factor of 4 , to a sample interval of 2.064 millisec .
the snr obtained for the peak of the average profiles is quite good in the range 200 to 600 resulting in single pulse peak snrs of the order to 5 to 10 and better , which are needed for using the w - t technique effectively .
all the pulsars show multi - component profiles , with at least 3 or more components easily visible .
the classification of each pulsar , according to the scheme proposed by rankin ( 1993 ) , is given in col .
6 of table [ tbl1 ] , where the triple ( t ) pulsars have 3 known components and the multiple ( m ) pulsars have 5 known components . from published results related to ( i ) frequency evolution of the profile components and ( ii ) polarization properties of average profiles ( e.g. rankin 1993 , lyne & manchester 1988 and references therein ) , it is clear for most of these pulsars that the central component is the core component .
further , in almost all cases , it is apparent from figure 1 that the conal components are located asymmetrically with respect to the core component , with the tendency for each trailing conal component to be closer to the core than its leading counterpart .
this is consistent with the expectations of retardation and aberration effects due to finite emission height of the conal beams ( e.g. paper i ) . to carry out a detailed modeling of the geometry of the emission cones for each pulsar , we need to identify all possible emission components in the profiles , that can be determined within the snr limitations of our data . for this
we have used the w - t technique described by us in paper i , where we employed this technique to identify as many as 9 emission components in psr b0329 + 54 . in this technique ,
we set a window in a desired pulse longitude range and employ an intensity threshold to select the single pulses that go to make an average profile , which we refer to as a `` w - t profile '' .
we consider all those pulses which have emissions above the threshold within the window . as a result of
this averaging of selected pulses , emission components within the window improve in snr compared to other parts of the profile and are thus more easily detected in the w - t profile .
as described in paper i , we use a few different checks to ensure that the detected components are genuine .
usually , the window size and location are varied by a few bins while checking that the component shows up at the same location ( this is not always possible to do for components that are located in close proximity to known , strong components ) .
in addition , the selection threshold is varied over a reasonable range to check the persistence of the component , and to minimize picking up contributions from random noise .
these checks are robust enough to avoid detection of spurious components by chance accumulations in the selected window .
this is supported by the fact that the w - t technique does _ not _ detect any new components when applied to off - pulse regions of the profile . using the w - t technique ,
we have been able to detect new emission components in most of the pulsars studied here , as well as get accurate estimates for the locations of the existing components .
the total number of _ new _ components detected for each pulsar is given in column 8 of table 1 .
some typical examples of detection of components ( new as well as known ones ) are illustrated in the sample of w - t profiles shown in figure 2 . here ,
2(a ) & 2(e ) show examples of clear and easy detections ( usually of already known components ) ; figs .
2(b),2(d ) & 2(g ) illustrate detection of new , weak components and figs .
2(c ) , 2(f ) & 2(i ) are cases of detections of new components located in close proximity to known , strong components . in some cases
, we were able to increase the confidence of our detections by integrating the data to a larger time constant ( 2 to 4 times the original sampling rate ) and repeating the w - t analysis , albeit with a loss in resolution of the location of the component peak .
this was particularly useful in cases where the original pulse profile is relatively broad and well sampled ( e.g. psr b2111 + 46 ) and the snr is marginal for the weaker emission components . in some cases
, we detect likely candidates for new emission components but can not confirm them with enough confidence from our data set .
this is due to limitations of either ( i ) inadequate snr , ( ii ) not enough pulses for the weak components to show up reliably often enough or ( iii ) very close proximity of the suspected component to a known component , especially when the latter is a strong component . in such cases ,
we label the component detection as `` tentative '' .
we mention this specifically for each pulsar when we discuss the individual results .
to determine accurate locations in pulse longitude for each component of a pulsar , we take the corresponding w - t profile and fit a gaussian curve to the selected component . the noise contribution to each data point
is estimated from the off - pulse noise r.m.s .
of the w - t profile .
the peak of the best - fit gaussian is taken as the best estimate of the location of the component and the 1-sigma error estimate on this peak is used to derive the error in the location of the component .
for the sample detections shown in figure 2 , we also show the best fit gaussian to the component detected within the w - t window . for most cases ,
we get quite good fits , with values for the reduced chi - square @xmath0 1.0 .
the error estimates are generally found to be proportional to the snr of the fitted component , which is as expected . from our determination of the total number of the emission components , we resolve the emission geometry for each pulsar as consisting of a central core component and one or more cones of emission around it . from our best estimates of the locations of these components
, we then solve for the emission heights of the cones , as well as for the transverse location of the emitting field lines on the polar cap , using the technique described in paper i. the polar cap location of the field lines is quantified by the ratio of the distance to the `` foot '' of the field line ( from the magnetic pole ) to the distance to the `` foot '' of the last open field line in the magnetosphere
the parameter of paper i ( in this description , the foot of the field line is the point where the field line intersects the neutron star surface ) .
the results from our analysis of the 6 pulsars are summarized in tables 2 to 7 .
the table for each pulsar gives , for every emission cone that we detect ( including the known ones ) , the following information : & , the location ( with respect to location of the core component ) of the leading and trailing emission components that constitute the cone ( columns 2 and 3 ) ; , the inferred retardation plus aberration angle ( column 4 ) ; , the inferred half angle of the conal beam ( column 5 ) ; , the computed emission height for the cone ( column 6 ) ; and , the polar cap location of the associated field lines ( column 7 ) .
the emission heights of the cones are given in kilometers , as well as fraction of the light cylinder radius ( ) in column 6 .
the errors for the derived quantities ( columns 4 to 7 ) are determined from appropriate propagation of the errors on the estimates of the emission component locations ( including that for the core component ) , and do not include any effects of possible errors in the values of @xmath2 and @xmath3 .
we first discuss the results for individual pulsars and present a general summary of the results at the end of this section .
this pulsar shows a 3 component profile for frequencies below about 1 ghz and is interpreted as a typical triple ( core and one conal pair ) pulsar ( e.g. rankin 1993 ) .
however , this classification is not completely unambiguous . though the frequency evolution of the components ( e.g. gould , 1994 ) appears to indicate that the central component is a core , the polarization characteristics do not show clear , unambiguous evidence of core radiation for this component . at our frequency
, the profile shows a fairly wide emission region around the central peak , especially on the trailing side ( fig . 1 ) .
the w - t analysis readily localises the leading and trailing components in the profile .
however , in the central 10 to 15 degrees of pulse longitude we get unclear and conflicting results .
the w - t analysis shows evidence for multiple emission peaks in this region ( e.g. at @xmath14.1and 3 ) , but these are generally not found to be stable against variations of window widths and thresholds .
hence , it is difficult to unambiguously identify distinct emission components in the central region , including a clear identification of the reference core component .
it is possible that what we are seeing in this region is a somewhat tangential cut through the edge of an inner emission cone , and that the core , if present at all , is a weak component in this region .
this would be consistent with the fact that this pulsar has a relatively large value of the impact parameter @xmath3 it is the largest amongst all the pulsars studied here ( table 1 ) .
the entry in column 8 of table 1 for this pulsar reflects our uncertainty in the identification of new emission components . using a tentative identification of the core as the peak produced by setting a wide window in the central region
, we get the locations of the main leading and trailing emission components with respect to the core as @xmath18.79and + 6.76 , respectively .
in addition , there is some evidence from our analysis for the presence of weak components in the leading and trailing wings of the profile , but these can not be confirmed with the quality of our data .
solving for a single emission cone for this pulsar , the emission height and polar cap location are obtained as 310 km ( 1.2% of ) and 0.56 respectively ( see table [ tbl_0450 ] ) .
we note that the results for this pulsar should be treated with caution , as the emission geometry is probably more complicated than what we present .
the typical average profile for this well studied pulsar clearly shows 5 emission components with the central one known to be the core component .
the centers of both cones are clearly offset to earlier longitudes than the core , as would be expected due to retardation and aberration .
our analysis reveals the presence of two more emission components , located between the core and the inner conal ring .
of these two , the component on the leading side ( at @xmath12.12 ) is seen relatively easily in our analysis ( fig .
2b ) , whereas the one on the trailing side ( at + 1.56 ) is somewhat harder to detect .
this is because it is located in the relatively narrower region between the core and the trailing component of the inner cone .
we therefore interpret this pulsar as having 3 emission cones around the core component , with each cone clearly showing retardation and aberration effects . solving for the locations of the emission cones at 318 mhz gives heights of 180 , 460 and 600 km ( 0.3% , 0.7% and 0.9% of respectively ) and transverse locations on the polar cap of 0.33 , 0.41 and 0.59 ( see table [ tbl_1237 ] ) .
the average profile for this pulsar shows a core dominated triple component emission beam at 318 mhz ( fig . 1 ) .
the conal components , located at @xmath113.75and + 10.94 , clearly exhibit retardation - aberration effects . even though the snr is not very good , there are indications of extra emission components between the core and the known conal components . with some difficulty
, our analysis picks up a component at @xmath16.07from the core and one at + 8.2on the trailing side ( figs .
2c & 2d ) .
in addition , we have marginal detections of two more conal components , at @xmath110.25and + 4.61 .
these last two are detected with fairly low thresholds in the w - t technique and are , at best , tentative detections , requiring confirmation with higher sensitivity observations .
analysis of the data after further integration by a factor of 2 were used to confirm the presence of these components .
our final result for this pulsar is a detection of 3 cones for which the emission heights at 318 mhz are obtained as 290 , 410 and 570 km ( 0.8% , 1.2% and 1.6% of respectively ) , and polar cap locations are obtained as 0.43 , 0.57 and 0.63 respectively ( see table [ tbl_1821 ] ) .
this is a well known multi - component ( m - type ) pulsar , where the typical profile at meter wavelengths readily shows 4 components and has clear indications of a fifth component located close to the leading edge of the central , core component ( see fig .
2e for our detection of this component ) .
our analysis easily gives the locations for these 5 known components .
in addition , we see evidence for two more outer components , located near the leading and trailing edges of the profile , at @xmath123.73and + 19.95with respect to the core ( see fig .
2f for one of these ) .
we label these two as tentative detections , as the threshold and number of contributing pulses are rather small ( though the analysis of data integrated by a factor of 2 supports these detections ) .
further , there are hints of emission components in between the known conal components , but these can not be confirmed with the quality of our data .
we thus have 5 confirmed components ( along with 2 tentative ones ) for this pulsar , leading to an interpretation of core and 2 emission cones ( and a tentative third outer cone ) . the emission heights ( at 318 mhz ) for these 3 cones
are found to be 220 , 480 and 690 km ( 0.8 , 1.7 and 2.4 % of ) and polar cap locations of 0.61 , 0.69 and 0.74 ( table [ tbl_1857 ] ) .
this well known pulsar shows a 3 component profile where the central core component is clearly asymmetrically located with respect to the centre of the conal ring .
this is an indication of significant aberration and retardation effects for the emission cone . on using the w - t technique , we are able to detect 3 more emission components , located in the `` saddle '' regions between the core and the known conal components , at longitudes of @xmath17.2 , @xmath15.5and + 3.06with respect to the core ( see fig . 2 g for an example of one of these )
this gives a total of 5 conal components , leaving us with the somewhat difficult situation of a probable undetected conal component , which needs to be located in the narrow valley between the core and the already known trailing conal component .
there are some indications of an emission component at + 4.0 , but we can not determine this unambiguously from our analysis , as it is too close to the known component at + 4.59 . lacking this , we have the problem of deciding which of the 3 new conal components is the odd one out .
if we take the components at @xmath15.5and + 3.06as forming the second conal ring , the solutions for the two rings are : emission heights at 1240 and 2230 km ( 1.3 and 2.4 % of ) and polar cap locations of 0.28 and 0.32 this is the solution listed in table [ tbl_2045 ] .
alternatively , taking the components at @xmath17.2and + 3.06as forming the second conal ring changes the emission height estimate for this cone from 1240 km to 2090 km and the polar cap location shifts from 0.28 to 0.22 .
we believe the first option is more likely to be the real situation .
this pulsar shows a very wide but clearly triple profile with a dominant core component at meter wavelengths ( fig . 1 ) .
the location of the conal components is clearly asymmetric with respect to the core , showing retardation - aberration effects . for this pulsar
two data sets , one taken at 318 mhz in december 2000 and another taken at 333 mhz in february 2002 were available . since the data from february 2002 had better snr and more number of pulses , all the results presented here are from the analysis of this data set .
analysis of the december 2000 data produced results which are similar to those reported here , and well within the errors of the measurements .
since the profile is very wide , the data were averaged to 4 times the original time resolution of 0.516 millisec , to improve the snr for the w - t analysis .
besides confirming the location of the main components , our w - t analysis picks up 2 additional conal components , located at @xmath118.71and + 14.91with respect to the core ( see figs .
2h & 2i ) .
in addition , there are hints of another pair of components located at the leading and trailing edges of the profile ( at approximately @xmath148.3and + 38.7with respect to the core ) , but we are unable to confirm them clearly . hence we confine ourselves to 2 cones of emission around the central core component , for this pulsar
. the estimates for the emission heights at 333 mhz are 1360 km and 2080 km ( 2.8 and 4.3 % of ) , while the values for the polar cap locations are 0.22 and 0.31 ( see table [ tbl_2111 ] ) .
our main results can be summarized as follows : \(i ) all the 6 pulsars show clear signatures of an asymmetry of the cones with respect to the core radiation , in that the centers of the cones occur at earlier pulse longitudes , with respect to the location of the core . for a given pulsar , the magnitude of the effect increases from inner to outer cones .
\(ii ) interpreting the shift of the cone centers as the effect of retardation and aberration on the emitted conal beam leads to reasonable values for the emission heights . these range from about 200 km to 2200 km over all the 6 pulsars . in terms of the light cylinder distance ,
this range is from 0.3% to 4.3% of . for a given pulsar ,
the heights are found to increase systematically from the inner to the outer cones ( though in some cases the sizes of the error bars prevent this from being made a firm conclusion ) .
\(iii ) combining the emission height estimates with a dipole field line geometry results in estimates of the transverse polar cap locations of the foot of the field lines that each cone is radiated from .
these are found to vary from about 0.22 to 0.74 of the radial distance to the last open field line from the magnetic axis .
furthermore , this value is found to increase systematically from inner to outer cones for a given pulsar ( though again , for some cases , the sizes of the errors reduce the significance of this conclusion ) .
crrrrrr cone & & & & & & + no .
& & & & & & + 1 & @xmath18.79 @xmath4 0.10 & 6.76 @xmath4 0.11 & @xmath11.01 @xmath4 0.11 & 5.25 @xmath4 0.02 & 310 @xmath4 30 ( 1.2% ) & 0.56 @xmath4 0.03 + crrrrrr cone & & & & & & + no . & & & & & & + 1 & @xmath12.12 @xmath4 0.07 & 1.56 @xmath4 0.09 & @xmath10.28 @xmath4 0.07 & 1.47 @xmath4 0.05 & 180 @xmath4 40 ( 0.3% ) & 0.33 @xmath4 0.04 + 2 & @xmath14.41 @xmath4 0.03 & 2.98 @xmath4 0.06 & @xmath10.72 @xmath4 0.05 & 2.95 @xmath4 0.03 & 460 @xmath4 30 ( 0.7% ) & 0.41 @xmath4 0.02 + 3 & @xmath16.97 @xmath4 0.02 & 5.11 @xmath4 0.02 & @xmath10.93 @xmath4 0.04 & 4.82 @xmath4 0.01 & 600 @xmath4 30 ( 0.9% ) & 0.59 @xmath4 0.01 + crrrrrr cone & & & & & & + no . & & & & & & + 1 & @xmath16.07 @xmath4 0.47 & 4.61 @xmath4 0.30 & @xmath10.73 @xmath4 0.31 & 3.36 @xmath4 0.13 & 290 @xmath4 120 ( 0.8% ) & 0.43 @xmath4 0.09 + 2 & @xmath110.25 @xmath4 0.30 & 8.20 @xmath4 0.43 & @xmath11.03 @xmath4 0.29 & 5.28 @xmath4 0.13 & 410 @xmath4 120 ( 1.2% ) & 0.57 @xmath4 0.08 + 3 & @xmath113.75 @xmath4 0.39 & 10.94 @xmath4 0.27 & @xmath11.41 @xmath4 0.27 & 6.90 @xmath4 0.12 & 570 @xmath4 110 ( 1.6% ) & 0.64 @xmath4 0.06 + crrrrrr cone & & & & & & + no . & & & & & & + 1 & @xmath19.78 @xmath4 0.25 & 8.51 @xmath4 0.25 & @xmath10.64 @xmath4 0.20 & 4.58 @xmath4 0.07 & 220 @xmath4 70 ( 0.8% ) & 0.61 @xmath4 0.10 + 2 & @xmath118.06 @xmath4 0.11 & 15.30 @xmath4 0.07 & @xmath11.38 @xmath4 0.11 & 7.64 @xmath4 0.03 & 480 @xmath4 40 ( 1.7% ) & 0.69 @xmath4 0.03 + 3 & @xmath123.73 @xmath4 0.46 & 19.95 @xmath4 0.34 & @xmath11.96 @xmath4 0.30 & 9.77 @xmath4 0.12 & 690 @xmath4 100 ( 2.4% ) & 0.74 @xmath4 0.06 + crrrrrr cone & & & & & & + no . & & & & & & + 1 & @xmath15.50 @xmath4 0.28 & 3.06 @xmath4 0.13 & @xmath11.22 @xmath4 0.16 & 2.78 @xmath4 0.08 & 1240 @xmath4 160 ( 1.3% ) & 0.28 @xmath4 0.02 + 2 & @xmath18.97 @xmath4 0.04 & 4.59 @xmath4 0.03 & @xmath12.19 @xmath4 0.05 & 4.18 @xmath4 0.01 & 2230 @xmath4 50 ( 2.4% ) & 0.32 @xmath4 0.004 + crrrrrr cone & & & & & & + no . & & & & & & + 1 & @xmath118.71 @xmath4 0.72 & 14.91 @xmath4 0.53 & @xmath11.90 @xmath4 0.45 & 3.14 @xmath4 0.07 & 1360 @xmath4 320 ( 2.8% ) & 0.22 @xmath4 0.03 + 2 & @xmath135.69 @xmath4 0.11 & 29.87 @xmath4 0.16 & @xmath12.91 @xmath4 0.11 & 5.61 @xmath4 0.02 & 2080 @xmath4 80 ( 4.3% ) & 0.31 @xmath4 0.006 +
there are several interesting aspects of pulsar emission geometry that can be addressed by our study .
firstly , it should be clear from this work ( and paper i ) that it is not sufficient to study only the average profiles to understand the emission geometry .
we have shown that a careful analysis of good quality single pulse data can provide valuable extra information about the emission geometry .
it would be almost impossible to detect some of the emission components we have detected , using other techniques such as fitting multiple gaussians to the average profile . here
, the window - thresholding method used by us comes in as a handy tool to find components that are weak and/or emit intermittently , or those which are located close to other strong components . using this , we find that most triple and multiple component pulsars can be shown to have 5 or 7 components .
it is then likely that most pulsars , when observed with sufficient quality data , are going to show multiple emission cones around the central core component .
further , a careful identification of the emission components appears , almost invariably , to support the picture of a core beam plus multiple , concentric conal beams of emission , in that the conal components turn out to be arranged in equal numbers on either side of the core .
what is even more interesting is that in all cases , there is an asymmetry in the location of the core with respect to the centers of the cones , and that too always in the same sense of having the cone centers advanced with respect to the core component location .
though we have presented the detailed analysis for only a handful of pulsars , this trend is easily visible in a vast majority of pulsars that have clearly defined core and conal emission components .
a check of the data from different sources in the literature ( e.g. @xcite ; @xcite ; @xcite ; the epn database ) reveals several clear cases : psrs b0105 + 68 , b1700@xmath132 , b1737 + 13 , b1826@xmath117 , b1916 + 14 , b1946 + 35 , b2002 + 31 & b2003@xmath108 ( detailed studies of several such pulsars have been initiated and will be reported upon shortly ) .
thus it would seem that this is a fairly universal trend among pulsars that show clear core and conal emission components , and our interpretation of this as being due to retardation - aberration effects in the magnetosphere , is a reasonable one .
the emission heights that we obtain for the different pulsars have very plausible values . for psrs
b0450@xmath118 , b1237 + 35 , b1821 + 05 and b1857@xmath126 , we obtain emission heights in the range of 200 to 700 km , while for psrs b2045@xmath116 and b2111 + 46 , the values range from 1200 to 2200 km .
typical estimates of emission heights reported in literature are a few 100 km for most pulsars ( e.g. @xcite , @xcite ) .
rankin ( 1993 ) estimates the emission heights for the cones from the measured half - power widths of the conal component pairs and making the assumption that the cones are radiated from near the last open field lines of the polar cap . this yields 1 ghz height estimates of about 130 and 220 km for the two cones .
our height estimates will naturally be larger than these as we find that the emission is located on field lines that are well inside of the last open field line : a cone of the same angular width radiated from inner field lines has to come from a greater height in the magnetosphere . specifically , for psrs b1237 + 25 and b1857@xmath126 , rankin(1993 ) gives estimates of emission heights for the cones at a range of frequencies . comparing with the values at the nearby frequency of 270 mhz ( table 3 of rankin , 1993 ) , we find that our estimates for the same cone at 318 mhz are more than twice as large .
if , however , the estimates of rankin ( 1993 ) are corrected for emitting field line with @xmath5 ( as postulated in @xcite ) , then the results are in better agreement with ours .
our emission height estimates are closer to those obtained by kijak & gil ( 1997,1998 ) , though still somewhat larger .
we note that the height estimates for psrs b2045@xmath116 and b2111 + 46 are somewhat unusually large . correspondingly , these two pulsars also have rather small values for the polar cap locations of the radiating field lines , as compared to the lower emission height pulsars .
the reason for these larger heights is not clear . as a fraction of the light cylinder radius distance ,
however , all these heights are still quite small , being @xmath0 5% of at the largest .
this also means that effects of magnetic field sweepback ( see paper i ) can safely be ignored for all these cases .
strictly speaking , all our height estimates are heights of the cones _ relative to that of the core radiation _ , as the retardation - aberration of the cone centers is measured with the core component as reference . since the core radiation is thought to originate very close to the neutron star surface ( e.g. rankin , 1993 & 1990 ) , these are good estimates for the actual emission heights of the cones above the neutron star surface ( the extra height of 10 km for the radius of the neutron star is much smaller than our typical error bars ) .
however , if this assumption about the height of the core emission beam is not correct , then the height of the core emission has to be added to the height estimates we have reported for the cones in this sense , our results can be taken as lower limits to the actual emission heights .
the other effect of a finite core height will be to cause the inferred polar location of the cones to become somewhat smaller ( see eqn .
15 of paper i ) , with the fractional change diminishing progressively from inner to outer cones . for example , an emission height of 100 km for the core radiation of psr b1237 + 25 would cause the polar cap locations of the cones to reduce to 0.26 , 0.37 and 0.55 from their present values of 0.33 , 0.41 and 0.59 . in any case
, it is quite clear from our results that the core emission originates at heights that are substantially ( @xmath0 100s of km ) less than that for the cones otherwise we would not see any relative retardation - aberration effects between the core and cones . another interesting feature of the results is that they are quite self - consistent with radio emission originating within the open field line region of the magnetosphere , in that we do not obtain results for polar cap locations that are greater than 1.0 .
this can happen in principle , as the emission height estimate is completely independent of the width of the cone , whereas both these parameters are relevant for the calculation of the polar cap location ( see eqns . 9 and 15 of paper i ) .
polar cap locations greater than 1.0 can result if the emission height estimates are less than the current results by factors of 4 to 16 . in a sense then , this argues that our interpretation of component location asymmetries as being due to retardation - aberration effects is a sensible interpretation .
perhaps the most interesting result from this work is that the different cones for a pulsar originate on different sets of field lines in the magnetosphere .
clearly , all the conal radiation does _ not _ originate near the last open field line region .
in fact , it appears that a significant fraction of the polar cap region may be active in the generation of the multiple conal beams .
furthermore , our results indicate that this activity is confined along distinct and concentric rings on the polar cap . generally ,
it is thought ( e.g. @xcite , @xcite ) that this activity may be produced by spark discharges in vacuum gaps formed immediately above the neutron star surface .
for such a model , our results indicate that the sparks are confined to concentric rings on the polar cap .
we note that there are recent reports of inference of such patterns of circulating sparks for pulsars with conal - single geometries showing drifting sub - pulses ( e.g. @xcite ) .
our results tend to support a similar conclusion for multi - component pulsars with more central traverses of the observing line - of - sight . for our pulsars
, we can actually constrain the radii of these rings , via the parameter . for a 10 km radius neutron star with a period of p seconds ,
the radial size of the polar cap is just @xmath6 meters . for three concentric conal rings anchored to field lines at = 0.33 , 0.41 and 0.59 ( as in psr b1237 + 25 )
, this translates to ring radii of 41 , 51 and 73 meters .
similar numbers ( @xmath0 a few to several tens of meters ) are obtained for other pulsars also .
it is indeed remarkable that we can infer the presence of features of this size on the polar cap .
@xcite have recently proposed a model for pulsar radiation that invokes concentric rings of sparks to explain the observed properties of pulsar radiation . in their theory , the separation between concentric rings of sparks , very close to the neutron star surface , is of the order of _ h _ , the height of the vacuum gap .
they estimate the value of _ h _ to be of the order of 10 meters for pulsars with period @xmath0 1 second , which is quite similar to the typical separation of rings that we infer .
for example , for psr b1237 + 25 , the separation between first and second rings is 10 meters and that between the second and third is 22 meters . for psr b1821
+ 05 , these numbers are 23 and 12 meters respectively ; for psr b1857@xmath126 , they are 15 and 9 meters .
thus our results make interesting connections with some of the existing theoretical models and support the idea of concentric rings of sparks circulating on the polar cap as being the cause of the radio radiation .
although the activity that is the ultimate source of the radio radiation that we observe may be occurring very close to the neutron star surface , the final radiation that we see appears to originate at significant heights in the magnetosphere .
further , we have shown that the inner cones are emitted at lower altitudes and outer cones at progressively higher altitudes .
in addition , it is generally accepted that the same cone is emitted at higher heights at lower frequencies .
clearly , all this is pointing to the requirement of the appropriate combination of parameters at specific points in the magnetosphere for the plasma disturbance to produce a beam of radiation that can be seen along our line of sight through the magnetosphere .
this could be a combination of physical and geometrical parameters that contrive to give the right conditions at specific points in the magnetosphere .
the localisation of the emitting regions achieved by the kind of work reported here and in paper i should help in better constraining the parameter space and thus help the theories trying to find the correct explanation for the generation of the radio waves .
using the window - thresholding technique to analyse single pulse data from observations in the 325 mhz band of the gmrt , we have conducted a detailed analysis of the emission geometry of 6 radio pulsars that have prominent , multiple - component profiles .
we have been able to detect new emission components in almost all of the pulsars . in some cases , we report tentative detections of weak components that need further confirmation studies .
it thus looks likely that many multiple component pulsars may have at least 5 to 7 components .
the total number and location of the emission components we detect clearly supports a picture of multiple conal beams around a central core beam for these pulsars ( and are a pointer to the possibility that most pulsars may have such an emission beam structure ) .
further , for all cases of conal emission , we find that the cone centre is offset to earlier longitudes with respect to the location of the core , and this effect increases in magnitude from the innermost to the outermost cones for every pulsar .
it appears this effect may be present in most pulsars that have clear core and conal emission geometries .
we interpret this effect as being due to retardation and aberration of the conal emission beams in the magnetosphere of the pulsar . from this
, we are able to solve for the emission geometry and estimate the emission height for each cone and also the polar cap location of the field lines associated with the cone , following the technique developed in gangadhara & gupta ( 2001 ) .
we find emission height estimates in the range of 200 to 700 km for four pulsars
psrs b0450@xmath118 , b1237 + 25 , b1821 + 05 and b1857@xmath126 and 1200 to 2200 km for two others
psrs b2045@xmath116 and b2111 + 46 . in terms of the light cylinder distance ,
the emission altitudes range from 0.3% to 4.3% .
the estimates for the polar cap location of the radiating field lines lines come out to be in the range 0.22 to 0.74 of the distance to the last open field line , from the magnetic axis .
further , for every pulsar with multiple cones , the emission altitude and the polar cap location are found to increase systematically from the innermost to the outermost cone .
mapping the locations of the field lines associated with each cone , on to the neutron surface , gives concentric rings with radii of the order of a few to several tens of meters and spacings between adjacent rings of about 10 to 20 meters .
our results thus tend to support a model of concentric rings of sparks produced in the vacuum gap region just above the neutron surface ( e.g. @xcite ) . the fairly precise localization of emission regions that we have reported should provide additional information for constraining the parameter space for theories and models of radio emission mechanisms .
we thank p. ahmadi and the staff of the gmrt for help with the observations .
the gmrt is run by the national centre for radio astrophysics of the tata institute of fundamental research .
we wish to thank the referee for useful suggestions that helped improve the quality of the results , and j. gil and t. hankins for comments on the manuscript .
in addition , yg would like to acknowledge the support of the arecibo observatory of the naic for a sabbatical visit , during which the bulk of the work on this manuscript was done .
cheng , a.f . , & ruderman , m. a. 1980 , , 235 , 576 deshpande , a.a . , & rankin , j. m. 1999 , , 524 , 1008 gangadhara , r. t. , & gupta , y. 2001 , , 555 , 31 ( paper i ) gil , j. a. , gupta , y. , gothoskar , p.b . , & kijak , j. 2002 , , 565 , 500 gil , j. a. , & sendyk , m. 2000 , , 541 , 351 gould , d. m. 1994 , ph.d .
thesis , university of manchester gupta , y. , gothoskar , p. b. , joshi , b. c. , vivekanand , m. , swain , r. , sirothia , s. , & bhat , n. d. r. , 2000 , in iau colloquium 177 : pulsar astronomy 2000 and beyond , eds . m. kramer , n. wex & r. wielebinski , asp conf .
series , 105 , 277 hankins , t. h. , & rickett , b. j. 1986 , , 311 , 684 kijak , j. , & gil , j. 1997 , , 288 , 631 kijak , j. , & gil , j. 1998 , , 299 , 855 lyne , a. g. , & manchester , r. n. 1988 , , 234 , 477 mitra , d. , & rankin , j. m. 2002 , , 577 , 322 rankin , j. m. 1990 , , 352 , 247 rankin , j. m. 1993 , , 85 , 145 ruderman , m. a. & sutherland , p.g .
1975 , , 196 , 51 weisberg , j.m . ,
cordes , j.m . , lundgren , s.c . ,
dawson , b.r . , despotes , j.t . ,
morgan , j.j . , weitz , k.a . ,
zink , e.c . & backer , d.c .
, 1999 , , 121,171
|
we have conducted a detailed analysis of the emission geometry of a handful of radio pulsars that have prominent , multiple - component profiles at meter wavelengths . from careful determination of the total number of emission components and their locations in pulse longitude , we find that all of the six pulsars show clear evidence for retardation and aberration effects in the conal emission beams . using this information , coupled with a dipolar field geometry , we obtain estimates of the height and transverse location in the magnetosphere , for each of the emitting cones in these pulsars .
these results support our earlier conclusions for psr b0329 + 54 in that we find successive outer cones ( in cases of multi - cone pulsars ) being emitted at higher altitudes in the magnetosphere .
the range of inferred heights is from @xmath0 200 to @xmath0 2200 km .
the set of `` active '' field lines from which the conal emissions originate are located in the region from @xmath0 0.22 to @xmath0 0.74 of the polar cap radius . at the neutron star surface , these conal rings map to radii of a few to several tens of meters and
the separation between successive rings is about 10 to 20 meters .
we discuss the implications of these findings for the understanding of the pulsar emission geometry and for current theories and models of the emission mechanism .
|
[ [ section ] ] * 1 . * + + + + the classical approach to the representation theory of the symmetric group involves the combinatorial notion of partitions and associated standard tableaux .
a classical explicit realisation of the irreducible representations is often referred to as the seminormal form " and was already given by a. young @xcite ( see also @xcite ) . in this realisation , a distinguished ( seminormal " ) basis of the irreducible representation spaces
is indexed by standard tableaux and is such that a certain set of commuting elements , the jucys murphy elements , acts as diagonal matrices @xcite .
analogues of the seminormal representations of the symmetric group have been given explicitly , using combinatorial constructions , in many other situations .
we refer to @xcite . in all these situations , analogues of jucys
murphy elements are present and act semisimply in the representations .
the combinatorial constructions originate in an identification of the sequences of eigenvalues of these jucys
murphy elements with combinatorial objects ( in the spirit of the work of a. okounkov and a. vershik for the symmetric group @xcite ) .
we note that all these situations have a type a flavour and that the combinatorics is related to the classical one ( one considers for example strict tableaux , multi - tableaux or up - down tableaux ) .
in particular , the group naturally acting on the set of combinatorial objects is the symmetric group .
the symmetric group is the weyl group of the irreducible root system of type a. in this paper , we will consider irreducible root systems of simply - laced types .
the simply - laced types are the types a , d and e , and we will call of type ade " any object associated to such a root system ( for examples : weyl group , finite hecke algebra , affine hecke algebra , ... ) .
[ [ section-1 ] ] * 2 . *
+ + + + in most of the situations quoted above , an
affine " version of the algebras under consideration controls the eigenvalues of these jucys
murphy elements , and in turn the combinatorics involved .
the affine " algebras controlling the jucys murphy elements of the symmetric group and of the finite hecke algebra of type a are actually particular examples of affine hecke algebras ( degenerate , or graded , for the symmetric group ) .
they are affine hecke algebras associated to @xmath0 . in general , affine
hecke algebras are associated to root data coming from algebraic reductive groups , and a well - known motivation is that their representation theory forms a part of the representation theory of reductive groups defined over @xmath1-adic fields .
an affine hecke algebra comes equipped with a distinguished commutative subalgebra . in the case of @xmath0 , the jucys
murphy elements of the symmetric group and of the finite hecke algebra of type a are obtained as images ( in a well - chosen quotient ) of well - chosen generators of this commutative subalgebra .
this is the reason for the fact that representations of affine hecke algebras of @xmath0 appear in the study of these jucys
murphy elements .
in fact , in this setting , only a part of the representation theory of affine hecke algebras is used .
namely , the representations involved are the ones in which the elements of the distinguished commutative subalgebra act semisimply . such representations are called _ calibrated _ ( sometimes also completely splittable @xcite , or in another context , homogeneous @xcite ) . for the affine hecke algebras of gl , the irreducible calibrated representations are classified and described combinatorially in terms of skew partitions @xcite ( see also @xcite ) . in the degenerate case , by restriction to the symmetric group ,
one recovers the so - called skew " representations .
they were introduced and studied for example in @xcite , originally in connections with the littlewood
richardson rule . in the non - degenerate case ,
one obtains by restriction the analogues of skew representations for the finite hecke algebra of type a , whose decompositions into irreducibles also involve littlewood richardson coefficients @xcite .
calibrated representations were classified in @xcite for affine hecke algebras associated to the weight lattice of a root system of arbitrary type . here , we will follow a different line of generalisation by considering different affine hecke algebras .
as far as affine hecke algebras are concerned , we generalise the construction for the affine hecke algebra of gl in terms of skew partitions . besides , our point of view on the affine hecke algebras considered here is also that , in the spirit explained above , they are algebraic structures controlling some sort of combinatorics , that we apply to representations , in this case , of finite hecke algebras and weyl groups of type ade .
[ [ section-2 ] ] * 3 . * + + + + one goal of this paper is to provide a whole class of combinatorial models for representations of finite hecke algebras and weyl groups of type ade .
we will mostly work at the level of some affine hecke algebras of type ade and actually construct irreducible ( calibrated ) representations of these affine hecke algebras .
representations of finite hecke algebras are then obtained simply by restriction , and representations of weyl groups are obtained by a limiting procedure .
main ingredients in the construction are new sets of combinatorial objects equipped with actions of the corresponding weyl groups . these combinatorial objects are built from and generalise the usual young tableaux associated to ( skew ) partitions .
representations are constructed by explicit formulas , generalising the seminormal representations of the symmetric group associated to ( skew ) partitions .
another goal of this paper is to introduce and study some affine hecke algebras ( of type ade ) .
the combinatorial constructions evoked above are controlled by those affine hecke algebras .
we emphasize their fundamental role by introducing them as the starting point of our study .
they generalise in a natural sense the affine hecke algebra of gl . as for gl
, they are associated to reductive groups ( of type ade ) which are not semisimple .
however , as an independent motivation to consider them , they admit certain quotients by a distinguished central element which are isomorphic to affine hecke algebras associated to semisimple groups ( the most studied cases ) .
this will be more detailed below . for each type ( a , d or e )
, our construction depends on a chosen vertex of the corresponding dynkin diagram .
the classical construction for the symmetric group with the usual partitions then corresponds to type a with the chosen vertex being an extremity of the dynkin diagram .
so , already for the symmetric group , we obtain new combinatorial constructions : those associated to a vertex which is not one of the two extremities of the dynkin diagram .
[ [ section-3 ] ] * 4 . *
+ + + + we will now describe in more details the main objects and the organisation of the paper .
we first introduce some affine hecke algebra of type ade ( section [ sec - def ] ) . given an irreducible root system @xmath2 of type ade , we take an arbitrary vertex @xmath3 of the dynkin diagram of @xmath2 , and construct an affine hecke algebra @xmath4 .
the algebra @xmath4 is associated to a free @xmath5-module , denote @xmath6 , obtained by extending the root lattice of @xmath2 by one dimension ; the action of the weyl group on it depends on the choice of @xmath3 . an important step is the choice of a basis of @xmath6
. this will be the key to the combinatorial approach , as explained later .
we also use this basis to work out explicitly a finite presentation by generators and relations of the algebra @xmath4 .
this presentation helps to see that the algebra @xmath4 contains several ( up to 3 ) distinguished subalgebras isomorphic to affine hecke algebras of gl , and further how they are glued " together .
it reflects a certain covering , depending on @xmath3 , of the dynkin diagram of @xmath2 by several ( up to 3 ) subdiagrams of type a. it turns out that the free @xmath5-module @xmath6 contains an distinguished element , denoted @xmath7 , invariant under the action of the weyl group .
furthermore , the quotient module @xmath8 turns out to coincide with the following one : the root lattice of @xmath2 extended by adding the fundamental weight corresponding to the vertex @xmath3 .
this @xmath5-module , included between the root lattice and the weight lattice , automatically gives rise to an affine hecke algebra , that we denote @xmath9 , which now corresponds to a semisimple group , unlike @xmath4 .
it follows from our construction that this algebra @xmath9 is isomorphic to the quotient of @xmath4 over a relation @xmath10 , where @xmath11 is a central element of @xmath4 .
therefore , we will obtain informations on the representation theory of @xmath9 immediately from our results on @xmath4 .
this could be an independent motivation for considering the algebras @xmath4 .
as @xmath3 varies among the set of vertices of the dynkin diagram of @xmath2 , we obtain different @xmath5-modules included between the root lattice of @xmath2 and the weight lattice of @xmath2 .
we note that all such @xmath5-modules ( for types ade ) are obtained this way with the following exceptions : the root lattice of type @xmath12 and the weight lattice of type @xmath13 with @xmath14 even . to sum up , the algebras @xmath9 , when @xmath2 and @xmath3 vary , exhaust the set of affine hecke algebra of type ade associated with semisimple root data , with these two exceptions .
[ [ section-4 ] ] * 5 . *
+ + + + a crucial notion in our study is that of an action and a truncated action of the weyl group on sequences of numbers ( section [ sec - act ] ) .
these sequences are identified with characters of the commutative subalgebra of @xmath4 , and this is done using the basis of @xmath6 chosen earlier .
one goal will be to determine the characters appearing in calibrated representations ( _ i.e. _ the calibrated spectrum ) and to interpret them as combinatorial objects .
the sequences form the intermediary step between characters and combinatorial objects .
the action of the weyl group of @xmath2 is explicitly calculated on the chosen basis and is then translated as an action of the weyl group on sequences ( and later on tableaux ) .
this action generalises the permutation action of the symmetric group .
roughly speaking , the truncated action on a given sequence consists in killing " the action of some simple reflections .
it is not an action of the weyl group anymore , however the important property is that the braid relations are still satisfied .
this allows us to define a notion of orbits under this truncated action .
some of these truncated orbits will parametrise the representations constructed later .
the new combinatorial objects , the tableaux of type @xmath15 , are defined in section [ sec - tab ] .
the action of the weyl group is described explicitly , in a combinatorial way .
a new feature of the action of the weyl group , compared to the usual action of the symmetric group on tableaux , is that not only boxes are exchanged , but some of them sometimes also move to a different position .
this results in the possibility that , starting from a tableau of type @xmath15 , one obtains , after applying an element of the weyl group , something which is not a tableau of type @xmath15 .
this leads us to the notion of admissibility for a standard tableau of type @xmath15 , the precise definition being that its truncated orbit must contain only standard tableaux of type @xmath15 .
examples are given for several choices of @xmath2 and @xmath3 .
representations of @xmath4 are constructed in section [ sec - rep ] with the help of standard tableaux of type @xmath15 .
the representations are parametrised by admissible truncated orbits and the representation spaces have a basis indexed by the elements of a given admissible truncated orbits ( by admissibility , these are standard tableaux of type @xmath15 ) .
the action of the generators is given explicitly ( in a
seminormal " form ) and the representations are irreducible and calibrated . as a corollary of the construction of representations associated to admissible truncated orbits , we obtain a complete description of the calibrated spectrum of @xmath4 . roughly speaking
, the calibrated spectrum is the set of characters of the commutative subalgebra of @xmath4 which support a calibrated representation .
the result is that the calibrated spectrum of type @xmath15 is in bijection with the set of admissible standard tableaux of type @xmath15 . in a sense , this sums up the interpretation of these combinatorial objects in terms of representation theory .
it is immediate to determine which representations pass to the quotient and become irreducible calibrated representations of @xmath9 .
this property is seen directly on the standard tableaux .
a description of the calibrated spectrum of the quotient algebra @xmath9 also follows immediately from its description as a quotient of @xmath4 by a central element , together with general considerations on the calibrated spectrum of a quotient affine hecke algebra .
it is described as a subset of the set of admissible standard tableaux of type @xmath15 , namely those satisfying a condition on their contents given explicitly , and related to the expression of the fundamental weight corresponding to @xmath3 in terms of simple roots .
[ [ section-5 ] ] * 6 . *
+ + + + let us make connections with the classical situation of @xmath0 . among the algebras @xmath4 ,
the affine hecke algebra of @xmath0 appears when one takes @xmath16 and @xmath3 an extremity of the dynkin diagram . in this case
, the @xmath5-module @xmath6 is the usual free @xmath5-module with basis @xmath17 equipped with the permutation action of the symmetric group ( the simple roots are @xmath18 and @xmath19 is the basis vector added to the root lattice ) .
the distinguished invariant is @xmath20 and the quotient algebra @xmath9 is the affine hecke algebra associated to the weight lattice ( that is , associated to @xmath21 ) . for @xmath0 ,
our notion of tableaux of type @xmath15 coincides evidently with the usual notion of young tableaux associated to skew shapes .
the truncated action in this case is defined as follows .
consider the simple transposition exchanging @xmath22 and @xmath23 .
then its action on a given standard tableau is killed " if @xmath22 and @xmath23 are adjacent .
a fact underlying the representation theory of the symmetric group is that this process respects the braid relations , a result that we generalise to any type @xmath15 .
we check that , with our definitions , a truncated orbit for @xmath0 contains all standard tableaux of a given skew shape .
moreover , here any standard tableau is obviously admissible . to sum up , our construction recovers , for @xmath0 , the constructions of @xcite .
in this situation , the general result concerning the quotients @xmath9 becomes that the representations passing to the affine hecke algebra of @xmath21 are those corresponding to shapes such that the product of the contents of all boxes is equal to 1 . from the point of view of representation theory , the notion of being of the same shape for two standard tableaux is replaced , for a general type @xmath15 , by the notion of belonging to the same truncated orbit . in this sense ,
the ( admissible ) truncated orbits replace the usual skew shapes .
we note that the notion of shape is not relevant anymore since the boxes can move under the action of the weyl group .
this emphasizes the importance of the truncated action which was used to define the truncated orbits , and whose properties ultimately rely on the fact that the braid relations are preserved .
furthermore , the truncated action is necessary to define the notion of admissibility which could not have been defined using the full action of the weyl group .
we note that the notion of admissibility was not present ( or , more precisely , is trivial ) in the classical @xmath0 situation .
[ [ section-6 ] ] * 7 . *
+ + + + finally , in a last part ( section [ sec - res - lim ] ) , the restriction to finite hecke algebras ( of type ade ) is considered .
the representations , indexed by admissible truncated orbits , are already constructed .
we also obtain representations of weyl groups ( of type ade ) by a limiting procedure in the formulas for the action of the generators .
we consider the obtained representations as analogues of the skew representations of the symmetric group and of the finite hecke algebra of type a. we note that , corresponding to a given type @xmath2 , we obtain several different families of combinatorial constructions ( depending of the vertex @xmath3 ) .
the obtained representations of finite hecke algebras and weyl groups are not irreducible in general .
nevertheless , we obtain a sufficient condition for these representations to be irreducible , which is expressed combinatorially in terms of tableaux .
we name this property as to be of level 1 " since it corresponds to considering representations which pass to a certain cyclotomic " quotient of level 1 .
the term cyclotomic for these quotients , and the ones of higher levels , is motivated by the remark below . again to clarify the connections with the @xmath0 situation , we note that the cyclotomic quotients ( of any level ) exactly coincide , in the @xmath0 situation , with the well - studied ariki koike algebras @xcite , also called cyclotomic hecke algebras .
then the property of being of level 1 for a classical tableau is simply that its shape must be a usual partition ( not skew ) .
it is well - known that in this case , the corresponding representations of the finite hecke algebra of type a ( or of the symmetric group ) are irreducible .
this is the property we generalise to any type @xmath15 .
we note that a further property in the @xmath0 situation is that all irreducible representations of the finite hecke algebra are obtained as such level 1 representations , or equivalently , the cyclotomic quotients of level 1 are isomorphic to the finite hecke algebra .
this is not true anymore for a general type @xmath15 and the understanding of the algebraic structure of the cyclotomic quotients of any level ( analogues of the ariki koike algebras ) would require a further study
. 0.5 cm
let @xmath2 be a finite reduced root system and fix @xmath24 a set of simple roots for @xmath2 .
we denote by @xmath25 the root lattice : @xmath26 we denote by @xmath27 the dual root system of coroots of @xmath2 and we identify coroots in @xmath27 with elements of @xmath28 via the natural pairing between roots and coroots .
let @xmath29 be a finitely - generated free @xmath5-module with a given inclusion @xmath30 of @xmath5-modules .
assume moreover that we have elements @xmath31 , for @xmath32 , such that the set of maps @xmath33 , restricted to @xmath25 , forms the dual root system @xmath27 . the data @xmath30 and @xmath34 form a root datum . by extension of scalars ,
the coroots @xmath35 are @xmath36-linear forms on @xmath37 .
to each root @xmath32 is associated the reflection @xmath38 in @xmath39 given by @xmath40 the group generated by @xmath38 , @xmath32 , is identified with the weyl group @xmath41 associated to @xmath2 .
let @xmath42 be the reflections in @xmath41 corresponding to the simple roots @xmath43 and let @xmath44 be the order of the element @xmath45 in @xmath41 ( we have @xmath46 ) .
the group @xmath41 is generated by @xmath47 and a set of defining relations is @xmath48 the weyl group @xmath41 acts on @xmath29 via formula ( [ act - w0 ] ) since @xmath49 if @xmath50 and @xmath30 .
[ ex - prel ] the free @xmath5-module @xmath51 is called the weight lattice and is naturally identified with a subset of @xmath52 .
namely , the weight lattice @xmath53 consists of elements @xmath54 such that @xmath55 for any @xmath32 .
the fundamental weights @xmath56 associated to the set of simple roots @xmath24 are defined by @xmath57 , @xmath58 .
we then have @xmath59 we have obviously @xmath60 .
let @xmath61 be a @xmath5-module such that @xmath62 . by definition of @xmath53 , the restrictions of the maps @xmath35 to @xmath61 belong to @xmath63 .
so finally , for such a @xmath5-module @xmath61 , the natural inclusions @xmath64 and @xmath65 form a root datum .
@xmath66 as we will only consider irreducible simply - laced types , we only give the definition of one - parameter ( affine ) hecke algebras . [ [ finite - hecke - algebras . ] ] * finite hecke algebras . *
+ + + + + + + + + + + + + + + + + + + + + + + + let @xmath67 be an indeterminate . the finite hecke algebra @xmath68 is the associative @xmath69$]-algebra with unit generated by elements @xmath70 with the defining relations : @xmath71 \underbrace{g_ig_jg_i\dots}_{m_{ij}\,terms}=\underbrace{g_jg_ig_j\dots}_{m_{ij}\,terms}\ \ \ & \text{for $ i , j\in\{1,\dots , n\}$ with $ i\neq j$}\ , . \end{array}\ ] ] for any element @xmath72 , let @xmath73 be a reduced expression for @xmath74 in terms of the generators @xmath42 .
we define @xmath75 .
this definition does not depend on the reduced expresion for @xmath74 and and it is a standard fact that the set of elements @xmath76 forms a @xmath69$]-basis for @xmath68 ( _ e.g. _ @xcite ) . the specialization of the algebra @xmath68 at @xmath77 is the group algebra @xmath78 . [ [ affine - hecke - algebras . ] ] * affine hecke algebras . *
+ + + + + + + + + + + + + + + + + + + + + + + + let @xmath79 be a root datum as before . the affine hecke algebra , denoted by @xmath80 , associated to the root datum is the associative @xmath69$]-algebra with unit generated by elements @xmath81 subject to the defining relations @xmath82 , @xmath83 for any @xmath84 , and relations ( [ rel - h0 ] ) together with @xmath85 for any @xmath50 and @xmath86 .
it follows from ( [ act - w0 ] ) that the fraction in the right hand side above is a polynomial in the ( commuting ) generators @xmath87 , @xmath88 .
explicitly , we have @xmath89 we record here the special cases of formula [ rel - lu ] that will be sufficient for the sequel : @xmath90 where we used that @xmath91 .
the following set of elements is a @xmath69$]-basis of @xmath80 ( _ e.g. _ @xcite ) : @xmath92 it follows in particular that the ( multiplicative ) group formed by elements @xmath93 , @xmath50 , of @xmath80 can be identified with the ( additive ) group @xmath29 .
it follows also that the subalgebra generated by @xmath70 is isomorphic to the finite hecke algebra @xmath68 .
[ rem - fin - pres ] an affine hecke algebra @xmath80 is finitely generated and finitely presented .
indeed , let @xmath94 be elements spanning @xmath29 over @xmath5 .
then , it is obvious that @xmath80 is generated by @xmath95 together with @xmath96 .
moreover , it is straightforward to check that , for @xmath86 , the set of relations ( [ rel - lu ] ) , where @xmath50 , is implied by the set of relations ( [ rel - lu ] ) , where @xmath97 . @xmath66 [ [ parabolic - root - subdata . ] ] * parabolic root subdata .
* + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath98 be a parabolic root subsystem of @xmath2 corresponding to a subset @xmath99 .
let @xmath100 be a free @xmath5-submodule of @xmath29 containing the simple roots @xmath101 for @xmath102 .
then @xmath103 contains the whole root lattice of @xmath98 and the restrictions to @xmath103 of the maps @xmath35 , @xmath104 , give an inclusion of the dual root system @xmath105 in @xmath106 .
thus @xmath107 is a root datum to which an affine hecke algebra @xmath108 is associated .
it follows at once from the definitions that the subalgebra of @xmath80 generated by @xmath109 , @xmath102 , and @xmath93 , @xmath110 , is a quotient of @xmath108 . from the knowledge of the basis ( [ base - aff ] )
, we have that this subalgebra is actually isomorphic to @xmath108 . [ [ quotient - root - data . ] ] * quotient root data . *
+ + + + + + + + + + + + + + + + + + + + + let @xmath111 be a @xmath5-submodule of @xmath29 such that @xmath112 for any @xmath32 ; in other words , any element of @xmath111 is fixed by the weyl group @xmath41 . in particular @xmath113 .
we note @xmath114 $ ] the class of an element @xmath50 in the quotient @xmath115 .
then @xmath116\in t / i$ ] gives an inclusion @xmath117 .
further , by assumption , the maps @xmath35 , @xmath32 , factor through @xmath115 and yield an inclusion @xmath118 .
assume moreover that @xmath115 is a free @xmath5-module .
thus @xmath119 is a root datum to which an affine hecke algebra @xmath120 is associated .
it follows from the definitions that @xmath120 is a quotient of @xmath80 , the surjective morphism being given on the generators by @xmath121 and @xmath122}\in h(r , t / i)$ ] .
again , from the knowledge of the basis ( [ base - aff ] ) , it follows that @xmath120 is isomorphic to the quotient of @xmath80 over the relations @xmath123 for @xmath124 .
note moreover that the elements @xmath93 , @xmath124 , are central in @xmath80 .
in all the paper , we will consider only finite - dimensional representations of affine hecke algebras .
let @xmath79 be a root datum and @xmath80 the associated affine hecke algebra .
we denote @xmath125}h(r , t)$ ] .
a representation @xmath126 of the affine hecke algebra @xmath127 is called _ calibrated _ if the elements @xmath93 , @xmath50 , act as diagonalizable endomorphisms of @xmath126 .
recall that the ( additive ) abelian group @xmath29 is identified with the ( multiplicative ) abelian group formed by elements @xmath93 , @xmath50 .
the set @xmath128 is the set of ( multiplicative ) group homomorphisms from @xmath29 to @xmath129 ( _ i.e. _ the set of @xmath130-characters of @xmath29 ) .
[ def - c - spec ] we denote by @xmath131 the set of elements @xmath132 such that there exist a calibrated representation @xmath126 of @xmath127 and a non - zero vector @xmath133 satisfying @xmath134 we call the set @xmath131 the _ calibrated spectrum _ of the affine hecke algebra @xmath80 . in other words ,
the calibrated spectrum @xmath131 is the subset of @xmath128 consisting of all the characters of @xmath29 appearing when we decompose the calibrated representations of @xmath127 as direct sums of characters of @xmath29 ( note that by definition a calibrated representation of @xmath127 is semisimple as a representation of @xmath29 ) .
we consider a quotient root datum as in ( [ quot - rd ] ) . for later use
, we formulate below the general result relating the calibrated spectrum of the quotient affine hecke algebra @xmath120 with the calibrated spectrum of @xmath80 .
[ prop - c - spec ] we have : @xmath135 we recall that @xmath120 is the quotient of @xmath80 by the relations @xmath123 for @xmath124 .
let @xmath126 be a calibrated representation of @xmath136 .
by composition with the canonical surjective morphism from @xmath80 to its quotient @xmath120 , the representation @xmath126 can be seen as a representation of @xmath127 , which is obviously calibrated .
the characters @xmath137 of @xmath29 appearing in this representation satisfy @xmath138 for @xmath124 .
this proves the inclusion from left to right .
next , let @xmath126 be a calibrated representation of @xmath127 and let @xmath139 be a basis of @xmath126 of common eigenvectors for elements @xmath93 , @xmath50 .
each vector @xmath140 corresponds to a character @xmath141 as in definition [ def - c - spec ] .
order the basis vectors such that , for some @xmath142 , the characters @xmath143 for @xmath144 satisfy @xmath145 for all @xmath124 , and the characters @xmath146 do not .
let @xmath147 be the subspace of @xmath126 spanned by the vectors @xmath148 .
it is easy to see that @xmath147 coincides with the subspace spanned by all vectors @xmath149 , where @xmath133 and @xmath124 .
this way , we see that @xmath147 is stable by @xmath80 since elements @xmath93 , @xmath124 , are central in @xmath80 . therefore ,
if non - zero , @xmath150 becomes a representation of @xmath127 , obviously calibrated ( a basis of common eigenvectors is formed by @xmath151 ) . moreover , the elements @xmath93 , @xmath124 , act as the identity on @xmath150 , so that @xmath150 becomes a calibrated representation of @xmath136 .
this proves the inclusion from right to left .
[ rem - cal ] the set of characters in @xmath152 is naturally a subset of @xmath128 .
the proposition can be formulated equivalently as @xmath153 .
besides , we note that the proof can be made shorter if we assume that all characters in @xmath131 appear in irreducible calibrated representations of @xmath80 .
this will turn out to be true for the algebras to which we will apply the proposition ( see proposition [ prop - rep - quot ] ) .
@xmath66 [ [ characters - and - sequences . ] ] * characters and sequences . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath154 be an ordered @xmath5-basis of the free @xmath5-module @xmath29 .
let @xmath155 be the set of sequences of @xmath156 elements in @xmath130 all different from @xmath157 .
as @xmath158 is a basis of @xmath29 , a character of @xmath29 is uniquely determined by its values on elements of @xmath158 .
therefore , the set of characters @xmath128 is identified with the set of sequences @xmath159 by the following bijection @xmath160 this identification depends on the choice of the ordered basis @xmath158 .
we denote @xmath161 the subset of @xmath159 consisting of sequences corresponding to characters in the calibrated spectrum of @xmath80 , via the identification ( [ bij - seq ] ) .
thus , for any ordered basis @xmath158 of @xmath29 , the identification ( [ bij - seq ] ) restricts by definition to @xmath162 and identifies the calibrated spectrum @xmath131 with the set @xmath161 . to save space
, we omit to indicate the algebra @xmath80 in the notation @xmath161 .
this should not cause any confusion as it will always be clear which affine hecke algebra we will be considering .
from now on , let @xmath2 be an irreducible root system of simply - laced type ( that is , @xmath163 , @xmath164 , or @xmath165 , @xmath166 , or @xmath167 , @xmath168 ) .
we consider the labelling of vertices of the dynkin diagram of @xmath2 as shown in figure [ cox - ade ] and refer to it as the standard " labelling .
( 1550,900)(0,380 ) ( 100,1180)@xmath12 ( @xmath164 ) : ( 500,1200 ) ( 490,1160 ) ( 500,1200)(1,0)150 ( 650,1200 ) ( 640,1160 ) ( 650,1200)(1,0)50 ( 750,1200)@xmath169 ( 850,1200)@xmath169 ( 950,1200)(1,0)50 ( 1000,1200 ) ( 970,1160 ) ( 1000,1200)(1,0)150 ( 1150,1200 ) ( 1140,1160 ) ( 100,930)@xmath13 ( @xmath170 ) : ( 470,1050 ) ( 500,1050 ) ( 500,1050)(1,-1)100 ( 470,840 ) ( 500,850 ) ( 500,850)(1,1)100 ( 600,950 ) ( 590,910 ) ( 600,950)(1,0)150 ( 750,950 ) ( 740,910 ) ( 750,950)(1,0)50 ( 850,950)@xmath169 ( 950,950)@xmath169 ( 1050,950)(1,0)50 ( 1100,950 ) ( 1070,910 ) ( 1100,950)(1,0)150 ( 1250,950 ) ( 1240,910 ) ( 100,680)@xmath171 ( @xmath168 ) : ( 500,700 ) ( 490,720 ) ( 500,700)(1,0)150 ( 650,700 ) ( 640,720 ) ( 650,700)(1,0)150 ( 800,700 ) ( 790,720 ) ( 800,700)(0,-1)150 ( 800,550 ) ( 800,510 ) ( 800,700)(1,0)150 ( 950,700 ) ( 940,720 ) ( 950,700)(1,0)50 ( 1050,700)@xmath169 ( 1150,700)@xmath169 ( 1250,700)(1,0)50 ( 1300,700 ) ( 1290,720 ) ( 200,420 ) [ cox - ade ] standard " labelling of dynkin diagrams of type ade [ [ new - labelling - of - the - dynkin - diagram . ] ] * new labelling of the dynkin diagram . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we choose a vertex of the dynkin diagram of the root system @xmath2 and denote it by @xmath3 . given @xmath2 and @xmath3 , we fix once and for all a new labelling of the set of vertices of the dynkin diagram of @xmath2 such that ( as shown in fig .
[ cox - label ] ) : * first , the distinguished vertex @xmath3 is labelled by @xmath172 . *
then , @xmath173 is the shortest path from @xmath3 to an extremity of the dynkin diagram . * if @xmath3 is not itself an extremity , then @xmath174 is the shortest path from @xmath3 to another extremity of the dynkin diagram . *
finally , in type @xmath175 or @xmath176 , there is a trivalent vertex .
up to interchanging the two first paths , we can assume that this trivalent vertex is already labelled by @xmath156 for @xmath177 .
then @xmath178 is the shortest path from @xmath3 to the third extremity of the dynkin diagram .
we make the following remarks / conventions : @xmath179 by convention , if @xmath163 , we consider that @xmath180 , that is , in the partition above , we have @xmath181 .
@xmath179 if @xmath3 is itself an extremity , then we consider that @xmath182 , that is , in the partition above , we have @xmath183 @xmath179 it can happen that @xmath184 .
this happens whenever @xmath3 is a trivalent vertex , that is , when @xmath165 ( respectively , @xmath167 ) and @xmath3 is the vertex labelled by @xmath185 ( respectively , by @xmath186 ) in the standard labelling .
we will make no distinction between a vertex and its label in the new labelling .
we denote by @xmath187 the set of vertices of the dynkin diagram of @xmath2 .
the new labelling results in the following partition of the set @xmath187 : @xmath188 ( 1550,600)(-150,-530 ) ( -10,30 ) ( 0,0 ) ( 0,0)(1,0)50 ( 100,0)@xmath169 ( 200,0)(1,0)50 ( 240,30 ) ( 250,0 ) ( 250,0)(1,0)150 ( 380,30 ) ( 400,0 ) ( 400,0)(1,0)150 ( 540,30 ) ( 550,0 ) ( 550,0)(1,0)50 ( 650,0)@xmath169 ( 750,0)(1,0)50 ( 790,30 ) ( 800,0 ) ( 800,0)(0,-1)150 ( 820,-155 ) ( 800,-150 ) ( 800,-150)(0,-1)50 ( 800,-280)@xmath189 ( 800,-350)(0,-1)50 ( 820,-405 ) ( 800,-400 ) ( 800,0)(1,0)150 ( 940,30 ) ( 950,0 ) ( 950,0)(1,0)50 ( 1050,0)@xmath169 ( 1150,0)(1,0)50 ( 1190,30 ) ( 1200,0 ) ( 200,-500 ) [ cox - label ] new labelling of the dynkin diagram it will be very convenient to use the following notations : @xmath190 { \underline{\underline{c}}}-1:={\underline{\underline{c-1}}}\,,\quad & \text{for $ c = k+1,\dots , l''$. } \end{array}\right.\ ] ] the new labelling of the dynkin diagram induces a new labelling of the simple roots of @xmath2 , of the simple coroots of @xmath2 , of the simple reflections of @xmath41 and of the generators of @xmath68 , by the set @xmath187 as in ( [ part ] ) .
they are now respectively denoted @xmath191 we recall that , for @xmath192 , we have @xmath193 and , for @xmath194 , @xmath195 if the vertices @xmath22 and @xmath196 are connected by an edge , while @xmath197 otherwise . [ [ the - mathbbz - module - l_rmathsfv . ] ] * the @xmath5-module @xmath6 . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we extend the root lattice @xmath25 into a larger free @xmath5-module by adding a vector denoted @xmath198 , which we set to be @xmath5-linearly independent from @xmath25 .
thus , we consider the following free @xmath5-module : @xmath199 the coroots are maps from @xmath25 to @xmath5 and we extend them to maps from @xmath6 to @xmath5 by setting : @xmath200 0 & \text{otherwise . }
\end{array}\right.\ ] ] this defines an inclusion @xmath201 which , together with the natural inclusion @xmath202 , forms a root datum .
[ [ the - basis - mathcalb_rmathsfv . ] ] * the basis @xmath203 . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we set , using notations ( [ notation ] ) , @xmath204 this defines recursively a set of vectors @xmath205 . more explicitly , we have : @xmath206 \delta_{{\underline{b}}}={\varepsilon}_{r,{\mathsf{v}}}+\alpha_{1}+\alpha_{{\underline{2}}}\dots+\alpha_{{\underline{b}}}\ , , & \text{for $ b=2,\dots , l'$,}\\[0.4em ] \delta_{{\underline{\underline{c}}}}={\varepsilon}_{r,{\mathsf{v}}}+\alpha_{1}+\dots+\alpha_{k}+\alpha_{{\underline{\underline{k+1}}}}+\dots+\alpha_{{\underline{\underline{c}}}}\,,\ \ \ & \text{for $ c = k+1,\dots , l''$. } \end{array}\ ] ] it follows at once that the set @xmath207 is a @xmath5-basis of the free @xmath5-module @xmath6
. we will always use in the following the basis @xmath207 that we will order as follows : @xmath208 we emphasize that we consider @xmath203 as an ordered basis . from formulas ( [ form - basis - inv])-([form - basis ] ) , it is straightforward to calculate the action of the simple reflections @xmath209 , @xmath210 , of the weyl group @xmath41 on elements of the basis @xmath203 .
it results in the following formulas : @xmath211 and @xmath212 r_{{\underline{2}}}(\delta_{a } ) & = & \delta_{a}+\delta_{{\underline{2}}}-\delta_{1}\qquad & \text{for $ a=2,\dots , l$\ , , } \end{array}\right.\ ] ] @xmath213 r_{{\underline{\underline{k+1}}}}(\delta_{a } ) & = & \delta_{a}+\delta_{{\underline{\underline{k+1}}}}-\delta_{k}\qquad & \text{for $ a = k+1,\dots , l$\ , , } \end{array}\right.\ ] ] @xmath214 r_{{\underline{\underline{2}}}}(\delta_{{\underline{b } } } ) & = & \delta_{{\underline{b}}}+\delta_{{\underline{\underline{2}}}}-\delta_{1 } & \text{for $ b=2,\dots , l'$\ \
( if $ k=1$)\ , .
} \end{array}\right.\ ] ] finally , if the action of a simple reflection @xmath209 on a basis element @xmath215 is not specified above , it means that it is trivial : @xmath216 note that the second line in ( [ act - w4 ] ) is present only when @xmath184 . if @xmath217 then @xmath218 acts trivially on @xmath219 for @xmath220 .
given @xmath2 and @xmath3 as before , we denote by @xmath4 the affine hecke algebra associated to the root datum @xmath221 defined in section [ subsec - span ] . from the general definition of affine hecke algebras , the algebra @xmath4 is generated by @xmath109 , @xmath210 , and @xmath93 , @xmath222 , subject to the defining relations as recalled in section [ sec - prel ] .
we will make more explicit the relations by considering generators of @xmath4 associated to the basis @xmath203 of the @xmath5-module @xmath6 .
we define @xmath223 since @xmath203 is a basis of @xmath6 , the affine hecke algebra @xmath4 is generated by the following elements : @xmath224 due to remark [ rem - fin - pres ] , the set of defining relations for @xmath4 can be reduced to a finite set of relations involving the generators above .
we first have the relations involving only the generators @xmath109 , @xmath210 , which are given in ( [ rel - h0 ] ) .
we also have the relations involving only the generators @xmath225 , @xmath226 : @xmath227 then we have the relations between generators @xmath109 , @xmath210 and @xmath225 , @xmath226 .
they are given in a uniform manner by relation ( [ rel - lu ] ) , where @xmath228 runs over the basis elements in @xmath203 .
using the action of @xmath41 on the basis @xmath203 calculated in ( [ act - w1])([act - w5 ] ) , we give them explicitly .
note that only the formulas in ( [ rel - lu2 ] ) are used .
first , corresponding to ( [ act - w1 ] ) , we have : @xmath229 then , corresponding successively to ( [ act - w2])([act - w4 ] ) , we have : @xmath230 g_{{\underline{2}}}x_{a}g_{{\underline{2 } } } & = & x_{a}x_{{\underline{2}}}x_{1}^{-1 } \qquad & \text{for $ a=2,\dots , l$\ , , } \end{array}\right.\ ] ] @xmath231 g_{{\underline{\underline{k+1}}}}x_{a}g_{{\underline{\underline{k+1 } } } } & = & x_{a}x_{{\underline{\underline{k+1}}}}x_{k}^{-1 } \quad & \text{for $ a = k+1,\dots , l$\ , , } \end{array}\right.\ ] ] @xmath232 g_{{\underline{\underline{2}}}}x_{{\underline{b}}}g_{{\underline{\underline{2 } } } } & = & x_{{\underline{b}}}x_{{\underline{\underline{2}}}}x_{1}^{-1 } \qquad & \text{for $ b=2,\dots , l'$\ \ ( if $ k=1$)\ , . }
\end{array}\right.\ ] ] again , the second relation in ( [ rel - h - gx4 ] ) is present only if @xmath184 .
lastly , there are cases where @xmath109 and @xmath225 commute
. it happens if and only if @xmath233 . to give the precise list ,
let @xmath234 be the set consisting of pairs @xmath235 , where @xmath210 and @xmath226 , such that : * first , @xmath236 ; * second , @xmath235 is not one of the pairs appearing in left hand sides of ( [ rel - h - gx2])([rel - h - gx4 ] )
. then we have : @xmath237 note in particular that @xmath238 if @xmath239 .
[ rem - rel ] we will see in the next subsection that the first set of relations ( [ rel - h - gx1 ] ) occurs inside subalgebras isomorphic to affine hecke algebras of gl
. vaguely speaking , the remaining relations ( [ rel - h - gx2])([rel - h - gx5 ] ) can be seen as mixed " relations expressing how these subalgebras are glued " together to form @xmath4 . besides , relations ( [ rel - h - gx1])([rel - h - gx5 ] ) ( or equivalently , relations ( [ act - w1])([act - w5 ] ) ) will be interpreted combinatorially in section [ sec - tab ] , in terms of an action of the weyl group on some combinatorial objects to be introduced later .
@xmath66 [ [ the - case - ra_n - and - mathsfv - an - extremity . ] ] * the case @xmath241 and @xmath3 an extremity . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath242 .
we consider the particular situation where @xmath241 and @xmath3 is an extremity of the dynkin diagram ( say , @xmath243 in the standard labelling of fig .
[ cox - ade ] ) . in this case , the new labelling of the dynkin diagram of fig .
[ cox - label ] coincides with the standard one : ( 1550,50)(150,1160 ) ( 500,1200 ) ( 490,1160 ) ( 500,1200)(1,0)150 ( 650,1200 ) ( 640,1160 ) ( 650,1200)(1,0)50 ( 750,1200)@xmath169 ( 850,1200)@xmath169 ( 950,1200)(1,0)50 ( 1000,1200 ) ( 970,1160 ) ( 1000,1200)(1,0)150 ( 1150,1200 ) ( 1140,1160 ) in this case , we have @xmath244 , that is , @xmath182 and @xmath245 .
the ordered basis @xmath246 in ( [ basis ] ) of the @xmath5-module @xmath247 reads @xmath248 where @xmath249 for @xmath250 .
the weyl group here is the symmetric group @xmath251 and it acts by permuting the @xmath252 s : the simple reflection @xmath209 , @xmath253 , exchanges @xmath252 and @xmath254 . with the presentation of the preceding subsection , the algebra @xmath255 is generated by elements ( to avoid confusion later
, we rename the generators : @xmath256 and @xmath257 ) @xmath258 and the relations between generators @xmath259 and @xmath260 are ( relations ( [ rel - h - gx2])([rel - h - gx4 ] ) are not present here ) : @xmath261 \sigma_iy_j = y_j\sigma_i & \text{for $ i\in\{1,\dots , n\}$ , $ j\in\{0,1,\dots , n\}$ such that $ j\neq i-1,i$\ , . }
\end{array}\ ] ] this algebra is the usual affine hecke algebras of @xmath262 and will thus be denoted in the following by @xmath240 . [
[ distinguished - subalgebras - of - hrmathsfv . ] ] * distinguished subalgebras of @xmath4 . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we return now to the general setting of @xmath4 .
we use below the general property on parabolic root subdata recalled in section [ sec - prel ] .
we set : @xmath263 let @xmath264 .
we consider the following subset of @xmath187 ( _ i.e. _ of simple roots ) and the following submodule of @xmath6 : @xmath265 and we denote by @xmath266 the corresponding subalgebra of @xmath4 : @xmath267 for @xmath268 , we have that @xmath266 is isomorphic to @xmath269 , where @xmath270 is equal to @xmath271 , @xmath272 and @xmath273 respectively .
the isomorphisms are given on the generators by @xmath274 g_{a } & \mapsto & \sigma_a\ \
\quad ( a=1,\dots , l)\,,\\[0.2em ] x_{a } & \mapsto & y_a\
\ \quad ( a=0,1,\dots , l)\ , , \end{array}\ ] ] @xmath275 g_{1}\ \mapsto\ \sigma_1\qquad\qquad\qquad\text{and}\quad\ g_{{\underline{b}}}\ \mapsto\ \sigma_b\ \ \quad ( b=2,\dots , l')\,,\\[0.2em ]
x_{a}\ \mapsto\ y_a\ \ \
( a=0,1)\ \quad\text{and}\quad\ x_{{\underline{b}}}\ \mapsto\ y_b\ \ \quad ( b=2,\dots , l')\ , , \end{array}\ ] ] @xmath276 g_{a}\ \mapsto\ \sigma_a\ \ \ ( a=1,\dots , k)\ \quad\text{and}\ \quad g_{{\underline{\underline{c}}}}\ \mapsto\ \sigma_c\ \ \ ( c = k+1,\dots , l'')\,,\\[0.2em ] x_{a}\ \mapsto\ y_a\ \ \ ( a=0,1,\dots , k)\ \quad\text{and}\ \quad x_{{\underline{\underline{c}}}}\ \mapsto\ y_c\ \ \ ( c = k+1,\dots , l'')\ , .
\end{array}\ ] ] [ [ invariant - for - the - action - and - quotient - root - datum . ] ] * invariant for the action and quotient root datum . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + recall that @xmath3 is a chosen vertex of the dynkin diagram of @xmath2 , which is labelled by @xmath172 in the new labelling .
the corresponding simple root is @xmath277 ( in the new labelling that we always use ) .
let @xmath278 be the fundamental weight corresponding to @xmath279 .
this is the element of @xmath52 determined by @xmath280 and @xmath281 for @xmath282 . as an element of @xmath52 ,
the fundamental weight @xmath278 can be written as @xmath283 where @xmath284 , @xmath226 , are integers such that @xmath285 is the smallest possible positive integer ( that is , @xmath286 and @xmath287 ) .
we then define : @xmath288 [ lem - quot ] 1 .
the element @xmath289 is invariant under the action of @xmath41 .
the quotient module @xmath8 is isomorphic as a @xmath41-module to @xmath290 .
\(i ) by definition of the fundamental weight @xmath278 and its expression ( [ weight ] ) , we have : @xmath291 0 & \text{otherwise . }
\end{array}\right.\ ] ] from the action of @xmath292 on @xmath198 set in ( [ def - eps ] ) , it follows that @xmath293 for any @xmath210 .
\(ii ) we define a @xmath5-linear map @xmath294 from @xmath6 to @xmath290 by : @xmath295 as @xmath296 for @xmath210 , we have that @xmath294 is an homomorphism of @xmath41-modules . moreover , it is obviously surjective .
it remains to check that the kernel of @xmath294 coincides with @xmath297 . by construction
, we have @xmath298 . on the other hand , let @xmath299 .
we have : @xmath300 in particular , it implies that @xmath301 for all @xmath210 , and thus that @xmath302 since by assumption , we have @xmath287 .
we conclude that @xmath303 .
[ quot - root ] as @xmath3 runs over the set of vertices of the dynkin diagram of @xmath2 , the @xmath5-module @xmath290 varies among the set of @xmath5-modules @xmath61 such that @xmath62 . in fact , all such @xmath5-modules @xmath61 are obtained this way , with two exceptions , which are the root lattice when @xmath163 , and the weight lattice when @xmath165 and @xmath14 is even ( this follows from the explicit formulas for the fundamental weights , see _ e.g. _
we also note that , for a given @xmath2 , different choices of @xmath3 can produce the same @xmath5-modules ; for example , in type @xmath304 , the @xmath5-module @xmath290 is the root lattice for any @xmath3.@xmath66 [ [ the - algebra - overlinehrmathsfv . ] ] * the algebra @xmath9 .
* + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the @xmath5-module @xmath290 satisfies @xmath305 , and so , as recalled in example [ ex - prel ] , it defines a root datum .
we define @xmath9 to be the corresponding affine hecke algebra associated to @xmath290 .
we set : @xmath306 the following is a corollary of lemma [ lem - quot ] , where we apply the general setting of quotient root data recalled in section [ sec - prel ] .
[ cor - quot ] 1 .
the element @xmath11 is central in @xmath4 .
the quotient of @xmath4 over the relation @xmath10 is isomorphic to @xmath9 .
it is desirable to express @xmath11 in terms of the generators @xmath225 , @xmath226 , of @xmath4 that we are using . to do this
, we need to decompose the element @xmath7 in the basis @xmath203 of @xmath6 .
recall that @xmath307 where @xmath284 , @xmath226 are as explained after ( [ def - eps0 ] ) .
it is straightforward , using ( [ form - basis - inv ] ) , to check that @xmath308 and where the sum in the expression for @xmath309 is taken over @xmath210 satisfying @xmath310 .
for example , @xmath311 , @xmath312 if @xmath217 and @xmath313 if @xmath184 . to sum up , the central element @xmath11 of @xmath4 is given in terms of the generators by : @xmath314 where the powers @xmath309 are given in ( [ form - delta ] ) , and the affine hecke algebra @xmath9 is the quotient of @xmath4 over the relation @xmath315 .
[ ex - c - gl ] let @xmath241 and @xmath243 in the standard labelling .
the resulting affine hecke algebra is @xmath240 ; see section [ subsec - gl ] . in this case , the application of the formulas above yields the following expression for the central element : @xmath316 since the fundamental weight @xmath278 is given by @xmath317 .
so here , the algebra @xmath318 is the quotient over @xmath319 .
it is the affine hecke algebra associated to the weight lattice @xmath320 , or in other words , the affine hecke algebra associated to @xmath321.@xmath66
we fix @xmath2 and @xmath3 as in the preceding section . we defined a free @xmath5-module @xmath6 , on which the weyl group @xmath41 acts , with ordered basis @xmath322 we recall that the ordering of the elements of @xmath203 is relevant .
the action of the simple reflections @xmath209 , @xmath210 , on the basis elements is given explicitly in ( [ act - w1])([act - w5 ] ) .
we define @xmath323 and we label the elements of a sequence @xmath324 in the following way , in accordance with the labelling of elements of the ordered basis @xmath203 : @xmath325 we will use the notation @xmath326 for @xmath226
. it will be sometimes visually convenient to write a sequence @xmath324 as an array as follows : @xmath327 s_{{\underline{\underline{k+1 } } } } \ , , & \dots\ , , & s_{{\underline{\underline{l '' } } } } & \end{array}\\ s_{{\underline{2 } } } \ , , & \dots\ , , & s_{{\underline{l ' } } } & \end{array}\right)\ .\ ] ] [ [ truncated - operators - overliner_i - iinmathrmvert_r - on - sequences . ] ] * truncated operators @xmath328 , @xmath210 , on sequences . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we recall from section [ subsec - prel3 ] that the choice of the ordered basis @xmath203 of @xmath6 determines a bijection between the set of characters @xmath329 and the set @xmath330 .
the labelling of elements of a sequence @xmath324 was made such that the corresponding character @xmath331 is given by @xmath332 the weyl group @xmath41 acts on @xmath6 and in turn on characters of @xmath333 .
namely , for @xmath334 and @xmath72 , the character @xmath335 is given by @xmath336 [ def - ovr ] let @xmath324 corresponding to the character @xmath337 of @xmath6 . *
( i ) * for @xmath72 , we define @xmath338 to be the sequence in @xmath330 corresponding to the character @xmath339 . *
( ii ) * let @xmath340 be a symbol . for @xmath210 , we define the _ truncated operator _ @xmath328 on @xmath341 by : @xmath342 r_i({\boldsymbol{s}})\ \ & \text{otherwise . }
\end{array}\right.\ ] ] we note that , for any @xmath343 and @xmath324 , we have by definition : @xmath344 [ lem - mon ] let @xmath192 and @xmath324 .
we have @xmath345 @xmath346 let @xmath192 and let @xmath324 .
let @xmath337 be the character corresponding to @xmath347 .
we recall that @xmath348 .
therefore , by definition , we have @xmath349 the lemma then follows from the fact that @xmath350 if @xmath351 and @xmath352 if @xmath353 .
[ prop - mon ] let @xmath192 and @xmath324 .
we have : * if @xmath351 : @xmath354 @xmath355 @xmath356 ; .2 cm * if @xmath353 : @xmath357 @xmath355 @xmath358 .
\(i ) assume @xmath351 . from the definitions
, we have @xmath354 if and only if @xmath359 and @xmath360 .
the assertion of item ( i ) follows then from ( [ lem - cont ] ) .
.2 cm ( ii ) assume @xmath353 . from the definitions
, we have @xmath357 if and only if @xmath361 from ( [ lem - cont2 ] ) , the last condition is equivalent to @xmath359 .
moreover , it also follows from ( [ lem - cont2 ] ) that , if @xmath362 and @xmath359 , then @xmath363 . as a consequence of the proposition
, we obtain that the braid relations are satisfied by the truncated operators .
[ cor - mon ] the operators @xmath328 , @xmath210 , on the set @xmath341 satisfy : @xmath364 by definition , the braid relations are satisfied by the operators @xmath209 , @xmath210 , on @xmath330 . due to the remark ( [ r - ovr ] ) after definition [ def - ovr ] , it remains to check that , for any @xmath192 with @xmath194 , we have @xmath365 this is an immediate consequence of proposition [ prop - mon ] . [ [ truncated - orbits - of - sequences . ] ] * truncated orbits of sequences . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for any element @xmath72 in the weyl group , let @xmath366 be a reduced expression for @xmath74 in terms of the simple reflections @xmath209 , @xmath210 .
we then define @xmath367 due to corollary [ cor - mon ] , the element @xmath368 does not depend on the reduced expression chosen for @xmath74 . by convention @xmath369 if @xmath370 .
[ def - oo ] for @xmath324 , we define @xmath371 and we call @xmath372 the _ truncated orbit _ of @xmath347 .
note that by definition @xmath373 and @xmath374 .
moreover , we have , for @xmath324 and @xmath72 , @xmath375 this is immediate from definition [ def - ovr ] if @xmath74 is a simple reflection , and in general , it follows at once by induction on the length of @xmath74 . as a direct consequence , for @xmath376
, we have either @xmath377 or @xmath378 .
[ rem - orb ] * ( i ) * for @xmath324 the truncated orbit @xmath372 is a subset of the orbit of @xmath347 under the action ( not truncated ) of @xmath41 .
more precisely , the @xmath41-orbit of @xmath347 is partitioned into truncated orbits . in particular , the truncated orbits are finite sets , namely , @xmath379 for any @xmath324 . *
( ii ) * the truncated orbit @xmath372 of a sequence @xmath324 as defined in definition [ def - oo ] coincides with the set of all @xmath380 which can be obtained by repeated applications of the truncated operators @xmath328 , @xmath210 .
in other words , an alternative definition of @xmath372 could be @xmath381 this is a direct consequence of corollary [ cor - mon ] together with the fact that if @xmath382 then @xmath383 .
@xmath66 the action of the simple reflections @xmath209 , @xmath210 , of the weyl group @xmath41 on a sequence @xmath347 can be read off directly from formulas ( [ act - w1])([act - w5 ] ) . for complete explicitness and due to their central role in the rest of the paper , we write down the formulas .
let @xmath324 written as an array as in ( [ string - array ] ) and let @xmath210 .
the sequence @xmath384 is given by the following procedure : * first , in @xmath347 , we exchange @xmath385 and @xmath386 .
the resulting sequence is @xmath384 whenever @xmath22 is not one of the following label : @xmath387 .
* now , if @xmath388 , it means that the sequence , in the array representation ( [ string - array ] ) , ramifies " after @xmath386 ( we read a sequence from left to right ) .
then the rule to calculate @xmath384 is as follows .
we still exchange @xmath385 and @xmath386
. then after the ramification , in the branch which contained @xmath385 nothing more happens , while all coefficients in the other branches , after the ramification , are multiplied by @xmath389 .
we show explicitly the outcome of the above procedure , when @xmath388 . for clarity ,
we consider first the situation @xmath217 .
then we have : @xmath390 s_{{\underline{\underline{k+1 } } } } \ , , & \dots\ , , & s_{{\underline{\underline{l '' } } } } & \end{array}\\ \displaystyle\frac{s_{2}}{s_{1}}s_{{\underline{2 } } } \ , , & \dots\ , , & \displaystyle\frac{s_{2}}{s_{1}}s_{{\underline{l ' } } } & \end{array}\right)\ , \ ] ] @xmath391 \displaystyle\frac{s_{{\underline{2}}}}{s_{1}}s_{{\underline{\underline{k+1 } } } } \ , , & \dots\ , , & \displaystyle\frac{s_{{\underline{2}}}}{s_{1}}s_{{\underline{\underline{l '' } } } } & \end{array}\\ s_{1 } \
, , & \dots\ , , & s_{{\underline{l ' } } } & \end{array}\right)\ , \ ] ] @xmath392 \displaystyle\frac{s_{k+1}}{s_{k}}s_{{\underline{\underline{k+1 } } } } \ , , & \dots\ , , & \displaystyle\frac{s_{k+1}}{s_{k}}s_{{\underline{\underline{l '' } } } } & \end{array}\\ s_{{\underline{2 } } } \ , , & \dots\ , , & s_{{\underline{l ' } } } & \end{array}\right)\ , \ ] ] and @xmath393 is obtained similarly to ( [ act - seq3 ] ) , by exchanging the role of @xmath394 and @xmath395 . if @xmath184 the situation is more symmetrical .
we have : @xmath396 \displaystyle\frac{s_{2}}{s_{1}}s_{{\underline{\underline{2 } } } } \ , , & \dots\ , , & \displaystyle\frac{s_{2}}{s_{1}}s_{{\underline{\underline{l '' } } } } \\[1em ] \displaystyle\frac{s_{2}}{s_{1}}s_{{\underline{2 } } } \ , , & \dots & \dots\ , , & \displaystyle\frac{s_{2}}{s_{1}}s_{{\underline{l ' } } } & \end{array}\right)\ , \ ] ] and @xmath397 , @xmath398 are obtained similarly , by permuting the role of @xmath399 , @xmath400 and @xmath401 .
in this section , let @xmath402 .
we define the combinatorial notion of placed skew shapes and their associated tableaux .
we note that we use the short terminology tableaux ( of size @xmath403 ) " for tableaux associated to placed skew shapes ( of size @xmath403 ) .
[ [ skew - diagrams . ] ] * skew diagrams . * + + + + + + + + + + + + + + + + let @xmath404 be a partition of @xmath403 , that is , @xmath405 is a family of integers such that @xmath406 and @xmath407 .
we say that @xmath408 is a partition _ of size _ @xmath403 and set @xmath409 .
a pair @xmath410 is called a _
node_. for a node @xmath411 , the classical content of @xmath412 is denoted by @xmath413 and is defined by @xmath414 .
the ( young ) diagram of @xmath405 is the set of nodes @xmath415 such that @xmath416 and @xmath417 .
the diagram of @xmath408 will be represented in the plan by a left - justified array of @xmath271 rows such that the @xmath196-th row contains @xmath418 nodes for all @xmath419 ( a node will be pictured by an empty box ) .
we number the rows from top to bottom .
we identify partitions with their diagrams and say that @xmath415 is a node of @xmath408 , or @xmath420 , if @xmath415 is a node of the diagram of @xmath408 .
a skew partition consists of two partitions @xmath421 such that , as sets of nodes , @xmath422 .
we denote it by @xmath423 .
the skew diagram of @xmath423 consists of the sets of nodes which are in @xmath408 and not in @xmath424 .
for example ,
@xmath425 \cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \end{array}\ ] ] is the skew diagram corresponding to @xmath423 with @xmath426 and @xmath427 .
we sometimes put dots at some empty positions and we mark with @xmath428 the point in @xmath429 with coordinate @xmath430 ( we omit the symbol @xmath428 only when @xmath424 is empty and when the top left box is in position @xmath430 , which corresponds to a usual diagram ) .
it will be convenient to also call a skew diagram any set of nodes ( possibly with non - positive coordinates ) which is the translated of a skew diagram along the diagonal .
the size @xmath431 of a skew diagram @xmath432 is the number of nodes in @xmath432 .
the connected components of a skew diagram @xmath432 are the minimal subsets of nodes @xmath433 such that @xmath434 ( disjoint union ) and two nodes @xmath435 and @xmath436 do not lie in the same diagonal nor in adjacent diagonals if @xmath437 and @xmath438 with @xmath194 . in the example above , the skew diagram consists of the two connected components @xmath439 \box & \hspace{-0.45cm}\box \end{array } $ ] and @xmath440 \box & \hspace{-0.45cm}\box \end{array } $ ] .
we say that two skew diagrams are equivalent if one is obtained from the other by translating some connected components along the diagonal .
for example , we have @xmath425 \cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \end{array } \ \ \sim\ \ \begin{array}{ccccc } \scriptstyle{\times } & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot \\[-0.2em ] \cdot & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \cdot & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \cdot & \hspace{-0.25cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \\[-0.2em ] \cdot & \hspace{-0.25cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \end{array } \ \ \sim\ \ \begin{array}{cccc } & \hspace{-0.35 cm } & & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] & \hspace{-0.35cm}\scriptstyle{\times } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \end{array}\ \ \ \ , \ ] ] and @xmath441 \cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & & \end{array } \
\ \sim\ \
\begin{array}{cccc } & & & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \cdot & \hspace{-0.35cm}\cdot & & \\[-0.2em ] \cdot & \hspace{-0.35cm}\cdot & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \end{array } \
\ \sim\ \
\begin{array}{cccccc } & \hspace{-0.35cm}\scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & & & \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & & & \end{array } \ \ \sim\ \
\begin{array}{ccccc } \scriptstyle{\times } & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \cdot & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2$}}}\\[-0.2em ] \cdot & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & & \\[-0.2em ] \cdot & \hspace{-0.25cm}\cdot & \hspace{-0.35cm}\cdot & & \\[-0.2em ] \cdot & \hspace{-0.25cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \\[-0.2em ] \cdot & \hspace{-0.25cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & & \end{array}\ \ \ \ \ .\ ] ] [ [ placed - skew - shapes . ] ] * placed skew shapes . *
+ + + + + + + + + + + + + + + + + + + + + let @xmath67 be an indeterminate .
a placed skew diagram @xmath442 consists of an integer @xmath443 , a @xmath444-tuple of non - empty skew diagrams @xmath445 and a @xmath444-tuple of elements @xmath446 such that @xmath447 for any @xmath194 and @xmath448 .
we write @xmath449 .
a placed skew diagram @xmath449 will be represented by a @xmath444-tuple of skew diagrams @xmath445 with each place @xmath450 written above each corresponding skew diagram .
the size @xmath451 of the placed skew diagram @xmath442 is the total number of nodes , that is , @xmath452 .
for example , @xmath453 \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{cc } \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \end{array}\\[-1.0em ] \
\end{array } \right\}\ ] ] represents a placed skew diagram of size 7 .
we say that two placed skew diagrams of the form @xmath454 and @xmath455 are equivalent if @xmath456 for some @xmath448 and @xmath457 is equivalent ( as a skew diagram ) to the translated of @xmath458 by @xmath459 steps to the left if @xmath460 , and by @xmath461 steps to the right if @xmath462 .
for example , we have @xmath463 \begin{array}{cc } \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \end{array } \end{array}\right\ } \ \sim\ \ \left\{\begin{array}{c } \gamma_1q^{-2 } \\[0.5em ] \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.25cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \cdot & \hspace{-0.25cm}\fbox{\phantom{\scriptsize{$2 $ } } } & \end{array } \end{array}\right\ } \ \sim\ \
\left\{\begin{array}{c } \gamma_1q^{2 } \\[0.5em ] \begin{array}{cc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \hspace{-0.35cm}\fbox{\phantom{\scriptsize{$2 $ } } } \\[-0.2em ] \fbox{\phantom{\scriptsize{$2 $ } } } & \end{array } \end{array}\right\}\ \ \ \ \ , \ ] ] where we noticed that , by equivalence under translation along the diagonal , translating a skew diagram to the left is equivalent to translating vertically downward . note also that if @xmath464 then @xmath454 and @xmath465 are equivalent if and only if @xmath458 and @xmath457 are equivalent as skew diagrams .
then two placed skew diagrams @xmath466 and @xmath467 are said to be equivalent if @xmath468 and there is some permutation @xmath469 such that @xmath470 for @xmath471 .
note that the size of placed skew diagrams is conserved by the equivalence relation .
[ pss - a ] a _ placed skew shape ( of size @xmath403 ) _ is an equivalence class of placed skew diagrams ( of size @xmath403 ) .
[ [ tableaux . ] ] * tableaux . *
+ + + + + + + + + + + a placed node is a pair @xmath472 , where @xmath473 is a node and @xmath474 .
a placed skew diagram can be seen as a set of placed nodes .
a placed skew shape can also be seen as a set of placed nodes , up to equivalence as described above .
a tableau ( of shape @xmath442 ) is a bijection between the set of placed nodes of some placed skew shape @xmath442 and the set @xmath475 .
it is represented by filling the placed nodes of the skew diagrams representing @xmath442 by numbers @xmath476 .
the size @xmath477 of a tableau @xmath478 is the size of its shape .
one could define a tableau as a filling of the placed nodes of a placed skew diagram . then
our notion of tableaux corresponds to considering this definition up to equivalence , the equivalence being naturally induced from the equivalence of placed skew diagrams ( the nodes are translated with their numbers in them ) .
we will in all the paper consider placed skew diagrams and associated tableaux up to equivalence , very often without mentioning it anymore .
for example , all our graphical representations of tableaux in later examples are to be considered up to equivalence.@xmath66 a tableau is standard if the numbers ascend along rows and down columns in every skew diagram .
the notion of being standard or not for a tableau is well - defined since compatible with the equivalence relation on placed skew diagrams .
the @xmath67-content , or simply the content , of a placed node @xmath479 is @xmath480 , where @xmath411 .
let @xmath478 be a tableau and , for @xmath481 , let @xmath482 be the placed node with number @xmath22 .
we set @xmath483 for @xmath481 .
the contents @xmath484 are well - defined since , for a placed node of a placed skew diagram , its content is conserved by the equivalence relation of placed skew diagram ( see example ( [ ex - eq ] ) ) .
we call @xmath485 the sequence of contents of @xmath478 and denote it by @xmath486 . for example , the tableau @xmath453 \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$5 $ } } \end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{cc } \fbox{\scriptsize{$3 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$6 $ } } \\[-0.2em ] \fbox{\scriptsize{$4 $ } } & \end{array } \end{array } \right\}\ ] ] is standard and its sequence of contents is @xmath487 . for a placed skew shape @xmath442 ,
we denote by @xmath488 the set of standard tableaux of shape @xmath442 , and we set @xmath489 recall that @xmath490 .
we recall the following characterization of standard tableaux in terms of their sequences of contents and we sketch a proof ( see , for example , ( * ? ? ?
* lemma 2.2 and parts of theorem 4.1 ) ) .
[ prop - tab - a ] the set @xmath491 is in bijection with the set of sequences @xmath492 satisfying : @xmath493 we define a map @xmath294 from the set @xmath491 to the set @xmath494 by @xmath495 for any @xmath496 , the sequence @xmath497 satisfies condition ( @xmath498 ) because the following configurations @xmath499 \,\cdot & \hspace{-0.45cm}\box \end{array } $ ] , @xmath440 \,\cdot & \hspace{-0.45cm}\box \end{array } $ ] and @xmath499 \box & \hspace{-0.45cm}\box \end{array } $ ] can not appear in a skew diagram and because , in a standard tableau , whenever @xmath500 \fbox{\scriptsize{$i_3 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$i_4 $ } } \end{array}$ ] appears , we must have @xmath501 . next , we construct by induction on @xmath403 a map @xmath502 from the set of sequences in @xmath494 satisfying condition ( @xmath498 ) to the set @xmath491 such that @xmath503 and @xmath504 are identity maps .
let @xmath505 .
we define @xmath502 by mapping @xmath506 to the standard tableau @xmath507\fbox{\scriptsize{$1$}}\end{array}\right\}$ ] .
we obviously have that @xmath504 is the identity map and moreover that @xmath508 is the unique element of @xmath509 with sequence of contents @xmath510 .
therefore , we also have that @xmath503 is the identity map of @xmath509 .
now we fix @xmath511 with @xmath512 satisfying ( @xmath498 ) .
assume by induction that we have a unique element @xmath513 with sequence of contents @xmath514 .
we construct below a standard tableau @xmath478 such that @xmath478 is obtained from @xmath515 by adding a placed node with the number @xmath403 in it and with content @xmath516 .
we treat several cases . for each case
, it is straightforward to check that @xmath478 is the unique element of @xmath517 with the required property .
this implies , by induction hypothesis , that @xmath478 is the unique element of @xmath491 with content string @xmath518 .
then , setting @xmath519 , we conclude that @xmath503 and @xmath504 are identity maps .
_ case ( 1 ) .
_ assume that , for any @xmath520 , we have @xmath521 . to construct @xmath478 , we add a new connected component @xmath522 in one of the diagram of @xmath515 if @xmath523 for some @xmath524 and some @xmath525 ; otherwise we set @xmath526\fbox{\scriptsize{$n$}}\end{array}\right\}\,$ ] . _ case ( 2 ) . _
assume that there are some @xmath527 such that @xmath528 and @xmath529 .
we choose the largest @xmath22 and @xmath196 with this property .
then one of the diagram of @xmath515 contains a part of the form @xmath530 \fbox{\scriptsize{$i$ } } & \hspace{-0.20cm}\cdot \end{array}$ ] , where there is no box in the position marked by the dot . to construct @xmath478
, we add a placed node with number @xmath403 in the position marked by the dot . _ case ( 3 ) . _
assume that , for any @xmath531 , we have @xmath532 , and that there is some @xmath524 such that @xmath528 .
then , by condition ( @xmath498 ) , we have @xmath533 for any @xmath531 .
we claim that we can assume that @xmath515 contains a part of the form @xmath534
\cdot & \hspace{-0.20cm}\cdot \\[-0.1em ] \fbox{\scriptsize{$i$ } } & \hspace{-0.20cm}\cdot \end{array}$ ] , where there is no box above @xmath535 in the same column and no box in the next column to the right of @xmath535 . indeed , as @xmath533 for any @xmath531
, the positions just above @xmath535 and just to the right of @xmath535 must be both empty . in turn , as @xmath515 is a tableau , every position above @xmath535 in the same column must be empty as well , and so must be every position below the line of @xmath535 in the column right to it .
as moreover @xmath532 for any @xmath531 , we have that , if there is a box , say @xmath536 , in the next column to the right of @xmath535 then it must lie not lower than the second line above @xmath535
. therefore we can translate the connected component containing @xmath536 one step to the south - east and thus we have @xmath515 of the desired form . finally , to construct @xmath478
, we add a placed node with number @xmath403 in the position just to the right of @xmath535 in the same line .
_ case ( 4 ) .
_ assume that there is some @xmath524 such that @xmath537 , and that , for any @xmath531 , we have @xmath538 .
this situation is equivalent to case ( 3 ) by reflection through the main diagonal .
[ cor - tab - a ] two standard tableaux are the same if and only if their sequences of contents coincide .
we consider the affine hecke algebra @xmath240 ( corresponding to @xmath241 and @xmath3 an extremity ) ; see section [ subsec - gl ] .
we use the ordered basis @xmath539 of the @xmath5-module @xmath247 to describe the calibrated spectrum . as explained in section [ subsec - prel3 ] , given the basis @xmath246 , a character in the calibrated spectrum @xmath540 is identified with a sequence of eigenvalues in @xmath541 .
we have the following description of the calibrated spectrum of the affine hecke algebras @xmath240 proved in @xcite ( see also @xcite ) .
we will use it as a first step to prove a generalisation for general @xmath15 in section [ subsec - cal ] .
[ theo - gl ] the calibrated spectrum @xmath540 is in bijection with the set of standard tableaux of size @xmath542 .
more precisely , we have @xmath543 the inclusion from right to left follows from an explicit construction @xcite of calibrated representations of @xmath544 .
this construction is going to be a special case of the construction in section [ sec - rep ] for arbitrary @xmath15 .
a proof of the other inclusion can be found in .
in this section , we come back to our general setting , where we fix @xmath2 and @xmath3 as in sections [ sec - def]-[sec - act ] .
let @xmath478 be a tableau of size @xmath542 and @xmath545 .
we denote by @xmath546 the tableau of size @xmath547 obtained from @xmath478 by keeping only the placed nodes containing @xmath548 .
[ def - tab - r ] a _ tableau of type @xmath15 _ is a triplet @xmath549 such that : * @xmath550 , @xmath551 and @xmath552 are tableaux of size , respectively , @xmath553 , @xmath554 and @xmath555 ; * @xmath556 and @xmath557 coincide ; * @xmath558 and @xmath559 coincide .
a tableau @xmath549 of type @xmath15 is standard if @xmath550 , @xmath551 and @xmath560 are standard .
we set : @xmath561 for a tableau @xmath562 of type @xmath15 , we define the contents of @xmath478 as follows : @xmath563 { \mathrm{c}}_{{\underline{b}}}({\boldsymbol{t}}):={\mathrm{c}}_b({\boldsymbol{t}}_2)\ , , & \text{for $ b=2,\dots , l'$,}\\[0.4em ] { \mathrm{c}}_{{\underline{\underline{c}}}}({\boldsymbol{t}}):={\mathrm{c}}_c({\boldsymbol{t}}_3)\ , , & \text{for $ c = k+1,\dots , l''$. } \end{array}\ ] ] we define the sequence of contents @xmath478 , and denote it by @xmath486 , to be the following ordered collection of the contents of @xmath478 : @xmath564 thus , the sequence of contents @xmath486 is naturally a sequence in @xmath330 as defined in section [ sec - act ] . in particular , if @xmath565 , then we have @xmath566 for any @xmath226 . [ [ graphical - representation . ] ] * graphical representation . * + + + + + + + + + + + + + + + + + + + + + + + + + + + we will represent a tableau @xmath562 of type @xmath15 by a triplet made of the usual representations of @xmath550 , @xmath551 and @xmath552 with the following convention : * in @xmath550 , the numbers in the placed nodes are denoted @xmath567 ; * in @xmath551 , the numbers in the placed nodes are denoted @xmath568 ; * in @xmath552 , the numbers in the placed nodes are denoted @xmath569 . with this convention and according to definition [ def - tab - r ] , if two placed nodes of @xmath478 contain the same label , they coincide .
thus , for any @xmath226 , the content @xmath570 is the content of the placed node of @xmath478 containing the label @xmath196 .
[ rem - conv ] * ( i ) * assume that we have @xmath183 , that is , @xmath182 . by definition ,
@xmath551 is completely determined by @xmath550 , namely @xmath571 .
in this situation , we will omit to represent @xmath551 . *
( ii ) * similarly , assume that we have @xmath181 , that is @xmath245 .
in this situation , @xmath572 by definition and we will omit to represent @xmath552 . *
( iii ) * assume that we are in the @xmath573 situation , that is , @xmath241 and @xmath3 is an extremity of the dynkin diagram
. then both assumptions of items * ( i ) * and * ( ii ) * are satisfied , and a tableau of this type is simply a tableau as in section [ sec - tab - cla ] .
@xmath66 let @xmath575 .
we define @xmath576 and call them the substrings of @xmath347 .
note that they appear naturally in the array representation ( [ string - array ] ) of @xmath347 .
assume that the sequence @xmath347 is a sequence of contents @xmath565 , for a tableau @xmath562 of type @xmath15 .
then , the substrings @xmath577 are , respectively , the sequences of contents of the tableaux @xmath578 .
according to proposition [ prop - tab - a ] and corollary [ cor - tab - a ] , we have : * a sequence @xmath324 is the sequence of contents of a standard tableau of type @xmath15 if and only if the three substrings @xmath577 satisfy the condition @xmath579 of proposition [ prop - tab - a ] ; * a standard tableau of type @xmath15 is uniquely characterised by its sequence of contents . from now on , we very often identify a standard tableau @xmath478 of type @xmath15 with its sequence of contents @xmath580 . recall that we have , for @xmath210 , operators @xmath209 and truncated operators @xmath328 on @xmath330 .
we will abuse notation and write @xmath581 and @xmath582 instead of @xmath583 and @xmath584 .
we will also write @xmath585 for @xmath586 and call it the truncated orbit of @xmath478 .
we stress that , by definition , @xmath581 ( and @xmath582 if not @xmath340 ) is a sequence in @xmath330 .
it does not have to be a sequence of contents of a tableau of type @xmath15 , and in general , it is not .
therefore , we make the following definition .
[ def - adm - tab ] a standard tableau @xmath587 of type @xmath15 is called _ admissible _ if its truncated orbit @xmath585 only consists of sequences of contents @xmath588 with @xmath589 .
we set : @xmath590 if @xmath587 is admissible , then @xmath585 is called an admissible truncated orbit of type @xmath15 . [
[ the - action - graphically . ] ] * the action graphically . * + + + + + + + + + + + + + + + + + + + + + + + + + the action of the operators @xmath209 ( or @xmath328 ) , @xmath210 , on a tableau @xmath562 of type @xmath15 is interpreted graphically as follows .
this is a direct reformulation of the explicit action given in subsection [ subsec - act - exp ] .
let @xmath264 . to give how @xmath591 is transformed by the action of @xmath209
, there are three different situations to consider : * if @xmath592 contains two placed nodes containing respectively @xmath22 and @xmath593 , then they are exchanged . *
if @xmath592 contains neither a placed node containing @xmath22 nor a placed node containing @xmath593 , then nothing happens in @xmath592 . * if @xmath592 contains a placed node containing @xmath593 but no placed node containing @xmath22 , then , first , the placed node with @xmath593 moves to the diagonal with content @xmath484 ; second , all the placed nodes @xmath594 with numbers greater than @xmath593 move such that each ratio @xmath595 ( _ i.e. _ the axial distance ) is conserved . note that the third situation only occurs if @xmath596 . finally , we note that the procedure above gives @xmath582 when it is different from @xmath340 , and moreover , it is very easy to determine when @xmath597 : this happens if and only if the placed nodes containing @xmath22 and @xmath593 are in adjacent diagonals in the same diagram .
[ rem - act ] starting from @xmath587 and its graphical representation , we note that there is no ambiguity in the graphical representation of @xmath581 if the sequence @xmath581 corresponds again to a standard tableau of type @xmath15 ( because then it corresponds to a unique one , as recalled earlier ) .
if @xmath581 is not a sequence of contents of a standard tableau of type @xmath15 , we can still interpret it and draw it as a sequence of placed nodes but there is no canonical way to do it ( because fixing the content of a placed node only fixes the diagonal to which it belongs ) .
in this situation , this ambiguity is not important since only the fact that @xmath581 does not correspond to a standard tableau is relevant and this is easily seen from any representation of @xmath581 ( see examples below).@xmath66 [ [ a - ra_n - and - mathsfv1-in - the - standard - labelling . ] ] * ( a ) @xmath163 and @xmath243 in the standard labelling . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this is the situation corresponding to @xmath598 as in section [ subsec - gl ] .
the weyl group is the symmetric group @xmath599 .
we already noted in remark [ rem - conv ] that , in this situation , a tableau of type @xmath15 is a usual tableau ( of size @xmath600 ) as in section [ sec - tab - cla ] .
for any @xmath601 , the action of @xmath209 on a tableau @xmath478 is given simply by exchanging the placed nodes with number @xmath22 and @xmath593 .
in this particular situation , the shape of a tableau remains invariant under the action of the weyl group @xmath599 . recall that , given a placed skew shape @xmath442 , the set @xmath488 is the set of standard tableaux of shape @xmath442 .
the following is a reformulation , in our setting , of classical combinatorial properties of tableaux .
[ prop - tab - a2 ] let @xmath442 be a placed skew shape and @xmath602 .
we have : @xmath603 in particular , every standard tableau is admissible and the admissible truncated orbits are parametrised by placed skew shapes .
let @xmath442 be a placed skew shape of size @xmath600 and @xmath478 a standard tableau of shape @xmath442 .
let @xmath604 such that @xmath605 .
we need to prove that @xmath606 is standard ( it is obviously a tableau of shape @xmath442 ) .
assume that this is not true .
write @xmath607 , where @xmath608 is a reduced expression of @xmath74 in terms of @xmath42 .
take @xmath1 such that @xmath609 is standard and @xmath610 is not .
then @xmath611 is a sequence satisfying condition @xmath579 in proposition [ prop - tab - a ] and @xmath610 is not .
this is possible only if @xmath612 .
however , this implies @xmath613 , and in turn @xmath614 , which is a contradiction .
now let @xmath615 .
it remains to show that @xmath616 .
we will use induction on @xmath14 . if @xmath617 this is immediate .
assume that @xmath618 .
let @xmath619 ( respectively , @xmath620 ) be the placed node of @xmath478 ( respectively , @xmath621 ) containing @xmath14 .
as @xmath478 is standard , @xmath619 must be the rightmost node of its line and the lowest node of its column , and similarly for @xmath620 .
in particular , @xmath619 and @xmath620 can not be in adjacent diagonals , and therefore @xmath622 . recall that @xmath623 denotes the standard tableau of size @xmath14 obtained from @xmath478 by removing @xmath619 , and @xmath624 the one obtained from @xmath621 by removing @xmath620 . * if @xmath625 then @xmath623 and @xmath624 are of the same shape and , by induction hypothesis , there is @xmath626 such that @xmath627 , and in turn such that @xmath628 . *
if @xmath629 then we proceed in three steps : * * first , note that @xmath620 is a placed node of @xmath623 which is the rightmost node of its line and the lowest node of its column .
so there exists a standard tableau of the same shape than @xmath623 with @xmath630 in @xmath620 . by induction hypothesis , we can obtain it from @xmath623 by applying @xmath631 for some @xmath632 . * * then , in @xmath633 , the number @xmath14 is contained in @xmath619 and @xmath630 in @xmath620 .
so @xmath634 , and in @xmath635 , the number @xmath14 is contained in @xmath620 .
* * finally , @xmath636 is of the same shape than @xmath624 and therefore , by induction hypothesis , we have @xmath637 for some @xmath638 .
+ we conclude that @xmath639 .
this concludes the verification of @xmath640 and the proof of the proposition .
[ [ b - rd_n - ngeq4-and - mathsfv1-in - the - standard - labelling . ] ] * ( b ) @xmath165 ( @xmath170 ) and @xmath243 in the standard labelling . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the new labelling of the vertices of the dynkin diagram is ( with @xmath641 ) : ( 1550,210)(0,850 ) ( 460,1050 ) ( 500,1050 ) ( 500,1050)(1,-1)100 ( 460,840 ) ( 500,850 ) ( 500,850)(1,1)100 ( 600,950 ) ( 590,910 ) ( 600,950)(1,0)150 ( 750,950 ) ( 740,910 ) ( 750,950)(1,0)50 ( 850,950)@xmath169 ( 900,950)@xmath169 ( 1000,950)(1,0)50 ( 1050,950 ) ( 1050,910 ) we show two examples of standard tableaux of type @xmath15 : @xmath642 \begin{array}{cccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \end{array } \end{array}\biggr\}\ , \ \biggl\{\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \end{array } \end{array}\biggr\ } \right)\ , \ \
\left(\biggl\{\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{cccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3$}}\\[-0.2em ] & & & \end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$4$}}\\[-0.2em ] \ \end{array } \end{array}\biggr\}\ , \ \biggl\{\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{cccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$}}\\[-0.2em ] & & & \end{array } \end{array}\biggr\ } \right)\ ] ] where @xmath643 are two places , and so satisfy by definition @xmath644 for any @xmath448 .
the associated sequences of contents are @xmath645 \gamma_1 q^2 & \end{array}\!\!\!\bigr)$ ] and @xmath646 \gamma_1 q^{6 } & \end{array}\!\!\bigr)\,$ ] .
let @xmath647 .
we show graphically an example of a truncated orbit of an element of @xmath574 , which turns out to be admissible ( see also example [ ex - res ] ) . to save space , we omit to indicate the place @xmath648 , which is constant here , and we omit the curly brackets .
the non - indicated actions of @xmath649 give @xmath340 .
@xmath650 \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & & \end{array } , \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \end{array } \right ) & \!\!\!\stackrel{\ \ { \overline{r}}_{2_{}}\ \,}{\longleftrightarrow}\!\!\ ! & \ !
\left ( \begin{array}{cccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & & \end{array } , \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \end{array } \right ) & \!\!\!\stackrel{\ \ { \overline{r}}_{1_{}}\ \,}{\longleftrightarrow}\!\!\!\!\ ! & \!\!\!\!\!\!\left ( \begin{array}{cccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & & \end{array } , \begin{array}{ccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \end{array } \right)\\[1.2em ] & & \ \text{\large{$\updownarrow$}\,\footnotesize{${\overline{r}}_{3}$ } } & & \ \text{\large{$\updownarrow$}\,\footnotesize{${\overline{r}}_{3}$}}\\[0.8em ] & & \!\!\ ! \!\!\ ! \!\!\ ! \!\!\ !
\!\!\!\left ( \begin{array}{cccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } & & \end{array } , \begin{array}{ccccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & & & & \end{array } \right)\!\!\ ! & \!\!\!\!\!\!\stackrel{\ \ { \overline{r}}_{1_{}}\ \,}{\longleftrightarrow}\!\!\!\!\ ! & \!\!\!\left ( \begin{array}{cccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } & & \end{array } , \begin{array}{ccccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & & & & \end{array } \right)\\[1.2em ] & & & & \
\text{\large{$\updownarrow$}\,\footnotesize{${\overline{r}}_{2}$}}\\[0.6em ] & & & & \!\!\!\!\!\left ( \begin{array}{cccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } & & \end{array } , \begin{array}{ccccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.2em ] \fbox{\scriptsize{$2 $ } } & & & & \end{array } \right)\\[1.2em ] & & & & \ \text{\large{$\updownarrow$}\,\footnotesize{${\overline{r}}_{{\underline{\underline{3}}}}$}}\\[0.6em ] & & & & \!\!\!\!\!\!\!\!\!\left ( \begin{array}{cccccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] & & & & & \end{array } , \begin{array}{ccccc } \scriptstyle{\times } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{\underline{3}}}$ } } & & & & \end{array } \right ) \end{array}\ ] ] [ [ c - re_n - n678-and - mathsfv3-in - the - standard - labelling . ] ] * ( c ) @xmath167 ( @xmath168 ) and @xmath651 in the standard labelling . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the new labelling of the vertices of the dynkin diagram is ( with @xmath652 ) : ( 1550,220)(100,510 ) ( 500,700 ) ( 490,720 ) ( 500,700)(1,0)150 ( 650,700 ) ( 640,720 ) ( 650,700)(1,0)150 ( 800,700 ) ( 790,720 ) ( 800,700)(0,-1)150 ( 800,550 ) ( 800,510 ) ( 800,700)(1,0)150 ( 950,700 ) ( 940,720 ) ( 950,700)(1,0)50 ( 1050,700)@xmath169 ( 1150,700)@xmath169 ( 1250,700)(1,0)50 ( 1300,700 ) ( 1290,720 ) we show an example of a standard tableau of type @xmath15 : @xmath653 \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] \fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \end{array } \end{array}\biggr\}\ , \ \biggl\{\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$ } } & \end{array } \end{array}\biggr\}\ , \
\biggl\{\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.4em ] \fbox{\scriptsize{$2 $ } } & & \end{array } \end{array}\biggr\ } \right)\ ] ] the associated sequence of contents is @xmath654 \gamma_1q^4 & \end{array}\\ \gamma_1q^{-2 } & \end{array}\!\bigr)$ ] .
the substrings are @xmath655 , @xmath656 and @xmath657 and they satisfy condition @xmath579 of proposition [ prop - tab - a ] .
we show graphically the action of the operators @xmath328 , @xmath658 , on the element @xmath478 considered above ( @xmath659 ) . in every tableaux below ,
the node with @xmath157 has coordinates @xmath430 .
again , the place @xmath648 , which is constant here , and the curly brackets are omitted . as can be seen graphically
, @xmath660 is a sequence of placed nodes which is not a tableau .
therefore , the truncated orbit @xmath585 is not admissible ( and thus neither are the standard tableaux appearing in it ) .
@xmath661 \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{$3 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \end{array } , \begin{array}{cc } & \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$ } } & \end{array } , \begin{array}{ccccc } & & & & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$}}\\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\cdot \\[-0.2em ] & & & & \end{array } \right ) & & & \hspace{-5cm}\left ( \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \\[-0.2em ] $ \ $ & & \end{array } , \begin{array}{cc } & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } \\[-0.2em ] & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$ } } & \end{array } , \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.4em ] \fbox{\scriptsize{$1 $ } } & & \\[-0.2em ] $ \ $ & & \end{array } \right ) \\[2em ] \hspace{2.5cm}\text{\footnotesize{${\overline{r}}_{3}$}}\!\!\searrow\mspace{-18.0mu}\nwarrow & & \hspace{-1.5cm}\swarrow\mspace{-18.0mu}\nearrow\!\text{\footnotesize{${\overline{r}}_{2}$ } } & \\[0.8em ] & \hspace{-2cm}\left ( \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \\[-0.2em ] \fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \end{array } , \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$ } } & \end{array } , \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.4em ] \fbox{\scriptsize{$2 $ } } & & \end{array } \right ) & \!\!\!\stackrel{\ \ { \overline{r}}_{4_{}}\ \,}{\longleftrightarrow}\!\!\!\ ! & \!\!\left ( \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } \\[-0.2em ] \fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } & \end{array } , \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$ } } & \end{array } , \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$ } } \\[-0.4em ] \fbox{\scriptsize{$2 $ } } & & \end{array } \right)\\[2em ] \hspace{3cm}\swarrow\mspace{-18.0mu}\nearrow\!\text{\footnotesize{${\overline{r}}_{{\underline{\underline{3}}}}$ } } & & \hspace{-2.5cm}\!\text{\footnotesize{${\overline{r}}_{{\underline{2}}}$}}\!\!\searrow\mspace{-18.0mu}\nwarrow & \\[0.8em ] \left ( \begin{array}{ccccc } & & & & \hspace{-0.35cm}\fbox{\scriptsize{$3$}}\\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \\[-0.2em ] & & & & \end{array } , \begin{array}{cc } & \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$ } } & \end{array } , \begin{array}{ccc } & & \\[-0.2em ] \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{\underline{3}}}$ } } & & \end{array } \right ) & & & \hspace{-4.5cm}\left ( \begin{array}{ccc } &
\hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \\[-0.2em ] & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3$}}\\[-0.2em ] \fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$4 $ } } & \end{array } , \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{2}}$ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & \\[-0.2em ] & \end{array } , \begin{array}{ccc } & \hspace{-0.35cm}\fbox{\scriptsize{$0 $ } } & \\[-0.2em ] & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{\underline{3}}}$}}\\[-0.4em ] \fbox{\scriptsize{$2 $ } } & & \end{array } \right ) \end{array}\ ] ] in terms of sequences , we have : @xmath662 \gamma_1q^4 & \end{array}\\ \gamma_1q^{-2 } & \end{array}\!\bigr)\stackrel{\ \ { \overline{r}}_{{\underline{2}}_{}}\ \,}{\longleftrightarrow}\bigl(\gamma_1,\gamma_1q^{-2},\begin{array}{cc}\gamma_1q^{-6 } \ , , & \begin{array}{cc } \gamma_1\ , , & \gamma_1q^{-4 } \\[0.2em ] \gamma_1 & \end{array}\\ \gamma_1q^{2 } & \end{array}\!\bigr)\ , \ ] ] and , in @xmath663 , the substrings @xmath664 and @xmath665 do not satisfy condition @xmath579 of proposition [ prop - tab - a ] and therefore do not correspond to any standard tableau . [
[ d - ra_3-and - mathsfv2-in - the - standard - labelling . ] ] * ( d ) @xmath666 and @xmath667 in the standard labelling . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in the following examples , we exhibit phenomenons which were obviously not present in the classical situation of @xmath598 ( example * ( a ) * ) .
namely , we first show in an example that the values of the places have an influence on the size or the admissibility of a truncated orbit . in a second example
, we show an admissible standard tableau @xmath478 such that @xmath668 for different @xmath669 .
the new labelling of the vertices of the dynkin diagram is : ( 1550,50)(-100,1160 ) ( 500,1200 ) ( 490,1160 ) ( 500,1200)(1,0)150 ( 650,1200 ) ( 640,1160 ) ( 650,1200)(1,0)150 ( 800,1200 ) ( 790,1160 ) we take two places @xmath643 so that , by definition , @xmath670 for any @xmath448 .
consider the sequence @xmath671
\gamma_2 \end{array}\!\bigr)$ ] ( it corresponds to a standard tableau of type @xmath15 shown below ) .
we have : @xmath672 \gamma_2 \end{array}\!\bigr)\ \ \stackrel{\ \ { \overline{r}}_{2_{}}\ \,}{\longleftrightarrow}\ \ \bigl(\gamma_1,\gamma_2,\begin{array}{c}\gamma_1q^{-2 } \\[0.2em ] \gamma_1^{-1}\gamma_2 ^ 2q^2 \end{array}\!\bigr)\ ] ] assume first that @xmath673 , which is equivalent , by assumption on @xmath674 , to @xmath675 for any @xmath448 .
then , a new place is created , and a graphical representation is : @xmath676 \begin{array}{c } \fbox{\scriptsize{$0 $ } } \\[-0.2em ] \fbox{\scriptsize{$1$}}\end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$2$}}\\[-0.2em ] \phantom{\fbox{\scriptsize{$2 $ } } } \end{array } \end{array}\!\!\biggr\}\ , \
\biggl\{\!\!\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$0 $ } } \\[-0.2em ] \fbox{\scriptsize{$1$}}\end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{${\underline{2}}$}}\\[-0.2em ] \
\end{array } \end{array}\!\!\biggr\ } \right ) \stackrel{\ \ { \overline{r}}_{2_{}}\ \,}{\longleftrightarrow } \left(\biggl\{\!\!\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$0 $ } } \\[-0.2em ] \fbox{\scriptsize{$2$}}\end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$1$}}\\[-0.2em ] \phantom{\fbox{\scriptsize{$1 $ } } } \end{array } \end{array}\!\!\biggr\}\ , \
\biggl\{\!\!\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$0 $ } } \\[-0.2em ] \
\end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$1$}}\\[-0.2em ] \phantom{\fbox{\scriptsize{$1 $ } } } \end{array } \end{array } , \begin{array}{c } \gamma_1^{-1}\gamma_2 ^ 2 \\[0.5em ] \begin{array}{c } \scriptstyle{\times}\ \fbox{\scriptsize{${\underline{2}}$}}\\[-0.2em ] \
\end{array } \end{array}\!\!\biggr\ } \right)\ ] ] in this case , it can be checked that this is part of an admissible truncated orbit containing @xmath677 elements .
now assume that @xmath678 for some @xmath448 ( @xmath459 is odd if @xmath679 ) .
then @xmath680 and therefore , in this situation , no new place is created . instead , the placed nodes with @xmath681 has to move in the diagram with place @xmath648 .
below , we only show how the second tableau ( with numbers 0,1,@xmath681 ) is transformed by @xmath682 in the situation , respectively , @xmath683 , @xmath684 , @xmath685 and @xmath686 .
@xmath687 \begin{array}{c } \fbox{\scriptsize{$0 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{2}}$}}\end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$1 $ } } \\[-0.2em ] \phantom{\fbox{\scriptsize{$1 $ } } } \end{array } \end{array}\!\!\biggr\ } \ , \ \ \ \
\biggl\{\!\!\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\cdot \\[-0.2em ] \cdot & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{2}}$}}\end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$1 $ } } \\[-0.2em ] \phantom{\fbox{\scriptsize{$1 $ } } } \end{array } \end{array}\!\!\biggr\ } \ , \ \ \ \
\biggl\{\!\!\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{2}}$ } } \\[-0.2em ] & \end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$1 $ } } \\[-0.2em ] \phantom{\fbox{\scriptsize{$1 $ } } } \end{array } \end{array}\!\!\biggr\ } \ , \ \ \ \
\biggl\{\!\!\begin{array}{c } \gamma_1 \\[0.5em ] \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\cdot & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{2}}$ } } \\[-0.2em ] & & \end{array } \end{array } , \begin{array}{c } \gamma_2 \\[0.5em ] \begin{array}{c } \fbox{\scriptsize{$1 $ } } \\[-0.2em ] \phantom{\fbox{\scriptsize{$1 $ } } } \end{array } \end{array}\!\!\biggr\}\ ] ] we see at once that we have a non - admissible situation for @xmath684
. it can be checked , as before , that for any other values of @xmath459 , we have an admissible truncated orbit .
however , now , the truncated orbit still contains 12 elements if and only if @xmath688 . if @xmath685 then the truncated orbit contains 6 elements while if @xmath683 the truncated orbit only contains @xmath185 elements . indeed , we have in general @xmath689 \gamma_1q^{-2 } \end{array}\!\bigr)$ ] and so it coincides with @xmath690 if @xmath683 .
we write the truncated orbit when @xmath683 , that is when @xmath691 ( the non - indicated actions of @xmath692 give @xmath340 ) : @xmath693 -\gamma_1q^{-2 } \end{array}\!\bigr)\ \ \stackrel{{\overline{r}}_{2_{}}\,,\,{\overline{r}}_{{\underline{2}}_{}}}{\longleftrightarrow}\ \
\bigl(\gamma_1,-\gamma_1q^{-2},\begin{array}{c}\gamma_1q^{-2 } \\[0.2em ] \gamma_1q^{-2 } \end{array}\!\bigr)\ \ \stackrel{{\overline{r}}_{1_{}}}{\longleftrightarrow}\ \ \bigl(-\gamma_1q^{-2},\gamma_1,\begin{array}{c}\gamma_1q^{-2 } \\[0.2em ] \gamma_1q^{-2 } \end{array}\!\bigr)\ ] ]
we recall that the generators of the affine hecke algebra @xmath4 are @xmath109 , @xmath210 , and @xmath694 , @xmath226 .
let @xmath695 be an admissible truncated orbit in @xmath330 . by definition [ def - adm - tab ] of admissibility ,
the orbit @xmath695 consists only of elements of the form @xmath486 with @xmath587 ; we identify @xmath478 and @xmath486 .
let @xmath696 be a @xmath130-vector space with basis @xmath697 and set @xmath698 .
we define linear operators @xmath109 , @xmath210 , and @xmath694 , @xmath226 , on @xmath696 by setting , for any @xmath699 , @xmath700 @xmath701 note that @xmath702 for any @xmath587 .
[ prop - rep ] formulas ( [ rep - x ] ) and ( [ rep - g ] ) define a representation of @xmath703 on @xmath696 .
we shall check that the defining relations of the algebra @xmath4 are satisfied by the operators @xmath109 , @xmath210 , and @xmath694 , @xmath226 , on @xmath696 .
first note that the commuting relation @xmath704 for any @xmath705 is obviously satisfied .
note also that the operators @xmath225 , @xmath226 , are invertible since @xmath706 for any @xmath587 .
we organize the remaining verifications in three steps . to save space ,
we denote during the proof @xmath707 for any @xmath699 and any @xmath226 .
recall the subalgebras @xmath266 , @xmath268 , of @xmath4 defined in section [ subsec - gl ] which are isomorphic to @xmath708 for some @xmath403 .
we consider first a defining relation of @xmath4 involving only generators lying inside one of these subalgebras @xmath266 .
. then formulas ( [ rep - x])-([rep - g ] ) only involve the contents of the standard tableaux @xmath591 inside @xmath478 and moreover , they coincide with the known formulas for the representations of @xmath709 given in @xcite .
we conclude that we already know that such defining relations are satisfied on @xmath696 .
in particular , this takes care of the relations @xmath710 for @xmath210 , and @xmath711 for @xmath192 such that @xmath353 .
therefore we do not repeat the calculations necessary to check these relations ( instead , we refer to @xcite ; see also @xcite for similar calculations ) .
now we will check the defining relations involving only generators @xmath109 , @xmath210 . due to step * 1 * , it is enough to check the ones which are not already relations inside one of the subalgebras @xmath266 , @xmath268 . in particular , only some defining relations of the form
@xmath712 remain to be checked .
we are going to check all defining relations of the form @xmath712 .
let @xmath192 such that @xmath351 , and let @xmath699 .
a direct calculation gives : @xmath713 & \displaystyle \hspace{-1.5cm}+\ \ \frac{q\,{\mathrm{c}}_{i}^{{\boldsymbol{t}}}-q^{-1}{\mathrm{c}}_{i-1}^{{\boldsymbol{t}}}}{{\mathrm{c}}_{i}^{{\boldsymbol{t}}}-{\mathrm{c}}_{i-1}^{{\boldsymbol{t}}}}\ , \biggl(\,\frac{(q - q^{-1}){\mathrm{c}}_{j}^{r_i({\boldsymbol{t}})}}{{\mathrm{c}}_{j}^{r_i({\boldsymbol{t}})}-{\mathrm{c}}_{j-1}^{r_i({\boldsymbol{t}})}}\cdot v_{{\overline{r}}_i({\boldsymbol{t}})}\ + \ \frac{q\,{\mathrm{c}}_{j}^{r_i({\boldsymbol{t}})}-q^{-1}{\mathrm{c}}_{j-1}^{r_i({\boldsymbol{t}})}}{{\mathrm{c}}_{j}^{r_i({\boldsymbol{t}})}-{\mathrm{c}}_{j-1}^{r_i({\boldsymbol{t}})}}\cdot v_{{\overline{r}}_j{\overline{r}}_{i}({\boldsymbol{t}})}\,\biggr)\ .
\end{array}\ ] ] note that this equation is always valid since , first , if @xmath597 then @xmath714 as well and the second line above is in this case equal to @xmath157 ; and second , thanks to lemma [ lem - mon ] , we have @xmath715 since @xmath351 ( in particular , the denominators are different from @xmath157 ) . using ( [ rep - eq2 ] ) in the above formula for @xmath716 , we can replace each @xmath717 by @xmath718 and each @xmath719 by @xmath720 .
moreover , from proposition [ prop - mon ] , we have @xmath721 . by exchanging @xmath22 and @xmath196 ,
it is now immediate that @xmath722 .
it remains to check the defining relation of @xmath4 involving both generators @xmath109 and @xmath225 .
they are given in general by relations ( [ rel - lu ] ) .
due to step * 1 * , it would be enough to check the ones which are not relations inside one of the subalgebras @xmath266 , @xmath268 .
nevertheless , we will check them in a uniform way by proving that the following relation is satisfied on @xmath696 , for @xmath210 and @xmath226 , @xmath723 first , recall that , by definition of the action of the weyl group on sequences ( section [ sec - act ] ) , we have @xmath724 for any @xmath699 . then recall that @xmath725 for any @xmath210 and thus that @xmath726 .
let @xmath699 .
we apply the right hand side of ( [ rep - eq1 ] ) on @xmath727 and find : @xmath728 on the other hand , we have @xmath729 while , @xmath730 we have @xmath731 and then , putting together the last three calculations , we check easily that relation ( [ rep - eq1 ] ) applied on @xmath727 is satisfied .
we use the ordered basis @xmath732 of the free @xmath5-module @xmath6 to identify the calibrated spectrum @xmath733 with the set @xmath734 , as explained in section [ subsec - prel3 ] .
we recall that @xmath735 and that an element of @xmath734 is a sequence of eigenvalues for these generators of @xmath4
. therefore an element of @xmath734 is naturally a sequence @xmath324 such that there is a calibrated representation @xmath126 of @xmath4 and a non - zero vector @xmath133 satisfying @xmath736 [ theo - spec ] the calibrated spectrum of @xmath4 is in bijection with the set @xmath737 of admissible standard tableaux of type @xmath15 .
more precisely , we have @xmath738 * 1 . *
we first show that @xmath739 .
let @xmath324 with substrings @xmath577 and assume that @xmath740 .
let @xmath126 be a calibrated @xmath703-module and @xmath741 a non - zero vector such that ( [ eq - proof1 ] ) is satisfied .
let @xmath264 and consider the subalgebra @xmath742 , which is isomorphic to @xmath743 , where @xmath744 , @xmath745 and @xmath746 ; see section [ subsec - gl ] .
we consider @xmath126 as a module for @xmath742 .
the module @xmath126 is calibrated as a @xmath742-module .
we now use theorem [ theo - gl ] , where the calibrated spectrum of @xmath742 is described .
equation ( [ eq - proof1 ] ) restricted to generators @xmath225 in @xmath266 involves only the substring @xmath747 of @xmath347 .
it follows immediately from the inclusion of @xmath266 in @xmath4 and theorem [ theo - gl ] that there is a standard tableau @xmath748 such that @xmath749 .
we thus obtained a triplet @xmath549 satisfying the conditions of definition [ def - tab - r ] , due to the intersections of the subalgebras @xmath266 .
moreover , @xmath750 , @xmath268 . we conclude that we have a standard tableau @xmath562 of type @xmath15 such that @xmath565 .
we prove now that @xmath478 is admissible .
to do so , we shall show that , for @xmath210 such that @xmath362 , we have @xmath751 .
indeed , due to the first part of the proof this will imply that @xmath384 is the sequence of contents of a standard tableau of type @xmath15 . in turn
, this implies that the truncated orbit of @xmath478 is admissible ( and so is @xmath478 ) .
so , let @xmath210 such that @xmath362 . for brevity , we set during the proof @xmath752 and @xmath753 for @xmath226 .
let @xmath754 be the following vector of @xmath126 : @xmath755 first , we have @xmath756 .
indeed , by the assumption @xmath362 , we have @xmath757 and therefore @xmath758 .
so @xmath759 would contradict the relation @xmath760 .
now , in order to show that @xmath384 is indeed in @xmath734 , we are going to prove that : @xmath761 let @xmath226 .
first , recall that , by definition of the action of @xmath41 on sequences ( section [ sec - act ] ) , we have @xmath762 , and also that @xmath726 since @xmath725 .
we calculate , using the defining relation between @xmath109 and @xmath225 , @xmath763 & = { \mathrm{c}}_j^{r_i({\boldsymbol{s}})}g_i(v_{{\boldsymbol{s } } } ) - ( q - q^{-1})\displaystyle\frac{({\mathrm{c}}_j^{r_i({\boldsymbol{s}})}-{\mathrm{c}}_j^{{\boldsymbol{s}}}){\mathrm{c}}_i^{{\boldsymbol{s}}}}{{\mathrm{c}}_i^{{\boldsymbol{s}}}-{\mathrm{c}}_{i-1}^{{\boldsymbol{s}}}}v_{{\boldsymbol{s } } } \\[1em ] & = { \mathrm{c}}_j^{r_i({\boldsymbol{s}})}v'_{{\boldsymbol{s } } } + ( q - q^{-1})\displaystyle\frac{{\mathrm{c}}_j^{{\boldsymbol{s}}}{\mathrm{c}}_i^{{\boldsymbol{s}}}}{{\mathrm{c}}_i^{{\boldsymbol{s}}}-{\mathrm{c}}_{i-1}^{{\boldsymbol{s}}}}v_{{\boldsymbol{s}}}\ .
\end{array}\ ] ] combining this calculation with formula ( [ eq - proof1a ] ) for @xmath754 , we obtain the desired formula ( [ eq - proof1b ] ) .
the first two steps show that @xmath764 .
the reverse inclusion is an immediate consequence of the construction of representations @xmath696 , for admissible truncated orbits @xmath695 of type @xmath15 , in section [ subsec - rep ] .
indeed , for any @xmath765 , the defining formula ( [ rep - x ] ) for the action of the generators @xmath225 shows that @xmath766 is calibrated and that @xmath767 . the algebra @xmath9 was defined in section [ subsec - quot ] .
we recall that we have a central element @xmath11 in @xmath4 : @xmath768 where @xmath7 is an element of the @xmath5-module @xmath6 invariant under the action of the weyl group @xmath41 .
the powers @xmath309 are the coefficients of the expansion of @xmath7 in the basis @xmath203 .
they are given in ( [ form - delta ] ) .
then the algebra @xmath9 is the quotient of @xmath4 over the relation @xmath10 .
we showed that in fact @xmath9 is an affine hecke algebra associated to the sublattice @xmath290 of the weight lattice .
[ delta - inv ] * ( i ) * let @xmath324 .
we define : @xmath769 * ( ii ) * we define a subset of admissible standard tableaux of type @xmath15 : @xmath770 as we show in the proposition below , it follows directly from the definitions that @xmath771 is actually an invariant for the action of the weyl group @xmath41 on sequences .
[ prop - delta - inv ] let @xmath324 .
we have @xmath772 for any @xmath72 .
in particular , the value @xmath771 is constant along the truncated orbits .
recall from section [ subsec - trunc ] that to @xmath324 corresponds a character @xmath773 of the abelian group @xmath6 which is identified with the group formed by elements @xmath93 , @xmath222 .
then , the definition of @xmath771 becomes @xmath774 from the definition of the action of @xmath41 on sequences , we have , for @xmath72 , @xmath775 the result then follows from the fact that @xmath7 is invariant under the action of @xmath41 .
let @xmath695 be a truncated orbit in @xmath330 .
the proposition allows to define : @xmath776 [ prop - rep - quot ] * ( i ) * let @xmath695 be an admissible truncated orbit of type @xmath15
. then @xmath696 passes through the quotient and becomes a representation of @xmath9 if and only if @xmath777 .
* ( ii ) * the calibrated spectrum of @xmath9 is in bijection with @xmath778 .
let @xmath695 be an admissible truncated orbit of type @xmath15 . from the definitions and formula ( [ rep - x ] ) giving the action of the generators @xmath225 on @xmath696
, it follows immediately that @xmath779 so the central element @xmath11 acts on @xmath696 by multiplication by @xmath780 .
item * ( i ) * follows .
item * ( ii ) * is a direct consequence of the general result stated in proposition [ prop - c - spec ] .
[ ex - c - gl2 ] we consider again the example of the affine hecke algebra @xmath240 ; see section [ subsec - gl ] .
we recall from section [ subsec - ex ] that in this situation , the admissible truncated orbits are parametrised by placed skew shapes of size @xmath542 and the notion of standard tableaux is the classical one .
moreover , we have seen in example [ ex - c - gl ] that we have @xmath781 .
therefore , in this case , @xmath782 is simply the product of the contents of the placed nodes of a standard tableau @xmath478 .
obviously here , @xmath782 only depends on the shape of @xmath478 .
the quotient algebra @xmath783 is the affine hecke algebra associated to the weight lattice , that is , the affine hecke algebra associated to @xmath321 .
propositions [ prop - rep ] and [ prop - rep - quot ] then provide a construction of representations of @xmath783 associated to placed skew shapes @xmath442 of size @xmath542 such that the product of the contents of the placed nodes of @xmath442 is equal to 1.@xmath66 we recall that standard tableaux of type @xmath15 are characterized by their sequence of contents .
let @xmath695 be an admissible truncated orbit in @xmath330 and let @xmath699 .
we define the following elements of @xmath703 : @xmath784 we have , on the representation @xmath696 , that : @xmath785 this follows from formulas ( [ rep - x ] ) defining the action of the generators @xmath225 on @xmath696 , together with the fact that for any two different @xmath786 , there is some @xmath226 such that @xmath787 . as operators on @xmath696 ,
the set @xmath788 is a complete set of pairwise orthogonal idempotents of @xmath789 .
[ prop - rep2 ] the representations @xmath696 , where @xmath695 runs over the set of admissible truncated orbits of type @xmath15 , are pairwise non - isomorphic irreducible representations of @xmath703 .
let @xmath695 and @xmath790 be two distinct admissible truncated orbits of type @xmath15 .
. from formulas ( [ rep - x ] ) , we have first that the vector @xmath727 in @xmath696 is a common eigenvector for the operators @xmath225 with eigenvalues given by @xmath486 , and second that , in @xmath791 , there is no common eigenvector of the operators @xmath225 with the same eigenvalues . indeed , as @xmath695 and @xmath790 are disjoint , we have that @xmath792 for every @xmath793 , and in turn , @xmath794 for every @xmath793 . we conclude that @xmath696 and @xmath791 are not isomorphic .
let @xmath695 an admissible truncated orbits of type @xmath15 and @xmath795 an invariant subspace of @xmath696 for the action of @xmath703 .
let @xmath796 be a non - zero vector of @xmath147 .
it is a linear combination of the basis vectors @xmath727 , @xmath699 , of @xmath696 with at least one non - zero coefficient . by applying the idempotents @xmath797 on @xmath796 and using that @xmath147 is stable by @xmath703
, we obtain that @xmath798 for some @xmath699 .
now let @xmath210 such that @xmath799 . recall that it means @xmath800 . from formula ( [ rep - g ] ) , we obtain that the coefficient in front of @xmath801 in @xmath802 is different from @xmath157 , and in turn that @xmath803 . as all elements of the truncated orbit @xmath695 can be obtained from @xmath478 by repeated applications of truncated operators
@xmath328 , @xmath210 , we conclude that @xmath804 for any @xmath805 .
this shows that @xmath806 and concludes the proof of the proposition .
let @xmath126 be a representation of @xmath703 .
we denote by @xmath807 the restriction of @xmath126 to the subalgebra @xmath808 . in particular , for @xmath695 an admissible truncated orbit of type @xmath15 , the representation @xmath809 is the restriction of the representation @xmath696 constructed in the preceding section .
the action of @xmath68 on @xmath809 is given explicitly by formulas ( [ rep - g ] ) .
in general , the representation @xmath809 is not irreducible .
we are going to formulate a sufficient condition on @xmath695 so that @xmath809 is an irreducible representation of @xmath808 and then interpret it combinatorially .
[ rem - pl ] a particular situation is when in all elements of @xmath695 , a single place @xmath648 appears ( this will be in particular satisfied for truncated orbits of level 1 to be defined below ) .
then the representation @xmath809 is given purely in terms of moving boxes in skew diagrams , and the value of the place is not relevant . indeed , in formulas ( [ rep - g ] ) giving the action of the generators @xmath109 , we can see that @xmath648 disappears by simplification @xmath66 let @xmath695 be an admissible truncated orbit of level 1 and set @xmath811 for @xmath699 .
the representation @xmath696 is an irreducible representation of @xmath703 and on @xmath696 , the generator @xmath812 acts as @xmath813 by assumption on @xmath695 .
let @xmath814 denote the quotient of @xmath703 by the relation @xmath815 .
then the representation @xmath696 factors through this quotient and becomes an irreducible representation of @xmath814 .
let @xmath816 be the canonical surjective map from @xmath703 to its quotient @xmath814 and consider the following diagram : @xmath817 where we denoted @xmath818 .
considering the restriction to @xmath808 , the representation @xmath809 factors through @xmath819 , which is a subalgebra of @xmath814 .
we claim that we have in fact @xmath820 .
indeed this follows immediately from @xmath821 and the relations @xmath822 for @xmath210 showing that @xmath823 for any @xmath226 .
therefore @xmath809 is irreducible for @xmath819 ( since @xmath696 is irreducible for @xmath814 ) and in turn , is irreducible for @xmath808 .
[ rem - res ] * ( i ) * as we just saw in the proof , the quotient @xmath814 of the affine hecke algebra @xmath4 by a relation of the form @xmath824 is in fact a quotient of the finite hecke algebra @xmath68 .
this is implied by the first set of defining relations ( [ rel - h - gx1 ] ) between the generators @xmath109 and @xmath225 .
the other defining relations ( [ rel - h - gx2])([rel - h - gx5 ] ) may imply other relations between the generators @xmath109 in @xmath814 .
so , in general , @xmath814 is a non - trivial quotient of @xmath68 .
this is related to the fact that , in general , the irreducible representations obtained in proposition [ prop - fin ] do not exhaust the set of irreducible representations of @xmath808 .
note however , that one could ask how much of the set of irreducible representations of @xmath808 is obtained by fixing @xmath2 and varying @xmath3 . *
( ii ) * the terminology level 1 " in definition [ def - lev ] refers to the fact that @xmath812 has only one eigenvalue in @xmath696 . admissible truncated orbits of higher levels would lead to the study of quotients of @xmath4 by characteristic equations for @xmath812 of higher degrees ( in some contexts , quotients of this sort are called cyclotomic quotients ) . *
( iii ) * a particular situation is when @xmath4 is the affine hecke algebra @xmath240 . in this case
, it is known that we have @xmath825 ( equivalently , all irreducible representations of @xmath808 are obtained via proposition [ prop - fin ] ) .
moreover , the quotients of higher levels in this case are well - known : they are the so - called cyclotomic hecke algebras , or ariki koike algebras @xcite . *
( iv ) * finally we note that the converse of proposition is not true .
it is already seen in the @xmath826 situation , where the representation associated to the skew diagram @xmath827 \box & \hspace{-0.45cm}\box \end{array}$ ] , restricted to @xmath828 , is an irreducible representation , while @xmath812 has two different eigenvalues .
@xmath66 [ [ combinatorial - characterisation - of - admissible - truncated - orbits - of - level-1 .
] ] * combinatorial characterisation of admissible truncated orbits of level 1 . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we will from now on make the following slight abuse of terminology .
a skew partition @xmath423 is a
usual partition " when @xmath424 is empty .
we will say that a placed skew shape is a usual partition if it is the equivalence class of a placed skew diagram of the form @xmath829 , where @xmath458 is the diagram of a usual partition ( in particular , there is a single place ; see remark [ rem - pl ] ) .
let @xmath830 be a skew diagram and @xmath412 a node of @xmath830 .
we call @xmath412 a _ top left node _ of @xmath830 if the positions above @xmath412 in the same column and the positions to the left of @xmath412 in the same line are all empty
. a placed node @xmath831 of a placed skew diagram is called a _
top left placed node _ if @xmath412 is a top left node of its diagram .
finally , we note that the notion of being a top left placed node is compatible with the equivalence relation of placed skew diagram , and therefore is well - defined for a placed skew shape .
let @xmath442 be a placed skew diagram with a unique top left placed node .
it means that there is only one place : @xmath832 , and moreover the skew diagram @xmath458 has only one top left node . using the equivalence relation
, we can translate @xmath458 such that the top left node moves to position @xmath430 ( the place @xmath648 may have to change ; see example ( [ ex - eq ] ) ) .
we thus see that @xmath442 is equivalent to a placed skew diagram of the form @xmath833 , where @xmath457 is the diagram of a usual partition .
[ prop - lev2 ] let @xmath695 be an admissible truncated orbit of type @xmath15 .
then @xmath695 is of level 1 if and only if for every @xmath834 , the shapes of @xmath835 and @xmath552 are usual partitions .
assume first that for every @xmath834 , the shapes of @xmath578 are usual partitions .
thus , for every @xmath834 , there is a unique top left placed node in the shapes of @xmath578 . by standardness of @xmath578 , the only possibility for @xmath157 is to be contained in this placed node
this shows that @xmath659 for every @xmath699 ( otherwise , in @xmath836 , @xmath157 would be contained in a placed node which is not top left ) . as @xmath837 is the only simple reflection in @xmath41 which modifies the position of @xmath157 , we conclude that @xmath838 remains constant along @xmath695 .
assume that we have @xmath834 such that , for some @xmath264 , the shape @xmath839 of @xmath591 is not a usual partition .
it means there are at least two top left placed nodes in @xmath839 .
so let @xmath619 be the one containing @xmath157 in @xmath591 and let @xmath620 be a different one .
it is clear that @xmath619 and @xmath620 can not lie in the same diagonal of the same diagram , and therefore @xmath840 .
let @xmath841 be a standard tableau of shape @xmath839 such that @xmath157 is contained in @xmath620 ( such a standard tableau exists as @xmath620 is a top left placed node ) .
let @xmath842 be the size of @xmath839 ( @xmath270 is @xmath271 , @xmath272 or @xmath273 depending on @xmath843 ) .
we have seen in proposition [ prop - tab - a2 ] that @xmath841 can be obtained from @xmath591 by repeated applications of truncated operators of the symmetric group @xmath844 .
so we have @xmath845 for some @xmath846 .
the symmetric group @xmath844 is a subgroup of @xmath41 , and so , considering @xmath74 as an element of @xmath41 , we set @xmath847 . from the definition of truncated operators , we have that @xmath848 since we already had @xmath849 .
so we obtained @xmath805 such that @xmath850 . from the fact that @xmath851
, we conclude that @xmath695 is not of level @xmath172 .
[ ex - res ] we take @xmath852 and @xmath243 in the standard labelling , as in section [ subsec - ex]*(b)*. we show an example of an admissible truncated orbits of level 1 : @xmath853 & & & \end{array } , \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{${\underline{3}}$ } } & & \end{array } \right ) \ \ \stackrel{\ \ { \overline{r}}_{3_{}}\ \,}{\longleftrightarrow}\ \ \left ( \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } \\[-0.2em ] \fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \end{array } , \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{3}}$ } } \\[-0.3em ] \fbox{\scriptsize{$2 $ } } & & \end{array } \right ) \ \ \stackrel{\ \ { \overline{r}}_{2_{}}\ \,}{\longleftrightarrow}\ \ \left ( \begin{array}{cc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } \\[-0.2em ] \fbox{\scriptsize{$1 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$3 $ } } \end{array } , \begin{array}{ccc } \fbox{\scriptsize{$0 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{$2 $ } } & \hspace{-0.35cm}\fbox{\scriptsize{${\underline{3}}$ } } \\[-0.3em ] \fbox{\scriptsize{$1 $ } } & & \end{array } \right)\ ] ] the non - indicated actions of @xmath854 give @xmath340 .
this truncated orbit gives rise to an irreducible representation of @xmath855 .
@xmath66 let @xmath478 be a tableau of type @xmath15 and , for @xmath210 , let @xmath856 be the placed node of @xmath478 with number @xmath22 . for @xmath192 , we write @xmath857 if @xmath858 and @xmath859 otherwise .
we recall that tableaux are only considered up to equivalence and we note that the values of the @xmath860 s are therefore not uniquely defined .
nevertheless , the property @xmath857 ( or its negation ) is well - defined since compatible with the equivalence relation .
let @xmath695 be an admissible truncated orbit of type @xmath15 . in this subsection
, we will make the following assumption .
if @xmath648 and @xmath861 are two different places which appear in an element @xmath699 then we assume that : @xmath862 where , for @xmath863 , @xmath864 denotes the evaluation of @xmath228 at @xmath77 . let @xmath126 be a representation of the finite hecke algebra @xmath808 and let @xmath158 be a basis of @xmath126
let @xmath865 be the matrix representing the generator @xmath109 , @xmath210 , in the basis @xmath158 .
let @xmath866 and assume that all coefficients of @xmath865 are non - singular when evaluated at @xmath867 , and this for all @xmath210 . then , sending the simple reflection @xmath209 , @xmath210 , to the evaluation of the matrix @xmath865 at @xmath867 , we obtain a complex representation of the weyl group @xmath41 on a complex vector space with basis @xmath158 .
in this situation , we say that the evaluation at @xmath867 of the representation @xmath126 in the basis @xmath158 exists .
[ prop - clas ]
let @xmath695 be an admissible truncated orbit of type @xmath15 satisfying ( [ hyp - pl ] ) and let @xmath866 .
the evaluation at @xmath867 of the representation @xmath809 in the basis @xmath697 exists , and the action of @xmath41 is given by the proof is a straightforward verification , consisting in the evaluation at @xmath867 of the coefficients appearing in formulas ( [ rep - g ] ) .
we only recall that @xmath872 if @xmath22 is contained in a placed node with place @xmath860 and that we use condition ( [ hyp - pl ] ) on the places .
let @xmath874 be the representation of the preceding proposition obtained when @xmath67 is evaluated at @xmath172 , and @xmath875 the one obtained at @xmath876 . from formulas ( [ rep - r1])-([rep - r2 ] )
, we see that : @xmath877 where @xmath878 denotes the signature representation of @xmath41 .
[ rem - hyp - pl ] * ( i ) * a remark similar to remark [ rem - pl ] applies here : when in all elements of @xmath695 , a single place @xmath648 appears , the representation @xmath874 is given purely in terms of moving boxes in skew diagrams , and the value of the place @xmath648 is not relevant . outside of this situation , even though the values of the places do not appear in formulas ( [ rep - r1])-([rep - r2 ] )
, these values determine the structure of @xmath695 ; see examples in section [ subsec - ex ] .
* ( ii ) * condition [ hyp - pl ] on @xmath695 was assumed for simplicity because it allowed us to perform the evaluation at @xmath77 and to obtain formulas ( [ rep - r1])-([rep - r2 ] ) as classical limits " of representations of @xmath68 .
therefore it followed immediately that we obtained representations of @xmath41 .
alternatively , one could define formulas ( [ rep - r1])-([rep - r2 ] ) for any truncated admissible orbit @xmath695 , and then check directly that the defining relations of @xmath41 are indeed satisfied , without any reference to @xmath68 .
@xmath66 m. geck and g. pfeiffer , _ characters of finite coxeter groups and iwahori
hecke algebras_. london mathematical society monographs .
new series , 21 . the clarendon press , oxford university press , new york ( 2000 ) .
a. okounkov and a. vershik , _ a new approach to representation theory of symmetric groups _ , selecta math .
( new series ) 2(4 ) ( 1996 ) 581605 .
_ a new approach to representation theory of symmetric groups ii _ , j. math .
( n. y. ) 131(2 ) ( 2005 ) 54715494 .
|
we introduce and study some affine hecke algebras of type ade , generalising the affine hecke algebras of gl .
we construct irreducible calibrated representations and describe the calibrated spectrum .
this is done in terms of new families of combinatorial objects equipped with actions of the corresponding weyl groups .
these objects are built from and generalise the usual standard young tableaux , and are controlled by the considered affine hecke algebras . by restriction and limiting procedure
, we obtain several combinatorial models for representations of finite hecke algebras and weyl groups of type ade .
representations are constructed by explicit formulas , in a seminormal form .
|
the theory of superconductivity , as first developed by bardeen , cooper , and schrieffer,@xcite and then generalized to strong coupling by migdal,@xcite eliashberg,@xcite , and others@xcite , has proven to be one of the most accurate theories of solid - state physics .
properties of conventional low-@xmath0 materials are typically explained to accuracies of one percent or better .
newly discovered materials , however , which have moderate @xmath0 s , do not fit into the parameter regimes studied so successfully in the 50 s and 60 s .
these materials , such as the a15 compounds , ba@xmath1k@xmath2bio@xmath3 , and k@xmath3c@xmath4 have large phonon energy scales relative to the inverse electronic density of states , so that both the effect of the energy dependence of the bare electronic density of states ( i. e. , nonconstant density of states ) , and the effect of vertex corrections may become important in their description .
we examine here a series of different weak - coupling perturbation theories that include both the effects of nonconstant density of states and of vertex corrections to ascertain what methods should be used for real materials calculations of these higher @xmath0 compounds . since the most popular implementation of migdal - eliashberg theory is manifestly particle - hole symmetric ( because of the neglect of the energy dependence of the electronic density of states ) it is not clear what the most accurate approximation scheme is when the electron filling is doped away from half - filling and the energy dependence of the electronic density of states becomes important .
our strategy is to solve a model system ( the holstein model ) in the infinite - dimensional limit via the dynamical mean field theory .
this allows us to compare numerically exact quantum monte carlo solutions ( in the thermodynamic limit ) with the different perturbative approximations .
the holstein model@xcite consists of conduction electrons that interact with local ( einstein ) phonons : @xmath5 in eq .
( [ eq : holham ] ) , @xmath6 ( @xmath7 ) creates ( destroys ) an electron at site @xmath8 with spin @xmath9 , @xmath10 is the electron number operator , and @xmath11 ( @xmath12 ) is the phonon coordinate ( momentum ) at site @xmath8 .
the hopping is chosen to be between the nearest neighbors @xmath13 and @xmath14 of a hypercubic lattice in @xmath15-dimensions and the unit of energy is the rescaled matrix element @xmath16 ( so that the kinetic energy remains finite as @xmath17 )
. the phonon has a mass @xmath18 ( chosen to be @xmath19 ) , a frequency @xmath20 , and a spring constant @xmath21 associated with it .
the deformation potential is denoted by @xmath22 and it governs the strength of the coupling of electrons to phonons .
the effective electron - electron interaction strength ( mediated by the phonons ) is then the bipolaron binding energy @xmath23 a chemical potential @xmath24 is employed to adjust the total electron filling with @xmath25 corresponding to half - filling in the exact solution . in the instantaneous limit where @xmath26 remains finite and @xmath22 and @xmath20 are large compared to the bandwidth @xmath27 , the holstein model maps onto the attractive hubbard model@xcite @xmath28 with @xmath26 defined by eq .
( [ eq : udef ] ) .
two cases of the holstein hamiltonian have well - established perturbative expansions . in the limit where the phonon frequency becomes small @xmath29 , but @xmath26 remains finite ,
migdal - eliashberg theory@xcite is known to be an accurate approximation .
migdal - eliashberg theory is a self - consistent hartree - fock approximation that employs fully dressed phonon propagators and neglects all vertex corrections .
typically , the energy dependence of the electronic density of states is also neglected , so the theory is evaluated with a constant density of states .
( this latter approximation is always particle - hole symmetric and maps onto the limit of half - filling when the nonconstant density of states is used . )
the second limit is the large phonon frequency limit @xmath30 with @xmath26 also remaining finite . in this case , a truncated perturbative expansion through second order@xcite ( which includes the lowest - order vertex correction ) is known to be accurate at half - filling for a large range of @xmath26 values , because it properly reproduces both the weak - coupling and strong - coupling limits of the hubbard model
. it would be nice to construct an approximation scheme that continuously connects these two limits as the phonon frequency is varied .
however , no simple approximation can be found , because the set of diagrams that dress the phonon propagator in the small-@xmath20 limit do not correctly dress the phonon propagator in the large-@xmath20 limit ; in the large - frequency limit the interaction is only between up and down spins , but in the small - frequency limit the interaction is between all spins .
but , if one is willing to examine perturbative expansions that are truncated with respect to the fluctuating diagrams , then approximations can be constructed that agree with the two known limits through the order of diagrams included in the expansion .
the description of these different methods is both subtle and technical , and will be covered in detail in section ii .
here we only want to comment that the previous calculations for the electron - phonon problem that were called the iterated perturbation theory@xcite ( ipt ) are actually based on a truncated perturbation theory about the hartree mean - field theory solution , which does not reproduce the migdal - eliashberg ( hartree - fock ) limit properly .
a more promising approach for all @xmath20 is to construct a truncated perturbation theory about the hartree - fock mean - field theory solution , which is done here . in addition
, we also examine some simple methods that can be used to repair inconsistencies that develop in the ipt as the system is doped off of half - filling .
this method entails self - consistently renormalizing the higher - order fluctuations of the hartree or fock diagrams to all orders , which allows for the electron filling on the lattice to be different from the electron filling of the mean - field theory solution when the higher - order fluctuations are included .
unfortunately , we find that all attempts to produce an accurate perturbation theory that reproduces the ipt in the instantaneous limit are not very accurate for moderate phonon frequencies . these perturbative approaches , and the many - body problem in general ,
simplify when the infinite - dimensional limit@xcite is taken @xmath31 .
then the many - body problem becomes a local problem that retains its complicated dynamics in time .
the hopping integral is scaled to zero in such a fashion that the free - electron kinetic energy remains finite while the self energy for the single - particle green s function and the irreducible vertex functions have no momentum dependence and are functionals of the local green s function@xcite .
this limit retains the strong - correlation effects that arise from trying to simultaneously minimize both the kinetic energy and the potential energy , and hence has relevance for three - dimensional materials .
of course , we can also solve the infinite - dimensional holstein model using a quantum monte carlo method , which contains all effects due to phonon renormalizations , vertex corrections , and nonconstant electronic density of states .
we use these solutions as a benchmark to test the accuracy of the different approximation methods and to determine what is the most fruitful approximation for the electron - phonon problem .
we employ a green s function formalism to solve the many - body problem .
the local green s function is defined to be @xmath32 and is calculated directly from a self - consistent quantum monte carlo procedure described elsewhere.@xcite static two - particle properties can also be determined since the irreducible vertex function is local@xcite . the static susceptibility for charge - density - wave ( cdw ) order is given by @xmath33,\cr & = & t\sum_{mn } \tilde\chi_{mn}^{cdw } ( { \bf q } ) , \label{eq : chicdw}\end{aligned}\ ] ] at each ordering wavevector @xmath34 [ the indices @xmath35 and
@xmath36 denote matsubara frequencies @xmath37 .
dyson s equation for the two - particle green s function becomes@xcite @xmath38 with @xmath39 the irreducible vertex function in the cdw channel .
the bare cdw susceptibility @xmath40 in eq .
( [ eq : cdwdys ] ) is defined in terms of the _ dressed _ single - particle green s function @xmath41 with all of the wavevector dependence described by the scalar@xcite @xmath42 . the integral for @xmath43 in eq .
( [ eq : chi0cdw ] ) can then be performed analytically@xcite for the `` checkerboard '' cdw phase @xmath44 .
the irreducible vertex function @xmath39 is either directly calculated in a perturbative expansion ( described below ) or is determined by inverting the dyson equation in eq .
( [ eq : cdwdys ] ) for the local charge susceptibility ( qmc ) .
a similar procedure is used for the singlet @xmath45-wave superconducting ( sc ) channel .
the corresponding definitions are as follows : the static susceptibility in the superconducting channel is defined to be @xmath46 for superconducting pairs that carry momentum @xmath34 ; dyson s equation becomes @xmath47 with @xmath48 the corresponding irreducible vertex function to describe the sc channel ; the bare pair - field susceptibility becomes @xmath49 with the special value @xmath50 for the sc pair that carries no net momentum ; and finally the irreducible vertex function is also either directly calculated in a perturbative expansion ( described below ) or is determined by inverting the dyson equation in eq .
( [ eq : scdys ] ) for the local pair - field susceptibility ( qmc ) .
the transition temperature of the infinite - dimensional holstein model is then found by calculating the temperature at which the relevant susceptibility diverges ( cdw or sc ) .
this transition temperature is found by locating the temperature where the scattering matrix ( in the relevant channel ) @xmath51 has unit eigenvalue@xcite ( note that the local green s functions are always used in the evaluation of the bare susceptibility @xmath52 ) .
the remainder of the paper is arranged as follows : in section ii we describe the formalism employed to derive the different truncated perturbative expansions for both the electronic self energy and for the irreducible vertex functions .
particular attention is paid to the mapping from the lattice problem onto an effective impurity problem , and how to extract a perturbative expansion that includes the lowest - order fluctuations beyond a mean - field theory solution . in section iii
we describe the details of the calculational procedure and present our results in section iv .
section v contains a summary and our conclusions .
the properties of the holstein model are calculated by mapping the lattice problem onto the single - impurity wolff - holstein model@xcite , and solving the impurity self energy by a truncated perturbation expansion . to define the mapping , we begin with the self - energy of the holstein model , which is momentum independent in infinite dimensions , and write the local green s function for the lattice as , @xmath53 where @xmath54 is the noninteracting electronic bandstructure for the lattice ( assuming spin degeneracy @xmath55 ) .
the local self - energy @xmath56 is defined by the functional , @xmath57 $ ] , where @xmath58 $ ] represents the sum of all the skeleton diagrams generated by the perturbation theory , with u as the expansion parameter ( in a baym - kadanoff expansion@xcite ) . on the other hand ,
the same set of skeleton diagrams appears in the self - energy of an impurity problem described by the wolff - holstein hamiltonian , @xmath59 where @xmath60 describes a band of non - interacting electrons , @xmath61 is the number operator for conduction electrons of spin @xmath9 at the impurity site and @xmath62 ( @xmath63 ) is the `` impurity '' phonon coordinate ( momentum ) .
the renormalized electron propagator at the impurity site can be written as , @xmath64 here @xmath65 is the free - electron propagator at the impurity site ( which is often called the effective - medium propagator ) , and @xmath56 describes the effects due to the coupling to the local phonon .
note that the effective - medium @xmath65 is not equal to the noninteracting lattice propagator except when @xmath66 .
since the impurity and the lattice self - energy functionals , @xmath67 $ ] and @xmath58 $ ] are the same , the self energies will also be the same if @xmath68 .
thus , the lattice problem maps onto the impurity problem , provided the effective medium propagator @xmath65 is adjusted so that the right hand sides of eqs .
( [ g_loc ] ) and ( [ g_imp ] ) are equal . in general , the self - energy functional @xmath58 $ ] is not known . but equivalent expansions can also be made for the impurity self - energy by rearranging the skeleton - diagram expansion .
for example , a functional defined on the effective - medium propagator @xmath69 can be employed , such that @xmath70 $ ] , where @xmath71 $ ] represents the sum of all connected graphs generated by wick s theorem ( without any resummations ) .
if the graphs for @xmath58 $ ] and @xmath71 $ ] are summed to all orders , then the self energies must agree . furthermore , the effective medium @xmath69 can be uniquely determined by either @xmath72 $ ] or @xmath73 $ ] .
since the exact self - energy for the wolff - holstein model with an arbitrary density of states is not known , our strategy is to make a truncated expansion for the impurity self - energy functional @xmath74 $ ] , and define the self - energy for the lattice problem via @xmath75 $ ] , the self consistency is then imposed on @xmath69 , so that @xmath76 .
the hope is , that a controlled expansion of the impurity self - energy will lead to an accurate solution of the impurity problem , and of the resultant lattice problem . here
, we use the lowest - order yosida - yamada expansion , which is known to provide reliable answers for the wolff model describing magnetic impurities@xcite .
the yosida - yamada expansion for the effective impurity problem ( with an unknown density of states ) is obtained by a partial resummation of diagrams that allows us to construct the self - energy functional in terms of diagrams that involve either hartree or hartree - fock green s functions .
that is , one considers the self - energy corrections with respect to the mean - field solution , and rewrites the dyson equation as , @xmath77^{-1 } - \sigma_{yy}^{mf}[g_{mf } ] } .\ ] ] where @xmath78 is the sum of all self - energy diagrams defined in terms of the mean - field propagator for the impurity problem , @xmath79 , and @xmath80 it is clear that functionals @xmath81 $ ] defined with different mean - field propagators are not the same , and it is not known _ a priori _ which truncated approximation for the impurity self energy leads to the most accurate result for the lattice problem .
the mean - field solution has an average electron filling @xmath82 defined by @xmath83 with @xmath84 $ ] the fermi factor ( the factor of 2 is from the spin summation ) .
this filling is usually different from the electron filling on the lattice which is determined by the same equation , but with @xmath79 replaced by @xmath85 . in this contribution
, we consider four different expansions to the self energy and compare the corresponding results to the quantum monte carlo solution .
our first approximation is called the hartree expansion ( h ) and it includes a truncated perturbation - theory expansion about the hartree mean - field solution , including all diagrams through second order .
this expansion should not be confused with the hartree approximation , which does not include any of the second - order vertex corrections .
these self - energy corrections with respect to the hartree mean - field solution are constructed by using the hartree self - energy diagram shown in figure 1 , to construct the mean - field propagator of eq .
( [ g_yy ] ) .
that is , we take @xmath86 which implies that the hartree self - energy insertions are included to all orders in the perturbative expansion . here
@xmath87 is the ( hartree ) mean - field particle number defined by eq .
( [ n_mf ] ) with @xmath88 .
next , the hartree mean - field green s function is used in evaluating the truncated yosida - yamada self - energy functional to determine the approximate self - energy .
this functional is given in figure 1 through second - order in @xmath26 where the electronic propagators ( solid lines ) are the hartree mean - field propagators from eq .
( [ g_mf ] ) , the phonon propagators ( wiggly lines ) are bare @xmath89 $ ] , and the impurity self - energy which renormalizes the effective field @xmath65 in eqn.([g_imp ] ) becomes @xmath90.\ ] ] note that this procedure is summing an infinite class of diagrams ( the hartree self - energy insertions ) , but is truncating the perturbation theory _ with respect to the hartree mean - field - theory solution _ to include both the fock and the second - order fluctuating terms .
it turns out that this procedure is identical in form to what was previously called the ipt approximation@xcite , since the renormalization of the effective medium by the hartree self energy ( to construct the hartree mean - field green s function ) can be absorbed by a redefinition of the chemical potential ( since the hartree self energy just provides a frequency - independent shift ) , with the exception being the inclusion of the fifth diagram , which renormalizes the hartree diagram by the fock self - energy insertion .
the omission of the fifth diagram in the older ipt paper@xcite may be thought to be innocuous , because that diagram vanishes at half - filling and should not affect the results much off of half filling .
however , when the coupling strength becomes large , it s effects do become strong , as shown in the next section .
to reiterate , the difference between the present expansion and the older ipt work is that the fluctuating diagrams in @xmath91 $ ] are evaluated with @xmath92 which does not include the frequency - independent shift from the fifth diagram of figure 1 .
the ipt calculation included the frequency - independent shifts to all orders , and hence evaluated the first four diagrams of figure 1 using a different mean - field propagator than @xmath92 .
the inclusion of the fock self - energy insertion into the hartree diagram is just one of the subtle , and often neglected diagrams that needs to be included in a truncated approximation that includes all diagrams up to a given order .
our second approximation is called the hartree - fock expansion ( hf ) and is obtained by using the hartree - fock self - energy @xmath93 to define the mean - field propagator in eqn.([g_mf ] ) , and including all second - order diagrams with respect to the hartree - fock mean - field solution ( once again , this should not be confused with the hartree - fock approximation which does not include any of the second - order vertex corrections ) .
the ( hartree - fock ) mean - field particle number @xmath94 is calculated by using @xmath95 in eqn.([n_mf ] ) , and the fock self energy @xmath96 satisfies @xmath97 with @xmath98 the bare phonon propagator again .
this mean - field solution sums both the hartree and fock self - energy insertions to all orders , and would be identical to migdal - eliashberg theory if the dressed phonon propagator was employed in eq .
( [ sigma_f ] ) rather than the bare propagator .
as described in the introduction , we are forced to use the bare propagator if we want to reproduce the ipt limit in the large - phonon - frequency limit .
the yosida - yamada functional defined on @xmath95 is given by the diagrams in figure 2 where the solid lines are now hartree - fock propagators .
the self - energy correction to @xmath65 for the hf expansion can then be written as @xmath99.\ ] ] the self - energy diagrams given in figs .
1 and 2 are evaluated on the imaginary axis by using the standard rules \{explicit expressions are given in eq .
( 16 ) of ref . [ and the suitable modifications for the hartree - fock case ] } , and the matsubara summations occurring in @xmath100 $ ] and @xmath101 $ ] are performed numerically for both the hartree and the hartree - fock expansions .
if the perturbative expansion for the self - energy is extended to higher - order in u we find additional frequency - independent diagrams [ which sum up to @xmath102 in addition to the frequency - dependent diagrams ( similar to the fifth diagram in figure 1 ) . here
@xmath36 is the fully renormalized particle number of the lattice which is calculated by eq .
( [ n_mf ] ) with @xmath85 replacing @xmath79 .
these diagrams arise simply from the fact that in the exact skeleton expansion the hartree diagram is evaluated with the fully dressed green s function ( yielding @xmath103 ) rather than with the mean - field green s function ( which yields @xmath104 . at half - filling ,
the two fillings @xmath36 and @xmath82 are usually equal , but they need not be the same ( and never are ) away from half filling . we introduce a new self - energy functional @xmath105 $ ] ( that is based on the yosida - yamada functionals defined above ) that has all of the higher - order renormalizations of the hartree diagram removed from it [ as shown , through third - order in figure 3 ( a ) ]
. then we have an exact relation @xmath106 $ ] for the self energy @xmath56 defined in eq .
( [ g_imp ] ) . in this contribution , the two self - energy functionals @xmath107 and @xmath108 differ by one diagram ( the fifth diagram in figure 1 ) .
we can employ this exact relation to formulate the renormalized - hartree expansions@xcite ( also called the n - consistent approximation ) , which enforces this self - consistency condition on the renormalized particle number .
that is , for a given @xmath24 , g , @xmath109 and @xmath110 , we solve the transcendental equation , @xmath111 to determine the impurity electron filling @xmath36 , and to obtain the correction to the real part of the total self energy . for the anderson and wolff impurity models , and for the two - dimensional hubbard model , this renormalized - hartree approach ( rh )
expands the region of the validity of the perturbation theory , and allows for the description of stronger correlations@xcite .
the hope is that this renormalized - hartree scheme will repair some of the inaccuracies of the truncated perturbation theories away from half filling .
it is a much simpler approximation to study than a scheme that tries to interpolate between the weak - coupling and atomic limits@xcite ( which is significantly more challenging to formulate for the electron - phonon problem ) .
a similar approach can be made for summing higher - order diagrams in the hartree - fock expansion .
in this case , however , an n - consistent approximation , that sums only the frequency - independent diagrams to all orders , is likely to be less accurate than one that sums the renormalizations based on both the hartree and fock diagrams .
hence , we form the exact relation for the self energy of eq .
( [ g_imp ] ) @xmath112,\ ] ] and employ it to evaluate the renormalized - hartree - fock expansion ( rhf ) where the functional @xmath113 is truncated at second order [ see figure 3(b ) ] .
this approximation is formed by evaluating both the hartree and fock diagrams with the local green s function ( instead of @xmath95 ) but the second - order diagrams continue to be evaluated with @xmath95 . in this sense
it is `` halfway '' between a conserving approximation ( which evaluates all diagrams with @xmath85 ) and the truncated expansions described above .
the only remaining objects that need to be determined are the irreducible vertices for the cdw and sc instabilities .
these vertices are calculated with the mean - field green s function ( either hartree or hartree - fock ) and have an identical functional form for all different approximation schemes .
the diagrams included are presented in figure 4(a ) for the cdw vertex and figure 4(b ) for the sc vertex .
the solid lines are @xmath79 ( @xmath92 or @xmath95 ) and the wiggly lines are the bare phonon propagator .
explicit expressions for these diagrams on the imaginary axis are given in eqs .
( 17 ) and ( 18 ) of ref . with @xmath65
replaced by @xmath79 .
the calculational methods used are straightforward .
the perturbation expansion is carried out on the imaginary axis .
we employ an iterational algorithm to solve the self - consistent equations , which is summarized in figure 5 : 1 .
start with an initial self - energy @xmath56 ; 2 . use eqn.([g_loc ] ) to calculate @xmath85 ; 3 . calculate the effective - medium propagator @xmath65 using eqn.([g_imp ] ) with @xmath114 ; 4 . calculate the mean - field propagator using eqn.([g_mf ] ) ; 5 . calculate the yosida - yamada self - energy diagrams as defined by fig . 1 or fig . 2 , with and without the renormalization of the first - order diagrams ( as shown in fig .
3 ) , to determine the total self - energy ; 6 .
if the calculation for the self - energy has not converged , then adjust the chemical potential to produce the target electron filling ; 7 .
repeat steps 2 - 6 until the calculation has converged ; 8 . once converged , calculate the irreducible vertex functions from fig . 4 and determine the maximal eigenvalue of the scattering matrix in both the cdw and sc channels from eq .
( [ eq : tmat ] ) ; 9 .
repeat 1 - 8 for another temperature @xmath115 until the transition temperature @xmath0 is found .
we use an energy cutoff of 256 matsubara frequencies , which provides accurate results for temperatures @xmath115 larger than @xmath116 , and our convergence criterion is that the green s functions do not change by more than one part in @xmath117 from one iteration to the next . we find that the perturbation theory typically converges in approximately 100 iterations of the main ring in figure 5 . sometimes the iterated equations develop limit cycles , whose oscillations can be suppressed by employing standard damping methods that average the @xmath118st and @xmath119th iterations to produce the starting point for the next iteration . we also found that if the coupling strength is large enough , then some of the approximate theories will have multiple solutions for @xmath120 near half filling . in this case , the symmetry point , with @xmath25 becomes unstable ( i. e. , the compressibility is negative ) , and the green s functions are no longer purely imaginary when the electron filling is half - filled .
this latter result is an indication of the breakdown of the approximation method at such a large coupling strength .
there also are some technical details that need to be discussed about the quantum monte carlo simulations@xcite .
the results for the transition temperatures were calculated with a @xmath121 , and sometimes also with a @xmath122 and then extrapolated to @xmath123 , when the correlations were large enough that the trotter error was noticeable .
the self - energies and irreducible vertex functions were calculated with a fixed @xmath124 to ensure a high accuracy . in general
, there are three sources of error to the quantum monte carlo simulation : ( i ) statistical error ; ( ii ) iteration error ; and ( iii ) systematic error .
the statistical error is the easiest error to control , and is the smallest of the three sources of error .
the iteration error arises from performing calculations with the wrong dynamical mean - field @xmath65 , because the calculation has not yet converged fully .
we typically iterate 8 or 9 times , which provides good convergence for the iterations .
the systematic error is more difficult to control .
it arises from the trotter error , and from other ( potentially unknown ) sources .
one surprising source of error arose from the choice of global updating moves for the phonon coordinate .
global moves that shift the phonon coordinate at every time slice by an amount randomly chosen between @xmath125 and @xmath126 were supplemented by global shifts by an amount ranging between @xmath127 and and @xmath128 .
these latter moves were chosen to allow the phonon coordinate to shift between the two minima of the effective phonon potential ( separated by @xmath129 $ ] ) , which form when the system enters the strong - coupling region , were preformed pairs exist above @xmath0 .
when one is in the weak - coupling regime , where the effective phonon potential possesses a single - well structure , we find that the self - energy can differ by a few percent depending on whether the global moves are all chosen randomly , or if the supplemental global shifts are chosen to move the phonon coordinate to the `` other '' well ( i.e. if the phonon coordinates lie in the right well then the global shift is chosen to move the system to the left , and vice versa ) .
the accuracy was best when the global moves were chosen completely randomly , which is the method used here .
these results differ by a few percent from those shown previously@xcite , where the global moves were coupled to the values of the phonon coordinates , and the trotter time slice was larger .
the parameters of the hamiltonian in eq .
( [ eq : holham ] ) are chosen to agree with previous theoretical work , and to represent a parameter range where the vertex corrections can cause large effects .
as such , we choose the phonon frequency to be on the order of one tenth of the bandwidth @xmath130 ( with the effective bandwidth of the gaussian being about two standard deviations above and below the center , or approximately @xmath131 ) .
most of our calculations are performed for three different values of the interaction strength ( i ) @xmath132 ( @xmath133 , one sixth of the bandwidth ) which corresponds to a fairly weak interaction where there are no preformed pairs , but the transition temperatures are large enough that the phase diagram can be determined reliably ; ( ii ) @xmath134 ( @xmath135 , one fourth of the bandwidth ) which is in the moderate interaction regime ; and ( iii ) @xmath136 ( @xmath137 , two - fifths of the bandwidth ) which is where the system enters the strongly coupled regime , where numerous preformed pairs are present above @xmath0 .
the quasiparticle renormalization factor , the irreducible cdw vertex function , and the irreducible sc vertex function are plotted in figs .
6(a - c ) .
the vertex functions are averaged to show just the symmetric frequency component , since that is all that contributes to the eigenvalue of the scattering matrix for the cdw at half filling , or for the sc at any filling ( this is so is because the eigenvector with maximal eigenvalue is symmetric with respect to a change in sign of the matsubara frequency ) .
the coupling strength is @xmath132 and the filling is @xmath138 , which yields a bare electron - phonon coupling strength of @xmath139 , or @xmath140 , which would naively be viewed as quite weak coupling .
however , this is not the case , as one can estimate the renormalized value of @xmath141 from @xmath142 which lies at about 0.7 .
these calculations indicate that higher - order diagrams , such as those that fully dress the phonon propagator , are important even when the bare coupling has @xmath143 .
what is interesting , is that the perturbative results seem to have the right shape , but need to have the frequency axis moved to the right , to line up with the monte carlo data .
furthermore , since the perturbation theory underestimates both the renormalization factor ( which will enhance @xmath0 ) and the vertex function ( which will reduce @xmath0 ) one might expect these results to tend to cancel each other out when calculating a phase diagram , which is indeed what we find below .
7(a - c ) show the same results for the stronger coupling case @xmath134 and @xmath138 ( @xmath144 and @xmath145 ) .
we can see that the error in the renormalization factor has grown , even though all four perturbative methods yield similar values .
the cdw vertex becomes more attractive at moderate coupling , which is missed by the perturbation theory .
the sc vertex has the right qualitative shape , but is underestimated at low frequencies . when the coupling strength is increased to @xmath136 ( not shown ) the situation becomes even worse , with the same qualitative features seen , a growing difference in the renormalization factors , and an underestimation of the size of the vertices .
we also include , in fig . 8 , a result at quarter filling @xmath146 and @xmath132 at the same temperature @xmath147 as in fig . 6 .
one might have thought that the quarter - filled case would be approximated much better by the perturbation theory because the effective coupling strength is lower due to the reduction of the electronic density of states at the fermi level , but we see that the shape of the curves is both qualitatively and quantitatively similar for figs . 6 and 8
. the quantitative agreement has improved , but the improvement is not dramatic .
what is surprising is that the calculated transition temperature agrees much better with the qmc here than at half filling , because of the cancellation of the errors to both the self energy and the vertex function .
note that we only plot the sc vertex here , because an asymmetric average is needed for the physically relevant piece of the cdw vertex .
in addition to calculating the single - particle and two - particle self energies , we can also investigate the phase diagrams of the holstein model .
based on our results for the self energy and the vertex functions , it does not appear likely that any approximation will be too accurate for the phase diagrams , but as theorized above , it is possible that the inaccuracies will cancel out to produce more accurate results for the transition temperatures . such behavior should be viewed as `` accidental '' agreement with the exact result .
we begin by calculating the transition temperature to the commensurate ( checkerboard ) cdw insulator at half - filling as a function of the interaction strength .
the qmc solution@xcite showed that the maximal @xmath0 was on the order of 1/25th of the bandwidth , occurring when @xmath26 was about two - fifths of the bandwidth .
the results for each of the four expansions shown here is presented in figure 9 .
the horizontal axis is the interaction strength and the vertical axis is the transition temperature .
one can see that the peak in @xmath0 versus @xmath26 is not produced by the perturbation theory , rather the transition temperature continues to increase .
this differs from what happened in the so - called ipt approximation where the approximate curve for @xmath0 did show a peak.@xcite the hartree and renormalized - hartree expansions agree with each other and the ipt approximation until @xmath26 is approximately @xmath148 , where they differ from each other because the thermodynamically stable h and rh expansions are no longer at the particle - hole symmetric point with @xmath25 , but rather have different values of @xmath24 ( and no longer have purely imaginary green s functions when evaluated along the imaginary axis ) .
this is the point where the kink appears in the @xmath149 curve .
the remarkable agreement of the ipt approximation in predicting both the peak position and the peak height accurately is really just a coincidence , since neither the self energy nor the vertex function is reproduced well at that value of the interaction strength ( see figures 68 ) .
there is a very delicate balance between the self - energy terms ( which tend to reduce @xmath0 as the interaction increases ) and the vertex - function terms ( which tend to increase @xmath0 as the interaction increases ) that causes @xmath0 to have a peak for the ipt approximation , but continue to increase for the approximations in this contribution .
we end with a calculation of the phase diagrams off of half - filling , which show the transition from the checkerboard cdw insulator to a sc .
these results are plotted in figure 10 .
the horizontal axis is the electron filling and the vertical axis is the transition temperature to either a cdw or sc phase .
the cdw - sc phase boundary lies at the points where the phase diagrams have a slope discontinuity .
three values of interaction strength are included , and both the approximate solutions and the qmc results are presented ( the lines through the qmc data are just a guide to the eye ) . the agreement for @xmath0 is quite good for the weak - coupling value @xmath132 , with all approximations going through the qmc data for the sc phase , and overestimating @xmath0 only slightly for the cdw phase . here , the hartree - fock expansion is actually worse than the hartree expansion for the cdw phase .
the calculation also seems to underestimate the location of the cdw - sc phase boundary .
the case with @xmath134 begins to show how the solutions develop qualitatively wrong behavior near half filling .
all but the hartree - fock expansion show a suppression of @xmath0 as the filling approaches @xmath150 .
such a suppression was never seen in the qmc simulations , and is likely an artifact of the non - conserving nature of the approximations@xcite .
it continues to be clear that the sc solutions are approximated better than the cdw solutions , and that the approximations behave worst near half - filling . in particular , it is important to note that the simple n - consistent scheme , which worked so well for the hubbard model , does not appear to work as well for the electron - phonon problem examined here , because the phase diagrams ( h and rh ) continue to have qualitatively incorrect behavior .
finally we examine the strongly coupled case @xmath136 . here
the renormalized expansions do a better job at repairing the qualitatively wrong behavior of a @xmath0 suppression near half - filling , but they do so by greatly enhancing the @xmath0 s at and near half - filling from the truncated expansions .
it is clear , that at this large a value of the coupling strength , none of these weak - coupling approaches is doing a good job at predicting the phase diagrams .
it is interesting to note that the close agreement of the hartree and renormalized hartree calculations shows that the majority of the shift of the chemical potential comes from the fock dressing of the hartree diagram .
surprisingly , the inclusion of this diagram has a large effect on the phase diagram when @xmath151 .
what are the conclusions that can be drawn from this work ? even at a weak value of the bare interaction strength ( @xmath132 ) both the self energy and the irreducible vertex functions are not approximated accurately by the perturbation theory .
this implies that third - order ( and higher - order ) diagrams are important and can not be neglected .
for example , it is possible that a scheme that uses an expansion with dressed phonon propagators would be more accurate , which is work in progress .
nevertheless , the perturbation theory does appear to produce a more accurate approximation to the phase diagrams than the individual self - energies .
on the other hand , these results show that there is no simple scheme that will allow one to approximate the electron - phonon problem accurately for all values of the phonon frequency .
so one should not necessarily rely on approximations that employ purely electronic models ( such as the hubbard model ) to carry over to the electron - phonon problem and work well when the phonon frequency is small to moderate . instead
, it is more fruitful to work on generalizations of the migdal - eliashberg theory that work with dressed phonons , but include higher - order nonadiabatic effects such as vertex corrections .
this should save a lot of time in formulating accurate approximation methods for real materials that have large enough phonon energy scales that the effects of nonconstant density of states and of vertex corrections can not be neglected . working with dressed phonons
actually makes a lot of sense for the electron - phonon problem , because experiments can directly measure the dressed phonon spectral function , so it is readily available for use in real materials calculations .
furthermore , the dressed phonon propagator can be extracted from the qmc simulation and used in the perturbation theory for the electrons , just as is done from experiments on real materials .
work along these lines is currently in progress .
we would like to thank paul miller and joe serene for useful discussions .
j.k.f . and w.c .
acknowledge support for the initial stages of this work from the office of naval research young investigator program n000149610828 and j.k.f . and
acknowledge support for the final stages of this work from the national science foundation int-9722782 and dmr-9627778 .
m.j . acknowledges support from the national science foundation dmr-9704021 and dmr-9357199 , and from the ohio supercomputer center .
p. b. allen and r. c. dynes , phys .
b * 12 * 905 ( 1975 ) ; p. b. allen and b. mitrovi ' c , solid state phys .
* 37 * , 1 ( 1982 ) ; j. p. carbotte , rev . mod . phys . * 62 * , 1027 ( 1990 ) ; h .- s .
y weng , g. ji , and z .- f .
zhou , j. phys .
solids * 48 * , 395 ( 1987 ) .
j. e. hirsch and r. m. fye , phys .
* 56 * , 2521 ( 1986),m .
jarrell , phys .
lett . * 69 * , 168 ( 1992 ) ; m. jarrell and th .
pruschke , z. phys .
b * 90 * , 187 ( 1993 ) ; th .
pruschke , d. l. cox , and m. jarrell , phys .
b * 47 * , 3553 ( 1993 ) ; j. k. freericks , m. jarrell , and d. j. scalapino , phys . rev .
b * 48 * , 6302 ( 1993 ) ; europhys . lett . *
25 * , 37 ( 1994 ) ; j. k. freericks and m. jarrell , in _
computer simulations in condensed matter physics _ , vol .
vii , edited by d. p. landau , k. k. mon , and h .- b .
schttler ( springer - verlag , heidelberg , berlin , 1994 ) .
k. yosida and k. yamada , prog .
* 46 * , 244 ( 1970 ) ; k. yamada , prog .
phys . * 53 * , 970 ( 1975 ) ; k. yosida and k. yamada , prog .
phys . * 53 * , 1286 ( 1975 ) ; k. yamada , prog .
theor . phys . *
54 * , 316 ( 1975 ) ; prog .
. phys . * 55 * , 1345 ( 1976 ) ; prog .
theor . phys . * 62 * , 354 ( 1979 ) .
|
a series of weak - coupling perturbation theories which include the lowest - order vertex corrections are applied to the attractive holstein model in infinite dimensions .
the approximations are chosen to reproduce the iterated perturbation theory in the limit of half - filling and large phonon frequency ( where the holstein model maps onto the hubbard model ) .
comparison is made with quantum monte carlo solutions to test the accuracy of different approximation schemes .
|
facilitating reuse on the semantic web requires support for effectively finding and exploring existing semantic resources such as ontologies or taxonomies . particularly in the biomedical domain , where ontologies are highly specialized , costly and often created collaboratively by large groups of domain experts ,
_ identifying and finding existing semantic vocabularies and terminologies _ is crucial . to support researchers and practitioners in this task ,
the national center for biomedical ontology ( ncbo ) has developed bioportal_the world s most comprehensive online repository of biomedical ontologies _ @xcite . [ cols= " < , < , < " , ] [ tab : cluster counts ]
in this paper we have analyzed , modeled and clustered the browsing behavior of users on bioportal .
in particular , we have grouped users according to their browsing behavior in section [ subsec : res : clusters ] and compared how users explore ontologies in section [ subsec : res : browsing_comparison ] . * differences in browsing - behavior types . *
we have shown that we were able to obtain multiple , clearly distinguishable browsing - behavior types , when using stationary distribution as input for @xmath2-means .
further , we have calculated the explained variance per cluster to determine the number of browsing - behavior types to investigate .
the manual inspection of the obtained clusters shows that users primarily concentrate on browsing specific classes ( _ class explorers _ and _ specific class browsers _ ) of biomedical ontologies on bioportal .
two strategies of users to explore ontologies involve the exploitation of the ontological hierarchy ( _ ontology tree explorers _ ; if available ) and the search functionality provided by bioportal ( _ search explorers _ ) .
users in the third - largest cluster ( @xmath3)bioportal experts use advanced functionality , such as the annotator or the recommender , to find classes and explore ontologies .
the comparison of our results with clusters obtained from static page views indicate that the inclusion of information about the dynamic nature of a user s interaction behavior with a website is important . in general , our results indicate that users exhibit a preference to explore and browse the content of the classes of biomedical ontologies .
however , this also indicates that additional efforts are warranted to promote and further enhance and evaluate the utility of the functionality provided by bioportal to explore and find ontologies .
for example , only a small portion of users ( the bioportal experts ) use the mappings between ontologies . *
differences when exploring ontologies . *
our results indicate that several characteristics of ontologies , such as the hierarchical structure ( or lack thereof ) , can influence how users explore and interact with these semantic vocabularies and terminologies on bioportal .
one particular example of such an ontology is rxnorm , which does not exhibit a structured hierarchy , and is thus not displayed in the class explorer on bioportal .
the only ways to interact with classes of this ontology on the website is to either use the search functionality , which requires prior knowledge about the content of the ontology , or to rely on mappings from other ontologies .
however , this also explains the strong focus of users / visitors of rxnorm to use the search functionality ( _ browse search _ ) to explore , investigate and find classes of this ontology on bioportal ( see figure [ fig : bp_browsing_comp : rel ] ) .
in contrast , the structural hierarchy of cpt is used by bioportal to visualize the class tree . hence , to explore the ontology
, users can ( and do)aside from searching ( _ browse search _ ) for specific classes use and explore the hierarchy ( _ browse ontology class tree _ ) to find a specific class .
note that we have only presented the differences in browsing - behavior types for two of more than @xmath0 ontologies ! yet
, we have shown that simple differences ( i.e. , the lack of a class hierarchy ) between projects cause users of bioportal to employ different strategies to find classes or explore the ontological content .
in this paper , we have presented an approach for identifying different types of user behavior in ontology exploration log data using stationary distributions from first order markov chains .
we provide evidence for a total of @xmath4 distinct browsing - behavior types on bioportal and offer an interpretation of their relevance and meaning .
in addition , we cluster ontologies by their user behavior and identify pragmatic commonalities among ontology projects .
our results advance our understanding of the ways in which ontological repositories , such as bioportal , are used and may guide the development of more user - oriented systems for ontology exploration and reuse on the web . for future work , we are particularly interested in further elaborating the comparison between the browsing behavior of different ontologies on bioportal . in doing that , we might be able to identify commonalities for groups of ontologies that trigger differences in the way users interact with ontologies on bioportal .
understanding the different ways in which semantic structures may influence ontology exploration behavior has the potential to help in the design of bioportal as well as in the development of more effective and more reusable ontologies .
this work was supported by grant gm086587 from the u.s
. national institute of general medical sciences ( nigms ) of the national institutes of health .
the protg project is supported by nigms grant gm103316 .
ncbo is funded by national institutes of health grant u54 hg004028 .
l. espn noboa , f. lemmerich , p. singer , and m. strohmaier .
discovering and characterizing mobility patterns in urban spaces : a study of manhattan taxi data . in _ proceedings of the 25th international conference companion on world wide web _ ,
pages 537542 , 2016 . c. lakshminarayan , r. kosuru , and m. hsu .
modeling complex clickstream data by stochastic models : theory and methods . in
_ proceedings of the 25th international conference companion on world wide web _ , pages 879884 , 2016 .
j. lehmann , c. mller - birn , d. laniado , m. lalmas , and a. kaltenbrunner .
reader preferences and behavior on wikipedia . in _ proceedings of the 25th acm conference on hypertext and social media _ , pages 8897 .
acm , 2014 .
n. f. noy , n. h. shah , p. l. whetzel , b. dai , m. dorf , n. griffith , c. jonquet , d. l. rubin , m .- a .
storey , c. g. chute , et al .
bioportal : ontologies and integrated data resources at the click of a mouse .
, page gkp440 , 2009 .
p. singer , d. helic , a. hotho , and m. strohmaier .
hyptrails : a bayesian approach for comparing hypotheses about human trails on the web . in _ proceedings of the 24th international conference on world wide web _ , www 15 , pages 10031013 , 2015 .
b. smith , m. ashburner , c. rosse , j. bard , w. bug , w. ceusters , l. j. goldberg , k. eilbeck , a. ireland , c. j. mungall , et al . the obo foundry : coordinated evolution of ontologies to support biomedical data integration .
, 25(11):12511255 , 2007 .
m. strohmaier , s. walk , j. pschko , d. lamprecht , t. tudorache , c. nyulas , m. a. musen , and n. f. noy .
how ontologies are made : studying the hidden social dynamics behind collaborative ontology engineering projects . , 20(0 ) , 2013 .
s. walk , p. singer , l. e. noboa , t. tudorache , m. a. musen , and m. strohmaier . understanding how users edit ontologies : comparing hypotheses about four real - world projects . in _ the semantic web - iswc 2015 - 14th international semantic web conference _ , pages 551568 , 2015 .
s. walk , p. singer , and m. strohmaier .
sequential action patterns in collaborative ontology - engineering projects : a case - study in the biomedical domain . in _ proceedings of the 23rd
acm international conference on conference on information and knowledge management , cikm 2014 , shanghai , china , november 3 - 7 , 2014 _ , pages 13491358 , 2014 .
s. walk , p. singer , m. strohmaier , d. helic , n. f. noy , and m. a. musen . how to apply markov chains for modeling sequential edit patterns in collaborative ontology - engineering projects .
, 84:51 66 , 2015 .
h. wang , t. tudorache , d. dou , n. f. noy , and m. a. musen .
analysis of user editing patterns in ontology development projects . in _ on the move to meaningful internet systems : otm 2013 conferences _ , pages 470487 .
springer , 2013 .
l. whetzel , n. f. noy , n. h. shah , p. r. alexander , c. nyulas , t. tudorache , and m. a. musen .
bioportal : enhanced functionality via new web services from the national center for biomedical ontology to access and use ontologies in software applications . , 39(suppl 2):w541w545 , 2011 .
|
ontologies in the biomedical domain are numerous , highly specialized and very expensive to develop .
thus , a crucial prerequisite for ontology adoption and reuse is effective support for exploring and finding _ existing _ ontologies . towards that goal ,
the national center for biomedical ontology ( ncbo ) has developed bioportal an online repository designed to support users in exploring and finding more than @xmath0 existing biomedical ontologies . in @xmath1
, bioportal represents one of the largest portals for exploration of semantic biomedical vocabularies and terminologies , which is used by many researchers and practitioners .
while usage of this portal is high , we know very little about how exactly users search and explore ontologies and what kind of usage patterns or user groups exist in the first place .
deeper insights into user behavior on such portals can provide valuable information to devise strategies for a better support of users in exploring and finding existing ontologies , and thereby enable better ontology reuse . to that end , we study and group users according to their browsing behavior on bioportal using data mining techniques .
additionally , we use the obtained groups to characterize and compare exploration strategies across ontologies . in particular , we were able to identify seven distinct browsing - behavior types , which all make use of different functionality provided by bioportal .
for example , _ search explorers _ make extensive use of the search functionality while _ ontology tree explorers _ mainly rely on the class hierarchy to explore ontologies .
further , we show that specific characteristics of ontologies influence the way users explore and interact with the website .
our results may guide the development of more user - oriented systems for ontology exploration on the web .
< ccs2012 > < concept > < concept_id>10002951.10003260.10003277.10003280</concept_id > < concept_desc > information systems web log analysis</concept_desc > < concept_significance>500</concept_significance >
< /concept > < concept > < concept_id>10002951.10003260.10003277.10003281</concept_id > < concept_desc > information systems traffic analysis</concept_desc > < concept_significance>500</concept_significance >
< /concept > < concept > < concept_id>10010147.10010341.10010346.10010348</concept_id > < concept_desc > computing methodologies network science</concept_desc > < concept_significance>300</concept_significance > < /concept >
< /ccs2012 >
|
on a phenomenological level , the most successful alternative to cosmic dark matter is the modified newtonian dynamics ( mond ) proposed by milgrom ( 1983a ) .
the basic idea is that the deviation from newtonian gravity or dynamics occurs below a fixed acceleration scale a proposal which is supported in a general way by the fact that discrepancy between the classical dynamical mass and the observable mass in astronomical systems does seem to appear at accelerations below @xmath5 cm / s@xmath6 ( sanders 1990 , mcgaugh 1998 ) .
the fact that this acceleration scale is comparable to the present value of the hubble parameter multiplied by the speed of light ( @xmath7 ) suggests a cosmological basis for this phenomenology .
mond , in a sense , is designed to reproduce flat extended rotation curves of spiral galaxies and a luminosity - rotation velocity relationship of the observed form , @xmath8 ( the tully - fisher relation ) .
but apart from these aspects which are built - in " , the prescription also successfully predicts the observed form of galaxy rotation curves from the observed distribution of stars and gas with reasonable values for the mass - to - light ratio of the stellar component ( begeman et al .
1991 , sanders 1996 , sanders & verheijen 1998 ) .
moreover , mond predicts that the discrepancy between the newtonian dynamical mass and the observed mass should be large in low surface brightness galaxies a prediction subsequently borne out by observations of these systems ( mcgaugh & de block , 1997 , 1998 ) .
the observational success of mond is most dramatically evident for spiral galaxies where often the rotation curve can be a rather precise tracer of the radial distribution of the effective gravitational force ; for hot stellar systems elliptical galaxies the predictions are less precise .
milgrom , in his original papers , ( 1983b ) pointed out that mond implies a mass - velocity dispersion relation for elliptical galaxies of the form @xmath0 .
if there were no systematic variation of m / l with mass , this , would become the observed faber - jackson relation ( faber & jackson 1976 ) . in his seminal paper on this subject , milgrom ( 1984 )
calculated the structure of isothermal spheres in the context of mond and drew several very general conclusions : first , all isothermal spheres , regardless of the degree of anisotropy in the velocity distribution , have finite mass . for a one - dimensional velocity dispersion of 100 to 200 km / s ,
this mass is inevitably on a galaxy scale .
second , at large radial distance the density decreases as @xmath9 where @xmath10 is in the vicinity of 4 .
third , there is an absolute maximum on the mean surface density which is on the order of @xmath11 where @xmath12 is the mond critical acceleration . for a mass - to - light ratio of three to four in solar units
, this would translate into a surface brightness in the v band of 20 to 20.5 mag/(arcsec)@xmath6 which is characteristic of hot stellar systems ranging from massive ellipticals to bulges of spiral galaxies to globular clusters ( corollo et al .
1997 ) . finally , the mass of an isothermal sphere with a specific anisotropy factor is exactly proportional to @xmath13 where @xmath14 is the space velocity dispersion .
all of these conclusions would apply to elliptical galaxies , in so far as these objects can be regarded as isothermal spheres . since this work
, it has come to light that the global properties of elliptical galaxies comprise a three parameter family ( dressler et al .
1987 , djorovski & davis 1987 ) ; that is to say , elliptical galaxies lie on a surface in a three - dimensional space defined by the luminosity ( l ) , the the central velocity dispersion ( @xmath15 ) , and the effective radius ( @xmath1 ) ; the mean surface brightness ( @xmath16 ) may be substituted for either luminosity or effective radius ( i.e. , @xmath17 ) .
this surface appears as a plane on logarithmic plots and , consequently , has been designated as the fundamental plane " of elliptical galaxies , with the form @xmath18 where @xmath19 and @xmath20 . because of the small scatter perpendicular to the fundamental plane , this three - parameter relationship supersedes the faber - jackson relation as a distance indicator .
the usual physical interpretation of the fundamental plane is that these relations result from the traditional virial theorem plus a dependence of mass - to - light ratio on galaxy mass ( van albada et al .
1995 ) ; although , one must also assume that elliptical galaxies comprise a near - homologous family . it is not immediately clear how the fundamental plane can be interpreted in terms of modified dynamics , or , indeed , if the fundamental plane is even consistent with modified dynamics .
the effective virial theorem for a system deep in the regime of mond ( low internal accelerations ) is of the form @xmath21 , with no length scale appearing . on the face of it
, this would imply that hot stellar systems should comprise a two parameter family as suggested by the older faber - jackson relation .
the purpose of this paper is to consider the dynamics of elliptical galaxies in terms of mond , particularly with respect to the relations between global properties .
i demonstrate that the mond isothermal sphere is actually not a good representation of high surface brightness elliptical galaxies because the implied average surface densities are too low .
elliptical galaxies , within the half - light radius ( the effective radius ) , are essentially newtonian systems with accelerations in excess of the critical mond acceleration .
this suggests that other possible degrees of freedom must be exploited to model elliptical galaxies using modified dynamics . to reproduce the observed properties of high surface - brightness elliptical galaxies ,
it is necessary to introduce small deviations from a strictly isothermal and isotropic velocity field in the outer regions .
a simple and approximate way of doing this is to consider high - order polytropic spheres ( all of which are finite in the context of mond ) with a velocity distribution which varies from isotropic within a critical radius to highly radial motion in the outer regions .
the structure of such objects is determined here by numerical solutions of the hydrostatic equation of stellar dynamics ( the jeans equation ) modified through the introduction of the mond formula for the gravitational acceleration .
i find that all models characterized by a given value of the polytropic index and the appropriately scaled anisotropy radius are homologous and exhibit a perfect mass - velocity dispersion relation of the form @xmath22 with no intrinsic scatter .
however a range of models over this parameter space is required to reproduce the dispersion in the observed global properties of elliptical galaxies , and strict homology is broken .
this adds considerable scatter to the mass - velocity dispersion relation and introduces a third parameter which is , in effect , the surface density ( or effective radius ) .
although these objects are effectively newtonian in the inner regions , mond imposes boundary conditions which restrict these newtonian solutions to a well - defined domain in the three dimensional space of dynamical parameters a dynamical fundamental plane .
combined with a weak dependence of m / l on galaxy mass , the fundamental plane in its observed form is reproduced .
not only the form but also the scaling of the @xmath23 relation is fixed over the relevant domain of parameter space ; this scaling is relatively independent of the detailed structure of the stellar system .
therefore , given the effective radius and velocity dispersion , the mass of any elliptical galaxy may be calculated and the mass - to - light ratio can be determined . for the galaxies in the samples of jrgensen , franx , & krgaard ( 1995a , b ) and jrgensen ( 1999 ) the mean m / l turns out to be 3.6 m@xmath24/l@xmath24 with about 30% scatter . with a weak dependence of m /
l with galaxy mass , the predicted form of the fundamental plane agrees with that found by jrgensen et al .
following milgrom ( 1984 ) , i calculate the structure of spherical systems by integrating the spherically symmetric hydrostatic equation ( the jeans equation ) : @xmath25 where @xmath26 is the density , @xmath27 is the radial component of the velocity dispersion , @xmath28 is the anisotropy parameter ( @xmath29 is the velocity dispersion in the tangential direction ) , and @xmath30 is the radial gravitational force which , in the context of mond , is given by @xmath31 here @xmath32 is the mass within radius r , @xmath12 is the mond acceleration parameter ( found to be @xmath33 cm / s@xmath6 from the rotation curves of nearby galaxies ) , and @xmath34 is the typically assumed function interpolating between the newtonian regime ( @xmath35 ) and the mond regime ( @xmath36 ) . because of the assumed spherical symmetry , eq .
2 is exact in the context of the lagrangian formulation of mond as a modification of newtonian gravity ( bekenstein & milgrom 1984 ) .
however , if viewed as a modification of inertia eq . 2
may only be an approximation in the general case of motion on non - circular orbits ( milgrom 1994 ) .
there are four unknown functions of radius , @xmath26 , @xmath27 , @xmath37 , and @xmath30 ; therefore , additional assumptions are required to close this set of 2 equations .
first , i take a definite pressure - density relation , that of a polytropic equation of state : @xmath38 where @xmath39 , for a particular model , is a constant which is specified by the given central velocity dispersion and density , and @xmath40 is the polytropic index ( for an isothermal sphere @xmath40 is infinite ) .
this is done simply as a convenient way of providing stellar systems which are somewhat cooler in the outer regions as is suggested by observations ( discussed below ) .
the anisotropy parameter is further assumed to depend upon radius as @xmath41 } } \eqno(4)\ ] ] where @xmath42 is the assumed anisotropy radius .
this provides a velocity distribution which is isotropic within @xmath42 , but which approaches pure radial motion when @xmath43 .
such behavior is typical of systems which form by dissipationless collapse ( van albada 1982 ) .
thus i will be considering a two - parameter set of models for elliptical galaxies characterized by a polytropic index @xmath40 and an anisotropy radius @xmath42 .
additional physical considerations , such as stability , may limit the range of these parameters . milgrom ( 1984 )
found that for an isothermal sphere with a given constant @xmath37 , there exits a family of mond solutions having different asymptotic behavior as @xmath44 .
in general , the density approaches a constant value near the center except for one particular limiting solution where @xmath45 .
the global properties ( e.g. , the mean surface density , the value of @xmath46 , the asymptotic density distribution at large r ) do not vary greatly within such a family of solutions , except for models which are characterized by an unrealistically low central density of @xmath47/pc@xmath48 ( see milgrom 1984 , figs . 1 & 2 ) .
i find the same to be true for the high - order polytropic spheres with an anisotropy factor given by eq .
4 ; i.e. , for a given value of @xmath40 and @xmath42 , there is a family of solutions with differing asymptotic behavior at small r. again , because the global parameters do not vary greatly within a family , i consider below only the limiting solution ; i.e. , that with the @xmath49 density cusp . before describing the results of the numerical integration of eqs . 1 - 4 , one important aspect of this system of equations should be emphasized .
the appearance of an additional dimensional constant , @xmath12 , in the equation for the gravitational force , eq . 2 , provides , when combined with a characteristic velocity dispersion of the system ( the central radial velocity dispersion , for example ) , a natural length and mass scale for objects described by these equations .
this differs from the case of newtonian systems where two system parameters ( a velocity dispersion and central density ) are required to define units of mass and density . for mond systems
the characteristic length and mass are given by @xmath50 @xmath51 there is , in addition , a natural scale for surface density given by @xmath52 which depends only upon fundamental constants .
these natural units imply that the properties of homologous systems described by modified dynamics should scale according to these relations ; i.e. , @xmath0 with a characteristic surface density which is independent of @xmath14 . however , for the system to be described by modified dynamics it must extend into the regime of low accelerations , i.e. , where @xmath53 .
this is not a necessary attribute of newtonian systems which have finite mass and radius ( @xmath54 ) .
therefore , we would expect that only those higher order polytropes ( @xmath55 including the isothermal sphere ) to be necessarily characterized by this scaling because the newtonian solution always has infinite extent and mass .
such polytropes would inevitably extend to the regime of low accelerations . for a velocity dispersion of 200 km / s , typical for elliptical galaxies ,
we find @xmath56 kpc and @xmath57 m@xmath24 characteristic galactic dimensions .
the manner in which these length and mass scales arise , in the context of a cosmological setting for the dissipationless formation of elliptical galaxies , will be discussed in future paper .
eq . 1 is numerically solved using a fourth - order runge - kutte technique . the integration proceeds radially outward by specifying a central radial velocity dispersion and choosing , at a particular radius , the density corresponding to that of the limiting solution given by milgrom ( 1984 ) . in solving for the structure of isothermal spheres
, milgrom used the natural units of length and mass ( eq . 5 ) to write eq . 1 in unitless form .
here , i use physical units : 1 kpc , @xmath58 , and 1 km / s ( in these units @xmath59 and @xmath60 .
although convenient for comparison with observations , these units do obscure the scaling of solutions . for comparison with milgrom s
results the first systems considered here are isotropic isothermal spheres ( @xmath40 , @xmath42 @xmath61 ) .
the limiting solution of an isothermal sphere with a specified value of @xmath37 can be scaled as implied by the natural units ( eqs .
i verify this by solving eqs . 1 - 4 for isothermal spheres with radial velocity dispersion ranging from 50 km
/ s to 350 km / s .
for all such spheres , the density at large radius falls off as @xmath62 .
this implies , directly from eq . 1 , that @xmath63 which is to say , the mass - velocity dispersion relation is exact for mond isotropic isothermal spheres . because the density distribution for the limiting solution falls as @xmath49 at small r and as @xmath62 in the outer regions , the density profile of the mond isotropic isothermal sphere resembles that of the jaffe model ( jaffe 1983 ) . the mean surface density inside the projected half - mass radius ( the effective radius ) , @xmath1 , is found to be @xmath64 where the characteristic mond surface density , @xmath65 , is given by eq .
given that @xmath66 , then from eqs .
6 and 7 it follows that @xmath67 for a system with @xmath68 km / s , one would find , with these formulae , that m = @xmath69 m@xmath24 and @xmath70 kpc .
the surface density distribution of such an object is well - described by the de vaucouleurs @xmath71 law which works well as an empirical fit to surface brightness profile of elliptical galaxies
. however , the global properties of mond isotropic isothermal spheres are inconsistent with those of high surface - brightness elliptical galaxies .
basically , for a given velocity dispersion , the mass is too large , the effective radius is too large , and the mean surface density is too small compared to that of actual galaxies .
this is evident from fig . 1 which is a logarithmic plot of the central velocity dispersion plotted against effective radius for a large homogeneously observed sample of galaxies
. the sample consists of 154 elliptical galaxies in several clusters ( open points ) observed by jrgensen et al .
( 1995a , b ) combined with 116 early - type galaxies ( crosses ) in the coma cluster ( jrgensen 1999 ) .
the connected solid points show the sequence of isotropic isothermal spheres ( eq .
we see that the distribution of observed properties is not matched with those of the isothermal spheres ; for example , for galaxies with a central line - of - sight ( l.o.s . )
velocity dispersion of about 150 km / s , the effective radii are typically 2.5 kpc , an order of magnitude lower than that of the mond isothermal spheres . assuming a mass - to - light ratio of 3 to 4 m@xmath24/l@xmath24 for the stellar population , the mean observed surface density in actual elliptical galaxies within @xmath1 would be about two or three times larger than @xmath65 ; which is to say , elliptical galaxies are well within the newtonian regime inside an effective radius .
in contrast the mean surface density of mond isotropic isothermal spheres are at least 10 times lower ( eq . 7 ) and hence fail as representations of luminous high surface brightness elliptical galaxies . the situation becomes worse for constant @xmath72 as is evident from the calculation of milgrom ( 1984 ) .
such systems are even deeper in the mond regime ; for @xmath73 , @xmath74 . in order to represent elliptical galaxies in the context of mond
, we must allow for the possibility that pressure - supported systems become cooler with a radial orbit anisotropy in the outer regions .
there is observational evidence that the stellar component of elliptical galaxies is not , in general , isothermal : the l.o.s .
velocity dispersion is observed to decrease with increasing projected radius . for reasons of stability
such a decrease can probably not be attributed entirely to velocity field anisotropy ( see below ) . describing this decline by a power law
, @xmath75 , franx ( 1989 ) finds that typically @xmath76
. a simple way of introducing this mild deviation from an isothermal state into the mond models is to consider the more general equation of state expressed by eq .
3 the polytropic gas assumption with large @xmath40 .
it is straight - forward to demonstrate from eq . 1 that all mond polytropic spheres of finite @xmath40 are finite in extent as well as in mass ,
unlike the newtonian case where only polytropes with @xmath54 have finite extent and mass . as @xmath77 , the radius of the edge of the sphere also approaches infinity ; for high order polytropes ( @xmath78 ) the outer radius is many times larger than the effective radius .
a polytropic sphere with @xmath55 will always have a mond regime which establishes boundary conditions for the inner newtonian solution .
therefore , the polytropic index , which must lie between 5 and infinity , is a free parameter of such models .
it is unlikely that the stars in elliptical galaxies have a completely isotropic velocity distribution . as noted above , a radial orbit anisotropy of the form given by eq .
4 emerges naturally in dissipationless collapse models .
this expression provides a second dimensionless parameter for characterizing mond models of elliptical galaxies : @xmath79 , the anisotropy radius in terms of the characteristic mond length scale a very general result is that such high order anisotropic polytropic spheres have a velocity dispersion - mass relation of the form , @xmath80 where @xmath15 is the _ central _ l.o.s .
velocity dispersion . that is to say
, the ratio of @xmath81 to @xmath82 , defined here as @xmath83 , depends upon the two dimensionless parameters @xmath40 and @xmath84 .
this follows from the general scaling relation for systems which extend into the regime of modified dynamics , eq .
for the pure isotropic isothermal sphere @xmath85 ( eq . 6 ) . for any given
polytropic index and scaled anisotropy radius , the structure of objects is self - similar over central radial velocity dispersion , and the @xmath86 relation is exact ; i.e. , such objects form a 2-parameter family in @xmath15 and @xmath87 .
the same is not true if we consider the set of such systems over a domain of the two - dimensional parameter space . in fig .
2 the connected solid points show the logarithm of @xmath83 vs. the logarithm of the mean surface density within the effective radius ( @xmath88 ) in units of the mond critical surface density @xmath65 for a sequence of isotropic polytropes ( @xmath89 ) .
each point is a set of models over a range of central @xmath15 but having a specific value of @xmath40 .
the locus of such points define a curve with the isothermal sphere ( denoted by an x ) at one extreme and , as displayed here , the n=7 polytrope at the other .
the sequence of isotropic polytropes approaches but does not exceed the mond critical surface density .
each polytrope is self - similar over velocity dispersion with a perfect mass - velocity dispersion relation , but its own mass - velocity relation .
the entire set of polytropes breaks homology , and for this ensemble of isotropic polytropes , @xmath83 can be considered as a function of the mean surface density ; indeed , over the range of n=30 to n@xmath90 , the curve is well approximated by a power law , i.e. , @xmath91 that is to say , a third parameter , the mean surface density , enters into the basic dynamical relationship , eq .
9 . substituting eq . 10 into eq .
9 we find a relation of the form @xmath92 this can be viewed as a generalized virial relationship for objects which extend into the regime of modified dynamics . for models with a specified value of @xmath40 and @xmath84 , @xmath93 and
we recover the mond mass - velocity dispersion relationship for homologous systems ; models covering a range of n and @xmath84 non - homologous models with @xmath94 comprise a three - parameter family . if structure of elliptical galaxies could be approximated by this set of high order isotropic polytropic spheres ( @xmath95 ) , then , from eq
. 11 , there would exist a theoretical fundamental plane relationship of the form @xmath96 where @xmath97 @xmath98 and @xmath99 for the sequence of isothermal spheres over this range of @xmath40 , @xmath100 which , from eqs .
12b and 12c , implies that @xmath101 and @xmath102 .
thus , in this generalized dynamical relation , the exponents may differ from the expected newtonian values ( @xmath3 , @xmath4 ) .
_ significantly , this one dynamical formula ( eq . 12a ) applies to a range of models which are non - homologous_. however , pure isotropic polytropic spheres also fail as acceptable models of elliptical galaxies . in the models , as in actual ellipticals , the l.o.s .
velocity dispersion declines with increasing projected radius .
this decline , mild though it is , is steeper than that typically observed in ellipticals if @xmath103 and too shallow if @xmath104 .
mond polytropes in the range of n=12 to n=16 exhibit roughly the observed form of @xmath105 . for polytropes in this range ,
the mean surface density within the effective radius is still significantly lower than that of true elliptical galaxies typically by about a factor of 5 assuming a mass - to - light ratio of 4 for stellar population of elliptical galaxies .
if such simple spherically symmetric models are to approximate real elliptical galaxies , it is clear an additional degree of freedom must enter into the structure equation ( eq . 1 ) .
here we assume that that degree of freedom is provided by the radial dependence of the anisotropy parameter as represented by eq .
4 . in fig . 2
we see the effect of introducing this second parameter , @xmath84 , on the position of polytropes in the log(@xmath83)-log@xmath106 plane . branching off of the principal curve defined by the sequence of isotropic polytropes are sequences of models with @xmath84 ranging from 200 to 0.1 for polytropes of n=12 and n=16 .
when @xmath107 the models , of course , are very similar to the mond isotropic polytropes .
the effect of increasing anisotropy ( lower @xmath84 ) is to increase the mean surface density of the polytropic spheres as is needed to match the observations of elliptical galaxies .
the mass and effective radius are decreased but the mean surface density is higher . for a given polytropic index , the branch defined by increasing anisotropy ( decreasing @xmath84 ) exhibits a maximum surface density ; for n=16 this maximum is about @xmath108 and occurs for @xmath109 ; that is , for lower @xmath84 ( higher anisotropy ) the surface brightness is again lower .
the sequence of anisotropic models is double - valued in surface brightness .
these models near the maximum surface density are quite anisotropic in the sense that the radial orbit anisotropy reaches within the effective radius ; all models with @xmath110 are designated by an open circle in fig . 2 .
the stability of such anisotropic models is questionable ( binney & tremaine 1987 ) .
given the fact that mond anisotropic polytropes between n=12 and n=16 can reproduce the approximate decline of the l.o.s .
velocity dispersion with projected radius observed in ellipticals , we can take this as an observational restriction upon the range of @xmath40 .
the range of the second parameter , @xmath84 , can also be restricted by excluding all highly anisotropic models ( with @xmath111 ) on the basis of possible radial orbit instability .
3 is the locus of a grid of such models on the log(@xmath83)-log@xmath106 plane .
there are 360 models with @xmath40 = 12 , 13 , 14 , 15 , 16 and @xmath84 = 0.2 , 0.4 , 0.8 , 1.6 , 3.2 , 6.4 and covering a range of the central @xmath27 between 75 km / s and 350 km / s in steps of 25 km / s .
each point in fig . 3 represents a particular value of @xmath40 and @xmath84 and exhibits its own perfect @xmath86 relation .
however , the ensemble of models is non - homologous and presents an ensemble of @xmath86 relations . because @xmath83 varies by a factor of 5 or 6 this would be the expected intrinsic scatter in the combined @xmath112 relation .
however , over this range of parameter space , the models lie in a restricted domain of the log(@xmath113)-log@xmath114 plane sufficiently restricted as to define a theoretical fundamental plane with scatter less than that of the @xmath112 relation .
a least - square fit to this distribution of points gives @xmath115 in eq .
thus , by eqs .
12 , this yields a dynamical fundamental plane relation near that implied by the newtonian virial theorem for homologous systems , i.e. , @xmath116 , @xmath117 .
because the scatter in @xmath83 about this power law relation is much less than the total range in @xmath83 , the scatter perpendicular to the dynamical fundamental plane is very much reduced . for comparison with observations
, it must be realized that both the @xmath118 and fundamental plane relations are altered by the way in which elliptical galaxies are actually observed .
specifically , it is not the velocity dispersion along the very central line - of - sight , @xmath15 , which is measured , but rather the l.o.s .
velocity dispersion , @xmath119 , within some finite - size aperture with radius @xmath120 .
the data of jrgensen et al .
have the advantage that all measured velocity dispersions are corrected to a circular aperture with a fixed linear diameter of 1.6 kpc for h@xmath121 = 75 .
the appearance of a fixed linear scale has the effect of introducing an additional dimensionless parameter into dynamical relationship , eq . 9 ; i.e. , @xmath83 also becomes a function of @xmath122 where @xmath123 is the mond length scale appropriate to the system .
however , this parameter can be absorbed if it is expressed as @xmath124 where @xmath125 km / s .
when we observe the polytropic models in the same way as real galaxies ( determining the volume emissivity - weighted l.o.s . velocity dispersion in the inner projected 0.8 kpc ) , the distributions of velocity dispersions and effective radii may be compared directly to the observations of jrgensen et al .
this is done in fig .
4 where we see that such models can account for the observed range in these properties provided that free parameters cover their allowed ranges : @xmath126 and @xmath127 .
that is to say , the set of models must be non - homologous in order to explain the dispersion in observed properties . for these
realistically `` observed '' models , the dependence of @xmath83 on @xmath119 is also found to be power law ; thus , we may rewrite eq .
9 as @xmath128^\lambda{gma_o } ; \eqno(13)\ ] ] i.e. , i explicitly write the dependence of @xmath83 on @xmath119 leaving the quantity @xmath129 as a pure function of surface density . in fig . 3
we see that the dependence of @xmath129 on surface brightness is well - represented by a power law with with roughly the same exponent as the @xmath83 dependence .
thus the @xmath118 relation becomes @xmath130 and the fundamental plane exponent in eq .
12a is @xmath131 from the models it is found via least - square fits that @xmath132 and @xmath115 implying @xmath133 for the mass - velocity dispersion relation and @xmath134^{1.76 } [ r_e({\rm kpc})]^{0.98}. \eqno(14b)\ ] ] these @xmath118 and dynamical fundamental plane relations are shown for the 360 models in fig . 5 .
the scatter about the dynamical fundamental plane is a factor of 10 smaller than that about the @xmath118 relation .
note in eqs .
14 that for this restricted set of models there is a definite scaling of both the @xmath118 and the dynamical fundamental plane .
using eq .
14b to calculate the mass of the galaxies in the jrgensen et al .
sample , one finds the distribution of m / l shown in fig . 6 which is a log - log plot of m / l against the calculated mass . here
it is found that @xmath135 in solar units ( h@xmath121 @xmath136 75 ) and and @xmath137 .
this , of course , ignores possibly important effects such as deviations from spherical symmetry and systematic rotation , and the fact that real galaxies are almost certainly not characterized by a pure polytropic velocity dispersion - density relation .
bearing these potential dangers in mind , one could also extrapolate eq .
14b down to globular clusters .
for a sample of globular clusters tabulated by trager , djorgovski & king ( 1993 ) and by pryor & meylen ( 1993 ) , i find that @xmath138 in solar units .
thus , mond polytropic spheres in the range n=12 to n=16 with radial anisotropy beyond the effective radius not only reproduce the observed distribution of galaxies in the @xmath139 plane but also provide a reasonable value for the mass - to - light ratio of ellipticals and a weak dependence of m / l on mass . as is evident from eq .
9 , the usual mond @xmath140 relation remains , albeit with large scatter due to the necessary deviations from homology . considering the manner in which the central velocity dispersion is actually measured ( within a fixed circular diaphragm ) the relation
is altered to @xmath141 .
further , considering the necessary dependence of m / l on m , the predicted faber - jackson relation becomes @xmath142 which is consistent with the data of jrgensen et al .
; i.e. , a least square fit to the log(l)-log(@xmath119 ) distribution for this sample of early - type galaxies yields a slope of @xmath143 .
the predicted dynamical fundamental plane eq .
14b can be converted into the more commonly used form by making use of the relation @xmath144 ; then one finds @xmath145 where @xmath113 is the mean _ mass surface density _ within @xmath1 . with @xmath146
we would then predict a fundamental plane of @xmath147 where @xmath16 is the mean surface brightness within @xmath1 . within the uncertainties
this is identical to the fundamental plane defined by the observations of jrgensen et al .
( 1995a , b ) . given the approximations involved in these calculations ( primarily the polytropic gas assumption and the specific radial dependence of the anisotropy parameter ) , the actual exponents of the @xmath118 and fundamental plane relationships are less important than the fact that mond predicts a fundamental plane relation with a scatter which is a factor of 10 less than that about the @xmath118 relation .
this is true in spite of the fact that the set of models must be non - homologous in order to explain the range of observed properties surface density and effective radius .
this arises as a natural aspect of the basic dynamics of systems which extend into the regime of modified dynamics and need not be accounted for by complicated conspiracies in the process of galaxy formation .
moreover various mechanisms for the subsequent dynamical evolution of ellipticals ( e.g. merging , canabalism ) would not effect this relationship .
all that is required is that stellar velocity field in ellipticals not deviate too dramatically from being isothermal and possess a radial orbit anisotropy similar to that described by eq .
the essential results of these calculations can be summarized as follows : .2 in 1 .
the dynamics of high surface brightness elliptical galaxies span the range from newtonian within the effective radius to mond beyond .
the mean surface density within @xmath1 is at least twice as large as the mond surface density , and the internal accelerations are too large to be within the domain of modified dynamics .
this is consistent with the fact that there is no large mass discrepancy or , viewed in terms of dark matter , no need for dark matter within the bright inner regions .
however , mond isothermal spheres have a mean surface density which is roughly one - tenth the critical surface density within @xmath1 ; they are almost entirely pure mond objects .
this effectively rules out these objects as models for elliptical galaxies . 0.2 in 2 .
in order to reproduce the observed global properties of high surface brightness elliptical galaxies in the context of mond , it is necessary to introduce deviations from a constant velocity dispersion and strict isotropy of the velocity field in the outer regions .
this may be done , in an approximate way , by considering mond polytropes in the range n=12 to n=16 with a radial orbit anisotropy beyond an effective radius ( @xmath148 ) .
these objects exhibit the mean decline of line - of - sight velocity dispersion with projected radius observed in elliptical galaxies .
moreover , such models provide reasonable representations of elliptical galaxies with respect to the distribution by velocity dispersion and effective radius ( fig . 4 ) .
in order to match these observations , the models must cover a range in the parameter space of polytropic index and scaled anisotropy radius which implies that strict homology is broken .
this breaking of homology leads to considerable scatter in the mass - velocity dispersion relation ( and the implied faber - jackson relation ) while introducing a third parameter which is the mean surface density or effective radius .
the intrinsic scatter about this dynamical fundamental plane is much lower than that about the mass - velocity dispersion relationship because of the relative insensitivity of this theoretical relationship to deviations from homology ( fig . 5 ) .
both the theoretical @xmath118 and fundamental plane relationships are modified when one considers that the central velocity dispersion is actually measured within a finite size aperture corrected , in the jrgensen et al .
observations , to a fixed diameter of 1.6 kpc for all galaxies in their samples . 0.2 in 3 .
these calculations are highly idealized and apply strictly only to spherical systems with a perfect polytropic equation of state .
moreover , the fact that the models cover a range of internal accelerations around @xmath12 means that the detailed structure is dependent upon the assumed form of the mond interpolation function , @xmath149 ( eq . 2 ) . nonetheless , when the derived dynamical fundamental plane relation is used to estimate the mass of elliptical galaxies from the observed central velocity dispersion and effective radius , one finds , for the galaxies in the samples of jrgensen et al .
, a mean mass - to - light ratio of 3.6 m@xmath24/l@xmath24 with a dispersion of 30% and a weak dependence of this m / l on galaxy mass ( as in the strictly newtonian case ) .
such a m / l would seem quite reasonable for the older stellar populations of elliptical galaxies . when the predicted dynamical fundamental plane relation is converted into an observed relationship ( on the @xmath1 , @xmath119 , and @xmath16 parameter space ) , the jrgensen et al .
result is recovered if @xmath150 .
0.2 in the principal conclusion is that the existance of a fundamental plane with lower intrinsic scatter than that of the faber - jackson relation is implied by modified dynamics , given that high surface brightness elliptical galaxies can not be represented by pure mond isothermal spheres .
it may be argued that newtonian dynamics also predicts a fundamental plane via the traditional virial theorem , and therefore the existance of such a relationship in no sense requires modified dynamics .
it is true that the fundamental plane by itself would not be a sufficient justification for modified dynamics .
however , a curious aspect of the newtonian basis for the fundamental plane is the small scatter about the observed relation in view of the likely deviations from homology in actual elliptical galaxies .
the advantage of mond in this respect is the existance of a single dynamical relationship ( eq . 12 or eq .
13 ) for a range of non - homologous models .
because an additional dimensional constant , @xmath12 , enters into the structure equation ( eq . 2 ) , mond self - gravitating objects are more constrained than their newtonian counterparts . in this respect , it is worthwhile to emphasize that a pure newtonian self - gravitating object with a central line - of - sight velocity dispersion of 200 km / s can have any mass .
but an object with this same velocity dispersion and which extends at least partially into the regime of modified dynamics can only have a galaxy - scale mass .
it is the proximity to being isothermal which requires that elliptical galaxies extend into the regime of modified dynamics .
if this one basic requirement is satisfied , mond inevitably imposes boundary conditions on the inner newtonian solution boundary conditions which restrict these objects to lie on such a well - defined plane in the three dimensional space of observed quantities in spite of detailed variations in the structure between individual objects .
structural variety does , however , lead to a large intrinsic scatter in the @xmath118 relation because each distinct class of objects characterized by an appropriately scaled radial dependence of velocity dispersion and degree of anisotropy exhibits it s own m-@xmath14 relation with a different normalization .
none - the - less , a faber - jackson relation does exist and remains as an imprint of modified dynamics on nearly isothermal hot stellar systems .
0.2 in i am grateful to moti milgrom and stacy mcgaugh for very useful comments on the original manuscript .
i thank especially inger jrgensen and marijn franx for providing their data on elliptical galaxies in convenient form and for helpful comments .
i am grateful to the referee , massimo stiavelli , for remarks and criticisms which greatly improved the content of this paper .
begeman , k.g . ,
broeils , a.h . & sanders , r.h .
1991 , mnras , 249 , 523 bekenstein , j.d . &
milgrom , m. 1984 , apj , 286 , 7 binney , j. & tremaine , s. 1987 , galactic dynamics , princeton univ . press ( princeton ) .
corollo , c.m .
, stiavelli , m. , de zeeuw , p.t . & mack , j. 1997 , aj , 114 , 2366 djorgovski , s. & davis , m. 1987 , apj , 313 , 59 dressler , a. , lynden - bell , d. , burstein , d. , davies , r.l . ,
faber , s.m . ,
terlevich , r.j . , wegner , g. 1987 , apj , 313 , 42 faber , s.m . , & jackson , r.e .
1976 , apj , 204 , 668 franx , m. 1988 , phd dissertation , univ . of leiden gerhard , o.e . & spergel , d.n .
1992 , apj , 397 , 38 jaffe , w. 1983 , mnras , 202 , 995 jrgensen , i. 1999 mnras , 306 , 607 jrgensen , i. , franx , m. & krgard , p. 1995a
, mnras , 273 , 1097 jrgensen , i. , franx , m. & krgard , p. 1995b
, mnras , 276 , 1341 mcgaugh , s. 1998 , astro - ph/9812327 mcgaugh , s.s . ,
de blok , w.j.g . 1998a , apj , 499 , 66 mcgaugh , s.s . ,
de blok , w.j.g . 1998b , apj , 508 , 132 milgrom , m. 1983a , apj , 270 , 365 milgrom , m. 1983b , apj , 270 , 371 milgrom , m. 1984 apj , 287 , 571 milgrom , m. 1994 , ann.phys , 229 , 384 pryor , c. & meylan , g. 1993 , in structure and dynamics of globular clusters , aps series v. 50 , p. 357 romatka , r. 1992 , phd .
thesis , max plank inst .
sanders , r.h .
, verheijen m.a.w .
1998 apj , 503 , 97 sanders , r.h .
1996 , apj , 473 , 117 sanders , r.h .
1990 , a&a rev . 2 , 1 trager , s.c . ,
djorgovski , s. & king , i.r .
1993 in structure and dynamics of globular clusters , aps series v. 50 , p. 347
van albada , t.s .
1982 , mnras , 201 , 939 van albada , t.s . ,
bertin , g. , stiavelli , m. 1995 , mnras , 276 , 1255
|
the modified newtonian dynamics ( mond ) , suggested by milgrom as an alternative to dark matter , implies that isothermal spheres with a fixed anisotropy parameter should exhibit a near perfect relation between the mass and velocity dispersion of the form @xmath0 .
this is consistent with the observed faber - jackson relation for elliptical galaxies a luminosity - velocity dispersion relation with large scatter .
however , the observable global properties of elliptical galaxies comprise a three parameter family ; they lie on a `` fundamental plane '' in a logarithmic space consisting of central velocity dispersion , effective radius ( @xmath1 ) , and luminosity .
the scatter perpendicular to this plane is significantly less than that about the faber - jackson relation .
i show here that , in order to match the observed properties of elliptical galaxies with mond , models must deviate from being strictly isothermal and isotropic ; such objects can be approximated by high - order polytropic spheres with a radial orbit anisotropy in the outer regions .
mond imposes boundary conditions on the inner newtonian regions which restrict these models to a dynamical fundamental plane of the form @xmath2 where the exponents may differ from the newtonian expectations ( @xmath3 , @xmath4 ) .
scatter about this plane is relatively insensitive to the necessary deviations from homology .
epsf galaxies : elliptical galaxies : kinematics and dynamics dark matter , gravitation
|
the level of beaming in grbs is one of the most interesting open questions in this subject .
the relativistic flow which drives a grb may range from isotropic to strongly collimated into a narrow opening angle .
the degree of collimation ( beaming ) of the outflow has many implications on most aspects grb research , such as the requirements from the `` inner engine '' , the energy release and the event rate . during the prompt emission
the lorentz factor of the flow is very high ( @xmath0 ) , and due to relativistic aberration of light , only a narrow angle of @xmath1 around the line of sight ( los ) is visible . during the afterglow stage , the flow decelerates and an increasingly larger angle becomes visible .
as long as @xmath2 an outflow collimated within an angle @xmath3 around the los produces the same observed radiation as if it were part of a spherical outflow .
once @xmath4 drops below @xmath5 , the observer can notice that radiation arrives only from within an angle @xmath3 around the los , instead of @xmath6 as in the spherical case .
sideways expansion is also expected to become important when @xmath7 @xcite .
these two effects combine to create a break in the light curve at @xmath7 , but it is not quite clear whether this break is sharp enough to be detected @xcite . to explore this question we have performed fully relativistic three dimensional simulations that follow the slowing down and the lateral expansion of a relativistic jet .
we then calculate the synchrotron light curve and spectrum using the conditions determined by the hydrodynamical simulation .
we use a fully relativistic three dimensional code for the hydrodynamical calculations .
the initial conditions are a wedge with an opening angle @xmath8 taken from the blandford mckee @xcite , bm hereafter , self similar spherical solution , embedded in a cold and homogeneous ambient medium . for this initial opening angle
, the jet is expected to show considerable lateral expansion when @xmath9 , where @xmath4 is the lorentz factor of the fluid .
the bm solution used for the initial conditions was therefore at the time when @xmath10 , where @xmath11 is the lorentz factor of the shock .
the total isotropic energy was @xmath12 , and the ambient number density was @xmath13 . in figures [ fig2]-[fig5 ] we present snapshots of the number density , internal energy density , lorentz factor and velocity field of the fluid , as the jet slows down and spreads sideways .
an explanation of what is seen in these snapshots is given in figure [ fig1 ] .
the snapshots are taken at consecutive times , ranging @xmath14 in the rest frame of the ambient medium ( corresponding roughly to observer times from one and a half to twenty days for an observer along the jet axis ) .
the results shown in this work are still preliminary .
the resolution is not sufficient to resolve the very thin initial shell .
for this reason the maximal lorentz factor of the matter is just over @xmath15 , instead of @xmath16 ( see figure [ fig4 ] ) .
higher resolution runs are in progress .
the local emission coefficient at a given space - time point is calculated directly from the hydrodynamical quantities there .
the magnetic field and electron energy densities are assumed to hold constant fractions , @xmath17 and @xmath18 , respectively , of the internal energy , while the electrons posses a power law energy distribution , @xmath19 .
the local emissivity is approximated by a broken power law : @xmath20 and @xmath21 , below and above the typical synchrotron frequency , respectively @xcite .
the light curves and spectra are calculated for several viewing angles with respect to the jet axis .
once the emission coefficient is determined in the local frame , it is transformed to the frame of each observer .
the time of arrival to each observer is then calculated , and the contributions are sumed over space - time , producing the various light curves .
this is done for several frequencies , simultaneously , so that the spectrum may be obtained , as well as the light curve at different frequencies .
a few light curves , calculated for @xmath22 and @xmath23 , are shown in figure ( [ fig6 ] ) .
these light curves serve to demonstrate the potential of this approach .
future simulations are expected to achieve sufficient resolution to produce realistic light curves which could be compared with afterglow observations .
blandford , r.d . & mckee , c.f .
1976 , phys . of fluids , * 19 * , 1130 .
granot , j , piran , t. & sari , r. 1999 , apj , * 513 * , 679 .
rhoads , j.e . , 1999 ,
apj submitted , astro - ph9903399 .
sari , r. , piran , t. & halpren , j.p . , 1999 ,
apj , * 519 * , l17 .
wei , d.m .
& lu , t. 1999 , astro - ph9908273 the region depicted in figures ( [ fig2]-[fig5 ] ) is within the shaded region , outlined in blue . @xmath24 and @xmath25 are the cylindrical coordinates .
the jet axis is in the @xmath24 direction .
two colored lines which indicate opening angles of @xmath26 and @xmath27 are added to help follow the lateral expansion.,width=529 ] the velocity field of the fluid .
each arrow points at the direction of the velocity at the point where it begins , while its length is proportional to the size of the velocity.,title="fig:",width=151 ] the velocity field of the fluid .
each arrow points at the direction of the velocity at the point where it begins , while its length is proportional to the size of the velocity.,title="fig:",width=151 ] the velocity field of the fluid .
each arrow points at the direction of the velocity at the point where it begins , while its length is proportional to the size of the velocity.,title="fig:",width=151 ] the velocity field of the fluid .
each arrow points at the direction of the velocity at the point where it begins , while its length is proportional to the size of the velocity.,title="fig:",width=151 ] the velocity field of the fluid .
each arrow points at the direction of the velocity at the point where it begins , while its length is proportional to the size of the velocity.,title="fig:",width=151 ] the velocity field of the fluid .
each arrow points at the direction of the velocity at the point where it begins , while its length is proportional to the size of the velocity.,title="fig:",width=151 ] + the green , red , blue and black curves represent the light curves at the observed frequencies @xmath28 and @xmath29 , respectively , as seen by an observer at a distance of @xmath30 along the jet axis.,width=302 ]
|
we describe fully relativistic three dimensional calculations of the slowing down and spreading of a relativistic jet by an external medium like the ism .
we calculate the synchrotron spectra and light curves using the conditions determined by the hydrodynamic calculations .
preliminary results with a moderate resolution are presented here .
higher resolution calculations are in progress .
|
there is substantial evidence to indicate that the central regions of seyfert 1 active galactic nuclei ( agn ) consist of a geometrically - thin , optically - thick accretion disk surrounding a supermassive ( 10@xmath3@xmath4 ) black hole . in the leading models ,
the inner disk is surrounded by a hot , optically - thin corona that compton - upscatters soft photons thermally emitted from the disk to produce an x - ray power - law continuum over the range 1100 kev , though the geometry of the corona remains uncertain ( e.g. , haardt , maraschi & ghisellini 1994 ) .
furthermore , some optically thick material , such as the accretion disk , intercepts and reprocesses some of the x - rays , producing the so - called compton reflection hump above @xmath110 kev and a strong fe k@xmath0 fluorescent emission line at 6.4 kev ( guilbert & rees 1988 , george & fabian 1991 ) . though the exact configuration of accreting material throughout the central engine is not known , the fe k@xmath0 line , as a primary signature of the reprocessing , serves as a potential tool to trace the distribution of accreting gas .
single - epoch spectral fitting has revealed a variety of profiles for the fe k@xmath0 line . in some seyfert 1
galaxies , the line is seen to be relativistically broadened ( e.g. , fwzi up to 100,000 km s@xmath5 in the well - studied broad line profile of mcg6 - 30 - 15 , tanaka et al .
1995 ) and redshifted .
this can be explained in terms of a line originating in the innermost regions of the accretion disk , in close proximity to the black hole , and affected by strong doppler and gravitational redshifts ( fabian et al .
however , the majority of seyfert 1 galaxies observed with _ xmm - newton _ ( reeves 2002 ) and _ chandra _ ( yaqoob , george & turner 2001 ) feature a narrow fe k@xmath0 component , modelled to be only mildly broadened , with fwhms of @xmath610,000 km s@xmath5 ( e.g. , kaspi et al .
2002 , nandra 2001 , obrien et al .
2001 , yaqoob et al .
2001 ) . for these objects , a line origin in the outer accretion disk , putative molecular torus , broad emission line region or narrow emission line region
may be more relevant . studying the time - resolved variability of the line and continuum components
provides a complementary analysis to single - epoch spectral fitting .
for instance , fe k@xmath0 line emission originating in close proximity to the central black hole is expected to respond rapidly to changes in the illuminating continuum flux ( stella 1990 , matt & perola 1992 ) .
the simplest models predict that if the reprocessing time scale is negligible , then variations in the fe k@xmath0 line should track continuum variations , modified only by a time delay equal to the light - crossing time across the inner disk .
however , recent studies have not been able to support this picture .
investigations into the behavior of the broad fe k@xmath0 line in mcg6 - 30 - 15 by iwasawa et al .
( 1996 , 1999 ) , lee et al .
( 1999 ) , reynolds ( 2000 ) , vaughan & edelson ( 2001 , hereafter ve01 ) , and fabian et al .
( 2002 ) have confirmed the presence of rapid fe k@xmath0 line variability in this object , but have found the line to vary considerably less strongly than the continuum , and have found no correlation between continuum and line variations .
yaqoob et al .
( 1996 ) found the red wing of the fe k@xmath0 profile in ngc 7314 to respond to rapid continuum variations , while the core did not respond as rapidly .
conversely , nandra et al .
( 1999 ) found that the narrow core of the line in ngc 3516 tracked the continuum flux , while the broad wings displayed uncorrelated variability .
petrucci et al .
( 2002 ) used _ xmm - newton _ to obtain two observations of mkn 841 separated by 15 hours ; while the continuum flux and shape stayed nearly constant between the two observations , the narrow line virtually disappeared .
this paper aims to study the broadband spectral variability of seyfert 1 galaxies on time scales of days to years .
this includes quantifying the variability of the fe k@xmath0 line and determining the relation between the line and the continuum flux . to this end
, samples have been assembled that probe spectral variability on time scales of days to weeks ( hereafter referred to as `` short '' time scales ) and time scales of months to years ( hereafter referred to as `` long '' time scales ) , using archival monitoring data from the _ rossi x - ray timing explorer ( rxte)_. _ rxte _ provides the largest effective area at 6 kev of any current x - ray instrument flown to date and thus offers the highest signal - to - noise of the fe k region of the x - ray spectrum . the data reduction and sampling
are described in @xmath72 . in order to track the line behavior , direct time - resolved spectral
fitting has been systematically performed and fractional variability amplitude spectra have been constructed , presented in @xmath73 .
the observational results are summarized in @xmath74 .
the physical implications are discussed in @xmath75 , followed by a brief summary of the main results in @xmath76 .
the goal of this project was to obtain uniform samples to survey line and continuum variability behavior .
continuous _ rxte _
archival monitoring data turning public before 2002 december , when these analyses were performed , were used , as well as the authors proprietary observations of three seyfert 1 galaxies observed during cycle 7 .
all _ rxte _ data considered in this paper were obtained with the proportional counter array ( pca ) .
318 kev count rates were estimated with the online w3pimms tool , using mean 210 kev archival count rates and photon indices obtained either from previously published fits ( e.g. , kaspi et al .
2001 ) or the tartarus database .
the goal of time - resolved spectral fitting was to obtain at least ten high - quality spectra with at least @xmath150,000 photons in each time bin ; sources that failed to meet this criterion were excluded .
this resulted in eight observations consisting of continuous monitoring on short time scales ( e.g. , sources observed once every 12 orbits for @xmath11 ks over a span of @xmath130 days ) and an additional seven observations covering long time scales ( e.g. , sources observed once every 3 or 4 days for several years ) .
there were six objects common to both lists .
this included two mcg6 - 30 - 15 observations on short time scales , an 8-day intensive scan henceforth denoted as short - term mcg6 - 30 - 15 ( i ) and a 53-day observation denoted as short - term mcg6 - 30 - 15 ( ii ) .
this also included two mcg6 - 30 - 15 observations on long time scales , one spanning 326 days denoted as long - term mcg6 - 30 - 15 ( i ) and a 2093-day observation denoted as long - term mcg6 - 30 - 15 ( ii ) .
additionally , ngc 7469 was monitored only on the short time scale .
table 1 lists the date ranges and source parameter information .
the pca consists of five identical collimated proportional counter units ( pcus ; swank 1998 ) . for simplicity ,
data were collected only from those pcus that did not suffer from repeated breakdown during on - source time ( pcus 0 , 1 , and 2 prior to 1998 december 23 ; pcus 0 and 2 from 1998 december 23 until 2000 may 12 ; pcu 2 only after 2000 may 12 ) .
count rates quoted in this paper are normalized to 1 pcu .
only pca standard-2 data were considered .
the data were reduced using standard extraction methods and ftools v5.2 software .
data were rejected if they were gathered less than 10@xmath8 from the earth s limb , if they were obtained within 30 min after the satellite s passage through the saa , if electron0 @xmath9 0.1 ( electron2 after 2000 may 12 ) , or if the satellite s pointing offset was greater than 0@xmath1002 .
as the pca has no simultaneous background monitoring capability , background data were estimated by using pcabackest v2.1e to generate model files based on the particle - induced background , saa activity , and the diffuse x - ray background .
this background subtraction is the dominant source of systematic uncertainty ( e.g. , in total broadband count rate ) in _ rxte _ agn monitoring data ( e.g. , edelson & nandra 1999 ) .
counts were extracted only from the topmost pcu layer to maximize the signal - to - noise ratio .
all of the targets were faint ( @xmath6 40 ct s@xmath5 pcu@xmath5 ) , so the most recent versions of the l7 - 240 background models were used .
source and background spectra were derived over the time intervals listed in table 1 .
these time bins were chosen to ensure adequate signal - to - noise in the fe k@xmath0 line in the time - resolved fits , but also to ensure adequate temporal resolution in determing the variability characteristics of both the continuum and the line .
response matrices were generated for each gain epoch using pcarsp v.8.0 .
the time - averaged fits , described in @xmath73.1 , were performed with the goal of minimizing the number of free parameters needed to produce an acceptable fit .
these model fits were then applied to the time - resolved data , as described in @xmath73.2 .
the errors on parameters derived from spectral fitting are discussed in @xmath73.3 .
the time - resolved spectral fitting is discussed in @xmath73.4 .
continuum and line variability amplitudes are presented in @xmath73.5 .
temporal cross - correlation analysis is described in @xmath73.6 .
finally , the fractional variability amplitude ( @xmath2 ) spectra are described in @xmath73.7 . for all analyses ,
data below 3.5 kev were discarded in order to disregard pca calibration uncertainties below this energy and also to reduce the effects of warm absorbers in sources ( e.g. , mcg6 - 30 - 15 , reynolds 1997 ; ngc 3783 , kaspi et al .
2002 ; ngc 5548 , nicastro et al .
data above 18 kev were discarded as the source count rate became too low at higher energies to obtain a good detection in each time bin .
all spectral fitting was done with xspec v.11.0.2 ( arnaud 1996 ) .
most long - term observations spanned multiple gain epochs , requiring simultaneous fitting of data sets . for all observations , the simple model of a power - law modified by cold absorption resulted in large residuals near 6 kev due to the fe k@xmath0 line ; the addition of a emission component proved significant at greater than 99.99@xmath11 confidence in an f - test .
given the limit of pca calibration ( data / model residuals @xmath122 - [email protected] ] , and the low energy resolution of the pca ( @xmath11 kev at 6 kev ) , the pca is largely insensitive to the detailed shape of the fe k@xmath0 line .
both a single gaussian and a laor model component ( diskline model for a maximally rotating kerr black hole ; laor 1991 ) were tested to parameterize the line emission .
more complex fits such as double gaussians or a laor component plus a narrow gaussian were either not required or did not yield sensible results ; the single - gaussian and laor fits alone gave sensible results .
for the laor model component , the inclination angle was kept fixed at 30@xmath8 , and the outer radius kept fixed at 400 gravitational radii ; in no case did thawing either or both of these parameters measurably improve the fits .
the emissivity index @xmath13 was constrained to fall in the range 0.09.9 .
laor models gave significantly better fits ( at @xmath999@xmath11 confidence in an f - test ) than the single - gaussian parameterization in seven of the fifteen time - averaged spectra .
an implicit assumption in these fits is that the emission regions in each source do not evolve substantially ( i.e. , major changes in the geometry of the emitting regions ) throughout the duration of the observation . for all sources except ngc 4151 ,
neutral absorption was kept fixed at the galactic value . for ngc 4151 ,
modelling the column density with a value much higher than galactic resulted in a significant ( @xmath999.99@xmath11 confidence in an f - test ) improvement in the model fit .
the column density values obtained , given the pca calibration , were in reasonable agreement with each other and with the value of [email protected] @xmath15 10@xmath16cm@xmath5 in the spectral fit by schurch & warwick ( 2002 ) .
for all model fits except the short time scale observation of 3c 390.3 and the laor model fit to the short time scale observation of ngc 3516 , there were residuals above 10 kev that signalled the need to incorporate a compton reflection hump .
including a pexrav component to model reflection of an underlying power law component off optically thick and neutral material ( magdziarz & zdziarski 1995 ) provided the necessary improvement to the fits , significant at greater than 99@xmath11 confidence compared to models with just a power - law plus fe k@xmath0 line , modified by cold absorption . in the pexrav component , the inclination angle of the reflector
was kept fixed at 30@xmath8 , and the power - law cutoff energy was kept fixed at 100 kev .
all element abundances were kept fixed at solar ; as noted by ve01 a change in fe abundance caused the absolute values of spectral fit parameters to change slightly , but the variability pattern , the focus of this paper , remained unchanged .
the basic model thus used , a power - law component plus pexrav ( where applicable ) plus a parameterization for the fe k@xmath0 line as described above , all modified by neutral absorption , gave reasonable fits to the data .
the exceptional signal - to - noise in these time - averaged spectra invariably meant that systematic errors associated with the background modelling and spectral response of the pca dominated the uncertainties .
consequently , finding a statistically acceptable time - averaged fit ( with @xmath18 @xmath1 1 ) was unlikely even if the model was representative of the instrinsic spectrum of the source .
the data / model residuals seen were consistent with the limit of the pca calibration .
tables 23 summarize the time - averaged fitting results for both the single - gaussian and laor models , respectively , on both time scales .
while higher - resolution x - ray instruments can detect one or both of the fe k@xmath0 broad and narrow components , the pca is sensitive only to the combined flux of the two components .
recent _ xmm - newton _ model fits can provide insight as to which component , the broad or narrow , dominates the _ rxte_view in each object .
( _ chandra - hetg _ can additionally constrain the narrow component but is mostly insensitive to the broad component . ) for instance , the equivalent width of the broad component has been measured to be much greater than that of the narrow component in both mcg6 - 30 - 15 ( wilms et al .
2001 ) and ngc 3516 ( turner et al .
the present _ rxte _ fits also indicated that the fe k@xmath0 lines in these two objects are among the broadest of the sample , consistent with the bulk of the line emission originating in the inner accretion disk .
in constrast , the narrow component seems to dominate in ngc 4151 ( schurch et al .
2002 ) , ngc 5548 ( pounds et al .
2003 , yaqoob et al .
2001 ) and ngc 7469 ( blustin et al .
2003 ) , with a small upper limit or no detection of the broad component .
the present _ rxte _ fits for ngc 4151 , for instance , indicated the narrowest line of the sample , consistent with emission far from the central black hole .
the equivalent widths of the broad and narrow components in ngc 3783 are approximately equal ( blustin et al .
2002 ) , indicating that emission from both the inner accretion disk and a more distant location is relevant in this source .
the above models were then applied to the time - resolved data .
spectral fits to time bins spanning gain epoch changes required simultaneous fitting of the two data sets .
the fe k@xmath0 line was always detected in each time bin , as the addition of a line emission component was significant at @xmath999@xmath11 confidence in an f - test for nearly every time bin ( in the single exception , the last time bin of the long time scale observation of ngc 5548 , it was significant only at 98.1@xmath11 confidence ) . while fitting the time - averaged models to their respective time - resolved spectra , it was not practical to thaw parameters such as the fe k@xmath0 line profile shape ( line width in the single - gaussian models ; inner radius and emissivity index @xmath13 in the laor models ) or line peak energy , given the pca resolution and sensitivity . the f - test was used to determine which free parameters to include in the fit .
the time - resolved spectral fits were repeated with each of photon index @xmath19 , reflection fraction @xmath20 ( ratio of the reflection normalization at 1 kev to the normalization of the power law ) and line flux normalization f@xmath21frozen at their time - averaged values , and the values of total @xmath22 were compared via an f - test to determine which parameters could justifiably be allowed to be free and which should be kept fixed in the fitting , as listed in table 4 .
there was no formal significant evidence ( @xmath999.7@xmath11 confidence ) of the need to thaw @xmath20 in any object . for most sources ( 6/8 short time scale and 6/7 long time scale observations )
, there was significant ( @xmath999.7@xmath11 confidence ) evidence of the need to thaw @xmath19 .
there was also significant ( @xmath999.7@xmath11 confidence ) evidence of the need to thaw f@xmath21 in the short time scale observations of mcg6 - 30 - 15 ( i ) , ngc 3516 and ngc 7469 as well as in the long time scale observations of mcg6 - 30 - 15
( ii ) , ngc 3516 and ngc 4151 .
these results initially suggest statistically significant variations in the line in at least some of these objects .
table 5 lists the mean values of the 210 kev flux @xmath23 , @xmath19 and f@xmath21 derived from the time - resolved spectral fitting .
@xmath23 was measured from the model fits . for the majority of the individual time - resolved fits , the resulting values of the fit statistic
@xmath24 were rather low ( @xmath24 @xmath6 1.0 for 243 of the 316 total time bins ) .
as discussed by nandra et al .
( 2000 ) and ve01 , this was most likely due to the standard software s overestimation of the errors in any given energy bin in the net spectrum .
these errors are produced by combining the errors in the source with those of the modelled background .
the software uses poisson statistics , estimated from a large quantity of background data , to estimate the background error . for short exposures , this results in an overestimate of background errors , which should typically be negligible compared to the error on the total ( source plus background ) rate . to better ascertain the variability of the measured spectral parameters ,
the standard errors produced by the software were discarded and new 1@xmath25 errors were estimated based on the variability properties of the derived light curves .
the point - to - point variance was used ; this is defined as @xmath26 where @xmath27 are each of the n total points in the light curve .
ve01 used simulations to validate the accuracy of these errors .
these are conservative error estimates ; only trends in the light curves sustained over several time bins will exceed the expected error using this formulation .
these errors are listed in table 5 for @xmath19 and @xmath28 and used hereafter .
errors on values of @xmath23 within a time bin were determined from the 210 kev continuum light curve , using the mean flux error on the @xmath11 ks exposures in that time bin .
figures 1 and 2 show light curves of @xmath23 , @xmath19 , and f@xmath21for the single - gaussian model on short and long time scales , respectively .
the results of the laor model time - resolved fits showed identical variability patterns and nearly identical time - averaged values to the single - gaussian fits , and they are not plotted .
this spectral fitting was performed keeping the value of @xmath20 fixed , i.e. with the normalization of the pexrav reflection component varying in sync with the normalization of the power - law .
this was done as there was no formal statistical requirement to allow the value of @xmath20 to vary in the fits ; freezing @xmath20 also minimized the influence of this free parameter on measurement of @xmath19 and f@xmath21 . however , as the reflection continuum and fe k@xmath0 line are likely to be produced in the same location ( e.g. , the accretion disk ) it would perhaps be expected that the reflection component tracks variations in the line rather than the power - law continuum .
time - resolved spectral fitting was repeated with the normalization of the reflector tied to the f@xmath21instead .
the resulting light curves of f@xmath21 were essentially the same as those with a fixed value of @xmath20 .
fractional variability amplitudes ( @xmath2 ; e.g. , vaughan et al . 2003 , edelson et al . 2002 ) were measured to quantify the instrinsic variability amplitude relative to the mean count rate and in excess of the measurement noise in the continuum and line light curves ; @xmath29 where @xmath30 is the total variance of the light curve , @xmath31 is the mean square error and @xmath32 is the mean count rate of @xmath33 total points .
the error on @xmath2 is @xmath34 as discussed in vaughan et al .
this error formulation estimates @xmath35 based on random errors in the data itself , and not due to random variations associated with red - noise processes .
furthermore , these estimates are only an approximation to the true values of @xmath35 since the data errors used here instead have been estimated from the point - to - point variance .
table 6 lists the continuum and line variability amplitudes .
temporal cross - correlation functions were calculated for f@xmath36f@xmath21 , f@xmath36@xmath19 , and @xmath19f@xmath21 using both the discrete correlation function dcf ; edelson & krolik 1988 ) and the interpolated correlation function ( icf , white & peterson 1994 ) .
icf lag uncertainties were calculated using the bootstrap method of peterson et al .
( 1998 ) ; however , this method often underestimates the true errors introduced by uneven sampling , e.g. , edelson et al .
figures 3 and 4 and tables 79 show the cross - correlation results , where a positive shift denotes the first quantity leading the second .
the dcf and icf time bins used were 1.0 and 0.5 times the time bin size in the time - resolved fits , respectively .
one must be skeptical of cross - correlation analysis in these circumstances , however , due to the red - noise nature of the continuum and line fluxes ( as speculated below ) .
cross - correlation between two unrelated red - noise noise light curves can randomly yield spurious correlations with higher than expected values of the correlation strength @xmath37 ( e.g. , welsh 1999 ) .
for instance , the @xmath37 = 0.7 optical - to - x - ray lag in ngc 3516 claimed by maoz et al .
( 2000 ) was not confirmed with additional data by maoz et al .
combined with the fact that most of these light curves only have 1020 points , spurious correlations and anticorrelations , including those based on very few points ( i.e. , trends much shorter than the total duration ) , can arise .
discarding peak lags greater than @xmath11/3 of the total duration can eliminate some of these spurious lags ; this has been done for the values reported in tables 79 and the functions plotted in figures 3 and 4 .
however , as has been done in the present work , it is more practical to not trust any lag where @xmath37 is not very close to 1.0 , applying relatively more skepticism for lower values of @xmath37 .
cross - correlation analysis of f@xmath36 and f@xmath21 in the short time scale observations yields peak correlation coefficients in the range 0.170.72 .
these values do not readily convince us that the analysis has yielded believable correlations .
for instance , the highest correlation coefficient , @xmath37 = 0.72 , arises in ngc 3783 , which only has 12 points and for which substantial variability in the line is not detected ( see table 4 ) .
ngc 5548 , for which cross - correlation analysis yields the second - highest value of @xmath37 in the short time scale sample , is a similar situation : the reported lag is far from zero , based on only a handful of points , and one can not conclude that this is a reliable correlation robustly detected in excess of the long time scale red - noise variations present .
it follows that the lag delays and lag errors listed in table 7 must also be regarded with a large dose of skepticism . for long time scale
f@xmath36f@xmath21 correlations , the values of @xmath37 lie in the range 0.450.87 .
despite the presence of higher values of @xmath37 , however , caution is again warranted .
for instance , as the continuum and line light curves are both dominated by large decreases in ngc 3516 and ngc 5548 , the resulting cross - correlation functions have moderately - high values of the correlation coefficient at nearly all lags tested , and the peak correlation coefficients do not provide evidence of a strong correlation `` in excess '' of this artefact . if there were an increase in the values of @xmath37 towards a particular time scale , this would indicate more of a physical relevance associated with that time scale .
however , only three objects show an increase in their respective values of average dcf and icf @xmath37 towards longer time scales .
@xmath19f@xmath21 cross - correlation functions show similar correlation coefficient values as the f@xmath36f@xmath21 cross - correlation functions . however , many cross - correlation functions between f@xmath36 and @xmath19 on both time scales yield higher values of @xmath37
. the values of @xmath37 on short time scales span 0.540.92 ; those on long time scales span 0.310.94 ( 0.620.94 if ngc 4151 is excluded ) .
all but one of the lag delays are consistent with zero .
@xmath2 spectra , measuring @xmath2 of the net ( background - subtracted ) count rate in each channel , were calculated as a relatively less model - dependent assessment of the broadband spectral variability .
such analysis allows one to study the behavior of the varying components .
source and background spectra were derived over the time intervals listed in table 10 with the goal of constructing light curves of the flux in each channel for at least @xmath120 points .
in addition to the observations used in the above analysis , @xmath2 spectra were measured for an additional four objects on short time scales and five objects on long time scales ; the total of twelve observations on each time scale included nine objects common to both lists . because the pcu gain settings changed three times since launch and
the channel energy boundaries differ between gain epochs , the longest duration that did not cross a gain epoch boundary was used .
the error on each light curve point was assigned using the point - to - point variance ( see @xmath73.3 ) .
the highest - energy bins were re - grouped as necessary to insure adequate signal - to - noise in all energy bins . for the soft - spectrum ( @xmath19 @xmath12.7 ; vaughan et al .
1999 ) source akn 564 , data above 9 kev were discarded to minimize the influence of systematic variability in the _ rxte _ background models .
similarly , in fairall 9 ( @xmath19 @xmath1 2.2 over 210 kev , tartarus database ) , data above 14.3 kev were discarded for the short time scale data .
@xmath2 and @xmath35 were calculated in each channel using the equations in @xmath73.5 .
the final @xmath2 spectra are presented in figures 5 and 6 for short and long time scales respectively .
for the seven sources with published power spectral densities ( markowitz et al .
2003 , uttley , mchardy & papadakis 2002 ) , the relative continuum variability amplitudes obtained on short and long time scales are in reasonable agreement with values predicted from their published power spectral density shapes .
a summary of the observed continuum broadband flux and spectral variability properties of the sample is given in @xmath74.1 .
the observed line variability properties are summarized in @xmath74.2 .
finally , a comparison between the continuum and line variability properties is given in @xmath74.3 .
notes on individual sources are deferred to appendix a. all objects display significant variability in continuum flux : in @xmath22-tests against the constant hypothesis , the constant hypothesis is rejected at @xmath999.99@xmath11 confidence for all continuum light curves .
visual inspection of figures 1 and 2 confirms this .
there is evidence for the continuum to vary more strongly towards longer time scales : of the six objects with both short- and long time scale monitoring , the value of continuum @xmath2 increases towards longer time scales for five objects .
the @xmath2 spectra also exhibit stronger continuum variability on longer time scales in most sources .
both these analyses support an increase in variability amplitude towards longer time scales that is consistent with the long - established red - noise nature of seyfert 1 continuum x - ray emission .
the majority of @xmath2 spectra show the softer energies to be the most variable , consistent with softening of the dominating broadband x - ray continuum as sources brighten ( e.g. , markowitz & edelson 2001 , nandra et al .
the only exception is akn 564 , which exhibits a much flatter dependence on @xmath2 with energy .
the presence of f@xmath36@xmath19 correlations with no measureable lags in many objects on both time scales supports this idea .
such behavior has been observed numerous times before ( e.g. , ve01 , done et al .
2000 , lee et al .
2000 , chiang et al . 1999 ,
nandra , pounds & stewart 1990 ) . above @xmath110
kev , unmodelled systematic variations in the background can contribute greatly to the total observed variability .
in most sources , therefore , the influence of the variability of the compton reflection hump on the broadband continuum above @xmath110 kev can not be reliably constrained from the @xmath2 spectra .
given that the 57 kev continuum flux greatly outshines the line flux and given the nature of spectral fitting , it is possible that there could exist some systematic effect that corrupts the determination of the continuum level at the @xmath1few per cent level , in turn corrupting determination of the line flux by @xmath110 percent or higher .
however , such an effect is not readily identifiable , and the line variations measured are hence assumed intrinsic to the sources .
time - resolved spectral fitting has revealed a wide range of variability in the line on both time scales . on short time
scales , mcg6 - 30 - 15 ( i ) shows evidence for line variability : a @xmath22-test against a constant hypothesis results in rejection of that hypothesis at @xmath999@xmath11 confidence .
similarly , the constant hypothesis is rejected in mcg6 - 30 - 15 ( ii ) and ngc 3516 , but only at 98@xmath11 and 95@xmath11 confidence , respectively . on long time scales , the number of objects showing line variability increases : in three of the seven observations ( ngc 3516 , ngc 4151 and ngc 5548 ) , the constant line flux hypothesis is rejected at @xmath999.99@xmath11 confidence . in two additional observations , mcg6 - 30 - 15 ( i ) and 3c 390.3
, the constant line flux hypothesis is rejected , but only at 97.9@xmath11 and 97.7@xmath11 confidence , respectively .
the most noteable trends in line flux are the gradual decreasing trends throughout the long time scale observations of ngc 3516 , ngc 4151 and ngc 5548 .
lack of line variability can not be ruled out in mcg6 - 30 - 15 ( ii ) and ngc 3783 .
overall , one can conclude that fe k@xmath0 line variability is more commonly detected towards longer time scales .
this tendency is confirmed using measured values of @xmath2 : the value of line @xmath2 increases towards longer time scales for four objects , the values of line @xmath2 on both time scales are consistent for the remaining two objects , and in no case does line @xmath2 decrease towards longer time scales .
it is therefore plausible to conclude that , at least on the time scales probed herein , the iron line variability is , like the continuum variability , a red - noise process . the average maximum peak - to - trough line flux variation detected is a factor of @xmath12 for all observations ; typically , this is a factor of @xmath14.5 greater than the typical error bar .
the largest peak - to - trough variations are @xmath13 , as in the short - term observation mcg6 - 30 - 15 ( ii ) and the long - term observations of ngc 4151 and ngc 5548 .
the smallest peak - to - trough variations are @xmath14050% , as with the short - term observations of ngc 3516 and ngc 3783 and the long - term obsrvations mcg6 - 30 - 15 ( ii ) and ngc 3783 ; typically this is a factor of @xmath12 greater than the typical error bar . with the time bins chosen in this experiment to maximize the variability - to - noise in the iron line light curves , this experiment may not be sensitive to smaller amplitudes of variability ( peak - to - trough variations of @xmath130@xmath11 or less )
. visual examination of the continuum and line light curves does not reveal any obvious relation between them in most objects .
large - amplitude trends in continuum often go unmatched by simultaneous changes in the line ; prominent examples of this are listed for individual sources below . in only a few observations
do the continuum and line appear to track each other : in the long time scale observation of ngc 5548 , both the continuum and line decrease by a factor of @xmath13 over the 4-year duration .
there are also somewhat similar decreasing trends dominating both the continuum and line in the long - time scale observations of ngc 3516 and ngc 4151 .
measured variability amplitudes , @xmath2 spectra , and the increased detection of line variability in the light curves towards longer time scales all indicate that both the continuum and line appear to be more strongly variable towards longer time scales in most objects .
however , there is evidence that the line does not vary as strongly as the continuum . in 11/15 observations ,
the value of continuum @xmath2 is greater than the line @xmath2 .
for the eight of these eleven observations with a defined value of line @xmath2 ( positive excess variance ) , the continuum and line excess variances are compared via an f - test . in six of these eight observations ,
the null hypothesis of the continuum and line excess variances being consistent is rejected at @xmath998.2@xmath11 confidence . in four of the fifteen observations ,
the values of continuum and line @xmath2 are consistent with each other . for three of these observations
, an f - test can not reject the null hypothesis of the continuum and line excess variances being consistent at greater than 79@xmath11 confidence ; it is rejected at @xmath999.99@xmath11 confidence in the fourth observation ( the long - term ngc 3783 observation ) .
there are no observations with a larger value of @xmath2 for the line compared to the continuum .
as seen in figures 5 & 6 , in many of the @xmath2 spectra , there is a reduction in variability at @xmath16 kev .
the average drop in @xmath2 at @xmath16 kev relative to the surrounding continuum is @xmath110@xmath11 , or @xmath15 times larger than the typical errors .
best fit lines to the @xmath2 spectra including and excluding the @xmath157 kev region indicate that the dip is significant at greater than 90@xmath11 confidence in an f - test for 11/24 observations .
these dips are consistent with dilution of the @xmath157 kev continuum variability by relatively reduced line variability .
such a feature has been noted previously in the @xmath2 spectra of ngc 4151 ( schurch & warwick 2002 ) and mcg6 - 30 - 15 ( fabian et al . 2002 ) .
the presence of this feature supports the argument for decoupling of the line and continuum variability properties .
additionally , cross - correlation analysis does not reveal any significant temporal correlations between continuum and line variations .
overall , there hence is little support for a simple or direct temporal relation between the observed continuum and fe k@xmath0 line variability properties .
this result broadly agrees with the study of weaver , gelbord & yaqoob ( 2001 ) , who did not find any evidence that fe k@xmath0 line variability on time scales of days to years was correlated with continuum variations ; their _ asca _ sample included all of the objects in this paper s time - resolved spectroscopy samples except 3c 390.3 , though with much less continuous sampling .
this paper examines the variability properties of both the continuum and fe k@xmath0 line on time scales from days to years .
the continuum variability properties observed in the current sample have been observed before in many seyfert 1 galaxies .
the data show an increase in broadband variability amplitude towards longer time scales , demonstrating the red - noise nature of the continuum variability .
additionally , there is an increase in variability amplitude towards the softer x - rays , consistent with sources softening as they brighten .
in contrast , the fe k@xmath0 line shows a divergence of behaviors .
most observations show no clear trend whatsoever .
only a few observations show temporal trends in the line flux that mimic the continuum to at least some degree , and only on time scales of years . as discussed in @xmath74.3
, however , the line varies less strongly than the continuum in many sources , as seen from a comparison of the spectral fit line and continuum flux light curve variability amplitudes as well as from a reduction in broadband @xmath2 near 6 kev seen in many sources @xmath2 spectra .
the relationship between continuum and line variability is examined further in @xmath75.1 .
the broadband spectral variability results are discussed in @xmath75.2 .
this paper has demonstrated the presence of a range of variability behavior in the fe k@xmath0 line .
the results permit tests of simple models in which the bulk of the continuum originates in a central corona and the line variations are driven solely by continuum flux fluctuations modified by light - travel time effects .
for instance , such models imply that there should exist a connection between the origin of the line emission ( which can be inferred from the resolved profile shape ) and the presence of rapid line variability . if the bulk of the line is emitted in the inner accretion disk ( as implied by the broad line profiles in mcg6 - 30 - 15 and ngc 3516 ) , then the continuum and line variations should be correlated on time scales comparable to the light crossing time of the inner accretion disk .
this is because the reprocessing times are likely very short compared to the light travel times , which dominate the correlation .
however , the time bins used in this paper are typically at least two orders of magnitude longer than the expected light crossing time scale of the putative inner accretion disk .
this means that any variations on this time scale would be grossly undersampled .
if the bulk of the line is emitted far from the central corona ( e.g. , as implied by the narrow line profiles for ngc 4151 and ngc 5548 ) , then one would expect the line variations to track the continuum variations only on sufficiently long time scales ( e.g. , longer than days
weeks if the line is emitted in a region coincident with the broad line region ; kaspi et al .
for instance , the presence of similar long - term trends in the continuum and line in ngc 5548 supports a model in which the bulk of the line emission originates far from the central corona in this object .
more specifically , it is consistent with the prediction of yaqoob et al .
( 2001 ) that the flux of the narrow core in ngc 5548 , estimated to originate in a region consistent with the broad line region in this object , should track the continuum flux on time scales of @xmath15 @xmath1100 days
. however , such trends are not ubiquitous . for fe k@xmath0 lines originating far from the central corona
, one would expect to see an increase in number or strength of continuum line correlations as one goes towards towards long time scales . however , there is no evidence from the current sample to support this .
the lack of correlation between the continuum and the line suggests that modification to this straightforward picture is required .
one possible explanation for the reduced variability in the line relative to the continuum is that a portion of line emission originating in the inner disk is absent or suppressed .
the inner portions of the accretion disk may be truncated , highly ionized , or radiatively inefficient ( e.g. , as in an advection - dominated accretion flow ; narayan & yi 1994 ) .
geometry may play a role : a relatively more concave disk , for instance , can reprocess a larger fraction of continuum photons in its outer regions ( blackman 1999 ) . for cases in which the bulk of the observed line flux originates in the outer disk , the direct response of the line peak to rapid continuum variations is reduced .
another possibility is that if the x - ray continuum emission is produced in a corotating flare some height above the reflecting disk , then the orbital motions of the flare and disk can also lead to complex temporal behavior in the observed line flux ( ruszkowski 2000 ) .
fabian & vaughan ( 2003 ) suggested that light - bending near the black hole can lead to substantial de - coupling of the observed continuum and line fluxes .
this happens because much of the emission from the x - ray source is focussed on to the disc whereas only a small fraction is able to reach the observer .
in such a scenario small changes in the geometry of the x - ray source can lead to relatively large changes in the amount of continuum radiation reaching the observer ( hence the strong continuum variations ) but relatively little change in the total amount of radiation illuminating the disk .
another intriguing possibility relies on the presence of an ionized skin on the disk surface .
thermal instabilities can lead to formation of such an ionized skin ( ross , fabian & young 1999 ; nayakshin , kazanas & kallman 2000 ) with most of the observed neutral line emission originating in the colder layers beneath .
the ionized skin can scatter incoming continuum photons and outgoing line photons , affecting the observed neutral line flux , as discussed by nayakshin ( 2000 ) and nayakshin & kallman ( 2001 ) .
however , changes in the ionization state of the skin are linked to the dynamic time scale for readjustment of hydrostatic equilibrium ( nayakshin & kazanas 2002 for a lamppost illuminating model ; see also collin et al .
2003 for a multiple - flare illuminating model ) .
this time scale is usually an order of magnitude or more larger than the light - crossing time ; this can be @xmath38minutes
hours if most of the line emission originates in the inner regions of a typical thin accretion disk ( e.g. , around a 10@xmath39 @xmath4 black hole ) .
the observed line flux hence may not respond to continuum variations shorter than this time scale , resulting in decorrelation of the observed continuum and line flux variations .
furthermore , the ionization state of the skin depends on the high - energy cutoff of the power - law continuum ( nayakshin & kallman 2001 ) ; this could be variable , but it is difficult to constrain with _
rxte_. interestingly enough , such a variation in this parameter could lead to a change in fe k@xmath0 line flux even without a change in 210 kev continuum flux .
however , recent studies have not been able to directly link the line flux behavior with the ionization state of the disk , e.g. , ballantyne , vaughan & fabian ( 2003 ) .
all of the above scenarios are still consistent with the prediction that both broad and narrow fe k@xmath0 lines should eventually track continuum variations on time scales of months to years .
the finding that the line decreased along with the continuum in the long time scale observation of ngc 5548 is consistent with this idea ; however , such correlated behavior is not seen throughout the entire sample . for line emission originating in the outer disk or molecular torus , ionization in the reflecting material
is negligible due to the low x - ray continuum flux .
however , other processes may be relevant for decorrelating long - term continuum and line variations .
for instance , it may be possible for an ionized wind ( e.g. , george et al .
1998 ) near the site of reflection to mimic an ionized skin in these regions .
most of these objects show a strong correlation with no measureable lag between 210 kev continuum flux and photon index .
additionally , @xmath2 spectra show that in nearly all the objects , the soft - spectrum ( narrow line ) source akn 564 being an exception , variability amplitudes tend to increase towards relatively softer energies .
these observations show that the broadband emission steepens as it brightens , in line with numerous similar observations ( e.g. , magdziarz et al .
1998 , done et al .
2000 ) .
the time - resolved spectral fitting of mcg6 - 30 - 15 performed from _ asca _
data by shih , iwasawa & fabian ( 2002 ) and from _ xmm - newton _ data by fabian & vaughan ( 2003 ) point towards a scenario in which flux - correlated changes in the overall 210 kev continuum slope are explained by the superposition of a soft , variable power - law likely associated with coronal emission and a harder , much less variable spectral component likely associated with the compton reflection hump .
taylor , uttley & mchardy ( 2003 ) use flux flux plots to show that such a description of spectral variability works in mcg6 - 30 - 15 and ngc 3516 . specifically , in this model , the intrinsic slope of the underlying power - law does not vary significantly ; rather , it is the _ relative contributions _ of the two spectral components that change the overall spectral shape as observed .
this model does not require physical properties of the power - law emitting corona such as electron temperature , optical depth , or flare size to vary significantly .
furthermore , this two - component model may have relevance to explaining the observed fe k@xmath0 variability . as the fe k@xmath0 line and compton hump are expected to originate together , a relatively constant compton hump
could be associated with variability in the fe k@xmath0 line that is reduced relative to the continuum variability , as observed ( e.g. , shih et al .
2002 ) .
alternate models to explain the observed spectral variability rely on flux - correlated changes in the power - law slope of the coronal component . in the context of popular thermal comptonization corona models ( haardt , maraschi & ghisellini 1997 , poutanen & fabian 1999 ) , a flux photon index correlation can arise if the x - ray variability is due to changes in the input seed photon population .
in contrast to the two - component model above , some physical property of the corona may vary with the 210 kev x - ray flux in these types of models .
the @xmath2 spectra of the soft - spectrum ( narrow - line ) seyfert 1 galaxy akn 564 display a much flatter dependence on energy in the 3.59 kev bandpass ( and may even be slightly increasing with energy on the long time scales ) relative to the other objects , most of which are hard - spectrum ( broad - line ) seyfert 1 galaxies .
this has been seen previously in akn 564 by edelson et al .
( 2002 ) as well as in other soft - spectrum sources ( vaughan et al . 2002 ) .
this could imply that , in the context of the spectral - pivoting model above , the broadband spectral shape changes very little in soft - spectrum sources , perhaps due to the lack of variation of some physical parameter that does vary in broad - line seyfert 1 galaxies .
alternatively , if the two - component model above is relevant , it is possible that soft - spectrum sources have much weaker hard / constant components , leading to relatively little change in the overall spectral shape with flux .
overall , however , it is difficult to present a simple , unified phemonological model to simultaneously explain the dependence on energy of the variability in broad - line seyfert 1 galaxies and lack thereof in narrow - line seyfert 1 galaxies .
time - resolved broadband spectral fitting has been systematically performed on near - continuous _ rxte _ monitoring data of seven seyfert 1 galaxies in order to probe the behavior of emission components on time scales of both @xmath1days to @xmath1weeks and @xmath1months to @xmath1years . fe k@xmath0 line variability is present at a variety of levels .
it is readily detectable in some observations ; for instance , the average peak - to - trough variations are a factor of @xmath12 and as high as @xmath13 in three cases ( the short - term observation mcg6 - 30 - 15 ( ii ) and the long - term observations of ngc 4151 and ngc 5548 ) . in other observations ,
the line does not display appreciably variability ; peak - to - trough variations are only @xmath14050@xmath11 .
the measured values of @xmath2 indicate that a higher fraction of the objects sampled show detectable line variability towards longer time scales .
of particular interest are systematic long - term decreases in line flux over durations of 34 years in ngc 3516 , ngc 4151 , and most noteably ngc 5548 .
furthermore , the fractional variability amplitude of the line increases towards longer time scales in many objects , `` tracking '' the red - noise continuum .
however , fractional variability amplitudes show that the line does not vary as strongly as the continuum .
@xmath2 spectra also generally show a dilution in @xmath157 kev continuum variability due to the relatively less variable fe k@xmath0 line . in ngc 5548
, the continuum and line simultaneously decrease by similar factors over a 4-year period ; the long - term observations of ngc 3516 and ngc 4151 also show somewhat similar trends in continuum and line flux . for the sample as a whole
, however , there appears to be no simple temporal relation between the continuum and the line , contradicting simple models in which the line flux is driven solely by continuum variations .
local effects such as an ionized skin may be relevant in decorrelating the continuum and line variations , particularly for lines originating in the inner accretion disk .
the strongest statement that can be made on the basis of these results is that for most of the seyfert 1 galaxies studied , the fe k@xmath0 emission is less variable than the continuum , and in those cases where there does appear to be evidence for changes in the line flux , these changes are not well correlated with the continuum .
there is evidence for a correlation with no measureable lag between continuum flux and measured photon index on both time scales in many sources .
relatively less model - dependent @xmath2 spectra generally show the broadband continuua to be more variable at relatively softer energies , supporting seyfert 1 galaxies softening as they brighten , though the soft - spectrum source akn 564 , with its flat @xmath2 spectra , is an exception .
the authors acknowledge the dedication of the entire _ rxte _ mission team .
the authors thank sergei nayakshin for suggestions and discussion on ionized skin models , as well as the referee for useful comments . this work 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 tartarus database , which is supported by jane turner and kirpal nandra under nasa grants nag 5 - 7385 and nag 5 - 7067 , and the nasa@xmath40ipac extragalactic database which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
a.m. & r.e .
acknowledge financial support from nasa grant nag 5 - 9023 .
@xmath41 3c 390.3 : the short time scale time - resolved and time - average fit results are in reasonable agreement with gliozzi et al .
( 2003 ) : the light curve for @xmath19 has absolute values slightly systematically lower by @xmath120.1 than gliozzi et al . (
2003 ) but the variability patterns are identical . on short time
scales , a decrease in f@xmath36 by a factor of @xmath12 over a span of 30 days does not appear to have any simultaneous effect on f@xmath21 , which remains nearly constant during the observation . on long time scales ,
there is a doubling in f@xmath36 over 300 days , and an increase of @xmath150@xmath11 in f@xmath21 over approximately the same period .
@xmath41 mcg6 - 30 - 15 : due to improved background models , the current analysis yields a lower absolute value for @xmath20 for the integrated spectrum of the short time scale observation mcg6 - 30 - 15
( i ) than in ve01 ( values of @xmath20 obtained from spectral fitting are highly sensitive to the shape of the modelled background spectrum above @xmath110 kev ) , but the time - resolved spectral analysis and the cross - correlation analysis on the short time scale observation mcg6 - 30 - 15 ( i ) is in excellent agreement with ve01 .
f@xmath21 shows variability on short time scales , as well as in the long time scale observation mcg6 - 30 - 15 ( i ) .
the long time scale observation mcg6 - 30 - 15 ( ii ) does not yield a significant detection of variability in the line ; one possibility is that on these time scales probed ( @xmath1100 d to @xmath12000 d ) the variability is `` averaged out '' to some degree .
there are examples in these observations of large increases or decreases in f@xmath36 ( in some cases by a factor of @xmath12 ) that do not have any obvious immediate effect on f@xmath21 .
the 6.4 kev dips in the @xmath2 spectra are prominent , especially for the short time scale observation mcg6 - 30 - 15 ( i ) .
the four observations , when considered in order of increasing duration , do not show any obvious pattern to the fe k@xmath0 line variability .
@xmath41 ngc 3516 : on short time scales , f@xmath36 and f@xmath21 exhibit uncorrelated variability .
persistent trends in f@xmath36 such as the rapid decrease from modified julian day ( mjd ) 5055050554 are not simultaneously matched by similar behavior in f@xmath21 .
rapid changes in f@xmath21 occur without sudden changes in f@xmath36 , e.g. , the decrease in f@xmath21 by @xmath120@xmath11 at mjd 50551 and also the increase in f@xmath21 by @xmath130@xmath11 at mjd 50558 as illustrated in figure 7 .
the 6.4 kev dip in the short time scale @xmath2 spectra is especially noticeable . on long time scales , both f@xmath36 and f@xmath21
show an overall decrease by a factor of @xmath12 over the 1000-day duration .
@xmath41 ngc 3783 : formally , f@xmath21 does not demonstrate appreciable variability on either time scale .
relatively rapid changes in f@xmath36 by @xmath3810@xmath11 towards the middle of the short time scale observation and @xmath120@xmath11
@xmath140@xmath11 towards the beginning and end of the long time scale observation do not appear to have any major simultaneous impact on f@xmath21 .
@xmath41 ngc 4151 : on short time scales , f@xmath21 is not formally variable .
one note of interest , for instance , is the doubling of f@xmath36 during the last 4 days of the short time scale observation while f@xmath21 remains virtually constant during that time .
the 6.4 kev dip in the short time scale @xmath2 spectra is especially noticeable . on the long time scale , f@xmath21 is much more variable .
there are similar decreasing trends by a factor of @xmath13 in both the continuum and line over the whole 800-day duration . of note
is the sudden dip and rise in the line around mjd 51620 in the long time scale observation while f@xmath36 remains nearly constant ; figure 7 demonstrates the dip .
@xmath41 ngc 5548 : on long time scales , a gradual decreasing trend by a factor of 34 dominates the behavior of both f@xmath36 and f@xmath21 .
the fe k@xmath0 line variability amplitude is substantially larger on the long time scale than the short due to this dominating long time scale trend .
@xmath41 ngc 7469 : all the time - average and time - resolved spectral fit results and cross - correlation results are in reasonable agreement with nandra et al .
the current analysis yields absolute values of @xmath19 that are slightly systematically lower by @xmath120.1 and the absolute values of f@xmath21 obtained in the current analysis are systematically @xmath130@xmath11 higher compared to nandra et al .
( 2000 ) , but the variability patterns are consistent .
arnaud , k. 1996 , in _ astronomical data analysis software and systems _ , jacoby , g. , barnes , j. , eds .
, asp conf .
series vol . 101 , p. 17
balalntyne , d.r . ,
vaughan , s. & fabian , a.c .
2003 , mnras , 342 , 239 blackman , e. 1999 , mnras , 306 , l25 blustin , a. et al . 2002 , a&a , 392 , 453 blustin , a. et al .
2003 , a&a , 403 , 473 chiang , j. et al . 1999 ,
apj , 528 , 292 collin , s. et al .
2003 , a&a , 400 , 437 done , c. et al .
2000 , apj , 536 , 213 edelson , r. & krolik , j. , 1988 , , 333 , 646 edelson , r. & nandra , k. 1999 , apj , 514 , 682 edelson , r. , griffiths , g. , markowitz , a. , sembay , s. , turner , m.j.l . ,
warwick , r. , 2001 , apj , 554 , 274 edelson , r. , turner , j. , pounds , k. , vaughan , s. , markowitz , a. , marshall , h. , dobbie , p. , warwick , r. 2002 , apj , 568 , 610 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 . , vaughan , s. , nandra , k. , iwasawa , k. , ballantyne , d.r . ,
lee , j.c . , de rosa , a. , turner , a. , & young , a. j. 2002 , mnras , 335 , l1 fabian , a. & vaughan , s. 2003 , mnras , 340 , l28 george , i.m . &
fabian , a.c .
1991 , mnras , 249 , 352 george , i.m . ,
turner , t.j . ,
netzer , h. , nandra , k. , mushotzky , r. & yaqoob , t. 1998 , apjs , 114 , 73 gliozzi , m. , sambruna , r. , eracleous , m. 2003 , apj , 584 , 176 guilbert , p. & rees , m. , 1988 , mnras , 233 , 475 haardt , f. , maraschi , l. & ghisellini , g. 1994 , , 432 , l95 haardt , f. , maraschi , l. & ghisellini , g. 1997 , apj , 476 , 620 iwasawa , k. et al .
1996 , mnras , 282 , 1038 iwasawa , k. et al . 1999 ,
mnras , 306 , l19 kaspi , s. et al .
2000 , apj , 533 , 631 kaspi , s. et al .
2001 , apj , 554 , 216 kaspi , s. et al .
2002 , apj , 574 , 643 laor , a. 1991 , apj , 376 , 90 lee , j.c . ,
fabian , a.c . ,
brandt , w.n .
, reynolds , c.s . , iwasawa , k. 1999 , mnras , 310 , 973 lee , j. c. , fabian , a.c . , brandt , w.n . , reynolds , c.s . ,
iwasawa , k. 2000 , mnras , 318 , 857 magdziarz , p. & zdziarski , a. 1995 , mnras , 273 , 837 magdziarz , p. et al .
1998 , mnras , 301 , 179 maoz , d. , edelson , r. & nandra , k. , 2000 , aj , 119 , 119 maoz , d. , markowitz , a. , edelson , r. & nandra , k. , 2002 , aj , 124 , 1988 markowitz , a. & edelson , r. 2001 , apj , 547 , 684 markowitz , a. et al . 2003 ,
apj , 593 , 96 matt , g. & perola , c. 1992 , mnras , 259 , 433 nandra , k. , pounds , k. & stewart , g. 1990 , mnras 242 , 660 nandra , k. , george , i.m . , mushotzky , r. , turner , t.j . ,
yaqoob , t. 1997 , apj , 476 , 70 nandra , k. , george , i.m .
, mushotzky , r. , turner , t.j . ,
yaqoob , t. 1999 , , 523 , l17 nandra , k. , le , t. , george , i.m . , edelson , r. , mushotzky , r. , peterson , b. , turner , t.j .
2000 , , 544 , 734 nandra , k. 2001 , in `` workshop on x - ray spectroscopy of agn with chandra and xmm - newton , '' mpe report 279 , p. 35
narayan , r. & yi , i. 1994 , apjl , 428 , l13 obrien , p. et al .
2001 , in `` workshop on x - ray spectroscopy of agn with chandra and xmm - newton , '' mpe report 279 , p. 77
nayakshin , s. 2000 , apj , 540 , l37 nayakshin , s. & kallman , t. 2001 , apj , 546 , 406 nayakshin , s. , kazanas , d. & kallman , t. 2000 , apj , 537 , 833 nayakshin , s. & kazanas , d. , 2002 , apj , 567 , 85 nicastro et al .
2000 , apj , 536 , 718 peterson , b. et al .
1998 , , 110 , 660 petrucci , p. et al .
2002 , a&a , 388 , l5 pounds , k. et al .
2003 , mnras , 341 , 953 poutanen , j. & fabian , a.c .
1999 , mnras , 306 , l31 reeves , j. 2002 , in _ active galactic nuclei : from central engine to host galaxy _ ,
eds . : s. collin , f. combes and i. shlosman , pasp ( astro - ph/0211381 ) reynolds , c.s .
1997 , apj , mnras , 286 , 513 reynolds , c.s .
2000 , apj , 533 , 811 ross , r. , fabian , a.c . &
young , a. 1999 , mnras , 306 , 461 ruszkowski , m. 2000 , mnras , 315 , 1 schurch , n. & warwick , r. 2002 , mnras , 334 , 811 schurch , n. , warwick , r. , griffiths , r. & ptak , a. 2002 , in ` new visions of the universe in the xmm - newton and chandra era ' conference , astro - ph/0202093 shih , d.c . , iwasawa , k. & fabian , a.c .
2002 , mnras , 333 , 687 stella , l. , 1990 , nature , 344 , 747 swank , j. 1998 , in nuclear phys .
suppl . ) : the active x - ray sky : results from bepposax and rossi - xte , rome , italy , 1997 october 21 - 24 , eds .
l. scarsi , h. bradt , p. giommi , & f. fiore , nucl .
b suppl .
( the netherlands : elsevier science b.v . ) , 69 , 12 tanaka , y. , nandra , k. , fabian , a.c . , inoue , h. , otani , c. , dotani , t. , hayashida , k. , iwasawa , k. , kii , t. , kunieda , h. , makino , f. , & matsuoka , m. 1995 , nature , 375 , 659 taylor , r. , uttley , p. , mchardy , i. 2003 , mnras , in press ( astro - ph/0304523 ) turner , t.j .
2002 , apj , 574 , l123 uttley , p. , mchardy , i. & papadakis , i. 2002 , mnras , 332 , 231 vaughan , s. , pounds , k. , reeves , j. , warwick , r. , edelson , r. 1999 , mnras , 304 , l34 vaughan , s. & edelson , r. , 2001 , , 548 , 694 vaughan , s. , boller , th . ,
fabian , a. , ballantyne , d.r .
, brandt , w. n. , trmper , j. , 2002 , mnras , 337 , 247 vaughan , s. , edelson , r. , warwick , r. & uttley , p. 2003 , mnras , submitted weaver , k. , gelbord , j. & yaqoob , t. 2001 , apj , 550 , 261 welsh , w. 1999 , pasp , 111 , 1347 white , r. & peterson , b.m .
1994 , pasp , 106 , 879 wilms , j. et al .
2001 , mnras , 328 , l27 yaqoob , t. , serlemitsos , p.j . ,
turner , t.j . , george , i.m .
& nandra , k. 1996 , apj , 470 , l27 yaqoob , t. et al .
2001 , , 546 , 759 yaqoob , t. , george , i.m . &
turner , t.j .
2001 , in proceedings of `` high energy universe at sharp focus : chandra science '' , eds .
eric m. schlegel and saequ vrtilek ( astro - ph/0111428 ) lccccccc 3c 390.3 & 44.26 & 0.056 & 50220.650278.0 & 2.3 & 6.0 d & 10 + mcg6 - 30 - 15 i & 42.80 & 0.008 & 50664.150672.5 & 4.1 & 0.13 d & 63 + mcg6 - 30 - 15 ii & 42.83 & 0.008 & 51622.751676.0 & 4.4 & 1.6 d & 34 + ngc 3516 & 42.90 & 0.009 & 50523.050657.4 & 4.2 & 6.4 d & 21 + ngc 3783 & 43.12 & 0.010 & 51960.151980.0 & 5.7 & 1.7 d & 12 + ngc 4151 & 42.24 & 0.003 & 51870.651904.9 & 7.4 & 1.4 d & 24 + ngc 5548 & 43.52 & 0.017 & 52091.652126.7 & 4.5 & 3.2 d & 11 + ngc 7469 & 43.25 & 0.016 & 50244.050275.7 & 2.8 & 1.3 d & 25 + 3c 390.3 & 44.40 & 0.056 & 51186.051964.8 & 3.2 & 51.9 d & 15 + mcg6 - 30 - 15 i & 42.81 & 0.008 & 51926.752252.9 & 4.2 & 23.4 d & 13 + mcg6 - 30 - 15 ii & 42.85 & 0.008 & 50159.852252.9 & 4.6 & 128.7 d & 16 + ngc 3516 & 42.86 & 0.009 & 50523.051593.4 & 3.8 & 53.5 d & 20 + ngc 3783 & 43.22 & 0.010 & 51180.552375.2 & 7.1 & 59.8 d & 20 + ngc 4151 & 42.55 & 0.003 & 51179.651964.6 & 14.9 & 43.6 d & 18 + ngc 5548 & 43.55 & 0.017 & 51183.952643.1 & 4.8 & 104.2
d & 14 + lccccccc 3c 390.3 & 4.28 & [email protected] & 6.00@xmath42 & [email protected] & [email protected] & @xmath6 0.03 & 61.0/34 + mcg6 - 30 - 15 i & 4.09 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 292.0/33 + mcg6 - 30 - 15 ii & 4.09 & [email protected] & 6.17@xmath43 & [email protected] & [email protected] & [email protected] & 23.0/27 + ngc 3516 & 2.94 & [email protected] & [email protected] & 0.59@xmath44 & 2.62@xmath45 & 0.20@xmath46 & 487.6/33 + ngc 3783 & 8.26 & 1.666@xmath47 & 6.27@xmath48 & [email protected] & [email protected] & [email protected] & 22.3/28 + ngc 4151 & 616 & [email protected] & [email protected] & 0.14@xmath48 & [email protected] & [email protected] & 101.0/26 + ngc 5548 & 1.75 & [email protected] & [email protected] & [email protected] & 0.73@xmath49 & [email protected] & 25.3/28 + ngc 7469 & 4.86 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 146.0/33 + 3c 390.3 & 4.28 & 1.646@xmath50 & [email protected] & 0.25@xmath51 & [email protected] & 0.03@xmath52 & 122.0/97 + mcg6 - 30 - 15 i & 4.09 & 2.020@xmath53 & 5.92@xmath49 & 0.75@xmath54 & [email protected] & 1.96@xmath55 & 30.75/28 + mcg6 - 30 - 15 ii & 4.09 & [email protected] & [email protected] & 0.59@xmath44 & [email protected] & 1.17@xmath56 & 726.9/98 + ngc 3516 & 2.94 & 1.642@xmath57 & [email protected] & 0.47@xmath58 & [email protected] & [email protected] & 1009.7/65 + ngc 3783 & 8.26 & 1.693@xmath59 & [email protected] & [email protected] & [email protected] & 0.47@xmath56 & 206.6/98 + ngc 4151 & 812 & 1.692@xmath60 & 6.40@xmath61 & 0.12@xmath62 & 3.72@xmath63 & 0.55@xmath64 & 581.6/94 + ngc 5548 & 1.75 & [email protected] & 6.28@xmath43 & 0.36@xmath65 & 0.76@xmath66 & 0.41@xmath67 & 105.0/99 + lcccccccc 3c 390.3 & 4.28 & 1.637@xmath68 & [email protected] & 3.8@xmath69 & 13@xmath70 & 0.68@xmath71 & @xmath6 0.02 & 55.6/33 + mcg6 - 30 - 15 i & 4.09 & 1.922@xmath72 & 6.47@xmath43 & [email protected] & 2.2@xmath73 & 2.76@xmath74 & [email protected] & 98.2/32 + mcg6 - 30 - 15 ii & 4.09 & 1.932@xmath75 & [email protected] & 3.6@xmath76 & 13@xmath77 & [email protected] & [email protected] & 19.6/25 + ngc 3516 & 2.94 & [email protected] & 6.48@xmath78 & 8.0@xmath79 & 7.1@xmath80 & [email protected] & @xmath6 0.03 & 101.5/32 + ngc 3783 & 8.26 & [email protected] & [email protected] & 2.5@xmath81 & 5.8@xmath82 & 1.95@xmath83 & 0.57@xmath84 & 20.4/27 + ngc 4151 & 618 & 1.524@xmath85 & [email protected] & 3.0@xmath86 & 300@xmath87 & 3.43@xmath88 & [email protected] & 101.0/25 + ngc 5548 & 1.75 & 1.725@xmath89 & 6.43@xmath90 & 2.8@xmath91 & 8.7@xmath92 & 0.78@xmath93 & 0.40@xmath54 & 24.5/27 + ngc 7469 & 4.86 & [email protected] & [email protected] & [email protected] & [email protected] & 0.79@xmath94 & 0.38@xmath95 & 117.4/32 + 3c 390.3 & 4.28 & 1.653@xmath96 & [email protected] & 8.2@xmath97 & 260@xmath98 & 0.44@xmath43 & 0.06@xmath99 & 122.1/96 + mcg6 - 30 - 15 i & 4.09 & 1.975@xmath100 & [email protected] & 9.9@xmath101 & 8.9@xmath102 & 1.70@xmath103 & 1.59@xmath104 & 27.1/27 + mcg6 - 30 - 15 ii & 4.09 & 1.919@xmath105 & 6.43@xmath44 & [email protected] & 1.8@xmath106 & 2.64@xmath107 & 0.81@xmath108 & 457.0/97 + ngc 3516 & 2.94 & [email protected] & [email protected] & [email protected] & 2.8@xmath109 & 2.73@xmath110 & [email protected] & 766.8/64 + ngc 3783 & 8.26 & 1.664@xmath111 & 6.43@xmath112 & 2.9@xmath113 & [email protected] & 2.37@xmath114 & [email protected] & 185.4/97 + ngc 4151 & 690 & 1.604@xmath115 & [email protected] & 0.0@xmath116 & 230@xmath117 & 4.04@xmath109 & 0.54@xmath43 & 450.5/91 + ngc 5548 & 1.75 & [email protected] & [email protected] & 8.6@xmath118 & 44@xmath119 & 0.73@xmath48 & 0.40@xmath120 & 120.5/97 + lcccccc 3c 390.3 & 5.8 & 1.2e-10 & 1.80 & 3.7 e-2 & n / a & n / a + mcg6 - 30 - 15 i & 1.78 & 1.9e-4 & 1.63 & 1.4e-3 & 0.76 & 0.92 + mcg6 - 30 - 15 ii & 2.4 & 1.6e-5 & 1.70 & 5.7e-3 & 1.31 & 0.11 + ngc 3516 & 24.5 & @xmath61@xmath1510@xmath121 & 3.20 & 2.9e-6 & 0.39 & 0.99 + ngc 3783 & 0.81 & 0.64 & 0.72 & 0.74 & 0.53 & 0.90 + ngc 4151 & 1.94 & 4.8e-3 & 0.82 & 0.71 & 1.04 & 0.41 + ngc 5548 & 1.5 & 0.26 & 0.99 & 0.45 & 0.47 & 0.92 + ngc 7469 & 3.80 & 1.7e-9 & 2.05 & 1.9e-3 & 1.58 & 3.5e-2 + 3c 390.3 & 5.84 & 3.2e-11 & 0.94 & 0.52 & 0.65 & 0.84 + mcg6 - 30 - 15 i & 2.21 & 2.9e-2 & 1.65 & 4.4e-2 & 1.80 & 2.3e-2 + mcg6 - 30 - 15 ii & 2.90 & 1.3e-4 & 2.29 & 2.9e-3 & 0.86 & 0.62 + ngc 3516 & 13.0 & 2.3e-36 & 4.83 & 3.8e-11 & 1.08 & 0.36 + ngc 3783 & 1.17 & 0.28 & 1.03 & 0.43 & 0.53 & 0.95 + ngc 4151 & 12.3 & 5.2e-31 & 7.74 & 2.2e-18 & 1.76 & 2.7e-2 + ngc 5548 & 1.77 & 4.11e-2 & 1.83 & 3.2e-2 & 0.85 & 0.61 + lccc 3c 390.3 & 2.58 @xmath14 0.05 & 1.633 @xmath14 0.023 & 0.639 @xmath14
0.174 + mcg6 - 30 - 15 i & 4.73 @xmath14 0.05 & 2.005 @xmath14 0.063 & 1.829 @xmath14 0.250 + mcg6 - 30 - 15
ii & 4.58 @xmath14 0.10 & 1.914 @xmath14 0.103 & 1.384 @xmath14 0.261 + ngc 3516 & 4.80 @xmath14 0.08 & 1.609 @xmath14 0.061 & 2.384 @xmath14 0.177 + ngc 3783 & 6.15 @xmath14 0.14 & 1.665 @xmath14 0.028 &
1.768 @xmath14 0.200 + ngc 4151 & 10.11 @xmath14 0.19 & 1.519 @xmath14 0.037 & 3.466 @xmath14 0.264 + ngc 5548 & 4.85 @xmath14 0.13 & 1.723 @xmath14 0.028 & 0.724 @xmath14 0.149 + ngc 7469 & 3.24 @xmath14 0.05 & 1.845 @xmath14 0.034 & 0.642 @xmath14 0.096 + 3c 390.3 & 3.44 @xmath14 0.08 & 1.644 @xmath14 0.044 & 0.453 @xmath14
0.066 + mcg6 - 30 - 15 i & 5.10 @xmath14 0.14 & 2.002 @xmath14 0.080 & 1.788 @xmath14 0.253 + mcg6 - 30 - 15 ii & 5.41 @xmath14 0.10 & 2.014 @xmath14 0.038 & 1.801 @xmath14 0.226 + ngc 3516 & 4.31 @xmath14 0.08 & 1.583 @xmath14 0.061 & 1.953 @xmath14 0.269 + ngc 3783 & 7.56 @xmath14 0.12 & 1.689 @xmath14 0.037 & 2.080 @xmath14 0.202 + ngc 4151 & 22.23 @xmath14 0.24 & 1.616 @xmath14 0.055 & 4.878 @xmath14 0.795 + ngc 5548 & 4.12 @xmath14 0.10 & 1.711 @xmath14 0.041 & 0.750 @xmath14 0.104 + lcccc 3c 390.3 & short & 29.3 @xmath14 0.6 & undef .
+ & long & 26.9 @xmath14 0.6 & 13.7 @xmath14 4.7 + mcg6 - 30 - 15 & short i & 18.5 @xmath14 0.1 & 9.4 @xmath14 2.5 + & short ii & 17.4 @xmath14 0.4 & 14.4 @xmath14 4.4
+ & long i & 16.1 @xmath14 0.7 & 14.1 @xmath14 4.8 + & long ii & 7.0 @xmath14 0.5 & undef .
+ ngc 3516 & short & 24.6 @xmath14 0.4 & 5.6 @xmath14 2.2 + & long & 32.9 @xmath14 0.4 & 14.9 @xmath14 3.7 + ngc 3783 & short & 4.7 @xmath14 0.7 & 6.0 @xmath14 5.5 + & long & 10.0 @xmath14 0.4 & 1.5 @xmath14 9.9 + ngc 4151 & short & 16.7 @xmath14 0.4 & 3.2 @xmath14 3.1 + & long & 30.5 @xmath14 0.2 & 24.0 @xmath14 4.3 + ngc 5548 & short & 24.8 @xmath14 0.8 & 8.2 @xmath14 12.6 + & long & 39.0 @xmath14 0.7 & 28.8 @xmath14 3.9 + ngc 7469 & short & 13.3 @xmath14 0.3 & undef .
+ lccccc 3c 390.3 & short & + 12.0 & 0.46 & + [email protected] & 0.61 + mcg6 - 30 - 15 i & & + 1.3 & 0.51 & + [email protected] & 0.53 + mcg6 - 30 - 15 ii & & 14.4 & 0.38 & [email protected] & 0.38 + ngc 3516 & & + 6.4 & 0.17 & + [email protected] & 0.18 + ngc 3783 & & + 3.3 & 0.59 & + [email protected] & 0.72 + ngc 4151 & & + 2.8 & 0.23 & + [email protected] & 0.27 + ngc 5548 & & + 6.4 & 0.63 & + [email protected] & 0.68 + ngc 7469 & & 0 & 0.57 & + [email protected] & 0.67 + 3c 390.3 & long & + 51.9 & 0.48 & + 25.9@xmath14159 & 0.48 + mcg6 - 30 - 15 i & & 0 & 0.75 & [email protected] & 0.73 + mcg6 - 30 - 15 ii & & 386 & 0.45 & 0@xmath14606 & 0.45 + ngc 3516 & & 161 & 0.66 & 134@xmath14262 & 0.71 + ngc 3783 & & 0 & 0.53 & 0@xmath14388 & 0.51 + ngc 4151 & & 0 & 0.56 & 0@xmath14258 & 0.61 + ngc 5548 & & 0 & 0.84 & 0@xmath14428 & 0.87 + lccccc 3c 390.3 & short & 0 & 0.91 & [email protected] & 0.92 + mcg6 - 30 - 15 i & & 0 & 0.82 & [email protected] & 0.82 + mcg6 - 30 - 15 ii & & 0 & 0.54 & [email protected] & 0.57 + ngc 3516 & & 0 & 0.90 & [email protected] & 0.90 + ngc 3783 & & 0 & 0.88 & [email protected] & 0.89 + ngc 4151 & & 0 & 0.69 & [email protected] & 0.73 + ngc 5548 & & 9.6 & 0.85 & [email protected] & 0.88 + ngc 7469 & & 3.8 & 0.64 & [email protected] & 0.73 + 3c 390.3 & long & 0 & 0.85 & 0@xmath14136 & 0.82 + mcg6 - 30 - 15 i & & 0 & 0.94 & [email protected] & 0.86 + mcg6 - 30 - 15 ii & & 0 & 0.62 & 0@xmath14555 & 0.64 + ngc 3516 & & 0 & 0.92 & 0@xmath14161 & 0.93 + ngc 3783 & & 0 & 0.81 & 0@xmath14357 & 0.81 + ngc 4151 & & 0 & 0.31 & + 21.8@xmath14191 & 0.46 + ngc 5548 & & 0 & 0.74 & 52.1@xmath14506 & 0.75 + lccccc 3c 390.3 & short & + 12.0 & 0.64 & + [email protected] & 0.66 + mcg6 - 30 - 15 i & & + 1.5 & 0.49 & + [email protected] & 0.54 + mcg6 - 30 - 15 ii & & + 9.6 & 0.38 & + [email protected] & 0.44 + ngc 3516 & & 32.0 & 0.25 & [email protected] & 0.33 + ngc 3783 & & + 3.3 & 0.70 & + [email protected] & 0.74 + ngc 4151 & & + 4.2 & 0.28 & + [email protected] & 0.33 + ngc 5548 & & + 9.6 & 0.43 & + [email protected] & 0.47 + ngc 7469 & & + 5.7 & 0.40 & + [email protected] & 0.45 + 3c 390.3 & long & + 51.9 & 0.52 & + 25.9@xmath14156 & 0.56 + mcg6 - 30 - 15 i & & 0 & 0.67 & [email protected] & 0.63 + mcg6 - 30 - 15 ii & & 386 & 0.47 & 386@xmath14667 & 0.44 + ngc 3516 & & 107 & 0.79 & 134@xmath14300 & 0.82 + ngc 3783 & & 0 & 0.47 & 0@xmath14390 & 0.46 + ngc 4151 & & 43.6 & 0.19 & 174@xmath14280 & 0.17 + ngc 5548 & & 0 & 0.55 & + 52.1@xmath14481 & 0.61 + lccccc 3c 120 & 50458.550515.3 & 2.0 & 29 + 3c 390.3 & 50220.650278.0 & 2.0 & 29 + akn 564 & 51694.851726.7 & 1.6 & 20 + fairall 9 & 52144.952179.0 & 1.6 & 21 + ic 4329a & 50665.850723.5 & 2.0 & 28 + mcg6 - 30 - 15 i & 50664.150672.5 & 0.13 & 63 + mcg6 - 30 - 15 ii & 51622.751676.0 & 1.6 & 34 + ngc 3516 & 50523.050657.4 & 4.8 & 28 + ngc 3783 & 51960.151980.0 & 0.66 & 30 + ngc 4151 & 51870.651904.9 & 1.2 & 29 + ngc 5548 & 52091.652126.7 & 1.3 & 27 + ngc 7469 & 50244.050275.7 & 1.1 & 30 + 3c 120 & 50812.451259.7 & 20.4 & 19 + 3c 390.3 & 51259.751676.0 & 20.8 & 20 + akn 120 & 50868.151259.7 & 16.3 & 19 + akn 564 & 51678.752683.7 & 50.2 & 20 + fairall 9 & 51679.852678.2 & 50.0 & 20 + mcg6 - 30 - 15 i & 51926.752252.9 & 13.1 & 23 + mcg6 - 30 - 15 ii & 50188.951259.7 & 53.6 & 20 + ngc 3516 & 50523.051259.7 & 28.3 & 26 + ngc 3783 & 51679.752375.9 & 29.0 & 24 + ngc 4051 & 51676.052228.3 & 27.6 & 20 + ngc 4151 & 51259.751676.0 & 17.3 & 24 + ngc 5548 & 51678.752681.4 & 45.6 & 22 +
|
direct time - resolved spectral fitting has been performed on continuous _ rxte _ monitoring of seven seyfert 1 galaxies in order to study their broadband spectral variability and fe k@xmath0 variability characteristics on time scales of days to years .
variability in the fe k@xmath0 line is not detected in some objects but is present in others , e.g. , in ngc 3516 , ngc 4151 and ngc 5548 there are systematic decreases in line flux by factors of @xmath125 over 34 years .
the fe k@xmath0 line varies less strongly than the broadband continuum , but , like the continuum , exhibits stronger variability towards longer time scales .
relatively less model - dependent broadband fractional variability amplitude ( @xmath2 ) spectra also show weaker line variability compared to the continuum variability .
comparable systematic long - term decreases in the line and continuum are present in ngc 5548 .
overall , however , there is no evidence for correlated variability between the line and continuum , severely challenging models in which the line tracks continuum variations modified only by a light - travel time delay .
local effects such as the formation of an ionized skin at the site of line emission may be relevant .
the spectral fitting and @xmath2 spectra both support spectral softening as continuum flux increases .
|
observational work during the past two decades has established the paradigm of interaction - induced activity in galaxies ( cf .
fricke & kollatschny 1989 ; sulentic 1990 ) .
galaxy interactions were initially studied theoretically by simplified n - body simulations ( toomre & toomre 1972 ) and investigated in more detail by n - body / sph models ( barnes & hernquist 1996 and references therein ) .
interactions are linked to a variety of physical processes such as nuclear starbursts , non - thermal nuclear activity , and morphological evolution into different classes of galaxies ( barnes & hernquist 1992 ) .
already , collisionless calculations can simulate the transformation of merged spirals into ellipticals ( schweizer 1982 ) and the formation of tidal dwarf galaxies ( duc & mirabel 1994 ; sanders & mirabel 1996 ) .
thereby , understanding galaxy interactions is a prerequisite for addressing issues like the morphological mix and the luminosity and number distribution of galaxies .
gravitational perturbations of the gas componentthrough encounters between spirals can provoke large - scale internal gas instabilities , loss of angular momentum , and inflow of gas into the nuclear regions .
accumulation of gas at high densities ( solomon et al .
1992 ) is thought to be a necessary condition for the onset of vigorous nuclear starbursts ( sanders & mirabel 1996 and references therein ) or even the development of agn activity . in the presence of dust
, the radiative output of massive stellar clusters is absorbed and reradiated in the far - infrared , a process which is thought to operate in the class of fir - luminous galaxies . at the same time , the collective output of energy and momentum from ob stars and supernovae explosions into the interstellar medium creates a hot ( @xmath8 few @xmath9 k ) x ray emitting gas phase . the thermal pressure exerted by this component on the ambient cold gas may terminate the gas - inflow and even disrupt the outlying -layer , thus giving rise to a fast ( @xmath10 km sec@xmath11 ) galactic wind ( heckman et al
. 1987;1996 , veilleux et al .
1994 ) . ' '' '' [ fig1 ] on the other hand , interactions between galaxies may effect a redistribution of a large fraction of the cold gaseous component on scales of hundreds of kpc .
-plumes or bridges emanating from the nuclear regions or protruding from the plane do not seem to be uncommon among interacting / merging galaxies .
gaseous features on scales of 50 to 150 kpc and mass ranges between 1.5 and @xmath12 were discovered in systems spanning a wide range across the merging sequence , such as arp 144 ( higdon 1988 ) , arp 215 ( smith 1991 ) , ngc 520 ( norman et al . 1996 ) and ngc 3256 ( hibbard & van gorkom 1996 ) .
the fate of this material , in particular the question , whether it is going to be recaptured in the gravitational field of the merged system , is uncertain .
it is , however , conceivable that after some time reinfall of tidally ejected matter onto a merger ( hibbard & mihos 1995 ) could lead to shock formation , thereby provoking extranuclear activity or feeding a subsequent burst of star formation .
+ ' '' '' in this paper we focus our attention on the x - ray properties of the interacting system arp 284 ( arp 1966 ) .
this system consists of the nuclear starburst galaxy ngc 7714 and its fainter inactive companion ngc 7715 .
weedman et al . (
1981 ) have first pointed out the extraordinary properties of ngc 7714 among starburst nuclei and inferred from _ einstein _
observations an intrinsic x - ray luminosity of @xmath13 in the 0.253.5 kev energy range . other observational support for the presence of an ongoing burst of star formation in ngc 7714
is provided by the detection of wr spectral features and strong h@xmath14 emission on scales of @xmath15 ( gonzlez - delgado et al .
1995 , smith et al .
1997 , telles & terlevich 1997 ) .
spectral synthesis models of bernlhr ( 1993 ) imply for ngc 7714 a burst age of @xmath6 myr .
the same author concluded that the less massive interaction counterpart ngc 7715 has undergone a burst of star formation roughly 70 myr ago ( bernlhr 1993 ) , presumably at the time of the initial impact between both galaxies ( smith et al .
arp 284 displays features predicted to evolve shortly after the first closest passage between two gas - rich spirals ( smith et al .
1997 , hernquist 1992 ) .
a striking one is a large asymmetric ring located @xmath6 eastwards of the nucleus of ngc 7714 .
narrow - band imaging of ngc 7714 has not revealed a notable contribution of h@xmath14 emission to the light of this feature ( gonzlez - delgado et al .
1995 , smith et al .
also high angular resolution 21 cm vla - maps of arp 284 ( smith et al . 1997 ) show that the ring is comparatively devoid of -gas .
furthermore , bushouse & werner ( 1990 ) have reported that the stellar ring possesses nir colours similar to those in the disk .
all latter facts lend support to the idea that its formation process is related to a collisionally induced dynamical perturbation of the stellar disk ( lynds & toomre 1976 ) rather than being the result of past star formation .
this hypothesis is in accord with numerical simulations of off - center encounters between spiral galaxies demonstrating the formation of similar morphological features ( mihos & hernquist 1996 , smith et al .
1992,1997 ) . in the following
we investigate the x - ray properties of arp 284 using recent rosat data , together with relevant optical and radio observations .
we shall adopt a distance of 40 mpc ( h@xmath16=75 km sec@xmath11 mpc@xmath11 ) to arp 284 ( bernlhr 1993 ) .
arp 284 was observed with the 2.2 m telescope at calar alto with a site - ccd detector attached to the cafos focal reducer giving an instrumental scale of 053pixel@xmath11 .
the target was observed for 3 min in the johnson r under non - photometric conditions with a seeing of 21@xmath17005 .
the image was calibrated on the basis of aperture photometry by bushouse & werner ( 1990 ) and bernlhr ( 1993 ) .
the detection limit of point sources at a s / n of 2 was estimated to @xmath18 rmag .
as shown in fig.1 , the nuclear regions of the interacting galaxies ngc 7714 and ngc 7715 are separated by @xmath19 , corresponding to a linear distance of 22.8 kpc .
both galaxies are apparently connected by a diffuse and faint ( @xmath20 ) stellar bridge protruding from ngc 7715 into the outer boundary of the stellar ring in ngc 7714 .
the point source indicated to the upper - left is the qso [ hb89]2333 + 019 ( @xmath21)@xmath22 : 23@xmath2336@[email protected],+02@xmath26 1042 ( hewitt & burbidge 1989 ) .
the apparent luminosity of this background source ( @xmath27 ) was determined to @xmath28 mag . for the sake of revealing faint compact structures we processed the central region of ngc 7715 and ngc 7714 with a hierarchical binning algorithm ( hereafter hb - method ) designed for extracting weak features of angular size smaller than a user - defined angular threshold .
results obtained by the hb - method for the central regions of ngc 7714 and ngc 7715 are displayed in the insets in fig.1 .
the dominant features in ngc 7714 within a region of @xmath29 in diameter are the compact starburst nucleus and an adjacent fainter knot .
as shown in fig.2 the intensity distribution of ngc 7714 can be approximated reasonably well by an exponential fitting law for @xmath30 .
the marked intensity depression by @xmath31 mag in the sbp of ngc 7714 at a photometric radius @xmath32 reflects the sharp transition from the ( bar@xmath33ring)dominated regime of ngc 7714 to its fainter disk population . ' '' ''
we observed arp 284 in the 0.12.4 kev energy band with both , the position sensitive proportional counter ( pspc ) and the high resolution imager ( hri ) on board rosat ( trmper 1983 , pfeffermann et al .
the analysis of the x ray data was carried out with the exsas package ( zimmermann et al .
1994 ) implemented in the midas .
arp 284 was observed with the pspc twice ( pspc_1 : 1.33 and pspc_2 : 12.68 ksec ) .
( 0,6 ) ' '' '' [ cols="<,^,^,^,^,^,^,^ " , ] @xmath34 : hri _ softness ratio _ determined as the ratio of the net counts received in the hri channels 15 and 611 ( cf .
wilson et al . 1992 ) .
+ @xmath35,@xmath36 : radii enclosing 50% and 80% of the flux , as determined from the intensity profiles ( fig.7 ) .
+ @xmath37;@xmath38 : intrinsic flux and luminosity in the rosat energy band ( 0.12.4 kev ) in units of 10@xmath39 and 10@xmath40 , respectively .
both quantities are derived on the assumption that the x - ray emission in either component can be approximated by a _
thbr _ model and shares the same spectral properties . according to bernlhr ( 1993 ) the photometric properties of the nuclear region in ngc 7714
can be accounted for by an ongoing burst of star formation of age @xmath41 myr .
scaling the model predictions of leitherer & heckman ( 1995 , hereafter lh95 ) to the observed nuclear luminosity and assuming continuous star formation over @xmath42 , we obtain an average star formation rate ( sfr ) of @xmath43 yr@xmath11 .
hereby , we adopt a salpeter initial mass function ( @xmath44 ) , solar metallicity and lower and upper stellar mass cutoffs of 1 and 100 , respectively .
we shall remark that the above model - dependent quantities are approximative only and represent a lower limit for the actual sfr in ngc 7714 as being derived by other authors from detailed spectrophotometric studies .
for instance , from the -luminosity of ngc 7714 storchi - bergman et al .
( 1994 ) deduced a sfr of 17.2@xmath45 yr@xmath11 .
furthermore , calzetti ( 1997 ) has inferred from uv - to - nir synthesis studies a sfr of [email protected]@xmath45 yr@xmath11 and deduced a much lower starburst age of @xmath47 myr .
scaling figs.2&6 of lh95 to the adopted sfr we obtain for the present evolutionary age of the nuclear starburst a number @xmath48 of o stars and a sn rate @xmath49 yr@xmath11 .
the corresponding production rate of _ mechanical _ energy injected into the ism by massive stars and sne is @xmath50 ( lh95 , fig.56 ) and the total _ mechanical _ energy generated since the onset of the starburst is @xmath51 erg ( lh95 , fig.58 ) .
+ the x - ray luminosity from o stars is @xmath52 @xmath53 . given an average x - ray luminosity of @xmath54 over a time scale @xmath55 yr for a sn remnant ( cowie et al .
1981 , williams et al . 1997 ) we estimate the contribution of such sources to @xmath56 or @xmath57% of @xmath58 . from the number ratio of @xmath59 of o - stars to high mass x - ray binaries ( hmxbs ; fabbiano et al .
1982 ) we obtain an upper limit of 140 currently active hmxbs in the nuclear starburst region of ngc 7714 . with a typical x - ray luminosity of a hmxb of @xmath60
this results in a maximum luminosity contribution of this type of objects of the order of @xmath61 . in summary ,
the direct contribution of x - ray emitting point sources produced by the nuclear starburst , such as o stars , snrs and hmxbs amounts to @xmath62 or @xmath63% of @xmath58 .
this percentage will not be greatly enhanced even when postulating the presence of a few _ super - eddington _ point - like sources with luminosities of @xmath64 as found in some nearby starburst galaxies ( dahlem et al .
1994 , vogler & pietsch 1997 , wang et al .
1995 , fabbiano et al .
1997 ) . as a result , @xmath65% of the x - ray luminosity confined to the starburst nucleus of ngc 7714 must be attributed to the emission by extended hot gas ( few @xmath66 k ) being heated up by the injection of _ mechanical _ energy from ob stars and sne .
equation3.2 of nulsen et al .
( 1984 ) yields for the required mass of the hot plasma in the nuclear region of ngc 7714 @xmath67 where @xmath68 is the radius of the volume in pc , @xmath69 the emission measure in @xmath70 , and @xmath71 the volume filling factor .
assuming for the emitting volume of component n a radius of 1 kpc and a mean plasma temperature @xmath72 k we obtain a nominal gas mass @xmath73 .
thus , the mass of the hot plasma required to explain the nuclear x - ray luminosity of ngc 7714 is comparable to the stellar mass formed in the burst .
the radiative output of the hot gas ( @xmath74 ) corresponds to roughly 20% of the current _ mechanical _ energy output provided by the nuclear stellar population .
thus , there is no fundamental energy problem for the starburst hypothesis .
the absence of signatures of recent star formation at the stellar ring close to component e ( sect.1 ) excludes the possibility that its x - ray luminosity of @xmath75 is due to a superposition of point sources such as hmxbs and snrs .
next we shall investigate whether the extended emission observed in region e is not intrinsic to ngc 7714 but due to the superposition of at least two foreground or background sources . as pointed out in sect.2 , the estimated detection limit of the ccd - frame ( fig.1 ) is @xmath18 rmag .
this limit can effectively be pushed down by @xmath76 mag when applying the hb - method . at this level ,
ground - based images do not reveal any optical point sources coincident with component e. the only conspicuous feature seen in an archival hst - wfpc2 exposure of ngc 7714 at a distance @xmath77=+167 , + 51 from the nucleus is a pair of two faint point sources being separated by @xmath78 .
their apparent r magnitude , estimated to 22.8 mag , is equivalent to that of a m5v star ( @xmath79 ; johnson 1966 ) at a distance of @xmath80 kpc .
the x - ray luminosity of such late - type stellar sources amounts to @xmath81 ( schmitt & snowden 1990 , hnsch & schrder 1996 ) . at the assumed distance of 2.9 kpc ,
a source of this type would produce an x - ray flux @xmath82 , much lower than the detection threshold of rosat .
analogously we can excluse other candidates of foreground and background sources as origins for component e. we have made detailed checks on the possibilities of the contamination by popii binaries , background clusters and diffuse background sources ( hasinger et al .
1993 ) .
we , therefore , conclude that the extended x - ray flux detected at the location of component e be a superposition of point sources of known type to be extremely improbable .
given that a physical intersection between the eastern -bridge ( smith et al .
1997 ) and ngc 7714 is likely , we consider below ( _ i _ ) a starburst wind interaction with the low - density halo or with the dense -bridge and ( _ ii _ ) collisions of clouds resulting from gas infall from the bridge including the interaction of the rotationally moving disk clouds with the material of the bridge .
theoretical and observational work in the past decade has advanced a picture where gas heated up by the central energy source is able to penetrate the ambient -layer and then to escape into the galactic halo with velocities of the order of @xmath83 kms@xmath11 ( chevalier & clegg 1985 , heckman et al .
1987 , norman & ikeuchi 1989 and references therein ) where it produces an extended shock region .
this mechanism could cause component e. the presence of wind of warm gas with a velocity @xmath84 kms@xmath11 with its axis pointing to the ne of the nuclear region is already suggested by optical spectroscopy ( taniguchi et al .
although the spectra do not prove a large - scale outflow of hot gas towards the galactic halo they show , however , the gas phase in ngc 7714 to be kinematically perturbed .
suchkov et al .
( 1994 , hereafter sbhl94 ) have shown that in a nuclear starburst - wind the bulk of the soft ( 0.22.2 kev ) x - ray photons will not be produced in the wind material itself but rather by its collisionally shocked interface with the ambient halo gas . by analogy to the model prediction of sbhl94 , component e could be interpreted as the shell of a hot cavity expanding from the nuclear starburst , with the wind material itself being too hot ( @xmath85 k ) to make a measurable flux contribution to the rosat energy band .
read & ponman ( 1998 ) have presented a thorough discussion on unipolar outflows of hot gas seeming not to be unusual among colliding galaxies in a mid - stage of merging .
an unipolar structure seems to readily evolve when injection occurs above the galaxy plane ( mac low & mccray 1988 ) or in the presence of a kinematically disturbed ism .
models a1 and a2 of sbhl94 are computed assuming starburst properties similar to those we have adopted for ngc 7714 , e.g. a sfr of 2 yr@xmath11 , salpeter imf and @xmath86 myr and yield a number of predictions in accord with the present observations .
for instance the luminosity of component e can be accommodated by a wind - halo interaction model when assuming a halo density @xmath87 @xmath88 .
however , in this model it is expected that the source will appear much more extended than the rather compact component e. alternatively , component e may be the result of the collision of a starburst - driven wind with the western boundary of the -bridge discovered by smith et al .
this would lead to the formation of a shocked interface at the radius where the ram pressure ( @xmath89 ) of the wind becomes equal to the ambient pressure of the -gas . as discussed in tenorio - tagle & muoz - tun ( 1997 ) , while the outer shock is going on sweeping up the ambient cold gas , the inner ( reverse ) shock will thermalize the wind to temperatures of @xmath90 k where @xmath91 is the wind velocity in units of @xmath92 . in this model , the adopted plasma temperature @xmath72 k corresponds to a post - shock temperature generated by a velocity of the order of @xmath93 kms@xmath11 .
pressumably , such a process is going to produce a relatively compact shocked interface with size comparable to the diameter of the western tip of the -bridge , yielding a better correspondence to the morphology of component e than the wind - halo interaction hypothesis . in summary
, the morphological and energetic properties of component e may be understood in terms of the collision of a starburst - driven wind with the eastern -bridge rather than collision with the tenuous halo - gas . a further hypothesis for the origin of the x - ray emission in region e invokes gas infall from the eastern -bridge on the -disk of ngc 7714 .
as shown by tenorio - tagle ( 1982 ) the energy - release caused by an -cloud impinging on the disk can provoke a supersonic expansion of a hot ( @xmath9@xmath94 k ) ring - shaped cavity in the ambient cold medium .
smith et al .
( 1997 ) were not able to confirm a systematic motion of the massive ( @xmath95 ) -bridge towards ngc 7714 .
nevertheless , they measured a projected velocity dispersion @xmath96 kms@xmath11 within the -bridge .
therefore , if the -bridge is intersecting the disk of ngc 7714 high - speed gas cloud collisions are likely to occur in the interaction zone . if the infalling gas medium is composed of clouds with a radius @xmath97 pc and a volume filling factor @xmath98 then the mean free path @xmath99 of an interloping cloud amounts to @xmath100 pc , thus less than the disk thickness of ngc 7714 .
given the low sound speed in such a cloud ( @xmath101 ; spitzer 1978 ) , a collision at a relative velocity @xmath102 few 100 would be characterized by a high mach number , thus , give rise to a strong shock wave propagating within either cloud with a velocity @xmath103 ( jog & solomon 1992 ) . from the latter considerations ,
a strip of shocked material with size comparable to @xmath104 ( @xmath105 kpc ) can be formed in a time span of few myr when @xmath106 . substituting the estimated intrinsic x - ray luminosity of component e and such a radius in eq.(1 ) we obtain for the mass of the hot interior @xmath107 . from eq.(3 ) of nulsen et al .
( 1984 ) the corresponding gas cooling time is @xmath108 myr .
an estimate of @xmath109 erg for the energy content of component e can be inferred assuming a constant radiation of @xmath110 over @xmath111 .
this is equivalent to the kinetic energy of a cloud with mass @xmath112 ( @xmath113% of the mass of the -bridge ) plunging onto the disk of ngc 7714 with a velocity @xmath114 . + from the virial mass of @xmath115 within a radius of 15 inferred for ngc 7714 by bernlhr ( 1993 ) and assuming rotational motions one obtains at the distance of component e a circular velocity of @xmath84 kms@xmath11 .
thus , the required collisional velocity @xmath114 is already comparable to the average circular gas velocity in the disk of ngc 7714 .
+ as a result , the release of kinetic energy at region e can be understood in a twofold manner : gas infall onto the disk by analogy to galactic _ worms _ ( heiles 1979 , mirabel 1982 ) and/or collision of rotationally moving -clouds of the disk of ngc 7714 with the western boundary of the -bridge .
clearly , a number of parameters must be specified to assess the energy output of these processes .
most important , the -kinematics and the thermalization efficiency during such a collision remains uncertain .
nevertheless , we argue that in view of the energetics such a mechanism powered by cloud collisions only can readily account for the observed x - ray luminosity in component e. reinfall of tidally ejected matter is likely to be a frequent occurrence in the merging of two gas - rich galaxies ( hibbard & mihos 1995 ) and may well be accompanied by processes discussed in sects.4.2.1&4.2.2 generating x - ray emission . however , it is difficult to analyze how such processes depend on the kinematical , dynamical and geometrical circumstances of the merging event . while read & ponman ( 1998 ) do not report the presence of any strong extranuclear features they find the clear presence of low - surface - brightness extended x - ray emission in their sample .
they discuss the possibility that such x - ray halos seen in evolved mergers do not form entirely through the starburst - driven ejection of hot gas from the nucleus , but also through infalling shock - heated tidal matter .
if component e is due to a systematic gas infall from the bridge one may speculate about the observability of this feature . among the fifteen thoroughly investigated colliding systems shown in fig.5 only arp 284 exhibits a high - surface - brightness extranuclear x - ray component .
assuming that the formation of such features in the interaction region must commonly occur at some stage during the collision lasting @xmath116 yr , then from a statistical point of view this feature can not survive much longer than @xmath117 yr .
this time - scale does not seem unreasonable : as discussed above , the initial impact between two cold media with a velocity of a few @xmath118 100 can readily give rise to the formation of a shocked interface .
subsequent infall of cold gas occurring with a velocity smaller than the internal sound speed in the hot cavity is not expected to add much to the shock - induced luminosity .
the period of visibility of such a shock - region ( identified with region e of arp 284 ) will be of the order of its cooling time @xmath119 yr . while mass - loading of the hot cavity enhances the cooling efficiency , differential rotation may smear the high - surface - brightness x - ray features ( again on a time - scale @xmath85 yr ) , thus contributing to the formation of a diffuse low - surface - brightness x - ray halo similar to those found by fricke & papaderos ( 1996 ) and read & ponman ( 1998 ) .
numerical simulations suggest that arp 284 is the result of an off - center encounter between the spiral galaxies ngc 7714/15 ( smith et al .
1997 ) . the less massive one , ngc 7715 , underwent a burst of star formation @xmath120 myr ago , presumably at the time of the closest passage between both systems . while ngc 7715 is at present in a quiescent post - starburst phase , its interacting neighbour ngc 7714 ( smith et al .
1997 ) being by a factor @xmath121 more massive harbours an ongoing nuclear starburst initiated @xmath122 myr ago ( bernlhr 1993 , calzetti 1997 ) .
morphological signatures of the interaction in arp 284 are a conspicuous stellar ring in ngc 7714 to the east of its nucleus and a faint ( @xmath123 ) stellar bridge probably connecting ngc 7715 and ngc 7714 .
smith et al .
( 1997 ) have discovered a massive ( @xmath124 ) -bridge parallel to the stellar one , likely being of tidal origin . both features intersect ngc 7714 roughly 20 eastwards of its nuclear region , near to the outer boundary of the stellar ring .
+ the surface brightness distribution of ngc 7714 shows two distinct luminosity parts in excess of its disk brightness . on small scales ( @xmath125 )
the light is being dominated by its compact high surface brightness ( @xmath126 ) starburst nucleus . at intermediate intensity levels and out to a radial extent
equivalent to two disk scale lengths the intensity distribution of ngc 7714 shows an initially flat and then sharply decreasing plateau .
this feature is partly due to the luminosity of the stellar ring mentioned above . in agreement to results from _
einstein _ , rosat pspc observations show that the entire emission in arp 284 is confined to ngc 7714 .
fits to the rosat pspc spectrum imply a thermal emission origin approximated best by a thermal bremsstrahlung model with @[email protected]@xmath460.6 kev ( @xmath128 k ) .
we deduce an intrinsic 0.12.4 kev luminosity of ngc 7714 ranging between 2 and @xmath129 , comparable to the luminosity reported for other interacting / merging starburst systems .
rosat hri maps reveal that the soft x - ray luminosity of ngc 7714 is being contributed by two distinct x - ray emitting sources both of which were found to be extended . the more luminous one ( n ) , for which we estimate an x - ray luminosity @xmath130 is coincident with the starburst nucleus of ngc 7714 .
its luminosity can be accounted for by the superposition of point- and diffuse x ray emitting sources expected to form in a starburst .
point sources evolving in a continuous nuclear starburst with an age @xmath6 myr and average sfr @xmath8 2 @xmath131 yr@xmath11 are estimated to provide only a minor fraction ( @xmath47% ) of the nuclear x
ray luminosity in ngc 7714 while the bulk of the soft x - ray emission is being produced by hot ism residing in a volume @xmath132 kpc in diameter .
the second x ray source ( component e ) , located roughly 20 eastwards of the starburst nucleus is of intriguing nature .
it lacks any optical counterpart and the probability for confusion with an assembly of x - ray point sources unrelated to arp 284 is low .
we estimate the intrinsic x - ray luminosity of this component to @xmath75 and favour a model based on _ in situ _ shock heated gas . with respect to the combined evidence of vla- and rosat hri maps different interpretations seem possible .
one plausible scenario invokes the collision of a starburst driven wind with the western boundary of the massive -bridge . in this case a relatively compact shocked layer with a temperature of few 10@xmath133 k may evolve .
alternatively , gas infall from the eastern -bridge onto the disk of ngc 7714 with a speed comparable to the intrinsic disk velocity dispersion can produce a collisional interface powering component e. collisional breaking of the disk - rotation in ngc 7714 by -clouds contributes an additional energy source .
we argue that the x - ray spot is a transient phenomenon and may eventually evolve into an extended low - surface - brightness x - ray halo as commonly observed in merging systems .
this work was supported by the deutsche agentur fr raumfahrtangelegenheiten ( dara ) grant 50 or 9407 6 .
we wish to thank an anonymous referee for useful comments which have helped to clarify various aspects of this work .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , caltech , under contract with the national aeronautics and space administration .
|
we discuss the x - ray properties of the interacting system arp 284 , consisting of the active nuclear starburst galaxy ngc 7714 and its post - starburst companion ngc 7715 .
a morphological signature of the interaction , thought to have started @xmath0 myr ago , is an asymmetric stellar ring dominating the intensity profile of ngc 7714 in the inner disk ( @xmath1 exponential scale lengths ) . in agreement to previous _
einstein_-data our rosat pspc exposure shows the x - ray emission of arp284 to be confined to ngc 7714 . the bulk of the intrinsic x - ray luminosity in the rosat 0.12.4 kev band can be accounted for by thermal emission from hot ( @xmath2 k ) gas and amounts to @xmath3 .
follow - up observations with the rosat hri revealed two distinct extended emitting regions contributing to the x - ray luminosity in ngc 7714 .
the more luminous of them ( @xmath4 ) coincides with the central starburst nucleus and can be explained this way .
the fainter one ( @xmath5 ) is located @xmath6 off - center and does not have any conspicuous optical counterpart .
it is , likely , located at the borderline between the stellar ring and a massive ( @xmath7 ) -bridge further to the east possibly intersecting ngc 7714 .
the and x - ray morphology and the extensive starburst nature of the nuclear energy source suggest different scenarios for the formation of the eastern emission spot . the possibilities of ( i ) collisional heating of the outlying gas by a starburst - driven nuclear wind and ( ii ) infall of -clouds from the bridge onto the disk of ngc 7714 are discussed .
|
line emission or absorption is a promising probe of the high - redshift universe .
considerable theoretical and instrumental effort has been devoted to the use of the redshifted 21 cm line of neutral hydrogen in such a fashion ( see @xcite for a recent review ) .
comparable observations using atomic or molecular lines at far - infrared wavelengths have seen less development .
recent advances in far - ir instrumentation , combining sensitive receivers with large collecting area , have led to the detection of far - ir lines in individual sources at cosmological distances , raising the prospects for a future generation of far - ir line surveys as a probe of the high - redshift universe .
infrared fine structure lines are important coolants of the neutral and ionized gas phases of normal and starburst galaxies .
they are easily excited , arise from the most abundant metals with low ionization potential , and are generally not affected by galactic or intergalactic attenuation .
the strongest line is the @xmath3 fine - structure line of singly - ionized carbon at 157.74 @xmath0 m rest wavelength .
the ubiquity of carbon in the interstellar medium combined with the relatively low ionization potential and modest excitation temperature make the [ cii ] line bright , with as much as 0.1 1% of the total bolometric luminosity of the host galaxy emitted in this single line @xcite .
the intrinsic line brightness and lack of significant attenuation from the galactic or intergalactic medium in the far infrared in turn render the [ cii ] line visible over cosmological distances @xcite .
the ability to detect [ cii ] emission to redshifts @xmath4 makes the [ cii ] line a promising candidate for cosmological surveys .
[ cii ] emission originates from multiple phases of the interstellar medium .
the combined emission from all phases serves as a marker for source redshift for surveys of large - scale structure @xcite .
the empirical relation between [ cii ] and the bolometric far - ir luminosity allows the observed [ cii ] intensity to serve as a tracer of star - formation activity @xcite , while the ratio of [ cii ] to other far - ir lines probes physical conditions within different phases of interstellar medium of the host galaxies @xcite .
several authors have discussed requirements for [ cii ] surveys at cosmological distances .
@xcite estimate that a 500-hour observation of the 2.4 diameter hubble deep field with the atacama large millimeter array could yield 15 [ cii ] detections over the redshift range @xmath5 .
the sub - arcsecond angular resolution available for such interferometric observations minimizes confusion from source superposition within each synthesized beam , but the large number of independent pointings needed to map an area comparable to the hdf minimizes the integration time spent on any individual source so that only the brightest sources are detected .
observations at coarser angular resolution reduce the need for multiple pointings , at the cost of introducing competing lines from foreground sources .
intensity mapping is one such technique , using fluctuations in the spatial and spectral distribution of line emission from a superposition of sources to derive the underlying power spectrum of the individual , unresolved sources ( see , _ e.g. _ , the discussion in @xcite ) .
several such surveys have been proposed .
for example , @xcite present a concept for [ cii ] intensity mapping using a 25-beam grating spectrometer each with beam width 10 , while @xcite propose a suite of single - beam spectrometers each with 30 beam width .
the larger beam size anticipated for such surveys facilitates mapping large areas but requires additional analysis to separate the target line emission from the foreground created by emission from other galaxies along the same line of sight .
several authors discuss cross - correlation techniques to minimize foreground confusion in line intensity mapping @xcite . for a given target redshift ,
the technique defines two observing wavelengths corresponding to emission from two different rest - frame lines .
maps of line emission taken at each observing wavelength contain spatially correlated structure corresponding to the two lines emitted by each source galaxy .
the contribution from other lines at different rest - frame wavelengths redshifted into the observing wavelengths must originate from sources at different redshifts .
spatial correlations from such distant sources will be weaker , so that the cross - correlation traces the power spectrum of the source galaxies at the target redshift .
intensity mapping surveys large volumes of the universe to derive statistical estimates of structure formation , but provides little information on individual sources . in this paper
, we investigate the complementary technique of direct line identification in narrow - beam spectroscopic surveys .
if the [ cii ] line were the only observable far - ir line , redshift identification would be straightforward . in practice , however , the [ cii ] line is merely one of several cooling lines .
the problem of sorting through an observed spectrum containing multiple lines from multiple galaxies at different redshifts presents both instrumental and analytical challenges .
we use a toy model of far - ir line emission to evaluate a simple algorithm for line identification and estimate relevant instrumental parameters ( angular resolution , sensitivity , and spectroscopic resolution ) required to derive cosmologically interesting results .
the results are applicable for spectral cleaning of intensity mapping surveys as well as pencil - beam surveys of more limited cosmological volume . , width=259 ]
we use a simple model of the brightest far - ir lines to simulate the observed line spectra from distant galaxies .
we assume that the total luminosity , @xmath6 emitted in a single line at rest frequency @xmath7 may be simply related to the total infrared luminosity , @xmath8 of a galaxy , @xmath9 the observed specific flux of a single line from a set of ir galaxies at redshift @xmath10 is then given by @xmath11 where @xmath12 is the galaxy ir luminosity function , @xmath13 is the luminosity distance , @xmath14 is the redshifted frequency , and @xmath15 is the intrinsic line profile , normalized so that @xmath16 . ignoring the effects of gravitational lensing , the total line intensity emitted from galaxies within a redshift interval @xmath17 to the present ( @xmath18 ) then becomes @xmath19 \nonumber\end{aligned}\ ] ] if the ratio of line intensity to ir luminosity is a constant , eq .
( [ obs_specific_intensity ] ) simplifies to @xmath20\left|{cdt\over dz}\right|{dz\over 1+z}\ ] ] where @xmath21 @xmath22 is the comoving number density of galaxies , and @xmath23 is their average ir luminosity .
we simulate the far - ir line spectra using a standard cosmology with @xmath24 , @xmath25 , and @xmath26 @xcite .
we divide the comoving volume into redshift bins and compute the total number of galaxies @xmath27 within each bin @xmath28 where @xmath29 is the comoving volume at redshift @xmath10 .
we use the resulting number density to generate monte carlo realizations of the far - ir line spectra .
the background cosmology fixes the mean number of galaxies within a fixed observing solid angle .
we assign each simulated galaxy a redshift and ir luminosity , using the @xcite luminosity function to obtain a representative luminosity distribution within each redshift bin .
each galaxy then contributes a set of far - ir lines .
figure [ line_fig ] shows the relative intensity of the rest - frame lines used for this study .
we include the [ cii ] line at 158 @xmath0 m , [ ci ] at 610 and 371 @xmath0 m , [ oiii ] at 88 and 52 @xmath0 m , [ oi ] at 145 and 63 @xmath0 m , [ niii ] at 57 @xmath0 m , and [ nii ] at 205 and 122 @xmath0 m , as well as the co ( @xmath30 ) series through @xmath31 .
the total luminosity in each line is related to that of the host galaxy by the factor @xmath32 ( eq . [ f_def ] ) .
we set the mean value of @xmath33 for each line using the ratios of values in table 1 of @xcite . for simplicity
, we do not explicitly calculate the star formation rate for each simulated galaxy , but relate the comoving star formation rate @xmath34 to the ir luminosity density @xmath35 as @xmath36 where @xmath37 is the comoving ir luminosity density , @xmath34 is the comoving cosmic star formation rate ( csfr ) in @xmath38 yr@xmath39 mpc@xmath40 , and @xmath41 is a conversion factor for a salpeter stellar initial mass function ( kennicutt 2012 ) .
the relative luminosities from different lines within a source are known to vary by roughly a factor of 10 .
we account for such variation by multiplying @xmath32 for each line in each host galaxy by a random number drawn uniformly from the range [ 0.3 3 ] .
we additionally account for the systematic variation in line luminosity and line ratios due to the increased cmb temperature at higher redshift @xcite .
we assume that the host galaxies are not spatially resolved and assign a random rotational linewidth to each galaxy uniformly drawn from the range [ 30 300 ] km s@xmath39 .
each galaxy contributes multiple lines ; all lines from a single galaxy have the same rotational line broadening . finally , we accumulate the lines from all source galaxies within the sampled volume and redshift each line to obtain a realization the observed far - ir spectrum .
the effects of instrument spectroscopic resolution are discussed in @xmath423.2 .
we ignore instrument noise and continuum emission from the superposed host galaxies to focus here on the problem of line identification . for the [ cii ] line .
( top ) individual line flux , shown at the central wavelength before rotational broadening .
lines are color - coded by species using the same color coding as figure [ line_fig ] . 48 lines from 48 individual galaxies at @xmath43 fall within the plotted wavelength range .
the assumption that the brightest lines are always [ cii ] lines produces a high error rate , with roughly a third of the brightest lines originating from species other than [ cii ] .
( bottom ) the same lines are shown after rotational broadening . the superposition of rotationally broadened lines produces a complicated spectrum ( black curve ) .
[ sample_spectrum ] , width=336 ] figure [ sample_spectrum ] shows a typical spectrum with 2 angular resolution at wavelengths corresponding to redshifts @xmath44 for the [ cii ] line .
two problems are evident .
the first problem is confusion . unlike radio observations , where the redshifted 21 cm line is the only bright line ,
a number of lines contribute to the observed far - ir spectra .
only 16% of the lines within the observed wavelength range correspond to the targeted [ cii ] line , with the remainder consisting of `` interloper '' lines originating at both higher and lower redshifts from species other than [ cii ] .
the second problem is blending : the superposition of rotationally broadened lines produces a complicated spectrum .
the fainter lines merge to form a near continuum spectrum , further hindering line identification .
although the [ cii ] line is bright , simply assuming that the brightest observed lines within a single field are always [ cii ] lines produces a high error rate for line identification .
we quantify this by taking the simulated spectra and selecting the brightest 10% of lines within a narrow range centered at different source redshifts .
we may then count how many of the brightest lines in each range are in fact [ cii ] lines as opposed to interloper lines from sources at other redshifts .
the probability that a given line within the brightest 10% is in fact a [ cii ] line falls from 0.8 for observations near 315 @xmath0 m ( @xmath45 ) to 0.4 for observations at 1100 @xmath0 m ( @xmath46 ) .
we address the problem of blind line identification using a version of the radio - astronomical clean algorithm .
although the brightest lines are not necessarily [ cii ] lines , the [ cii ] lines are generally bright .
this suggests an iterative process .
we first select the brightest line within the observed spectrum . assuming that this `` parent '' line is [ cii ] fixes the corresponding source redshift @xmath47 .
we then predict the wavelengths at which other far - ir lines originating from the same source should be observed , @xmath48 and examine the spectra at those wavelengths .
if the expected `` companion '' lines are found , the redshift identification is confirmed and both the candidate [ cii ] line and the detected companion lines are flagged as known and removed from the list of lines to search .
if no companion lines are found , we repeat the test assuming that the parent line is the next - most - common far - ir line ( typically a co line ) and continue down the list of far - ir lines until we identify a parent / companion combination or exhaust the list of potential parent lines .
if no companion lines are found , the bright parent line is flagged as unknown and removed from the list of possible parent lines , although it may later be identified as a companion line .
we then proceed to the brightest remaining un - flagged line and iterate until all lines are tested .
m to 1160 @xmath0 m containing 390 peaks ( see text ) .
( top ) cumulative number of correct ( solid line ) and incorrect ( dashed ) [ cii ] line identifications as a function of the relative line intensity , sorted from brightest ( @xmath49 ) to faintest ( @xmath50 ) .
( bottom ) probability that the @xmath51-brightest peak in the observed spectra will be correctly identified .
the number of false identifications is small for the brightest half of the observed lines .
[ error_fig ] , width=336 ] we test the algorithm using monte carlo realizations of the observed spectra and compare the derived identifications to the known monte carlo inputs to determine the detection efficiency and error rate . figure [ error_fig ] shows the results obtained from a single 2 beam observing one octave in wavelength from 580 @xmath0 m to 1160 @xmath0 m , corresponding to [ cii ] redshifts @xmath52 .
the input spectrum contains 1600 spectral lines from 550 distinct galaxies , including 260 [ cii ] lines . after rotational broadening ,
the observed superposed spectrum contains 390 peaks 3.2 ] , 110 of which result from a [ cii ] line .
blending is significant : of the 390 peaks in the superposed spectra , 290 are dominated by a single line and 100 are blends of multiple lines .
the un - blended peaks comprise 175 unique host galaxies .
the line identification algorithm starts with the brightest such peak and works to progressively fainter peaks .
the algorithm correctly identifies 200 lines corresponding to 80 unique parent redshifts and 120 companion lines .
40 of the correctly identified parent lines result from [ cii ] emission , with the remainder primarily from lower - redshift co lines .
the detection efficiency is thus 15% of the 260 [ cii ] lines input to the simulation , or 36% of the 110 potentially detectable [ cii ] lines in the superposed spectra .
the algorithm correctly identifies half of the 390 peaks in the superposed spectra .
the remaining 190 peaks are either incorrectly identified ( 160 peaks ) or not identified ( 30 peaks ) .
false identifications typically result from blended lines and are spread uniformly across all 17 candidate lines ( i.e. with 510 false identifications for each line in figure [ line_fig ] ) .
the error rate is a function of relative line intensity , with the probability of a correct identification near unity for the brightest 1015 lines and remaining above 75% for lines in the brightest half of the observed spectrum ( 3040 lines for the parameters in figure [ error_fig ] ) .
lines in the faintest third of the observable spectrum are more likely to generate a false identification than a true identification . restricting the algorithm to searching only the brightest half of the observed set of lines reduces the number of correctly - identified [ cii ] lines from 40 to 38 while yielding only 3 false identifications , for an error rate of 7% .
incorrect line identification results from the chance spectral alignment of one line from one galaxy with a different line from a different galaxy .
the spectral density of lines increases at fainter intensities ( fig [ sample_spectrum ] ) , increasing the probability for such alignment . restricting potential
`` parent '' lines to the brightest half of observed lines largely eliminates this problem without sacrificing correct identifications .
incorrectly identified lines may also be flagged through anomalous line ratios , although in this work we do not apply such a line ratio test . .
results are shown for angular resolution 0.3 , 1 , 3 , and 10 arc - minutes ( left to right ) .
observations at modest spectral resolution ( @xmath53 ) resolve as much of the blended spectrum as possible .
the superposition of rotationally broadened lines from multiple sources creates an effective ceiling at @xmath54 .
observations at higher spectral resolution do not distinguish additional lines .
[ spectral_res ] , width=259 ] the clean algorithm identifies [ cii ] lines by searching for predicted `` companion '' lines among the set of observed spectral lines .
as the number of lines increases , confusion from overlapping or blended lines can create errors in the position of the fitted line centers to limit the efficiency of the algorithm .
we use simulated spectra to assess the instrumental parameters required to mitigate line confusion .
the number of lines depends on the number of galaxies within a spatially unresolved observation . decreasing the angular resolution increases both the number of galaxies observed and the spectral density of emission lines .
observing more spectral lines in turn increases the probability that lines from different source redshifts will overlap in wavelength , requiring higher spectral resolution to distinguish individual lines .
we quantify the spectroscopic resolution as a function of the angular resolution . for a given angular resolution
, we simulate the galaxy population and generate the superposed spectra ( including rotational broadening ) from all sources using the species in figure [ line_fig ] .
we identify all peaks in the resulting ( noiseless ) spectra and compare the separation @xmath55 between adjacent peaks to the wavelength @xmath56 to determine the spectral resolution @xmath57 needed to distinguish the two as separate lines .
figure [ spectral_res ] shows the fraction of the observed spectral peaks whose separation @xmath55 from the neighboring peak is greater than than @xmath58 , where the factor of 2 accounts for the need to observe two distinct peaks . at low spectral resolution , the lines
are heavily blended so that only the few brightest lines can be distinguished .
as the instrument spectral resolution increases , progressively more of the lines can be distinguished .
the spectral resolution needed to distinguish separate peaks depends on the angular resolution : decreasing the angular resolution reduces the number of galaxies observed , producing correspondingly greater separation between adjacent peaks in the observed spectra .
spectral resolution @xmath59 ( velocity resolution 430 km s@xmath39 ) successfully resolves 90% of the peaks in the blended spectra for a 20 resolution . a higher resolution @xmath60
( 75 km s@xmath39 ) is required to resolve 90% of the peaks for 10 resolution .
rotational line widths of a few hundred km s@xmath39 effectively limit the resolution to values @xmath61 independent of the angular resolution . increasing
the spectral resolution above @xmath62 does not distinguish additional lines , as the observed spectra are then limited by intrinsic source blending and not by the instrument . line ( solid line ) as a function of observing wavelength .
the mean flux of the brighter companion lines from each [ cii ] source is shown as a dotted line , while the dashed line shows the flux of competing emission from unrelated interloper lines .
[ sensitivity_fig ] , width=240 ] the line identification algorithm requires sufficient instrumental sensitivity to detect both the [ cii ] line as well as one or more companion lines .
figure [ sensitivity_fig ] shows the typical line flux as a function of observing wavelength .
for each [ cii ] source in the simulation , we compute the [ cii ] line intensity as well as the mean intensity of the four companion lines ( [ oi ] 146 @xmath0 m , [ oiii ] 88@xmath0 m , [ nii ] 205@xmath0 m , and [ nii ] 122@xmath0 m ) most often used by the algorithm for successful [ cii ] identifications .
to the extent that physical conditions in the interstellar medium remain constant , the ratio of companion to [ cii ] line intensity remains constant so that the observed line fluxes simply fall at the longer observing wavelengths in accordance with the higher source redshift .
interloper lines , however , include sources at multiple redshifts unrelated to the [ cii ] source redshift and thus fall more slowly with wavelength .
the companion lines are roughly a factor of 10 fainter than the [ cii ] line .
observations using the clean algorithm to determine source redshift must retain sufficient sensitivity to observe these fainter companion lines in addition to the targeted [ cii ] lines .
interloper lines become brighter than the companion lines for wavelengths @xmath63 m corresponding to [ cii ] source redshift @xmath64 .
observations at longer wavelengths targeting higher source redshifts require sufficient spectral resolution to distinguish the companion lines from the brighter interloper lines ( figure [ spectral_res ] ) .
the requirement to distinguish at least one identifiable companion line for each [ cii ] line implies that the sensitivity and integration times for a blind survey are determined by the fainter companion lines and not the targeted [ cii ] lines . to reduce the total observing time needed to map a given field , surveys are likely to employ multiple spectrometers simultaneously observing independent lines of sight .
such an arrangement has been envisioned for future surveys @xcite
the ubiquity and intensity of line emission from ionized carbon makes the [ cii ] line a promising probe of the high - redshift universe . we may thus anticipate future surveys using the [ cii ] line for redshifts @xmath65 to probe large scale structure , galaxy evolution , and the epoch of reionization @xcite .
instrumental requirements for such a survey , however , are not clear . observations from ground - based platforms ( e.g. alma or pdbi ) allow large collecting areas but are limited to the available atmospheric windows . even within these windows ,
atmospheric emission is 6 orders of magnitude brighter than the astrophysical foregrounds .
the resulting photon noise limits sensitivity , requiring integration times of 500 hours to detect a handful of sources within a cosmologically interesting volume @xcite .
observations from a balloon or space platform can mitigate atmospheric emission at the cost of limiting the size of the beam - forming optics .
the smaller collecting area reduces sensitivity to individual unresolved sources , while the coarser angular resolution increases confusion from multiple sources within each angular resolution element . @xcite
discuss intensity mapping from a balloon platform as one way to cope with these problems .
here we consider the complementary approach of a pencil - beam survey from a small millimeter - wave telescope .
a pencil - beam survey allows a single instrument to measure multiple galaxies within each line of sight , using spectroscopy to separate individual sources and determine source redshifts .
figure [ mjy_fig ] shows the number of [ cii ] lines that could be identified within a single pointing for a noiseless instrument .
we simulate a single 2 beam observed with spectral resolution @xmath66 over one octave in wavelength and use the clean algorithm to identify lines within the resulting spectra .
we examine 3 cases for spectral bands 300600 @xmath0 m , 500 - 1000 @xmath0 m , and 7001400 @xmath0 m corresponding to targeted [ cii ] redshifts @xmath10 = 2 , 4 , and 6 .
the upper panel shows the cumulative number count @xmath67 for all [ cii ] lines brighter than threshold @xmath68 that were identified by the algorithm .
the broad spectral band allows the number counts to include sources over a range of redshifts .
the bottom panel shows the number counts for the subset of detected lines within 20% of the target redshift .
a single pointing at 5@xmath69 threshold of 0.3 mjy can potentially identify 2040 redshifted [ cii ] lines ( plus a comparable number of companion lines ) within a 2 beam .
lines as a function of line intensity .
solid lines show the cumulative [ cii ] number counts derived from simulations of a single 2 beam observed over one octave in wavelength centered at redshifts 2 ( top curve , blue ) , 4 ( green ) , or 6 ( bottom curve , red ) . dashed lines show the number of false identifications .
companion lines are not shown .
( top ) total number of [ cii ] detections over the full wavelength band .
( bottom ) subset of detected lines with redshift lying within 20% of the target redshift .
[ mjy_fig ] , width=336 ] we illustrate the potential utility of a pencil - beam survey using a toy model of the star formation history .
we compare a model with no luminosity evolution to a second model for which the typical luminosity increases by a factor of 5 at redshift @xmath70 . figure [ luminosity_evolution ] shows the result .
the recovered luminosity distributions accurately reflect the toy model luminosity evolution .
such a pencil - beam survey easily distinguishes a model in which the luminosity peaks at @xmath71 from a model with no evolution .
the [ cii ] line is a principal coolant of the interstellar medium and is observable at cosmological distances .
however , the [ cii ] line is observed against a forest of competing lines from sources at multiple redshifts .
we use monte carlo simulations of the galaxy distribution at redshifts @xmath72 to generate realizations of the far - ir spectra including emission from 17 of the brightest far - ir lines .
the resulting spectra provide a test bed to evaluate the use of the [ cii ] line as a cosmological probe in blind pencil - beam surveys .
the superposition of rotationally - broadened lines from multiple galaxies at different redshifts creates a complex spectrum . the simplest assumption , that the brightest lines are always [ cii ] lines , is at best a coarse approximation to the simulated spectra .
the fraction of [ cii ] lines among the brightest 10% of all observed lines falls from 0.8 for observations near 315 @xmath0 m ( @xmath45 ) to 0.4 at 1100 @xmath0 m ( @xmath46 ) .
line identification based solely on relative intensity produces a high error rate .
a variant of the radio - astronomical clean algorithm yields superior results .
the algorithm selects the brightest line from an observed set of spectral lines , tentatively identifies the line as [ cii ] , and uses the resulting source redshift to predict the observed wavelengths for redshifted line emission from other bright far - ir lines from the same source .
if the predicted lines are observed , the source identification is confirmed and the [ cii ] line and redshifted companion lines are flagged and removed from the list of unknown lines . if the predicted lines are not observed , the source identification is not confirmed and the bright source line remains on the list of unknown lines for possible later identification as a companion line for a source at a different redshift .
the algorithm then proceeds to the next brightest remaining line and iterates until all lines have been processed .
a typical spectrum for 2 angular resolution contains some 1600 spectral lines within one octave , originating from 550 different galaxies .
the superposition of rotationally broadened lines reduces this to roughly 390 distinct peaks in the observed spectrum , 110 of which result from a single un - blended [ cii ] line .
the line identification algorithm correctly identifies 40 of the 110 [ cii ] line peaks for an efficiency of order 36% ( not counting companion lines identified from the same host galaxies ) .
chance alignments of lines lead the algorithm to incorrectly identify 3 peaks as [ cii ] lines ; however , the incorrect identifications result almost exclusively from fainter candidate peaks . restricting the algorithm to the brightest third of the observed peaks ( corresponding roughly to threshold 0.3 mjy ) increases the percentage of correct identifications above 90% .
correct line identification requires detection of both the targeted `` parent '' line and one or more `` companion '' lines from the same source .
survey sensitivity thresholds are thus determined by the expected intensities of the companion lines and not solely by the brighter [ cii ] line .
line over one octave in wavelength at spectral resolution @xmath66 .
the recovered luminosity distribution ( solid line ) accurately follows the toy model ( dotted curve ) .
the dashed line shows the recovered luminosity distribution for a model without luminosity evolution .
[ luminosity_evolution ] , width=259 ] the ability to distinguish individual peaks in the superposed spectra depends on the spectral resolution and angular resolution . decreasing the angular resolution increases the number of lines in the observed spectra , requiring higher spectral resolution to minimize blending of neighboring lines .
the required resolutions are modest .
spectral resolution @xmath59 resolves 90% of the peaks for 20 angular resolution , while @xmath60 resolves 90% of the peaks for 10 angular resolution or coarser .
the [ cii ] line is a promising cosmological probe .
the large number of potential sources observed within a single pencil beam combined with an efficient and accurate line identification algorithm allow rapid characterization of the galaxy luminosity distribution to redshifts encompassing reionization .
a simulated pencil - beam survey demonstrates the ability to distinguish toy models of galaxy evolution within a single angular resolution element .
the required angular resolution , sensitivity , wavelength range , and spectroscopic resolution appear achievable , suggesting that far - ir lines may soon provide a new tool to map the large - scale galaxy distribution and evolution for redshifts @xmath73
|
far infrared cooling lines are ubiquitous features in the spectra of star forming galaxies
. surveys of redshifted fine - structure lines provide a promising new tool to study structure formation and galactic evolution at redshifts including the epoch of reionization as well as the peak of star formation . unlike neutral hydrogen surveys , where the 21 cm line is the only bright line , surveys of red - shifted fine - structure lines suffer from confusion generated by line broadening , spectral overlap of different lines , and the crowding of sources with redshift .
we use simulations to investigate the resulting spectral confusion and derive observing parameters to minimize these effects in pencil - beam surveys of red - shifted far - ir line emission .
we generate simulated spectra of the 17 brightest far - ir lines in galaxies , covering the 150 to 1300 @xmath0 m wavelength region corresponding to redshifts @xmath1 , and develop a simple iterative algorithm that successfully identifies the 158@xmath0 m [ cii ] line and other lines .
although the [ cii ] line is a principal coolant for the interstellar medium , the assumption that the brightest observed lines in a given line of sight are always [ cii ] lines is a poor approximation to the simulated spectra once other lines are included .
blind line identification requires detection of fainter companion lines from the same host galaxies , driving survey sensitivity requirements .
the observations require moderate spectral resolution @xmath2 with angular resolution between 20 and 10 , sufficiently narrow to minimize confusion yet sufficiently large to include a statistically meaningful number of sources .
|
a core - collapse supernova is a colossal explosion resulting from the death of a massive star . at its peak
, a supernova can outshine the combined output of the billion or so stars in its host galaxy .
the star s demise is brought about by an accumulation of iron the final product of stellar burning in the star s core . collapsing catastrophically
, the core increases in average density by about a factor of a million in only half a second .
this usually results in a hot and dense neutron star with a mass about the same as the sun , but compressed to a size of tens of kilometers .
( the result of collapse in the most extreme cases is a black hole . )
the gravitational potential energy released during collapse is initially stored as heat in the nascent neutron star , but a small fraction of this energy eventually gives rise to the spectacular disruption of the entire star . about 99% of the energy escapes directly in the form of neutrinos , ghostly particles with very weak interactions .
the explosion is caused by the remaining 1% of this energy , which is transferred to the outer layers of the star through some yet - to - be - elucidated combination of heating by neutrinos and magnetohydrodynamic processes .
the high dimensionality of the core - collapse supernova problem is a result of the need to do neutrino transport . inside the dense neutron star ,
neutrinos diffuse slowly out of the core , and their angular distribution is isotropic ; but after escaping the neutron star , their angular distribution becomes more and more strongly forward - peaked .
the transition between these two regimes occurs in the crucial region between the neutron star and the shock resulting from the halt of collapse of the inner core ; this is the region in which neutrino heating may reenergize the stalled shock and allow the explosion to proceed . because the heating rate depends on the energy and angle distributions of the neutrinos
, one should ideally do `` neutrino radiation transport , '' in which these distributions of the neutrinos are tracked at every point in space a six - dimensional problem .
the computational difficulty of six - dimensional neutrino radiation hydrodynamics has spawned a variety of approximations , provoking a long history of uncertainty in the core - collapse supernova explosion mechanism @xcite .
recognizing the possible impact of convection and rotation on the efficiency and distribution of neutrino heating , some groups have focused on these while treating neutrino transport in a crude , averaged manner , seeing explosions as a result .
other groups have focused on doing more accurate neutrino transport , which required a restriction to spherical symmetry ; in this case no explosions were seen .
only now are computations with three total dimensions becoming routine ; soon we hope to push the envelope , with simulations in five dimensions : two space dimensions , together with a modest discretization of all three momentum space dimensions .
the computational demands of radiation hydrodynamics can be ameliorated by ` adapative mesh refinement ' ( amr ) .
the basic idea of amr is to employ high resolution only where needed , in order to conserve memory and computational effort .
our current expectation is to allow for refinement only in the space dimensions .
this will help with the management of two difficulties : the large dynamic range in length scales associated with the density increase of six orders of magnitude that occurs during core collapse , and adequate resolution of particular features of the flow ( the shock , for instance ) . with eulerian codes in multiple space dimensions ,
these tasks require high resolution ; and particularly for radiation hydrodynamics , the savings achievable by reducing the number of zones is considerable , since an entire three - dimensional momentum space is carried by each spatial zone . in implementing the zone - by - zone - refinement approach to the representation of spacetime
, we have tried to follow object - oriented design principles to the extent allowed by fortran 90/95 .
figure [ zone ] outlines the basic data structures we use to model the ideal of a continuous spacelike slice with a discretized approximation .
( the hierarchy of structures , and our operations on them with well - controlled interfaces , are instances of the object - oriented principles of _ inheritance _ and _ encapsulation_. ) a region of a spacelike slice is represented by an object of zonearraytype .
each such object contains an array of objects of zonetype , along with information about the coordinates of the zones and pointers to neighboring zone arrays .
each zone , an object of zonetype , contains various forms of stress - energy , each of which is a separate object .
figure [ zone ] shows a perfect fluid and a radiation field ; in the code we have an electromagnetic field as well .
each zone has a pointer to another object of zonearraytype , whose allocation constitutes refinement of that zone ; this structure can be extended to arbitrarily deep . a two - dimensional hydrodynamics test problem computed with our adaptive mesh code
is shown in figure [ initialfinal2d ] . the word `` general '' that goes into the name of our code , _
genasis_for _ gen_eral _ a_strophysical _ si_mulation _ s_ystem refers to the use of fortran 90 s facility for function overloading .
( this is an instance of the object - oriented principle of _
polymorphism_. ) this allows a generic function name to have several different implementations , providing for extensibility of the physics : different equations of state , hydrodynamic flux methods , coordinate systems , gravity theories , and so forth can be employed by adding new implementations of generic function names , without having to go back and change basic parts of the code to implement new physics .
in solving the equations of self - gravitating magnetohydrodynamics , we use a conservative formulation of hydrodynamics and a constrained transport approach to the magnetic induction equation .
the physical meaning of a conservative equation is that ( modulo source terms ) the time rate of change of a conserved quantity in a volume is equal to a flux through the volume s enclosing surface .
this meaning is built into our finite - difference representation of the hydrodynamics equations ; divergences are represented in discrete correspondence to their mathematical definition , using zone volumes and face areas .
our discretization of the poisson equation for the gravitational potential takes a similar form , since the laplacian operator can be expressed as the divergence of a gradient . in our method of constrained transport , the curl of the electric field in the induction equation is represented in discrete correspondence to the mathematical definition of the curl , using zone face areas and edge lengths .
the stokes and divergence theorems then ensure a divergence - free magnetic field as required by maxwell s equations .
our solution methods are second order in space and time .
accurate computation of fluxes at zone faces and electric fields at zone edges is a key feature .
we employ so - called ` central schemes , ' which are able to capture shocks with an accuracy comparable to riemann solvers , but with much greater simplicity .
we use a second - order runge - kutta time stepping algorithm , made possible by the semi - discrete formulation of the central scheme . in doing
so we evolve all levels of the mesh synchronously , avoiding problems we encountered in self - gravitating systems with ` asynchronous ' evolution .
parallelization is achieved by giving each processor its share of spatial zones at each level of refinement .
we rely on the petsc library for the distributed matrix inversion required by the gravity solver .
we have been working with a number of hydrodynanic and magnetohydrodynamic test problems , one of which is shown here .
the rotor problem which consists of a rapidly rotating dense fluid , initially cylindrical , threaded by an initially uniform magnetic field was devised to test the onset and propagation of strong torsional alfvn waves into the ambient fluid .
we have computed a version of the rotor problem with initial data identical to a so - called second rotor problem, @xcite and display the results in figure [ rotorcontours ] .
@xmath0{fastrotordensity40lines.eps } & \includegraphics[width=4in]{fastrotorpressure40lines.eps } \end{array}$ ]
neutrino radiation transport is the greatest computational challenge in supernova simulations .
neutrino distributions must be tracked in order to compute the transfer of lepton number and energy between the neutrinos and the fluid .
one challenge is constructing a formalism @xcite and discretization that allows both energy and lepton number to be conserved to high precision .
two other challenges are associated with the limits of computational resources : the solution of a very large nonlinear system of equations @xcite , and neutrino interaction kernels of high dimensionality .
a large system of nonlinear equations requiring inversion ( as opposed to explicit updates ) results from the disparity between hydrodynamic and particle interaction time scales , which motivates implicit time evolution .
the nonlinear solve is achieved with the newton - raphson method . a fixed - point method employing a preconditioner that splits the space and momentum space couplings
is used for the linear solve required within each newton - raphson iteration .
because neutrino interactions are expensive to compute on - the - fly , we have implemented interpolation tables . particularly for neutrino scattering and pair interactions which depend on neutrino states before and after the collision the interaction kernels are of high dimensionality , requiring a globally distributed table .
we have made a promising start on _ genasis , _ a new code being developed to study the explosion mechanism of core - collapse supernovae .
our plan is to include all the relevant physics including magnetohydrodynamics , gravity , and energy- and angle - dependent neutrino transport in a code with adaptive mesh refinement in two and three spatial dimensions @xcite .
parallelization and implementation on the adaptive mesh are not yet complete , and the physics components have not yet been fully integrated ; but steady progress and the successful completion of test problems give us confidence that we are well on our way towards a tool that will provide important insights into the supernova explosion mechanism .
we gratefully acknowledge s. w. bruenn s contribution of subroutines for the computation of neutrino interaction kernels .
we thank r. d. budiardja and m. w. guidry for discussions on the poisson solver , and r. d. budiardja for helping with that solver s interface to the petsc library . this work was supported by scientific discovery through advanced computing ( scidac ) , a program of the office of science of the u.s .
department of energy ( doe ) ; and by oak ridge national laboratory , managed by ut - battelle , llc , for the doe under contract de - ac05 - 00or22725 .
9 cardall c y , razoumov a , endeve e , and mezzacappa a 2005 in _ open issues in core collapse supernova theory _ ed a mezzacappa and g m fuller ( singapore : world scientific ) in press ( _ preprint _ astro - ph/0510704 )
|
the computational difficulty of six - dimensional neutrino radiation hydrodynamics has spawned a variety of approximations , provoking a long history of uncertainty in the core - collapse supernova explosion mechanism . under the auspices of the terascale supernova initiative
, we are honoring the physical complexity of supernovae by meeting the computational challenge head - on , undertaking the development of a new adaptive mesh refinement code for self - gravitating , six - dimensional neutrino radiation magnetohydrodynamics .
this code called _ genasis , _ for _ gen_eral _ a_strophysical _ si_mulation _ s_ystem is designed for modularity and extensibility of the physics .
presently in use or under development are capabilities for newtonian self - gravity , newtonian and special relativistic magnetohydrodynamics ( with ` realistic ' equation of state ) , and special relativistic energy- and angle - dependent neutrino transport including full treatment of the energy and angle dependence of scattering and pair interactions .
|
the study of variations in solar activity is important for understanding the mechanism behind the solar activity , and because of solar activity affects space weather and it may also have a contribution in earth s climate change ( e.g. , * ? ? ? * ; * ? ? ?
although it is well believed that the solar cycle results from generation of strong toroidal magnetic fields in the sun , with a main period of about 22-years , by combined effects of convection and differential rotation in the sun ( @xmath7 , * ? ? ?
* ) , it is not known exactly what causes solar cycle .
because of the long record of observations , sunspot measurements constitute a primary source of information to better understand the level and nature of solar activity .
it is well - known that the dynamic properties , such as rotation rate and meridional motion , of sunspot groups depend on their sizes .
this may imply the dynamics of the sun s subsurface layers in which the different size sunspot groups anchor / generated @xcite .
therefore , studies on the variations in the sizes of sunspot groups are important , which may provide important clues on the underlying mechanism of solar activity and its variations . in some studies it is found that the sunspot area distribution does not depend on the zrich or international sunspot number ( @xmath3 ) ( @xmath7 , * ? ? ?
recently , we have analyzed the combined greenwich and soon sunspot group data during the period 18742011 and studied variations in the annual numbers ( counts ) of the small , large , and big sunspot groups , by classifying the sunspot groups on the basis of their maximum areas ( * ? ? ?
* hereafter paper i ) .
although , it is found that the number of the sunspot groups in each class shows the 11-year periodic variation of solar cycle , some noticeable differences are seen in the variations of the numbers of the sunspot group in different classes . in the present paper we extended that analysis and studied long - term variations in the mean maximum size of each of the three classes of sunspot groups and
also that of all the sunspot groups in the combined class of the three classes . in the next section we will describe the data and the method of analysis . in section 3 we will present the results and in section 4 we will present conclusions and a brief discussion .
here the data and the method of analysis are the same as in paper i. we have used the combined greenwich and soon sunspot group data during the period may-1874 to may-2011 ( taken from ) .
these data include the observation time ( the greenwich data contain the date with the fraction of a day , in the soon data the fraction is rounded to 0.5 day ) , heliographic latitude and longitude , central meridian distance ( cmd ) , and corrected umbra and whole - spot areas ( in msh ) , _ etc .
_ , of sunspot groups for each day of observation .
the positions of the sunspot groups are geometrical positions of the centers of the groups .
the greenwich data ( may-1874 to december-1976 ) have been compiled from the majority of the white light photographs which were secured at the royal greenwich observatory and at the royal observatory , cape of good hope .
the gaps in their observations were filled with photographs from other observatories , cape town , south africa ; kodaikanal , india ; and mauritius .
the soon data ( january-1977 to may-2011 ) include measurements made by the united states air force ( usaf ) based on sunspot drawings obtained by a network of the observatories in boulder , hawaii , and so on .
david hathaway scrutinized the greenwich and soon data and produced a reliable continuous data series from 1874 up to date . in the case of soon data , we increased the corrected whole - spot areas by a factor of 1.4 .
david hathaway found this correction was necessary to have a combined homogeneous greenwich and soon data .
however , the agreement between these two data sets may be still not 100% @xcite .
the combined greenwich and soon sunspot group data are the largest available , reliable data that include the positions and the areas of sunspot groups . we have used the data on only those sunspot groups whose birth and death occurred within a disk passage .
that is , we have not used the sunspot groups whose central meridian distance @xmath8 in any day of their respective life times .
this reduces the foreshortening effect and helps to obtain the maximum area of a sunspot group unambiguously . if @xmath9 , @xmath10, ,@xmath11 denote the areas ( corrected for the foreshortening effect ) of all the sunspots in a sunspot group observed at times @xmath12 , @xmath13,
,@xmath14 during the life time of the sunspot group @xmath15 ( days ) , then the maximum area is defined as follows : @xmath16 where @xmath17 .
each appearance of a recurrent group is treated as an independent group .
thus , @xmath18 days .
we have used here only the sunspot groups which had @xmath19 day ( soon data do not contain the data on the sunspot groups whose @xmath20 day ) . on the basis of @xmath21 values we have classified sunspot groups into three classes as small sunspot groups ( ssgs : @xmath0 msh ) , large sunspot groups ( lsgs : @xmath22 msh ) , and big sunspot groups ( bsgs : @xmath2 msh ) .
we determined the mean maximum sizes ( sum of the maximum areas divided by the number of groups ) @xmath23 , @xmath24 , and @xmath25 of ssgs , lsgs , and bsgs respectively , for each year during the period 18742011 as follows : @xmath26 @xmath27 where nsg , nlg , and nbg are the numbers ( counts ) of ssgs , lsgs , and bsgs , respectively , in a given year ( @xmath28 ) .
the mean maximum size ( @xmath29 ) of all sunspot groups ( asgs ) of the same year is calculated as follows : @xmath30 we have also calculated the mean maximum size of each of the above classes of sunspot groups over a solar cycle ( @xmath31 , the sum of the yearly means divided by the number of years of a cycle ) as follows : @xmath32 where wcn and @xmath33 represent the waldmeier cycle number and length ( in years ) of a cycle , respectively ; c represents any of the above classes of sunspot groups , and @xmath34 is a year of the corresponding cycle . the uncertainty ( standard error ) in @xmath31
is also determined .
we determined the variations in @xmath23 , @xmath24 , @xmath25 , @xmath29 , and @xmath31 .
1 shows the variations in @xmath23 , @xmath24 , @xmath25 , and @xmath29 ( @xmath35 , annual variations in the mean maximum sizes of ssgs , lsgs , bsgs , and asgs , @xmath36 , eqs .
( 2 ) and ( 3 ) ) during the period 18742011 ( a value equal to zero implies the absent of sunspot groups in the corresponding class ) . in the same figure we have also shown the variation in the annual mean @xmath3 , whose values are taken from the website , ftp://ftp.ngdc.noaa.gov/stp/solar_data/sunspot_numbers/international/yearly/year.plt ) . as can be seen in this figure the mean maximum size of the sunspot groups in each class varies considerably , but the patterns of the variations considerably differ with the patterns of the corresponding 11-year cycles in @xmath3 .
there is no significant correlation between the mean maximum size of any class of sunspot groups and @xmath3 ( in the case of big groups the value of the correlation coefficient is largest and equal to 0.315 ) .
the mean maximum size of the sunspot groups in any class does not have a 11-year solar cycle pattern
2 and 3 show the comparisons of @xmath23 and @xmath24 with @xmath3 during different solar cycles , using superposed epochs related to the respective maximum epochs of the cycles ( corresponding figures of @xmath25 and @xmath29 are not shown ) . as can be seen in fig . 3
, there is a pattern of an approximate 9-year cycle in @xmath24 .
the similar pattern is also found in the case of @xmath25 , but in @xmath23 it can not be seen clearly ( see fig . 2 ) .
no clear such pattern seen in @xmath29 also .
that is , in both of these cases the variations during different solar cycles are highly different .
the mean variation ( represented by the solid curve ) is statistically highly insignificant , @xmath35 the errors bars are too large . figs .
4(a)4(d ) show the variations in @xmath31 ( @xmath36 , eq .
( 4 ) ) of the sunspot groups in each class ( @xmath35 , cycle - to - cycle variations in the mean maximum sizes of ssgs , lsgs , bsgs and asgs ) . as can be seen in this figure @xmath31 of ssgs varies systematically on a time scale of about 12 cycles , suggesting the existence of a strong 130 or more years cycle in @xmath31 , whose minimum was taken place during cycles 1516 .
the long - term variations in @xmath31 of lsgs and bsgs suggest the existence of @xmath4 44-year cycles in these parameters , but the variations are not well defined ( the error bars are large ) . a 44-year periodicity was seen in sunspot activity @xcite . from cycle 12 to 20
the pattern of the variation in @xmath31 of asgs ( see fig .
4(d ) ) is largely same as the corresponding variations in @xmath31 of ssgs , whereas it is drastically different during the last three cycles in which it seems the sunspot groups were relatively small ( this could be due the areas of sunspot groups of soon data may not be increased adequately , see section 3 ) .
the sunspot groups in cycles 12 and 17 were relatively large , whereas in cycles 14 and 20 ( relatively small and long cycles ) they were relatively small , suggesting the existence of a 5560 year cycle in @xmath31 of asgs .
incidentally , the values of @xmath31 of ssgs of cycles 12 and 23 are equal ( see fig .
4(a ) ) . in the case of cycle 23 ,
the high value of @xmath31 of ssgs was largely contributed by the sunspot groups in the decline phase , whereas in case of cycle 12 the high value of @xmath31 of ssgs was largely contributed by the sunspot groups in the rise phase ( see fig
in fact , in the former case the dominance in the mean maximum sizes of ssgs was continued from the decline phase of cycle 11 . during this period of cycle 11
the mean maximum sizes of all the three classes of sunspot groups were relatively larger , whereas in the case of the decline phase of cycle 23 only the mean maximum sizes of ssgs were relatively larger .
that is , during this period of cycle 23 ( and some extent also during the rise of cycle 12 ) there was a scarcity in the sunspot groups whose areas @xmath3737 msh .
this can be seen in fig .
5 , in which solar cycle variations in the annual numbers of the small sunspot groups whose @xmath38 msh and @xmath39 msh are shown . during 20002006
the values that are represented by the red squares ( connected by dotted curves ) in 5(a ) are lower than the corresponding values in 5(b ) .
6 shows the cycle - to - cycle variations in the mean annual numbers of these two classes of the small sunspot groups .
as can be see in this figure in cycle 23 on the average the annual mean number of sunspot groups whose @xmath38 msh is relatively less than those of whose @xmath39 msh .
cycle 23 was unusually long ( about 12.4 years ) and followed by the unusually deep and prolonged minimum . when the sun reached the minimum activity between cycles 22 and 23 in 1996
, it seems the low - latitude coronal holes were almost completely absent , whereas in 2007 and 2008 , in spite of the extremely low sunspot activity , coronal holes at low - latitudes were still relatively large ( e.g. * ? ? ? * ) .
the occurrence of bsgs and lsgs stopped in 2004 and 2006 , respectively , whereas the occurrence of ssgs continued ( see fig . 1
, also see * ? ? ?
7 shows the plot of yearly mean sizes of ssgs during the preceding minimum epochs of cycles 1223 and around the minimum between cycles 23 and 24 ( @xmath35 , in years 2006 , 2007 , 2008 and 2009 ) versus the corresponding years .
as can be seen in this figure in 2007 the mean size of ssgs whose @xmath39 msh is relatively large , and in 2008 the overall mean size of all ssgs are relatively large . in 2007 the mean sizes of both the two classes of ssgs are relatively larger than corresponding values of in 2008 , whereas in the case of the combined class of ssgs the value at 2008 is relatively large ( this is because the number of sunspot groups with @xmath38 are relatively less in 2008 ) .
the presence of relatively large coronal holes at low - latitudes in 2007 and 2008 may be related to the presence of relatively large size ssgs in these years . since the sunspots block the the sun s radiative output , the existence of the relatively large size ssgs in the minimum between cycles 23 and 24 ( the evolution rate of ssgs is also relatively low ,
* ) my be also responsible for the existence of very low total solar irradiance during this period @xcite .
from the above analyses of a large set of sunspot group data the following conclusions can be drawn : 1 .
there is no significant correlation between the yearly mean maximum size of the sunspot groups in any class and @xmath3 .
a pattern of @xmath4 9-year cycle is seen in the variations of the mean maximum sizes of lsgs and bsgs with phases of solar cycles .
3 . there exists a strong 130 or more years cycle in the variation of the mean maximum size of ssgs , whereas there is a hint on the existence of @xmath4 44-year cycles in the variations of the mean maximum sizes of lsgs and bsgs .
4 . during the decline phase of cycle 23
, there was a scarcity in the sunspot groups whose areas @xmath3737 msh .
5 . during the minimum between cycles 23 and 24
may be due to the presence of a number of small sunspot groups whose @xmath6 was larger than that of whose @xmath537 , the relatively large size coronal holes were present at low - latitudes and the total solar irradiance was very low .
it is believed that the area of a sunspot or a sunspot group has a better physical significance than @xmath3 because the area is a better measure ( proxy ) of solar magnetic flux than @xmath3 @xcite ; an area of 130 msh ( 1 msh @xmath40 km@xmath41 ) corresponds approximately to @xmath42 mx ( maxwell ) @xcite .
the annual sum of the areas of sunspot groups reasonably well correlate to the annual average @xmath3 @xcite .
the sum of the areas of sunspots and sunspot groups may be representing the total magnetic flux of the sunspots and sunspot groups rather than @xmath3 .
it may be worth to note that @xmath3 is not weighted with the areas of sunspots and sunspot groups , @xmath35 it gives an equal weight to all sunspots and to all sunspot groups regardless of their actual size ( @xmath7 * ? ? ? * ) .
the yearly mean @xmath3 is determined from the corresponding monthly means , @xmath35 , the sum of monthly means divided by number of months ( it seems to be always 12 , in spite of in some months the values of the sums are zero , @xmath35 absent of sunspots and sunspot groups ) . here
the yearly mean size of sunspot groups is directly determined ( @xmath36 , eq .. ( 2 ) ) .
that is , there is a difference in the methods of determinations of the yearly mean size of sunspot groups and the yearly mean of @xmath3 .
the annual number of asgs reasonably well correlate to @xmath3 ( the variations in nsg , nlg and nbg considerably differ with that of @xmath3 during some solar cycles , see figs .
24 in * ? ? ?
the sum of the areas of any class of sunspot groups reasonably correlate to @xmath3 .
the good correlations of both the number and the sum of the areas of asgs with @xmath3 implies that in a given time interval ( year ) the areas of a large number of sunspot groups are not drastically different ( a few or no outliers ) .
it is well - known that the relative frequency of small sunspot groups is larger during solar minimum than during solar maximum @xcite .
however , the smallest regions dominate the global flux emergence rate @xcite , and small sunspots and small sunspot groups outnumber the large ones @xcite .
this reduces the overall mean sizes of the sunspots groups in the maximum years ( for example , there exits an high anticorrelation between the variations that are represented by the solid curves in figs .
4(a ) and 6 ) .
all these together may be constitute a possible reason for there is no statistical significant correlation between the yearly mean maximum size of sunspot groups in any class and @xmath3 ( conclusion 1 above ) . may be for the same reason the annual mean daily rates of growth and decay of sunspot groups do not correlate to the annual @xmath3 @xcite , whereas the annual rates of the growth and decay of sunspot groups very well correlate to the annual @xmath43 @xcite .
the existence of approximate 90-year ( gleissberg cycle ) and 160270 year @xcite periodicities in sunspot activity is known . in paper
i these periodicities were seen ( visualized ) in the variations of the numbers ( counts ) of lsgs / bsgs and ssgs ( see also fig . 5 ) , respectively .
the existence of approximate 130-year periodicity ( which is seen here in the variations of the mean maximum size of ssgs , conclusion 3 above ) in some solar activity indices is also known .
in addition , the period of the well - known gleissberg cycle is reported to be very wide , 60140 years @xcite .
the aforementioned all periodicities seem to be related to the subharmonics of the 22-year solar magnetic cycle @xcite . @xcite and @xcite , from the spherical fourier analysis of a large set of sunspot group data , found that solar variability may be the resultant of the superpositions of global modes of solar magneto - hydrodynamic oscillations . @xcite
connected @xmath4 180-year period of solar magnetic activity to the 179-year period of the sun s motion around the solar system barycenter .
@xcite found that violations of the gnevyshev and ohl rule or g - o rule @xcite in sunspot cycles may be related to the sun s retrograde motion around the solar system barycenter , and he has suggested a role of the sun s spin - orbit coupling in solar variability .
nevertheless , the origin of long - term variations in solar cycle is still an unsolved problem ( @xmath7 , * ? ? ?
the large / big / strong sunspot groups / magnetic - field - regions may be associated with a deep dynamo mechanism , whereas the small / weak sunspot groups / magnetic - field - regions may be associated with a near surface dynamo mechanism @xcite .
the small sunspot groups may be the fragmented or the branched parts of the large / big sunspot groups @xcite .
the 130-year and 44-year periodicities seen in the variations of the mean maximum sizes of ssgs and lsgs / bsgs ( conclusion 3 above ) may be related to the near surface ( local ) and deep ( global ) dynamo mechanisms , respectively .
the conclusion 4 above , is consistent with the result that deficiency of the very small sunspots since the activity maximum around 2000 @xcite , and it may be related to a very low growth rate of sunspot groups during this period of cycle 23 @xcite .
the area of a large / small sunspot group ( magnetic flux ) decreases in a fast / slow rate and in fact , it seems the amounts of growth and decay of the magnetic flux of the sunspot groups in a given time interval depends on ( proportional to ) the total amount of the flux in that interval @xcite .
the growth of sunspot groups ( largely contributed by the emergence of new magnetic flux ) may be mostly related to the generation mechanism of solar activity , whereas the decay of sunspot groups may be related to the magnetic flux diffusion and magnetic reconnection processes .
the total amount of flux of cycle 23 is relatively small ( compared to the recent cycles ) and it may be related to the slow growth of sunspot groups during this cycle ( relatively less emergence of new magnetic flux , which is in addition due to the low efficiency of dynamo , my be due to a strong effect of the coriolis force on the rising / emerging magnetic flux , * ? ? ?
the average size of bsgs / lsgs is relatively small in this cycle ( see figs
. 4(b ) and 4(c ) ) .
@xcite found that the average area of a sunspot that belongs to a large sunspot group decreased in solar cycle 23 , and interpreted this result as the large sunspot groups of cycle 23 could be more complex and may have a large number of small sunspots .
the arguments above imply that on the average the sunspot groups of cycle 23 are small due to a low supplement of newly emerging magnetic flux , a contradiction to the aforementioned conclusion of @xcite .
the relatively long length of this cycle may be related to the slow growth rate ( and also relatively slow decay / fragmentation of bsgs / lsgs of this cycle because of their relatively small size , * ? ? ?
the relatively slow rate of the fragmentation may be also contributed for the less number of small sunspot groups in this cycle .
it may be interesting to note here that the cycle pair ( 22 , 23 ) violated the g - o rule mainly due to a large scarcity of ssgs in cycle 23 @xcite .
all these may be related to the different behavior of the mean merdional motion of the sunspot groups , @xmath35 during this cycle the mean meridional motion of sunspot groups was high and its behaviour was quite different , south - bound during minimum years and north - bound during maximum years @xcite .
ahluwalia , h.s . ,
ygbuhay , r.c .
sunspot cycle 24 and the advent of dalton - like minimum . adv .
2012 , article i d 126516 ( 5 pages ) , 2012 .
attolini , m.r . ,
cecchini , s. , galli , m. , nanni , t. the subharmonics of the 22-year solar cycle .
il nov s cimento 13c , 131138 , 1990 .
babij , v.p . ,
efimenko , v.m . ,
lozitsky , v.g .
statistical characteristics of large sunspots in solar activity cycles 1723 . kinematics and physics of celestial bodies 27 , 191196 , 2011 .
clette , f. , lefvre , l. are the sunspots really vanishing ? .
anomalies in solar cycle 23 and implications for long - term models and proxies
. j. space weather space clim .
2 , a06 ( 9 pages ) , 2012 . de toma , g. evolution of coronal holes and implications for high - speed solar wind during the minimum between cycles 23 and 24 .
solar phys .
, 274 , 195217 , 2011 .
dikpati , m. , gilman , p.a . simulating and predicting solar cycles using a flux - transport dynamo .
j. 649 , 498514 , 2006 .
fan , y. , fisher , g.h .
, deluca , e.e . the origin of morphological asymmetries in bipolar active regions .
j. 405 , 390401 , 1993 .
frhlich , c. evidence of a longterm trend in total solar irradiance .
501 , l27l30 , 2009 .
gilman , p.a . ,
foukal , p.v .
angular velocity gradients in the solar convection zone .
j. 229 , 11791185 , 1979 .
gnevyshev , m.n . ,
ohl , a.i .
about 22-year cycle of solar activity .
25(1 ) , 1820 , 1948 .
gokhale , m.h . ,
javaraiah , j. sunspot activity as originating in interference of the sun s global mhd oscillations .
mnras 243 , 241251 , 1990 .
gokhale , m.h . ,
javaraiah , j. global modes constituting the solar magnetic cycle .
ii - phases , ` geometrical eigenmodes ' , and coupling of field behaviour in different latitudes . solar phys .
138 , 399410 , 1992 .
gokhale , m.h . , javaraiah , j. global modes constituting the solar magnetic cycle .
3 : ` shapes ' and ` sizes ' of the sunspot cycles and maintenance of mhd spectrum by energy cascade . solar phys . 156 , 157177 , 1995 .
gokhale , m.h . ,
javaraiah , j. , kutty , k.n . ,
varghese , b.a .
global modes constituting the solar magnetic cycle .
i - search for ` dispersion relations ' .
solar phys .
138 , 3547 , 1992 .
gokhale , m.h . ,
sivaraman , k.r .
the two components in the distribution of sunspot groups with respect to their maximum areas .
j. astrophys .
2 , 365377 , 1981 .
hathaway , d.h .
the solar cycle .
living reviews in solar physics 7 ( 1 ) ( 65 pages ) , 2010 .
hathaway , d.h . ,
choudhary , d.b .
sunspot group decay .
solar phys .
350 , 269278 , 2008 .
hiremath , k.m .
change of rotation rates of sunspot groups during their lifetimes : clues to the sites of origin of different flux tubes .
386 , 674680 , 2002 .
howard , r.f .
solar active regions as diagnostics of subsurface conditions .
astron . astrophys .
34 , 75110 , 1996 .
javaraiah , j. dependence of the meridional motions of sunspot groups on life spans and age of the groups , and on the phase of the solar cycle .
solar phys .
189 , 289304 , 1999 .
javaraiah , j. relationship between magnetic structures of small and large sunspot groups . in : pevtsov , a.a .
, uitenbroek , h. ( eds . ) , current theoretical models and future high resolution solar observations : preparing for atst .
cs-286 , astro .
sanfrancisco , pp . 325332 , 2003 .
javaraiah , j. sun s retrograde motion and violation of even - odd cycle rule in sunspot activity .
mnras 362 , 13111318 , 2005 .
javaraiah , j. predicting the amplitude of a solar cycle using the north - south asymmetries in the previous cycle : ii .
an improved prediction for solar cycle 24 .
solar phys . , 252 , 419439 , 2008 .
javaraiah , j. long - term variations in the mean meridional motion of the sunspot groups .
astron . astrophys .
509 , a30 ( 10 pages ) , 2010 .
javaraiah , j. long - term variations in the growth and decay rates of sunspot groups .
solar phys . , 270 , 463483 , 2011 .
javaraiah , j. the g - o rule and waldmeier effect in the variations of the numbers of large and small sunspot groups .
solar phys . , 281 , 827837 ( paper i ) , 2012a .
javaraiah , j. solar cycle variations in the growth and decay of sunspot groups .
space sci .
338 , 217226 , 2012b .
javaraiah , j. , gokhale , m.h .
estimation of the depths of initial anchoring and the rising - rates of sunspot magnetic structures from rotation frequencies of sunspot groups .
327 , 795 - 799 , 1997 .
jose , p.d .
sun s motion and sunspots .
j. 70 , 193200 , 1965 .
kilcik , a. , yurchyshyn , v.b . ,
abramenko , v. , goode , p.r . ,
ozguc , a. , rozelot , j.p . ,
cao , w. time distributions of large and small sunspot groups over four solar cycles .
j. 731 , 30 ( 8 pages ) , 2011 .
lefvre , l. , clette , f. a global small sunspot deficit at the base of the index anomalies of solar cycle 23 .
536 , l11 ( 4 pages ) , 2011 .
nagovitsyn , y.a . a nonlinear mathematical model for the solar cyclicity and prospects for reconstructing the solar activity in the past .
astron . lett .
23 , 742748 , 1997 .
mandrykina , t.l .
distribution of sunspot groups from maximum areas .
lvov , 49 , 2427 , 1974 .
meunier , n. magnetic network dynamics : activity level , feature size and anchoring depth .
astron . astrophys .
436 , 10751086 , 2005 .
ogurtsov , m. g. , nagovitsyn , yu .
a. , kocharov , g. e. , and jungner , h. long - period cycles of the sun s activity recorded in direct solar data and proxies .
solar phys .
211 , 371394 , 2002 .
schove , d. j. sunspot cycles .
hutchinson ross publ .
co. , stroudsburg , pennsylvania , 1983 .
pevtsov , a.a . ,
nagovitsyn , y.a . ,
tlatov , a.g . ,
rybak , a.l .
long - term trends in sunspot magnetic fields .
j. 742 , l36 ( 4 pages ) , 2011 .
sivaraman , k.r . ,
gokhale , m.h .
anchor depths of flux elements and depths of flux sources in relation to the two rotation profiles of the sun s surface magnetic fields .
solar phys .
221 , 209223 , 2004 .
sivaraman , k.r . ,
gokhale , m.h . ,
sivaraman , h. , gupta , s.s . ,
howard , r.f . measurement of kodaikanal white - light images : relaxation of tilts of spot groups as indicator of subsurface dynamics of parent flux loops . astrophys .
j. 657 , 592599 , 2007 .
sivaraman , k.r . ,
sivaraman , h. , gupta , s.s . , howard , r.f . measurement of kodaikanal white - light images - vi . variation of rotation rate with age of sunspot groups .
solar phys .
214 , 6587 , 2003 .
sivaraman , k.r . ,
sivaraman , h. , gupta , s.s . ,
howard , r.f .
return meridional flow in the convection zone from latitudinal motions of umbrae of sunspot groups .
solar phys .
266 , 247259 , 2010 .
tan , b. multi - timescale solar cycles and the possible implications .
space sci .
332 , 6572 , 2011 .
wang , y .- m . , sheeley , n.r .
average properties of bipolar magnetic regions during sunspot cycle 21 .
solar phys .
124 , 81100 , 1989 .
zirin , h. weak solar fields and their connection to the solar cycle .
solar phys .
110 , 101107 , 1987 . ,
@xmath24 , @xmath25 , and @xmath29 , @xmath35 the annual mean maximum sizes of the small ( cross - dashed curve , green ) , large ( filled circle - solid curve , black ) , big ( triangle - three dotted - dashed curve , red ) , and all ( square - one dotted - dashed curve , blue ) spot groups divided by 1 , 5 , 20 , and 4 , respectively , during the period 18742011 ( note : a value equal to zero implies the absent of sunspot groups in the corresponding class , the data in 1874 and 2011 are incomplete ) . the open circles connected by the dotted lines represents the values of @xmath3 divided by 2 .
the horizontal lines represent the respective mean values . near the maximum of each cycle
the corresponding waldmeier cycle number is indicated .
( for interpretation of the references to color in this figure legend , the reader is referred to the web version of this article . ) ] ) _ versus _ the year from the maximum epoch of the solar cycle .
the blue ( dashed curves ) and red ( dotted curves ) colors are used for even- and odd - numbered cycles , respectively .
( for the sake of clarity black color is used for cycle 11 . )
different symbols are used for different cycles ( numbers are given in the parentheses ) : asterisks ( 11 ) , pluses ( 12 and 13 ) , open - circles ( 14 and 15 ) , crosses ( 16 and 17 ) , diamonds ( 18 and 19 ) , triangles ( 20 and 21 ) , and squares ( 22 and 23 ) . the filled circles connected by the solid lines represent the mean solar cycle variation determined from the values of annual means .
the error bars represent the standard error .
there is only one data point at years -6 ( beginning of cycle 14 ) , 8 ( end of cycle 23 ) and 911 ( first three years of cycle 24 ) .
( for interpretation of the references to color in this figure legend , the reader is referred to the web version of this article . ) ] ) .
the blue square ( connected by dashed line)at year @xmath44 of cycle 22 and the red square ( connected by dotted line ) at year @xmath45 of cycle 23 ca nt be seen because the corresponding values are equal to zero due to the absent of lsgs .
( for interpretation of the references to color in this figure legend , the reader is referred to the web version of this article . ) ] , @xmath35 cycle to cycle variations in the mean maximum sizes ( in msh ) of ( a ) small sunspot groups ( ssgs ) , ( b ) large sunspot groups ( lsgs ) , ( c ) big sunspot groups ( bsgs ) , and ( d ) all sunspot groups .
the error bars represent the standard error.,title="fig:",width=207,height=226 ] , @xmath35 cycle to cycle variations in the mean maximum sizes ( in msh ) of ( a ) small sunspot groups ( ssgs ) , ( b ) large sunspot groups ( lsgs ) , ( c ) big sunspot groups ( bsgs ) , and ( d ) all sunspot groups .
the error bars represent the standard error.,title="fig:",width=207,height=226 ] , @xmath35 cycle to cycle variations in the mean maximum sizes ( in msh ) of ( a ) small sunspot groups ( ssgs ) , ( b ) large sunspot groups ( lsgs ) , ( c ) big sunspot groups ( bsgs ) , and ( d ) all sunspot groups .
the error bars represent the standard error.,title="fig:",width=207,height=226 ] , @xmath35 cycle to cycle variations in the mean maximum sizes ( in msh ) of ( a ) small sunspot groups ( ssgs ) , ( b ) large sunspot groups ( lsgs ) , ( c ) big sunspot groups ( bsgs ) , and ( d ) all sunspot groups .
the error bars represent the standard error.,title="fig:",width=207,height=226 ] msh and ( b ) @xmath46 msh .
the different symbols and colors represent the different cycles , the same as in figure 2 .
( for interpretation of the references to color in this figure legend , the reader is referred to the web version of this article.),width=302 ] msh and ( b ) @xmath46 msh .
the different symbols and colors represent the different cycles , the same as in figure 2 .
( for interpretation of the references to color in this figure legend , the reader is referred to the web version of this article.),width=302 ] msh ( open circle - dotted curve ) and @xmath39 msh ( filled circle - solid curve ) , and all ssgs whose @xmath47 msh ( squares - dashed curve ) during the preceding minimum epochs of cycles 1223 and around the minimum between cycle 23 and 24 versus the corresponding years , 1878 , 1889 , 1901 , 1913 , 1923 , 1933 , 1944 , 1954 , 1964 , 1976 , 1986 , 1996 , 2006 , 2007 , 2008 , and 2009 . ]
|
recently , javaraiah ( 2012a ) analyzed the combined greenwich and solar optical observing network ( soon ) sunspot group data during the period 18742011 and studied variations in the annual numbers ( counts ) of the small ( maximum area @xmath0 millionth of solar hemisphere , msh ) , large ( @xmath1 msh ) , and big ( @xmath2 msh ) sunspot groups . here that analysis is extended and studied variations in the mean maximum sizes ( the mean values of maximum areas ) of the aforementioned three classes of sunspot groups and also their combination .
it is found that there is no significant correlation between the mean maximum size of any class of sunspot groups and the international sunspot number ( @xmath3 ) , probably due to in a given time interval small sunspot groups / sunspots outnumber the large ones .
a pattern of an approximate 9-year period cycle is seen in the variations of the mean maximum sizes of the large and the big sunspot groups during a solar cycle . on
long - time scales it is found that there exists a strong 130 or more years cycle in the variation of the mean size of the small sunspot groups , whereas there is a hint on the existence of @xmath4 44-year cycles in the variations of the mean maximum sizes of the large and the big sunspot groups . during the decline phase of cycle 23 , there was a scarcity in the sunspot groups whose @xmath537 msh , which may be related to the slow growth of sunspot groups during this period . during the minimum between cycles 23 and 24
may be due to the presence of a number of small sunspot groups whose @xmath6 was larger than that of whose @xmath537 , the relatively large size coronal holes were present at low - latitudes and the total solar irradiance was very low . solar magnetic field , solar activity , solar cycle = 0.5 cm
|
the advanced technology solar telescope ( atst , @xcite @xcite ) , with its 4-meter aperture , will be the largest solar telescope in the world .
a suite of advanced instrumentation will allow for unprecedented resolution and precision imaging , spectroscopy and polarimetry at visible and infrared wavelengths . in order to address the most challenging problems in modern solar physics , atst s
highest priorities are oriented towards high - resolution and coronal science .
obviously , these are aspects in which the site selection will have a strong impact and therefore should be considered very carefully in the decision of the telescope location .
the atst site survey working group ( sswg ) was created in 2000 with an open call to the international community .
its goal was essentially to test the best solar observatories in the world and produce a recommendation to maximize the scientific capabilities of the atst .
more especifically , a set of quantitative goals were set for the atst site : * clear time of 70% , or 3000 annual hours of sunshine . * 1800 annual hours of seeing with a fried parameter @xmath1 better than 7 cm , including at least 100 continuous 2-hour blocks . * 200 annual hours of @xmath2 cm , including at least 10 continuous 2-hour blocks . * large isoplanatic angle for adequate performance of the adaptive optics system over a large field of view .
this goal is equivalent to demanding minimal turbulence at high altitude . * 480 annual hours with a sky brightness smaller than 25 millionths of the disk value at a distance of 1.1 solar radii .
the radial profile curve should follow a power law with an exponent @xmath3 or steeper .
these conditions should be met in at least 40 continuous 4-hour block per year .
* 600 annual hours with precipitable water vapor below 5 mm , including at least 40 continuous 4-hour blocks .
only two comparative studies of solar observatories had been conducted prior to this work .
the joso organization studied nearly 40 sites in europe to conclude that observatorio del teide in tenerife ( canary islands ) provided the better conditions recorded in their data ( @xcite @xcite ; @xcite @xcite ) .
the other work is that of the caltech survey , which considered 34 sites in the southern california area to locate what is now the big bear solar observatory ( bbso ; @xcite @xcite ) .
initially , a list of 72 well - established observatories and some promising undeveloped sites was compiled . since the atst sswg had only sufficient resources for extensive testing of up to six sites , a first decision had to be made in the absence of any actual measurements .
this decision was made based on previous observing experience of the group members .
a first round of discussion quickly resulted in the selection of bbso , la palma ( canary islands ) and sacramento peak ( new mexico ) , which have an outstanding record in producing high - quality solar observations .
the sswg also selected hawaii , and in particular haleakala ( on the island of maui ) was preferred over mauna loa and mauna kea after taking feasibility issues into account .
diverse other considerations later resulted in the choice of panguitch lake ( utah ) and san pedro m ' artir ( mexico ) for the final two slots .
the instrumentation deployed at the six candidate sites includes a scintillometer array ( shabar ) and a solar differential image motion monitor ( s - dimm ) .
both instruments have been described in a recent paper by @xcite ( @xcite ) .
the shabar and s - dimm are used to determine the atmospheric turbulence as a function of height .
it must be noted that daytime seeing conditions are often strongly dependent on height above the ground .
thus , the atst site survey needed to determine the observing conditions at the telescope aperture height , approximately 35 meters above the ground ( the actual height is still to be decided ) .
the site survey instrumentation was placed at a height of 8 meters ( 6-meter stand plus a 2-meter telescope pier ) , which is high enough to avoid the lower part of the surface boundary layer but not so high that wind jitter could become a significant problem .
data from the shabar and s - dimm instruments have been gathered and analyzed until august 2004 . in october 2004 ,
the atst sswg presented a final recommendation and produced an exhaustive document with a detailed report of the site survey process .
the sswg report is publicly available for download from the atst website ( http://atst.nso.edu/ ) .
that document contains abundant technical documentation and the reader is referred to it for further details .
the present series of papers is a formal publication of the most relevant results .
this first paper describes the inversion methods used to infer the atmospheric turbulence and the fried parameter from the shabar / s - dimm data .
the methods have been extensively tested by means of simulations , comparison between different algorithms and also with _ in situ _ measurements .
one month of data ( may 2003 ) was used for testing and fine tuning the algorithms before being applied systematically to the entire site survey period .
the final objective is to determine the fried parameter as a function of height , @xmath5 .
this parameter depends not only on the turbulence at a height @xmath6 but also everything above it . with some assumptions
( see @xcite @xcite for a complete derivation ) , @xmath1 can be obtained as : @xmath7 where @xmath8 is a constant , @xmath9 is the zenith angle and @xmath10 ( which represents the amplitude of the refraction index fluctuations ) characterizes the local atmospheric turbulence at height @xmath6 .
notice that @xmath6 denotes height above the instrument , which in our case is located 8 meters above the ground .
the covariances ( @xmath11 ) of the brigthness fluctuations measured by two detectors , separated by a distance @xmath12 in the shabar array , are related to @xmath10 by : @xmath13 where the @xmath14 are known functions with analytical expressions .
if @xmath15 is the detector diameter and @xmath16 is twice the tangent of the solar radius ( which we consider approximately constant though the year ) : @xmath17 ^ 2 j_0(2\pi f s ) f^{-2/3 } df \ , , \end{aligned}\ ] ] with @xmath18 and @xmath19 being the bessel functions .
the total scintillation can be obtained from the expressions above as @xmath20 .
the kernels @xmath14 are shown in fig [ fig : kernels ] .
it can be deduced from this figure that the @xmath11 corresponding to further separations ( larger values of @xmath12 ) have their sensitivity peak at higher atmospheric layers .
fig [ fig : bis ] shows average values of @xmath21 measured at the six candidate sites .
not surprisingly , the covariances decrease with the detector separation @xmath12 .
the inversion strategy that we employed to solve for @xmath22 in eq ( [ eqbi ] ) is a non - linear least - squares iterative fitting combined with singular - value decomposition ( svd ) of the covariance matrix .
the details of the implementation can be found , e.g. in @xcite ( @xcite ) , but the basic idea is the following .
one starts with an initial guess for the @xmath23 ( where @xmath24 are the gridpoints of the atmospheric discretization ) . with this initial model
we calculate the 15 different covariances @xmath25 ( where @xmath26 is the separation between each pair of detectors @xmath27 and @xmath28 in the shabar instrument ) , the total scintillation @xmath20 and the fried parameter at the base of the atmosphere @xmath29 ( measured by the s - dimm instrument ) , using eqs ( [ eqr0 ] ) and ( [ eqbi ] ) .
the derivatives of these parameters with respect to the @xmath22 ( the so - called response functions ) are also computed , as this information is needed by the fitting procedure .
the synthetic @xmath11 and @xmath1 are then compared to those actually measured by the shabar and s - dimm . in general
, the synthetic values will not match the measurements .
the algorithm then uses the information in the response functions to calculate the first - order corrections @xmath30 that minimize the difference ( in a least - squares sense ) between the synthetic and measured values of @xmath21 and @xmath29 . the numerical solution of an inverse problem such as the one described here is always challenging , but even more is to ensure that the results obtained are physically meaningful .
understanding the issues and uncertainties involved in the inversion is critical for the success of the site survey and , ultimately , the atst .
it is equally important to ensure that the numerical codes employed are exhaustively tested and debugged .
based on previous experience with conceptually similar problems , the sswg decided that two different codes should be independently developed and applied to the data .
the two codes , based on the same minimization scheme described above , exhibit some differences in the actual implementation . by comparing the results provided by two different codes
, we can learn about the reliability of the procedure and the limitations of our seeing measurements . since there is no reason to prefer one above the other , we take the difference between their respective results as an estimate of the error bar .
the first code was developed at the instituto de astrof ' isica de canarias ( iac ) and run by national solar observatory ( nso ) staff .
the spatial grid is equispaced in @xmath31 with 68 points ranging from 20 cm to 40 km .
however , not all of these points are inverted for ( the problem would be ill - posed due to the large number of free parameters ) .
instead , @xmath22 is determined at a subset of 17 special locations , the so - called inversion nodes .
the rest of the atmosphere is then reconstructed via splines interpolation from the values at the nodes . in order to ensure that @xmath22 is always positive
, we used the following change of variable .
the code solves for a variable @xmath32 such that @xmath33 $ ] . in this manner , regardless of whether @xmath32 is positive or negative , @xmath22 will always be positive . before inverting ,
the measured @xmath11 and @xmath29 are processed with a 5-minute running median filter to smooth out fluctuations due to instrumental issues or problems such as birds or insects flying by the detectors .
after an initial application of the code to a sample dataset , it was found that a significant number of points could not be successfully fitted by the procedure .
the observed @xmath29 provided by the s - dimm instrument was inconsistent with the observed scintillation values @xmath21 from the shabar . in all cases
the s - dimm @xmath1 was smaller than would be expected from the scintillation measurements . to overcome this difficulty we introduced an additional parameter @xmath34 in the inversion that accounts for this `` missing scintillation '' , so that the synthetic @xmath11 are now given by : @xmath35 while eq ( [ eqr0 ] ) is still used to compute @xmath5 .
the necessity to include @xmath34 points towards the existence of a a source of image degradation that does not produce scintillation .
this may be due to a finite outer turbulence scale , to `` windfiltering '' ( due to the finite time amount it takes for the turbulence to cross the field - of - view at high altitudes ) or to high - altitude seeing ( higher than @xmath01 km , where the kernels converge ) . in either case
, the inclusion of @xmath34 would be appropriate .
while we have not been able to pinpoint the exact nature of the missing scintillation , the statistical distribution of @xmath34 ( see fig [ fig : deltas ] ) provides an interesting clue . the two island sites ( la palma and haleakala ) require a relatively small amount of missing scintillation , whereas the lake site ( big bear ) frequently demands a substantially larger value .
as we discuss below , the seeing in the island sites is typically dominated by a boundary layer near the ground and the turbulence decreases rapidly as we move upwards . the lake sites , on the other hand , are frequently dominated by high - altitude seeing .
thus , it seems likely that the @xmath34 is related to the turbulence above the sensitivity range of the shabar .
the second inversion code was developed independently at the high altitude observatory ( hao ) .
one of the essential differences with the iac code is that all the grid - points in the atmospheric discretization are now free parameters . in order to make the algorithm stable ,
a regularization scheme was implemented in such a way that the code prefers smooth solutions whenever possible .
basically , in each iteration we add a penalization term to the @xmath36 .
this term is proportional to the quadratic deviation of @xmath22 from a straight line on a @xmath37-@xmath37 plot .
the fact that each grid - point is a free parameter represents an important burden that slows down the execution of the code .
this is compensated , to some extent , by the more accurate integration scheme described below .
the model atmosphere only goes up to a maximum height @xmath38 where the inversion kernels merge ( fig [ fig : kernels ] ) .
we still consider the effect of turbulence above @xmath39 by breaking down the integrals as : @xmath40 and @xmath41 note that @xmath42 does not depend on @xmath12 .
as before , @xmath16 and @xmath42 are free parameters that account for the scintillation and image degradation coming from high - altitude turbulence .
another important difference with the iac code is the integration scheme used to solve eqs ( [ eqbi2 ] ) and ( [ eqr02 ] ) .
the integral over each finite grid interval is computed assuming a parabolic dependence of @xmath43 with @xmath31 ( see the appendix for details ) .
while this integration introduces some additional complexity in the algorithm , it also alleviates the computational expense of the inversions by allowing a coarser grid without loss of accuracy .
the data pre - processing includes a 5-minute block average and a test to look for clouds in the averaged interval . if the derivative of the intensity measured by the shabar detectors changes sign two or more times , the entire period is flagged as cloudy and the @xmath5 is set to zero at all heights .
a sample subset of data ( those taken in may 2003 ) were subject to inversion with both codes .
fig [ fig : testcomp1 ] shows the cumulative distributions produced by each code at la palma , haleakala and big bear .
the results are fairly similar for most of the height range represented in the figure .
as one would expect , the discrepancy increases with height due to the limited sensitivity of the shabar measurements to turbulence at higher layers .
the same behavior is observed with the in situ measurements ( see section [ sec : insitu ] below ) .
a more detailed representation is given in figs [ fig : testcomp2 ] and [ fig : testcomp3 ] , which show the average stratifications of @xmath10 and @xmath1 with height .
the @xmath10 produced by the hao code has more spatial structure ( small - scale fluctuations ) than the iac code .
however , the overall dependence is actually very similar and the iac results resemble a smoothed - out version of those from hao ( at least down to the first meter above the detectors ) . whether the fine structure inferred by the hao code is real or not is practically irrelevant for our purposes here , as the @xmath5 parameter , which is our ultimate goal , is barely sensitive to it .
the @xmath1 stratifications from both codes are very similar , at least up to @xmath0 40 m above the instrument . in this small sample
we can already see the different behavior of the lake sites ( big bear and panguitch ) as opposed to the island ( la palma and haleakala ) and continental sites ( san pedro m ' artir and sac peak ) .
the big bear plot exhibits a larger @xmath1 near the ground , but it does not improve significantly with height .
the island and continental sites , on the other hand , have a lower value at the ground but the seeing improves dramatically as we go to higher layers , suggesting the presence of a boundary layer near the ground .
the slope is particularly steep in the @xmath030 - 40 m range , precisely where the atst aperture will be .
these results emphasize the importance of analyzing the height dependence of @xmath1 , as we have done in this work , since the site and height of the telescope may end up being critical factors in the performance of the atst .
perhaps the most reliable test of the atst site survey analysis is to compare its results directly with actual in situ measurements .
the sswg had access to a tower with meteorological instrumentation including hygrometers and sonic anemometers at various heights ( dash , @xcite @xcite ) . with this instrumentation
it is possible to make local measurements of temperature @xmath44 and humidity @xmath45 up to a height of @xmath0100 m. these parameters are monitored at a rate of 30 hz , from which @xmath10 can be derived . a complete shabar / s - dimm system was deployed in erie ( colorado ) , at the base of the measuring tower .
simultaneous measurements from the tower and the atst instrumentation have been used to produce the scatter plots in fig [ fig : erie ] .
each panel shows the diagonal of the plot , as well as the pearson correlation between the atst inversions and the shabar instrumentation . before analyzing the discrepancies between the two ,
a brief explanation is in order . while the tower measurements are strictly local , the shabar detectors receive solar light that has passed through a cone of earth atmosphere .
thus , the atst instrumentation is sensitive to a field of view that becomes larger for higher heights . for this reason ,
the @xmath10 measured in situ is expected to exhibit larger temporal fluctuations than that from the atst instrumentation , which is a spatial average over the shabar field of view .
this effect , which can be seen in fig [ fig : erie2 ] , increases the scatter especially in the higher layers .
moreover , the shabar field of view is displaced with respect to the tower instrumentation by an amount that increases with height . in any case
, fig [ fig : erie2 ] shows that the overall trends are in general well retrieved by the atst analysis .
the tests indicate good agreement between both types of measurements up to a height between 22 and 50 m. it is important to note that the height range where the atst measurements are accurate is lower in this case than at the actual candidate sites , simply because the tests were carried out during mid - winter at the considerably higher latitude of erie ( the @xmath46 $ ] angle was always larger than 2.0 ) .
while the erie tests reassured our confidence in the atst seeing measurements , there was still some concern that the atmospheric structure above the continental colorado is probably very different from the actual sites .
the sswg then decided to mount similar in - situ probes on a portable crane to perform similar tests on big bear and haleakala .
new measurements ( this time considering lower heights due to the limitations of the crane ) were conducted at the two sites , with the results shown in fig [ fig : tests2 ] .
the haleakala tests show excellent agreement between the inversions and the measurements . in big bear
, however , the correlation is very weak . among the differences between the two sites
are the level of humidity , the local topography and environment , and the probable lack of a ground boundary layer in big bear
. a possible explanation may be that the weaker scintillation in big bear makes these measurements more sensitive to noise . a closer look at the data shows that the correlation is improved by removing outlier points and clouds ( fig [ fig : bbtest1 ] ) .
particularly interesting is the dataset gathered on may 12 2004 , which reveals the dependence of the data with the wind direction ( fig [ fig : bbtest2 ] ) .
when the wind blows from the lake ( west ) , the observing conditions are optimal and the seeing is dominated by high - altitude turbulence . on that particular day
the wind changed from west to variable directions , resulting in the development of a ground turbulence layer and improving the correlation coefficient to @xmath00.77 .
note also that , in the absence of a surface turbulence layer at big bear , @xmath1 will essentially be independent of @xmath6 near the ground and the s - dimm measurements at 8 m can be taken to represent @xmath1 at 25 - 40 m.
the atst project has imposed stringent requirements on the telescope site , both in terms of seeing and sky brightness .
the seeing analysis has proven very challenging due to the strong dependence of @xmath1 with height exhibited by most sites .
this paper describes the analysis employed by the atst sswg to find the site that provides optimal conditions ( @xmath1 ) at the height of the atst primary aperture .
various tests have been used to demonstrate the validity of our analysis , at least up to @xmath040 m above the ground .
the use of two independent inversion codes are helpful to understand the uncertainties in the procedure and to set error bars on the measurements . moreover , in situ measurements proved the validity of the approach in the height range of interest .
the only caveat identified by our work is that the errors increase when the seeing is dominated by high - altitude turbulence ( i.e. , in the absence of a ground boundary layer ) .
let @xmath47 and @xmath48 $ ] .
if we approximate @xmath32 and @xmath49 as parabolic functions of @xmath50 : @xmath51 we can solve analytically the integral in eq ( [ eqbi2 ] ) between two adjacent grid - points @xmath52 and @xmath53 , as : , s. l. , rimmele , t. , keller , c. u. , hill , f. , radick , r. r. , oschmann , j. m. , warner , m. , dalrymple , n. e. , briggs , j. , hegwer , s. l. , & ren , d. 2003 , in innovative telescopes and instrumentation for solar astrophysics .
edited by stephen l. keil , sergey v. avakyan .
proceedings of the spie , volume 4853 , pp .
240 - 251 ( 2003 ) .
|
the site survey for the advanced technology solar telescope concluded recently after more than two years of data gathering and analysis .
six locations , including lake , island and continental sites , were thoroughly probed for image quality and sky brightness .
the present paper describes the analysis methodology employed to determine the height stratification of the atmospheric turbulence .
this information is crucial because day - time seeing is often very different between the actual telescope aperture ( @xmath030 m ) and the ground .
two independent inversion codes have been developed to analyze simultaneously data from a scintillometer array and a solar differential image monitor .
we show here the results of applying them to a sample subset of data from may 2003 , which was used for testing .
both codes retrieve a similar seeing stratification through the height range of interest .
a quantitative comparison between our analysis procedure and actual _ in situ _ measurements confirms the validity of the inversions .
the sample data presented in this paper reveal a qualitatively different behavior for the lake sites ( dominated by high - altitude seeing ) and the rest ( dominated by near - ground turbulence ) .
|
laser cooling and manipulation of neutral atoms is one of the priority field of atom optics . in the recent time
the major developments and success have been obtained in atom lithography and direct deposition of atoms utilizing light fields as an immaterial optical mask for atomic beam @xcite . in most nanofabrication experiments ,
atomic structures are realized by far off detuned periodical conservative potential created by intense laser fields acting as an array of immaterial light lenses for the atoms .
the influence of the spontaneous emission on the focusing is considered to be negligible because of the large light detuning and short interaction times .
in essence , the atom trajectory affected by conservative dipole force without any loses ( or dissipation ) of energy in the atomic beam . in this case
the atomic beam focusing has a classical analogy and can be described with methods developed for particle optics @xcite . as in normal optics ,
the feature size is limited by a combination of chromatic aberration caused by the broad longitudinal and transverse velocity distribution of an atomic beam .
therefore an additional laser cooling field is required to prepare a well collimated and transversely cooled atomic beam to minimize deleterious effects . additionally , because of spherical aberration some atoms do not focus well and contribute to pedestal background .
these factors are dominant and do not allow one to reach the theoretically predicted diffraction limit for atom optics determined by de broglie wavelength of atoms , only a few picometres . therefore the new alternative methods for atom lithography are intensively investigated .
recently , the idea of combining the traditional focusing method with the well known concept of laser cooling was suggested for a blue detuned intensive light field @xcite and was mentioned earlier in @xcite . here the intense light field , commonly used for focusing , was used to create a deep optical potential .
additional dissipative light force provides additional cooling of the atoms to the minimum of optical potential at blue detuning .
the characteristic time when the dissipation processes take effect is a few inverse recoil frequencies @xmath0 ( where @xmath1 is recoil energy , gained by an atom with mass @xmath2 at rest after emission of a photon with momentum @xmath3 ) .
this time is several tens of microseconds for the number of elements with closed dipole optical transitions suitable for laser cooling .
thus for commonly used atomic beams with thermal longitudinal velocity it might be difficult to realize this type of dissiaptive optical mask experimentally due to power limitations of laser system used . in the following we consider the alternative regime of dissipative optical mask , created by red detuned low intensity light field with nonuniform polarization .
it is well known that low intensity light field with polarization gradients can be used for sub - doppler laser cooling of neutral atoms .
this mechanism of laser cooling is well understood and has been thoroughly studied by a number of authors @xcite especially with respect to the temperature of laser cooling , the atomic momentum distribution , and the localization @xcite that can be measured by spectroscopy methods @xcite . due to the extremely complex master equation for the quantum description of atomic motion in light fields the secular approximation @xcite was initially suggested is valid in the limit @xcite @xmath4 this limit assumes the energy separation between different energy bands of atoms in optical potential are much greater then their width due to optical pumping and tunneling effects . here
the light - shift well depth @xmath5 define the optical potential depth .
@xmath6 is the detuning between the laser @xmath7 and atomic transition @xmath8 frequencies , and @xmath9 is the radiative decay rate .
this approximation is valid , for a given potential depth , in the limit of large detuning . on the contrary , it might failed in a deep potential at given detuning .
more over , even if the secular approximation is well fulfilled for the lowest vibrational levels it might break down for upper ones where the separation between energy bands becomes smaller due to potential anharmonicity effects , and especially for above - barrier motioned atoms . in the present work we investigate the applicability of laser cooling in a deep optical potential created by light field with nonuniform polarization for generating spatially localized atom structures with high contrast for atom lithography .
we consider the conditions far from the situation of extremely low sub - doppler cooling cases .
thus , to describe the localization of atoms more correctly in here we do not restrict our consideration by secular approximation . rather we perform a full quantum numerical analysis of generalized optical bloch equation for atomic density matrix elements . in particular ,
we consider the light field parameters beyond the secular approximation limit .
finally we analyze the structure width and contrast of localized atoms , important parameters for technological applications .
let us consider one - dimensional ( along the z axis ) motion of atoms with total angular momenta @xmath10 in the ground state and @xmath11 in the excited state in a field of two oppositely propagating waves with the same frequency and intensity @xmath12 here @xmath13 is the amplitude of each of the oppositely propagating waves .
the unit vectors @xmath14 and @xmath15 determine their polarizations with components @xmath16 in cyclic basis @xmath17 . in this work
we restrict our consideretion by the weak - field limit , i.e. with small saturation parameter @xmath18 here @xmath19 is the single - beam rabi frequency characterize the coupling between the atomic dipole @xmath20 and light field . in the weak - field limit
, the atomic exited - state can be adiabatically eliminated , and atom motion is describing by reduced equation for the ground - state density matrix elements @xcite : @xmath21+{\hat \gamma}\{{\hat \rho}\}\ ] ] where the hamiltonian @xmath22 is given by @xmath23 the last term in ( [ ham ] ) describes atoms interaction with the light field in resonance approximation , where @xmath24 and operator @xmath25 is written through the clebsch - gordan coefficients : @xmath26 written in the basis of sublevel wave functions for exited @xmath27 and the ground @xmath28 atomic states .
in addition , the relaxation part of kinetic equation for atomic density matrix ( [ equation ] ) has the following form @xmath29 where @xmath30 standard anticommutator definition and @xmath31 is position operator .
this term describes redistribution of the atom on the ground state energy sublevels with taking into account the recoil effects in spontaneous photon emission .
the functions @xmath32 and @xmath33 is determined by the probability of emission of a photon with polarization @xmath34 into direction @xmath35 ( relative to the @xmath36 axis ) .
there are number of approaches developed for calculation of the evolution of the atomic density matrix .
the quantum problem is more difficult because it incorporate evolution of numbers of internal and external components of the density matrix .
the majority of works are based on secular approximation for density matrix elements @xcite .
it consists in the following : first , the eigenstates and eigenenergies of the hamiltonian @xmath22 should be found .
then we can consider only the evolution of diagonal elements of atomic density matrix in this eigenstatas basis .
it can be written in a form of balance equation with the rates characterizing the relaxation part of the master equation .
the secular approximation for the general master equation ( [ equation ] ) is valid for the lowest vibrational energy levels when the the energy separation between different bands are much greater than the effective width due to optical width and tunneling effects .
this condition implies very large detuning .
however , the density of energy states increases for the upper energy levels and the energy difference between adjacent states could be very small @xcite .
this circumstances therefore make very hard to use the secular approximation for describing of hot and nonlocalized atomic fraction . in order to take into account effects of localization in optical potential and modulation depth of the spatial distribution of atoms
more correctly we utilize a different approach for the quantum calculation of the master equation ( [ equation ] ) . in the wigner representation for atomic density matrix @xmath37 the general master equation ( [ equation ] ) takes the following form : @xmath38 - \frac{\gamma \,s}{2 } \left\{\hat{v}^{\dagger } \hat{v } , \hat{\rho}\right\ } + \hat{\gamma}\{\hat{\rho}\}\ ] ] where commutator of density matrix with kinetic part of hamiltonian ( [ ham ] ) reduces to partial deviation on @xmath36 @xmath39 the field part of hamiltonian ( [ ham ] ) has only the zeroth and the second spatial harmonics : @xmath40 thus the commutator and anticommutator in the right hand side of equation ( [ wigner ] ) could be written as : @xmath41 the last term describing relaxation due to spontaneous emission of photons has well - known form in wigner representation : @xmath42 { \hat t}_q \ , .
\nonumber\end{aligned}\ ] ] equation ( [ wigner ] ) admits solution that is periodic in the position variable .
we further introduce a fourier - series expantion for atomic density matrix in spatial coordinates @xmath43 and rewrite the master equation for discrete fourier components of density matrix @xmath44 : @xmath45 in steady - state problem ( @xmath46 ) such a recursion may often be solved by the method of continued fraction .
this approach is used for solution of the optical - bloch equations in different spectroscopy tasks as well as for the calculation of the force on atom in the light field ( see for example @xcite ) .
the major distinction here is that these equations for density matrix contain the recoil effects that makes them more complicated for calculation .
additionally we note , the similar approach was described in @xcite where the authors analyzed the laser cooling ( velocity distribution ) of two - level atom in the recoil limit and thus restrict their consideration only by the zeroth spatial harmonics for the ground state density matrix . in our case
the number of the spatial harmonics depends on the light field and atomic parameters .
typically we use less than @xmath47 harmonics that is enough to obtain the spatial solution for equilibrium atomic density matrix in considered range of light field parameters .
the spatial and momentum steady - state distribution of atoms with @xmath48 optical transition for @xmath49 , @xmath50 and chromium mass are shown in fig.[fig1](a ) and ( b ) .
the fig.[fig1](c ) represents the total population of spatial harmonics of atomic ground - state density matrix integrated over momentum space @xmath51 .
the zeroth harmonic is equal to @xmath52 that is the normalization condition . as it seen here the population of higher harmonics are rapidly decrease with number @xmath53 .
in this section we turn our attention to the steady - state spatial distribution of the atoms in the optical potential created by the light field with @xmath54 configuration .
we choose this configuration as a brightest example of the light field with nonuniform polarization .
only light field ellipticity varies in position space while the other parameters ( intensity , phase , orientation of polarization vector ) stay unchanged .
more over , the optical potential created by this field configuration has a period of @xmath55 that makes it very attractive for deposition of atomic structures with high spatial periodicity .
ccccccccc + element & & cooling & & @xmath56 & & @xmath57 & & @xmath58 + & & transition & & & & ( nm ) & & @xmath59 + + + @xmath60li & & @xmath61 & & 46 & & 671 & & 2.56 + @xmath62na & & @xmath63 & & 198 & & 589 & & 6.34 + @xmath64k & & @xmath65 & & 358 & & 766 & & 1.81 + @xmath66rb & & @xmath67 & & 770 & & 780 & & 1.63 + @xmath68cs & & @xmath69 & & 1270 & & 852.3 & & 1.06 + @xmath70cr & & @xmath71 & & 115 & & 425.6 & & 8.49 + @xmath72al & & @xmath73 & & 85 & & 309.4 & & 57 + @xmath74ga & & @xmath75 & & 382 & & 294.4 & & 127 + @xmath76 in & & @xmath77 & & 634 & & 325.7 & & 78 + @xmath78ag & & @xmath67 & & 601 & & 328 & & 76.8 + there are several physical parameters that are required to characterize a given laser cooling situation .
these are the atomic mass @xmath2 , the wavelength @xmath57 , and the natural line - width @xmath9 .
in addition , there are two light - field parameters in the low - field limit : detuning @xmath79 , and saturation parameter @xmath80 .
it is possible to choose reduced dimensionless units ( @xmath81 , @xmath82 , @xmath83 ) , so that the dimensionless atomic mass @xmath56 can be defined from relation @xmath84 @xcite .
this is so called a quasiclassical parameter that characterizes an atom s kinetic in a light field . in particular
, it describes the rate of kinetic processes and evolution of atomic distribution function in momentum space , thus the typical cooling time is order of @xmath85 .
we perform an analysis of the effects of atom localization in optical potential .
contrast can be defined as the ratio of spatial distribution modulation depth to its amplitude @xmath86 ( fig.[fig1 ] ( a ) ) as in @xcite . as it can be seen directly from ( [ wigner ] )
, there are only two parameters that characterize the stationary solution for an atomic density matrix .
this is the detuning , that can be measured in the units of natural line - width @xmath87 and the second one is non - dimensional parameter @xmath88 : @xmath89 this scale parameter makes it possible to implement the results of calculation for the elements with allowed closed dipole optical transition having degeneration over the angular momentum energy levels ( table i ) .
note , that in the secular approximation @xcite the stationary solution is characterized only by the ratio of optical potential depth to recoil energy @xmath90 that is proportional to @xmath91 in our notations .
thus we first express the results in parameters @xmath92 and @xmath91 . as it seen in the fig .
[ fig2 ] the differences between the curves become more significant with the optical potential depth @xmath5 increasing , corresponding to out of framework from secular approach .
the width of the localized structures and the contrast tends to an asymptotic constant value with increasing of the light field intensity , however that depends on light field detuning . in order to find the asymptotic values for the width we fit curves with an empirical low @xmath93 .
the results for asymptotic values @xmath94 are shown in fig .
[ fig3 ] . the fig .
[ fig4 ] represents the fwhm width and contrast of the spatial structures as a function of parameter @xmath88 at different detunings for the atomic model with @xmath95 optical transition .
the fwhm width of the spatial structures is monotone decrease for a large detuning and parameter @xmath88 . in spite of the fact that the equilibrium temperature is growing with the depth of optical potential ,
the localization of the atoms becomes stronger with growth of these parameters .
additionally the contrast tends towards its maximum value fig . [
fig4](b ) .
the dashed vertical lines on the fig.[fig4 ] underline the limitations of the weak - field theory for different elements from the table i in assumption @xmath96 .
note , that this is qualitative assumption . for thorough analysis of the weak - field theory limits the solution of the quantum equation for the total atomic density matrix is required with taking into account exited state population .
however , the width and contrast curves fig.[fig4 ] have a very strong dependence on @xmath88 , thus the localization effects remain rather significant for enough small saturation parameter especially for `` heavy '' atoms table i.
we performed the fully quantum analysis of atomic localization obtained by laser cooling in nonuniformly polarized low - intensity light field .
generally , conditions for a deep laser cooling mismatch conditions for strong atomic localization that require deeper optical potential and consequently leads to higher laser cooling temperature . additionally , in a deep optical potential the secular approximation ( [ sec ] ) restricted by relation on light field detuning and the depth of optical potential .
in our treatment we had no such limitation allowing a more correctly describe the spatial solution for atomic distribution function including localized and above - barrier motioned atoms .
the stationary solution of atoms is a function of the light field detuning @xmath79 and the dimensionless parameter @xmath88 ( [ mu ] ) .
we analyzed the width and contrast of localized atomic structures as a function of these parameters .
we showed that the atomic structure width and contrast have a strong dependence on @xmath88 and tend to constant values with an increasing optical potential dept that depends on the light field detuning .
we demonstrated the applicability of laser cooling in far - off detuned deep optical potential , created by a light field with polarization gradient , as a dissipative optical mask for the purposes of atom lithography and nanofabrication for generation of spatially localized atomic features with high contrast .
this type of the light mask can be an alternative method for creation of spatially localized atomic structures .
the remarkable distinction of this method from previous non - dissipative light mask is that the suggested one is not sensitive to any aberration effects .
more over , this type of optical mask has no classical analog and can not be described by methods for classical optics .
the width and the contrast of localized atomic structures here is determined by atomic energy dissipation mechanisms in the light field .
finally , we perform analysis of the possible limits for the width and the contrast of localized atomic structures that can in principle be reached by this type of the light mask .
this work was partially supported by rfbr ( grants 05 - 02 - 17086 , 04 - 02 - 16488 , 05 - 08 - 01389 ) and o.n.p .
was supported by the grant mk-1438.2005.2 .
99 d. meschede and h. metcalf , j. phys .
* 36 * r17-r38 , ( 2003 ) m.k .
oberthaler and t. pfau , j. phys . : condens .
matter * 15 , * r233-r255 , ( 2003 ) j.j .
mcclelland and m.r .
scheinfein , j.opt .
b , * 8 , * 1974 , ( 1991 ) r. stutzle , d. jurgens , a. habenicht , m.k .
oberthaler , j. opt .
b. : quantum semiclass . opt .
* 5 , * s164 , ( 2003 ) o.n .
prudnikov and e. arimondo , j. opt .
b : quantum semiclass . opt .
* 6 , * 336 ( 2004 ) a. p. kazantsev , g. i. surdutovich , and v. p. yakovlev , _ `` the mechanical action of light on atoms '' _ ( nauka , moscow , 1991 ) d.j .
wineland , wayne m. itano , phys .
a * 20 , * 1521 ( 1979 ) j. dalibard and c. cohen - tannoudji , j. opt .
b * 6 , * 2023 ( 1989 ) .
yoo and j. javanainen , phys .
rew . a * 45 , * 3071 ( 1992 ) y. castin and j. dalibard , europhys
. lett . , * 14 , * 761 - 766 ( 1991 ) k. berg - sorensen , y. castin , k. molmer and j. dalibard , europhys .
lett . , * 22 , * 663 - 668 ( 1993 ) j.guo and p.berman , phys . rew .
a * 48 , * 3225 ( 1993 ) y. castin , k. berg - sorensen , j. dalibard , and k. molmer , phys .
a * 50 , * 5092 ( 1994 ) i.h.deutsch , j. grondalski , and p.m. alsing , phys .
a * 56 , * r1705 ( 1997 ) m. gatzke , g. birkl , p. s. jessen , a. kastberg , s. l. rolston , and w. d. phillips , phys .
a * 55 , * r3987 ( 1997 ) s. marksteiner , r. walser , p. marte , p. zoller , appl . phys .
b * 60 , * 145 ( 1995 ) p. s. jessen , c. gerz , p. d. lett , w. d. phillips , s. l. rolston , r. j. c. spreeuw , and c. i. westbrook , phys .
rew . lett .
* 69 , * 49 ( 1992 ) g. raithel , g. birkl , a. kastberg , w. d. phillips , and s. l. rolston , phys .
* 78 , * 630 ( 1997 ) h.risken , _ `` the fokker - plank equation '' , _ ( springer , berlin 1989 ) s.m . tan , j. opt .
b : quantum semiclass . opt .
* 1 , * 424 ( 1999 ) s.m .
yoo and j. javanainen , j. opt .
b * 8 , * 1341 ( 1991 )
|
generally , the conditions for deep sub - doppler laser cooling do not match the conditions for the strong atomic localization that takes a place in deeper optical potential and , in consequence , leads to larger temperature . moreover , for a given detuning in a deep optical potential the secular approximation which is usually used for quantum description of laser cooling becomes no more valid .
here we perform an analysis of atomic localization in optical potential based on a full quantum approach for atomic density matrix .
we also show that the laser cooling in a deep far - off detuned optical potential , created by a light field with a polarization gradient , can be used as an alternative method for forming high contrast spatially localized structures of atoms for the purposes of atom lithography and atomic nanofabrication .
finally , we perform an analysis of the possible limits for the width and the contrast of localized atomic structures that can in principle be reached by this type of the light mask .
|
the extraordinary electric fields available from laser systems make laser - driven charged particle acceleration an exciting possibility .
practical vacuum laser acceleration requires a guided - mode structure capable of confining a speed - of - light ( sol ) mode and composed entirely of dielectric materials , and photonic crystals provide a means to achieve this capability .
a photonic crystal is a structure with permittivity periodic in one or more of its dimensions . as described in @xcite , optical modes in a photonic crystal form bands , just as electronic states do in a crystalline solid .
similarly , a photonic crystal can also exhibit one or more photonic band gaps ( pbg s ) , with frequencies in the gap unable to propagate in the crystal .
confined modes can be obtained by introducing a defect into a photonic crystal lattice . since frequencies in the bandgap
are forbidden to propagate in the crystal , they are confined to the defect . a linear defect
thus functions as a waveguide .
a significant benefit of photonic crystal accelerators is that only frequencies within a bandgap are confined . in general , higher order modes , which can be excited by the electron beam , escape through the lattice .
this benefit has motivated work on matallic pbg structures at rf frequencies @xcite .
in addition , an accelerating mode has been found in a pbg fiber structure @xcite . after discussing 2d planar structures we consider the fiber geometry in more generality .
the geometries we consider in this section are two - dimensional : we take them to be infinite in the vertical ( @xmath0 ) direction , while the electron beam and the accelerating optical field copropagate in the @xmath1-direction , transverse to the direction of symmetry .
while such structures are not immediately suitable for charged particle acceleration , 2d structures can be analyzed with much less cpu time than can 3d structures , thereby allowing rapid exploration of multiple sets of geometric parameters .
the computational technique is discussed further below .
our underlying photonic crystal lattice is a triangular array of vacuum holes in a silicon substrate . assuming an operating wavelength of in the telecom band
, silicon has a normalized permittivity of @xmath2 @xcite .
such a lattice exhibits a wide te bandgap , as desired since the accelerating field component is transverse to the direction of symmetry . for lattice constant @xmath3 the nearest - neighbor center - to - center hole spacing ,
we choose the hole radius @xmath4 to maximize the relative width of the bandgap .
the accelerator structure consists of a vacuum guide in this lattice , as shown in fig .
[ fig : mode ] . , the pad width is @xmath5 , and the wavelength is @xmath6 . ]
the guide width @xmath7 is defined such that the distance between the centers of the holes adjacent to the waveguide are @xmath8 . also , dielectric material can be added to the sides of the guide , and we let @xmath9 denote the total width of the dielectric `` pad '' added to both sides of the guide . fig . [
fig : mode ] also shows an accelerating mode of this geometry , i.e. @xmath10 is nonzero on axis and @xmath11 .
in fact , for a general selection of @xmath7 and @xmath9 , there will be a @xmath12 for which this waveguide mode is synchronous .
this is because the dispersion properties of this pbg waveguide are similar to a metallic guide in that @xmath13 throughout most of the bandgap , but at the upper edge of the gap the dispersion curve reduces in slope and meets the sol line .
the padding can be added in order to bring the sol frequency into the center of the gap where the dispersion curve is more linear .
several parameters characterize the performance of an accelerating mode .
the relationship between the input laser power and the accelerating gradient is described by the characteristic impedance @xcite .
since our 2d structures only confine modes in one transverse dimension , we normalize the impedance to that of a structure one wavelength high , so @xmath14 , where @xmath15 is the accelerating gradient and @xmath16 is the laser power per unit height .
we find an empirical power - law scaling of the impedance , with @xmath17 .
next , there is the _ damage factor _ @xmath18 , where @xmath19 is the maximum electric field magnitude anywhere in the dielectric material .
since laser power is ultimately limited by the damage threshold of the material , the damage factor is an important measure of the maximum possible accelerating gradient a structure can sustain .
the damage threshold exhibits a dependence on laser pulse width which becomes favorable at very short pulse widths , as examined in @xcite and paramaterized in @xcite .
thus these accelerator structures are transmission - mode , and a high group velocity @xmath20 is desired so that short pulses may be used .
the qualitative behavior of these parameters presents a trade - off .
as the guide is widened , the damage factor decreases . on the other hand , the group velocity increases , allowing shorter laser pulses to be used , for which the material damage threshold is at a higher field . to find the optimum parameters we plot the maximum possible accelerating gradient taking both effects into account in fig .
[ fig : maxgradient ] .
the geometry of this structure is again a triangular array of vacuum holes , this time in silica ( @xmath21 ) . here
the structure is considered to be a fiber drawn infinitely in the beam propagation direction , with the electrons and laser pulse copropagating along the fiber . for these structures
@xmath22 , and the defect consists of a larger central hole .
modes were found for three different mode radii , and are shown in fig .
[ fig : fibermode ] .
-1.8 in the frequencies of the three modes are given by @xmath23 of 8.20 , 8.12 , and 8.20 and group velocities @xmath24 , @xmath25 , and @xmath26 ( left to right ) .
the structure was simulated using periodic boundary conditions , and the fields in the lattice region are due to crosstalk between neighboring defects . by increasing the size of the supercell
, this crosstalk can be minimized , however the computational time significantly increases .
also , the 6-fold azimuthal symmetry of the structure implies that sol modes in vacuum contain only @xmath27 azimuthal modes for @xmath28 an integer , reducing the emittance blowup from higher - order modes .
finally , we find that the characteristic impedance decreases with guide radius , as is the case with metallic waveguide structures .
we use the mit photonic bands ( mpb ) package , a public - domain code using an iterative eigensolver technique @xcite . for a given geometry and bloch wavevector
, mpb computes frequencies and field configurations of supported modes .
mpb can compute group velocities of modes as well by applying the feynman - hellmann theorem @xcite . using the frequencies and group velocities ,
we can inductively converge on the longitudinal wavenumber for which a speed - of - light mode exists .
having found a mode for a particular wavenumber , we can use its frequency and group velocity to obtain a linear approximation for its dispersion curve near that wavenumber .
the intersection of that approximation with the sol line gives the wavenumber for the next computation , which yields a mode whose phase velocity is closer to @xmath29 .
since the iterative eigensolver for each step can be seeded with the field configuration from the result of the previous step , successive steps are quite fast , and convergence to an sol mode is a computationally light task once the initial mode has been found .
the 2d structures discussed above are amenable to photolithography , with 50 : 1 apect ratios available from current reactive ion etching equipment .
some investigation into fabrication of these structures has taken place in the past @xcite .
fabrication of 3d photonic crystals with omnidirectional bandgaps , such as the `` woodpile '' structure @xcite , is an active area of research .
a number of techniques are being developed , including multilayer lithography , wafer fusion , stacking by micromanipulation , self - assembly , and others @xcite .
pcf manufacturing is a large and growing area of research in industry , since photonic crystals allow for tailoring optical properties to specific applications , from nonlinearity for wavelength conversion in telecommunications to large mode area for materials processing @xcite .
photonic crystals have great promise as potential laser accelerator structures .
not only do they support accelerating modes , but such modes exist for a wide range of geometric parameters . while the basic accelerator parameters have been examined , much remains to be done to understand the properties of these structures .
wakefield computations as well as coupling structure design have yet to be explored .
in addition , there are many photonic crystal lattices for which accelerating modes have not been computed , including 3d geometries .
however , manufacturing technology , numerical simulation capability , and theoretical understanding continue to progress at an extraordinary rate , driven by industry forces .
we therefore expect a bright future for photonic crystals as an accelerator technology .
99 j. d. joannopoulos , r. d. meade , and j. n. winn , _ photonic crystals : molding the flow of light _ ( princeton university press , princeton , nj , 1995 ) .
m. a. shapiro et al . , phys . rev .
beams * 4 * , 042001 ( 2001 ) .
x. e. lin , phys . rev .
beams * 4 * , 051301 ( 2001 ) .
d. f. edwards , in _ handbook of optical constants _ , edited by e. d. palik ( academic press , 1985 ) , vol . 1 , p. 547 .
l. schchter , r. l. byer , and r. h. siemann , in _ advanced accelerator concepts : tenth workshop , mandalay beach , ca , 2002 _ , edited by c. e. clayton and p. muggli , u.s .
department of energy ( american institute of physics , melville , ny , 2002 ) , no . 647 in aip conference proceedings , pp .
b. c. stuart et al .
74 * , 2248 ( 1995 ) .
s. g. johnson and j. d. joannopoulos , optics express * 8 * , 173 ( 2001 ) .
see for instance j. j. sakurai , _ modern quantum mechanics _
( addison - wesley , reading , ma , 1995 ) s. g. johnson et al . , phys .
e * 65 * , 066611 ( 2002 ) .
wendt et al . , j. vac .
b * 11 * , 2637 ( 1993 ) . s. y. lin et al . ,
nature * 394 * , 251 ( 1998 ) .
s. g. johnson , _ fabrication of three - dimensional crystals : those clever experimentalists _ , from lecture series _ photonic crystals : periodic surprises in electromagnetism _ , ` http://ab-initio.mit.edu/photons/tutorial/ ` ren engel kristiansen , _ guiding light with holey fibers _ , oe magazine june 2002 , p. 25 .
|
photonic crystals have great potential for use as laser - driven accelerator structures .
a photonic crystal is a dielectric structure arranged in a periodic geometry . like
a crystalline solid with its electronic band structure , the modes of a photonic crystal lie in a set of allowed photonic bands .
similarly , it is possible for a photonic crystal to exhibit one or more photonic band gaps , with frequencies in the gap unable to propagate in the crystal .
thus photonic crystals can confine an optical mode in an all - dielectric structure , eliminating the need for metals and their characteristic losses at optical frequencies .
we discuss several geometries of photonic crystal accelerator structures .
photonic crystal fibers ( pcfs ) are optical fibers which can confine a speed - of - light optical mode in vacuum
. planar structures , both two- and three - dimensional , can also confine such a mode , and have the additional advantage that they can be manufactured using common microfabrication techniques such as those used for integrated circuits .
this allows for a variety of possible materials , so that dielectrics with desirable optical and radiation - hardness properties can be chosen .
we discuss examples of simulated photonic crystal structures to demonstrate the scaling laws and trade - offs involved , and touch on potential fabrication processes .
_ submitted to particale accelerator conference ( pac 2003 ) , may 1216 , 2003 , portland , oregon ( ieee ) _
|
the entanglement characteristics of the ground states of many body hamiltonians describing condensed matter systems constitute an important area of study in quantum information theory .
entanglement is an essential resource in quatum computation and communication protocols .
condensed matter , specially , spin systems have been proposed as candidate systems for the realization of some of the protocols .
entanglement provides a measure of non - local quantum correlations in the system and it is of significant interest to determine how the correlations associated with the ground state of the system change as one or more than one parameter of the system is changed .
the focus on ground state characteristics arises from the possibility of quantum phase transitions ( qpts ) which occur at temperature @xmath8 ( when the system is in its ground state ) and are driven solely by quantum fluctuations @xcite .
a qpt is brought about by tuning a parameter , either external or intrinsic to the hamiltonian , to a special value termed the transition point . in thermodynamic critical phenomena ,
the thermal correlation length diverges and the thermodynamic quantities become singular as the critical point is approached . in the quantum case ,
the correlation length diverges in the vicinity of the qcp and the ground state properties develop non - analytic features .
an issue of considerable interest is whether the quantum correlations , like the usual correlation functions , become long - ranged near the qcp . in a wider perspective
, the major goal is to acquire a clear understanding of the variation in entanglement characteristics as a tuning parameter is changed .
qpts have been extensively studied in spin systems both theoretically and experimentally . in recent years
, several theoretical studies have been undertaken to elucidate the relationship between qpts and entanglement in spin systems @xcite .
in particular , a number of entanglement measures have been identified which develop special features close to the transition point .
one such measure is concurrence which quantifies the entanglement between two spins @xmath9 . at a qcp , as illustrated by a class of exactly - solvable spin models @xmath9 , the derivative of the ground state concurrence has a logarithmic singularity though the concurrence itself is non - vanishing upto only next - nearest - neighbour - distances between two spins @xcite .
discontinuities in the ground state concurrence have been shown to characterize first order qpts @xcite .
later , wu et al .
@xcite showed that under some general assumptions a first order qpt , associated with a discontinuity in the first derivative of the ground state energy , gives rise to a discontinuity in a bipartite entanglement measure like concurrence and negativity .
similarly , a discontinuity or a divergence in the first derivative of the same entanglement measure is the signature of a second order phase transition with a discontinuity or a divergence in the second derivative of the ground state energy .
another measure of entanglement , studied in the context of qpts , is the entropy of entanglement between a block of @xmath10 adjacent spins in a chain with the rest of the system @xcite . at the qcp ,
the entropy of entanglement diverges logarithmically with the length of the block .
there is , however , no direct relation with the long range correlations in the system .
a number of entanglement measures have recently been proposed which are characterized by a diverging length scale , the entanglement length , close to a qcp .
the localizable entanglement ( le ) between two spins is defined as the maximum average entanglement that can be localized between them by performing local measurements on the rest of the spins @xcite .
the entanglement length sets the scale over which the le decays .
the two - body entanglement @xmath3 is a measure of the entanglement between two separated spins , at sites @xmath11 and @xmath12 , and the rest of the spins @xcite .
let @xmath13 be the reduced density matrix for the two spins , obtained from the full density matrix by tracing out the spins other than the ones at sites @xmath11 and @xmath12 .
the two body entanglement @xmath3 is given by the von neumann entropy @xmath14 in a translationally invariant system , @xmath15 depends only on the distance @xmath16 .
as pointed out in @xcite , the spins that are entangled with one or both the spins at sites @xmath11 and @xmath12 contribute to @xmath15 .
the following results have been obtained in the case of the @xmath17 exactly solvable anisotropic @xmath18 model in a transverse magnetic field .
the model , away from the isotropic limit , belongs to the universality class of the transverse ising model .
the two - body entanglement @xmath3 has a simple dependence on the spin correlation functions in the large @xmath19 limit .
away from the critical point , @xmath3 is found to saturate over a length scale @xmath20 as @xmath19 increases . near the qcp , one obtains @xmath21 the entanglement length ( el ) , @xmath20 , has an interpretation similar to that in the case of le .
the el diverges with the same critical exponent as the correlation length at the qcp .
@xmath3 thus captures the long range correlations associated with a qpt . at the critical point itself
, @xmath22 has a power - law decay , i.e. , @xmath23 . in the limit of large @xmath19 , the first derivative of @xmath3 w.r.t .
a hamiltonian parameter develops a @xmath24like cusp at the critical point .
the universality and a finite - size scaling of the entanglement have also been demonstrated .
the one - site von neumann entropy @xmath25 is also known to be a good indicator of a qpt @xcite .
it provides a measure of how a single spin at the site @xmath11 is entangled with the rest of the system .
the reduced density matrix @xmath26 is obtained from the full density matrix by tracing out all the spins except the one at the site @xmath11 .
oliveira et al .
@xcite have proposed a generalized global entanglement ( gge ) measure @xmath4 which quantifies multipartite entanglement ( me ) .
@xmath4 for a translationally symmetric system is given by @xmath27_{lm}\mid^{2}]\label{4}\ ] ] where @xmath28 is the reduced density matrix of dimension @xmath29 .
the factor 2 in @xmath4 indicates that the reduced density matrix is that for a pair of particles .
@xcite considered qpts characterized by non - analyticities in the derivatives of the ground state energy .
these arise from the non - analyticities in one or more of the elements of the reduced density matrix . in terms of the gge ,
a discontinuity in @xmath4 signals a first order qpt , brought about by a discontinuity in one or more of the elements , @xmath30_{lm}$ ] of the reduced density matrix @xcite . a discontinuity or divergence in the first derivative of @xmath4 w.r.t .
the tuning parameter occurs due to a discontinuity or divergence in the first derivetives of one or more of the elements of the reduced density matrix .
the associated qpt is of second order .
non - analyticities in @xmath4 and its derivatives thus serve as indicators of qpts . in the case of the @xmath18 @xmath17 spin chain ,
the gge measure shows a diverging el as the qcp is approached .
the el @xmath31 where @xmath32 is the usual correlation length .
thus , both the length scales diverge with the same critical exponent near the qcp .
the relationship between entanglement and qpts has mostly been explored for spin-@xmath33 systems .
the entanglement properties of the ground states of certain spin@xmath34 hamiltonians have been studied using different measures @xcite .
numerical studies show that the le has the maximal value for the ground state of the spin-1 heisenberg antiferromagnet with open boundary conditions ( obc ) @xcite . in the case of the spin-1 affleck - kennedy - lieb - tasaki ( aklt ) model @xcite ,
the result can be proved exactly .
a class of spin-1 models , the @xmath35-deformed aklt models , is characterized by an exponentially decaying le with a finite el @xmath20 .
the length @xmath20 diverges at the point @xmath36 though the conventional correlation length remains finite @xcite .
a recent study @xcite shows that in the case of spin-1 systems , the use of le for the detection of qpts is not feasible .
an example is given by the @xmath37 xxz heisenberg antiferromagnet with single - ion anisotropy .
the model has a rich phase diagram with six different phases .
the le is found to be always 1 in the entire parameter region and hence is insensitive to qpts .
the ground states of certain spin-1 models have an exact representation in terms of matrix product states ( mps ) @xcite .
the ground state of the spin-1 aklt model , termed a valence bond solid ( vbs ) state , is an example of an mps .
the ground state is characterized by short - ranged spin - spin correlations and a hidden topological order known as the string order .
the excitation spectrum of the model is further gapped . in the mps formalism ,
ground state expectation values like the correlation functions are easy to calculate .
this has made it particularly convenient to study phase transitions in spin models with mp states as exact ground states @xcite .
the transitions identified so far include both first and second order transitions and are brought about by the tuning of the hamiltonian parameters .
the second order transition in the class of finitely correlated mp states , however , differs from the conventional qpt in one important respect .
the spin correlation function is always of the form @xmath38 for large @xmath19 .
the correlation length @xmath32 diverges as the transition point is approached .
the pre - factor @xmath39 , however , vanishes at the transition point @xcite .
this is in contrast to the power - law decay of the correlation function at a conventional qcp .
some distinct features of qpts in mp states have recently been identified @xcite .
one of these relates to the analyticity of the ground state energy density for all values of the tuning parameter . in a conventional qpt ,
the energy density becomes non - analytic at the qcp .
the mp states appear to provide an ideal playground for exploring novel types of qpts . in this paper
, we consider a spin-1 model , the exact ground state of which is given by an mp state @xcite .
the model has a rich phase diagram with a number of first order phase transitions and a critical point transition .
we study the entanglement properties of the ground state with a view to pinpoint the special features which appear close to the critical point .
this is done by using three different entanglement measures , namely , the single - site , two - body and generalized global entanglement defined earlier .
we consider a spin-1 chain hamiltonian proposed by klmper et al .
@xcite which describes a large class of antiferromagnetic ( afm ) spin-1 chains with mp states as exact ground states .
the hamiltonian satisfies the symmetries : ( i ) rotational invariance in the @xmath40 plane , ( ii ) invariance under @xmath41 and ( iii ) translation and parity invariance .
the hamiltonian has the general form @xmath42 @xmath43 @xmath44 where @xmath10 is the number of sites in the chain and periodic boundary conditions ( pbc ) hold true .
the parameters @xmath45 are real and @xmath46 is a constant .
the nearest - neighbour ( n.n . ) interactions are @xmath47 the constant @xmath46 in eq . ( 5 ) may be adjusted so that the ground state eigenvalue of @xmath48
. hence @xmath49 i.e. , @xmath50 has only non - negative eigenvalues . in the afm case , the @xmath51-component of the total spin of the ground state @xmath52 .
klmper et al .
showed that in a certain subspace of the @xmath53parameter space , the afm ground state has the mp form .
let @xmath54 and @xmath55 be the eigenstates of @xmath56 with eigenvalues @xmath57 , @xmath58 and @xmath34 respectively . define a @xmath59 matrix at each site @xmath12 by @xmath60 with non - vanishing parameters @xmath61 .
the global afm state is written as @xmath62 where ` @xmath63 ' denotes a tensor product .
one can easily check that @xmath64 , i.e. , the state is afm .
one now demands that the state @xmath65 is the exact ground state of the hamiltonian @xmath50 with eigenvalues @xmath57 . for this
, it is sufficient to show that @xmath66 eq .
( 3 ) and ( 10 ) are satisfied provided the following equalities @xmath67 and inequalities @xmath68 hold true .
the state @xmath65 is the ground state of the hamiltonian @xmath69 with ground state energy zero provided the equalities in @xmath70 are satisfied .
the inequalities constrain the other eigenvalues of @xmath71 to be positive .
if the inequalities are satisfied , the ground state can be shown to be unique for any chain length @xmath10 . also , in the thermodynamic limit @xmath72 , the excitation spectrum has a gap @xmath73 . with equality signs in the inequalities @xmath74 ,
the state @xmath65 is still the ground state but is no longer unique .
the spin@xmath34 model has the typical feature of a haldane - gap ( hg ) antiferromagnet .
in fact , the aklt model is recovered as a special case with @xmath75 and @xmath76 .
the state @xmath77 now represents the vbs state . using the transfer matrix method @xcite , the ground state correlation functions can be calculated in a straightforward manner .
the results are @xmath78 : longitudinal correlation function @xmath79 transverse correlation function@xmath80\left(\frac{-\sigma}{1+|a|}\right)^{r}\label{14}\ ] ] the correlations @xmath81 and @xmath82 decay exponentially with the longitudinal and transverse correlation lengths given by@xmath83 furthermore , the string order parameter has a non - zero expectation value in the ground state .
one finds that the correlation lengths diverge as @xmath1 . at the point @xmath0 ,
the correlation functions given by eq . @xmath81 and @xmath82 are zero . at a conventional qcp ,
the correlation functions have a power - law decay .
we will , however , refer to the point as a qcp since the correlation lengths diverge as the point is approached .
a consequence of the diverging correlation length is that the excitation spectrum of the spin-1 model , which is gapped ( the haldane phase ) for @xmath84 , becomes gapless at the critical point @xmath0 @xcite .
the presence or absence of a gap in the excitation spectrum of a system is reflected in the low temperature thermodynamic properties of the system .
furthermore , the string order parameter has a non - zero expectation value in the ground state for @xmath84 and becomes zero at @xmath0 indicating the appearance of a new phase .
@xcite provide several other examples of spin-1 models with finitely correlated mp states as exact ground states .
all these models exhibit critical point transitions with features similar to those in the case of the spin-1 model described by the hamiltonian in eq .
we now focus on the entanglement properties of the mp ground state ( eq .
@xmath77 ) .
we consider @xmath5 to be @xmath85 and @xmath86 in eq . @xmath70 .
the one - site reduced density matrix @xmath26 ( eq . @xmath87 ) obtained by tracing out all the spins except the @xmath11-th spin from the ground state density matrix @xmath88@xmath89 , can be calculated using the transfer matrix method @xcite .
the density matrix , from eq .
@xmath77 , is@xmath90 the summation is over all the indices , @xmath91 , @xmath92 , @xmath93 . we define a @xmath94 matrix @xmath95 ( the elements of which are operators ) at any lattice site as @xmath96 the convention of the ordering of the multi - indices is @xmath97 .
thus , @xmath95 can be written as @xmath98 also,@xmath99 where the trace is over all the spins except the @xmath11-th one .
the transfer matrix @xmath100 at a site @xmath101 is obtained by taking the trace over @xmath95 at the same site , i.e. , @xmath102 where the states @xmath103 are the states @xmath54 , @xmath104 .
the transfer matrix @xmath100 is obtained as@xmath105 the eigenvalues are @xmath106 the corresponding eigenvectors are@xmath107 from eq . @xmath108,@xmath109
the factor in the denominator takes care of the condition @xmath110 .
on taking the thermodynamic limit @xmath72 , we get @xmath111 in the @xmath112 basis , the reduced density matrix becomes@xmath113 the calculation of the two - site reduced density matrix @xmath13 follows in the same manner .
@xmath13 is given by@xmath114 where the trace is taken over all the spins except the @xmath11-th and @xmath12-th ones .
@xmath115 in the thermodynamic limit @xmath72 , @xmath13 reduces to @xmath116 where @xmath117 .
the matrix @xmath13 is a @xmath118 matrix and defined in the two - spin basis states @xmath119 with the ordering @xmath120 the non - zero matrix elements , @xmath121 @xmath122 , of @xmath13 are : @xmath123 it is easy to check that @xmath13 has a block - diagonal form .
we now determine the entanglement content of the ground state @xmath125 ( eq .
@xmath77 ) of the hamiltonian ( eq .
@xmath69 ) using the entanglement measures @xmath2 , @xmath3 , and @xmath4 . the calculations are carried out for different values of the parameter @xmath5 in eq . @xmath70 .
the ultimate aim is to probe the special features , if any , of entanglement in the vicinity of the qcp at @xmath0 . from eq . @xmath87 and
@xmath126 , the one - site entanglement @xmath127\label{32}\ ] ] figure @xmath128 ( top ) shows the variation of @xmath2 w.r.t .
@xmath5 . the one - site entanglement has the maximum possible value @xmath129 .
this is attained at the aklt point @xmath130 .
the vbs state is in this case the exact ground state . in the vbs state ,
each spin-1 at a specific lattice site can be considered as a symmetric combination of two spin-@xmath33 s @xcite . in the vbs state ,
each spin-@xmath33 at a particular lattice site forms a spin singlet with a spin-@xmath33 at a neighbouring lattice site .
@xmath2 has the value zero at the qcp @xmath0 .
figure @xmath128 ( bottom ) shows the variation of @xmath131 with the parameter @xmath5 .
the derivative diverges as the qcp is approached .
this is the expected behaviour at the qcp of a conventional qpt . in the latter case , however , @xmath2 has the maximum value at the qcp @xcite . from eq .
@xmath132 and @xmath133 , the two - body entanglement @xmath3 is @xmath134 where @xmath135 s are the eigenvalues of the reduced density matrix @xmath13 .
these are given by@xmath136 knowing the reduced density matrix @xmath13 , the correlation functions @xmath137
( @xmath138 ) can be calculated in the usual manner .
one then recovers the expressions in eq . @xmath81 and @xmath82 ( @xmath139 , where @xmath117 ) .
figure @xmath140 ( top ) shows the variation of @xmath3 as a function of @xmath5 for @xmath141 .
figure @xmath140 ( bottom ) shows the variation of the derivative @xmath142 w.r.t .
@xmath5 for the same value of @xmath19 .
the maximum of @xmath3 is at the aklt point @xmath130 and has the value zero at @xmath0 . for large @xmath19 ,
the derivative @xmath142 diverges near the qcp at @xmath0 .
the last fearure is characteristic of a conventional qpt @xcite .
we next calculate the gge @xmath4 ( eq .
@xmath143 ) .
this is easily done as the matrix elements of the reduced density matrix ( eq .
@xmath87 ) are known .
figure @xmath144 ( top ) shows the variation of @xmath4 versus @xmath5 for @xmath141 .
figure @xmath144 ( bottom ) shows the plot of @xmath145 against @xmath5 .
again @xmath4 has the maximum value at the aklt point and is zero at @xmath0 .
the derivative @xmath145 does not diverge as @xmath1 in contrast to the case of a conventional qpt @xcite .
the derivative , however , attains the maximum value at the qcp @xmath0 .
figure @xmath146 ( top ) shows the plots of @xmath2 , @xmath3 , and @xmath4 versus @xmath5 for @xmath141 .
figure @xmath146 ( bottom ) shows the plots of the first derivatives of the same quantities w.r.t .
@xmath5 for @xmath141 .
the plots are shown for comparing the different entanglement measures .
we next determine the el @xmath20 and its variation w.r.t .
the parameter @xmath5 .
we consider the entanglement measure @xmath3 for this purpose .
close to the qcp @xmath0 and in the limit of large @xmath19 , one can write@xmath147 the longitudinal and transverse correlation functions , @xmath148 and @xmath149 are given by eq . @xmath81 and @xmath82 with @xmath139 . for @xmath150 , @xmath151 decays faster than @xmath152 with @xmath19 . the eigenvalues
@xmath135 s , @xmath153 , can be expressed in terms of the correlation functions @xmath151 and @xmath152 . for large @xmath19
, the contributions from @xmath151 can be ignored .
the eigenvalues now reduce to the expressions @xmath154 from eq . @xmath132 and
@xmath155 , one ultimately arrives at the expressions@xmath156 the pre - factor @xmath157 at the qcp @xmath158 .
the el @xmath159 is given by @xmath160 where @xmath161 is the transverse correlation length ( eq .
@xmath162 ) . in the case of the @xmath17 anisotropic xy model in a transverse field , an expression similar to that in eq .
@xmath163 is obtained close to the qcp in the limit of large @xmath19 @xcite . the pre - factor in this case , however , does not vanish at the qcp but has a power - law dependence on @xmath19 .
figure @xmath164 shows the variation of @xmath20 w.r.t .
@xmath5 based on the entanglement measure @xmath3 .
the total correlations , with both classical and quantum components , between two sites @xmath11 and @xmath12 are quantified in terms of the quantum mutual information @xcite@xmath165 as explained in @xcite , a comparison of the singular behaviour of @xmath2 with that of @xmath166 allows one to determine whether two - point ( @xmath167 ) or multipartite ( @xmath168 ) quantum correlations are important in a qpt .
figure @xmath169 ( top ) shows a plot of @xmath166 versus @xmath5 for @xmath141 .
figure @xmath169 ( bottom ) shows the variation of @xmath170 versus @xmath5 for @xmath141 . the derivative does not diverge as @xmath1 , a behaviour distinct from that of @xmath131 close to @xmath0
the difference in the singular behaviour of quantities associated with @xmath2 and @xmath166 indicates that multipartite quantum correlations are involved in the qpt . *
fig . 1 : * plot of @xmath2 ( top ) and @xmath131 ( bottom ) vs. @xmath5 .
* fig . * * 2 : * plot of @xmath3 ( top ) and @xmath142 ( bottom ) as a function of @xmath5 .
* fig . 3 : * plot of @xmath4 ( top ) and @xmath145 ( bottom ) as a function of @xmath5 . * fig . 4 : * plots of @xmath2 ( @xmath171 ) , @xmath3 ( @xmath172 ) , and @xmath4 ( @xmath173 ) ( top ) and the corresponding first derivatives w.r.t .
@xmath5 ( bottom ) as a function of @xmath5 .
@xmath6 represents the entanglement measure . *
fig . * * 5 : * plot of el as a function of @xmath5 .
* 6 : * plots of @xmath166 ( top ) and @xmath170 ( bottom ) as a function of @xmath5 .
the mp states provide exact representations of the ground states of several spin models in low dimensions @xcite .
the remarkable features of such states arises from the fact that complicated many body states have a simple factorized form .
the simplicity in structure makes the calculation of the ground state expectation values particularly easy to perform .
the spin-1 aklt model is a well - known example of a spin model in 1d the exact ground state of which ( a vbs state ) has an mp representation .
the aklt model and the spin-1 heisenberg afm chain belong to the same universality class @xcite .
the insight gained from the study of models in the mp formalism is expected to be of relevance in understanding the properties of more physical systems .
the mp states also serve as good trial wave functions for standard spin models .
the mp representation lies at the heart of the powerful density matrix renormalization group ( dmrg ) method and provides the basis for several interesting developments in quantum information @xcite .
studies of the entanglement characteristics of the mp states have begun only recently . the qpts which occur in such states have characteristics different from those of conventional qpts .
it is thus of considerable interest to determine whether the entanglement content of mp states develops special features close to a qcp . in this paper
, we consider a spin-@xmath128 model the exact ground state of which is of the mp form over a wide range of parameter values .
the model exhibits a novel qpt in that the longitudinal and transverse correlation lengths diverge as the qcp is approached but the correlation functions vanish at the qcp . in a conventional qpt ,
the correlation functions have a power - law decay at the qcp . in the spin-1 model ,
the divergence of correlation lengths is accompanied by the excitation gap going to zero .
the string order parameter , which has a non - zero expectation value in the mp state for @xmath84 , vanishes at the qcp @xmath0 .
the distinctive signatures indicate the appearance of a new phase .
we study the entanglement properties of the mp state for different values of the parameter @xmath5 .
the measures used are @xmath2 ( one - site von neumann entropy ) , @xmath3 ( two - body entanglement ) and @xmath4 ( gge ) .
all the entanglement measures have zero value at the qcp so that the ground state is disentangled at that point .
as seen from the different plots , figures @xmath132 , @xmath174 , @xmath87 , and @xmath143 , the entanglement content , as measured by @xmath175@xmath2 , @xmath3 and @xmath4 , has a slow variation w.r.t .
@xmath5 for @xmath176 . at the aklt point
@xmath130 , @xmath6 reaches its maximum value and as @xmath5 is reduced further , the magnitude of @xmath6 falls rapidly to approach zero value at @xmath0 .
the study of conventional qpts shows that @xmath6 is maximum at a qcp @xcite .
also , @xmath177 diverges as the qcp is approached .
the el , @xmath20 , as calculated from @xmath3 and @xmath4 for large @xmath19 , also diverges with @xmath31 where @xmath32 is the usual correlation length . in the case of the spin-1 model under consideration
, @xmath177 diverges as @xmath1 when @xmath178 and @xmath3 .
the el , @xmath20 , as calculated from @xmath3 in the large @xmath19 limit , also diverges with @xmath31 .
one now has the interesting situation that the entanglement content of the mp state decreases as @xmath1 but the entanglement is spread over larger distances .
the derivative @xmath145 , however , does not diverge as @xmath1 but attains a maximum value at the qcp .
the results can be understood by noting that in all the three cases , @xmath179 and @xmath4 , the reduced density matrices @xmath26 and @xmath13 smoothly approach the forms associated with pure states as the parameter @xmath5 tends to zero .
the matrix elements of the reduced density matrices do not develop non - analyticities in the parameter region of interest .
thus , the energy density , calculated from the reduced density matrix @xmath13 , does not develop a non - analyticity at the qcp .
the derivative @xmath145 depends upon the first derivatives of the matrix elements of @xmath13 .
since the latter is analytic for @xmath180 , @xmath145 does not diverge in the whole parameter regime including the point @xmath0 . in the cases of the entanglement measures @xmath178 and @xmath3 , the derivative @xmath177 diverges as @xmath1 due to the divergence of @xmath181 in the same limit .
the divergence is thus due to the special form of the von neumann entropy and occurs for @xmath182 .
a recent work @xcite provides another example of such a singularity . though @xmath145 does not diverge or become discontinuous at the qcp @xmath0 , it attains its maximum value at the point .
the first derivative of the string order parameter w.r.t .
@xmath5 also attains its maximum value at @xmath0 though the order parameter itself vanishes at the point .
figure @xmath143 shows that amongst the three entanglement measures @xmath175@xmath2 , @xmath3 and @xmath4 , used in this study to obtain a quantitative estimate of the entanglement content of the mp ground state , the measure @xmath3 yields the largest value of the entanglement at different values of @xmath5 . the difference in the singular behaviour of the measures @xmath2 and the mutual information entropy @xmath166 as @xmath1 indicates that multipartite quantum correlations are involved in the qpt . in summary ,
the present study identifies the entanglement characteristics of the mp ground state of a spin-1 model close to the critical point @xmath0 .
the features are distinct from those associated with conventional qpts .
several spin models are known for which the mp states are the exact ground states @xcite .
some of these models have interesting phase diagrams exhibiting both first and second order phase transitions .
it will be of interest to extend the present study to other spin models ( both @xmath17 and @xmath128 ) in order to identify the universal characteristics of qpts in mp states .
a. t. is supported by the council of scientific and industrial research , india , under grant no .
9/15 ( 306)/ 2004-emr - i .
10 s. sachdev , quantum phase transitions ( cambridge university press , cambridge , 2000 ) ; s. l. sondhi et al . ,
69 , 315 ( 1997 ) .
a. osterloh , l. amico , g. falci and r. fazio , nature 416 , 608 ( 2002 ) .
t. j. osborne and m. a. nielsen , phys .
a 66 , 032110 ( 2002 ) .
g. vidal , j. i. latorre , e. rico and a. kitaev , phys .
90 , 0279011 ( 2004 ) .
wu , m. s. sarandy and d. a. lidar , phys .
lett 93 , 250404 ( 2004 ) .
t. r. de oliveira , g. rigolin , m. c. de oliveira and e. miranda , phys .
97 , 170401 ( 2006 ) .
d. chen , cond - mat/0606126 .
i. bose and e. chattopadhyay , phys .
a 66 , 062320 ( 2002 ) .
f. c. alcaraz , a. saguia and m. s. sarandy , phys .
a 70 , 032333 ( 2004 ) .
j. vidal , r. mosseri and j. dukelsky , phys .
a 69 , 054101 ( 2004 ) .
f. verstraete , m. popp and j. i. cirac , phys .
92 , 027901 ( 2004 ) .
h. fan , v. korepin and v. roychoudhury , phys .
93 , 227203 ( 2004 ) .
f. verstraete , m. a. martin - delgado and j. i. cirac , phys .
92 , 087201 ( 2004 ) .
i. affleck , t. kennedy , e. h. lieb and h. tasaki , phys .
59 , 779 ( 1987 ) ; commun .
115 , 447 ( 1988 ) .
l. campos venuti and m. roncaglia , phys .
94 , 207207 ( 2005 ) .
a. klmper , a. schadschneider and j. zittartz , z. phys .
b 87 , 281 - 287 ( 1992 ) .
a. k. kolezhuk and h. j. mikeska , int
b 12 , 2325 ( 1998 ) and references therein . m. fannes , b. nachtergaele and r. f. werner , europhys . lett . 10 , 633 ( 1989 ) ; commun . math .
144 , 443 ( 1992 ) .
m. m. wolf , g. ortiz , f. verstraete and j. i. cirac , phys .
97 , 110403 ( 2006 ) .
a. klmper , a. schadschneider and j. zittartz , europhys .
24 , 293 ( 1993 ) .
a. anfossi , p. giorda , a. montorsi and f. traversa , phys .
95 , 056402 ( 2005 ) .
a. anfossi , p. giorda , a. montorsi and f. traversa , cond - mat/0611091 .
s. brehmer , h. -j .
mikeska and u. neugebauer , j. phys . : condens .
matter 8 , 7161 ( 1996 ) .
d. prez - garca , f. verstraete , m. m. wolf and j. i. cirac , quant - ph/0608197 .
m. cozzini , r. ionicioiu and p. zanardi , cond - mat/0611727 .
|
in this paper , we study the entanglement properties of a spin-1 model the exact ground state of which is given by a matrix product state .
the model exhibits a critical point transition at a parameter value @xmath0 .
the longitudinal and transverse correlation lengths are known to diverge as @xmath1 .
we use three different entanglement measures @xmath2 ( the one - site von neumann entropy ) , @xmath3 ( the two - body entanglement ) and @xmath4 ( the generalized global entanglement ) to determine the entanglement content of the mp ground state as the parameter @xmath5 is varied .
the entanglement length , associated with @xmath3 , is found to diverge in the vicinity of the quantum critical point @xmath0 .
the first derivative of the entanglement measure @xmath6 @xmath7 w.r.t .
the parameter @xmath5 also diverges .
the first derivative of @xmath4 w.r.t .
@xmath5 does not diverge as @xmath1 but attains a maximum value at @xmath0 . at the qcp itself
all the three entanglement measures become zero .
we further show that multipartite correlations are involved in the qpt at @xmath0 .
department of physics bose institute 93/1 , acharya prafulla chandra road kolkata - 700 009 , india
|
dwarf starburst galaxies are the more numerous and fainter counterparts of traditional starburst galaxies
. dwarf starbursts will have a particularly important role in the energizing and polluting of the intergalactic medium through galactic winds , as , because of their low escape velocities , even a moderate starburst event in a dwarf galaxy can result in the expulsion of a large fraction of the nuclear processed material in a superwind or outflow ( dekel & silk 1986 ; mac low & ferrara 1999 ; strickland & stevens 2000 ) .
understanding the complexities of starburst processes in dwarf galaxies is therefore an important step in understanding the feedback processes between star - formation and galactic evolution .
radio continuum observations of starburst galaxies are particularly useful in studying the star - formation processes , suffering little of the extinction that complicates matters in the optical and uv , which may appear to be more suitable wavebands for studying young massive star populations ( see hopkins et al .
2001 for a discussion of the merits of different wavelength regimes for finding star formation rates ) .
radio observations can also be used to study populations of supernovae remnants ( snrs ) in starburst galaxies ( e.g. kronberg et al .
2000 , greve et al . 2002 ) , and recent radio observations have revealed objects inferred to be optically obscured super star clusters ( see below ) .
earlier radio observations of dwarf starbursts , at low spatial resolution , have shown them to have a higher radio to optical luminosity ratio than normal spiral galaxies and spectral indices significantly flatter than a sample of spiral galaxies ( klein , wielebinski & thuan 1984 ) .
this was attributed to lower magnetic fields in the dwarf galaxies effectively suppressing non - thermal emission .
however , more recent and higher resolution observations are showing a more complex and interesting picture .
vla observations of the dwarf starburst henize 2 - 10 by kobulnicky & johnson ( 1999 , see also kobulnicky & johnson 2000 ) found several compact radio sources in the central regions .
these sources have positive spectral indices ( i.e. @xmath2 , with the radio flux defined as @xmath3 ) , indicative of optically thick bremsstrahlung emission .
these radio knots are believed to be unusually dense regions , with sizes of @xmath4pc and densities @xmath5 , associated with optically observed super star clusters .
further , vla observations of ngc5253 ( another dwarf starburst ) showed unusually high levels of free - free emission , most of it from a single source ( turner , ho & beck 1998 ) . several regions with a positive spectral index were also detected , which were inferred to be very young regions of star - forming activity .
these authors also found regions of non - thermal emission , probably associated with snrs .
further observations showed what was inferred to be a dense compact region , with a size of only 12pc , but containing several thousand o - stars , and which was termed a radio supernebula ( turner , beck & ho 2000 ) .
beck , turner & kovo ( 2000 ) have presented vla observations of nine wolf - rayet galaxies ( a particular class of starburst galaxies , containing very young and massive wolf - rayet stars , conti 1991 ) , including seven dwarf starbursts . in the majority of these galaxies
the global spectral indices were much flatter than in spiral galaxies ( the only exception being ngc3049 which is a large spiral galaxy as well as a wolf - rayet galaxy , which has a spectral index of @xmath6 ) .
this flatter index indicates a higher fraction of free - free emission in these galaxies , and in fact many of the galaxies had regions with positive spectral index , indicative of self - absorbed free - free emission and hence very recent and compact star - formation .
these objects are probably related to the optical super star clusters ( sscs ) observed with _ hst _ in many starbursts and galaxy mergers ( see whitmore 2002 for a review ) , but which are viewed at a younger age , and which are still enclosed within the parental cloud , obscuring them at optical wavelengths .
other examples of optically obscured sscs seen at radio wavelengths are found in ngc2146 ( a spiral starburst galaxy ) , where a number of radio sources with positive spectral indices are seen , and that have been interpreted as being due to massive ultra - compact regions , indicative of extremely young star - formation ( tarchi et al.2000 ) .
kobulnicky & johnson ( 2001 ) and johnson et al . (
2001 ) summarise the status of radio observations of these young starburst regions , and coins the phrase ultra - dense regions ( or udregions ) .
it is worth noting that _ merlin _ observations of the nearby ( 2.2mpc ) dwarf starburst ngc1569 found no emission associated with the optically visible sscs in this galaxy , but did find a number of radio supernovae or supernovae remnants ( greve et al .
2002 ) .
in this paper we report on radio observations , taken with the australia telescope compact array ( atca ) , of two dwarf starburst galaxies , ngc3125 and ngc5408 , which are also wolf - rayet galaxies .
because wolf - rayet galaxies contain very young massive stars and are thus sites of very recent star - formation , they are the obvious place to look for optically obscured star clusters .
radio maps with good spatial resolutions are therefore very important in understanding the global properties of these objects . the paper is organized as follows : in section 2 the characteristics of the target galaxies are discussed , in section 3 the radio observations and analysis are described and the results are discussed in section 4 .
.parameters for ngc3125 and ngc5408 . [ cols="<,^,^",options="header " , ] [ tab3 ] the observed radio morphologies of these two dwarf starbursts shows several discrete regions of emission , in which much of the observed emission is coincident with massive star - clusters in both galaxies .
leaving aside the possibility of an agn and emission mechanisms associated with a massive black - hole ( for which there appears to be no evidence of in these galaxies ) the likely emission mechanisms are ( c.f .
forbes & norris 1998 ) : 1 .
supernovae and supernova remnants : cosmic rays accelerated by supernovae and supernova remnants , which then interact with the interstellar magnetic field to emit synchrotron emission .
the typical spectral index of this synchrotron emission will be @xmath7 . in starburst and normal galaxies
this component will tend to be distributed throughout the disk .
however , in ngc3125 and ngc5408 we are dealing with dense stellar clusters , which could contain many sn remnants in a small volume , giving rise to a compact region of synchrotron emission .
regions : hot electrons in regions generate free free emission . at cm wavelengths
most regions are optically thin , resulting in a flat spectrum ( @xmath8 ) .
however , some compact hii regions are seen to be optically thick , giving rise to a rising spectrum at cm wavelengths ( ngc5253 turner , ho & beck 1998 ; hen 2 - 10 kobulnicky & johnson 1999 ; ngc2146 tarchi et al .
200 ; m82 mcdonald et al .
2002 ) . in these cases , which are probably associated with very young super star clusters , the gas densities are sufficient to cause a turn - over in the emission at low frequencies . in all regions
the radio spectrum does turnover at some sufficiently low frequency ( for instance , for orion this occurs at @xmath9 m ) .
the radio emission from ultra luminous infrared galaxies is optically thick to free free absorption , so that the typical synchrotron spectrum of these galaxies is flattened at low frequencies ( condon et al .
1991 ) .
the combined result of these effects in seyfert and starburst galaxies is to produce a typical radio spectral index of @xmath10 ( from the extended synchrotron emission ) with a flattening at low frequencies in some starburst sources because of free free absorption . for ngc3125 and ngc5408 ,
the morphology and spectral indices of the radio knots imply that the observed emission is primarily associated with type ii supernovae activity which is associated with star - formation activity within the star - clusters .
the exception is ngc5408:a , which has a flatter index indicative of optically thin free - free emission .
the obvious interpretation is that in this source , which is associated with an ssc , the emission is dominated by the @xmath11 region , and that the cluster is perhaps younger than the others in ngc5408 and supernova activity has not yet begun to dominate the radio emission .
the next step is to see if we can develop a coherent picture of the star - forming behaviour of these galaxies .
it is worth noting that in these galaxies the optical light may be dominated ( or have a substantial contribution ) from the star - clusters , but it is likely that the stellar content these clusters only comprise a few percent of the total stellar mass in the galaxies .
however , these clusters may provide a substantial fraction of the ongoing star - formation in the galaxy .
we can estimate the expected supernova rate in the radio emitting regions .
assuming that the 4.80ghz flux is non - thermal emission , then the implied sn rate in the various radio emitting regions can be estimated from ( condon & yin 1990 ) : @xmath12 for ngc3125 this corresponds to sn rate of @xmath13 for ngc3125:a and @xmath14 for ngc3125:b .
the corresponding sn rates for the ngc5408 regions are in the range @xmath15 .
similarly , we can estimate the star - formation rates from both far - infrared ( fir ) observations and these radio observations .
the fir rates will apply to the whole galaxy while the radio ones will apply to the observed emission regions ( see hopkins et al .
2001 for a comparison of the various methods for estimating star - formation rates in galaxies ) the fir luminosities of @xmath16 ( ngc3125 ) and @xmath17 ( ngc5408 ) .
from kennicutt ( 1998 ) we can estimate the total star - formation rate ( sfr ) in each galaxy from the following relationship : @xmath18 where the mass range for the star - formation rate is assumed to be @xmath19 and @xmath20 is the far infrared luminosity .
this means that the estimated total sfr in the two galaxies are @xmath21 ( ngc3125 ) and @xmath22 ( ngc5408 ) .
note , that assuming a salpeter imf , the star - formation rate of high mass , that is stars with @xmath23 , will be about 10 per cent of this value . using the results of condon ( 1992 , see also haarsma et al .
2000 ) we can estimate the star - formation rates in the radio regions from these radio observations .
the implied total star - formation rate in the two regions in ngc3125 is @xmath24 and in ngc5408 the figure is @xmath25 , somewhat higher than the fir values . as an aside
, we can compare the estimated sn rates with the theoretical expectations from leitherer & heckman ( 1995 ) . for an ongoing starburst with a star - formation rate of @xmath26 , a salpeter imf and solar abundance , the expected supernova rate is @xmath27 . for the sfr rates estimated above
, this corresponds to supernova rates of @xmath28 for ngc3125 and @xmath29 for ngc5408 . comparing these values to the estimated sn rates above
, there is reasonable agreement between the integrated sn rates in the various radio emitting regions and these expectations . as discussed earlier ,
both galaxies are also x - ray luminous .
there are several potential mechanisms for generating the x - rays ; an agn , x - ray binaries , snrs and emission from superbubbles .
there is no evidence for an agn , and while individual supernovae or snrs can be very radio and x - ray luminous ( van dyk et al .
1993 ; fabian & terlevich 1996 ; immler & lewin 2002 ) , their duration as high luminosity object is short - lived .
consequently , high mass x - ray binaries , associated with the active star - formation in these galaxies , are likely to provide much of the observed x - ray emission . extrapolating from the star - formation rate in our own galaxy ( @xmath30 )
we can estimate the number of luminous x - ray binaries in each system .
the galaxy has an estimated population of luminous high mass x - ray binaries ( hmxbs ) of @xmath31 ( dalton & sarazin 1995 ) .
based on this , these two dwarf starbursts should have a high mass x - ray binary population of @xmath32 for ngc3125 and @xmath33 for ngc5408 .
note , that this refers only to the most luminous hmxbs , namely those with @xmath34 .
the observed x - ray luminosities of these galaxies are @xmath35 ( ngc3125 ) and @xmath36 ( ngc5408 ) , meaning that ngc5408 is substantially more x - ray luminous than ngc3125 , although having a smaller star - formation rate . the x - ray emission from ngc3125
could be dominated by x - ray binaries , however , this is unlikely to be the case for ngc5408 and we may need an additional very x - ray luminous source , such as the objects seen by the satellite in other star - forming galaxies ( for instance , kaaret et al .
2001 ) believed to be intermediate mass black - holes .
consequently , from these observations , we can conclude that the radio emission in ngc3125 and ngc5408 is predominantly non - thermal in nature and is most likely dominated by processes associated with type ii snrs which are associated with recent starburst activity in the massive stellar clusters .
only one source in ngc5408 has a flatter spectral index indicative of optically thin free - free emission .
it will remain to be seen if the cluster associated with this radio source is younger than the other clusters in this galaxy . in the case of
ngc3125 we have a coherent picture , whereby the expected star - formation rate , supernovae rate and massive x - ray binaries populations can give rise to the observed properties , while in the case of ngc5408 although the fir and radio luminosities are broadly consistent the galaxy is substantially x - ray overluminous .
this galaxy may well be the home of a single super - luminous system such as has been seen in several other galaxies .
one problem with the simple analysis presented here is that it has mostly assumed a constant star - formation rate . in many of these stellar clusters
there may be a coeval population of stars and in this situation the sn rate can be very time variable .
a more detailed optical analysis is needed for these galaxies , to investigate the masses / ages of the star - clusters .
however , these new observations are throwing some more light on the radio evolution of massive star - clusters in dwarf starbursts . in conclusion ,
we have presented new radio observations of two dwarf starburst galaxies , ngc3125 and ngc5408 .
several discrete regions of radio emission are seen in these galaxies , most of them associated with massive young star clusters .
the radio spectral indices indicate the emission is dominated by synchrotron emission associated with supernova activity , though in one case a flatter spectrum is observed , possibly indicative of a younger cluster .
the australia telescope is funded by the commonwealth of australia for operation as a national facility funded by csiro .
ian stevens is supported by a pparc advanced fellowship .
|
we present new radio continuum observations of two dwarf starburst galaxies , ngc3125 and ngc5408 , with observations at 4.80ghz ( 6 cm ) and 8.64ghz ( 3 cm ) , taken with the australia telescope compact array ( atca ) .
both galaxies show a complex radio morphology with several emission regions , mostly coincident with massive young star clusters .
the radio spectral indices of these regions are negative ( with @xmath0 ) , indicating that the radio emission is dominated by synchrotron emission associated with supernova activity from the starburst .
one emission region in ngc5408 has a flatter index ( @xmath1 ) indicative of optically thin free - free emission , which could indicate it is a younger cluster .
consequently , in these galaxies we do not see regions with the characteristic positive spectral index indicative of optically obscured star - formation regions , as seen in other dwarf starbursts such as hen 2 - 10 .
galaxies : individual : ngc3125 , ngc5408 galaxies : starburst galaxies : radio emission
|
standard machine learning algorithms require a huge number of training examples to uncover hidden features , while humans and other animals can learn new concepts from only a few examples without any supervision signal @xcite .
learning hidden features in unlabeled training examples is called unsupervised learning . understanding how the number of examples confines the learning process is of fundamental importance in both cognitive neuroscience and machine learning @xcite .
as already observed in training of deep neural networks , unsupervised pretraining can significantly enhance the final performance , because the unsupervised pretraining provides a good initial region in parameter space from which the final fine - tuning starts @xcite . however , there are few theoretical works addressing how unsupervised learning extracts hidden features .
one potential reason is that the unsupervised learning process in a deep neural network is typically very complicated .
hence , understanding the mechanism of unsupervised learning in simple models is of significant importance .
this topic is recently studied based on the bayesian inference framework @xcite . in this recent work
, the authors treated each example ( data ) as a constraint on the factor graph , and reformulated the learning of model parameters as a bayesian inference problem on graphical models , and derived the message passing equations to infer the hidden feature from a finite amount of data .
they observed an entropy crisis in a simple restricted boltzmann machine ( rbm ) model , and predicted a discontinuous phase transition .
however , in an approximate hopfield model obtained by a high - temperature expansion of the rbm model , the entropy crisis is absent , and instead , a continuous transition is observed .
these properties observed in studies of single instances capture key characteristics of the unsupervised feature learning . here
, we further demonstrate that the message passing equation derived in the recent work @xcite agrees with the statistical analysis of the system in the thermodynamic limit via replica theory , a standard theoretical tool in spin glass theory of disordered systems @xcite .
the replica computation predicts the location of phase transition separating an impossible - to - infer regime to inferable regime . this transition can be continuous depending on the embedded feature strength .
a discontinuous phase transition always exists in a restricted boltzmann machine learning , but absent in an approximate hopfield model where only continuous phase transition is observed .
we also develop an iterative equation to infer the hyper - parameter ( temperature ) hidden in the data , which in physics corresponds to iteratively imposing nishimori condition .
this iterative scheme can even quantitatively predict how apparent features embedded in a real dataset are .
our analysis gives a thorough understanding of novel properties of the restricted boltzmann machine within replica symmetric approximation .
this paper is structured as follows . in sec .
[ srbmm ] , we introduce a simple rbm model for unsupervised feature learning , and propose the bayesian inference framework to derive the message passing equation on factor graph representation of the learning process , and this equation is then statistically analyzed and compared with replica computation under the replica symmetric assumption
. a more efficient approximate message passing equation is also derived .
we also derive an iterative equation based on bayes rule to predict the unknown temperature ( feature strength ) in the data . in sec .
[ ahopf ] , we approximate the rbm with the hopfield model where the stored pattern is interpreted as the feature vector . similar statistical analysis is carried out , and its physical implications are discussed .
we end the paper with a summary in sec .
restricted boltzmann machine is a basic unit widely used in building a deep belief network @xcite .
it consists of two layers of neurons .
the visible layer receives the input examples while the other hidden layer builds an internal representation of the input .
no lateral connections exist within each layer for computational efficiency .
the symmetric connections ( synapses ) between visible and hidden neurons are considered as features the network tries to learn from a large number of training examples .
it is a common strategy to use sampling - based gradient - decent method to learn features in the data @xcite , however , the gradient - decent learning is complicated and not amenable for analytical studies .
recent work showed that learning features can also be studied within a bayesian learning framework @xcite , which has the advantage of accounting for the uncertainty ( about the features ) caused by noises in the data @xcite , and furthermore can be analytically studied on probabilistic graphical models .
we focus on an unsupervised learning of finite samplings generated by a simple rbm , where a single hidden neuron is considered .
the task is to uncover an unknown rule embedded in the unlabeled data .
the rule is represented by a binary feature vector defined as @xmath0 where @xmath1 goes from @xmath2 to @xmath3 , the number of neurons in the visible layer .
we assume components of the true hidden feature vector connecting the visible neurons and the hidden neuron can take only two values , i.e. , @xmath4 or @xmath5 , with equal probabilities .
then this feature vector is used to generate independent random samples according to the joint probability @xmath6 , where @xmath7 , and @xmath8 is the visible configuration and @xmath9 is the hidden neuron s state . both @xmath9 and components of @xmath8 take binary values ( @xmath10 ) as well .
a rescaled feature factor by the system size is assumed .
the feature vector is also multiplied by an inverse - temperature parameter @xmath11 to investigate effects of the feature strength on the unsupervised learning . for simplicity ,
we consider the case of neurons without any external biases ( fields ) .
generalization to the case of neurons with external fields is straightforward ( appendix [ gener ] ) .
suppose we have @xmath12 independent samples or examples @xmath13 to learn the true hidden feature vector @xmath14 , using the bayes formula , we have the posterior distribution of the feature vector as @xmath15 where @xmath16 is the partition function of the model , @xmath17 goes over all examples and @xmath18 denotes a vector transpose operation .
a uniform prior probability for the feature vector is assumed for simplicity .
a large @xmath11 indicates the feature in the data is strong , and expected to be revealed by a few examples , while a weak feature vector may not be revealed by a huge number of examples .
each example serves as a constraint to the learning process .
once @xmath19 , the model becomes non - trivial as the partition function could not be computed exactly for a large number of visible neurons .
this @xmath12 can be finite or proportional to the system size , and in the latter case we define a data density as @xmath20 .
hereafter , we omit the conditional dependence of @xmath21 on @xmath22 . indicates the strength ( @xmath10 ) with which the feature component @xmath23 is related to @xmath17-th example .
right panel : the top panel shows constraint @xmath17 collects information from its neighboring feature nodes other than @xmath1 and produces an output message to node @xmath1 . the bottom panel shows node @xmath1 collects information from its neighboring constraints other than @xmath24 and produces an output message to node @xmath24 . the figure is taken from ref @xcite . ]
we call optimizing the marginal posterior probability of feature vectors given the data as bayesian learning in the current unsupervised learning context
. that is , we compute the maximizer of the posterior marginals ( mpm ) estimator @xmath25 @xcite .
we define the overlap between the inferred feature vector and the true one as @xmath26 , where @xmath27 is the inferred feature vector .
the mpm estimator maximizes the overlap .
if @xmath28 , the examples do not give any information about the feature vector . if @xmath29 , the feature vector is perfectly determined .
the statistical inference now is simplified to the computation of marginal probabilities , e.g. , @xmath30 , which is still a hard problem due to the interaction among example constraints . however , by mapping the model ( eq .
( [ pobs ] ) ) onto a factor graph @xcite , the marginal probability can be estimated by message passing ( fig .
[ rbm ] ) . for simplicity , we give the final simplified message passing equations ( smp ) as follows ( detailed derivations are given in ref @xcite , also in appendix [ smp - rbm ] ) : [ bp0 ] @xmath31 where @xmath32 .
the cavity magnetization is defined as @xmath33 .
@xmath34 can be interpreted as the message passing from feature @xmath1 to the data constraint @xmath17 , while @xmath35 can be interpreted as the message passing from data constraint @xmath24 to its feature @xmath1 .
if the weak correlation assumption ( also named bethe approximation @xcite ) is self - consistent , the smp would converge to a fixed point corresponding to a stationary point of the bethe free energy function with respect to the cavity messages @xmath36 @xcite . from this fixed point , one can extract useful information about the true feature vector of the data , by calculating the marginal probability as @xmath37 where @xmath38 . note that we perform the feature inference using the same inverse - temperature as used to generate the data , thus the inference is bayes - optimal and satisfies the nishimori condition @xcite . in the simulation section
, we also perform the bayesian inference using slightly different temperatures . next , we compute the bethe free energy , which is obtained by @xmath39 .
the free energy contribution of a feature node reads @xmath40+\ln\bigl(1+\prod_{a\in\partial i}\mathcal{g}_{a\rightarrow i}\bigr),\ ] ] and the free energy contribution of a data node reads @xmath41 where we define @xmath42 , @xmath43 , @xmath44 , and @xmath45 .
another important quantity is the number of feature vectors consistent with the presented random samplings , characterized by the entropy per neuron @xmath46 . in the presence of a larger dataset
, the generative machine should have less uncertainty about the underlying feature , corresponding to small or vanishing entropy .
the entropy can be derived by using the standard thermodynamic formula @xmath47 . under the bethe approximation ,
@xmath48 is evaluated as summing up contributions from single feature nodes and example nodes : @xmath49 , where single feature node contribution is expressed as @xmath50+\ln\left(1+\prod_{a\in\partial i}\mathcal{g}_{a\rightarrow i}\right)\\ -\left[\sum_{a\in\partial i } \mathcal{h}_{a\rightarrow i}(+1)+\prod_{a\in\partial i}\mathcal{g}_{a\rightarrow i}\sum_{a\in\partial i}\mathcal{h}_{a\rightarrow i}(-1)\right]/\left(1+\prod_{a\in\partial i}\mathcal{g}_{a\rightarrow i}\right ) , \end{split}\ ] ] and single example contribution reads @xmath51 where we define @xmath52 .
one iteration of the smp equation ( eq . ( [ bp0 ] ) ) requires the time complexity of the order @xmath53 and memory of the order @xmath53 .
the smp equation can be further simplified by reducing computational complexity .
the final equation in physics is called thouless - anderson - palmer ( tap ) equation @xcite , and in information theory is named approximate message passing ( amp ) equation @xcite .
one strategy is to use large-@xmath3 limit .
we first get the cavity bias in this limit as @xmath54 . then by applying the same large-@xmath3 expansion ,
we obtain @xmath55 .
therefore , we get the first amp equation as follows : @xmath56 where @xmath57
. then we define the local field @xmath58 , and note that @xmath59 , we can obtain an approximate @xmath60 in the large-@xmath3 expansion , @xmath61 the last term in the expression of @xmath60 serves as an onsager reaction term in a standard tap equation .
finally we arrive at the second amp equation : @xmath62 now we have only @xmath63 equations to solve rather than @xmath64 equations in the smp equation ( eq . ( [ bp0 ] ) ) . to make amp equations converge in a parallel iteration , the time indexes for the variables are important @xcite .
here we write down the closed form of amp equation with correct time indexes : [ amprbm ] @xmath65 where @xmath66 denotes the time index for iteration .
these time indexes just follow the temporal order when we derive the amp equation from the smp equation .
next , we give a statistical analysis of the smp equation .
we first define the cavity field @xmath67 . under the replica symmetric assumption
, @xmath68 follows a gaussian distribution with mean zero and variance @xmath69 in the large-@xmath3 limit .
we define @xmath70 .
similarly , @xmath71 also follows a gaussian distribution with mean zero but variance @xmath72 .
therefore , we arrive at the following thermodynamic equation : [ rbmrm ] @xmath73 where @xmath74 .
one can expect that when @xmath75 is small , only one solution of @xmath76 exists for the above equation , however , at some critical @xmath77 , there is a nontrivial solution of @xmath78 , which signals the fixed point of smp or amp starts to contain information about the underlying true feature vector .
@xmath77 can be determined by expanding the above equation around @xmath76 .
the expansion leads to @xmath79 , which implies that when @xmath80 , @xmath76 is the stable solution of the thermodynamic equation , but as long as @xmath81 , the @xmath76 is not the stable solution any more .
however , as we compare this solution with smp result on single instances , @xmath78 solution does not match the numerical simulation very well .
to explain this , a replica computation is required .
now , we perform a replica computation of the free energy function . instead of calculating a disorder average of @xmath82 , the replica trick computes the disorder average of an integer power of @xmath16 , then the free energy density ( multiplied by @xmath83 ) can be obtained as @xcite @xmath84 where the limit @xmath85 should be taken first since we can apply the saddle - point analysis @xcite , and the disorder average is taken over all possible samplings ( data ) and the random realizations of true feature vector .
the explicit form of @xmath86 reads @xmath87 where @xmath88 indicates the replica index .
we leave the technical details to the appendix [ replica - rbm ] , and give the final result here .
the free energy function reads , @xmath89 and the associated saddle - point equations are expressed as [ rbmreplica ] @xmath90 we make some remarks about the above saddle - point equations .
@xmath91 indicates the typical value of the overlap between the true feature vector and the estimated one , while @xmath92 indicates the typical value of the overlap between two estimated feature vectors selected from the posterior probability ( eq . ( [ pobs ] ) ) .
according to the nishimori condition , @xmath93 , implying that the embedded true feature vector follows the same posterior distribution in bayesian inference .
we verify this point later in numerical solution of the saddle point equations .
assuming @xmath91 and @xmath92 are both small values close to zero , eq .
( [ rbmreplica ] ) in this limit implies that a critical @xmath94 , above which @xmath28 is not a stable solution any more . by expanding eq .
( [ rbmreplica ] ) around @xmath28 up to the second order @xmath95 , we find @xmath96 when @xmath75 approaches @xmath77 from above . note that by assuming @xmath28 in eq .
( [ rbmreplica ] ) , we obtain eq .
( [ rbmrm ] ) ; this is the reason why eq .
( [ rbmrm ] ) can predict the correct threshold for transition but could not describe the property of @xmath97 .
the entropy can be derived from the free energy , and the result ( appendix [ replica - rbm ] ) is given by @xmath98 when @xmath28 , @xmath99 , which coincides with that obtained with cavity method ( appendix [ srbm ] ) .
given @xmath28 , that is the data still do not contain information about the hidden feature , the entropy will become negative once @xmath100 .
this suggests that the entropy crisis can even occur within @xmath28 regime .
alternatively , at a fixed @xmath75 , the entropy crisis occurs at a temperature @xmath101 . setting @xmath102 ,
one obtains @xmath103 , which distinguishes two cases : @xmath104 for @xmath105 , the transition of @xmath91 from zero to non - zero value takes place after the entropy crisis ; @xmath106 for @xmath107 , the transition takes place before the crisis .
this has clear physical implications .
if the transition occurs before the entropy crisis , the location of transition identified by the replica - symmetric theory is correct .
however , the transition after the crisis is incorrect under the replica - symmetric assumption , because although the replica - symmetric solution is stable , the entropy is negative , violating the fact that for a system with discrete degrees of freedom , the entropy should be non - negative .
according to arguments in refs .
@xcite , there exists a discontinuous transition before the entropy crisis takes place , since the transition can not continuously emerge from a stable replica - symmetric solution . without further solving complex more - steps replica symmetric breaking equations
, we adopt an alternative explanation of this entropy crisis . under the current context , if the data size is large enough , the data would shrink the feature space to a sub - exponential regime where the number of candidate features is not exponential with @xmath3 any more . in this case
, we encounter the entropy crisis .
therefore , the entropy crisis separates an exponential regime from a sub - exponential regime .
the transition for @xmath91 occurs within the exponential regime if the feature strength is strong enough ( large @xmath11 ) .
instances of size @xmath108 .
( a ) entropy per neuron versus data density @xmath75 .
the lines are replica result , while the symbols are results obtained on single instances .
the error bars are smaller than the symbol size .
( b ) order parameters ( @xmath109 ) versus @xmath75 . @xmath93 as expected .
the numerical simulations on single instances approximately match the theoretical predictions of replica computation .
, title="fig : " ] .05 cm instances of size @xmath108 .
( a ) entropy per neuron versus data density @xmath75 .
the lines are replica result , while the symbols are results obtained on single instances .
the error bars are smaller than the symbol size .
( b ) order parameters ( @xmath109 ) versus @xmath75 . @xmath93 as expected .
the numerical simulations on single instances approximately match the theoretical predictions of replica computation .
, title="fig : " ] .05 cm we use the above mean field theory to analyze single realizations ( instances ) of the unsupervised learning model .
random samplings are first generated according to a rbm distribution @xmath6 @xcite , where the energy is rescaled by the system size and the inverse temperature , which tunes difficulty level of the learning task .
these random samplings then serve as the quenched disorder specifying the interaction between the feature vector and the example constraint ( fig . [ rbm ] ) . finally , by initializing the message on each link of the factor graph ( fig . [ rbm ] )
, we run the smp equation ( eq .
( [ bp0 ] ) ) until it converges within a prefixed precision . from the fixed point , we compute the entropy of consistent feature vectors and the overlap between the inferred feature vector and the true one . we first compare results of the message passing algorithm with those obtained by replica computation . in fig .
[ rbm - mf ] ( a ) , we show the entropy density versus the data density @xmath75 .
the entropy characterizes how the number of candidate feature vectors compatible with the given data changes with the network size @xmath3 .
the replica result predicts an entropy crisis , i.e. , the entropy becomes negative at some data size , but the negativity of the entropy is not allowed in a system with discrete degrees of freedom ( @xmath110 ) .
this implies that , a discontinuous phase transition should be present before the crisis @xcite .
the results obtained by running smp coincide perfectly with the replica result .
the entropy density decreases more rapidly with @xmath75 at larger @xmath11 .
this is expected , because large @xmath11 indicates strong feature , thus to shrink the feature space to the same size , less data is required compared to the case of detecting weak feature ( small @xmath11 ) . in fig .
[ rbm - mf ] ( b ) , we show how order parameters change with @xmath75 .
the simulation results agree with the replica prediction , in spite of the deviation caused by finite size effects . for @xmath111 ,
a first continuous transition occurs at @xmath112 , which should be correct since the replica computation is stable and the entropy is positive there . at @xmath113 ,
the system starts to have information about the embedded feature , and therefore , the overlap @xmath91 starts to increase even for a finite - size system .
the asymptotic behavior of @xmath91 at a slightly larger @xmath75 ( @xmath114 ) is captured by @xmath115 , as already derived in the theory section ( sec .
[ replicasrbm ] ) .
as predicted by replica computation , at a larger value of @xmath116 , the entropy becomes negative .
equivalently , at this @xmath75 , the entropy vanishes at a critical temperature , and thus the equilibrium is dominated by a finite number of lowest energy states ( so - called condensation phenomenon @xcite ) .
but the smp is still stable , therefore a one - step replica symmetry breaking solution should grow discontinuously from the replica symmetric solution , and this second discontinuous glass transition is expected before or at the crisis @xmath75 to resolve the entropy crisis . intuitively , we expect that the increasing data will freeze the value of synapses , thus vanishing entropy indicates that the feature space develops isolated configurations : each configuration forms a single valley in the free energy profile , while the number of these valleys is not exponential any more ( but sub - exponential ) . to prove this picture
, one needs to go beyond the replica symmetric assumption . for @xmath117 , the continuous transition takes place after the entropy crisis , which is incorrect .
therefore , according to the above argument , a discontinuous transition should be expected at or before the crisis data size .
our simulation result also confirms the nishimori condition ( @xmath93 ) , that is , when the temperature used to generate data is equal to that used to infer the true feature , the true feature follows the posterior distribution as well . as a consequence ,
the overlap between a typical feature configuration and the embedded one is equal to the overlap between two typical configurations . , @xmath118 .
@xmath119 random instances are considered . ] secondly , the inference can also be carried out by using amp with less requirements of computer memory and time .
the result is compared with that obtained by smp , which is shown in fig .
[ amprbp0 ] .
we also study the effects of temperature deviation .
if the inference is carried out in a different temperature from that used to generate the data , is the performance degraded ?
we address this question by considering two different temperatures : one is slightly larger than the data temperature ( @xmath120 ) ; the other is slightly below the data temperature ( @xmath121 ) , where @xmath122 denotes the inverse data temperature . as shown in fig .
[ rbmnopt ] , when @xmath123 , the performance is optimal in the inferable regime , compared to other inference temperatures , as expected from the nishimori condition @xcite .
large fluctuations around the transition point may be caused by finite size effects .
finally , we explore the effect of network size keeping the identical feature strength ( fig .
[ netsize ] ) . at a given number of examples , larger network size yields better performance in terms of prediction overlap
. however , the performance seems to get saturated when @xmath124 at a relatively large @xmath12 in the current context .
using a larger network seems to make the unsupervised learning better , but further increasing the network size has a little effect on the performance . , @xmath125 .
@xmath119 random instances are considered . ] .
feature strength @xmath126 is kept constant ( @xmath127 ) . ] in fig .
[ rbmnopt ] , we have showed the inference performance with slightly different inference temperatures .
is it possible to infer the true temperatures used to generate the data itself ?
if we can learn the temperature parameters , we can know the typical properties of phase transitions intrinsic in the system .
this is possible by applying the bayesian rule once again .
the posterior probability of @xmath11 given the data @xmath13 is given by @xmath128 where we used the uniform prior probability @xmath129 for the hyper - parameters .
note that @xmath130 is the same partition function as in eq .
( [ pobs ] ) .
we maximize the posterior probability with respect to @xmath11 , and obtain the self - consistent equation @xmath11 should satisfy : @xmath131 the left hand side of the above equation is exactly the negative energy ( @xmath132 ) , which can be evaluated by smp equation ( eq . ( [ bp0 ] ) ) . under the bethe approximation , the energy per neuron @xmath133 can be computed by @xmath134 , where @xmath135 and @xmath136 are given respectively by [ energy ] @xmath137/\left(\beta+\beta\prod_{a\in\partial i}\mathcal{g}_{a\rightarrow i}\right),\\ \delta\epsilon_a&=\beta\xi_{a}^2+g_a\tanh(\beta g_a).\end{aligned}\ ] ] starting from some initial value of @xmath11 , one can iteratively update the value of @xmath11 until convergence within some precision .
after one updating , the messages in smp equation are also updated . to avoid numerical instability , we used the damping technique , i.e. , @xmath138 , where @xmath66 denotes the iteration step and @xmath139 $ ] is a damping factor .
it is not guaranteed that there exists unique maximum of the posterior ( eq .
( [ pobsem ] ) ) @xcite , but if necessary , one can choose the hyper - parameter corresponding to the global maximum of the posterior by running the smp from different initial conditions . in statistics , this iterative scheme is named expectation - maximization algorithm @xcite , where the message updates are called e - step , and the temperature update is called m - step . in physics ,
( [ betaem ] ) corresponds to the nishimori condition ( @xmath93 , see also appendix [ replica - rbm ] for derivation of the energy function ) .
this means that , in principle , the hyper - parameter can be learned by iteratively imposing the nishimori condition @xcite . on the nishimori condition ,
the state space of the model is simple @xcite , and thus smp yields informative information about the dominant feature vector .
we know that the temperature parameter is related to the feature strength embedded in the data .
once we learn the temperature , we are able to know how apparent the hidden feature is in a dataset , and determine the critical data size for unsupervised feature learning .
we first test our method in synthetic dataset as already studied in sec .
[ simusrbm ] , where the true value of hyper - parameter is known .
we then infer the embedded feature strength in the real dataset ( mnist handwritten digit dataset @xcite ) , where we do not have any knowledge about the true feature strength .
results are shown in fig .
[ rbm - em ] . for the synthetic data at @xmath140 , as the data size grows , inferred value of @xmath11 gets closer to the true value as expected ( fig .
[ rbm - em ] ( a ) ) .
as shown in the inset , the time ( iteration steps ) dependent inferred value first drops to a lower value , and then gradually approaches the true value . with a larger data size ( e.g. , @xmath141 ) , @xmath11 increases more rapidly after a sudden drop .
for the real dataset ( fig .
[ rbm - em ] ( b ) ) , we observe that the final fixed point of @xmath11 is quite large , implying that the feature strength in the handwritten digits is very strong ( @xmath142 ) . in synthetic ( rbm ) and real ( mnist , digits @xmath143 and @xmath2 ) dataset .
( a ) deviation of inferred @xmath11 from the true value decreases with the data size . in simulations ,
we consider @xmath144 instances of size @xmath145 , and use @xmath146 and initial value of @xmath147 .
two examplar trajectories of @xmath148 are shown in the inset .
( b ) examplar trajectories of @xmath148 are shown for the real dataset .
we use @xmath149 and @xmath150 .
the fixed point does not change when we use @xmath151 .
, title="fig : " ] .05 cm in synthetic ( rbm ) and real ( mnist , digits @xmath143 and @xmath2 ) dataset .
( a ) deviation of inferred @xmath11 from the true value decreases with the data size . in simulations , we consider @xmath144 instances of size @xmath145 , and use @xmath146 and initial value of @xmath147 .
two examplar trajectories of @xmath148 are shown in the inset .
( b ) examplar trajectories of @xmath148 are shown for the real dataset .
we use @xmath149 and @xmath150 .
the fixed point does not change when we use @xmath151 .
, title="fig : " ] .05 cm
it is interesting to show that one can also perform the same unsupervised learning task by using an associative memory ( hopfield ) model defined by @xmath152 where @xmath153 .
this posterior distribution of feature vectors given the input examples can be obtained by a small-@xmath11 expansion of eq .
( [ pobs ] ) @xcite .
this relationship implies that one can infer the feature vector of a rbm by an approximate hopfield model , and the feature vector is interpreted as the stored pattern in the hopfield model , encoding memory characteristics of the input data .
note that , @xmath154 are still governed by a rbm distribution , whereas , by applying the associative memory framework , we show many similar interesting properties of the unsupervised learning model . in analogous to the derivation of eq .
( [ bp0 ] ) , we have the smp corresponding to the posterior probability ( eq . ( [ hopf ] ) ) : @xmath155 where @xmath156 , @xmath157 in which @xmath158 .
details to derive eq .
( [ bp5 ] ) are given in appendix [ smp - hopf ] .
in the approximate hopfield model , the bethe free energy can be constructed similarly , i.e. , @xmath39 , where [ hopfz ] @xmath159+\ln2\cosh\hb h_i,\\ \ln z_a&=\frac{\hb}{2}\hg^2_{a}f_a-\frac{1}{2}\ln(1-\hb c_a),\end{aligned}\ ] ] where we define @xmath160 , @xmath161 , @xmath162 , and @xmath163
. similar to the case in rbm , the entropy for the approximate model can be evaluated as @xmath49 , where single feature node contribution reads @xmath164\\ +\ln\left(2\cosh(\hb
h_i)\right)-(\hb h_i+\hb h'_i)\tanh(\hb h_i ) , \end{split}\ ] ] and single example contribution reads @xmath165 where @xmath166 , @xmath167 , and @xmath168 .
we also derive amp equations for the hopfield model .
note that @xmath169 .
therefore @xmath170 , we thus derive the first amp equation for the hopfield model as @xmath171 where @xmath57 . from the definition of the local field @xmath60
, we have @xmath172 in the large-@xmath3 expansion .
finally , we derive the second amp equation : @xmath173 taking the time index into account , the amp equations can be summarized in a parallel update scheme as follows : [ amphopf ] @xmath174 in this section , we derive the thermodynamic equation .
similarly , the local field defined as @xmath175 follows a gaussian distribution with mean zero and variance @xmath176 .
the variance can be derived by noting that @xmath177 .
therefore , we have the following thermodynamic equation for the hopfield model : @xmath178 @xmath76 is a solution of eq .
( [ hopfrm ] ) , however , it is stable only when @xmath179 ^ 2 $ ] .
this threshold can be derived by expanding eq .
( [ hopfrm ] ) around @xmath76 to the first order .
interestingly , this equation matches the mean - field equation without ferromagnetic part ( related to retrieval phase ) derived in standard hopfield model @xcite .
gaussian assumption for messages does not generally hold , particularly for those messages related to the memorized patterns @xcite .
we also perform replica computation for the approximate hopfield model .
the free energy function is given in the appendix [ replica - hopf ] .
the associated saddle - point equations are given as follows : [ hopfreplica ] @xmath180 note that , the data is generated by the rbm , but the inference is carried out in an approximate hopfield model .
although we use the same temperature , the model mismatching leads to @xmath181 .
the threshold for the transition of @xmath92 can be determined by studying the linear stability around @xmath182 , and the result is @xmath183 ^ 2 $ ] , consistent with the cavity prediction .
when @xmath75 approaches @xmath184 from above , @xmath92 behaves like @xmath185 .
the transition for @xmath91 can only be determined numerically , since @xmath92 could not be assumed a small value , and it follows @xmath186 where @xmath187 .
the recursive equation for @xmath92 in the regime of @xmath28 is exactly the equation derived from the smp equation ( eq . ( [ hopfrm ] ) ) .
the entropy of the model can also be similarly computed , and reads as follows : @xmath188 the thermodynamic properties of the approximate hopfield model are shown in fig .
first , we show that the entropy crisis is absent in the hopfield model , although the inference is carried out by the smp equation of hopfield model .
this is quite interesting , because within the associative memory framework , the inference is improved smoothly and there does not exist condensation in the feature space . secondly , the replica computation predicts @xmath181 , as expected from the fact that by applying hopfield model approximation , the nishimori condition does not hold .
thirdly , the simulation results obtained by running smp equation agree with the theoretical predictions , in spite of observed fluctuations caused by finite size effects .
the asymptotic behavior of @xmath92 near @xmath189 can be analytically determined by small-@xmath92 expansion of the saddle - point equation ( eq . ( [ hopfreplica ] ) ) . as already derived in the theory section ( sec .
[ replicaahopf ] ) , @xmath185 when @xmath75 tends to @xmath189 from above .
the transition for @xmath91 can only be determined by numerically solving the saddle - point equation .
it seems that @xmath91 changes smoothly to a non - zero value at the same data size as that of the rbm .
due to the model mismatching , transition for @xmath92 takes place much more earlier than that for @xmath91 .
instances of size @xmath108 . the feature strength @xmath190 .
, title="fig : " ] .05 cm . @xmath119
random instances are considered . ] finally , the inference can also be carried out by using amp with less requirements of memory storage and computer time .
the result is compared with that obtained by smp , which is shown in fig .
[ amprbp1 ] .
we also study the effects of temperature deviation . as shown in fig .
[ hopfnopt ] , even when @xmath123 , it is not guaranteed that the performance is optimal in the inferable regime , compared to other inference temperatures . , @xmath125 .
@xmath119 random instances are considered . ]
in conclusion , we build a physical model of unsupervised learning from a finite number of examples in the framework of rbm . here , we consider binary features rather than real - valued ones ; this is because binary features are more robust and efficient in large - scale neuromorphic applications @xcite , yet it remains open to figure out an efficient algorithm .
we show that physics method can inspire an efficient ( fully - distributed ) message passing procedure not only to infer the hidden feature embedded in a noisy data , but also to estimate the entropy of candidate features .
distinct from conventional slow sampling - based methods , each example in this work is treated as a constraint on the factor graph , and the message passing carries out a direct bayesian inference of the hidden feature , which marks an important step implementing unsupervised learning in neural networks . in particular , the approximate message passing equation has low requirements of computer space and time in practical applications . we show that , the results obtained by the cavity method are consistent with the statistical analysis by replica theory .
the replica theory describes the thermodynamic properties of the unsupervised learning system .
it first predicts a discontinuous phase transition in a restricted boltzmann machine , signaled by the entropy crisis before the message passing equation loses its stability .
however , if the feature strength is strong enough , there exists another phase transition which is continuous , i.e. , the order parameter ( the overlap between the true feature vector and the inferred one ) smoothly changes from zero to non - zero value .
this continuous transition will be followed by an additional discontinuous transition at a larger data size .
interestingly , in an approximate hopfield model , the entropy crisis is absent , and the entropy decreases much more slowly towards zero .
therefore , there exists a continuous transition from impossible - to - infer to inferable regime . unlike the rbm ,
inference in the hopfield model does not satisfy the nishimori condition , and thus the statistics of metastable states would be very interesting , and its relationship with the dynamics of inference deserves further investigation .
our work not only derives in detail various kinds of message passing algorithms in a bayesian framework for practical applications , but also statistically characterizes the thermodynamic properties of the restricted boltzmann machine learning with binary synapses , and its connection with associative memory networks .
many interesting properties related to phase transitions are also revealed .
in addition , we derive an iterative equation to infer the unknown temperature in the data , providing a quantitative measure of how cold a dataset is .
this method corresponds to expectation - maximization algorithm in statistics @xcite , and in physics iteratively imposing nishimori condition @xcite .
therefore , our study forms a theoretical basis of unsupervised feature learning in a single simple rbm , and are expected to be helpful in constructing a deep architecture for hierarchical information processing , which is currently under way .
we first assume feature components on the factor graph are weakly correlated , then by using the cavity method @xcite , we define a cavity probability @xmath191 of @xmath23 on a modified factor graph with example node @xmath17 removed . due to the weak correlation assumption , @xmath191 satisfies a recursive equation ( namely belief propagation ( bp ) in computer science @xcite ) : [ bp00 ] @xmath192 where the symbol @xmath193 indicates a normalization constant , @xmath194 defines the neighbors of node @xmath1 except constraint @xmath17 , @xmath195 defines the neighbors of constraint @xmath24 except visible node @xmath1 , and the auxiliary quantity @xmath196 represents the contribution from constraint @xmath24 to visible node @xmath1 given the value of @xmath23 @xcite .
( [ bp00 ] ) has been similarly derived to understand rbm in a recent paper @xcite . here
, we exchange the role of the observed data @xmath22 and that of synaptic interaction ( feature vector here ) , and predict feature vector given the data .
therefore the data ( random samplings ) , rather than the synaptic interaction in the previous work @xcite , becomes a quenched disorder .
note that in eq .
( [ bp1 ] ) , the sum inside the hyperbolic cosine function with the @xmath1-dependent term excluded is a random variable following a normal distribution with mean @xmath71 and variance @xmath197 @xcite , where @xmath32 and @xmath198 . the cavity magnetization is defined as @xmath33 .
thus the intractable sum over all @xmath199 ( @xmath200 ) can be replaced by an integral over the normal distribution . using the magnetization representation @xcite , the bp equation ( eq . ( [ bp00 ] ) )
could be reduced to the simplified message passing equations ( see eq .
( [ bp0 ] ) in the main text ) . in physics , the contribution from a single feature node to the partition function ,
@xmath201 is obtained via cavity method as @xmath202 ; the contribution of a single data node reads @xmath203 , which can be further computed by applying the central - limit theorem as well .
this calculation is exact only when the underlying factor graph is a tree .
however , it is approximately correct when correlations among feature components are weak .
it needs to be compared with numerical simulations and replica computations .
first , we compute @xmath204 . by noting that @xmath205 , @xmath206 , and @xmath207 , we have @xmath208 where we have used the fact that @xmath209 and @xmath210 .
analogously , @xmath211 is simplified to be @xmath212 collecting the above results , we arrive at the final simplified entropy as @xmath99 .
for the approximate hopfield model , we similarly define the auxiliary quantity @xmath196 as @xmath213 ^ 2\right)\\ & = \frac{1}{\sqrt{1-\hb c_{b\rightarrow i}}}\exp\left(\frac{\hb f_{b\rightarrow i}}{2}\bigl(\frac{1}{n}+\hg_{b\rightarrow i}^{2}\bigr)\right)\exp\left(\frac{\hb\hg_{b\rightarrow i}\xi_i\sigma_i^{b}f_{b\rightarrow i}}{\sqrt{n}}\right ) , \end{split}\ ] ] where @xmath214 .
@xmath191 is the same as that in rbm .
using eq .
( [ bp0hopf ] ) , the cavity magnetization @xmath34 can thus be derived as eq .
( [ bp5 ] ) .
we first define @xmath215 , and @xmath216 . both @xmath217 and @xmath218 are random variables subject to the covariance structure : @xmath219 , @xmath220 , @xmath221 , @xmath222 , @xmath223 , @xmath224 , where we have dropped off the data index @xmath17 because of independence among data samples , and defined the overlap between true feature vector and the estimated one as @xmath225 , and the overlap between two estimated feature vectors as @xmath226 . under the replica symmetric assumption , @xmath227 and @xmath228 , after introducing the definition of @xmath229 ( and @xmath230 ) as a delta function , @xmath86 can be estimated as @xmath231\\ \times\exp\left[\alpha n\ln \left\{e^{-\beta^2/2}\int dy\int dt\cosh\beta t(\cosh\beta(qt+\sqrt{r - q^2}y))^n\right\}\right ] , \end{split}\ ] ] where we have written @xmath232 ( @xmath233 and @xmath234 are standard gaussian random variables ) .
finally , we arrived at the following free energy function : @xmath235 the saddle - point equation for the order parameters @xmath236 can be derived from @xmath237,@xmath238 , @xmath239 , and @xmath240 .
note that to derive the entropy formula , we used @xmath241 , where @xmath242 is the inverse temperature at which the data is generated , and @xmath11 the temperature at which the bayesian inference is carried out .
for the approximate hopfield model , we replace @xmath243 with @xmath244 in eq .
( [ zn ] ) .
the subsequent calculation proceeds similarly to the appendix [ replica - rbm ] .
analogously , we have @xmath245\\ \times\exp\left[\alpha n\ln \left\{e^{-\beta^2/2}\int dy\int dt\cosh\beta t\bigl(\frac{1}{\sqrt{1-\hb(1-r)}}e^{\frac{\hb(qt+\sqrt{r - q^2}y)^2}{2(1-\hb(1-r))}}\bigr)^n\right\}\right ] . \end{split}\ ] ] using the replica trick defined in eq .
( [ replica ] ) , we obtain the free energy function @xmath246 where we used the identity @xmath247
^ 2\cosh\beta t = e^{\hb/2}(\hb q^2+r)$ ] .
the saddle - point equations can be derived similarly .
the smp for rbm can be easily generalized to take into account external fields of visible neurons and the hidden neuron . here , for simplicity , we consider only the case of hidden neuron with external field .
the external field has binary values as well , defined by @xmath248 ( @xmath249 ) .
the only modification to the factor graph in fig .
[ rbm ] is to add one additional variable node named by @xmath250 for the unknown external field .
the additional node @xmath250 is connected to all data nodes . following the similar procedure as in appendix [ smp - rbm ] , we obtain the following four kinds of messages : once @xmath252 , the original smp for the rbm in the main text is recovered .
to derive smp for the case of visible neurons with external fields , an additional central limit theorem applies to the interacting external fields ( @xmath253 ) , which leads to introducing a joint cavity probability @xmath254 as well as @xmath255 where @xmath256 .
i am very grateful to taro toyoizumi , lukasz kusmierz , alireza goudarzi and roberto legaspi for attending a series of lectures about this work and their useful feedback .
i thank lukasz kusmierz for a careful reading of the manuscript and his useful feedback .
this work was supported by the program for brain mapping by integrated neurotechnologies for disease studies ( brain / minds ) from japan agency for medical research and development , amed .
|
revealing hidden features in unlabeled data is called unsupervised feature learning , which plays an important role in pretraining a deep neural network . here
we provide a statistical mechanics analysis of the unsupervised learning in a restricted boltzmann machine with binary synapses .
a message passing equation to infer the hidden feature is derived , and furthermore , variants of this equation are analyzed .
a statistical analysis by replica theory describes the thermodynamic properties of the model .
our analysis confirms an entropy crisis preceding the non - convergence of the message passing equation , suggesting a discontinuous phase transition as a key characteristic of the restricted boltzmann machine .
continuous phase transition is also confirmed depending on the embedded feature strength in the data .
the mean - field result under the replica symmetric assumption agrees with that obtained by running message passing algorithms on single instances of finite sizes .
interestingly , in an approximate hopfield model , the entropy crisis is absent , and a continuous phase transition is observed instead .
we also develop an iterative equation to infer the hyper - parameter ( temperature ) hidden in the data , which in physics corresponds to iteratively imposing nishimori condition .
our study provides insights towards understanding the thermodynamic properties of the restricted boltzmann machine learning , and moreover important theoretical basis to build simplified deep networks .
|
stoichiometrically stable complexes are formed by proteins that _ physically _ interact to achieve biological functions within the cell .
these complexes interact with individual proteins or other complexes to form functional modules and pathways that drive the cellular machinery .
therefore , a faithful reconstruction of the entire set of complexes is essential to not only understand complex formations , but also the higher level organization of the cell .
recent advances in high - throughput techniques have enabled to catalogue enormous amounts of physical interaction data particularly in organisms such as _ saccharomyces cerevisiae _ ( budding yeast ) . typically these interactions are arranged in the form of a protein interaction network ( or ppi network ) and mined for complexes using computational techniques . from a topological perspective
, these complexes are typically interpreted as regions in the network where proteins are densely connected to each other than to the rest of the network @xcite .
accordingly , several computational methods have been proposed that depend primarily on the _ topologies _ of ppi networks , and model _ dense _ regions as complexes ; for a survey , see .
for example , mcl @xcite simulates a series of random walks ( called a _ flow _ ) , the principle being that when the walks reach a dense region , with high probability , they will continue to remain in that region . by repeated iterations of inflation ( thickness ) and expansion ( spread ) of the flow , mcl identifies complexes .
mcode @xcite , on the other hand , identifies seed " proteins in the network using clustering coefficients and greedily expands in the neighborhood of these seeds to build complexes .
cmc @xcite first generates maximal cliques from the network , and then merges highly interconnected cliques to assemble complexes .
haco @xcite performs agglomerative clustering by generating small clusters and hierarchically merging them into complexes .
haco improves upon the traditional hierarchical agglomerative clustering ( hac ) by allowing for overlaps among the generated complexes .
finally , mcl - caw @xcite produces initial clusters using mcl and then refines these clusters by incorporating core - attachment structure to generate complexes .
we performed comprehensive evaluations @xcite of these methods , particularly mcl , mcl - caw , cmc and haco , on a variety of yeast ppi networks ranging from raw to highly - filtered and under varying levels of natural as well as artificial noise , and found that these methods failed to detect many known complexes catalogued in the mips @xcite database .
for example , mcl missed 65 out of the 123 mips complexes present in the consolidated@xmath5 network from @xcite .
even the union " of these methods missed 52 out of the 123 complexes .
since the goal here is to study genome - wide compositions of complexes ( the complexosome " ) , failure to detect even the known subset of complexes reflects severe limitations in current methods . in order to understand the characteristics of these missed complexes , we superimposed " yeast complexes taken from the mips benchmark @xcite onto the high - confidence consolidated@xmath5 yeast ppi network @xcite ( # proteins : 1622 , # interactions : 9704 , average node degree : 11.187 ) .
this superimposition " involves identifying the proteins of the benchmark complex in the ppi network , and extracting out the subnetwork induced by those proteins .
figure 1 in supplementary materials shows this superimposition " visualized using _ cytoscape
_ @xcite .
the immediate observation , which is of course typical to most ppi networks , was that the network comprised of one main large component and multiple _ disjoint _ smaller components of sizes 2 to 50 . out of the 123 mips complexes containing at least four proteins in the network ,
89 were completely embedded in the main component , and the remaining 34 were scattered " among more than one components .
when we ran mcl on this network , it was able to recover only 58 of these 123 complexes .
of the 65 undetected complexes , 27 complexes were the ones that were scattered " , and 34 complexes , though intact , had very low interaction densities ( @xmath6 ) in the network .
in fact , some of these complexes lacked internal connectivities to an extent that it was impossible for _ any _ algorithm to assemble back these disconnected pieces into whole complexes solely based on topological information .
for example , the mips complex 510.190.110 ( ccr4 complex ) had seven proteins in the network scattered among four disjoint components .
( shown within ellipses in figure 1 in supplementary materials ) .
this complex remained disconnected with a low density of 0.1905 , and naturally went undetected by all the four algorithms ( a few more examples are available from supplementary materials ) .
further , most mips complexes being small ( sizes @xmath7 10 - 15 ) , lacking in just a few proteins or interactions easily rendered many complexes disconnected or with low interaction densities , resulting in them going undetected .
all these findings revealed that a potentially strong correlation existed between the network constitution " of a complex ( the number of member proteins in the network and their connectivities ) and the possibility of it being detected using existing algorithms .
this work is strongly motivated by the limitations in existing complex detection methods in successfully detecting complexes , and the aforesaid revelations on the topologies of these undetected complexes within ppi networks .
the purpose of our work therefore is two - fold : ( i ) to _ characterize _ these undetected complexes , that is , to quantitatively measure their network constitution " ; and ( ii ) to propose a novel algorithm employing functional interactions to enhance the derivability " of sparse complexes , which in turn empowers existing methods in detecting these complexes satisfactorily .
we represent our ppi network as @xmath8 , where @xmath9 is the set of proteins and @xmath10 is the set of interactions between the proteins .
each interaction @xmath11 is assigned a weight @xmath12 that reflects the confidence of the interaction , which is usually determined using an affinity weighting scheme ( the weight it is set to 1 if no scheme is used ) . for any @xmath13
, @xmath14 refers to the set of neighbors of @xmath15 .
let @xmath16 be the set of benchmark complexes .
we propose the term _ sparse complexes _ for the undetected complexes and very broadly " define them as follows : [ sp1 ] sparse complexes : given a ppi network @xmath17 and a set of benchmark complexes @xmath18 known to be embedded in @xmath17 , the subset @xmath19 of complexes that can not be satisfactorily detected from @xmath17 by existing methods are called sparse complexes .
we next propose _ indices _ that measure the derivability " of a benchmark complex from a given ppi network .
these indices capture whether or not a benchmark complex is derivable from a given ppi network , and if so , to what extent .
we propose two kinds of indices here .
the first kind defines definitive criteria to categorize a given benchmark complex as derivable or not from the ppi network , and provides _ derivability bounds _ on the number of such complexes in the network .
the second kind does not strictly categorize the benchmark complex as derivable or not , but instead assigns a _
derivability score _ to the complex . to begin with , a naive yet natural way to categorize a benchmark complex as _ derivable _ from a ppi network
is if it satisfies two criteria : ( i ) it has sufficient number of proteins in the network ; and ( ii ) it is connected within the network .
we consider a benchmark complex @xmath20 to be _
@xmath21-protein - derivable _ from @xmath17 if at least @xmath22 of its member proteins are present in @xmath17 .
we consider a @xmath21-protein - derivable complex to be _
@xmath21-network - derivable _ from @xmath17 if these member proteins form a connected subnetwork within @xmath17 .
@xmath21-protein - derivable complex : a benchmark complex @xmath20 is @xmath21-protein - derivable from network @xmath8 if @xmath23 , for some @xmath22 .
the set of @xmath21-protein - derivable complexes in @xmath17 is represented by @xmath24 , and the _ @xmath21-protein - derivability index _ of @xmath17 is @xmath25 .
@xmath21-network - derivable complex : a benchmark complex @xmath20 is @xmath21-network - derivable from @xmath8 if @xmath23 for some @xmath22 , and @xmath26 forms a connected subnetwork in @xmath17 .
the set of @xmath21-network - derivable complexes in @xmath17 is represented by @xmath27 , and the _ @xmath21-network - derivability index _ of @xmath17 is @xmath28 . from our extensive experiments ( details omitted due to lack of space ) , we found that two factors strongly contributed to the derivability " of a given complex from the network - the presence of a significant fraction of complex proteins within the same connected component , and the density of the complex relative to its local neighborhood .
based on these two factors we next define indices that assign _ derivability scores _ to each benchmark complex to reflect the confidence or extent to which the complex is derivable from the network . _
component score _
@xmath29 : in the network @xmath17 , let any @xmath21-protein - derivable complex @xmath30 be decomposed into several connected components , @xmath31 , ordered in non - increasing order of size .
we define @xmath29 as the fraction of proteins within the maximal component @xmath32 among all _ non - isolated proteins _ in @xmath30 : @xmath33 where @xmath34 .
_ edge score _
@xmath35 : we define @xmath35 as the ratio of the weight of interactions within @xmath30 to the total weight of interactions within @xmath30 and its immediate neighborhood in @xmath17 : @xmath36 the denominator is the weight of interactions in the subnetwork of @xmath17 induced by the member proteins of @xmath30 and their direct neighbors , given by : @xmath37 and @xmath38 .
note that the edge score is different from the absolute _ edge density _ of @xmath30 , which is defined as : @xmath39 .
we define the _ component - edge score _ @xmath40 as the product of the component and edge scores of @xmath30 : @xmath41
@xmath21-ce - derivable complex : given a threshold @xmath42 , a @xmath21-protein - derivable complex @xmath30 is @xmath21-@xmath43-derivable if @xmath44 .
therefore , the set of @xmath21-@xmath43-derivable complexes in @xmath17 is given by : @xmath45 , and the _ @xmath21-ce - derivability index _ of @xmath17 is @xmath46 . for any @xmath22 , by definition @xmath47 . given a threshold @xmath42 , the relationships between @xmath24 and @xmath27 with @xmath48 are as follows .
when @xmath49 , all @xmath21-@xmath43-derivable complexes are also @xmath21-protein - derivable , but because they may not be connected we can say , @xmath50 .
when @xmath51 , all @xmath21-@xmath43-derivable complexes are connected complexes that are disjoint , therefore @xmath52 . intuitively , @xmath53 can be varied in the entire range @xmath54 $ ] to include the hardest " complexes to detect ( without any internal connectivities ) to only the easiest " complexes to detect ( disjoint connected complexes ) .
these hardest " complexes to detect can form holes " in the network by having zero interactions among their member proteins but having interactions with their immediate neighbors ( see supplementary materials for a visual representation of these complex sets ) .
we now validate the derivability scores ( @xmath55 , @xmath56 , @xmath43 scores and absolute edge density ) of benchmark complexes with respect to the ppi network against the accuracies with which these complexes are actually derived using existing methods .
this will reveal how effective each of these indices are in capturing actual complex derivability using existing methods .
l || c || c | c | c + & + & + & & + method & edge density & @xmath43 & @xmath55 & @xmath56 + mcl & _ 0.101 _ & _ 0.719 _ & 0.511 & 0.518 + mcl - caw & _ 0.196 _ & _ 0.785 _ & 0.492 & 0.628 + cmc & _ 0.174 _ & _ 0.649 _ & 0.471 & 0.477 + haco & _ 0.159 _ & _ 0.786 _ & 0.472 & 0.608 + [ table : derivability_indices_correlation_consol ] we use two ppi networks for this validation , the consolidated@xmath5 network ( a weighted network ) from @xcite , and the ` filtered yeast interaction ' ( fyi ) network ( a literature - validated but unweighted network ) from @xcite .
we use complexes from the mips and wodak catalogues as our benchmark complexes .
table [ table : derivability_indices_correlation_consol ] shows the pearson correlation values between the derivability scores and the _ jaccard _ accuracies obtained from four complex detection methods , mcl , mcl - caw , cmc and haco ( the complete set of results are available from the supplementary materials ) .
the results show the @xmath43-scores and jaccard accuracies are _ strongly correlated _
( pearson : 0.719 using mcl ) , better than the correlation between absolute edge densities and jaccard accuracies ( pearson : 101 using mcl ) .
this means our proposed @xmath43-score is a _ stronger _ indicator of actual complex derivability compared to the traditionally adopted indicators like edge density .
( even the individual scores , @xmath55 and @xmath56 show reasonable correlation with jaccard accuracies .
also , there are a few other indices like global and local modularity @xcite , but these do not capture the notion of proteins being part of the same connected component , and they perform similar to our edge - score @xmath56 )
. we can now employ our proposed @xmath43-score to give a more quantitative definition for sparse complexes .
[ sp2 ] sparse complexes : given a ppi network @xmath17 , a benchmark complex @xmath30 and a threshold @xmath42 , the complex @xmath30 is called sparse with respect to @xmath17 if @xmath57 . notice how the two definitions [ sp1 ] and [ sp2 ] can be linked " using our @xmath43-score and threshold @xmath53 , which offer a quantitative value to the derivability of complexes .
if this value is less than a certain threshold , the complex is highly likely to go undetected from existing methods and therefore it is _ sparse _ , else it is highly likely to be detected and therefore it is _ dense_. in general , for the benchmark complexes @xmath18 , the set of sparse complexes is given by @xmath58 , and its complementary set @xmath59 forms the dense complexes .
the threshold @xmath53 defines this boundary " between the sparse and dense benchmark complexes in the network .
since we do not know at which value of @xmath53 existing methods operate , we propose an approach that packs " higher number of dense complexes for all values of @xmath60 $ ] or at least for the larger values of @xmath53 .
we noted in section [ sec_insights ] that existing methods are severely constrained by gaps " in crucial topological information required to ensure the two required criteria for complex derivability namely , component - based connectivity and relative edge density .
in fact , any new method based solely on ppi networks would also face these constraints . due to these reasons ,
a natural approach to aid existing methods or devise new methods would be to first fill these topological gaps " in existing ppi networks .
even though this seems like a simple enough solution to pursue , we are severely lacking in the interaction data required to fill these gaps .
current estimates on yeast @xcite , put the verified fraction of the physical interactome to @xmath470% , which means we are still lacking in @xmath430% reliable interaction data , mainly due to limitations in existing experimental and computational techniques . consequently , a novel solution is to look beyond physical interactions to fill these topological gaps . in our work
, we propose to use _ functional interactions _ for this purpose , specifically aimed at improving complex prediction .
functional interactions or associations are logical interactions among proteins that share similar functions @xcite .
these interactions can be inferred among proteins participating in the same multi - protein assemblies ( complexes , functional modules and pathways ) , or annotated to similar biological functions and processes , or encoded by genes maintained and regulated together or genes having the same ` phylogenetic profile ' ( present or absent together across several genomes ) , etc .
therefore , these interactions encode " information beyond just direct physical interactions .
in fact many of the computational methods developed to predict protein interactions mainly manage to predict functional interactions .
functional interactions can be considered more general " or a superset " of direct physical interactions : two proteins involved in a stable physical interaction are functionally related , but two proteins involved in a functional interaction may not necessarily interact physically .
this means functional interactions have a potential to effectively _ complement _ physical interactions .
we capitalize on this complementarity by non - randomly adding functional interactions to ensure the two required criteria : ( i ) some functional interactions may be direct physical interactions missing in the physical datasets - these are directly useful to pull - in " disconnected proteins ; and ( ii ) even if some functional interactions do not correspond to direct physical interactions , if they fall within the same complex , they can artificially " increase the density of that complex . here , we propose a post - processing based algorithm sparc to empower existing methods in detecting sparse complexes by using functional interactions .
sparc works as follows .
let @xmath61 be the ppi network and @xmath62 be the functional network .
_ step 1 : _ the input to the algorithm is the set of physical clusters @xmath63 from network @xmath64 generated using an existing method .
it then calculates the @xmath43-score @xmath65 for each cluster @xmath66 .
all clusters with @xmath43-scores above a threshold @xmath67 , that is , @xmath68 , are output as predicted complexes , while the remaining are reserved for further processing .
_ step 2 : _ we then add - in the interactions of @xmath69 to @xmath64 to produce a larger network @xmath70 , where @xmath71 and @xmath72 .
_ step 3 ( iterative ) : _ for each reserved cluster @xmath73 , the @xmath43-score is recalculated with respect to @xmath74 .
if for the cluster @xmath73 , the @xmath43-score improves beyond @xmath67 , that is , @xmath75 , it is output as a predicted complex .
if not , we explore in the neighborhood of @xmath73 to include proteins that can potentially improve @xmath76 .
we consider the set of direct neighbors @xmath77 , and sort them in non - increasing order of their interaction weights to @xmath73 .
we then repeatedly consider a protein @xmath78 in that order such that @xmath79 and add it to @xmath73 , till the @xmath43-score can not be improved any further .
if the improved @xmath43-score manages to cross @xmath67 , we output the cluster @xmath73 as a predicted complex .
the key idea behind sparc is as follows .
many complexes have low @xmath43-scores in the ppi network . if adding functional interactions can either increase their internal connectivities or
pull in " the disconnected proteins , we can increase the @xmath43-scores of these complexes .
however , blindly adding functional interactions can result in many false positive predictions .
therefore , here we selectively utilize functional interactions only to improve the @xmath43-scores of clusters predicted out of the physical network . those clusters that show
the improvement correspond to real complexes .
we gathered physical interactions from _ saccharomyces cerevisiae _ ( budding yeast ) inferred from the following yeast two - hybrid and affinity purification experiments , deposited in biogrid @xcite : uetz ( 2000 ) , ito ( 2001 ) , gavin ( 2002 , 2006 ) , krogan ( 2006 ) , collins ( 2007 ) and yu ( 2008 ) , to build the protein interaction network , which we call the _ physical network _ @xmath80 .
the interactions of @xmath80 are not scored .
next , high - confidence functional interactions from yeast were gathered from the string database @xcite to build the _ functional network _
these functional interactions showed confidence scores @xmath82 0.90 in at least two of the following evidences : gene neighborhood , co - occurrence , co - expression and text mining ( these scores are available from string ) .
we combined the two networks to generate a larger network which we call the _ augmented _ physical+functional network @xmath83 .
table [ table : ppi_network_properties ] shows some properties of these networks .
the overlaps between @xmath80 and @xmath81 are as follows : @xmath84 and @xmath85 .
l | c | c | c & # proteins & # interactions & avg node degree + physical @xmath86 & 4113 & 26518 & 12.89 + functional @xmath87 & 3960 & 18683 & 10.12 + augmented @xmath88 & 5145 & 43905 & 17.07 + [ table : ppi_network_properties ] the presence of _ noise _ ( false positives ) is a severe limiting factor in publicly available interaction datasets in spite of gathering only high - confidence datasets .
therefore , we further _ filtered _ these datasets , which involves assigning each interaction a confidence score ( between 0 and 1 ) that reflects its reliability , and discarding interactions with low scores ( @xmath89 ) . here , we ( re)scored the networks using three scoring schemes , two of which were based on network topology namely , _
fs - weight _ devised by @xcite and _ iterative - cd _ devised by @xcite , while the third was based on evidences from gene ontology ( go ) @xcite , called _ tcss _ devised by @xcite . the _ benchmark _ or reference set of complexes was assembled from two sources : 313 complexes of mips @xcite and 408 complexes of the wodak lab cyc2008 catalogue @xcite .
the properties of these benchmark sets are shown in table [ table : size_distribution_benchmarks ] .
for the evaluation , we considered only the 4-protein - derivable complexes out of these sets .
this is because it is typically difficult to predict very small complexes ( size @xmath90 ) with high accuracy by using primarily topological information .
.properties of hand - curated ( benchmark ) yeast complexes from the mips and wodak cyc2008 catalogues .
[ cols="^,^,^,^,^,^ " , ] [ table : size_distribution_benchmarks ] the go annotations for yeast proteins were downloaded from the _ saccharomyces _ genome database ( sgd ) @xcite , which include the annotations ( not considering the inferred from electronic annotations or iea ) for three ontologies - cellular component ( cc ) , biological process ( bp ) and molecular function ( mf ) .
these annotations were used as evidences in the tcss scheme @xcite .
we excluded the branch corresponding to the go term ` macromolecular complex ' ( go:0032991 ) to avoid any bias coming from the go complexes .
we used four complex detecting algorithms mentioned previously , mcl @xcite , cmc @xcite , haco @xcite and mcl - caw @xcite .
some of their properties and the preset parameter values are summarized in table [ table : evaluation_algorithms ] .
these methods are different from one another in the algorithmic techniques employed , and therefore form a good mix of methods for our evaluation .
c || c | c | c | c & * mcl * & * mcl - caw * & * cmc * & * haco * + & flow & core - attach & maximal & hier agglo + & simulation & refinement & clique & cluster with + & & over mcl & merging & overlaps + * parameters * & @xmath91 & @xmath91 , @xmath92 , @xmath93 & merge @xmath94 , & upgma + * ( preset values ) * & ( 2.5 ) & ( 2.5 , 1.5 , 0.75 ) & overlap @xmath95 , & cutoff + & & & min clust size & ( 0.2 ) + & & & ( 0.5 , 0.4 , 4 ) & + [ table : evaluation_algorithms ] usually , recall @xmath96 ( coverage ) and precision @xmath97 ( sensitivity ) are used to evaluate the performance of methods against benchmark complexes . here
, we use previously reported @xcite definitions for these measures .
let @xmath98 and @xmath99 be the sets of benchmark and predicted complexes , respectively .
we use the jaccard coefficient @xmath100 to quantify the overlap between a @xmath30 and a @xmath73 : @xmath101 .
we consider @xmath30 to be covered by @xmath73 , if @xmath102 _ overlap threshold _ @xmath103 . in our experiments , we set the threshold @xmath104 , which requires @xmath105 .
for example , if @xmath106 , then the overlap between @xmath30 and @xmath73 should be at least 6 . based on this
the recall @xmath96 is given by : @xmath107 here , @xmath108 gives the number of _ derived benchmarks_. and the precision @xmath97 is given by : @xmath109 here , @xmath110 gives the number of _ matched predictions_. to begin with , we measured the number of derivable benchmark complexes from the physical @xmath86 , functional @xmath87 , augmented @xmath88 networks and their scored versions , @xmath111 , @xmath112 and @xmath113 , using our proposed derivability indices .
table [ table : derivability_indices1 ] shows the number of protein - derivable and network - derivable benchmark complexes from these networks .
the findings can be summarized as follows : ( a ) the network - derivable complexes were significantly fewer than the protein - derivable complexes further supporting the claim ( section [ sec_intro ] ) that many benchmark complexes remained disconnected within the networks .
( b ) the number of protein - derivable and network - derivable complexes were higher for the @xmath83 network than the individual @xmath80 and @xmath81 networks .
the significance of this increase was gauged against a random network @xmath114 built using the same set of proteins and the average node degree in @xmath81 .
the @xmath115 network showed fewer network - derivable complexes compared to @xmath83 .
this indicated that @xmath81 added more interactions to complexed " regions in @xmath80 compared to what the @xmath114 network added .
( c ) the number of protein - derivable and network - derivable complexes in the scored networks , @xmath111 , @xmath112 and @xmath113 , were fewer than the @xmath83 network .
this is not a concern because filtering usually discards interaction data leading to smaller networks .
( d ) even though protein - derivable complexes in the scored networks were fewer than the @xmath83 network , the corresponding decrease in network - derivable complexes was relatively marginal .
this indicated that the scoring schemes retained most interactions among complexed proteins , and discarded mainly the noisy ones .
l || c | c + & # protein- & # network- + network & derivable & derivable + physical @xmath80 & 155 & 59 + functional @xmath81 & 153 & 28 + p+random & 164 & 61 + p+f & 164 & 68 + icd(p+f ) & 122 & 64 + fsw(p+f ) & 119 & 64 + tcss(p+f ) & 158 & 68 + [ table : derivability_indices1 ] c || c | c | c | c | c | c + & + & p & f & p+f & icd(p+f ) & fsw(p+f ) & tcss(p+f ) + 0.00 & 155 & 153 & 164 & 152 & 119 & 162 + 0.10 & 153 & 151 & 162 & 148 & 116 & 160 + 0.20 & 149 & 136 & 158 & 145 & 113 & 157 + 0.30 & 140 & 108 & 149 & 142 & 110 & 154 + 0.40 & 129 & 81 & 135 & 137 & 108 & 148 + 0.50 & 101 & 54 & 102 & 112 & 101 & 126 + 0.60 & 81 & 21 & 70 & 93 & 87 & 101 + 0.70 & 62 & 9 & 55 & 71 & 69 & 86 + 0.80 & 39 & 0 & 34 & 44 & 42 & 59 + 0.90 & 19 & 0 & 14 & 21 & 21 & 35 + 1.00 & 6 & 0 & 3 & 11 & 10 & 18 + [ table : ce_derivability_mips ] next , table [ table : ce_derivability_mips ] shows the number of @xmath43-derivable benchmark complexes from these networks for all threshold values @xmath60 $ ] .
this table does a more fine - scale dissection of the improvement shown before . for lower values of @xmath53 ,
the number of @xmath43-derivable complexes was higher for @xmath83 compared to @xmath80 .
but , for higher values of @xmath53 , the number was lower compared to @xmath80 .
similarly , for lower values of @xmath53 , the number of @xmath43-derivable complexes was higher for @xmath83 compared to the three scored networks .
but , for higher values of @xmath53 , the three scored networks showed considerably higher @xmath43-derivable complexes than both the @xmath80 and @xmath83 networks .
these findings indicate that noise had a sizable impact on the @xmath43-scores of complexes : the improvement obtained by adding functional interactions was completely canceled out by noise , leading to lower performance of the @xmath83 network .
but , affinity scoring ( filtering ) considerably alleviated this impact of noise , thereby improving the @xmath43-derivability of the networks .
table [ table : methods1_mips ] shows the performance of the four methods mcl , mcl - caw , cmc and haco on the raw physical and scored physical networks ( we do not show the results on @xmath81 because functional interactions are only used to improve the physical clusters , and not for complex detection by themselves - many of the functional clusters do not correspond to physical complexes ) .
it shows that scoring helped to reconstruct significantly more complexes and with better accuracies ( also noted in ) .
next , table [ table : methods2_mips ] shows the performance after refining the physical clusters using functional interactions by applying sparc ( @xmath116 ) .
it shows that post - processing using raw functional interactions @xmath88 led to many noisy clusters , resulting in lower precision and recall .
but , using filtered ( scored ) functional interactions helped to reconstruct significantly more complexes out of the physical clusters . one interesting point to note
is that the compositions of predicted complexes vary based on the scoring scheme used ( also noted in @xcite ) , and therefore we had to construct a _ consensus set _ of complexes from the three scoring schemes for each of the methods . to do this , we employed a three - way agreement scheme based on jaccard overlaps .
let @xmath117 be a complex triplet , each complex predicted from a different scored network by the same method .
if at least two complex pairs from @xmath118 achieve significant jaccard overlaps @xmath119 , then the proteins of @xmath120 , @xmath121 and @xmath122 are merged together into a single consensus complex @xmath123 . only the proteins originating from at least two complexes
are included in @xmath123 .
we noticed that this consensus operation further improves the accuracies of the predictions leading to better reconstruction of benchmark complexes .
l || l || c | c || c | c || c | c & & + & network & # predicted & # matched & # derivable & # derived & @xmath97 & @xmath96 + & physical p & 294 & 29 & 155 & 38 & 0.098 & 0.245 + mcl & fsw(p ) & 156 & 31 & 102 & 40 & 0.198 & 0.333 + & icd(p ) & 167 & 32 & 109 & 40 & 0.191 & 0.293 + & tcss(p ) & 172 & 39 & 112 & 41 & 0.226 & 0.366 + & physical p & 297 & 39 & 155 & 49 & 0.131 & 0.316 + mcl & fsw(p ) & 149 & 38 & 102 & 51 & 0.255 & 0.392 + -caw & icd(p ) & 162 & 41 & 109 & 52 & 0.253 & 0.376 + & tcss(p ) & 168 & 41 & 112 & 54 & 0.244 & 0.366 + & physical p & 156 & 41 & 155 & 56 & 0.263 & 0.361 +
cmc & fsw(p ) & 144 & 31 & 102 & 59 & 0.215 & 0.313 + & icd(p ) & 165 & 43 & 109 & 60 & 0.260 & 0.394 + & tcss(p ) & 128 & 39 & 112 & 59 & 0.304 & 0.357 + & physical p & 414 & 34 & 155 & 41 & 0.082 & 0.264 + haco & fsw(p ) & 221 & 32 & 102 & 44 & 0.144 & 0.313 + & icd(p ) & 248 & 37 & 109 & 45 & 0.149 & 0.339 + & tcss(p ) & 253 & 46 & 112 & 45 & 0.181 & 0.410 + [ table : methods1_mips ] l || l || c | c | c || c | c || c | c & & + & network & # predicted & size & # matched & # derivable & # derived & @xmath97 & @xmath96 + & p & 294 & 7.96 & 29 & 155 & 38 & 0.098 & 0.245 + & p+f & 338 & 8.66 & 19 & 164 & 23 & 0.056 & 0.140 + mcl & fsw(p+f ) & 102 & 15.88 & 29 & 119 & 38 & 0.284 & 0.319 + & icd(p+f ) & 138 & 17.14 & 33 & 122 & 44 & 0.239 & 0.361 + & tcss(p+f ) & 261 & 10.52 & 42 & 158 & 54 & 0.161 & 0.342 + & consensus & 429 & 13.01 & 57 & 164 & 56 & 0.133 & 0.341 + & p & 297 & 7.94 & 39 & 155 & 49 & 0.131 & 0.316 + & p+f & 342 & 8.34 & 25 & 164 & 29 & 0.073 & 0.177 + mcl & fsw(p+f ) & 136 & 9.46 & 41 & 119 & 57 & 0.301 & 0.479 + -caw & icd(p+f ) & 141 & 7.44 & 48 & 122 & 61 & 0.340 & 0.500 + & tcss(p+f ) & 296 & 9.98 & 49 & 158 & 61 & 0.166 & 0.386 + & consensus & 484 & 8.72 & 81 & 164 & 71 & 0.167 & 0.432 + & p & 156 & 11.42 & 41 & 155 & 56 & 0.263 & 0.361 + & p+f & 306 & 14.39 & 33 & 164 & 41 & 0.108 & 0.250 + cmc & fsw(p+f ) & 136 & 12.44 & 36 & 119 & 48 & 0.265 & 0.403 + & icd(p+f ) & 252 & 8.91 & 51 & 122 & 63 & 0.202 & 0.516 + & tcss(p+f ) & 127 & 11.66 & 45 & 158 & 60 & 0.354 & 0.380 + & consensus & 429 & 9.80 & 80 & 164 & 66 & 0.186 & 0.402 + & p & 414 & 5.98 & 34 & 155 & 41 & 0.082 & 0.264 + & p+f & 510 & 6.68 & 28 & 164 & 34 & 0.055 & 0.207 + haco & fsw(p+f ) & 111 & 10.17 & 39 & 119 & 54 & 0.351 & 0.454 + & icd(p+f ) & 131 & 8.90 & 43 & 122 & 60 & 0.328 & 0.492 + & tcss(p+f ) & 269 & 7.49 & 55 & 158 & 67 & 0.204 & 0.424 + & consensus & 419 & 7.61 & 79 & 164 & 74 & 0.189 & 0.451 + [ table : methods2_mips ] finally , table [ table : methods3_mips ] compares the number of benchmark complexes successfully reconstructed by sparse clusters before and after the sparc - based post - processing .
it clearly demonstrates that many physical clusters were in fact sparse ( @xmath43-score @xmath124 ) , many of which underwent post - processing by sparc .
these post - processed clusters were able to reconstruct significantly higher number of benchmark complexes .
figure 5 in supplementary materials correlates the improvement in @xmath43-scores of these sparse clusters with the improvement in their jaccard accuracies when matched to benchmark complexes .
l || l || c || c | c || c || c | c & & & + & & & sparse & & final & derived & derived + method & network & initial & ( @xmath125 ) & processed & ( size @xmath126 ) & ( before ) & ( after ) + & @xmath80 & 638 & 269 & 8 & 338 & 0 & 2 + mcl & fsw(p+f ) & 188 & 42 & 16 & 102 & 1 & 9 + & icd(p+f ) & 258 & 57 & 18 & 138 & 2 & 9 + & tcss(p+f ) & 380 & 102 & 19 & 261 & 2 & 10 + & @xmath80 & 472 & 212 & 8 & 342 & 0 & 2 + mcl- & fsw(p+f ) & 255 & 37 & 19 & 136 & 2 & 11 + caw & icd(p+f ) & 258 & 39 & 21 & 141 & 2 & 13 + & tcss(p+f ) & 408 & 97 & 26 & 296 & 3 & 16 + & @xmath80 & 424 & 186 & 20 & 306 & 0 & 8 + cmc & fsw(p+f ) & 251 & 32 & 23 & 136 & 2 & 18 + & icd(p+f ) & 354 & 44 & 36 & 252 & 2 & 21 + & tcss(p+f ) & 224 & 56 & 41 & 127 & 4 & 27 + & @xmath80 & 389 & 25 & 510 & 338 & 1 & 10 + haco & fsw(p+f ) & 53 & 29 & 111 & 102 & 2 & 21 + & icd(p+f ) & 59 & 31 & 131 & 138 & 3 & 23 + & tcss(p+f ) & 66 & 43 & 269 & 261 & 6 & 36 + [ table : methods3_mips ] we performed in - depth analysis of some of the predicted complexes using _ cytoscape _ @xcite .
for example , the ccr4-not complex is a multifunctional complex that regulates transcription , plays a role in mrna degradation , and also regulates cellular functions in response to changes in environmental signals in yeast @xcite .
this complex was scattered " among multiple disjoint components of the physical network , and therefore went undetected from all four methods .
the addition of functional interactions facilitated linking together of these components , enabling the methods to detect it successfully ( see figure 1 ) .
-score = 0.1905 . after :
the four components were linked together into a single component ; @xmath43-score = 0.623.,title="fig : " ] [ figure:510.190.110.before_after_shrink ] while many additional complexes were detected using sparc - based refinement , there were a few complexes that were missed as well ( see supplementary materials for a list ) .
for example , the rna polymerase complexes i , ii and iii , that are involved in the formation of rna chains during transcription @xcite , were bundled into a large dense module together with some of the tbp - associated factors and tfiid complexes , which are also involved in transcription @xcite . due to the functional similarity between the subunits of all these complexes , several functional interactions were added among them . consequently , the methods recovered a large dense module housing all these complexes from which the individual complexes could not be segregated .
the same was the case with the multi - eif complexes and the saga - slik - ada - tfiid complexes .
the increase in the average cluster sizes in table [ table : methods2_mips ] further depict this effect .
functional interactions can be considered a superset " of physical interactions .
however , the low overlaps between the @xmath80 and @xmath81 networks seems to be projecting a suprisingly different picture ( @xmath84 and @xmath85 ) .
the differential curation of the two datasets - the physical dataset is inferred predominantly from experimental techniques while the functional dataset is inferred predominantly from computational techniques - along with the presence of many missing ( true negatives ) and spurious ( false positives ) interactions , give rise to these low overlaps .
though this is an observation from only the two yeast datasets considered here , it may be worthwhile investigating how far away are we from the ideal " picture of physical interactions being a proper subset of functional interactions in order to make most effective use of the two .
in this work , we attempt to reconstruct sparse " complexes from ppi networks , a problem which has not been explored in previous works ( see the recent survey by @xcite ) mainly because of the overused assumption that complexes form dense " regions within the networks .
though this assumption might be valid , relying too much on it in the wake of insufficient ppi data makes it ineffective to detect sparse complexes . to counter this ,
we employ functional interactions , which again has not been tried before
. this approach will particularly be effective in detecting complexes where a significant portion of the physical interactions are unknown or unreliable , for instance , in human .
in addition to these , we also develop some theory around complex derivability " ( the ce score ) that could be useful for developing new computational methods .
for example , in the future we will looking at devising a new computational approach that selectively uses functional interactions by treating them differently from physical interactions .
the authors would like to thank limsoon wong ( nus ) , hufeng zhou ( nus ) and gary bader ( utoronto ) for insightful discussions , guimei liu ( nus ) and shobhit jain ( utoronto ) for providing the scoring softwares , and the reviewers for their detailed comments and suggestions .
ashburner , m. , ball c.a . , blake j.a . ,
botstein , d. , butler , h. , cherry , m. , davis , a.p . , dolinski , k. , dwight , s.s . ,
epigg , j. , harris , m.a . , hill , d.p .
, issel - tarver , l. , kasarkis , a. , lewis , s. , matase , j.c . ,
richardson , j. , ringwald , m. , rubin , g.m . , sherlock , g. ( 2010 ) ` gene ontology : a tool for the unification of biology ' , _ nature genetics _ , vol . 25 , pp.2529 .
cherry , j.m . , adler , c. , chervitz s.a . , dwight s.s .
, jia , y. , juvik , g. , roe , t. , schroeder , m. , weng , s. , botstein , d. ( 1998 ) ` sgd : saccharomyces genome database ' , _ nucleic acids research _ , vol .
26 , pp.7379 .
chua , h. , ning , k. , sung , w. , leong , h. , wong , l. ( 2008 ) ` using indirect protein - protein interactions for protein complex prediction ' , _ j. bioinformatics and computational biology _ ,
vol . 6 , pp.435466 .
collins , s.r .
, kemmeren p. , zhao , x.c .
, greenbalt , j.f . , spencer f. , holstege , f. , weissman , j. , krogan , n.j .
( 2007 ) ` toward a comprehensive atlas of the physical interactome of saccharomyces cerevisiae ' , _ mol . cell .
proteomics _ , vol . 6 , pp.439450 .
cusick , m.e .
, yu , h. , smolyar , a. , venkatesan , k. , carvunis , a - r . , simonis , n. , rual , j - f . , borick , h. , braun , p. , dreze , m. , vandenhaute , j. , galli , m. , yazaki , j. , hill , d.e .
, ecker , j.r . ,
roth , f.p . , vidal , m ( 2008 ) `
literature - curated protein interaction datasets ' , _ nature methods _ , vol .
6(1 ) , pp.3946 .
han , j.d . , bertin , n. , hao , t. , goldberg , d. , berriz , g. , zhang , l.v . , dupuy , d. , walhout , a. , cusick , m.e . , roth , f. , vidal , m. ( 2004 ) ` evidence for dynamically organized modularity in the yeast protein - protein interaction network ' , _ nature _ , vol 430 , pp.88 - 93 .
mewes , h.w . , amid , c. , arnold , r. , frishman , d. , guldener , u. , mannhaupt , g. , munsterkotter , m. , pagel , p. , strack , n. , stumpflen , v. , warfsmann , j. , ruepp , a. ( 2006 ) ` mips : analysis and annotation of proteins from whole genomes ' , _ nucleic acids research _ ,
34 , pp.d169d172 .
panasenko , o. , landrieux , e. , feuermann , m. , finka , a. , paquet , n. , collart , m. ( 2006 ) ` the yeast ccr - not complex controls ubiquitination of the nascent - associated polypeptide ( nac - egd ) complex ' , _ j. biological chemistry _ , vol .
281 , pp.3138931398 .
shannon , p. , markiel , a. , ozier o. , baliga , n.s . , wang , j. , ramage , d. , amin , n. , schwikowski , b. , ideker , t. ( 2003 ) ` cytoscape : a software environment for integrated models of biomolecular interaction networks ' , _ genome research _ , vol . 13 , pp.24982504 .
srihari , s. , ning , k. , leong , h.w .
( 2010 ) ` mcl - caw : a refinement of mcl for detecting yeast complexes from weighted ppi networks by incorporating core - attachment structure ' , _ bmc bioinformatics _ ,
11 , pp.504 .
von mering , c. , huynen , m. , jaeggi , d. , schmidt , s. , bork , p. , snel , b. ( 2003 ) ` string : a database of predicted functional associations between proteins ' , _ nucleic acids research _ , vol .
12(1 ) , pp.258261 .
wang , h. , kakaradov b. , collins s.r . , karotki , l. , fiedler , d. , shales m. , shokat , k.m . ,
walter , t. , krogan n.j .
, koller , d. ( 2009 ) ` a complex - based reconstruction of the saccharomyces cerevisiae interactome ' , _ mol .
proteomics _ , vol . 8 , pp.13611377 .
|
over the last few years , several computational techniques have been devised to recover protein complexes from the protein interaction ( ppi ) networks of organisms .
these techniques model dense " subnetworks within ppi networks as complexes .
however , our comprehensive evaluations revealed that these techniques fail to reconstruct many ` gold standard ' complexes that are sparse " in the networks ( only @xmath0 recovered out of @xmath1 known yeast complexes embedded in a network of @xmath2 interactions among @xmath3 proteins ) . in this work
, we propose a novel index called component - edge ( ce ) score to quantitatively measure the notion of complex derivability " from ppi networks . using this index ,
we theoretically categorize complexes as sparse " or
dense " with respect to a given network .
we then devise an algorithm sparc that selectively employs functional interactions to improve the ce scores of predicted complexes , and thereby elevates many of the sparse " complexes to dense " .
this empowers existing methods to detect these sparse " complexes .
we demonstrate that our approach is effective in reconstructing significantly many complexes missed previously ( 104 recovered out of the 123 known complexes or @xmath447% improvement ) .
availability : http://www.comp.nus.edu.sg/@xmath4leonghw/mcl-caw/
|
because of the still lacking rigorous explanation of confinement , the nonperturbative nature of quarks and gluons inside hadrons can be explored by extracting information from distribution ( df ) and fragmentation functions ( ff ) in hard scattering processes .
there are three fundamental df that completely characterize the quark inside hadrons at leading twist with respect to its longitudinal momentum and spin : the momentum distribution @xmath1 , the helicity distribution @xmath2 and the transversity distribution @xmath3 . at variance with the first two ones
, @xmath3 is difficult to address because of its chiral - odd nature .
a complementary information can come from the analysis of the hadrons produced by the fragmentation process of the final quark , namely from ff . so far , only the leading unpolarized ff , @xmath4 , is partly known , which is the counterpart of @xmath1 .
the basic reason for such a poor knowledge is related to the difficulty of measuring more exclusive channels in hard processes .
the new generation of machines ( hermes , compass , rhic ) and planned projects ( elfe , epic ) allow for a much more powerful final - state identification and , therefore , for a wider and deeper analysis of ff , particularly when final state interactions ( fsi ) are considered . in this context , naive `` t - odd '' ff naturally arise because the existence of fsi prevents constraints from time - reversal invariance to be applied to the fragmentation process @xcite .
this new set of ff includes also chiral - odd objects that become the natural partner needed to isolate @xmath3 .
the presence of fsi allows that in the fragmentation process there are at least two competing channels interfering through a nonvanishing phase .
however , this is not enough to generate naive `` t - odd '' ff . excluding _
ab initio _ any mechanism breaking factorization , there are basically two ways to describe the residual interactions of the leading hadron inside the jet : assume the hadron moving in an external effective potential , or model microscopically independent interaction vertices that lead to interfering competing channels .
in the former case , introduction of an external potential in principle breaks the translational and rotational invariance of the problem .
further assumptions can be made about the symmetries of the potential , but at the price of loosing interesting contributions to the amplitude such as those coming from naive `` t - odd '' ff @xcite . in the latter case ,
the difficulty consists in modelling a genuine interaction vertex that can not be effectively reabsorbed in the soft part describing the hadronization .
this poses a serious difficulty in modelling the quark fragmentation into one observed hadron because it requires the ability of modelling the fsi between the hadron itself and the rest of the jet @xcite .
therefore , here we will consider a hard process , semi - inclusive deep - inelastic scattering ( dis ) , where the hadronization leads to two observed hadrons inside the same jet .
a new set of interference ff arise at leading twist @xcite and their symmetry properties are briefly reviewed . for the hadron pair being a proton and a pion with invariant mass equal to the roper resonance , we have already estimated these ff using an extended version of the diquark spectator model @xcite . in this case ,
fsi come from the interference between the direct production of the two hadrons and the decay of the roper resonance . here
, we present results for the hadron pair being two pions with invariant mass around the @xmath0 mass .
in analogy with semi - inclusive hard processes involving one detected hadron in the final state @xcite , the simplest matrix element for the hadronisation into two hadrons is the quark - quark correlation function describing the decay of a quark with momentum @xmath5 into two hadrons @xmath6 , namely @xmath7 where the sum runs over all the possible intermediate states involving the two final hadrons @xmath8 . since the three external momenta @xmath9 can
not all be collinear at the same time , we choose for convenience the frame where the total pair momentum @xmath10 has no transverse component . by generalizing the collins - soper light - cone formalism @xcite for fragmentation into multiple hadrons ,
the cross section for two - hadron semi - inclusive emission is a linear combination of projections @xmath11}$ ] of @xmath12 by specific dirac structures @xmath13 , after integrating over the ( hard - scale suppressed ) light - cone component @xmath14 and , consequently , taking @xmath15 as light - like @xcite . at leading order
, we get @xmath16}(z_h,\xi,{\bf k}_{{\scriptscriptstyle t}}^2,{\bf r}_{{\scriptscriptstyle t}}^2 , { \bf k}_{{\scriptscriptstyle t}}\cdot { \bf r}_{{\scriptscriptstyle t } } ) & \equiv & d_1 { \nonumber}\\ \delta^{[{\gamma}^- { \gamma}_5]}(z_h,\xi,{\bf k}_{{\scriptscriptstyle t}}^2,{\bf r}_{{\scriptscriptstyle t}}^2,{\bf k}_{{\scriptscriptstyle t}}\cdot { \bf r}_{{\scriptscriptstyle t } } ) & \equiv & \frac{{\epsilon}_{{\scriptscriptstyle t}}^{ij } \,r_{ti}\,k_{tj}}{m_1\,m_2}\ ; g_1^\perp { \nonumber}\\ \delta^{[i{\sigma}^{i- } { \gamma}_5]}(z_h,\xi,{\bf k}_{{\scriptscriptstyle t}}^2,{\bf r}_{{\scriptscriptstyle t}}^2 , { \bf k}_{{\scriptscriptstyle t}}\cdot { \bf r}_{{\scriptscriptstyle t } } ) & \equiv & { \epsilon_{{\scriptscriptstyle t}}^{ij}r_{tj}\over m_1+m_2}\ , h_1^{{{<\kern -0.3 em{\scriptscriptstyle ) } } } } + { \epsilon_{{\scriptscriptstyle t}}^{ij}k_{tj}\over m_1+m_2}\ , h_1^\perp \ ; , \label{e : mradici : ff } \end{aligned}\ ] ] where @xmath17 . the functions @xmath18 are the interference ff that depend on how much of the fragmenting quark momentum @xmath5 is carried by the hadron pair @xmath19 , on the way this momentum is shared inside the pair @xmath20 , and on the `` geometry '' of the pair , namely on the transverse relative momentum of the two hadrons @xmath21 and on the relative orientation between the pair plane and the quark jet axis @xmath22 , see also fig .
[ f : mradici:1 ] ) .
the different dirac structures @xmath13 are related to different spin states of the fragmenting quark and lead to the nice probabilistic interpretation at leading order @xcite : @xmath4 is the probability for an unpolarized quark to produce a pair of unpolarized hadrons ; @xmath23 is the difference of probabilities for a longitudinally polarized quark with opposite chiralities to produce a pair of unpolarized hadrons ; @xmath24 and @xmath25 both are differences of probabilities for a transversely polarized quark with opposite spins to produce a pair of unpolarized hadrons .
@xmath23 , @xmath24 and @xmath25 are ( naive ) `` t - odd '' and do not vanish only if there are residual interactions in the final state . in this case
, the constraints from time - reversal invariance can not be applied .
@xmath23 is chiral even ; @xmath24 and @xmath25 are chiral odd and can , therefore , be identified as the chiral partners needed to access the transversity @xmath3 . given their probabilistic interpretation , they can be considered as a sort of `` double '' collins effect @xcite .
in order to make quantitative predictions , we adopt the formalism of the spectator model , specializing it to the emission of a hadron pair . the basic idea is to replace the sum over the complete set of intermediate states in eq .
( [ e : mradici : delta ] ) with an effective spectator state with a definite mass @xmath26 , momentum @xmath27 .
consequently , the correlator simplifies to @xmath28 where the additional @xmath29 function allows for a completely analytical calculation of the dirac projections ( [ e : mradici : ff ] ) . for the hadron pair being a proton and a pion with invariant mass the mass of the roper resonance ,
results have been published in ref .
@xcite . in this case
, the spectator state has the quantum numbers of a scalar or axial diquark .
fsi arise from the interference between the direct production and the decay of the roper resonance . here
, we show results for the hadron pair being two pions with invariant mass in the range @xmath30 $ ] , @xmath31 mev and @xmath32 mev .
the spectator states becomes an on - shell quark with mass @xmath33 mev .
the quark decay is specialized to the set of diagrams shown in fig .
[ f : mradici:3 ] , and their hermitean conjugates , where the naive `` t - odd '' ff now arise from the interference between the direct production of the two @xmath34 and the decay of the @xmath0 . diagram [ f : mradici:3]a accounts for almost all the strength of @xmath38 production in the relative @xmath39-channel .
we have explicitly checked that diagram [ f : mradici:3]b reproduces the experimental transition probability for @xmath38 production in the relative @xmath40-channel .
hence , we believe this choice represents most of the @xmath38 strength for invariant mass in the considered interval .
we define feynman rules for the @xmath41 , @xmath42 and @xmath43 vertices introducing cut - offs to exclude large virtualities of the quark while keeping the asymptotic behaviour of ff at large @xmath44 consistent with the quark counting rule .
we infer the vertex form factors from previous works on the spectator model @xcite .
however , there numbers should be taken as indicative , since the ultimate goal is to verify that nonvanishing `` t - odd '' ff occur , particularly when integrating on some of the kinematical variables and possibly washing all interference effects out . results of analytical calculation of eq.([e : mradici : specdelta ] ) show that @xmath45 and @xmath46 . after integrating over @xmath47 while keeping @xmath48 in the horizontal plane of the lab ( usually identified with the scattering plane )
, we still get nonvanishing ff . as an example , @xmath49 is shown in fig .
[ f : mradici:4 ] for the fragmentation @xmath50 , where @xmath51 is the invariant mass of the two pions .
the cross section for the deep - inelastic scattering of an unpolarized electron on a polarized proton target where two pions are detected in the final state , contains , after integrating all the transverse dynamics ( @xmath52 ) , an unpolarized contribution proportional to @xmath53 and a term proportional to @xmath49 which depends on the transverse target polarization @xmath54 .
therefore , by flipping the polarization of the target , it is possible to build the following azimuthal asymmetry @xmath55 where @xmath56 are the azimuthal angles of @xmath57 with respect to the scattering plane , respectively .
the asymmetry shows indeed the familiar sinusoidal azymuthal dependence .
noteworthy is the factorization of the chiral - odd , naive `` t - odd '' @xmath24 from the chiral - odd transversity @xmath3 .
therefore , such asymmetry measurement allows for the extraction of @xmath3 using a model input for the ff .
de rjula a. , kaplan j.m . , and de rafael e. , _ nucl . phys . _ * b35 * , 365 ( 1971 ) .
bianconi a. , boffi s. , jakob r. , and radici m. , _ phys .
_ d * 62 * , 034008 ( 2000 ) .
bianconi a. , boffi s. , jakob r. , and radici m. , _ phys . rev .
_ d * 62 * , 034009 ( 2000 ) .
mulders p.j . , and tangerman r.d . ,
_ nucl . phys .
_ b * 461 * , 197 ( 1996 ) .
collins j.c . , and soper d.e . , _ nucl
_ b * 194 * , 445 ( 1982 ) .
collins j.c . , _ nucl .
_ b * 396 * , 161 ( 1993 ) .
jakob r. , mulders p.j .
, and rodrigues j. , _ nucl .
phys . _ a * 626 * , 937 ( 1997 ) .
|
we investigate the properties of interference fragmentation functions arising from the emission of two leading hadrons inside the same jet for inclusive lepton - nucleon deep - inelastic scattering .
using an extended spectator model for the mechanism of the hadronization , we give a complete calculation and numerical estimates for the examples of a proton - pion pair produced with invariant mass on the roper resonance , and of two pions produced with invariant mass close to the @xmath0 mass .
we discuss azimuthal angular dependence of the leading order cross section to point up favourable conditions for extracting transversity from experimental data .
= cmr8 1.5pt
|
the lulu operators remove impulsive noise before a signal extraction from a sequence .
they are computationally convenient and conceptually simpler compared to the the median smoothers usually considered to be the `` basic '' smoothers .
the lulu operators have particular properties , e.g. they are fully trend preserving , @xcite , preserve the total variation , @xcite , etc . , which make them an essential tool for multiresolution analysis of sequences .
furthermore , it was demonstrated during the last decade or so that these operators , being specific cases of morphological filters , @xcite , have a critical role in the analysis and comparison of nonlinear smoothers , @xcite .
we extend the lulu theory from sequences to functions on a continuous domain , namely , a real interval @xmath7 .
the existing lulu theory can be considered as a particular case in this new development , since the discrete lulu operators can be equivalently formulated for splines of order 1 or order 2 on the integer partition of real line , the one - to - one mapping being given by the b - spline basis .
given a sequence @xmath8 and @xmath9 the operators @xmath10 and @xmath11 are defined as follows @xmath12 in analogy with the above discrete lulu operators , for a given @xmath1 the basic smoothers @xmath2 and @xmath3 in the lulu theory are defined for functions on @xmath7 through the concepts of the so called lower and upper @xmath6-envelopes of these functions . these definitions are given in section 2 , where it is also shown that the operators @xmath2 and @xmath3 preserve essential properties of their discrete counterparts .
in particular , the operators @xmath2 and @xmath3 generate through composition a fully ordered four element semi - group also called a strong lulu structure .
this issue is dealt with in section 3 . in section 4
we define the concept of local @xmath6-monotonicity and show that the compositions @xmath4 and @xmath5 are smoothers in the sense that the resulting functions are locally @xmath6-monotone .
the errors of approximation of real functions @xmath13 by the these compositions are estimated in terms of the modulus of nonmonotonicity @xmath14 .
let @xmath15 denote the set of all bounded real functions defined on the real interval @xmath0 .
let @xmath16 denote the closed @xmath6-neighborhood of @xmath17 in @xmath18 , that is , @xmath19 .
the pair of mappings @xmath20 , @xmath21 defined by @xmath22are called lower baire , and upper baire operators , respectively , sendov .
we consider on @xmath23 the point - wise defined partial order , that is , for any @xmath24 @xmath25 then the lower and upper baire operators can be defined in the following equivalent way . for every @xmath26
the function @xmath27 is the maximal lower semi - continuous function which is not greater than @xmath13 .
hence , it is also called lower semi - continuous envelope . in a similar way , @xmath28 is the smallest upper semi - continuous function which is not less than @xmath13 and is called the upper semi - continuous envelope of @xmath13 . in analogy with @xmath27 and @xmath28
we call the functions @xmath29a lower @xmath6-envelope of @xmath13 and an upper @xmath30-envelope of @xmath13 , respectively .
it is easy to see from ( [ ldenvelop ] ) and ( [ udenvelop ] ) that for every @xmath31 @xmath32 furthermore , the operators @xmath33 and @xmath34 , @xmath1 , as well as @xmath20 and @xmath35 are all monotone increasing with respect to the order ( [ forder ] ) , that is , for every @xmath24 @xmath36 the following operators can be considered as continuous analogues of the discrete @xmath37 operators given in the introduction : @xmath38 we will show that these operators have similar properties to their discrete counterparts .
let us note that they inherit monotonicity with respect of the functional argument from the operators @xmath33 and @xmath34 , ( [ idsdmon ] ) , that is , for @xmath24 @xmath39 [ tldfleqf ] for every @xmath40 and @xmath41 we have @xmath42 , @xmath43 .
let @xmath44 , @xmath1 .
for any @xmath45 it follows from the definition of @xmath33 that @xmath46 . therefore @xmath47 , @xmath45 . the second inequality in the theorem is proved in a similar way . [ tldudmon ]
the operator @xmath2 is monotone increasing on @xmath6 while the operator @xmath3 is monotone decreasing on @xmath6 , that is , for any @xmath44 and @xmath48 we have @xmath49 , @xmath50 .
let @xmath51 . using the properties ( [ iscomp ] )
the operator @xmath52 can be represented in the form @xmath53 .
it follows from theorem [ tldfleqf ] that for every @xmath44 we have @xmath54 . hence using the monotonicity of the operator @xmath34 given in ( [ idsdmon ] ) we obtain @xmath55 , @xmath44 . the inequality @xmath56 is proved in a similar way .
the next lemma is useful in dealing with compositions of @xmath33 and @xmath34 .
[ tidsdid ] we have @xmath57 , @xmath58 . using the monotonicity of @xmath33 , see ( [ idsdmon ] ) , and theorem [ tldfleqf ] for @xmath59 we have @xmath60 . on the other side ,
applying theorem [ tldfleqf ] to @xmath61 we obtain @xmath62 .
therefore @xmath63 , @xmath44 .
the second equality is proved similarly .
[ tldudabsorb ] for every @xmath31 we have @xmath64 and @xmath65 .
we will only prove the first equality since the proof of the second one is done in a similar manner .
let first @xmath51 . using the property ( [ iscomp ] ) and lemma [ tidsdid ]
we obtain @xmath66 if @xmath67 in a similar way we have @xmath68 the proof in the case when @xmath69 follows from either of the above identities where @xmath70 or @xmath71 respectively are replaced by the identity operator .
important properties of smoothing operators are their idempotence and co - idempotence .
hence the significance of the next theorem .
[ tldudidemp ] the operators @xmath2 and @xmath3 are both idempotent and co - idempotent , that is , @xmath72 , @xmath73 , @xmath74 , @xmath75 , where @xmath76 denotes the identity operator .
the idempotence of @xmath2 and @xmath3 follows directly from theorem [ tldudabsorb ] .
the co - idempotence of the operator @xmath2 is equivalent to @xmath77 .
using the first inequality in theorem [ tldfleqf ] one can easily obtain @xmath78 .
hence , for the co - idempotence of @xmath2 it remains to show that @xmath79 .
assume the opposite .
namely , there exists a function @xmath44 and @xmath45 such that @xmath80 .
let @xmath81 be such that @xmath82 .
using the definition of @xmath2 the above inequality implies that there exists @xmath83 such that for every @xmath84 we have @xmath85 , or equivalently @xmath86 for every @xmath87 we also have @xmath88 .
hence there exists @xmath89 such that @xmath90 , @xmath87 .
taking @xmath91 in the above inequality we obtain @xmath92 , which contradicts ( [ ineq3tidemp ] ) .
the co - idempotence of @xmath3 is proved in a similar way .
the figures below illustrate graphically the smoothing effect of the operators @xmath2 , @xmath3 and their compositions .
the graph of function @xmath13 is given by dotted lines .
+ the functions @xmath93 and @xmath94 + + the functions @xmath95 and @xmath96 the operator @xmath2 smoothes the function @xmath13 from above by removing sharp picks while the operator @xmath3 smoothes the function @xmath13 from below by removing deep depressions . the smoothing effect of the compositions @xmath4 and @xmath5 can be described in terms of the local @xmath6-monotonicity discussed in the section 4 . note that @xmath4 and @xmath5 resolve ambiguities in a different way ; @xmath4 treats oscillations of length less then @xmath6 as upward impulses and removes them while @xmath5 considers such oscillations as downward impulses which are accordingly removed .
the inequality @xmath97 which is observed here will be proved in the next section for @xmath44 .
in this section we consider the set of the operators @xmath2 and @xmath3 and their compositions . for operators on @xmath23
we consider the point - wise defined partial order .
namely , for operators @xmath98 , @xmath99 on @xmath23 we have @xmath100 then the inequalities in theorem [ tldfleqf ] can be represented in the form @xmath101 where @xmath76 denotes the identity operator on @xmath23 .
[ tulleqlu ] for any @xmath1 we have @xmath102 .
let @xmath44 and let @xmath45 .
denote @xmath103 .
let @xmath104 be an arbitrary positive .
for every @xmath105 we have @xmath106 there exists @xmath107 such that @xmath108 .
then @xmath109 , which implies that @xmath110 .
hence @xmath111 . then @xmath112 for @xmath113 .
since @xmath114 , see the case assumption , from the above inequality we have @xmath115 .
+ for every @xmath116 we have @xmath117 .
denote @xmath118 .
we will show that for every @xmath119 we have @xmath120 due to the inequality ( [ ineqtulleqlu ] ) we have that ( [ intersecttulleqlu ] ) holds for every @xmath116 .
let @xmath119 and let @xmath121 .
this implies that @xmath122 . using the inequality ( [ ineqtulleqlu ] ) for @xmath123 as well as the case assumption
we obtain that the set @xmath124\cap d$ ] is not empty .
then @xmath125\cap d\neq \emptyset$ ] . for @xmath126 condition ( [ intersecttulleqlu ] )
is proved in a similar way .
hence ( [ intersecttulleqlu ] ) holds for all @xmath119 .
let @xmath119 and @xmath127 .
since @xmath128 we have @xmath129 , for @xmath130 . using that @xmath131 we obtain that @xmath132 , @xmath119
. therefore @xmath133 . then @xmath134 combining the results of case 1 and case 2 we have @xmath135 .
since @xmath104 is arbitrary this implies that @xmath136 .
[ tluulidemp ] for a given @xmath1 the operators @xmath4 and @xmath5 are both idempotent .
the proof is an immediate application of lemma 1 .
[ tulu = lu ] we have @xmath137 , @xmath138 , @xmath1 . using the inequalities ( [ lleqidlequ ] ) and the monotonicity of the operators @xmath2 , @xmath3 , see ( [ ldudfmon ] )
, we obtain @xmath139 . for the proof of the inverse inequality we use theorem [ tulleqlu ] and the idempotence of @xmath3 as follows : @xmath140 therefore @xmath137 .
the second equality is proved in a similar way .
it follows from theorems [ tluulidemp ] and [ tulu = lu ] that for a fixed @xmath1 every composition involving finite number of the operators @xmath2 and @xmath3 is an element of the set @xmath141 . hence the operators @xmath2 and @xmath3 form a semi - group with a composition table as follows : [ cols="^,^,^,^,^",options="header " , ] furthermore , an easy application of theorem [ tulleqlu ] shows that this semi - group is completely ordered .
namely , we have @xmath142 .
let @xmath1 and a function @xmath143 be given .
\(i ) the function @xmath13 is called upwards @xmath6-monotone if for every interval @xmath144\subset\omega$ ] with @xmath145 we have @xmath146 } f(z)=\max\{f(x),f(y)\}$ ] .
\(ii ) the function @xmath13 is called downwards @xmath6-monotone if for every interval @xmath144\subset\omega$ ] with @xmath145 we have @xmath147 } f(z)=\min\{f(x),f(y)\}$ ] .
\(iii ) the function @xmath13 is called locally @xmath6-monotone if it is both downwards @xmath6-monotone and upwards @xmath6-monotone .
the name locally @xmath6-monotone reflects the following characterization : @xmath148\!\subseteq\ !
\omega$ , $ y\!-\!x\!\leq\!\delta$ , $ f$ is either\\ monotone increasing or monotone decreasing \end{tabular}\ ] ] for every @xmath1 and @xmath44 the function @xmath149 is upwards @xmath6-monotone while the function @xmath150 is downwards @xmath6-monotone .
let @xmath1 , @xmath44 and @xmath144\subseteq\omega$ ] , @xmath145 .
denote @xmath151 .
it is easy to see that for every @xmath152 $ ] we have @xmath153 .
therefore @xmath154 which shows that function @xmath155 is upwards @xmath6-monotone .
the downwards @xmath6-monotonicity of @xmath94 is proved in a similar way .
[ tldupmon ] for every @xmath1 and @xmath44 the function @xmath93 is upwards @xmath6-monotone while the function @xmath94 is downwards @xmath6-monotone .
[ tullocdmon ] for every @xmath1 and @xmath44 the functions @xmath156 and @xmath95 are both locally @xmath6-monotone .
the proof follows from corollary [ tldupmon ] and the composition table in the preceding section .
theorem [ tullocdmon ] shows that the operators @xmath5 and @xmath4 provide locally @xmath6-monotone approximations to the functions in @xmath23 .
the error of the approximation can be estimated in terms of the modulus of nonmonotonicity .
let us recall the definition .
[ defmodulus ] let @xmath40 .
the mapping @xmath157 given by @xmath158$\end{tabular } } \!\!(|f(x_1)\!-\!f(x)|\!+\!|f(x_2)\!-\!f(x)|\!-\!|f(x_1)\!-\!f(x_2)|)\ ] ] is called modulus of nonmonotonicity of @xmath13 .
the locally @xmath6-monotone functions can be conveniently characterized through the modulus of nonmonotonicity . for any @xmath44 and @xmath1 we have @xmath159 we will derive
error estimates first in the case when @xmath160 .
it will prove useful to consider the upper semi - continuous envelope of the modulus of nonmonotonicity .
let @xmath44 . using that @xmath14 is monotone increasing with respect to @xmath6 the upper semi - continuous envelope of @xmath161
can be represented as @xmath162 .
[ testim1 ] let @xmath163 and @xmath1 .
then @xmath164 , @xmath165 , @xmath45 .
let @xmath166 .
denote @xmath167 . if @xmath168 the first inequality of the theorem holds .
assume that @xmath169 .
let @xmath170 be such that @xmath171 .
then we have @xmath172 denote @xmath173 , @xmath174 , @xmath175 , @xmath176 . using the inequality ( [ ineq1testim1 ] ) with @xmath177 and @xmath123
we obtain that @xmath178 and respectively @xmath179 are not empty and that @xmath180 . therefore @xmath181
= b_\frac{\delta}{2}(x)$ ] . then the inequality ( [ ineq1testim1 ] ) implies that @xmath182 .
hence @xmath183 .
let @xmath81 be arbitrary .
the neighborhood @xmath184 has nonempty intersections with both @xmath178 and @xmath179 .
let @xmath185 and @xmath186 .
we have @xmath187 and @xmath188 $ ] . from the definition of the modulus of nonmonotonicity we have @xmath189 . on the other side @xmath190
. therefore @xmath191 . going with @xmath104 to @xmath192 we
obtain @xmath193 . since @xmath194 is arbitrary small
this implies the first inequality of the theorem .
the second inequality is proved in a similar manner .
using theorem [ testim1 ] as well as ( [ dmonchar ] ) and corollary [ tldupmon ] we have the following characterization of the fixed points of operators @xmath2 and @xmath3 . for any @xmath163 @xmath195 [ testim2 ] let @xmath1 and @xmath163 .
then @xmath196 , @xmath197 , @xmath198 , @xmath199 .
we will prove the inequalities for @xmath93 since the ones for @xmath94 are proved in a similar way .
denote @xmath200 .
let @xmath201 $ ] be an arbitrary interval of length at most @xmath6 and let @xmath202 $ ] .
we consider the number @xmath203 .
according to corollary [ tldupmon ] the function @xmath155 is upper @xmath6 monotone , which implies that @xmath204 . if we also have @xmath205 , then the number @xmath206 is zero and the first inequality of the theorem is trivially satisfies .
let @xmath207 and let @xmath170 be such that @xmath208 .
the number @xmath206 can then be represented in the form @xmath209 .
if we assume that @xmath210 for all @xmath105 , then @xmath211 , which is a contradiction . therefore , there exists @xmath83 such that @xmath212 . if @xmath213 then using that @xmath214 we obtain @xmath215 for all @xmath216 which implies @xmath217 .
this contradiction shows that @xmath218 . in a similar way
we show that @xmath219 .
using also the first inequality of theorem [ tldfleqf ] we have @xmath220 . using
that @xmath194 can be arbitrary small we obtain @xmath221 .
since the interval @xmath201 $ ] of length at most @xmath6 and @xmath202 $ ] are arbitrary this implies that @xmath222 .
the inequality @xmath223 follows from the monotonicity of the operator @xmath35 , see ( [ idsdmon ] ) . in the next theorem
we give estimates for the error of approximation of a function @xmath163 in terms of the supremum norm denoted here by @xmath224 .
[ testim3 ] let @xmath1 and @xmath163 .
then @xmath225 , @xmath226 .
applying theorems [ testim1 ] and [ testim2 ] we obtain @xmath227 and @xmath228 , which implies the first inequality of the theorem .
the second inequality is proved in a similar way .
error estimates similar to theorem [ testim3 ] can be derived in case of @xmath7 being finite or semi - finite interval using a modification of the modulus of nonmonotonicity . for simplicity we will only consider the case
@xmath229 $ ] , @xmath230 .
let @xmath40 .
the mapping @xmath231 given by @xmath232}(|f(x_1)\!-\!f(x_2)| , \sup_{x_{1,2}\in[b-\frac{\delta}{2},b]}(|f(x_1)\!-\!f(x_2)|\right\}\ ] ] is called modified modulus of nonmonotonicity of @xmath13 .
this modulus is similar to the corrected modulus of nonmotonicity in @xcite .
[ testim4 ] let @xmath163 and @xmath1 .
then @xmath233 , @xmath234 , @xmath45 .
the proof is similar to the proof of theorem [ testim1 ] .
it is easy to see that for any @xmath235 $ ] the functions @xmath93 and @xmath94 are constants on each of the intervals @xmath236 $ ] and @xmath237 $ ] .
therefore , using also theorem [ testim2 ] we have @xmath238 . in the same way we obtain @xmath239 .
hence the modified modulus satisfies the similar inequalities to the ones given in theorem [ testim2 ] for @xmath161 and @xmath240 . using these inequalities and theorem [ testim4 ]
we obtain the error estimates in the next theorem , which are similar to the ones in theorem [ testim3 ] .
[ testim5 ] let @xmath1 and @xmath163
. then @xmath241 , @xmath242 .
in this paper we extended the lulu operators from sequences to real functions defined on a real interval using the lower and upper @xmath6-envelopes of functions . the obtained structure , although more general than the well known lulu structure of the discrete operators , retains some of its essential properties . for a fixed @xmath1 the compositions @xmath4 and @xmath5 provide locally @xmath6-monotone approximations for real functions ,
the error of approximation being estimated in terms of the modulus of nonmonotonicity of the functions .
further properties of the lulu operators for functions on continuous domains , e.g. trend preservation , will be investigated in the future .
generalizing the theory to functions on multidimensional domains is still an open problem .
|
the lulu operators , well known in the nonlinear multiresolution analysis of sequences , are extended to functions defined on continuous domain , namely , a real interval @xmath0 .
similar to their discrete counterparts , for a given @xmath1 the operators @xmath2 and @xmath3 form a fully ordered semi - group of four elements .
it is shown that the compositions @xmath4 and @xmath5 provide locally @xmath6-monotone approximations for the bounded real functions defined on @xmath7 .
the error of approximation is estimated in terms of the modulus of nonmonotonicity .
|
computational mechanics ( cm ) @xcite has a formalism @xcite that has been proved to construct the minimal model capable of statistically reproducing all the resolvable causal structure of any infinite sequence of discrete measurements ( be they scalar , vector , tensor , or descriptive)@xcite .
the size of a model so defined , measured by a quantity termed statistical complexity , @xmath0 , @xcite is a reliable and falsifiable indication of the amount of structure the data contain@xcite .
the particular strengths of this approach are that it enables the complexities and structures of different sets of data to be quantifiably compared and that it directly discovers detailed causal structure within those data . by examining data in this way it is possible to appreciate , in a well - defined abstract sense , how a system actually functions and what scales are most important to it .
this information can then be used to optimise the efficiency of physically plausible models@xcite . as with all other analytical tools , cm has some limitations in the face of certain real - world problems that affect the information content of the signal under study .
these problems may include : 1 .
gaps in the data 2 .
noise 3 .
restricted sequence length 4 .
correlations at a very wide range of scales .
the problem of correlations at a wide range of scales is particularly interesting and relevant to geophysical and other natural time series because of their typically power law ( coloured noise ) fourier spectra @xcite .
theoretically , the minimum resolvable scale will be constrained by the data sampling interval and the maximum resolvable scale by the length of the data series . in practice
, the range of resolvable scales will also be set by the available computational resources .
thus , the range of resolvable scales may be less than those necessary to evaluate correlations on all relevant scales .
consequently , it is important to understand how structural analysis is affected by unresolved structure due to correlation . in this paper
, we address these issues in detail . in section [ method ] ,
we discuss how the ideal formalism of cm can be adapted to apply to a non - infinite series of corrupted and correlated data .
in particular , three concepts are defined and discussed : ( 1 ) a tolerance parameter @xcite to account for the statistical uncertainty introduced by a non - infinite series that destroys the exact equivalence of different causal states sharing the same outcome .
( 2 ) a new expletive filter that removes signal corruption by assuming that corruption creates rare causal states or words that are not in the dictionary of the true signal . (
3 ) the new concept of effective soficity in which a data series has a finite set of equivalent causal states that is stable to small changes in the effective memory of those states .
the latter concept distinguishes between unresolvable random " structure and resolvable structure whose discovery is only prevented by the effective memory being used and by the length of the data series . in section [ examples ]
, we apply the cm algorithm with these additional concepts to the analysis of structure in four simulated time series : ( 1 ) uncorrelated , white noise .
( 2 ) periodic signal with white noise corruption . ( 3 ) a biased poisson switch ( i.e. , a sequence of pulses whose pulse durations and inter - pulse intervals are determined by stationary poisson processes ) . ( 4 ) a sequence of bursts similar to ( 3 ) but with fixed pulse duration .
the structure of the time series is analysed by searching for regions of effective soficity in maps of statistical complexity over the parameter space of the cm model .
the simulated time series represent four types of signal thought to be present in time series measurements of the geomagnetic field . in section
[ application ] , we use the cm algorithm to examine a real geomagnetic time series measured at halley , antarctica , in which deflections of the earth s magnetic field are due mainly to electrical currents in the ionosphere .
the cm analysis yields a structural model that comprises a diurnal component corresponding to the oscillation of the measuring apparatus with the rotation of the earth and a poisson - switched , fixed - duration , pulse component that is likely associated with the magnetospheric substorm @xcite . in section [ discussion ] , we discuss some general principles that have been learnt in applying cm to the analysis of structure in real data , and draw conclusions in section [ conclusion ] .
here we give an introduction to the practical use of cm in the analysis of real data .
we concentrate only on describing in detail the formalism for the parsing structure that we have used in the analyses . for a fuller description of the potential intricacies of the method see reference@xcite . defining some new terminology ,
we highlight the difficulties associated with analysing experimental data in this way , and explain solutions to these problems . to start with , one has a set of measurements - either a spatial or temporal series where the separation between each point is known .
the total time or length for which data exist is their span , @xmath1 .
after coarse - graining at a fixed scale @xmath2 , the series has @xmath3 equally spaced measurements .
next , we digitise the signal amplitude . for reasons that will be apparent later
, the number of possible digits should be low unless the series length is extremely large .
the digitised sequence is then a concatenation of @xmath4 letters @xmath5 , where there are @xmath6 types of such letters , ranging from @xmath7 up to @xmath8 . in order to maximize the prior probable information content of the processed sequence , digitisation
should normally be performed such that there are equal numbers of each letter present .
for example , in the case of binarisation ( where @xmath9 ) , this would mean that the threshold for letter @xmath10 would be the median value of the data .
it should be noted , though , that the best way to digitise the sequence is that which _ actually _ maximises the information content of the result ; but that this can not usually be guessed .
another approach that has been suggested@xcite is to use the formalism of maximum entropy .
the next step is to parse the sequence .
one begins by composing words from each group of @xmath11 consecutive letters ; the @xmath12th word , @xmath13 , is defined by : @xmath14 thus there are @xmath15 possible words , each represented by a unique scalar , @xmath16 : @xmath17 the total number of words generated from the sample is @xmath18 .
we now introduce some terminology ; any word @xmath13 may be called a _ proword _ , @xmath19 when followed by any word @xmath20 .
this latter is called the _ epiword _ , @xmath21 .
for this sentence we digress slightly to note that it may sometimes be beneficial to perform the initial digitisation on each separate block of data @xmath22 letters long rather than the entire dataset .
we now proceed to capture causal structure in the word sequence by compiling a tally of epiwords following each proword .
this means going through the sequence incrementing an array @xmath23 accordingly . representing summation over an index by its omission
, we see that the total tally is @xmath24 . thus , contracting over epiwords gives a tally of prowords only : @xmath25 and the fractional prevalence of each proword in the sequence is therefore contained in the vector @xmath26 finally , the fractional _ profile _ of each proword by epiword is given by the array @xmath27 where the repeated indices in the division are not summed over .
given a particular proword , this tells us the likelihoods of transitions to the various epiwords .
the crux of the technique now lies in identifying prowords with equivalent epiword profiles .
such prowords are said to belong to the same `` equivalence class '' or `` causal state '' - i.e. they share statistically equivalent probabilistic futures ( at the level of analysis one has been pursuing ) .
the identification is made via an equivalence relation , denoted by @xmath28 .
for an infinite sequence , @xmath28 can demand exact correspondence between profiles , in which case it is always transitive ( meaning @xmath29 ) . in a practical situation , where even the finite length of the sequence introduces fluctuations in the calculated profiles@xcite ,
it is not possible to be so exact .
we therefore introduce a tolerance parameter , @xmath30 , within the bounds of which the profiles of words in the same equivalence class are allowed to vary : two prowords , @xmath31 and @xmath32 , are in the same equivalence class if , @xmath33 @xmath34 ; @xmath35 where the large vertical bars signify absolute magnitude .
although this appears to destroy the formal transitive property of @xmath28 , it is postulated here that after some finite sequence length is surpassed , there always exists a definite range of values for @xmath30 within which the transitive property is observed correctly .
the transitive property can be enforced where @xmath28 is not transitive by grouping equivalence classes defined by @xmath28 that share at least one word .
having identified the words lying within each equivalence class , a model which outputs a series of letters statistically equivalent to the original can be constructed .
it is a particular strength of the technique that the model generated is always a minimal representation of the data s statistical structure for the amount of memory the analysis employs@xcite .
the model is easiest to describe in terms of its representation as a labelled diagraph " .
two very simple labelled diagraphs , with extracts from their outputs , are presented in figure [ twzrcmplxtdgrphs ] .
a more complicated labelled diagraph , also representing a minimal model , is shown in figure [ mrcmplctddgrph ] .
each diagraph comprises a node or nodes , indicated by a circle with a number in it , and lines joining one node to another or to itself .
each numbered node of a diagraph represents a causal state corresponding to each of the model s equivalence classes , while each line ( unidirectionally ) joining two nodes is labelled with the string of letters ( the word ) that is output when that line is followed .
in addition , each line is associated with a probability .
the word output on going from one causal state to another is in the equivalence class of the future state .
the probability of each word s output may therefore be trivially given by @xmath36 .
it should be held in mind that only a subset of all possible labelled diagraphs represent minimal models .
even so , an arbitrary diagraph s output can naturally be used to construct the appropriate minimal model .
models like this are useful for three reasons : 1 .
their minimality allows the structure of two sets of data to be directly compared .
once a model has been synchronized with current data it optimizes one s ability to forecast the behaviour of the system in the future .
3 . the information concerning scales of causal structure in the data can be used to optimize the performance of more physically plausible models .
if a recursive decomposition is employed@xcite , diagraphs labelled with words to output can be reformed into equivalent diagraphs labelled with single letters .
this is perhaps a mathematically aesthetic thing to do .
however , because any real analysis is performed with an effective memory of only @xmath11 symbols , only the last @xmath11 symbols are of any use for prediction .
this fact is not manifest in the single - letter diagraph obtained through the decomposition of the transition matrix , @xmath37 .
moreover , it is necessary to synchronise a single - letter diagraph to an input stream of data before it is of any use for the purpose of prediction . in this paper
we concentrate only upon the identification of proword equivalence classes , because it is a powerful tool for pattern discovery in its own right . a measure of the structure of such models is given by the statistical complexity@xcite : @xmath38 where logarithms are canonically taken to base @xmath39 and the prevalence @xmath40 of equivalence class @xmath12 is given by the sum of the prevalences of the words in that class . for example
, the models represented by the labelled diagraphs in figure [ twzrcmplxtdgrphs ] both have a statistical complexity of zero because they only have one causal state ( and therefore one equivalence class ) each .
this is sensible because they both output noise .
the model represented in figure [ mrcmplctddgrph ] , though , has four causal states with equal prevalences and a correspondingly higher statistical complexity of two bits : @xmath41 @xmath0 is extremely important , not only because it reflects the complexity of the system , but also because it does not converge until the data have been fully characterized .
it is a hard fact that if the sequence length @xmath4 is too small , full characterization will not be possible .
this is because fluctuations in the proword profiles will corrupt the identification of equivalence classes . in this text the resultant unresolvable structure
is called _ noise_. in contrast , resolvable but as yet unresolved structure is described as _ random _ @xcite .
such random structure is likely to be encountered in analysing data sets with correlation lengths comparable to or exceeding the maximum wordlength .
making this distinction is very important , even though it is not possible to discern whether unresolved structure is noisy or random until further computation has resolved it . in other words ,
the data appear to be noisy until the series is found to be _ effectively sofic _ , at which point @xmath0 attains its correct value and the model is complete .
we note that sofic sequences are those which still have a finite number of equivalence classes when @xmath4 is infinite and @xmath11 is semi - infinite ( see badii and politi@xcite , page 80 for a longer explanation ) .
_ effective _ soficity is here defined to mean that a sequence has equivalence classes that are stable to an increase in wordlength .
thus , a sequence could be effectively sofic at one range of wordlengths but not at another where either more or less structure is in the process of being identified .
both random structure and noise will redistribute the original tally from what would be expected if only resolved structure was present , raising @xmath0 from the value corresponding to resolved structure alone and increasing the complexity of its model .
a simple one - parameter model for the corruption process is to assume that the probability that any letter is corrupted to any other letter is @xmath42 .
then the probability any letter stays as it is is @xmath43 and the corruption will have been governed by the redistribution function @xmath44 where @xmath45 reads , _ the pure word @xmath46 , when corrupted by noise in a certain way , is identical to the corrupt word @xmath47_. the label `` pure '' implies effective soficity .
thus , assuming the corruption of prowords and epiwords are independent we have @xmath48 where @xmath49 where @xmath50s is a kronecker delta and the corruption of letters is assumed to be independent .
it happens that arbitrarily corrupted distributions can be uniquely deconvolved as long as one knows @xmath42 , but this is not usually the case in an experimental situation .
we have two alternative options .
the first is to scan through @xmath42 , deconvolving the proword prevalences each time .
this will produce a drastic decrease in the statistical complexity at some point , signifying correct parameterization of @xmath42 . a good guess for @xmath42
might be the first value which results in a single proword having a prevalence of zero . whilst the assumption of independent corruption of letters is likely to be a good model of noise , it is unlikely to be a good model of the uncharacterised correlated structure that we call random .
consequently , a second option is to ignore the details of any corruption and simply assume that the prevalence of any expletive ( corrupted word ) is below a certain expletive prevalence , @xmath51 .
we scan through @xmath51 , eradicating any prowords whose prevalence is less than @xmath51 , and recalculate @xmath0 each time .
this procedure can alternatively be performed after the identification of preliminary equivalence classes to eradicate expletive equivalence classes . in any case
, the approach can only work when the actual structure - to - noise ratio ( snr ) is high enough to ensure that expletives are eradicated before meaningful words are .
if the pure proword prevalence distribution is very uneven this method can not work .
in general , a combined method would probably be most successful - that is , where one first attempts the deconvolution after making some bold assumptions and then removes the resulting low - prevalence words completely . it is always possible to determine all the resolvable structure of a sequence for which the snr is arbitrarily small , so @xmath0 is independent of snr .
of course though , if the snr is zero , so is @xmath0 , because the model suddenly collapses to a single equivalence class .
note that deconvolution can always be achieved by inversion of an assumed convolution matrix , but that this is not always easy .
in particular , if one knew the actual matrix then the `` noise '' would not be noise at all , but resolved structure .
the only deconvolution that is strictly necessary is that which removes the noise ( unresolvable structure ) from the signal .
it should therefore assume that the redistribution is gaussian .
in practice though , some random ( resolvable ) structure may be so computationally difficult to identify that a messy deconvolution is required to remove it , allowing the analysis of more easily resolvable structure to proceed .
it is admissible to remove expletives from the prevalence distribution because they destroy the effective soficity of the data .
it is instructive at this point to go through the uncertainties present in the profile and prevalence distributions . when the sequence length is large compared to @xmath15 the probability that any individual word has been corrupted is approximately @xmath52 .
following the definition of the prevalence distribution , we find that the uncertainty in the prevalence of a proword @xmath53 is governed by an inequality : @xmath54 where the lower limit corresponds to uncorrelated errors and the upper limit to systematic errors .
we indeed expect the uncertainty to be somewhere in this range because the errors are due to unresolved structure .
the uncertainty in the prevalence of a single epiword within a particular proword s profile is expected to be greater : @xmath55 these inequalities go some way to justifying the use of the blanket tolerance @xmath30 to identify the equivalence classes , because we know nothing about the nature of the errors . in some cases
it is conceivable that @xmath30 would have to be scaled by @xmath56 , where @xmath57 in order to correctly identify equivalence relations between profiles . in such cases @xmath58 is an extra parameter .
we now turn to the analysis of test sets of data by the algorithm described in detail above .
the test data represent signal types thought to be present in time series measurements of the geomagnetic field that we shall study in the next section . a white noise ( temporally uncorrelated ) signal was generated by a sequence of 5000 independent samples from a uniform distribution and converted to binary by setting those values above the median to unity and those below the median to zero .
figure [ rndmsttstclcmplxt ] shows the variation of statistical complexity , @xmath0 , versus word length , @xmath11 , and tolerance , @xmath30 , for this signal .
the absence of a plateau in this graph indicates that , for the range of memories ( wordlengths ) tested , the analysis does not discern any structure at all in the signal .
the linear variation of @xmath0 with @xmath11 for @xmath59 represents models with as much arbitrariness as is possible at each level of memory used in the analysis .
these models collapse to a single equivalence class as the tolerance parameter is increased .
an increasing amount of tolerance is required for this collapse for increasing wordlength , as expected from equation [ uncertainty ] .
thus , no complex models at all were constructed for this noisy sequence at any time during this analysis .
this was expected ; we would have been disappointed with the random number generator that was used to construct the sequence ( the idl
randomu " function , see also@xcite ) if we had easily found correlations .
figure [ 10prcntnsbnrprd4sttstcclcmplxt ] shows the result of the analysis on a binary period four sequence ( i.e. , 00110011 ... ) of length 5000 , where 10% of the bits have been randomly flipped .
this graph has a stable , but rather jagged , plateau at @xmath60 which begins at wordlength 4 for tolerances in the range @xmath61 .
this plateau corresponds to a group of models that capture the essential structure in the signal . in the absence of noise the statistical complexity of a binary period
four signal should be @xmath62 .
the apparently anomalously high level of the plateau is caused by both the noise and the finite sequence length corrupting the identification of the equivalence classes .
it is not entirely flat because the corruption is different at each value of wordlength and tolerance .
in fact , there is a gentle downward trend which would converge to @xmath63 in the limit of the extra , spurious , states decreasing in prevalence at longer and longer wordlengths , if the sequence was long enough .
note that the gradient of the increase of statistical complexity with wordlength changes at a memory equal to half the period of the structure in this signal .
this is the point at which the structure is first discovered : it is important to note that the convergence of @xmath0 is not immediate , suggesting an analogue of the nyquist sampling theorem for cm .
note also steep drops in @xmath0 where previously distinguishable equivalence classes have suddenly collapsed together as the tolerance parameter @xmath30 exceeds some critical value , supporting our earlier postulate that there is a definite range of @xmath30 within which a true transitive property is observed .
figure [ 10prcntnsbnrprd4sttstcclcmplxtx075 ] shows results of a similar analysis on the same sequence , excepting that this time , words of prevalence less than @xmath51 ( the expletive prevalence parameter ) were eradicated from the probability distributions .
@xmath51 was chosen to be 0.075 for this graph .
as we can see , this approach was entirely successful in the respect that @xmath0 converges to a plateau for a broad range of the tolerance parameter @xmath30 .
a minimal model that was capable of outputting sequences with statistical structure identical to that characterized from the input was effectively constructed at every point on this plateau .
the value of @xmath0 for periodic sequences was always found to directly reflect the amount of memory required by the system to produce such data : a sequence with a sole period , @xmath64 , has a statistical complexity of @xmath65 ( in this case the period four signal has a statistical complexity of @xmath66 ) .
moreover , we can appreciate that the analysis only yields a convergent value after the wordlength has exceeded at least half the period of the sequence .
more generally ; convergence begins when an analysis first has greater memory than a system .
if the system has certain structure with greater memory than it may be feasible to analyse , for example , a red noise signal that consists of many fourier modes with a power law distribution of amplitudes and random phases , @xmath0 will not ever truly converge
. however , there may be stages where the analysis has enough memory to identify _ some _ structure , and this is indicated by approximately flat regions , or , at the very least , dips in the gradient of @xmath0 with increasing wordlength .
we next turned to more detailed analyses of two other illustratively important diagraph s outputs .
the first we considered was the biased poisson switch@xcite , represented as a labelled diagraph in figure [ pssnswtchdgrph ] .
the circled states , @xmath67 and @xmath10 may each generate either a one or a zero with the probabilities shown . in the figure , @xmath68 and @xmath69 .
note that when @xmath70 the output sequence is no longer biased .
it turns out that the values of @xmath0 we can derive for different values of @xmath71 and @xmath72 provide some nice insights into the nature of information and the optimization of measurement processes .
the measure has two distinct regimes : where @xmath73 , and where they do not .
since the diagraph only has two states , it is clear that as far prediction of the next epiword is concerned , only the last bit of any proword can ever matter .
therefore , all words usually separate into two equivalence classes ( corresponding to odd and even words ) .
if , however , @xmath73 then @xmath74 and @xmath75 .
this always results in the two equivalence classes collapsing into one , giving a statistical complexity of zero , corresponding to pure noise .
this is appropriate because in this degenerate situation the possible outcomes of node 0 in figure [ pssnswtchdgrph ] are identical to those of node 1 and the diagraph collapses to a single state too ( see inset ) , and can only produce noise anyway .
if the diagraph does not collapse in this way there will always be two equivalence classes .
their prevalences are found to be @xmath76 and @xmath77 .
thus , when @xmath78 , we have : @xmath79 } \log_{2}\left ( \frac{\alpha}{\beta}+1\right ) + \frac{1}{\left [ \frac{\beta}{\alpha}+1\right ] } \log_{2}\left ( \frac{\beta}{\alpha}+1\right)\end{aligned}\ ] ] and if @xmath80 , @xmath0 is always zero . a graph of this function is shown in figure [ pssnswtchsttstcclcmplxtvrssbs1 ] .
note that it always evaluates to unity when @xmath81 , except when @xmath71 = @xmath82 .
if @xmath71 does not equal @xmath72 ( and @xmath78 ) then it is less than unity .
in fact , as the switch becomes more and more biased the statistical complexity goes down and down , reaching zero when only one digit is ever output .
this is to be expected because a biased data set ( e.g. more ones than zeros ) is a symptom of an inefficient measurement apparatus : if one symbol is more prevalent than any other then the system is under - characterised by the alphabet in use . in the parlance of shannon s
theory of communication this statistical complexity is equivalent to the maximum rate of information . given that the collapse discussed above takes a slice out of the graph in figure [ pssnswtchsttstcclcmplxtvrssbs1 ] , we would expect sequences generated by certain poisson switches to be more difficult to characterize .
for example , sequences produced by a switch with @xmath83 have a statistical complexity of unity , but it is difficult to distinguish them from noise ( where @xmath84 ) because they are so close to the collapse at @xmath85 .
such a sequence , 5000 symbols long , was analysed up to a wordlength of @xmath86 at 100 equal intervals between @xmath87 and @xmath88 , without assuming any noise was present ( i.e. @xmath89 and @xmath90 ) .
the results are presented in figure [ ab49pssnswtchsttstcclcmplxt ] .
the plateau corresponding to the optimal model is that which has a statistical complexity of unity .
we can see that it is difficult to construct this model because the plateau is radically constricted at higher wordlengths .
on one side of it @xmath30 is too small to identify the equivalence classes , so every proword occupies its own equivalence class and @xmath91 , its maximum value at any wordlength @xmath11 . on the other side
, @xmath30 is too large , so the two equivalence classes collapse together , producing degenerate models that would output noise .
the @xmath92 plateau has a distinct end at @xmath93 because the sequence is not long enough to support analysis at a wordlength of @xmath94 .
at the latter wordlength statistical fluctuations in every proword profile mean that the correct classification of equivalence classes is no longer possible at any range of @xmath30 .
we are not too concerned about this here because we have already identified the optimal model which was stable from @xmath95 to @xmath93 .
in fact , a `` more optimal '' model would be able to predict the flipping of the switch itself to some extent .
the construction of such a model would probably need a lot of computation and would probably require @xmath4 to be very large .
these things depend on how random or noisy the switch is .
if a set of data is very complicated , no stable model might be identified before the wordlength becomes too large to be statistically supportable by the sequence length .
the only solution is to gather more data .
the alternative is to settle with models that are either inadequate or arbitrarily complicated .
although the latter models reproduce structure well ( and are therefore most useful to engineers ) , studying them can reveal little about underlying processes .
they are scientifically unaesthetic .
in contrast , one can tell a lot about the intricacies of a system from the minimal adequate model associated with it at a certain level of analysis .
this is the concern of scientists .
the next class of labelled diagraphs we consider produce binary sequences that are simple models of a process with bursts .
these sequences have the structure of sustained switches - that is , when the switch is down it has a constant probability of switching up , and when up , it stays up for a fixed count , @xmath96 .
when the sequence is unbiased the up - switching probability is @xmath97 .
see figure [ 4smblsstndswtchdgrph ] for an example of this kind of labelled diagraph .
the exact statistical complexities of such unbiased sustained switches are given by @xmath98 \nonumber \\ & = & \frac{1}{\ln(2)}\left[\ln(2u)-\left(\frac{u+1}{2u}\right)\ln(u+1)\right]\end{aligned}\ ] ] we now investigate the practical analysis of a sequence one million binary symbols long that was produced by a sustained switch with @xmath99 .
the statistical complexities of the models constructed by the analysis are shown in figure [ 4smblsstndswtchsttstcclcmplxt ] .
it can be seen that the first convergent values are at wordlengths one greater than @xmath96 .
that is to say , good models can be constructed when the analysis first has a greater memory than the system .
the plateau identifiable with a model of the form shown in figure [ 4smblsstndswtchdgrph ] begins at a wordlength of @xmath100 and extends laterally from @xmath101 to @xmath102 .
the remarkable thing about this plateau is that , although it is very flat , it is not _
entirely _ flat .
it begins at @xmath103 which is significantly higher than the theoretical statistical complexity of : @xmath104\approx1.54879 $ ] and subsequently oscillates around this value while it converges to it ( e.g. @xmath105 ) .
this behaviour is caused by the phase ambiguity due to the absence of information concerning the synchronisation of a burst when a word is composed entirely of `` up '' symbols .
for example , at wordlength six , the profile of word @xmath106 , ( i.e. , @xmath107 in binary ) , is a superposition of the profiles of sequences like @xmath108 $ ] , @xmath109 $ ] and @xmath110 $ ] , from each of which it can not be distinguished at that level of analysis .
therefore , in this case , the profile of word @xmath106 does not match that of any other word , and is allocated its own equivalence class . although the prevalence of this word , and thence its class , is very low , it is sufficient to distort the statistical complexity . in an analysis with recourse to infinite memory ,
the prevalence of an infinite sequence of `` up '' symbols is zero .
thus , the @xmath96 causal states of such a sequence would be correctly identified , and the statistical complexity of the model constructed would match exactly with the theoretical value .
of course , in practice no analysis can have infinite memory .
if one wishes to retain optimal predictability of future data then it is necessary to accept whatever model is actually constructed by an analysis with finite memory .
the test data examples analysed in the previous section represent signal types thought to be present in time series measurements of the geomagnetic field .
if this is true , we may expect to see similar structure emerging from a cm analysis of a real geomagnetic time series . the cm analysis detailed in section [ method ]
was performed on 3-hour averaged measurements of the variation of the east - west component of the geomagnetic field at halley , antarctica , from three separate years : 24 february - 16 december , 1995 , 26 january - 28 december , 1998 , and 2 january - 30 december 2000 . a graph of the data from 26 january to 28 december 1998 is shown in figure [ mgntcdflctn1998 ] .
it can be seen that the magnetic deflections have both a linear trend and a high frequency signal with an annual amplitude modulation that maximises in the austral summer .
the linear trend is caused by the movement of the ice shelf upon which halley is situated and was removed by subtracting the result of a linear regression for each of the three years .
the detrended time series was then binarised with respect to the median , giving three sequences of 2352 , 2688 , and 2896 symbols , respectively .
these series were then analysed up to a wordlength of 10 and with tolerances varying in 80 equal steps from 0.05 to 0.25 .
words with prevalences less than @xmath111 were eradicated .
the graph of statistical complexity , @xmath0 , is shown in figure [ 180mntmgntcdflctnsttstcclcmplxt ] .
two plateaus are evident , one at @xmath112 , covering a wide range of tolerances and between word lengths of 1 and 3 , and the other plateau at @xmath113 , at the top left - hand corner of the graph , between tolerances of about 0.05 and 0.07 and at word lengths of 8 or more .
the convergence of statistical complexity at a word length of 8 corresponds to a time scale of 8 x 3 = 24 hours .
such a diurnal variation is well known and is primarily caused by the rotation of the observing station with the earth under the so - called sq ionospheric current system that is driven by pressure gradients caused by solar heating and is thus fixed in the sun - earth frame @xcite .
the variation can be seen in the raw data , as illustrated by plotting a typical month of halley geomagnetic data in figure [ mgntcdflctn199802 ] .
the associated ground magnetic variation has neither a pure sinusoidal shape nor a fixed period of exactly 24 hours , that is likely to account for the higher observed statistical complexity of @xmath114 compared to the expected @xmath115 for a pure binary period 8 signal .
the other plateau in figure [ 180mntmgntcdflctnsttstcclcmplxt ] at word lengths of 1 to 3 indicates the presence of significant structure at 3 to 9 h time scales .
this plateau has a statistical complexity of approximately 0.9 and an overall structure similar to that of figure [ ab49pssnswtchsttstcclcmplxt ] , suggesting the possibility of some random pulse - like process .
such a possibility is intriguing because pulse - like geomagnetic perturbations on hour time scales ( known as magnetic bays ) are particularly prominent during the nighttime at high ( auroral zone ) latitudes and are associated with magnetospheric substorms @xcite whose occurrence has been argued to be a stationary poisson process with mean recurrence time of 5 h @xcite .
figure [ mgntcdflctn19980616 ] shows a single day of halley geomagnetic data that illustrates the presence of such pulse - like disturbances on hour time scales sitting on top of the diurnal variation . to investigate this further
, an analysis was made of a 40-minute averaged time series of the east - west component of the geomagnetic field at halley from 00:00 ut , 25 january , 1998 to 00:00 ut , 26 december , 1998 .
after removing the linear trend in the data due to the movement of the ice shelf , the time series was binarised with respect to the median , giving a sequence of 12011 symbols .
the series was analysed up to a word length of 11 for tolerances in 60 equal steps between 0.00 and 0.15 .
words with prevalences less than @xmath116 were eradicated .
the graph obtained for @xmath0 is shown in figure [ 40mntmgntcdflctnsttstcclcmplxt ] .
the plateaus in this graph are stable to variation of @xmath51 .
the higher plateaus have models that are more useful for prediction of future data , if they are stable to an increase in the amount of data available to the analysis .
the lower plateaus have models that show the most dominant structures and are easier to understand and interpret physically .
it can be seen from the graph that , at a tolerance between @xmath117 and @xmath118 , more structure is identified between wordlengths six and eight than it was possible to resolve with a memory of only five symbols . the model which corresponds to this plateau
is represented , for a wordlength of seven , in figure [ 40mntmgntcdflctndgrph ] .
the details of the model are in the appendix .
comparing with figure [ pssnswtchdgrph ] , the transitions between states 0 and 1 of this diagraph are an approximately poisson - switched process with a timescale of about five hours .
this value is given by the range of wordlengths capable of resolving this structure from the sequence within this range of @xmath30 ( @xmath119 ) ; at @xmath86 the characteristic timescale is @xmath120 .
it was thought that the other states and transitions in figure [ 40mntmgntcdflctndgrph ] would be caused by the diurnal variation of the data alone .
this was investigated by analysing , in exactly the same way , a pure binary sequence with a period of 36 symbols corresponding to a period of one day if each symbol were to represent a 40-minute average . the principal transitions of the model constructed for this sequence are shown in the diagraph drawn in figure [ prd36dgrph ] .
the structural similarities and differences between this diagraph and the one in figure [ 40mntmgntcdflctndgrph ] are obvious , and support the idea that the transitions between states 0 and 1 of figure [ 40mntmgntcdflctndgrph ] are due to substorm activity , rather than merely being an artifact of a partially characterised 24-hour period .
in the previous sections , we have demonstrated how cm can measure the statistical complexity of linear data sequences and construct the minimal model necessary to describe the data .
the reader may have noticed that there are seven degrees of freedom in making such a model : 1 .
digitisation method ( binary , trinary , etc ) 2 .
coarse - graining scale , @xmath2 3 . sequence length , @xmath6 4 .
wordlength , @xmath11 5 .
tolerance , @xmath30 6 .
corruption frequency , @xmath42 7 .
expletive frequency , @xmath51 these degrees of freedom express the level of information in the data and the depth of knowledge with which the model is probing the system from which the data are measured .
for example , increasing the sequence length , @xmath6 , reducing the coarse - graining scale , @xmath2 , or increasing the digitisation from binary to trinary , all provide increased information and thereby increased knowledge of the system that the data represent .
conversely , increasing the tolerance or the expletive frequency reduces information by admitting different states to be equivalent or to be omitted , respectively , thereby reducing knowledge of the system .
consequently , we might anticipate that the best model of the system is the model corresponding to the region of the multi - dimensional parameter space in which information is maximised .
whilst such a model is the most accurate description of the data sequence with the greatest information content , it is not necessarily the optimal model of the system .
this is because any data sequence is not a complete representation of the system it is measured from . in particular , it is limited in two important respects : first , there is structure in a data sequence , that we have termed noise , that can not be resolved under any amount of computation .
this will create differences in the profiles of words that are statistically insignificant and should be ignored by allowing some non - zero value of tolerance , corruption frequency or expletive frequency .
second , there is structure in a data sequence , that we have termed random , that has not been resolved at a certain level memory or wordlength but that is resolvable at a greater wordlength .
recognising these sources of structure , we advance a hypothesis about model construction : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ given enough data relevant to a system , there exists at least one connected region of finite size in the space of parameters used to construct models from those data within which an optimal model of the system is constructed .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in other words , meaningful models of the data can only be found within certain , usually finite , zones of the parameter space@xcite . within each zone , @xmath0 is constant and the model is both stable and minimal . outside this zone , the model is either too degenerate or overly complicated .
for example , it will be degenerate ( and @xmath0 will be too low ) if @xmath30 is set too large .
this is because equivalence classes will collapse into one another .
similarly , the model will be unnecessarily complicated ( and @xmath0 will be too high ) if @xmath30 is set too small .
this is because distinctions will be made between words on the basis of insignificant differences in their profiles .
an analogy is the construction of a vocabulary for the structure of speciation of feline animals .
if one is too fussy about the tail , manx cats can not be classed as domestic cats .
if one s sole criterion is purring or a meow , a lion cub may be misclassed as a domestic cat .
the correct classification of feline animals needs a finite amount of information to fall within the boundaries of a finite number of provisos . in the case of computational mechanics
we interpret effectively sofic models to be optimal .
thus we seek plateaus in the multi - dimensional parameter space .
generally , this space can contain many plateaus , the heights of which are the corresponding models statistical complexities .
if we want to forecast the data most accurately , we are looking for the highest plateau , which has the most stringent conditions@xcite .
more physically understandable models may exist on some lower plateaus where only the more dominant causal structures are preserved .
thus , in the end , the success of the analysis depends upon the existence of effectively sofic plateaus of statistical complexity in the multi - dimensional parameter space and our ability to discover them .
this is contingent upon the data that are supplied and how much computing power is available .
it is important to bear in mind that the data are not only a function of the physical system s behaviour , but also of the measurement apparatus and any pre - processing .
there are four main pitfalls ( represented by corresponding model parameters ) : 1 .
mischaracterization of the system by the measurement apparatus ( @xmath121 ) 2 .
degradation of data prior to the analysis by processing ( @xmath2 ) 3 .
insufficient data to resolve all structure present ( @xmath6 ) 4 .
insufficient computing power to resolve random structure ( @xmath122 ) the apparatus may easily mischaracterize the system , either by introducing structure to the data which is foreign to the system s behaviour , or by neglecting to transcribe structure that should be present .
this situation is most apparent when the apparatus is clearly only taking measurements from a cross - section of the system .
nevertheless , if it is reasonable to assume in a particular case that the apparatus is capable of providing a good representation , then identified structure can be attributed to the system . in such cases we would also expect the statistical complexity to scale with the system s true complexity .
naturally , this is not valid when the cross - section happens to be an exact sub - system .
although all processing degrades data , it may still be possible to correctly characterize all the structure present .
this is because the degradation will usually produce noise , which can be ignored .
a graver problem is when ( uncharacterizable ) noise represents some of the system s structure .
the only solution may be to collect more data , but other preliminary approaches are to use a finer scale when coarse - graining and/or to digitise more finely .
however , it is always necessary to choose sensible margins for the parameter search because some regions of the parameter space are computationally very costly to explore .
for example , a trinary sequence is about seven hundred times as hard to fully analyse at a word length of eight than a binary sequence .
you must have a good reason not to use binary .
an alternative approach may be useful when the data have a number of widely separated scales with structure ; it may be more computationally efficient to construct higher - level equivalence classes than to persist with using longer and longer words .
if this is the case there will be a change in the constant increase of @xmath0 with wordlength at the wordlength where the lower scale of structure is found , but the gradient will thereafter remain constant ( until the next scale of structure is reached ) .
classes on the next highest level are found by applying the same analysis method to the sequence expressed in terms of a set of primary level causal states for which @xmath0 has not yet converged .
all information between the scales @xmath123 and @xmath124 is lost in this process .
even so , it is a more preferable approach than simply further coarse - graining the data to intervals of @xmath123 if one has reason to believe that the system s degrees of freedom at the two scales are coupled .
the total statistical complexity is the sum of those calculated at each level , so it is in fact possible to test for such coupling by comparing the coarse - grained @xmath0 with the hierarchical value .
there is actually no reason why the prowords and epiwords should not come from different sequences , enabling the direct causal correlation of two systems , such as the solar wind and the magnetosphere .
computational mechanics is an intuitive and powerful way to study complicated linear outputs from physical systems .
this is because the analysis identifies causal structure from data presented to it and constructs the minimal adequate model that fits these data .
the information about this structure , and in particular its scales , can then be used to optimise more physically plausible models . in this paper
we have discussed in detail how the original formalism has to be used when applied to non - infinite sequences .
the main conclusion is that models constructed by computational mechanics are good if , and only if , they are stable to the variation of the parameters used to construct them from the data . in addition , various more general postulates and definitions are made .
these concern the general constructibility of models from a set of observations : 1 .
structure which can not be resolved from a set of data under any amount of computation is most usefully called _ noise _ 2 .
structure which has not been resolved at a certain level of computation or memory , but which is resolvable from the set of data is usefully called _ random _ 3 . given enough data relevant to a system , there exists at least one connected region of finite size in the space of parameters used to construct models from those data within which an optimal model of the system is constructed .
the prior undecidability of whether unresolved structure is noise or randomness is a direct parallelism of gdel s famous theorem . for a proof relating the two fields , but in a slightly different context , see g.j . chaitin@xcite .
the method developed in this paper was applied to magnetometer measurements of ionospheric currents for the years 1995 , 1998 and 2000 .
the technique successfully constructed models , the simplest of which comprised a diurnal component and a poisson - switched process with a timescale of about five hours that likely relates to the occurrence of magnetic substorms .
the most complicated model could be used to forecast space weather .
a similar method was also proposed to characterize the causal relationship of any two systems , such as the solar wind and the magnetosphere .
details of the simple stable model at wordlength seven , @xmath125 .
@xmath116 , at which value 114 of 128 words are cut .
+ word number ( class ) & word & probability + 0 ( 0 ) & 0000000 & 0.481382 + 1 ( 1 ) & 0000001 & 0.0759726 + 3 ( 1 ) & 0000011 & 0.0729505 + 7 ( 1 ) & 0000111 & 0.0641128 + 15 ( 1 ) & 0001111 & 0.0540634 + 31 ( 1 ) & 0011111 & 0.0546018 + 63 ( 1 ) & 0111111 & 0.0532191 + 64 ( 0 ) & 1000000 & 0.0338069 + 96 ( 0 ) & 1100000 & 0.0234980 + 127 ( 2 ) & 1111111 & 0.0514929 + word number ( class ) & word & probability + 0 ( 0 ) & 0000000 & 0.0423272 + 64 ( 0 ) & 1000000 & 0.0291209 + 96 ( 0 ) & 1100000 & 0.0209377 + 112 ( 0 ) & 1110000 & 0.0209523 + 120 ( 0 ) & 1111000 & 0.0293505 + 124 ( 0 ) & 1111100 & 0.0442261 + 126 ( 0 ) & 1111110 & 0.0822315 + 127 ( 2 ) & 1111111 &
0.684994 + word number ( class ) & word & probability + 0 ( 0 ) & 0000000 & 0.0789801 + 64 ( 0 ) & 1000000 & 0.0907960 + 96 ( 0 ) & 1100000 & 0.0945274 + 112 ( 0 ) & 1110000 & 0.103234 + 120 ( 0 ) & 1111000 & 0.108831 + 124 ( 0 ) & 1111100 & 0.105721 + 126 ( 0 ) & 1111110 & 0.101990 + 127 ( 2 ) & 1111111 & 0.268035 + class & average word & probability + 0 & [ 0.161 , 0.102 , 0.061 , 0.040 , 0.027 , 0.009 , 0.000 ] & 0.573587 + 1 & [ 0.000 , 0.142 , 0.288 , 0.432 , 0.603 , 0.797 , 1.000 ] & 0.374920 + 2 & [ 1.000 , 1.000 ,
1.000 , 1.000 , 1.000 , 1.000 , 1.000 ] & 0.051493 + class & average word & probability + 0 & [ 0.843 , 0.735 , 0.657 , 0.579 , 0.470 , 0.306 , 0.000 ] & 0.269146 + 1 & [ 0.000 , 0.309 , 0.444 , 0.590 , 0.714 , 0.906 , 1.000 ] & 0.045860 + 2 & [ 1.000 , 1.000 , 1.000 , 1.000 , 1.000 , 1.000 , 1.000 ] & 0.684994 + class & average word & probability + 0 & [ 0.885 , 0.752 , 0.614 , 0.463 , 0.304 , 0.149 , 0.000 ] & 0.684080 + 1 & [ 0.000 , 0.117 , 0.221 , 0.325 , 0.532 , 0.701 , 1.000 ] & 0.047886 + 2 & [ 1.000 , 1.000 , 1.000 , 1.000 ,
1.000 , 1.000 , 1.000 ] & 0.268035 + for the unbiased poisson switch , see the discussions of the random telegraph " in j.s .
bendat , principles and applications of random noise theory , john wiley and sons , new york , 1958 , and of the poisson switch in h.j .
jensen , self - organized criticality , cambridge university press , 1998 .
if we had been considering infinite sofic sequences this zone would be always be infinitely big because all structure would be resolved after a certain wordlength , and would continue to be resolvable .
this does not usually apply to finite sofic sequences because beyond a certain threshold wordlength statistical fluctuations in the profiles are able to destroy the equivalence classes .
engineers would usually settle for that model constructed by the highest level of analysis , whether of not it is on a plateau .
this is not acceptable from a scientific perspective because there is no way to justify that such models are not totally arbitrary .
|
we discuss how the ideal formalism of computational mechanics can be adapted to apply to a non - infinite series of corrupted and correlated data , that is typical of most observed natural time series
. specifically , a simple filter that removes the corruption that creates rare unphysical causal states is demonstrated , and the new concept of effective soficity is introduced .
we believe that computational mechanics can not be applied to a noisy and finite data series without invoking an argument based upon effective soficity .
a related distinction between noise and randomness is also defined : noise can only be eliminated by increasing the length of the time series , whereas the resolution of random structure only requires the finite memory of the analysis to be increased .
the benefits of these new concepts are demonstrated on simulated times series by ( a ) the effective elimination of white noise corruption from a periodic signal using the expletive filter and ( b ) the appearance of an effectively sofic region in the statistical complexity of a biased poisson switch time series that is insensitive to changes in the wordlength ( memory ) used in the analysis . the new algorithm
is then applied to analysis of a real geomagnetic time series measured at halley , antarctica .
two principal components in the structure are detected that are interpreted as the diurnal variation due to the rotation of the earth - based station under an electrical current pattern that is fixed with respect to the sun - earth axis and the random occurrence of a signature likely to be that of the magnetic substorm . in conclusion ,
a hypothesis is advanced about model construction in general .
|
the large radio nebula w50 , also referred to as g39.72.0 and the `` manatee nebula '' , is both unusual and of an undetermined origin .
the nebula appears to be interacting with the central compact source and microquasar ss433 the first known stellar source of relativistic jets @xcite .
the w50/ss433 system was first identified at radio wavelengths by @xcite .
initial suggestions for the system s origin were made by , e.g. @xcite , which identified ss433 as a new class of stellar remnant ( the first - known galactic microquasar ) and proposed that the surrounding radio nebula of w50 was the remnant from a supernova explosion .
the nebula is widely catalogued and considered as a supernova remnant ( snr ; * ? ? ?
* ) , however this origin is still uncertain and debated .
while the circular shell of w50 is reminiscent of shell - type snrs , the additional elongated morphology is completely anomalous .
this elongation is at least partly due to the interaction between ss433 and w50 , with `` blobs '' of material being ejected at @xmath7 from the central engine of ss433 by precessing jets that are aligned along the east west axis of elongation ( e.g. * ? ? ?
the elongated morphology of the radio nebula could therefore have been formed by ( i ) _ a wind+jets model _ : with the ss433 jets expanding into a swept - up interstellar medium ( ism ) a stellar - wind bubble that was itself created either entirely by the jets of ss433 or even by a continuous wind from the ss433 binary @xcite , or alternatively by ( ii ) _ a snr+jets model _ : with an encounter between the jets from ss433 and an already formed snr , with the ram pressure from the jets ` punching ' through the shell to form the ears . what is certain is that the w50/ss433 system whether a wind - blown bubble or a snr is an anomalous object
. it could be a one - of - a - kind , and could also form part of a rare community of other peculiar galactic radio nebulae @xcite . to further complicate matters , the morphology of the system is also affected by interactions between the swept - up media and the surrounding environment which consists of a dense surrounding interstellar medium ( ism ) and an ambient interstellar magnetic field .
these environmental factors are important for w50 , as the west of the nebula appears in projection as dipping into the galactic plane likely resulting in a density gradient across the source and expansion into an inhomogeneous ism .
in addition , previous attempts at modelling w50 have largely neglected the impact of magnetic fields on the evolution of the nebula .
the large - scale magnetic fields have likely played a crucial role in the development of w50 independent of the chief mechanism that drives the object s formation as the interaction of magnetic fields is known to influence the evolution of wind - blown bubbles ( e.g. * ? ? ?
* ) , snrs ( e.g. * ? ? ?
* ) , and other radio nebulae ( e.g. * ? ? ?
moreover , previous observations have revealed regular magnetic fields , including a strong field in the rim to the north .
the polarization angle of this region is indicative of the ordered field strength being dominated by standard compression of an interstellar field that is oriented approximately east west , with the radio ears that are generally believed to be punched out by ram pressure from the jets of ss433 also approximately aligned along this same axis .
this further complicates understanding the evolution of w50 s morphology since , for example , a snr evolving in a strong ambient field is known to become elongated along the field direction @xcite , while in a wind - driven nebula a stellar wind with frozen - in toroidal fields can evolve into collimated outflows @xcite .
the presence of magnetic fields may also influence the collimation of the jets of ss433 ( e.g. * ? ? ?
* ) , which itself has likely had a strong influence on the nebula s evolution .
furthermore , more complex models essentially consisting of either _ snr+wind+jets _ or _ wind+snr+jets _ could also influence the shape of w50 ( e.g. * ? ? ?
models have recently been proposed for generating asymmetric supernova remnants ( e.g. * ? ? ?
* ) , and even off - centred or asymmetric supernova explosions within a pre - existing wind - driven bubble could give rise to a morphology that resembles that of w50 .
attempting to separate the magnitude of each of these multiple effects on the elongation of w50 ultimately requires a multi - wavelength study .
for example , diffuse h@xmath1 emission is a reliable tracer of the ionised hydrogen content / thermal electrons in the surrounding environment , which can depolarize synchrotron radio emission emitted at a similar or greater distance along the line - of - sight and can thereby conceal the presence of ordered magnetic fields in the nebula ( e.g. * ? ? ? * ) .
furthermore , the presence of optical filamentary emission associated with the central region of the nebula would most likely originate from shock - excited material in a snr with the surrounding ism being heated by the outgoing shock wave , and producing a set of filamentary shell - like structures that occur due to cooling behind the shock front .
this shock - excited optical emission results in both filamentary and diffuse faint h@xmath1 emission , while the presence of diffuse optical emission can also be associated with a wind - blown bubble @xcite .
a filamentary shell - like structure in h@xmath1 is therefore one way to distinguish a snr from a wind - driven bubble .
however , neither of these scenarios are mutually exclusive .
nevertheless , the discovery of such optical emission would allow for the optical spectra to be obtained thereby providing a direct test of shock physics in the object , and potentially allowing for a definitive experiment to determine the origin of w50 ( e.g. * ? ? ?
in addition , if new optical filaments can be found , they would help us to understand the connections between the filamentary emission and the charged particles and ordered magnetic fields within w50 .
if shocked optical filaments exist in w50 , do any of them have coincident synchrotron emission ?
are the magnetic fields more strongly ordered in such filaments ? to what extent does the creation of optical filaments influence the orientation of the magnetic fields ? and
conversely , to what extent do the magnetic fields influence the creation of the optical filaments ? in this paper , we set out to understand the origin and formation of w50 , by performing a needed multi - wavelength study
obtaining new broadband 1 to 3 ghz radio data and complementary 1 arcsec spatial resolution h@xmath1 images that allow us to investigate both the large - scale magnetic fields and the ionised hydrogen content .
the paper is structured as follows : in section [ backgroundw50 ] we provide extensive background information and review of the much - studied w50/ss433 system , in section [ datareduction ] we discuss the technical details of the data processing , in section [ w50 ] we present and analyse our new multi - wavelength results , in section [ discussion ] we provide a discussion of our findings , and in section [ conclusions ] we provide our conclusions . throughout , we assume a distance to ss433 ( and by association w50 ) of @xmath8 kpc @xcite .
this distance is still a source of some debate ( e.g. * ? ? ?
the total intensity spectral index , @xmath1 , is defined in the sense @xmath9 . when referring to ` polarization ' , we refer to linear polarization unless otherwise specified .
the radio nebula of w50 ( g39.72.0 ) has been studied thoroughly since its identification in the radio catalogue of @xcite , with a comprehensive literature dedicated to the object .
w50 is a non - thermal galactic source of large angular extent ( @xmath10 ) .
the shape of the object has been compared to that of a conch shell ( e.g. * ? ? ? * ) and of an interstellar manatee . at radio wavelengths ,
the morphology of the nebula of w50 is seen in projection as an approximately - circular shell that is @xmath11 in diameter with the bright compact point source ss433 located close to the geometric centre .
there are two bright radio ears ( or ` ansae ' ) to the east and west of the nebula , and the entire radio structure appears elongated and aligned along this east west axis with the entire nebula covering @xmath10 on the sky . the western edge of the nebula dips into the galactic plane , as the system is at a galactic latitude of @xmath12 .
the surrounding field of view contains several bright extragalactic radio sources , and a nearby hii region s74 to the nw .
w50 is one of the largest galactic supernova remnants .
such objects are poorly studied at high angular resolution due to their overall angular size and the considerable number of pointings required for full imaging at radio wavelengths .
it has been shown that the nebula can be reproduced to a good approximation by the combined effects of the evolution of a snr shell and its interaction with the precessing jets from ss433 @xcite . while an accumulating body of evidence tends to be in favour of a snr origin ( see e.g. * ? ? ?
* and references therein ) , this is still a source of debate with some studies suggesting it is more likely a stellar - wind / interstellar bubble ( see section [ intro ] ) .
the object is certainly in a unique class of outflow - driven objects ( e.g. * ? ? ?
the progenitor that formed ss433 is typically presumed to have exploded @xmath13@xmath14 years ago @xcite .
the morphology of w50 hints at an approximately circular explosion , with the `` ears '' having been `` punched '' through the rim by the relativistic jets moving outwards from ss 433 at 0.26@xmath15 .
importantly , the contributions from an initial sn explosion , that probably preceded formation of the compact object ss433 , have not been convincingly distinguished from the impact of the jet and wind activity of the central compact system @xcite .
recent studies in the optical and x - ray wavebands possibly suggest at least two distinctly different states of jet activity @xcite . optical emission was first associated with w50 in the 1980 s @xcite , with filaments located @xmath16 arcmin to the e and w of ss433 , although none were located within the central region of the nebula itself .
diffuse x - ray emission is also coincident with the optical filaments .
the extended x - ray lobes or jets extend to the e and w along the precession axis of the radio jets , with the x - ray emission peaking in the region of the optical filaments @xcite .
there has also been additional soft x - ray emission discovered in the ear to the e , coincident with the radio ear and associated with the terminal shock of the ss433 jets @xcite .
it has been found that the jet terminates in a ring - like termination shock .
no x - ray emission is found in the ear to the w , suggesting an inhomogeneous surrounding medium .
both the x - ray emission discovered in the lobes , and the presence of the optical filaments , strongly implies a density enhancement to the e and w at a radius of @xmath16 arcmin from ss433 .
this is consistent with the traced out shell that is visible at radio wavelengths to the n and s , and together with the optical filaments being oriented perpendicularly to the radio jets , implies an interaction between the jets with a preexisting shell .
@xcite reported the first faint optical emission associated with the central region of the nebula itself , albeit in only a small rectangular region of @xmath17 arcmin width and height .
the identification of more extensive optical filaments , especially if there was good correlation between the radio structure and the filaments , would lend increasing support to a _
snr+jets _ hypothesis .
such emission is challenging to observe , as it is expected to be faint , and possibly not visible at all due to dust extinction . to observe faint h@xmath1 emission
also requires a continuum - correction .
new sensitive and high - resolution surveys such as the isaac newton telescope photometric h@xmath1 survey of the northern galactic plane ( iphas ) may feasibly allow for their detection @xcite . as mentioned in section
[ intro ] , the large - scale magnetic fields in w50 are likely playing an influential role in the development of the object .
the chief method with which to study cosmic magnetic fields is via radio polarimetry .
observations of the linear polarization in radio nebulae can provide information on the physical processes driving the emission and propagation of synchrotron radiation . in the case of w50 , such observations have traditionally been complicated by the object s large angular size , although contemporary analyses have now been applied to numerous angularly - large astrophysical systems ( see e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) .
as polarization is a direct tracer of magnetic field structure , its detection can enable investigations of the order and orientation of fields .
this is particularly important in shocked regions which compress and order magnetic fields . in particular ,
the interaction between snrs and magnetic fields is a crucial one , as the shocked material from a sn explosion compresses fields that are frozen - in to the plasma .
furthermore , magnetohydrodynamic turbulence can give rise to rayleigh taylor instabilities in the sn shell that allows for a distinction between younger and older objects @xcite .
in the case of w50 , the magnetic fields may also be influencing the jets of ss 433 .
measurements of the faraday rotation , in the form of rotation measures ( rm ; see for a full review ) , are strong probes of the magnetic field structure in any foreground medium along the line - of - sight and across the face of a radio nebula , allowing us to distinguish between the presence of coherent and random fields . furthermore , detection of a rm gradient across the source and any resulting depolarization can allow us to infer the geometry and also the anisotropy of the field which could be caused by either turbulent or systematically varying regular fields .
such scenarios are presently unconstrained for w50 .
while previous studies at radio wavelengths have shown that there is an inhomogeneous distribution of polarization from w50 , there is no understanding of whether this inhomogeneity could be the result of a very turbulent and randomly - distributed magnetic field internally to the nebula itself , depolarization in the surrounding environment of w50 , or due to unassociated magnetoionic material that is located in the foreground .
furthermore , the previous rm measurements of @xcite used only two data points at different frequencies and were also subject to the well - known @xmath18-ambiguity .
in addition , the limited sensitivity of these observations resulted in gaps where the polarization was too low to calculate a reliable rm across the entire eastern part of the central region , and several additional large gaps in the northern , western , and southern parts of the shell , and also the eastern ear .
no spectropolarimetry of w50 using new less - ambiguous techniques , such as rm synthesis , has ever been performed despite considerable technical advances in the handling of such data .
the rms can be determined using the high sensitivity , broadband , and improved resolution observations now available with contemporary instruments such as the australia telescope compact array ( atca ) allowing us to understand the magnetic field geometry within the object , and how this varies across different parts of the nebula .
such a magnetic field study is a useful tool for separating different contributions to the evolution of the object , as distinct patterns are expected to be imprinted on the observed rm in both the case of a _
wind+jets _ @xcite or a _ snr+jets _ origin @xcite .
there have been numerous previous studies of the w50 radio nebula at radio wavelengths .
and @xcite observed the total intensity and the polarized emission of w50 at 1.7 ghz , 2.7 ghz , and 5 ghz with the 100-m effelsberg radio telescope .
the brightest polarization was found to be associated with the northern and eastern edges at 1.7 ghz , with fractional linear polarizations up to @xmath19% . at 2.7 ghz , these areas are @xmath16% polarized , with the brightest polarization emanating within the roughly circular region of size @xmath20 degree centred on ss433 , and from the extreme eastern end of the source . at 5 ghz , the northern edge is @xmath21% polarized , with a magnetic field direction approximately tangential to the shell typical of old snrs @xcite .
this bright northern edge therefore appears to be a typical example of compression of the large - scale ambient interstellar magnetic field .
the region that connects the central region and the eastern ear , and several other areas within the object were all found to be completely unpolarized .
the entire western ear was also found to be completely unpolarized even at a resolution of 30 arcsec @xcite .
what creates this inhomogeneous distribution of polarization across the radio nebula is still unknown , and it could be the result of a very turbulent and randomly - distributed magnetic field internally to the nebula itself , depolarization in the surrounding environment of w50 , or due to unassociated magnetoionic material that is located in the foreground .
using the urumqi survey at 5 ghz , also found strong polarized emission along the northeastern shell of the central part of w50 .
they derived an integrated spectral index of @xmath22 , consistent with that found by @xcite .
they find that measurements between 6 cm and 11 cm gave @xmath23 for the central part , and @xmath24 for the eastern ear .
the eastern ear therefore appears to have a slightly steeper spectrum than the central region , consistent with the finding of @xcite .
meanwhile , interferometer - based studies such as that by @xcite tend to find that the spectra flattens towards the western ear with @xmath25 and steepens towards the eastern ear with @xmath26 .
these spectral index variations are likely due to stronger shocks and higher compression to the west where the shock is encountering a denser ism , and consistent with the location of the galactic plane .
this interpretation is also consistent with the depolarization , which is less in the eastern part of the source and greater across the rest of the object possibly implying variations on smaller angular scales due to increased structure in the magnetic field towards the west .
the flattest spectra of all appear to be located in the southwestern rim , possibly as low as @xmath27 .
in addition , @xcite also find other small features such as a `` chimney '' radiating away from the eastern ear of w50 .
many of these features are also still not fully understood .
the spectral index and rm distributions of the very brightest emission were studied in detail by @xcite , and two - frequency estimates were made of the depolarization and faraday rotation ( albeit subject to @xmath18-ambiguities , with low sensitivity , and with an angular resolution of 3 arcmin ) .
rms were found to vary from 0 rad m@xmath3 to as high as 240 rad m@xmath3 .
the rm in the eastern part of the source is either @xmath28 rad m@xmath3 or ( @xmath29@xmath30)@xmath31 rad m@xmath3 .
in addition to @xmath18-ambiguities , the limited sensitivity of these observations resulted in gaps where the polarization was too low to calculate a reliable rm across the entire eastern part of the central region , and several additional large gaps in the northern , western , and southern parts of the shell , and the eastern ear .
the rms of these regions can be determined using the high sensitivity , broadband , and improved resolution observations now available with contemporary instruments such as the atca ( e.g. * ? ? ?
* ) allowing us to understand the magnetic field configuration within the object , and how this varies across different parts of the nebula .
sensitive polarimetry has the potential to separate the extent to which the morphology of w50 is dominated by extrinsic effects or coupling between the jet / remnant .
if the ear - like structures to the e & w do originate due to the jets of ss433 , then there should in principle be consistent variations in the spectral index , fractional polarization , and rm in both sides of the source allowing us to distinguish between competing scenarios for the remnant s formation such as the _ wind+jets _ and the _ snr+jets _ models , as described in section [ intro ] . as discussed in section
[ intro ] , filamentary h@xmath1 emission associated with large - scale radio nebulae can allow distinction between competing models for an object s formation .
an optical filamentary nebulosity has already been identified in the w50/ss433 system @xcite , that is visible in both h@xmath1 and sii .
these bright filaments are located in the eastern and western ears , and are aligned perpendicular to the axis of the jets from ss433 . as the relativistic jet from ss433 appears to have broken through the edge of the central region of w50 , it is of interest that this bright optical emission lies at these breakout regions and is therefore likely a consequence of the jet activity .
these breakout regions have been discussed further by @xcite . in an ideal case ,
the breakouts are rings of emission that occur where the expanding jet envelope shocks a dense shell of swept - up and compressed material at the boundary of the circular region of w50 .
optical emission should then not be observed throughout the circular region , which is too hot and rarefied , and with the ambient medium being of too lower density in front of the jet . as the jet axis of ss433
is tilted with the eastern edge pointing towards us @xcite , this orientation modifies the rings of optical emission into ellipses .
it has been argued that the observed arc morphology is due to heavy , patchy foreground interstellar absorption , so that only the component of the ring nearer to us along the line - of - sight remains visible @xcite .
however , such a mechanism would likely require absorption within the nebula , as purely foreground material should affect both the front and back of the ring .
while this overall picture is consistent with the observed arc morphology , it does not account for the known ` corkscrew ' structure of the jet close to ss433 , which is also markedly imprinted onto the radio features of the eastern ear ( e.g. * ? ? ?
until recently , there was no known optical emission associated with the central region of w50 . however , a small rectangular region of @xmath17 arcmin width and height was observed by @xcite , who found faint optical emission associated with this central region of w50 .
the observed region was not large enough to make definitive statements about the formation of the object , or to connect this optical emission with the radio morphology .
these results suggest that a deep optical mosaic across the entirety of w50 would help to locate additional optical filaments .
filamentary h@xmath1 emission tends to be produced by shocks ( see section [ intro ] ) , and has been identified in several snrs .
the _ snr+jets _ model would suggest that there would be faint filaments that cover all of the circular portion of the object , and which due to different pressures would be expected to be considerably fainter than the previously identified , jet - driven , optical emission . identifying a network of such optical filaments would lend significant support to the _ snr+jets _ model , and future studies would be able to use any new filaments to obtain in - depth information on the kinematics throughout the shell thereby allowing an experiment to determine the formation history of w50 .
such observations are complicated in that they require a large optical mosaic that deals with continuum - contamination , and are also hindered by possible interstellar dust extinction towards the galactic plane .
nevertheless , many of the observational challenges have already been overcome ( see section [ halpha ] ) . from the h@xmath1/h@xmath32 ratios
, it has also been inferred that there is likely patchy absorption of optical emission across w50 @xcite consistent with the distance and galactic latitude .
while the presence of foreground dust that heavily obscures the optical emission is potentially problematic , both the patchiness and the small piece of faint emission in the interior of w50 suggest there may be new fainter filaments waiting to be discovered .
we obtained full spectropolarimetric observations of the w50/ss433 system using the atca .
we collected six days of data , using three different hybrid array configurations ( h75 , h168 , and h214 ) which are all compact and well - suited for observing at low declinations , with baseline lengths distributed between 31 to 230 m. we used the cabb system @xcite , which provided 2 ghz of instantaneous bandwidth centred at 2.1 ghz in 2048 spectral channels .
as w50 is above the elevation limit of atca for @xmath339h 49 m per day , we used six full - track observations consisting of 66 hours in total which included overheads for standard calibration , full parallactic angle coverage of a leakage calibrator , and antenna slew times ( i.e. @xmath34 hours in each hybrid configuration ) .
our @xmath0 mosaic consisted of 198 pointings . in order to obtain an approximately uniform sensitivity pattern across w50 , and to limit instrumental effects ,
the pointings were arranged in a hexagonal pattern such that they nyquist - sample the sky at the high - frequency end of the band .
details of the observations are provided in table [ table1 ] .
@xmath35 in order to measure the large - scale magnetic fields , we only require the detection of diffuse extended linearly polarized emission .
this requires only low - angular resolution data that have sensitivity to large angular scales , i.e. short baselines .
note that observations of a large galactic object such as w50 can be well - suited to a radio interferometer , rather than a single dish , as the former naturally filters out the very large - scale galactic emission that would otherwise require subtraction and would thereby affect our interpretation .
the observing strategy here is appropriate for our scientific questions .
the hybrid configurations at the atca are ideal for this , as they are nominally sensitive to a largest angular scale of @xmath36 arcmin in mosaiced total intensity , while simulataneously providing a synthesised beam of @xmath37 arcsec to @xmath38 arcsec across the band .
mosaicing is also a standard procedure with the atca .
the excision of narrow band radio frequency interference ( rfi ) and calibration were carried out using standard techniques in the miriad package .
data were loaded into miriad and flagged for known bad channels and edge effects using the automated flagging routine pgflag .
the default settings were found to discard significant quantities of good data , and were adjusted to provide several iterative loops of flagging and calibration using customised parameters allowing for lighter flagging in each loop . due to the compact configurations used ,
flags were also applied to remove shadowed antennas . during the data reduction ,
the fixed antenna ca06 was also flagged throughout the observations and was thereby not included during the imaging stage in order to avoid a very significant gap in the @xmath39-coverage .
a standard initial calibration was performed to correct for the effects of the bandpass and the complex antenna gains using the tasks mfcal @xmath40 gpcal @xmath40 gpcopy to transfer solutions from the flux calibrator ( pks1934 - 638 ) to the phase calibrator ( j1859 + 129 ) .
the frequency - dependent instrumental leakages were then calculated in 16 bins across the full bandwidth using gpcal and the phase calibrator which was observed over a large range in parallactic angle .
the absolute flux scale was bootstrapped using gpboot , and the final solutions then linearly interpolated across the 198 pointings of the target source ( w50 ) .
all of the sources ( including calibrator scans and target pointings ) were then gently flagged ( being careful to avoid over - flagging ) before iteratively performing a new calibration loop ( which provided improved solutions due to the removal of bad data ) .
once flagging was complete , a final calibration loop was then performed to provide the complete dataset .
no self - calibration was performed .
channels at the edges of the bandpass were discarded . due to the rfi environment , which at atca is selectively worse at lower - frequencies , it was found that the usable region of the band extended from 1.4 ghz to 3.1 ghz .
channels outside of this range tended to have very limited @xmath39-coverage , particularly on shorter baselines .
high - quality mosaic imaging of this field of view is complicated , as it contains diffuse extended emission from w50 , compact sources embedded in diffuse emission , and large - scale diffuse emission from the galaxy .
the field of view is also at a low declination ( @xmath41 ) , has imperfect and limited @xmath39-coverage ( as shown in fig .
[ figuvcoverage ] ) , and was observed with a broad 2 ghz bandwidth .
two techniques were therefore used for separate wide - field , wide - band imaging of both the total intensity ( stokes @xmath42 ) and the linearly polarized intensity ( stokes @xmath43 and @xmath44 ) emission . to image stokes @xmath42 ,
the common astronomy software applications ( casa v4.2.0 ) package was used to process the wide - field , wide - band data , as the package implements the full - mosaic , multi - frequency synthesis algorithms that are now available .
the mosaic imaging was not able to simultaneously solve for the spectral index in the current implementation of casa , as mosaics can currently only be processed with nterms=1 .
however , the incorporation of the large 2 ghz bandwidth substantially improved the @xmath39-coverage available for imaging the complex field of view albeit with some frequency - dependent artefacts appearing around very bright sources such as ss433 .
as the mosaic imaging process employed in casa combines the individual pointings in @xmath39-space , we are essentially imaging a single @xmath45 field - of - view ; we therefore also used the @xmath46-projection algorithm with 256 terms .
subtraction during deconvolution using the clean algorithm in casa also takes place in the @xmath39-plane , allowing for a more accurate clean , and the hogbom - type deconvolution was performed slowly with a very - low loop gain , 0.005 , in order to avoid confusing emission and sidelobes when extended emission was present .
a high cyclefactor of 5 was also used in order to complete major clean - cycles more often .
the final image has a resolution of @xmath47 arcmin with a position angle of @xmath48 . to create stokes @xmath43 and @xmath44 image datacubes ,
the miriad package was used , which is specifically designed to handle data with linear feeds ( i.e. @xmath49 , @xmath50 , @xmath51 , and @xmath52 ) such as that from the atca .
deconvolution of the images from the synthesised dirty beam took place in the image - plane .
the data were imaged using multi - frequency synthesis of both 10 mhz and 100 mhz channel - subsets in order to balance maintaining good @xmath39-coverage while also minimising bandwidth depolarization . due to the large observational frequency range ,
the bandwidth depolarization varies across the band . for the 10 mhz channel dataset , significant bandwidth depolarization , where @xmath53/@xmath54 are the intrinsic polarization
] is expected for rms @xmath551,367 rad m@xmath3 at 1.35 ghz and for rms @xmath5519,965 rad m@xmath3 at 3.3 ghz .
for the 100 mhz channel dataset , significant bandwidth depolarization is expected for rms @xmath55137 rad m@xmath3 at 1.35 ghz and for rms @xmath551,997 rad m@xmath3 at 3.3 ghz .
note that the 100 mhz dataset has improved @xmath39-coverage relative to the 10 mhz dataset , whereas the 10 mhz dataset is more sensitive to high rms .
the data were imaged with a robust setting of @xmath56 , giving moderately uniform weights and avoiding excessive weight being given to visibilities in relatively sparsely - filled regions of the @xmath39-plane .
an image of size 168 pixels was made for each pointing before the mosaic operation was performed , and each image was over - sampled , with a cell - size of 15 arcsec .
the images were deconvolved from the synthesised dirty beam using mossdi , and a gain of 0.01 .
a constant restoring beam was used for all images across the band , using the best angular resolution at the low - frequency end of the band of @xmath57 arcmin with a position angle of @xmath58 .
the effects of the primary beam were included in both casa and miriad .
the mosaicing process in casa includes the effects of the frequency - dependent total intensity primary beam . in miriad ,
a frequency - dependent primary beam model ( the ` atca16 ' model ) for the cabb feeds was used .
in addition , the stokes @xmath43 and @xmath44 images are also affected by the off - axis instrumental polarization beam . with cabb
, this instrumental leakage has been estimated to be as large as @xmath59% at the half - power point @xcite .
we here make the critical assumption that this instrumental polarization is oriented radially outwards from the phase - centre so that the altitude azimuth mounts will cancel out instrumental effects via vector averaging over the large parallactic angle range .
the mosaicing procedure will further reduce the wide - field polarimetric aberrations by a factor of at least @xmath60 , and will be aided further as we have closely - packed ( better than nyquist - sampled ) mosaic spacings .
wide - field polarization effects are therefore expected to be very small within the central portion of the mosaic , while strong linear polarization at the edges of the mosaic is likely spurious .
for the linear polarization data , we do not need to image the entire band simultaneously into one image we only need to produce a stokes @xmath43 and @xmath44 datacube of narrow - band channels as a function of frequency , so that we can fourier transform the data and measure the peak faraday depth using the technique of rotation measure synthesis .
the observations allow for rm synthesis across a bandwidth that to our knowledge is unprecedented for any large radio nebulae .
the large bandwidth provides a rm spread function ( rmsf ; the point spread function in faraday space ) of fwhm 107.2 rad m@xmath3 , as shown in fig .
[ figrmsf ] .
no weights were applied during rm synthesis .
the few gaps in the frequency - coverage also provide low sidelobe levels within the rm range in which we are primarily interested .
we do not deconvolve the data from the rmsf using the technique of rm - clean @xcite .
as we are sensitive to different angular scales in our @xmath42 and @xmath43/@xmath44 images respectively , we can not calculate accurate polarized fractions or correct for the effects of spectral index ( see section [ limitations ] ) .
this prevents accurate rm - deconvolution , but is only an amplitude effect , i.e. our estimates of rms are unaffected .
in addition , the technique of rm - deconvolution has been found to have several limitations and makes many simplifying assumptions @xcite we therefore avoid all issues associated with deconvolution techniques .
the full faraday structure of the remnant can likely only be obtained with @xmath61-fitting , to include the effects of depolarization across the band although this has not yet been applied to extended objects ( e.g. * ? ? ?
* ; * ? ? ?
full @xmath61-fitting is beyond the scope of this current work , and would likely be required to fully separate the magnetic field contributions that arise both within the local environment of w50 and along the line - of - sight in any intervening faraday rotating media .
nevertheless , other properties such as the depolarization and correlations with other structural features of the radio nebula can also allow elucidation of these properties .
using our data , we are in principle also able to detect faraday thick structures along the line - of - sight ( i.e. as @xmath62 ) . we define a faraday thick source as those with @xmath63 , where the extent of the source in faraday depth , @xmath64 , is given by @xmath65 .
our atca data are therefore sensitive to thicknesses @xmath66 rad m@xmath3 at 3.1 ghz , and to @xmath67 rad m@xmath3 at 1.4 ghz
. a line - of - sight can therefore appear to be subtly extended in faraday space , relative to the fwhm of the rmsf@xmath68 rad m@xmath3 , however we can not _ resolve _ faraday thick structures .
furthermore , with a noise level in our faraday cubes of 0.125 mjy beam@xmath69 rmsf@xmath69 , this is the most sensitive mosaic of w50 that has been created in linear polarization . due to the large observational bandwidth , there is substantial variation in the @xmath39-coverage across the band , as shown in fig .
[ figuvcoverage ] .
this is problematic for consistent imaging of the very extended emission in the total intensity images .
as shorter @xmath39-spacings are better sampled at lower observational frequencies , the extended emission is both better reconstructed and has less missing flux at these low frequencies .
this tends to provide overly steep spectral indices that vary with position in the image depending on the scale of the emission .
this overestimation of a steep spectral index due to frequency - dependent @xmath39-coverage is an inherent property of ultra broadband observations that is potentially generally overlooked . due to the limited number of baselines , attempts at imaging with a constant @xmath39-range across the band produced artefacts in the final images .
in addition , the capability to perform both mosaic imaging and to simultaneously obtain a spectral index is not yet available ( see section [ imaging ] ) , and our data can not therefore be easily used to recover the spectral index .
the recovery of spectral indices is particularly difficult for this field of view as the western edge of the nebula dips into the galactic plane , with increased diffuse background emission .
this background emission changes across the band , along with dc offsets in the images , and also causes low - level frequency - dependent changes in the antenna system temperature .
in addition , inaccuracies in the primary beam model and its frequency dependence can also affect the spectral index
although the closely - packed mosaic and the applied primary beam model should minimise such an effect . furthermore , there is also increased rfi at the low - frequency end of the cabb band which selectively tends to have a larger effect on shorter baselines .
the flagging of these rfi affected data can therefore tend to remove larger scale structures at the low - frequency end of the band .
the ability to use techniques such as @xmath70@xmath70 plots ( e.g. * ? ? ?
* ; * ? ? ?
* ) and spectral tomography @xcite is limited as we have multiple narrow channels across our band .
attempts to reimage our data as two separate channels and to apply these techniques did not provide consistent results as the @xmath39-coverage still varies significantly .
additional observations to more fully sample shorter spacings in the @xmath39-plane would assist in recovering all of the extended total intensity emission across the band however another facility is likely required , as with the atca hybrid arrays we are essentially sampling the shortest spacings that the interferometer can provide ( although rfi preferentially affects the very shortest baselines ) .
these issues also prevent us from calculating reliable polarization fractions , @xmath71 , since different scales , negative bowls , and background emission are retrieved in both the @xmath72 and @xmath42 images .
fortunately , these effects are much less significant for our polarization images , as the emission is only ever a fraction of the total intensity emission , and is also broken up into much smaller scale structures with the polarized emission being separated into individual @xmath43 and @xmath44 components , and also being broken up in the plane of the sky due to polarization angle variations ( and faraday rotation ) .
since we use rm synthesis to recover the peak in the faraday spectra , any such residual amplitude effect will not affect the derived rms and will only affect the measured polarized intensity at a given rm , i.e. it is only an amplitude effect in faraday space . in cases of multiple faraday components , the highest amplitude component could feasibly be misidentified , although we have not identified such multiple components in our data .
we can therefore obtain reliable rms despite variations in @xmath39-coverage across the band and without having to assume a spectral index during rm synthesis .
we created a large continuum - corrected h@xmath1 mosaic covering the full @xmath73 region surrounding w50/ss433 , using data release 2 ( dr2 ) observations from iphas @xcite .
iphas is a 1800 deg@xmath74 ccd survey of the entire northern galactic plane in the latitude range
@xmath75 < @xmath76 < @xmath77 .
the data obtained are wide field camera images in the h@xmath1 narrow - band , and sloan @xmath78 and @xmath79 broad - band filters .
the survey is particularly well - suited to studies of spatially resolved nebulae , which indicate ionisation of circumstellar gas , as demonstrated by previous studies of a continuum - subtracted mosaic image of shajn 147 ( s147 ) , a @xmath80 diameter supernova remnant @xcite .
the observations were carried out between 2003 and 2012 with a median seeing of 1.1 arcsec ( sampled at 0.33 arcsec pixel@xmath69 ) and to a mean @xmath81 depth of 21.2 ( @xmath82 ) , 20.0 ( @xmath83 ) , and 20.3 ( h@xmath1 ) in the vega magnitude system @xcite .
imaging of extended nebulae requires a well - behaved background on the ccds , and the limitations of such a technique are largely due to background variations .
the sky generally subtracts fairly well in an h@xmath84 image , albeit with a few imaging artefacts that can occur on occasion @xcite . to create the mosaic , ccd images were initially downloaded from the iphas data repository , alongside the corresponding confidence maps of each field .
the confidence maps were used to create an image mask , and to filter out confidence values below 0.9 .
this tended to remove bright corner regions that were present in ccd3 , bad pixels streaks across the images , and also removed some additional bad data around the ccd edges .
the images were all then individually regridded into the ` sin ' world - coordinate system projection corresponding to that used for the atca images .
large mosaics of h@xmath1 and @xmath85 were then made using montage . using this software ,
the initial mosaic images were corrected for background differences between the individual images in order to produce a final mosaic of the sky .
we are interested in imaging both the fine filamentary @xmath86 features , as well as the diffuse emission .
one challenge associated with this is the uneven illumination of the ccds , which is partially due to moonlight scattering off the inside of the telescope dome onto the ccds ( n. wright , priv .
this is only at the level of a few counts in each image , and so is not visible in single - pointing mosaics , but can dominate the emission at larger scales .
it was found that by limiting the size of the mosaic to @xmath87 pixels , that the features in the images were still clearly visible with no obvious large - scale background variation .
attempts to increase the mosaic size beyond this would likely require an alternative approach to remove any gradient , but this was not necessary to cover the angular extent of w50 .
as we are only interested in the morphology of optical emission associated with w50 , rather than attempting to obtain any quantitative flux measurements , we were able to provide a continuum - correction by subtracting the median image value individually from the h@xmath1 and @xmath85 mosaics , and then creating a final image of h@xmath88 . in combination with the lack of gradient removal ,
this method has the advantage that it also highlights the diffuse emission rather than subtracting it .
the result is a continuum - corrected mosaic covering the entire w50/ss433 system .
we are missing part of the shell from dr2 of iphas , due to a few missing pointings , but these will likely be covered by future data releases .
a wide - field , wide - band , multi - frequency synthesis mosaiced image of the total intensity emission ( stokes @xmath42 ) of the w50/ss433 system is shown in fig . [ fig1 ] as observed using the atca .
the image covers the frequency range from 1.4 ghz to 3.1 ghz , with an effective bandwidth of 1.7 ghz .
there are artefacts around bright point sources due to the assumption of zero spectral index during simultaneous imaging ( see section [ imaging ] ) .
the surrounding field of view contains several bright extragalactic radio sources , the nearby hii region s74 to the northwest , and one half of the shell of the galactic supernova remnant candidate g38.71.4 to the southwest .
the bright compact source and microquasar , ss433 , is clearly visible at the approximate geometric centre of the nebula .
there is emission throughout the field - of - view , including diffuse extended emission from the nebula itself , and also background emission from the galaxy which is brighter towards the west . as seen in projection
, there are some compact sources embedded in the extended emission associated with w50 , although these are believed to be extragalactic and therefore in the background of the object itself .
the extended emission associated with w50 forms an approximately circular ring that is equidistant from ss433 , giving an appearance consistent with that of a shell - type snr . to the east and west are two ` ears ' that appear to have been ` punched ' through the shell by pressure from the jets of ss433 itself ( see section [ intro ] ) .
there are several small - scale filamentary structures that appear to be associated with these ears , including two filaments that are oriented approximately along north south in either ear ( roughly perpendicular to the presumed impact angle of the jets ) .
the eastern ear appears to extend much further from ss433 than the western ear , likely due to a denser environment to the west . at the very faintest levels of the image , an additional feature that corresponds to the `` chimney '' seen in previous studies
@xcite is visible as originating from the north
south oriented filament in the eastern ear and extends @xmath89 arcmin towards the north .
this has previously been interpreted as a rayleigh taylor instability originating from the material associated with the ears accelerating into a denser ambient medium .
a potentially new second chimney is seen @xmath90 arcmin to the west of the first chimney , also located in the eastern ear , and also extending towards the north in the direction of a bright compact background source . when using the atca to observe at a declination of @xmath77 ,
the @xmath39-coverage is sparse in the north south direction even when using the hybrid arrays .
this creates an asymmetric dirty beam and can cause low - level cleaning artefacts that occur around bright compact sources the feature is therefore likely spurious .
although limited by the imaging artefacts , a significant number of radio filaments are visible .
one bright filament extends from near to ss433 and curves down towards the southwest of the shell , terminating below the western ear .
other faint filaments are present to the east of ss433 , with one filament extending away from ss433 towards the east , at a slight angle ( @xmath91 ) to the east west axis .
another filament extends inwards from the northeast of the shell , towards the central northern region of the shell .
numerous other radio filaments are also present in the eastern and western ears , all of which have been identified in previous studies @xcite .
the linearly polarized intensity images of the mosaic surrounding w50 are shown in fig .
[ fig2 ] at frequencies from 1.4 ghz to 3.1 ghz , with each image having a 100 mhz bandwidth .
there is clear structure in the polarization towards the interior of the mosaic , which is spatially uncorrelated with total intensity .
this indicates that it is real , rather than instrumental polarization .
there is also apparent polarization towards the very outer edges of the mosaic , which is consistent with our assumptions about the off - axis instrumental polarization in section [ imaging ] . using the sensitivity of our atca observations , we are able to see a ` ring ' of linearly polarized emission surrounding ss433 and corresponding to the shell of the central region of w50 .
the brightest polarized emission in the north and east of the central ring has been identified previously .
strong polarization is also associated with a radio filament that extends from ss433 down towards the south - west of the central region of w50 .
in general , there appears to be a divide between the polarization in the east and the west of the image .
there is clear linear polarization associated with the eastern ear of w50 , while there is no polarization associated with the western ear .
polarized emission that is presumably galactic in origin , and unrelated to w50 itself , is also visible across the entire eastern half of the image , while there is conversely almost no polarization in the western half .
there is a polarized feature at the most western part of the image that appears as a faint semi - circle of polarized emission of @xmath89 arcmin radius with an additional extension to the north ( with any additional morphology located outside of the fov ) .
even this faint polarized emission is at the @xmath92 level .
all of this feature is located further to the west than the ear , and again appears to be galactic polarization .
no linear polarization was detected from either the nearby hii region s74 or the supernova remnant candidate g38.71.4 .
all of the visible polarized emission exhibits faraday depolarization , with the polarization having decreased significantly at 1.4 ghz compared to at 3.1 ghz . in section [ linpol ]
, we were able to see a ` ring ' of linearly polarized emission surrounding ss433 and corresponding to the shell of the central region of w50 .
however , this does not take depolarization effects into account ( see section [ depol ] ) .
it is not obvious whether the rest of the object should be linearly polarized at these observational frequencies , or whether faraday effects could be expected to have depolarized the emission elsewhere .
it is known that faraday rotation is related to the presence of thermal electrons which can be traced by h@xmath1 emission .
combined red green blue images of w50 , showing the radio total intensity , polarized intensity at a low faraday depth ( see section [ faradayrotation ] ) , and the h@xmath1 emission are shown in fig .
[ figrgb ] .
the figure shows a strong correlation between bright diffuse h@xmath1 and complete depolarization of the region surrounding w50 , or conversely there is an anticorrelation between bright h@xmath1 and linearly polarized intensity .
this anticorrelation is highlighted further in fig .
[ figrgb2 ] , which overlays polarized intensity contours at a low faraday depth over the h@xmath1 image . in the presence of the diffuse h@xmath1 , the polarized emission from w50 itself , and also some of the galactic polarized emission , both exhibit complete depolarization .
the type of depolarization can be influenced if ( i ) the synchrotron - emitting and thermal medium are mixed , or ( ii ) the thermal medium is in the foreground of the nebula . in the former case
, we expect broad faraday thick features in our faraday spectra . in the latter case
, we expect narrow faraday thin peaks .
our faraday spectra are almost completely dominated by a single faraday thin peak , with the sole exception of a slight indication of two faraday components from the southern part of the polarized emission associated with the eastern ear ( see section [ discussion ] ) .
a measurement of the fwhm of the strongest faraday components , provides values between @xmath93@xmath94 rad m@xmath3 , consistent with our rmsf@xmath68 rad m@xmath3 ( see section [ rmsynth ] ) . an extra constraint on the depolarization can be obtained by comparing our data with the 5 ghz observations of .
these data are shown in fig .
[ fig7 ] . even at higher radio frequencies of 5 ghz
, there is clearly polarization associated with the central region of w50 , and complete depolarization in the regions in which we have identified diffuse h@xmath1 emission .
we use the large continuum - corrected h@xmath1 mosaic from iphas , as described in section [ halpha ] .
we use this to investigate the connection between the optical emission and both the total intensity and linearly polarized radio counterparts , as described in section [ radiodata ] .
the complete iphas h@xmath1 mosaic , overlaid with atca total intensity contours , is shown in fig .
[ figregions ] .
the discrete rectangular regions used are shown alongside their respective numerical labels from 1 to 7 . to allow the reader to skip the qualitative descriptions , we provide additional detail of the morphology and features of each of these regions within appendices [ 1stregion ] to [ 7thregion ] .
we identify new previously - undiscovered filamentary structures that are associated with the w50/ss433 system , and which are far more extensive than has been previously reported .
most of these have radio continuum counterparts , and in a few cases there is a possible correspondence to the polarized emission .
the region showing the eastern breakout region , where the eastern ear punches through the eastern boundary of the circular region of w50 , is shown in fig .
[ figreg1 ] .
the images show the pseudo - colour iphas continuum - corrected h@xmath1 image , overlaid with the total intensity radio continuum contours from the vla @xcite image at 1.4 ghz , and with the linearly polarized intensity contours from the atca images at both 3.1 ghz and 2.2 ghz .
the region showing the eastern ear , and where the radio emission from the eastern ear terminates , is shown in fig .
[ figreg2 ] .
the region showing the western ear , the western breakout region where the jet punches through the western boundary of the circular region of w50 , and where the radio emission from the western ear terminates , is shown in fig .
[ figreg3 ] .
the region showing the northeastern interior of w50 , including ss433 within the field of view , is shown in fig .
[ figreg4 ] .
the region showing a small section from the northern rim of w50 is shown in fig .
[ figreg5 ] .
the region showing a small section from the northwestern rim of w50 is shown in fig .
[ figreg6 ] .
the region surrounding the snr candidate g38.71.4 is shown in fig .
[ figreg7 ] . for all images from fig .
[ figreg2 ] to fig .
[ figreg7 ] , the representation of various emission by different colours and contours is the same as in fig .
[ figreg1 ] .
the exact position of each region relative to the full extent of the w50/ss433 system is shown in fig .
[ figregions ] .
the stokes @xmath43 and @xmath44 datacubes as a function of frequency were converted into faraday - space using the technique of rm synthesis , as per section [ rmsynth ] .
this was done pixel - by - pixel in order to reconstruct a full image of the faraday rotation across the radio nebula .
the rm corresponding to the peak in faraday space for each pixel was recorded , and pixels with a signal to noise ( s / n ) @xmath95 were blanked making our final image very reliable .
this procedure implicitly assumes that a single faraday component is responsible for the faraday rotation ( e.g. * ? ? ?
* ) - although the situation can be more complicated , @xmath61-fitting was not possible using these data ( see section [ limitations ] ) .
nevertheless , rm synthesis will still retrieve an image showing spatial variations of the mean rm which can inform us of the average magnetic field structure , order , and geometry .
slices from the faraday depth cube of w50 are shown in fig .
the response from @xmath96 rad m@xmath3 to @xmath97 rad m@xmath3 is clearly shown .
there is clearly an rm gradient across w50 , with both positive and negative rms , and diffuse linearly polarized emission that fills the entire field of view .
note that the faraday spectra could not be cleaned , as explained in section [ rmsynth ] .
an image of the peak rm at each pixel is shown in fig .
[ fig4 ] , with the atca stokes @xmath42 contours overlaid to guide the eye .
further zoomed - in regions of the central region of w50 , and of the eastern ear , are shown in fig .
[ fig4a ] .
the image is broadly consistent with the results of @xcite albeit with sensitivity to much fainter polarized emission , larger spatial scales , and without @xmath18-ambiguities . within the circular central region of w50
, there is clearly a strong asymmetry in the rm , with positive rms to the north and south east , and negative rms to the west .
the highest positive rm is in the northern rim with @xmath98 rad m@xmath3 , while the highest negative rm to the west is @xmath99 rad m@xmath3 .
both the western ear and the funnel leading to the eastern ear are completely depolarized as described in section [ depol ] .
the polarized emission from the eastern ear also displays a strong asymmetry , with positive rms ( @xmath100 to @xmath101 rad m@xmath3 ) to the north and negative rms ( @xmath102 to @xmath103 rad m@xmath3 ) along the southern rim .
note that the negative region to the north - west of the ear does not originate from the bright polarization in the ear , but is rather from the surrounding diffuse galactic emission .
the rm foreground estimate further strengthens the suggestion that there is an rm asymmetry in the ear ( see section [ rmforeground ] ) , and an rm asymmetry is also consistent with the expected physics of the region ( see section [ earfield ] ) .
there is a sharp transition between the rm of the eastern ear in contrast to the surrounding galactic emission , which mostly has an rm of @xmath104 rad m@xmath3 in this area , and shows how the environment associated with w50 is modifying the line - of - sight magnetic field .
away from w50 , the diffuse polarized galactic emission also fills the majority of the field of view , and separately also displays a strong rm gradient that encircles the radio nebula .
the strongest emission is visible along the east of w50 , with negative rms up to @xmath105 rad m@xmath3 in the north which gradually increase to become positive in the south with positive rms up to @xmath106 rad m@xmath3 .
additionally , to the west of the field - of - view , the semi - circular gap in the diffuse h@xmath1 emission that also has polarized emission ( see section [ depol ] ) also displays an rm gradient , with a positive rm of @xmath107 rad m@xmath3 nearest to w50 and changing to @xmath108 rad m@xmath3 further from w50 in the north ( and at the edge of the field ) . interpreting the magnetic field structure from the rms
requires an understanding of the foreground galactic faraday screen .
this is a particularly complicated field of view , due to the large angular extent of w50 , and due to the source s location which dips into the galactic plane at a latitude of @xmath109 .
the best model of the galactic foreground currently available is .
such foreground models are not fully applicable to galactic objects , as the model is derived using extragalactic rms , whereas w50 is immersed within the galaxy .
consequently , the line - of - sight magnetic field towards w50 does not sample the full magnetoionic content between us and the periphery of the milky way .
the integrated magnetic field between us and w50 could have been modified ( and could even have changed sign ) several times before exiting the galaxy , particularly due to magnetic field reversals in the spiral arms of the milky way @xcite . another technique is therefore required in order to estimate the foreground galactic contribution to the faraday rotation . the use of pulsars as galactic magnetic field probes is viable as the faraday rotation from a pulsar magnetosphere is expected to be negligible and to not exhibit significant variations @xcite .
pulsars therefore provide a very clean rm signal , which measures the galactic magnetic field , and which is not contaminated by strong intrinsic variations .
following @xcite , we use v1.50 of the atnf pulsar catalogue @xcite which collates the properties of more than 2300 rotation - powered pulsars and is continually revised as new discoveries are made .
the atnf pulsar catalogue contains rms , dispersion measures ( dms ) , and pulsar distance estimates . within a @xmath110 radius of ss433
, we find 13 pulsars with measured rms , with most located to the east of w50 within the galactic plane .
it is difficult to obtain accurate distances to pulsars , with only a few precise measurements determined using annual parallaxes for relatively nearby pulsars .
all of the pulsars used here have a distance determined from the dm .
the dm - derived distance relies on a model of the galactic free - electron distribution ( e.g. * ? ? ?
* ) , which may be unreliable .
we here use the ne2001 model for our thermal - electron density estimates @xcite . in order to contrast the pulsar rm measurements in the foreground of w50 with the entire integrated galactic rm contribution
, we also use the extragalactic rm catalogue of @xcite . within a @xmath110 radius of ss433
, we find 12 extragalactic sources with measured rms . determining the foreground rm to w50 is challenging , as in combination with any possible errors on the distance to w50 itself , there are also additional uncertainties due to the derived electron model distances .
a plot showing the rm of the pulsars and extragalactic sources , as a function of distance , is shown in the top panel of fig .
[ figforeground ] . the distance estimate and @xmath111-uncertainty for ss433
is indicated by the red - coloured vertical band .
the pulsars in front of w50 have considerable scatter , with a median foreground rm of @xmath112 rad m@xmath3 thus consistent with zero foreground faraday rotation or a low positive rm ( on average across the field - of - view ) . the scatter in rm of pulsars behind w50 ( and
hence behind the region of the milky way in which w50 is situated ) appears to be larger than the scatter in rm of pulsars in front of w50 . as an overview of the general distribution of pulsars relative to w50
, the pulsar with nearest angular distance to ss433 is located at a proximity of @xmath113 , with an rm@xmath114 rad m@xmath3 .
the dm distance to this pulsar is 2.95 kpc , also placing it in front of the radio nebula .
the foreground rm as determined using the pulsars is evidently different from the median extragalactic rm towards w50 of @xmath115 rad m@xmath3 thus consistent with an integrated magnetic field pointing towards us in this region of the galaxy . as both the rm and the dm are _ integrated _ quantities along the line - of - sight , it is also possible to plot one against the other to measure their relative variation as a function of dm - derived distance .
such rm versus dm techniques have been discussed before ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . under reasonable assumptions for such a plot
, the gradient is related to the average local magnetic field strength as @xmath116 .
the direction of the local field , at a given dm - distance , is determined by the slope of the rm / dm plot .
consequently , maxima and minima reveal the location of field reversals , and it has been argued these reversals correspond to the magnetic fields associated with different spiral arms @xcite . there is always a constraint on such a plot at @xmath117 .
this rm versus dm method , could feasibly be affected by magnetic fields in intervening hii regions and snrs along the line - of - sight .
a plot showing the rm and dm of the pulsars is shown in the bottom panel of fig .
[ figforeground ] . the dm - distance estimate ( @xmath118 pc @xmath119 ) and @xmath111-uncertainty for ss433 , as calculated using ne2001 ,
is indicated by the red - coloured vertical band . to guide the eye
, a dashed line shows a smooth interpolation of the data .
the pulsars appear to show regular and consistent fluctuations of rm and dm .
the data suggest that multiple magnetic field reversals are visible along the line - of - sight , with two field reversals occurring in the foreground of w50 . at the distance corresponding to the location of w50
, the local rm gradient appears to be positive in sign and therefore the parallel component of the magnetic field is pointing towards us .
our best estimate of the foreground faraday rotation is @xmath120 rad m@xmath3 thus again consistent with zero foreground faraday rotation .
however , due to the large angular size of w50 , the dm varies across the source .
these variations could change the dm - distance towards w50 .
the dm across the face of w50 is clearly expected to vary from east to west , i.e. with higher dms closer to the galactic plane .
there is also possibly some , but considerably less , variation expected from north to south across the shell of w50 .
as shown in table [ table:2 ] , our data suggests a @xmath121% dm gradient from east to west , with no detectable dm gradient from north to south .
the dms measured at different locations towards the radio nebula , as calculated using ne2001 , do not change sufficiently to change our foreground faraday rotation estimate .
note that this does not rule out angular variations in the faraday rotating foreground itself as it tells us nothing about the foreground magnetic fields , which could have both large- and small - scale fluctuations across the face of the source .
we are therefore unable to constrain variations in foreground faraday rotation across w50 , and to do so would likely require a more finely - sampled pulsar grid .
.variations in the dispersion measure for various lines - of - sight across the face of the w50/ss433 system , according to the ne2001 model and corresponding to a distance of @xmath8 kpc . [ cols="^,^,^,^",options="header " , ]
in section [ linpol ] , we were able to see a ` ring ' of linearly polarized emission surrounding ss433 and corresponding to the shell of the central region of w50 .
this ring of ordered magnetic fields would naively appear to be consistent with field compression from an outgoing shock wave , and therefore consistent with a snr hypothesis for the origin of the nebula .
however , this does not take depolarization effects into account ( see section [ depol ] ) . in the presence of the diffuse h@xmath1 , the polarized emission from w50 itself , and also some of the galactic polarized emission , both exhibit complete depolarization .
this indicates that the warm ionised gas / thermal electrons that are traced by the diffuse h@xmath1 emission must be acting as a faraday screen that appears to be located in the foreground of w50/ss433 .
similar anticorrelations between h@xmath1 and polarization have been observed before in the galactic plane ( e.g. * ? ? ? * and references therein ) .
we interpret the faraday thin nature of the depolarizing screen as being consistent with the thermal medium being located in the foreground .
it is possible that the medium is faraday thick , however the thickness must be at a level below the sensitivity of our observations . even for reasonable estimates of magnetic field and electron density ,
e.g. @xmath122 @xmath119 , @xmath123 @xmath124 g , and @xmath125 pc , the total faraday thick contribution through the nebula would be only @xmath126 rad m@xmath3 ( which falls below our faraday thickness sensitivity limits ) .
a thickness of @xmath126 rad m@xmath3 therefore can not be ruled out ( also see section [ rmsynth ] ) .
however , a faraday thin hypothesis is also consistent with the expected strong extinction in the galactic plane which would preferentially allow us to observe only nearby ionised gas .
this suggests that the diffuse galactic emission to the east and west of the radio nebula also originates in the foreground , from a region nearer to us than the ionised gas .
one may argue that the source is unpolarized , rather than depolarized , however the w50 radio nebula has been shown to have steep optically - thin radio spectral indices see section [ backgroundw50 ] .
this is typical of synchrotron radiation , which is intrinsically polarized . in combination with the 5 ghz observations of ( shown in fig .
[ fig7 ] ) , we interpret the observed strong depolarization as being due to spatially - varying faraday rotation fluctuations across the face of the radio nebula . we can measure the magnitude of these fluctuations in the two most significantly depolarized regions : ( i ) the western ear , and ( ii ) the ` funnel ' leading towards the eastern ear . in the western ear , we find a polarized fraction of @xmath127% at 5 ghz . from our 2.1 ghz faraday cubes and band - averaged total intensity image , by measuring the noise , @xmath128 , in the @xmath43 and @xmath44 faraday depth cubes , we obtain @xmath129 mjy beam@xmath69 , and the brightest total intensity in the region reaches 51.54 mjy beam@xmath69 .
this allows us to place a @xmath130 upper limit on the polarized fraction within the western ear of @xmath131% .
similarly , within the eastern funnel , we find a polarized fraction of @xmath132% at 5 ghz . from our 2.1 ghz faraday cubes and band - averaged total intensity image
, we obtain @xmath133 mjy beam@xmath69 , and the brightest total intensity in the region reaches 25.93 mjy beam@xmath69 .
this allows us to place a @xmath130 upper limit on the polarized fraction within the eastern funnel of @xmath134% .
note that due to the limited spatial - scales sampled by the interferometer , there can be significant missing flux in the total intensity image at 2.1 ghz ( see section [ limitations ] ) .
our limits on the polarized fractions are therefore likely conservative .
we assume that the depolarizing faraday screen traced by the diffuse h@xmath1 consists of a magnetoionic region that has a negligible amount of relativistic particles and that exists somewhere along the line - of - sight between the observer and the source .
if this faraday screen contains a constant regular field and has a homogeneous distribution of free electrons , then the region causes faraday rotation of the polarized emission from background sources , but does not cause any physical depolarization ( remember that for depolarization , the magnetic field along the line - of - sight is nt necessarily relevant , but rather the magnetic field in the plane of the sky as seen in projection )
. nevertheless , any deviation from a constant field within the synthesised beam , or any density fluctuations in the number of thermal electrons , will create a rm gradient and subsequently cause depolarization . in the case of magnetic fields ,
such anisotropy is caused by either turbulent , or systematically varying regular fields .
we assume that the depolarization can be well - described by the model proposed by @xcite ( for a full review of external faraday depolarization effects , see @xcite ) . in this case , @xmath135 where @xmath136 is the measured fractional polarization , @xmath137 is the intrinsic fractional polarization , @xmath138 is the observing wavelength , and @xmath139 characterises the rm fluctuations within the observing beam , where we assume a gaussian distribution of rms within the beam .
we can therefore derive an estimate of the rm fluctuations using , @xmath140 } \ln \left ( \frac{\pi_{2.1~\textrm{ghz}}}{\pi_{5.0~\textrm{ghz } } } \right ) \\.\ ] ] in the western ear , this corresponds to faraday rotation measure fluctuations of @xmath14148 rad m@xmath3 on scales smaller than 4.5 to 6 pc . in the eastern funnel
, this corresponds to fluctuations of @xmath14161 rad m@xmath3 on scales smaller than 4.5 to 6 pc . for the case of the western ear
, the depolarization could possibly be the result of both rm fluctuations due to variations in the number of intervening thermal electrons , and also possibly a quasi - laing
garrington effect ( e.g. * ? ? ?
* ) as the ear lies further along the line - of - sight79@xmath6 to the line - of - sight , or equivalently 11@xmath6 to the plane of the sky @xcite , with the eastern ear pointing towards us .
] ( and is also dipping into the galactic plane ) .
this could mean that our estimate of the variations in the foreground screen could be overestimated , with a contribution to this depolarization also occurring in the local ism immediately surrounding w50 .
nevertheless , our estimate still reliably constrains the overall fluctuations in the plane of the sky towards the nebula which we have shown is substantially affected by the ionised gas distribution .
we present a toy model interpretation of the location of the diffuse h@xmath1 clouds relative to w50 itself in fig .
[ figrgb3 ] .
clouds are shown in purple , and are located between w50 and the observer . ] it could be argued that the distribution of linear polarization from w50 also appears to show a correlation with the location of a diffuse molecular cloud , as inferred from co observations ( e.g. * ? ? ?
* ; * ? ? ?
however , there is no plausible mechanism that would allow such a cloud to preferentially allow the transmission of linearly polarized radio emission in contrast to the surrounding regions .
more recent observations also suggest there is no convincing evidence for an interaction of the w50/ss433 system with this molecular cloud @xcite , and so the cloud is most likely simply coincidentally aligned with w50 .
we here follow the results presented in section [ faradayrotation ] , and consider the foreground faraday rotation determined in section [ rmforeground ] .
the median foreground rm of the @xmath110 radius patch towards the w50/ss433 system is consistent with @xmath142 rad m@xmath3 .
however , we are unable to constrain any spatial foreground variations at scales less than @xmath110 .
the location of w50 near to the galactic plane , and situated in an area with rapidly - varying galactic emission , and with a known depolarizing faraday screen ( see section [ depol ] ) , suggests that the observed variation in rm across the face of w50 could be due to foreground variations .
indeed , the rms of the diffuse polarized emission that surrounds w50 changes sign in both the east and the west , which shows that the foreground magnetic field structure is complicated . nevertheless , the variation in the rms of the polarized emission of w50 itself shows no obvious relation to the depolarizing screen , e.g. there are no enhancements or variations of the rm near to the diffuse h@xmath1 clouds .
this suggests either that the rm variations are intrinsic and related to the local environment , or that there are additional intervening screens of significant angular extent and of unknown origin .
the observed rms towards the w50/ss433 system appear to show two key morphological features , as discussed in section [ atcarms ] : ( i ) the central region of w50 appears to have an axis of symmetry that separates two large regions of oppositely - signed rm .
this axis of symmetry appears to be located somewhere between @xmath143 to @xmath144 relative to the north south axis that intersects ss433 , and ( ii ) the eastern ear of w50 appears to have an axis of symmetry that lies parallel to the jet - axis , with positive rms to the north and negative rms to the south . as both of these features appear to be related to other morphological features of w50 , and in combination with the average foreground rm being consistent with @xmath142 rad m@xmath3 , we interpret the observed rm variations to be intrinsic and directly related to the environment of w50 .
while such an interpretation is strongly dependent on our foreground estimate , we would require a very large error in our estimate to remove the rm - axis associated with the central region with a foreground offset of @xmath145 rad m@xmath3 required to completely remove any axis across the source .
this further suggests that the foreground is likely not responsible for this feature .
similarly , the galactic polarized emission that surrounds the eastern ear has a positive rm that is continuous across both the positive and negative regions of the emission associated with w50 s ear .
the sharp transition from the rm of @xmath104 rad m@xmath3 associated with the galactic emission to the weaker @xmath146 rad m@xmath3 associated with the ear , also suggests that the galactic foreground can not be responsible for this feature .
however , this transition from @xmath104 rad m@xmath3 to @xmath146 rad m@xmath3 suggests that the galactic emission is located behind the ear .
if the galactic emission were originating in the foreground of the ear , one would expect the @xmath104 rad m@xmath3 to be imprinted on part of the eastern ear , which is clearly not the case .
nevertheless , the observed rm variations across some parts of the source could be due to galactic foreground variations , which hinders enabling an accurate determination of the intrinsic properties of the magnetised plasma within the radio nebula especially at a galactic latitude of @xmath147 .
the rm of the foreground diffuse galactic emission , visible in both the east and west of the atca mosaic , very clearly varies in both sign and magnitude on angular scales smaller than that of w50 .
it is likely that much more sophisticated techniques for modelling the foreground magnetic field structure in the galaxy will need to be developed before a comprehensive understanding of the magnetic field structure in the object can be ascertained .
the faraday rotation of the eastern ear can be interpreted as evidence that there is a toroidal- or helical - shaped magnetic field that loops about the axis from the jets , with the magnetic field facing away from us in the southern part of the ear , and facing towards us in the northern part of the ear .
the magnetic field is not an exact loop , with positive rms to the north , negative rms to the south , and intermediately positive rms in the conjoining intermediate region of @xmath148 rad m@xmath3 .
interestingly , the axis of symmetry of the rms would become much more strongly aligned with the jet - axis after subtracting our foreground rm estimate of @xmath112 rad m@xmath3 .
this rm - feature is also consistent with x - ray observations , which find that the eastern ss433 jet terminates in the eastern ear in a _ ring - like _ ` terminal shock ' .
the spatial coincidence of x - ray and radio emission in this region suggests that the physical conditions of the terminal shock region are very similar to those found at the outer shocks of ordinary supernova remnants , and indicates the interaction between the terminal shock of the jet and the circumstellar medium .
the x - ray emission is very strongly correlated with the radio - continuum emission and seems to form a ring - like structure confined by the outer boundary of the radio remnant , and indicative of the braking of a hollow - cone flow .
suggest this indicates that the final jet - flow may retain a hollow - cone morphology .
the thermal x - ray component has a low - fitted temperature which implies that the terminal shock must not be very strong , and that the jet - flow must have been considerably slowed down from its initial @xmath7 .
where and how this deceleration happens , remains unclear ( e.g. * ? ? ?
the faraday rotation of the central region can be interpreted as evidence that there is a shell - like magnetic field looping around the central region of the radio nebula .
this is consistent with observations of snrs ( e.g. * ? ? ?
it has been shown by @xcite that the rm of a uniform ambient medium that is swept - up into a shell by an expanding snr can leave a strong imprint on the rm variations of the resulting shell itself .
following @xcite , the resulting variations in rm can therefore be used to infer the in - situ ambient magnetic field geometry of the gas into which the snr has exploded . however , there is a known strong density gradient running e
w in w50 , identified by both the spectral index and morphology of the western ears ( see section [ backgroundw50 ] ) , and yet this gradient is not identified in the rm structure . similarly , as the shell of w50 is only clearly visible to the n and s , this provides an axis of symmetry for the radio nebula .
if the snr had swept up a uniform magnetic field it would be expected to leave identical rm gradients along each edge , with the gradient aligned with the objects axis of symmetry @xcite .
such a gradient is not observed . as an alternative explanation
, a shell could have embedded an imprint in the rms of the ambient magnetic field left behind by a progenitor wind .
again following @xcite , at large distances from a stellar surface , stellar winds are expected to have largely toroidal fields if such a field geometry is preserved after material was swept up by an outgoing shock front , this could produce a rm pattern that was negative on one limb of the snr and positive on the other .
this does qualitatively appear to fit the observations of w50 .
nevertheless , such an analysis is non - trivial for the w50/ss433 system , which could have a wind - blown bubble origin . in models of wind - blown bubbles
in which a fast - wind ( or supernova ejecta ) overtakes and `` sweeps - up '' a slower magnetised wind from a prior state of stellar evolution , in the case of an azimuthal field , the resultant radio nebula displays two arc - like features with oppositely - signed rms ( e.g. * ? ? ?
in particular , the negative rms along the ridge to the sw of the central region below ss433 are also visible in the image of @xcite , who find that unlike the rest of the object , this region does not have a tangential distribution of magnetic field relative to the shell but rather that the magnetic fields run directly along the ridge .
this may imply that we are seeing polarization from the front layers of the shell in this region , with polarization further along the line - of - sight having already been depolarized by faraday effects within the nebula .
this would be consistent with the total intensity spectral index variations across w50 , and with the diffuse depolarizing foreground screen near to this area which both imply a denser environment with more small - scale structure in the magnetic field towards the galactic plane .
the analysis is complicated further due to the action of the ss433 jets that could have entirely remodified the magnetic field structure of w50 .
indeed , one could argue that there is an axis of symmetry in the rms of the central portion of w50 that are aligned perpendicularly to the axis of the ss433 jets : with negative rms to the west and positive rms to the east .
however , this assessment does not incorporate the contradictory large positive rms to the north / north west of the central region .
nevertheless , the northern portion of the shell has other features that suggest it should not be incorporated in an assessment of any symmetry axis and that this region could have an independently large rm .
these anomalous features include its very high polarized fraction of @xmath21% ( see section [ radiopolbackground ] ) , which indicates that the northern rim has the most strongly - ordered magnetic fields across the whole object .
furthermore , the linear polarization of the northern rim is coincident with some of the optical filaments this is the only region in w50 in which this is the case .
these features suggest that this region is dominated by classic compression of an interstellar field by a sn shock that is oriented approximately east west .
the most consistent interpretation therefore appears to be that there is a symmetry axis running approximately north south ( although the precise angle of this axis of symmetry is ambiguous ) , with a loop magnetic field that mostly runs east west in the central region as seen in projection .
this explanation also fits with the known orientation of w50 along the line - of - sight ( with the eastern lobe oriented slightly towards us , and the western lobe directed away from us ) , and with other multiwavelength evidence that implies a preferential east west axis in w50 ( see section [ backgroundw50 ] ) . in combination
, this therefore implies that the current magnetic field geometry is predominantly oriented parallel to the axis of the ss433 jets .
furthermore , as the symmetry axis of the rm - gradient in the circular region of w50 may be oriented perpendicularly to the jet - axis of ss433 , this suggests that the central region of w50 is not a distinct object from the jets , and that the central region evolved in tandem with the jet - features .
there must therefore be a strong degree of jet / remnant coupling in the w50/ss433 system , which includes coupling of the respective magnetic fields . in combination ,
we interpret our data as evidence in favour of a loop magnetic field existing in the shell of w50 , which is oriented almost exactly east west . in this scenario ,
the polarization from the northern rim probes a separate and distinct region of the shell with a large positive rm .
this would naively appear to provide an intangible three - dimensional magnetic field geometry , however this is only as seen in projection and does not include e.g. depolarization effects along the line - of - sight .
there is other evidence to suggest the northern rim could be distinct from the rest of the shell it is brighter in total intensity , and is considerably stronger in linear polarization compared to the rest of the shell .
furthermore , unlike the rest of the shell , the polarization angles in the northern rim are typically noted as a classic case of compression of frozen - in magnetic fields ( see section [ faradayrotation ] ) .
this is consistent with the northern rim being the only region in the w50/ss433 system that has been unmodified by the jets of ss433 , with the northern rim therefore providing an archaeological glimpse into the initial snr that compressed the ambient magnetic field of the surrounding ism .
alternatively , we could invoke the presence of a pre - existing cavity that had been carved out prior to the supernova explosion , i.e. a stellar - wind bubble , and with the cavity being an ovoid that is preferentially larger along the east west jet axis .
it is known that an ambient ism magnetic field oriented perpendicularly to the direction of the expansion of a bubble can lead to the formation of an ovoid , rather than spherical bubble .
strong magnetic fields , can even completely stop the expansion of the bubble in the direction perpendicular to the field , leading to the formation of a tube - like bubble .
this effect would have strong consequences on the shape and evolution of nebulae such as w50 , which may have been formed within a main wind - blown bubble . in this case , the supernova shock would then propagate undisturbed through the ovoid cavity , leaving behind a loop magnetic field structure . in the case of the northern rim ,
the shock interacted strongly compressing the denser ambient ism .
both interpretations clearly require a supernova progenitor , and therefore favour a _ snr+jets model_. in fact , this latter scenario suggests the possibility of a more complex scenario than has typically been considered in the literature , a _
wind+snr+jets model _
, with a stellar - wind bubble driven by the jets of ss433 expanding into a swept - up ism that was itself created either entirely by the jets of ss433 or even by a continuous wind from the ss433 binary , followed by a supernova explosion with the snr rapidly taking on the dimensions of the wind - blown bubble and finally an encounter between the jets from ss433 and the already - formed snr , with the ram pressure from the jets ` punching ' through the shell to form the ears .
wind+snr+jets model _ has been partially suggested before @xcite .
this model could reconcile why the snr has been previously interpreted as displaying features consistent with both a wind - blown bubble and a snr .
significant questions remain concerning the association of w50 and ss433 ( see e.g. * ? ? ?
* and references therein ) . while a consensus has emerged that the jets of ss433 have punched holes in a preceding , expanding , supernova shell , the formation of the whole of w50 s structure by ss433 s jets alone has not been completely ruled out ( see section [ intro ] )
the lack of optical - emitting filaments in the shell of w50 has previously presented a challenge for a supernova origin of the shell ( e.g. * ? ? ? * ) .
if the shell is an old snr , then we would expect optical emission from some of the shell filaments , and for optical emission to be observed all along the shell and to trace out the region of radiative cooling of the shock wave in a relatively dense region of the ism ( e.g. * ? ? ?
although non - radiative shocks are also seen as faint filaments in the h@xmath1 image of younger snrs such as kepler s snr and rcw 86 @xcite , and ionised h@xmath1 filaments have also been detected in systems such as the orion - eridanus superbubble @xcite .
previously proposed explanations for this discrepancy in w50 are that ( i ) there is a rift of higher extinction crossing the center of w50 @xcite , that could be obscuring any filaments , and ( ii ) alternatively , that the radio emission from a large snr , like w50 , of @xmath149 pc radius would come from the crushing of low - density clouds which would produce only weak optical emission ( as predicted by the model of @xcite ) .
we have resolved this dilemma by identifying weak optical filaments distributed throughout the shell .
we have identified counterparts in the radio morphology for nearly all of the newly - identified optical filaments .
an alignment is known for the brightest optical features at the breakout regions of the shell @xcite , while an alignment between optical and radio filaments in the shell itself would be expected for an old snr . while it could be argued that the newly discovered optical filaments are suggestive of a sn shock ,
this is by no means certain , and it is not possible using the current data to directly rule out an origin of the h@xmath1 filaments due to a wind - blown bubble .
nevertheless , the combined weight of the multiwavelength evidence does typically suggest a snr origin @xcite . either way , future studies will be able to use these new filaments to obtain in - depth information on the kinematics of the shell of w50 potentially allowing a definitive experiment to determine how w50 was formed . in w50 ,
the lack of a detection of an additional , more extensive , network of filamentary nebulosity throughout the shell is consistent with previous deep - images and high - resolution spectroscopy of the northern radio ridge of w50 that suggested that patchy foreground dust along the 5.5 kpc line - of - sight is inhibiting the detection of all of the optical nebulosity associated with w50 @xcite .
our detection of the faintest filaments , in parts of the shell , lend support to a very patchy foreground .
the discovery of a more extensive network of optical filaments lends additional support to the _ snr+jets model _ for the formation of the w50/ss433 system .
however , it does not rule out a mixed - type origin for the nebula , and could also imply a _ wind+snr+jets model _ with a snr expanding in a pre - formed wind - blown bubble ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
there is also the possibility of most of the expansion of the shell being driven by a wind that follows , rather than precedes , the snr a _
snr+wind+jets model _
( or perhaps a _ wind+snr+wind+jets model _ as it is not mutually exclusive from there being a pre - existing cavity before the supernova explosion ) . in this scenario ,
a preceding snr left a shell , which was then further inflated by a wind from the binary system , and with the eastern and western regions later being strongly distorted by outflows from the jets .
this potentially better explains the morphology of the filled - centre almost plerion - like central component which has a steep - spectrum interior , however this scenario would appear to be inconsistent with the observed magnetic field structure in the northern rim that differs from the rest of the central region .
what is clear is that the magnetic contributions of the shell and the jets are coupled , and that either ( i ) the field in the shell has helped to form ss433 s axis and to collimate the jets in this direction , ( ii ) the axis of the jets has modified the magnetic fields in the surrounding medium , ( iii ) there is a strong ambient field that has influenced both the intrinsic axes of the shell and the jets , or ( iv ) some combination of the above .
numerical simulations of explosions inside pre - existing wind - driven bubbles , particularly of non - spherical snrs , and predictions of the effect on the observed rm geometry will help substantially in unpicking these various interpretations . while beyond the scope of this paper , demonstrating the application of @xmath61-fitting @xcite to an extended object such as w50 would form a useful future study .
we would require the total intensity short - spacing data in order to perform reliable @xmath61-fitting on our atca data , as we are currently unable to retrieve reliable polarized fractions due to the limited and changing @xmath39-coverage across the very broad band .
future observations would therefore ideally include broadband data from a single - dish in order to recover all of the spatial scales in the object
. nevertheless , @xmath61-fitting should be able to identify separate signatures from the northern rim and the east west aligned ring - like magnetic field .
this would allow one to test the various scenarios for w50 s formation , although it appears a supernova explosion is likely a necessary feature in the object s evolution .
w50 has implications for how snrs which contain compact objects fit into the overall picture of snr evolution .
the combined evidence of our own and previous observations is consistent with w50 being the reanimated corpse of a shell - like snr , driven back to life by the powerful outflows from its central engine .
this places w50 into a unique class of zombie supernova remnants .
the study of such outflow - dominated objects may provide unique methods in which to identify remnants with massive star progenitors , and may also assist in revealing why the fraction of snrs in which a compact source is identified is much less than that expected by supernovae type alone .
this is doubtlessly partially due to observational selection effects .
nevertheless , future studies of w50 , and other possible zombie snrs , may allow us to begin recognising more subtle signatures that allow the presence of a central source to be inferred rather than directly detected .
an extensive sample of similar zombie remnants are already known , and a number of potential candidates are already visible in the most snr catalogue .
future observations of w50 and these other candidates may therefore be able to help us understand the larger picture of where snrs with compact objects fit on the evolutionary ladder , and to create an even more comprehensive catalogue of similar objects for future study .
observations of similar ununsual large - angular size nebulae are already underway ( e.g. * ? ? ?
* ) , and the use of facilities such as the atca and the vla will continue to improve upon galactic snr imaging techniques ( e.g. * ? ? ? * ) .
it is clear that the w50/ss433 system is in a special and unique class of outflow - driven snrs .
using the atca , we have obtained a 198 pointing mosaic of a @xmath150 degree region of the sky surrounding w50 , with continuous radio frequency coverage from 1.4 to 3.1 ghz , and in full - stokes . following rm synthesis , we have obtained a noise level in our faraday cubes of 0.125 mjy beam@xmath69 rmsf@xmath69 .
this is the most sensitive mosaic of w50 that has been created in linear polarization , and provides an unrivalled view of the large - scale magnetic fields in the radio nebula . to our knowledge , these data are nominally also the most sensitive images of w50 in the radio continuum to date , although we are limited in total intensity by our @xmath39-coverage .
we have found the limitation that the frequency - dependent @xmath39-coverage inherent to ultra broadband radio observations tends to provide an overly steep spectral index of extended sources .
we have complemented our radio data by also creating a large mosaic of iphas continuum - corrected h@xmath1 data , which maps out the ionised gas distribution surrounding the w50/ss433 system thereby providing a unique multi - wavelength perspective of the object .
our conclusions are summarised as follows : 1 .
we have identified new , faint , optical , filamentary emission across the nebula .
conversely to previous studies , we find optical emission coincident with both the northern shell and the bright radio filaments in the western ear .
we also find additional new filaments within the filled - centre of the w50 nebula . unlike previously discovered filaments that are most likely associated with breakout regions from the microquasar jets , the newly discovered faint filaments are most consistent with being associated with the initial snr shell itself .
all of these optical filaments are either directly traced out , or have a nearby filamentary counterpart , in the radio continuum image of w50 .
2 . we have mapped the linearly polarized radio emission across the nebula . unlike the radio continuum data
, the linear radio polarization appears to be largely independent of the brightest optical filaments . nevertheless , the polarization is coincident with some of the fainter filaments in the northern ridge also consistent with classic field compression by a supernova shock .
the linear radio polarization also shows a strong anti - correlation with the surrounding diffuse h@xmath1 emission , demonstrating that the foreground ionised gas traced by the h@xmath1 emission is depolarizing the observed emission due to spatially - varying faraday rotation fluctuations across the face of the radio nebula .
the simultaneous depolarization of both w50 and the galactic polarized emission indicates that these clouds of thermal electrons must be located in the foreground of w50/ss433 .
the structure of the faraday foreground towards w50 is complicated and features many large - scale changes of sign of the magnetic field both in the plane of the sky ( as traced by the rm from linearly polarized emission ) and along the line - of - sight ( as traced by the rm and dm of nearby pulsars ) .
as the median foreground rm towards the nebula is consistent with @xmath142 rad m@xmath3 , the pattern of rms across w50 appears to be best described by a loop magnetic field surrounding the central region itself .
the symmetry axis of this loop - field is oriented perpendicularly to , and appears to be connected to , the symmetry axis of the jets / ears suggesting that both the shell and the ears are related and dependent phenomena . the evolution of magnetic fields in both the jet and remnant components of the w50/ss433 system must be intertwined .
the faraday rotation data suggest that there is a ring - like magnetic field that is threaded through the termination shock associated with the eastern ear , at the location where the shock of the jet interacts with the circumstellar medium .
this is consistent with the ring - like structure previously inferred by x - ray observations .
finally , it should be noted that the enhanced radio and x - ray emitting eastern ear structure is possibly similar to a hotspot in frii radio galaxies that are also the physical manifestation of a jet - termination shock .
thus w50 , and similar objects of its zombie class , present a unique opportunity to study structures similar to hotspots using galactic jets . curiously , a similar magnetic field structure to that seen in w50 s eastern ear has also been implicated in the magnetic field structure of jets from active galactic nuclei ( e.g. * ? ? ?
this further suggests that investigations of the magnetic field structure using tools such as @xmath61-fitting , will be important in unravelling the full faraday structure along the line - of - sight towards galactic supernova remnants .
in particular , future radio observations with for example the square kilometre array ( ska ) will allow us to learn about magnetic field structures in snrs through detection of the emission itself , and through a background grid of polarized sources , due to the vast increase in the available sample size @xcite .
such studies will help us further understand the ism itself and features such as snrs , and to therefore refine and remove our estimates for the galactic foreground for a large number of extragalactic studies .
we thank martin bell , ilana feain , and craig anderson for useful discussions during the preparation of this project , and janet drew and nick wright for discussions on mosaicing data from the iphas survey .
we are also grateful to naomi mcclure - griffiths for helpful advice on observing with the atca .
is very grateful to filippo mannucci and the staff of the arcetri astrophysical observatory for hosting him as a visitor during the writing of this paper . j.s.f . & b.m.g .
acknowledge the support of the australian research council through grant dp0986386 .
parts of this research were conducted by the australian research council centre of excellence for all - sky astrophysics ( caastro ) , through project number ce110001020 .
s.p.o . acknowledges support from unam through the papiit project ia103416 .
the australia telescope compact array is part of the australia telescope national facility which is funded by the commonwealth of australia for operation as a national facility managed by csiro .
this paper makes use of data obtained as part of the int photometric h@xmath1 survey of the northern galactic plane ( iphas , www.iphas.org ) carried out at the isaac newton telescope ( int ) .
the int is operated on the island of la palma by the isaac newton group in the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias .
all iphas data are processed by the cambridge astronomical survey unit , at the institute of astronomy in cambridge .
the bandmerged dr2 catalogue was assembled at the centre for astrophysics research , university of hertfordshire , supported by stfc grant st / j001333/1 .
this research made use of montage , which is funded by the national science foundation under grant number aci-1440620 , and was previously funded by the national aeronautics and space administration s earth science technology office , computation technologies project , under cooperative agreement number ncc5 - 626 between nasa and the california institute of technology .
999 abell g. o. , margon b. , 1979 , nature , 279 , 701 . abolmasov p. , maryeva o. , burenkov a. n. , 2010 , ast .
nach . , 331 , 412 . akahori t. , kumazaki k. , takahashi k. , ryu d. , 2014 , pasj , 66 , 65 .
anderson m. c. , rudnick l. , 1993 , apj , 408 , 514 .
anderson c. s. , gaensler b. m. , feain i. j. , franzen t. m. o. , 2015 , apj , 815 , 49 .
anderson c. s. , gaensler b. m. , feain i. j. , 2016 , preprint ( arxiv:1604.01403 ) .
barentsen g. , et al .
, 2014 , mnras , 444 , 3230 .
begelman m. c. , hatchett s. p. , mckee c. f. , sarazin c. l. , arons j. , 1980 , apj , 238 , 722 .
bhatnagar s. , rau u. , green d. a. , rupen m. p. , 2011 ,
apj , 739 , l20 .
bisnovatyi - kogan g. s. , silich s. a. , 1995 , rev .
phys . , 67 , 661 .
blandford r. d. , cowie l. l. , 1982 , apj , 260 , 625 . blundell k. m. , bowler m. g. , 2004 , apj , 616 , l159 . blundell k. m. , hirst p. , 2011
, apj , 735 , l7 .
boumis p. , et al .
, 2007 , mnras , 381 , 308 . brentjens m. a. , de bruyn a. g. , 2005 , a&a , 441 , 1217 . brinkmann w. , aschenbach b. , kawai n. , 1996 , a&a , 312 , 306 .
brinkmann w. , pratt g. w. , rohr s. , kawai n. , burwitz v. , 2007 , a&a , 463 , 611 .
brown j. c. , haverkorn m. , gaensler b. m. , taylor a. r. , bizunok n. s. , mcclure - griffiths n. m. , dickey j. m. , green a. j. , 2007 , apj , 663 , 258 . burn b. j. , 1966 , mnras , 133 , 67 .
chu y .- h .
, 2008 , iaus , 250 , 341 .
ciotti l. , dercole a. , 1989 , a&a , 215 , 347 .
cordes j. m. , lazio t. j. w. , 2002 , preprint ( arxiv:0207156 ) .
dickel j. r. , milne d. k. , 1976 , aujph , 29 , 435 .
downes a. j. b. , pauls t. , salter c. j. , 1986 , mnras , 218 , 393 .
downes a. j. b. , salter c. j. , pauls t. , 1981 , a&a , 103 , 277 . drew j. e. , et al .
, 2005 , mnras , 362 , 753 .
dubner g. m. , holdaway m. , goss w. m. , mirabel i. f. , 1998 , aj , 116 , 1842 .
durouchoux p. , sood r. , safi - harb s. , lu f. j. , oneill p. , flohic h. , lefevre f. , 2001 , apsss , 276 , 139 .
elston r. , baum s. , 1987 , aj , 94 , 1633 .
fabian a. c. , rees m. j. , 1979 , mnras , 187 , 13p .
fabrika s. , 2004 , astrophysics and space physics reviews , 12 , 1 .
farnes j. s. , gaensler b. m. , carretti e. , 2014 , apjs , 212 , 15 .
forbes d. , 1984 , jrasc , 78 , 210 .
gabuzda d. c. , knuettel s. , reardon b. , 2015 , mnras , 450 , 2441 .
gaensler b. m. , green a. j. , manchester r. n. , 1998 , mnras , 299 , 812 .
gao x. y. , han j. l. , reich w. , reich p. , sun x. h. , xiao l. , 2011 , a&a , 529 , a159 .
gao x. y. , reich w. , reich p. , han j. l. , kothes r. , 2015 , a&a , 578 , a24 .
geldzahler b. j. , pauls t. , salter c. j. , 1980 , a&a , 84 , 237 .
goodall p. t. , alouani - bibi f. , blundell k. m. , 2011 , mnras , 414 , 2838 .
goodall p. t. , blundell k. m. , bell burnell s. j. , 2011 , mnras , 414 , 2828 .
green d. a. , 1990 , aj , 100 , 1927 .
green d. a. , 2009 , basi , 37 , 45 .
green d. a. , 2014 , basi , 42 , 47 .
green d. a. , downes a. j. b. , 1987 , mnras , 225 , 221 .
han j. , 2013 , iaus , 291 , 223 .
han j. l. , manchester r. n. , qiao g. j. , 1999 , mnras , 306 , 371 .
han j. l. , manchester r. n. , lyne a. g. , qiao g. j. , van straten w. , 2006 , apj , 642 , 868 .
harvey - smith l. , gaensler b. m. , kothes r. , townsend r. , heald g. h. , ng c .- y .
, green a. j. , 2010 , apj , 712 , 1157 .
harvey - smith l. , madsen g. j. , gaensler b. m. , 2011 , apj , 736 , 83 .
haverkorn m. , gaensler b. m. , brown j. c. , bizunok n. s. , mcclure - griffiths n. m. , dickey j. m. , green a. j. , 2006 , apj , 637 , l33 .
haverkorn m. , gaensler b. m. , brown j .- a .
c. , mcclure - griffiths n. m. , dickey j. m. , green a. j. , 2006 , ast .
nach . , 327 , 483 .
haverkorn m. , brown j. c. , gaensler b. m. , mcclure - griffiths n. m. , 2008 , apj , 680 , 362 .
haverkorn m. , et al .
, 2015 , proceedings of advancing astrophysics with the square kilometre array ( aaska14 ) , 913 june , 2014 , giardini naxos , italy .
heald g. , 2009 , iaus , 259 , 591 .
heald g. , braun r. , edmonds r. , 2009 , a&a , 503 , 409 .
helder e. a. , vink j. , bamba a. , bleeker j. a. m. , burrows d. n. , ghavamian p. , yamazaki r. , 2013 , mnras , 435 , 910 .
hester j. j. , 1987 , apj , 314 , 187 .
hill a. s. , mao s. a. , benjamin r. a. , lockman f. j. , mcclure - griffiths n. m. , 2013 , apj , 777 , 55 .
holden d. j. , caswell j. l. , 1969 , mnras , 143 , 407 .
huang r. h. h. , wu j. h. k. , hui c. y. , seo k. a. , trepl l. , kong a. k. h. , 2014 , apj , 785 , 118 .
y .- l . , dame t. m. , thaddeus p. , 1983 ,
apj , 272 , 609 .
ignace r. , 2014 , astra proceedings , 1 , 1 .
ignace r. , pingel n. m. , 2013 , apj , 765 , 19 .
insertis f. m. , rees m. j. , 1991 , mnras , 252 , 82 .
jowett f. h. , spencer r. e. , 1995 , the xxviith young european radio astronomers conference , ed .
green and w. steffen , in association with cambridge university press .
katz - stone d. m. , kassim n. e. , lazio t. j. w. , odonnell r. , 2000 , apj , 529 , 453 .
kesteven m. j. , caswell j. l. , 1987 , a&a , 183 , 118 .
kesteven m. j. , caswell j. l. , milne d. k. , haynes r. f. , roger r. s. , 1987 , aujph , 40 , 855 . konigl a. , 1983 , mnras , 205 , 471 .
koo b .- c . , heiles c. , 1995 , apj , 442 , 679 .
kothes r. , 2015 , hia , 16 , 395 .
kothes r. , brown j .- a . , 2009 , iaus , 259 , 75 .
laing r. a. , 1988 , nature , 331 , 149 .
landecker t. l. , routledge d. , reynolds s. p. , smegal r. j. , borkowski k. j. , seward f. d. , 1999 , apj , 527 , 866 .
li f. , brown s. , cornwell t. j. , de hoog f. , 2011 , a&a , 531 , a126 . linden t. , hooper d. , yusef - zadeh f. , 2011 , apj , 741 , 95 .
lockman f. j. , blundell k. m. , goss w. m. , 2007 , mnras , 381 , 881 .
longmore a. j. , clark d. h. , murdin p. , 1977 ,
mnras , 181 , 541 .
manchester r. n. , hobbs g. b. , teoh a. , hobbs m. , 2005 , aj , 129 , 1993 .
margon b. , ford h. c. , katz j. i. , kwitter k. b. , ulrich r. k. , stone r. p. s. , klemola a. , 1979 , apj , 230 , l41 .
margon b. , 1984 , ara&a , 22 , 507 .
martin p. g. , murdin p. g. , clark d. h. , milone e. f. , 1979 , iauc , 3358 , 1 .
meyer d. m .- a . , langer n. , mackey j. , velzquez p. f. , gusdorf a. , 2015 , mnras , 450 , 3080 .
milne d. k. , 1987 , aujph , 40 , 771 .
mitra d. , wielebinski r. , kramer m. , jessner a. , 2003 , a&a , 398 , 993 .
murata k. , shibazaki n. , 1996 , pasj , 48 , 819 .
osullivan s. p. , et al .
, 2012 , mnras , 421 , 3300 .
osullivan s. p. , et al .
, 2013 , apj , 764 , 162 .
oppermann n. , et al .
, 2015 , a&a , 575 , a118 .
panferov a. a. , 2010 , preprint ( arxiv:1001.5097 ) .
pingel n. m. , ignace r. , 2012 , journal of the southeastern association for research in astronomy , 7 , 23 .
pon a. , johnstone d. , bally j. , heiles c. , 2014 , mnras , 441 , 1095 .
purcell c. r. , et al .
, 2015 , apj , 804 , 22 .
ransom r. r. , kothes r. , geisbuesch j. , reich w. , landecker t. l. , 2015 , apj , 799 , 198 .
rees m. j. , 1978 , symposium on quasars and active nuclei of galaxies , copenhagen , denmark , june 27july 2 , 1977 , physica scripta , 17 , 193 .
rozyczka m. , tenorio - tagle g. , 1985 , a&a , 147 , 209 .
rozyczka m. , tenorio - tagle g. , franco j. , bodenheimer p. , 1993 ,
mnras , 261 , 674 .
rozyczka m. , tenorio - tagle g. , 1995 , mnras , 274 , 1157 . rozyczka m. , franco j. , 1996 , apj , 469 , l127 .
ryle m. , hine g. , shakeshaft j. , caswell j. l. , 1978 , nature , 276 , 571 .
sadowski a. , narayan r. , 2015 , mnras , 453 , 3213 . sabin l. , et al . , 2013 , mnras , 431 , 279 .
safi - harb s. , gelman h. , 1997 , apj , 483 , 868 .
safi - harb s. , petre r. , 1999 , apj , 512 , 784 .
sankrit r. , blair w. p. , borkowski k. j. , long k. s. , patnaude d. , raymond j. c. , reynolds s. p. , williams b. j. , 2015 , baas , 225 , 140.20 .
schaudel d. , becker w. , voges w. , aschenbach b. , reich w. , weisskopf m. , 2002 , neutron stars in supernova remnants , asp conference series , held in boston , ma , usa , 14 - 17 august 2001 , ed .
patrick o. slane and bryan m. gaensler , san francisco : asp , 2002 , 271 , 391 .
schnitzeler d. , banfield j. , emonts b. , hales c. , lenc e. , norris r. , mcclure - griffiths n. , mcconnell d. , osullivan s. , stevens j. , 2011 , `` calibrating linear polarization data at 16 cm with atca / cabb '' , at technical memo series at/39.9/129 , retrieved from http://www.atnf.csiro.au/observers/memos/at39.9_129.pdf .
seward f. , grindlay j. , seaquist e. , gilmore w. , 1980 , nature , 287 , 806 .
sun x. h. , gaensler b. m. , carretti e. , purcell c. r. , staveley - smith l. , bernardi g. , haverkorn m. , 2014 , mnras , 437 , 2936 .
sun x. h. , et al .
, 2015 , aj , 149 , 60 . taylor j. h. , cordes j. m. , 1993 , apj , 411 , 674 . tenorio - tagle g. , bodenheimer p. , franco j. , rozyczka m. , 1990 , mnras , 244 , 563 .
tenorio - tagle g. , rozyczka m. , franco j. , bodenheimer p. , 1991 ,
mnras , 251 , 318 .
tribble p. c. , 1991 , mnras , 250 , 726 .
turtle a. j. , pugh j. f. , kenderdine s. , pauliny - toth i. i. k. , 1962 , mnras , 124 , 297 .
van den bergh s. , 1980 , apj , 236 , l23 .
van den bergh s. , tammann g. a. , 1991 , ara&a , 29 , 363 .
van eck c. l. , et al .
, 2011 , apj , 728 , 97 .
van marle a. j. , meliani z. , marcowith a. , 2015 , a&a , 584 , a49 .
velzquez p. f. , raga a. c. , 2000 , a&a , 362 , 780 . velusamy t. , kundu m. r. , 1975 , a&a , 41 , 307 .
wang c. , han j. l. , lai d. , 2011 , mnras , 417 , 1183 .
watson m. g. , willingale r. , grindlay j. e. , seward f. d. , 1983 , apj , 273 , 688 .
weaver r. , mccray r. , castor j. , shapiro p. , moore r. , 1977 , apj , 218 , 377 .
weisberg j. m. , cordes j. m. , kuan b. , devine k. e. , green j. t. , backer d. c. , 2004 , apjs , 150 , 317 . westerhout g. , 1958 , bulletin of the astronomical institutes of the netherlands , 14 , 215 .
whiteoak j. b. z. , green a. j. , 1996 , a&as , 118 , 329 .
wilson w. e. , et al . , 2011 , mnras , 416 , 832 .
zavala j. , velzquez p. f. , cerqueira a. h. , dubner g. m. , 2008 , mnras , 387 , 839 . zealey w. j. , dopita m. a. , malin d. f. , 1980 , mnras , 192 , 731 .
region 1 shows the eastern breakout region , where the eastern ear punches through the eastern boundary of the circular region of w50 , and is shown in fig .
[ figreg1 ] .
the images show the pseudo - colour iphas continuum - corrected h@xmath1 image , overlaid with the total intensity radio continuum contours from the vla image at 1.4 ghz @xcite , and with the linearly polarized intensity contours from the atca images at both 3.1 ghz and 2.2 ghz . for the regions discussed in appendices [ 2ndregion ] to [ 7thregion ]
, the representation of various emission by different colours and contours is the same as in fig .
[ figreg1 ] . the exact position of all the regions relative to the full extent of the w50/ss433 system
is shown in fig .
[ figregions ] .
the most striking features are the very brightest optical filaments , that have been identified before @xcite , although the quality of the iphas data allows us to observe them with unprecedented resolution .
the network of filaments is very extensive , and includes numerous previously - unidentified smaller filamentary structures adjacent to the main optical arcs .
as seen in projection , the entanglement of the filaments is highly pronounced . from 19@xmath15115@xmath15200@xmath153 , + 5@xmath64@xmath152 to 19@xmath15114@xmath15220@xmath153 , + 5@xmath64@xmath152 the edge of the radio continuum emission is traced out by an h@xmath1 filament . just southwest of this edge , near 19@xmath15114@xmath15215@xmath153 , + 5@xmath62@xmath152 is a bright compact radio source which is likely extragalactic . surrounding the compact source is diffuse radio emission associated with the very brightest h@xmath1 filaments .
the emission appears continuous and extended , and is brighter at radio wavelengths in the areas with brighter optical emission .
it is likely that improved angular resolution observations would be able to identify individual radio counterparts to each optical filament within the observing beam . to the east , at 19@xmath15114@xmath15240@xmath153 , + 5@xmath60@xmath152 another region of extended radio emission is associated with another group of optical filaments .
the edge of this extended emission is traced out by several filaments , notably in the southwest of the region , where the boundary of the radio emission drops off rapidly at the location of the optical filament .
there is another bright optical filament oriented almost exactly along north south and centred at 19@xmath15113@xmath15250@xmath153 , + 4@xmath652@xmath152 .
the filament approximately traces out the missing portion of the circular central part of w50 , where the jet punches through the periphery .
the filament has a counterpart in the radio continuum image , also aligned north
south , with a circular compact ` knot ' located near the centre of the vertical feature .
further southeast , near 19@xmath15114@xmath15235@xmath153 , + 4@xmath648@xmath152 there is another group of optical filaments
fainter than the main filaments that again appear to be preferentially oriented north south .
these features have faint diffuse radio emission associated with them , that again approximately traces out the boundary of the optical filaments .
to the west of this region , at 19@xmath15114@xmath15225@xmath153 , + 4@xmath646@xmath152 are four small optical filaments @xmath154 arcmin in length and again oriented north south .
these features have no radio counterpart , but are not related to any specific imaging artefact they are most likely real features associated with w50 itself .
there are no apparent counterparts to any of the optical filaments in the linearly polarized radio emission , either at 3.1 ghz or 2.2 ghz .
this is due to the foreground of diffuse h@xmath1 emission that is depolarizing the radio nebula , as discussed in section [ depol ] .
a significant number of the features in the west of this image have already been described in section [ 1stregion ] , although this image provides a zoomed - out perspective of the same region while also including views further south and east . to the south of the brightest optical nebulosity , even more faint optical filaments are identified .
the filaments described in section [ 1stregion ] at 19@xmath15114@xmath15235@xmath153 , + 4@xmath648@xmath152 extend even further south than was previously seen , with the southern extension curving steadily towards the east at lower declinations . at the most southern end of these filaments , at 19@xmath15114@xmath15250@xmath153 , + 4@xmath640@xmath152 , and where the optical filaments are almost below the detection threshold , there is diffuse radio emission that is slightly extended along the length of the filaments and which we therefore interpret as a radio counterpart . further west , at 19@xmath15114@xmath15210@xmath153 , + 4@xmath640@xmath152 is another optical filament that is again largely oriented north south albeit with several curves such that the filament appears to have an ` s'-shape .
there is nearby diffuse radio emission , that is brightest towards the southwest of the ` s ' , although there are no obvious morphological similarities between the optical and radio continuum emission
. the radio emission may be unrelated to the ` s ' feature , which may therefore lack a counterpart at radio wavelengths .
further south from the ` s ' feature , are two more small filaments .
the brightest of the two is located at 19@xmath15114@xmath15210@xmath153 , + 4@xmath636@xmath152 and is @xmath154 arcmin in length and oriented at @xmath155 towards the east .
a fainter filament is located at 19@xmath15114@xmath15220@xmath153 , + 4@xmath634@xmath152 .
neither of these two filaments appear to have a radio counterpart .
there are an additional two optical filaments nearby to where the radio continuum emission begins to increase rapidly in brightness towards the eastern ear , and located at 19@xmath15115@xmath15215@xmath153 , + 4@xmath653@xmath152 and 19@xmath15115@xmath15210@xmath153 , + 4@xmath647@xmath152 .
these features also have no obvious radio continuum counterpart .
there are no optical filaments associated with the end of the eastern ear itself .
the radio continuum chimney feature described by @xcite is visible extending to the north from the outskirts of the ear , and has no optical counterparts .
the foreground of diffuse h@xmath1 emission described in section [ depol ] is visible running through the image .
the diffuse h@xmath1 appears to decline in brightness towards the eastern ear , with the boundary approximately aligned with the north
south radio filament at 19@xmath15115@xmath15230@xmath153 , + 4@xmath654@xmath152 . in combination with the brightest radio filament further towards the east , and also with the mentioned optical filaments
, there appears to be a significant number of features that are approximately aligned north south .
we interpret these features as being related to the jet axis of ss433 which is aligned approximately perpendicularly along east west .
the optical and radio filaments are then all consistent with shocked material due to ram pressure from the jets .
there are no apparent counterparts to any of the southern optical filaments in the linearly polarized radio emission , either at 3.1 ghz or 2.2 ghz . similarly to the filaments described in section [ 1stregion ] , this is due to the foreground of diffuse h@xmath1 emission that is depolarizing the radio nebula .
however there is very strong linearly polarized emission from the bright radio filaments in the eastern ear .
the linear polarization is almost entirely associated with the termination shock of the eastern jet , indicating a well - ordered magnetic field and a weakly depolarizing foreground .
this polarized emission appears to terminate relatively abruptly along the edge of the diffuse h@xmath1 emission , as discussed in section [ depol ] . due to
spectral index effects , there is a region to the north of the ear where the polarized intensity is brighter at lower radio frequencies , although this places no constraint on whether the polarized fraction is decreasing .
there are other regions clearly showing polarized intensity , particularly located along the eastern and southern edges of the image .
this is most likely diffuse galactic polarized emission , as described in section [ linpol ] .
region 3 shows the western ear , the western breakout region where the jet punches through the western boundary of the circular region of w50 , and where the radio emission from the western ear terminates , and is shown in fig .
[ figreg3 ] .
the most striking features are the very brightest optical filaments , that have been identified before @xcite , although the quality of the iphas data allows us to observe them with unprecedented resolution .
the pre - identified filaments are those that trace out the apparent circular edge of the central region of w50 , extending from 19@xmath15110@xmath15215@xmath153 , + 5@xmath610@xmath152 down to 19@xmath1519@xmath15230@xmath153 , + 4@xmath655@xmath152 .
we also identify an additional filament that traces out the periphery of the southern central region , which is located at 19@xmath1519@xmath15225@xmath153 , + 4@xmath644@xmath152 .
all of these filaments are located along the approximately circular boundary of w50 .
there are radio continuum contours associated with almost all of these features , with another north
south feature seen tracing out the brightest optical filament .
this is consistent with shocked material due to ram pressure from the jets of ss433 , and is similar to the north
south features also seen in the eastern ear ( see section [ 2ndregion ] ) .
a small h@xmath1 filament of @xmath154 arcmin in length is located at 19@xmath15110@xmath15240@xmath153 , + 4@xmath643@xmath152 , and is both coincident with , and oriented in the same direction as the bright radio filament that extends from near to ss433 and curves down towards the southwest .
the full extent of this radio filament lies outside of fig .
[ figreg3 ] , but is more clearly visible in fig . [ figrgb ] .
we find no other optical features associated with this radio filament . along the northern edge of the western radio ear , a bright h@xmath1 filament traces out the entire northern boundary that is visible in the radio continuum images .
a network of filaments extend further south into the interior of the western ear itself , with several bright knots of emission
. the northeastern part of the ear , located at 19@xmath1519@xmath15245@xmath153 , + 5@xmath615@xmath152 also contains fainter filaments that connect the outer ear to the central region of w50 .
the filaments very closely follow the boundary of the radio structure .
there are no visible filaments associated exclusively with the circular portion of w50 the location is most consistent with jet activity .
interestingly there is a further filament located to the far west , approximately @xmath156 pc further than the western edge of the radio ear .
the alignment of the filament with the general shape of the western ear suggests that it is related to the w50/ss433 system and implies that ss433 is influencing the surrounding ism at greater distances than is typically deduced from the radio morphology .
there is also considerable radio linear polarization associated with the central portion of w50 .
however , any intrinsic linear polarization associated with the ear is depolarized by the foreground diffuse h@xmath1 ( see section [ depol ] ) .
there is a clear difference between the 3.1 ghz and 2.2 ghz distributions of polarized emission , with the 3.1 ghz distribution reaching out to a larger radius from ss433 and consistent with w50 being a magnetised spherical structure . at lower frequencies , the depolarization effects are more considerable
, such that only polarization on the near side of w50 can reach us . in projection , this appears as a shrinking radius of the polarized emission at lower radio frequencies .
this suggests that the depolarization happens inside w50 itself .
this is further justified as the inner part of the 2.1 ghz data is more rapidly beginning to be depolarized than the outer ring , as would be expected due to the longer line - of - sight through the turbulent inner gas in the central region of w50 .
this trend becomes more apparent at the even lower frequencies down to 1.4 ghz , as shown in fig .
[ fig2 ] , with the ring both shrinking in radius , and also becoming more ` ring - like ' , as the central portion becomes a polarized hole .
there are largely no connections between the linear polarization and the optical filaments , with the notable exception of the @xmath154 arcmin in length filament located at 19@xmath15110@xmath15240@xmath153 , + 4@xmath643@xmath152
. the full polarized distribution in this region can be seen from rm synthesis in fig .
[ figrgb ] . the optical filament coincides with the edge of a bright peak in the linearly polarized emission .
the shocking of material in this filament has therefore possibly given rise to a compressed and well - ordered magnetic field , although it may also be a coincidental alignment .
region 4 shows the northeastern interior of w50 , including ss433 within the field of view , and is shown in fig .
[ figreg4 ] . in this region , there is no optical emission visible in the area in closest proximity to ss433 .
this is consistent with the previously inferred patchy absorption of optical emission across w50 ( see section [ discussion ] and @xcite ) . to the far southeast of the region
are the bright optical filaments associated with the eastern breakout region and which were discussed in section [ 1stregion ] . from the east , there are several fainter h@xmath1 filaments that appear to stretch inwards towards the interior of the radio nebula . the largest two of these filaments are oriented approximately along the east west axis , and become increasingly wispy towards the east ( possibly implying that they are associated with the jets of ss433 ) . of these two larger filaments , the brighter one to the north will be referred to as filament ` a ' , and the fainter one to the south will be referred to as filament ` b ' .
filament a has a direct radio continuum counterpart .
this radio counterpart is a linear filamentary feature and appears to be offset very slightly so that the emission originates along the southern edge of the h@xmath1 emission .
while filament b has no obvious radio counterpart , to the west there is an associated linear radio feature leading directly between the end of the h@xmath1 filament and that terminates at ss433 itself .
the h@xmath1 filament and the associated radio filament both appear to be part of one , aligned , physical structure . by interpreting this feature as a single filament
, this further suggests that filament b is associated with the jets of ss433 .
the implication is that this filamentary feature traces the magnetic field lines .
the lack of optical emission near to ss433 and the abruptness of the transition is consistent with the inferred patchy absorption @xcite . for only the western edge of the filament to be bright in radio continuum
suggests that the charged particles in the eastern edge are : either being swept - away so as not being constantly replenished ( although ram - pressure stripping would also leave an imprint on the h@xmath1 filament ) , or that the magnetic field is increasingly ordered further from ss433 such that incident electrons are expelled before they are able to significantly contribute to the synchrotron emission , in a manner similar to that theorised for linear radio filaments near to the galactic centre @xcite .
this would imply that either the filament rapidly transitions to a region that contains a very strong magnetic field , on the order of 100 @xmath124 g , or that the filaments are possibly aligned as seen in projection . in linear radio polarization ,
both filament a and b appear to be depolarizing the diffuse polarized emission from the radio nebula , with even the small extension to the north - west from filament a having a possibly offset counterpart in the radio polarization .
note the very different intrinsic angular resolution of both the radio polarization and h@xmath1 data .
this implies that the filaments depolarize the radio shell , and so the filaments must be located in the foreground of the magnetoionic layer in the shell from which the polarization is being emitted .
this possibly suggests that the polarization in this region emanates from the rear - side of the shell at higher - frequencies .
the brightest radio polarization throughout the region is dominated by the emission in the northern and northwestern rim as discussed in sections [ 5thregion ] and [ 6thregion ] .
region 5 shows a small section from the northern rim of w50 and is shown in fig .
[ figreg5 ] . in this small region
, diffuse h@xmath1 appears to fill the field of view .
there are again several h@xmath1 filaments , near to the sensitivity limit of the observations . to the very top of the region
are three parallel filaments located at 19@xmath15112@xmath15215@xmath153 , + 5@xmath627@xmath152 , each running tangentially to the shell of the circular region of w50 . at the bottom of the region
is another thicker filament , located at 19@xmath15112@xmath15210@xmath153 , + 5@xmath621@xmath152 , with some indication of a bifurcation at the western edge of the filament . in the radio continuum ,
the three parallel filaments are coincident with a peak in the total intensity and must therefore also contain a peak in the density of charged particles and/or a localised strong magnetic field .
these filaments are located at the northern edge of the radio nebula , and trace out the periphery of w50 .
the filament at the bottom of the region very approximately traces out a boundary where the radio continuum rapidly increases in brightness , however the filament is offset at an angle unrelated to the radio emission , and we interpret the approximate alignment as likely being coincidental . at the noisiest levels of the image , there is a very faint tentative indication of a possible bridge filament that links the parallel filaments with the filament further to the south . in radio polarization ,
there are numerous linearly polarized features throughout the region , with a strong variation from 3.1 ghz to 2.2 ghz that indicates significant faraday depolarization effects .
there is no obvious connection between the optical and the linearly polarized radio features , suggesting that while the polarized emission is originating from within the radio nebula , it is from nearer to the observer along the line - of - sight .
region 6 shows a small section from the northwestern rim of w50 and is shown in fig .
[ figreg6 ] . throughout the h@xmath1 image
, the field appears to be full with diffuse and patchy emission .
the region indicated by the white box in fig .
[ figreg6 ] shows the approximate area in which h@xmath1+nii emission has been previously reported , albeit without continuum - subtraction @xcite .
there are no identified h@xmath1 filaments within this box . due to the location of the diffuse nebulosity ,
it is not clear whether the diffuse emission measured by @xcite is associated with w50 itself or with the surrounding ism .
the optical image again shows numerous optical filaments .
most appear to have radio continuum counterparts , although the field is essentially full of emission .
interestingly , an h@xmath1 filament located at 19@xmath15111@xmath15210@xmath153 , + 5@xmath613@xmath152 , appears to trace out a boundary of the 3.1 ghz polarized radio emission , and which has depolarized at 2.2 ghz .
region 7 shows the snr candidate g38.71.4 and is shown in fig .
[ figreg7 ] . at this outer edge of the large optical mosaic ,
there are several imaging artefacts as shown by the ` blocky ' features and bright horizontal / vertical stripes .
this supernova remnant candidate was first identified in the rosat all - sky survey @xcite .
the presence of optical emission indicates that this object is indeed most likely an old supernova remnant , and this has previously been reported using iphas data in @xcite ( which reports the snr as g038.7 - 1.3 ) .
the work of @xcite also agrees that it is an old snr .
optical filamentary emission clearly traces out the rim between the edge of the radio continuum emission and the surrounding ambient medium , and additional diffuse optical emission is coincident along the entire eastern part of the radio shell .
no linearly polarized radio emission is associated with g38.71.4 at these sensitivity levels .
as the supernova remnant is completely depolarized across the entire atca band , the depolarization could be associated with the h@xmath1 emission itself , or alternatively , g38.71.4 could be depolarized due to bandwidth depolarization which would imply a very high rm@xmath551,997 rad m@xmath3 ( see section [ imaging ] ) .
it is most likely that depolarization effects are strong due to the same foreground variations that are affecting detection of linear polarization from w50 ( see section [ depol ] ) . referring back to fig .
[ figreg7 ] , there is also clearly h@xmath1 emission associated with s74 ( sh2 - 74 ) , located in the galactic plane to the northwest of w50 ( e.g. * ? ? ?
sh2 - 74 is a partially - obscured hii region that is situated at a distance of more than 3.1 kpc , on the outer edge of the sagittarius arm .
it is therefore in the foreground of w50 .
the diffuse h@xmath1 ( see section [ depol ] ) that depolarizes the western ear is likely associated with this complex , and is therefore feasibly in front of g38.71.4 ( hence the snr candidate s complete depolarization ) .
this places a weak distance estimate of @xmath157 kpc on the distance to g38.71.4 .
|
the w50/ss433 system is an unusual galactic outflow - driven object of debatable origin . we have used the australia telescope compact array ( atca ) to observe a new 198 pointing mosaic , covering @xmath0 , and present the highest - sensitivity full - stokes data of w50 to date using wide - field , wide - band imaging over a 2 ghz bandwidth centred at 2.1 ghz .
we also present a complementary h@xmath1 mosaic created using the isaac newton telescope photometric h@xmath1 survey of the northern galactic plane ( iphas ) .
the magnetic structure of w50 is found to be consistent with the prevailing hypothesis that the nebula is a reanimated shell - like supernova remnant ( snr ) , that has been re - energised by the jets from ss433 .
we observe strong depolarization effects that correlate with diffuse h@xmath1 emission , likely due to spatially - varying faraday rotation measure ( rm ) fluctuations of @xmath2 to 61 rad m@xmath3 on scales @xmath4 to 6 pc .
we also report the discovery of numerous , faint , h@xmath1 filaments that are unambiguously associated with the central region of w50 .
these thin filaments are suggestive of a snr s shock emission , and almost all have a radio counterpart . furthermore , an rm - gradient is detected across the central region of w50 , which we interpret as a loop magnetic field with a symmetry axis offset by @xmath590@xmath6 to the east west jet - alignment axis , and implying that the evolutionary processes of both the jets and the snr must be coupled .
a separate rm - gradient is associated with the termination shock in the eastern ear , which we interpret as a ring - like field located where the shock of the jet interacts with the circumstellar medium .
future optical observations will be able to use the new h@xmath1 filaments to probe the kinematics of the shell of w50 , potentially allowing for a definitive experiment on w50 s formation history .
= 1 [ firstpage ] magnetic fields polarization ism : bubbles ism : jets and outflows
ism : supernova remnants radio continuum : ism
|
most of the higgs efforts of the cdf and collaborations are centered on finding a standard model higgs boson .
however the standard model is incomplete and there are proposed extensions to the standard model , many of which feature one or more higgs bosons as well . these are often similar to the standard model higgs , but differ in their couplings to other particles . for the purposes of this presentation ,
i divide the higgs searches into two categories , fermiophobic higgs searches and susy higgs searches at high @xmath0 .
in the fermiophobic higgs models , the higgs is assumed not to couple to fermions . at high masses
it then decays to a w pair and at low masses to a photon pair .
the high mass searches are done in the channel , searching for a higgs boson produced in association with a w boson .
this is done in order to avoid the higher backgrounds associated with the channel . the search for @xcite looks for two same sign leptons and vetos on a third lepton .
the sample is then divided according to lepton types .
the final limit is extracted from the dilepton invariant mass distribution .
the results can be seen in figure [ p03 ] .
the cdf search for @xcite also looks for exactly two leptons of the same sign and divides the sample by lepton type , but it performs a simple counting experiment instead of extracting the limit from the dilepton invariant mass .
the results can be seen in figure [ p04 ] . performed a search for @xcite . that search employs a neural network to distinguish between photons and jets .
the diphoton mass spectrum is then scanned for an excess and a mass dependent limit is set .
the results can be seen in figure [ p05 ] . also performed a search for @xcite .
the basic idea is that the vertex @xmath1 is suppressed and instead the vertex @xmath2 is employed , where @xmath3 is another higgs boson .
the search only looks for three photons , allowing one photon to be lost .
a cut on the transverse momentum of the photons is employed to separate out the background , and the remaining events are treated as a counting experiment .
the results can be seen in figure [ p06 ] .
the susy higgs searches exploit the fact that at high @xmath0 the coupling of the higgs to bottom quark and tau leptons is enhanced with @xmath4 , which has a number of effects that are exploited in the search .
the higgs cross section increases from femtobarns to picobarns .
the higgs is often produced by the fusion of two bottom quarks producing a final state with two extra bottom quarks , which can be tagged to reduce backgrounds . furthermore the higgs decay changes and is in 90% of cases a pair of bottom quarks and in the remaining 10% into a pair of tau leptons .
cdf performed a search for @xcite , where the choice of one associated bottom jet was due to the better signal to background ratio compared to zero or two associated bottom jets . in the search the two and three jet distributions are fit to obtain the signal and background distributions . from this , mass dependent cross section limits are obtained , and @xmath0 limits which take the higgs width effects at high @xmath0 .
the results can be seen in figure [ p09 ] . also performed a search for @xcite . for the search
the data is split into three , four and five jet bins . within these bins
the two and three jet distributions are fit to obtain the signal and background contributions . from these the mass dependent @xmath0 limit is extracted .
the results can be seen in figure [ p10 ] .
cdf performed a search for @xcite . due to the higher purity of the ditau sample
no additional bottom jets are required .
the sample is split according to the tau decay mode .
the signal template is fit to the ditau visible mass spectrum to obtain the mass and model dependent @xmath0 limits .
the results can be seen in figure [ p11 ]
. performed two searches for higgs decaying into a tau pair @xcite .
the one search is in the channel . for this
the sample is divided by tau decay type and run period .
the signal template is then fit to the visible mass spectrum to extract the model and mass dependent @xmath0 limit .
the results can be seen in figure [ p12 ] .
the other search is in the search , which tags an additional bottom quark to reduce the background rate .
one tau is required to decay to a muon , while the other is required to decay hadronically .
the sample is then divided according to the hadronic tau decay type . in each decay category
a counting experiment is performed to extract the model and mass dependent @xmath0 limits .
the results can be seen in figure [ p13 ] .
the cdf and collaborations have performed a large number of searches for signs of a fermiophobic or susy higgs . in the absence of signal
, limits have been set on the various models . for the near future ,
updates to several of these searches can be expected , as well as a combination of these searches into an overall tevatron limit .
|
we present an overview of the full range of higgs searches in models beyond the standard model at the tevatron .
this includes both searches for fermiophobic higgs and for susy higgs at high @xmath0 .
no excess is seen in the data , so model dependent limits are set .
|
we use the data compilation by caldwell & bergvall ( 2006 ) consisting of a sample of 1199 close to edge - on ( @xmath0 ) low surface brightness ( lsb ) galaxies in the fourth data release of the sloan digital sky survey ( sdss - dr4 ) .
all galaxies in the sample were selected in order to obtain a sample of isolated , bulge - less systems , which is ideal for comparison with simple cems since the merger histories of such galaxies have less influence on the evolution of elemental ratios .
this total sample is divided by @xmath1 colour limits into a blue sample a ( 377 galaxies ) , a green ( intermediate ) sample b ( 436 galaxies ) and a red sample c ( 386 galaxies ) . from these 1199 galaxies we pick out the ones with a relative error in the h@xmath2-flux less than @xmath3 .
we have computed a large grid ( 5000 models ) of one - zone cems , testing time - scales for infall from @xmath4 gyr to @xmath5 gyr , baryon densities ranging between @xmath6 pc@xmath7 and @xmath8 pc@xmath7 , which should cover most of the possible variations within dwarf and late - type galaxies .
star formation is prescribed by a simple schmidt - law , @xmath9 , where the constant @xmath10 is varied between @xmath11 to @xmath12 .
furthermore , we have tried several power - law initial imfs @xmath13 in order to test the bottom - heavy imf hypothesis ( @xmath14 ) suggested by lee et al .
the nucleosynthesis prescriptions are taken from chieffi & limongi ( 2004 ) for high mass stars and van den hoek & groenewegen ( 1997 ) for low and intermediate stars and all model tracks where evolved over 13.7 gyr ( one hubble - age )
. observed abundances and model results ( grey dots ) for different imfs . slowly evolving tracks
populate the @xmath15-plateau , while rapidly evolving tracks populate the `` secondary branch '' .
the abundances of the late - type lsbgs are shown as blue diamonds ( sample a ) , green triangles ( sample b ) and red squares ( sample c ) .
black dots and circles show data for dwarf lsbgs from rnnback & bergvall ( 1995 ) , van zee et al .
( 1996 ) and van zee & haynes ( 2006 ) . , title="fig:",width=163 ] observed abundances and model results ( grey dots ) for different imfs . slowly evolving tracks
populate the @xmath15-plateau , while rapidly evolving tracks populate the `` secondary branch '' .
the abundances of the late - type lsbgs are shown as blue diamonds ( sample a ) , green triangles ( sample b ) and red squares ( sample c ) .
black dots and circles show data for dwarf lsbgs from rnnback & bergvall ( 1995 ) , van zee et al .
( 1996 ) and van zee & haynes ( 2006 ) . , title="fig:",width=163 ] observed abundances and model results ( grey dots ) for different imfs . slowly evolving tracks
populate the @xmath15-plateau , while rapidly evolving tracks populate the `` secondary branch '' .
the abundances of the late - type lsbgs are shown as blue diamonds ( sample a ) , green triangles ( sample b ) and red squares ( sample c ) .
black dots and circles show data for dwarf lsbgs from rnnback & bergvall ( 1995 ) , van zee et al .
( 1996 ) and van zee & haynes ( 2006 ) .
, title="fig:",width=163 ]
in fig . 1 we show a comparison between our abundances derived from spectral data from sdss - dr4 using the so - called p - method ( see , e.g. , pilyugin 2003 ) and our model results .
clearly , the late - type lsbgs do not deviate from the general , rising trend found in hsb galaxies .
based on the results from our model grid we draw the following conclusions : + - the @xmath16-ratios in dwarf irregular and late - type lsbgs follow the same basic pattern as their hsb counterparts .
+ - the chemical evolution , as revealed by the hii - regions , is not compatible with an extremely bottom - heavy imf as suggested by lee et al ( 2004 ) , nor is it compatible with a top - heavy imf .
+ - the lsb property of these objects is most likely due to a low global star formation density , although the efficiency of star formation in star - forming regions must be as high as in hsbgs in order to explain the high @xmath15-ratios .
+ - given these results , it is unlikely ( although not impossible ) that the ages of lsbgs are significantly different from the ages of corresponding hsbgs .
caldwell b. & bergvall n. , 2006 , in the proceedings of the iau symposium no .
235 chieffi a. & limongi m. , 2004 , , 608 , 405 van den hoek l.b . & groenewegen m.a.t .
, 1997 , , 123 , 305 lee h - c .
, gibson b.k . ,
flynn c. , kawata d. & beasley m.a . , 2004 , , 353 , 113 pilyugin l.s . , thuan t.x .
& vilchez j.m . , , 2003 , 397 , 487 rnnback j. & bergvall n. , 1995 , , 302 , 353 van zee l. , haynes m. & salzer j.j . , 1997 , , 114 , 2479 van zee l. & haynes m. , 2006 , , 636 , 214
|
we have derived oxygen and nitrogen abundances of a sample of late - type , low surface brightness ( lsb ) galaxies found in the sloan digital sky survey ( sdss ) .
furthermore , we have computed a large grid ( 5000 models ) of chemical evolution models ( cems ) testing various time - scales for infall , baryon densities and several power - law initial mass functions ( imfs ) as well . because of the rather stable n / o - trends found both in cems ( for a given imf ) and in observations , we find that the hypotheses that lsb galaxies have stellar populations dominated by low - mass stars , i.e. , very bottom - heavy imfs ( see lee et al .
2004 ) , can be ruled out .
such models predict much too high n / o - ratios and generally too low o / h - ratios .
we also conclude that lsb galaxies probably have the same ages as their high surface brightness counterparts , although the global rate of star formation must be considerably lower in these galaxies .
|
the internal structure of the nucleon has long been studied with inclusive deep - inelastic lepton scattering . from these studies , the spin - dependent and spin - independent nucleon structure functions and corresponding parton distribution functions
have been extracted .
recently , the formalism of generalized parton distributions ( gpd ) @xcite-@xcite was introduced . in this framework ,
non - forward degrees of freedom are probed in exclusive processes such as meson production or deeply virtual compton scattering ( dvcs ) , the electroproduction of a single real photon on the nucleon , which has been studied at desy @xcite-@xcite and jefferson lab @xcite .
this opens the way to access the distribution of partons in the nucleon perpendicular to its momentum @xcite , and to access the contribution of the orbital angular momenta of the quarks to the nucleon spin @xcite .
the hera storage ring at desy provides an 820 - 920 gev proton beam and a 27.6 gev electron or positron beam to the two collider experiments , h1 and zeus . at collider energies , the leading order handbag diagram ( fig .
[ fig : lo+nlo ] a ) is accompanied by a sizable contribution of colour singlet two - gluon exchange ( fig .
[ fig : lo+nlo ] b ) . in both cases ,
the final state is indistinguishable from that of the bethe - heitler ( bh ) process , in which the real photon is radiated off the lepton .
the bh process constitutes a background to the dvcs process .
since it is purely electromagnetic , its amplitude can be calculated exactly .
this contribution is subtracted from the combined dvcs+bh data by means of a monte carlo simulation that is cross checked with a bh - dominated sample of the experimental data @xcite .
dependence of dvcs cross section for the h1 data , gpd predictions ( light shade ) , a colour dipole model ( dark shade ) .,height=215 ] dependence of dvcs cross section for the 1997 h1 data , overlays as in fig .
.,height=215 ] dependence of dvcs cross section for the zeus positron data , overlays as in fig .
.,height=215 ] dependence of dvcs cross section for the zeus positron data , overlays as in fig .
.,height=215 ] the cross section from the h1 experiment , shown in figs .
[ fig : h1q2 ] and [ fig : h1w ] , was measured in the 1997 running period with an 820 gev proton energy .
the sample corresponds to an integrated luminosity of 8 pb@xmath5 .
the data cover a kinematic range in four - momentum transfer squared @xmath1 from 2 to 20 gev@xmath2 , invariant mass of the photon - nucleon system @xmath3 from 30 to 120 gev and the square of the momentum transfer to the proton @xmath6@xmath01 gev@xmath2 , over which the cross section is integrated .
the data in figs .
[ fig : h1q2 ] and [ fig : h1w ] have been scaled to central values of @xmath3=75 gev and @xmath1=4.5 gev@xmath2 , respectively . here , also theoretical predictions for dvcs at h1 kinematics from a gpd - based model ( ffs , @xcite ) , as well as from a colour dipole model ( dd , @xcite ) , are shown . in both models ,
the exponential @xmath7-dependence factorizes from the @xmath1 and @xmath3 dependences .
since the @xmath7-dependence of the cross section was not measured , upper and lower bounds are determined by taking values of the @xmath7-slope parameter @xmath8 of 5 and 9 gev@xmath9 .
a newer gpd - based approach can be found in @xcite .
the zeus results are based on data taken between 1996 and 2000 , and , for the first time , also make use of the hera electron beam .
the integrated luminosities are 95 pb@xmath5 for @xmath10 and 17 pb@xmath5 for @xmath11 collisions .
the data were analyzed in the kinematical region defined by 5@xmath0@xmath1@xmath0100 gev@xmath2 and 40@xmath0@xmath3@xmath0140 gev , with no cut on @xmath7 .
the data were found to agree for both lepton charges .
the @xmath12 data shown in figs . [ fig : zeus - q2 ] and [ fig : zeus - w ] are scaled to central values of @xmath3 = 89 gev and @xmath1 = 9.6 gev@xmath2 , respectively .
the data are overlaid with theoretical predictions from the same models as the h1 data . due to the larger kinematical range and integrated luminosity ,
the zeus data have enough leverage and statistical accuracy to discriminate between different theoretical predictions and can be used to constrain the predicted @xmath1 dependence of theoretical models . here , the zeus data are seen to favour the less steep slope in @xmath1 of the colour dipole model .
the @xmath3-dependence of the data is described well in both cases , reflecting the steep rise in gluon density at lower @xmath13 , the momentum fraction of the struck parton in the nucleon .
the zeus and h1 data can not be compared directly due to the different kinematic ranges . however ,
if the observed functional behaviour in @xmath1 is used to scale the h1 data to the zeus central kinematics , the two data sets agree . moment of the beam charge asymmetry from hermes data from 1998 and 2000.,height=238 ] moment of the beam spin asymmetry from hermes data from 1996/1997 ( empty symbols ) and 2000 ( filled symbols).,height=238 ] the fixed target experiment hermes makes use of the 27.6 gev polarized hera positron beam incident on a storage cell that is filled with unpolarized gas ( hydrogen , deuterium , or heavier nuclei ) . at hermes kinematics , the bh amplitude largely dominates the dvcs amplitude .
however , the dvcs and bh amplitudes interfere , giving rise to a single - spin asymmetry , which depends on beam charge , helicity , and the azimuthal angle @xmath14 between the lepton scattering plane and the reaction plane of the real photon and recoil proton . by extracting the @xmath15 and @xmath16 moments of the exclusive data
, access is obtained to certain combinations of dvcs amplitudes @xcite .
the @xmath15 moment of the beam spin asymmetry is calculated as @xmath17 where @xmath18 is the beam polarization for each event .
the @xmath16 moment is determined similarly from the beam charge asymmetry with an unpolarized beam .
the hermes forward spectrometer detects the scattered lepton and the real photon , but not the recoiling proton .
therefore the exclusivity of an event is measured through the missing mass ( @xmath19 ) of the event , which in the exclusive case is equal to the proton mass .
examples are shown in figs .
[ fig : hermes - bca ] and [ fig : hermes - bsa-96 - 2000 ] , in which the @xmath20 moment of the beam charge asymmetry for the 1998 and 2000 data , and the @xmath15 moment of the beam spin asymmetry for the 1996/1997 data @xcite and the 2000 data are shown .
asymmetries are observed at values of @xmath19@xmath01 - 2 gev , in the exclusive region . at higher missing masses ,
where non - exclusive processes dominate , asymmetries vanish .
the two regions overlap slightly , because the resolution in @xmath19 is limited by the energy resolution of the electromagnetic calorimeter .
the @xmath15 measurements were confirmed at jefferson lab @xcite .
gpd - based predictions can be found in refs .
@xcite . , @xmath21 and @xmath7-dependences of the @xmath15 moment of the beam spin asymmetry from hermes data from 1996 and 1997 . ] from the exclusive sample ( @[email protected] gev ) of the 1996/97 data the kinematic dependences of the beam spin asymmetry were extracted as functions of @xmath1 , @xmath13 and @xmath7 ( preliminary results shown in fig . [
fig : hermes - bsa - kine ] ) . in the case of @xmath7 , as with @xmath19 , the reconstruction relies on the photon energy measured in an electromagnetic calorimeter
. therefore some events are reconstructed at @xmath22@xmath00 .
the hera run ii starting in 2002 will be used by all three experiments to improve the dvcs measurements . in the 2001 shutdown ,
spin rotators were installed before and after the collider experiments h1 and zeus , thus allowing for spin - dependent measurements .
it is planned to run alternately with electron and positron beams .
the h1 experiment will install a very forward proton spectrometer in 2003 @xcite , allowing it to measure @xmath6 at small values of @xmath3 .
in addition , the precise tracking data from the backward silicon tracker will be used for a measurement of azimuthal asymmetries .
the measurement of differences in the cross section for the available beam polarizations and charges will be the main goal of the zeus dvcs analysis .
the hermes experiment will improve the exclusivity of its dvcs sample , as well as the @xmath7-resolution , by installing a large acceptance recoil detector around the hermes target @xcite . +
* acknowledgements * 9 d. mller _ et al .
_ , fortschr .
* 42 * ( 1994 ) 101 .
a. v. radyushkin , phys .
* b 385 * ( 1996 ) 333 .
x. ji , phys .
* d 55 * ( 1997 ) 7114 . c. adloff _ et al .
_ , h1 collaboration , phys .
b 517 ( 2001 ) 47 - 58 .
zeus collaboration , proceedings of the international europhysics conference on high energy physics , budapest , 2001 .
a. airapetian _ et al . _ ,
. lett . 87 ( 2001 ) 182001
. s. stepanyan _ et al .
_ , phys .
lett . 87 ( 2001 ) 182002 . j. p.
ralston , b. pire , cpht - s038 - 0901 ( 2001 ) , hep - ph/0110075 .
x. ji , phys . rev .
* 78 * ( 1997 ) 610 .
l. l. frankfurt , a. freund and m. strikman , phys .
* d*58 ( 1998 ) 114001 .
a. donnachie and h. g. dosch , phys .
b 502 ( 2001 ) 74 .
a. freund and m. mcdermott , eur .
j. c 23 ( 2002 ) 651 , phys .
rev . * d*65 ( 2002 ) 091901 ; a. freund , m. mcdermott and m. strikman , hep - ph/0208160
. k. goeke , m. v. polyakov , m. vanderhaeghen , progr .
47 , no . 2 , 401 ( 2001 ) , hep - ph/0106012 . l. favart _ et al .
_ , proposal desy - prc-01/00 and h1 - 5/00 - 582 .
r. kaiser _ et al .
_ , technical design report , desy - prc 02 - 01 , hermes 02 - 003 .
|
deeply virtual compton scattering has recently been studied by three hera experiments , h1 , zeus and hermes , covering a broad range of kinematic regimes .
we present cross section measurements of the two collider experiments in the kinematic region 2@xmath0@xmath1@xmath0100 gev@xmath2 and 30@xmath0@xmath3@xmath0140 gev , and compare them to qcd - based calculations .
hermes measurements of azimuthal asymmetries and their kinematical dependences are presented for @xmath1@xmath41 gev@xmath2 and 2@xmath0@xmath3@xmath07 gev .
|
we consider the propagation of @xmath1 nucleons [ @xmath2 neutrons ( @xmath3 ) and @xmath4 protons ( @xmath5 ) ] in a nuclear medium described by the particle densities @xmath6 and @xmath7 as well as the temperature @xmath8 .
using a thermodynamic green function approach , the following in - medium wave equation can be derived ( see @xcite ) @xmath9\psi_{a \nu p}(1\dots a ) \nonumber \\ & & + \sum_{1'\dots a'}\sum_{i < j}[1- f(i)- f(j)]v(ij , i'j')\prod_{k \neq i , j } \delta_{kk'}\psi_{a \nu p}(1'\dots a')=0\ , . \label{wavea}\end{aligned}\ ] ] this equation contains the effects of the medium in the quasiparticle shift @xmath10 as well as in the pauli blocking terms @xmath11 $ ] .
the single - nucleon state @xmath12 is abbreviated by @xmath13 etc .
the single - nucleon quasiparticle shift @xmath14 is obtained from the single - nucleon self - energy which can be taken , for instance , in mean - field approximation .
standard expressions give a rigid shift in the energy and , if the momentum dependence is taken into account , the replacement of the nucleon mass @xmath15 by an effective mass .
the treatment of the single - nucleon shift in solving eq .
( [ wavea ] ) is trivial @xcite and will not be investigated here .
in particular , the rigid energy shifts can be absorbed by a shift @xmath16 of the cluster quasiparticle energy @xmath17 and will not be considered here any more , the same also with the shift of the coulomb energy .
note that the in - medium few - nucleon schrdinger equation ( [ wavea ] ) describes the medium modification not only of possible bound states that may be considered as quasi - particles , but also the modification of the scattering states . for exploratory calculations we use a simple separable interaction @xcite with gaussian form factor .
@xmath18 the parameter values can be fixed by experimental data . in the case of the deuteron ( @xmath19 ) , with potential parameters @xmath20 mev @xmath21 and @xmath22 @xmath23 in eq .
( [ sepa ] ) , the binding energy and rms radius of the free deuteron are reproduced .
our main concern are the effects of the pauli blocking terms where the phase space occupation is described by the fermi distribution function @xmath24^{-1}$ ] . here
, the chemical potential @xmath25 is related to the nucleon density @xmath26 . in the case of zero temperature ( @xmath27 ) , the fermi function becomes the step function @xmath28 where the chemical potential coincides with the fermi energy @xmath29 .
the pauli blocking term leads to a shift in the energy @xmath30 @xcite and the in - medium modification of the wave function or the scattering phase shifts .
the properties of bound states have been investigated elsewhere @xcite .
in particular we investigate the phase shifts and the consequences for the second virial coefficient .
different approaches are discussed .
we first consider the ( deuteron like ) case @xmath19 .
furthermore we will use the tamm - dancoff ( td ) expression @xmath11 \to [ 1- f(i)][1- f(j)]$ ] which neglects the hole - hole contributions that are of relevance in deriving the gap equation if pairing is considered . for the separable potential , the t - matrix can be solved . with the integral @xmath31 ^ 2\ ] ]
we can now calculate the scattering phase - shift depending on the energy @xmath32 of relative motion ( zero angular momentum ) : @xmath33 = \frac{{\rm im}\,{t(12,1'2',e ) } } { { \rm re}\,{t(12,1'2',e ) } } = \frac { -{\rm im}\,{i(e)}}{1+{\rm re}\,{i(e)}}\ ] ] note that for negative values of @xmath32 , the solution of @xmath34 gives the in - medium binding energy , for instance , of the deuteron ( isospin singlet channel ) .
obviously the bound state wave functions and energy eigenvalues @xmath35 as well as the scattering phase shifts become dependent on temperature and density because of the fermi distribution functions in the pauli blocking term .
we focus on the scattering phase shifts @xmath36 . for @xmath37 and @xmath38 ,
some results are shown in fig .
[ fig.1 ] .
= 0.1 mev scattering phase shifts for different densities , td blocking.,scaledwidth=60.0% ] = 5 mev scattering phase shifts for different densities , td blocking.,scaledwidth=60.0% ] having the scattering phase shifts to our disposal , we can calculate the second virial coefficient according to the generalized beth - uhlenbeck formula @xcite .
in addition to the bound state contribution @xmath39 $ ] ( @xmath40 dependence of the in - medium binding energy neglected ) we have the contribution of scattering states , @xmath41\right]\ ] ] if there exist a bound state ( below the critical density @xmath42 where , e.g. , @xmath43 @xmath44 , @xmath45 @xmath44 ) , it is possible to combine the bound state and the scattering state contribution to the second virial coefficient extending the scattering phase shift to the bound state energy , see fig .
[ fig.1 ] .
there is a smooth behavior near the mott point where the bound state is dissolved because of pauli blocking .
we give also a simple interpolation formula that can be used to estimate the contribution of the continuum correlations to the second virial coefficient and , furthermore , to the density fraction of correlated states : @xmath46 + n^2]}{0.008709 \exp(-0.15403 t ) + n^2/[-0.1012 + 0.08611 \exp(0.09432 t ) ] \exp[(26.3 + 483.8/t)n]}\ ] ] for @xmath47 mev . at lower temperatures , @xmath48 $ ] .
an important feature is that at very low temperatures the phase shifts are constant , @xmath49 , not only up to the edge of the continuum of scattering states but up to the fermi energy ( at @xmath27 , energy conserving nucleon - nucleon scattering for energies below @xmath50 is not possible because of pauli blocking ) .
thus , the bound states merge with the continuum of scattering states at zero temperature only if the bound state energy per nucleon becomes larger than the fermi energy . at higher temperatures these sharp structures in the phase shifts shown in fig .
[ fig.1 ] are washed out , see fig .
[ fig.2 ] . the bound state will be dissolved already when it merges with the continuum of scattering states ( @xmath51 when the self - energy rigid shifts are absorbed by the chemical potential ) .
coming back to the so - called particle - particle random - phase approximation ( pprpa ) @xcite , the pauli blocking term reads @xmath11 $ ] which takes the hole - hole contributions into account .
the in - medium phase shifts are shown in fig .
[ fig.3 ] for moderate temperatures of warm dense matter . at very low temperatures ,
sharp structures arise similar to the tamm - dancoff case , see fig .
[ fig.4 ] .
= 0.1 mev ; pprpa blocking . ] = 0.1 mev ; pprpa blocking . ] as in the tamm - dancoff case , we obtain a jump of the phase shifts at energies that are near to the energies of bound state formation , but there is also a general negative shift what resembles the effect of repulsion .
we will not discuss here the consequences due to the use of different expressions for the pauli blocking term ( the critical densities are now @xmath52 @xmath44 , @xmath53 @xmath44 ) . the difference is the term @xmath54 that is of higher order in the density , and one may argue that there are also further terms missing if we consider this order of density . however , there is a more fundamental difference between both expressions for the pauli blocking .
the tamm - dancoff expression is used , for instance , in the brueckner theory to calculate the quasiparticle energy shifts of the nucleons . on the other hand
, pairing may occur in nuclear matter at low temperatures .
the corresponding gorkov equation follows only if the pprpa expression for the pauli blocking is used .
a detailed discussion is given in ref .
because of the relatively high binding energy ( 28.3 mev ) , the @xmath55-particle is of special interest in describing correlations and cluster formation in warm dense matter .
the wave function of the free @xmath55-particle can be approximated solving the four - particle schrdinger equation with a pair potential ( [ sepa ] ) , where the potential parameters @xmath57 mev @xmath21 and @xmath58 @xmath23 are determined by the binding energy and the rms radius of the @xmath55-particle .
the medium modifications of the bound state energy and of the continuum states are of interest . within a variational approach
, results are obtained as described elsewhere @xcite .
as example , we give here the shift of the binding energy of the @xmath55-particle , see fig .
[ fig:4nuc ] . .
the continuum edge of free single - particle states is given by @xmath59 . at zero baryon density ,
the four - nucleon energy is given by the binding energy of the @xmath55 particle , @xmath60 mev .
with increasing density , the binding energy @xmath61 is decreasing due to the pauli blocking .
the four - nucleon bound state disappears at @xmath62 @xmath44 .
a fit to the calculated values is given in ref .
@xcite.,scaledwidth=60.0% ] using a variational approach , we considered the shift of the @xmath55 particle in zero temperature matter @xcite .
whereas at finite temperatures the bound state becomes dissolved already when the binding energy becomes zero , at zero temperature the bound state can disintegrate only when the free particle states are not pauli blocked . above the fermi energy the decay is possible . this way
we can understand why at finite temperatures the mott point is given by the zero binding energy condition , whereas at zero temperature the fermi energy is of relevance .
we conclude that at zero temperature the @xmath55-like bound state is not dissolved at @xmath63 @xmath44 where it merges in the continuum of unbound scattering states but at @xmath64 @xmath44 where it can decay into unbound , free states above the fermi energy because they are not occupied .
this value @xmath64 @xmath44 is in agreement with other approaches , in particular the fully antisymmetrized thsr formalism , see @xcite .
it should be mentioned that we may calculate the virial coefficient for the four - nucleon case by rescaling the two - nucleon result .
another important issue is the treatment of correlations in the medium .
already the pprpa expression for the pauli blocking leads to the formation of cooper pairs in the medium .
the thsr ansatz allows to treat four - nucleon correlations also in the medium , taking into account full antisymmetrization .
the author acknowledges the contributions of p. schuck , y. funaki , h. horiuchi , zhongzhou ren , a. tohsaki , chang xu , t. yamada , and bo zhou to this work .
|
the properties of few - body clusters ( mass number @xmath0 ) are modified if they are immersed in a nuclear medium .
in particular , pauli blocking that reflects the antisymmetrization of the many - body wave function is responsible for the medium modification of light clusters and the dissolution with increasing density . a more consistent description is given with takes also the contribution of correlations in the continuum into account .
the relation between cluster formation in warm dense matter and in nuclear structure is discussed .
|
let us start with the lagrangian @xmath2 \nonumber \\ & & + \sum_{\alpha , i } \
, h^\ast_{\alpha i } \ , \overline{n_{ri } } \ , \phi^{\dagger } \ , \ell_{l}^{\alpha } + \sum_{\alpha , i } h_{\alpha i } \ , \overline{\ell_{l}^{\alpha } } \phi \ , n_{ri } \label{eq : lag}\\ & & + \sum_{\alpha , i } \ , h^\ast_{\alpha i } \ , \overline{(\ell_{l } ^{\alpha}})^c \ , \phi^\dagger \ , ( n_{ri})^c + \sum_{\alpha , i } h_{\alpha i } \overline{(n_{ri})^c } \ , \phi \ , ( \ell_{l}^{\alpha})^c \nonumber \ ] ] as in ref.@xcite .
@xmath3 and @xmath4 .
@xmath5 are light leptons and @xmath6 is a scalar field . for concreteness the @xmath5 can be the left - handed lepton doublet of the standard model , @xmath6 can be the standard model higgs doublet .
we supress the su(2 ) indices carried by the lepton and the higgs fields which are contracted by the @xmath7 tensor . @xmath8 and @xmath9 are generation indices and we assume one heavy majorana field per generation of light leptons .
for simplicity we shall work with two generations .
the lagrangian may also be rewritten in terms of @xmath10 . in what follows
we shall replace @xmath11 by @xmath12 . since the right handed neutrinos @xmath13 decay to leptons while the left - handed neutrinos @xmath12 decay to antileptons , the lepton number violation is in the mass term for the heavy neutrinos .
note that the last two lines in the lagrangian above are equivalent as @xmath14 .
one loop effects give rise to mixing between heavy neutrino flavours as shown in fig . 1 .
4.25in.2 in the one loop effective mass term due to the diagrams in fig .
1 is given by @xmath15 ignoring the masses of the light leptons and of the scalar field , the expressions for @xmath16and @xmath17 are given in ref.@xcite as @xmath18 ( g_{\alpha ij}^{dis } - \frac{i}{2 } g_{\alpha ij}^{abs})\ ] ] @xmath19 ( g_{\alpha ij}^{dis } - \frac{i}{2 } g_{\alpha ij}^{abs})\ ] ] and @xmath20 ( g_{\alpha ii}^{dis } - \frac{i}{2 } g_{\alpha ii}^{abs})\ ] ] where @xmath21 .
the effective mass matrix @xmath22 in eq .
( [ eq : meff ] ) can be written as @xmath23 where the dispersive part of @xmath24 ( i.e. terms proportional to @xmath25 ) is included in @xmath26 and the absorptive part ( proportional to @xmath27 ) is included in @xmath28 .
the absorptive part reflects the fact that the particles in the loop have gone on shell . both @xmath26 and @xmath28
are hermitian but not @xmath29 .
this is analogous to the formulation used to obtain an effective mass matrix for neutral kaons@xcite .
the one loop corrections in @xmath26 can be absorbed away by renormalisation .
( the off - diagonal elements of @xmath26 are renormalised by non - diagonal elements of the wave function counterterms@xcite . )
however , there are no absorptive counterterms@xcite , nor are they needed for the absorptive parts are finite
. it will be the mixing via the absorptive terms that will be important for creating the lepton asymmetry .
since the effective mass matrix is non - hermitian it can not be diagonalised by a unitary matrix , i.e. @xmath30 , where @xmath31 is diagonal , but @xmath32 . the columns of @xmath33 are the eigenkets of @xmath29 and the rows of @xmath34 are the suitably normalised eigenbras of @xmath29 , i.e. , @xmath35 .
( the subtleties of normalisation are discussed below . )
if @xmath29 were hermitian the eigenbras of @xmath29 would be the hermitian conjugates of the eigenkets and @xmath33 would equal @xmath36 .
this is not true in our case , though the bra - eigenvalues and ket - eigenvalues are the same
. therefore we rewrite the mass term of eq .
( [ eq : meff ] ) as @xmath37 where @xmath38 and @xmath39 , and only @xmath31 is diagonal above . besides the choice mentioned above one can diagonalise @xmath29 by a bi - unitary transformation , i.e , @xmath40 , where @xmath41 and @xmath34 are unitary .
however in this case the kinetic energy term will not be diagonal in the new @xmath42 basis . if one can expand the @xmath43 and @xmath44 fields in terms of annihilation and creation operators ( later we comment on this assumption ) , then @xmath45 , where @xmath46 represents a particle created by the action of the creation operator in @xmath47 on the vacuum and @xmath48 represents a particle created by the creation operator in @xmath49 acting on the vacuum .
note that while @xmath50 , @xmath51 where @xmath52 .. they state that @xmath53 still represents a majorana field and hence we sum only over particle states . in models involving charged scalars
one certainly needs to consider both particle and antiparticle states . in that case
the sum over all states will include particle and antiparticle states .
] instead @xmath54 and @xmath55 where @xmath56 represents a particle created by the action of the destruction operator in @xmath57 acting to the left on the vacuum states has also been discussed elsewhere , as for example , in ref.@xcite and references thereof . ] .
refs.@xcite and @xcite differ in their choice of the initial state relevant for the calculation of the lepton asymmetry .
let @xmath58 be a measure of cp violation due to the self - energy correction . in refs.@xcite
the authors define@xmath59 \label{eq : epspsi}\ ] ] while the definition of the asymmetry parameter in ref.@xcite is @xmath60 \label{eq : epsn}\ ] ] where , as stated earlier , the definitions of the asymmetry parameter in refs.@xcite have been modified for a scenario involving majorana neutrinos .
below we compare the approaches of refs.@xcite and @xcite .
before we proceed let us reiterate here the mechanism of the creation of the lepton asymmetry .
starting with two relativistic neutrino species the second ( heavier ) species becomes non - relativistic at @xmath61 .
after @xmath61 the physical states are the eigenstates of an effective hamiltonian for one relativistic neutrino species with a lorentz suppressed mass and one non - relativistic neutrino species .
they can be represented by @xmath62 and @xmath63 fields , where @xmath63 is a linear combination of the @xmath1 ( but is not @xmath64 ) .
we assume the heavier @xmath63 decay while the @xmath62 neutrinos are in equilibrium . at time
@xmath65 the lighter neutrinos @xmath62 go out of equilibrium and later at @xmath66 they become non - relativistic , i.e. , we assume @xmath67 , where @xmath68 is the temperature at @xmath65 . thus at @xmath66 one will have a universe with only @xmath62 neutrinos with equal numbers of right - handed and left - handed neutrinos , @xmath69 and @xmath70 .
after @xmath66 the effective hamiltonian is as in eq .
( [ eq : meff ] ) and the @xmath71 are the eigenstates of the effective hamiltonian .
the state of the universe with equal numbers of @xmath69 and @xmath70 can now be described as a linear combination of the @xmath71 .
more precisely , a state @xmath72 ( or @xmath73 ) at @xmath66 can be rewritten in terms of @xmath74 . as in the standard neutrino flavour oscillation scenario ,
the right- and left - handed content of an originally @xmath72 or @xmath73 state will now start to oscillate .
one can see from eq .
( [ eq : meff ] ) that the amplitude for @xmath75 is proportional to @xmath76 while the amplitude for @xmath77 is proportional to @xmath78 .
( the amplitudes for @xmath79 are equal and proportional to @xmath80 . ) because of cp violation @xmath81 . therefore ,
though one starts with an equal number of @xmath69 and @xmath70 , more @xmath70 may transform into @xmath82 than @xmath69 into @xmath83 and one ends up with an universe with an unequal content of @xmath69 plus @xmath82 and @xmath70 plus @xmath83 . when the neutrinos subsequently decay one obtains a lepton number asymmetry since right - handed neutrinos decay to leptons and left - handed neutrinos decay to anti - leptons .
( we ignore here asymmetry due to the vertex correction . )
the formalism of ref.@xcite starts with a universe with equal numbers of @xmath69 and @xmath70 at @xmath66 , or , equivalently , with states @xmath72 and @xmath73 with equal probabilities of existing at @xmath66 .
the asymmetry obtained reflects the oscillations in the right - left @xmath44 content that takes place prior to decay . on the other hand ,
the final asymmetry is obtained in ref.@xcite by considering only @xmath84 .
one may argue that typically one does consider the asymmetry generated by only the lightest neutrino .
but what that entails in this scenario is that the heavier @xmath63 neutrino decays while the interactions of the lighter @xmath62 are still in equilibrium , leading to the erasure of any asymmetry generated in the decay of @xmath63 .
subsequently the lightest neutrino @xmath62 goes out of equilibrium , becomes non - relativistic and is re - expressed as a linear combination of both @xmath85 states .
thus one has to consider the lightest @xmath1 species and not the lightest @xmath71 species as the initial state at @xmath66 .
nevertheless , since @xmath86 is an asymmetric mixture of @xmath87 and @xmath88 its decay gives a lepton asymmetry .
ref.@xcite also argues against starting with the @xmath71 states as these are propagation " states which the authors claim are not physical .
however our observations are based on a cosmological argument rather than a field theoretic argument .
( we shall also state the field theoretic argument later . )
the importance of the initial state is also discussed in ref.@xcite .
our arguments above do not apply to scenarios where @xmath89 , i.e. , when the neutrinos are non - relativistic when they go out of equilibrium . in such cases , from a cosmological standpoint , @xmath90 will be the relevant asymmetry parameter . in the limit of large mass splittings ,
i.e. , @xmath91 , the asymmetry obtained in ref.@xcite is similar to that obtained in ref.@xcite .
we find that the approximations needed to obtain this result , i.e. , ignoring terms proportional to @xmath92 , @xmath93 and @xmath94 in eq .
( 17 ) of ref.@xcite and assuming a small mixing angle is equivalent to ignoring the asymmetry from the interference between @xmath95 and @xmath64 and from @xmath64 decay and to setting @xmath96 .
however in the small mass splitting case the asymmetry obtained in refs.@xcite are not similar , as we show later .
our arguments imply that one has a mixed state ( as opposed to a pure state ) of @xmath72 and @xmath73 with equal probabilities at @xmath65 .
thereafter one evolves the @xmath97 states independently . since the mass term is not significant when @xmath62 is relativistic there will be no significant mixing till @xmath66 .
after @xmath66 both @xmath97 can be rewritten as pure states of @xmath86 and @xmath98 and hence one gets interference between @xmath86 and @xmath98 states which is necessary for left - right oscillations .
however if one argues that one instead has a mixed state of @xmath74 at @xmath66 then one would evaluate @xmath90 as in ref.@xcite . in that case the initial number densities of the @xmath71 particles that enter into the solution of the botzmann equation would be weighted by the probabilities of obtaining the @xmath71 at @xmath66 , given by @xmath99 .
one would then include the contributions of both @xmath71 to the total asymmetry since the asymmetry generated by the decay of @xmath64 will not be erased as @xmath95 interactions are out of equilibrium at @xmath66 .
we now raise more fundamental issues with both the approaches using the effective hamiltonian .
the asymmetry parameter is defined in refs.@xcite using decay rates for rotated fields @xmath71 with complex masses .
it is shown in ref.@xcite ( in the context of kaon physics ) that the lack of orthonormality of the @xmath74 states raises ambiguities in the normalisation of states making it difficult to interpret @xmath100 as the probability amplitude for @xmath101 .
if one normalises @xmath74 as @xmath102 then the state is scale dependent .
this is because @xmath103 implies @xmath104 since @xmath105 and @xmath106 . on the other hand ,
if one normalises the @xmath53 states as @xmath107 then @xmath108 \langle f|f\rangle\ne1 $ ] .
this can be better understood by considering the quantum mechanical example of kaon mixing with @xmath74 corresponding to @xmath109 ( or @xmath110 ) and @xmath111 corresponding to @xmath112 and @xmath113 .
a similar argument can be extended to field theoretic scenarios .
if @xmath71 are intermediate particles , i.e. , if one includes a production amplitude for the @xmath71 , the above concerns do not apply @xcite .
if one does include a production amplitude for the @xmath71 in @xmath90 it will cancel out in the numerator and denominator leaving behind the asymmetry parameter as defined in refs.@xcite .
in ref.@xcite the asymmetry parameter is defined in terms of decay rates of the mixed @xmath1 fields .
it has been pointed out in ref.@xcite that in general there does not exist a fock space for mixed fields .
if one assumes that the annihilation operators of the @xmath71 fields and the creation operators of the @xmath114 fields obey canonical anticommutation relations then the annihilation and creation operators of the @xmath1 fields satisfy , for example,@xcite @xmath115 where @xmath116 represent helicity , @xmath117 , the sum is over the mass eigenstates , @xmath118 and @xmath119 include spinor elements as in ref.@xcite and we have used @xmath120 and @xmath121 .
these anticommutation relations are non - diagonal in the @xmath122 indices and are hence not canonical . therefore if @xmath121 , @xmath123 is not a well defined statement in general .
furthermore , the two cases in which one can define an approximate fock space for the @xmath44 fields , namely , the extreme relativistic limit and almost degenerate masses with real @xmath34 @xcite do not apply to our problem . in the limit that all the neutrino species are non - relativistic , as is the case for @xmath124 , the above commutation relations are diagonal in the @xmath125 indices .
however , as pointed out in ref.@xcite , the annihilation and creation operators of the @xmath44 fields are no longer time - independent in this case .
but if one starts with an @xmath1 state as the initial state at @xmath66 and evolves it as a linear combination of @xmath71 states then this may not be a problem since one requires that the relation @xmath123 hold at only one instant , namely , at @xmath66 .
note that in ref.@xcite the states @xmath126 and @xmath127 are in co - ordinate space .
the creation and annihilation operators for the mixed fields in co - ordinate space , namely , @xmath128 and @xmath129 , do satisfy diagonal anticommutation relations .
it was pointed out by veltman that it is inappropriate to have unstable particles as asymptotic in or out states@xcite . to avoid this problem
different effective approaches have been adopted to obtain the one loop decay amplitude . in refs.@xcite
the authors use an one loop effective hamiltonian . in ref.@xcite the decay amplitude
is obtained by splitting a two particle scattering amplitude with the unstable particle as the intermediate particle . in ref.@xcite
one obtains the decay amplitude from an effective lsz formulation .
we discuss here the differences and similarities in the above approaches to obtain the asymmetry . broadly speaking , one can further divide the different approaches into those that use unrotated fields @xmath1 to define the initial state , such as refs.@xcite and those that work with rotated fields to define the initial state , such as refs.@xcite fields may not be possible . ] . though refs.@xcite deal with rotated fields their approaches are somewhat different .
below we first compare the rotating matrices for all these approaches in the small and large mass difference limits .
we also comment on ref.@xcite .
we finally re - evaluate the asymmetry in the small mass difference limit .
in ref.@xcite the authors obtain the eigenvalues and eigenvectors of the effective mass matrix in eq .
( [ eq : meff ] ) . in ref.@xcite
the authors obtain the eigenvalues and eigenkets of the two off - diagonal blocks of @xmath29 . in general , the eigenvalues of the two off - diagonal blocks of the matrix do not equal each other or the eigenvalues of the full matrix and the eigenkets of the off - diagonal blocks describe particles different from those related to the eigenkets of the full matrix . one can show that the eigenvalues of the two off - diagonal blocks are the same as that of the larger matrix , and the eigenvectors of the larger matrix are trivial combinations of corresponding eigenvectors of the smaller blocks when the two off - diagonal blocks are the same . in the limit of a small mass difference @xmath130 , @xmath131 and we find that the diagonalising matrices are equivalent and one gets the same asymmetry .
for example , the matrix relating the right handed field projections @xmath132 and @xmath133 is given by @xmath134 for both approaches .
the above matrix is obtained from @xmath33 of section 1 and we have not shown the normalisation .
however in the limit of a large mass difference @xmath136 the relation between @xmath132 and @xmath133 is given by @xmath137 and @xmath138 for refs.@xcite and @xcite respectively .
one can see that they are not equivalent .
we find that working with the 2x2 matrices in this limit gives an asymmetry that is approximately 1/2 of the asymmetry obtained using the 4x4 approach .
this can be traced to the absence of the contribution of the term proportional to @xmath139 . once again , if @xmath140 are real @xmath141 and the diagonalising matrices of the two approaches become identical . in ref.@xcite
the authors also consider the eigenvalues of the two blocks of a 4x4 effective mass matrix .
however our above arguments do not apply to this case as the two blocks are along the diagonal and are transposes of each other , necessitating that the eigenvalues of the two 2x2 blocks are the eigenvalues of the larger matrix and that the eigenvalues of each block are the same , and the eigenvectors of the larger matrix are a trivial combination of the eigenvectors of the blocks .
ref.@xcite diagonalises the mass matrix and obtains the relationship between @xmath74 and @xmath48 states given by @xmath142 and @xmath143 .
using this to factorise the amplitude for @xmath144 and @xmath145 they obtain the amplitudes for @xmath146 in the large mass difference limit . in the limit of a large mass difference ,
@xmath147 , the matrix relating @xmath132 and @xmath133 fields for ref.@xcite is given by @xmath148 though the matrix above differs from that for ref.@xcite the ultimate asymmetry is the same as that of ref.@xcite because the factor of 2 in the off - diagonal entries above compensates for the presence of the terms proportional to @xmath76 in eq .
( [ eq : vus4wk ] ) . ) is twice that of ref.@xcite .
] we have also obtained the matrix relating @xmath149 and @xmath150 fields in the limit of a small mass difference , @xmath151 , by diagonalising the projections of the resummed propagator as in ref.@xcite .
we obtain the same matrix as refs.@xcite , as given in eq .
( [ eq : vusres ] ) and @xmath152 do not have the same poles making it non - trivial to define the particle mass .
we do not address such field theoretic subtleties .
we only mention that in the limit that @xmath153 @xmath154 and @xmath152 do have the same pole . ] .
in ref.@xcite the author obtains the resummed propagator as in ref.@xcite but does not diagonalise it and obtains the amplitudes for @xmath155 by by truncating the green s function in momentum space with @xmath156^{-1}\hat\sigma_{21}$ ] , where the self - energy corrections to the propagator @xmath157 are defined in eq .
( 4.2 ) of ref.@xcite . in the lsz formalism
the truncating operator defines the corresponding incoming or outgoing particle .
this implies that the incoming neutrino has a mass that is the ( complex ) pole of the component @xmath158 and not of the full resummed propagator @xmath159 , i.e. , one is including only one loop diagonal corrections to define the neutrino mass and excluding off - diagonal corrections .
excluding mixing in the incoming state allows one to use the classical boltzmann equation as discussed later .
the derivation of the lsz formula in momentum space used in ref.@xcite involves integrations over time in intermediate steps , whose range of integration must be modified from @xmath160 to @xmath161 if the neutrino fields have a complex mass .
however the decay amplitude obtained in ref.@xcite can be embedded in an amplitude involving only stable particles in asymptotic states ( analogous to the approach in section 9 of ref.@xcite ) , if one equates @xmath162 $ ] with @xmath163 and again @xmath164 and @xmath165 satisfy only a diagonal dirac equation , @xmath166 , and not @xmath167 , equals @xmath168 . ] .
( ... " is approximately equal to @xmath169 where @xmath170 includes width effects close to @xmath62 production . )
we reiterate here the nature of the incoming particle in various approaches . in ref.@xcite
the incoming particle is an unmixed particle corresponding to the original lagrangian of the theory , i.e. , @xmath171 of eq .
( [ eq : lag ] ) . in ref.@xcite
the incoming particle is a particle corresponding to the mixed field of the one loop effective theory , i.e , @xmath171 of eq .
( [ eq : meff ] ) . in ref.@xcite
the incoming particle is a particle corresponding to the rotated field of the one loop effective theory , i.e. , @xmath95 of eq .
( [ eq : md ] ) or its equivalent . in ref.@xcite
the incoming particle corresponds to the rotated field obtained by diagonalising projections of the resummed propagator .
the incoming particle has a one loop effective mass . in ref.@xcite
the incoming particle is unrotated but has a one loop ( complex ) effective mass corresponding to the pole of the @xmath158 component of the resummed propagator . in the limit that mixing is small compared to the mass difference , i.e , @xmath172 the amplitudes for decay in refs.@xcite are similar though the spinor @xmath173 in the amplitude satisfies @xmath174 , or @xmath175 , with a different @xmath176 for each approach .
as implied earlier the amplitude for ref.@xcite differs in form from the others but refs.@xcite give approximately the same asymmetry in this limit .
however this does not affect the asymmetry which involves a ratio of decay rates . ] . in the opposite limit that @xmath177 the rotating matrices of refs.@xcite
are equivalent and give the same asymmetry .
it was first shown in ref.@xcite that there is an enhancement in the asymmetry when @xmath178 and @xmath179 .
we have reevaluated the asymmetry in this limit and obtained @xmath180 + 4(g_{\alpha ij}^{abs})^2 re^2(h_{\epsilon 1}^ * h_{\epsilon 2 } )
|h_{\delta1}+h_{\delta2}|^2 } \label{eq : e1}\ ] ] where @xmath181 and the sum over @xmath182 and @xmath58 is implied in ref.@xcite in this limit . ] .
the expression for the asymmetry obtained in ref.@xcite in the limit @xmath183 is @xmath184 where @xmath185 and @xmath186 . in obtaining the asymmetry above , @xmath187 in the invariant amplitude
was substituted by @xmath188 . to be consistent with the truncation of the greens s function @xmath187
should be replaced by the pole of @xmath158 .
it is not trivial to solve for the pole . using a naive substitution of @xmath187 by @xmath189^{-1}\hat\sigma_{21}$ ] , where @xmath190 are evaluated at @xmath191 and ignoring terms of order @xmath192 at the amplitude level , as in ref.@xcite , one obtains @xmath193 where @xmath194 $ ] . in the case of hierarchical yukawa couplings
the above result reduces to the asymmetry in eq .
( [ eq : e2 ] ) .
in the limit that @xmath195 , or @xmath196 , and @xmath197 , which is comparable to the limit in which the enhancement was seen in ref.@xcite , @xmath198 . then the asymmetry reduces to @xmath199 one can see that the asymmetry in eq .
( [ eq : e1 ] ) is different from that in eq .
( [ eq : e4 ] ) .
this is not surprising as eq .
( [ eq : e1 ] ) represents the asymmetry in the decay of a @xmath53 particle while eq .
( [ eq : e4 ] ) reflects the asymmetry in the decay of a particle represented by an unrotated field with a complex mass . in the limit of large mixing these need not be equivalent .
we have also obtained the asymmetry in the small mass difference limit for the model involving sneutrino decays in ref.@xcite . in the limit
that @xmath183 and @xmath200 ( in the notation of ref.@xcite ) , we find @xmath201 } \ , , \label{eq : e5}\ ] ] where @xmath202 and @xmath203 are defined as above . while comparing this result with the asymmetry obtained in other works one must keep in mind that eq .
( [ eq : e5 ] ) is the asymmetry obtained from sneutrino decays and that the model in ref.@xcite has twice as many decay channels for the sneutrinos as compared to the neutrinos in other works .
we would also like to add a note of caution when one uses the effective hamiltonian approach and the boltzmann equation to obtain the lepton asymmetry .
the boltzmann equation which is used to calculate the number densities of various species assumes that the nature of the species does not change over time .
the l.h.s . of the boltzmann equation is the time derivative of the number density of a species and one has to be careful as to exactly what one means by @xmath204 or @xmath205 at times when @xmath1 or @xmath71 may not be the physical states .
this comment applies to scenarios where one starts with @xmath44 states at @xmath66 which evolve subsequently as @xmath53 states , as well as to ref.@xcite where one obtains the asymmetry by evolving the @xmath53 number density from when the neutrinos are relativistic .
it has been argued in ref.@xcite that decoherence effects due to interactions with the thermal universe justify excluding mixing effects in the incoming neutrino state .
excluding mixing in the incoming state allows one to work with a classical ( diagonal ) boltzmann equation .
in conclusion , we have argued that in leptogenesis scenarios where the heavy neutrinos or scalars are relativistic when they go out of equilibrium one should calculate the asymmetry in the decay of the lightest unrotated species .
we have pointed out certain issues related to the definition of the decay amplitude for eigenkets of the non - hermitian effective hamiltonian and the existence of a fock space for mixed fields . as different authors have used various approaches to rotate the mixed fields we have compared the rotating matrices of refs.@xcite and discussed their similarities and differences .
finally we have re - evaluated the asymmetry in the small mass difference limit using different approaches . e. w. kolb and m. s. turner , ann .
nucl . part .
sci . * 33 * ( 1983 ) 645 ; a. d. dolgov , phys .
* 222 * ( 1992 ) 309 ; v. a. rubakov and m. e. shaposhnikov , usp .
nauk * 166 * ( 1996 ) 493 , phys .
* 39 * ( 1996 ) 461 ; a. riotto , hep - ph/9807454 .
m. fukugita and t. yanagida , phys
* b 174 * ( 1986 ) 45 ; p. langacker , r.d .
peccei and t. yanagida , mod .
. lett . * a 1 * ( 1986 ) 541 ; m.a .
luty , phys . rev . *
d 45 * ( 1992 ) 455 ; c. vayonakis , phys .
b286 * ( 1992 ) 92 ; r.n .
mohapatra and x. zhang , phys . rev . *
d 46 * ( 1992 ) 5331 one loop self - energy corrections were included in baryogenesis models in a. yu .
ignatev , v. a. kuzmin and m. e. shaposhnikov , jetp lett . * 30 * ( 1979 ) 688 and in f. j. botella and j. roldan , phys . rev . *
d 44 * ( 1991 ) 966 .
asymmetry due to mixing is also mentioned in section 2.4.4 of e. w. kolb and s. wolfram , nucl .
* b 172 * ( 1980 ) 224 .
self - energy corrections were first included in leptogenesis models in ref.@xcite .
|
the mass term for majorana neutrinos explicitly violates lepton number .
several authors have used this fact to create a lepton asymmetry in the universe by considering cp violating effects in the one loop self - energy correction for the decaying heavy majorana neutrino .
we compare and comment on the different approaches used to calculate the lepton asymmetry including those using an effective hamiltonian and resummed propagators .
we also recalculate the asymmetry in the small mass difference limit . plus 1pt
@=12 -0.5 in 0.0 in 0.0 in 8.5 in 6.5 in [ 1 ] # 10= 0=0 1= 1=1 0>1 # 1 / 12h_12 12 t august 1999 + hep - ph/9908417 * leptogenesis with heavy majorana neutrinos reexamined + * * e - mail : [email protected] @xmath0 e - mail : [email protected] after sakharov s seminal paper on the conditions for obtaining a matter asymmetric universe several authors have attempted to realise these conditions in the context of different models@xcite .
one approach is to create a lepton asymmetric universe by the asymmetric decay of heavy scalars or majorana neutrinos and to convert the lepton asymmetry into a baryon asymmetry by b+l violating sphaleron processes . in the original papers on leptogenesis cp violation
was realised by the interference between the tree level diagram and the one loop vertex correction@xcite .
later it was realised that there was an additional contribution from the one loop self - energy correction to the external heavy scalar or neutrino@xcite .
the self - energy contribution has been calculated differently by different authors .
refs.@xcite obtain the asymmetry by considering the interference of the tree level diagram for decay and the diagram containing the one loop self - energy correction .
however this approach breaks down in the degenerate heavy particle mass limit . in refs.@xcite
the authors obtain the one loop effective hamiltonian , or more precisely , the one loop effective mass matrix for the heavy species . because particles in the self - energy loop go on shell
the effective mass matrix is non - hermitian .
this , coupled with cp violation , results in different evolution of the scalars and their antiparticles , or of the right - handed and left - handed neutrinos , and ultimately gives rise to an asymmetry when the heavy particles decay . in refs.@xcite
the authors find an enhancement in the degenerate scalar or neutrino mass limit@xcite .
the authors of refs.@xcite and @xcite differ in their use of the effective hamiltonian . in this paper
we compare and comment on both approaches and on the approach in refs.@xcite where the authors use resummed propagators to evaluate cross sections and decay rates
. we shall discuss below models involving decays of heavy majorana neutrinos .
however our comments also apply to models with heavy scalars . in refs.@xcite
the authors consider models with charged scalar fields ; to facilitate discussion we have modified their definitions to suit the majorana fields @xmath1 . in section 1
we discuss some subtleties associated with a non - hermitian hamiltonian , particularly with respect to the normalisation of eigenstates of the hamiltonian . in section 2
we compare the approaches of refs.@xcite and @xcite both of whom use the effective hamiltonian . in section 3
we discuss some field theoretic concerns regarding the definition of the decay amplitude for eigenkets of a non - hermitian hamiltonian and the existence of a fock space for mixed fields . in section 4
we compare the matrices used to diagonalise the effective hamiltonian or projections of the resummed propagator in refs.@xcite .
we also comment on the approach of ref.@xcite .
we re - evaluate the asymmetry in the small mass difference limit . in section 5
we summarize our results .
|
empirical likelihood ( el ) is a powerful and currently widely used approach for developing efficient nonparametric statistics .
the el was introduced by owen ( 1988 ) and since then many papers have appeared on this topic making varied contributions to different inferential problems . in this approach , the parameters are usually defined as functionals of the unknown population distribution . the first purpose of this paper is to introduce a new family of empirical test statistics as an alternative to the likelihood ratio test statistic proposed by qin and lawless ( 1994 ) for testing a simple null hypothesis . as an extension of the empirical likelihood ratio test of qin and lawless ( 1995 ) , a new family of empirical test statistics
is also considered here for composite hypothsis .
this new family of empirical test statistics is based on divergence measures .
consider @xmath0-variate i.i.d .
random vectors @xmath1 with unknown distribution function @xmath2 , and a @xmath3-dimensional parameter , @xmath4 , associated with @xmath2 having finite mean and non - singular covariance matrix .
we assume that all the information about @xmath5 and @xmath2 is available in the form of @xmath6 functionally independent unbiased estimating functions , through the functions @xmath7 , @xmath8 , such that @xmath9 = 0 $ ] . in vector notation
, we have @xmath10 , such that@xmath11 = \boldsymbol{0}_{r}. \label{estf}\ ] ] we shall assume that for each realization @xmath12 of @xmath13 , @xmath14 @xmath15 is a vector - valued function and the @xmath16 matrix @xmath17 exists .
this formulation is as in qin and lawless ( 1994 ) , but a little different from that of owen ( 1988 , 1990 ) .
the essential difference is that owen considered @xmath18 instead of @xmath6 .
for example , by taking into account that @xmath19 = \theta$ ] , we can adopt steps similar to the method of moments to obtain estimators through the estimating function @xmath20 of a univariate distribution .
in addition , if we assume @xmath21 = 2\theta^{2}+1 $ ] , the other estimating function is @xmath22 , and in this case we have @xmath23 .
let @xmath24 be a realization of @xmath25 .
the empirical likelihood function is then given by @xmath26 where @xmath27 .
only distributions with an atom of probability at each @xmath28 have non - zero likelihood , and without consideration of estimating functions , the empirical likelihood function @xmath29 is seen to be maximized , at @xmath30 , by the empirical distribution function@xmath31 which is associated with the @xmath32-dimensional discrete uniform distribution@xmath33 now , let@xmath34 be an empirical distribution function associated with the probability vector@xmath35 and @xmath36 be the kernel of the empirical log - likelihood function , @xmath37 . if we are interested in maximizing @xmath38 subject to the restrictions defined by the estimating functions based on @xmath39 given by @xmath40 we obtain , by applying the lagrange multipliers method , @xmath41 where @xmath42 is an @xmath43-dimensional vector to be determined by solving the non - linear system of @xmath43 equations,@xmath44 subject to ( [ cond ] ) and ( [ empf2 ] ) . thus , the kernel of the empirical log - likelihood function is@xmath45 .
\label{elf}\ ] ] one of the important results of qin and lawless ( 1994 ) is that the empirical likelihood ratio test statistic for testing@xmath46 is given by @xmath47 where @xmath48 is the empirical maximum likelihood estimator of the parameter @xmath5 obtained by maximizing @xmath36 in ( [ elf ] ) . in particular ,
if @xmath18 , it can be seen that @xmath49 , @xmath50 and @xmath48 is the solution of the system of @xmath43 equations @xmath51 , where@xmath52 subject to ( [ cond ] ) and ( [ empf2 ] ) .
furthermore , if @xmath18 , @xmath53 and @xmath54 . the asymptotic properties of @xmath48 , when @xmath6 , were studied by qin and lawless ( 1994 ) .
in fact , the asymptotic distribution of @xmath55 in ( [ 1.1 ] ) is chi - square with @xmath3 degrees of freedom . here , we propose a new family of test statistics for testing the hypotheses in ( [ h ] ) , based on @xmath56-divergence measures , and then derive their asymptotic distribution .
this new family of empirical test statistics is referred to hereafter as empirical @xmath56-divergence test statistics . in section [ sec2 ] , the asymptotic null distribution of the empirical @xmath56-divergence test statistics is derived .
then , two power approximations of the empirical @xmath56-divergence test statistics are presented in section [ sec3 ] .
an illustrative example is presented in section [ example ] . in section [ simulation ] , a monte carlo simulation study is carried out to compare its performance with respect to the empirical likelihood ratio test when confidence intervals are constructed based on the respective statistics for small simple sizes .
the results show that the empirical @xmath56-divergence test statistic is competitive in terms of power when compared to the empirical likelihood ratio test statistic , and moreover is more robust than the empirical likelihood ratio test statistic in the presence of contamination in the data .
in section [ seccomp ] , we propose empirical phi - divergence test statistics for testing a composite null hypothesis and present some asymptotic as well as simulation results to evaluate the performance of these test procedures .
finally , in section [ lastsec ] , we make some concluding remarks .
the kullback - leibler divergence measure between the probability vectors , @xmath57 and @xmath58 , is given by @xmath59 in the sequel , we shall denote it by @xmath60 , since the above expression is the distance , in the sense of the kullback - leibler divergence , between the distributions functions @xmath61 and @xmath62 . using this notation ,
it is clear that the empirical likelihood ratio test statistic in ( [ 1.1 ] ) can be expressed as @xmath63 or equivalently @xmath64 where @xmath65 with @xmath66 , @xmath67 .
now , we shall denote by @xmath68 the class of all convex functions @xmath69 such that at @xmath70 , @xmath71 , @xmath72 , and at @xmath73 , @xmath74 and @xmath75 . instead of @xmath66 ,
if we consider a function @xmath56 belonging to @xmath68 , we obtain a new family of test statistics for testing ( [ h ] ) given by @xmath76 we will refer to this family of test statistics as empirical @xmath56-divergence test statistics . for every @xmath77 that is differentiable at @xmath70
, the function @xmath78 also belongs to @xmath68 .
then , we have @xmath79 and @xmath80 has the additional property that @xmath81 since the two divergence measures are equivalent , we can consider the set @xmath68 to be equivalent to the set @xmath82 . in
what follows , we shall assume that @xmath83 .
the statistics in ( [ f1 ] ) have been considered recently by broniatowski and keziou ( 2012 ) to give some empirical test statistics , generalizing an important subfamily of test statistics introduced by baggerly ( 1998 ) .
we present more details in section [ simulation ] in this regard .
the main purpose of this paper is to present a new family of test statistics for testing the hypotheses in ( [ h ] ) based on the @xmath56-divergence measure between @xmath84 and @xmath85 , namely , @xmath86 .
we shall consider the empirical family of @xmath56-divergence test statistics given by @xmath87 where @xmath56 is a function satisfying the same conditions as function @xmath56 used to construct @xmath88 .
observe that ( [ f1 ] ) and ( [ f2 ] ) are equivalent only when @xmath18 .
it is well - known that the family of test statistics based on @xmath56-divergence has some nice and optimal properties for different inferential problems , and especially in relation to robustness ; see pardo ( 2006 ) and basu et al .
in this section , we derive the asymptotic distribution of @xmath89 . for this purpose ,
the asymptotic distribution of the maximum empirical likelihood estimator of the parameter @xmath5 , @xmath90 as well as the asymptotic distribution of @xmath91 are important@xmath92 these asymptotic distributions are given in qin and lawless ( 1994 ) , for example .
under some regularity assumptions ( see lemma 1 and theorem 1 of qin and lawless ( 1994 ) ) , we have@xmath93 where @xmath94 denotes convergence in law and@xmath95
with@xmath96 , \label{s11}\\ \boldsymbol{s}_{12}\left ( \boldsymbol{\theta}_{0}\right ) & = e_{f}\left [ \boldsymbol{g}_{\boldsymbol{x}}(\boldsymbol{\theta}_{0})\right ] , \quad\boldsymbol{s}_{21}\left ( \boldsymbol{\theta}_{0}\right ) = \boldsymbol{s}_{12}\left ( \boldsymbol{\theta}_{0}\right ) ^{t}. \label{s12}\ ] ] this result is derived from : a ) @xmath97 where @xmath98 is given by ( [ g_bar ] ) .
it is clear from the central limit theorem that @xmath99 and so @xmath100;b ) @xmath101 , where @xmath102 is given by ( [ v2 ] ) and so @xmath103.in addition , @xmath48 and @xmath104 are asymptotically uncorrelated . [ lem1]the influence function of the empirical maximum likelihood estimator of parameter @xmath5 , @xmath48 , is given by@xmath105 it follows from the expression given in ( [ thetahat ] ) and taking into account the definition of the influence function given in formula ( 20.1 ) of van der vaart ( 2000 , page 292 ) .
the empirical maximum likelihood estimator of parameter @xmath5 , @xmath48 , is obtained maximizing the kernel of empirical log - likelihood function given in ( [ elf ] ) or equivalently minimizing the function @xmath106 , subject to the restrictions given in ( [ res ] ) .
this expression can be written as the @xmath56-divergence measure between the probability vectors @xmath57 and @xmath107 , i.e. @xmath108 with @xmath109 .
therefore,@xmath110 subject to the restrictions given in ( [ res ] ) .
if we consider a general function @xmath83 , defined in section [ sec2a ] , instead of considering @xmath66 , then we can define the empirical minimum @xmath56-divergence estimator by @xmath111 subject to the restrictions given in ( [ res ] ) .
the empirical exponential tilting estimator ( et ) , considered for instance in schennach ( 2007 ) , is defined by @xmath112 , subject to the restrictions given in ( [ res ] ) .
the et is another member of this family of estimators since @xmath113 with @xmath114 .
the asymptotic properties of @xmath115 and @xmath48 are the same ( for more details , see ragusa ( 2011 ) , broniatowski and keziou ( 2012 ) and schennach ( 2007 ) ) .
the fisher consistence of @xmath115 was established in ragusa ( 2011 )
. therefore , all the asymptotic results obtained for the test statistics considered in this paper are valid replacing @xmath48 by @xmath115 .
the expression given in ( [ if ] ) , for the influence function of the empirical maximum likelihood estimator of parameter @xmath5 , can be found for the empirical minimum @xmath56-divergence estimators , @xmath116 , in proposition 2.3 of toma ( 2013 ) .
hence , all the estimators based on @xmath56-divergence measures , independently of the @xmath56 function , share the same influence function .
let @xmath117 denote any vector or matrix norm .
we shall assume the following regularity conditions : 1 .
@xmath118 in ( [ s11 ] ) is positive definite , and for @xmath119 in ( [ s12 ] ) , @xmath120 .
2 . there exists a neighbourhood of @xmath121 , in which @xmath122 is bounded by some integrable function .
there exists a neighbourhood of @xmath121 , in which @xmath123 , given in ( [ g ] ) , is continuous and @xmath124 is bounded by some integrable function .
there exists a neighbourhood of @xmath121 in which @xmath125 is continuous and @xmath126 is bounded by some integrable function . the asymptotic distribution of the empirical @xmath56-divergence test statistics , @xmath127 , is given in the following theorem .
[ th1]under @xmath128 in ( [ h ] ) and the assumptions i)-iv ) above , we have@xmath129 based on the taylor expansions of ( [ f2 ] ) , as well as the asymptotic properties of @xmath48 , given in qin and lawless ( 1994 ) , it holds@xmath130 with @xmath131 , according to ( [ dist ] ) . there are some measures of divergence which can not be expressed as a @xmath132-divergence measure such as the divergence measures of bhattacharya ( 1943 ) , rnyi ( 1961 ) , and sharma and mittal ( 1977 ) .
however , such measures can be written in the form@xmath133 where @xmath134 is a differentiable increasing function mapping from @xmath135 onto @xmath136 , with @xmath137 , @xmath138 , and @xmath83 . in table
[ t1 ] , these divergence measures are presented , along with the corresponding expressions of @xmath134 and @xmath56 . 0.8pt @xmath139{ccccc}\hline divergence & \hspace*{0.5 cm } & $ ] h ( x ) @xmath140 ( x ) @xmath141^{\frac{b-1}{a-1}}-1\right\ } , \quad b , a\neq1 $ } & & \multicolumn{1}{l}{$\frac{x^{a}-a\left ( x-1\right ) -1}{a\left ( a-1\right ) } , \quad a\neq0,1$}\\ \multicolumn{1}{l}{battacharya } & & \multicolumn{1}{l}{$-\log\left ( -x+1\right ) $ } & & \multicolumn{1}{l}{$-x^{1/2}+\frac{1}{2}\left ( x+1\right ) $ } \\\hline \end{tabular } \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ $ ] in the case of rnyi s divergence , we have @xmath142 and its limit cases corresponding to @xmath143 and @xmath144 as @xmath145 and @xmath146 respectively .
the @xmath147-divergence measures were introduced in menndez et al .
( 1995 ) and some associated asymptotic results were established in menndez et al .
( 1997 ) .
[ th3]under the assumptions of theorem [ th1 ] , the asymptotic null distribution of the family of empirical test statistics @xmath148 is chi - squared with @xmath3 degrees of freedom . [ remark]note that we can also consider the family of empirical @xmath147-divergence test statistics given by@xmath149 and their asymptotic distribution is chi - squared with @xmath3 degrees of freedom as well .
test - statistics @xmath150 and @xmath151 , include as particular cases @xmath88 and @xmath127 , taking @xmath152 . in the same way done for @xmath127 in ( [ taylors ] ) , it may be concluded that all of them have a taylor expansion of second order which does not depend on @xmath56 or @xmath134,@xmath153 with @xmath154 . with regard to their influence function , a second order influence function
have to be considered since the first order one vanishes ( see van der vaart ( 2000 , section 20.1.1 ) ) , and its expression is given by@xmath155 the first and second equality come from the previous second order taylor expansion , the third one from heritier and ronchetti ( 1994 ) and the last one from lemma [ lem1 ] .
it suggests that under the simple null hypothesis , all members of both family of test - statistics have the same infinitesimal robustness .
this is not a strange result , since in lemma 1 of toma ( 2009 ) a similar conclusion has been reached based on power divergence measures for parametric models .
based on the null distribution presented in theorem [ th1 ] , we reject the null hypothesis in ( [ h ] ) in favour of the alternative hypothesis , if @xmath156 , where @xmath157 is the @xmath158-th quantile of the chi - squared distribution with @xmath3 degrees of freedom . in most cases , the power function of this test procedure can not be derived explicitly . in the following theorem , we present an asymptotic result , which provides an approximation of the power of the test . [
th5]let @xmath159 be the true parameter value with @xmath160 . under the assumptions of theorem [ th1 ] and @xmath161 < \infty$ ]
, we have@xmath162 where @xmath163 with @xmath164 and the matrix @xmath165 as defined in ( [ v ] ) . a first - order taylor expansion gives@xmath166 but , @xmath167 and so @xmath168
thus , the random variables @xmath169 have the same asymptotic distribution , and hence the desired result .
[ rempostth5]it is straightforward to extend theorem [ th5 ] for @xmath147-divergence measures @xmath170@xmath171 , as follows . since @xmath172
, we have @xmath173 and so@xmath174 where@xmath175 @xmath176 and @xmath177 as defined in theorem [ th5 ] . [ beta1]from theorem [ th5 ]
, we can present a first approximation of the power function , at @xmath160 , of the test based on @xmath56-divergence measure , as@xmath178 with @xmath179 being the standard normal distribution function.if some alternative @xmath160 is the true parameter , then the probability of rejecting @xmath121 with the rejection rule @xmath156 , for fixed significance level @xmath180 , tends to one as @xmath181 .
thus , the test is consistent in the sense of fraser ( 1957 ) .
similarly , from theorem [ th5 ] , we obtain a first approximation of the power function , at @xmath182 , of the test based on @xmath183-divergence measure as @xmath184 to produce some less trivial asymptotic powers that are not all equal to @xmath185 , we can use a pitman - type local analysis , as developed by le cam ( 1960 ) , by confining attention to @xmath186-neighborhoods of the true parameter values . a key tool to get the asymptotic distribution of the statistic @xmath89 under such a contiguous hypothesis is le cam s third lemma , as presented in hjek and sidk ( 1967 ) .
instead of relying on these results , we present in the following theorem a proof which is easy and direct to follow . this proof is based on the results of morales and pardo ( 2001 ) . specifically , we consider the power at contiguous alternatives of the form @xmath187 where @xmath188 is a fixed vector in @xmath189 such that @xmath190 .
[ th1b]under the assumptions of theorem [ th1 ] and @xmath191 in ( [ cont ] ) , we have the asymptotic distribution of the empirical @xmath132-divergence test statistic @xmath127 to be non - central chi - squared with @xmath3 degrees of freedom and non - centrality parameter@xmath192 where @xmath193 we can write@xmath194 under @xmath191 , we have @xmath195 and @xmath196 in theorem [ th1 ] , it has been shown that @xmath197 on the other hand , we have@xmath198 we thus obtain@xmath199 with @xmath200 as in ( [ ncp ] ) .
notice that @xmath201 and @xmath202 have the same asymptotic distribution under the contiguous alternative hypothesis in ( [ cont ] ) .
hence , it is straightforward to extend theorem [ th1b ] for @xmath150 and to conclude that both test statistics , @xmath150 and @xmath127 , have the same asymptotic distribution under the contiguous alternative hypothesis .
[ beta2]using theorem [ th1b ] and the preceding remark , we obtain a second approximation to the power function , at @xmath203 , as @xmath204 where @xmath205 is the distribution function of @xmath206.if we want to approximate the power at the alternative @xmath207 , then we can take @xmath208 .
it should be noted that this approximation is independent of the function @xmath56 in @xmath209 , as well as of the @xmath134 function .
in this section , we make use of a well - known data to illustrate the application of the empirical phi - divergence test statistics for testing a simple null hypothesis . in 1882 ,
the astronomer and mathematician simon newcomb , measured the time required for a light signal to pass from his laboratory on the potomac river to a mirror at the base of the washington monument and back , a distance of @xmath210 meters .
table [ tnewcomb ] contains these measurements from three samples , as deviations from @xmath211 nanoseconds .
for example , for the first observation , @xmath212 represents the @xmath213 nanoseconds that were spent for the light to travel the required @xmath210 meters .
the data contain three samples , of sizes @xmath214 , @xmath214 and @xmath215 , respectively , corresponding to three different days .
they have been analyzed previously by a number of authors including stigler ( 1973 ) and voinov et al .
( 2013 ) .
our interest here is in obtaining confidence intervals for the mean , @xmath216 , of the passage time of light , @xmath217 .
this means that a unique estimating function , @xmath218 , is required for a univariate random variable ( @xmath219 ) .
note that @xmath220 , and so @xmath221 .
based on the results of theorem [ th3 ] , we shall use the empirical rnyi test statistics@xmath222{ll}\frac{2n}{a\left ( a-1\right ) } \log{\textstyle\sum\limits_{i=1}^{n } } \frac{1}{n}\left ( 1+t(\mu_{0})(x_{i}-\mu_{0})\right ) ^{a-1 } , & a\neq0,1,\\ -2{\textstyle\sum\limits_{i=1}^{n } } \dfrac{\log\left ( 1+t(\mu_{0})(x_{i}-\mu_{0})\right ) } { 1+t(\mu_{0})(x_{i}-\mu_{0 } ) } , & a=0,\\ 2{\textstyle\sum\limits_{i=1}^{n } } \log\left ( 1+t(\mu_{0})(x_{i}-\mu_{0})\right ) & a=1 , \end{array } \right . \label{renyiex}\ ] ] where @xmath223 is the solution of the equation @xmath224 subject to @xmath225 and @xmath226 .
2.8pt @xmath139{ccccccccccccccccccccccccccc}\cline{1 - 21}day 1 & $ ] 28@xmath22726@xmath22733@xmath22724@xmath22734@xmath227 - 44@xmath22727@xmath22716@xmath22740@xmath227 - 2@xmath22829@xmath22722@xmath22724@xmath22721@xmath22725@xmath22730@xmath22723@xmath22729@xmath22731@xmath22719@xmath22924@xmath22720@xmath22736@xmath22732@xmath22736@xmath22728@xmath22725@xmath22721@xmath22728@xmath22729@xmath22837@xmath22725@xmath22728@xmath22726@xmath22730@xmath22732@xmath22736@xmath22726@xmath22730@xmath22722@xmath23036@xmath22723@xmath22727@xmath22727@xmath22728@xmath22727@xmath22731@xmath22727@xmath22726@xmath22733@xmath22826@xmath22732@xmath22732@xmath22724@xmath22739@xmath22728@xmath22724@xmath22725@xmath22732@xmath22725@xmath22729@xmath22827@xmath22728@xmath22729@xmath22716@xmath22723@xmath231 a confidence interval , with confidence level @xmath232 , based on ( [ renyiex ] ) is given by@xmath233 with @xmath234 being the ( right hand ) quantile of order @xmath180 for the @xmath235 distribution . by taking different choices for the parameter @xmath236 as @xmath237
, we obtain different confidence intervals , and this is done for each of the three days .
table [ tci ] summarizes these results obtained at @xmath238 confidence level .
if the confidence interval based on the likelihood ratio test statistic , i.e. , @xmath239 are compared with @xmath240 , we observe that narrower confidence intervals are obtained for @xmath241 , and wider confidence intervals are obtained for @xmath242 .
this behavior is more evident for the first day , due to the presence of some outliers in the sample ; see for example , bernett and lewis ( 1993 ) .
next , we approximate the power function , assuming that @xmath243 is a value very close to @xmath244 . despite the fact that the exact significance may not be approximated well
( since the true value @xmath243 is unknown ) , the power values can provide meaningful information on the quality of performance of the different rnyi test statistics .
2.8pt @xmath139{clcccccc}\hline & $ ] ci_0.95 ^ -1()@xmath227ci_0.95 ^ -0.5()@xmath227ci_0.95 ^ 0()@xmath228ci_0.95 ^ 0.5()@xmath227ci_0.95 ^ 1()@xmath227ci_0.95 ^ 1.5()@xmath228ci_0.95 ^ 2.5()@xmath245(13.65,27.25)@xmath228(10.85,27.11)@xmath227(12.00,27.24)@xmath227(12.92,27.24)@xmath227(9.46,26.97)@xmath228(6.90,26.60)@xmath246(26.47,30.70)@xmath228(26.40,30.78)@xmath227(26.43,30.74)@xmath227(26.45,30.71)@xmath227(26.35,30.83)@xmath228(26.24,30.97)@xmath247(26.27,29.48)@xmath228(26.00,29.70)@xmath227(26.10,29.61)@xmath227(26.20,29.54)@xmath227(25.88,29.79)@xmath228(25.62,30.01)@xmath248 in order to calculate the power function by using the first approximation@xmath249 we need to obtain @xmath250 in terms of@xmath251{ll}\frac{1}{a\left ( a-1\right ) d_{\phi_{a}}\left ( f_{n,\mu^{\ast}},f_{n,\mu_{0}}\right ) + 1 } , & a\left ( a-1\right ) \neq0,\\ 1 , & a\left ( a-1\right ) = 0 , \end{array } \right . \\ & = \left ( \sum\limits_{i=1}^{n}p_{i}^{a}(\mu^{\ast})p_{i}^{1-a}(\mu _ { 0})\right ) ^{-1},\end{aligned}\]]@xmath252@xmath253{ll}\dfrac{n}{1-a}{\displaystyle\sum\limits_{i=1}^{n } } \left [ t^{\prime}(\mu^{\ast})(x_{i}-\mu^{\ast})-t(\mu^{\ast})\right ] p_{i}^{a+1}(\mu^{\ast})p_{i}^{1-a}(\mu_{0 } ) , & a\neq1,\\ n{\displaystyle\sum\limits_{i=1}^{n } } \left [ t^{\prime}(\mu^{\ast})(x_{i}-\mu^{\ast})-t(\mu^{\ast})\right ] p_{i}^{2}(\mu^{\ast})\left ( \log\frac{p_{i}(\mu_{0})}{p_{i}(\mu^{\ast})}-1\right ) , & a=1 , \end{array } \right.\]]@xmath254 and@xmath255 .\ ] ] since @xmath256 is unknown , it can be replaced by a consistent estimator@xmath257 so that@xmath258 where@xmath259@xmath260{ll}-\frac{1}{a\left ( a-1\right ) } \log\kappa_{\text{\textrm{r\'{e}nyi}}}^{a}\left ( \mu_{0},\mu^{\ast}\right ) , & a\left ( a-1\right ) \neq0,\\ \sum_{i=1}^{n}p_{i}(\mu_{0})\log\frac{p_{i}(\mu_{0})}{p_{i}(\mu^{\ast } ) } , & a=0,\\ \sum_{i=1}^{n}p_{i}(\mu^{\ast})\log\frac{p_{i}(\mu^{\ast})}{p_{i}(\mu_{0 } ) } , & a=1 . \end{array } \right.\ ] ] in figure [ fig000 ] , the power functions are plotted for the three days , for three different values of the parameter @xmath261 .
we do observe from these plots that the most powerful test statistics are for the case when @xmath242 , which is in fact the case when the confidence intervals are wider .
notice that there are two outliers on day 1 ( @xmath262 and @xmath263 ) and one possible outlier on day 3 ( @xmath264 ) ; see stigler ( 1973 ) and barnett and lewis ( 1994 ) . by dropping these outliers , we reproduced the plots of approximated power functions for days 1 and 3 , and we found that in the new plots the approximated power functions are more similar to that of day 2 . in the presence of outliers ( days 1 and 3 )
, the comparison of shape of the power function for different values provides an interesting information : the power function is flatter for @xmath265 , with @xmath266 in comparison with @xmath144 , @xmath267 .
this tends to be associated with higher capacity of the empirical likelihood ratio test ( @xmath268 , with @xmath144 ) for detecting samples not fulfilling the null hypothesis , but a lack of robustness in the presence of outliers . in section [ simulation2 ]
, a simulation study is carried out to examine this robustness aspect .
@xmath139{cc}\raisebox{-0cm}{\includegraphics [ natheight=2.491500 in , natwidth=3.746400 in , height=4.815 cm , width=7.2071 cm ] { fig3.pdf } } & \raisebox{1.0032in}{\fbox{\includegraphics [ natheight=0.677200 in , natwidth=1.152800 in , height=0.7083 in , width=1.1865 in ] { fig1.pdf } } } \\{\includegraphics [ natheight=2.491500 in , natwidth=3.746400 in , height=1.8957 in , width=2.8374 in ] { fig2.pdf } } & { \includegraphics [ natheight=2.491500 in , natwidth=3.746400 in , height=1.8957 in , width=2.8374 in ] { fig6.pdf } } \\{\includegraphics [ natheight=2.491500 in , natwidth=3.746400 in , height=1.8957 in , width=2.8374 in ] { fig4.pdf } } & { \includegraphics [ natheight=2.491500 in , natwidth=3.746400 in , height=1.8957 in , width=2.8374 in ] { fig7.pdf } } \end{tabular } $ ]
in this section , we will pay special attention to a subfamily of @xmath132-divergence measures in ( [ eqdiv ] ) , the so - called power divergence measures ( see cressie and read ( 1984 ) ) for which @xmath269 is given by@xmath270 , \quad\text{if } \lambda(1+\lambda)\neq 0,\label{pd0}\\ \phi_{0}(x ) & = \lim_{\lambda\rightarrow0}\phi_{\lambda}(x)=x\log x - x+1,\nonumber\\ \phi_{-1}(x ) & = \lim_{\lambda\rightarrow-1}\phi_{\lambda}(x)=-\log x+x-1.\nonumber\end{aligned}\ ] ] in this case , ( [ f1 ] ) can be expressed as@xmath271 this family of test statistics include several commonly used tests such as the empirical modified kullback - leibler statistic ( @xmath272 , with @xmath273 ) , the empirical freeman - tukey statistic ( @xmath274 , with @xmath275 ) , the empirical likelihood ratio test statistic ( @xmath272 , with @xmath276 ) , the empirical cressie - read statistic ( @xmath272 , with @xmath277 ) , and the empirical pearson s chi - square statistic ( @xmath272 , with @xmath278 ) .
similarly , ( [ f2 ] ) can be expressed as @xmath279 the @xmath43-dimensional vectors @xmath280 and @xmath104 are obtained by solving ( [ ec ] ) with @xmath121 and @xmath48 being specified .
we now examine , through monte carlo simulations , the performance of the confidence intervals ( cis ) obtained through the empirical power divergence test statistics in ( [ pd1 ] ) and ( [ pd2 ] ) for the choices @xmath281 , when the sample sizes are small , @xmath282 , and the nominal confidence levels are @xmath283 and @xmath238 . even though the asymptotic distribution is equivalent for the different empirical divergence based statistics , in practice we need to work with finite sample sizes and it is therefore important to compare the cis based on @xmath272 and @xmath284 , in different scenarios .
we focus on the coverage probability and average width of the cis of a mean based on @xmath285 simulated samples under the same setting as example 1 of qin and lawless ( 1994 ) . in this case , a sample of univariate ( @xmath219 ) i.i.d .
random variables , @xmath286 , of size @xmath32 was considered , with mean @xmath287 and variance @xmath288 , i.e. , @xmath289 and@xmath290 0.8pt @xmath139{llllccccccc } & & & & \multicolumn{7}{c}{$\mathcal{n}(\theta,\theta^{2}+1)$ , $ \theta _ { 0}=0$}\\\hline & \hspace*{0.35 cm } & & \hspace*{0.35 cm } & \multicolumn{3}{c}{$t_{n}^{\lambda } ( \widehat{\theta}_{e , n},\theta_{0})$ } & \hspace*{0.35 cm } & \multicolumn{3}{c}{$s_{n}^{\lambda}(\widehat{\theta}_{e , n},\theta_{0})$}\\\cline{4 - 7}\cline{9 - 11 } & & & & cov ( \% ) & \hspace*{0.35 cm } & avw & & cov ( \% ) & \hspace*{0.35 cm } & avw\\\hline & & \multicolumn{1}{r}{$\lambda$ } & & \multicolumn{7}{c}{$n=30$}\\ $ ] 90%@xmath29195%@xmath29290%@xmath29395%@xmath29490%@xmath29595%@xmath296{llllccccccc } & & & & \multicolumn{7}{c}{$\mathcal{n}(\theta,\theta^{2}+1)$ , $ \theta _ { 0}=1$}\\\hline & \hspace*{0.35 cm } & & \hspace*{0.35 cm } & \multicolumn{3}{c}{$t_{n}^{\lambda } ( \widehat{\theta}_{e , n},\theta_{0})$ } & \hspace*{0.35 cm } & \multicolumn{3}{c}{$s_{n}^{\lambda}(\widehat{\theta}_{e , n},\theta_{0})$}\\\cline{4 - 7}\cline{9 - 11 } & & & & cov ( \% ) & \hspace*{0.35 cm } & avw & & cov ( \% ) & \hspace*{0.35 cm } & avw\\\hline & & \multicolumn{1}{r}{$\lambda$ } & & \multicolumn{7}{c}{$n=30$}\\ $ ] 90%@xmath29795%@xmath29890%@xmath29995%@xmath30090%@xmath30195%@xmath302 in the present study , we have considered an additional sample of size @xmath303 , compared to those of qin and lawless ( 1994 ) . for constructing the test statistic ,
an unknown distribution of @xmath304 is assumed , but for this simulation study , normally distributed random variables were taken with @xmath305 .
it is important to mention here that baggerly ( 1998 ) studied the coverage probabilities for the normal distribution , but only for the family of statistics in ( [ pd1 ] ) .
the results of the simulation study are summarized in table [ t2 ] and figure [ fig1 ] for the empirical power divergence test statistics in ( [ pd1 ] ) and ( [ pd2 ] ) for two different samples when the nominal confidence level is @xmath238 . in figure [ fig1 ]
, it can be seen that the confidence intervals based on both statistics , @xmath306 and @xmath307 , with small values of @xmath281 tend to be narrower in width , and this is also seen in the simulation results of table [ t2 ] .
in fact , the simulation study of baggerly showed that @xmath308 had the worst coverage probability , and this is also seen in the present simulation study .
this is in accordance with the results obtained for two specific samples analyzed in figure [ fig1 ] .
the smallest average width of confidence intervals is obtained for @xmath309 , which is the so - called empirical modified likelihood ratio test statistic or empirical minimum discrimination information statistic ( see gokhale and kullback , 1978 ) , and this might explain , when @xmath310 , the low coverage probability as well . often , the coverage probability closest to the nominal level is also obtained for the same test statistics , @xmath309 when @xmath311 but the empirical likelihood ratio test has a closer coverage probability when @xmath310 .
when both characteristics of a ci , viz . , average width and coverage probability , are taken into account , the empirical modified likelihood ratio test statistic , @xmath309 , is a good compromise and it turns out to be the best for @xmath310 in all the cases considered .
finally , another important feature seen in figure [ fig1 ] is that even though the support of the asymptotic distribution of the test statistics is strictly positive , it is possible to find samples for which @xmath306 is negative , while @xmath312 is always strictly positive .
this usually happens when @xmath313 is very close to @xmath314 .
@xmath139{cc}\multicolumn{2}{c}{{\fbox{\includegraphics [ natheight=1.218500 in , natwidth=1.540200 in , height=1.0637 in , width=1.337 in ] { fig9.pdf } } } } \\{\includegraphics [ natheight=2.965400 in , natwidth=4.483200 in , height=2.252 in , width=3.3909 in ] { fig8.pdf } } & { \includegraphics [ natheight=2.965400 in , natwidth=4.483200 in , height=2.252 in , width=3.3909 in ] { fig12.pdf } } \\{\includegraphics [ natheight=2.965400 in , natwidth=4.483200 in , height=2.252 in , width=3.3909 in ] { fig10.pdf } } & { \includegraphics [ natheight=2.965400 in , natwidth=4.483200 in , height=2.252 in , width=3.3909 in ] { fig14.pdf } } \end{tabular } \ $ ] we conducted a simulation study using the same design as in the preceding subsection , but by considering @xmath315 of shifted observations in each sample : @xmath267 shifted observations out of @xmath316 in the sample , @xmath317 shifted observations out of @xmath318 , @xmath319 shifted observations out of @xmath320 .
the underlying distribution for shifted observations is taken to be @xmath321 , given that @xmath322 is the true distribution .
therefore , the shifted observations follow @xmath323 distribution when the true distribution is @xmath324 , while they follow @xmath325 distribution when the true distribution is @xmath326 . in table
[ t2b ] , the results of introducing @xmath315 of shifted observations is shown .
our interest is focused on identifying the statistics that are less sensitive ( robust statistics with respect to shifted observations ) as well as those that are quite sensitive ( non - robust statistics with respect to shifted observations ) . as pointed out in remark [ remark ] ,
all the test - statistics have the same infinitesimal robustness , so the differences in robustness could be atributable to the magnitude of the shift .
when @xmath310 , the best statistics with shifted observations are exactly the same as ones without shifted observations , but when the dispersion of the data is higher , i.e. @xmath311 , the closest coverage probability to the nominal level is not for the empirical likelihood ratio test ; we further observe that @xmath306 and @xmath307 not only have the narrowest cis but also the highest coverage probabilities .
this means that the empirical modified likelihood ratio test statistic , @xmath327 , is a robust statistic with respect to shifted observations in comparison to the empirical likelihood ratio test , with respect to the coverage probability and width of cis .
this also agrees with the conclusion obtained from table [ tci ] .
0.8pt @xmath139{llllccccccc } & & & & \multicolumn{7}{c}{$\mathcal{n}(\theta,\theta^{2}+1)$ , $ \theta _ { 0}=0$}\\\hline & \hspace*{0.35 cm } & & \hspace*{0.35 cm } & \multicolumn{3}{c}{$t_{n}^{\lambda } ( \widehat{\theta}_{e , n},\theta_{0})$ } & \hspace*{0.35 cm } & \multicolumn{3}{c}{$s_{n}^{\lambda}(\widehat{\theta}_{e , n},\theta_{0})$}\\\cline{4 - 7}\cline{9 - 11 } & & & & cov ( \% ) & \hspace*{0.35 cm } & avw & & cov ( \% ) & \hspace*{0.35 cm } & avw\\\hline & & \multicolumn{1}{r}{$\lambda$ } & & \multicolumn{7}{c}{$n=30$}\\ $ ] 90%@xmath32895%@xmath32990%@xmath33095%@xmath33190%@xmath33295%@xmath333{llllccccccc } & & & & \multicolumn{7}{c}{$\mathcal{n}(\theta,\theta^{2}+1)$ , $ \theta _ { 0}=1$}\\\hline & \hspace*{0.35 cm } & & \hspace*{0.35 cm } & \multicolumn{3}{c}{$t_{n}^{\lambda } ( \widehat{\theta}_{e , n},\theta_{0})$ } & \hspace*{0.35 cm } & \multicolumn{3}{c}{$s_{n}^{\lambda}(\widehat{\theta}_{e , n},\theta_{0})$}\\\cline{4 - 7}\cline{9 - 11 } & & & & cov ( \% ) & \hspace*{0.35 cm } & avw & & cov ( \% ) & \hspace*{0.35 cm } & avw\\\hline & & \multicolumn{1}{r}{$\lambda$ } & & \multicolumn{7}{c}{$n=30$}\\ $ ] 90%@xmath33495%@xmath33590%@xmath33695%@xmath33790%@xmath33895%@xmath339
now , let us consider the problem of testing the composite null hypothesis@xmath340 the rest of this section , proceeds as follows . in section [ seccomp1 ] , we introduce the empirical phi - divergence test statistic for the composite null hypothesis in ( [ 2.1.3 ] ) .
section [ seccomp2 ] is devoted to the asymptotic results . in section [ seccomp3 ] , we present some results regarding the power function of the family of empirical test statistics proposed here . in section [ seccomp4 ] ,
a simulation study is carried out to evaluate the performance of the proposed test procedure in comparison to some other tests . in the following
, we shall assume as in qin and lawless ( 1995 ) , that the unknown parameter vector @xmath5 is defined through @xmath18 estimating functions , given in ( [ estf ] ) .
with regard to ( [ 2.1.3 ] ) , we shall assume that @xmath341 @xmath342 is a vector - valued function such that the @xmath343 matrix @xmath344@xmath345 exists and is continuous at @xmath5 and that @xmath346@xmath344@xmath347 ( @xmath348 ) .
we denote by @xmath349 the empirical restricted maximum likelihood estimator of @xmath5 , obtained by minimizing @xmath350 $ ] , subject to @xmath351 and @xmath352@xmath353 , i.e. , ( [ ec ] ) or@xmath354 where @xmath355 provides @xmath42 in terms of @xmath356 . upon using the lagrange multiplier method , once again
, we have@xmath357 where @xmath358 is a @xmath359-dimensional vector of lagrange multipliers , and differentiating @xmath360 with respect to @xmath5 and @xmath358 , we obtain@xmath361 therefore , @xmath362 is obtained as the solution of ( [ seq2 ] ) and @xmath363 and conditions ( [ cond ] ) and ( [ empf2 ] ) must be also satisfied .
the empirical likelihood ratio test for testing ( [ 2.1.3 ] ) has the expression @xmath364 where @xmath48 is the empirical maximum likelihood estimator of the parameter @xmath5 , defined in section [ sec1 ] ( case @xmath18 ) , for which @xmath365 .
taking into account the results stated in section [ sec1 ] for @xmath18 , we have @xmath366 , and so it is easy to show that the expression in ( [ lrt2 ] ) can be written as @xmath367 defining a family of empirical phi - divergence test statistics for testing the hypothesis in ( [ 2.1.3 ] ) , we have@xmath368 where the @xmath56-divergence measures are as defined in ( [ eqdiv ] ) . since @xmath366 , both families of empirical phi - divergence test statistics are equivalent and taking into account ( [ f2 ] ) and @xmath369 , we have@xmath370 under the same conditions of theorem [ th1 ] and the conditions given in section [ seccomp1 ] for @xmath371 , it is assumed that there exists a neighbourhood of @xmath121 in which @xmath372 is bounded by some some constant , we have @xmath373{c}\widetilde{\boldsymbol{\theta}}_{e , n}-\boldsymbol{\theta}_{0}\\ \widetilde{\boldsymbol{\nu}}_{e , n}\end{array } \right ) \underset{n\rightarrow\infty}{\overset{\mathcal{l}}{\rightarrow}}\mathcal{n}\left ( \boldsymbol{0}_{p+q},\left ( \begin{array } [ c]{cc}\boldsymbol{p}\left ( \boldsymbol{\theta}_{0}\right ) & \boldsymbol{0}_{p\times q}\\ \boldsymbol{0}_{q\times p } & \boldsymbol{q}\left ( \boldsymbol{\theta}_{0}\right ) \end{array } \right ) \right ) , \label{2.5}\ ] ] where @xmath121 is the true value of the parameter @xmath5 and@xmath374 this result is derived by taking into account , the facts that@xmath375 for a complete proof , one may refer to qin and lawless ( 1995 ) .
[ th2]under @xmath128 in ( [ 2.1.3 ] ) and the assumptions above , we have@xmath376 let us consider the function@xmath377 a second - order taylor expansion of @xmath378 around @xmath379 , with @xmath380 , gives@xmath381 but , @xmath382 and @xmath383 by the strong law of large numbers , @xmath384 , and so ( [ o1 ] ) can be rewritten as@xmath385 consequently,@xmath386 with @xmath387 , which means that ( [ th2r ] ) holds .
assume that @xmath388 is the true value of the unknown parameter so that @xmath389 , and that there exists a @xmath390 such that the restricted empirical maximum likelihood estimator satisfies @xmath391
. then , we have @xmath392{c}\widehat{\boldsymbol{\theta}}_{e , n}\\ \widetilde{\boldsymbol{\theta}}_{e , n}\end{array } \right ) -\left ( \begin{array } [ c]{c}\boldsymbol{\theta}^{\ast}\\ \boldsymbol{\theta}_{0}\end{array } \right ) \right ) \underset{n\rightarrow\infty}{\overset{\mathcal{l}}{\rightarrow}}\mathcal{n}\left ( \left ( \begin{array } [ c]{c}\boldsymbol{0}_{p}\\ \boldsymbol{0}_{p}\end{array } \right ) , \left ( \begin{array } [ c]{cc}\boldsymbol{v}\left ( \boldsymbol{\theta}_{0}\right ) & \boldsymbol{a}_{12}\left ( \boldsymbol{\theta}^{\ast},\boldsymbol{\theta}_{0}\right ) \\ \boldsymbol{a}_{12}\left ( \boldsymbol{\theta}^{\ast},\boldsymbol{\theta}_{0}\right ) ^{t } & \boldsymbol{p(\theta}_{0 } ) \end{array } \right ) \right ) , \label{a2}\ ] ] where @xmath393 is a @xmath394 matrix . [ thth]under the conditions stated above , we have @xmath395 where @xmath396 with @xmath397 given by ( [ tao ] ) and@xmath398 based on theorem [ thth ] , we get an approximation of the power function at @xmath159 as @xmath399 if some alternative @xmath160 is the true parameter , then the probability of rejecting @xmath121 with the rejection rule @xmath400 , for a fixed significance level @xmath180 , tends to one as @xmath401 thus , the test is consistent in the sense of fraser ( 1957 ) .
we may also find an approximation to the power of @xmath402 at an alternative hypothesis close to the null hypothesis .
let @xmath403 be a given alternative , and let @xmath121 be the element in @xmath404 closest to @xmath405 in terms of euclidean distance .
in order to introduce contiguous alternative hypotheses , we may consider a fixed @xmath188 @xmath406 and allow @xmath405 to tend to @xmath121 as @xmath32 increases in the following manner : @xmath407 [ th2b]under the assumptions of theorem [ th2 ] and @xmath191 given in ( [ c1 ] ) , we have@xmath408 where @xmath409 .
if we substitute ( [ o2 ] ) into ( [ est2 ] ) , we obtain@xmath410 since @xmath411 , the taylor expansion @xmath412 yields @xmath413 by following the same steps as in the proof of theorem [ th1b ] , we have that under @xmath191 , @xmath414 .
hence , we obtain @xmath415 where @xmath416 .
we thus obtain@xmath417 which completes the proof .
a second way to consider contiguous alternative hypotheses is to relax the condition @xmath418 defining the null hypothesis @xmath404 .
let @xmath419 be such that @xmath420 .
consider the following sequence @xmath405 of parameters approaching @xmath404:@xmath421 a taylor expansion of @xmath352@xmath422 around @xmath121 yields @xmath423 upon substituting @xmath203 in the previous formula and taking into account that @xmath352@xmath424 , we have@xmath425 then , the equivalence between @xmath426 and @xmath191 is obtained for @xmath427 .
we thus have the following result . under the contiguous alternative hypothesis in ( [ c2 ] ) ,
we have @xmath428 with @xmath429 and @xmath430 as defined in ( [ bm ] ) . from a sample @xmath286 of i.i.d .
random variables with @xmath431=\mu$ ] and @xmath432=\sigma^{2},$ ] we wish to test that the coefficient of variation is @xmath185 , i.e. , @xmath433 against@xmath434 . in order to make a decision
, we need to obtain the maximum likelihood estimator under the restriction @xmath435 , with @xmath436 .
the estimating equations in this case are @xmath437 , @xmath438 .
if we establish a bijective transformation between @xmath436 and @xmath439 , and are able to obtain the empirical restricted maximum likelihood estimator of @xmath5 , @xmath349 , due to the invariance property , we can obtain the empirical restricted maximum likelihood estimator of @xmath440 , @xmath441 , by taking the inverse of the transformation .
let @xmath442 and @xmath443 .
then@xmath444 with @xmath445 .
the estimating equations under the new parameterization are @xmath446 in general , we have the system of @xmath447 equations as follows:@xmath448 which are equivalent to @xmath449 for obtaining @xmath450 .
observe that@xmath451 since the sum of the probabilities is 1 , and consequently the third and fourth equations become @xmath452 , @xmath453 , and we also know that @xmath454 .
so , the optimization problem is reduced to simply to@xmath455 the solution @xmath349 must satisfy that the @xmath32 probabilities are not less than zero , i.e. , @xmath456 and if such a solution exists , it is unique ; see qin and lawless ( 1994 ) . in our empirical study
, we solved the system of these two equations by using the nag subroutine in fortran , ` c05pbf ` . to compare the exact coverage probabilities of the confidence intervals for the coefficient of variation based on some empirical phi - divergence test statistics ,
a simulation study was conducted separately for continuous and discrete distributions since we found that the rate of convergence to the asymptotic distribution is much faster for discrete distributions ( poisson ) than for continuous distributions ( normal and student @xmath457 ) .
based on @xmath458 samples of sizes @xmath316 , @xmath318 , @xmath320 , @xmath459 , @xmath460 , @xmath461 from @xmath323 and @xmath462 distributions and sample sizes @xmath463 , @xmath214 , @xmath464 , @xmath316 from poisson , @xmath465 , the so - called power - divergence measures ( see cressie and read ( 1984 ) ) were considered to construct the phi - divergence test statistics , @xmath466 that is , for each @xmath467 , we have a different divergence measure by taking @xmath468 , where @xmath469 is as defined earlier in section [ simulation ] . in addition , the empirical generalized wald test statistic , the empirical generalized score test statistic , and the empirical lagrange multiplier test statistic were also obtained . for this purpose ,
the following auxiliary matrices are necessary:@xmath470 where @xmath471 $ ] and @xmath472 $ ] are unknown .
the matrices @xmath473 and @xmath474 can be replaced by any consistent estimator @xmath475 and also @xmath476 and @xmath477 by@xmath478 the unconstrained maximum likelihood estimators were calculated as the solution of the system of @xmath479 equations:@xmath480 that is , @xmath481 first , the expression of the empirical generalized wald test statistic is obtained as @xmath482 next , the empirical generalized score test statistic is obtained as @xmath483 ^{2}}{\frac{1}{n}\sum_{i=1}^{n}x_{i}^{4}-8\widetilde{u}\frac{1}{n}\sum_{i=1}^{n}x_{i}^{3}-2\widehat{v}\widetilde{v}+\widetilde{v}^{2}+8\widehat{u}\widetilde{u}\widetilde{v}+24\widetilde{u}^{2}\widehat{v}-8\widetilde{u}^{2}\widetilde{v}-32\widehat{u}\widetilde{u}^{3}+16\widetilde{u}^{4}}\\ & = \frac{n\left [ \widehat{v}+\widetilde{v}-4\widetilde{u}\widehat{u}\right ] ^{2}}{\frac{1}{n}\sum_{i=1}^{n}x_{i}^{4}-8\widetilde{u}\frac{1}{n}\sum _ { i=1}^{n}x_{i}^{3}-8\widehat{u}\widetilde{u}\widetilde{v}+10\widehat{v}\widetilde{v}+\widetilde{v}^{2}},\end{aligned}\ ] ] where the second equality is obtained by taking into account that @xmath484 ; finally , the empirical lagrange multiplier test statistic is obtained as @xmath485 [ c]ccccccccccc@xmath32 & nom . level & @xmath486 & @xmath487 & @xmath488 & @xmath489 & @xmath490 & @xmath491 & @xmath492 & @xmath493 & @xmath494 + 30 & 0.90 & 0.8537 & 0.8615 & 0.8672 & 0.8717 & * 0.8723 * & 0.8683 & 0.8676 & 0.8671 & 0.8689 + 45 & 0.90 & 0.8682 & 0.8745 & 0.8789 & 0.8828 & 0.8836 & 0.8811 & 0.8778 & 0.8780 & * 0.8842 * + 60 & 0.90 & 0.8775 & 0.8824 & 0.8860 & 0.8895 & 0.8908 & 0.8899 & 0.8852 & 0.8854 & * 0.8930 * + 75 & 0.90 & 0.8815 & 0.8856 & 0.8892 & 0.8923 & 0.8932 & 0.8922 & 0.8870 & 0.8880 & * 0.8962 * + 90 & 0.90 & 0.8851 & 0.8892 & 0.8919 & 0.8947 & 0.8955 & 0.8951 & 0.8900 & 0.8900 & * 0.8996 * + 105 & 0.90 & 0.8870 & 0.8905 & 0.8938 & 0.8965 & 0.8976 & 0.8976 & 0.8916 & 0.8917 & * 0.9009 * + 30 & 0.95 & 0.9062 & 0.9143 & 0.9210 & 0.9251 & * 0.9258 * & 0.9211 & 0.9228 & 0.9169 & 0.9149 + 45 & 0.95 & 0.9199 & 0.9265 & 0.9312 & 0.9352 & * 0.9358 * & 0.9321 & 0.9308 & 0.9275 & 0.9285 + 60 & 0.95 & 0.9286 & 0.9342 & 0.9392 & 0.9425 & * 0.9426 * & 0.9401 & 0.9374 & 0.9336 & 0.9360 + 75 & 0.95 & 0.9322 & 0.9379 & 0.9417 & * 0.9446 * & * 0.9446 * & 0.9427 & 0.9399 & 0.9367 & 0.9398 + 90 & 0.95 & 0.9350 & 0.9398 & 0.9432 & 0.9457 & * 0.9460 * & 0.9443 & 0.9415 & 0.9387 & 0.9419 + 105 & 0.95 & 0.9372 & 0.9417 & 0.9449 & 0.9477 & * 0.9480 * & 0.9464 & 0.9428 & 0.9403 & 0.9447 + @xmath495 [ c]ccccccccccc@xmath32 & nom . level & @xmath486 & @xmath487 & @xmath488 & @xmath489 & @xmath490 & @xmath491 & @xmath492 & @xmath493 & @xmath494 + 30 & 0.90 & 0.8003 & 0.8086 & 0.8145 & 0.8175 & 0.8174 & 0.8130 & * 0.8302 * & 0.8148 & 0.8208 + 45 & 0.90 & 0.8232 & 0.8306 & 0.8356 & 0.8387 & 0.8384 & 0.8343 & * 0.8465 * & 0.8333 & 0.8417 + 60 & 0.90 & 0.8354 & 0.8424 & 0.8473 & 0.8489 & 0.8488 & 0.8446 & * 0.8561 * & 0.8446 & 0.8526 + 75 & 0.90 & 0.8440 & 0.8500 & 0.8541 & 0.8568 & 0.8573 & 0.8535 & * 0.8621 * & 0.8531 & 0.8597 + 90 & 0.90 & 0.8517 & 0.8571 & 0.8610 & 0.8631 & 0.8629 & 0.8586 & * 0.8683 * & 0.8603 & 0.8645 + 105 & 0.90 & 0.8577 & 0.8627 & 0.8657 & 0.8668 & 0.8665 & 0.8621 & * 0.8737 * & 0.8654 & 0.8665 + 30 & 0.95 & 0.8584 & 0.8696 & 0.8776 & 0.8825 & 0.8822 & 0.8740 & * 0.8897 * & 0.8665 & 0.8709 + 45 & 0.95 & 0.8807 & 0.8906 & 0.8976 & 0.9012 & 0.9008 & 0.8938 & * 0.9037 * & 0.8847 & 0.8923 + 60 & 0.95 & 0.8925 & 0.9007 & 0.9074 & 0.9103 & 0.9098 & 0.9026 & * 0.9107 * & 0.8958 & 0.9004 + 75 & 0.95 & 0.9013 & 0.9088 & 0.9140 & 0.9165 & 0.9159 & 0.9087 & * 0.9169 * & 0.9040 & 0.9072 + 90 & 0.95 & 0.9080 & 0.9150 & 0.9199 & 0.9218 & 0.9209 & 0.9146 & * 0.9220 * & 0.9108 & 0.9112 + 105 & 0.95 & 0.9135 & 0.9201 & 0.9234 & 0.9244 & 0.9235 & 0.9160 & * 0.9265 * & 0.9163 & 0.9125 + @xmath495 [ c]ccccccccccc@xmath32 & nom . level & @xmath486 & @xmath487 & @xmath488 & @xmath489 & @xmath490 & @xmath491 & @xmath492 & @xmath493 & @xmath494 + 15 & 0.90 & 0.8856 & 0.8878 & 0.9010 & 0.9089 & 0.9089 & 0.9071 & 0.8962 & 0.9075 & * 0.9099 * + 20 & 0.90 & 0.8814 & 0.8849 & 0.8875 & 0.8904 & 0.8925 & 0.8923 & * 0.9074 * & 0.8962 & 0.9013 + 25 & 0.90 & 0.8758 & 0.8793 & 0.8874 & 0.8934 & 0.8930 & 0.8923 & * 0.9048 * & 0.8882 & 0.8974 + 30 & 0.90 & 0.8790 & 0.8848 & 0.8861 & 0.8916 & 0.8923 & 0.8901 & * 0.9027 * & 0.8933 & 0.8928 + 15 & 0.95 & 0.9329 & 0.9392 & 0.9529 & 0.9606 & 0.9613 & 0.9639 & 0.9546 & 0.9440 & * 0.9670 * + 20 & 0.95 & 0.9359 & 0.9452 & 0.9486 & 0.9522 & 0.9539 & * 0.9544 * & 0.9501 & 0.9505 & 0.9519 + 25 & 0.95 & 0.9336 & 0.9388 & 0.9409 & 0.9434 & 0.9438 & 0.9442 & * 0.9502 * & 0.9468 & 0.9472 + 30 & 0.95 & 0.9318 & 0.9336 & 0.9406 & 0.9442 & 0.9449 & 0.9446 & * 0.9519 * & 0.9408 & 0.9466 + @xmath495 focusing on the coefficients of variation of the two continuous distributions , the exact coverage probabilities of the confidence intervals based on the empirical power divergence test statistics with @xmath496 are presented in tables [ t2.1 ] and [ t2.2 ] when the nominal coverage probability based on the asymptotic distribution is either @xmath283 or @xmath238 . in addition , the empirical generalized wald test statistic , the empirical generalized score test statistic and the empirical lagrange multiplier test statistic are also considered for comparative purposes . from these results , we note that the empirical likelihood ratio test is not satisfactory and that among the empirical power divergence test statistics there is a good choice , for the underlying normal distribution , in the empirical chi - squared test statistic ( @xmath278 ) , and for the non - normal underlying distribution in the empirical cressie - read test statistic ( @xmath497 ) , but there is not much difference between their performance . if we consider other test statistics , for the normally distributed observations , with theoretical asymptotic coverage as @xmath238 , the empirical chi - squared test statistic ( @xmath278 ) while for @xmath283 coverage probability , the empirical lagrange multiplier test statistic are seen to be the best ones but there is very little difference with respect to the empirical chi - squared test statistic ( @xmath278 ) .
for the non - normal distribution , the empirical wald test statistic is slightly superior than the empirical cressie - read test statistic , while the opposite seems to be the case for the normal distribution . since in practice we do not know the form of the underlying distribution , based on this simulation study
, we would recommend the use of either the empirical cressie - read test statistic or the empirical wald test statistic . in table
[ t2.3 ] , the same study has been carried out but for the case of poisson distribution .
the empirical likelihood ratio test is once again unsatisfactory and that the empirical cressie - read test statistic and the empirical chi - squared test statistic have a better coverage probability close to the nominal level .
further , the empirical wald test statistic seems to be slightly superior than the empirical cressie read test statistic since it has a greater coverage probability in 5 of the 8 cases .
in a non - parametric setting , we have proposed here a broad family of empirical test statistics based on @xmath56-divergence measures , first for a simple null hypothesis and then for a composite null hypothesis . through numerical examples and simulations
, it has been shown that confidence intervals constructed thought the empirical @xmath56-divergence based statistics improve the coverage probability of the empirical likelihood ratio test slightly . however , the most promising advantage of this new family of test statistics is that some members outperform the empirical likelihood ratio test in the presence of some shifted observations in the data .
the approximation of the power function based on a specific sample provides an insight about the most appropriate robust @xmath56-divergence statistic .
these robust statistics tend to yield narrower confidence intervals in comparison to the empirical likelihood ratio test .
the development of these empirical @xmath56-divergence tests in the two - sample and multi - sample situations will be of great interest .
we are currently working in this direction and hope to report these findings in a future paper .
menndez , m. l. , morales , d. , pardo , l. and salicr , m. ( 1995 ) .
asymptotic behavior and statistical applications of divergence measures in multinomial populations : a unified study .
_ statistical papers , _ * 36 , * 1 - 29 .
menndez , m. l. , pardo , j. a. , pardo , l. and pardo , m. c. ( 1997 ) .
asymptotic approximations for the distributions of the @xmath183-divergence goodness - of - fit statistics : applications to rnyi s statistic .
_ kybernetes , _ * 26 , * 442 - 452 .
|
the main purpose of this paper is to introduce first a new family of empirical test statistics for testing a simple null hypothesis when the vector of parameters of interest are defined through a specific set of unbiased estimating functions .
this family of test statistics is based on a distance between two probability vectors , with the first probability vector obtained by maximizing the empirical likelihood on the vector of parameters , and the second vector defined from the fixed vector of parameters under the simple null hypothesis .
the distance considered for this purpose is the phi - divergence measure .
the asymptotic distribution is then derived for this family of test statistics . the proposed methodology is illustrated through the well - known data of newcomb s measurements on the passage time for light .
a simulation study is carried out to compare its performance with respect to the empirical likelihood ratio test when confidence intervals are constructed based on the respective statistics for small sample sizes .
the results suggest that the empirical modified likelihood ratio test statistic provides a competitive alternative to the empirical likelihood ratio test statistic , and is also more robust than the empirical likelihood ratio test statistic in the presence of contamination in the data .
finally , we propose empirical phi - divergence test statistics for testing a composite null hypothesis and present some asymptotic as well as simulation results to study the performance of these test procedures .
= 1 * :* 62e20 * :* empirical likelihood , empirical phi - divergence test statistics , influence function , phi - divergence measures , power function , empirical likelihood ratio , empirical modified likelihood ratio .
|
[ [ background ] ] background + + + + + + + + + + problem , known as a thermodynamical model of liquid - solid phase transition , has been widely studied since jeseph stefan s work in 1889@xcite .
typical applications include sea ice melting and freezing @xcite , continuous casting of steel @xcite , crystal - growth @xcite , thermal energy storage system @xcite , and lithium - ion batteries @xcite .
for instance , time evolution of the arctic sea ice thickness and temperature profile was modeled in @xcite using the stefan problem , and the correspondence with the empirical data was investigated .
apart from the thermodynamical model , the stefan problem has been employed to model a type of population growth , such as in @xcite for tumor growth process and in @xcite for information diffusion on social networks .
while phase change phenomena described by the stefan condition appear in various kinds of science and engineering processes , their mathematical analysis remains quite challenging due to the implicitly given moving interface that reflects the time evolution of a spatial domain , so - called free boundary problem " @xcite .
physically , the classical one - phase stefan problem describes the temperature profile along a liquid - solid material , where the dynamics of the liquid - solid interface is influenced by the heat flux induced by melting or solidification phenomena .
mathematically , the problem involves a diffusion partial differential equation ( pde ) coupled with an ordinary differential equation ( ode ) . here
, the pde describes the heat diffusion that provokes melting or solidification of a given material and the ode delineates the time - evolution of the moving front at the solid - liquid interface .
while the numerical analysis of the one - phase stefan problem is broadly covered in the literature , their control related problems have been addressed relatively fewer .
in addition to it , most of the proposed control approaches are based on finite dimensional approximations with the assumption of an explicitly given moving boundary dynamics @xcite,@xcite,@xcite .
diffusion - reaction processes with an explicitly known moving boundary dynamics are investigated in @xcite based on the concept of inertial manifold @xcite and the partitioning of the infinite dimensional dynamics into slow and fast finite dimensional modes .
motion planning boundary control has been adopted in @xcite to ensure asymptotic stability of a one - dimensional one - phase nonlinear stefan problem assuming a prior known moving boundary and deriving the manipulated input from the solutions of the inverse problem .
however , the series representation introduced in @xcite leads to highly complex solutions that reduce controller design possibilities . for control objectives ,
infinite - dimensional frameworks that lead to significant challenges in the process characterization have been developed for the stabilization of the temperature profile and the moving interface of the stefan problem .
an enthalpy - based boundary feedback control law that ensures asymptotical stability of the temperature profile and the moving boundary at a desired reference , has been employed in @xcite .
lyapunov analysis is performed in @xcite based on a geometric control approach which enables to adjust the position of a liquid - solid interface at a desired setpoint while exponentially stabilizing the @xmath1-norm of the distributed temperature .
however , the results in @xcite are stated based on physical assumptions on the liquid temperature being greater than melting point , which needs to be guaranteed by showing strictly positive boundary input .
backstepping controller design employs an invertible transformation that maps an original system into a stable target system . during the past few decades
, such a controller design technique has been intensely exploited for the boundary control of diffusion pdes defined on a fixed spatial domain as firstly introduced in @xcite for the control of a heat equation via measurement of domain - average temperature . for a class of one - dimensional linear parabolic partial integro - differential equations ,
a state - estimator called backstepping observer was introduced in @xcite , which can be combined to the earlier state - feedback boundary controller designed for an identical system @xcite to systematically construct output feedback regulators . over the same period ,
advection - reaction - diffusion systems with space - dependent thermal conductivity or time - dependent reactivity coefficient were studied in @xcite , and parabolic pdes containing unknown destabilizing parameters affecting the interior of the domain or unknown boundary parameters were adaptively controlled in @xcite combining the backstepping control technique with passive or swapping identifiers .
further advances on the backstepping control of diffusion equations defined on a multidimensional space or involving in - domain coupled systems can be found in @xcite .
results devoted to the backstepping stabilization of coupled systems described by a diffusion pde in cascade with a linear ode has been primarily presented in @xcite with dirichlet type of boundary interconnection and extended to neuman boundary interconnection in @xcite . for systems relevant with stefan problem
, @xcite designed a backstepping output feedback controller that ensures the exponential stability of an unstable parabolic pde on _ a priori _ known dynamics of moving interface which is assumed to be an analytic function in time .
moreover , for pde - ode cascaded systems under a state - dependent moving boundary , @xcite developed a local stabilization of nonlinear odes with actuator dynamics modeled by wave pde under time- and state - dependent moving boundary .
such a technique is based on the input delay and wave compensation for nonlinear odes designed in @xcite and its extension to state - dependent input delay compensation for nonlinear odes provided by @xcite .
however , the results in @xcite and @xcite that cover state - dependence problems do not ensure global stabilization due to a so - called feasibility condition that needs to be satisfied _
a priori_. [ [ results - and - contributions ] ] results and contributions + + + + + + + + + + + + + + + + + + + + + + + + + our previous result in @xcite is the first contribution in which global exponential stability of the stefan problem with an implicitly given moving interface motion is established without imposing any _ a priori _ given restriction under a state feedback design of neumann boundary control , and in @xcite we developed the design of an observer - based output feedback control .
this paper combines the results of our conference papers @xcite and @xcite and extends them by providing the robustness analysis of the closed - loop system with the state feedback control to the plant parameters mismatch and the corresponding results under dirichlet boundary control .
first , a state feedback control law that requires the measurement of the liquid phase temperature profile and the moving interface position is constructed using a novel nonlinear backstepping transformation .
the proposed state feedback controller achieves exponential stabilization of the temperature profile and the moving interface to the desired references in @xmath0-norm under the least restrictive condition on the setpoint imposed by energy conservation law .
robustness of the state feedback controller to thermal diffusivity and latent heat of fusion mismatch is also characterized by explicitly given bounds of the uncertainties magnitude .
second , an exponentially state observer which reconstruct the entire distributed temperature profile based solely on the measurement of only the interface position , namely , the length of the melting zone is constructed using the novel backstepping transformation . finally , combining the state feedback law with the state estimator , the exponential stabilization of the estimation error , the temperature profile , and the moving interface to the desired references in the @xmath2-norm under some explicitly given restrictions on the observer gain and the setpoint .
[ [ organizations ] ] organizations + + + + + + + + + + + + + the one - phase stefan problem with neumann boundary actuation is presented in section [ model ] , and its open loop - stability is discussed in section [ open ] .
the full - state feedback controller is constructed in section [ state - feedback ] with a robustness analysis to parameters perturbations .
section [ sec : estimation ] explains the observer design and in section [ sec : output ] is presented the observer - based output - feedback control . in section [ sec : tempcontrol ] , both state feedback and output feedback are derived for a dirichlet type of boundary actuation and robustness analysis is provided for the state feedback case .
simulations to support the theoretical results are given in section [ simu ] .
the paper ends with final remarks and future directions discussed in section [ sec : conclusion ] .
[ [ notations ] ] notations + + + + + + + + + throughout this paper , partial derivatives and several norms are denoted as @xmath3 part i : neumann boundary actuation ---------------------------------- in this first part , we design a boundary controller for the stefan problem with a neumannn boundary actuation .
the one - dimensional stefan problem is defined as the physical model which describes the melting or the solidification mechanism in a pure one - component material of length @xmath4 in one dimension as depicted in fig .
[ fig : stefan ] .
the dynamics of the process depends strongly on the evolution in time of the moving interface ( here reduced to a point ) at which phase transition from liquid to solid ( or equivalently , in the reverse direction ) occurs .
therefore , the melting or solidification mechanism that takes place in the physical domain @xmath5 $ ] induces the existence of two complementary time - varying sub - domains , namely , @xmath6 $ ] occupied by the liquid phase , and @xmath7 $ ] by the solid phase .
assuming a temperature profile uniformly equivalent to the melting temperature in the solid phase , a dynamical model associated to a melting phenomena ( see fig .
[ fig : stefan ] ) involves only the thermal behavior of the liquid phase . at a fundamental level ,
the thermal conduction for a melting component obeys the well known fourier s law @xmath8 where @xmath9 is a heat flux profile , @xmath10 is the thermal conductivity , and @xmath11 is a temperature profile . considering a melting material with a density @xmath12 and heat capacity @xmath13 , in the open domain @xmath14 , the energy conservation law is defined by the following equality @xmath15 + therefore , a heat flux entering the system at the boundary at @xmath16 of the liquid phase influences the dynamics of the solid - liquid interface @xmath17 . assuming that the temperature in the liquid phase is not lower than the melting temperature of the material @xmath18 and combining and
, one can derive the heat equation of the liquid phase as follows @xmath19 with the boundary conditions @xmath20 and the initial conditions @xmath21 moreover , for the process described in fig .
[ fig : stefan ] , the local energy balance at the position of the liquid - solid interface @xmath22 leads to the stefan condition defined as the following ode @xmath23 where @xmath24 denotes the latent heat of fusion .
equation expresses the velocity of the liquid - solid moving interface . for the sake of brevity , we refer the readers to @xcite , where the stefan condition is derived for a solidification process . for an homogeneous melting material ,
the stefan problem presented in section [ model ] exhibits some important properties that are stated in the following remarks .
_ as the moving interface @xmath17 governed by is unknown explicitly , the problem defined in is a nonlinear problem .
note that this non - linearity is purely geometric rather than algebraic .
_ [ assumption ] _ due to the so - called isothermal interface condition that prescribes the melting temperature @xmath18 at the interface through , this form of the stefan problem is a reasonable model only if the following condition holds : @xmath25,\end{aligned}\ ] ] _ the model validity requires the liquid temperature to be greater than the melting temperature and such a condition yields the following property on moving interface .
[ monoinc ] if the model validity holds , then the moving interface is always increasing , i.e. @xmath26 corollary [ monoinc ] is established by hopf s lemma as shown in @xcite .
hence , it is plausible to impose the following assumption on the initial values .
[ assumini ] the initial position of the moving interface satisfies @xmath27 and the lipschitz continuity of @xmath28 holds , i.e. @xmath29 assumption [ assumini ] is physically consistent with remark [ assumption ] .
we recall the following lemma that ensures , for the validity of the model . [ lemma1 ] for any @xmath30 on the finite time interval @xmath31 , the condition of the model validity holds .
the proof of lemma [ lemma1 ] is based on maximum principle as shown in @xcite . in this paper , we focus on this melting problem which imposes the constraint to keep the manipulated heat flux positive .
the steady - state solution @xmath32 of the system - with zero manipulating heat flux @xmath33 yields a uniform melting temperature @xmath34 and a constant interface position given by the initial data .
hence , the system - is marginally stable . in this section ,
we consider the asymptotical stabilization of the interface position @xmath17 at a desired reference setpoint @xmath35 , while the equilibrium temperature profile is maintained at @xmath18 .
thus , the control objective is formulated as @xmath36 the positivity of the manipulated heat flux in lemma [ lemma1 ] imposes a restriction on the setpoint due to an energy conservation of the system - .
the energy conservation law is given by @xmath37 the left hand side of denotes the growth of internal energy and its right hand side denotes the external work provided by the injected heat flux .
integrating the energy balance in @xmath38 from @xmath39 to @xmath40 and substituting and , the condition to achieve the control objective is given by @xmath41 from relation , one can deduce that for any positive heat flux control action which imposes @xmath30 , the internal energy for a given setpoint must be greater than the initial internal energy .
thus , the following assumption is required .
[ assumsetpoint ] the setpoint @xmath35 is chosen to satisfy @xmath42 therefore , assumption [ assumsetpoint ] stands as a least restrictive condition for the choice of setpoint .
for any given open - loop control law @xmath43 satisfying , the asymptotical stability of system at @xmath44 can be established and the following lemma holds .
[ rec * ] consider an open - loop setpoint control law @xmath45 which satisfies .
then , the interface converges asymptotically to the prescribed setpoint @xmath35 and consequently , conditions and hold .
the proof of lemma [ rec * ] can be derived straightforwardly from . to illustrate the introduced concept of open - loop
energy shaping control " action , we define @xmath46 as the difference of total energy given by the left - hand side of . hence , @xmath47 for instance , the rectangular pulse control law given by @xmath48\\ 0 & { \rm for}\quad t >
k\delta e/\bar{q } \end{array } \right \}\end{aligned}\ ] ] satisfies for any choice of the boundary heat flux @xmath49 and thereby , ensures the asymptotical stability of to the setpoint @xmath50 .
it is remarkable that adopting an open - loop control strategy such as the rectangular pulse , does not allow to improve the convergence speed .
moreover , the physical parameters of the model need to be known accurately to avoid robustness issues .
more generally , the practical implementation of open - loop control laws in engineering process is limited by performance and robustness issues , thus closed - loop control laws have to be designed to deal with such limitations .
in this section , we focus on the design of closed - loop backstepping control law for the one - phase stefan problem in order to achieve faster exponential convergence to the desired setpoint @xmath50 while ensuring the robustness of the closed - loop system to the uncertainty of the physical parameters .
we recall that from a physical point of view , for any positive heat flux @xmath51 , the irreversibility of the process restrict _ a priori _ the choice of the desired setpoint @xmath35 to satisfy . assuming that the temperature profile in the melt zone @xmath11 and the position of the moving interface @xmath17 are measured @xmath52 $ ] and @xmath53 , the following theorem holds : [ theo-1 ] consider a closed - loop system consisting of the plant and the control law @xmath54 where @xmath55 is an arbitrary controller gain .
assume that the initial conditions @xmath56 are compatible with the control law and satisfies .
then , for any reference setpoint @xmath35 which satisfies , the closed - loop system remains the condition of model validity and is exponentially stable in the sense of the norm @xmath57 is established through following steps : * a backstepping transformation for moving boundary pdes and the associated inverse transformation are constructed for a reference error system ( see section [ fbktarget ] ) .
* physical constraints that guarantee the positivity of the boundary heat flux under closed - loop control are derived ( see section [ positivness ] ) . *
the stability analysis of the target system that induces the stability of the original reference error system is performed ( see section [ fbkstability ] ) . following a standard procedure , for a given reference setpoint @xmath50
, we define the reference errors as @xmath58 respectively .
then , the reference error system associated to the coupled system is written as @xmath59 next , we introduce the following backstepping transformation @xmath60 which transforms system into the following target system " @xmath61 the derivation of the explicit gain kernel function , which enables to map into , is given in appendix [ appdbst ] . basically , the target system - is obtained by taking the derivatives of with respect to @xmath38 and @xmath62 respectively along the solution of - with the control law .
the invertibility of the transformation guarantees that the original system and the target system have equivalent stability properties .
the inverse transformation of is given by @xmath63 with an explicit gain kernel function @xmath64 as for the direct transformation , the derivation of the inverse tranformation is detailed in appendix [ appibst ] .
therefore , there exists a unique inverse transformation and the stability properties between @xmath65-system and @xmath66-system are identical .
as stated in remark [ assumption ] and lemma [ lemma1 ] , a strictly positive heat flux @xmath51 is required due to the fact that a negative heat flux may lead to a freezing process , which violates the condition of the model validity . to achieve the control objective @xmath67 , the aforementioned constraint for
a melting stefan problem does not allow the overshoot beyond the reference setpoint @xmath35 due to monotonically increasing property of the moving interface dynamics stated in . in this section
, we establish that the state feedback control law guarantees the following physical constraints " @xmath68 under assumption [ assumsetpoint ] , the closed - loop responses of the plant with the control law satisfies the physical constraints and , and hence the required conditions for the model validity , namely , and hold . by taking the time derivative of along the solution of , we have @xmath69 the differential equation leads to the dynamics of the control law to be exponentially decaying function , i.e. @xmath70 since the setpoint restriction implies @xmath71 , the dynamics of the state - feedback control law defined in verifies condition
. then by lemma 1 and corollary [ monoinc ] , the model validity and increasing property on interface are satisfied . applying and to the control law
, the following inequality is obtained @xmath72 in addition , inequality leads to @xmath73 .
thus , we have @xmath74 in the next section , inequalities and are used to establish the lyapunov stability of the target system - . in the following we establish the exponential stability of the closed - loop control system in @xmath0-norm of the temperature and the moving boundary based on the analysis of the associated target system . we consider a functional @xmath75 where @xmath76 is a positive parameter to be determined .
taking the derivative of along the solution of the target system - and applying young s , cauchy - schwartz , pointcare s , agmon s inequalities , with the help of and , we have @xmath77 where @xmath78 , @xmath79 .
the detailed derivation of is given in appendix [ stabilitytarget ] .
however , the second term of the right hand side of does not enable to directly conclude the exponential stability . to deal with it
, we introduce a new lyapunov function @xmath80 such that @xmath81 the time derivative of is written as @xmath82 and the following bound can be deduced for @xmath83 using @xmath84 hence , @xmath85 and using and , we arrive at @xmath86 from the definition of @xmath87 in the following holds @xmath88 such that @xmath89 . the detailed derivation from to
is described in appendix [ stabilityoriginal ] .
next , we show the robustness of the closed - loop system in this section , we investigate robustness of the backstepping control design to perturbations on the plant s parameters @xmath90 and @xmath91 .
physically , these perturbations are caused by the uncertainty of the thermal diffusivity and the latent heat of fusion .
hence , we consider the closed - loop system @xmath92 with the control law , where @xmath93 and @xmath94 are perturbation parameters such that @xmath95 and @xmath96 . defining the vector @xmath97 where @xmath98 , the following theorem is stated .
[ robust - thm ] consider a closed - loop system and the control law under assumption [ assumini ] and [ assumsetpoint ] .
then , for any pair of perturbations such that @xmath99 where @xmath100 the closed - loop system is exponentially stable in the sense of the norm . using the backstepping transformation , the target system associated to is defined as follows @xmath101 next we prove that the control law applied to the perturbed system , satisfies the physical constraints and . taking time derivative of along with , we arrive at @xmath102 the positivity of is shown using a contradiction argument .
assume that there exists @xmath103 such that @xmath30 for @xmath104 and @xmath105 .
then , lemma [ lemma1 ] and hopf s lemma leads to @xmath106 for @xmath104 . since @xmath99 with , implies that @xmath107 for all @xmath99 . using comparison principle , and
assumption [ assumsetpoint ] leads to @xmath108 .
thus @xmath109 which is in contradiction with the assumption @xmath105 .
consequently , holds by this contradiction argument .
accordingly , is established using the positivity of the heat flux @xmath43 defined in and the control law .
now , consider a functional @xmath110 where the parameters @xmath111 and @xmath112 are chosen to be @xmath113 , @xmath114 . taking the time derivative of along the solution of , and applying the aforementioned inequalities in the derivation of
, we get @xmath115 from we deduce that for all @xmath99 , there exists positive parameters @xmath116 and @xmath117 such that @xmath118 the exponential stability of the target system - can be straightforwardly established following the steps in deriving the proof of exponential stability of - , which completes the proof of theorem [ robust - thm ] .
generally , the implementation of a full - state feedback controller is rarely achievable due to the unavailability of full state measurements .
the computation of the controller requires a full measurement of the distributed temperature profile @xmath11 along the domain @xmath119 $ ] and the moving interface position @xmath17 which relatively limits its practical relevance . here
, we extended the full - state feedback results considering moving interface position @xmath17 as the only available measurement and deriving an estimator of the temperature profile based on this only available measurement @xmath120 . denoting the estimates of the temperature @xmath121 ,
the following theorem holds : [ observer ] consider the plant with the measurement @xmath120 and the following observer @xmath122 where @xmath123 $ ] , and the observer gain @xmath124 is @xmath125 with a gain parameter @xmath126 .
assume that the two physical constraints and are satisfied .
then , for all @xmath127 , the observer error system is exponentially stable in the sense of the norm @xmath128 is proved later in section [ sec : estimation ] . for the reference error system , namely , the @xmath129-system , we consider the following observer : @xmath130 where @xmath124 is the observer gain to be determined .
defining estimation error variable of @xmath129-system as @xmath131 combining with , the @xmath132-system is written as @xmath133 as for the full state feedback case , the following backstepping transformation for moving boundary pdes , whose kernel function is the observer gains @xmath134 is constructed to transform the following exponentially stable target system @xmath135 into @xmath132-system defined in - . taking the derivative of with respect to @xmath38 and @xmath62 along the solution of - respectively , the solution of the gain kernel and the observer gain are given by @xmath136 where @xmath137 is a modified bessel function of the first kind . also , one can derive the inverse transformation that maps the @xmath138-system - into the @xmath132-system - as @xmath139 where @xmath140 is a bessel function of the first kind . to show the stability of the target @xmath138-system in -
, we consider a functional @xmath141 taking time derivative of along the solution of the target system - , we obtain @xmath142 using and applying pointcare s inequality , a differential inequality in @xmath143 is derived as @xmath144 hence , the origin of the target @xmath138-system is exponentially stable .
since the transformation is invertible as in , the exponential stability of @xmath138-system at the origin induces the exponential stability of @xmath132-system at the origin with the help of , which completes the proof of theorem [ observer ] .
an output feedback control law is constructed using the reconstruction of the estimated temperature profile through the exponentially convergent observer - and the measured interface position @xmath120 as shown in fig .
[ statediag2 ] and the following theorem holds : [ outputthm ] consider the closed - loop system with the measurement @xmath120 and the observer - under the output feedback control law @xmath145 assume that the initial values @xmath146 are compatible with the control law and the initial plant states @xmath147 satisfy .
additionally , assume that the upper bound of the initial temperature is known , i.e. the lipschitz constant @xmath148 in is known
. then , for any initial temperature estimation @xmath149 , any gain parameter of the observer @xmath150 , and any setpoint @xmath35 satisfying @xmath151 respectively , where the parameters @xmath152 and @xmath153 satisfy @xmath154 , the closed - loop system is exponentially stable in the sense of the norm @xmath155 the proof of theorem [ outputthm ] is derived by * introducing an appropriate backstepping transformation together with a suitable target system , * verifying the constraints and , * and establishing the lyapunov stability proof . by equivalence , the transformation of the variables @xmath156 into @xmath157 leads to the gain kernel functions defined by the state - feedback backstepping transformation given by @xmath158 with an associated target system given by @xmath159 where @xmath160 . evaluating the spatial derivative of at @xmath16
, we derive the output feedback controller as @xmath161 by the same procedure as in appendix [ appa ] , an inverse transformation is given by @xmath162 in this section , we derive sufficient conditions to guarantee the two physical constraints and .
first , we state the following lemma .
[ wmaximum ] suppose that @xmath163 .
then , the solution of - satisfies @xmath164 , @xmath165 , @xmath166 . the proof of lemma [ wmaximum ] is provided by maximum principle as in @xcite .
next , we state the following lemma .
[ maximum ] for any initial temperature estimate @xmath149 and , any observer gain parameter @xmath150 satisfying and , respectively , the following properties hold : @xmath167 lemma [ wmaximum ] states that if @xmath168 , then @xmath164 .
in addition , by the direct transformation , @xmath164 leads to @xmath169 due to the positivity of the solution to the gain kernel .
therefore , with the help of , we deduce that @xmath169 if @xmath170 satisfies @xmath171 considering the bound of the solution under the condition of , the sufficient condition for to hold is given by which restricts the gain parameter @xmath150 .
thus , we have shown that conditions and lead to @xmath169 for all @xmath172 .
in addition , from the boundary condition and hopf s lemma , it follows that @xmath173 . the final step is to prove that the output feedback closed - loop system satisfies the physical constraints .
[ proposition ] suppose the initial values @xmath174 satisfy and the setpoint @xmath35 is chosen to satisfy .
then , the physical constraints and are satisfied by the closed - loop system consisting of the plant , the observer - and the output feedback control law . taking the time derivative of along with the solution , with the help of the observer gain
, we derive the following differential equation : @xmath175 from the positivity of the solution and the neumann boundary value , the following differential inequality holds @xmath176 hence , if the initial values satisfy @xmath71 , equivalently is satisfied , from and , we get @xmath177 then , with the relation given in lemma [ maximum ] and the positivity of @xmath178 provided by lemma [ lemma1 ] , the following inequality is established : @xmath179 finally , substituting the inequalities and into , we arrive at @xmath180 , @xmath166 , which guarantees that the second physical constraint is satisfied .
we consider a functional @xmath181 where @xmath111 is chosen to be large enough and @xmath112 is appropriately selected . taking time derivative of along the solution of target system - , and applying young s
, cauchy - schwarz , poincare s , agmon s inequalities , with the help of and , the following holds : @xmath182 where , @xmath183 , @xmath184 .
hence , the origin of the @xmath185-system is exponentially stable .
since the transformation and are invertible as described in and , the exponential stability of @xmath185-system at the origin guarantees the exponential stability of @xmath186-system at the origin , which completes the proof of theorem [ outputthm ] .
in part i , the neumann boundary actuation of the manipulated heat flux has been considered to design state and output feedback controllers for the one - dimensional stefan problem . however , some actuators require dirichlet boundary control design , such as a thermo - electric cooler actuation controlling a boundary temperature @xcite . in the following section ,
the boundary temperature is used for the control design .
we define the control problem consisting of the same diffusion equation , dirichlet boundary condition , and initial conditions : @xmath187 where @xmath188 is a controlled temperature .
analogously to lemma [ lemma1 ] , the following lemma is stated .
[ lemma3 ] for any @xmath189 on the finite time interval @xmath31 , @xmath190 and @xmath191 .
and then @xmath192 , @xmath191 .
similarly to lemma [ lemma1 ] , the proof of lemma [ lemma3 ] is based on maximum principle and hopf s lemma @xcite .
therefore , the following physical constraint holds @xmath193 for boundary temperature control , the energy - like conservation law is described as @xmath194 considering the same control objective as in section [ open ] , taking the limit of from @xmath39 to @xmath40 yields @xmath195 hence , by imposing the physical constraint , the following assumption on the setpoint is required .
[ tempsetpoint ] the setpoint @xmath35 is chosen to satisfy @xmath196 again , assumption [ tempsetpoint ] stands as a least restrictive condition for the choice of setpoint , and the open - loop stabilization is presented in the following lemma .
[ rec ] consider an open - loop setpoint control law @xmath197 which satisfies .
then , the control objectives and are satsified .
hence , the similar rectangular pulse of energy shaping for temperature control as in section [ sec : pulse ] can be derived as @xmath198\\ 0 & { \rm for}\quad t >
k\delta e/\bar{q } \end{array } \right \},\end{aligned}\ ] ] where @xmath199 firstly , we suppose that the physical parameters are known and state the following theorem .
[ thm3 ] consider a closed - loop system consisting of the plant and the control law @xmath200 where @xmath55 is the controller gain .
assume that the initial conditions @xmath56 are compatible with the control law and satisfies .
then , for any reference setpoint @xmath201 and control gain @xmath202 which satisfy @xmath203 respectively , the closed - loop system is exponentially stable in the sense of the norm .
the backstepping transformation leads to the following target system @xmath204 and the control law .
next , we show that the physical constraints and are insured if holds .
taking a time derivative of , we have @xmath205 assume that @xmath206 such that @xmath189 for @xmath207 and @xmath208 .
then , by maximum principle , we get @xmath209 and @xmath210 for @xmath207 . hence , @xmath211 .
from , the following equality is deduced @xmath212 thus , @xmath213 for @xmath207 and verifies a differential inequality @xmath214 comparison principle and yield @xmath215 in contradiction to @xmath208 .
therefore , @xmath216 such that @xmath189 for @xmath207 and @xmath208 , which implies @xmath189 for @xmath166 assuming .
finally , we consider a functional @xmath217 with an appropriate choice of positive parameters @xmath111 and @xmath112 , time derivative of yields @xmath218 therefore , choosing the controller gain to satisfy , it can be verified that there exists positive parameters @xmath117 and @xmath116 such that @xmath219 similarly , in the neumann boundary actuation case , under the physical constraint , the exponential stability of the target system - can be established , which completes the proof of theorem [ thm3 ] .
next , we investigate robustness of the boundary temperature controller to perturbations on the plant s physical parameters @xmath220 and @xmath91 . again , the perturbed system is described as @xmath221 where @xmath93 and @xmath94 are perturbation parameters such that @xmath95 and @xmath96 .
[ robust - temp ] consider a closed - loop system , and the control law .
then , for any perturbations @xmath222 which satisfy @xmath223 , there exists @xmath224 such that for @xmath225 the closed - loop system is exponentially stable in the norm .
note that the transformation and the system , - are identical to the ones considered in section [ sec : robust ] .
moreover , only the boundary condition of the target system at @xmath16 is modified as @xmath226 , in order to match the temperature control problem .
condition and need to be satisfied under the parameter perturbations . taking time derivative of along the system , - , with the boundary condition
, we obtain @xmath227 the inequality @xmath228 enables to state the positivity of the controller @xmath229 .
hence , the physical constraints and are verified .
finally , we introduce the following functional @xmath230 with an appropriate choice of @xmath111 and @xmath112 , we have @xmath231 where @xmath232 , @xmath233 . let @xmath224 be a positive root of @xmath234 .
then , for @xmath225 , there exists positive parameters @xmath116 and @xmath117 which verifies @xmath118 hence , it concludes theorem [ robust - temp ] . with respect to the boundary temperature control introduced in section [ sec : tempcontrol ] instead of the heat control ,
the observer design is replaced by the following .
[ obsvtemp ] consider the following closed - loop system of the observer @xmath235 where @xmath123 $ ] , and the observer gain @xmath236 is @xmath237 with an observer gain @xmath126 .
assume that the two physical constraints and are satisfied .
then , for all @xmath127 , the observer error system is exponentially stable in the sense of the norm @xmath238 is established by the same procedure as in section [ obsvtarget ] . for boundary temperature controller ,
the output feedback controller is designed using the state observer presented in corollary [ obsvtemp ] .
the following corollary is stated for a controlled boundary temperature .
[ outputthmtemp ] consider the closed - loop system consisting of the plant , the measurement @xmath120 , the observer - , and the output feedback control law @xmath239 with @xmath202 , @xmath149 , @xmath150 satisfying , , and , respectively , and , the setpoint @xmath35 satisfying @xmath240 the closed - loop system is exponentially stable in the sense of the norm .
the proof of corollary [ outputthmtemp ] can be established following the steps of the proof corollary [ obsvtemp ] detailed in section [ sec : output ] .
simulation results in the case of the neumann boundary actuation developed in part i are performed considering a strip of zinc as in @xcite whose physical properties are given in table 1 ( the consistent simulations to illustrate the feasibility of the backstepping controller with dirichlet boundary actuation are easily achievable but due to space limitation , these are not provided ) .
here , we use the well known boundary immobilization method combined with finite difference semi - discretization . the setpoint and the initial values
are chosen as @xmath35 = 0.35 m , @xmath241 = 0.01 m , and @xmath242 with @xmath148 = 10000 k@xmath243 . then , the setpint restriction is satisfied .
[ fig : robust1 ] shows the responses of the plant with the open - loop pulse input ( dashed line ) and the backstepping control law ( solid line ) .
the time window of the open - loop pulse input is set to be 50 min .
the gain of the backstepping control law is chosen as @xmath202=0.001 to be small enough to avoid numerical instabilities .
[ fig : robust1 ] ( a ) shows the response of @xmath17 without the parameters perturbations , i.e. @xmath244 and clearly demonstrates that @xmath17 converges to @xmath35 applying both rectangular pulse input and backstepping control law , and the convergence speed with the backstepping control is faster .
moreover , from the dynamics of @xmath17 under parameters perturbations @xmath245 shown in fig .
[ fig : robust1 ] ( b ) , it can be seen that the convergence of @xmath17 to @xmath35 is successfully achieved with the backstepping control law but not obtained with the pulse input . on both fig .
[ fig : robust1 ] ( a ) and ( b ) , the responses with the backstepping control law show that the interface position converge faster without the overshoot beyond the setpoint , i.e. @xmath210 and @xmath246 for @xmath166 .
the dynamics of the controller @xmath51 and the temperature at the initial interface @xmath247 with the backstepping control law are described in fig .
[ fig : robust2 ] ( a ) and ( b ) , respectively , for the system without parameter s uncertainties , i.e. , @xmath244 ( red ) and the system with parameters mismatch @xmath245 ( blue ) . as presented in fig .
[ fig : robust2 ] ( a ) , the boundary heat controller @xmath51 remains positive , i.e. @xmath30 in both cases . moreover , fig .
[ fig : robust2 ] ( b ) shows that @xmath247 converges to @xmath18 with @xmath248 for both the system with accurate parameters and the system with uncertainties on the parameters . physically , fig .
[ fig : robust2 ] ( b ) means that the temperature at the initial interface warms up from the melting temperature @xmath18 , to enable melting the solid - phase to the setpoint @xmath35 before settling back to @xmath18 .
these physical constraints and phenomena holds even with the parameter uncertainty as long as is satisfied .
therefore , the numerical results are consistent with our theoretical result .
.physical properties of zinc [ cols="<,<,<",options="header " , ] the initial estimation of the temperature profile is set to @xmath249 with @xmath250 = 1000 k@xmath243 , and the observer gain is chosen as @xmath150 = 0.001 .
then , the restriction on @xmath149 , @xmath150 , and @xmath35 described in - are satisfied , which are conditions for theorem [ observer ] and theorem [ outputthm ] to remain valid .
the dynamics of the moving interface @xmath17 , the output feedback controller @xmath51 , and the temperature at the initial interface @xmath247 are depicted in fig .
[ fig : simulation ] ( a)(c ) , respectively . fig .
[ fig : simulation ] ( a ) shows that the interface @xmath17 converges to the setpoint @xmath35 without overshoot which is guaranteed in proposition [ proposition ] .
[ fig : simulation ] ( b ) shows that the output feedback controller remains positive , which is a physical constraint for the model to be valid as stated in lemma [ lemma1 ] and ensured in proposition [ proposition ] .
the model validity can be seen in fig .
[ fig : simulation ] ( c ) which illustrates @xmath247 increases from the melting temperature @xmath18 to enable melting of material and settles back to its equilibrium . the positivity of the backstepping controller shown in fig .
[ fig : simulation ] ( b ) , results from the negativity of the distributed estimation error @xmath251 as shown in lemma [ maximum ] and proposition [ proposition ] . fig . [
fig : simulation ] ( d ) shows the dynamics of estimation errors of distributed temperature at @xmath16 ( red ) , @xmath252 ( blue ) , and @xmath253 ( green ) , respectively .
it is remarkable that the estimation errors at each point converge to zero and remains negative confirming the theoretical results stated in lemma [ maximum ] and proposition [ proposition ] .
[ 6a ] + +
this paper presented a control and estimation design for the one - phase stefan problem via backstepping method .
the system is a diffusion pde on a moving interface domain which has the dynamics described by ode .
the novelties of this paper are summarized as followings . 1 .
we developed a new approach for the global stabilization of nonlinear parabolic pde via a nonlinear backstepping transformation .
the closed - loop responses guarantee the physical constraints imposed by the validity of the model .
3 . a new formulation of the lyapunov function for moving boundary pde was applied and it showed the exponential stability of the closed loop system .
we emphasize that the proposed controller does not require the restriction imposed in @xcite regarding the material properties . in that sense
, our design can be applied to a wider class of melting materials .
also , even if the controller is same as the one proposed in @xcite , we ensure the exponential stability of the sum of interface error and @xmath0-norm of temperature error which is stronger than the asymptotical stability result presented in @xcite .
recall that the reference error system defined in section [ state - feedback ] is @xmath254 suppose that the backstepping transformation for moving boundary pde is formulated as @xmath255 which transforms the reference error system - into the following target system @xmath256 the stability of the target system - is guaranteed with the assumptions such that @xmath210 , @xmath246 as shown in appendix [ stabilitytarget ] . then , taking first and second spacial derivatives of and the first time derivative of , we obtain the following relation @xmath257 evaluating and its spacial derivative at @xmath22 , we have @xmath258 in order to hold - for any continuous functions @xmath259 , by - , @xmath260 and @xmath261 satisfy @xmath262 the solutions to the gain kernels are given by @xmath263 suppose that the inverse transformation is formulated as @xmath264 where @xmath265 , @xmath266 are the kernel functions . taking derivative of with respect to @xmath38 and @xmath62 , respectively , along the solution of - ,
the following relations are derived @xmath267 @xmath268 in order to hold - for any continuous functions @xmath269 , by - , @xmath266 and @xmath265 satisfy @xmath270 using and , the solution of @xmath266 is given by @xmath271 the conditions - yields @xmath272 then , the solutions and satisfy the condition as well .
in this section , we prove the exponential stability of @xmath66 system defined in - via lyapunov analysis , which induces the stability of the original @xmath65 system .
the following assumptions on the interface dynamics @xmath273 which are shown in section [ positivness ] are stated .
firstly , we show the exponential stability of the target system - .
consider the lyapunov function @xmath87 such that @xmath274 with a positive number @xmath76 to be chosen later .
then , by taking the time derivative of along the solution of the target system - , we have @xmath275 where @xmath276 .
the boundary condition yields @xmath277 by chain rule @xmath278 . substituting into ,
we get @xmath279 define @xmath280 . from the assumption , young s , causchy schwartz , pointcare s , agmon s inequality , and choosing @xmath281
, we have @xmath282 where @xmath283 the norm equivalence between the target system and original system is shown from the direct transformation and the inverse transformation . taking the square of the inverse transformation and applying young s and cauchy schwartz inequality
, we have @xmath285 integrating from @xmath39 to @xmath17 and applying cauchy schwartz inequality , we have @xmath286 taking the spatial derivative of and by the same manner , we have @xmath287 in this case , we have @xmath288 .
thus , all of the inequality are written as @xmath289 where @xmath290 , @xmath291 , @xmath292 , @xmath293 , @xmath294 , @xmath295 , @xmath296 , @xmath297 .
adding to and to , we derive the following inequality @xmath298 where @xmath299 , @xmath300 .
define a parameter @xmath301 as @xmath302 .
then , with the help of , , we deduce that there exists @xmath301 and @xmath303 such that @xmath304 acknowledgment -------------- n. petit , control problems for one - dimensional fluids and reactive fluids with moving interfaces , " in _ advances in the theory of control , signals and systems with physical modeling _ , volume 407 of _ lecture notes in control and information sciences _ , pages 323337 , lausanne , dec 2010 .
|
this paper develops a control and estimation design for the one - phase stefan problem .
the stefan problem represents a liquid - solid phase transition as time evolution of a temperature profile in a liquid - solid material and its moving interface .
this physical process is mathematically formulated as a diffusion partial differential equation ( pde ) evolving on a time - varying spatial domain described by an ordinary differential equation ( ode ) .
the state - dependency of the moving interface makes the coupled pde - ode system a nonlinear and challenging problem .
we propose a full - state feedback control law , an observer design , and the associated output - feedback control law via the backstepping method .
the designed observer allows estimation of the temperature profile based on the available measurement of solid phase length .
the associated output - feedback controller ensures the global exponential stability of the estimation errors , the @xmath0-norm of the distributed temperature , and the moving interface to the desired setpoint under some explicitly given restrictions on the setpoint and observer gain .
the exponential stability results are established considering neumann and dirichlet boundary actuations .
stefan problem , backstepping , distributed parameter systems , moving boundary , nonlinear stabilization .
|
statistics of binaries and hierarchical stellar systems trace conditions of star formation and serve as an excellent diagnostic for testing theoretical predictions and numerical simulations .
only recently have observational techniques such as adaptive optics ( ao ) and radial velocities ( rv ) reached the maturity and productivity needed for a high quality binary census over a wide range of periods .
the latest results on binary statistics are reviewed by @xcite .
reaching completeness for triple and higher - order hierarchies is even more challenging .
the fraction of known hierarchical systems in the nearby 25-pc volume has doubled recently @xcite in comparison with the earlier work by @xcite .
large space volume and large samples are needed for their statistical study as the fraction of triples is only 14% @xcite . however , observing thousands of targets by complementary techniques would require prohibitive amounts of telescope and calendar time unless an intelligent strategy is applied . as noted by @xcite ,
new observations mostly convert binaries into triples .
so , focusing on known binaries and quantifying the presence of additional components ( hence hierarchies ) is a productive approach .
it was adopted in the massive multiplicity survey of solar - type stars conducted in 20122013 with the robo - ao instrument at the palomar 60-inch telescope @xcite ; hereafter r15 .
that survey looked for subsystems in the faint secondary components and for distant tertiary companions to known binaries with fg primaries within 67 pc .
the full sample is described in @xcite .
we report here the beginings of follow - up observations to the r15 survey at the palomar 5 m telescope .
our goal is to both confirm the newly discovered subsystems with infrared ( ir ) imagery and to combine that ir imagery with the robo - ao observations in @xmath3 band .
multi - color photometry allows us to place the secondaries on a color - magnitude diagram ( cmd ) to test their physical relation to the main target . for systems with high proper motion ( pm )
, the second - epoch astrometry within a year allows discrimination between physical binaries and unrelated optical companions .
this is most critical for wide binaries because the probability of interlopers is proportional to the square of the separation .
however , many new pairs with separations on the order of 5 also need confirmation , particularly near the galactic plane where the density of stars is high . at this time
, we have only been able to collect data on a portion of the companions detected in r15 ; future observations will cover the other systems .
quite unexpectedly , observations with palm-3000 resolved several subsystems missed by the robo - ao instrument in r15 because their companions were either too close or too faint .
the old rule about new techniques bringing new discoveries is once again verified .
the observations and data reduction are covered in section [ sec : obs ] . in section [ sec : ptm ] we derive individual magnitudes and colors of the components and place them on the cmd .
each multiple system is commented on in section [ sec : ind ] .
discussion and conclusions are in section [ sec : disc ] .
we observed the stars on 2013 september 28 ut with the palomar observatory hale 5 m telescope using the palm-3000 ao system and the pharo near - ir camera .
the palm-3000 ao system is a natural guide - star system using two deformable mirrors @xcite .
one corrects low - amplitude high spatial frequency aberrations , while the other corrects the higher - amplitude low spatial frequency aberrations .
it is optimized for high contrast observations and routinely produces strehl ratios greater than 80% in _ ks _ band .
the pharo camera uses a hgcdte hawaii detector for observations between 1 and 2.5 wavelength @xcite .
the camera has multiple filters in two filter wheels .
the filter wheels contain spectral band filters and neutral density filters .
these neutral density filters are used for observing bright stars , which saturated the detector .
unfortunately , both the 1% and the 0.1% nd filter cause ghost reflections which appear as stellar companions .
these ghosts always appear in the same locations and during analysis they are considered to be part of the point spread function .
fifty frames were collected of each object with an exposure time of 1.416 s , the minimum exposure for pharo .
after the observing run , the individual frames were reduced by debiasing , flat fielding , bad pixel correction and background subtraction .
then we created five images by co - adding 10 frames into each image . creating multiple images allowed us to evaluate the precision of the measurements .
the _ fitstars _
algorithm was used to measure the astrometry and photometry of the objects @xcite .
we observed six calibration binaries on the same night as the science targets .
we compared their measured astrometry with the ephemeris predicted from their orbits and used the results to compute the plate scale and the position angle offset , [email protected] mas pixel@xmath5 and [email protected]@xmath2 error .
the largest error term is the position angle offset . in speckle and ao work on binaries , the measurement errors in tangential and radial directions are usually the same or similar . for the position angle error bar , we quadratically added the position angle offset error with the product of one radian and the ratio of the measurement error to the measured separation .
the separation error bar was computed simarily , with the measurement error summed quadratically with the product of the plate scale error and the separation .
photometric error bars were assigned using the technique described in @xcite .
the resulting photometry and astrometry are presented in table [ results ] .
the table gives the washington double star ( wds ; @xcite ) catalog designation for the system as are the hd and hip numbers for the primary star .
this is followed by the discoverer designation , the observation epoch , the astrometry and the photometry .
new discoveries are marked with a footnote to the wds ddesignation .
lrrlccccc 00487 + 1841 & 4655 & 3795 & rao 5 ca , cb & 2013.8371 & [email protected] & [email protected] & ... & [email protected] + 01027 + 0908 & 6152 & 4878 & rao 39 ab & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 01075 + 4116 & 6611 & 5276 & rao 40 & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 02462 + 0536 & 17250 & 12925 & rao 9 & 2013.8370 & [email protected] & [email protected] & [email protected] & [email protected] + 03390 + 4232 & 22521 & 17022 & rao 47 & 2013.8370 & [email protected] & [email protected] & [email protected] & [email protected] + 03401 + 3407 & 22692 & 17129 & rbr 26 aa , ab & 2013.8370 & [email protected] & [email protected] & [email protected] & [email protected] + 03401 + 3407 & 22692 & 17129 & stf 425 ab & 2013.8370 & [email protected] & [email protected] & [email protected] & [email protected] + 03413 + 4554 & 22743 & 17217 & rao 48ac & 2013.8370 & [email protected] & [email protected] & ... & [email protected]
+ 04363 + 5502 & 28907 & 21443 & rao 35 & 2013.8370 & [email protected] & [email protected] & [email protected] & [email protected] + 05247 + 6323 & 34839 & 25300 & stf 677 a , ba & 2013.8370 & [email protected] & [email protected] & ... & [email protected] + 05247 + 6323 & 34839 & 25300 & rbr 27 ba , bb & 2013.8370 & [email protected] & [email protected] & ... & [email protected] + 05247 + 6323 & 34839 & 25300 & rao 36 ac & 2013.8370 & [email protected] & [email protected] & ... & [email protected] + 06335 + 4822 & 46013 & 31267 & rao 80 ac & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 06562 + 4032 & 50720 & 33355 & rbr 28 aa , ab & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 06562 + 4032 & 50720 & 33355 & rao 56 ac & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 17422 + 3804 & 161163 & 86642 & rbr 29 aa , ab & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 17422 + 3804 & 161163 & 86642 & rao 20 ab & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 19359 + 5659 & 185414 & 96395 & rao 87 & 2013.8371 & [email protected] & [email protected] & [email protected] & [email protected] + 20312 + 5653 & 195872 & 101234 & rao 22 & 2013.8367 & [email protected] & [email protected] & [email protected] & [email protected] + 20333 + 3323 & 195992 & 101430 & hj 1535 a , ba & 2013.8368 & [email protected] & [email protected] & [email protected] & [email protected] + 20333 + 3323 & 195992 & 101430 & rbr 29 ba , bb & 2013.8368 & [email protected] & [email protected] & [email protected] & [email protected] + 20333 + 3323 & 195992 & 101430 & rao 71 ae & 2013.8368 & [email protected] & [email protected] & [email protected] & [email protected] + 20577 + 2624 & 199598 & 103455 & rao 24 & 2013.8368 & [email protected] & [email protected] & [email protected] & [email protected] + 21102 + 2045 & 201639 & 104514 & rao 25 & 2013.8368 & [email protected] & [email protected] & [email protected] & [email protected] + 21585 + 0347 & 208776 & 108473 & rao 73 & 2013.8368 & [email protected] & [email protected] & ... & [email protected] + 23588 + 3156 & 224531 & 118213 & rbr 30 aa , ab & 2013.8369 & [email protected] & [email protected] & [email protected] & [email protected] + 23588 + 3156 & 224531 & 118213 & rao 76 & 2013.8369 & [email protected] & [email protected] & [email protected] & [email protected] + 23588 + 3345 & 224543 & 118225 & rao 77 & 2013.8372 & [email protected] & [email protected] & [email protected] & [email protected] +
using the differential photometry in the @xmath0 band from table [ results ] and the @xmath3 band data from r15 , we calculated the individual magnitudes of stellar components . the combined @xmath6 and @xmath0 magnitudes of close binaries were taken from the 2mass
. however , the sdss @xmath3 magnitudes @xcite are not available for all our targets .
we interpolated them from the @xmath6 and @xmath7 magnitudes using the colorindex @xmath8 as argument .
taking as a template the 1 gyr isochrone from the dartmouth stellar models @xcite , we approximated the color - color relations for dwarfs less massive than 1.5@xmath9 by quadratic polynomials @xmath10 the two estimates of @xmath3 for our targets derived from @xmath7 and @xmath6 agree very well , with an rms difference of only 0.026mag .
we then averaged these estimates to get the combined @xmath3 magnitudes .
= @xmath110.5 , respectively @xcite .
the green line connects the ca and cb components of hip 3795
. some primary components undoubtably contain unresolved binaries .
the hipparcos names and binary star designations are listed for some of the points . see section [ sec : ind ] for discussion of the individual systems.,height=226 ] figure [ fig : cmd ] places primary and secondary components of multiple systems resolved both here and in r15 in the @xmath12 cmd .
the _ hipparcos
_ parallaxes of the primary targets @xcite are used .
the lines are the 1 gyr isochrones for solar metallicity and [ fe / h][email protected] from the dartmouth stellar models @xcite retrieved from the web interface .
the errors of the color indices of secondary components are shown .
they are computed by assuming realistic errors of the differential @xmath3-band photometry in r15 , typically 01 , and combining them with the errors reported in table 1 .
note that all primary components ( as well as other f- and g - type stars within 67pc ) are located above the main sequence , indicating potential bias in the dartmouth models .
some primaries are in fact close binaries .
stellar systems with three or more components require individual analysis . we provide below comments on each object , identified by the _ hipparcos _ number of the primary component . the location of faint components on the cmd is used jointly with the astrometric information and field crowding to evaluate whether the binary is physical or optical ( chance alignment ) .
for some new pairs we compare in table [ motion ] the differential motion of the binary with the proper motion ( pm ) of the main target @xcite .
differential astrometry from table [ results ] is combined with the first epoch taken either from r15 ( assuming positional errors of half a pixel or 22mas when they are not listed ) or from 2mass ( accuracy of 01 is assumed ) .
columns ( 1 ) and ( 2 ) identify the pair by its _ hipparcos _ number and components .
the time difference @xmath13 in column ( 3 ) exceeds 10 years when the first epoch is taken from 2mass . columns ( 4 ) and ( 5 ) give the computed relative motion in right ascension and declination , while columns ( 6 ) and ( 7 ) list the pm of the main target .
column ( 8) gives the number of 2mass stars expected within the binary separation .
these estimates do not take into account the brightness of the companion and the magnitude distribution of background sources , and are just a crude indication of the odds that a binary is optical .
the last column ( 9 ) summarizes the status of the companion ( o for optical , p for physical , p ?
for likely physical and ?
for status unknown ) .
background stars usually have a small pm , and in such cases the differential pm is almost opposite to pm(a ) .
on the other hand , a detectable relative motion can be orbital or can be caused by motions in inner subsystems and does not necessarily mean that the binary is optical .
rcr ccr rcl 3795 & ca , cb & 1.2 & @xmath114@xmath420 & @xmath1118@xmath420 & @xmath1113 & @xmath1149 & 0.000 & p + 4878 & ab & 1.2 & @xmath1132@xmath425 & @xmath11101@xmath423 & 66 & @xmath1123 & 0.004 & p ?
+ 5276 & ab & 1.2 & 56@xmath462 & 17@xmath436 & 118 & @xmath1155 & 0.047 & p + 12925 & ad & 1.2 & 5@xmath49 & 28@xmath416 & 73 & @xmath1143 & 0.001 & p + 17022 & ab & 1.1 & 12@xmath424 & @xmath1111@xmath422 & @xmath11194 & @xmath11132 & 0.008 & p ? + 17217 & ac & 1.1 & 48@xmath432 & 21@xmath448 & @xmath1149 & 2 & 0.059 & p
? + 21443 & ab & 1.1 & 29@xmath467 & @xmath1133@xmath452 & @xmath115 & 20 & 0.155 & p ? + 25300 & ac & 14.8 & 15@xmath47 & @xmath1120@xmath47 & @xmath11127 & @xmath1158 & 0.087 & p + 31267 & ac & 0.2 & 201@xmath4182 & 766@xmath4175 & 44 & 31 & 0.032 & ? + 33355 & ac & 0.8 & @xmath11125@xmath464 & @xmath11149@xmath438 & 3 & @xmath1137 & 0.048 & ? + 86642 & ab & 1.1 & @xmath1111@xmath423 & 7@xmath426 & @xmath11109 & 94 & 0.007 & p + 96395 & ab & 14.8 & @xmath116@xmath47 & 22@xmath49 & 0 & @xmath11200 & 0.197 & p + 101234 & ab & 1.1 & 88@xmath421 & 4@xmath422 & @xmath11155 & @xmath11142 & 0.000 & p ? + 101430 & ab & 13.8 & 8@xmath48 & 6@xmath411 & 154 & 140 & 2.610 & p + 101430 & ae & 13.8 & @xmath11167@xmath410 & @xmath11140@xmath410 & 154 & 140 & 1.321 & o + 103455 & ab & 1.1 & 50@xmath49 & 17@xmath410 & 273 & 95 & 0.002 & p + 104514 & ab & 1.1 & @xmath1133@xmath429 & 43@xmath424 & @xmath1191 & 67 & 0.021 & p + 108473 & ab & 13.8 & 0@xmath47 & @xmath114@xmath412 & @xmath11248 & @xmath11133 & 0.122 & p + 118213 & ab & 1.2 & @xmath114@xmath420 & @xmath1151@xmath441 & 76 & @xmath11134 & 0.014 & p + 118225 & ab & 1.2 & @xmath1151@xmath449 & @xmath1113@xmath426 & 272 & @xmath11130 & 0.022 & p for some of the systems , we estimated the orbital period , @xmath14 , of the binary from the measured separations , @xmath15 , distances , and mass sum @xmath16 using kepler s third law , @xmath17 .
the median ratio between the projected separation , @xmath18 @xmath19 , and orbital semi - major axis , @xmath20 , is close to one @xcite , with scatter by a factor of two caused by orbital phase , orbit orientation , and eccentricity .
the strict lower limit is @xmath21 .
statistical period estimates using the assumption @xmath22 are denoted as @xmath23 .
_ hip 3795 ( hd 4655 = wds 00487 + 1841 ) _ is a quadruple system .
the inner ab pair , bu 495 , has an orbit with a [email protected] yr period @xcite .
the cpm component c at 152from ab was found by @xcite , who estimate the probability of it being physical as 0.99 .
the component c was resolved by r15 into a 067 pair ca , cb .
the pm is too small to determine if the ca and cb share common proper motion , but the low density of background stars and the small separation imply that the pair ca , cb is physical . _
hip 4878 ( hd 6152 = wds 01027 + 0908 ) _ is a double - lined spectroscopic binary with @xmath24d @xcite .
the tertiary companion b at 28 did not move substantially in a year since its discovery by r15 , but the pm of the main component a is small and the astrometry is not conclusive .
b is located below the main sequence ( ms ) on the cmd , although still within the errors of its photometry .
considering the small field crowding , we count b as likely physical . _
hip 5276 ( hd 6611 = wds 01075 + 4116 ) _ is a triple system with the inner 74.1 d spectroscopic pair @xcite and the tertiary from r15 confirmed as physical by our photometry and astrometry .
our observations demonstrate that the photometry of b given in 2mass is biased by the proximity of the bright primary , as in many other similar pairs .
their `` unusual '' colors derived from the 2mass photometry are thus wrong .
_ hip 12925 ( hd 17250 = wds 02462 + 0536 ) _ is a quadruple system with a 3-tier hierarchy : a close spectroscopic pair with yet unknown period , the robo - ao companion d at 189 , and the cpm companion c at 494 @xcite .
yet another star , hip 12862 , at 09 , also shares the common pm and parallax .
the stars are young and belong to the tucana - horlogium moving group @xcite .
our photometry places d slightly above the ms .
the companion d was independently discovered by @xcite by high - contrast imaging at subaru .
they measured it on 2012.0 at 2529 and 1893 , in excellent agreement with r15 .
_ hip 17022 ( hd 22521 = wds 03390 + 4232 ) _ is a triple system consisting of an astrometric binary with period of 3yr @xcite and the robo - ao tertiary b at 176 . considering the fast pm(a ) , the fixed position of ab during 1 year would appear to confirm it as a physical binary .
b is located well below the ms , although its photometry has a large uncertainty . if it is a physical component with unusual colors , the system merits further study . _
hip 17129 ( hd 22692 = wds 03401 + 3407 ) _ is a known binary , stf 425 , which shows apparently a non - keplerian motion .
@xcite proposed that the motion was caused by an unseen companion and calculated an astrometric orbit for it with @xmath25yr , eccentricity 0.61 , and photocenter semi - major axis 0179 .
our observations were able to resolve the predicted subsystem aa , ab . for the time of our observations the orbit predicts a companion position angle of 433 , in excellent agreement with the position angle of [email protected]@xmath2 measured here .
the ratio of the predicted displacement to the measured separation of 050 is @xmath26 .
hence , the mass ratio @xmath27 ( the contribution of the light from ab to the photo - center is neglected ) .
the mass of aa is estimated at 1.16@xmath28 from its absolute magnitude , leading to 0.54@xmath9 for ab .
the absolute magnitude of ab corresponds to a smaller mass of 0.33@xmath9 .
this discrepancy , if confirmed , can be explained by ab being a close pair of m - dwarfs , as happens in other known multiple systems , e.g. in @xmath29 for @xcite .
see figure [ new_binaries]a for the discovery image . [
cols="^,^,^ " , ] _ hip 17217 ( hd 22743 = wds 03413 + 4554 ) _ is a known triple system . the inner binary has been known for over a century @xcite and its first orbit was computed by r15 .
r15 also detected a third component c which was confirmed by examination of 2mass images .
the close binary is detectable in our @xmath0 images , but the psfs of the stars are overlapping and we are unable to make a consistent measurement with @xmath30 .
we are able to extract the astrometry of the ac pair and consider it as likely physical .
_ hip 21443 ( hd 28907 = wds 04363 + 5502 )
_ is triple , consisting of a 2.6-d spectroscopic binary @xcite and the robo - ao and 2mass companion at 57 .
the field is crowded and pm(a ) is small .
our photometry places b slightly above the ms .
the star is young and active according to @xcite and we retain b as likely physical . _
hip 25300 ( hd 34839 = wds 05247 + 6323 ) _ is now a 3-tier quadruple system since we resolved the secondary component of the binary ab = stf 677 into a new 018 pair ba , bb and confirmed that the robo - ao tertiary c at 7 is physical .
we find c to be slightly below the ms , within errors .
although the relative motion of ac is fast , it is not directed away from pm(a ) as would be the case for a distant background star ; it is produced by the orbital motion of the inner binary ab .
the presence of an inner subsystem in the binary ab was suspected previously from its variable rv @xcite .
the estimated masses of ba and bb , 0.98 and 0.65@xmath9 respectively , remove the discrepancy between the mass sum of 3.05@xmath9 calculated for the ab system from its grade-4 orbit by @xcite and the expected mass sum .
although such discrepancies are not uncommon for low - quality orbits , in this case it serves as an indirect confirmation of ba , bb .
the estimated period of ba , bb is on the order of 20yrs .
see figure [ new_binaries]b for the discovery image .
_ hip 31267 ( hd 46013 = wds 06335 + 4822 ) _ is another triple where the inner spectroscopic binary with a period of 1.3yr ( d. latham , 2012 , private communication ) has a tertiary companion discovered with robo - ao and seen in the archival 2mass image .
this component is located above the ms , but its photometry has large errors . its status remains indeterminate because the pm(a ) is only 0053 yr@xmath5 and we do not have a sufficiently long time base . a wider pair ( ab = uc 1450 at 53 " )
was identified as a common proper motion pair @xcite , however , the colors we find here implies that it is optical .
the crowding is moderate .
possibly the preferred motion of background stars accidentally matches the slow pm of a. _ hip 33555 ( hd 50720 = wds 06562 + 4032 ) _ has a slow pm of 0037 yr@xmath5 .
the r15 companion at 56 is located below the ms and could be optical , despite moderate crowding .
we discovered another similarly faint star at 19 which was not spotted in the robo - ao @xmath3-band image .
little can be said about its status . the main target itself is a close spectroscopic binary ( d. latham , 2012 , private communication ) and an x - ray source .
see figure [ new_binaries]c for the discovery image .
_ hip 86642 ( hd 161163 = wds 17422 + 3804 ) _ is a triple system .
the r15 companion at 22 is confirmed as physical by its fixed position and its location on the cmd .
the main star is a double - lined spectroscopic binary with @xmath31yr ( d. latham , 2012 , private communication ) and estimated semi - major axis of 01 .
our standard algorithm , @xmath30 , produced subpar results on this stars and instead we used the deconvolution technique used in @xcite to analyze this star .
it is resolved here at 007 .
the magnitude difference of aa , ab ( @xmath32 ) matches the spectroscopic mass ratio of 0.52 and corresponds to the ab mass of 0.60@xmath9 . the orbital motion of aa
, ab can be followed with ao and speckle interferometry .
see figure [ new_binaries]d for the discovery image .
_ hip 96395 ( hd 185414 = wds 19359 + 5659 ) _ is triple .
its inner system has a preliminary spectroscopic period of 14yr and a low amplitude ( d. latham , 2012 , private communication ) . despite the estimated semi - major axis of 025 ,
the spectroscopic secondary is too faint for its direct resolution without a high - contrast coronagraph .
the 2mass companion at 10 , noted first by @xcite , shares the large pm of a and is located on the ms .
_ hip 101234 ( hd 195872 = wds 20312 + 5653 ) _ is an acceleration binary @xcite and was first resolved by r15 .
the position angle in this paper measured on 2013.8367 differs by 27@xmath2 from the 2012.7592 measurement .
that is not the only discrepancy between the two data sets .
the photometry with p3k shows large magnitude differences ( @xmath33 and @xmath34 ) , while robo - ao detected a binary of equal brightness .
this system is not plotted in figure [ fig : cmd ] .
the separation between the two components is fairly small and in the p3k data , the companion lies on the airy ring of the primary , which could increase the errors .
the primary is a high proper motion star , the pair ab is most likely a physical binary .
additional observations are necessary to understand this system .
_ hip 101430 ( hd 195992 = wds 20333 + 3323 ) _ is a new 2 + 2 quadruple system .
the outer 17 binary hj 1535 ab has been known since 1828 @xcite . if b were a background star , the pm of a , 0208 yr@xmath5 , would moved it by 38 in 185yr .
the pair ab is therefore undoubtedly physical .
the discrepant pm of b reported in the wds could be caused by the subsystem ba , bb discovered here .
the main component a has an astrometric subsystem aa , ab with @xmath35yr @xcite , confirmed spectroscopically by d. latham .
the component b was tentatively resolved by robo - ao , but was thought to be too uncertain to publish in r15 . now ba , bb is clearly resolved at 017 .
its estimated period is on the order of 30yr .
the field is extremely crowded and another faint star e is detected at 121 from a by both robo - ao and palm-3000 .
comparison with 2mass shows relative motion opposite to pm(a ) , so ae is optical .
see figure [ new_binaries]e for the discovery image .
_ hip 103455 ( hd 199598 = wds 20577 + 2624 ) _ is a new 06 binary detected by robo - ao .
it was targeted because of its variable velocity @xcite and astrometric acceleration @xcite .
the binary is confirmed here , helped by the large pm of the primary .
our photometry places b above the main sequence . according to @xcite ,
the target is young , so its companion could indeed be a pms star .
the object was already targeted by @xcite in their survey of young stars , but the binary was not detected .
after learning of our discovery , metchev was able to extract a measurement of the star from the archival data ( private communication s. metchev ) .
_ hip 104514 ( hd 201639 = wds 21102 + 2045 ) _ has a variable rv @xcite , so the 33 companion b discovered by robo - ao makes it a triple system ( the estimated period of ab is 2000yr , too long to cause the rv variation ) .
we found the component b to be on the [ fe / h]=@xmath36 isochrone , in agreement with [ fe / h]=@xmath37 measured for the star a @xcite . _
hip 108473 ( hd 208776 = wds 21585 + 0347 ) _ is a single - lined spectroscopic binary with @xmath38yr @xcite and a large pm of 028 yr@xmath5 .
the distant component b at 123 is confirmed as physical by 2mass , robo - ao , and this work . it is located slightly below the ms , just like hip 104514b . the object is also metal - deficient relative to the sun , [ fe / h]=@xmath39 @xcite . _ hip 110574 ( hd 212426 = wds 22240 + 0612 ) _ is a triple system consisting of a close binary aa , ab with a separation of 009 detected by r15 and a cpm companion with a separation of 171 .
the aa , ab system is unresolved in the @xmath0 images , but is elongated in the @xmath6 images . we were unable to extract astrometry from the data .
the separation has decreased since it was discovered in 2012 ; @xcite was able to resolve it with visible speckle interferometry at the 4.1 m soar telescope on 2013.73 at 41mas @xcite .
_ hip 118213 ( hd 224531 = wds 23588 + 3156 ) _ was previously thought to be a binary , but our observations reveal that it is actually a triple system .
the outer component b at 48 was discovered by robo - ao and confirmed by the 2mass image and by the new data presented here .
it is located near the low end of the ms in figure [ fig : cmd ] .
the main star with astrometric acceleration @xcite is now resolved at 04 .
this separation corresponds to an orbital period on the order of 100yrs . the new component ab is too faint for its detection in the @xmath3 band with robo - ao
. see figure [ new_binaries]f for the discovery image . _ hip 118225 ( hd
224543 = wds 23588 + 3345 ) _ is a triple system .
the inner spectroscopic binary has @xmath40d @xcite , the robo - ao companion at 5 is seen at a constant position despite pm(a)= 03 yr@xmath5 , while its color places it on the ms .
this study shows the power of high - resolution ao imaging for the study of hierarchical multiplicity .
we resolved for the first time inner subsystems in six binaries .
four of those ( hip 17129 , 86642 , 118213 , 25300 ) had previous indications of subsystems from variable rv and/or astrometric acceleration .
the ao discovery space overlaps with these alternative techniques , but direct resolution of subsystems allows their characterization in terms of period estimated from separation and mass ratio estimated from photometry .
as the orbital periods of these subsystems range from several years to several decades , such characterization by rv or astrometry would necessarily involve a long - term monitoring , which is not available in most cases .
the exception is hip 17129 , where a subsystem with @xmath25yr , first resolved here , was inferred from the strange motion of the visual pair observed for almost two centuries .
the new subsystem in the secondary component of hip 101430 was totally unexpected . this 17visual pair known for 185 years
is now revealed as a 2 + 2 quadruple system , where each of the visual components is in turn a close binary .
the fainter secondary components of visual binaries generally have much less information on subsystems compared to the brighter primaries @xcite , and here high - resolution ao imaging makes a large difference . our strategy of observing binaries to discover subsystems has been proven successful .
the emerging statistics @xcite indicates that a third to a half of visual binaries harbor subsystem(s ) .
the simple fact that binaries can contain more than just two stars is often forgotten or ignored , for example when comparing pms of components in wide binaries or calculating their orbits from short arcs while assuming single - star masses .
similarly , rv monitoring of wide binaries in search of exo - planets is incomplete without full characterization of subsystems @xcite .
this work contributes to the statistical characterization of hierarchical multiplicity in the solar neighborhood .
it will help to understand the formation mechanisms that create the panoply of single , binary , and multiple stars , and thus to understand the origin of stars and planets in general .
there is additional work that can be done with these systems .
additional astrometric measurements will show orbital motion confirming that the systems are physically bound .
in addition , future astrometric measures can be combined with the rv measurements that have been made on several of the systems and will allow for the computation of the full orbital solution .
though there are strong indications that the newly observed companions are physical , with the exception of hip 33555 , the new companions need to be confirmed with additional observations .
additional observations of hip 101234 are need to resolve the discrepancy between the visible and the near - ir differential magnitude . with a period of only six years measured by rv , hip 86642 aa ,
ab is a prime target for frequent follow up astrometric measurements that can be combined with the double lined spectroscopic orbit .
we thank d. latham for the insight he provided on a number of these systems .
this paper is based on observations obtained at the hale telescope , palomar observatory .
a portion of the research in this paper was carried out at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration ( nasa ) .
this research made use of the washington double star catalog maintained at the u.s .
naval observatory , the simbad database , operated by the cds in strasbourg , france and nasa s astrophysics data system .
this publication made use of data products from the two micron all sky survey , which is a joint project of the universit of massachusetts and the infrared processing and analysis center / california institute of technology , funded by nasa and the national science foundation .
|
twenty multiple stellar systems with solar - type primaries were observed at high angular resolution using the palm-3000 adaptive optics system at the 5 m hale telescope . the goal was to complement the knowledge of hierarchical multiplicity in the solar neighborhood by confirming recent discoveries by the visible robo - ao system with new near - infrared observations with palm-3000 . the physical status of most , but not all , of the new pairs is confirmed by photometry in the @xmath0 band and new positional measurements . in addition , we resolved for the first time five close sub - systems : the known astrometric binary in hip 17129ab , companions to the primaries of hip 33555 , and hip 118213 , and the companions to the secondaries in hip 25300 and hip 101430 .
we place the components on a color - magnitude diagram and discuss each multiple system individually
. = @xmath1 @xmath2
|
we start out with a few definitions from group theory .
let @xmath4 be a group .
we say that _ @xmath4 splits over the subgroup @xmath5 _ if @xmath4 admits an hnn decomposition with base group @xmath6 and amalgamating subgroup @xmath5 .
more precisely , @xmath4 splits over the subgroup @xmath5 if there exists an isomorphism @xmath7 where @xmath8 are subgroups of @xmath4 and @xmath9 is a monomorphism . in this notation ,
relations of @xmath6 are implicit .
we will write such a presentation more compactly as @xmath10 in this paper we are interested in splittings of knot groups .
given a knot @xmath11 we denote the knot group @xmath12 by @xmath2 .
we denote by @xmath13 the genus of the knot , the minimal genus of a seifert surface @xmath14 for @xmath0 .
it follows from the loop theorem and the seifert - van kampen theorem that we can split the knot group @xmath2 over the free group @xmath15 of rank @xmath16 .
the _ rank _
@xmath17 of a group @xmath18 is the minimal size of a set of generators for @xmath18 .
it is well known that if @xmath0 is a fibered knot , that is , the knot complement @xmath19 fibers over @xmath20 , then the group @xmath2 splits only over free groups of rank @xmath16 .
( see , for example , lemma [ lem : fibsplit ] . )
we show that this property characterizes fibered knots .
in fact , we can say much more .
[ thm : fibsplitintro ] let @xmath0 be a non - fibered knot
. then @xmath2 splits over non - free groups of arbitrarily large rank .
neuwirth ( * ? ? ?
* problem l ) asked whether there exists a knot @xmath0 such that @xmath2 splits over a _
free _ group of rank other than @xmath16 . by the above
, such a knot would necessarily have to be non - fibered .
lyon ( * ? ? ?
* theorem 2 ) showed that there does in fact exist a non - fibered genus - one knot @xmath0 with incompressible seifert surfaces of arbitrarily large genus .
this implies in particular that there exists a knot @xmath0 for which @xmath2 splits over free groups of arbitrarily large rank .
we give a strong generalization of this result .
[ freesplitting ] let @xmath0 be a non - fibered knot .
then for any integer @xmath21 there exists a splitting of @xmath2 over a free group of rank @xmath22 .
note that an incompressible seifert surface gives rise to a splitting over a free group of _ even rank_. the splittings over free groups of _ odd rank _ in the theorem are therefore not induced by incompressible seifert surfaces .
feustel and gregorac @xcite showed that if @xmath23 is an aspherical , orientable @xmath24-manifold such that @xmath25 splits over the fundamental group of a _ closed _
surface @xmath26 , then this splitting can be realized topologically by a properly embedded surface .
( more splitting results can be found in ( * ? ? ?
* proposition 2.3.1 ) . )
the fact that fundamental groups of non - fibered knots can be split over free groups of odd rank shows that the result of feustel and gregorac does not hold for splittings over fundamental groups of surfaces with boundary .
theorems [ thm : fibsplitintro ] and [ freesplitting ] can be viewed as strengthenings of stallings s fibering criterion .
we refer to section [ section : stallings ] for a precise statement .
our third main theorem shows that theorem [ freesplitting ] is optimal .
[ mainthm][mainthm3 ] if @xmath0 is a knot , then @xmath2 does not split over a group of rank less than @xmath16 . the case @xmath27 follows from the kneser conjecture and work of waldhausen @xcite , as we show in section [ section : genusone ] .
however , to the best of our knowledge , the classical methods of 3-manifold topology do not suffice to prove theorem [ mainthm ] in the general case .
we use the recent result @xcite that wada s invariant detects the genus of any knot .
this result in turn relies on the seminal work of agol @xcite , wise @xcite , przytycki
wise @xcite and liu @xcite .
theorem [ mainthm ] is of interest for several reasons : it gives a completely group - theoretic chararcterization of the genus of a knot , namely @xmath28 a different group - theoretic characterization was given by calegari ( see the proof of proposition 4.4 in @xcite ) in terms of the ` stable commutator length ' of the longitude .
theorem [ mainthm ] fits into a long sequence of results showing that minimal - genus seifert surfaces ` stay minimal ' even if one relaxes some conditions .
for example , gabai @xcite showed that the genus of an _ immersed _ surface cobounding a longitude of @xmath0 is at least @xmath13 .
furthermore , minimal - genus seifert surfaces give rise to surfaces of minimal complexity in the 0-framed surgery @xmath29 ( see @xcite ) and in most @xmath20-bundles over @xmath29 ( see @xcite ) . given a closed @xmath24-manifold @xmath23 it is obvious that @xmath30 is a lower bound for the heegaard genus @xmath31 of @xmath23 . in light of theorem [ mainthm ] one might hope that this is in an equality ; that is , that @xmath32 .
this is not the case , though , as was shown by various authors ( see @xcite and @xcite ) .
the paper is organized as follows . in section [ section : hnnsplittings ]
we discuss several basic facts about hnn decompositions of groups . in section [ section : splitk ] we recall that incompressible seifert surfaces give rise to hnn decompositions of knot groups and we characterize in lemma [ lem : fibsplit ] the splittings of fundamental groups of fibered knots . in section [ section:52 ] we consider the genus - one non - fibered knot @xmath33 .
we give explicit examples of splittings of the knot group over a non - free group and over the free group @xmath34 of rank 3 , and inequivalent splittings of the knot group over @xmath35 .
section [ section : splitnonfree ] contains the proof of theorem [ thm : fibsplitintro ] , and in section [ section : splitfree ] we give the proof of theorem [ freesplitting ] . in section [ section : stallings ] we show that these two theorems strengthen stallings s fibering criterion . in section [ section : genusone ]
we give a proof of theorem [ mainthm ] for genus - one knots .
the proof relies mostly on the kneser conjecture and a theorem of waldhausen . in section [ section : wada ] we review the definition of wada s invariant of a group .
finally , in section [ section : proof ] we prove theorem [ thm : technical ] , which combined with the main result of @xcite provides a proof of theorem [ mainthm ] for all genera .
we conclude this introduction with two questions .
the precise notions are explained in section [ section : hnnsplittings ] .
let @xmath4 be a word hyperbolic group and let @xmath36 be an epimorphism such that @xmath37 is not finitely generated .
does @xmath38 admit splittings over ( infinitely many ) pairwise non - isomorphic groups ? ( the group @xmath39 satisfies these conditions if @xmath0 is a non - fibered knot . ) let @xmath0 be a non - fibered knot of genus @xmath1 .
does @xmath2 admit ( infinitely many ) inequivalent splittings over the free group @xmath40 on @xmath3 generators ?
all groups are assumed to be finitely presented unless we say specifically otherwise .
all @xmath24-manifolds are assumed to be connected , compact and orientable . given a submanifold @xmath41 of a @xmath24-manifold @xmath23 , we denote by @xmath42 an open tubular neighborhood of @xmath41 in @xmath23 . given @xmath43 we denote by @xmath44 the free group on @xmath22 generators .
the first author wishes to thank the university of sydney for its hospitality .
we are also very grateful to eduardo martinez - pedroza , saul schleimer and henry wilton for very helpful conversations .
an _ hnn decomposition _ of a group @xmath4 is a 4-tuple @xmath45 consisting of subgroups @xmath46 of @xmath4 , a
_ stable letter _
@xmath47 , and an injective homomorphism @xmath9 , such that the natural inclusion maps induce an isomorphism from @xmath48 to @xmath4 .
alternatively , a hnn - decomposition of @xmath4 can be viewed as an isomorphism @xmath49 where @xmath50 is an injective map .
we will frequently go back and forth between these two points of view .
we need a few more definitions : given an hnn - decomposition @xmath45 we refer to the homomorphism @xmath36 that is given by @xmath51 and @xmath52 for @xmath53 as the _
canonical epimorphism_. let @xmath4 be a group and let @xmath54 be an epimorphism .
a _ splitting of @xmath55 over a subgroup @xmath5 ( with base group @xmath6 ) _ is an hnn decomposition @xmath45 of @xmath4 such that @xmath56 equals the canonical epimorphism . with the alternative point of view
explained above , a splitting of @xmath38 is an isomorphism @xmath49 such that the following diagram commutes : @xmath57_\e \ar[rr]^-{f } & & \ll a , t\,|\ , \varphi(b)=tbt^{-1}\rr\ar[dl]^\psi\\ & \z&}\ ] ] where @xmath58 denotes the canonical epimorphism .
two splittings @xmath45 and @xmath59 of @xmath38 are called _ weakly equivalent _ if there exists an automorphism @xmath60 of @xmath4 with @xmath61 .
if @xmath60 can be chosen to be an inner automorphism of @xmath4 , then the two hnn decompositions are said to be _
strongly equivalent_. we conclude this section with the following well - known lemma of @xcite .
it appears as theorem b * in @xcite where an elementary proof can be found .
[ lem : splitexists ] let @xmath4 be a finitely presented group and let @xmath54 be an epimorphism . then there exists a splitting @xmath62 of @xmath38 where @xmath6 and @xmath5 are finitely generated .
the following lemma characterizes splittings of pairs @xmath38 for which @xmath37 is finitely generated .
[ lem : splitkerfg ] let @xmath4 be a finitely presented group , @xmath36 an epimorphism , and @xmath63 an element of @xmath4 with @xmath51 . if @xmath37 is finitely generated , then there exists a canonical isomorphism @xmath64 where @xmath65 and where @xmath66 is given by conjugation by @xmath63 .
furthermore , any other splitting of @xmath38 is strongly equivalent to this splitting .
let @xmath4 be a finitely presented group and let @xmath36 be an epimorphism such that @xmath67 is finitely generated .
we have an exact sequence @xmath68 let @xmath69 with @xmath51 .
the map @xmath70 defines a right - inverse of @xmath56 , and we see that @xmath5 is canonically isomorphic to the semi - direct product @xmath71 where @xmath72 acts on @xmath5 by conjugation by @xmath72 . that is , we have a canonical isomorphism @xmath73 we now suppose that we have another splitting @xmath74 of @xmath38 . by our hypothesis the group @xmath67
is finitely generated .
on the other hand , it follows from standard results in the theory of graphs of groups ( see @xcite ) that @xmath75 where @xmath76 and @xmath77 for all @xmath78 and each map @xmath79 is given by @xmath58 . as in @xcite , the fact
that the infinite free product with amalgamation is finitely generated implies that @xmath80 for all @xmath78 .
this , in turn , implies that each @xmath81 and @xmath82 is isomorphic to @xmath83 .
it is now clear that the identity on @xmath4 already has the desired property relating the two splittings of @xmath38 .
let @xmath84 be an hnn - extension .
given @xmath85 we denote by @xmath86}$ ] the result of amalgamating the groups @xmath87 , @xmath88 along the subgroups @xmath89 , @xmath90 . in our notation , @xmath91}= \langle \ast_{i = m}^n t^i a t^{-i}\mid t^j\varphi(b)t^{-j } = t^{j+1}bt^{-j-1}\ ( j = n , \ldots , m-1 ) \rr.\ ] ] given any @xmath92 , we have a canonical map @xmath93}\to a_{[k , l]}$ ] which is a monomorphism ( see , for example , @xcite for details ) .
if @xmath94 is the canonical epimorphism , then it is well known that @xmath37 is given by the direct limit of the groups @xmath95}$ ] , @xmath96 ; that is , @xmath97}.\ ] ] the following well - known lemma shows that a splitting of a pair @xmath98 gives rise to a sequence of splittings .
[ lem : moresplittings ] let @xmath84 be an hnn - extension
. for any integer @xmath99 , let @xmath100})\to a_{[1,n+1]}\ ] ] be the map that is given by conjugation by @xmath63 .
then the obvious inclusion maps induce an isomorphism @xmath101},t\,|\ , \varphi_n(a_{[0,n]})=ta_{[0,n]}t^{-1}\rr\xrightarrow{\i } \pi=\ll a , t\,|\ , \varphi(b)=tbt^{-1}\rr.\ ] ] we write @xmath102},t\,|\ , \varphi_n(a_{[0,n]})=ta_{[0,n]}t^{-1}\rr.\ ] ] we denote by @xmath103 ( respectively @xmath104 ) the kernel of the canonical map from @xmath4 ( respectively @xmath105 ) to @xmath106 .
it is clear that it suffices to show that the restriction of @xmath107 to @xmath108 is an isomorphism . for @xmath78
, we write @xmath109 and @xmath110 .
note that @xmath104 is canonically isomorphic to @xmath111 and @xmath103 is canonically isomorphic to @xmath112 it is now straightforward to see that @xmath113 does indeed restrict to an isomorphism @xmath114 .
note that the isomorphism in lemma [ lem : moresplittings ] is canonical . throughout the paper
we will therefore make the identification @xmath115},t\,|\ , \varphi_n(a_{[0,n]})=ta_{[0,n]}t^{-1}\rr.\ ] ] in the paper we will also write @xmath116}$ ] .
now let @xmath11 be a knot , that is , an oriented embedded simple closed curve in @xmath117 .
we write @xmath118 and @xmath119 the orientation of @xmath0 gives rise to a canonical epimorphism @xmath120 sending the oriented meridian to 1 .
let @xmath14 be a seifert surface of genus @xmath1 for @xmath0 ; that is , a connected , orientable , properly embedded surface @xmath14 of genus @xmath1 in @xmath121 such that @xmath122 is an oriented longitude for @xmath0 .
note that @xmath14 is dual to the canonical epimorphism @xmath56 .
suppose that @xmath14 is incompressible .
( recall that a surface @xmath14 in a @xmath24-manifold @xmath23 is called _ incompressible _ if the inclusion - induced map @xmath123 is injective . )
we pick a tubular neighborhood @xmath124 $ ] .
the manifold @xmath125 is the result of _ cutting along @xmath126_. the seifert
van kampen theorem gives us a splitting @xmath127 of @xmath128 , where @xmath129 is induced by the canonical homeomorphism @xmath130 .
we thus see that @xmath2 splits over the free group @xmath15 of rank @xmath3 .
given a knot @xmath11 , we denote by @xmath131 the minimal genus of a seifert surface .
it follows from the loop theorem ( see , for example , ( * ? ? ?
* chapter 4 ) ) that a seifert surface of minimal genus is incompressible .
hence @xmath2 splits over a free group of rank @xmath16 .
if two incompressible seifert surfaces of a knot @xmath0 are isotopic , then it is clear that the corresponding splittings of @xmath2 are strongly equivalent .
there are many examples of knots that admit non - isotopic minimal genus seifert surfaces ; see e.g. @xcite .
we expect that these surfaces give rise to splittings that are not strongly equivalent .
on the other hand , if a knot is fibered , then it admits a unique minimal genus seifert surface up to isotopy ( see e.g. ( * ? ? ?
* lemma 5.1 ) ) .
it is therefore perhaps not entirely surprising that @xmath2 admits a unique splitting up to strong equivalence .
more precisely , we have the following well - known lemma , which is originally due to neuwirth @xcite . [
lem : fibsplit ] let @xmath0 be a fibered knot of genus @xmath1 with fiber @xmath14 .
then any splitting of @xmath2 is strongly equivalent to @xmath132 in particular @xmath2 only splits over the free group of rank @xmath3 .
if @xmath126 is a fiber surface for @xmath121 , then the infinite cyclic cover of @xmath121 is diffeomorphic to @xmath133 . put differently , @xmath134 which implies in particular that @xmath135 is finitely generated .
the lemma is now a straightforward consequence of lemma [ lem : splitkerfg ] .
in this section we give several explicit splittings of the knot group @xmath2 where @xmath33 , the first non - fibered knot in the alexander - briggs table .
we construct : three splittings of @xmath136 over the free group @xmath35 , no two being weakly equivalent ; a splitting of @xmath136 over the free group @xmath34 on three generators ; a splitting of @xmath136 over a non - free group .
note that neither the second nor the third splitting is induced by an incompressible surface .
we will also see that at least two of the three splittings over @xmath35 are not induced by an incompressible surface .
since @xmath0 is a knot of genus one , a minimal - genus seifert surface gives rise to a splitting of @xmath2 over a free group of rank 2 . in the following
we will consider an explicit splitting that comes from a wirtinger presentation of the knot group : @xmath137 here the knot group has an hnn decomposition @xmath45 , where @xmath6 is the free group on @xmath138 while @xmath5 is the subgroup freely generated by @xmath139 and @xmath140 .
the isomorphism @xmath129 sends @xmath141 and @xmath142 . for the remainder of this section
we identify @xmath2 with @xmath48 .
[ prop:52splittings ] consider the splittings : @xmath143},t\,|\ , \varphi_1(a)=tat^{-1}\rr,\\ \pi(k)&=&\ll a_{[0,2]},t\,|\ , \varphi_2(a_{[0,1]})=ta_{[0,1]}t^{-1}\rr\ea\ ] ] where the latter two splittings are provided by lemma [ lem : moresplittings ] . then the following hold .
a. each is a splitting over a free group of rank two .
b. no two of the splittings of @xmath128 are weakly equivalent .
c. at least two of the splittings are not induced by an incompressible seifert surface . in the proof of proposition [ prop:52splittings ]
we will make use of the following lemma which is perhaps also of independent interest .
[ lem : propersubgroup ] let @xmath144 be a hyperbolic @xmath24-manifold with empty or toroidal boundary , and let @xmath18 be a subgroup of @xmath145 . if @xmath146 is an automorphism with @xmath147 , then @xmath148 .
we do not know whether the conclusion of the lemma holds for any @xmath24-manifold .
let @xmath146 be an automorphism with @xmath147 .
since @xmath144 is hyperbolic , it is a consequence of the mostow rigidity theorem that the group of outer automorphisms of @xmath4 is finite .
( see , for example , ( * ? ? ? * theorem c.5.6 ) and @xcite for details . )
consequently , there exists a positive integer @xmath149 and an element @xmath150 such that @xmath151 .
it follows from ( * ? ? ?
* theorem 4.1 ) that @xmath152 .
the assumption that @xmath147 implies inductively that @xmath153 .
hence @xmath148 .
we can now turn to the proof of proposition [ prop:52splittings ] .
it is clear that the first and the second splitting are over a free group of rank two .
it remains to show that @xmath154}$ ] is a free group of rank two .
first note that @xmath155}\cong \ll a_0 , b_0 , a_1 , b_1\,|\ , a_1=b_0,\ , b_1^{-1 } a_1b_1^{-1}=(b_0^{-1}a_0)^2\rr,\ ] ] where @xmath156 and @xmath157 denote @xmath158 and @xmath159 , respectively .
using the first relation to eliminate the generator @xmath160 , we obtain @xmath154}\cong \ll a_0 , a_1 , b_1\,|\ , r \rr,$ ] where @xmath161 we let @xmath162 and @xmath163 . clearly @xmath164 is a basis for the free group on @xmath165 .
hence @xmath154}\cong \ll c , d , r\ , |\ , r \rr \cong \ll c , d\ , |\ , \rr$ ] is indeed a free group of rank 2 .
this concludes the proof of ( i ) .
we turn to the proof of ( ii ) . since @xmath0 is not fibered it follows from stallings s theorem ( see theorem [ thm : st62 ] ) that @xmath166}$ ]
is not finitely generated .
it follows that easily that for any @xmath167 the map @xmath168}\to a_{[0,l]}$ ] is a proper inclusion .
in particular , we have proper inclusions @xmath169 } \varsubsetneq a_{[0,2]}$ ] .
since @xmath170 is hyperbolic , the desired statement now follows from lemma [ lem : propersubgroup ] . we prove ( iii ) .
it is well known ( see , for example , @xcite ) that any two minimal - genus seifert surfaces of @xmath171 are isotopic .
this implies , in particular , that any two splittings of @xmath2 induced by minimal - genus seifert surfaces are strongly equivalent .
it follows from ( ii ) that at least two of the three splittings are not induced by a minimal genus seifert surface .
we show that @xmath2 admits a splitting over a free group of rank 3 . in order to do
so we note that there exists a canonical isomorphism [ equ1 ] & & a , b , t| tat^-1=b , tb^-1ab^-1t^-1=(b^-1a)^2 + & & a , b , c , t| tat^-1=b , tb^-1ab^-1t^-1=(b^-1a)^2 , tb^-2ab^-2t^-1=c.let @xmath172 be the free group generated by @xmath173 .
let @xmath174 be the subgroup of @xmath172 generated by @xmath175 .
the fundamental group of the covering graph in figure [ graph ] is free on @xmath176 , and @xmath177 , and so @xmath174 is a free rank-3 subgroup of @xmath172 .
the elements @xmath178 of @xmath172 also generate a free group of rank 3 , since they are free in the abelianization of @xmath172 .
there exists therefore a unique homomorphism @xmath179 such that @xmath180 , @xmath181 and @xmath182 .
it follows that @xmath183 is in fact a monomorphism .
hence from ( [ equ1 ] ) , @xmath184 defines a splitting of @xmath2 over the free group @xmath174 of rank three .
finally we give an explicit splitting of @xmath2 over a subgroup that is not free .
recall that by lemma [ lem : moresplittings ] the group @xmath2 admits an hnn decomposition with the hnn base @xmath185}$ ] defined as the amalgamated product of @xmath186 and @xmath187 .
it suffices to prove the following claim .
the group @xmath185}$ ] is not free .
note that @xmath185}$ ] has the presentation @xmath188 using the first and third relations , we eliminate the generators @xmath160 and @xmath189 .
thus @xmath190 } \cong \ll a_0 , a_1 , a_2 , b_2\,|\ , r_1,\ , r_2 \rr,\ ] ] where @xmath191 and @xmath192 .
let @xmath193 and @xmath194 .
one checks that @xmath195 is a basis for the free group @xmath196 using the substitutions @xmath197 we see @xmath190 } \cong \ll \e , f , b_2 \,|\ , r_2 \rr \cong \ll \e , f ,
b_2 \,|\ , f^{-2}\e^{-2}(b_2 \e^2 ) f(b_2\e^2)\rr.\ ] ] we perform two more changes of variables
. first we let @xmath198 and eliminate @xmath199 to obtain @xmath190 } \cong \ll \e , f , g \,|\ , \e^{-2}(gf)^2f^{-3 } \rr , .\ ] ] second , we let @xmath200 and we eliminate @xmath1 : @xmath190 } \cong \ll \e , f , h \,|\ , \e^{-2}h^2 = f^3 \rr.\ ] ] we thus see that @xmath185}$ ] is a free product of two free groups amalgamated over an infinite cyclic group .
by lemma 4.1 of @xcite ( see example 4.2 ) , if the group @xmath185}$ ] is free , then either @xmath201 or @xmath202 is a basis element in its respective factor .
since neither element is a basis element ( seen for example by abelianizing ) , the group @xmath185}$ ] is not free .
this concludes the proof of the claim .
in section [ section:52 ] we saw that we can split the knot group @xmath136 over a group that is not free .
we will now see that this example can be greatly generalized .
we recall the statement of our first main theorem .
[ thm : fibsplit ] if @xmath0 is a non - fibered knot , then @xmath2 admits splittings over _ non - free _ subgroups of arbitrarily large rank .
let @xmath203 be a seifert surface of minimal genus .
we write @xmath204 and @xmath205 , and we consider the corresponding splitting @xmath206 of @xmath128 over @xmath207 .
given @xmath208 we consider , as in section [ section : anm ] , the group @xmath91}= \langle \ast_{i = m}^n t^i a t^{-i}\mid t^j\varphi(b)t^{-1 } = t^{j+1}bt^{-j-1}\ ( j = n , \ldots , m-1 ) \rr.\ ] ] by lemma [ lem : moresplittings ] the group @xmath2 splits over the group @xmath209}$ ] for any non - negative integer @xmath149 .
there exists an integer @xmath210 such that @xmath209}$ ] is not a free group for any @xmath211 .
as we pointed out in section [ section : anm ] , we have an isomorphism @xmath212}\ ] ] where the maps @xmath213}\to a_{[-k , k]}$ ] for @xmath214 are monomorphisms .
it follows from ( * ? ? ?
* theorem 3 ) that @xmath135 is not locally free ; that is , there exists a finitely generated subgroup of @xmath135 which is not a free group . but
this implies that there exists @xmath43 such that @xmath215}$ ] is not a free group .
we have a canonical isomorphism @xmath215}\cong a_{[0,2k]}$ ] , and for any @xmath216 we have a canonical monomorphism @xmath217}\to a_{[0,n]}$ ] .
it now follows that @xmath209}$ ] is not a free group for any @xmath216 .
this concludes the proof of the claim . to complete the proof of theorem [ thm : fibsplit ] it remains to prove the following claim : writing @xmath218}$ ] we have @xmath219 since @xmath220 is not a fiber it follows from ( * ? ? ?
* theorem 10.5 ) that there exists an element @xmath221 . by work of przytycki
wise ( see ( * ? ? ?
* theorem 1.1 ) ) the subgroup @xmath222 is separable .
this implies , in particular , that there exists an epimorphism @xmath223 onto a finite group @xmath18 such that @xmath224 .
then @xmath225 given @xmath226 we denote by @xmath227 the restriction of @xmath228 to @xmath229 and we write @xmath230 .
note that in @xmath231 the groups @xmath232 , viewed as subgroups of @xmath2 , are conjugate .
it follows that the groups @xmath233 are conjugate in @xmath18 .
in particular , each of the groups @xmath233 has order @xmath234 . the same argument
shows that each of the groups @xmath235 has order @xmath236 .
standard arguments about fundamental groups of graphs of groups ( see , for example , @xcite ) imply that @xmath237 is the fundamental group of a graph of groups , where the underlying graph @xmath238 is a connected graph with @xmath239 vertices and @xmath240 edges . from the reidemeister - schreier theorem ( see , for example , ( * ? ? ?
* theorem 2.8 ) and from the fact that @xmath237 surjects onto @xmath241 it then follows that @xmath242 but this sequence diverges to @xmath243 since @xmath244 .
lyon ( * ? ? ?
* theorem 2 ) showed that there exists a non - fibered knot @xmath0 of genus one that admits incompressible seifert surfaces of arbitrarily large genus ( see also @xcite for related examples ) . by the discussion in section [ section : splitk ] , this implies that @xmath2 splits over free groups of arbitrarily large rank .
splitting along incompressible seifert surfaces is a convenient way to produce knot group splittings .
yet there are many non - fibered knots that have unique incompressible seifert surfaces ( see , for example , @xcite ) .
for such a knot , seifert surfaces gives rise to only one type of knot group splitting .
in section [ section:52 ] we saw an example of a splitting of a knot group over a free group that is not induced by an embedded surface .
we generalize the example in our second main theorem .
we recall the statement .
[ thm : splitfreelarge ] let @xmath0 be a non - fibered knot . then for any integer @xmath21 there exists a splitting of @xmath2 over a free group of rank @xmath22 .
the key to extending the result in section [ section:52 ] is the following theorem , which we will prove in the next subsection .
[ thm : extendfree ] let @xmath0 be a non - fibered knot .
then there exists a seifert surface @xmath126 of minimal genus such that for a given base point @xmath245 there exists a nontrivial element @xmath246 such that the subgroup of @xmath2 generated by @xmath247 and @xmath1 is the free product of @xmath247 and the infinite cyclic group @xmath248 .
theorem [ thm : splitfreelarge ] is now a consequence of theorem [ thm : extendfree ] and the following proposition about hnn decompositions .
assume that @xmath55 splits over a free group @xmath249 of rank @xmath149 with base group @xmath6 .
if there exists an element @xmath250 such that the subgroup of @xmath4 generated by @xmath249 and @xmath1 is the free product @xmath251 , then @xmath55 splits over free groups of every rank greater than @xmath149 . by hypothesis
we can identify @xmath4 with @xmath252 where @xmath253 generate the group @xmath249 and where @xmath56 is given by @xmath51 and @xmath254 .
the kernel of the second - factor projection @xmath255 is an infinite free product @xmath256 .
let @xmath257 be any positive integer .
choose a nontrivial element @xmath258 and define @xmath259 , for @xmath260 .
then @xmath261 is a free subgroup of @xmath262 with rank @xmath263 . by hypothesis @xmath264
is then also a free subgroup of @xmath6 of rank @xmath265 .
note that @xmath4 is canonically isomorphic to @xmath266 we denote by @xmath172 the free product of @xmath6 and @xmath267 , and we denote by @xmath183 the unique homomorphism @xmath268 that extends @xmath129 and that maps each @xmath269 to @xmath270 .
since @xmath183 is the free product of two isomorphisms , it is also an isomorphism .
we then have a canonical isomorphism @xmath271 we have thus shown that @xmath55 splits over the free group @xmath264 of rank @xmath265 . to prove theorem [ thm : extendfree ] we will need to discuss the jsj pieces of knot complements .
( see @xcite for exposition about jsj decompositions . )
it is therefore convenient to generalize a few notions for knots to more general 3-manifolds .
given a @xmath24-manifold @xmath23 , we can associate to each class @xmath272 its thurston norm @xmath273 , which is defined as the minimal ` complexity ' of a surface dual to @xmath56 .
we say that a class @xmath272 is _ fibered _ if there exists a fibration @xmath274 such that the induced map @xmath275 agrees with @xmath276 .
it is well known that given a non - zero @xmath277 , the class @xmath56 is fibered if and only if @xmath278 is fibered .
note that given a non - trivial knot @xmath11 we have @xmath279 , and @xmath280 is fibered if and only if @xmath0 is fibered .
we refer to @xcite for background and more information .
we will need the following theorem , which in particular implies theorem [ thm : extendfree ] in the case that @xmath170 is hyperbolic .
[ thm : extendfreehyp ] let @xmath23 be a hyperbolic @xmath24-manifold and let @xmath126 be a properly embedded , connected thurston norm - minimizing surface that is not a fiber surface .
we write @xmath281 and we pick a base point @xmath282 on @xmath283 . then there exists a nontrivial element @xmath284 such that the subgroup of @xmath285 generated by @xmath286 and @xmath1 is the free product of @xmath286 and @xmath287 .
let @xmath23 be a hyperbolic @xmath24-manifold .
we denote by @xmath288 the boundary components of @xmath23 .
let @xmath126 be a properly embedded , connected thurston norm - minimizing surface that is not a fiber surface .
we write @xmath281 and we pick a base point @xmath282 on @xmath283 .
we now take all fundamental groups with respect to this base point .
it follows again from the loop theorem and the fact that @xmath126 is thurston norm - minimizing that the inclusion - induced map @xmath289 is a monomorphism .
we will henceforth view @xmath290 as a subgroup of @xmath291 .
we first suppose that @xmath126 hits all boundary components of @xmath23 .
since @xmath126 is not a fiber surface , it follows from the tameness theorem of agol @xcite and calegari
gabai @xcite that @xmath291 is word - hyperbolic and that @xmath290 is a quasi - convex subgroup of @xmath291 .
( we refer to ( * ? ? ?
* sections 14 and 16 ) for more details . )
it then follows from work of gromov @xcite ( see also ( * ? ? ?
* theorem 1 ) ) that there exists an element @xmath292 such that the subgroup of @xmath291 generated by @xmath105 and @xmath1 is in fact the free product of @xmath105 and @xmath287 .
we now suppose that there exists a boundary component @xmath293 that is not hit by @xmath126 .
we pick a path in @xmath144 connecting @xmath293 to the chosen base point and we henceforth view @xmath294 as a subgroup of @xmath291 .
note that @xmath295 is hyperbolic relative to the subgroups @xmath296 .
since @xmath126 is not a fiber surface , it follows from the tameness theorem and from work of hruska ( * ? ? ?
* corollary 1.3 ) that @xmath105 is a relatively quasi - convex subgroup of @xmath295 . since @xmath105 is a non - abelian surface group we can find an element @xmath297 such that @xmath298 is trivial .
we see again from the tameness theorem that @xmath299 is a relatively quasi - convex subgroup of @xmath295 . summarizing
, we have shown that @xmath207 and @xmath299 are two relatively quasi - convex subgroups of @xmath295 which have trivial intersection with the parabolic subgroup @xmath294 .
it now follows from martinez - pedroza ( * ? ? ?
* theorem 1.2 ) that there exists a @xmath300 such that the subgroup of @xmath295 generated by @xmath105 and @xmath301 is the free product of @xmath105 and @xmath302 .
the proposition now follows from the observation that according to our choices , both @xmath105 and @xmath302 lie in @xmath291
. we can now prove theorem [ thm : extendfree ] .
for the reader s convenience we recall the statement .
[ thm : infinitecyclic ] let @xmath0 be a non - fibered knot .
then there exists a seifert surface @xmath126 of minimal genus and a nontrivial element @xmath303 such that the subgroup generated by @xmath304 and @xmath1 is the free product of @xmath305 and the infinite cyclic group @xmath248 .
let @xmath0 be a non - fibered knot .
we write @xmath306 .
we denote by @xmath307 , the jsj components , and we denote by @xmath308 , the jsj tori of @xmath41 .
we let @xmath309 be the generator that corresponds to the canonical homomorphism @xmath310 . for each @xmath311 , we denote by @xmath312 the restriction of @xmath56 to @xmath313 .
the pair @xmath314 has a natural graph structure , since each jsj torus cobounds two jsj components .
since @xmath41 is a knot complement , this graph is a based tree , where the base is the vertex @xmath315 for which @xmath316 contains the boundary torus .
we now denote by @xmath317 the boundary torus of @xmath41 , and for each @xmath318 we denote by @xmath319 the unique jsj torus which is a boundary component of @xmath313 and which separates @xmath313 from @xmath316 .
there exists an element @xmath320 such that @xmath321 is hyperbolic and such that @xmath322 is not a fibered class .
we say that a vertex @xmath311 is non - fibered if @xmath312 is not a fibered class .
since @xmath323 is by assumption not fibered , it follows from ( * ? ? ?
* theorem 4.2 ) that some vertex is not fibered .
let @xmath320 be a non - fibered vertex of minimal distance to @xmath324 .
note that if @xmath311 is fibered and if @xmath325 is non - trivial , then the restriction of @xmath325 to any boundary torus is also non - trivial .
since @xmath326 is non - trivial and since @xmath320 is a non - fibered vertex of minimal distance to @xmath324 , we conclude that the restriction of @xmath327 to @xmath328 is non - trivial .
it follows from the geometrization theorem and from ( * ? ? ? * lemma vi.3.4 ) that @xmath321 is one of the following : the exterior of a torus knot ; a ` composing space ' , that is , a product @xmath329 , where @xmath330 is the result of removing @xmath149 open disjoint disks from @xmath331 ; a ` cable space ' , that is , a manifold obtained from a solid torus @xmath332 by removing an open regular neighborhood in @xmath333 ) of a simple closed curve @xmath334 that lies in a torus @xmath335 , where @xmath336 is a simple closed curve and @xmath334 is non - contractible in @xmath332 ; a hyperbolic manifold . as we argued above , the restriction of @xmath322 to one of the boundary tori , namely @xmath328 , is non - trivial . it is well known that in each of the first three cases , this would imply that @xmath327 is a fibered class
. hence @xmath321 must be hyperbolic .
this concludes the proof of the claim . in the following ,
given a vertex @xmath337 with @xmath325 non - zero , we denote by @xmath338 the divisibility of @xmath312 . for all other vertices
we write @xmath339 .
there exists a minimal genus seifert surface @xmath126 for @xmath0 with the following properties : @xmath126 intersects each @xmath340 transversally ; each intersection @xmath341 consists of a possibly empty union of parallel , non - null - homologous curves ; for each @xmath337 with @xmath342 the surface @xmath343 is the union of @xmath344 parallel copies of a surface @xmath345 . for each @xmath337 with @xmath342
we pick a properly embedded thurston norm - minimizing surface @xmath345 that represents @xmath346 .
after possibly gluing in annuli and disks , we may assume that at each boundary torus @xmath347 of @xmath348 , all the components of @xmath349 are parallel as oriented curves and no component of @xmath349 is null - homologous .
we now pick a tubular neighborhood @xmath350 $ ] of @xmath345 and we denote by @xmath351 the union of @xmath352 where @xmath353 with @xmath354 . for each @xmath337 with @xmath339
we denote by @xmath355 the empty set .
the surfaces @xmath351 are chosen such that at each jsj torus the boundary curves are parallel .
since at a jsj edge the adjacent surfaces have to represent the same homology class , at each jsj torus the adjacent surfaces have exactly the same number of boundary components which furthermore represent the same homology class in the jsj torus . after an isotopy in the neighborhood of the tori we can therefore glue the surfaces @xmath351 together to obtain a properly embedded surface @xmath126 .
since the thurston norm is linear on rays , it follows from ( * ? ? ? * proposition 3.5 ) that @xmath126 is a connected thurston norm - minimizing surface representing @xmath56 . by construction ,
the intersection of @xmath126 with @xmath356 consists of one curve , which is necessarily a longitude for @xmath0 .
we thus see that @xmath126 is indeed a genus - minimizing seifert surface for @xmath0 .
it is now clear that @xmath126 has the desired properties .
this concludes the proof of the claim .
recall that @xmath322 is not a fibered class . by the discussion at the beginning of this section
, this implies that @xmath357 is also not a fibered class , and so @xmath358 is not a fiber surface .
we pick a base point @xmath359 on @xmath360 , which is then also a base point for @xmath321 .
it follows from theorem [ thm : extendfreehyp ] that there exists an element @xmath361 such that the subgroup of @xmath362,p_w)$ ] generated by @xmath363 and @xmath1 is in fact the free product of @xmath363 and @xmath364 .
it now remains to prove the following claim .
the subgroup of @xmath365 generated by @xmath366 and @xmath1 is the free product of @xmath366 and @xmath364 .
we may pick an oriented simple closed curve @xmath334 in @xmath367 $ ] that intersects @xmath360 in precisely the base point @xmath359 and that represents @xmath368,p_w)$ ] . note that @xmath369 is precisely the free product of @xmath366 and @xmath364 .
it thus suffices to show that the inclusion - induced map @xmath370 is injective .
let @xmath371 be an element in the kernel of this map .
we pick a representative curve @xmath372 which intersects the jsj tori transversally .
we will show that @xmath371 represents the trivial element in @xmath369 by induction on @xmath373 if @xmath374
, then @xmath372 lies in the component of @xmath375 that contains @xmath359
. then @xmath334 lies completely in @xmath376 .
but the map @xmath377 is injective , and the map @xmath378 is also injective .
it thus follows that @xmath371 is the trivial element .
we now consider the case that @xmath379 .
we then think of @xmath380 as the fundamental group of the graph of groups @xmath381 .
we can view the curve @xmath372 as a concatenation of curves @xmath382 such that each curve @xmath383 lies completely in some @xmath384 . recall that we assume that @xmath372 represents the trivial element . a standard argument in the theory of fundamental groups of graph of groups ( see e.g. @xcite ) implies that there exists a @xmath383 with the following two properties : the two endpoints of @xmath383 lie on the same boundary torus @xmath347 of some @xmath384 , @xmath383 is homotopic in @xmath384 rel endpoints to a curve @xmath385 that lies completely in @xmath347 .
note that the two endpoints of @xmath383 lie on @xmath386 .
in fact we can prove a stronger statement .
the two endpoints of @xmath383 lie on the same component of @xmath386 .
we first make the following observation .
let @xmath387 be a properly oriented embedded surface @xmath387 in an oriented 3-manifold @xmath144 and let @xmath139 be an oriented embedded arc that does not intersect @xmath387 at the endpoints .
we can then associate to @xmath387 and @xmath139 the algebraic intersection number @xmath388 , which has in particular the following two properties : for any properly oriented embedded arc @xmath324 homotopic to @xmath139 rel base points we have @xmath389 , if @xmath139 lies completely in a boundary component @xmath5 of @xmath144 , then @xmath390 equals the algebraic intersection number of the oriented curve @xmath391 with the oriented arc @xmath139 in @xmath5 .
we now turn to the proof of the claim .
we first note that there exists a homeomorphism @xmath392 which is the identity on @xmath393 , which has the property that for any @xmath394 with @xmath395 and @xmath396 $ ] we have @xmath397 and which is isotopic to the identity on @xmath398 .
more informally , @xmath399 is a map that pushes everything on @xmath400 $ ] slightly to the left .
note that @xmath399 pushes everything on @xmath401 off @xmath401 .
furthermore , if @xmath402 , then the intersection of @xmath403 with @xmath404 is also empty . since @xmath385 and @xmath383 are homotopic rel base points and since @xmath399 is homotopic to the identity , the curves @xmath405 and @xmath406 are homotopic rel base points .
it follows from the above that @xmath407 .
but the latter is clearly zero , since @xmath406 does not intersect @xmath401 .
we now conclude that @xmath408 .
since the curves @xmath409 are all parallel it now follows that @xmath405 does not intersect @xmath410 at all . but
this means that the two endpoints of @xmath385 , and thus also the two endpoints of @xmath383 , have to lie on the same component of @xmath386 .
this concludes the proof of the claim .
we then make the following claim .
the curve @xmath383 is homotopic in @xmath384 rel end points to a curve @xmath411 that lies completely in @xmath386 . by the previous claim
we know that the two endpoints of @xmath383 lie on the same component of @xmath386 .
we denote the initial point of @xmath383 by @xmath412 , and the terminal point by @xmath413 .
we denote by @xmath399 the component of @xmath414 that contains @xmath412 .
we endow @xmath399 with an orientation .
note that @xmath399 is homologically essential on @xmath347 .
the curve @xmath399 thus defines a subsummand @xmath415 of @xmath416 .
we also pick a curve @xmath417 in @xmath418 from @xmath412 to @xmath413 .
the concatenation @xmath419 lies in @xmath347 , and also lies in @xmath420 .
the curve @xmath419 thus represents an element in @xmath421 .
but the group @xmath422 is free ( regardless of whether @xmath334 lies on the @xmath412-component of @xmath420 or not ) whereas @xmath416 .
the two groups thus intersect in an infinite cyclic subgroup .
furthermore , the intersection contains the subsummand @xmath415 .
it follows that the intersection equals @xmath415 .
in particular , @xmath423 is homotopic rel @xmath412 to @xmath424 for some @xmath22 .
it now follows that relative to the end points we have the following homotopies : @xmath425 but the curve @xmath426 lies completely in @xmath386 .
this concludes the proof of the claim .
we can thus replace @xmath427 by @xmath428 and push @xmath411 slightly into the adjacent jsj component of @xmath41 .
we have found a representative of @xmath371 of smaller length than @xmath372 .
the claim that @xmath371 represents the trivial element now follows by induction .
this concludes the proof that the subgroup of @xmath429,p_w)$ ] generated by @xmath430 and @xmath1 is the free product of @xmath366 and @xmath364 .
we are therefore done with the proof of theorem [ thm : infinitecyclic ] .
let @xmath0 be a knot . recall that we denote by @xmath120 the unique epimorphism that sends the oriented meridian to 1 .
stallings @xcite proved the following theorem .
[ thm : st62 ] if @xmath0 is not fibered , then @xmath135 is not finitely generated .
it follows from lemma [ lem : splitkerfg ] that if @xmath135 is finitely generated , then there exists precisely one group @xmath5 such that @xmath2 splits over @xmath5 . thus stalling s
theorem follows as a consequence of either theroem [ thm : fibsplit ] or theorem [ thm : splitfreelarge ] . on the other hand ,
a group @xmath4 with an epimorphism @xmath94 such that @xmath37 is not finitely generated may still split over a unique group .
the baumslag - solitar group , the semidirect product @xmath431 $ ] where @xmath432 acts on @xmath433 $ ] by multiplication by @xmath434 , has abelianization @xmath106 .
the kernel of the abelianization @xmath435 is the infinitely generated subgroup @xmath433 $ ] .
since every finitely generated subgroup of @xmath433 $ ] is isomorphic to @xmath106 , @xmath431 $ ] splits only over subgroups isomorphic to @xmath106 .
( in fact , any two splittings are easily seen to be strongly equivalent . )
this shows that the conclusions of theorems [ thm : fibsplit ] and [ thm : splitfreelarge ] are indeed stronger than the conclusion of theorem [ thm : st62 ] .
stallings s fibering criterion has been generalized in several other ways . for example , if @xmath0 is not fibered , then @xmath37 can be written neither as a descending nor as an ascending hnn - extension @xcite , @xmath37 admits uncountably many subgroups of finite index ( see ( * ? ? ?
* theorem 5.2 ) , @xcite and ( * ? ? ?
* theorem 3.4 ) ) , the pair @xmath128 has ` positive rank gradient ' ( see ( * ? ? ? * theorem 1.1 ) ) and @xmath135 admits a finite index subgroup which is not normally generated by finitely many elements ( see ( * ? ? ? * theorem 5.1 ) ) .
in this section we will prove theorem [ mainthm ] , i.e. we will show that if @xmath0 is a knot , then @xmath2 does not split over a group of rank less than @xmath16 .
we will first give a ` classical ' proof for genus - one knots before we provide the proof for all genera .
the main ingredients in the proof are two classical results from 3-manifold topology .
first , we recall the statement of the kneser conjecture , which was first proved by stallings @xcite in the closed case , and by heil @xcite in the bounded case . * _ ( kneser conjecture)_*[thm : kneserconj ] let @xmath23 be a @xmath24-manifold with incompressible boundary .
if there exists an isomorphism @xmath436 , then there exist compact , orientable @xmath24-manifolds @xmath437 and @xmath438 with @xmath439 , @xmath440 and @xmath441 . in the following ,
we say that a properly embedded 2-sided annulus @xmath6 in a 3-manifold @xmath23 is _ essential _ if the inclusion map @xmath442 induces a @xmath443-injection and if @xmath6 is not properly homotopic into @xmath444 . the second classical result we will use
is the following , which is a direct consequence of a theorem of waldhausen @xcite ( see corollary 1.2(i ) of @xcite ) .
let @xmath0 be a genus - one knot .
since @xmath0 is non - trivial , the loop theorem implies that @xmath445 is incompressible .
since knot complements are prime 3-manifolds , it now follows from the kneser conjecture that @xmath2 can not split over the trivial group , i.e. @xmath2 can not split over a free group of rank zero .
now suppose that @xmath446 is a non - trivial knot such that @xmath447 splits over a free group of rank one , that is , over a group isomorphic to @xmath106 . from theorem [ thm : annulus ]
we deduce that @xmath448 contains an essential , properly embedded , 2-sided annulus @xmath6 .
lemma 2 of @xcite ( an immediate consequence of @xcite ) implies that the knot @xmath446 is either a composite or a nontrivial cable knot .
if @xmath446 is a composite knot , then it follows from the additivity of the knot genus ( see , for example , @xcite ) that the genus of @xmath446 is at least two .
moreover , a well - known result of schubert @xcite ( see proposition 2.10 of @xcite ) implies that the genus of any cable knot is greater than one .
thus in both cases we see that @xmath449 .
we introduce the following convention .
if @xmath4 is a group and @xmath450 a representation over a ring , then we denote by @xmath451 also the @xmath106-linear extension of @xmath451 to a map @xmath452 \to m(k , r)$ ] .
furthermore , if @xmath6 is a matrix over @xmath452 $ ] then we denote by @xmath453 the matrix given by applying @xmath451 to each entry of @xmath6 .
let @xmath4 be a group , @xmath94 an epimorphism , and @xmath454 a representation .
first note that @xmath228 and @xmath56 give rise to a tensor representation @xmath455 now let @xmath456 be a presentation of @xmath4 . by adding trivial relations if necessary
, we may assume that @xmath457 .
we denote by @xmath458 the free group with generators @xmath459 .
given @xmath460 we denote by @xmath461\to \z[f_{k}]$ ] the fox derivative with respect to @xmath462 , i.e. the unique @xmath106-linear map such that @xmath463 for all @xmath464 and @xmath465 .
we denote by @xmath466 the @xmath467-matrix over @xmath452 $ ] of all the fox derivatives of the relators . given subsets @xmath468 and @xmath469 we denote by @xmath470 the matrix formed by deleting the columns @xmath471 and by deleting the rows @xmath472 of @xmath144 . note that there exists at least one @xmath473 such that @xmath474 .
it follows that @xmath475 we define @xmath476 ( note that each @xmath477 is a @xmath478-matrix . )
it is worth considering the special case that @xmath479 ; that is , the case of a presentation of deficiency one . then the only choice for @xmath446 is the empty set , and hence @xmath480 wada @xcite introduced the following invariant of the triple @xmath481 .
@xmath482 a priori , wada s invariant depends on the various choices we made . the following theorem proved by wada ( *
* theorem 1 ) shows that the indeterminacy is well controlled .
[ thm : wadadefined ] let @xmath4 be a group , let @xmath94 be an epimorphism , and let @xmath454 be a representation . then @xmath483 is well - defined up to multiplication by a factor of the form @xmath484 , where @xmath485 and @xmath486 .
finally , let @xmath11 be a knot and let @xmath487 be a representation .
as before , we denote by @xmath488 the epimorphism that sends the oriented meridian of @xmath0 to 1 .
we write @xmath489 if @xmath490 is the trivial one - dimensional representation , then wada s invariant is determined by the classical alexander polynomial @xmath491 .
more precisely , we have @xmath492 wada s invariant equals the twisted reidemeister torsion of a knot , and is closely related to the twisted alexander polynomial of a knot , which was first introduced by lin @xcite .
we refer to @xcite for more details about wada s invariant , its interpretation as twisted reidemeister torsion and its relationship to twisted alexander polynomials .
before we provide the proof of theorem [ mainthm ] we need to introduce two more definitions .
first , given a non - zero polynomial @xmath493 with @xmath494 and @xmath495 , we write @xmath496 if @xmath497 is a non - zero rational function , we write @xmath498 note that if wada s invariant of a triple @xmath481 is non - zero , then the degree of wada s invariant @xmath483 is well defined .
[ thm : technical ] let @xmath4 be a group and let @xmath499 be a splitting .
we denote by @xmath500 the canonical epimorphism which is given by @xmath51 and @xmath52 for @xmath53 . if @xmath501 is a representation such that @xmath502 , then @xmath503 in @xcite ( see also @xcite ) it was shown that if @xmath0 is a knot and @xmath504 is a representation such that @xmath505 , then [ equ : genuslowerbound ] _
k^k(2(k)-1).in light of the discussion in section [ section : splitk ] , we can view theorem [ thm : technical ] as a generalization of ( [ equ : genuslowerbound ] ) .
let @xmath4 be a group and let @xmath506 be a splitting , where @xmath50 is a monomorphism and @xmath5 is a rank-@xmath372 subgroup of @xmath507 .
we pick generators @xmath508 for @xmath5 .
note that @xmath509 we write @xmath510 . for any @xmath473 and @xmath516 we have @xmath517
the same argument as in ( 1 ) shows that the first term lies in @xmath515 $ ] , and one can similarly see that the second term is of the form @xmath518 , where @xmath519 $ ] .
thus @xmath520 , the matrix obtained from @xmath144 by deleting the @xmath521-st column , is of the form @xmath522 where @xmath412 and @xmath413 are matrices over @xmath515 $ ] , and where all but the last @xmath523 rows of @xmath413 are zero .
let @xmath501 be a representation and @xmath524 a subset with @xmath525 .
it follows from the above that @xmath526 where @xmath527 and @xmath528 are matrices over @xmath515 $ ] and where at most @xmath372 rows of @xmath528 are non - zero .
we then see that @xmath529 where at most @xmath530 rows of @xmath531 are non - zero .
if @xmath532 is non - zero , then it follows from an elementary argument that @xmath533 we now consider @xmath534 by the above , if @xmath535 , then @xmath536 .
the last ingredient in the proof of theorem [ mainthm ] is the following result from @xcite .
the proof of the theorem builds on the virtual fibering theorem of agol @xcite ( see also @xcite ) , which applies for knot complements by the work of liu @xcite , przytycki - wise @xcite and wise @xcite . in ( * ? ? ?
* theorem 1.2 ) an analogous statement is formulated for twisted reidemeister torsion instead of wada s invariant .
the theorem , as stated , now follows from the interpretation ( see , for example , @xcite ) of wada s invariant as twisted reidemeister torsion . note that @xmath544 is an epimorphism .
in particular , it sends the meridian to either @xmath545 or @xmath546 . by possibly changing the orientation of the knot
, we can assume that @xmath544 sends the meridian to @xmath545 .
by theorem [ thm : fv12a ] , there exists a representation @xmath539 such that @xmath547 and such that @xmath540 by definition , we have @xmath548 theorem [ thm : technical ] implies that @xmath549 s. friedl and s. vidussi , _ a survey of twisted alexander polynomials _ , the mathematics of knots : theory and application ( contributions in mathematical and computational sciences ) , editors : markus banagl and denis vogel ( 2010 ) , p. 45 - 94 .
k. johannson , _ on the mapping class group of simple 3-manifolds _ , topology of low - dimensional manifolds ( proc .
second sussex conf .
, chelwood gate , 1977 ) , lecture notes in math .
722 , springer , berlin , 1979 , 4866 .
|
let @xmath0 be a knot of genus @xmath1 .
if @xmath0 is fibered , then it is well known that the knot group @xmath2 splits only over a free group of rank @xmath3 .
we show that if @xmath0 is not fibered , then @xmath2 splits over non - free groups of arbitrarily large rank . furthermore ,
if @xmath0 is not fibered , then @xmath2 splits over every free group of rank at least @xmath3 .
however , @xmath2 can not split over a group of rank less than @xmath3 .
the last statement is proved using the recent results of agol , przytycki wise and wise .
|
since the first results of the _ kepler _ mission were published , the number of known planet candidates has enlarged tremendously .
most _ hot jupiters _ have been found in single planetary systems and it was believed that those kind of giant , close - in planets are not accompanied by other planets ( see e.g. * ? ? ?
this result has been obtained analysing 63 _ kepler _ hot jupiter candidates and is in good agreement with inward migration theories of massive outer planets , and planet planet scattering that could explain the lack of additional close planets in hot jupiter systems .
nonetheless , wide companions to hot jupiters have been found , as shown e.g. in @xcite for the hat - p-13 system .
one has to state , though , that the formation of hot jupiters is not yet fully understood ( see @xcite and references therein for some formation scenarios , and e.g. @xcite for possible tests ) .
recently @xcite reanalysed a larger sample of 159 _ kepler _ candidates and in some cases found dynamically induced _ transit timing variations ( ttvs)_. if the existence of additional planets in hot jupiter systems can be confirmed , planet formation and migration theories can be constrained .
since , according to @xcite , there is only a small fraction of hot jupiters believed to be part of a multiplanetary system , it is important to analyse those systems where an additional body is expected .
in contrast to e.g. the kepler mission , where a fixed field on the sky is monitored over a long time span , our ongoing study of ttvs in exoplanetary systems only performs follow - up observations of specific promising transiting planets where additional bodies are suspected .
the targets are selected by the following criteria : * the orbital solution of the known transiting planet shows non - zero eccentricity ( though the circularization time - scale is much shorter than the system age ) and/or deviant radial velocity ( rv ) data points both indicating a perturber . *
the brightness of the host star is @xmath6mag to ensure sufficient photometric and timing precision at 1 - 2 m telescopes . *
the target location on the sky is visible from the northern hemisphere .
* the transit depth is at least 10 mmag to ensure a significant detection at medium - sized , ground - based telescopes . *
the target has not been studied intensively for ttv signals before .
our observations make use of the yeti network ( young exoplanet transit initiative ; * ? ? ?
* ) , a worldwide network of small to medium sized telescopes mostly on the northern hemisphere dedicated to explore transiting planets in young open clusters . this way
, we can observe consecutive transits , which are needed to enhance the possibility to model ttvs as described in @xcite , and @xcite .
furthermore , we are able to obtain simultaneous transits observations to expose hidden systematics in the transit light curves , like time synchronization errors , or flat fielding errors . in the past , the transiting exoplanets wasp-12b @xcite , wasp-3b @xcite , wasp-10b @xcite , wasp-14b @xcite and tres-2 ( raetz et al . 2014 , submitted ) have been studied by our group in detail . in most cases , except for wasp-12b , no ttvs could be confirmed .
recently , also @xcite claimed to have found possible ttv signals around qatar-1 .
however , all possible variations should be treated with reasonable care . in this project
we monitor the transiting exoplanet hat - p-32b .
the g0v type @xcite host star hat - p-32 was found to harbour a transiting exoplanet with a period of @xmath7d by @xcite .
having a host star brightness of @xmath8mag and a planetary transit depth of @xmath4mmag the sensitivity of medium - sized telescopes is sufficient to achieve high timing precision , therefore it is an optimal target for the yeti telescopes .
the rv signal of hat - p-32 is dominated by high jitter of @xmath9ms@xmath10 .
@xcite claim that a possible cause of the jitter is the presence of one or more additional planets. @xcite also analysed the rv signature of hat - p-32 and found a long term trend indicating a companion with a minimum mass of @xmath11m@xmath12 at separations of @xmath13au .
however , such a companion could not yet explain the short time - scale jitter as seen in the hartman data .
besides the circular orbit fit , an eccentric solution with @xmath14 also fits the observed data .
though @xcite mention that the probability of a real non - zero eccentricity is only @xmath15 , it could be mimicked or triggered by a second body in the system .
thus , hat - p-32b is an ideal candidate for further monitoring to look for transit timing variations induced by a planetary companion .
between 2011 october and 2013 january we performed 30 complete and 15 partial transit observations ( see tables [ tab : h32completeobservations ] and [ tab : h32partialobservations ] ) from 10 different observatories : our own one in jena , as well as from telescopes at torun ( poland ) , trebur ( germany ) , gettysburg and swarthmore ( usa ) , tenerife and sierra nevada ( spain ) , antalya and ankara ( turkey ) , and rozhen ( bulgaria ) mostly throughout the yeti network @xcite .
in addition , three literature data points from @xcite , and two observations from @xcite are available .
the telescopes and abbreviations used hereafter are summarized in table [ tab : h32telescopes ] , a short description of each observing site can be found below , sorted by the number of observations . [ cols="<,<,<,<,^",options="header " , ] we presented our observations of hat - p-32b planetary transits obtained during a timespan of 24 months ( 2011 october until 2013 october ) .
the data were collected using telescopes all over the world , mainly throughout the yeti network .
out of 44 started observations we obtained 24 light curves that could be used for further analysis .
21 light curves have been obtained that could not be used due to different reasons , mostly bad weather .
in addition to our data , literature data from @xcite , @xcite , and @xcite were also taken into account ( see fig .
[ fig : h32oc ] , [ fig : h32ar ] , [ fig : h32k ] , and [ fig : h32i ] and table [ tab : fitresults ] ) . the published system parameters @xmath3 , @xmath2 and @xmath1 from the circular orbit fit of @xcite were confirmed . in case of the semimajor axis over the stellar radius and the inclination
, we were able to improve the results due to the number of observations .
as for the planet - to - star radius ratio , we did not achieve a better solution for there is a spread in the data making constant fits difficult .
in addition , @xcite found an m - dwarf @xmath16 away from hat - p-32 and hence a possible cause for this spread .
regarding the transit timing , a redetermination of the planetary ephemeris by @xmath4ms can explain the obtained mid - transit times , although there are still some outliers .
of course , having 1@xmath17 error bars , one would expect some of the data points to be off the fit .
nevertheless , due to the spread of data seen in the o c diagram , observations are planned to further monitor hat - p-32b transits using the yeti network .
this spread in the order of @xmath5min does not exclude certain system configurations .
assuming circular orbits even an earth mass perturber in a mean motion resonance could still produce such a signal .
ms would like to thank the referee for the helpful comments on the paper draft .
all the participating observatories appreciate the logistic and financial support of their institutions and in particular their technical workshops .
ms would also like to thank all participating yeti telescopes for their observations .
mmh , jgs , ap , and rn would like to thank the deutsche forschungsgemeinschaft ( dfg ) for support in the collaborative research center sonderforschungsbereich sfb tr 7 `` gravitationswellenastronomie '' .
re , mk , and rn would like to thank the dfg for support in the priority programme spp 1385 on the _ first ten million years of the solar system _ in projects ne 515/34 - 1 & -2 .
gm and dp acknowledge the financial support from the polish ministry of science and higher education through the luventus plus grant ip2011 031971 .
rn would like to acknowledge financial support from the thuringian government ( b 515 - 07010 ) for the stk ccd camera ( jena 0.6 m ) used in this project .
the research of dd and dk was supported partly by funds of projects do 02 - 362 , do 02 - 85 and ddvu 02/40 - 2010 of the bulgarian scientific foundation , as well as project rd-08 - 261 of shumen university .
we also wish to thank the tbitak national observatory ( tug ) for supporting this work through project number 12bt100 - 324 - 0 using the t100 telescope .
99 bakos g. . , et al . , 2009 , apj , 707 , 446 broeg c. , fernndez m. , neuhuser r. , 2005 , an , 326 , 134 carter j. a. , winn j. n. , 2009 , apj , 704 , 51 carter j. a. , winn j. n. , 2010 , apj , 716 , 850 chambers j. e. , 1999 , mnras , 304 , 793 claret 2000 , a&a 363 , 1081 eastman j. , gaudi b. s. , agol e. , 2013 , pasp , 125 , 83 eastman j. , siverd r. , gaudi b. s. , 2010 , pasp , 122
, 935 etzel p. b. , 1981 , psbs.conf , 111 fulton b. j. , shporer a. , winn j. n. , holman m. j. , pl a. , gazak j. z. , 2011 , aj , 142 , 84 gazak j. z. , johnson j. a. , tonry j. , dragomir d. , eastman j. , mann a. w. , agol e. , 2012 , adast , 2012 , 30 gibson n. p. , aigrain s. , barstow j. k. , evans t. m. , fletcher l. n. , irwin p. g. j. , 2013 , arxiv , arxiv:1309.6998 hartman j. d. , et al . , 2011 ,
apj , 742 , 59 hoyer s. , rojo p. , lpez - morales m. , 2012 , apj , 748 , 22 kipping d. m. , 2009 , mnras , 392 , 181 kipping d. m. , 2010 , mnras , 408 , 1758 knutson h. a. , et al . , 2013 , arxiv , arxiv:1312.2954 lloyd j. p. , et al . , 2013 , arxiv , arxiv:1309.1520 maciejewski g. , et al . , 2010 , mnras , 407 , 2625 maciejewski g. , errmann r. , raetz s. , seeliger m. , spaleniak i. , neuhuser r. , 2011 , a&a , 528 , a65 maciejewski g. , et al . , 2011 , mnras , 411 , 1204 maciejewski g. , et al . , 2011 , a&a , 535 , 7 maciejewski g. , et al .
, 2013a , a&a , 551 , a108 maciejewski g. , et al .
, 2013b , aj , 146 , 147 mandel k. , agol e. , 2002 , apj , 580 , l171 meschiari s. , wolf a. s. , rivera e. , laughlin g. , vogt s. , butler p. , 2009 ,
pasp , 121 , 1016 mugrauer m. , berthold t. , 2010 , an , 331 , 449 neuhuser r. , et al . , 2011 , an , 332 , 547 nesvorn d. , morbidelli a. , 2008 , apj , 688 , 636 payne m. j. , ford e. b. , 2011 , apj , 729 , 98 pickles a. , depagne . , 2010 , pasp , 122 , 1437 popper d. m. , etzel p. b. , 1981 , aj , 86 , 102 raetz st . , 2012 , phd thesis university jena sada p. v. , et al . , 2012 , pasp , 124 , 212 salaris m. , chieffi a. , straniero o. , 1993 , apj , 414 , 580 southworth j. , et al . , 2009 , mnras , 396 , 1023 southworth j. , 2008 , mnras , 386 , 1644 steffen j. h. , et al . , 2012 , pnas , 109 , 7982 szab r. , szab g. m. , dlya g. , simon a. e. , hodosn g. , kiss l. l. , 2013 , a&a , 553 , a17 von essen c. , schrter s. , agol e. , schmitt j. h. m. m. , 2013 , a&a , 555 , a92 weber m. , granzer t. , strassmeier k. g. , 2012 , spie , 8451 ,
|
we present the results of 45 transit observations obtained for the transiting exoplanet hat - p-32b .
the transits have been observed using several telescopes mainly throughout the yeti network . in 25 cases , complete
transit light curves with a timing precision better than @xmath0min have been obtained .
these light curves have been used to refine the system properties , namely inclination @xmath1 , planet - to - star radius ratio @xmath2 , and the ratio between the semimajor axis and the stellar radius @xmath3 .
first analyses by @xcite suggest the existence of a second planet in the system , thus we tried to find an additional body using the transit timing variation ( ttv ) technique .
taking also literature data points into account , we can explain all mid - transit times by refining the linear ephemeris by @xmath4ms .
thus we can exclude ttv amplitudes of more than @xmath5min .
[ firstpage ] stars : individual : hat - p-32 planets and satellites : individual : hat - p-32b planetary systems
|
osiris ( optical system for imaging and low / intermediate - resolution integrated spectroscopy ) is an instrument designed to obtain images and low - resolution spectra in the optical domain ( from 3650 through 10,000 ) .
it will be installed at the nasmyth focus of the 10 m gtc telescope @xcite at la palma observatory , although it can be also mounted at the cassegrain focus .
a detailed account of the observing modes , optical design , and performance is given in @xcite .
osiris is a versatile instrument offering many observing modes : ( a ) direct imaging with broad - band and narrow - band using tunable filters , ( b ) narrow - band with charge shuffling , ( c ) fast photometry , ( d ) long - slit spectroscopy , ( e ) multi - object spectroscopy , ( f ) nod - and - shuffle multi - object spectroscopy , and ( g ) fast spectroscopy .
the main driver of the optical and mechanical design of osiris has been the inclusion of tunable filters ( tf ) in order to obtain narrow - band imaging at an arbitrary wavelength with effective resolutions of around 5 - 60 . in the spectroscopic modes
, osiris provides spectral resolutions @xmath0 of 250 , 500 , 1500 , and 2500 covering the full spectral range from 3650 to 10,000 by using 11 grisms .
grisms can be rotated by @xmath1 to change the dispersion direction and allow for fast spectroscopic observations .
the detector consists of a mosaic of two 2k@xmath24k marconi mat-44 - 82 ccds of 15@xmath3m / pixel providing a large field of view ( @xmath4 ) and a pixel size of @xmath5 .
there will be a gap of 8.3 arcsec parallel to the normal dispersion direction .
osiris will be the first instrument using the technique of tunable filters in large telescopes .
the potential of this observational mode has been described widely in the literature ( e.g. @xcite ; @xcite ; @xcite ; @xcite ) . the next observational mode in complexity is the multi - object spectroscopy ( mos ) .
this technique arrived at its maturity in the eighties , increasing significantly the number of objects that can be observed simultaneously .
together with advances in optical design , new coatings for the camera elements , large - format detectors , and new technologies in grisms and diffraction gratings , their use in large telescopes is opening new windows in astronomical research , especially in the developments of large scale surveys .
recently , the use of the technique of nod - and - shuffle allows to obtain spectra with optimal sky subtraction , making possible to observe deeper @xcite .
most of the observing modes need to block part of the fov to allow either charge shuffling or slit and multislit spectroscopy .
this blocking is done by using a mask located at the telescope focal plane .
masks are fabricated by cutting the necessary holes or slits in aluminium plates which are of @xmath6 mm in size , being the size of the imaging field of @xmath7 mm .
in standard mos mode , the number of slits ( 8 arcsec long ) that can fit in a single mask is around 60 and 100 at the highest and lowest spectral resolution , respectively . in this paper
we describe the software that has been developed to assist the observer to design osiris masks .
the osiris mask system is the system in charge of storing and positioning masks and it is fully described in @xcite .
figure 1 shows a general view of the mask system .
a cassette provides storage for 13 masks which are positioned at the focal plane by a mask loader mechanism .
the position accuracy of the masks over the telescope focal plane is of the order of 0.01 mm .
masks are slightly curved in cylindrical form in order to adapt to the geometry of the focal plane .
masks will be tooled in a flat plate and afterwards bent to fit the cylindrical shape over the mask frame .
for each 10m - class telescope where an instrument with mos capabilities is operating , a specific software for mask design has been developed .
examples are the fors instrument mask simulator @xcite , the vimos mask preparation software @xcite , and the mask design program of focas @xcite .
our software has been designed to provide similar capabilities to these programs , adapted to the special needs of osiris .
the mask designer is a subsystem of the osiris observing program manager which is , in turn , integrated into the gtc control system @xcite .
the software has been designed using unified modelling language ( uml ) and coded in java programming language .
the osiris mask designer is a plug - in application of the jskycat package . in this way
, the software uses classes from jskycat to manage fits files , to perform coordinate transformations , and for graphical user interface .
the mask designer performs all the calculations needed to transform sky ( ra and dec ) and ccd ( @xmath8 ) coordinates to coordinates over the mask .
the final output of the mask designer is therefore a set of instructions for the machine in charge of cutting the holes in the mask plate .
this means that accurate coordinate transformations are needed to convert from sky to telescope focal plane positions , and from ccd to focal plane positions .
also , atmospheric refraction and thermal expansion corrections should be taken into account .
figure 2 shows all the coordinate transformations needed to correct a sky position onto a ccd position which include : ( 1 ) atmospheric refraction , ( 2 ) gnomonic projection and geometrical distortion of the telescope focal plane , ( 3 ) transformation from telescope focal plane onto mask surface , ( 4 ) mapping from mask to ccd plane . atmospheric refraction is computed following @xcite and depends on wavelength , air pressure and temperature , as well as on zenith distance of the targets .
transformations between different surfaces are made using 5th - order polynomials whose coefficients will be obtained in the instrument commissioning phase . taking into account the errors due to mask fabrication , mechanical flexures , thermal expansion , astrometric accuracy , and the fact that the mask shape does not match exactly the focal plane , we expect that the position of the slits over the targets ( after acquiring the field using fiducial stars ) , will be better than a 6% of the slit width . for a nominal
slit of 1.4 arcsec , this corresponds to a positional accuracy of 0.08 arcsec .
two types of masks can be generated depending of the osiris observing mode . in the standard mos mode
, slits can be located at any position over the physical area of the mask .
the second observing mode is nod and shuffle which allows astronomers to subtract the sky emission by nodding the telescope between targets and sky while simultaneously ccd charge is shuffled up and down ( e.g. @xcite ) . in this mode
slits can be smaller than in standard mos mode .
the system performs the necessary checks for mask integrity , slit collision and spectra overlapping . in case of overlapping slits or spectra
the software can assign slits to additional masks using a priority - based algorithm .
the software also takes into account the gap in the detector plane and the edges of the mask plate which are disabled by the mask frame that holds the mask plate .
figure 3 is a diagram showing the high - level classes in which the mask designer system has been decomposed .
these classes ( userinputhandler , filehander , and maskhandler ) are in turn decomposed into individual classes .
the class osiris_mask_designer represents the whole system and contains all the functionality to fully produce masks .
composite class userinputhandler holds the functionality for handling the user interaction with the system .
filehander class is in charge of all actions regarding files .
the maskhandler composite class contains the algorithms for processing the slit data , transforming coordinates , and producing valid mask configurations .
its decomposition diagram is shown in fig .
3 . algorithms for processing the slits and producing valid mask configurations belong to classes slitprocessor and maskprocessor .
classes skytomaskmapping and ccd_tomaskmapping contain the functions needed to transform coordinates between the different systems involved .
not all the operations are shown in this diagram as all classes are decomposed into many simpler functions .
the mask designer has been designed as a plug - in of jskycat application , which contains many other classes and packages required by the system .
the most relevant jsky utilities and classes used by the osiris mask designer are : wcstransform to convert between image pixel and wcs ( world coordinate system ) coordinates ; fitscodec for handling of fits files ; gui classes for image display and related tasks .
the osiris mask designer is launched by starting the jskycat application and therefore , its functionality initially is identical to that of jskycat .
all operations are invoked through the main jskycat menu bar ( see fig .
4 ) . for the different parameters related to a particular observation , such as date , hour angle ,
central coordinates of the field , grism , position of the grism ( 0 or 90 degrees for normal or fast modes , respectively ) , position angle of the mask onto the sky , expected observation temperature and pressure , expected fabrication temperature , maximum number of masks per pointing , the system implements a configure observation menu to read an xml file ( observation configuration file ) .
other xml configuration files are needed by the system but they are not user configurable .
these contain telescope and instrument parameters and information relevant to computations , such as the coefficients for coordinate transformations and grism properties .
the main advantage of using xml files is that full independence from target platform is guaranteed .
the software can process slits defined interactively by using an image displayed on the screen , via cursor .
this image should either have absolute astrometry information on its header or it can be a raw frame taken with osiris in broad - band mode . alternatively , the user can provide a list of coordinates and slit properties ( type , width , priority ) that can be , in turn , either absolute equatorial coordinates or pixel coordinates from a raw broad - band image taken with osiris .
these options allow processing object lists obtained with , for instance , image detection software .
the mask designer can be run in four different modes : * defining slit positions interactively using a raw image taken with osiris * defining slit positions interactively using an image containing absolute astrometric information in its header * using a list of ra and dec coordinates * using a list of @xmath9 , @xmath10 positions from a raw image taken with osiris three different types of slits can be defined : ( a ) rectangular , ( b ) circular , and ( c ) curved slits .
circular slits are used mainly ( but not exclusively ) to define fiducial holes .
a minimum of three fiducial slits must be defined , normally to include bright stars in the field , in order to acquire the field .
the type of the slit is defined by launching the slit window ( see fig . 5 ) and can be changed at any moment during the process of defining the slits .
also , the slit dimensions can be changed by selecting a slit already defined .
slits can be deleted either one by one or all at once .
a special feature of this software is the ability of defining curved slits ( available only in interactive processing ) .
these are arc - shaped slits and may be useful for taking spectra of special targets such as , for instance , galaxy arcs produced by gravitational lensing .
integrity checks are made to avoid placing slits too close to mask borders or to other slits which may cause the mask to break at the fabrication process .
once all the slits are successfully configured the system processes the information .
several masks can be generated up to a maximum value specified by the user .
the system will try to place as many slits as possible in the first mask .
slits are placed in additional masks depending on their priority , which is set by the user .
overlapping slits or slits producing integrity conflicts are placed onto the second mask and so forth until the maximum number of masks is reached .
a residual mask is produced containing conflicting or out - of - bounds slits .
this mask will not be sent to fabrication , being its only purpose to inform the user .
conflicting and out - of - bounds slits are highlighted on the display using a color code to help the user to re - design the mask .
the output consists of several files containing all the information required to build the masks .
the user can save the work after masks have been processed .
saved files can be retrieved for later processing if needed .
there are three output files , two of them in fits format and the third in a format that will be read by the cutting machine software .
a fits file called odf ( object definition file ) includes all the slits defined by the user , including invalid ones .
the file contains the observation and telescope data , and , for all the generated masks , slit type and slit positions .
a separate storage is provided in the odf file to keep the fiducial slit information .
this file can be handled by the mask designer to allow the user to modify the design in new sessions .
a second fits file , called mdf ( mask definition file ) , is generated for each mask .
there is one file per mask containing only the slit information corresponding to that mask .
a mdf can be saved also for the residual mask , although in that case , fiducial slits information is lost .
the third file is the so - called mask cutter file and contains instructions to be read by the cutting machine to fabricate physically the masks .
the user has the option of printing a graphic view of each mask , containing the defined slits , the fiducial holes and , if wanted , rectangles marking where the spectra will appear in the image .
an example is shown in figure 6 .
we have designed and developed an application to design focal plane multislit masks for the instrument osiris . its functionality has been tested and the user requirements have been achieved .
the software takes into account all the effects that could affect an accurate slit positioning over the mask plates , and allows different input scenarios to design osiris masks .
it has been designed as a jsky plug - in offering an easy user interface . only during the instrument commissioning phase
will be possible to test the software in real conditions .
a critical issue is the fine tuning of the parameters involved in coordinate transformations . in particular ,
the coefficients to convert from telescope focal plane to detector and from sky to telescope focal plane will be obtained using pin - hole masks and observations of astrometric fields .
commisioning phase for osiris at the gtc is planned to start in 2006 .
figure 1 has been kindly provided by lorenzo peraza .
this work has been supported by the spanish plan nacional de astronoma y astrofsica under grants aya2000 - 2688-e , aya2002 - 03326 , and aya2002 - 01379 .
|
osiris is a day one instrument that will be available at the 10 m gtc telescope which is being built at la palma observatory in the canary islands .
this optical instrument is designed to obtain wide - field narrow - band images using tunable filters and to do low - resolution spectroscopy in both long - slit and multislit modes . for the multislit spectroscopy mode ,
we have developed a software to assist the observers to design focal plane masks . in this paper
we describe the characteristics of this mask designer tool .
we discuss the main design concepts , the functionality and particular features of the software .
|
active galactic nuclei ( agns ) emit intense x - ray radiation by converting the gravitational energy of matter accreted onto the supermassive black holes ( smbhs ) . according to the so - called agn unified model @xcite , the central engine is obscured by a dusty torus , which affects the observed x - ray spectrum via photoelectric absorption and compton reflection . in addition , a relativistically blurred reflection component from the innermost accretion disk
would be expected when the disk is extended down to a vicinity of the smbh .
hence , observations of broadband x - ray spectra of agns are useful to investigate their structure including the torus and accretion disk .
low - luminosity agns ( llagns ) , which we define by their low x - ray luminosities in the 14195 kev band ( @xmath10 erg s@xmath2 ) , are important objects to understand the evolution of agns .
llagns contain two extreme types of agns : those with a small smbh mass and a high mass - accretion rate ( i.e. , with a high eddington ratio ) , and those with a large smbh mass and a low mass - accretion rate ( a low eddington ratio ) .
the former type is expected in the early growing phase of a smbh , whereas the latter corresponds to the fading phase of an agn into a quiescent smbh .
broadband x - ray studies of llagns have been limited , however , because of their low fluxes even in the local universe ( @xcite ) .
theories and observations suggest that llagns are not a simple scaled - down version of luminous agns in terms of their nucleus structure .
when the accretion rate falls below a critical level , it is predicted that the state of the inner accretion disk changes from the standard accretion disk @xcite to an optically - thin radiatively inefficient accretion flow ( riaf ; @xcite ; @xcite ) .
studies of spectral energy distributions ( seds ) from the radio to x - ray bands showed that the standard disk is truncated at a radius much larger than a few schwarzschild radii ( e.g. , @xcite ; @xcite ) .
this is supported by the lack of the big blue bump in llagns , which arises from the thermal emission from the inner standard disk ( e.g. , @xcite ) . moreover , most of llagns do not show a broad iron - k@xmath5 line feature , suggesting that the standard disk does not extend to the inner most region around the smbh , except for some llagns ( e.g. , ngc 4051 ; @xcite ) .
also , @xcite found a negative correlation between the x - ray photon index and eddington ratio for low eddington ratio agns ( @xmath11 ) , whereas the positive one for luminous agns was reported by @xcite .
@xcite suggested that the negative correlation can be explained with a riaf model .
for the detailed investigations of the broad iron line feature and continuum components , broadband x - ray spectra are highly important .
x - ray surveys showed that the absorbed agn fraction , which reflects the torus covering factor , decreases with x - ray luminosity ( e.g. , @xcite , 2014 ; @xcite ; @xcite ; @xcite ; @xcite ) .
these results would be consistent with receding torus models ( e.g. , @xcite ) , where the inner wall of the torus recedes with luminosity by keeping its scale - height more constant .
however , recent surveys have found evidence that the absorbed agn fraction decreases towards lower x - ray luminosities from a peak around @xmath12 erg s@xmath2 ( e.g. , @xcite ; @xcite ; @xcite ) .
these facts suggest that , at low luminosities , other physical mechanisms than agn radiation must be responsible to determine the torus structure .
main goals of this paper are ( 1 ) to best constrain the x - ray spectral properties of llagns by taking advantage of their `` broadband '' ( 0.5200 kev ) data observed with _
suzaku _ and _ swift_/bat , ( 2 ) to reveal the nuclear structure including the torus and accretion disk , and ( 3 ) to find key parameters that determine them .
most of previous studies are limited to soft energy bands below @xmath1310 kev ( e.g. , @xcite ) . from the _ swift_/bat 70-month catalog @xcite , we select all ten llagns ( @xmath14 in the 14195 kev band ) whose _ suzaku _ archival data are public .
we exclude ngc 4051 , which is known to exhibit large spectral variation among different epochs ( e.g. , @xcite ; @xcite ; @xcite ) .
this paper is organized as follows .
section [ sec : obs ] describes the details of the observation and the overview of the data .
we show the results of the spectral analysis in section [ sec : spec_ana ] .
the discussion and conclusions are presented in sections [ sec : dis_con ] and [ sec : con ] , respectively . unless otherwise noted , all errors are quoted at the 90% confidence level for a single parameter of interest .
ccccccccc ngc 2655 & j0856.0 + 7812 & seyfert 2 & 08h55m37.7s & + 78d13m23s & 4.670@xmath15 & 24.4 & 7.7@xmath16 & ( 1,1 ) + ngc 3718 & j1132.7 + 5301 & xbong & 11h32m34.9s & + 53d04m05s & 3.312@xmath15 & 17.0 & 8.0@xmath17 & ( 1,2 ) + ngc 3998 & j1157.8 + 5529 & seyfert 1/liner & 11h57m56.1s & + 55d27m13s & 3.496@xmath15 & 19.4 & 8.9@xmath18 & ( 2,3 ) + ngc 4102 & j1206.2 + 5243 & liner & 12h06m23.0s & + 52d42m40s & 2.823@xmath15 & 19.1 & 7.9@xmath16 & ( 3,1 ) + ngc 4138 & j1209.4 + 4340 &
seyfert 1.9 & 12h09m29.8s & + 43d41m07s & 2.962@xmath15 & 15.9 & 7.2@xmath16 & ( 4,1 ) + ngc 4258 & j1219.4 + 4720 & seyfert 1.9/liner & 12h18m57.5s & + 47d18m14s & 1.494@xmath15 & 7.6 & 7.61@xmath19 & ( 5,4 ) + ngc 4395 & j1202.5 + 3332 & seyfert 1.9 & 12h25m48.9s & + 33d32m49s & 1.064@xmath15 & 3.9 & 5.6@xmath20 & ( 3,5 ) + ngc 4941 & j1304.3 - 0532 & seyfert 2 & 13h04m13.1s & -05d33m06s & 3.696@xmath15 & 19.7 & 6.9@xmath17 & ( 2,6 ) + ngc 5273 & j1341.9 + 3537 & seyfert 1.9 & 13h42m08.3s & + 35d39m15s & 3.619@xmath15 & 13.1 & 6.7@xmath16 & ( 2,7 ) + ngc 5643 & j1432.8 - 4412 & seyfert 2 & 14h32m40.7s & -44d10m28s & 3.999@xmath15 & 16.9 & 6.3@xmath21 & ( 1,6 )
the basic information of our targets is summarized in table [ tab : info_srcs ] .
the distances are taken from the literature or are calculated from the redshift .
the black hole masses are compiled from the literature or are calculated from the stellar velocity dispersion by using the relation of @xcite .
@xcite observations of our targets were performed with exposures of @xmath13 4080 ksec .
table [ tab : info_obs ] gives the observation log .
suzaku _ carries x - ray ccd cameras that cover the energy range below @xmath13 10 kev called the x - ray imaging spectrometer ( xis ) , and a non - imaging instrument sensitive to hard x - rays above @xmath13 15 kev called the hard x - ray detector ( hxd ) .
two front - illuminated xiss ( fi - xiss ; xis-0 and xis-3 ) and one back - illuminated xis ( bi - xis ; xis-1 ) were available during the observations .
all targets were too faint above @xmath13 50 kev to be detected with hxd / gso .
hxd / pin was not operated during the observation of ngc 2655 , and no hxd / pin data of ngc 4138 were available because of a telemetry saturation problem .
ngc 2655 , ngc 3998 , ngc 4258 , ngc 4941 , and ngc 5273 were observed at the xis nominal position , whereas the others were observed at the hxd nominal position .
ccccccc ngc 2655 & 2014 may 12 & 709003010 + ngc 3718 & 2009 oct .
24 & 704048010 + ngc 3998 & 2014 may 03 & 709002010 + ngc 4102 & 2009 may 30 & 704057010 + ngc 4138 & 2009 nov .
02 & 704047010 + ngc 4258 & 2010 nov .
11 & 705051010 + ngc 4395 & 2007 june 02 & 702001010 + ngc 4941 & 2012 june 22 & 707001010 + ngc 5273 & 2013 july 16 & 708001010 + ngc 5643 & 2007 aug .
19 & 702010010 ftools ( version 6.15.1 ) and xspec ( version 12.8.1.g ) are used for the data reduction and spectral analysis .
the caldb released on 2015 jan 5 are utilized .
we reprocess the xis event files according to the abc guide and extract source events from a circular region centered on the target centroid .
the background events are taken from an off - source region within the xis field - of - view .
the fi - xiss data are merged to increase the signal - to - noise ratio .
we utilize the cleaned event files of hxd / pin provided by the _
suzaku _ team . to reproduce the background spectra
, we utilize the `` tuned '' non x - ray background model @xcite and the cosmic x - ray background spectrum simulated by assuming the model given in @xcite . from eight targets except ngc 2655 and ngc 4138
, we detect significant signals with hxd / pin well above the systematic uncertainties in the non x - ray background model ( 1% for the exposure time above 40 ksec ; @xcite ) . in the spectral analysis
, we also use the _ swift_/bat spectra averaged for 70 months @xcite .
+ + + + figure [ fig : lc ] shows the background - subtracted light curves of our sample in the 210 kev band ( xis-0+xis-3 ) .
the bin size is set to 5760 sec ( the orbital period of _ suzaku _ ) to exclude any modulations that depend on the orbital phase .
strong time variability is seen in ngc 4395 and ngc 5273 . to quantify it
, we calculate the mean fractional variation ( @xmath22 ) defined in equation ( 3 ) of @xcite : @xmath23 where @xmath24 and @xmath25 here , for the sample size @xmath26 , @xmath27 and @xmath28 is an observed flux and the mean value , respectively , while @xmath29 an error .
larger @xmath22 means that the object is more variable .
also , by fitting a constant flux model to the light curves , we calculate the probability of obtaining the observed data ( p - value ) based on chi - square distribution . @xmath22 and p - values of the 210 kev light curves
are summarized in table [ tab : info_val ] .
judging from @xmath30 and p - value @xmath31 , ngc 4395 and ngc 5273 were highly variable .
this would reflect the small spatial scale of the x - ray emitting region inferred from their low smbh masses ( @xmath32 , where @xmath33 is the solar mass ) .
although the smbh mass of ngc 5643 is also low , its variability may be suppressed due to heavy obscuration of the direct transmitted emission ( @xcite ; @xcite ) . as we will discuss in section [ sec : broad_ana_2 ] , in these low - mass llagns
the flux of the reflection component from the torus follows that of the primary emission on time scales shorter than 70 months .
p3cmp2cmcccc ngc 2655 & 0.064 & 0.02 + ngc 3718 & 0.031 & 0.08 + ngc 3998 & 0.003 & 0.40 + ngc
4102 & 0.138 & 0.07 + ngc 4138 & 0.040 & 0.24 + ngc 4258 & 0.083 & @xmath31 + ngc 4395 & 0.424 & @xmath31 + ngc 4941 & 0.090 & 0.02 + ngc 5273 & 0.303 & @xmath31 + ngc 5643 & 0.106 & 0.06
+ + + + + we simultaneously analyze the x - ray spectra of the fi - xiss , bi - xis , hxd / pin , and _
swift_/bat , which cover the 112 kev , 0.58 kev , 1460 kev , and 15200 kev bands , respectively .
we exclude the 1.71.9 kev band of the xis spectra to avoid systematic uncertainties of the energy response around the si k - edge region .
the cross - normalization factor of the hxd / pin spectrum to the fi - xiss one is set to 1.16 ( 1.18 ) for the xis ( hxd ) nominal position observation , and that of the bi - xis spectrum is allowed to vary . on the basis of the previous works ( @xcite ; @xcite ) , we start with a base - line model , @xmath34 in the xspec terminology .
this model includes absorbed primary x - ray emission ( i.e. , a cut - off power law ) , a scattered component , and a reflection continuum from distant cold matter accompanied with a narrow iron - k@xmath5 line .
optically - thin thermal emission from the host galaxy ( apec in xspec ) and other emission / absorption lines ( zgauss ) are also added if they are significantly required with a confidence level above 90% in terms of @xmath35 . because it is difficult to determine the cut - off energy from our data
, we fix it at 300 kev , a typical value measured in nearby agns @xcite . the first constant factor , @xmath36 ,
is applied to the primary power - law component in the _ suzaku _ spectra to absorb possible time variability between the _ suzaku _ ( one epoch ) and _ swift_/bat ( averaged for 70 months ) observations .
the second constant term represents the scattered fraction , @xmath37 . as a reflection component from the torus
, we employ the pexrav model , which calculates a reflected spectrum from an optically thick slab with a solid angle of @xmath38 irradiated by a point source @xcite .
we set the reflection strength , @xmath39 , as a free parameter , and fix the inclination angle at 60@xmath40 .
it is confirmed that even if 30@xmath40 is adopted , best - fit parameters do not significantly change .
the shape of the incident spectrum is assumed to be the same as the power - law component .
we basically assume that the reflection and scattered components did not vary between the _ suzaku _ and _ swift_/bat observations , considering that the size of the reflector has most likely a pc scale . in the three low mass llagns ( ngc 4395 , ngc 5273 , and ngc 5643 ) , however , we assume that the reflection component varied in accordance with the primary emission because of a smaller size of the emitting regions . thus , in these targets , @xmath41 is defined with respect to the primary component in the _ suzaku _ data .
the zgauss component represents an iron - k@xmath5 fluorescence line .
the line width is fixed at @xmath42 ev , which corresponds to a typical velocity dispersion of @xmath13 2000 km s@xmath2 measured with _
chandra_/hetgs in local seyfert galaxies @xcite .
we always consider the galactic absorption @xmath43 , which is calculated with the nh command @xcite in ftools .
the details of the spectral analysis for individual targets are described in section [ sec : broad_ana_2 ] .
the best - fit parameters , observed fluxes , and absorption corrected luminosities are summarized in tables [ tab : para ] and [ tab : flux ] .
the unfolded spectra and best - fit models are plotted in figure [ fig : unf_spec ] and [ fig : unf_model ] .
the eddington ratios in table [ tab : flux ] are calculated from the luminosity in the 210 kev band measured with _ suzaku _ ( _ swift _ ) for the low ( high ) mass llagns , by assuming a bolometric correction factor of 10 ( @xcite ; @xcite ) . for ngc 2655
, the base - line model alone does not give an acceptable fit , leaving residuals in the soft x - ray band below 2 kev .
although one apec component is insufficient to improve the fit ( @xmath44 ) , adding a second apec model gives an acceptable fit ( @xmath45 ) . we obtain a hydrogen column density of @xmath46 @xmath47 and @xmath48 , which are consistent with the _ asca _ result by @xcite .
the temperatures of the apec models are @xmath49 kev and @xmath50 kev .
the higher temperature is consistent with the _ xmm - newton _
result @xcite .
although the lower temperature component has not been reported for this source , the value is within a range observed in nearby seyfert galaxies ( e.g. , @xcite ) .
we significantly detect an iron - k@xmath5 line at 6.4 kev , which was not detected in the previous _ asca _ and _ xmm - newton _ observations ( @xcite ; @xcite ) .
the iron - k@xmath5 line fluxes observed with _ asca _ , _ xmm - newton _ , and _ suzaku _ were @xmath51 photons s@xmath2 @xmath47 , @xmath52 photons s@xmath2 @xmath47 , and @xmath53 photons s@xmath2 @xmath47 , respectively .
hence , the iron line flux slightly increased in the _ suzaku _ observation ( performed in 2014 ) compared with the _ xmm - newton _ epoch ( 2005 ) .
this indicates that the spatial size of the reflector ( torus ) is less than a few pc . the base - line model well reproduces the x - ray spectra of ngc 3718 , yielding @xmath54 with @xmath55 @xmath47 and @xmath56 .
no significant iron - k@xmath5 line is detected .
@xcite also reported a similar result by analyzing the _ chandra _ and _ xmm - newton _ spectra . the base - line model without a scattering component
well reproduces the spectra with @xmath57 = 950.8/982 .
the obtained photon index , @xmath58 , is consistent with a previous study @xcite .
we find that the reflection strength is very weak with an upper limit of @xmath59 .
no significant iron - k@xmath5 line emission is detected , either .
these suggest that there is little surrounding matter around the nucleus .
the spectra of ngc 4102 are complex , and can not be fit with the base - line model alone . by including two apec models , we obtain a better fit with @xmath60 .
further , adding three narrow emission lines other than the neutral iron - k@xmath5 line largely improves the fit ( @xmath61 ) .
each line component improves the fit by @xmath62 for two degrees of freedom ( center energy and normalization ) .
we obtain the line energies of @xmath63 kev , @xmath64 kev , and @xmath65 kev , which possibly correspond to mg xi , slightly blue - shifted fe xxv , and fe k@xmath66 lines , respectively .
the reflection strength derived from the continuum fit is small , @xmath67 , whereas the observed equivalent width of the iron - k@xmath5 line is large ( @xmath68 ev ) .
this implies that the reflector may not be compton thick .
the base - line model combined with a single apec model well reproduces the x - ray spectra of ngc 4138 with @xmath69 .
the addition of the apec model improves the fit by @xmath70 for two degrees of freedom ( temperature and normalization ) .
we obtain @xmath71 @xmath47 and @xmath72 , which are consistent with the result of the _ xmm - newton _ observation in 2001 @xcite .
the bi - xis spectrum of ngc 4258 below 0.9 kev shows highly complex line features , and therefore we exclude this energy band for our analysis . the base - line model combined with two apec models gives a good fit ( @xmath73 ) , which is better than that with a single apec model ( @xmath74 ) .
the obtained @xmath75 , @xmath76 , and temperatures of the apec components are consistent with the results by @xcite based on the _ asca _ observation in 1993 .
a significant narrow iron - k@xmath5 line at 6.4 kev is detected in our _ suzaku _ spectra .
the observations with _
chandra _ in 2000 and 2001 and _ xmm - newton _ in 2006 revealed that the observed iron - k@xmath5 line fluxes were variable ( e.g. , @xcite ; @xcite ) .
indeed , the _ chandra _ spectra did not show the iron - k@xmath5 emission line @xcite . also , analyzing the _ suzaku _ and _ xmm - newton _ spectra , @xcite concluded that the narrow iron - k@xmath5 emission was originated from the accretion disk because of its flux variability on time scales of 160 days .
hence , the iron line emitting region of ngc 4258 may not be distant cold matter or a dusty torus .
this fact is also recognized when we statistically discuss the torus structure of llagns in section [ sec : lumedd2iron ] .
the spectra of ngc 4395 are well reproduced with the base - line model plus a narrow absorption line at @xmath77 kev ( @xmath78 ) .
the absorption line improves the fit by @xmath79 , and is consistent with fe xxvi k@xmath5 .
we obtain @xmath80 by assuming that the fluxes of the reflected emission varied between the _ suzaku _ and _ swift_/bat spectra ( see above ) .
we note that without this assumption we would derive an unphysically large reflection strength , @xmath81 , which is much larger than typical values of @xmath82 observed in local agns @xcite .
the base - line model plus an apec model well reproduces the spectra with @xmath83 .
the broadband x - ray spectra were already analyzed by @xcite , and the difference from the previously used model is the absence of the disk - reflection component .
the obtained parameters are consistent with each other .
the base - line model gives a good fit with @xmath57 = 586.4/563 .
the observed equivalent width of the iron - k@xmath5 line , @xmath84 ev , is smaller than that estimated from the _ xmm - newton _ data taken in 2002 , @xmath13 230 ev @xcite , even though the observed continuum flux are almost the same .
this suggests that the line emitting region must be smaller than a few pc and that the averaged past activity before each observation , which determines the iron - k@xmath5 line flux , was higher in the earlier epoch .
similarly to the case of ngc 4395 , we would obtain an unphysically large reflection strength of @xmath85 if we assumed that the reflection component did not change between the _ suzaku _ and _ swift_/bat observations .
_ xmm - newton _ observations ( @xcite ; @xcite ) showed that ngc 5643 contains a ulx , which is located at @xmath86 0.9 from the nucleus and can not be resolved with the _ suzaku _ beamsize ( @xmath87 ) . because the nucleus is heavily obscured
, the ulx emission may contaminate the xis spectra at energies below 10 kev .
hence , we take into account the ulx spectrum ( see appendix ) in our spectral analysis . the base - line model plus the ulx component and two apec models
well reproduces the spectra ( @xmath57 = 109.5/114 ) .
the large hydrogen column density of @xmath88 @xmath47 and strong reflection strength of @xmath89 confirm that ngc 5643 has a heavily obscured agn .
we obtain an ( apparent ) scattered fraction of @xmath90 , although it strongly depends on the assumed flux level of the ulx .
when we increase the ulx flux by a factor of 2 to make it consistent with that measured by the _ xmm - newton _ observation in 2003 @xcite , we obtain only an upper limit of @xmath91 .
our data can be used to constrain the variability of the ulx if we assume that the agn is less variable .
the summed 210 kev flux from the nucleus and ulx as observed with _ suzaku _ in 2007 was @xmath92 erg s@xmath2 @xmath47 , while the fluxes from the nucleus measured with _ xmm - newton _ in 2003 and 2009 were both @xmath93 erg s@xmath2 @xmath47 ( @xcite ; @xcite ) . by assuming that the flux from the nucleus remained constant , the ulx flux during the _ suzaku _ observation is estimated to be @xmath94 erg s@xmath2 .
this value is between that during the _ xmm - newton _ observation in 2003 ( @xmath95 erg s@xmath2 @xmath47 ) and that in 2009 ( @xmath96 erg s@xmath2 @xmath47 ) , implying a consistent decay of the ulx flux from 2003 to 2009 .
+ + + + cccccccccccccccccccccccc ngc 2655 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath49 & @xmath104 & @xmath105 + ngc 3718 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath112 & - & - & @xmath113 & + ngc 3998 & @xmath114 & @xmath115 & @xmath116 & @xmath117 & - & @xmath118 & @xmath112 & - & - & @xmath119 & + ngc 4102 & @xmath120 & @xmath121 & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 + ngc 4138 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 & @xmath137 & - & @xmath138 & + ngc 4258 & @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143 & @xmath144 & @xmath145 & @xmath146 & @xmath147 & @xmath148 + ngc 4395 & @xmath149 & @xmath150 & @xmath150 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 & - & @xmath156 + ngc 4941 & @xmath157 & @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & - & @xmath165 + ngc 5273 & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath172 & - & - & @xmath173 & + ngc 5643 & @xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 cccccccccccccccccccccccc ngc 2655 & @xmath184 & @xmath185 & & @xmath186 & @xmath187 & @xmath188 & & @xmath189 & @xmath190 + ngc 3718 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 + ngc 3998 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 + ngc 4102 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath204 & @xmath215 & @xmath216 + ngc 4138 & @xmath217 & @xmath218 & & @xmath219 & @xmath220 & @xmath221 & & @xmath222 & @xmath223 + ngc 4258 & @xmath224 & @xmath225 & @xmath226 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 + ngc 4395 & @xmath232 & @xmath233 & @xmath234 & @xmath226 & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 + ngc 4941 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 + ngc 5273 & @xmath184 & @xmath249 & @xmath250 & @xmath251 & @xmath252 & @xmath204 & @xmath253 & @xmath254 & @xmath255 + ngc 5643 & @xmath256 & @xmath92 & @xmath257 & @xmath258 & @xmath259 & @xmath196 & @xmath260 & @xmath261 & @xmath262 in this section , we investigate possible contribution of a relativistically broadened iron - k@xmath5 line in our spectra .
we exclude ngc 4102 , ngc 4941 , and ngc 5643 from this study , which show heavy absorptions ( @xmath263 cm@xmath2 ) preventing detailed studies of the broad iron - k@xmath5 line feature .
first , we fit the broadband x - ray spectra with the best - fit model obtained in section [ sec : broad_ana_2 ] plus a disk - reflection component ( rdblur*pexmon ) .
the rdblur model calculates the relativistic blurring around a schwarzschild black hole @xcite , and the pexmon model reproduces the same reflection continuum as pexrav together with emission lines of fe k@xmath5 , fe k@xmath66 , and ni k@xmath5 computed in a self - consistent way @xcite .
this disk component is corrected for the same absorption and time variability as those for the primary x - ray emission .
because the two reflection components from the disk and distant matter are strongly coupled in a spectral fit , we fix the disk reflection strength @xmath265 . for this purpose
, we refer to the study by @xcite , who calculated the predicted equivalent width of the iron - k@xmath5 line from an annulus slab irradiated by a point source with a power - law spectrum .
we thus determine @xmath265 in the pexmon model that reproduces the predicted equivalent width of the iron - k@xmath5 line , by assuming an inclination angle ( @xmath266 ) , a photon index , and a ratio between the scale height of the x - ray source and the inner radius of the annulus slab ( @xmath267 ) . here
@xmath266 is set to 30@xmath40 and 60@xmath40 for type-1
( @xmath268 @xmath47 ) and type-2 ( @xmath269 @xmath47 ) agns , respectively . for ngc 4258
, we adopt @xmath270 , which was precisely estimated by maser observations @xcite .
the photon index is tied to that of the primary x - ray component .
we adopt the scale height of @xmath271 ( @xmath272 is the gravitational radius ) , a typical value measured in some agns ( e.g. , @xcite ; @xcite ) .
then , we consider two cases , @xmath273 and @xmath274 , which correspond to @xmath275 and @xmath276 , respectively . in the rdblur model , the outer radius and radial emissivity index are set to @xmath277 and @xmath278 , respectively .
we find that adding the disk reflection component with the above parameters does not significantly improve the fit in all 7 objects ( @xmath279 ) .
thus , our data do not require a cold standard disk extending down to 10100 @xmath272 , although its presence can not be strictly ruled out within the quality of our data ( see below ) . to derive upper limits of the flux of a broad iron - k@xmath5 line , we replace pexmon*rdblur with diskline+pexrav*rdblur and fit the xis spectra in the 39 kev ( 38 kev for bi - xis ) band .
the parameters ( @xmath266 , @xmath66 , @xmath267 , and @xmath280 ) of the diskline model are set to the same as in the rdblur model .
we leave the normalizations of the zgauss , diskline , and zpowerlw models as free parameters .
table [ tab : broad_line ] summarizes the line fluxes and equivalent widths ( ew@xmath281 ) of the broad iron - k@xmath5 line in the case of @xmath273 , which gives more conservative upper limits than the case of @xmath282 . to check whether the assumed equivalent widths based on @xcite are consistent with the observed upper limits , we also calculate the corrected equivalent width ( ew@xmath283 ) with respect to the continuum composed of the primary power - law and disk - reflection components .
we find that in ngc 3998 the upper limit of ew@xmath284 ev is inconsistent with the assumed one ( @xmath86 120 ev for @xmath273 and @xmath86 2030 ev for @xmath282 ) , implying that the inner disk radius may be much larger than @xmath274 . in the other objects ,
the assumed equivalent widths are consistent within the observed upper limits , and hence the possible presence of a cold standard disk with @xmath285 10100 @xmath272 is not ruled out .
cllllll ngc 2655 & @xmath286 & @xmath287 & @xmath288 + ngc 3718 & @xmath289 & @xmath290 & @xmath290 + ngc 3998 & @xmath291 & @xmath292 & @xmath292 + ngc 4138 & @xmath293 & @xmath294 & @xmath287 + ngc 4258 & @xmath295 & @xmath296 & @xmath297 + ngc 4395 & @xmath298 & @xmath299 & @xmath300 + ngc 5273 & @xmath301 & @xmath302 & @xmath302 similarly , we obtain the flux and observed equivalent width of the narrow iron - k@xmath5 line ( ew@xmath303 ) by fitting the xis narrow band ( 39 kev ) spectra with the best - fit models obtained in section [ sec : broad_ana_2 ] .
only the normalizations of the zgauss and zpowerlw components are allowed to vary in this analysis .
figure [ fig : narrow_line ] plots the spectra with the best - fit model . for easier comparison with model predictions ( see section [ sec : ikeda ] )
, we also calculate the corrected equivalent width ( ew@xmath304 ) with respect to the continuum composed of the primary power - law and torus - reflection components .
the results are summarized in table [ tab : narrow_line ] .
cccccccc ngc 2655 & @xmath103 & @xmath305 & @xmath306 + ngc 3718 & @xmath112 & @xmath292 & @xmath292 + ngc 3998 & @xmath112 & @xmath307 & @xmath307 + ngc 4102 & @xmath308 & @xmath309 & @xmath310 + ngc 4138 & @xmath311 & @xmath312 & @xmath313 + ngc 4258 & @xmath314 & @xmath315 & @xmath316 + ngc 4395 & @xmath317 & @xmath318 & @xmath313 + ngc 4941 & @xmath163 & @xmath319 & @xmath320 + ngc 5273 & @xmath321 & @xmath322 & @xmath323 + ngc 5643 & @xmath324 & @xmath325 & @xmath326
we have systematically analyzed the broadband x - ray spectra in the 0.5200 kev band of ten nearby llagns observed with _
suzaku _ and _ swift_/bat .
the spectra are basically well reproduced with the base - line model composed of absorbed primary x - ray emission , a scattered component , and a reflection component from distant matter with a fluorescence iron - k@xmath5 line .
some spectra require additional components , such as optically - thin thermal emission and emission / absorption lines .
the spectra of ngc 3718 and ngc 3998 show no narrow iron - k@xmath5 line , whereas it is significantly detected in the other objects . in the following discussions , we refer to the x - ray luminosities measured with _
suzaku _ and _
swift_/bat for the three low mass llagns ( ngc 4395 , ngc 5273 , and ngc 5643 ) and the other llagns , respectively . in section [ subsec : broad ]
, we show that a relativistically broadened iron - k@xmath5 line from a standard disk extending down to 10100 @xmath272 is not required from the spectral fit , even though we can not , in most of the targets , rule out its presence within the statistical errors
. only upper limits of the line flux are derived . in the case of ngc 3998 ,
the tight constraint on the equivalent width of the broad iron - k@xmath5 line ( @xmath292 ev ) indicates that the standard disk must be truncated at a radius larger than @xmath274 .
this is consistent with the result by @xcite , who reported that a riaf model fit to the sed of ngc 3998 requires a truncation radius larger than @xmath327 .
the broadband coverage of our data including the hard x - ray band above 10 kev has enabled us to most reliably determine the photon index of the intrinsic power - law continuum even for obscured agns .
figure [ fig : index_vs_edd - ratio ] plots the best - fit photon index against the eddington ratio for our sample .
we obtain a significant negative correlation with a form of @xmath328 by a @xmath329 fit ( errors are 1@xmath29 ) .
the slope is consistent with that found by @xcite , who derived a form of @xmath330 for low eddington - ratio ( @xmath331 ) agns .
@xcite also derived a negative but steeper slope of @xmath332 for lower eddington - ratio agns ( @xmath333 ) . to compare our result with that of @xcite in figure [ fig : index_vs_edd - ratio ] , we convert the 210 kev luminosity into the bolometric one with a 210 kev bolometric correction factor of 10 ( @xcite ; @xcite ) . it is known that the slope between the photon index and eddington ratio turns into a positive value at high eddington - ratio agns . for instance , @xcite reported a positive slope of @xmath334 for agns with @xmath335 .
a similar result was also obtained by @xcite . as suggested by some authors ( e.g. , @xcite ; @xcite ) ,
the @xmath76-@xmath8 correlations for high and low eddington - ratio agns imply that the state of the accretion disk changes according to the eddington ratio .
the positive correlation at the high eddington - ratio regime can be explained in terms of standard disk as follows .
a hot corona located above the disk produces comptonized emission of seed photons provided from the disk .
hence , as the disk luminosity ( or accretion rate ) increases , electrons in the corona are more effectively cooled through inverse compton scattering , leading to a smaller compton y - parameter , and hence a softer spectrum .
there are some numerical studies on the @xmath76-@xmath8 correlation for low eddington - ratio agns under a riaf assumption .
for instance , the convection dominated accretion flow model developed by @xcite predicts a negative correlation , whereas the advection dominated accretion flow model by @xcite predicts a positive one , which is inconsistent with the observations .
thus , this relation can be used to test riaf models at low mass - accretion rates .
the flux ( or equivalent width ) of a narrow iron - k@xmath5 line at 6.4 kev is useful to constrain the geometry and column density of circumnuclear matter , which we call the `` torus '' . to constrain the torus opening angle with respect to the polar axis
, we utilize the ikeda torus model @xcite , which is a monte - carlo based numerical spectral model .
the torus structure has a nearly spherical shape with two conical - shaped holes along the rotation axis , and is defined by three parameters : the hydrogen column density at the equatorial plane ( @xmath336 ) , the half opening angle ( @xmath337 ) , and the inclination angle ( @xmath266 ) .
hence , @xmath338 for type-1 agns and @xmath339 for type-2 agns .
as done in previous works ( @xcite ; @xcite ) , we fix @xmath336 at the observed line - of - sight column density ( @xmath75 ) and a photon index of the incident x - ray spectrum according to the best - fit value listed in table [ tab : para ] .
figure [ fig : ikeda_model ] shows the predicted equivalent width of the iron - k@xmath5 line as a function of @xmath337 for several different inclinations . because simulated spectra for @xmath340 are not available in the ikeda torus model , for type-1 agns , we simply extrapolate the result at @xmath341 towards larger opening angles by assuming that the equivalent width is proportional to the volume of the torus .
the observed range of ew@xmath304 at the 90% confidence limits is represented with the dot - dot - dot - dashed horizontal lines ( magenta ) .
as noticed from figure [ fig : ikeda_model ] , the half opening angles of ngc 3718 and ngc 3998 can not be constrained if we assume @xmath342 , although their weak iron - k@xmath5 line fluxes and low line - of - sight absorptions suggest that their tori are little developed , having a small covering fraction and/or a small averaged column density . by contrast , ngc 2655 , ngc 4102 , ngc 4138 , ngc 4258 , and ngc 4941 have moderately developed tori , fulfilling @xmath343 @xmath47 and/or @xmath344 . for ngc 4395 , ngc 5273 , and ngc 5643 , the ikeda model largely underpredicts the equivalent widths under the assumption of @xmath342 .
we find that their equatorial hydrogen column densities must be larger than @xmath345 @xmath47 , @xmath346 @xmath47 , and @xmath347 @xmath47 , respectively , to consistently explain the observed values of ew@xmath304 . to summarize
, we reveal that there are at least two types of llagns in terms of torus structure : those with moderately developed tori ( @xmath348 corresponding to @xmath344 , or @xmath349 @xmath47 ) or those with little developed tori .
the variety in the torus structure of llagns is a new result obtainable only by detailed spectral analysis of individual objects .
previous works based on the type-2 agn fraction have given only the `` averaged '' torus covering factor , which decreases towards lower luminosities ( @xcite ; @xcite ) .
@xcite pointed out that , in order to discuss the torus structure of type-2 agns , the luminosity ratio between the iron - k@xmath5 line and 1050 kev continuum is a better indicator of the torus covering factor than the iron - k@xmath5 equivalent width because the continuum flux above 10 kev is less affected by absorption than that at 6.4 kev . using an agn sample observed with _
suzaku _ , they found that the @xmath350 ratio decreases with luminosity in both type-1 and type-2 agns at @xmath351 erg s@xmath2 .
these results can be explained if the torus covering fraction becomes smaller at higher luminosities , and are consistent with the anti - correlation between the absorbed - agn fraction and luminosity in this luminosity range ( e.g. , @xcite ) .
we plot @xmath350 against luminosity of our llagn sample together with the original sample of @xcite in figure [ fig : line_vs_lum ] .
here we exclude compton - thick agns ( @xmath352 @xmath47 ) because the heavy obscuration also strongly affects the iron - k@xmath5 line flux and the relation may become complex ( see * ? ? ?
the dashed lines in the figure are the best - fit linear regression forms for type-1 and type-2 agns obtained by @xcite .
figure [ fig : line_vs_lum ] shows a trend that the @xmath350 ratio has a peak around @xmath353 erg s@xmath2 , from which the average value rapidly declines towards lower luminosities .
this agrees with the implication from hard x - ray ( @xmath0 kev ) surveys that the absorbed agn fraction peaks at @xmath354 erg s@xmath2 and decreases with decreasing luminosity ( @xcite , @xcite ) .
thus , our study using the narrow iron - k@xmath5 line supports the picture that the averaged solid angle of agn tori is not a monotonically decreasing function with luminosity , and that irradiation is not the only key factor that determines the torus structure .
we note that there is a large scatter of @xmath350 in the llagns , as discussed in the previous subsection . figure [ fig : ew_vs_edd - rat ] shows the eddington ratio dependence of @xmath350 based on our llagn sample .
a positive correlation is seen between the @xmath350 ratio and @xmath8 .
if we exclude ngc 4258 ( the second left point ) , whose narrow iron - k@xmath5 line may not be originated from the torus @xcite , there is a trend that all llagns with @xmath355 have @xmath350 smaller than @xmath356 .
we thus infer that the eddington ratio is a key parameter that affects the torus structure of llagns : tori are hardly developed below a critical value of @xmath8 . because the sample size is limited , however , we would need a larger number of llagns to draw a robust conclusion .
the lack of the torus may lead to the shortage of mass supply through it , hence reducing the mass - accretion rate . in an opposite way
, the torus is not formed because of the lack of powerful outflow from the inner accretion disk at low mass - accretion rates .
thus , to understand the reason of the @xmath350 versus @xmath8 correlation , we need to reveal the physical origin(s ) of agn tori .
many studies found strong luminosity correlations between the x - ray and mid - infrared ( mir ) bands ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
such correlation is expected because the dust torus heated by an agn emits blackbody radiation in the mir band .
@xcite find no eddington - ratio dependence of the x - ray / mir ( 12 @xmath357 m ) luminosity ratio using a sample that covers a wide eddington ratio range of @xmath358 .
figure [ fig : x - ray_vs_ir ] shows the @xmath359 versus @xmath360 correlation obtained from our llagn sample .
the regression line derived by @xcite from a large local agn sample is plotted by the solid line .
@xcite took only the nuclear mir emission by separating from that in the host galaxy , using subarcsecond resolution imaging data .
we refer to the photometric data in the 12 @xmath357 m band of @xcite for our objects whenever available , and to the wide - field infrared survey explorer ( _ wise _ ) catalog @xcite for the rest . due to the poor angular resolution of _
wise _ ( 6.5 arcsec ) , however , its flux may be contaminated from that of the host galaxy and hence should be regarded as an upper limit of the nuclear emission .
we find that our llagns also approximately follow the same correlation as found by @xcite for higher luminosity agns , regardless of agn types and eddington ratios .
this is expected if a ( moderate size of ) torus is present in llagns .
there is a possibility that synchrotron emission from the jets or blackbody radiation from the truncated disk may contribute to the mir luminosity in some llagns ( e.g. , @xcite ; @xcite ) .
our result in section [ sec : lumedd2iron ] suggests that the low eddington ratio llagns ( @xmath355 ) do not have developed tori . in these objects ,
the mir luminosity would be indeed dominated by these emission mechanisms other than thermal emission from agn - heated dust .
mir seds are often used to identify agns by detecting a power - law component originating in agn - heated dust ( e.g. , @xcite ) .
the spectral index in the mir band of luminous agns are typically @xmath361 in the form of @xmath362 ( e.g. , @xcite ) .
accordingly , @xcite defined the so - called agn - wedge region in the mir color - color diagram based on the three _ wise _ bands ( 3.4 @xmath357 m , 4.6 @xmath357 m , and 12.0 @xmath357 m ) in order to efficiently select luminous agns .
figure [ fig : agn_wedge ] shows the color - color plot of our llagn sample based on the _ wise _ data together with the agn - wedge region .
as noticed , all objects are not located within this region .
this is attributable to significant contamination from the host galaxy in the _ wise _ mir fluxes .
this is an example demonstrating difficulty in identifying agns using mir data alone .
we have performed a systematic broadband ( 0.5200 kev ) x - ray spectral analysis of ten local llagns ( with intrinsic 14195 kev luminosities of @xmath363 erg s@xmath2 ) observed with _
suzaku _ and _ swift_/bat .
the main conclusions are summarized as follows . 1 .
the broadband x - ray spectra can be basically reproduced with an absorbed cut - off power law often accompanied by a reflection component from distant cold matter and a narrow fluorescence iron - k@xmath5 line , and optically - thin thermal emission from the host galaxy .
2 . in all objects ,
a relativistically blurred reflection component from a standard disk is not required from the spectra .
we find the negative correlation between the photon index and eddington ratio .
these results are consistent with a picture that an optically - thin riaf is formed at innermost radii in llagns .
3 . applying a torus model by @xcite
, we find eight objects showing a significant narrow iron - k@xmath5 emission line have a moderately developed torus with an equatorial column density of @xmath364 or a half opening - angle of @xmath344 .
the tori of the two llagns without a narrow iron - k@xmath5 line are likely little developed .
4 . the luminosity ratio between the iron - k@xmath5 line and 1050 kev continuum ( @xmath350 ) has a peak around @xmath353 erg s@xmath2 , from which the average value rapidly declines towards lower luminosities .
this result indicates that agn irradiation is not the only key factor that determines the torus structure .
low eddington - ratio llagns with @xmath355 show small @xmath350 ratios .
we infer that the eddington ratio is a key parameter that affects the torus structure of llagns .
5 . regardless of eddington
ratio , our llagns follow the same luminosity correlation between the hard x - ray and mir bands as found for more luminous agns .
this implies that other emission mechanisms than agn - heated dust are responsible for the mir emission in low eddington - ratio llagns without developed tori .
we thank the referee for the comments , which helped us to improve the quality of the manuscript .
we are grateful to c. ricci for kindly providing us with the data .
part of this work was financially supported by the grant - in - aid for jsps fellows for young researchers ( t.k ) and for scientific research 26400228 ( yu ) .
this research has made use of the nasa/ ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
to model the spectrum of the ulx in ngc 5643 , we use the _ xmm - newton _ data ( obsid = 0601420101 ) obtained in 2009 , because the exposure ( @xmath365 ksec ) is longer than that of the other _ xmm - newton _ observation ( @xmath366 ksec ) in 2003 .
the standard analysis software ( version 13.5.0 ) is utilized to reduce the data of the epic cameras ( mos-1 , mos-2 , and pn ) .
for all cameras , the source events are extracted from a circular region of a 26 radius , and the background events are taken from a source - free circular region with the same radius .
the spectra of mos-1 and mos-2 are summed .
we find that a cut - off power law plus a disk - blackbody component well reproduces the spectra in the 0.37 kev band ( @xmath57 = 183.4/154 ) . here
the cut - off energy is fixed at 6 kev , a value observed from a typical ulx , ngc 1313 x-1 @xcite ; the fit is not significantly improved by leaving the cut - off energy as a free parameter ( @xmath57 = 182.4/153 ) .
the best - fit parameters are a photon index of @xmath367 , a disk - blackbody normalization of 3.7 @xmath368 erg s@xmath2 kpc@xmath369 , and a disk temperature of 0.35 kev .
when we analyze the _ suzaku _ spectra of ngc 5643 , we always include this best - fit ulx model .
|
we systematically analyze the broadband ( 0.5200 kev ) x - ray spectra of hard x - ray ( @xmath0 kev ) selected local low - luminosity active galactic nuclei ( llagns ) observed with _
suzaku _ and _ swift_/bat .
the sample consists of ten llagns detected with _
swift_/bat with intrinsic 14195 kev luminosities smaller than @xmath1 erg s@xmath2 available in the _ suzaku _ archive , covering a wide range of the eddington ratio from @xmath3 to @xmath4 .
the overall spectra can be reproduced with an absorbed cut - off power law , often accompanied by reflection components from distant cold matter , and/or optically - thin thermal emission from the host galaxy . in all objects ,
relativistic reflection components from the innermost disk are not required .
eight objects show a significant narrow iron - k@xmath5 emission line . comparing their observed equivalent widths with the predictions from the monte - carlo based torus model by @xcite , we constrain the column density in the equatorial plane to be @xmath6 or the torus half opening angle @xmath7 .
we infer that the eddington ratio ( @xmath8 ) is a key parameter that determines the torus structure of llagns : the torus becomes large at @xmath9 , whereas at lower accretion rates it is little developed .
the luminosity correlation between the hard x - ray and mid - infrared ( mir ) bands of the llagns follows the same one as for more luminous agns .
this implies that other mechanisms than agn - heated dust are responsible for the mir emission in low eddington ratio llagns .
|
i would like to express thanks to g.v.domogatsky , j - a.m .
djilkibaev , s.i.klimushin , b.k.lubsandorzhiev , e.a.osipova , v.v.prosin , c.spiering and v.b.tsvetkov for useful discussion .
i.a.belolaptikov et al .
, + the baikal underwater neutrino telescope : design , performance , and first results .
+ astroparticle physics 7(1997 ) 263 - 282 .
+ 2.e.andres et al . , + the amanda neutrino telescope : principle of operation and first results .
+ astroparticle physics 13 ( 2000 ) 1 - 20 .
+ 3.b.monteleoni et al . , + nestor a deep sea physics laboratory for the mediterranean + proc . of the 17th international conference on neutrino physics and astrophysics , ( neutrino 96 ) + helsinki , finland , june 13 - 19 , 1996 .
+ 4.ph.amram et al . , + the antares project + nucl.physics b ( proc . suppl . ) 75a ( 1999 ) 415 .
+ 5.v.s.berezinsky and g.t.zatsepin + sov.phys.usp . 20
( 1977 ) 361 .
+ 6.dumand-75 .
proceeding of the 1975 dumand summer study , bellingham , + edited by p.kotzer.published by western washington state college .
+ 7.l.d.landau , e.m.lifshiz .
electrodynamics of continuous media .
+ moscow , nauka , 1982 , p 345 .
+ 8.a.sommerfeld , optik , wiesbaden 1950 .
+ 9.physical quantities , handbook , ed .
i.s.grigorev and e.z.meylihoba , + energoatomizdat , moscow 1991 , p. 790 .
a.i.stogarov and n.f.timofeeva , + trudi goi , 1963 , v 31 , n 160 ( in russian ) .
|
during the last few years deep underwater neutrino telescopes of a new generation with dimensions close to 100 m or more were taken into operation .
for the correct track reconstruction of events and for the interpretation of light pulses from calibration lasers one has to use the group velocity for light signals . the difference between group velocity and the traditionally used phase velocity leads to an additional delay of about 10 ns for a distance of 100 m between light source and photomultiplier . from the time of the appearance of the first projects of deep underwater neutrino telescopes in the middle of 70th
this fact was never mentioned in the literature . on the velocity of light signals in the deep underwater neutrino experiments l.a.kuzmichev + skobeltsyn institute of nuclear physics , moscow state university + 119899 , moscow , russia , e - mail : [email protected] [ [ section ] ] during the last few years the new large - scale neutrino telescopes nt-200 /1/ and +
have been taken into operation .
two deep underwater arrays , and antares /4/ , are under construction .
the cherenkov light from high energy muons , electromagnetic and hadronic showers can be recorded at distances of 20 100 m depending upon the light absorption of water or ice .
impulse laser light sources are widely used in these arrays for calibration .
light pulses from such sources can be recorded over even larger distances . at such distances ,
the difference between group and phase velocity of light in water or ice is essential .
as far as i know , this fact hasnt been mentioned in the literature from the time of the first neutrino telescope projects in the middle of the seventies /5,6/ and hasnt been taken into account for the data analysis and mc calculations .
the velocity of light pulses in matter is given by the group velocity /7,8/ : @xmath0 where @xmath1 is the frequency , @xmath2 is the wave vector .
the connection between @xmath1 and @xmath2 can be written in the following form ( dispersion equation ) : @xmath3 where @xmath4 is the velocity of light in vacuum , @xmath5 is the refraction index .
as it is known , refraction indices given in handbooks like /9/ correspond to the phase velocity due to the method they are measured /10/ : @xmath6 where @xmath7 is the phase velocity .
it is reasonable to accept a group refraction index for the group velocity similar to the phase index of refraction : @xmath8 where @xmath9 is the group velocity . using ( 2 ) and replacing the @xmath1 dependence of @xmath5 by a wavelength dependence
, the relation between @xmath10 and @xmath5 can be derived : @xmath11 where @xmath12 is the wavelength .
@xmath13 is negative for water over the visible light range , and so @xmath10 is larger than @xmath5 .
fig.1 shows the wavelength dependence of @xmath5 ( curve 1 ) for destilled water at 20@xmath14c /9/ , and the same for @xmath10 ( curve 2 ) .
the density of water influences the refraction index at large depths of neutrino telescopes . but
this correction is rather small due to the comparatively small compression coefficient of water . for the conditions of nt-200 ( depth 1100 m , water temperature the variation of the refraction index is about 0.002 only
this is more than 10 times smaller than the difference between @xmath10 and @xmath5 .
this correction of @xmath5 would reach 0.01 for the arrays to be placed in the ocean depths of 4 -5 km , but even in this case the depth correction is significantly smaller than that due to the difference between group and phase velocities .
the curve 3 on fig.1 shows the dependence of @xmath10 on @xmath12 for the conditions of the telescope nt-200 .
all curves at fig.1 were calculated for destilled water .
the light absorption of natural water differs from that of destilled water , and so there is a difference in the image part of their dielectric penetration . as kramers - kroning equation connects the real part of the dielectric penetration with it s imaginary part /7/ , the real refraction index differs , in principle , from that of destilled water .
such a correction would be non - negligible only if there are the intensive lines of absorption in the visible light absorption spectrum .
fig.2 shows the dependencies of delays on the distance between light source and receiver which emerge from the difference between the phase and group velocities for different wavelengths . for arrays with angular resolution 1 - 2@xmath14 replacement of the phase velocity by the group velocity probably would not lead to essential changes in track reconstruction . for the projects which claim an angular accuracy 0.1 - 0.2@xmath14 and absorption length of @xmath15 m /4/ the use of group velocity in track reconstruction procedures seems to be absolutely necessary
|
alloys are among the most relevant materials for modern technologies .
conventional alloys typically consist of one principal element , such as the iron in steel , plus one or more dopant elements in small proportion ( e.g. carbon in the case of steel ) that enhance a certain property of interest ; the properties are based on the modification of those of the principal element . in sharp contrast , high - entropy alloys ( hea ) are comprised of multiple principal elements that are all present in major proportion , with the simple structures observed attributed to the high configurational entropy of the random mixing of the elements on their lattice sites @xcite .
thus , the concept of a `` principal element '' becomes irrelevant .
the elements in heas arrange on simple lattices with the atoms stochastically distributed on the crystallographic positions ; heas are commonly referred to as metallic glasses on an ordered lattice ( see figure [ fig:1](a ) and ( b ) ) .
the properties of heas arise as a result of the collective interactions of the randomly distributed constituents @xcite .
there is no strict definition , but heas are typically composed of four or more major elements in similar concentrations . by applying this concept , several new alloys with simple body - centered cubic ( bcc ) , hexagonal - closest packing ( hcp ) , or face - centered cubic ( fcc ) structures have been realized @xcite .
the heas compete for thermodynamic stability with crystalline intermetallic phases with smaller numbers of elemental constituents @xcite .
therefore , one central concept of designing these alloys is to understand the interplay between mixing entropy @xmath0 and phase selection . considering the large number of metals in the periodic table ,
the total number of possible hea compositions is virtually unlimited . + + . [ fig:1 ] in addition to their structural and chemical diversity , heas can display novel , highly - tunable properties such as for example excellent specific strength @xcite , superior mechanical performance at high temperatures @xcite , and fracture toughness at cryogenic temperatures @xcite , making them promising candidates for new applications .
simple niobium - titanium based binary alloys are nowadays still the most often and widely used materials for superconducting magnets , such as e.g. in nmr and mri devices @xcite or the large hardron collider @xcite , and thus the discovery of bulk superconductivity with a single well defined phase transition on a highly disordered bcc lattice in the nb - ti
related ta - nb - hf - zr - ti hea is of considerable interest @xcite .
this multicomponent phase , stabilized by the high mixing entropy , appears to fall between an ordered solid and a glass , and thus allows for study of the chemical composition and structure - property relations of a superconducting material part - way between an ordinary alloy and an amorphous material on a fundamental level .
here , we report the results of our investigations of the influence of electron count and alloy complexity on superconductivity in the ta - nb - hf - zr - ti hea .
we find that the variation in superconducting transition temperature with electron count is intermediate to those displayed by simple alloys and amorphous materials , and that the elemental make - up of the hea superconductor is critical for determining its properties , in spite of the fact that the materials system is very highly disordered .
zone , is shown.,scaledwidth=40.0% ]
all samples were prepared from pieces of the pure metals .
stoichiometric amounts of niobium ( purity 99.8 % ) , tantalum ( purity 99.9 % ) , zirconium ( purity 99.6 % ) hafnium ( purity 99.6 % ) , and titanium ( purity 99.95 % ) pieces were arc melted in high currents ( @xmath1c ) in an argon atmosphere , and rapidly cooled on a water - chilled copper plate .
a zirconium sponge was co - heated to purify the reaction atmosphere from remaining oxygen .
the samples were melted five times , and turned over each time to ensure optimal mixing of the constituents , the weight loss during melting was found to be insignificant .
x - ray diffraction patterns were obtained from mechanically flattened pieces ( in liquid nitrogen ) of the very hard alloys , measured in a bragg - bretano reflection geometry .
the patterns were obtained on a bruker d8 advance eco with cu @xmath2 radiation and a lynxeye - xe detector .
the resistivity , magnetization and specific heat were studied using a _ quantum design _ physical property measurement system ( ppms ) dynacool with a 9 t magnet , equipped with a vibrating sample magnetometer option ( vsm ) . for the resistivity measurements ,
a standard 4-probe technique was employed with 20 @xmath3 m diameter platinum wires attached with silver epoxy .
the applied current for these measurements was @xmath4 = 2 ma .
specific - heat measurements were performed with the quantum design heat - capacity option using a relaxation technique .
electron diffraction measurements were performed at brookhaven national laboratory on a jeol arm200f transmission electron microscope with double - cs correctors .
the powder x - ray diffraction ( xrd ) patterns of the high - entropy alloys ( heas ) for _ x _ = 0.2 , 0.25 , 0.3 , 0.33 , 0.35 , 0.4 , 0.45 , 0.5 , 0.6 , 0.7 , 0.8 , and 0.84 which were synthesized by arcmelting , can all be indexed with a simple bcc unit cell .
all prepared alloys fall within the definition for hea compositions ( see , e.g. , reference ) , with no constituent element of less than 5 mol-% and/or more than 40 mol-% . in figure [ fig:1]c
, we show three representative xrd patterns of the members @xmath5 0.3 , 0.45 , and 0.65 .
the patterns are found to shift only slightly with composition .
therefore a shifting of the cell parameter @xmath6 is observed , but its change between the different heas is only minor .
all alloys are found to be single phase with broad reflections , which we attribute to both the high degree of disorder present in the heas and also the non - ideal diffraction sample preparation ( the alloys are too hard to crush by our methods , and so fine particle size powders could not be made for the diffraction experiment ) .
the observed unit cell change results from the large difference of atomic radii of the different constituent atoms .
the unit cell parameter for the bcc lattice observed is found to vary from @xmath7 to @xmath8 within the solid solution .
the earlier reported cell parameter for variants of this ta - nb - hf - zr - ti hea follow this trend accordingly @xcite .
thus the observed physical properties reported below are those of the bulk , since no impurity phases are observed .
an earlier reported minor hexagonal phase impurity is not present in the samples of this study @xcite .
+ + 2 mt . ] in figure [ fig:1b ] , we show a representative high resolution transmission electron microscope ( hrtem ) image of a nearly optimally doped superconducting hea sample _ x _ = 0.33 .
the hrtem image is taken along the [ 111 ] zone axis .
this image of the nano - structure of the alloy reveals the arrangement of the atoms on a simple , homogeneous bcc lattice , despite the presence of five constituent atoms with very different atomic radii .
critically , no nanoscale chemical phase separation was observed for any of the materials investigated . in the inset of figure [ fig:1b ]
, we show the fourier - transform of the observed atom - positions in the real space image . in the fourier - transform pattern of the hrtem image ,
the six reflections close to the center spot represent 110 planes , clearly support the bcc structure of the hea even at the nanoscale .
+ + and the blue line is a trend line ) in comparison with transition metals and their alloys in the crystalline form @xcite ( yellow dashed line ) and as amorphous vapor - deposited films @xcite ( red dashed line ) .
the superconducting transition temperatures @xmath9 are plotted as function of the electron / atom ratio . ] the elemental metals in this pentinary superconducting hea , when taken by themselves , order on either hcp and bcc lattices : while hafnium , zirconium , and titanium crystallize on a hcp lattice , niobium and tantalum crystallize on a bcc lattice at room temperature . for conventional alloys between metals with a valence electron count ( vec ) of 5 ( niobium or tantalum ) and with a vec of 4 ( titanium , zirconium , or hafnium ) a structural transition from a hcp to bcc lattice is observed @xcite with decreasing electron count . due to their electron count
, the heas prepared here with @xmath10 = 0.8 and 0.84 would be expected to order on a hcp lattice .
this polymorphic transition is , however , not observed in the hea .
the high entropy of the system therefore stabilizes the structure of this hea preferentially on a bcc lattice ( see also , for example references ) .
0 t to 9 t in steps of 0.5 t. the dotted line displays the 50 % criterion , which is commonly used for the determination of the upper critical field @xmath11 .
( b ) temperature dependence of the upper critical field @xmath11 , determined by the 50 % criterion , of the heas with _ x _
= 0.30 , 0.33 , 0.35 , 0.5 , and 0.84 .
the lines are linear fits for determination of @xmath12 . ]
.critical temperatures @xmath9 , slopes of the upper critical fields @xmath13 , and upper critical fields at zero temperature @xmath14(0 ) of with @xmath10 = 0.3 , 0.33 , 0.4 , 0.5 , and 0.84 [ cols="^,^,^,^",options="header " , ] with the sommerfeld constant @xmath15 and @xmath16 , where @xmath17 is the number of atoms per formula unit , @xmath18 = 8.31 j mol@xmath19 k@xmath19 is the gas constant , and @xmath20 the debye temperature . for comparability reasons the numbers of atoms per formula unit @xmath17
was fixed to be one hundred .
the obtained values for @xmath15 , @xmath21 , and @xmath20 are summarized in table [ tab:2 ] .
the obtained ratios @xmath22 all exceed the standard weak - coupling bcs value , which is @xmath23 1.43 , indicating intermediate to strong - coupling superconductivity .
the sommerfeld constant is found to decrease substantially from @xmath24 797 mj mol@xmath19 k@xmath25 to 645 mj mol@xmath19 k@xmath25 with a decreasing electron count within the series .
thereby , the density of states at the fermi - level is reduced , since @xmath15 is proportional to the density of states at the fermi - level ( @xmath26 ) .
thus we attribute the decreasing of the critical temperature @xmath9 with an increasing electron count to a significant decrease in the density of states .
it should be noted that simultaneously also the electron - phonon coupling is lowered , which also might contribute to the lowering of the critical temperature ( see below ) . and have nominally the same electron count , and experimentally we find the same density of states at the fermi - level , with almost identical values for @xmath15 .
the debye temperature is found to increase only slightly .
thus although the critical temperatures differ by almost a factor of 2 , there is not much difference in the fundamental quantities that determine the transition temperature : @xmath15 and @xmath27 .
we therefore tentatively attribute the decrease in @xmath9 to the difference in the electron - phonon coupling @xmath28 that must occur on going from to .
the electron - phonon coupling @xmath29 can be estimated using the approximated mcmillan formula , which is based on the phonon spectrum of niobium @xcite and is valid for @xmath30 @xcite : @xmath31 the parameter @xmath32 is the effective coulomb repulsion which arises from coulomb - coupling propagating much more rapidly than phonon - coupling . here
, we are using a value of @xmath32 = 0.13 , which is an average value used commonly for intermetallic superconductors ( see , e.g. , reference ) . having the sommerfeld parameter and the electron - phonon coupling , the noninteracting density of states at the fermi energy can be calculated according to : @xmath33 from the electronic low temperature specific heat data , we have estimated the value of the superconducting gap of all three compounds , according to @xmath34 the obtained values for the electronic and phononic contributions to the superconductivity in heas are summarized in table [ tab:2 ] .
the values for @xmath35(0 ) are similar to comparable intermetallic superconductors and for all three samples the value for 2@xmath35(0)/k@xmath36 is close to the expected value of 3.52 , which is expected for s - wave superconductors according to the bcs model .
the estimated values for the electron - phonon coupling @xmath29 further support that the density of states at the fermi level @xmath37 remains the same for and , while the electron - phonon coupling constant @xmath29 is strongly reduced for the latter material .
this finding supports the general concept that specific elements are essential for optimized superconductivity in compounds .
here we find that the elemental make - up is crucial even in the case of a highly disordered multi - compontent hea superconductor .
we have synthesized the hea for _ x _ = 0.2 , 0.25 , 0.3 , 0.33 , 0.35 , 0.4 , 0.45 , 0.5 , 0.6 , 0.7 , 0.8 , and 0.84 by arcmelting of the elements under argon and by subsequent quenching .
we found from x - ray powder diffraction measurements that all these alloys arrange on a simple bcc crystal lattice ( @xmath38 ) , with unit cell parameters between @xmath39 to @xmath40 within the solid solution .
all prepared samples are found to be bulk superconductors with critical temperatures between @xmath41 4.49 k and 7.92 k. by comparison of the critical temperatures of with the critical temperatures of the transition metals and their alloys in the crystalline form and as amorphous vapor - deposited films , we find the superconducting hea to display characteristics intermediate to both of them .
the valence electron dependence of the transition temperatures for is clearly less pronounced than that seen for crystalline alloys .
however , a maximum is reached right around e / a(d - electron ) = 4.7 , which is an essential feature of the matthias rule for crystalline transition metal superconductors .
therefore , we find that this system neither follows a crystalline nor an amorphous - like trend for this collective electron state .
we find the temperature dependent electrical resistivity @xmath42 of the heas to be metallic and decreasing linearly with decreasing temperature and that the slopes of the upper critical field @xmath13 increase with increasing mixing entropy of the system .
it is , therefore , not the member of this series with the highest critical temperature @xmath9 that has the largest @xmath43(0 ) .
rather , the sample with _ x _
= 0.5 has the largest upper critical field , with a large negative slope of the upper critical field @xmath44 -2.618 t / k and an overall @xmath43(0 ) @xmath45 11.67 t. by reducing the mixing entropy @xmath0 the critical temperatures are found to decrease only slightly from the binary alloy with a critical temperature of @xmath41 9.2 k to the hea .
thus the disorder introduced by the increasing number of constituent atoms does not lead to a loss of the superconductivity or a large decrease of the critical temperature @xmath9 .
we do find however that the effect of elemental make - up is significant for the physical properties even for the highly disordered atoms on the simple lattice in this superconducting hea .
the general interplay of chemical structure , disorder , and superconductivity are topics of fundamental interest .
many known superconductors are posed near structural instabilities , for example , the bismuth oxide superconductors @xcite , the tungsten bronzes @xcite , and also many intermetallic superconductors @xcite .
the superconducting hea studied here offers the unique opportunity to investigate superconductivity on one of the three most fundamental crystal lattices stabilized by high - entropy mixing . for future work , it will be of interest to determine the electronic and phononic densities of states of these alloys in order to understand their interplay .
our results suggest that heas are versatile model systems for the investigation of structure - property relations , as well as for the understanding of the change of electronic properties , on going from crystalline to amorphous superconducting materials .
this work was primarily supported by the gordon and betty moore foundation , epiqs initiative , grant gbmf-4412 . the research performed at the gdansk university of technology
was financially supported by the national science centre ( poland ) grant no .
dec-2012/07/e / st3/00584 .
the electron microscope work done at brookhaven national laboratory was supported by the doe bes , by the materials sciences and engineering division , under contract de - ac02 - 98ch10886 .
99 j - w yeh , s - k chen , s - j lin , j - y gan , t - s chin , t - t shun , c - h tsau , and s - y chang ( 2004 ) nanostructured high - entropy alloys with multiple principal elements : novel alloy design concepts and outcomes . _ adv eng mater _ 6:299 . for a review see : y f ye , q wang , j lu , ct liu , and y yang ( 2015 ) high - entropy alloy : challenges and prospects . _ mater today _ doi:10.1016/j.mattod.2015.11.026 and j - w , yeh ( 2013 ) alloy design strategies and future trends in high - entropy alloys .
_ jom _ 65:12 .
o n senkov , j d miller , d b miracle , and c woodward ( 2015 ) accelerated exploration of multi - principal element alloys with solid solution phases .
_ nat comm _ 6:6529 .
m c troparevsky , j r morris , p r c kent , a r lupini , and g m stocks ( 2015 ) criteria for predicting the formation of single - phase high - entropy alloys .
_ phys rev x _ 5:011041 .
k m youssef , a j zaddach , c niu , d l irving , and c c koch ( 2016 ) a novel low - density , high - hardness , high - entropy alloy with close - packed single - phase nanocrystalline structures .
_ materials research letters _ 3:95 .
h kou , j lu , and y li ( 2016 ) high - strength and high - ductility nanostructured and amorphous metallic materials .
_ adv mater _ 26:5518 .
y zou , h ma , and r spolenak ( 2015 ) ultrastrong ductile and stable high - entropy alloys at small scales .
_ nat comm _ 6:7748 .
b gludovatz , a hohenwarter , k v s thurston , h bei , z wu , e p george , and r o ritchie ( 2015 ) exceptional damage - tolerance of a medium - entropy alloy crconi at cryogenic temperatures .
_ nat comm _ 7:10602 .
b gludovatz , a hohenwarter , d catoor , e h chang , e p george , r o ritchie ( 2014 ) a fracture - resistant high - entropy alloy for cryogenic applications . _
science _ 345:1153 .
r g sharma ( 2015 ) _ superconductivity - basics and applications to magnets _
( springer international publishing , london ) and references therein .
l rossi , superconductivity : its role , its success and its setbacks in the large hadron collider of cern ( 2010 ) _ supercond sci technol _ 23:034001 .
p kozelj , s vrtnik , a jelen , s jazbec , z jaglici , s maiti , m feuerbacher , w steurer , and j dolinsek ( 2014 ) discovery of a superconducting high - entropy alloy .
_ phys rev lett _ 113 , 107001 .
k jasiewicz , b wiendlocha , p korbe , s kaprzyk , and j tobola ( 2016 ) superconductivity of high entropy alloy from first principles calculations . _ phys status solidi rrl _ 10:415 .
o n senkov , j m scott , s v senkova , d b miracle , and c f woodward ( 2011 ) microstructure and room temperature properties of a high - entropy tanbhfzrti alloy , _ j alloys compd _ 509:6043 .
e bucher and j mller ( 1961 ) supraleitung in hexagonalen ti - v - and ti - nb - legierungen .
_ helv phys acta _ 34:410 .
t graf , c felser , ssp parkin ( 2011 ) simple rules for the understanding of heusler compounds .
_ progress in solid state chemistry _ 39:1 .
b t matthias ( 1955 ) empirical relation between superconductivity and the number of valence electrons per atom . _ phys rev _ 97:74 . a simon ( 1997 ) superconductivity and chemistry . _ angewandte chemie _ 36:1788 . for
a review see : e w collings ( 1983 ) _ a sourcebook of titanium alloy superconductivity _ ( plenum press , new york and london ) , and references therein .
m m coilver and r h hammond ( 1973 ) superconductivity in amorphous transition - metal alloy films .
_ phys rev lett _ 30:92 .
m m coilver and r h hammond ( 1977 ) reduced superconducting transition temperatures in amorphous transition metal alloys . _
solid state commun _ 22:55 .
g a mansoori , n f carnahan , k e starling , and t w leland jr ( 1971 ) equilibrium thermodynamic properties of the mixture of hard spheres .
j chem phys 54:1523 .
h takagi , b batlogg , h l kao , j kwo , r j cava , j j krajewski , and w f peck , jr ( 1992 ) systematic evolution of temperature - dependent resistivity in , _ phys rev lett _ 69:2975 .
o gunnarsson , m calandra , and j e han ( 2003 ) colloquium : saturation of electrical resistivity . _
rev mod phys _ 75:1085 .
t p orlando , e j mcniff , jr , s foner , and m r beasley ( 1979 ) critical fields , pauli paramagnetic limiting , and material parameters of and .
_ phys rev b _ 19:4545 .
f von rohr , r nesper , and a schilling ( 2014 ) superconductivity in rubidium - substituted . _ phys rev b _ 89:094505 .
s weyeneth , m bendele , f von rohr , p dluzewski , r puzniak , a krzton - maziopa , s bosma , z guguchia , r khasanov , z shermadini , a amato , e pomjakushina , k conder , a schilling , and h keller ( 2012 ) superconductivity and magnetism in : impact of thermal treatment on mesoscopic phase separation .
_ phys rev b _ 86:134530 .
n r werthamer , e helfand , and p c hohenberg ( 1966 ) temperature and purity dependence of the superconducting critical field .
_ physical review _ 147:295 .
c c koch , and d s easton ( 1977 ) a review of mechanical behaviour and stress effects in hard superconductors .
_ cryogenics _ 7:391 .
w l mcmillan ( 1968 ) transition temperature of strong - coupled superconductors .
_ phys rev _ 167:331 . r c dynes ( 1972 ) mcmillan s equation and the @xmath9 of superconductors .
_ solid state commun _ 10:615 .
p b allen , r c dynes ( 1975 ) transition temperature of strong - coupled superconductors reanalyzed .
_ phys rev b _ 12:3 .
t klimczuk , c h wang , k gofryk , f ronning , j winterlik , g h fecher , j - c griveau , e colineau , c felser , j d thompson , d j safarik , and r j cava ( 2012 ) superconductivity in the heusler family of intermetallics .
_ phys rev b _ 85:174505 .
r j cava , b batlogg , j j krajewski , r farrow , l w rupp jr , a e white , k short , w f peck and t kometani ( 1988 ) superconductivity near 30 k without copper : the perovskite .
_ nature _ 332:814 .
g hinks , d r richards , b dabrowski , d t marx , and a w mitchell ( 1988 ) the oxygen isotope effect in .
_ nature _ 335:419 .
h r shanks ( 1974 ) enhancement of the superconducting transition temperature near a phase instability in . _ solid state commun _ 15:753 .
l r testardi ( 1975 ) structural instability and superconductivity in a-15 compounds . _ rev mod phys _ 47:637 .
d hirai , f von rohr , and r j cava ( 2012 ) emergence of superconductivity in at a structural instability .
_ phys rev b _
86:100505(r ) .
|
high - entropy alloys are made from random mixtures of principal elements on simple lattices , stabilized by a high mixing entropy .
the recently discovered bcc ta - nb - hf - zr - ti high entropy alloy superconductor appears to display properties of both simple crystalline intermetallics and amorphous materials , e.g. it has a well defined superconducting transition along with an exceptional robustness against disorder .
here we show that the valence - electron count dependence of the superconducting transition temperature in the high entropy alloy falls between those of analogous simple solid solutions and amorphous materials , and test the effect of alloy complexity on the superconductivity .
we propose high - entropy alloys as excellent intermediate systems for studying superconductivity as it evolves between crystalline and amorphous materials .
|
according to the standard model of the primordial star formation process , the very first stars , the so - called population iii ( or pop .
iii ) stars form within dark matter ( dm ) halos with virial temperature @xmath1 k and masses @xmath2 @xmath3 that collapsed at redshift @xmath4 25 - 30 or above ( @xcite ) .
the hydrogen atoms in the dm halos combine with the small abundances of free electrons via @xmath5 , followed by @xmath6 , where the free electrons act as catalysts and are present as residue from the epoch of recombination at @xmath4 1100 ( @xcite ) .
the typical fractional abundances of h@xmath0 of @xmath7 are sufficient to permit the primordial gas in the minihalos to cool as it collapses ( see , e.g. @xcite ) via h@xmath0 rotational and vibrational line emission .
this occurs until the gas reaches a temperature @xmath8 200k @xcite at a density of @xmath9@xmath10 , at which point the h@xmath0 energy levels move into the local thermodynamical equilibrium ( lte ) with the kinetic temperature of the gas , and the resulting cooling time becomes longer than the free - fall of the gas .
the gas now heats up slightly as the collapse proceeds .
this transition from cooling to heating with increasing density sets a characteristic jeans length , allowing the gas to fragment with jeans mass of @xmath11 @xmath8 1000@xmath3 at typical temperatures of @xmath12k and number densities @xmath13@xmath10 ( @xcite ) .
once the gas reaches a number density of @xmath14@xmath10 , further h@xmath0 formation is possible via the 3-body reactions @xcite : @xmath15 @xmath16 unfortunately , the rate coefficients for the above set of reactions are very uncertain in the temperature regime which is applicable to pop .
iii star formation , with the current estimates in the literature spanning two orders of magnitude ( see @xcite for discussion ) . however , independent of the choice of rate coefficient , calculations of the gravitational collapse of primordial gas show that by a number density of @xmath14@xmath10 , almost all the atomic hydrogen is converted to h@xmath0 ( @xcite ) .
the sudden formation of @xmath17 has two interesting effects .
first , the process is accompanied by heating , since 4.4 ev is released for every h@xmath0 molecule that is formed .
if the formation rate is fast enough , this heating can momentarily stall the collapse against collapse . on the other hand ,
the ability of the gas to cool is proportional to ( among other things ) the fractional abundance of h@xmath0 , so the more h@xmath0 is formed , the more the gas can cool . in practise ,
the interplay between these two phenomena is complex , and depends crucially on the adopted rates for the 3-body reactions @xcite . under certain conditions ,
a chemo - thermal instability can also occur that may promote the gas to fragment @xcite .
in addition to any fragmentation induced by the chemo - thermal instability , recent simulations conclude that the disk around the first primordial protostar becomes gravitationally unstable and forms a multiple system with low - mass protostars instead of a single protostar @xcite .
this can occur on timescales as short as 10 to 100 years after the formation of the first object .
@xcite show that the fragmentation of this disc results from the fact that the disc is unable to accrete onto the protostar faster than it accretes from the in - falling envelope .
the study by @xcite used both sph and amr simulations to conclude that the choice of 3-body - rate coefficients can introduce significant uncertainty into the radial velocity , temperature and accretion rate , and therefore can affect the rate at which the protostellar disc is fed .
this has recently been verified by @xcite , who used both old rates and newly calculated improved rate given by @xcite , to reach similar conclusion as @xcite .
although they found compressed spiral arms , with strong density contrast between the arms and embedding medium , these simulations were unable to follow the evolution beyond the formation of the protostar , and so could not address the long - term behaviour of the disc structure .
however the degree to which the disc fragmentation depends on the chemical uncertainties in the 3-body h@xmath0 formation rate has never been systematically tested .
previous studies did not include sink particles , and could not study the fragmentation that occurs once the first object is formed . in this work we focus on the effect that the choice of 3-body reaction rate has on the fragmentation of the primordial gas .
we investigate two minihaloshalos obtained from the cosmological simulations of @xcite , which used the hydrodynamic moving mesh code _ arepo _
we use the snapshots from these simulations at a point when the central number density is just below @xmath18 @xmath10 comfortably before the onset of the 3-body reactions as the starting point for this project . interpreting the mesh - generating points of _ arepo _ as lagrangian fluids particles
, we then use the _ arepo _ output as the initial conditions for our gadget-2 @xcite implementation .
the modifications to the standard gadget-2 include a time - dependent chemical network for primordial gas , and a sink particles to capture the formation of collapsing protostellar cores @xcite . a fuller description of the code and its features can be found in @xcite .
the conversion from the moving mesh to smoothed - particle hydrodynamics ( sph ) formalism is the same as that performed by @xcite , and indeed the minihalos are identical , both in terms of the baryonic and dark matter content .
the original _ arepo _ simulations resolve the jeans length with 128 cells @xcite , at which point further refinement is deactivated , resulting in roughly constant mass particles within central @xmath19 1000 au .
however , to investigate the effect of 3-body h@xmath0 reactions , here we use the intermediate _ arepo _ snapshots ( when the central number density @xmath20 @xmath10 ) as the starting point .
although the initial size of the minihalos are @xmath19 3 pc , all the fragmentation and accretion take place in the central region of the halos where the sph particle masses are roughly 10@xmath21 @xmath3 . in gadget-2 simulation
the mass resolution for 100 sph particles ( e.g. @xcite ) is @xmath19 10@xmath22 @xmath3 .
the chemical and thermodynamical treatment of the gas in the _ arepo _ simulations of @xcite is also identical to those in our gadget-2 implementation for the range of densities that are studied in this paper .
more details of the initial conditions for this study are given below .
of the various rates available in the literature for the 3-body h@xmath0 formation reactions , we choose the extreme rate coefficients : the slowest of which is @xcite ( hereafter abn02 ) alongside that of @xcite which provides the fastest of the rates ( hereafter fh07 ) .
we have used the collisional - dissociation rate in the lte limit as described in @xcite . for our sph calculation
we used a temperature - dependent value calculated via the principle of detailed balanced . for
the abn02 runs , data from @xcite are used for low density ( and low temperature , @xmath23 k ) cases , and a power law ( described in table 1 ) for high density and temperatures . for fh07 runs , we used the rates given in @xcite . for the radiative cooling rate
, we use sobolev approximation to calculate it in the optically thick region , as described in detail in @xcite and @xcite .
full details of these rates are given in table 1 .
[ tab:3bh2rates ] @xmath24 & temp range + & ( @xmath25 s@xmath26 ) & + abn02 & 1.3 @xmath27 10@xmath28 & ( @xmath29 ) + abn02 & 1.3 @xmath27 10@xmath30 & ( @xmath31 ) + fh07 & 1.44 @xmath27 10@xmath32 & + & h@xmath0 collisional dissociation rates ( @xmath33 s@xmath26 ) & + abn02 & @xmath34 & + + fh07 & 1.38 @xmath27 10@xmath35 & + the simulations with halo1 start with a maximum central cloud number density ( @xmath36 ) 10@xmath37@xmath10 and minimum @xmath36 of 71@xmath10 .
the initial temperature of the atomic hydrogen is 470k ( max ) and 60k ( min ) .
halo1 contains 1030@xmath38 of gas and 690855 sph particles .
the maximum and minimum mass of sph particles are 0.16 @xmath38 and 1.3 @xmath39 @xmath38 .
this means that the numerical resolution of our simulations for halo1 for 100 sph particles is 1.3 @xmath40 @xmath38 .
the simulations with halo2 starts with a maximum @xmath36 of 10@xmath37@xmath10 and minimum @xmath36 of 85@xmath10 .
the initial temperature of the atomic hydrogen is 436k ( max ) and 54k ( min ) .
halo2 contains 1093@xmath38 of gas and 628773 sph particles .
the maximum and minimum mass of sph particles are 0.05@xmath38 and 1.4 @xmath39 @xmath38 .
this means that the numerical resolution of our simulations for halo2 for 100 sph particles is 1.4 @xmath40 @xmath38 . with altogether 6 million particles in the simulation
, we resolve the halo gas and its velocity structure significantly better than in the sph models of our previous study @xcite .
however , we note that none of the existing numerical calculations are able to fully resolve the turbulent cascade in the halo gas and do justice to the large reynolds numbers ( of @xmath41 or above ) that characterise the accretion flow ( see , e.g. chapter 4 in * ? ? ? * ) .
each of the minihalos are simulated with the two contrasting 3-body reaction rates , resulting in 4 simulations in total .
the different minihalos allow us to distinguish between those features of the high - density collapse / evolution that are caused by the chemo - thermal differences arising from the reaction rates , and those features which result from cosmic variance ( i.e. differences caused by the formation of the minihalos ) .
as collapse progresses in the central region of the cloud , sink particles are created once the number density of the gas reaches 5 @xmath42@xmath10 , at which point the gas has a temperature of around 1000k .
the corresponding jeans mass at this density and temperature is 0.06@xmath38 , so our calculations for both of the halos are well resolved .
we replace a candidate particle by a sink particle ( see , e.g. @xcite ) that can accrete gas particles within its accretion radius @xmath43 that we fix at 6 au , which is the jeans radius at the density threshold for sink creation .
we also prevent sink particles from forming within @xmath44 of one another in order to avoid spurious formation of new sink particles out of gas that , in reality , would by accreted by a neighbouring sink particle .
lastly , gravitational interactions between sinks , and between the sinks and the gas , are softened using a fixed softening parameter of 1.2 au for the sinks .
we run our simulations until a mass of @xmath45 m@xmath46 has been accreted by all the sink particles.the final time scales range between @xmath47 yr due to the physical differences between the minihalos and the 3-body h@xmath0 formation rates .
[ sec : instability ] radially binned , mass - weighted averages of physical quantities for different 3-body h@xmath0 formation rate ( red : abn02 , green : fh07 ) of two different cosmological halos ( dotted : halo1 , solid : halo2 ) , are compared just before the first protostar forms .
the enclosed mass ( a ) , h@xmath0 fraction ( b ) , temperature ( c ) , ratio of cooling time to free - fall time ( d ) and ratio of free - fall time to sound - crossing time ( e ) are plotted as a function of number density . ,
in this section we examine the physical conditions in the gas once the central region has collapsed to a density of @xmath48 @xmath10 .
the radial profiles of the gas for the two different halos and two different 3-body rates are shown in figure [ fig : gas_props ] .
the panels show mass - weighted averages of the properties of individual sph particles within radial logarithmic bins .
panel a shows the mass enclosed as a function of the number density , and we see that it displays a rough power - law over most of the density range covered here .
this simply follows from the radial power - law in the density which is typical for primordial gas : @xmath49 ( e.g. @xcite ) .
we see that there are slight differences between the profiles for the two rates , but these are also comparable to the differences between the two halos . in panel b , we see the rapid rise in the h@xmath0 fraction with increasing density that arises from the 3-body reactions . here
we see a clear difference between the two rates , with the runs with fh07 producing most of the h@xmath0 by a density of @xmath41 @xmath10 , while the abn02 runs are only starting to produce h@xmath0 by the time this density is reached . as the rates are a strong function of density , we see very little difference in the h@xmath0 fractions between the two halos . the mean temperature as a function of the number density is shown in panel c. we find that the gas is significantly cooler , up to a factor of 2 , around density @xmath50 @xmath10 , in the fh07 halos than in those using abn02 .
the differences in temperature - density evolution start around a number density of @xmath51 @xmath10 , once the h@xmath0 abundance in the fh07 case has increased by an order of magnitude .
again , we find that there is little difference between evolution of halos 1 and 2 in this panel , implying that the chemical state of the gas is more important for the temperature evolution than any differences in the halo properties . curves in panels a , b and c are consistent with the results of @xcite .
in this section we study the difference in mass accretion associated with the cloud collapse , that can arise due to the choice of 3-body reaction rates . in what follows
, we will first study the accretion rate as predicted by the properties of the gas in the clouds just before the first sink particle forms .
we will then examine the accretion rate as measured by the sink particles themselves .
the accretion rate can be estimated from the radial gas profiles using , @xmath52 this has been shown to be a good estimate of the accretion rate @xcite .
in addition , one can define an accretion time from , @xmath53 since accretion is related to infall velocity , we also discuss the corresponding radial and rotational velocities .
figure [ vel ] shows these physical quantities as the function of enclosed mass for all the simulations , using the radial profiles in the gas just before the formation of the first sink particle .
the left panel of figure [ vel ] in the clockwise order , shows the accretion rate , radial velocity , radial velocity over sound speed and accretion time respectively .
similarly the right panel of figure [ vel ] shows the rotational velocity , radial velocity over rotational velocity , rotational velocity over keplerian velocity and rotational velocity over sound speed respectively in the clockwise order .
we note that the mass accretion rate ( @xmath54 ) for all the simulations has a maximum at an enclosed mass @xmath8 12 @xmath55 . given that h@xmath0 line - cooling becomes optically thick at the corresponding densities for this enclosed mass , it makes sense that @xmath54 for all the simulations converge at this mass scale , as the gas loses its ability to cool efficiently . in general the differences in accretion rates caused by the choice of 3-body formation rate
are small , around 25 percent .
the differences between the rates in different halos are much larger , at around a factor of 2 for masses below a few tens of solar mass .
however we see from the top left panel of figure [ vel ] that once the mass approaches 100@xmath56 , the difference in the accretion rates can be large , approaching an order of magnitude .
we conclude that halo - to - halo variations will play a larger role in the accretion rates than chemical uncertainties .
the choice of 3-body rate coefficient introduces difference in radial velocity of @xmath57 km s@xmath26
. we also note that the radial velocities for both rates are substantially lower than the rotational velocities .
the gas particles spiral inwards with trajectories following a combination of gravitational attraction and angular momentum conservation .
all clouds are sub - keplerian .
however , there is a slight tendency for the gas to be more rotationally supported in the fh07 runs at least for the inner 10@xmath56 of gas .
we can compare these results with both the amr and sph simulations of @xcite .
our high - resolution study also concludes that the difference in the 3-body @xmath58 formation rates significantly affect the dynamical behaviour of the collapsing gas .
the uncertainty in the velocity structure has also been investigated by @xcite using hydrodynamics code enzo .
they also found that there are relatively large differences in the radial velocities of about @xmath59 km s@xmath26 depending on the halo distribution as well as 3-body h@xmath0 formation rates .
although there is scatter in the velocities for abn02 and fh07 rates , the scatter also exists in the velocities for both halo1 and halo2 .
we again conclude that the halo - to - halo differences in the velocity structure is , in general , more pronounced than that caused by the differences in the chemical reaction rates .
the number of bonnor - ebert masses ( @xmath60 ) in the central dense volume is plotted as a function of enclosed mass , just before formation of first protostar .
the degree of gravitational instability for fh07 within the central dense regime is larger than abn02 , indicating larger opportunity for the gas to fragment in the fh07 case , than in the abn02 simulations.,width=307 ] we have shown in section [ sec : instability ] that the temperature of the gas at high densities , such as those in disc , depends on the choice of adopted 3-body h@xmath0 formation rate . in this section ,
we study how the gas fragments as it evolves beyond the formation of the first collapsing core .
our discussion will first focus on the properties of the gas , to see whether there is any hint of the future evolution already present before the first sink particle forms .
we will then compare the properties of the sink particles in the simulations and discuss the differences .
one useful measure of the ability of the gas to fragment is the ratio of the thermal to gravitational energy .
this can be represented by the number of bonnor - ebert masses ( @xmath61 ) contained in the central dense volume ( @xcite ) .
we therefore investigate the changes in the number ( i.e , @xmath62 ) for each of our simulations and how it varies with the enclosed gas mass , similar to the analysis performed by @xcite .
we computed the bonnor - ebert mass as the mass - weighted average within the logarithmic radial bin , @xmath63 where @xmath64 is the sound speed , @xmath65 is the external pressure that we assume to be equal to the local gas pressure , @xmath66 is the mean molecular weight and @xmath67 is the adiabatic index respectively .
the results are shown in fig .
[ mbe ] . note that the ratio @xmath62 is equivalent to the ratio of fragmentation timescale to that of accretion , @xmath68 . once again
, these snapshots were taken immediately before the formation of the first protostar .
the dashed lines represent the case when fragmentation timescale equals the accretion timescale .
while we see that there are differences between the mass profiles of the fh07 and abn02 runs , the effects are , once again , not substantial .
however , in the case of @xmath62 we see that the fh07 runs have a differently shaped profile from the abn02 runs ( top panel of fig .
[ mbe ] ) . in the fh07 case
, the profiles are peaked at characteristic mass of 1 - 2@xmath69 .
in contrast , the abn02 runs do not display such a peak , and it is less clear whether these runs would favour a particular mass for fragmentation .
if the ratio @xmath70 the gas enclosed in the shells are accreted faster than it can fragment . as a result
fewer new protostars are formed and the available mass contributes to the mass growth of the existing ones @xcite . in our case
we find @xmath71 , i.e. , the gas in the shells can fragment faster than it is accreted by the central dense clump , favouring low - mass protostars . therefore by considering the radial density profiles before the onset of sink formation we conclude that gas in fh07 case is perhaps more likely to fragment than in the abn02 case .
next we study how the gas fragments at later epochs after formation of the first sink particle .
we follow the simulations until @xmath72 have been converted into , or accreted onto , sink particles .
we have calculated the total mass accretion rate by all the sinks particles .
although fragmentation may depend on the way sink particles are modelled , our implementation ( section 2.2 ) takes considerable care to avoid what is often called artificial fragmentation and the spurious formation of sink particles ( further details are found in @xcite , see also @xcite , @xcite , and @xcite ) .
the column density and column - weighted temperature distribution in the inner 2000 au at the end of the simulations are shown in figure [ disk_image ] . in all cases
we see that the simulations exhibit disc structure on several scales within this central region .
it is not surprising , given the high levels of rotational support seen in right - hand panel in figure [ vel ] .
we also see that the temperatures in the disc are slightly less in the case with the fh07 rate , due to the increased h@xmath0 fraction .
however from these images it is clear that all simulations fragment to form a small-@xmath73 system within the time that @xmath74 of material is accreted onto sink particles .
figure [ imf]a shows the time evolution of total mass accretion rate onto the sinks ( same as figure [ vel ] ) .
the plot shows that mass accretion rate for abn02 is larger than the fh07 rate . for both rates ,
dm@xmath75/dt decreases with time .
after @xmath76 years , further sink particles form , and the total accretion rate increases again , however , now with large temporal variations . in figure [ imf]b
, we plot the mass accretion rate as a function of total sink mass .
figure [ imf]c shows the time evolution of the total mass in sinks and figure [ imf]d shows the maximum sink mass in the period over which the sink formation occurs .
we see that the simulations employing the abn02 rate produce a maximum sink mass that is consistently above the fh07 rate , for a given halo . however , the difference is less than a factor of 2 and is similar to the inter - halo difference .
when we examine figure [ imf]e , which shows the total number of sink particle versus the total mass in all sinks , we see a similar behaviour : the differences between the runs with different rates can be a factor of two .
although in this case the halo - to - halo differences are less pronounced .
we caution the reader that our simulations do not include feedback effects from nascent protostars .
we therefore restrain ourselves from following the dynamical evolution of the embedded stellar system over very long timescales .
we end our calculations before we expect that radiative and mechanical feedback changes the thermal as well as chemical state and as a consequence also modifies the fragmentation behavior of the remaining gas ( see , e.g. , the discussion in @xcite , @xcite , @xcite , or @xcite ) .
we have investigated the effects of different proposed 3-body h@xmath0 formation rates during the collapse of primordial star - forming clouds and analysed their influence on the resulting fragmentation behaviour of the gas .
we compared the rates proposed by @xcite and by @xcite , which span two orders of magnitude .
we follow the chemical and thermal evolution of the primordial gas in two different minihalos in order to assess differences introduced by varying halo parameters .
we find that the uncertainty in the 3-body h@xmath0 formation rates leads to the differences in the chemical state of the gas which is more important for the thermal evolution of the clouds than any differences in the halo properties .
the halo to halo variation affects the dynamical evolution of the gas and mass - accretion more than the chemical uncertainties .
we also find that the simulations employing the @xcite rate to produce on average fewer and more massive fragments compared to those calculations using the @xcite rate ( consistent with the results of @xcite which however did not use sink particles ) .
this difference of roughly a factor of two , however , is comparable to the halo - to - halo variation .
this is similar to the results reported by @xcite with added advantage of using sink particles . in summary ,
we find that the variations introduced to the mass distribution of primordial stars introduced by the uncertainties in the three - body h@xmath0 formation rate is of similar order to the fluctuations introduced by difference in the halo properties .
the authors wish to thank thomas greif for providing the data of the minihalos on which our analysis is based
. the authors also acknowledge simon glover , andrea ferrara , kazukai omukai , mahavir sharma for helpful comments .
j.d . is grateful to the raman research institute for financial support and hospitality .
all the computations described here were performed on the hpc - gpu cluster kolob ( funded by heidelberg university and deutsche forschungsgemeinschaft ) .
rsk acknowledges support from the european research council under the european community s seventh framework programme ( fp7/2007 - 2013 ) via the erc advanced grant ` starlight : formation of the first stars ' ( project number 339177 ) .
|
it has been shown that the behaviour of primordial gas collapsing in a dark matter minihalo can depend on the adopted choice of 3-body h@xmath0 formation rate . the uncertainties in this rate span two orders of magnitude in the current literature , and
so it remains a source of uncertainty in our knowledge of population iii star formation .
here we investigate how the amount of fragmentation in primordial gas depends on the adopted 3-body rate .
we present the results of calculations that follow the chemical and thermal evolution of primordial gas as it collapses in two dark matter minihalos .
our results on the effect of 3-body rate on the evolution until the first protostar forms agree well with previous studies .
however , our modified version of gadget-2 sph also includes sink particles , which allows us to follow the initial evolution of the accretion disc that builds up on the centre of each halo , and capture the fragmentation in gas as well as its dependence on the adopted 3-body h@xmath0 formation rate .
we find that the fragmentation behaviour of the gas is only marginally effected by the choice of 3-body rate co - efficient , and that halo - to - halo differences are of equal importance in affecting the final mass distribution of stars .
[ firstpage ] stars : formation stars : early universe hydrodynamics
instabilities
|
the multiple - descriptions ( md ) source coding problem arises naturally in a number of applications such as transmission of video , audio and speech over packet networks and fading channels @xcite@xcite .
the multiple - descriptions ( md ) source coding setup describes a communications setting consisting of one encoder and several decoders .
the encoder receives a discrete memoryless source and wishes to compress it into several _
descriptions_. each decoder receives a specific subset of these descriptions through noiseless links , and produces a reconstruction of the source vector with respect to its own distortion criterion .
the parameters of interest are the rates required for transmitting the description and the resulting distortions at the decoders .
the objective is to design communications schemes which result in the optimal asymptotic trade - off between these two groups of parameters .
the problem has been studied extensively @xcite@xcite@xcite@xcite@xcite@xcite , however , the optimal asymptotically achievable rate - distortion ( rd ) is not known even for the most elementary case when only two descriptions are considered .
the two - descriptions setup is depicted in figure [ 2desc ] .
evidently , for the individual decoders ( which receive only one description ) to perform optimally the encoder must transmit the two - descriptions according to the optimal point - to - point ( ptp ) source coding schemes .
this may require the two - descriptions to be similar to each other . on the other hand , if the descriptions are similar , one of them would be redundant at the central decoder ( which receive two descriptions ) .
in fact , this decoder requires the two - descriptions to be different from one another in order to yield a better reconstruction .
the main challenge in the md problem is to strike a balance between these two situations .
the best known achievable region for the this communications setting is due to zhang and berger @xcite . in the zhang - berger ( zb ) strategy
, the encoder in the first step sends a common and coarsely quantized version of the source on both descriptions , then in the next step , the encoder sends individual refinements for each decoder on the corresponding descriptions .
the zb coding strategy was generalized in @xcite for the case where there are more than two descriptions . in this strategy , first , a common coarsely quantized version of the source is sent to all the decoders , then in the next step , several refinement layers are transmitted . for the symmetric @xmath1descriptions problem ,
a coding scheme based on random binning was considered in @xcite which outperforms the vkg scheme .
this involves generation of independent codebooks followed by random binning . although the md problem has a centralized encoder , the strategy involving random binning
was proved to be useful .
this was further improved upon by a new coding scheme in @xcite based on certain parity - check codes .
however all the three schemes do not fully exploit the common - information among every subset of individual descriptions .
for example in the three - descriptions problem , there can be common - information between the first and second descriptions which is not common with the third description . a new coding scheme called combinatorial message sharing with binning ( cmsb )
was considered in @xcite which provided a unified achievable rd region for the general @xmath0-descriptions problem .
this scheme provided a grand unification of the schemes based on conditional codebooks and the schemes based on random binning , which in turn results in the largest achievable rd region for the problem and subsumes all previous coding schemes .
the name is due to the combinatorial number of common - component codebooks present .
it can be noted that cmsb scheme is based on a construction of random codes where the codewords are mutually independent , and where the codebooks do not have any algebraic structure . in this paper
, we provide a new coding strategy for the general @xmath0-descriptions problem which strictly subsumes cmsb strategy which is the best known in the literature till now .
the coding strategy is based on the common - information perspective . taking a cue from the two - descriptions zb strategy
, we propose that for the general @xmath0-descriptions problem the encoder constructs a common constituent codebook for each subset of the @xmath2 decoders .
so , for each subset of the decoders there is one common component in the overall coding scheme .
this implies that the number of constituent codebooks grows double - exponentially in @xmath0 .
however , we prove that only an asymptotically exponential number of the codebooks are necessary in terms of contributing to the rate - distortion region , and the rest are redundant .
this significantly simplifies the coding strategy .
as an example , for the @xmath3 case , there are @xmath4 possible common code components , but only @xmath5 of the corresponding codebooks are non - redundant .
it turns out that one can identify all of the non - redundant codebooks by associating them with the sperner families of sets @xcite . as a result
, we call the new scheme the sperner set coding ( ssc ) scheme
. the cmsb scheme utilizes @xmath6 codebooks for the @xmath7-descriptions problem .
we prove analytically that the addition of the @xmath7 new codebooks in the ssc scheme results in an improved achievable rd region .
in other words , we show analytically that the cmsb scheme is not complete .
additionally , we propose a generalized binning approach which improves upon the cmsb scheme and further enhances the ssc scheme .
we characterize the asymptotic performance of this coding scheme using computable single - letter information quantities .
this forms the first part of the paper .
similar to the coding scheme of cmsb , the ssc scheme uses random unstructured codes .
it has been observed in several other multi - terminal communications settings such as the broadcast channel ( bc ) @xcite , interference channel ( ic ) @xcite , variations of the mac channel @xcite@xcite and the distributed source coding ( dsc ) problem @xcite , that the application of algebraic structured codes results in improvements over random unstructured codes in the asymptotic performance limits .
based on the inherent dualities between the multi - terminal communication problems and the corresponding coding schemes , these observations suggest that one may get such gains in performance even in the md problem . in the second part of the paper
we show that ssc coding scheme which is based on unstructured codes as mentioned above is not complete .
we provide several specific examples of @xmath7- and @xmath8-description problems and example - specific coding schemes based on random linear codes that perform strictly better than the above ssc coding scheme .
subsequently , we supplement the above ssc scheme with new coding layers which have algebraic structure . we restrict our attention to the algebraic structure associated with finite fields .
we present a unified coding scheme which works for arbitrary sources and distortion measures .
we characterize the asymptotic performance of this coding scheme using computable single - letter information quantities .
we interpret the ssc coding as capturing the common information components among @xmath9 decoders using univariate functions , and the algebraic coding supplement as capturing common information among @xmath9 decoders using bivariate and multivariate functions .
the rest of the paper is organized as follows .
section [ sec : notation ] explains the notation used in the paper .
section [ sec : prelim ] provides an overview of the ideas developed in previous works and provides the groundwork for the next sections . in section [ sec : random ] , we present a new unstructured coding strategy which improves upon the cmsb scheme . we show that there are two different types of gains compared to the previous scheme : the first is due to the addition of several common - component codebook layers , the second is due to a more generalized binning method . in section [ sec : linexamples ] , we identify examples where improvements due to structured coding materialize in the md setup . in this section ,
we investigate three different examples . in two of the examples the achievable rd region
is improved via using linear quantizers , and in the other example the gains are due to linear binning . in section [ sec : rdregion ] we generalize the ideas in the previous section and provide an achievable rd region for the general @xmath0-descriptions problem .
since the characterization of rd region is involved and complicated we provide the final rd region through several steps , adding new coding layers in each step .
section [ sec : conclusion ] concludes the paper .
in this section we introduce the notation used in the paper . we restrict ourselves to finite alphabet random variables .
we denote random variables by capital letters such as @xmath10 and their corresponding alphabets ( finite ) by sans - serif typeface @xmath11 , @xmath12 , respectively .
numbers are denoted by small letters such as @xmath13 .
sets of numbers are also denoted by the sans - serif typeface such as @xmath14 .
specifically , we denote the set of natural numbers by @xmath15 , and the field of size @xmath16 by @xmath17 .
the set of numbers @xmath18 is also denoted by @xmath19 $ ] .
@xmath20 is used to express the vector @xmath21 where @xmath22 . a collection
whose elements are sets is called a family of sets and is denoted by the calligraphic typeface @xmath23 . for a given family of sets @xmath23
we define a set @xmath24 as the set of numbers which are the elements of the sets in @xmath23 .
the family of sets containing all subsets of @xmath25 is denoted by @xmath26 . a collection whose elements are families of sets is denoted by the bold typeface @xmath27 .
the collection of families of sets @xmath28 is also represented by @xmath29 .
random variables are indexed by families of sets as in @xmath30 .
for the purposes of brevity we will write @xmath31 instead of @xmath30 where @xmath32 wherever the notation does nt cause ambiguity .
@xmath33 denotes a vector of length @xmath34 of random variables , each distributed according to the distribution @xmath35 . for @xmath36 and @xmath37
, we denote the set of @xmath34-length vectors which are @xmath38-typical with respect to @xmath39 by @xmath40 .
we use the definition of frequency typicality as given in @xcite in this paper .
we denote a set of random variables as follows @xmath41 . for two collections of families @xmath42 and @xmath43 , we write @xmath44_{(\mathbf{m}_1,\mathbf{m}_2)}$ ] to denote the unordered collection of random variables @xmath45 .
let @xmath46 , and define @xmath47 .
we express this as @xmath48 .
unions , intersections and complements are defined for @xmath49 in the same manner .
a family of sets is called a sperner family of sets if none of its elements is a subset of another element . in other words
a family of sets @xmath50 is a sperner family if @xmath51 . for
any given set @xmath25 , the three families @xmath52 and @xmath53 are all sperner families . for a set @xmath25
, we define the collection of families of sets @xmath54 as the set of all sperner families whose elements are subsets of @xmath25 except for the three trivial sperner families mentioned above .
so we have @xmath55 . for the general @xmath0-descriptions problem
, we define the set @xmath56 $ ] , and this set represents the set of all descriptions .
each decoder receives a subset of these descriptions .
let @xmath57 $ ] for some @xmath34 .
we denote the decoder which receives descriptions @xmath58 by the set @xmath59 .
define the family of sets @xmath60 .
this family of sets corresponds to the set of all possible decoders .
we further explain the notation through an example .
consider the three - descriptions problem . in this case
we have @xmath3 , the set of descriptions are @xmath61 .
there are seven possible decoders .
the set of all decoders is @xmath62 .
consider the two families of sets @xmath63 and @xmath64 . in this case ,
define the set @xmath66 .
the set of random variables @xmath67 is denoted by @xmath68 .
here @xmath69 is a sperner family , but @xmath70 is not a sperner family since @xmath71 and @xmath72 , furthermore @xmath73 but @xmath74 .
the second part of the paper is involves application of linear codes and their cosets .
the following gives a formal definition for such codes , let @xmath16 be a prime number .
a @xmath75 linear code @xmath76 is characterized by its generator matrix @xmath77 defined on @xmath17 .
@xmath76 is defined as follows : @xmath78 .
a coset code @xmath79 is a shifted version of a linear code and is characterized by a generator matrix @xmath77 and a dither @xmath80 defined on @xmath17 .
@xmath79 is defined as follows : @xmath81 .
we will make frequent use of nested linear codes .
a pair of nested linear codes is defined as follows , [ def : pnlc ] for natural numbers @xmath82 , let @xmath83 , and @xmath84 be matrices on @xmath17 . define @xmath85 as the linear codes generated by @xmath86 , @xmath87 $ ] , respectively .
( @xmath88 is called a pair of nested linear codes with the inner code @xmath89 and the outer code @xmath90 .
nested coset codes are defined as shifted versions of nested linear codes .
the general @xmath0-descriptions problem is described in this section .
the setup is characterized by a discrete memoryless source with probability distribution @xmath91 , where @xmath11 is a finite set , and the distortion functions @xmath92 , where @xmath93 is the reconstruction alphabet .
we assume that the distortion functions are bounded , and that the distortion for the @xmath34-length sequence @xmath94 is given by the average distortion of the components @xmath95 .
the discrete , memoryless source @xmath96 is fed into an encoder . the encoder upon receiving a block of length @xmath34 of source symbols produces @xmath0 different indices called _ descriptions _ of the source .
these descriptions are sent to the decoders .
each decoder receives a specific subset of the descriptions .
decoder @xmath97 receives description @xmath98 for all @xmath99 .
based on the descriptions it has received , the decoder produces a reconstruction of the source vector .
an @xmath100 multiple - descriptions code consist of an encoder and @xmath101 decoders : @xmath102 , i \in\mathsf{l},\\ & f_\mathsf{n}:\prod_{i\in \mathsf{n } } [ 1,\theta_i]\to \hat{\mathsf{x}}_\mathsf{n}^n , \mathsf{n}\in \mathcal{l}.\end{aligned}\ ] ] the achievable rate - distortion ( rd ) region is defined as follows , the rd vector @xmath103 is said to be achievable if for all @xmath36 and sufficiently large @xmath34 , there exists an @xmath100 multiple - descriptions code such that the following constraints are satisfied : 1 .
@xmath104 , 2 .
@xmath105\leq d_\mathsf{n}+\epsilon ,
\forall \mathsf{n}\in \mathcal{l}$ ] .
the achievable rd region for the @xmath1descriptions problem is the set of all achievable rd vectors .
although the reconstruction alphabet can be different from the source alphabet , throughout this paper we assume that the two alphabets are the same for the ease of notation .
the results hold for the general case . in this section
we present a brief description of some of the previous known schemes , and state the corresponding inner bounds developed for the achievable rd region .
one of the early strategies for coding over two descriptions was the el gamal - cover ( egc ) strategy @xcite .
similar to all the other strategies explained in this section , the egc scheme relies on random , unstructured codebook generation .
the following theorem describes the corresponding inner bound to the achievable rd region which results from the egc scheme .
note that this is an alternative way to characterize the inner bound described in @xcite . for a joint distribution @xmath106 on random variables @xmath107 and a set of reconstruction functions @xmath108 , the set @xmath109 is defined as the set of rd vectors satisfying the following bounds : @xmath110 the rd vector @xmath111 is achievable for the two descriptions problem , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath113 . in the egc scheme , two codebooks @xmath114 and
@xmath115 are generated independently based on the marginals @xmath116 and @xmath117 .
the two codebooks should be large enough so that the encoder can find a pair of jointly typical codevectors in the two codebooks .
if the codebooks were generated jointly based on the joint distribution @xmath118 , @xmath119 would ensure the existence of such jointly typical codevectors , however in the egc scheme , since the codebooks are generated independently , a _ rate - penalty _ is inflicted on the encoder .
the term @xmath120 in is a manifestation of this _ rate - penalty_. towards reducing the _ rate - penalty _ a new coding strategy was introduced .
the resulting achievable rd region is called the zhang - berger [ zb ] region .
the region is given in the following theorem : for a joint distribution @xmath106 on random variables @xmath121 and set of reconstruction functions @xmath108 , the set @xmath122 is defined as the set of rd vectors satisfying the following bounds : @xmath123 [ thm : zb ] the rd vector @xmath111 is achievable for the two descriptions problem , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath124 .
the closure of the union of all the achievable vectors is called the zb rate - distortion region and is denoted by @xmath125 : @xmath126 the scheme differs from the egc strategy in the introduction of the random variable @xmath127 .
the random variable @xmath127 is called the _ common - component _ between the two descriptions . in the egc scheme , in order to send @xmath128 and @xmath129 , one has to pay the following _ rate - penalty _ :
@xmath130 but in the zb scheme the _ rate - penalty _ is reduced to : @xmath131 the following definition provides a characterization of the common - component between two random variables , let @xmath132 and @xmath133 be two random variables .
@xmath134 is called a common - component between @xmath132 and @xmath133 , if there exist functions @xmath135 such that @xmath136 with probability one , and the entropy of @xmath134 is positive .
it was shown in @xcite that in a certain two - descriptions setup , the addition of @xmath127 enlarges the rd region .
we call such a random variable non - redundant .
the following definition gives a formal description of a non - redundant random variable : in a given achievable rd region for the @xmath1descriptions setup , characterized by a collection of auxiliary random variables , an auxiliary random variable @xmath137 is called non - redundant if the rd region strictly reduces when @xmath137 is set as constant .
[ ex : zb ] we provide an overview of the example in @xcite where the zb rate - distortion region is strictly better than egc rate - distortion region , since it is used extensively in the following sections .
consider the two - descriptions setting . here
@xmath96 is a binary symmetric source ( bss ) , and the side decoders intend to reconstruct @xmath96 with hamming distortion .
the central decoder needs a lossless reconstruction of the source . in @xcite
, it is shown that the rate distortion vector @xmath138 is achievable using the zb scheme but not the egc scheme .
typically , in a given rd region , a codebook is associated with each random variable .
we call a codebook non - redundant if it is associated with a non - redundant random variable . in zb coding scheme , the codebook corresponding to @xmath127 is non - redundant .
the idea of constructing a codebook carrying the common - component between the two random variables is the foundation of most of the schemes proposed for the general @xmath1descriptions problem .
one can even interpret the main difference between these schemes to be the way the common - component between different random variables are exploited .
as explained in the introduction , the best known achievable rd region for the @xmath1descriptions problem is the cms with binning ( cmsb ) strategy . in this strategy a combinatorial number of _ common - component _ random variables are considered .
we explain the coding scheme for the three - descriptions case .
the codebook structure is shown in figure [ fig : cms ] .
there are two layers of codebooks , a layer of maximum - distance separable ( mds ) codes and a layer of source channel erasure codes ( scec s ) . the codebook @xmath139 is decoded at decoder @xmath140 if @xmath141 .
the codebooks are binned independently , and the bin numbers for the mds code @xmath139 are carried by description @xmath142 if @xmath143 . whereas the bin number for each scec is carried by only one description i where @xmath144 .
let @xmath145 denote the resulting rd region achievable using cmsb strategy ( see @xcite ) .
our objective is to provide a new achievable rd region for the @xmath0-descriptions problem , which improves upon the rd region given by the cmsb strategy .
this is based on a new coding scheme involving both unstructured and structured codes .
the achievable rd region and the corresponding coding scheme is presented pedagogically in two steps . in the first step , presented in this section , we provide an rd region achievable using unstructured codes .
this region is strictly better than the cmsb region . in other words
this is an improvement upon the cmsb region using only unstructured codes . in the second step , presented in the next two sections ,
this is enhanced with a structured coding layer which improves the performance even further .
in other words we show that the codebooks associated with the structured coding layer are non - redundant .
we describe the key ideas for the case @xmath3 .
there are @xmath146 distinct decoders , one associated with every non - empty subset of @xmath61 .
that is , we identify the set of decoders with @xmath147 .
the new achievable rd region that we provide improves upon the cmsb rate - distortion region on two factors .
the first comes by adding extra codebooks , and the second comes by a more general binning method . using the common - component perspective , we associate with every non - empty subset @xmath23 of these @xmath146 decoders an auxiliary random variable and a corresponding codebook . that is , we identify the collection of auxiliary variables ( and their codebooks ) with @xmath148 .
each codebook is binned multiple times .
if a description is received by at least one decoder in @xmath23 , then a bin index of the codebook associated with @xmath23 is sent on that description .
although it appears that the strategy involves the generation of a doubly - exponential number of codebooks ( in @xmath0 ) , we show that most of these codebooks are redundant , leaving only an asymptotically exponential number of non - redundant codebooks . while the remaining codebooks are generally non - redundant , only a small number of them are such in most of the examples we consider in this paper .
it turns out that a codebook is non - redundant if and only if it is associated with a a family of sets in @xmath149 .
so , instead of @xmath150 codebooks , we have just @xmath5 . since the indices of the codebooks are associated with the sperner families of sets , we call the scheme the sperner set coding ( ssc ) scheme .
a schematic of the codebook collection is shown in figure [ codebooks ] .
we start from the left and from the top .
the first two codebooks can be identified as @xmath151 and @xmath152 mds codes .
the next six codebooks can be identified as three @xmath153 mds codes , and three @xmath154 mds codes associated with decoders which get two descriptions .
the next three can be identified as @xmath151 source - channel erasure codes ( scec ) .
the next three can be identified as @xmath152 scec s ( similar to the codebooks used in the egc rate region ) .
all these @xmath6 codebooks are considered in deriving the cmsb rate region .
the final set of codebooks are new .
they can be identified as three @xmath153 mds codes associated with decoders that receive disjoint subsets of descriptions .
the following theorem characterizes the achievable rd region for the ssc scheme : for a joint distribution @xmath106 on random variables @xmath155 and @xmath96 and a set of reconstruction functions @xmath108 , the set @xmath156 is defined as the set of rd vectors satisfying the following bounds for some non - negative real numbers @xmath157 : @xmath158 where @xmath159 is the set of all codebooks decoded at decoder @xmath140 , that is @xmath160 , and @xmath161 denotes the set of all codebooks decoded at decoders @xmath162 which receive subsets of descriptions received by @xmath140 , that is @xmath163 .
[ thm : ssc ] the rd vector @xmath164 is achievable for the @xmath1descriptions problem , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath165 . the closure of the union of all such achievable vectors is called the ssc achievable rate - distortion region and is denoted by @xmath166 , @xmath167 in order to clarify the notation we explain the random variables decoded at each decoder in the three - descriptions problem . when @xmath3 , we know @xmath168 has 17 elements . in the formulas , @xmath159 corresponds to the set of random variables decoded at decoder @xmath140 , whereas @xmath169 corresponds to the set of random variables which are decodable if we have access to strict subsets of the descriptions received by @xmath140 . here
are the random variables decoded at decoders @xmath170 and @xmath171 : @xmath172 so as an example @xmath173 which are all the codebooks decoded at decoder @xmath170 .
also @xmath174 , @xmath175 , @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 , @xmath181 , and these are all the codebooks which are decoded at decoders @xmath182 and @xmath183 . [ lem : sscconvex ] the ssc rate - distortion region is convex .
see section [ ap : sscconvex ] in the appendix . for every decoder @xmath184
, we have defined the reconstruction as a function of the random variable @xmath185 . however , decoder @xmath140 decodes all random variables @xmath30 where @xmath186 .
the following lemma shows that the rd region does not improve if the reconstruction function is defined as a function of @xmath187 instead .
[ lem : recfunc ] the rd region in theorem [ thm : ssc ] does not change if the reconstruction function at decoder @xmath140 is defined as a function of @xmath187 .
see section [ ap : recfunc ] in the appendix . in the scheme proposed in theorem
[ thm : ssc ] there are @xmath188 codebooks . we know that the size of @xmath168 is the number of sperner families on @xmath189 minus three .
the number of sperner families is called the dedekind numbers @xcite .
there has been a large body of work in determining the values of dedekind numbers for different @xmath0 .
it is known that these numbers grow exponentially in @xmath0 . as an example
the number of codebooks necessary for @xmath190 and @xmath8 are @xmath7 , @xmath5 and @xmath191 .
however in all of the examples in this paper it turns out that many of the codebooks become redundant and only a small subset are used in the scheme . before proceeding to a more detailed description of the coding strategy we provide a brief outline . for each family of sets
@xmath192 the encoder generates a codebook @xmath139 based on the marginal @xmath35 independently of the other codebooks . intuitively , this codebook is the common - component among all the decoders @xmath140 such that @xmath193 , and it is decoded in all decoders @xmath194 .
codebook @xmath139 is binned independently and uniformly for each description @xmath98 if @xmath195
. the description will carry the corresponding bin number for the codewords in each of the corresponding codebooks .
each decoder reconstructs its corresponding codewords by finding a unique set of jointly typical codevectors in the bins it has received .
the existence of the jointly typical set of codewords is ensured at the encoder by the way of satisfaction of , whereas at the decoder unique reconstruction is warranted by . * codebook generation : * fix blocklength @xmath34 and positive reals @xmath196 . for every @xmath192 ,
generate a codebook @xmath139 based on the marginal @xmath39 with size @xmath197 .
for the @xmath98th description , if @xmath198 , bin the codebook @xmath139 randomly and uniformly into @xmath199 bins ( i.e. randomly and uniformly assign an index @xmath200 $ ] to each codeword in @xmath139 , and the index is called the _ bin - index_. ) .
* encoding : * upon receiving the source vector @xmath201 , the encoder finds a jointly - typical set of codewords @xmath202 .
each description carries the _ bin - indices _ of all the codewords corresponding to its own binning function .
* decoding : * having received the bin - indices from descriptions @xmath99 , decoder @xmath140 tries to reconstruct @xmath203 if @xmath204 . in other words
the decoder finds a unique vector @xmath205 of jointly typical sequences in the corresponding bins .
if the vector does not exist or is not unique , the decoder declares error . * covering bounds : * since codebooks are generated randomly and independently , to find a set of vectors @xmath206 that is jointly typical with the source vector @xmath201 , the mutual covering bounds are necessary based on the mutual covering lemma @xcite .
* packing bounds : * for decoder @xmath140 , description @xmath98 is received if @xmath99 .
since binning is done independently and uniformly , to find a unique set of jointly typical sequences @xmath205 , the mutual packing bounds are required by the mutual packing lemma @xcite .
there are two main differences between the new scheme and the previous cmsb scheme .
first there are additional codebooks present .
as an example in figure [ codebooks ] , the three codebooks in the right column are not present in the cmsb scheme .
second , description @xmath98 bins all of the codebooks @xmath23 such that @xmath207 .
we will show in the next sections that these additional codebooks contribute to an enlargement of the achievable rd region .
in other words we prove that all of the additional codebooks are non - redundant .
also we show that the new binning strategy improves the achievable rd region .
consider the general @xmath0-descriptions problem . in this section
we prove that a codebook @xmath203 is non - redundant if @xmath192 .
it is straightforward to see that addition of a codebook @xmath139 where @xmath208 is not going to result in a larger achievable rd region . to see this consider the three descriptions problem and
assume we add the codebook @xmath209 . by our definition
this new codebook is decoded if we either receive description 1 or both descriptions 1 and 2 . in this case
the codebook is decoded in exactly those decoders where @xmath114 is decoded .
this means that merging these two codebooks does not change the packing bounds whereas it may relax the covering bounds .
so such a codebook would be redundant .
this is the reason why we consider only those codebooks which are associated with sperner families .
there are three sperner families for which we do not construct codebooks : @xmath210 .
it is clear that @xmath211 and @xmath212 are not necessary since they are not decoded at any decoder .
furthermore one can use the proof provided in @xcite to show that @xmath213 is also redundant .
the next lemma proves that the random variables considered in theorem [ thm : ssc ] are non - redundant .
the random variable @xmath30 is non - redundant for every @xmath214 .
[ theoremrandnonred ] [ redthe ] is not redundant.,width=336 ] we provide the proof for the @xmath3 case and give an outline of how the proof is generalized for @xmath215 .
the codebooks @xmath114 , @xmath115 , @xmath216 , @xmath217 , @xmath218 , @xmath219 , @xmath220 , @xmath221 , @xmath222 , @xmath223 , @xmath224 , @xmath225 , @xmath226 , @xmath227 are all present in the cmsb scheme and it was shown that they are non - redundant .
the new codebooks are @xmath228 , @xmath229 and @xmath230 .
we prove that @xmath231 is non - redundant using the following example , the two other codebooks are non - redundant by symmetry .
we build on example [ ex : zb ] to construct a three - descriptions example as shown in figure [ nonred ] .
as explained in the previous section , it is known that @xmath127 is non - redundant .
let @xmath232 , and @xmath233 .
let @xmath234 let @xmath235 be the set of probability distributions @xmath236 , such that @xmath237 @xmath238 @xmath239 belongs to @xmath240 for some @xmath241 as given in theorem [ thm : zb ] .
define the joint distribution @xmath242 as follows : @xmath243 let @xmath244 be the marginal distribution of @xmath127 and @xmath96 .
define a random variable @xmath134 that is correlated with @xmath96 such that @xmath245 .
let @xmath246 be a binary random variable independent of @xmath96 and @xmath134 with @xmath247 . define @xmath248 where @xmath249 denotes the logical and function .
let @xmath250 be the induced joint and conditional distributions , respectively .
[ ex : binning ] we proceed by explaining the new example .
the source @xmath96 is a bss , decoders @xmath251 and @xmath183 want to reconstruct the source with respect to hamming distortion and the central decoder wants to reconstruct the source losslessly .
decoder @xmath170 wants to reconstruct the source with respect to the distortion function given by : @xmath252 [ lemma1 ] the following rd vector does not belong to @xmath145 , where @xmath253 is constant .
the vector belongs to @xmath166 given in theorem [ thm : ssc ] which is achievable using the ssc scheme : @xmath254 where @xmath255 .
we provide the intuition behind the proof first . in the coding scheme in theorem
[ thm : ssc ] , the only codebooks capable of carrying the common - component between decoders @xmath251 and @xmath183 are @xmath221 , @xmath222 , @xmath226 and @xmath228 .
we have set the distortion constraint at decoder @xmath170 such that this common message ca nt be carried exclusively on either of the descriptions @xmath256 and @xmath257 , but rather both descriptions are necessary for the reconstruction of the common codebook .
so the codebook @xmath228 ca nt be empty .
the proof is provided in section [ ap : lemma1 ] in the appendix .
so far we have shown that the additional codebooks are non - redundant when @xmath3 .
the argument can be extended to the case when @xmath215 , an outline of the general argument is provided in appendix [ app : addcodegen ] .
the second factor contributing to the gains in the ssc rate - distortion region is the binning method . in the ssc scheme
all descriptions @xmath258 carry independent bin indices of codebook @xmath139 .
this is different from the cmsb strategy where each codebook is binned by a specific subset of the descriptions based on whether the codebook is a scec or an mds codebook .
we prove through a three - descriptions example that the rd region enlarges due to binning in the ssc scheme , even with the three additional codebooks .
we show in the following example that the bin indices of @xmath223 should be carried by all descriptions .
[ ex : binning ] the example is generated by modifying example [ ex : binning ] and is illustrated in figure [ binningexample ] .
the source @xmath96 is bss .
@xmath259 is defined as in example [ ex : binning ] .
decoders @xmath251 and @xmath260 want to reconstruct the source with hamming distortion and decoder @xmath261 wants to reconstruct the source losslessly .
[ theorem5 ] in order to achieve @xmath262
@xmath263 @xmath264 @xmath265 we must have @xmath266 .
see section [ ap : theorem5 ] in the appendix .
before providing a unified rd region which uses both unstructured and structured codes ( step 2 ) , in this section , for pedagogical reasons , we look at three examples of @xmath1descriptions problems and provide example - specific coding schemes based on linear codes that perform strictly better than the ssc scheme which is based on unstructured codes .
this shows that the ssc region is not complete and a structured coding layer is necessary .
these coding schemes are unified and presented in the next section .
we create a three - descriptions setting where reconstructions of bivariate functions are necessary .
[ ex : vecsource ] consider the three - descriptions example in figure [ 3d ] . here
@xmath96 and @xmath267 are independent bss .
decoder @xmath268 and @xmath183 wish to reconstruct @xmath96 , @xmath267 and @xmath269 , respectively , with hamming distortion .
decoders @xmath270 , and \{2,3 } wish to reconstruct the pair @xmath271 with distortion function @xmath272 we are interested in achieving the following rd vector : @xmath273 first we argue that in this example , description 3 should carry a bivariate function of descriptions 1 and 2 .
decoders @xmath170 and @xmath182 operate at the optimal ptp rate - distortion function .
so the corresponding descriptions have to allocate all of their rates to satisfy their individual decoder s distortion criteria .
since the distortion constraint at decoder @xmath170 only relates to @xmath96 , this description only carries a quantization of @xmath96 , and by the same argument description 2 carries a quantization of @xmath267
. then description 3 has to carry the sum of these two quantizations so that the joint decoders distortion constraints are all satisfied .
since structured codes are efficient for transmitting bivariate summations of random variables , we expect that using structured codes would give gains in this example as opposed to unstructured codes .
first , we prove that the rd vector is achievable using linear codes .
the rd vector in ( [ rdv3 ] ) is achievable . *
encoding : * construct a sequence of random linear codes @xmath274 of rate @xmath275 , where @xmath276 is going to 0 .
it is well known that such a sequence of linear codes can be used to quantize a bss to hamming distortion @xmath277 .
define the following : @xmath278 since @xmath279 and @xmath280 are codewords and the codebook is linear , @xmath281 is also a codeword .
description 1 carries the index of @xmath279 , description 2 carries the index of @xmath280 and description 3 carries the index of @xmath281 . *
decoding : * decoders @xmath170 and @xmath182 , receive @xmath279 and @xmath280 , respectively ,
so they satisfy their distortion constraints .
decoder @xmath183 reconstructs @xmath281 .
lemma [ distsum ] shows that the distortion criteria at this decoder is satisfied .
[ lemma2 ] in the above setting , we have @xmath282 . [ distsum ]
see section [ ap : lemma2 ] in the appendix .
decoder @xmath251 receives @xmath279 and @xmath280 , so it satisfies its distortion requirements .
also decoders @xmath260 and @xmath171 can recover @xmath279 and @xmath280 by adding @xmath281 to @xmath279 and @xmath280 , respectively .
this shows that the rd vector in ( [ rdv3 ] ) is achievable using linear codes .
next we show that the ssc scheme can not achieve this rd vector .
[ theorem7 ] the rd vector in ( [ rdv3 ] ) does not belong to @xmath166 , i.e. , it is not achievable using the ssc scheme .
see section [ ap : theorem7 ] in the appendix .
in the ssc scheme , there are two stages in the codebook generation phase . in the first stage unstructured codebooks are generated randomly and independently , and in the second stage these codebooks are binned randomly in an unstructured fashion for each description . in the previous example it was shown that in the first stage , it is beneficial to generate codebooks with a linear structure .
however in that example there was no need for binning . in the next example , we show that the binning operation needs to be carried out in a structured manner as well .
this is analogous to the gains observed in the distributed source coding problem @xcite where the bin structure needs to be linear .
consider the four - descriptions example in figure [ 4d ] .
[ ex : vecsourcebin ] @xmath96 and @xmath267 are bss s . @xmath96 and @xmath267 are not independent , and they are related to each other through a binary symmetric channel with bias @xmath283 . in other words
@xmath284 where @xmath285 is independent of @xmath96 and @xmath267 .
decoders @xmath170 and @xmath286 wish to decode @xmath96 and @xmath267 , respectively , with hamming distortion .
decoders @xmath251 , @xmath287 and @xmath171 require a lossless reconstruction of @xmath96 , @xmath267 and @xmath269 , respectively .
we are interested in achieving the following rd vector : @xmath288 we show that the rd vector in ( [ rd4 ] ) is achievable using structured codebooks and linear binning in the next lemma .
the rd vector in ( [ rd4 ] ) is achievable . * codebook generation : * take an arbitrary sequence of positive numbers @xmath276 , where @xmath289 as @xmath290 . for any large @xmath291 , fix @xmath292 and @xmath293 . construct a family of nested coset codes @xmath294 where @xmath295 such that the rate of the outer code is @xmath296 and the rate of the inner code is @xmath297 .
choose @xmath298 such that it is a good channel code for a bsc(@xmath299 ) , and choose @xmath300 such that it is a good source code for quantizing a bss to hamming distortion @xmath299 .
the existence of such nested coset codes is well - known from random coding arguments @xcite .
next we bin the space @xmath301 into shifted versions ( cosets ) of @xmath302 .
let @xmath303 be the voronoi region of the codeword @xmath304 in @xmath305 .
any vector @xmath306 can be written in the form @xmath307 .
define the @xmath98th bin as @xmath308 .
this operation bins the space into @xmath309 bins .
the bin number associated with an arbitrary vector @xmath310 determines exactly the quantization noise resulting from quantizing the vector using @xmath311 with the minimum hamming distortion criterion .
we denote the bin number of @xmath310 as @xmath312 .
a similar binning operation can be performed using @xmath300 .
denote the bin number of @xmath310 obtained using shifted versions of @xmath300 by @xmath313 . * encoding : * the encoder quantizes @xmath310 and @xmath314 using @xmath315 to @xmath316 , and @xmath317 , respectively .
it also finds the bin number of the two source sequences @xmath312 and @xmath318 .
@xmath316 is transmitted on the first description , @xmath312 is transmitted on the second description , @xmath318 is transmitted on the third description , and @xmath317 is transmitted on the fourth description . * decoding : * since the outer codes are good source codes , the distortion constraints at decoders @xmath170 and @xmath286 are satisfied .
we argue that the voronoi region of @xmath304 in @xmath300 is a subset of the one for @xmath305 .
this is true since @xmath319 .
hence , having @xmath312 , decoders @xmath251 and @xmath287 can calculate @xmath313 . as mentioned above
the bin number determines the quantization noise , so the decoders can reconstruct the source losslessly using the bin number and the quantization vector .
decoder @xmath171 receives @xmath312 and @xmath318 .
we have @xmath320 and @xmath321 , so @xmath322 . since @xmath305 is linear
, @xmath323 is a codeword , and @xmath324 can be thought of as the noise vector for a @xmath325 .
we constructed @xmath305 such that it is a good channel code for bsc(p ) , so the decoder can recover @xmath326 from @xmath327 .
then by subtracting the two vectors it can get @xmath324 .
although we have used linear codes for quantization as well as binning , the linearity of the binning codebook @xmath305 is critical in this example .
in fact , it can be similarly shown that one can achieve the rd vector in ( [ rd4 ] ) with @xmath328 chosen to be a union of random cosets of @xmath305 .
this is in contrast with the previous example where the quantizing codebook was required to be linear .
[ theorem9 ] the rd vector in ( [ rd4 ] ) is not achievable using the ssc scheme . see section [ ap : theorem9 ] in the appendix .
it can be noted that in the case of ssc scheme , the unstructured quantizers are generated randomly and independently . as observed in these two examples , in order to efficiently reconstruct the bivariate summation , it is beneficial to use the same linear code for quantizing the source . however , in the two examples the source was a vector with two components which were separately quantized using identical linear codes , and the analysis of the coding scheme required only standard ptp covering and packing bounds for linear codes . in the more general case , evaluation of the performance of identical , and more generally , correlated linear codes for md quantization , requires new covering and packing bounds .
this is illustrated through the following scalar source example which is depicted in figure [ scalarfig ] .
the setup is constructed based on the no - excess rate example described in @xcite for the two - descriptions problem . in the two - descriptions example , the source @xmath96 is bss , and the distortion functions at all decoders is hamming distortion .
for the special case , called no - excess rate regime , when @xmath329 , it was shown that the egc region is tight .
here @xmath330 is the distortion @xmath331 at decoder @xmath251 , and the minimum side distortion
@xmath332 achievable was shown to be @xmath333 .
the three - descriptions example is given as follows .
[ ex : scalar ] the source @xmath96 is bss , the distortion functions at decoders @xmath170 , @xmath182 , @xmath251 , @xmath260 and @xmath171 are hamming distortions , and the distortion function at decoder @xmath334 is the following general distortion function , @xmath335 where @xmath336 and @xmath337 are positive real numbers .
we are interested in achieving the rd vectors with the following projections : @xmath338 our objective is to evaluate the optimal @xmath339 trade - off .
the following lemma provides the rd vectors achievable using linear codes .
the rd vector in ( [ rdscalar ] ) is achievable using linear codes , as long as the following constraints are satisfied : @xmath340 [ scalartheorem ] consider the following definition .
let @xmath17 be a field .
consider 3 random variables @xmath96 , @xmath137 and @xmath341 , where @xmath96 is defined on an arbitrary finite set @xmath11 , and @xmath137 and @xmath341 are defined on @xmath17 .
fix a pmf @xmath342 on @xmath343 . a sequence of code pairs @xmath344 , where @xmath345 for @xmath346 , is called @xmath347-covering if @xmath348 , @xmath349 [ p_cov ] first , we derive new covering and packing bounds for joint quantization of a general source @xmath96 ( i.e. not necessarily binary ) , using two pairs of nested coset codes .
let @xmath350 and @xmath351 be two pairs of nested coset codes with generator matrices @xmath352 and @xmath353 shown in figure [ codebookssum ] which share the inner code @xmath89 . if @xmath354 , the two codebooks are generated independently . on the other hand ,
if @xmath355 , the two codebooks are the same , so this construction generalizes the previous constructions .
, height=201 ] [ thm : same_code ] for any @xmath347 on @xmath343 and rates @xmath356 and @xmath357 satisfying ( [ newe])-([neweq ] ) , there exists a sequence of two pairs of nested coset codes @xmath358 and @xmath359 which are @xmath347-covering .
@xmath360 see section [ ap : thm : same_code ] in the appendix .
the only difference between the new mutual covering bounds and the ones for independent codebook generation is the presence of the constraint ( [ neweq ] ) . if @xmath354 , ( [ neweq ] ) is redundant , so we recover the mutual covering bounds for independent codebook generation as expected .
if @xmath361 , ( [ neweq ] ) is non - redundant .
there is an intuitive explanation for this additional bound .
define @xmath362 .
@xmath363 is a coset code with generator matrix @xmath364^t$ ] , and the size of this codebook is @xmath365 .
suppose there are codevetors @xmath366 and @xmath367 jointly typical with @xmath368 with respect to @xmath369 , then @xmath370 is jointly typical with @xmath368 with respect to @xmath371 .
this implies that @xmath363 should have size at least @xmath372 by the converse source coding theorem .
let @xmath373 and @xmath96 be as in definition [ p_cov ] .
a sequence of code pairs @xmath344 and bin functions @xmath374 , i\in \{1,2\}$ ] is called @xmath347-packing if for all @xmath36 , @xmath375 [ thm : same_codepack ] for any @xmath347 on @xmath343 , there exists a sequence of two pairs of nested coset codes @xmath376 and @xmath359 and bin function @xmath377 which are @xmath347-packing , if @xmath378 and @xmath379 satisfy @xmath380 see section [ ap : thm : same_codepack ] in the appendix .
we proceed with explaining the achievability scheme .
define the joint distribution in table [ tb : jointdist ] on random variables @xmath381 and @xmath96 .
x2cm|x2cm|x2cm|x2cm|x2cm| & & & & + 0&@xmath382&@xmath383&@xmath383&@xmath384 + 1&@xmath385&@xmath386&@xmath386&@xmath387 + * codebook generation : * set @xmath388 , and @xmath389 and @xmath390 .
construct a family of coset codes @xmath76 with rate @xmath391 . also , construct three binning functions @xmath392 , i\in \{1,2,3\}$ ] .
* encoding : * upon receiving source sequence @xmath310 , the encoder finds @xmath393 and @xmath394 in the codebook , such that they are jointly typical with @xmath310 with respect to @xmath395 .
such a pair of codewords exists as long as the covering bounds in lemma [ thm : same_code ] are satisfied . in the case at hand
it can be readily checked that @xmath356 and @xmath357 satisfy the bounds .
description 1 carries the bin index of @xmath393 using @xmath396 , description 2 carries the bin index of @xmath394 using @xmath397 and description 3 carries the bin index of @xmath398 using @xmath399 .
* decoding : * decoder @xmath400 receives the bin index carried by description 1 , and reconstructs @xmath401 as long as there is a unique codeword in the bin which is typical with respect to @xmath402 .
the following packing bound ensures correct decoding with arbitrarily small error : @xmath403 by the same arguments decoder @xmath404 reconstructs @xmath405 correctly .
decoder @xmath334 reconstructs @xmath398 with arbitrarily small error since the following packing bound is satisfied : @xmath406 we conclude that all the decoders which receive two descriptions would have access to @xmath393 and @xmath394 .
decoders @xmath170 , @xmath182 and @xmath183 announce their decoded codewords as their reconstruction of the source .
the reconstruction function at the decoders which receive two descriptions is given as follows :
@xmath407 this implies that the rd vector stated in the lemma is achieved from strong typicality .
the following lemma shows that some of the rd vectors in lemma [ scalartheorem ] are not achievable using the ssc scheme .
[ theorem11 ] the rd vector in ( [ rdscalar ] ) is not achievable using the ssc scheme for the following values of @xmath336 and @xmath337 and when the equality holds in ( [ eq : nonredconstraint ] ) : @xmath408 for example , @xmath409 and @xmath410 satisfy the above constraints , where we have rounded the parameters up to the third decimal place .
see section [ ap : theorem11 ] in the appendix .
in this section , we provide a new achievable rd region for the general @xmath1descriptions problem by enhancing the ssc coding scheme with a structured coding layer .
we present this region in four stages . in the first stage , we prove that the ssc region can also be achieved using structured codes .
in particular , we use independent nested coset codes for each auxiliary random variable , and exploit the pairwise independence of the codewords to show the achievability of the ssc region . in the subsequent stages , we add coding layers that facilitates the reconstruction of multi - variate functions of the auxiliary random variables . the improvements due to these additional layers comes from exploiting the algebraic structure of the codebooks .
in the second stage , we only allow the reconstruction of a bivariate summation of codewords . in the third stage
we extend this to a multi - variate summation of the codewords . in the fourth stage , we consider the general case involving the reconstruction of an arbitrary number of multi - variate summations at the decoders . for a joint distribution @xmath106 on random variables @xmath411 and @xmath96 , and a set of reconstruction functions @xmath108 , the set @xmath412 is defined as the set of rd vectors satisfying the following bounds for some non - negative real numbers @xmath413 : @xmath414}\!\ ! \rho_{\mathcal{m},j}\!\!-r_{o,\mathcal{m}}),\!\!\!\!\!\ ! \quad \forall \mathbf{l}\subset \mathbf{m}_\mathsf{n } , \forall \mathsf{n}\in \mathcal{l}\\ & r_i=\sum_{\mathcal{m } } \rho_{\mathcal{m},i } , \qquad d_{\mathsf{n}}=e\big\{d_{\mathsf{n}}(h_{\mathsf{n}}(u_{\mathsf{n}},x))\big\}.\nonumber\end{aligned}\ ] ] where @xmath415 .
[ linind ] [ theorem12 ] the rd vector @xmath164 is achievable for the @xmath1descriptions problem using nested coset codes , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath416 .
the encoding and decoding steps are exactly the same as the ones in the proof of theorem [ thm : ssc ] .
the only difference is in the codebook generation phase . in this phase , for every @xmath192
, we generate a coset code @xmath139 with rate @xmath417 , generator matrix @xmath418 , and dither @xmath419 .
@xmath418 and @xmath419 are generated randomly and uniformly for every @xmath23 .
the bounds in ( [ lc1 ] ) are the mutual covering bounds for independently generated coset codes .
these bounds ensure encoding can be carried out without error .
the bounds in ( [ lc2 ] ) are the mutual packing bounds in each decoder .
they ensure errorless decoding .
[ lemma5 ] the rd region in theorem [ theorem12 ] is equal to the ssc rd region . see section [ ap : lemma5 ] in the appendix . in the first stage we constructed one codebook for each subset of the decoders .
however , only the codebooks corresponding to the sperner families of sets are shown to be non - redundant .
we interpret this using the notion of common - information as defined by gacs , krner , witsenhausen @xcite @xcite .
let @xmath420 denote the common information between any two random variables @xmath421 and @xmath422 .
the common information among @xmath423 random variables @xmath424 is a vector of length @xmath425 of information that is common among every subset of @xmath423 random variables of size at least two .
when @xmath426 , the common information is given by @xmath427.\ ] ] this was referred to as univariate common information in @xcite , as each of these components are characterized using univariate functions .
we interpret the scheme in the first stage ( ssc scheme ) as capturing the common - information components among the random variables associated with @xmath9 decoders using univariate functions . for @xmath426 , this notion of common information was generalized using bivariate functions to the following seven - dimensional vector in @xcite : @xmath428.\ ] ] there are seven degrees of freedom in having information common among @xmath7 random variables .
the latter three are called bivariate common information components as they are characterized using bivariate functions of random variables . in this sense
, the addition of the structured coding layers in the next stages can be thought of as capturing the common - information among @xmath9 decoders using bivariate and , more generally , multivariate functions .
we extend the notion of bivariate common information to @xmath429 random variables as follows . to characterize a bivariate common information component
, we consider three subsets of @xmath430 and @xmath431 of @xmath432 .
define @xmath433 as a bivariate common information component among @xmath424 , where @xmath434 is the information that is common among @xmath435 .
for example for @xmath436 , let @xmath437 , @xmath438 and @xmath439 .
this characterizes the information in @xmath440 that can be computed by a conference via a bivariate function of ( i ) the information common between @xmath421 and @xmath422 , and ( ii ) the information in @xmath441 .
this concept can be extended to define multivariate common information among @xmath423 random variables .
we return to our discussion on the achievable rd region for the md problem , where @xmath442 . in the second stage ,
we aim to capture the bivariate common information among random variables associated with @xmath2 decoders .
in particular , we reconstruct a summation of two codebooks . from the above arguments , instead of one codebook for each subset of decoders as in the first stage , in this stage we need to construct one codebook for every triple of subsets of the decoders . for a given triple of sets of decoders ,
the third set of decoders reconstruct a bivariate summation of a random variable corresponding to the first subset and a random variable corresponding to the second subset of decoders .
this is explained in more detail next .
we add two new codebooks to the ssc scheme .
the underlying random variables for these two codebooks are denoted by @xmath443 and @xmath444 where @xmath445 , @xmath446 .
we construct two pairs of nested coset codes for these two random variables .
the two nested coset codes have the same inner code .
the codebook corresponding to @xmath447 is decoded at decoder @xmath140 if @xmath448 , furthermore , the sum of the two codebooks is decoded at decoder @xmath140 if @xmath449 , where @xmath450 is an element of @xmath168 .
for example , let us choose @xmath451 . in this case
the first codebook is decoded whenever description 1 is received , the second codebook is decoded if description 2 is received , and the sum is decoded whenever description 3 is received .
this corresponds to the coding schemes we presented for example [ ex : vecsource ] , where @xmath452 and @xmath453 .
the following theorem describes the achievable rd region using this scheme . for any three distinct families @xmath454 , and for a joint distribution @xmath106 on random variables @xmath411 , @xmath455 , and @xmath96 , where the underlying alphabet for all auxiliary random variables is the field @xmath17 , and a set of reconstruction functions @xmath108 , the set @xmath456 is defined as the set of rd vectors satisfying the following bounds for some non - negative real numbers @xmath457 and @xmath458 and @xmath459 : @xmath460_{\overline{{\mathbf{m}}}_\mathsf{n}}|[u , v , w]_{\widehat{\mathbf{m}}_\mathsf{n}\cup \overline{\mathbf{l}}})\leq \!\!\!\!\!\!\!\!\!\sum_{\mathcal{m}\in { { \mathbf{m}}}_\mathsf{n}\backslash\widetilde{\mathbf{m}}_\mathsf{n}\cup { \mathbf{l}}}\!\!\!\!\!\!\ ! ( \log q + \!\ !
\sum_{j\in \widetilde{\mathcal{m } } } \rho_{\mathcal{m},j}-r_{o,\mathcal{m}})+\!\!\!\!\!\!\!\!\!\!\!\!\sum_{\substack{\mathcal{m}\in{\mathbf{m}}_\mathsf{n}\backslash\widetilde{\mathbf{m}}_\mathsf{n}\cup \overline{\mathbf{l}}\\\bigcap\{\mathcal{a}_i|i\in[1,3]\ } } } \!\!\!\!\!\!\!\!\!\!(\log{q}+\sum_{j\in \widetilde{\mathcal{m}}}\rho_{o,\mathcal{m},j}-r'_{o,\mathcal{m } } ) , \quad \forall \overline{\mathbf{l}}\subset { \overline{\mathbf{m}}}_\mathsf{n } \label{sec1pack1 } \\&r_i=\sum_{\mathcal{m } } \rho_{\mathcal{m},i } , \qquad d_{\mathsf{n}}=e\big\{d_{\mathsf{n}}(h_{\mathsf{n}}(u_{\mathsf{n}},x))\big\}.\label{sec1rd } \ ] ] where ( a ) @xmath461 , ( b ) @xmath462 , ( c ) @xmath463 , ( d ) @xmath464 , ( e ) @xmath465 , and ( f ) @xmath466 to denote subscripts of random variables throughout the paper .
however , the collection @xmath467 is used as the subscript for @xmath134 since the random variable is defined using @xmath336 and @xmath337 . ] .
[ linoneadd ] the rd vector @xmath164 is achievable for the @xmath1descriptions problem , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath468 . before providing the proof we explain the bounds in the new rd region .
( [ sec1cov1 ] ) and ( [ sec1pack1 ] ) are the mutual covering and packing bounds which are also present in the theorem [ thm : ssc ] , respectively .
( [ sec1cov2 ] ) is a generalization of the additional covering bound derived in the lemma in [ thm : same_code ] .
note that the common component among decoders @xmath469 is the pair @xmath470 , and similarly for @xmath471 .
the common component among decoders @xmath472 is the pair @xmath473 , and observe that @xmath474 . given a joint distribution @xmath475 , and codebook and binning rates satisfying the bounds in the theorem we prove achievability of the rd vector in ( [ sec1rd ] ) .
* codebook generation : * fix blocklength @xmath34 . for every @xmath192 , independently generate a linear code @xmath139 with size @xmath476 .
also generate two nested coset codes @xmath477 where the inner code has rate @xmath459 and the outer codes have rates @xmath478 .
define the set of codewords @xmath479 .
the size of @xmath480 is @xmath481 , where @xmath482 . for the @xmath98th description bin
the codebook @xmath139 randomly and uniformly with rate @xmath199 . *
encoding : * upon receiving the source vector @xmath201 , the encoder finds a jointly - typical set of codewords @xmath483 .
each description carries the _ bin - indices _ of all of the corresponding codewords .
the encoder declares an error if there is no jointly typical set of codewords available . *
decoding : * having received the bin - indices from descriptions @xmath99 , decoder @xmath140 tries to find a set of jointly typical codewords @xmath484 .
if the set of codewords is not unique , the decoder declares error .
in order for the encoder to find a set of jointly typical codewords , the mutual covering bounds ( [ sec1cov1 ] ) and ( [ sec1cov2 ] ) should hold .
this is a generalization of the result in lemma [ thm : same_code ] and we omit the proof for brevity .
the bounds in ( [ sec1pack1 ] ) are the mutual packing bounds at each decoder . here
we have considered the general case where @xmath485 are chosen arbitrarily from @xmath168 .
it turns out that only certain choices of @xmath485 would give non - redundant codebooks and thus provide improvements over the ssc scheme .
one can show that the codebooks are redundant if @xmath486 .
for example take @xmath487 , @xmath488 and @xmath489 . in this section
we reconstruct a multi - variate summation of an arbitrary number @xmath423 of random variables at one decoder where @xmath490 and the summation is with respect to a finite field @xmath17 .
following the steps in the previous section , we add @xmath423 new codebooks to the original ssc scheme .
let @xmath491 $ ] .
the underlying random variables for these codebooks are denoted by @xmath492 .
the random variable @xmath493 is decoded at decoder @xmath140 if @xmath494 .
we take the families @xmath495 to be distinct .
the random variable @xmath496 is decoded at decoder @xmath140 if @xmath497 , where @xmath498 is an element of @xmath168 .
the following theorem describes the achievable rd region : for any @xmath499 , and @xmath500 distinct families @xmath501 $ ] , and for a joint distribution @xmath106 on random variables @xmath411 , @xmath492 and @xmath96 , where the underlying alphabet for the auxiliary random variables is the field @xmath17 , and a set of reconstruction functions @xmath108 , the set @xmath502 is defined as the set of rd vectors satisfying the following bounds for some non - negative real numbers @xmath457 and @xmath503 $ ] and @xmath504 : @xmath505_{\overline{{\mathbf{m}}}_\mathsf{n}}|[u , v , w]_{\widehat{\mathbf{m}}_\mathsf{n}\cup \overline{\mathbf{l}}})\leq \!\!\!\!\!\!\!\!\!\sum_{\mathcal{m}\in { { \mathbf{m}}}_\mathsf{n}\backslash\widetilde{\mathbf{m}}_\mathsf{n}\cup { \mathbf{l}}}\!\!\!\!\!\!\ ! ( \log q + \!\ ! \sum_{j\in \widetilde{\mathcal{m } } } \rho_{\mathcal{m},j}-r_{o,\mathcal{m}})+\!\!\!\!\!\!\!\!\!\!\!\!\sum_{\substack{\mathcal{m}\in{\mathbf{m}}_\mathsf{n}\backslash\widetilde{\mathbf{m}}_\mathsf{n}\cup \overline{\mathbf{l}}\\\bigcap\{\mathcal{a}_i|i\in[1,m+1]\ } } } \!\!\!\!\!\!\!\!\!\!(\log{q}+\sum_{j\in \widetilde{\mathcal{m}}}\rho_{o,\mathcal{m},j}-r'_{o,\mathcal{m } } ) , \quad \forall \overline{\mathbf{l}}\subset { \overline{\mathbf{m}}}_\mathsf{n } \label{sec2pack1 } \\&r_i=\sum_{\mathcal{m } } \rho_{\mathcal{m},i } , \qquad d_{\mathsf{n}}=e\big\{d_{\mathsf{n}}(h_{\mathsf{n}}(u_{\mathsf{n}},x))\big\}\label{sec2rd}. \ ] ] where ( a ) @xmath506 , ( b ) @xmath507 , ( c ) @xmath508 , ( d ) @xmath509 , ( e ) @xmath510 , ( f ) @xmath511 , ( g ) @xmath465 , and ( h ) @xmath512 . [ linend ]
the rd vector @xmath164 is achievable for the @xmath1descriptions problem , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath513 . toward proving the theorem
we need the following definition . a set of @xmath423 coset codes @xmath514 is called an ensemble of nested coset codes with parameter @xmath515 if the size of the intersection @xmath516 is equal to @xmath517 for all @xmath518 .
it is straightforward to show that one can always generate an ensemble of nested coset codes @xmath519 with parameter @xmath520 as long as @xmath521 .
it is enough to choose the rows of the generator matrices of @xmath522 such that they have @xmath523 common rows , similar to the case of figure [ codebookssum ] .
we provide an outline of the proof .
the codebook generation for codebooks @xmath524 is similar to the previous scheme . for random variables
@xmath525 we construct an ensemble of nested coset codes @xmath519 with parameter @xmath520 .
the encoder chooses a set of codewords from all the codebooks that is jointly typical with the source sequence .
the following is a generalized covering lemma which shows that if ( [ sec2cov1 ] ) and ( [ sec2cov2 ] ) is satisfied such a set of codewords exists .
[ p_covmut ] let @xmath17 be a field and define @xmath526 .
consider @xmath500 random variables @xmath96 , @xmath527 , where @xmath96 is defined on an arbitrary finite set @xmath11 and @xmath528 are defined on @xmath17 .
fix a pmf @xmath529 on @xmath530 . a sequence of m - tuples of codebooks @xmath531
is called @xmath532-covering if : @xmath533 [ thm : same_codemany ] for any @xmath529 on @xmath534 and rates @xmath535 satisfying ( [ sec2newe])-([sec2neweq ] ) , there exists a sequence of ensemble of nested coset codes @xmath536 with parameter @xmath537 which are @xmath538-covering .
@xmath539 where , ( a ) @xmath540 , ( b ) @xmath541 and ( c ) @xmath542 .
the proof of the lemma follows the same steps as in lemma [ thm : same_code ] .
we provide the intuition behind the proof .
given that there is a set of codewords in the codebooks @xmath543 which are jointly typical with the source sequence , for any linear combination @xmath544 there is a codeword which is jointly typical with the random variables @xmath545 . from a ptp perspective , the rate of codebook @xmath76 must satisfy ( [ sec2newe ] ) and ( [ sec2newe2 ] ) .
this rate can be calculated by counting the number of rows in the generator matrix of @xmath76 which is @xmath546 .
the packing bounds at each encoder can be written in the same way as in the previous section and are given in ( [ sec2pack1 ] ) .
@xmath547 is defined such that @xmath548 is the set of all possible linear combinations of @xmath493 s . in this section for completeness
, we provide a coding scheme where we reconstruct multi - variate summations of random variables at an arbitrary number of decoders , and these summations each have arbitrary lengths .
of course , due to the large number of random variables the coding scheme becomes extremely complicated .
let the number of the summations be @xmath549 , and for each summation , let the length of the summation be denoted by @xmath550 $ ] . define the sets @xmath551 $ ] and @xmath552 , i\in \mathsf{s}$ ] .
following the steps in the previous sections , we add @xmath553 new codebooks for each summation .
the underlying random variables for these codebooks are denoted by @xmath554 .
the random variable @xmath555 is decoded at decoder @xmath140 if @xmath556 .
fix the prime number @xmath557 .
the random variable @xmath558 is decoded at decoder @xmath140 if @xmath559 , where the summation is carried out in the finite field @xmath560 .
the following theorem describes the achievable rd region . for a joint distribution @xmath106 on random variables @xmath411 , @xmath554 and @xmath96 , where the underlying alphabet the auxiliary random variables is the field @xmath17 , and a set of reconstruction functions @xmath108 , the set @xmath561 is defined as the set of rd vectors satisfying the following bounds for some non - negative real numbers @xmath457 and @xmath562 , i\in \mathsf{s}$ ] and @xmath563 : @xmath564_{\overline{{\mathbf{m}}}_\mathsf{n}}|[u , v ,
w]_{\widehat{\mathbf{m}}_\mathsf{n}\cup \overline{\mathbf{l}}})\leq \!\!\!\!\!\!\!\!\!\sum_{\mathcal{m}\in { { \mathbf{m}}}_\mathsf{n}\backslash\widetilde{\mathbf{m}}_\mathsf{n}\cup { \mathbf{l}}}\!\!\!\!\!\!\ ! ( \log q + \!\ !
\sum_{j\in \widetilde{\mathcal{m } } } \rho_{\mathcal{m},j}-r_{o,\mathcal{m}})+\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\sum_{\substack{\mathcal{m}\in{\mathbf{m}}_\mathsf{n}\backslash\widetilde{\mathbf{m}}_\mathsf{n}\cup \overline{\mathbf{l}}\\\bigcap\{\mathcal{a}_{\{i\},s}|i\in[1,m_s+1 ] , s\in\mathsf{s}\ } } } \!\!\!\!\!\!\!\!\!\!\!\!\!\!(\log{q}+\sum_{j\in \widetilde{\mathcal{m}}}\rho_{o,\mathcal{m},j}-r'_{o,\mathcal{m } } ) , \quad \forall \overline{\mathbf{l}}\subset { \overline{\mathbf{m}}}_\mathsf{n } \label{sec3pack1 } \\&r_i=\sum_{\mathcal{m } } \rho_{\mathcal{m},i } \label{sec3rd } , \quad d_{\mathsf{n}}=e\big\{d_{\mathsf{n}}(h_{\mathsf{n}}(u_{\mathsf{n}},x))\big\}. \ ] ] where ( a ) @xmath565 , ( b ) @xmath566 , ( c ) @xmath567 , ( d ) @xmath568 , ( e ) @xmath569 , ( f ) @xmath570 , ( g ) @xmath465 and ( h ) @xmath571 .
[ linfinal ] the rd vector @xmath164 is achievable for the @xmath1descriptions problem , if there exists a distribution @xmath106 and reconstruction functions @xmath112 such that @xmath572 .
this is a straightforward generalization of the previous step , since the proof is similar , it is omitted .
similar to theorem [ theoremrandnonred ] one can identify the non - redundant codebooks in the above scheme .
one can show that a large number of possible codebooks become redundant in this case as well .
we provided several improvements over previous coding strategies for the md problem .
first , we showed that the cmsb strategy can be enhanced using additional unstructured quantizers and a new unstructured binning approach .
we demonstrated these gains using examples involving binary sources and three descriptions .
we provided the resulting rd region for the @xmath0-descriptions problem with arbitrary sources and distortion functions .
additionally , we proved that all of the new codebooks in our scheme are non - redundant for the @xmath0-descriptions problem . in the second part of the paper , we introduced structure in both the quantizer construction as well as in the binning functions .
we showed through several examples that the improvements derived in the first part can be extended if structured quantizers and binning functions are utilized .
the rd region in the first part of the paper was further improved upon by introducing additional linear coding layers .
lastly , we combined the ideas in the two parts to provide a new and strictly improved achievable rd region for the @xmath0-descriptions problem .
[ sec : conclusion ]
let @xmath573 and @xmath574 . without loss of generality , assume @xmath575 .
let @xmath576 be defined on @xmath577 . also let @xmath578 if @xmath579 . for @xmath580 $ ] , define a new distribution @xmath581 as follows : we provide an outline of the proof .
fix @xmath584 .
consider a new scheme where the reconstruction function at decoder @xmath585 is defined as @xmath586 with the rest of the reconstruction functions defined as in theorem [ thm : ssc ] .
let the rd vector @xmath164 be achievable in the new scheme using the distribution @xmath587 and reconstruction functions @xmath588 .
we provide a new probability distribution @xmath589and reconstruction functions @xmath590 to shows that the rd region given in theorem [ thm : ssc ] contains @xmath164 . to construct the probability distribution define @xmath591 , and @xmath592 . as for the reconstruction functions define @xmath593 and @xmath594 .
it is straightforward to check that with these parameters , the rd region in theorem [ thm : ssc ] contains @xmath164 . intuitively ,
since the reconstruction functions are the same , the same distortion is achieved by both schemes . as for the rates , in the first scheme , wherever @xmath595 is decoded , all of the random variables @xmath596 are also decoded .
so , adding a function of these random variables to @xmath595 does not require additional rate .
let @xmath597 , @xmath598 , @xmath599 , @xmath600 , where @xmath601 are the reconstructions at decoder @xmath602 in the two user problem in example [ ex : zb ] .
then it is straightforward to check that the rd vector is achievable from theorem [ thm : ssc ] .
next , assuming the codebook @xmath603 is empty , we consider all of the remaining 16 codebooks in the ssc scheme and show that the rd vector is not achievable . * step 1 : * in this step , we argue that the only non - trivial codebooks are @xmath114 , @xmath216 , @xmath604,@xmath605 and @xmath606 . due to the structure of the problem ,
a number of the codebooks are functionally equivalent , meaning they are decoded at exactly the same decoders .
so we can merge these codebooks without any loss .
for example , description @xmath182 is only received by decoders @xmath251 and @xmath261 , hence we can merge @xmath115 into @xmath217 without any loss .
@xmath607 , @xmath608 , @xmath219 and @xmath609 are only decoded at decoder @xmath261 so they are redundant from the results in @xcite .
@xmath610 can be merged into @xmath114 since decoder @xmath182 is not present .
@xmath611 is equivalent to @xmath604 and can be eliminated .
@xmath612 , @xmath613 and @xmath614 can be merged into @xmath217 .
finally @xmath615 can be merged with @xmath114 .
also @xmath217 can be merged with @xmath609 and is eliminated .
so we are left with four codebooks @xmath114 , @xmath216 , @xmath604,@xmath605 and @xmath606 .
* step 2 : * in this step , we show that if we set @xmath616 and @xmath617 , there would be no loss in terms of rd function . the codebooks @xmath114 and @xmath221 are decodable using description @xmath256 .
since decoder @xmath170 is at ptp optimality , these codebooks only carry @xmath618 .
to be more precise there is a markov chain @xmath619 , which we prove in the following lemma .
since @xmath618 is decoded both at decoder @xmath170 and @xmath183 , if we replace @xmath624 with @xmath625 , the decoders decode the same random variables as before , so no extra rate is required . also , from the lemma @xmath619 .
hence , we conclude that we can set @xmath616 and @xmath617 without any loss in terms of distortion . *
step 3 : * assume there are random variables @xmath624 and @xmath626 such that the rd vector is achievable in the ssc scheme .
from the markov chain @xmath627 , description 1 is not used in the reconstruction in decoders @xmath251 , @xmath183 and @xmath261 .
if we set @xmath617 , the distortions constraint in decoders @xmath251 , @xmath183 and @xmath261 are satisfied .
so we have constructed a scheme to send the descriptions at a lower rate ( by setting @xmath617 ) without any loss in terms of distortion in these three decoders .
this contradicts optimality of the random variables chosen for the two user scheme .
we have proved that if @xmath628 , the rd vector is not achievable but if the constraint is lifted the scheme can achieve this rd vector , so the codebook is non - redundant . for the general @xmath0-descriptions problem
, we provide an outline of the non - redundancy proof for @xmath629 .
let @xmath630 $ ] be the elements of @xmath631 .
then to construct an example where @xmath632 is non - redundant , first consider a set up where for any @xmath98 , each set of three decoders @xmath633 and @xmath634 and @xmath635 are as in the two user setup in example [ ex : zb ] . then there should be a common component between each two of the descriptions .
it is straightforward to show that the common components must be the same for all of the decoders , otherwise since the codebooks are independent there would be a rate - loss as explained in the previous section .
we ensure that the common component can be decoded only when all descriptions @xmath636 are received and not when a subset of the descriptions is received .
this is done by adding decoders @xmath637 , @xmath638 through @xmath639 such that each of them would be at ptp optimality by receiving a refined version of @xmath134 ( i.e @xmath637 would receive @xmath618 and @xmath638 would receive a refinement of @xmath618 and so on ) . in this way
the only codebook that can carry @xmath134 without rate - loss is @xmath640 .
let @xmath266 , description 1 carries @xmath618 to decoder @xmath170 with rate @xmath641 .
descriptions @xmath257 and @xmath7 send @xmath134 to decoders @xmath251 and @xmath260 by sending a refinement on @xmath223 . in other words @xmath597 , @xmath642 , @xmath599 and @xmath643 similar to the proof of lemma [ lemma1 ]
. then one can check that the rd vector is achievable using the ssc scheme .
next , assume @xmath644 , then @xmath645 . as in the previous section ,
we begin by eliminating the redundant codebooks for this communications setting . *
step 1 : * in this step we argue that only the codebooks @xmath608 , @xmath217 @xmath114 and @xmath223 are non - trivial .
due to the structure of this communications setting many of the codebooks are functionally the same and can be merged together .
the codebooks @xmath646 , @xmath604 , @xmath610 , @xmath647 are decoded at all four of the decoders and can be merged with @xmath114 .
@xmath648 can be merged with @xmath608 since decoder @xmath171 is not present , by the same argument @xmath612 is concatenated with @xmath613 , also @xmath614 and @xmath649 are merged with @xmath217 .
@xmath650 and @xmath606 are combined with @xmath614 .
lastly since decoders @xmath257 and @xmath7 are not present , @xmath115 and @xmath216 can be merged into @xmath217 and @xmath608 , respectively .
so only the four codebooks @xmath608 , @xmath217 @xmath114 and @xmath223 remain .
* step 3 : * by assumption , the codebook @xmath223 is only carried by the first description . however , the codebook is not decoded at decoder @xmath170 .
since the decoder is at ptp optimality , @xmath223 ca nt be sent through the first description either ( i.e @xmath651 and @xmath223 can be eliminated . ) . by the definition of @xmath618 we have @xmath627 and @xmath653 ,
so the bound above is strictly larger than the case when @xmath618 is replaced by @xmath134 ( i.e. when @xmath642 . ) .
this concludes the proof .
note that @xmath655 is the quantization noise of quantizing @xmath201 and @xmath656 is the quantization noise of quantizing @xmath657 .
since the source vectors are independent , the noise vectors are also independent and the summation converges to @xmath658 ( the arguments are similar to the ones given in @xcite . ) .
we assume that there exists a probability distribution @xmath106 on @xmath96 and @xmath659 for which the rd vector is achievable using the ssc scheme and arrive at a contradiction . since all of the decoders are present in this setup , we need to consider the ssc with all the codebooks present , so the proof is more involved than the proofs in the previous section .
* step 1 : * in this step we show that description @xmath98 , where @xmath660 , does not carry any bin indices for codewords from codebook @xmath139 if @xmath661 .
descriptions 1 and 2 only carry indices which are used in the reconstruction at decoders @xmath170 and @xmath182 , respectively .
this is true since these two decoders are receiving information at optimal ptp rate - distortion .
note that this does not mean the corresponding codebooks are empty , we can only conclude that no bin indices for the codewords are sent through these descriptions .
for example if @xmath662 and @xmath663 , then @xmath664 .
* step 2 : * in this step , we show that there are no common codebooks decoded at decoders @xmath170 and @xmath182 .
since decoder @xmath251 receives descriptions 1 and 2 at optimal rd from a ptp perspective , the random variables decoded at decoder @xmath170 must be independent of those decoded at decoder @xmath182 . from the next lemma we have that if @xmath675 then @xmath676 .
consider the setup in figure [ 2desc ] , let @xmath677 be such that @xmath678 , where @xmath679 is shannon s optimal ptp rd function for distortion function @xmath680 at point @xmath681 .
for any distribution @xmath682 which achieves this rd vector , the following conditions must hold : [ indeplem ] 1)@xmath683 and @xmath684 + 2)if in addition @xmath685 then , @xmath686 . where the right - hand side of the second equality is the sum - rate of the two - descriptions problem .
using the conditions @xmath693 , we have : @xmath694 so @xmath695 , which gives the desired markov chain in @xmath696 . assuming the original scheme achieves the rd vector in the theorem
, we give a new scheme which also achieves the rd vector .
we propose that the encoder operates as before , but decoder @xmath251 decodes @xmath697 only if @xmath698 or @xmath699 .
it needs to be shown that the rd vector is the same .
first we consider the resulting rates .
the covering bounds are not changed .
the packing bounds are the same at all decoders other than decoder @xmath251 since the same variables are being decoded at those decoders .
let @xmath701 and @xmath702 be subsets of @xmath703 and @xmath704 .
we need to show that the following packing bound is satisfied : @xmath705 let the random variables @xmath710 be such that @xmath711 .
then for an arbitrary distortion function @xmath712 , there is an optimal reconstruction of @xmath96 using @xmath137 and @xmath341 which is a only function of @xmath341 . by these arguments ,
codebook @xmath697 is eliminated if @xmath714 .
also in the new scheme , @xmath715 and @xmath716 are functionally similar since by the same arguments as in this step @xmath715 is not used in the reconstruction in decoder @xmath251 , so we can eliminate @xmath717 . in summary , thus far we have eliminated 7 codebooks .
+ * step 3 : * we have the following lemma : first we argue that @xmath720 .
if this is not true , it contradicts optimality at decoder @xmath260 .
@xmath721 is not decoded at decoder @xmath260 , but its bin index is carried through description 3 .
so if the bin index is non - zero , one could reduce @xmath722 by setting the bin index equal to 0 without increasing distortion at decoder @xmath260 , this contradicts optimality at that decoder .
by the same arguments @xmath723 .
now assume @xmath724 .
we show that this contradicts optimality at decoder @xmath260 .
@xmath725 is decodable using description @xmath256 ( since it is decodable at decoder @xmath170 ) . hence , if we set @xmath726 to @xmath727 ( i.e. do not send the bin index on description 3 ) , then decoder @xmath260 can still decode @xmath728 using description 1 .
so the distortion is the same at this decoder , but the rate @xmath722 is reduced which contradicts optimality . by the same arguments , @xmath729 . * step 4 : * we proceed by showing that @xmath730 .
so far we have shown that none of the descriptions carry the bin indices for these codebooks .
consider the following packing bounds in decoders @xmath170 , @xmath171 and @xmath260 : @xmath731 we add the above inequalities and subtract the mutual covering bound on all rvs , we get : @xmath732 this imposes the markov chain @xmath733 @xmath734 .
hence by the same arguments as in step 2 , we can eliminate @xmath735 . also by the same arguments
@xmath219 can be eliminated .
* step 5 : * in this step we eliminate @xmath736 and @xmath737 .
assume @xmath739 .
we claim this contradicts optimality at decoder @xmath260 , since @xmath624 can readily be decoded from the bin number carried by description @xmath7 , so setting @xmath740 to @xmath727 would decease rate without increasing distortion . the rest of the proof follows by the same argument . particularly @xmath743 , by the same arguments @xmath744 . +
* step 6 : * so far we have shown that only @xmath745 and @xmath746 can be non - trivial . from optimality at decoders @xmath170 and @xmath260 we have the following equalities : hence we have : @xmath748 define the following : @xmath749 where @xmath750 is the reconstruction of @xmath96 at decoder @xmath170 , @xmath751 is the reconstruction of @xmath269 at decoder @xmath183 , and @xmath752 is the reconstruction of @xmath267 at decoder @xmath260 .
then from : @xmath753 all the above inequalities must be equality .
in particular we have : @xmath754 note that from ( b ) , we can conclude that @xmath267 is independent of @xmath755 , we have @xmath755 and @xmath756 are independent .
define @xmath757 .
we have : @xmath758 ( a ) holds since @xmath755 and @xmath756 are independent . in ( b ) we have replaced @xmath759 by @xmath760 .
define @xmath761 , then by the same calculations @xmath762 , so @xmath763 .
note @xmath764 , hence using @xmath765 , we get @xmath766 with equality if and only if @xmath767 . also note that @xmath768 is available at decoder @xmath260
so @xmath769 and @xmath767 , otherwise there is a contradiction with optimality of @xmath752 . if @xmath767 , then @xmath770 is equal to @xmath771 .
so by the same arguments we have : @xmath772 where @xmath773 since @xmath774 , @xmath775 and @xmath776 , we have : @xmath777 we argue that @xmath778 and @xmath746 can be taken eliminated without any loss in rd . to prove this assume we have a scheme with @xmath779 .
construct new random variables @xmath780 , @xmath781 and @xmath782 and eliminate the rest of the codebooks . from the independence relations above , the packing bounds would stay the same .
since we have merged codebooks , the covering bounds would loosen , and it is straightforward to see that the reconstructions at each decoder are still the same .
we are left with four codebooks , @xmath783 and @xmath216 .
note that since decoder @xmath170 is only decoding @xmath114 we must have @xmath784 .
this is deduced from the packing bound in decoder @xmath170 : @xmath785 but @xmath786 so they are equal .
the same argument gives @xmath787 , and @xmath788 .
also , from optimality at the joint decoders and lemma [ indeplem ] , we have @xmath789 .
@xmath790 note that @xmath791 and @xmath792 from optimality at decoders @xmath170 and @xmath260 .
so @xmath793 .
replacing @xmath722 into , we get @xmath794 .
so we have the markov chain @xmath795 . by the same arguments we can derive the markov chain @xmath796 . using lemma [ markovindeplem ] and the previous two markov chains we get @xmath797 .
take the markov chain @xmath798 , along with @xmath799 we get @xmath800 . also from the optimality of the reconstruction of x at decoders
@xmath170 and @xmath260 , we have : @xmath801 from the above and @xmath799 , we conclude @xmath802 .
applying lemma [ markovindeplem ] we get @xmath803 .
let a , b , c and d be rvs such that @xmath804 and @xmath805 , and also assume there is no @xmath806 for which given @xmath807 there are non - constant functions @xmath808 and @xmath809 with @xmath810 with probability 1 .
then @xmath811.[markovindeplem ] this lemma is a generalization of the one in @xcite .
we need to show that @xmath812 for any @xmath813 .
note since functions @xmath814 and @xmath815 do not exist , it is straightforward to show that there is a finite sequence of pairs @xmath816 such that @xmath817 and @xmath818 with the property that either @xmath819 or @xmath820 and that @xmath821 . then from the first markov chain if @xmath820 , we have @xmath822 , also if @xmath823 the second markov chain gives this result .
so @xmath824 is constant on all of the sequence particularly @xmath812 .
let @xmath825 be the reconstruction of @xmath267 at decoder @xmath260 .
we have : @xmath826 so there is at least one @xmath827 such that @xmath828 .
let @xmath829 be the reconstruction of z using @xmath830 . by the same argument we can find a reconstruction of x using @xmath830 , then @xmath831 from a ptp perspective which is a contradiction .
we provide an outline of the proof here , the arguments are similar to the ones in the previous proofs .
+ * step 1 : * i any codebook which is not decoded at decoders @xmath170 , @xmath251 , @xmath171 , @xmath287 and @xmath286 is redundant .
this implies that there are at most only 17 codebooks which are non - redundant .
these codebooks are @xmath832 , @xmath833 , @xmath834 , @xmath835 , @xmath836 , @xmath837 , @xmath838 , @xmath839,@xmath840,@xmath841 , @xmath842 , @xmath843 , @xmath844 , @xmath845 and @xmath846 . +
* step 2 : * in this step we prove that the only non - trivial codebook decoded at decoder @xmath602 is @xmath847 for @xmath848 .
all possible codebooks decoded at decoder @xmath170 are @xmath832 , @xmath849 , @xmath850,@xmath851 , @xmath852 and @xmath853 .
from optimality at decoder @xmath251 , @xmath854 is redundant .
the reason is @xmath855 otherwise we can set it to zero without any loss in distortion at decoder @xmath251 which contradicts optimality , also any random variable that description @xmath183 carries must be used in reconstructing @xmath267 at decoder @xmath287 because that decoder is at optimality , which means @xmath856 so the codebook is decoded at decoder @xmath171 but not sent through either description @xmath182 or @xmath183 , from similar arguments as before the codebook is redundant .
same arguments can be provided to deduce redundancy of @xmath850 , @xmath857 , @xmath852 and @xmath853 .
this implies that only @xmath832 is decoded at decoder @xmath170 and @xmath858 at decoder @xmath286 . +
* step 3 : * we proceed with eliminating @xmath859 and @xmath860 . using the ptp optimality of decoder @xmath251 we have : @xmath861 where ( a ) follows from the usual ptp source coding results . comparing the lhs with the rhs
we conclude the markov chain @xmath862 .
in particular we are interested in @xmath863 . by the same arguments and using the optimality at decoder @xmath287 , we get @xmath864 .
these two markov chains along with lemma [ markovindeplem ] prove @xmath865 .
so these two variables are not used in reconstructing the source and the corresponding codebooks are eliminated . + * step 4 : * the only remaining codebooks are @xmath832 , @xmath858 , @xmath866 , @xmath867 , @xmath843 and @xmath868 . from optimality at decoders @xmath251 and @xmath287 we must have @xmath869 , also @xmath870 and @xmath871 . from lemma [ mclemma ] , we get @xmath872 .
we only need to show that @xmath877 , the rest of the implications of the long markov chain are either direct results of the three short markov chains or follow by symmetry . for arbitrary @xmath878
we have : @xmath879 we get an inner bound for @xmath880 at decoder @xmath171 : @xmath881 where the minimum is taken over all @xmath882 for which the long markov chain @xmath883 is satisfied and @xmath884 produce a lossless reconstruction of @xmath269 .
this resembles the distributed source coding problem in @xcite .
so the rd vector ca nt be achieved using random codes . in this proof
we use bold letters to denote vectors and matrices .
fix integers @xmath885 and @xmath886 .
choose the elements of the matrices @xmath887 , @xmath888 and @xmath889 and vectors @xmath890 and @xmath891 randomly and uniformly from @xmath17 . the codebooks @xmath892 and @xmath893 are defined as follows : @xmath894 for a typical sequence @xmath368 with respect to @xmath895 , we define @xmath896 as the function which counts the number of codewords in @xmath90 and @xmath897 jointly typical with respect to @xmath347 : @xmath898 for @xmath900 and @xmath901 , we denote the corresponding codeword as @xmath902 . similarly define @xmath903 for any @xmath904 and @xmath905 .
the following lemma proves several results on the pairwise independence of the codewords . 1 .
@xmath906 and @xmath907 are distributed uniformly uniform over @xmath908 .
2 . if @xmath909 , then @xmath906 is independent of @xmath910 .
3 . if @xmath911 , then @xmath907 is independent of @xmath912 .
if @xmath547 and @xmath913 are chosen independently and uniformly over @xmath908 , then @xmath914 and @xmath915 are independent .
we calculate the expected value of @xmath921 for any @xmath918 : @xmath922 also : @xmath923 using lemma [ lem : indep ] : @xmath924 at this point we have to consider several different cases for the values of @xmath925 .
cases when @xmath941 and @xmath942 are similarly considered but the derivations are omitted for brevity . considering cases
@xmath943 : @xmath944\end{aligned}\ ] ] consequently : @xmath945 where we have used lemma 8 in @xcite to get the fourth term .
after considering all the cases , the only non - redundant bounds are the ones mentioned in the lemma .
the proof follows the same arguments as that of lemma [ thm : same_code ] .
we provide an outline of the proof .
define the probability of error @xmath948 as follows : @xmath949 we define a new conditional probability of error for any triple @xmath950 : @xmath951 clearly if @xmath952 goes to 0 for all @xmath950 as @xmath953 , then @xmath948 goes to 0 .
also define : @xmath954 , and @xmath955 .
we have : @xmath956 @xmath957}\sum_{b_2\in [ 1,2^{n\rho_2}]}p(\mathbf{x } ) \\\nonumber & p\{g(\mathbf{a},\mathbf{m})=\mathbf{u } , g(\mathbf{\tilde{a}},\mathbf{\tilde{m}})=\mathbf{\tilde{u } } , g'(\mathbf{b},\mathbf{m}')=\mathbf{v } , g'(\mathbf{\tilde{b}},\mathbf{\tilde{m}}')=\mathbf{\tilde{v}}\}p\{b_1(\mathbf{u})=b_1({\mathbf{\tilde{u}}})=b_1 , b_2(\mathbf{u})=b_2({\mathbf{\tilde{u}}})=b_2\}\\ \end{aligned}\ ] ] note that the binning is done independently and uniformly , so @xmath958 .
the rest of the summations are the ones which were present in the proof of lemma [ thm : same_code ] .
again we have to do a case by case investigation of the summation .
the only new bond comes from the case when @xmath959 and @xmath960 , @xmath961 and @xmath962 .
we have : @xmath963 dividing this last term by @xmath964 : @xmath965 which goes to @xmath727 if the following is satisfied : @xmath966 however as shown in the next lemma the new bound in ( [ neweqpackred ] ) is redundant .
assume there is a distribution @xmath342 for which ( [ neweqpackred ] ) is violated , we show that either ( [ neweq ] ) or ( [ sumpack ] ) is also violated .
conversely , as long as ( [ neweq ] ) and ( [ sumpack ] ) are satisfied , ( [ neweqpackred ] ) is also satisfied .
assume we have : @xmath967 adding the two bounds we get : @xmath968 which contradicts ( [ neweq ] ) .
the proof follows the same arguments as in the previous two examples .
first we assume there exists a joint distribution @xmath969 such that the ssc scheme achieves the rd vector , then we arrive at a contradiction by eliminating all codebooks .
first note that from our definition of @xmath970 , direct calculation shows that @xmath971 .
this means that decoder @xmath251 is at ptp optimality . also by the definition of the distortion function @xmath972 ,
decoder @xmath183 is at optimal rd .
* step 2 : * from optimality of decoder @xmath183 , description 3 ca nt carry the bin number of any codebook which is not decoded at that decoder . also description 1 and 2
ca nt carry the bin numbers of codebooks which are not decoded at @xmath251 because of optimality at this decoder .
so codebooks @xmath975 , @xmath735 and @xmath857 are not sent on any description and are redundant . * step 3 : * the codebook @xmath849 is not binned by description 2 or 3 .
description 3 ca nt bin the codebook since it is not decoded at decoder @xmath183 , and that decoder is at ptp optimality .
note @xmath849 can be decoded using description 1 , so any bin information for this codebook that is carried by description 2 is not used at decoder @xmath251 , since decoder @xmath251 is at ptp optimality we must have @xmath855 .
the codebook is not sent on description 2 or 3 , so by the same arguments as in the previous proofs it ca nt help in the reconstruction at decoder @xmath171 and is redundant . by the same arguments @xmath976 is redundant .
* step 4 : * in this step we show that there is no refinement codebook decoded at decoder @xmath251 .
this would eliminate @xmath977 and @xmath978 .
more precisely we show that the reconstruction at decoder @xmath251 is a function of the reconstructions at decoders @xmath170 and @xmath182 .
this means that sending a refinement codebook to decoder @xmath251 will not help in the reconstruction , so the codebook is redundant . to prove this claim we consider the two user example depicted in figure @xcite . here
all distortions are hamming distortions .
we are interested in achieving the rate distortion vector @xmath111 given in ( [ rdscalar ] ) .
let @xmath979 be a distribution on the random variables in the two user ssc achieving this rd vector .
define @xmath980 and @xmath981 as the reconstructions at the corresponding codebooks . as in step 1 , from optimality of decoder @xmath251
, @xmath866 is redundant .
also @xmath985 and @xmath986 are independent from lemma [ indeplem ] .
note that @xmath983 is a function of @xmath985 and @xmath984 is a function of @xmath986 , so @xmath987 .
we proceed by characterizing @xmath988 .
note that decoder @xmath251 is at ptp optimality .
it is well - known result that when quantizing a bss to hamming distortion @xmath330 with rate @xmath989 , the reconstruction is uniquely given by @xmath990 where @xmath991 .
@xmath992 and @xmath981 are available at decoder @xmath251 , from optimality at this decoder we must have : @xmath993 so the inequality must be equality , which means @xmath994 .
in other words the markov chain @xmath995 must hold . using the three facts 1 ) @xmath996 ,
2 ) @xmath987 and 3 ) @xmath995 , we can characterize all possible distributions on @xmath997 .
let @xmath998 and @xmath999 . then from @xmath987 , @xmath1000 is fixed .
assume the distribution @xmath1001 is as given below : as shown on the table there are 5 independent linear constraints on @xmath1011 s .
we have : @xmath1012 , a_2\in[0,1 ] , p_{000}\in [ 0,(1-a_1)(1-a_2 ) ] , p_{001}\in [ 0,(1-a_1)a_2 ] , p_{010}\in [ 0,a_1(1-a_2)]\\ & p_{000}+p_{001}+p_{010}\in [ \frac{1}{2}-a_1a_2,\frac{1}{2}]\end{aligned}\ ] ] using the markov chain @xmath995 , we have @xmath1013 .
so @xmath1014 is as follows : we can minimize the resulting distortion at decoders 1 and 2 by choosing @xmath1023 and @xmath1004 optimally .
let @xmath1024 be the optimal joint distribution , we will show that there are two choices for @xmath1024 .
we have : @xmath1025 this is an optimization problem on @xmath1026 with respect to the constraints : @xmath1027 , a_2\in[0,1 ] , p_{000}\in [ 0,(1-a_1)(1-a_2 ) ] , p_{001}\in [ 0,(1-a_1)a_2 ] , p_{010}\in [ 0,a_1(1-a_2)]\\ & p_{000}+p_{001}+p_{010}\in [ \frac{1}{2}-a_1a_2,\frac{1}{2}].\end{aligned}\ ] ] also note that for fixed @xmath1028 and @xmath1029 the problem becomes a linear optimization problem ( otherwise the constraints are not linear ) .
so we fix @xmath1028 and @xmath1029 and optimize @xmath1023 and @xmath1004 for each value of @xmath1028 and @xmath1029 . in this case
the simplex algorithm provides a straightforward solution .
we investigate the solution in several different cases : * case 1 : * @xmath1030 : note that in the simplex algorithm , the variable with smallest ( most negative ) coefficient takes its maximum possible value first.since @xmath1031 , @xmath1032 , so the algorithm would first maximize the value of @xmath1002 . since @xmath1030
, we have @xmath1033 .
this along with constraint @xmath1034 $ ] sets @xmath1035 and @xmath1036 .
so in this case : @xmath1037 now we optimize on @xmath1038 such that @xmath1030 . increasing @xmath1028 or @xmath1029 decreases the distortion so the optimal value is achieved when @xmath1039 , so @xmath1040 .
we have : @xmath1041 optimizing the value of @xmath1028 , we get @xmath1042 .
these values give @xmath1043 . also replacing the values in @xmath1001
, we get : which shows that @xmath981 is a function of @xmath983 and @xmath984 .
* case 2 : * @xmath1046 , @xmath1047 : in this case the simplex method yields the following set of optimal distributions : @xmath1048 where @xmath1049 $ ] is an auxiliary variable that does not play a role in the distortion since the coefficients of @xmath1050 and @xmath1051 are equal in the distortion formula .
we get : note that since @xmath1053 , the term @xmath1054 is decreasing with @xmath1029 , so the distortion is increasing with @xmath1029 and the optimal values are @xmath1055 , since @xmath1047 , @xmath1056 , replacing @xmath1057 we have : @xmath1058 solving for @xmath1028 we get @xmath1059 and in tun @xmath1060 as in the previous case .
* case 3 * : @xmath1061 : the probabilities are as in the last case with @xmath1062 $ ] .
the distortion is similar to the last case . since @xmath1063
, the distortion is decreasing in @xmath1029 .
so @xmath1064 . which yields : @xmath1065 this would have no solution for optimizing @xmath1028 at the given range .
( [ mc50 ] ) holds from lemma [ indeplem ] . from the optimality at decoder @xmath251 and step 4
we have : @xmath1075 which proves ( [ mc51 ] ) .
next we prove ( [ mc52 ] ) : @xmath1076 in @xmath1077 we have used ( [ mc50 ] ) and the markov chain ( [ mc51 ] ) , in @xmath1078 , we have used ( [ mc50 ] ) . ( [ mc53 ] ) follows by symmetry .
( [ mcmiss ] ) can be proved using optimality at decoder @xmath183 and the argument given in the proof of ( [ mc50 ] ) .
we proceed with the proof of ( [ mc54 ] ) .
consider the following packing bounds at decoder @xmath251 and @xmath183 : @xmath1079 and the following covering bounds : @xmath1080 adding all the bounds and simplifying we get : @xmath1081 this resembles the two user sum - rate bound when the first user is sending descriptions 1 and 2 while the second user transmits description 3 .
from optimality at decoder @xmath1702 , @xmath1082 and optimality at decoder @xmath183 yields @xmath1083 .
so @xmath1084 .
this proves ( [ mc54 ] ) .
we have : where @xmath1077 follows from ( [ mc52 ] ) and lemma [ mcmove ] given below .
@xmath1078 follows from ( [ mc54 ] ) .
so we have shown that @xmath1086 , using lemma [ mcmove ] we conclude ( [ mc55 ] ) .
( [ mc56 ] ) follows by symmetry .
lastly we prove ( [ mc57 ] ) : @xmath1087 where @xmath1077 follows form [ mc54 ] .
@xmath1078 holds because of ( [ mcmiss ] ) .
( [ mc57 ] ) follows from lemma [ mcmove ] .
next we argue that if we set @xmath985 to be equal to @xmath1092 there would be no change in distortion and the rate does not increase .
first consider decoder @xmath260 .
the optimal reconstruction function is given by + @xmath1093 .
we have : @xmath1094 where in @xmath1077 we used the fact that @xmath1095 is a function of @xmath1096 and in @xmath1078 we use ( [ mc55 ] ) .
so the distortion wo nt change at decoder @xmath260 .
also the reconstruction at decoder @xmath170 is @xmath1092 so setting @xmath1097 wo nt change the reconstruction at this decoder . at decoder @xmath251
we showed in step 4 that @xmath1098 is a function of @xmath1099 where @xmath1100 is a function of @xmath1101 , so setting @xmath1097 does not change the distortion at this decoder either .
the rest of the decoders do not receive @xmath985 .
as for the rate , note that @xmath1092 was reconstructed at all decoders reconstructing @xmath985 .
so replacing @xmath985 with @xmath1092 does not require sending any extra information .
so we set @xmath1097 without any loss in distortion and with a potential gain in rate .
the same argument combined with the markov chains ( [ mc56 ] ) sets @xmath1102 , also using markov chains ( [ mc57 ] ) and ( [ mc58 ] ) we set @xmath1103 .
( [ pxx1x2 ] ) was proved in the step 4 .
( [ pxx3 ] ) follows from ptp optimality at decoder @xmath183 .
( [ x1mark ] ) follows from ( [ mc52 ] ) , ( [ x2mark ] ) follows from ( [ mc53 ] ) .
( [ x3mark ] ) follows from ( [ mcmiss ] ) .
( [ 123mark ] ) follows from ( [ mc54 ] ) .
we proceed by bounding the cardinality of @xmath1105 and @xmath1106 . using lemma [ imlemma ] ,
the joint distribution between the random variables is given as follows : @xmath1107 also note that we have the following equality : @xmath1108 denote @xmath1109 and @xmath1110 and @xmath1111 .
we have : @xmath1112 @xmath1113 using the values given in table ( [ tb : jointdist ] ) , we solve the system of equations : @xmath1114 hence the distribution in [ joint ] is completely determined by @xmath1115 and @xmath1116 .
based on the previous lemma it is enough to show that for every @xmath1115 and @xmath1117 , @xmath1120 , in that case we have a contradiction .
we need to maximize @xmath1121 as a function of @xmath1115 and @xmath1117 .
we use the following lemma : @xcite let @xmath11 be a finite set and @xmath12 be an arbitrary set .
let @xmath1122 be a set of pmfs on @xmath11 and @xmath1123 be a collection of pmfs on @xmath11 for every @xmath1124 .
let @xmath1125 $ ] be real - valued continuous functions on @xmath1122 .
then for every @xmath1126 defined on @xmath12 , there exists random variable @xmath1127 with cardinality @xmath1128 and a collection of conditional pmfs @xmath1129 on @xmath11 for every @xmath1130 such that for every @xmath1131 $ ] : @xmath1132 we want to use the lemma to bound cardinality of @xmath1133 .
take @xmath1134 and @xmath1135 .
note that fixing the expectation on @xmath1136 fixes the joint distribution in ( [ joint ] ) and fixing the expectation of @xmath1137 fixes the term we want to minimize .
so for any @xmath624 minimizing @xmath1121 , there exists @xmath1138 with cardinality at most 2 , such that the joint distribution and @xmath1121 are the same .
so it is enough to search over @xmath624 with cardinality 2 .
the same arguments hold for bounding the cardinality of @xmath1139 . for this size of random variables
, computer - assisted calculation shows that @xmath1140 as shown in figure [ figd ] .
so we have a contradiction and the ssc does not achieve the rd vector .
index the inequalities in the ssc from 1 to @xmath1141 . for every inequality in the linear coding region ( lcr )
, there exists a unique inequality in the ssc with the same left hand side , index this inequality with the same index used in the rcr .
let @xmath1142 be a bound resulting from applying fme on the ssc .
assume the bound results from adding inequalities indexed @xmath1143 , it is straightforward to show that adding inequalities with the same indices in the lcr gives the same bound .
the reason is that by our construction , the left - hand sides would be the same . in the right - hand side , due to the fme
, the terms involving @xmath1144 would be eliminated .
define @xmath1145 and @xmath1146 , eliminating @xmath1144 is equivalent to eliminating @xmath1147 or @xmath1148 .
videoaudio v. k. goyal , multiple description coding : compression meets the network , in signal processing magazine , ieee , vol.18 , no.5 , pp.74 - 93 , sep 2001 y. wang , a. r. reibman , and s. lin , multiple description coding for video delivery , in proceedings of the ieee , vol.93 , no.1 , pp.57 - 70 , jan .
2005 l. ozarow , on a source - coding problem with two channels and three receivers , bell systems tech .
journal , 59(10):1909 - 1921 , dec 1980 .
r. ahlswede , the rate - distortion region for multiple descriptions without excess rate , ieee trans . on inf .
theory , 31(6):721 - 726 , nov 1985 .
a. el gamal , and t. cover , achievable rates for multiple descriptions , ieee trans . in inf .
theory , vol .
it-28 , pp .
851 - 857 , 1982 .
z. zhang , and t. berger , new results in binary multiple - descriptions , ieee trans . on inf .
theory , vol.33 , no.4 , pp.502,521 , jul 1987 r. venkataramani , g. kramer , and v. k. goyal , multiple description coding with many channels , ieee trans . on inf .
theory , vol .
9 , pp . 2106 - 2114 , 2003 .
c. tian , and j. chen , new coding schemes for the symmetric k -description problem , ieee trans . on inf .
theory , vol.56 , no.10 , pp.5344,5365 , oct .
2010 s. s. pradhan , r. puri , and k. ramchandran , n - channel symmetric multiple descriptions - part i:(n , k ) source - channel erasure codes , ieee trans . on inf .
theory , vol .
47 - 61 , 2004 .
e. akyol , k. viswanatha , and k. rose , combinatorial message sharing and random binning for multiple description coding , in ieee international symp . on inf .
ieee , pp.1371,1375 , 1 - 6 july 2012 .
j. krner , and k. marton , how to encode the modulo - two sum of binary sources , ieee trans . on inf .
theory , vol .
219 - 221 , 1979 .
a. padakandla , and s.s .
pradhan , achievable rate region for three user discrete broadcast channel based on coset codes , in ieee international symp . on inf .
ieee , pp.1277,1281 , 7 - 12 july 2013 a. padakandla , a.g .
sahebi , and s.s .
pradhan , a new achievable rate region for the 3-user discrete memoryless interference channel , in ieee international symp . on inf .
theory , july , 2012 ( also to appear in ieee trans . on information theory 2016 ) .
b. nazer , and m. gastpar , computation over gaussian multiple - access channels , in ieee international symp . on inf .
ieee , pp.2391,2395 , 24 - 29 june 2007 t. philosof , and r. zamir , on the loss of single - letter characterization : the dirty multiple access channel , ieee trans . on inf .
theory , vol.55 , no.6 , pp.2442,2454 , june 2009 d. krithivasan , and s.s .
pradhan , distributed source coding using abelian group codes : a new achievable rate - distortion region , ieee trans . on inf . theory , vol.57 , no.3 , pp.1495,1519 , march 2011 a. el gamal , and y. h. kim , network information theory , cambridge university press , 2011 a.b .
wagner , b.g .
kelly and y. altu , distributed rate - distortion with common components , ieee trans . on inf . theory , vol.57 , no.7 , pp.4035 - 4057 , july 2011 d. kleitman , g , markowsky , on dedekind
s problem : the number of isotone boolean functions .
ii , transactions of the american mathematical society 213 : 373 - 390 , 1975 j. wang , j. chen , l. zhao , p. cuff , and h. permuter , on the role of the refinement layer in multiple description coding and scalable coding , ieee trans . on inf .
theory , vol.57 , no.3 , pp.1443 - 1456 , march 2011 s. s. pradhan , j. chou , k. ramchandran , duality between source coding and channel coding and its extension to the side information case , ieee trans . on inf .
theory , vol.49 , no.5 , pp.1181 - 1203 , may 2003 r. g. gallager , information theory and reliable communication .
, new york : wiley , 1968 .
f. s. chaharsooghi , a. g. sahebi , s. s. pradhan , distributed source coding in absence of common components , in ieee international symp . on inf . theory .
ieee , pp.1362 - 1366 , 7 - 12 july 2013 k. vinodh , v. lalitha , n. prakash , p. v. kumar , s. s. pradhan , on the achievable rates of sources having a group alphabet in a distributed source coding setting , in communication , control , and computing ( allerton ) , 2010 48th annual allerton conference on , pp.479 - 486 , sept .
. 1 2010 p. gacs and j. krner , common information is far less than mutual information , problems of control and information theory , vol .
119 - 162 , 1972 .
h. s. witsenhausen , on sequences of pairs of dependent random variables , siam journal of applied mathematics , vol .
100 - 113 , january 1975 .
|
we consider the problem of multiple descriptions ( md ) source coding and propose new coding strategies involving both unstructured and structured coding layers .
previously , the most general achievable rate - distortion ( rd ) region for the @xmath0-descriptions problem was the combinatorial message sharing with binning ( cmsb ) region .
the cmsb scheme utilizes unstructured quantizers and unstructured binning . in the first part of the paper , we show that this strategy can be improved upon using more general unstructured quantizers and a more general unstructured binning method . in the second part , structured coding strategies are considered .
first , structured coding strategies are developed by considering specific md examples involving three or more descriptions .
we show that application of structured quantizers results in strict rd improvements when there are more than two descriptions .
furthermore , we show that structured binning also yields improvements . these improvements are in addition to the ones derived in the first part of the paper .
this suggests that structured coding is essential when coding over more than two descriptions . using the ideas developed through these examples we provide a new unified coding strategy by considering several structured coding layers .
finally , we characterize its performance in the form of an inner bound to the optimal rate - distortion region using computable single - letter information quantities .
the new rd region strictly contains all of the previous known achievable regions .
|
in the presence of extraordinarily strong gauge fields , we encounter essentially new phenomena that are not observed in the vacuum .
such phenomena are collectively called strong - field physics , " which has been attracting the attention of many researchers in various fields in physics @xcite .
there are intense laser facilities planned around the world ; in addition , compact stars and relativistic nucleus - nucleus collisions offer unique opportunities to study the strong - field phenomena . even for a system with a small coupling constant ,
the physics becomes nonperturbative because the smallness of the coupling constant is compensated by the strong background field strength , which requires some sort of resummations of higher - order contributions .
such nonperturbative nature of interactions between particles and fields is one of the outstanding properties of strong - field physics . at the same time
, it is an important issue to understand the transition from the perturbative to the nonperturbative regime with increasing the strength of the fields and/or with changing other parameters .
this is relevant also to experiments for realizing the setups for strong - field physics .
the present paper is devoted to clarifying the interplay between perturbative and nonperturbative aspects of the phenomena under strong fields . to be more specific ,
let us consider a system described by qed in a very strong electric field @xmath14 where @xmath15 is the electron mass @xcite .
in such a system , propagation of an electron significantly differs from that of the bare one because it receives large corrections from the strong field . from a naive dimensional argument
, the propagation of an electron acquires corrections of the order of @xmath16^n)$ ] if it receives @xmath17 kicks from the strong field .
therefore , for @xmath14 , the higher - order interactions are not suppressed and the propagation of an electron bears nonperturbative nature .
typical examples of the nonperturbative phenomena include the schwinger mechanism @xcite , i.e. , spontaneous production of @xmath18 pairs from the vacuum ( see ref .
@xcite for a review ) .
intuitively it is understood as the process where a virtual @xmath18 pair produced as a vacuum fluctuation is kicked many times by the strong field so that they obtain enough energy to become a real pair .
formally , an imaginary part appears in the effective action of the background fields ( the euler - heisenberg action @xcite ) only after summing up the electron s one - loop diagrams with infinitely many insertions of the external fields .
the total number of created electrons in case of a constant electric background field with infinite duration is given by @xmath19 whose dependence on @xmath20 clearly indicates nonperturbative nature of the schwinger mechanism @xcite .
another aspect of strong fields which becomes important in actual physical situations is the temporal dependence of the fields . as far as a strong electric field
is concerned , it is not realistic to keep it for a long time compared to the typical time scale of the system .
for example , electric fields produced in high - energy heavy - ion collisions decay quite fast .
those in intense lasers are also time dependent . based on the intuitive picture that charged particles receive quantum corrections by kicks from the strong fields
, one may expect that the number of kicks from the fields depends on the field lifetime .
indeed , perturbative calculation may be justified for a short - lived pulse field like a shock wave or a spike .
thus , we need to study carefully whether the process should be described in a perturbative or nonperturbative way for the pair creation in the presence of time - dependent strong fields .
this is the problem we are going to address in the present paper . before going into the details ,
let us briefly explain here how the finite duration of the fields modifies our intuitive picture for the pair creation in the background field . in a static and homogeneous electric field ,
the system has only two dimensionful parameters , @xmath20 and @xmath15 . as mentioned above ,
a criterion for pair creation from the vacuum is then given by @xmath21 .
this is also understood in the following way .
pair creation will become possible if the work , @xmath22 , done by the field on a virtual electron / positron for a distance @xmath23 is at least of the same order as the electron mass . since this must happen within the lifetime of pair fluctuation , the distance @xmath23 is identified with the compton length @xmath24 .
however , when the external electric field is time dependent , we have another dimensionful parameter @xmath2 , namely , a typical duration of the field . in the static field case ,
we care only the lifetime of the fluctuation , but now we need to deal with the two time scales : @xmath2 and @xmath25 .
we will find that two dimensionless parameters @xmath26 and @xmath27 are relevant for this time dependent case , and can discuss the interplay between perturbative and nonperturbative physics with the values of @xmath28 and @xmath29 .
in particular , we can study the interplay in detail by using the sauter - type electric field @xmath30 which allows analytic calculation for the schwinger mechanism .
the present paper is organized as follows : in the next section , we provide perturbative and nonperturbative formulations for the @xmath7 pair creation in a time - dependent electric field without specifying any temporal profile . in section 3 , we take the sauter - type field as an example of pulsed electric fields and compute the number densities of produced electrons in both formulations .
then , we compare the two results .
a summary is given in the last section .
the purpose of this section is to present a general expression for the number of electrons created from the vacuum in the presence of a time - dependent electric field . to this aim , we consider the following qed lagrangian : @xmath31\psi - e\bar{\psi}{{\ooalign{\hfil/\hfil\crcr$\bar{a}$ } } } \psi \equiv { \mathcal l}_0 + { \mathcal l}_{\rm bg } \ , .
$ ] here @xmath32 is the coupling constant , @xmath33 is the electron field , and @xmath34 is the background gauge field .
we assume a background electric field directed to the @xmath35 axis , which is homogeneous in space but depends on time , and we set the background gauge field four - potential @xmath34 to @xmath36 below we first derive a formula for the number of produced electrons in the lowest - order perturbation theory , and then briefly describe how to obtain the same quantity from the nonperturbative expression for the schwinger mechanism that includes all order interactions with the background field .
we treat the interaction of the electron with the background @xmath37 as perturbation and compute an @xmath38-matrix element for the @xmath7 pair creation from the vacuum , @xmath39 \ket{\rm vac;in}$ ] , in the lowest order perturbation theory .
the diagrammatic expression for the lowest - order contribution @xmath40 reads @xmath41{fig1.eps}}\ , .
\end{aligned}\ ] ] a straightforward calculation yields @xmath42 \big ( \bar{u}(\mbox{\boldmath $ p$ } , s ) \gamma^3 v(-\mbox{\boldmath $ p$ } , s ' )
\big ) \delta^3({\mbox{\boldmath $ p$}}+{\mbox{\boldmath $ p$ } } ' ) \nonumber \\ & = \frac { e \tilde{e}(2p_0)}{2p_0 } \ ; \big ( \bar{u}(\mbox{\boldmath $ p$ } , s ) \gamma^3 v ( -\mbox{\boldmath $ p$ } , s ' ) \big ) \delta^3(\mbox{\boldmath $ p$}+\mbox{\boldmath $ p$ } ' ) \ , .
\label{eq4}\end{aligned}\ ] ] note that we have expanded the unperturbed operator @xmath33 as @xmath43 \end{aligned}\ ] ] to get the second expression .
here @xmath44 , the annihilation operators @xmath45 for an electron and @xmath46 for a positron satisfy the following anticommutation relations : @xmath47 and the dirac spinors @xmath48 are normalized as @xmath49 in the last line of eq . ( [ eq4 ] )
we have introduced the fourier transform of the background electric field @xmath50 .
now , we can compute the number density of electrons @xmath51 created from the vacuum in the lowest - order perturbation theory , @xmath52 where integration over the positron momentum @xmath53 gives a volume factor @xmath54 .
the physical meaning of the formula ( [ eq15 ] ) is evident .
for an electric field oscillating in time @xmath55 , @xmath56 is proportional to @xmath57 . for the on - shell electron energy @xmath58 ,
the number of produced electrons vanishes if @xmath59 , which means that the pair creation does not occur when the energy supplied by a single photon is below this threshold .
this is certainly true for a constant electric field @xmath60 , no matter how strong the background electric field is ( within the perturbation theory ) . for a general time - dependent background field
the number of produced electrons is nonvanishing even for a single photon as long as the background electric field has a nonzero fourier spectrum @xmath61 above the threshold @xmath62 .
the total number of produced electrons , @xmath63 , is obtained after integration over : @xmath64 the @xmath65 integral is not possible in general unless we specify the background electric field @xmath66 .
creation of @xmath7 pairs from the vacuum is possible in the presence of an electric field as a nonperturbative process , which is characterized by the critical field strength @xmath67 .
schwinger @xcite computed formulas for the number of created pairs and for the vacuum persistent probability in case of the electric fields which is homogeneous in space and constant in time . it is also known that we can equally formulate the case with temporal dependence . here
, we briefly explain such a case .
notice that this is a nonperturbative calculation because we include the interaction of electrons or positrons with the electric fields up to infinite order .
we employ a formalism based on the canonical quantization in the presence of external fields .
namely , we expand the field operator @xmath33 by the exact mode functions of an electron @xmath68 ( as = in / out ) and a positron @xmath69 ( as = in / out ) under the given background field @xmath70 as @xmath71\ , , \\ & \left [ i{{\ooalign{\hfil/\hfil\crcr$\partial$ } } } - e\bar{{{\ooalign{\hfil/\hfil\crcr$a$ } } } } - m\right ] { } ^{}_{\pm}\ ! \psi^{\rm as}_{\mbox{\boldmath { \small $ p$}}s}(x)=0\quad { \rm ( as = in / out ) } \ ; . \label{eq23}\end{aligned}\ ] ] noting that any linear combinations of @xmath72 satisfy the equation of motion ( [ eq23 ] ) , we identify the electron ( positron ) mode as the positive ( negative ) frequency mode in the asymptotic in- and out - region , respectively .
the important point is that the mode function of the in - state @xmath73 and that of the out - state @xmath74 do not coincide with each other in the presence of the background field and the electron ( positron ) mode function of the in - state @xmath75 ( @xmath76 ) becomes a linear combination of the mode functions of the out - state : @xmath77 the bogoliubov coefficients , @xmath78 and @xmath79 , satisfy the relation @xmath80 .
one may simply understand this fact ( [ 2bog ] ) in analogy with an one - dimensional barrier scattering problem where we always have a mixture of in - coming wave and its reflection on one side of the barrier , but the other side consists of out - going wave only .
this difference between in- and out - state mode functions results in the difference between the in- and out - state annihilation operators , @xmath81 and @xmath82 . from the orthonormality of the mode functions ,
one can obtain @xmath83 using eqs .
( [ eq24 ] ) and ( [ eq25 ] ) , we can construct a nonperturbative formula to compute the number of electrons created via the schwinger mechanism : @xmath84 thus , the problem now is reduced to computing the coefficient @xmath85 or solving the dirac equation ( [ eq23 ] ) .
notice that we have not specified the time dependence of the background field @xmath70 so far and thus eq . ( [ eq26 ] ) is a general formula .
however , since there are only a few cases where analytic solutions for the dirac equation ( [ eq23 ] ) are available , we are usually forced to use numerical methods to evaluate eq .
( [ eq26 ] ) .
when the background electric field is constant and homogeneous , one can easily solve the dirac equation ( [ eq23 ] ) @xcite .
the total number of produced electrons per unit volume and time is given by @xmath86 , which clearly shows the nonperturbative nature of the formula ( [ eq26 ] ) .
in this section , we consider a special case where the background electric field is applied as a pulse in time .
in particular , we work with a sauter - type pulse field with height @xmath1 and width @xmath2 : @xmath87 as mentioned in the introduction , the advantage of the sauter - type background field is that the analytic solution for the dirac equation is known so that we can explicitly compute the particle number created via the schwinger mechanism @xcite . thus , by comparing this nonperturbative result with that of our perturbative computation obtained in the sauter - type background field , we can discuss which picture , perturbative or nonperturbative , is appropriate for studying the pair production in a strong field with finite duration .
we substitute the sauter - type background field ( [ eq21 ] ) into eqs .
( [ eq15 ] ) and ( [ eq20 ] ) to get the electron number density @xmath88 and the total electron number @xmath63 , respectively , in the lowest - order perturbation theory . by using @xmath89
we find that the electron number density @xmath88 is given by @xmath90\right|^2 } \ , .
\label{eq233}\end{aligned}\ ] ] we obtain a nonvanishing result because the fourier spectrum of the sauter - type field @xmath91 is nonzero at any value of @xmath65 , in particular in the region @xmath62 .
also , the total electron number @xmath63 is given by @xmath92 here we have introduced the function @xmath93 by @xmath94 which behaves asymptotically as ( see the appendix ) @xmath95 figure [ fig1 ] shows the momentum ( @xmath96 ) dependence of the electron number density @xmath97 ( [ eq233 ] ) for @xmath98 and @xmath99 .
we see that the peak is located at @xmath100 , which reflects the fact that the energy threshold for creating one @xmath18 pair @xmath101 takes its minimum @xmath102 at @xmath100 .
we also find that the distribution decays exponentially for large @xmath103 .
actually , one can check this by taking the limit of @xmath104 : @xmath105 figure [ fig2 ] shows the @xmath2 dependence of the total electron number @xmath106 . as is seen in fig .
[ fig2 ] , @xmath63 increases monotonically for small @xmath107 while it decreases exponentially for large @xmath108 .
this tendency can be roughly understood as follows : since the threshold energy of the pair creation is @xmath109 , the background field must supply energy @xmath110 larger than @xmath111 for the pair creation to occur . in our computation
based on the lowest - order perturbation theory , such energy @xmath110 is supplied by a single ( virtual ) photon from the background electric field @xmath112 . since the typical energy @xmath65 of a photon which forms the sauter - type background field @xmath112 is @xmath113 ( see eq .
( [ spect ] ) and the left panel of fig . [ spectrum ] ) , we find @xmath114 ( number of photons ) @xmath115 ( typical photon energy ) @xmath116 .
thus , @xmath117 i.e. , @xmath118 is required for the pair creation in the lowest order perturbation theory .
the upper limit @xmath119 is shown as a vertical line in fig .
pair creation from a single photon occurs when the pulse duration @xmath2 is short enough . on the other hand , as shown in the right panel of fig .
[ spectrum ] , the strength of the fourier component @xmath61 decreases with decreasing @xmath2 .
this essentially explains the decrease of electron number density for @xmath120 as seen in fig .
[ fig2 ] .
one can obtain analytic solutions to the dirac equation in the presence of the sauter - type background field @xcite ( see also ref .
@xcite for discussion in @xmath121 case ) , which enables us to compute the number of produced electrons ( [ eq26 ] ) .
after some calculations , one finds @xmath122 \sinh\left [ \frac{\pi\tau}{2 } \left ( 2ee_0 \tau -p_{0}^{(- ) } + p_{0}^{(+ ) } \right ) \right ] } { \sinh\left [ \pi \tau p_{0}^{(- ) } \right ] \sinh\left [ \pi \tau p_{0}^{(+ ) } \right ] } \ , , \label{eqa16}\end{aligned}\ ] ] where @xmath123 are the energy of the electron ( positron ) with the transverse momentum @xmath124 and the _ canonical _ longitudinal momentum @xmath125 .
note that the corresponding electron mode originally has the _ kinetic _ longitudinal momentum @xmath126 in the infinite past and @xmath127 in the infinite future .
we stress that this result is clearly nonperturbative with respect to the coupling constant @xmath128 , while the lowest order perturbation gave the result proportional to @xmath129 [ see eq .
( [ eq233 ] ) ] .
the total number of produced electrons , @xmath106 , is obtained after integration over the momentum @xmath96 .
we compare the nonperturbative result ( [ eqa16 ] ) with the perturbative one ( [ eq233 ] ) and ( [ eq244 ] ) in figs . [ fig3 ] , [ fig3_taudep ] , and [ fig4 ] .
figure [ fig3 ] shows the comparison of the momentum ( @xmath125 , @xmath103 ) dependence of the number density of electrons @xmath130 .
the peak strength of the field is taken as @xmath131 .
three lines are for different values of @xmath2 .
we have shown only relatively short pulse cases : @xmath132 and 0.6 .
we immediately observe that the nonperturbative result ( [ eqa16 ] ) and the perturbative result ( [ eq233 ] ) coincide with each other for the _ short _ pulse . the deviation becomes larger as @xmath2 increases , which can be explicitly seen in fig .
[ fig3_taudep ] .
there , @xmath2 dependence is shown for different values of the peak strength : the left panel is for the subcritical and @xmath133 , respectively , but precisely speaking , the condition @xmath131 ( valid for a constant electric field ) does not play the same role for finite pulses . ]
field strength @xmath134 and 0.8 and the right panel is for the supercritical field strength @xmath135 and 20 .
we again observe the agreement of the two results for _ short _ pulses @xmath136 no matter how large the field strength @xmath137 is .
however , the size of the agreement region in @xmath2 heavily depends on the field strength @xmath138 . for subcritical field strength @xmath139
, perturbative result dominates the nonperturbative result even when the pulse is not _ very short _ @xmath140 .
on the other hand , for supercritical field strength @xmath141 , perturbative description is applicable only for _ very short _ pulse region @xmath142 .
we will clarify the reason for this behavior in the later discussion .
the important point here is that for any field strength @xmath137 there surely exists a region ( _ short _ pulse region ) where pair creation can be understood as a purely perturbative phenomenon .
we can also observe that there is a clear deviation between the two in the _ long _ pulse region where the nonperturbative result approaches schwinger s result ( horizontal lines ) .
in particular , the deviation is larger for supercritical field @xmath143 .
this can be understood as follows .
notice first that the perturbative result always approaches 0 in the long pulse limit @xmath144 because the typical energies of a ( virtual ) photon which forms the sauter - type field @xmath145 for large @xmath2 and thus not enough to create a pair . on the other hand
, schwinger s formula valid in the long pulse region says that pair creation for subcritical field strength is exponentially suppressed @xmath146 .
therefore , the deviation between the two is almost negligible for weak field strength @xmath139 , while it increases with increasing peak strength @xmath138 in the supercritical regime @xmath141 .
the same tendency is found in the comparison of the total number of electrons @xmath63 as shown in fig .
we note that the peak structure in the _ short _ pulse region is reproduced by the perturbative result quite well .
although for small @xmath125 the perturbative value of the density @xmath130 is somewhat larger than the nonperturbative one , while it becomes smaller for large @xmath125 ( see the left panel of fig .
[ fig3 ] ) , these differences cancel out with each other in integration over @xmath125 .
thus we have a nice agreement in the total number @xmath63 as displayed in fig .
[ fig4 ] .
figure [ fig3_taudep ] also shows an interesting behavior .
for relatively short pulses @xmath147 with subcritical field strength @xmath148 ( left panel ) , the results of the sauter - type field are enhanced as compared to schwinger s value ( horizontal lines ) .
since the pair creation in this region is dominated by the perturbative contribution , this enhancement should be understood as a purely perturbative effect .
it shows up because schwinger s nonperturbative result @xmath146 is exponentially small for subcritical field strength @xmath139 , while the perturbative result is only power - suppressed as @xmath149 ( see eq .
( [ eq15 ] ) ) . by using the perturbative formula for @xmath130 ( [ eq233 ] )
, we immediately find that the peak position @xmath150 is given by @xmath151 $ ] or @xmath152 , which does not depend on the field strength @xmath153 as is seen in fig .
[ fig3_taudep ] .
accordingly , the peak value is given by @xmath154 . the peak value , @xmath155 $ ] , at @xmath156 is displayed in fig . [ fig7 ] as a function of the field strength @xmath138 , together with schwinger s value and the peak value of the perturbative contribution . the extrapolation of schwinger s value to the weak field case @xmath139 underestimates the pair creation in the sauter - type pulsed field ; the pair creation from the vacuum in the region @xmath157 is actually more abundant than schwinger s value , owing to the perturbative contribution with a single photon .
indeed , the compact formula for the perturbative peak nicely describes the enhancement for the subcritical fields , which is explicitly depicted with a dashed line in fig .
similar behavior was found in refs .
@xcite , who however regarded this peak as a result of nonperturbative physics .
now we return to the question : to what extent are we able to say a pulse is _ short _ ?
to answer this , we expand the nonperturbative result ( [ eqa16 ] ) by the pulse duration @xmath2 .
more precisely , we expand ( @xmath158 ) by the following two dimensionless parameters , @xmath159 because there are two dimensionful quantities @xmath160 in addition to @xmath2 . the result is @xmath161\ ! \sinh\!\!\left[\pi
\nu \!\left ( 1\!-\!\frac{1}{2\gamma } \!\sqrt{\gamma^2 \!-\ ! 2 \frac{p_z}{p_0 }
\gamma \!+\ ! 1 } \!+\ !
\frac{1}{2\gamma } \!\sqrt{\gamma^2 \!+\ !
2 \frac{p_z}{p_0 } \gamma \!+\ ! 1 } \right)\ ! \right]\ ! } { \sinh\left[\pi\nu \sqrt{1 - 2\frac{p_z}{p_0 } \gamma^{-1 } + \gamma^{-2 } } \right ] \sinh\left[\pi\nu \sqrt{1 + 2\frac{p_z}{p_0 } \gamma^{-1 } + \gamma^{-2 } } \right ] } \nonumber\\ & \longrightarrow \left\ { \begin{array}{ll } \displaystyle \frac{1}{(2\pi)^3}\left ( 1-\frac{p_z^2}{p_0 ^ 2 } \right ) \left| \frac{e e_0}{p^2_0}\right|^2 \frac{(\pi p_0 \tau)^4}{\pi^2 \left|{\rm sinh}[\pi p_0 \tau]\right|^2 } & \quad ( \nu , \gamma \ll 1 ) \\ \displaystyle { \rm exp } \left [ -\frac{\pi\left ( m^2+\mbox{\boldmath $ p$}^2_{\perp } \right)}{|ee_0| } \right ] & \quad ( \nu , \gamma \gg 1)\ , .
\end{array } \right .
\label{eq28}\end{aligned}\ ] ] notice that the asymptotic forms ( [ eq28 ] ) exactly reproduce the perturbative result ( [ eq233 ] ) for @xmath162 and the nonperturbative expression for the schwinger mechanism in a constant electric background field @xcite for @xmath163 .
thus , we conclude that pulses , such that the condition @xmath162 i.e. , @xmath164 is satisfied , are so _ short _ that pair creation becomes purely perturbative , where the _ lowest _ order perturbation theory works very nicely . on the other hand , pulses ,
such that the condition @xmath165 i.e. , @xmath166 is satisfied , are so _ long _ that pair creation becomes nonperturbative , where perturbation theory completely breaks down .
we can also say that for _ middle _ pulses , such that neither condition @xmath165 nor @xmath162 is satisfied , perturbation theory is still applicable ; however , the lowest - order perturbation theory does not work because higher - order corrections @xmath167 ( @xmath168 ) become important .
we summarize our picture in fig .
these considerations clearly show that in order to investigate the nonperturbative nature of the schwinger mechanism we must require not only the strength @xmath141 but also a sufficient duration @xmath169 ; otherwise pair creation from the vacuum can be understood simply as a perturbative phenomenon . and width @xmath2 .
@xmath170 is the electron mass .
, width=604 ] the discussion given above is a natural result if we consider the meaning of the dimensionless parameters @xmath171 .
recall the fact that the work @xmath172 done by a pulsed electric background field with height @xmath1 and width @xmath2 is given by @xmath173 and that the typical energy @xmath65 of a photon that forms the pulsed background field is given by @xmath113 .
then , we can understand the physical meaning of @xmath171 as follows : @xmath174 is the number of ( virtual ) photons of the background field involved in a scattering process .
@xmath175 is the work done by the background field scaled by the typical energy scale of the system @xmath170 .
keeping these in mind , we can interpret that the perturbative condition @xmath176 corresponds to the case where both the number of photons involved in a scattering process @xmath29 and its correction to the system @xmath28 are very small .
this is obviously a natural criterion for the lowest order perturbation theory to work .
we can also interpret the nonperturbative condition @xmath163 in the same way .
it is interesting to compare our discussion with ref .
@xcite , which claims that the keldysh parameter @xmath177 , where @xmath65 is the typical frequency of the background field , discriminates whether the system is perturbative or nonperturbative .
note that their discussion is limited to the case where ( i ) an oscillating electric field background @xmath178 , ( ii ) @xmath65 is sufficiently small compared to the electron mass @xmath179 , and ( iii ) the background field is sufficiently weak @xmath180 .
if we assume that the typical frequency @xmath65 of a pulsed background field is given by the inverse of the pulse duration @xmath113 , we find that our discussion obtained in a pulsed background field ( see fig .
[ fig5 ] ) agrees with ref . @xcite
as long as the limitation ( iii ) is satisfied . in such a condition
, @xmath28 determines the perturbativeness " of the system in our discussion and is equivalent to the keldysh parameter because @xmath181 .
we have explicitly demonstrated that , by using the sauter - type electric field , an interplay between perturbative and nonperturbative effects for the @xmath7 pair creation from the time - dependent electric field is controlled by two dimensionless parameters @xmath182 and @xmath27 .
perturbative pair creation occurs when @xmath183 is satisfied , while nonperturbative pair creation ( the schwinger mechanism ) occurs when @xmath184 is satisfied . in particular ,
an enhancement of the electron number density seen when the pulse duration and the field strength is relatively short @xmath147 and weak @xmath139 , respectively , can be understood as the lowest order perturbative process with a single photon . throughout this paper
, we considered the case where the background field is described by the sauter - type pulse in order to explicitly perform an analytic calculation .
however , we stress that our qualitative discussion should be valid for more general pulse fields smoothly characterized by its height @xmath1 and width @xmath2 .
it is also interesting to note that , though we focused on the pulse in _ time _ in this paper , our analysis implies that the finite _ space _ extension would also affect the interplay between perturbative and nonperturbative aspects of the phenomena under strong fields .
our analysis is instructive when we consider the effects of time - dependent strong fields in actual physical situations .
let us briefly discuss the case in high - energy heavy - ion collisions as an example .
it is estimated that a very strong field @xmath185 is generated in heavy - ion collisions operated in the relativistic heavy ion collider ( rhic ) at bnl and the large hadron collider ( lhc ) at cern .
( a ) in noncentral collisions such that two nuclei can touch each other , numerical simulations@xcite have shown that its strength is of the order of @xmath186 for rhic and @xmath187 for lhc , where @xmath188 is the pion mass .
( b ) in the ultraperipheral collisions where two nuclei do not touch each other , the electric field is still very strong and is estimated as @xmath189 , where @xmath190 is the impact parameter and @xmath191 is the lorentz factor . with modest parameters , the strength of the field reaches @xmath192 for rhic ( @xmath193 ) and @xmath194 for lhc ( @xmath195 ) . at first sight
, it might be natural to expect there exists nonperturbative strong field effects such as the schwinger mechanism because the field is extremely strong @xmath196 .
however , from the analysis of the present paper , we have learned that we need to be careful about the finite lifetime of the strong fields .
indeed , the duration @xmath2 of the strong field is extremely short when compared to the typical energy scale of the system @xmath15 : @xmath197 .
for instance , a rough estimate yields @xmath198 ( rhic , lhc ) for case ( a ) and @xmath199 ( rhic ; @xmath200 ) and @xmath201 ( lhc ; @xmath202 ) for case ( b ) . with such short durations ,
the important parameter @xmath28 can be large @xmath203 ; however , @xmath29 is always so small @xmath204 that the pair creation in these processes is no longer nonperturbative and the perturbative treatment would be sufficient . however , as suggested in ref .
@xcite for case ( a ) , the matter created in the collisions could let the electric field survive longer than the naive estimation .
if this is the case , there is a possibility that pair creation could be nonperturbative .
this work was supported in part by the center for the promotion of integrated sciences ( cpis ) of sokendai and grants - in - aid for scientific research of mext [ ( c)24540255 ] .
let us find an asymptotic expression of @xmath205 ( [ eq222 ] ) for @xmath206 and @xmath207 .
the diversity of strong - field physics can be seen in the proceedings and web pages of a series of workshops , physics in intense fields ( pif ) : * pif2010 * , november 2010 , kek , japan , edited by k.itakura , et al . , http://ccdb5fs.kek.jp/tiff/2010/1025/1025013.pdf and http://atfweb.kek.jp/pif2010/ ; * pif2013 * , july 2013 , desy hamburg , https://indico.desy.de/conferencedisplay.py?confid=7155 .
|
we investigate electron - positron pair creation from the vacuum in a pulsed electric background field . employing the sauter - type pulsed field @xmath0 with height @xmath1 and width @xmath2 , we demonstrate explicitly the interplay between the nonperturbative and perturbative aspects of pair creation in the background field .
we analytically compute the number of produced pairs from the vacuum in the sauter - type field , and the result reproduces schwinger s nonperturbative formula in the long pulse limit ( the constant field limit ) , while in the short pulse limit it coincides with the leading - order perturbative result .
we show that two dimensionless parameters @xmath3 and @xmath4 characterize the importance of multiple interactions with the fields and the transition from the perturbative to the nonperturbative regime .
we also find that pair creation is enhanced compared to schwinger s formula when the field strength is relativity weak @xmath5 and the pulse duration is relatively short @xmath6 , and reveal that the enhancement is predominantly described by the lowest order perturbation with a single photon .
ut - komaba/14 - 1 kek - th-1737 * finite pulse effects on @xmath7 pair creation from strong electric fields * h. taya@xmath8 , h. fujii@xmath9 , and k. itakura@xmath10 + @xmath9 _ institute of physics , university of tokyo , komaba 3 - 8 - 1 , tokyo 153 - 8902 , japan _
+ @xmath11 _ department of physics , the university of tokyo , hongo 7 - 3 - 1 , bunkyo - ku , tokyo 113 - 0033 , japan _
+ @xmath12 _ theory center , ipns , high energy accelerator research organization ( kek ) , tsukuba , oho 1 - 1 , ibaraki 305 - 0801 , japan
_ + @xmath13 _ department of particle and nuclear studies , graduate university for advanced studies ( sokendai ) , oho 1 - 1 , tsukuba , ibaraki 305 - 0801 , japan _
|
the first experimental observation of a new , manifestly exotic baryon ( b=1 , s=1 , originally called the @xmath3 but now denoted as the @xmath4 ) has been reported by leps collaboration @xcite . the experiment was carried out at the laser - electron photon facility at spring-8 in japan .
a sharp baryon resonance peak for the strangeness quantum number s=+1 was found at @xmath5 gev in the @xmath6 missing mass spectrum of the @xmath7 reaction on @xmath8 .
the width of the resonance was estimated to be smaller than 25 mev and gaussian significance of the peak was 4.6 @xmath9 .
soon after the confirmation of this observation was received by several experimental groups @xcite from different experiments where sharp peaks were observed in the @xmath10 and @xmath11 invariant mass spectra at the mass near 1540 mev . as a rule , in all experiments
a width was limited by the experimental resolution .
diana collaboration @xcite reported results from analysis of bubble - chamber data for the charge - exchange reaction @xmath12 where the spectrum of @xmath13 effective mass shows a resonant enhancement with m = [email protected] gev and @xmath15 9 mev .
the statistical significance of the enhancement is near 4.4 @xmath9 .
a narrow peak in the @xmath16 invariant mass spectrum that can be attributed to exotic baryon with strangeness s=+1 was seen by clas collaboration @xcite in exclusive measurement of the reaction @xmath17 .
this peak is at [email protected] gev with a measured width of 0.021 gev , which is largely determined by experimental mass resolution , and has a statistical significance 5.3 @xmath140.5 @xmath9 .
the positive - strangeness baryon resonance @xmath4 was observed in photoproduction of the @xmath18 final state with the saphir detector at the bonn electron stretcher accelerator elsa @xcite .
it was seen as a peak in the @xmath10 invariant mass distribution with a 4.8@xmath9 confidence level at the mass @xmath19=1540@xmath144@xmath142 mev with an upper limit of the width @xmath2025 mev at 90% c.l .
itep group @xcite reported an evidence for formation of a narrow @xmath21 resonance with the mass estimated as 1533@xmath145 mev in neutrino and antineutrino collisions with nuclei .
the observed width is less than 20 mev being entirely due to experimental resolution , and statistical significance of the signal is near 6.7 standard deviations .
it is supposed that this resonance arises from neutrino production of the @xmath4 pentaquark baryon .
the performed analysis was based on the data obtained in past neutrino experiments with big bubble chambers : wa21 , wa25 , wa59 , e180 , and e632 .
a new , more comprehensive study of the @xmath4 production on a proton target has been done by clas collaboration @xcite .
two reactions , @xmath22 and @xmath23 , were studied at jefferson lab using a tagged photon beams with energy range of 3.5 - 4.7 gev .
a narrow baryon state with strangeness s=+1 and mass m=1555@xmath1410 mev was observed in the @xmath10 invariant mass spectrum .
the peak s width is close to the experimental mass resolution ( fwhm = 26 mev ) of the clas detector , and its statistical significance is [email protected] @xmath9 . besides , the @xmath24 invariant mass distribution was analyzed in the reaction @xmath25 too , and no resonance structure was found there .
hermes collaboration @xcite presented the results of a search for the @xmath4 in quasi - real photoproduction on a deuterium target through the decay channel @xmath26 . a peak was observed in the @xmath11 invariant mass spectrum at 1528 @xmath14 2.6(stat ) @xmath14 2.1(syst ) mev .
depending on the background model , the naive statistical significance of the peak is 4 - 6 standard deviations and its width may be somewhat larger than the experimental resolution of @xmath9=4.3 - 6.2 mev .
in addition , no signal for the @xmath27 baryon was observed in the @xmath24 invariant distribution .
the most recent experimental evidence for the @xmath4 came from zeus collaboration @xcite and svd collaboration @xcite .
zeus collaboration presented preliminary results of an evidence for exotic baryon decaying to @xmath28-(anti)proton : a signal at 1527@xmath142(stat ) mev with a gaussian width of 10 @xmath14 2 mev and statistical significance 4 - 5 @xmath9 ( from gaussian fit ) has been observed for both @xmath29-proton and @xmath29-antiproton channels .
svd collaboration also reported an experimental observation a resonant structure with m=1526 @xmath14 3(stat ) @xmath14 3(syst ) mev and @xmath15 24 mev in the @xmath11 invariant mass spectrum in the reaction @xmath30 .
it should be emphasized that the search for baryon resonances with the strangeness s=+1 , that can not be built by three quarks , has a long and interesting history , and first experimental observation of a new , manifestly exotic baryon reported by leps collaboration @xcite was motivated in part by theoretical studies @xcite where baryon s anti - decuplet has been constructed in the framework of chiral soliton model exploiting the old and deep idea of skyrme @xcite .
taking the mass of @xmath31(1710 ) nucleon resonance as input identifying that resonance with the member of the constructed anti - decuplet , it was predicted a mass of @xmath321530 mev and a total width of less than 15 mev for the @xmath3 ( spin 1/2 , isospin 0 and strangeness + 1 ) baryon and pointed out that this region of the masses has avoided thorough searches before . in particular , the s=+1 baryon resonances with a relatively low mass of about 1530 mev have not been searched for in the @xmath33 scattering in the past , probably , because momenta of kaons were too high .
nevertheless , some claims for observing baryon states decaying into @xmath16 and @xmath13 have been made in the past ( see e.g. old particle data group booklets in the middle of 70th ) but they were all with substantially higher mass and larger width than the prediction in @xcite .
since the simplest quark assignment consistent with the quantum numbers of the @xmath4 is ( @xmath34 ) , this baryon is often called pentaquark .
so , the first problem arises how to describe pentaquark states in qcd .
this problem is extremely complicated even though on a pure empirical base e.g. what quark configuration is favorable for the @xmath4 : is it diquark - triquark pentaquark state or this is the state of antiquark and two diquarks or something else .
it is well known that a description of new , recently discovered charmonium meson states with unexpectedly low masses and narrow total decay widths is extremely problematic in the framework of phenomenology based on conventional quark potential models .
now , it appears on a more high level with pentaquark states .
here , a similar story repeats itself : the discovery of exotic baryon with low mass and narrow width brings new , yet more serious problems for that phenomenology .
concerning pure theory , up to now our theoretical understanding of ( especially low - energy ) qcd is far from what is desired
. we do nt know in quantum field theory how to construct even the states for protons and pions though from fundamental quark and gluon degrees of freedom in qcd , it appears to be very complicated and unresolved problem so far .
the best currently performed lattice computations in qcd can not help us to make calculations accurately at low energy too .
nevertheless , there is a hope that powerful computers will allow to overcome many significant technical problems in the future . in ref .
@xcite , where some of our previous studies were partially summarized , it has been claimed that existence of the extra dimensions in the spirit of kaluza and klein together with some novel dynamical ideas may provide new conceptual issues for the global solution of the spectral problem in hadron physics to create a unified picture for hadron spectra .
the question arises : what place in the unified picture for hadron spectra the @xmath0 baryon takes up ?
this paper presents the answer to that question .
recently a new , very simple and at the same time quite general physical law concerning the structure of hadron spectra has been found ; see @xcite and references therein . the developed theoretical conception allowed to construct the global solution of the spectral problem in hadron spectroscopy and provided quite new scheme of systematics for hadron states .
our approach to hadron spectroscopy has been verified with a large amount of experimental data on meson s states and received an excellent agreement . what is remarkable that all new meson states experimentally discovered last year
have been observed just at the masses predicted in our approach , and those states appeared to be narrow as predicted too .
the main advantage of our developed theoretical conception is that all calculated numbers for masses and widths do not depend on a special dynamical model but follow from fundamental hypothesis on existence of the extra dimensions with a compact internal extra space .
one very important fact has been established now in a reliable way : the size of the internal compact extra space determines the global characteristics of the hadron spectra while the masses of the constituents are the fundamental parameters of the compound systems being the elements of the global structure .
here we apply our approach to analyse the hadron system under consideration i.e. kaon - nucleon system .
first of all , according to the general law , the kaluza - klein tower of kk - excitations for the kaon - nucleon system has been built by the formula @xmath35
where ( @xmath36 ) and ( @xmath37 ) , @xmath38 is the fundamental scale established before from analysis of nucleon - nucleon dynamics at low energies ; see @xcite and references therein for the details .
the such built kaluza - klein tower is shown in table 1 .
some experimental information known from rpp booklet has been presented in this table as well . for completeness
we have also calculated the kaluza - klein tower of kk - excitations for the @xmath39 system by similar formula like ( [ kkexcit ] ) . that kaluza - klein tower is shown in table 2 together with experimental information extracted from the same rpp source .
we very hope that both tables 1 and 2 might serve as a guide for the experimenters in particle physics .
now we would like to compare the theoretically calculated spectra with experimental material presented in above mentioned papers ; see introduction .
so then , let s see one after another .
diana collaboration investigated low - energy @xmath40xe collisions by the bubble chamber filled with liquid xenon which has been exposed to a separated @xmath40 beam with momentum of 850 mev / c from the itep proton synchrotron .
the spectrum of @xmath41 effective mass was analyzed in the charge - exchange reaction @xmath42 where the neutron was bound in a xenon nucleus .
the events of this reaction were fully measured and reconstructed in space using specially designed stereo - projectors ; the details on the experimental procedure can be found in original paper @xcite .
effective mass of the @xmath41 system formed in the charge - exchange reaction is plotted in fig .
1a for all measured events where an enhancement is clear seen at the mass @xmath43 mev / c@xmath44 . to estimate the level of background
, the effective mass spectrum was fitted to a linear combination of two regular distributions : one distribution was obtained by simulation that took into account many real conditions of the experiment , and a distribution which was obtained by the method of random stars .
the result of the fit is depicted by the dashed line in fig .
1a . after that
, additional topological selections have been applied : see @xcite for details . of the 1112 measured events , nearly a half ( 541 events ) survived the additional selections . in the @xmath41 mass spectrum for these events , that is shown in fig .
1b , the enhancement near 1540 mev / c@xmath44 became more prominent .
in summary it has been concluded that a baryon resonance with mass @xmath45 mev and width @xmath46 mev has been observed in the @xmath41 effective - mass spectrum .
the statistical significance of the signal was estimated as @xmath47 .
it is suggested that the observed resonance is a strong indication for formation of the exotic pentaquark @xmath3 baryon , but the work is still in progress .
vertical , spectral lines have been plotted in figs . 1a and 1b
as well to compare the theoretically calculated spectrum of kk excitations in the @xmath41 system with the picture observed at the experiment . even though we found a strong correlation of the spectral lines with the peaks on the histograms the reported resonance peak appeared approximately in the middle between two spectral lines corresponding to @xmath48(1527)-storey and @xmath49(1558)-storey in kk tower for @xmath50 system
; see table 1 .
let s assume that further , thorough analysis would clarify a matter of dispute .
clas collaboration @xcite has performed an exclusive measurement on deuterium for the reaction @xmath52 where the final state neutron was reconstructed from the missing momentum and energy .
the data have been obtained at the thomas jefferson national accelerator facility with the clas detector and the photon tagging system : photon beams were produced by 2.474 and 3.115 gev electrons incident on a bremsstrahlung radiator , the maximum tagged photon energy was @xmath53 of the electron beam energy , the photons struck a liquid - deuterium target ; see @xcite and references therein for details .
the details on the experimental procedure as well as a discussion of the explicit cuts which have been made to remove the main background sources from the final event sample in order to enhance the signal relative to background can be found in original paper @xcite too .
final result is presented as @xmath10 invariant mass spectrum which is shown in fig .
2 . a fit ( solid line ) to the peak and a gaussian plus constant term fit to the background ( dashed line ) are also depicted in this figure .
the peak is at @xmath54 gev with a width ( fwhm ) of 0.021 gev .
the width is consistent with the instrumental resolution .
the uncertainty of 0.005 gev in the mass is due to calibration uncertainties of the photon tagging spectrometer , the electron beam energy , and the momentum reconstruction in clas .
the statistical significance of the peak is @xmath55 .
the spectrum of events removed by the @xmath1(1520 ) cut is shown in fig . 2 by the dashed - dotted histogram , and does not appear to be associated with the peak at 1.542 gev .
the invariant mass spectrum of @xmath24 system was also examined by clas collaboration using the same event selection as in the case of @xmath10 system .
the statistics were limited , but there was no clear peak in the signal region .
it was noted that the clas acceptance for the @xmath24 system is not the same as for the @xmath10 system , so the two spectra are not directly comparable . the absence of peak in @xmath56 spectrum suggests that @xmath4 is probably an isosinglet , though there is no a firm conclusion based on the current data .
we have plotted in fig .
2 the spectral lines corresponding to kk excitations in the @xmath10 system taken from table 1 . here
it is also found a strong correlation of the spectral lines with the peaks on the histogram .
the reported resonance peak appeared between two spectral lines corresponding to @xmath57(1525)-storey and @xmath58(1556)-storey in kk tower for @xmath10 system though a little bit closed in second spectral line of @xmath58(1556 ) .
saphir collaboration @xcite presented evidence for the @xmath4 in photoproduction of the @xmath18 final state off protons .
actually , the reaction @xmath59 with the decay @xmath60 was measured with the saphir detector at elsa .
the elsa electron beam produced photons via bremsstrahlung in a copper foil radiator .
the tagged photon energies were from 31% to 94% of the incident electron energy which was 2.8 gev for the data presented in this experiment .
the photon beam passed through a liquid hydrogen target .
interacting photons were detected in a photon counter .
the coincidences of tagger and photon counter determined the photon flux .
additional , needed details concerning the experimental procedure can be found in original paper @xcite .
the observation of the @xmath4 in the reaction chain @xmath61 has been presented by fig .
3 showing the @xmath10 invariant mass distribution after cuts ; original paper @xcite contains a comprehensive discussion of all cuts which have been made to arrive at this figure . as is seen from fig .
3 there is a clear peak at @xmath62mev .
the statistical significance of the peak is @xmath63 .
the photoproduction of the @xmath64 final state has been studied by saphir collaboration as well to search for the doubly charged @xmath27 in the reaction chain @xmath65 fig .
4 shows the resulting @xmath24 invariant mass spectrum .
it should be emphasized that the data are consistent with a small structure in the @xmath24 invariant mass distribution .
however , since the saphir acceptance was considerably larger for the fully constrained @xmath64 events they expected a peak with more than 5000 @xmath27 , far above the observed level .
the absence of a signal in the @xmath24 invariant mass distribution at the expected strength resulted in conclusion that the @xmath4 must be isoscalar .
as before we have plotted in figs .
3 and 4 the spectral lines corresponding to kk excitations in the @xmath10 and @xmath24 systems taken from table 1 .
again we found a strong correlation of the spectral lines with the peaks on the histograms .
the reported resonance peak of @xmath4 in fig.3 appears just in the middle between two spectral lines corresponding to @xmath57(1525)-storey and @xmath58(1556)-storey in kk tower for @xmath10 system ( see table 1 ) , and this peak is broad enough .
however , we would like to point out a remarkable fact that the spectral line @xmath66(1555 ) in fig.4 exactly coincided with the experimentally observed structure in the @xmath24 invariant mass distribution . in our opinion further , scrupulous experimental studies of kaon - nucleon system are very desirable , no doubt they would welcome .
itep group reported a search for formation of the @xmath4 baryon in neutrino and antineutrino collisions with protons , deuterons , and neon nuclei @xcite .
the data collected by several neutrino experiments with big bubble chambers bebc at cern and the 15-foot chamber at fermilab have been analyzed .
compiled database included about 120 000 @xmath67 and @xmath68 induced charged - current events , and covered the bulk of neutrino data collected with bebc ( wa21 , wa25 , and wa59 ) and a significant fraction of data collected with the 15-foot bubble chamber at fermilab ( e180 and e632 ) .
neutral - current interactions were not systematically included in the database , and therefore the analysis was restricted to charged - current events only .
it has been stressed that neutrino data from big bubble chambers are still unrivaled in quality and completeness of physics information .
further details on the neutrino experiments can be found in original paper @xcite and references therein .
figure 5 shows the @xmath21 invariant mass distribution for the neon and deuterium data combined ( top panel ) .
in the bottom panel the same @xmath21 distribution with bins shifted by 5 mev has been plotted .
a fit of the bottom histogram yielded @xmath69 mev and @xmath70 mev for the position and r.m.s .
width of the resonance .
the statistical significance of the peak appeared to be near 6.7 standard deviations .
we have plotted in fig .
5 ( top panel ) the spectral lines corresponding to kk excitations in the @xmath50 system taken from table 1 and found a striking picture : all spectral lines without any exception coincided with the peaks on the histogram .
the reported resonance peak of @xmath4 just corresponds to the spectral line @xmath48(1527 ) in kk tower for @xmath50 system ( see table 1 ) .
really , analyzing the excellent neutrino data resulted in excellent agreement with the theory .
recently a more comprehensive study of the @xmath4 production on a proton target which includes data from three distinct runs under different experimental conditions in clas has been reported by clas collaboration @xcite .
two reactions , @xmath72 and @xmath73 , have been analyzed in the three runs having identical geometrical acceptance and trigger requirement but with a slightly different tagged photon beams in the energy range of 3.23.95 gev , 35.25 gev , and 4.85.47 gev , respectively ( details described in @xcite ) .
the combined analysis of these three runs offered access to a wider range of acceptance and energies . the final @xmath74 invariant mass spectrum calculated from missing mass in the reaction @xmath75 , combining data from all three data runs ,
is shown in fig .
the authors concluded that no obvious structure is seen in this spectrum .
figure 7 shows the resulting @xmath10 mass spectrum prepared by applying a several explicit angular cuts . as is seen the @xmath4 peak
was clearly observed in that figure . the @xmath10 effective mass distribution shown in fig .
7 was fitted by the sum of a gaussian function and a background function obtained from the simulation .
the fit parameters are : @xmath76 , @xmath77 mev , and @xmath78 mev ( fwhm ) , where the errors are statistical .
the systematic mass scale uncertainty was estimated to be @xmath79 mev .
it was also estimated the significance to be 7.8 @xmath14 1.0 @xmath9 .
in addition , a search for a manifestly exotic baryon ( @xmath80 , @xmath81 ) was performed in the reaction @xmath82 . using the similar procedures as in previous case ,
the @xmath24 invariant mass distribution was built but there were no resonance structures evident in that distribution .
however , a more detailed analysis will be presented in a future . as usual ,
we have plotted in figs . 6 and 7
the spectral lines corresponding to kk excitations in the @xmath10 system taken from table 1 . here , in addition to strong correlation of the spectral lines with the peaks on the histograms as in previous cases , we found out new , quite a remarkable fact : the reported resonance peak of @xmath4 just corresponds to the spectral line @xmath58(1556 ) in kk tower for the @xmath10 system ( see table 1 ) .
the results of a search for the @xmath4 in quasi - real photoproduction on deuterium have been presented by hermes collaboration @xcite .
the data were obtained by the hermes experiment with the 27.6gev positron beam of the hera storage ring at desy .
the analysis searched for inclusive photoproduction of the @xmath4 followed by the decay chain @xmath83 .
events selected contained at least three tracks : two oppositely charged pions in coincidence with one proton .
the event selection included constraints on the event topology to maximize the yield of the @xmath84 peak in the @xmath85 spectrum while minimizing its background .
however , no constraints were optimized to increase the significance of the signal visible in the final @xmath86 spectrum , as such optimization would have produced a spectrum to which standard statistical tests do not apply . to search for the @xmath4 , events were selected with a @xmath85 invariant mass within @xmath87 about the centroid of the @xmath28 peak .
more details on the experimental procedure described in original paper @xcite where the interested reader referred to .
the resulting spectrum of the invariant mass of the @xmath88 system is displayed in fig .
8 . as is seen a narrow peak is observed , and this peak is identified with the @xmath4 .
figure 9 shows the spectra of invariant mass @xmath89 ( top ) and @xmath90 ( bottom ) . the event selection for the spectra in this figure was the same as for the @xmath91 analysis , except the reconstructed @xmath84 track
was replaced by that of the observed charged kaon .
the @xmath92 mass peak fitted with a gaussian is shown in the @xmath89 spectrum of fig . 9 ( details of the fit can be found in the original paper ) .
a clear peak is seen for the @xmath92 in the @xmath89 invariant mass distribution .
however , no peak structure is seen for the hypothetical @xmath27 in the @xmath90 invariant mass distribution near 1.53gev .
following our general strategy , as before , we have plotted in figs . 8 and 9 the spectral lines corresponding to kk excitations in the @xmath50 system ( fig . 8) and @xmath93 system ( fig .
9 ) taken from table 1 . no doubt , this is quite a remarkable fact that the reported resonance peak of @xmath4 exactly coincided with the spectral line @xmath48(1527 ) in kk tower for the @xmath50 system , and @xmath92 peak just corresponds to the spectral line @xmath94(1523.6 ) in kk tower for the @xmath93 system ( see table 1 ) .
moreover , we observe a barely visible structures in the @xmath90 invariant mass distribution around the spectral lines depicted in fig .
9 . clearly , a much more careful experimental studies with a higher statistics and resolution are utterly desired .
svd collaboration @xcite presented the preliminary results of a search for the @xmath4 in proton - nuclei(c , si , pb ) interactions at 70 gev(@xmath95=11.5 gev ) ihep accelerator ( protvino ) with the svd-2 setup in the reaction chain : @xmath96 ( svd-2 experimental setup described in the paper ) . the data analysis was done in the inclusive reactions with the limited multiplicity in the proton beam fragmentation region .
the resulting @xmath97 effective mass distribution before the cuts is shown in fig .
the authors point out that no obvious structure in this spectrum is seen , however there is a small enhancement in the 1530 mev mass region . after applying a special cuts ( details of cuts procedure
can be found in original paper ) a narrow peak in the @xmath97 effective mass distribution was seen at the mass @xmath98 mev with a @xmath99 mev .
the statistical significance of this peak was estimated to be of @xmath100 .
as it should be , we have also plotted in fig .
10 the spectral lines corresponding to kk excitations in the @xmath50 system taken from table 1 .
apart from strong correlation of the spectral lines with the peaks on the histogram in fig . 10 , here is found out too that the above mentioned small enhancement in the 1530 mev mass region " just corresponds to the spectral line @xmath48(1527 ) in kk tower for the @xmath50 system ( see table 1 ) .
unfortunately , we did nt have in hand the published paper of zeus collaboration .
however , we can arrive at ref .
@xcite by @xmath101www@xmath102 where the preliminary results of an evidence for exotic baryon decaying to @xmath28-(anti)proton were presented . a signal at 1527@xmath142(stat )
mev with a gaussian width of 10 @xmath14 2 mev and statistical significance 4 - 5 @xmath9 ( from gaussian fit ) has been observed .
the measured mass is in excellent agreement with the predicted state @xmath48(1527 ) from kk tower for the @xmath50 system ( see table 1 ) .
what is remarkable here that zeus collaboration observed this state for both @xmath29-proton and @xmath29-antiproton channels as we predicted too .
recently zeus collaboration papers arrived at lanl e - print archive @xcite where we found that a resonance search has been made in the @xmath21 and @xmath103 invariant mass spectrum measured with the zeus detector at hera using an integrated luminosity of 121 pb@xmath104 , taken between 1996 and 2000 .
the @xmath105 invariant mass spectrum has been studied in the central rapidity region of inclusive deep inelastic scattering at an @xmath106 center - of - mass energy of 300318 gev for a large range in the exchanged - photon virtuality , @xmath107 , above 1 gev@xmath44 .
the results presented there support the existence of exotic baryon decaying to @xmath28-(anti)proton , and these results constitute first evidence for the production of such a state in a kinematic region where hadron production is dominated by fragmentation .
the peak position , determined from a fit to the mass distribution in the kinematic region @xmath108 gev@xmath44 , is @xmath109 mev , and the measured gaussian width of @xmath110 mev is above , but rather close to the experimental resolution of @xmath111 mev .
the signal is visible at high @xmath107 and , for @xmath112 gev@xmath44 , contains @xmath113 events .
the statistical significance , estimated from the number of events assigned to the signal by the fit , varied between @xmath114 and @xmath115 depending upon the treatment of the background .
the mass of the @xmath4 baryon resonance reported now lies somewhat below the average mass of the previous preliminary results @xcite .
this is because some additional cuts ( energy and k * cuts ) were removed in the final analysis compared to the previous preliminary results , and only de / dx cuts were saved to get a good purity for the proton sample .
the zeus final results @xcite agree still with the zeus preliminary results @xcite ( without systematics ) though .
the result of the fit using two gaussians is shown in fig .
it was found that the second gaussian significantly improved the fit in the low mass region .
it has a mass of @xmath116 mev and a width of @xmath117 mev and may correspond to the @xmath118 .
however , as pointed out , the parameters and significance of any state in this region are difficult to estimate due to the steeply falling background close to threshold .
the probability of the 1522 mev signal anywhere in the range 15001560 mev arising from statistical fluctuation of the background was below @xmath119 . for a more realistic case , when the starting background distribution included the 1465 mev gaussian the probability was found to be about a factor of ten lower . in order to determine the natural width of the state ,
the breit - wigner function convoluted with the gaussian was used in the fitting procedure to describe the peak near 1522 mev . if the width of the gaussian was fixed to the experimental resolution , the estimated intrinsic width of the signal turned out @xmath120 mev .
the systematical error on this width was expected to be smaller than the statistical error , but due to low statistics , complicated background and the narrowness of the peak leading to unstable fits , full systematical uncertainty was difficult to estimate @xcite .
the invariant mass spectrum was also investigated for the @xmath21 and @xmath103 samples separately .
this result is shown as an inset in fig . 11 for @xmath121 gev@xmath44 .
the results for two decay channels are compatible , though the number of @xmath103 candidates is systematically lower .
the mass distributions were fitted using the same function as the combined sample and gave statistically consistent results for the peak position and width .
the number of events in the @xmath103 channel is @xmath122 .
if one considers the signal in the @xmath21 channel as the @xmath4 then the signal in the @xmath103 channel might be identified with the @xmath123 .
this is quite a new , remarkable observation made by zeus collaboration . besides
, the @xmath124 invariant mass spectra were investigated for a wide range of minimum @xmath107 values as well , identifying proton and charged kaon candidates using @xmath125 in a kinematic region similar to that used in the @xmath105 analysis . for @xmath126 gev@xmath44 , no peak
was observed near 1522 mev in the @xmath127 spectrum , while a clean @xmath128 signal was observed in the @xmath129 channel at @xmath130 mev , corresponding to the pdg @xmath131 state ; see fig . 5 in @xcite .
the @xmath127 and @xmath132 mass spectra were especially investigated at @xmath133 gev@xmath44 , where the 1522 mev peak is clearly pronounced for the @xmath105 channel .
again , no sign of a peak was found in the @xmath127 decays ; see fig . 6 in @xcite .
however , we would like to emphasize , looking at the figures 5 and 6 in @xcite , one can not definitely conclude that no structures are seen in the @xmath127 mass spectra .
certainly , more statistics and higher mass resolution are needed to make a physically adequate statement .
we have plotted in fig .
11 the spectral lines corresponding to kk excitations in the @xmath50 system taken from table 1 . as seen in fig . 11 , the spectral lines of the @xmath134 , @xmath135 and @xmath136 from kaluza - klein tower for kaon - nucleon system may correspond to the observed structures in invariant mass spectra for the @xmath105 channel .
unfortunately , the statistics is small and mass resolution is not so high to make a more reliable statement concerning the structures seen from the left side of the 1522 mev peak .
we also found a strong correlation of the calculated spectral lines with the enhancements on the right side from the 1522 mev peak in the invariant mass spectrum . in this respect , it would be extremely important to improve the statistics and mass resolution of the experiment , below and above of the 1522 mev peak , to confirm the lower and higher new exotic states predicted .
quite an interesting results came just recently from dubna experiment @xcite where the 2 m propane bubble chamber experimental data have been analyzed to search for the exotic baryon decaying to @xmath28-proton in the reaction p+@xmath137 at 10 gev .
the observation of three peaks in the @xmath97 invariant mass spectrum has been reported with the masses @xmath138 = ( [email protected] , [email protected] , [email protected] ) mev and the measured widths of @xmath139 = ( [email protected] , [email protected] , [email protected] ) mev .
the statistical significance of these peaks were estimated to be of ( 5.5@xmath140,5)@xmath9 , ( [email protected])@xmath9 and ( [email protected])@xmath9 respectively . the effective mass distribution for 2300 @xmath140 combination is shown in fig .
the background events were analyzed using the same experimental condition .
the background distribution was fitted by six - order polynomial . the solid curve in fig .
12 represents the sum of the background and 4 breit - wigner resonances . as seen in fig .
12 , there are significant enhancements in 1539 , 1610 and 1810 mev mass regions .
there is also a small peak in 1980 mev mass region .
the total ( 5554 combination ) @xmath140 effective mass distribution is shown in fig . 13 .
the total experimental background was obtained with the same experimental condition .
the significant enhancements are clear seen in fig . 13 in 1545 , 1616 and 1811 mev mass regions .
a small peak in 2.0 gev mass region is pointed out too .
the calculated spectral lines corresponding to kk excitations in the @xmath50 system taken from table 1 have been plotted in figs .
12 and 13 as well . here ,
as in previous cases , we also found a remarkable coincidence of the calculated spectral lines with the enhancements in the invariant mass spectra .
we have presented an analysis of the recent results from several experimental groups reported observation of a new @xmath0 baryon .
our analysis was motivated by the wishes to understand a real origin of the @xmath0 baryon and to find its place in the unified picture for hadron spectra developed early @xcite .
a thorough analysis of all experiments mentioned in introduction taken together shows that , in fact , two different @xmath0 baryons have been discovered : one @xmath141 baryon in the @xmath142 system and other @xmath4 baryon in the @xmath10 system .
it s our first claim .
we have established that both @xmath0 baryons are excellently incorporated in the unified picture for hadron spectra : the @xmath143 baryon lives on the six storey in kaluza - klein tower for @xmath50 system but the @xmath144 baryon lives on the seven storey in kaluza - klein tower for @xmath10 system ; see table 1 .
as is seen from table 1 , almost all experimentally filled storeys in kaluza - klein tower for kaon - nucleon system contain pairs of @xmath2 and @xmath1 particles with negative strangeness .
one might think that @xmath143 baryon represents an isovector with positive strangeness partner of isoscalar @xmath1(1520 ) particle with negative strangeness , and @xmath145 baryon is isoscalar with positive strangeness partner of isovector @xmath2(1560 ) particle with negative strangeness . in this respect
, it would be quite an important to set up purposeful experiments to search for the states @xmath146 and @xmath147 .
it would also be very interesting an attempt to seek a rare decay mode @xmath148 with an isospin violation ( see 8-storey in table 2 ) . from strong correlation between spectral lines corresponding to kaluza - klein tower for kaon - nucleon system and the peaks on the histograms established above
, one can conclude that there exist positive srangeness @xmath0 partners for all @xmath2 and @xmath1 particles with negative strangeness as we predicted this is our second claim
. the most impressive manifestation and confirmation of that claim we found in neutrino interactions data set @xcite .
our predictions should be tested in the future experimental studies .
the properties of the observed @xmath0 states , such as spin , isospin , and parity are not established in the above mentioned experiments .
so , it is important to measure the quantum numbers of the @xmath0 particles in the future experiments . in our opinion , a classification of baryons by soliton configurations in chiral field theories is more preferable compared to the standard quark model classification . in particular ,
the su(3 ) version of the chiral soliton model @xcite predicts the @xmath4 state of having spin 1/2 , isospin 0 , and positive parity .
attractive in the soliton classification scheme that the @xmath0 baryon is considered on an equal ground with other baryons .
this is ideologically close to our scheme of systematics .
however , our description significantly differs in many important things from the chiral soliton description .
our conservative estimate for the widths of kk excitations looks like @xmath149 mev , where @xmath150 is the number of kk excitation .
this gives @xmath151 mev and @xmath152 mev .
thus , a more careful experimental studies with a higher statistics and mass resolution are extremely important . here
, we should like to bring up some comments on the problems under discussion .
if one takes the structures below and above of the @xmath0 peak in the @xmath21 invariant mass spectra observed in @xcite seriously , then we come to the conclusion that the experimentally observed picture is not at all what is predicted by the chiral soliton model . as an additional argument in favor of that claim we would like to refer to the recent , interesting article @xcite where a careful analysis of @xmath153 total cross section data was undertaken to explore possible manifestations of the resonance in the s=+1 hadronic system with mass around 1.55 gev .
it was found that a structure corresponding to the resonance is visible in the data .
the width consistent with the observed deviation from background was found to be [email protected] mev and the mass is [email protected] gev for the spin parity @xmath154 and [email protected] gev for @xmath155 .
perhaps this is the first experimental verdict on a crucial problem concerning the quantum numbers of the @xmath0 baryon .
the obtained mass [email protected] gev for the spin parity @xmath154 is in excellent agreement with the @xmath156(1556mev)-storey in kaluza - klein tower for kaon - nucleon system ( see table 1 ) and far enough from accurate prediction 1530 mev in @xcite .
the mass [email protected] gev is closer to the prediction 1530 mev but the negative parity ( @xmath155 ) in that case contradicts to the predicted @xcite positive parity ( @xmath154 ) .
we would also like to point out , once again , a noticeable spread among the masses of the @xmath0 states observed in different experiments .
however , it should be noted that the clas collaboration @xcite reported a very small statistical error on the mass of the @xmath4 which is 1555@xmath141(stat . )
the large error in the quoted mass @xcite was due to possible systematic errors .
the mass value 1555@xmath141(stat . )
mev is in excellent agreement with the @xmath156(1556mev)-storey in the kaluza - klein tower for kaon - nucleon system as well , and this might be considered as an additional argument in favor of the spin parity @xmath154 .
certainly , this is a remarkable fact that it may be possible to infer the parity of the state from a comparison of its mass values @xcite . in this respect
a more accurate experimental measurement and precise determination of the masses of the @xmath0 states is a question of vital importance as well .
no doubt , further measurements to obtain a more accurate kaon - nucleon scattering data would be very useful to better establish all possible states in kaon - nucleon system with positive and negative strangeness simultaneously .
we have already claimed that the year 2003 will enter the history of particle physics as a year of fundamental discoveries @xcite .
this concerned a series of new narrow mesons states discovered at the masses which are surprisingly far from the predictions of conventional quark potential models .
exciting experimental measurements in baryons spectroscopy in the same year and next with discovery of the @xmath0 baryon states provided an additional excellent confirmation of that claim .
all of these discoveries in meson and baryon spectroscopy clearly open a new page in hadron physics which is certainly a starting page of a new era in particle physics .
one might think that this is an era of the extra dimensions .
i would like to thank a.v .
razumov for friendly table conversations and useful professional advices . * * t. nakano _
et al . _ ,
lett . * 91 * , 012002 ( 2003 ) ; arxiv : hep - ex/0301020 .
diana coll .
barmin _ et al .
_ , arxiv : hep - exp/0304040 ( 2003 ) .
clas coll .
, s. stepanyan _ et al . _ ,
arxiv : hep - ex/0307018 ( 2003 ) .
saphir coll .
, j. barth _ et al .
_ , arxiv : hep - ex/0307083 ( 2003 ) .
a. e. asratyan _ et al .
_ , arxiv : hep - ex/0309042 .
clas coll .
, v. kubarovsky _ et al .
_ , arxiv : hep - ex/0311046 .
hermes coll .
, a. airapetian _ et al .
_ , desy 03 - 213 , december 2003 ; arxiv : hep - ex/0312044 .
zeus coll .
, s. chekanov , http://www.desy.de/f/seminar/sem_schedule.html svd coll .
, a. aleev _ et al .
_ , arxiv : hep - ex/0401024 .
d. diakonov _ et al .
_ , z. phys . * a 359 * , 305 ( 1997 )
. t. h. r. skyrme , nucl . phys . * 31 * , 556 ( 1962 ) .
arkhipov , arxiv : hep - ph/0309327 ( 2003 ) ; preprint ihep 2003 - 37 , protvino , 2003 , available at http://dbserv.ihep.su/~pubs/prep2003/ps/2003-37.pdf zeus coll . , s. chekanov _ et al .
_ , arxiv : hep - ex/0403051 . s. v. chekanov ( for the zeus collaboration ) , arxiv : hep - ex/0404007 .
p. zh .
aslanyan , v. n. emelyanenko , g. g. rikhkvitzkaya , arxiv : hep - ex/0403044 . w. r. gibbs , arxiv : nucl - th/0405024 . system formed in the reaction @xmath157 extracted from ref .
@xcite : ( a ) for all measured events , ( b ) for events that pass additional selections ; see @xcite .
the vertical ( spectral ) lines correspond to kk tower for @xmath41 system ; see table 1.,title="fig:"][fig1 ] mass distribution after cuts in the @xmath158 mass distribution and in the @xmath84 production angle extracted from ref .
the solid line represents a fit using a breit wigner distribution ( convoluted with a resolution function ) plus polynomial background ; see @xcite for details .
the vertical ( spectral ) lines correspond to kk tower for @xmath10 system ; see table 1.,title="fig:"][fig3 ] invariant mass distribution in the @xmath64 final state extracted from ref .
the solid line represents a fit using a breit wigner distribution ( convoluted with a resolution function ) plus polynomial background ; see @xcite for details .
the vertical ( spectral ) lines correspond to kk tower for @xmath24 system ; see table 1.,title="fig : " ] [ fig4 ] invariant mass distribution for the neon and deuterium data combined ( top panel ) .
the dots depict the random - star background .
the vertical ( spectral ) lines correspond to kk tower for @xmath50 system ; see table 1 .
a fit of the same @xmath21 distribution but plotted with shifted bins is shown in the bottom panel ; see @xcite for details.,title="fig:"][fig5 ] invariant mass spectrum in the reaction @xmath160 with the cut @xmath161 and @xmath162 . @xmath163 and @xmath164 are the angles between the @xmath165 and @xmath40 mesons and photon beam in the center - of - mass system .
the vertical ( spectral ) lines correspond to kk tower for @xmath10 system ; see table 1 .
the inset shows the @xmath10 invariant mass spectrum with only the @xmath161 cut ; see @xcite for details .
, title="fig:"][fig7 ] system subject to various constraints described in ref . @xcite . in top panel
the phythia6 monte carlo simulation is represented by the gray shaded histogram ; see @xcite for details .
the vertical ( spectral ) lines in top panel correspond to kk tower for @xmath13 system ; see table 1.,title="fig:"][fig8 ] invariant mass spectrum in the reaction @xmath167 before the cuts extracted from ref .
the vertical ( spectral ) lines correspond to kk tower for @xmath41 system ; see table 1.,title="fig:"][fig10 ] channel for @xmath121 gev@xmath44 extracted from ref .
the solid line is the result of a fit to the data using a three - parameter background function plus two gaussians ( see @xcite ) .
the dashed lines show the gaussian components and the dotted line the background according to this fit .
the histogram shows the prediction of the ariadne mc simulation normalized to the data in the mass region above 1650 mev .
the inset shows the @xmath103 ( open circles ) and the @xmath21 ( black dots ) candidates separately , compared to the result of the fit to the combined sample scaled by a factor of 0.5 .
the vertical ( spectral ) lines correspond to kk tower for @xmath41 system ; see table 1.,title="fig:"][fig11 ] invariant - mass spectrum with the momentum of @xmath168gev / c for identified protons in the reaction p+@xmath169 + x extracted from ref .
the solid curve represents the sum of the experimental background and 4 breit - wigner resonances .
the experimental background was approximated by six - order polynomial .
the vertical ( spectral ) lines correspond to kk tower for @xmath41 system ; see table 1.,title="fig:"][fig12 ] effective mass distribution combined for protons with the momenta @xmath170 gev / c@xmath44 and @xmath171 extracted from ref .
the curve is the experimental background taken in the form of six - order polynomial .
the vertical ( spectral ) lines correspond to kk tower for @xmath41 system ; see table 1.,title="fig:"][fig13 ]
|
an analysis of the recent results from several experimental groups reported observation of a new @xmath0 baryon has been presented from a view point of the unified picture for hadron spectra developed early @xcite .
it is shown that , in fact , two different @xmath0 baryons have been discovered .
we have also established that both @xmath0 baryons are excellently incorporated in the unified picture for hadron spectra . it is argued that the presented experimental material revealed an existence of the positive srangeness @xmath0 partners for the observed @xmath1 and @xmath2 states with negative strangeness as we predicted .
* what can we learn about @xmath0 baryon + from unified picture for hadron spectra ?
* + a.a .
arkhipov + _ state research center institute for high energy physics " + 142281 protvino , moscow region , russia _ +
|
the success of lcls @xcite motivated the planning of further development . according to this planning
, the two new undulators will be installed in a new undulator tunnel , and their radiation be directed to novel underground experimental halls .
the lcls baseline undulator will remain operative at least during the next decade @xcite . as a result
, enhancing the operating lcls baseline capability becomes a challenging problem , subject to many constraints including low cost , little available time to perform changes ( and to be subtracted from operations ) and almost no risk i.e. guarantee of safe return to the baseline mode of operation . in a previous paper of us
@xcite we began to analyze possibilities for enhancing the lcls baseline subject to the above constraints .
the setup proposed in that paper is based on the insertion of a single , short magnetic chicane setup , which allows one to implement a novel single - bunch self - seeding scheme exploiting a single crystal monochromator , which is extremely compact and can be straightforwardly installed in the lcls baseline undulator .
the setup can be installed in little time , and is not expensive . with the radiation beam
monochromatized down to the fourier transform limit , a variety of very different techniques leading to further improvements of the lcls performance become feasible .
in particular , in @xcite we described an efficient way for obtaining full polarization control at the lcls hard x - ray fel .
we also proposed to exploit crystals in the bragg reflection geometry as movable deflectors for the lcls x - ray transport systems .
the hard x - ray beam can be deflected of an angle of order of a radian without perturbations .
the monochromatization of the output radiation constitutes the key for reaching such result . with reference to fig .
[ lclsbm1 ] one can appreciate the flexibility of our setup . if the chicane is switched off , the baseline mode of operation is immediately restored , fig .
[ lclsbm1 ] ( a ) .
the self - seeding scheme can be implemented by introducing a wake monochromator @xcite , fig .
[ lclsbm1 ] ( b ) . by retracting the crystal , inserting x - ray mirrors in its place , and
keeping the chicane switched on , one may still implement a method for ultrafast photon pulse measurements @xcite , fig .
[ lclsbm1 ] ( c ) .
the same setup can be used to implement @xcite a pump - probe scheme which is not affected by the jitter between pump laser and x - ray pulse , fig .
[ lclsbm1 ] ( d ) .
in this paper we will discuss how to further exploit the installation of the chicane .
in particular , the setup with mirrors and chicane can be used to provide a method to serve more users simultaneously , fig .
[ lclsbm1 ] ( e ) .
moreover , we will discuss how to further exploit the self - seeding scheme , in combination with a second harmonic after burner ( shab ) .
the lcls team plans @xcite to convert the last five undulator segments of the lcls baseline undulator to shab , fig .
[ lclsbm4 ] .
we propose to remove the 14th undulator segment , and to substitute it with a weak chicane , fig .
[ lclsbm2 ] . one extra use of the setup in fig .
[ lclsbm2 ] , aside for those already described in fig .
[ lclsbm1 ] ( a)-(d ) , consist in increasing the user access at the lcls baseline , which constitutes a very important goal . in its essence , the method is very simple .
we propose to generate two sase x - ray beams in different parts of the baseline undulator , and spatially separate them with the help of x - ray mirrors introduced in the offset provided by the chicane .
this will allow to simultaneously supply two experiments .
the lcls baseline undulator is long enough to provide two sase x - ray beams at saturation , and very small x - ray mirrors can be used to spatially separate the two x - ray beams .
the gap of the lcls baseline undulator can only be adjusted within a few percent , and therefore the x - ray beam doubler can only operate with coupled wavelengths . normally , if the sase fel operates at saturation , the quality of the electron bunch is too bad for the generation of sase radiation in a subsequent part of undulator , which is resonant at nearly the same wavelength .
we propose to use a method for activating or deactivating sase undulators , based on a rapid switching of the fel amplification process in order to obtain a user doubler @xcite . in the lcls baseline case
a betatron sase switcher @xcite may be practically advantageous .
the method is discussed in the following section [ increase ] .
the presence of the shab can be further exploited together with the self - seeding technique according to the scheme in fig .
[ lclsbm3 ] . in this setup , the first two segment of the first undulator part , before the chicane are deactivated to allow lasing in the linear regime .
we performed simulations and studied the properties of the 2nd harmonic of the output radiation , which is also monochromatic due to the implementation of the self - seeding technique .
the setup output will therefore consist of two monochromatic x - ray beams , at the first and at the second harmonic .
monochromaticity is the key for implementing multi - user operation in the hard x - ray range , which can be granted by using crystal deflectors and small absorption of the radiation at @xmath5 kev in crystals .
we suggest to flip crystals for switching of reflectivity very much like polarization - switching techniques with x - ray phase retarders at synchrotron radiation facilities , based on the use of piezoelectric components .
the method is discussed in section [ increase ] .
after section [ increase ] , in section [ feas ] and in section [ shab ] , we study feasibility with the help of simulations based on the code genesis @xcite .
finally , in section [ conc ] , we come to conclusions .
as discussed in the introduction , substituting a single undulator segment in the lcls baseline with a weak chicane yields a dramatic increase of the lcls capabilities , allowing for many different schemes to be realized without risks for the baseline mode of operation . in this section we discuss the principles of a method to increase user access in the full photon energy range , based on sase radiation , and to obtain a multi - user facility in hard x - ray range , based on self - seeded radiation .
the method to increase the user access in the full photon energy range , based on sase radiation , is sketched in fig .
[ lclsbm6]-[lclsbm7 ] , and is justified by the length of lcls baseline undulator , which is about two times longer than what is needed to reach saturation at @xmath6 nm wavelength @xcite . the basic idea is to exploit this extra - available length and to obtain two user simultaneous operation at lcls baseline .
the two sase x - ray beams , generated from two different parts of the baseline undulator are spatially separated with the help of x - ray mirrors .
the idea is sketched in fig .
[ lclsbm6 ] .
a sketch of the two x - ray mirrors installed within the magnetic chicane after the first ( 13 cells - long ) part of undulator is shown in fig .
[ lclsbm5 ] .
this insertable mirrors pair can be used to generate a deflection angle of @xmath7 mrad , which does not lead to interception of the fel beam with the vacuum chamber within the undulator , but creates a separation of about @xmath1 mm in the front end enclosure ( fee ) tunnelmm , the footprint on a single mirror at glancing angle of @xmath8 mrad would be about @xmath9 m , too long to be accomodated within the chicane .
the two mirrors in fig .
[ lclsbm5 ] are at a glancing angle of @xmath10 mrad .
this yields a footprint of just @xmath9 cm .
then , in order to introduce a deflection of @xmath7 mrad , one only needs to rotate one mirror with respect to the other of such small amount . ] .
fee mirror systems will subsequently direct the two spatially separated x - ray beams into two experimental stations .
change of the the active part of the undulator is based on rapid switching of the fel amplification process . for the lcls baseline undulator
, we propose to use a method to control the sase amplification process based on betatron switcher @xcite .
using a kicker installed upstream of the lcls baseline undulator and existing correctors located at the end of the first part of the undulator , it is possible to rapidly switch the x - ray beam from one user to another , thus providing two active beamlines at any time , fig .
[ lclsbm7 ] .
the location of the kicker must be chosen such that the betatron phase advance between kicker and corrector is @xmath11 , or a multiple thereof .
the fast kicker operates at a repetition rate of @xmath12 hz , yielding a kick of about 0.01 mrad to every second bunch .
such kick is compensated by using a corrector ( which always gives the same , static kick ) .
after the corrector , any bunch which was initially kicked goes into a straight section , and lases up to saturation in the second part of undulator . in the first part of undulator instead , the trajectory of such any bunch strongly deviates from the straight path , thus inhibiting the lasing process .
when the fast kicker is off , the bunch lases in the first part of the undulator up to saturation . as a result
, two sase beams can be generated from different parts of the undulator with repetition rate @xmath12 hz for each independent user . when the kicker is off , in the second undulator we have a perturbation of the orbit due to the corrector , but this is not important because in this case we already have saturation in the first undulator . in reference
@xcite , a setup with three or more undulators was considered . in our case , where we discuss only about two undulator parts , we deal with much simpler situation .
we also need to have a betatron phase shift of @xmath11 between kicker and corrector , but in our scheme this is not a constraint for the undulator lengths , because the switcher scheme can always be implemented with the proper choice of the kicker position upstream of the baseline undulator .
an idea for obtaining a user doubler at lcls ii was presented in @xcite .
there it is propose to tilt the second part of the undulator for obtaining spatial separation between the x - ray beams generated in the first and in the second part of the new soft x - ray undulator .
our idea to use an insertable mirrors pair within a short magnetic chicane for the same purpose , has the advantages that the first and the second part of the undulator can be kept straight , and that can therefore be applied to the lcls baseline .
obviously , such method can be applied to the lcls ii design as well . if the x - ray mirrors are substituted by a wake monochromator , the self - seeding scheme described in @xcite is applied , fig [ lclsbm1 ] . in this case
, there exists an efficient way for obtaining a multi - user facility at lcls baseline in hard x - ray range , based on self - seeded radiation . to this end
, we propose a photon beam distribution system based on the use of crystals in the bragg reflection geometry as deflectors .
it can be possible to deflect the full radiation pulse of an angle of the order of a radian without perturbations .
the proposed photon beam distribution system would allow to switch the hard x - ray beam quickly between @xmath1 experimental stations in the near and far experimental halls in order to make an efficient use of the lcls source .
the high monochromatization of the output radiation constitutes the key for reaching such results . in @xcite
we described self - seeding setup for lcls baseline . in its simplest configuration ,
a self - seeded fel consists of an input and an output undulator separated by a monochromator .
a distinguishing feature of our method is that it uses a single crystal in bragg transmission geometry as monochromator , installed within the magnetic chicane .
the method has the potential for a decrease of the bandwidth of the output radiation down to the fourier transform limit , given by the finite duration of the pulse . as discussed in the introduction , the lcls baseline undulator design with magnetic chicane proposed in fig .
[ lclsbm1 ] is very flexible and can be used for monochromatization purpose too , fig .
[ lclsbm3 ] .
successful operation of the self - seeding xfel requires the fulfillment of several requirements .
the first undulator part must operate in the deep linear regime , and not at saturation .
calculations show that in order not to spoil the electron beam quality , the number of cells in the first part of undulator should be decreased from @xmath13 down to @xmath14 .
this can be done simply by detuning @xmath10 cells , using the @xmath15-adjustable gap of the lcls baseline undulator . in section [ shab ]
we will demonstrate that for the undulator equipped with shab it is possible decrease the bandwidth of the second harmonoc radiation down to the transform limit as well .
the deflector can be simply switched off by tilting , which results in a change in the angle of incidence of the radiation .
the angular change necessary for switching on and off the reflected beam is less than @xmath16 mrad .
only @xmath17 of the incoming 2 nd harmonic radiation is absorbed in this case.,scaledwidth=100.0% ] dealing with monochromatic radiation one may use crystals as deflector , for both the first and the second harmonic , fig .
[ lclsbm9 ] and fig .
[ lclsbm10 ] .
the first harmonic deflector can be constituted of a silicon crystal , exploiting the 111 reflection in the symmetric bragg geometry .
the thickness of the first harmonic deflector is 0.05 mm .
the second harmonic deflector is diamond plate of 0.05 mm thickness .
it is used in bragg reflection geometry with the c(111 ) reflection plane .
one of the most advantages of crystal deflector in bragg geometry is that the reflected beam is switched only by changing angle .
the typical angular change necessary for the switching is less than @xmath16 mrad .
this opens up a new possibility of fast switching of the reflectivity , which is necessary for multi - user operation . in order to achieve a stable photon beam deflection the rotation error must be less than @xmath18 mrad .
existing technology enables rotating crystals to satisfy these requirements .
for example , at synchrotron radiation facilities , x - ray phase retarder crystals are driven by a piezoelectric device operated at hundred hz repetition rate , which flips the crystals with a rotation error of about a fraction of a microradian @xcite .
we propose to consider a multi - user distribution system based on ( flipping ) crystal deflectors , which will allow all existing user stations to work in parallel with the same high - quality x - ray beams , fig .
[ lclsbm8 ] . the initial second harmonic photon beam at 0.075 nm wavelength will thus be transformed into 5 different beams .
the switching crystals need to flip at a frequency of @xmath19 hz , so that each user ( except the first , which receives the full @xmath20 hz x - ray beam at the first harmonic ) receives 12 x - ray beams per second .
we first consider the scheme in fig .
[ lclsbm6 ] .
a pair of x - ray mirrors is inserted in the transverse offset generated by the magnetic chicane , as described in section [ increase ] .
this feasibility study is performed with the help of the fel code genesis 1.3 @xcite , running on a parallel machine , for the low - charge mode of operation ( @xmath8 nc ) corresponding to a bunch length of about @xmath1 fs .
parameters are presented in table [ tt1 ] .
.parameters for the low - charge mode of operation at lcls used in this paper . [ cols="<,^,^",options="header " , ] [ tt1 ] we study , separately , the output from the first undulator , from the second , and finally , from the shab , constituted by the last five undulator segments .
we first let the electron beam through the first undulator .
results are summarized in fig . [ us1rms ] ,
[ us1ene ] , fig .
[ us1pow ] and fig .
[ us1spec ] . fig .
[ us1spec ] shows a typical sase spectrum at 0.15 nm .
powers are in the @xmath21 gw - level , fig .
[ us1pow ] , with energies per pulse in the order of a few tens of microjoules . the poor longitudinal coherence of the sase pulse results in small energy fluctuations , as can be seen from fig .
[ us1rms ] .
[ us1rms ] also shows that the end of the linear regime is reached since , at it can be seen by inspection , the energy fluctuations drop .
in the proposed scheme for the x - ray beam doubler we assume a fresh bunch at the entrance of the second undulator .
we only account for an increase of the energy spread in the electron bunch due to quantum fluctuations during the pass through the first part of the undulator .
the output characteristics of the first harmonic for the second user are calculated at 0.15 nm for a 14 cells - long undulator .
the results , as can be expected , are similar as for the first user . the output characteristics , still calculated at @xmath6 nm , are shown in fig .
[ us2rms ] , fig
. [ us2ene ] , fig .
[ us2pow ] and fig .
[ us2spec ] . as for the first part of the undulator , fig .
[ us2spec ] shows a typical sase spectrum , far from the fourier limit .
powers are in the @xmath21 gw - level , fig .
[ us2pow ] , with energies per pulse in the order of a few tens of microjoules .
the poor longitudinal coherence of the sase pulse results in small energy fluctuations , as can be seen from fig .
[ us2rms ] also shows that the end of the linear regime is reached since , at it can be seen by inspection , the energy fluctuations drop .
the last five segments in the second undulator part are used as a shab , and tuned to the second harmonic of the fundamental .
the shab takes advantage of the density modulation at the second harmonic developed along the first undulator part .
the output characteristics , calculated at @xmath22 nm , are shown in fig .
[ shabrms ] , fig .
[ shabene ] , fig .
[ shabpow ] and fig .
[ shabspec ] . as before , fig .
[ shabspec ] shows a typical sase spectrum , far from the fourier limit .
powers are in the gw - level , fig .
[ shabpow ] , with energies per pulse in the order of ten microjoules .
the poor longitudinal coherence of the sase spectrum results in small energy fluctuations , as can be seen from fig .
[ shabrms ] .
as explained before , an alternative scheme to that considered in section [ feas ] consists in exploiting the self - seeding scheme .
this amounts to deactivating the first two segments in the first undulator part , in order to let the first undulator part to lase in the linear regime , so that the self - seeding scheme can be implemented .
the mirror setup is now substituted by a wake monochromator , see @xcite .
the second ( output undulator ) can be partly tapered in order to increase the output power .
the output power is optimized when the output undulator is tapered , segment by segment , according to the law in fig .
[ taplaw ] .
the output characteristics , calculated at @xmath6 nm , are shown in fig .
[ taprms ] , fig .
[ tapene ] , fig .
[ tappow ] and fig .
[ tapspec ] .
[ tapspec ] shows the effect of using the wake monochromator , which yields an nearly fourier limited spectrum .
powers are in the tens of gw - level , fig .
[ tappow ] , with energies per pulse in the order of hundreds microjoules .
the self - seeding scheme results in the typical fluctuation pattern shown in fig .
[ taprms ] .
as discussed in previous papers ( see , e.g. @xcite ) , at the beginning of the undulator the fluctuations of energy per pulse apparently drop . this can be explained considering the fact that the genesis output consists of the total power integrated over the full grid up to an artificial boundary , i.e. there is no spectral selection .
therefore , our calculations include a relatively large spontaneous emission background , which has a much larger spectral width with respect to the amplification bandwidth and which fluctuates negligibly from shot to shot .
since there is a long lethargy of the seeded radiation at the beginning of the fel amplifier , one observes an apparent decrease of fluctuations .
then , when lethargy ends , the seed pulse gets amplified and fluctuations effectively return to about the same level as after monochromator . .
grey lines refer to single shot realizations , the black line refers to the average over a hundred realizations.,scaledwidth=100.0% ] .,scaledwidth=100.0% ] finally , the output from the second - harmonic undulator is shown in fig .
[ 2habrms ] , fig .
[ 2habpow ] and fig .
[ 2habspec ] .
the spectrum still shows a good level of monochromatization , fig .
[ 2habspec ] , and the power level is still in the gw level , fig . [
2habpow ] .
the fluctuation pattern typical of self - seeding schemes is shown in fig . [ 2habrms ] .
in this paper we continue to exploit the flexibility of the magnetic chicane setup at lcls baseline proposed in @xcite discussing additional options for multiple user operations . the possibility to increase
user access is follows from the length of the lcls baseline undulator , which is about two times longer than what is needed to reach saturation at @xmath6 nm wavelength .
quite surprisingly , the user capacity can be increased with the help of an almost trivial setup , composed by two components : a short magnetic chicane , and an insertable x - ray mirrors pair for x - ray beam separation .
it can be straightforwardly installed in the baseline undulator and is safe , in the sense that it guarantees the baseline mode operation .
we propose to install the chicane at a special position , after @xmath13 cells . in this case
we keep all previous options and add , at the same time , the possibility of serving two users with sase radiation originating from two different parts of the undulator .
this paper also describes an efficient way to extend the lcls capabilities to @xmath1 users working in parallel in the hard x - ray wavelength range , both in the near and far experimental halls .
the high monochromatization of the output radiation constitutes the key for reaching such result and is achieved with self - seeding . in order to exploit the self - seeding option we simply activate @xmath14 cells out of @xmath13
thus , our scheme is based , as for beam doubler , on the introduction of a short magnetic chicane , which is extremely flexible and can be straightforwardly adopted for self - seeding with single crystal monochromator .
for the first time we simulated self - seeding with shab , and studied the properties of the second harmonic in this case .
we demonstrated that this method can produce intense monochromatic radiation both at the first and at the second harmonic .
we optimized the first harmonic power and demonstrated that even with a small number of cells it is still possible to reach 40 gw power . for the second harmonic we obtained a gw level with a narrow bandwidth 0.007 @xmath23 .
we propose to use the self - seeding mode of operation with shab for enabling multi - user mode of operation .
for this we propose to use rapidly and frequently switching crystal deflectors .
this is possible due to the relatively large photon energy ( 16 kev ) , resulting in a small absorption .
switching of deflectors will be provided by a small tilting of the crystal .
such rotation can be realized similarly as for phase retarders for polarization switching in synchrotron radiation facilities .
routine parameters of existing synchrotron radiation facilities ( e.g. petra iii ) allow for a 100 hz repetition rate for the rotation cycle , and a rotation error of @xmath24rad . as a result ,
in this paper we discuss about two separate options for multi - user operation .
the first deals with the possibility of two users in whole wavelength range .
the second allows for @xmath1 users , but for the hard x - ray range only .
the proposed photon beam distribution system would allow to switch the x - ray beam quickly between 6 experiment in order to make an efficient use of the lcls source .
we are grateful to massimo altarelli , reinhard brinkmann , serguei molodtsov and edgar weckert for their support and their interest during the compilation of this work .
j. wu , `` lcls - ii : an upgrade for the linac coherent light source '' proceeding of the 32nd fel conference , malmoe , sweden , 2010 , to be published .
slides available at http://fel2010.maxlab.lu.se/programme/presentations.html
|
although the lcls photon beam is meant for a single user , the baseline undulator is long enough to serve two users simultaneously . to this end
, we propose a setup composed of two simple elements : an x - ray mirrors pair for x - ray beam deflection , and a short ( @xmath0m - long ) magnetic chicane , which creates an offset for mirrors pair installation in the middle of the baseline undulator .
the insertable mirrors pair can be used for spatial separation of the x - ray beams generated in the first and in the second half of the baseline undulator .
the method of deactivating one half and activating another half of the undulator is based on the rapid switching of the fel amplification process . as proposed elsewhere , using a kicker installed upstream of the lcls baseline undulator and an already existing corrector in the first half of the undulator , it is possible to rapidly switch the x - ray beam from one user to another , thus providing two active beamlines at any time .
we present simulation results dealing with the lcls baseline , and show that it is possible to generate two saturated sase x - ray beams in the whole 0.8 - 8 kev photon energy range in the same baseline undulator .
these can be exploited to serve two users .
implementation of the proposed technique does not perturb the baseline mode of operation of the lcls undulator .
moreover , the magnetic chicane setup is very flexible , and can be used as a self - seeding setup too .
we present simulation results for the lcls baseline undulator with shab ( second harmonic afterburner ) and show that one can produce monochromatic radiation at the 2nd harmonic as well as at the 1st .
we describe an efficient way for obtaining multi - user operation at the lcls hard x - ray fel . to this end
, a photon beam distribution system based on the use of crystals in the bragg reflection geometry is proposed .
the reflectivity of crystal deflectors can be switched fast enough by flipping the crystals with piezoelectric devices similar to those for x - ray phase retarders at synchrotron radiation facilities .
it is then possible to distribute monochromatic hard x - rays among @xmath1 independent experiments , thereby enabling @xmath1 users to work in parallel in the near and far experimental halls .
* deutsches elektronen - synchrotron * * + * desy 10 - 165 september 2010 @xmath2 @xmath3 * way to increase the user access at the lcls baseline * @xmath4 gianluca geloni , _ + european xfel gmbh , hamburg _ vitali kocharyan and evgeni saldin _ + deutsches elektronen - synchrotron desy , hamburg _
@xmath3 issn 0418 - 9833 @xmath3 * notkestrasse 85 - 22607 hamburg *
|
droplet emulsions are key to many natural and industrial processes . in presence of an external flow ,
droplets undergo deformation , breakup , and coagulation . in a turbulent flow ,
breakup of droplets larger than the kolmogorov scale is governed by the interplay between surface tension and turbulent pressure fluctuations .
balancing surface tension and pressure fluctuations allows to estimate the minimum unstable droplet diameter that can undergoes breakup @xcite .
turbulent pressure - driven and surface - tension stresses across a droplet of typical diameter @xmath0 and with surface tension @xmath1 can be estimated as : the initially spherical droplet ( left panel ) get strechted ( central panel ) by switching on the stirring force and generation of a turbulent flow .
the righ panel shows the initial droplet immediately after its first breakup.,title="fig : " ] the initially spherical droplet ( left panel ) get strechted ( central panel ) by switching on the stirring force and generation of a turbulent flow .
the righ panel shows the initial droplet immediately after its first breakup.,title="fig : " ] the initially spherical droplet ( left panel ) get strechted ( central panel ) by switching on the stirring force and generation of a turbulent flow .
the righ panel shows the initial droplet immediately after its first breakup.,title="fig : " ] different phases in the time evolution of a large droplet in a turbulent flow ( @xmath2 ) .
initially the droplet deforms until it is broken it two droplets .
this event is rapidly followed by several other multiple fragmentation events until a stationary droplet dispersion is attained . in the final stages the droplets continuously undergo deformation , break - up , and coagulation . here
the time is non dimensionalized by the large eddy turnover time @xmath3 . ]
@xmath4 where with @xmath5 we indicate the density of the continous phase and with @xmath6 we indicate a typical velocity fluctuations across the droplet .
the ratio between the two gives the weber number estimated on the droplet diameter : @xmath7 the maximum stable _ mean _ droplet diameter can be estimated by assuming k41 turbulent statistics for velocity increments : @xmath8 where @xmath9 is the energy dissipation . from the condition of weber number order unity , one obtains : @xmath10 corrections to finite viscosity of the dispersed phase can also be added @xcite .
the present simulation allows to study droplets breakup conditions e.g. testing the kolmogorov - hinze criteria at changing the reynolds number and the viscosity ratio between the dispersed and continuos phase . specifically it is also possible to study the fluctuactions around the mean average stable droplet radius ( [ kh ] ) induced by intermittent fluctuations of the velocity field . in the movie various stages of droplet breakup process
are shown .
the turbulent flow is homogeneous and isotropic and it is numerically integrated by means of a multi - component lattice - boltzmann ( mclb ) simulation integrated in order to evolve consistently droplets and the advecting fluid @xcite .
the multicomponent flow is simulated using a standard shan - chen model @xcite , the turbulence is generated by applying a large scale forcing , and periodic boundary conditions are used .
the parameter used in our simulation are given in table [ table1 ] .
we start the simulation with the fluid at rest and all the droplet volume fraction residing into a large , single spherical droplet . from the beginning of simulation ,
the large scale stirring start to be applied and the turbulence level grows continuosly during roughly one large scale eddy turnover time . in the initial phase the large drop gets mildly deformed , then stretched , and finally broken into smaller droplets ( see figure [ gialla ] ) . at this point
the fragmentations process continues for about another @xmath11 eddy turn over times .
after this period the distribution of droplets is almost stationary and is characterized by continuos coagulation , deformation and breakups events due to the continuous presence of underlying turbulent fluctuactions .
for very long simulation one may experience some droplets evaporation .
the different phases in the time evolution of droplets are schematically illustrated in fig .
[ phases ] .
interesting questions concern the rate of collision and the nature of deformation and breakup events .
it is also of interest to quantify the evolution of the statistical properties of droplets diameter and the correlations between stresses on the droplet and break - up dynamics .
these results will be reported in a forthcoming article @xcite .
.[table1 ] parameters of the numerical simulation from which the movie was produced .
@xmath12 is the number of grid points along each direction , @xmath1 is the surface tension , @xmath13 is the taylor - microscale reynolds number , @xmath14 is the taylor - microscale , @xmath15 is the average fluid kinetic energy and @xmath16 is the average enstrophy .
@xmath17 is the initial droplet diameter , @xmath18 is the critical droplet diameter obtained from the theoretical estimate ( see text ) and @xmath19 is the critical droplet diameter as measured from our simulation .
the reynolds number was kept low on purpose in order to keep the dissipative scale relatively large : this allow us to have an interface thickness smaller than the viscous scale . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] 99 kolmogorov an .
nauk sssr . *
* 66 825 ( 1949 ) .
aiche j. * 1 * 289 ( 1955 ) .
calabrese rv et al .
aiche j. * 32 * 657 ( 1986 ) .
qian et al .
* 193 * 1038 - 1063 ( 2006 ) shan x , chen h. phys rev e * 47 * 1815 ( 1993 ) ; phys rev e * 49 * 2941 ( 1994 ) ;
shan x , yuan x .- f . , and chen h. j. fluid mech . *
550 * 413441 ( 2006 ) .
l. biferale , p. perlekar , m. sbragaglia and f. toschi . in preparation ( 2010 ) .
|
this fluid dynamics video shows the breakup of a droplet in a stationary homogeneous and isotropic turbulent flow .
we consider droplets with the same density of the transporting fluid .
the droplets and the fluid are numerically modelled by means of a multicomponent lattice - boltzmann method .
the turbulent fluid is maintained through a large scale stirring force and the radius of stable droplets , for the parameters in our simulation , is larger than the kolmogorov scale .
events of droplet deformation , break - up and aggregation are clearly visible from the movie . with the present database droplet evolution
can be studied from both an eulerian and lagrangian point of view .
the kolmogorov - hinze criteria for droplets break - up can be tested also by means of simulations with different viscosity contrast between the two components .
|
the two - photon processes may lead to production of two charged leptons and therefore compete with other sources of dileptons , such as continuum drell - yan processes or resonant production of vector quarkonia or @xmath3 boson , which produce dilepton pairs of large invariant masses .
earlier studies of lepton pair production via @xmath4 fusion in inelastic proton - proton collisions can be found in @xcite . for a general review of the @xmath5-fusion mechanism ,
see @xcite .
inelastic processes are also included in the monte - carlo generator lpair based on @xcite
. at high energies and small dilepton transverse momenta also semi - leptonic decays of pair - wise produced charmed @xmath6 mesons may be an important ingredient of dileptons @xcite .
actual contribution of different processes depends strongly on the details of experimental cuts .
the color singlet exchange of photons naturally leads to rapidity gaps .
if the rapidity veto on particles close to the @xmath7 vertex is imposed in addition one can enhance the relative contribution of the @xmath5 processes compared to the qcd drell - yan mechanism @xcite .
the invariant mass distribution of dileptons produced in the drell - yan processes can be calculated in collinear - factorization approach ( see e.g. the textbook @xcite ) .
if one wants to address more differential distributions , say in transverse momentum of the lepton pair , one can turn to @xmath8-space resummation @xcite or , especially in the small-@xmath9 kinematics , @xmath1-factorization ( see @xcite for instance ) . in this paper
we wish to concentrate on the photon - photon induced production of charged leptons .
realistic estimation of these processes requires more attention . in general
, there are three types of such processes which can be classified according to whether the proton remnants appearing `` after '' photon emission are just protons or baryon resonances or a complicated continuum ( see fig.[fig : diagrams ] ) . in principle , the elastic - elastic , processes with one elastic and one inelastic or double inelastic processes can be distinguished by detailed studies of the final state .
however , in practice this separation may be not easy and all of them should be considered . here
we wish to concentrate rather on inelastic - inelastic processes .
+ there are two approaches in the literature in this context . in one of the approaches one can treat photons as collinear partons in the proton .
the application of this approach requires presence of a hard scale ( e.g. a large photon virtuality or large lepton transverse momenta ) .
such photon partonic distributions were discussed in @xcite . in some of these approaches
the photon pdf enters the dglap evolution equations .
the treatment in @xcite is somewhat simplified , here only the @xmath10 splitting is taken into account .
below we wish to comment on the interrelation between the two approaches .
the photon pdf approach was applied in many phenomenological studies , e.g. to a number of photon - photon processes in @xcite , and to dilepton production in @xcite . in another approach one
parametrizes the @xmath11 vertices in terms of the proton s structure functions .
one can assume that the photons are either collinear or allow them to have transverse momenta and nonzero virtualities @xcite .
recently we used a slightly simplified approach @xcite , which takes advantage of the high - energy limit and is formulated in an analogous way as the @xmath1-factorization approach often used in the context of two - gluon processes . in this approach
one uses unintegrated photon distributions , in contrast to collinear distributions in the previous approach and off - shell matrix elements for the @xmath12 subprocess .
we shall use this approach also in the present paper .
the unintegrated photon distributions can be expressed in terms of the proton structure functions .
the structure functions were measured in some different corners of very rich phase space .
in particular they were studied in so - called deep - inelastic regime with large @xmath13 where perturbative treatment embedded in the dglap evolution equation applies . in this corner of the phase space
the structure functions are very well known . when going outside of the perturbative regime the situation is less clear .
several parametrizations were presented in the literature @xcite .
the applicability of the different parametrizations is limited and not well tested .
thus in two - photon processes one may need structure functions in very different corners of the @xmath14 space .
it is not clear a priori which regions are needed for particular experiments i.e. specific kinematical cuts .
we wish to discuss some examples related to particular past and modern experiments .
this paper is organized as follows : in section [ sec : fluxes ] , we briefly review the different formalism employed in our calculations .
we also discuss the different structure functions used as an input in the @xmath1-factorization approach . in section [ sec : results ] we show our numerical results of various dilepton distributions for the kinematics , and cuts , relevant for different experiments .
these are , at the presently highest available energies , atlas and cms , which measure central rapidities , and lhcb with coverage at forward rapidities .
we also discuss examples for the lower energies of rhic , as well as data taken in 1980 s at the isr at still lower energy .
we summarize our results in conclusions section .
production of lepton pairs at large transverse momenta is a hard process , to which standard arguments for factorization apply , and collinear factorization should be an appropriate starting point to calculate e.g. rapidity or transverse momentum spectra of leptons .
in fact , the dominant contribution to large - invariant mass dilepton pairs is of course the well known drell - yan process , but nothing prevents us from also including photon as partons along with quarks and gluons .
then the photon parton distribution , @xmath15 , of photons carrying a fraction @xmath16 of the proton s light - cone momentum , obeys the dglap equation , @xmath17 \nonumber \\ & & + p_{\gamma \leftarrow \gamma}(y ) \gamma\big({z
\over y},q^2\big ) \big \ } \ , .\end{aligned}\ ] ] in the complete set of dglap equations this photon density is then again coupled to the quark and antiquark distributions : @xmath18 due to the smallness of @xmath19 one would expect that the effect of photons on the quark and antiquark densities can be safely neglected , unless one is interested in high order perturbative corrections to the qcd splitting functions themselves .
accordingly , we find two different approaches to dglap photons in the literature . a first one , by glck et al .
@xcite asserts , that we can neglect the photon density on the right hand side of the evolution equations .
then , at sufficiently large virtuality @xmath20 , the photon parton density can be calculated from the collinear splitting of quarks and antiquarks @xmath21 .
@xmath22 \ ; . \label{eq : dortmund}\end{aligned}\ ] ] this equation is easily integrated , and gives the photon parton density as @xmath23 + \gamma(z , q_0 ^ 2 ) \nonumber \\ & = & { { { \alpha_{\mathrm{em}}}}\over 2 \pi } \int_{q_0 ^ 2}^{q^2 } { d \mu^2 \over \mu^2 } \int_z^1 { dx \over x } p_{\gamma \leftarrow q}\big({z \over x}\big ) { f_2(x,\mu^2 ) \over x } + \gamma(z , q_0 ^ 2 ) \ , .\end{aligned}\ ] ] one is left to specify from some model considerations the photon density at some low scale @xmath24 , but one may hope that at very large @xmath25 the part predicted perturbatively from quark and antiquark distributions dominates .
in addition to the above contribution from dglap splitting , glck et al . also add the weizscker - williams flux from the coherent emission @xmath26 without proton breakup as found in @xcite .
more recently , the durham @xcite and nnpdf @xcite groups have given a more involved treatment , in which the photon distribution is fully incorporated into the coupled dglap evolution equation . as usual with dglap evolution , the photon parton density at a starting scale @xmath24 needs to be specified . while @xcite present model approaches , in ref.@xcite an ambitious attempt to obtain @xmath24 from a fit to experimental data is found .
preliminary work by the cteq collaboration @xcite is also based on qed corrected dglap equations , and attempts to fit the photon distribution from the prompt photon production @xmath27 at hera where in part of the phase space the compton subprocess @xmath28 contributes .
it should be noted , that in the approach of @xcite , the input distribution @xmath24 contains the coherent or elastic contribution with an intact proton in the final state .
notice that due to the proton form factors the integral over virtualities in the elastic case quickly converges , and the elastic contribution is basically independent of @xmath20 , as soon as @xmath29 . in the collinear approach the photon - photon contribution to inclusive cross section for dilepton production can be written as : @xmath30 here @xmath31 above indices @xmath32 and @xmath33 denote @xmath34 , i.e. they correspond to elastic or inelastic components similarly as for the @xmath1-factorization discussed in section [ sec : fluxes ] below , see also the diagrams in fig.[fig : diagrams ] . the factorization scale is chosen as @xmath35 .
the elastic photon distributions can be calculated with the help of elastic electromagnetic proton form factors .
the functions @xmath36 are precisely the dglap evolved distributions of section [ subsec : dglap - photons ] above . in the mrst2004(qed ) approach the initial photon distribution is parametrized as @xcite : @xmath37 \cdot \frac{1+(1-y)^2}{y } \;.
\label{photon_initial}\ ] ] above @xmath38 and @xmath39 are valence - like distributions at the initial scale @xmath20 . in actual calculation
mrst2004(qed ) uses current quark masses which causes that the @xmath40 s and in the consequence also the initial photon distributions are artificially large ( the consequences for lepton production will be discussed when showing corresponding cross sections ) .
it would seem more reasonable to use rather constituent quark masses than the current ones .
we will show that this leads to large differences in photon distributions at finite running scales @xmath13 . before discussing results for cross sections for @xmath7 production
we wish to concentrate for a while on the collinear photon distributions . to illustrate the effect of the initial input in fig.[fig : collinear_photon_pdf ] we show both original mrst2004(qed ) photon distribution and similar result obtained by ignoring the initial input which , as discussed above , may be questionable .
the results are shown for different evolution scales @xmath41 , 10 , 100 , 1000 , 10000 gev@xmath42 .
we observe a sizable difference between resulting photon distributions obtained within the two approaches . because in calculating the cross section the photon distributions enter twice in the cross section formula , for first and second proton , respectively
, one can expect that the cross section obtained with the different pdfs may differ considerably .
we will return to this issue in the result section .
= 2 gev@xmath42 is set to zero , i.e. completely neglected .
, title="fig:",width=302 ] = 2 gev@xmath42 is set to zero , i.e. completely neglected . ,
in this approach we start from the feynman diagrams shown in fig.[fig : diagrams ] , and exploit the high - energy kinematics .
let the four - momenta of incoming protons be denoted @xmath43 . at high energies
the proton masses can be neglected , so that @xmath44 .
the photon - fusion production mechanism in leptonic and hadronic reactions is in great detail reviewed in @xcite , where also many original references can be found . in the most general form
, the invariant cross section is written as a convolution of density matrices of photons in the beam particles , and helicity amplitudes for the @xmath45 process . in a high energy limit , where dileptons carry only a small fraction of the total center - of - mass energy , the density - matrix structure can be very much simplified , and there emerges a @xmath1-factorization representation of the cross section @xcite .
the unintegrated photon fluxes introduced in @xcite can be expressed in terms of the hadronic tensor as @xmath46 these unintegrated fluxes enter the cross section for dilepton production as @xmath47 where the indices @xmath48 denote elastic or inelastic final states .
the longitudinal momentum fractions of photons are obtained from the rapidities and transverse momenta of final state leptons as : @xmath49 the explicit form of the off - shell cross section @xmath50 can be found in ref .
@xcite .
we now first concentrate on inelastic processes with breakup of a proton .
then the hadronic tensor is expressed in terms of the electromagnetic currents as : @xmath51 and its elements can be measured in inclusive electron scattering off the target .
we wish to express it in terms of the virtual photoabsorption cross section of transverse and longitudinal photons . to this end
we introduce the covariant vectors / tensors @xmath52 and @xmath53 here @xmath54 projects on photons carrying helicity @xmath55 in the @xmath56-cms frame , and @xmath57 plays the role of the polarization vector of the longitudinal photon .
notice that @xmath58 , so that the hadronic tensor has the convenient gauge invariant decomposition @xmath59 the virtual photoabsorption cross sections are defined as @xmath60 it is customary to introduce dimensionless structure function @xmath61 as @xmath62 where @xmath63 then , our structure functions @xmath64 are expressed through the more conventional @xmath65 as @xmath66 in the literature one often finds rather @xmath67 structure functions , which are related to @xmath65 through @xmath68 now , performing the contraction with @xmath69 , we get @xmath70 in the deep inelastic region @xmath71 , and using @xmath72 in the second term , we can write more succinctly @xmath73 with @xmath74 \ , .\end{aligned}\ ] ] let us now isolate the elastic contribution to the hadronic tensor , which we need to describe the photon flux in processes in which the proton stays intact . in this case , the structure functions @xmath64 are most conveniently written in terms of the electric and magnetic form factors @xmath75 and @xmath76 of the proton : @xmath77 the contribution to the photon flux is then again obtained by contracting @xmath78 \nonumber \\\end{aligned}\ ] ] let us now give explicit formulas for the unintegrated fluxes in a form which makes it easy to compare them for example with fluxes of virtual photons given by budnev et al . @xcite .
the quantity to compare is the differential equivalent photon spectrum @xmath79 the fluxes in @xcite are given differentially in the virtuality @xmath13 , instead of the transverse momentum @xmath80 .
we therefore substitute @xmath81 so that we obtain @xmath82 \ ; dm_x^2\end{aligned}\ ] ] and for the elastic piece @xmath83 \ ; . \nonumber
\\\end{aligned}\ ] ] it is also interesting to convert the integration over @xmath84 into one over @xmath85 . to this end , we note that @xmath86 furthermore @xmath87 then we obtain for the photon flux @xmath88 .\end{aligned}\ ] ] in the deep inelastic limit , @xmath89 , and assuming @xmath90 , this obtains the form @xmath91 with the splitting function @xmath92 the `` parton densities of photons '' , which can be compared to the collinear factorization fluxes are @xmath93 here we show a few different parametrizations of the proton structure function @xmath94 .
the different parametrizations taken from the literature are labeled as : * allm @xcite .
this parametrization gives a very good fit to @xmath94 in most of the measured region .
* fjllm @xcite .
this parametrization explicitly includes the nucleon resonances and gives an excellent fit of the clas data .
* bdh @xcite .
this parametrization concentrates on the low-@xmath9 , or high mass region .
it features a froissart - like behaviour at very small @xmath9 . *
sy @xcite .
this paramerization of suri and yennie from the early 1970 s does not include qcd - dglap evolution .
it is still today often used as one of the defaults in the lpair event generator .
* su @xcite .
a parametrization which concentrates to give a good description at smallish and intermediate @xmath13 at not too small @xmath9 .
we also show @xmath94 calculated from the cteq6l parametrization @xcite . in fig.[fig
: f2 ] we show the proton structure function @xmath95 obtained from the various fits at @xmath96 as a function of bjorken-@xmath9 in fig .
[ fig : f2 ] . in fig .
[ fig : f2_lowx ] , we show the structure function @xmath95 at @xmath97 , but this time with a logarithmic abscissa to emphasize the low-@xmath9 behaviour of different parametrizations . also shown
are the hera data at low-@xmath9 .
experimental data on the figures are taken from the compilation @xcite and from @xcite .
here we see that the suri - yennie fit corresponds to a unit - intercept pomeron and does not describe the small-@xmath9 rise of the proton structure function .
a surprising lesson is , that the old suri - yennie @xcite fit , still gives a reasonable description of @xmath94 except of very small @xmath9 . for explicit account of resonances
it would be recommended to use the fiore et al .
@xcite , but care has to be taken to stay within the resonance region , as the quality of the fit beyond this region quickly deteriorates .
the overall best description appears to be given by the allm @xcite fit .
in contrast to our previous studies @xcite , all calculations performed within the @xmath1-factorization approach were performed with a monte carlo event generator , where the formulae presented above ( see also @xcite ) were implemented .
this monte carlo program is used to generate events ( four - momenta of leptons and outgoing protons / excited systems ) which are then transformed to distributions with the help of the standard software root @xcite .
the typical number of events generated in our studies is a few millions .
a more detailed description of the event generator will be presented elsewhere @xcite .
most of the experiments for the dilepton production concentrate on determination of dilepton invariant mass distributions . in fig.[fig : dsig_dmll_ine_ine ] we show invariant mass distributions of dilepton pairs produced in the photon - photon inelastic - inelastic mechanism for kinematical conditions relevant for different experiments . we show results obtained with different parametrizations of the structure functions known from the literature . surprisingly the different structure functions give quite different results .
for completeness in some cases ( when possible ) we also show the result obtained in the collinear approach with the mrst2004(qed ) photon distribution with ( solid black line ) and similar one when ignoring the initial input ( long - dashed black line ) .
the result obtained within the collinear approach with the mrst2004(qed ) distribution is much above the results obtained within the @xmath1-factorization approach . in our opinion this is mainly related to the large input photon distribution at the initial scale @xmath20 = 2 gev@xmath42 ( see eq.([photon_initial ] ) ) discussed in the context of fig.[fig : collinear_photon_pdf ] .
if the input is discarded ( long - dashed black line ) the collinear result is similar to the results obtained within the @xmath1-factorization .
the inelastic - inelastic contribution gives only a small fraction of the measured cross section for most experimental conditions ( atlas , lhcb , phenix ) .
for the isr experiment it is relatively larger . in fig.[fig :
dsig_dmll_ela_ine ] we show dilepton invariant mass distributions for elastic - inelastic and inelastic - elastic ( added together ) contributions . as for inelastic - inelastic contribution
the results strongly depend on the parametrization of the structure functions used .
the spread of results for different @xmath94 from the literature is now , however , significantly smaller than in the case of inelastic - inelastic contributions where the structure functions enter twice ( into both photon flux factors ) . as for the double inelastic case
we also show a result for the collinear approach .
the mixed components give similar contribution to the dilepton invariant mass distributions as the inelastic - inelastic one .
it is very interesting to understand which regions of ( @xmath98 ) and ( @xmath99 ) space contribute in the measured spectra .
we start our review from distributions in @xmath100 ( or @xmath101 ) .
the corresponding results for the inelastic - inelastic component are shown in fig.[fig : dsig_dmx_ine_ine ] .
again we show results for the isr ( left top panel ) , phenix ( right top panel ) atlas ( left bottom panel ) and lhcb ( right bottom panel ) experiment .
the dominant contributions come from the region of very small missing masses @xmath100 ( or @xmath101 ) .
this is not necessarily the region where standard evolution equation applies for the description of @xmath94 structure function . in general , the fiore et al .
@xcite and suri - yennie @xcite parametrization give much bigger cross section in the region of small missing masses .
in this plot the resolution in missing mass is rather coarse ( @xmath102 = 2.5 gev ) .
if the resolution of the distribution ( binning ) was improved one could observe even peaks corresponding to nucleon resonances excited by virtual photons for the fiore et al .
parametrization . as seen in fig.[fig : f2 ] the suri - yennie parametrization extremely well averages the structures in more detailed fiore et al .
clearly the fiore et al .
parametrization is not adequate for large @xmath100 ( @xmath101 ) masses .
all this demonstrates how important is using a `` proper '' structure function . in fig.[fig :
dsig_dmxdmy ] we show some examples of two - dimensional distributions @xmath103 for different parametrizations of the structure functions as an example for the phenix kinematics . here
we focus on small values of @xmath100 and @xmath101 to resolve apparent differences .
clearly the different parametrizations give very different results . in the case of fiore
parametrization one can observe now ( with better resolution ) resonance lines for @xmath100 or @xmath101 slightly bigger than 1 gev . for different structure functions : szczurek - uleshchenko ( upper - left ) , fiore et al .
( upper - right ) , suri - yennie ( lower - right ) and allm ( lower - left ) for @xmath104 = 200 gev and @xmath105 0.2 gev . , scaledwidth=100.0% ] for different structure functions : szczurek - uleshchenko ( upper - left ) , fiore et al .
( upper - right ) , suri - yennie ( lower - right ) and allm ( lower - left ) for @xmath104 = 200 gev and @xmath105 0.2 gev . ,
scaledwidth=100.0% ] for different structure functions : szczurek - uleshchenko ( upper - left ) , fiore et al .
( upper - right ) , suri - yennie ( lower - right ) and allm ( lower - left ) for @xmath104 = 200 gev and @xmath105 0.2 gev . , scaledwidth=100.0% ] for different structure functions : szczurek - uleshchenko ( upper - left ) , fiore et al .
( upper - right ) , suri - yennie ( lower - right ) and allm ( lower - left ) for @xmath104 = 200 gev and @xmath105 0.2 gev . ,
scaledwidth=100.0% ] in fig.[fig : dsig_dq12dq22 ] we show two - dimensional distributions @xmath106 for four different experimental conditions specified in the figure caption . in most of the cases rather large
photon virtualities contribute .
this is especially true for the atlas experiment with large cuts on lepton transverse momenta @xcite . in the case of the old isr experiment @xcite or more recent phenix experiment
@xcite the situation is very different and clearly contributions from @xmath94 nonperturbative regions come into game and should be carefully analyzed .
for the case of lhcb , in contrast to other cases , the distribution in @xmath107 is not symmetric along the diagonal which is related to asymmetric forward coverage of the lhcb experiment . for different experiments : isr ( upper - left ) , phenix ( upper - right ) , atlas ( lower - right ) and lhcb ( lower - left ) for allm structure function . ,
scaledwidth=100.0% ] for different experiments : isr ( upper - left ) , phenix ( upper - right ) , atlas ( lower - right ) and lhcb ( lower - left ) for allm structure function .
, scaledwidth=100.0% ] for different experiments : isr ( upper - left ) , phenix ( upper - right ) , atlas ( lower - right ) and lhcb ( lower - left ) for allm structure function .
, scaledwidth=100.0% ] for different experiments : isr ( upper - left ) , phenix ( upper - right ) , atlas ( lower - right ) and lhcb ( lower - left ) for allm structure function . , scaledwidth=100.0% ] summarizing this part , we have shown that with typical experimental cuts the contribution of photon - photon fusion is much smaller than dilepton experimental data and constitutes typically less than 1 % of the measured cross sections . in most of the cases considered so far
drell - yan processes dominate @xcite . however , the two - photon processes are interesting by themselves .
can they be measured experimentally ? in order to reduce the drell - yan contribution and relatively enhance the two - photon contribution one can impose an extra condition on lepton isolation .
first trials have been already done by the cms collaboration @xcite . in their analysis
an extra lepton isolation cuts were imposed in order to eliminate the otherwise dominating drell - yan component . in figs .
[ fig : cms_dn_dmll],[fig : cms_dn_dptsum],[fig : cms_dn_dphi ] we show our results for two different ( sy and allm ) parametrizations of the structure functions for distributions in dimuon invariant mass , in transverse momentum of the pair and in relative azimuthal angle between @xmath108 .
only invariant mass distribution can be obtained in the collinear approach . in the collinear approach
the second and the third distributions are just dirac delta functions in @xmath109 and @xmath110 , respectively . sy and allm parametrizations give almost the same contributions to all the distributions considered . in the first evaluation we have taken into account integrated luminosity of the experiment ( @xmath111 = 63.2 pb@xmath112 ) as well as experimental acceptances given in table 5 in ref.@xcite .
rather good agreement with the low statistics cms experimental data is achieved ( for both parametrizations of structure functions used in the figures ) without including any extra corrections due to absorption effects leading to destroying the rapidity isolation of leptons and a damping of corresponding cross section for the photon - photon mechanisms .
this result is interesting by itself .
it may mean that the absorption effects are small or alternatively that a contamination of the drell - yan contribution is still not completely removed .
both effects should be therefore studied in more detail in a future .
this can be done by full monte carlo simulations of both processes and clearly goes beyond the scope of the present analysis . for completeness and comparison in fig.[fig : cms_dn_dmll_collinear ] we show invariant mass distribution obtained within collinear factorization approach with @xmath113 .
we present results for the case when initial input at @xmath20 = gev@xmath42 ( see eq.([photon_initial ] ) ) is included ( thick red lines ) as well as when it is discarded ( thin blue lines ) as discussed in subsection [ subsec : dglap - photons ] .
the results obtained in the letter case are slightly larger than those obtained within the @xmath1-factorization approach ( see fig.[fig : cms_dn_dmll ] ) , especially when the mrst(qed ) input is included .
in the present paper we have discussed in detail production of dilepton pairs ( @xmath2 or @xmath108 ) in photon - photon processes in proton - proton scattering at high energies .
we have compared two different distinct theoretical approaches . in the first approach photon
is treated as a collinear parton in the proton and included into generalized ( qcd , qed ) dglap equations .
we have discussed and demonstrated that it is not necessary to keep photon distribution in the evolution equation .
it is sufficient to couple photon to other partons ( quarks / antiquarks ) in the proton , that undergo usual dglap evolution equations .
we have discussed also the issue of initial condition for the photon distribution at the initial scale . in this context
we have discussed parametrization / prescription proposed by mrst04(qed ) @xcite with their initial input as well as when starting evolution from zero input .
the two prescriptions lead to quite different results for photon distributions and in the consequence also for charged lepton observables for finite scales . in the second approach
we take into account the fact that photons are off shell and include their transverse momenta and/or virtualities .
we have shown that for typical kinematical conditions of modern experiments , especially at the lhc , the photon virtualities are fairly large , which puts doubts on the standard ( collinear ) parton model treatment .
the @xmath1-factorization approach uses unintegrated photon distributions which are expressed in terms of @xmath94 structure functions @xcite .
different model parametrizations known from the literature have been used in the present study .
the final results depend strongly on the choice of the parametrization .
we have identified regions of the @xmath114 space which give significant contribution to the cross section for different experimental conditions .
for example for the experimental cuts of the recent atlas experiment @xcite mostly perturbative region ( @xmath115 4 gev and @xmath116 3 gev ) contributes .
therefore a reliable predictions with accuracy better than 10 % are possible . in contrast , for the old isr @xcite and more recent phenix @xcite experiments substantial contributions come from the regions @xmath117 3 gev and @xmath118 1 gev . in this case
one should use explicit parametrizations which fit experimental data in this corner of the space . the calculation should take into account also resonance contributions . in the present paper
we have discussed production of dileptons .
similar analysis may be repeated e.g. for photon - photon induced production of @xmath119 pairs .
so far only the first approach was applied there @xcite .
we would like to thank juan rojo for remarks that initiated the present study , and laurent forthomme for help with a monte carlo code .
this study was partially supported by the polish national science centre grants dec-2013/09/d / st2/03724 and dec-2014/15/b / st2/02528 .
g. g. da silveira , l. forthomme , k. piotrzkowski , w. schfer and a. szczurek , `` central @xmath123 production via photon - photon fusion in proton - proton collisions with proton dissociation , '' jhep * 1502 * , 159 ( 2015 ) [ arxiv:1409.1541 [ hep - ph ] ] .
v. m. budnev , i. f. ginzburg , g. v. meledin and v. g. serbo , `` the two photon particle production mechanism . physical problems . applications .
equivalent photon approximation , '' phys .
rept . * 15 * , 181 ( 1975 ) .
s. chatrchyan _ et al . _
[ cms collaboration ] , `` search for exclusive or semi - exclusive photon pair production and observation of exclusive and semi - exclusive electron pair production in @xmath124 collisions at @xmath125 tev , '' jhep * 1211 * , 080 ( 2012 ) [ arxiv:1209.1666 [ hep - ex ] ] .
r. k. ellis , w. j. stirling and b. r. webber , `` qcd and collider physics , '' camb .
monogr . part .
cosmol .
* 8 * ( 1996 ) 1 .
m. a. nefedov , n. n. nikolaev and v. a. saleev , `` drell - yan lepton pair production at high energies in the parton reggeization approach , '' phys .
d * 87 * ( 2013 ) 1 , 014022 [ arxiv:1211.5539 [ hep - ph ] ] .
s. p. baranov , a. v. lipatov and n. p. zotov , `` drell - yan lepton pair production at the lhc and transverse momentum dependent quark densities of the proton , '' phys .
d * 89 * ( 2014 ) 9 , 094025 [ arxiv:1402.5496 [ hep - ph ] ] .
h. abramowicz and a. levy , `` the allm parameterization of sigma(tot)(gamma * p ) : an update , '' hep - ph/9712415 .
r. fiore , a. flachi , l. l. jenkovszky , a. i. lengyel and v. k. magas , `` explicit model realizing parton hadron duality , '' eur .
phys . j. a * 15 * , 505 ( 2002 ) [ hep - ph/0206027 ] . m. m. block , l. durand and p. ha , `` connection of the virtual @xmath127 cross section of @xmath128 deep inelastic scattering to real @xmath129 scattering , and the implications for @xmath130 and @xmath128 total cross sections , '' phys . rev .
d * 89 * , no .
9 , 094027 ( 2014 ) [ arxiv:1404.4530 [ hep - ph ] ] .
j. pumplin , d. r. stump , j. huston , h. l. lai , p. m. nadolsky and w. k. tung , `` new generation of parton distributions with uncertainties from global qcd analysis , '' jhep * 0207 * ( 2002 ) 012 [ hep - ph/0201195 ] .
m. osipenko _ et al . _
[ clas collaboration ] , `` a kinematically complete measurement of the proton structure function f(2 ) in the resonance region and evaluation of its moments , '' phys .
d * 67 * ( 2003 ) 092001 [ hep - ph/0301204 ] .
a. adare _ et al .
_ [ phenix collaboration ] , `` dilepton mass spectra in p+p collisions at s**(1/2 ) = 200-gev and the contribution from open charm , '' , phys .
b * 670 * ( 2009 ) 313 [ arxiv:0802.0050 [ hep - ex ] ] .
[ the lhcb collaboration ] , `` inclusive low mass drell - yan production in the forward region at sqrt(s)=7 tev , '' lhcb - conf-2012 - 013 ; conference report prepared for xx international workshop on deep - inelastic scattering and related subjects , 26 - 30 , march 2012 , bonn , germany .
g. aad _ et al . _ [ atlas collaboration ] , `` measurement of the low - mass drell - yan differential cross section at @xmath104 = 7 tev using the atlas detector , '' jhep * 1406 * , 112 ( 2014 ) [ arxiv:1404.1212 [ hep - ex ] ] .
g. aad _ et al .
_ [ atlas collaboration ] , `` measurement of the high - mass drell yan differential cross - section in pp collisions at sqrt(s)=7 tev with the atlas detector , '' phys .
b * 725 * , 223 ( 2013 ) [ arxiv:1305.4192 [ hep - ex ] ] .
v. khachatryan _ et al . _
[ cms collaboration ] , `` measurements of differential and double - differential drell - yan cross sections in proton - proton collisions at 8 tev , '' eur .
j. c * 75 * , no . 4 , 147 ( 2015 ) [ arxiv:1412.1115 [ hep - ex ] ] .
c. kourkoumelis , l. k. resvanis , t. a. filippas , e. fokitis , a. m. cnops , j. h. cobb , r. hogue and s. iwata _ et al .
_ , `` study of massive electron pair production at the cern intersecting storage rings , '' phys .
b * 91 * ( 1980 ) 475 .
s. chatrchyan _ et al . _
[ cms collaboration ] , `` search for exclusive or semi - exclusive photon pair production and observation of exclusive and semi - exclusive electron pair production in @xmath124 collisions at @xmath125 tev , '' jhep * 1211 * , 080 ( 2012 ) [ arxiv:1209.1666 [ hep - ex ] ] .
|
we investigate different methods to incorporate the effect of photons in hard processes .
we compare two different approaches used for calculating cross sections for two - photon @xmath0 process . in one of the approaches photon
is treated as a collinear parton in the proton . in the second approach
recently proposed a @xmath1-factorization method is used .
we discuss how results of the collinear parton model depend on the initial condition for the qcd evolution and discuss an approximate treatment where photon is excluded from the combined qcd - qed evolution .
we demonstrate that it is not necessary to put photon into the evolution equation as often done recently but it is sufficient to use a simplified approach in which photon couples to quarks and antiquarks which by themselves undergo dglap evolution equations .
we discuss sensitivity of the results to the choice of structure function parametrization and experimental cuts in the @xmath1-factorization approach . a new optimal structure function parametrization is proposed .
we compare results of our calculations with recent experimental data for dilepton production and find that in most cases the contribution of the photon - photon mechanism is rather small .
we discuss how to enhance the photon - photon contribution .
we also compare our results to those of recent measurements of exclusive and semi - exclusive @xmath2 pair production with certain experimental data by the cms collaboration .
|
magnetoresistance has long been used as a tool to study the fermi surface of metals @xcite .
the observation of quantum oscillations provides a measure of the cross - sectional area of closed pockets ( due to the quantization of the landau levels ) .
the background magnetoresistance also contains information about all the electronic orbits induced by the magnetic field .
for example , a transverse magnetoresistance saturating in high field can be evidence for the presence of closed orbits , while a magnetoresistance that increases quadratically with field can indicate open orbits .
a number of effects can complicate this picture and one such is magnetic breakdown , a concept first introduced to explain a giant orbit observed in magnesium @xcite . because of the periodic potential in a crystal , small gaps in the dispersion sometimes open up at the brillouin zone edge , splitting the fermi surface into distinct sections . in low magnetic fields ,
electrons traverse the fermi surface under the action of the magnetic field and bragg scatter at the brillouin zone edge . in a sufficiently strong magnetic field it is possible for an electron to tunnel from one segment of the fermi surface to another @xcite .
this is because in high magnetic field , mixing of states between different pieces of the fermi surface can increase the likelihood of magnetic breakdown , so that the resulting orbit more closely resembles that which would be obtained in the absence of the periodic potential @xcite .
the presence of magnetic breakdown presents a challenge for the simulations of magnetoresistance .
this is because at the points on the brillouin zone edge intersected by the fermi surface ( points which we call magnetic breakdown junctions ) there is a probability @xmath0 for magnetic breakdown to occur and this is given by @xmath1 , where @xmath2 is the characteristic breakdown field ; the problem is that under usual experimental conditions @xmath0 is _ between _ 0 and 1 .
thus there is a hierarchy of complex trajectories that must be summed to account for all possible contributions to the conductivity in which magnetic breakdown either does or does not occur at each magnetic breakdown junction .
if a quasiparticle crosses @xmath3 magnetic breakdown junctions , one has to consider @xmath4 possible trajectories with their correct probabilistic weightings , and this complicates a direct computation of magnetoresistance since one has to sum over trajectories with arbitrarily long path lengths and hence arbitarily large values of @xmath3 . in order to calculate how magnetic breakdown can change the connectivity of the orbits @xcite , and
thus the resistivity , falicov and sievert in 1965 provided a model @xcite that generalised the chambers path integral method @xcite to include magnetic breakdown at a finite number of points .
this was used to evaluate the magnetoresistance in a number of model fermi surfaces , although the calculations were performed numerically .
more recently , the problem of the effect of magnetic breakdown on the magnetoresistance in a network of coupled orbits was treated analytically @xcite , though the focus of this work was on interlayer transport .
this is because the newer calculations are motivated by recent experiments on quasi - two - dimensional organic metals @xcite in which magnetotransport experiments have been particularly illuminating @xcite . in this paper
we use the techniques developed in ref .
( see also ref . ) to derive an analytical expression for the in - plane components of the resistivity tensor in the falicov - sievert model .
this expression allows the underlying physics of the magnetotransport to be more clearly extracted and will allow a more convenient comparison with experimental data
. -@xmath5 plane showing the points where magnetic breakdown can occur which are at @xmath6 ( these are called magnetic breakdown junctions ) .
the azimuthal coordinate of a quasiparticle at the point labelled qp is @xmath7 . (
b ) an open orbit ( comprising the repeated traversal of the @xmath8 section ) .
( c ) closed orbit ( the @xmath9-orbit , comprising the repeated traversal of @xmath10 and @xmath11 ) .
( d ) breakdown orbit ( the @xmath12-orbit , comprising @xmath10-@xmath8-@xmath11-@xmath13 ) .
, width=340 ]
in this paper , we consider the quasi - two - dimensional fermi surface shown in fig .
[ fig : fs](a ) in which the dispersion is given by @xmath14 , where @xmath15 is the effective mass , @xmath16 is the fermi wave vector , @xmath17 is the interlayer spacing and the interlayer hopping @xmath18 is small ( @xmath19 ) .
the fermi surface consists of a cylinder with volume @xmath20 and thus the number density of electrons is @xmath21 .
the brillouin zone edges are placed at @xmath22 .
quasiparticles orbit around the fermi surface with constant @xmath23 when @xmath24 lies along the interlayer direction . in very low @xmath24 , because of bragg reflection , only open orbits [ fig . [ fig : fs](b ) ] and small closed orbits [ the @xmath9 orbit , fig .
[ fig : fs](c ) ] occur around the distinct sections of the fermi surface . in high @xmath24 , mixing between the states on the two fermi surface sections leads to magnetic breakdown at the four magnetic breakdown junctions . at these junctions a quasiparticle `` tunnels '' in @xmath25-space between the fermi surface sections , resulting in a single large closed orbit [ the @xmath12 orbit , fig .
[ fig : fs](d ) ] .
this model is equivalent to the situation outlined in fig . 1 of ref . but
the notation has been chosen to link with the physical situation relevant to a particular family of organic metals @xcite .
in this section we review the standard calculation of magnetoresistance in the case of full magnetic breakdown ( i.e. @xmath26 ) which is equivalent to assuming a very high magnetic field and ignoring any bragg scattering . in this case
we consider only the effect of a single closed breakdown orbit [ i.e. the case of fig .
[ fig : fs](d ) ] .
the boltzmann equation gives the conductivity tensor @xmath27 as an average of velocity correlations over the fermi surface , where @xmath28 is given by @xmath29\,{\rm d}t,\ ] ] and @xmath30 is the density of states .
the quasiparticle orbits lie in a plane perpendicular to the magnetic field @xmath31 and orbit the fermi surface with angular frequency @xmath32 given by @xmath33 if @xmath31 is perpendicular to the quasi - two - dimensional planes .
writing @xmath34 , we have that @xmath35 and so the conductivity tensor can be written as @xmath36 where @xmath37 and @xmath38 neglecting the small effect of the interplane warping @xmath18 on orbit size , and considering only the in - plane components of the conductivity tensor @xmath39 , we proceed as follows . setting @xmath40 and @xmath41 , and assuming full magnetic breakdown ( so all the orbits just go round the cylinder with no bragg scattering ) , the integral
can be evaluated to give @xmath42 and inverting this tensor yields the resistivity tensor @xmath43 or equivalently @xmath44 this result is exactly what one expects from the simple theory of metals : no magnetoresistance ( because @xmath45 is not field - dependent ) and a simple hall effect @xmath46 where @xmath47 . for later reference
, we will denote the simple form of the resistivity tensor with no magnetic breakdown in eq .
( [ eq : simplerho ] ) as @xmath48 .
a treatment fully including both magnetic breakdown and bragg scattering can be performed using the method described in ref . .
setting @xmath49 with @xmath50 and @xmath51 , eq .
( [ eq : sigma1 ] ) becomes @xmath52 and with the definitions of the functions @xmath53 @xmath54 the conductivity can be written @xmath55 what makes eq .
( [ eformula ] ) challenging to evaluate is that the integrand changes depending on the path taken by the quasiparticle which , at each magnetic breakdown junction of the orbit , can either undergo magnetic breakdown tunneling ( with probability @xmath56 ) or bragg reflection ( with probability @xmath57)@xcite .
the strategy for solving this problem follows the earlier approach @xcite of separating the motion of electrons in the four different segments of the orbit and constructing recursive equations which encode all the information about the behaviour at the magnetic breakdown junctions . in this way we can write the conductivity tensor as @xmath58 where the vectors @xmath59 and @xmath60 are given by @xmath61 and @xmath62 and the matrix @xmath63 which encodes all the breakdown probabilities and scattering
is given by @xmath64 where @xmath65 and @xmath66 .
the integrals can be performed analytically and after much simplification ( see appendix ) the conductivity can be obtained as @xmath67 where the parameter @xmath68 is given by @xmath69 and the matrix @xmath70 is given by @xmath71 equation ( [ conduc ] ) demonstrates that the conductivity is a sum of the familiar expression assuming no bragg scattering [ i.e. eq . ( [ eq : simplesigma ] ) ] and an additional term . this additional term is included by an amount controlled by @xmath68 , and it is therefore useful to plot the field dependence of this quantity .
this is shown in fig .
[ fig : delta](a ) for the case in which @xmath72 ( chosen to match fig . 1 of ref . ) for a range of different breakdown fields ( parametrized by @xmath73 ) and plotted as a function of magnetic field ( parametrized by @xmath74 ) .
( [ eq : delta ] ) ] , ( b ) @xmath75 [ eq . ( [ eq:2 ] ) ] and ( c ) @xmath76 [ eq . ( [ eq:3 ] ) ] . ,
width=340 ] for low field @xmath77 where @xmath78 and is independent of breakdown field @xmath79 . for intermediate fields , and for @xmath80
, @xmath68 rises to a plateau given by @xmath81 = \alpha/ ( \pi-2\xi)$ ] . in our example in which @xmath72 ,
the plateau is at @xmath82 [ see fig . [
fig : delta](a ) ] . for large fields ,
@xmath68 decreases again and follows @xmath83 .
thus we expect that at very low and very high fields the approximation of no bragg scattering ( the single breakdown orbit ) in eq .
( [ eq : simplesigma ] ) will work well .
it works at very low fields because electrons travel only a very short distance around the fermi surface before scattering and so orbit connectivity is largely irrelevant .
it works at very high fields because then the magnetic breakdown probability is close to unity . at intermediate fields @xmath68 is significant and
the effect of magnetic breakdown and bragg scattering is important in determining the magnetotransport . )
calculated according to eqs .
( [ eq:1][eq:3 ] ) for the case of @xmath72 and @xmath84 .
, width=340 ] we now invert the conductivity tensor in eq .
( [ conduc ] ) to obtain an expression for the resistivity tensor , and find @xmath85 where @xmath86 and @xmath87 equations ( [ eq:1][eq:3 ] ) are the main results of the paper .
the quantities @xmath75 and @xmath76 are plotted in fig .
[ fig : delta](b ) and ( c ) . in low field
( @xmath88 ) @xmath89 and @xmath90 . in large field
( @xmath91 ) @xmath92 and @xmath93 .
these results can be used to show that in low field the resistivity tensor is @xmath94 while in high fields the resistivity tensor is @xmath95 where the matrix @xmath96 contains only geometric factors and is given by @xmath97
the components of the resistivity tensor obtained using equations ( [ eq:1][eq:3 ] ) are shown in fig .
[ fig : basic ] for the case of @xmath72 and @xmath84 . at high magnetic field
the diagonal components of the resistivity tensor saturate ( to a value linear in @xmath79 ) , while the off - diagonal terms approach the asymptotic values proportional to @xmath98 [ in agreement with eq .
( [ highfieldlimit ] ) ] . as the magnetic field
is decreased all resistivity tensor elements go through a maximum that becomes sharper with increasing breakdown field @xmath79 . at lower fields they decrease linearly .
the position of the maximum is controlled predominantly by the behavior of @xmath68 plotted in fig .
[ fig : delta ] .
the components of the resistivity tensor are plotted again in fig .
[ fig : rho ] , though this time for a range of values of breakdown fields ( parametrized by @xmath99 ) .
[ the plot of @xmath100 is identical to that of fig .
1(c ) in ref .
, demonstrating the agreement of the analytical expressions in eqs .
( [ eq:1][eq:3 ] ) and the earlier numerical work . ] the difference between @xmath100 and @xmath101 reflects the fact that for our considered geometry in fig . [ fig : rho ] ( @xmath72 ) the open orbits which occur due to bragg scattering are efficient at carrying current in the @xmath102-direction .
this difference vanishes when @xmath103 .
it is worthwhile to show that the off - diagonal components of the resistivity tensor do indeed obey onsager symmetry , i.e. @xmath104 .
onsager symmetry is dependent upon perfect microscopic reversibility ( with the appropriate sign change in the magnetic field)@xcite , and we find that this is obeyed even with the presence of magnetic breakdown junctions , despite the apparent `` randomization '' of the electron trajectory which occurs at magnetic breakdown junctions
. reversal of the magnetic field changes the way in which the sections of the fermi surface are connected at the magnetic breakdown junctions and the result of this is that the connectivity matrix @xmath105 transforms into its transpose .
this has the effect of making @xmath68 invariant under a sign change of the magnetic field [ and is the origin of the modulus sign of the factor of @xmath106 in the numerator of eq .
( [ eq : delta ] ) ] .
the net consequence of this is that onsager symmetry is preserved . in summary ,
in this paper we have derived an analytical expression for the in - plane components of the resistivity tensor that includes the effect of all the orbits induced by a combination of bragg scattering and magnetic breakdown .
( [ conduc ] ) demonstrates that approach shows that the conductivity tensor can be written as a sum of a matrix representing the effect of total magnetic breakdown and one representing a combination of complex electronic trajectories .
our main result in eqs .
( [ eq:1][eq:3 ] ) provides a compact form for the in - plane components of the resistivity tensor for this problem which are in a convenient form to compare with experimental data .
we are grateful to epsrc ( uk ) and mansfield college , oxford for financial support .
the expression for conductivity from eq .
( [ eq : mainsigma ] ) is @xmath107 and as in ref . the matrix @xmath108 is given by @xmath109 where @xmath110 , @xmath111 , @xmath112 , @xmath113 and @xmath114 .
the function @xmath115 can be simplified because @xmath50 , @xmath51 , so that @xmath116 the vector in eq.([eq : lambdaalpha ] ) can then be written @xmath117 and that in eq.([eq : lambdainit ] ) as @xmath118 with @xmath119 .
the two first and two last entries of each vector differ by a minus sign because these parts of the orbit are exactly @xmath120 apart .
with this identity we can readily calculate : the initial contribution can be simplified as follows : @xmath122 integration of the second term then leads to @xmath123 and hence @xmath124\nonumber\\ & + & { 2\sigma_0\over\pi\gamma}\left [ \frac{\lambda^{1}_{\alpha+}\lambda^{1}_{\beta-}(1-bq)+\lambda^{2}_{\alpha+}\lambda^{2}_{\beta-}(1+aq)+\lambda^{1}_{\alpha+}\lambda^{2}_{\beta-}ap-\lambda^{2}_{\alpha+}\lambda^{1}_{\beta-}bp}{1-bq+a(q+b(p^2-q^2))}\right].\end{aligned}\ ] ] the final result in eq .
( [ conduc ] ) is obtained by substituting expressions for @xmath125 and @xmath126 from eq .
( [ lambdaalphapm ] ) and for @xmath127 and @xmath128 from eq .
( [ lambdaalphainit ] ) and simplifying . 1 d. shoenberg , _ magnetic oscillations in metals _ , ( cambridge university press , cambridge , 1984 ) .
a. b. pippard , _ magnetoresistance in metals _ , ( cambridge university press , cambridge , 1984 ) .
m. h. cohen and l. m. falicov , phys .
* 7 * , 231 ( 1961 ) .
e. i. blount , phys . rev .
* 126 * , 1636 ( 1962 ) .
a. b. pippard , proc .
roy . soc .
london a * 270 * , 1 ( 1962 ) .
l. m. falicov and p. r. sievert , phys .
* 12 * , 550 ( 1964 ) .
l. m. falicov and p. r. sievert , phys . rev . * 138 * , a88 ( 1965 ) .
r. g. chambers , proc .
a * 65 * , 458 ( 1952 ) .
a. nowojewski , p. a. goddard and s. j. blundell , phys .
b * 77 * , 012402 ( 2008 ) .
p. a. goddard , s. j. blundell , j. singleton , r. d. mcdonald , a. ardavan , a. narduzzo , j. a. schlueter , a. m. kini and t. sasaki , phys .
b * 69 * , 174509 ( 2004 ) .
w. kang , t. osada , y. j. jo , and h. kang , phys .
99 , 017002 ( 2007 ) .
a. f. bangura , p. a. goddard , j. singleton , s. w. tozer , a. i. coldea , a. ardavan , r. d. mcdonald , s. j. blundell and j. a. schlueter , phys .
b * 76 * , 052510 ( 2007 ) .
w. kang , y. j. jo , d. y. noh , k. i. son and o .- h .
chung , phys .
b * 80 * , 155102 ( 2009 ) .
m. v. kartsovnik , chem . rev . * 104 * , 5737 ( 2004 ) .
s. j. blundell , a. nowojewski and p. a. goddard , physica b * 405 * , s134 , ( 2010 ) .
l. onsager , phys . rev . * 37 * , 405 ( 1931 ) ; l. onsager , phys . rev . * 38 * , 2265 ( 1931 ) .
|
we derive an analytical expression which allows efficient computation of the effect of all the fermi surface trajectories induced by a combination of bragg scattering and magnetic breakdown on the in - plane components of the resistivity tensor .
the particular network of coupled orbits which we consider was first formulated by falicov and sievert , who studied the problem numerically .
our approach , based upon a method used previously to derive an analytical solution for interlayer transport , allows us to show that the conductivity tensor can be written as a sum of a matrix representing the effect of total magnetic breakdown and one representing a combination of complex electronic trajectories , and we find a compact expression for the in - plane components of the resistivity tensor that can be evaluated straightforwardly .
|
galaxy - hosting dark matter systems and the galactic stellar systems themselves act collisionlessly over hubble timescales . understanding the evolution of self - gravitating , collisionless systems is a foundational element to the larger picture of galaxy evolution .
much of the advancement in understanding collisionless evolution has come from @xmath0-body simulations that focus on the motions of large numbers of individual mass elements .
such simulations allow astrophysicists to determine the density and velocity distributions that result from a variety of initial conditions ( _ e.g. , _ * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . while @xmath0-body simulations have given the astrophysics community a powerful method for making predictions that can be tested against observations , the modeled evolutions are sufficiently complex that a full physical picture of the evolution remains lacking
effects such as dynamical friction , collisionless relaxation processes , and even numerical artifacts can all play roles of varying importance in these simulations .
astrophysicists now have substantial empirical evidence that collisionless systems in cosmological contexts have `` universal '' equilibrium distributions of mass and velocities , _
e.g. , _ the radial power - law behavior of density divided by velocity dispersion cubed @xcite .
we are interested in gaining a better understanding of the physical origin of these seemingly special distributions . before beginning to build a physical picture of collisionless evolution ,
let us quickly review some of the ideas that are involved .
thermal equilibrium is the state that systems relax towards when energy can be exchanged between the components of the system .
a system in thermal equilibrium is characterized by a minimum internal energy or , equivalently , a maximum entropy . in a thermodynamic sense
, entropy is linked to the amount of energy transferred as heat , bringing with it a connotation of randomness . in a statistical sense , entropy is simply a redefinition of the accounting of energy states in a system .
relaxation is a generic term for any process that erases the memory of initial conditions in a system .
for example , gases reach the completely relaxed state of thermal equilibrium through collisions .
@xcite demonstrated that non - degenerate collisionless self - gravitating systems with finite mass and energy do not have a state of maximum entropy when one works with the standard distribution function and assumes that the occupation numbers of phase - space volumes is large enough that stirling s approximation applies ( for an alternative view see * ? ? ? * ) .
this lack of entropy extremum implies that there is no thermal equilibrium state for such systems . and yet
, mechanical equilibrium , a state with no net force at any location , is possible . in stellar interiors , for example
, the lack of thermal equilibrium evidenced by gradients in temperature and pressure is required for mechanical equilibrium .
note that the thermal non - equilibrium description refers to the entire star .
it is common to treat the gas in a star as being in local thermodynamic equilibrium , where small regions of the star are modeled as having uniform temperatures . in this work ,
we consciously make a distinction between full thermal equilibrium and stationary ( time - independent ) situations .
an example that clarifies this difference is familiar to those of us nearer the poles . on a cold winter day ,
a house is heated to a temperature that is well above the outside air temperature .
heat flows out of the house , and entropy is generated in a wall during the energy transfer .
as long as the home s heating system runs and the outdoor temperature remains constant , this process will be steady - state , yet it does not represent a thermal equilibrium as the entropy of the home - outside air system continually increases .
dark matter halos and galactic stellar systems are similar to the above , in the sense that they have temperature gradients and are thus thermally non - equilibrium systems , even though they are steady - state and are in mechanical equilibrium .
thermal non - equilibrium systems are common in nature and have been the subject of numerous studies . for certain non - astrophysical , non
- gravitating systems it was found that stationary states , like mechanical equilibrium , coincide with states of extreme entropy production ( * ? ? ? * ; * ? ? ?
* and references therein ) . in some instances
, it appears that steady - states are reached by minimizing entropy production , while in other cases maximizing entropy production is the requirement .
the most common modern formulation of the principle of maximum entropy production is due to @xcite .
the principle of miminum entropy production @xcite appears to have a smaller range of applicability , but does find uses in physics , chemistry and biology .
the theoretical interpretation , and the relation between the two seemingly contradictory principles is still being worked out ( _ e.g. , _ , * ? ? ?
* ; * ? ? ?
a further , and more general reason to explore entropy production in self - gravitating systems is because irreversability and entropy production are closely related concepts .
the `` microscopic '' equation of motion of gravity ( and of other microscopic physics ) is time - reversible , while the global evolution of the system is not , in the sense that it has a definite arrow of time . a natural agent that can make the transition from reversability to irreversability in macroscopic systems is the thermodynamic entropy production .
this ability of thermodynamics to pick the arrow of time through non - zero ( positive ) entropy production has been utilized in a cosmological context by @xcite .
these authors propose a scenario where in the early universe some of the energy of the space - time ( gravity ) was transferred to matter .
because matter production is an irreversible , entropy creating process , the evolution of the universe proceeds with an arrow of time . in this paper
we investigate the consequences of applying the entropy production extremization principle to self - gravitating systems .
we do not prove that such a principle should be applicable here ; we merely explore what this principle implies .
in particular , we want to find what conditions are necessary to extremize entropy production and how those conditions differ from maximizing the total entropy .
we begin with a review of relevant ideas regarding entropy from both a thermodynamic and statistical point of view in section [ back ] . in
[ kinetic ] , entropy production equations for the two statistical families we are interested in are then derived . for these cases
, we also develop the specific forms for collisionless relaxation processes required to extremize the entropy production .
we present a discussion of our results and conclude in section [ discuss ] .
much of what follows is a distilled derivation of the outcome of extemizing entropy production in collisionless systems . as the mathematical details might muddy the overall flow of the paper , we present an executive summary to draw the reader s attention to the main equations and the connections between them .
* we start by developing a continuity equation for entropy density ( @xmath1 ) , which has an entropy production source term @xmath2 ( see equation [ s6 ] ) . * statistical descriptions of entropy and entropy density
are derived from enumerating states accessible to the system .
entropy density expressions for both maxwell - boltzmann and lynden - bell statistical families may be found in equations [ mbsdens ] and [ lbsdens ] , respectively .
these expressions depend on the coarse - grained distribution function , @xmath3 . * taking the partial time derivative of @xmath1 in equations [ mbsdens ] and [ lbsdens ] produces equations [ mbk1 ] and [ lbk1 ] , respectively .
these diffentiations result in @xmath4 terms , which we replace using the boltzmann equation ( [ boltz ] ) .
this introduces the relaxation function @xmath5 , as shown in equations [ mbk2 ] and [ lbk2 ] , respectively .
* expressions of the entropy production @xmath2 for both statistical families are identified by comparing terms resulting from equations [ mbk2 ] and [ lbk2 ] to those in equation [ s6 ] . *
finally , we set the variation of the @xmath2 expressions equal to zero .
this procedure then yields the @xmath5 functions found in equations [ mbexts ] and [ lbexts ] , respectively . following ( * ? ? ?
3 ) , we argue that there are two ways to change the total entropy of a system ; 1 ) entropy can enter or leave through a flux , or 2 ) entropy may be created within the system .
note that to be in accord with the second law of thermodynamics during the system s evolution , the system s entropy should not decrease .
a straightforward approach to dealing with entropy changes in time is , @xmath6 where the total entropy is , @xmath7 @xmath8 is the flux entropy , and @xmath9 is the created entropy . unless noted otherwise , all integrations are assumed to be over all configuration and/or velocity space .
the value of the flux entropy depends on how much entropy is entering or leaving the system through a bounding surface , @xmath10 where @xmath11 is an entropy flux normal to the surface s , @xmath12 is the entropy convected through the surface by mean motions , the minus sign is convention , and the last equality results from the divergence theorem .
the flux @xmath11 is the result of random motions within a system . in as much as random motions can be linked to thermal energy
, one can think of the random flux as being associated with heat .
the subscripts on the integrals simply draw attention to the change in integration range . the entropy creation term is likewise related to a quantity associated with the system , @xmath13 where @xmath2 is the entropy production per unit volume per unit time . substituting equations [ s2 ] , [ s3 ] , and [ s4 ] into equation [ s1 ] , @xmath14 as this expression must be true for an arbitrary volume ,
@xmath15 this is a continuity equation for the entropy density .
the source term , @xmath2 , represents the entropy production within any infinitesimal volume of the system . from a statistical point of view , entropy is simply a measurement of the number of states accessible to a particular system . from the statistical definition of temperature , entropy can be calculated using , @xmath16 where @xmath17 is the number of accessible states and @xmath18 is the boltzmann constant which serves to give entropy the correct thermodynamic units . as a result , counting procedures are key to determining specific realizations of entropy .
@xcite discusses how there are four counting styles that lead to physically relevant situations .
bose - einstein statistics follow from counting states for indistinguishable particles that can co - habitate . when indistinguishable particles are not allowed to share states ,
fermi - dirac statistics emerge .
classical maxwell - boltzmann ( mb ) statistics result from counting states available to distinguishable particles that can share states . completing the symmetry , systems where distinguishable particles can not co - occupy states obey what has become known as lynden - bell ( lb ) statistics ( statistics ` iv ' in lynden - bell s original notation ) .
each type of statistics will produce different representations of entropy , but we will focus on mb and lb here , both of which deal with classical distinguishable particles .
we briefly recap the notation used in @xcite before proceeding with our discussion . the usual six - dimensional position - velocity phase - space ( @xmath19,@xmath20 )
is the setting for determining the statistics .
imagine phase - space to be divided into a very large number of nearly infinitesimal volumes @xmath21 called micro - cells .
each micro - cell can either be occupied or unoccupied by one of the @xmath0 phase elements of the system .
these phase elements can be thought of as being represented by the fine - grained distribution function .
this fine - grained distribution function is assumed to have a constant value @xmath22 . at this level , the phase elements can not co - habitate .
if phase - space is also partitioned on a coarser level so that some number of micro - cells @xmath23 occupy a macro - cell , then we can discuss a coarse - grained distribution function .
the volume of a macro - cell is then @xmath24 and the @xmath25th macro - cell contains @xmath26 phase elements .
we assume that while the volume of a macro - cell is much larger than that of a micro - cell , it is still very small compared to the full extent of phase - space occupied by the system .
the number of ways of organizing the @xmath26 elements into the @xmath23 micro - cells without co - habitation is , @xmath27 if the elements were allowed to multiply occupy micro - cells ( as in mb statistics ) this number would be @xmath28 . to get the total number of accessible states , the possible ways to distribute the @xmath0 phase elements into @xmath26 chunks must also be included .
@xcite derives , @xmath29 in the mb case , the only change is that the factorial ratio in the final product term is replaced by @xmath28 . with equation [ lbw ] ,
the entropy can now be expressed as , @xmath30 \right\},\end{aligned}\ ] ] where we have assumed @xmath31 , @xmath32 , @xmath33 , and used stirling s approximation . in a maxwell - boltzmann situation , the familiar entropy expression is , @xmath34 where @xmath35 .
note that @xmath36 reduces to an expression similar to @xmath37 ( except for the constant ) when @xmath38 . in this
dilute , non - degenerate limit , the difference between allowing and denying co - habitation of states becomes irrelevant .
the number of elements occupying a macro - cell @xmath26 is now replaced by a coarse - grained distribution function of position and velocity . the coarse - grained function @xmath3 is a continuous function of @xmath19 and @xmath20 and represents the number of occupied micro - cells within macro - cells , @xmath39 where the integration is taken over all of phase - space .
@xcite goes on to investigate the results of minimizing equation [ lbs0 ] with the constraints of constant system mass and energy .
the well - know result of this work is that non - degenerate self - gravitating systems do not have a finite mass and energy thermodynamic equilibrium .
we will now describe our investigation into the entropy production that occurs in these systems , keeping the assumption that stirling s approximation is accurate .
note that entropy production takes place because we are considering coarse - grained distributions ; in other words , particles are now allowed to occupy states not accessible within the fine - grained world view . at the fine - grained level
the distribution function does not change and there is no change in entropy .
we now proceed to combine the fluid picture of entropy presented in section [ entropy ] with the statistical picture from section [ stat ]
. the general procedure will be to find the time rate of change of the statistical versions of entropy and equate terms to those in equation [ s6 ] . to do this
, we will rely on the boltzmann equation .
note that the coarse - grained distribution function @xmath3 does not obey the collisionless boltzmann equation , rather its evolution is defined by , @xmath40 where @xmath41 is the acceleration and @xmath42 is the velocity space gradient operator ( @xmath43 ) . in standard ,
ideal gas - like situations @xmath5 describes how inter - particle , two - body collisions can affect the number of particles in a given phase - space volume . in the systems that we are dealing with , @xmath5 represents the effect of collisionless relaxation processes , like violent relaxation and phase mixing , which may be aspects of landau damping @xcite
. we will refer to @xmath5 as the relaxation function in subsequent sections . as a simple introduction to the procedure , we begin by discussing a system in the maxwell - boltzmann regime .
discussion of lynden - bell - type systems will follow in section [ lb ] . from the entropy form given in equation [ mbs0 ] , our goal is to transform the summation over particles into an integration over phase space .
first , we re - write equation [ mbs0 ] as @xmath44 by taking the constant term into the summation .
next , we use the fact that @xmath45 , where @xmath46 , to produce , @xmath47 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{x}}}{\mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] the entropy density can now be written as , @xmath48 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] note that this expression for entropy density differs slightly from other kinetic theory derivations in the constant term ( @xmath49 versus 1 ) .
this is due to a different choice for the additive constant in the definition of entropy . as such
, it does not affect the behavior we are going to investigate . taking a partial time derivative of equation [ mbsdens ] results in , @xmath50 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\end{aligned}\ ] ] substituting @xmath4 from the boltzmann equation into equation [ mbk1 ] results in a lengthy expression that we will deal with term by term .
for reference , the expression after substitution is , @xmath51 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \nonumber \\ & & { \ensuremath{k_{\rm b}}}\int ( -{\ensuremath{\boldsymbol{a}}}{\ensuremath{\boldsymbol{\cdot}}}{\ensuremath{\boldsymbol{\nabla}}}_v f ) \left [ \ln{\left(\frac{f}{\eta}\right ) } + 1 - \ln{n } \right ] \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \nonumber \\ & & { \ensuremath{k_{\rm b}}}\int \gamma \left [ \ln{\left(\frac{f}{\eta}\right ) } + 1 - \ln{n } \right ] \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\end{aligned}\ ] ] the first integral on the right - hand side of equation [ mbk2 ] can be recast using integration by parts and the fact that @xmath52 in phase - space , @xmath53 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}= { \ensuremath{\boldsymbol{\nabla}}}{\ensuremath{\boldsymbol{\cdot}}}\int f \left ( \ln{\left(\frac{f}{\eta}\right ) } - \ln{n } \right ) { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] since @xmath41 is velocity - independent , @xmath54 , and we can again use integration by parts to transform the second term on the right - hand side of equation [ mbk2 ] to , @xmath55 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \ln{n}\int { \ensuremath{\boldsymbol{a}}}{\ensuremath{\boldsymbol{\cdot}}}{\ensuremath{\boldsymbol{\nabla}}}_v f \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] using the divergence theorem and the fact that any physical distribution function must vanish for large velocities , these integrals vanish . equation [ mbk2 ]
now has the form , @xmath56 \right\ } \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] assuming the velocity field to be composed of mean and peculiar components @xmath57 , we can re - cast equation [ mbk4 ] as , @xmath58 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}\bigg\}.\end{aligned}\ ] ] we now equate the terms in equation [ mbk5 ] to those in equation [ s6 ] , @xmath59 the entropy flux @xmath11 is given by the integral in the second term on the right - hand side of equation [ mbk5 ] and represents randomly fluxed entropy .
the remaining term is the entropy production for the system , @xmath60 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] this equation explicitly demonstrates how the non - collisionless nature of the coarse - grained distribution function leads to changes in entropy . as mentioned in
[ intro ] , thermodynamic non - equilibrium systems can have steady states described by extrema of entropy production .
note that we do not extremize the entropy itself ; that procedure implies a system in thermodynamic equilibrium ( but not necessarily with constant kinetic temperature , see * ? ? ?
what we wish to do is to set , @xmath61 .
taking the variation of equation [ mbsprod ] , we find @xmath62 + \frac{\gamma}{f } \right\ } \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] since @xmath63 is arbitrary , the variation disappears only when the term in curly braces is zero .
note that we did not impose any additional constraints while carrying out the extremization . in the standard entropy extremization procedure langrange multiplier terms
are introduced to assure constant mass and energy . the condition for an extremum in entropy production is , @xmath64
this can be rewritten as , @xmath65 with the following solution for the relaxation function , @xmath66 where the value of the fine - grained distribution function @xmath67 is the maximum value allowed for @xmath3 . in a collisional system that follows the maxwell - boltzmann distribution , like an ordinary gas
, one could imagine @xmath3 reaching any value due to collisions .
however , for the collisionless case , mb statistics are the non - degenerate limit , implying that @xmath68 . figure [ mbgplot ] contains plots of @xmath69 for a variety of values of @xmath0 . as @xmath0 increases ,
the relaxation function becomes more and more uniform .
we now investigate a system that obeys lynden - bell statistics .
the entropy of the system can be transformed from the summation in equation [ lbs0 ] to , @xmath70 + ( \eta - f)\ln{\left ( \frac{\nu}{\eta}(\eta - f ) \right ) } \ : { \mathrm{d}}{\ensuremath{\boldsymbol{x}}}{\mathrm{d}}{\ensuremath{\boldsymbol{v}}},\ ] ] where @xmath71 .
the entropy density can now be written as , @xmath72 + ( \eta - f)\ln{\left ( \frac{\nu}{\eta}(\eta - f ) \right ) } \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] taking a partial time derivative of equation [ lbsdens ] results in @xmath73 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}\nonumber \\ & = & -{\ensuremath{k_{\rm b}}}\int { \ensuremath{\frac{\partial f}{\partial t}}}\ln{\left(\frac{f}{\eta - f}\right ) } \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}+ { \ensuremath{k_{\rm b}}}c \int { \ensuremath{\frac{\partial f}{\partial t}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}},\end{aligned}\ ] ] where @xmath74 is a constant . substituting @xmath4 from the boltzmann equation into equation [ lbk1 ] again results in a lengthy expression that we will deal with term by term . for reference ,
the expression after substitution is , @xmath75 the first integral on the right - hand side of equation [ lbk2 ] can be transformed to , @xmath76 we now re - write the first term on the right - hand side of equation [ lbk3 ] as , @xmath77 { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] adding and subtracting terms of the form @xmath78 } { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}$ ] to equation [ lbk4 ] gives us a term that is very reminiscent of the entropy density in equation [ lbsdens ] , @xmath79 { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \nonumber \\ & & { \ensuremath{\boldsymbol{\nabla}}}{\ensuremath{\boldsymbol{\cdot}}}\int \eta \ln{\left(\frac{\nu}{\eta}(\eta - f ) \right ) } { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \int \frac{\eta}{\eta - f } { \ensuremath{\boldsymbol{v}}}{\ensuremath{\boldsymbol{\cdot}}}{\ensuremath{\boldsymbol{\nabla}}}f \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\end{aligned}\ ] ] looking at the second term on the right - hand side of equation [ lbk5 ] , we see that @xmath80 we can now simplify equation [ lbk5 ] to @xmath81 { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] next , we deal with the second integral on the right - hand side of equation [ lbk2 ] . since @xmath82 is velocity - independent , we can write @xmath83 using @xmath84 - \frac{\eta}{\eta - f}{\ensuremath{\boldsymbol{\nabla}}}_v f,\ ] ] and taking advantage of the velocity independence of @xmath41 , we can re - write the right - hand side of equation [ lbkk1 ] as , @xmath85 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \int \frac{\eta}{\eta - f } { \ensuremath{\boldsymbol{a}}}{\ensuremath{\boldsymbol{\cdot}}}{\ensuremath{\boldsymbol{\nabla}}}_v f \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] the first term disappears when the divergence theorem is applied , as the distribution function goes to zero at large velocities . to deal with the second term , we substitute @xmath86 to get , @xmath87 let us investigate one component of this vector integral , @xmath88 since a well - behaved distribution function is expected to disappear at the largest speeds possible .
so , the second integral on the right - hand side of equation [ lbk2 ] is zero .
we now re - combine the terms on the right - hand side of equation [ lbk2 ] using the results of equations [ lbk6 ] and [ lbkk1 ] to produce our final representation of the time rate of change of the entropy density , @xmath89 + ( \eta - f)\ln{\left ( \frac{\nu}{\eta}(\eta - f)\right ) } \right\ } { \ensuremath{\boldsymbol{v}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \nonumber \\ & & { \ensuremath{k_{\rm b}}}\int \gamma \left[\ln{\left(\frac{f}{\eta - f}\right)}-c \right ] \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\end{aligned}\ ] ] again assuming the velocity field to be composed of mean and peculiar components @xmath57 , we draw a correspondence between , @xmath90 + ( \eta - f)\ln{\left ( \frac{\nu}{\eta}(\eta - f)\right ) } \right\ } { \ensuremath{\boldsymbol{v_p}}}\ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}- \nonumber \\ & & { \ensuremath{k_{\rm b}}}\int \gamma \left[\ln{\left(\frac{f}{\eta - f}\right)}-c \right ] \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}},\end{aligned}\ ] ] and equation [ s6 ] @xmath59 the entropy flux due to random motions @xmath11 is given by the integral in the second term on the right - hand side of equation [ lbkk4 ] . the remaining term
then makes up the entropy production for the system , @xmath91 \ : { \mathrm{d}}{\ensuremath{\boldsymbol{v}}}.\ ] ] again , there is a relaxation function - dependent term in the entropy production .
following the same steps as in the maxwell - boltzmann case , we find that the condition for the entropy production term in equation [ lbsprod ] to be extremized is , @xmath92}-c},\ ] ] where @xmath93 is an integration constant and @xmath94 . if we set @xmath95 , @xmath96 and @xmath97}-c}.\ ] ] like @xmath98 , this function behaves more like a constant as the @xmath0 value increases ( @xmath99 ) .
@xmath100 curves for a variety of @xmath101 values are presented in figure [ lbgplot]a . the increase in @xmath100 seen as
@xmath102 is due to @xmath103}$ ] approaching @xmath101 .
once @xmath3 is sufficiently close to @xmath67 that subtracting @xmath101 produces a negligible change , @xmath104 .
the complex behavior of @xmath100 for @xmath105 is shown in figure [ lbgplot]b and is discussed further in the next section .
the seminal paper by @xcite has shown that , under certain assumptions , there is no maximum entropy state for self - gravitating systems .
alternative descriptions of the same systems which yield self - consistent results under entropy extremization may be found in @xcite and @xcite .
lynden - bell s conclusion is exemplified by the order - of - magnitude calculation in ( * ? ? ?
* ; * ? ? ?
* s. 4.7 ) , using maxwell - boltzmann statistics .
the result of this calculation is that when a collisionless system rearranges its mass to become more centrally concentrated in the core , the entropy of the contracting core mass decreases .
the outer envelope , expanding in response to the contracting core to satisfy the virial theorem , should have an increase in entropy . as a result , the entropy for the entire system increases .
since a system can always increase its entropy by contracting the core and expanding the envelope , there is no maximum entropy state , and hence maximizing entropy will not lead to a satisfactory description of a steady state .
yet , we know from numerous high - resolution @xmath0-body simulations that long - lived steady states do exist .
how does one find these theoretically ?
apparently , one has to resort to means other other than entropy extremization . in this paper
we try one alternative approach .
we apply a principle of extremizing entropy production rate to self - gravitating systems .
this principle has been used widely to describe thermal non - equilibrium , but not in systems that are self - gravitating .
our basic hypothesis is that a steady state is obtained by extremizing the entropy production .
we present expressions for the entropy production rates for two types of statistics , maxwell - boltzmann ( mb ) and lynden - bell ( lb ) , as equations [ mbsprod ] and [ lbsprod ] , respectively .
we then find expressions ( equations [ mbexts ] and [ lbexts ] ) for the relaxation term that forms the right hand side of the coarse - grained boltzmann equation .
the meaning of these expressions and the interpretation of our results , under the assumption that this idea is applicable to self - gravitating systems , are discussed below .
the development of the expressions for entropy production rates @xmath2 is a central result of this work .
the descriptions of @xmath2 for the mb and lb statistical families are given in equations [ mbsprod ] and [ lbsprod ] , respectively .
since we are dealing with collisionless systems exclusively , one might expect these to be zero .
in fact , if we were considering the fine - grained distribution function , there would be no entropy change , no thermodynamic evolution , as the collisionless @xmath106 boltzmann equation would apply .
however , we are considering the coarse - grained function .
as a system evolves , the fine - grained distribution function becomes stretched and twisted in phase - space . because the coarse - grained function averages the fine - grained function with nearby empty regions of phase - space , the coarse - grained function changes as the system evolves .
now , recall that entropy represents the number of accessible states . on the level of the fine - grained function
, the number of accessible states stays the same .
however , going from a fine - grained to coarse - grained description implies that there are now regions of phase - space not occupied by the fine - grained function that are accessible to the coarse - grained function .
this implies that there are more microscopic ways of realizing a given macroscopic state , leading to more possible states and larger entropy .
thus , coarse - graining an evolving system results in entropy production even in a collisionless system . in terms of physical processes , the evolution is due to the larger - scale phase - space evolution of the system driven by collisionless relaxation processes , like violent relaxation and phase mixing .
the above argues that entropy production takes places during evolution of collisionless systems .
but our analysis shows that entropy production takes place even during the steady state .
let us start by writing down the expression for the entropy production during the steady - state by combining equations [ mbsprod ] and [ mbexts ] for the mb case , @xmath107 where @xmath108 is the volume of occupied velocity space .
this is consistent with the second law of thermodynamics since @xmath109 and all other terms are positive .
a similar expression can be found for the lb case , using equations [ lbsprod ] and [ lbexts ] . since a steady state
is described by an unchanging value of @xmath2 , any non - zero value of @xmath2 persists even when a system has reached mechanical equilibrium .
it is interesting to think of the source of this continued entropy production .
after a system stops evolving on the macroscopic scale , it still continues to evolve on ever decreasing microscopic scales as the fine - grained function continues to stretch and twist almost indefinitely .
the corresponding continual coarse - graining of the ever evolving fine - grained function on smaller and smaller scales , results in constant , non - zero entropy production .
our expression for the boltzmann equation states that the relaxation function , @xmath5 , determines the rate of change of the coarse - grained distribution function , @xmath110 despite the right hand side being non - zero ( given by equations [ mbexts ] and [ lbexts ] for the mb and lb statistics , respectively ) , the above equation does not contradict the assumption of a stationary state . a stationary , or steady , state is the eulerian viewpoint , _ i.e. , _ @xmath111 , while the boltzmann equation above is a lagrangian viewpoint . @xmath111
does not imply @xmath112 .
in other words , the relaxation function @xmath5 does not determine the explicit time - dependence of @xmath3 , which must be zero for stationary states , but rather describes a flux of occupied cells through phase - space , as do the velocity - driven ( @xmath113 ) and acceleration - driven ( @xmath114 ) flux terms ( c.f .
equation [ boltz ] ) . in the context of the lagrangian derivative , we can think of @xmath115 simply as a parameter that indicates the location along a particle s or cell s path through phase - space . in collisional systems , the relaxation function
is called the collision term and is usually dealt with in a fokker - planck approximation scheme . in these systems the particles , through two - body encounters , gradually disperse over the whole available phase - space , and so @xmath116 following any given particle in an evolving system does not stay constant , but generally decreases with time . in a general collisionless system
, this term represents the processes like violent relaxation and phase mixing , on the coarse - grained scale . in a steady - state collisionless system ,
the large scale processes like violent relaxation no longer operate and the only changes happen on microscopic scales . in this context , the left hand side of equation [ boltzfull ] describes how a particle , or a cell moves through the system ( and @xmath115 is the parameter ) .
therefore it is not unexpected that over some portions of its motion the coarse - grained density around it will be increasing and over others , it will be decreasing .
for the mb case , @xmath117 for @xmath118 , implying that @xmath3 should continually grow .
however , this is impossible as the coarse - grained distribution function is limited to a maximum value @xmath67 .
this contradiction arises as mb statistics are valid only when @xmath68 so that macro - cells are not multiply occupied . on the other hand
, the lb case does not present any contradictions with the boltzmann equation .
@xmath100 is zero when @xmath119 and @xmath120 , and is positive over the vast majority of the intervening range .
this behavior guarantees that when the coarse - grained density reaches its maximum value , the relaxation term disappears and the system behaves collisionlessly even at the coarse - grained level . to sum up ,
if extremizing entropy production in self - gravitating systems does lead to steady - state configurations , then equation [ boltzfull ] with the appropriate expressions for @xmath5 ( as given by equations [ mbexts ] and [ lbexts ] ) describes the steady state of self - gravitating collisionless systems .
the relaxation term @xmath5 describes the continual evolution of the coarse - grained distribution function , which is due to the combimation of the dynamical evolution of the fine - grained df on microscopic scales , and coarse - graining .
|
certain thermal non - equilibrium situations , outside of the astrophysical realm , suggest that entropy production extrema , instead of entropy extrema , are related to stationary states . in an effort to better understand the evolution of collisionless self - gravitating systems
, we investigate the role of entropy production and develop expressions for the entropy production rate in two particular statistical families that describe self - gravitating systems . from these entropy production descriptions ,
we derive the requirements for extremizing the entropy production rate in terms of specific forms for the relaxation function in the boltzmann equation .
we discuss some implications of these relaxation functions and point to future work that will further test this novel thermodynamic viewpoint of collisionless relaxation .
|
it is generally accepted today that the fermi surfaces of metallic systems obtained with density functional theory ( dft ) , even at the level of local density approximation ( lda ) , are in good agreement with experiments .
unfortunately , this is not the case with the band gaps of insulators and semiconductors which are highly underestimated by most of the exchange - correlation ( xc ) functionals within dft .
even with the exact xc functional of dft , the kohn - sham ( ks ) gap is not expected to reproduce the experimental gap@xcite .
this deviation from experiment is most dramatic for mott - insulators , most of which are predicted by their ks spectrum to be metallic while they are experimentally known to be insulating in nature . in this regard reduced density matrix functional theory ( rdmft ) has shown great promise in improving on dft results for a wide class of systems in that it not only improves the ks - band gaps for insulators in general , but also predicts the correct insulating nature for mott insulators @xcite . within rdmft
the total energy of a system of interacting electrons is expressed in terms of the one - body reduced density matrix ( 1-rdm ) , @xmath0 .
this energy functional is then minimized with respect to @xmath1 under the @xmath2-representability conditions @xcite which restrict the minimization to the domain of 1-rdms that correspond to ensembles of @xmath2-electron wave functions .
a major advantage of rdmft comes from the fact that the _ exact _ kinetic energy is easily expressed as a functional of the 1-rdm of the ground state .
in addition , due to the departure from the idempotent single - determinant solution , static electronic correlations are well described@xcite .
the total ground - state energy as a functional of @xmath1 reads ( atomic units are used throughout ) : @xmath3=&-\frac{1}{2 } \int\lim_{{\bf r}\rightarrow{\bf r } ' } \nabla_{\bf r}^2 \gamma({\bf r},{\bf r}')\,d^3{\bf r } ' + \int\rho({\bf r})\ , v_{\rm ext}({\bf r})\,d^3{\bf r } \nonumber \\ & + \frac{1}{2 } \int \frac{\rho({\bf r})\,\rho({\bf r } ' ) } { |{\bf r}-{\bf r}'|}\,d^3{\bf r}\,d^3{\bf r}'+e_{\rm xc}[\gamma],\end{aligned}\ ] ] where @xmath4 .
@xmath5 is a given external potential , and @xmath6 we call the xc energy functional . in practice ,
the xc functional is an unknown functional of the 1-rdm and needs to be approximated .
a milestone in the development of approximate functionals of the 1-rdm is the mller functional@xcite , which has the following form : @xmath7=e_{\rm xc}[\{\phi_{j}\},\{n_{j}\ } ] = \\
-\frac{1}{2}\int \int \frac{|\gamma^{1/2}({\bf r},{\bf r}')|^2}{|{\bf r}-{\bf r}'| } \ , d^3{\bf r}\ , d^3{\bf r}'\end{gathered}\ ] ] where @xmath8 is an exponent in the operator sense .
diagonalization of @xmath1 produces a set of natural orbitals ( the eigenvectors of @xmath1 ) , @xmath9 , and occupation numbers ( the eigenvalues of @xmath1 ) , @xmath10 .
the mller functional is known to over correlate@xcite , however , there exist several other approximations most of which are modifications of this functional and are known to improve results for finite systems @xcite .
several of these rdmft functionals reproduce the discontinuity of the chemical potential at integer number of electrons which is a measure of the fundamental gap of the system@xcite .
more precisely , it was demonstrated @xcite that the complete removal of the self - interaction ( si ) terms leads to discontinuities in the chemical potential that are in good agreement with the fundamental gap for finite systems .
unfortunately , this removal has no effect in the total energy for infinitely extended natural orbitals in periodic systems since their contribution vanishes in the limit of the size of system going to infinity . to overcome this problem , sharma _ et al . _
@xcite introduced the power functional@xcite that reproduces discontinuities without requiring the removal of si terms .
this functional has the form @xmath11=e_{\rm xc}[\{\phi_{j}\},\{n_{j}\ } ] = \\ -\frac{1}{2}\int \ , \int \frac{|\gamma^{\alpha}({\bf r},{\bf r}')|^2}{|{\bf r}-{\bf r}'|}\,d^3{\bf r}\ , d^3{\bf r}'\end{gathered}\ ] ] where @xmath12 is an exponent in the operator sense .
the power functional was applied in the calculation of the fundamental gap of various systems@xcite including transition metal oxides @xcite .
an optimal value of @xmath12 between @xmath13 and @xmath14 was found to reproduce gaps of all systems in close agreement with experiments .
these gaps were obtained from the discontinuity of the chemical potential , @xmath15 , at an integer total number of electrons @xmath2 .
a problem of this method of predicting the gap is that the shape of @xmath15 differs substantially from a step function leading to large error - bars in the prediction of the gap .
a second problem is that one needs to calculate the total energy ( and @xmath16 ) for several values of @xmath2 making the calculation time consuming .
finally , a third problem is that this method does not allow for the calculation of quantities other than the gap , for direct comparison with experiments , like for example the density of states for extended systems and ionization potentials ( ips ) and electron affinities ( eas ) for finite systems .
an advantage of dft is that the ks eigenvalues can be used as an approximate single - electron spectrum of the system .
thus , quantities like the ip and ea can be easily estimated using the ks spectrum .
a fundamental difference between rdmft and dft is the lack of a ks system within rdmft and the lack of eigenvalue equation makes it difficult to obtain ( even approximately ) the ips and the eas .
one way to calculate ips in rdmft , is to use extended koopman s theorem ( ekt ) , as was proposed by pernal and cioslowski@xcite .
they demonstrated that the lagrangian matrix in rdmft is identical with the generalized fock matrix entering ekt .
thus , ips can be calculated by diagonalization of this matrix .
they used this idea in the calculation of ips for small molecular systems using the so called buijse baerends corrected ( bbc ) @xcite and goedecker umrigar ( gu)@xcite functionals and showed that the error in the obtained ips is of the order of 4 - 6% .
the same idea was employed in combination with yet another xc functional , namely the pnof1@xcite functional , for the calculation of the first ips ( fips ) as well as higher ips ( hips ) of molecular systems yielding results of similar quality@xcite .
however , the application of this method is restricted to finite systems , since , for solids , it would require the diagonalization of a large matrix in wave - vector space .
sharma _ et al .
_ in ref .
[ ] introduced an alternative technique to obtain spectral information .
we refer to this technique as the `` derivative '' ( der ) method as it entails for each natural orbital , @xmath17 , the associated energy , @xmath18 , be obtained as the derivative of the total energy with respect to the occupation number , @xmath19 , at @xmath20 and with the rest of the occupation numbers set equal to their ground - state optimal values .
this technique has been applied for the calculation of densities of states of transition - metal oxides ( nio , feo , coo and mno ) and the results were found to be in excellent agreement with experiments @xcite and other state - of - the - art many - body techniques like dynamical mean - field theory and the @xmath21 method . in the present work
this technique is applied to finite systems .
we discuss the validity of the approximations necessary for the accuracy of the method .
we present results for the fip as well as hip , for atoms and molecules as well as the eas of atoms , molecules , and radicals adopting several present - day functionals of the 1-rdm .
we compare the results with ekt , qci(t ) , and experiment .
.[tab : errors ] average absolute errors @xmath22 , @xmath23 , and @xmath24 , in the calculation of ip , fips and hips , for a set of atoms and molecules with calculations performed using various xc functionals in conjunction with def , dif , der methods respectively . [ cols="<,^,^,^ " , ]
our results for the average absolute errors in the calculation of ips with the three methods are included in table [ tab : errors ] .
average errors in the results obtained with ekt method are also included in the table .
the actual values for ips obtained using the der method are compiled in table [ tab : ip ] .
( full results for all methods as well as ekt can be found in the supplementary material@xcite . )
it is clear from table [ tab : errors ] that all functionals in combination with the der method give reasonable results for ips with errors ranging from 4 to 13% .
ml , ac3 and power functionals perform slightly better by giving an average error of only 4 - 6% . for the systems considered here ,
the ml functional with the der method is the most accurate for the fips ( with an error of only 2% ) .
it is important to note that the errors from the der and ekt method are of the same order ( using the same functionals and basis set ) , while there is less computational effort involved in der method .
the comparison of the dif and der methods allows us to assess the validity of the linear approximation the linearity of the total energy with respect to variation of one occupation number while keeping the rest of the occupation numbers as well as natural orbitals frozen ( this was demonstrated for solids in ref .
as we see in table [ tab : ip ] , the der method gives good results also for finite systems and for the best performing functionals the average difference between the dif and der methods results is in the range of 2 - 7% .
this percentage may be regarded as the magnitude of non - linearity of the total energy with respect to variation of a single occupation number . as mentioned in sec .
[ sec : method ] , the def method suffers from the over correlation error due to the approximate nature of the xc functional and the difference in variational freedom between closed and open - shell systems .
since the der method is less prone to this error , it is not a surprise that for the power and ml functionals the der method improves the results over the def method .
this indicates that the dependence of the total energy on a particular occupation number deviates from linearity but this deviation works in favor of the der method by further improving the results . in order to understand this , in fig .
[ fig : e_nip](top ) , we show the tangents at 1/2 of the total energy as a function of the occupation number of the homo , while keeping the rest frozen .
the plots are made for various functionals .
it is clear from fig .
[ fig : e_nip ] that although the total energy functional itself deviates from linearity the tangent at 1/2 is very close to the one that reproduces the experimental results .
one reason for this improvement over the dif method might be that the extension of the theory to open shells in the case the der method is minimal since only half of an electron is unpaired and der method reduces the error introduced by the extension of functionals to open shells .
the average percentage errors and the values of hips for various atoms and molecules are also presented in table [ tab : errors ] and table [ tab : ip ] respectively .
the average error in hips obtained using the mller , gu , power and pnof1 functionals , is substantially lower ( 4 - 7% ) than for the fips .
the rest of the functionals are less accurate for hips with average absolute errors slightly higher than those for the fips ( 5 - 10% ) .
the average absolute errors in the calculation of eas with the def , dif , and der methods are shown in table [ tab : eaerr ] .
the actual values for eas obtained using der method are included in table [ tab : ea ] .
as already mentioned , eas are more difficult quantities to calculate the errors are introduced by describing negative ions with localized basis which are usually optimized for the description of the ground states of neutral systems .
in addition , being a small quantity , eas are more prone to errors in the differences of total energies corresponding to two different shell structures ( for example the difference in energy between a doublet and a singlet state ) .
thus , it is not surprising that the errors in tables [ tab : eaerr ] and [ tab : ea ] are substantially higher than for the ips .
however , state - of - the art quantum chemical methods like qci(t ) also exhibit large errors within the adopted basis set . under these considerations
the ac3 , pnof1 and ml functionals perform surprisingly well for eas with average errors of 21.4% , 28.6% , and 21.5% respectively .
these errors are close to that of qci(t ) ( 17.3% ) .
it is interesting to note that ( see table [ tab : ea ] ) there are only a few cases ( 5 of 60 ) for which mller , gu , bbc3 and power functionals give a zero ea [ @xmath25 , i.e. , the system is not predicted by the corresponding functional to bind an extra electron . for the best performers like the ac3 , pnof1 , and ml functionals , no such case exists . in order to compare the dif and der methods for the determination of eas , fig .
[ fig : e_nip](bottom ) shows the tangents at 1/2 to the dependence of the total energy on the occupation of the homo of the negative ions f@xmath26 and cl@xmath26 . the exact tangent that reproduces the experimental eas is also shown in the figure .
again , as in the case of ips , the der method not only looks like a reasonable approximation but it also improves on the results of the def method ( for the functionals considered and in all cases studied in the present work ) . in particular , for the case of cl@xmath26 [ see fig .
[ fig : e_nip](bottom ) ] the tangents at 1/2 are in very good agreement with the exact tangent that reproduces experimental ea , although the dependence of the total energy on the homo occupation number deviates significantly from linearity . a striking example of the pathological behavior mentioned in sec .
[ sec : method ] is the negative ion f@xmath26 .
this system is found experimentally to be energetically lower than the neutral f atom by 3.34 ev ( see table [ tab : ea ] ) .
all functionals underestimate this energy substantially ( see supplementary material @xcite ) , as a result of the enhanced variational freedom of the open - shell , neutral f atom compared to the closed - shell f@xmath26 . in two extreme cases ( ml and ac3 functionals )
f@xmath26 is found energetically above the neutral f atom .
in summary we examined the performance of the derivative - method proposed in ref . [ ] for calculation of the ips and eas of finite systems .
the accuracy of ips and eas calculated using the derivative - method are compared to the ips and eas calculated using the definition of these quantities [ which involves two total energy minimizations for the system and the positive ( for ip ) or negative ion ( for ea ) ] . in order to have a complete analysis we have also considered an intermediate method ( difference method ) ,
in which the ips and eas are determined by the difference of the total energies with fixing one occupation number to 1 and/or zero .
all these results are further compared to the state - of - the - art ci method as well as experiments .
we find that , for ips both the difference and derivative methods are good approximations to the definition of this quantity .
furthermore , it was found that the derivative method results , obtained using mller , power , and ml functionals , are better than the values obtained using the definition method itself ( with errors of the order of 4 - 8% only ) . among these functionals
the ml functional in conjunction with derivative method is most accurate with errors of only up to 2% . for the eas
the errors are significantly larger , with the ml functional in conjunction with the derivative method being the most accurate ( with an error of 21% ) .
the errors in eas were found to be comparable to the errors in the ci results . from the present study
we conclude that the results of the derivative method for ips and eas are in good agreement with experiments and this method is a promising technique to obtain the single - electron spectrum for systems where state - of - the - art quantum chemical methods can not be applied and dft results deviate significantly from experiment .
|
in the recent work of s. sharma _
et al . _
, ( arxiv.org : cond - matt/0912.1118 ) , a single - electron spectrum associated with the natural orbitals was defined as the derivative of the total energy with respect to the occupation numbers at half filling for the orbital of interest .
this idea reproduces the bands of various periodic systems using the appropriate functional quite accurately . in the present work
we apply this approximation to the calculation of the ionization potentials and electron affinities of molecular systems using various functionals within the reduced density - matrix functional theory .
we demonstrate that this approximation is very successful in general and in particular for certain functionals it performs better than the direct determination of the ionization potentials and electron affinities through the calculation of positive and negative ions respectively .
the reason for this is identified to be the inaccuracy that arises from different handling of the open- and closed - shell systems .
|
in the spring of 2013 , a novel avian influenza , h7n9 virus , has broken out @xcite .
this new type of influenza has affected at least 132 people ( the total reported human cases until june 30 , 2013 ) and resulted in a huge social panic and billions dollars economic losses , even though there is no evidence of human - to - human transmission has been found so far @xcite . and also , several virological researches indicated that h7n9 virus has potential to be a novel disease with strong human - to - human infection @xcite . comparing with other avian influenzas without the capability of sustained human - to - human transmission ( e.g. h5n1 ) ,
a notable difference between them is that the birds symptom of h7n9 usually is inapparent @xcite , even though the spreading of h7n9 virus in birds and poultry is quite fast .
some typical warning signals for the outbreak of avian influenza , such as the death of large number of wild and domestic birds , can not be observed in the process of the h7n9 s spreading .
it causes many serious troubles in the monitor of the epidemic situation and the estimation of public health risk on h7n9 , due to the fact that usually the spreading within birds and poultry is not visible until the outbreak of human cases .
people have tried several measures , e.g. the wider virus test for wild birds and poultry in order to detect the spreading situation .
however , these measures usually are costly to be taken . due to the absence of the capability of sustained human - to - human transmission , human h7n9 cases
generally are occurred dispersedly in the spreading background of bird cases , and the region with higher epidemic level on bird cases usually has a higher risk to occur the bird - to - human infection .
therefore , we may has a chance to find out some useful information which is related to outbreak patterns of human cases . in this paper , based on the empirical analysis for the spatio - temporal patterns of the human cases of h7n9and the association with the comparison of modeling results , we find that , the spatio - temporal network that connects each of two nearest human cases along the order of outbreak timing can efficiently reflect the pattern of the macro spreading situation .
we collect the spatio - temporal information of all the reported human cases shown in the real - time reports of chinese center for disease control and prevention ( http://www.chinacdc.cn/jkzt/crb/rgrgzbxqlg_5295/rgrqlgyp/ ) during the term from feb .
19 , 2013 ( the onset date of the first human case ) , to april 30 , 2013 .
the total time - length of our recording is 70 days , which covers the term from the first onset of human case to the almost end of intensive outbreaks . in this term , the total of recorded human cases and cities with human cases respectively are 128 and 36 . in may and june of 2013 ,
four human cases are reported , however , not included in the dateset , because they occurred sporadically and out of the main period of disease outbreaks .
for each case , the dataset records the location ( the living city of each human case ) and the date of symptom onset of the case .
the total of cases grows exponentially before april 15 , 2013 ( fig .
[ growth](a ) ) , as well as the number of the infected cities ( fig .
[ growth](b ) ) . after the peak in the middle of april ( the inset of fig .
[ growth](a ) ) , the growth on both total of cases and infected cities tends to slow . besides the patterns of growth , the spatial distributions of cases are also investigated .
we find that the total of human cases in different cities matches power - law - like form in _
zipf s ranking plot _ ( see fig .
[ growth](c ) and its inset ) , indicating the heterogeneity in the outbreak of the new avian influenza for different regions : most of cases intensively outbreak in the region of a few cities , and most of cities only have one or two cases .
the introduction of zipf s ranking plot can be found in _
appendix a_. of the inset is the normalized total cases in each city , and the coordinate @xmath0 of the main panel is the cumulative number of @xmath1 , namely , @xmath2 , and here @xmath3 is the total of infected cities . the dashed line in each panel shows the fitting exponent of each curve .
, title="fig:",width=510 ] + moreover , both of the exponential growth and scaling regional distributions are the typical features in the early outbreaks of many infectious diseases ( e.g. sars , the pandemic h1n1 in 2009 and the bird cases of avian influenza ( h5n1))@xcite , indicating the outbreak of human cases of h7n9 shows some similarities to infectious diseases , even though it has no the capability of sustained human - to - human transmission .
since birds - to - human transmission is the major way for human cases , the spreading - like patterns would reflect that the spreading mode of h7n9 in birds and poultry is similar to the one of some human infective diseases . due to the large difference between the organization of human society and avian ecosystem
, the possibility that this similarity is resulted from human - activity - driven spreading ( e.g. poultry trade and transportation ) in birds / poultry can not be ruled out .
for reasons of the difficulties in the wide detection of bird cases of h7n9 virus , the method that digs out useful information from the outbreak patterns of human cases would be of great value to the estimation of spreading situation of h7n9 .
nevertheless , the sparsity of original data on human cases also rises the uncertainty and difficulty in the estimation by using the traditional methods ( e.g. spatial density analysis ) .
we therefore have to find a novel way to analyze the outbreak patterns of human cases . as the discussions below
, the spatio - temporal network can be used here . for a long time , network - based description
is risen to be an important type of basic framework in the researches of epidemic spreading @xcite .
the nodes of the network usually is to be the unit ( e.g. individuals , cities , populations ) in the spreading process , and the edges can express the paths that can transmit epidemic .
most of these network - based studies run the spreading process on networks , and few works use the inverse method that detect the real - world spreading patterns by using spatio - temporal networks , as we will do in this paper .
this type of inverse method is firstly proposed by small et al @xcite for the onset of bird cases of avian influenza ( h5n1 ) .
the basic principle in the construction of the spatio - temporal networks is : to link each pair of outbreak locations that the distance shorter than a threshold relating to the time difference of the two outbreaks .
obviously , these links denote possible path in spreading , and the topology of the network reflects the basic infection relationships for different locations in the outbreak . because the absence of direct transmission between human cases , our analysis focuses on the detection of higher - risk areas , and small s method
can not be used here directly . generally speaking ,
if we connect each infected city to the nearest one , the shorter links will more frequently appear in the areas with higher onset risks ( the density of infected cities is higher ) . also considering the onset time of different cities , our method in the construction of the network
is : to connect the new city with human cases to the nearest one from the cities that have broken out human cases before ( so called _ connecting nearest - existing - node network _ , and cnenn for short ) .
the topology of cnenn therefore must be tree - like , as shown in fig .
[ spatial](a ) . in a sense cnenn can be treated as the minimum spanning tree that connects the infected cities in space - time . noticed that the links in cnenn do not denote the possible path in virus transmission ,
nevertheless , they reflect both the temporal relationships on onset time and the geographical density of infected cities . and shown in fig .
[ spatial](b ) , the edge lengths of cnenn of infected cities in the zipf s ranking plot generally match the two - section power law : @xmath4 where @xmath5 is the rank of edge lengths by descending order , and @xmath6 .
the edges in the two sections are shown in fig .
[ spatial](a ) respectively using blue and red lines , in which the short - range edges usually gather into several small fields that are connected by the long - range edges .
the transition on the two section implies the factors driving the outbreaks in these small fields would be different to the one in the global level . and @xmath7 ( @xmath6 ) .
( b ) the two - section edge - length distribution of the real - world cnenn of h7n9 human cases in zipf s ranking plot .
( c ) the edge - length distributions of modeling cnenn in zipf s ranking plot for three situations on risk distribution : the homogenous case ( blue cycles ) , the heterogeneous situation ( green rectangles ) , and the mixed scenario ( the pink rectangles ) .
each of them runs 100 time steps ( @xmath8 ) , and all the results averaged over @xmath9 independent runs . to be comparable , the average edge length is fixed for different settings of the model .
the gray and red lines respectively denote the power law with zipf s exponent @xmath10 and @xmath11 in the empirical findings .
, title="fig:",width=298 ] and @xmath7 ( @xmath6 ) .
( b ) the two - section edge - length distribution of the real - world cnenn of h7n9 human cases in zipf s ranking plot .
( c ) the edge - length distributions of modeling cnenn in zipf s ranking plot for three situations on risk distribution : the homogenous case ( blue cycles ) , the heterogeneous situation ( green rectangles ) , and the mixed scenario ( the pink rectangles ) .
each of them runs 100 time steps ( @xmath8 ) , and all the results averaged over @xmath9 independent runs . to be comparable , the average edge length is fixed for different settings of the model .
the gray and red lines respectively denote the power law with zipf s exponent @xmath10 and @xmath11 in the empirical findings .
, title="fig:",width=211 ] + ) and the situation with homogeneous risk . to be comparable ,
the average edge length is fixed for different settings of the model .
( b ) the edge - length distributions of cnenn of the mixed scenario for different @xmath12 and different @xmath13 ( the inset ) when @xmath14 .
the gray and red dashed lines in the two panels respectively denote the power law with zipf s exponent @xmath10 and @xmath11 in the empirical findings .
all the results averaged over @xmath9 independent simulations , and each simulation runs 100 time steps ( @xmath8 ) .
, title="fig:",width=510 ] + to understand the origin of the two - section power law of cnenn , the empirical results are compared with the cnenn that is constructed by a simple model . in this model , each location ( with coordinate @xmath15 , say ) in a two - dimension area has a given risk value @xmath16 to breakout cases .
initially ( @xmath17 ) , none of locations have cases . at each time step ( @xmath18 ) , a new location ( @xmath15 , say)that is randomly chosen with probability @xmath16 , breakouts cases . for @xmath19 ,
the new location creates an edge to connect the nearest one of the existed locations with cases .
after @xmath20 time steps evolution , we have the cnenn of the modeling system that connects all the locations with cases .
the most simple situation of the model is that , all the locations have the same risk value @xmath21 . in this situation , we must set a size @xmath22 of the area .
this situation is equivalent to the case that each location with cases is randomly placed on the area one by one
. numerical simulations and analytical results ( see _ appendix b _ ) indicate that the edge - length distribution of the modeling cnenn in the homogeneous situation obeys the power law with zipf s exponent @xmath10 in zipf s ranking plot ( fig .
[ spatial](c ) ) , which is same to the front section in fig .
[ spatial](b ) , implying this section relates to the homogeneous nature in the risk of outbreak in the global level .
that is to say , it reflects the background risk pervading in most regions of east china . considering the zipf s law in regional distributions of human cases ( fig .
[ growth](c ) ) , another situation of the model is assumed as that the spatial distribution of the risk @xmath21 is heterogeneous : @xmath23 , where the parameter @xmath24 presents the heterogeneity of risk ( small @xmath24 corresponds to large heterogeneity ) . as shown in fig .
[ model](a ) , the edge - length distributions of the cnenn in the heterogeneous situation also obey power law in zipf s plot , and the fitting zipf s exponent @xmath25 reduces along the growth of @xmath24 ( reduce on heterogeneity of risk ) and trends to @xmath10 .
the zipf s exponent @xmath11 in empirical findings corresponds to the situation @xmath26 ( see fig .
[ model](a ) ) , showing a quantified real - world risk heterogeneity .
a natural inference from above discussions is that , the origin of the observed two - section power law would relate to the co - effect of the above two situations . to insure it ,
the scenario mixing with both of the two situations is considered . in the mixed scenario , on the @xmath22 area with homogeneous risk , we randomly place @xmath13 islands " , and the radius of each island is @xmath27 ( @xmath28 ) , where @xmath12 is the relative radius of islands .
the risk distribution inside each of the islands obeys the heterogeneous form @xmath29 , where @xmath30 is the position of the center of the island , and we set the risk @xmath16 is higher than the one outside these islands . here the overlap of several islands is allowed , and the risk in the overlapping area is the total of the corresponding risks in different islands .
for example , for the position @xmath15 in the overlapping area of @xmath31 islands , its risk is @xmath32 , where @xmath33 is the center of the i - th island .
simulations find that , the cnenn created by the mixed model shows two - section power - law - like edge - length distributions ( fig .
[ model](b ) ) .
the front section with zipf s exponent @xmath10 extends along the decay on the relative radius of islands @xmath12 and the growth on the number of islands @xmath13 .
when @xmath14 , the result can well fit the empirical findings ( fig .
[ spatial](c ) and fig .
[ model](b ) ) . above all ,
the empirical findings and the modeling results show us the global outbreak picture on the infection risk of h7n9 virus : several islands with heterogeneous outbreak risk straggle in a large area with lower background risk in east china .
these islands are the regions that have higher risk and need focus on in the control of the novel avian influenza ( h7n9 ) . due to
the transition on the edge - length distribution of cnenn is @xmath6 , @xmath34 would close to the upper limitation on the range of these islands .
using the dataset that only contains the information of 128 human cases and 36 cities with cases , the above empirical and modeling results uncover several basic properties in the outbreak of the novel influenza ( h7n9 ) .
firstly , several basic spatial and temporal characteristics in the growth of human cases , including the exponential growth of total human cases and total cities with human cases , and the regional distributions of human cases , are mush similar to that of several infective diseases , reflecting the fast spread of the virus in birds and poultry , and implying the possible human - activity - driven mechanism in the spreading of h7n9 virus in birds and poultry .
more importantly , on the discussion of the two - section power law of the edge - length distribution of cnenn of cities with human cases , the picture that several islands with higher and heterogeneous risk in the global spatial patterns of h7n9 are found out , even with the maximum range of the high - risk islands can be quantitatively estimated , indicating that the method using cnenn to detect the global spreading situation efficiently works in facing of the small - size dataset .
the core of cnenn method is the transformation from the density distribution of cities with human cases to the edge - length distribution of cnenn , which indeed introduces an averaged effect in the analysis : that is why the method is efficient here , although it still can not accurately locate the peak with the highest risk in each island .
it also implies that the cnenn - like methods can be extended to be useful in the analysis of other systems using small - size datasets . in summary
, the cnenn method would be much useful in the detection and surveillance of spreading situation under the paucity of available data that we usually have to face in the early outbreak of many diseases .
this work was funded by the national natural science foundation of china ( no . 11205040 , 11105024 , 11305043 , 11275186 , 10975126 ) , and the zhejiang provincial natural science foundation of china under the grant no .
ly12a05003 , and the research startup fund of hangzhou normal university .
_ zipf s ranking plot _ is widely used in the statistical analysis of the small - size sample @xcite , which can be obtained by first rearranging the data by decreasing order and then plotting the value of each data point versus its rank .
actually , zipf s ranking plot shows the inverse form of the cumulative distribution of the data .
_ zipf s law _
describes a scaling relation in zipf s ranking plot , @xmath35 , between the value of data point @xmath36 and its rank @xmath5 , where the exponent @xmath37 usually is called zipf s exponent " . as a signature of complex systems ,
the zipf s law is widely observed in much social and natural systems @xcite .
it corresponds to a power - law probability density function @xmath38 with @xmath39 @xcite .
for the situation with homogeneous risk , an equivalent model is that , we randomly place a location with cases in the area at each time step .
assuming there has been @xmath40 locations appear cases in a square area with size @xmath22 , for the @xmath41-th location , it should choose the nearest one from the @xmath40 existed locations to create a new edge ( the length is @xmath42 ) .
so the probability density of the relative length @xmath43 of the new edge is equal to the probability that appears such situation : @xmath44 existed locations are placed outside the ring centered by the new location with radius @xmath42 , and one existed location is exactly right on the ring . ignoring the boundary effects of the square
, we easily write out the normalized probability density of @xmath45 @xcite , @xmath46 where @xmath47 is the normalization factor of the distribution . from @xmath48
the maximum likelihood edge - length is @xmath49^{-1/2}.\ ] ] for large @xmath41 , @xmath50 . due to the value
@xmath41 denotes the current number of locations with cases , @xmath51 reduces along the growth of @xmath41 , and thus @xmath41 approximately corresponds to the rank of edge - length . accordingly
, we have the edge - length distribution of cnenn in zipf s ranking plot @xmath52 , as shown in fig . [ spatial](c ) , in agreement with the numerical simulations
. therefore the zipf s exponent @xmath10 of cnenn corresponds to the situation with homogeneous risk .
y. chen , w. liang , s. yang , et al .
human infections with the emerging avian infl uenza a h7n9 virus from wet market poultry : clinical analysis and characterisation of viral genome .
lancet 381 , ( 2013 ) 1916 - 1925 .
q. zhuang , s. wang , m. wu , et al .
epidemiological and risk analysis of the h7n9 subtype influenza outbreak in china at its early stage .
bulletin , ( 2013 ) doi : 10.1007/s11434 - 013 - 5880 - 5 .
d. m. skowronski , n. z. janjua , t. l. kwindt , g. de serres .
virus - host interactions and the unusual age and sex distribution of human cases of influenza a(h7n9 ) in china , april 2013 .
euro surveill .
18 , ( 2013 ) 20465 .
r. gao , b. cao , y. hu , et al .
human infection with a novel avian - origin influenza a ( h7n9 ) virus .
& j. med . 368 , ( 2013 ) 1888 - 1897 .
t. m. uyeki , n. j. cox .
global concerns regarding novel influenza a ( h7n9 ) virus infections .
368 , ( 2013 ) 1862 - 1863 . c. xu , f. havers , l. wang , et al .
monitoring avian influenza a(h7n9 ) virus through national influenza - like illness surveillance , china .
( 2013 ) doi : 10.3201/eid1908.130662 .
b. j. cowling , l. jin , e. h. y. lau , et al .
comparative epidemiology of human infections with avian influenza a h7n9 and h5n1 viruses in china : a population - based study of laboratory - confirmed cases .
lancet 382 , ( 2013 ) 129 - 137 .
q. li , l. zhou , m. zhou , et al .
epidemiology of the avian influenza a ( h7n9 ) outbreak in china .
( 2013 ) doi : 10.1056/nejmoa1304617 .
h. zhu , d. wang , d. j. kelvin , et al .
infectivity , transmission , and pathology of human - isolated h7n9 influenza virus in ferrets and pigs .
science 341 , ( 2013 ) 183 - 186 .
q. zhang , j. shi , g. deng , et al .
h7n9 influenza viruses are transmissible in ferrets by respiratory droplet .
science ( 2013 ) doi : 10.1126/science.1240532 .
han , b. -h .
wang , c. zhou , t. zhou , j. -f .
zhu . scaling in the global spreading patterns of pandemic in uenza a ( h1n1 ) and the role of control : empirical statistics and modeling , arxiv:0912.1390 .
l. wang , x. li , y. q. zhang , y. zhang , k. zhang . evolution of scaling emergence in large scale spatial epidemic spreading .
plos one 6(2011 ) e21197 . c. moore , m. e. j. newman .
epidemics and percolation in small - world networks .
e 61 , ( 2000 ) 5678 - 5682 . m. e. j. newman .
spread of epidemic disease on networks .
e 66 , ( 2002 ) 016128 .
r. pastor - satorras , a. vespignani . epidemic spreading in scale - free networks .
86 , ( 2001 ) 3200 - 3203 .
y. moreno , a. vzquez .
disease spreading in structured scale - free networks .
j. b 31 , ( 2003 ) 265 - 271 .
v. colizza , a. barrat , m. barth lemy , a. vespignani .
the role of the airline transportation network in the prediction and predictability of global epidemics .
103 , ( 2006 ) 2015 - 2019 .
d. balcan , v. colizza , b. goncalves , h. hu , j. j. ramasco , a. vespignani .
multiscale mobility networks and the spatial spreading of infectious diseases .
usa 106 , ( 2009 ) 21484 - 21489 .
t. gross t , c. j. d. dlima , b. blasius .
epidemic dynamics on an adaptive network .
96 , ( 2006 ) 208701
. g. k. zipf .
_ human behaviour and the principle of least effort : an introduction to human ecology _ ( addison - wesly , cambridge , 1949 ) . m. e. j. newman
. power laws , pareto distributions and zipf s law .
contemporary phys .
46 , ( 2005 ) 323 - 351 .
a. clauset , c. r. shalizi , m. e. j. newman .
power - law distributions in empirical data .
51 , ( 2009 ) 661 - 703 .
l. l , z. -k .
zhang , t. zhou .
zipf s law leads to heaps law : analyzing their relation in finitesize systems .
plos one 5 , ( 2010 ) e14139 .
l. l , z. -k .
zhang , t. zhou .
deviation of zipf s and heaps laws in human languages with limited dictionary sizes .
rep . 3 , ( 2013 ) 1082 . k. gong , m. tang , m. -s .
shang , t. zhou . empirical study on spatiotemporal evolution of online public opinion .
acta phys . sin .
61 , ( 2012 ) 098901 .
|
the attack of novel avian influenza ( h7n9 ) in east china caused a serious health crisis and public panic . in this paper , we empirically analyze the onset patterns of human cases of the novel avian influenza and observe several spatial and temporal properties that are similar to other infective diseases . more deeply , using the empirical analysis and modeling studies , we find that the spatio - temporal network that connects the cities with human cases along the order of outbreak timing emerges two - section - power - law edge - length distribution , indicating the picture that several islands with higher and heterogeneous risk straggle in east china .
the proposed method is applicable to the analysis on the spreading situation in early stage of disease outbreak using quite limited dataset .
avian influenza ( h7n9 ) , outbreak patterns , connecting nearest - existing - node network , edge - length distribution , two - section power law + 89.75.fb , 05.40.fb , 89.75.da
|
gravity makes objects cluster .
therefore , the motions of objects towards each other may provide information about the background cosmology .
of course , different subsets of the clustering particles may trace the underlying streaming motions differently .
the scale dependence of the mean streaming @xmath0 of dark matter particles has been understood for some time now ( hamilton et al .
1991 ; nityananda & padmanabhan 1994 ) .
but there has been little study of how this statistic depends on trace - particle type . to do this ,
we build a model in which gravitational clustering is viewed as the combination of two processes .
the first arises from the fact that gravity causes matter to stream towards local minima of the gravitational potential .
this requires a model of how matter which was initially distributed rather smoothly around the centre of collapse becomes redistributed into a more centrally concentrated density profile as the collapse proceeds .
the second process is that these centres around which local collapses are occurring , these clusters , are themselves moving towards each other : clusters cluster .
it is the combination of these two types of motions which gives rise to the spatial distribution and streaming motions of objects today .
section [ v12old ] summarizes useful results which follow from linear theory .
section [ 2halo ] shows how the streaming motions of collapsed dark matter haloes depend on halo mass .
section [ dm ] uses this to model the streaming motions of particles , rather than haloes .
it shows what fraction of a particle s streaming motion arises from the motion of its parent halo , and what fraction must arise from motions within the halo .
these smaller scale motions are essentially a consequence of the collapse around the halo centre we referred to earlier .
it then presents measurements from numerical simulations which show that the model predictions are reasonably accurate .
it also shows that the model provides a reasonable description of how the second moment of the pairwise velocity distribution of the dark matter differs from that of haloes .
section [ gals ] shows how to extend the model to study the mean streaming motions and the pairwise velocity dispersion of galaxies and presents measurements from semianalytic galaxy formation simulations which show that the model predictions are reasonably accurate .
section [ discuss ] discusses what this model implies if one wishes to use measurements of the streaming motions of galaxies to make inferences about cosmology .
we will begin by reviewing the strategy which led to the derivation of how @xmath0 depends on scale .
the relevant starting equation is the pair conservation equation in peebles book ( peebles 1980 ) , but we will start with the equation in the form presented by nityananda & padmanabhan ( 1994 ) : @xmath1 \label{pairs}\ ] ] where @xmath2 is the volume averaged correlation function on proper ( rather than comoving ) scale @xmath3 at the time when the expansion factor is @xmath4 , and the hubble constant is @xmath5 .
this says that if we know the correlation function for all scales @xmath3 and all times @xmath4 , then the assumption that the number of pairs is conserved allows us to compute how @xmath6 depends on scale today .
an approximate solution to this expression can be got as follows ( peebles 1980 ) .
assume that @xmath7 evolves according to linear theory : @xmath8 ^ 2 \bar\xi(r , a_0)$ ] , where @xmath9 is the linear theory growth factor at @xmath4 , and @xmath10 is the growth factor at the present time when @xmath11 .
in an einstein de - sitter cosmology , @xmath12 . then the left hand side is @xmath13 , where @xmath14 .
so , in this approximation we get @xmath15 on large scales , @xmath16 , and so this is just the usual linear theory expression with an extra factor of @xmath17 in the denominator . while this approximation is fine on large scales ( @xmath18 mpc@xmath19 )
, it underestimates the exact solution by a factor of 3/2 or so on smaller scales ( juszkiewicz , springel & durrer 1998 ; sheth et al .
2000 ) .
hamilton et al .
( 1991 ) showed they could compute a good estimate of the evolution of @xmath20 , if the initial correlation function is known ( also see nityananda & padmanabhan 1994 ) .
hamilton et al .
also showed that by inserting their expression for the evolution of @xmath20 into equation ( [ pairs ] ) above , they were able to describe the shape of @xmath0 well on all scales .
while this approach is very useful for studying the statistics of dark matter particles , it is not obvious that it can be used to estimate the streaming motions of galaxies .
this is because one usually assumes that galaxies form at different times .
this means that the number of galaxies is not conserved , so the number of galaxy pairs is not conserved .
this means , for example , that the correlation function of galaxies refers to different sets of particles at different times .
therefore , there is little reason to expect that inserting the correlation function of galaxies into the pair conservation equation should provide a good estimate of @xmath21 today .
we show below that , provided one makes the correct choice of what one uses for @xmath22 , the pair conservation equation can be used to provide an accurate estimate of the streaming motions of galaxies .
this subsection is concerned with the first moment of the pairwise velocity distribution of haloes identified at the present time .
every halo will be represented by one particle , say , the one at the halo centre of mass today .
imagine tracing these centre - of - mass particles back in time . by definition
the number of these particles is conserved , since all we re doing is following them back to high redshift .
of course , at high redshift , few if any of the haloes would actually have collapsed around these centre - of - mass particles . nevertheless , we will use the motions of these particles to represent the motions of the halo centre of mass .
peebles pair conservation equation , combined with the assumption that the motion of a halo today is the same as that of its associated centre - of - mass particle , says that if we knew @xmath20 for these tracer particles , then we can compute @xmath23 today .
so , to compute @xmath24 , we are stuck with the problem of studying the spatial distribution ( i.e. , the bias factor ) of a special marked set of particles at earlier times .
the case in which the marked particles ( in this case , the halo centres - of - mass ) are observed at a later epoch than when they were marked is familiar : e.g. this is like the mo & white ( 1996 ) simple model for galaxies , in which galaxies formed in haloes at @xmath25 but we only observe them today . here , we are interested in the spatial distribution of the special particles at earlier epochs than when they were marked . the halo centre - of - mass particles are biased tracers of the dark matter distribution .
the large scale bias factor is the square root of the ratio of the correlation function of these particles to that of the dark matter correlation function on large scales .
it depends on halo mass : @xmath26 and a similar equality holds for @xmath27 . here
@xmath28 denotes the initial correlation function of the dark matter extrapolated using linear theory to the present time . in equation ( [ xihhb ] )
we use the linearly extrapolated @xmath28 rather than the present day @xmath29 as a practical way of taking into account the volume exclusion effects of haloes at small scales . see section 2 in sheth et al .
( 2000 ) for a detailed discussion . to a good approximation , @xmath30 where @xmath31 is a function which increases with decreasing halo mass , and
@xmath9 and @xmath10 were defined earlier . at the present time , @xmath32 and
this is the familiar eulerian bias formula from mo & white ( 1996 ) .
the lagrangian bias factor is usually expressed as the ratio of @xmath33 at the initial time to the linearly extrapolated @xmath34 .
this means that the lagrangian bias factor is @xmath35 where @xmath36 denotes the expansion factor at the initial time . since @xmath37 , @xmath38 , which is another familiar expression from mo & white ( 1996 ) .
so , in this approximation , @xmath39 .
\label{ba}\ ] ] it is straightforward to insert these expressions for the halo correlation function and its evolution into the pair conservation formula ( equation [ pairs ] ) to see how different @xmath24 is from @xmath40 .
if we study the streaming motions of haloes of two different masses , then we must replace @xmath41 .
this gives @xmath42 \over [ 1 + b_1b_2\,\xi^{\rm lin}_{\rm dm}(r ) ] } \nonumber\\ & & \ - { f(\omega)\bar\xi^{\rm lin}_{\rm dm}(r)\over 3 } { [ b_1(1-b_2 ) + b_2(1-b_1)]\over[1 + b_1b_2\,\xi^{\rm lin}_{\rm dm}(r)]}. \label{v12h}\end{aligned}\ ] ] if we insert the linear evolution approximation for the relation between the correlation function and @xmath43 ( equation [ v12lin ] ) , then this becomes @xmath44 notice that when @xmath45 , then @xmath46 .
also , in the large separation ( small @xmath7 ) limit , @xmath47 times @xmath0 .
so , on average and on large separations , relative to the dark matter , massive haloes @xmath48 stream towards each other whereas less massive haloes @xmath49 stream away from each other .
this makes some physical sense ; clusters cluster , so they are moving towards each other , whereas smaller clumps are in or at the edges of expanding voids , so they are separating from each other .
this linear bias of the streaming velocities at large separation is consistent with the linear theory analysis of fisher et al .
( 1994 ) .
notice that @xmath43 scales with the sum of the bias factors .
if one ignored the evolution of the bias factor when using equation ( [ v12lin ] ) , one would have concluded that the scaling was with the product of the bias factors including the evolution of the bias factor is essential to getting the correct answer .
finally , notice that on smaller scales where @xmath50 , this analysis suggests that @xmath43 of less massive haloes should be larger than that of the dark matter , with the opposite trend being true for massive haloes .
of course , the linear theory and linear evolution approximations we used to obtain equation ( [ v12hlin ] ) are not accurate on small scales .
nevertheless , this provides at least some indication of the small scale behaviour of the halo streaming motions .
[ gifhaloes ] compares this model with measurements in the scdm and @xmath51cdm gif ( kauffmann et al .
1999 ) simulations which were run by the virgo collaboration ( jenkins et al . 1999 ) and are now available to the public .
the open circles show the streaming motions of all the haloes with @xmath52 and @xmath53 in the scdm and the @xmath51cdm simulations , and the solid line shows what our model predicts .
specifically , it shows @xmath54\over \bar n^2_{\rm halo}[1 + b^2_{\rm halo}\xi^{\rm lin}_{\rm dm}(r)]}\nonumber \\ & = & { 2f(\omega)\over 3 } { b_{\rm halo}\bar\xi^{\rm lin}_{\rm dm}(r)\over 1 + b^2_{\rm halo}\xi^{\rm lin}_{\rm dm}(r)}. \label{v12hav}\end{aligned}\ ] ] where @xmath55 is the average number density of haloes , @xmath56 is their average bias factor , the weighting factor in the second line is the ratio of the number of @xmath57 and @xmath58 halo pairs at @xmath3 to the total number of halo pairs at @xmath3 , and the final expression follows from inserting equation ( [ v12hlin ] ) for @xmath24 and using equation ( [ v12lin ] ) for @xmath40 . our model , which we only expect to be accurate on large scales because our approximation for the halo correlation function , equation ( [ xihhb ] ) , breaks down on small scales , is reasonably accurate down to scales of order a mpc@xmath19 or so .
for comparison , the dotted curves show the hubble velocity .
the crosses show the streaming motions of the dark matter particles in the simulations , and the dashed curve shows the prediction associated with the model described in the next section .
the large scale net streaming motion of the dark matter can be got from our expression for the halo motions by integrating up the contribution to the streaming motion from pairs in different mass haloes , weighting by the fraction of the total number of pairs which are in such haloes , and weighting by the halo mass function : @xmath59\over 1 + \xi_{\rm dm}(r)}\,{v_{12}^{\rm halo}(r)\over hr}\nonumber\\ & = & -{v_{12}^{\rm dm}(r)\over hr}\,{1 + \xi^{\rm lin}_{\rm dm}(r)\over 1 + \xi_{\rm dm}(r)}. \label{v12twoh}\end{aligned}\ ] ] the final equality follows from inserting equation ( [ v12h ] ) , noting that @xmath60 , and using the fact that the bias factors are defined so that @xmath61 .
we can now make two important points .
the first is that , at large separations , this expression equals @xmath62 ; in this regime the streaming motions of the dark matter particles are entirely due to the fact that the haloes which contain the particles are moving .
moreover , in this regime , @xmath63 , where @xmath64 is got from equation ( [ v12lin ] ) by using the linear theory values of @xmath65 and @xmath7 .
the second is that this expression exactly equals that in sheth et al .
( 2000 ) for the contribution to @xmath6 from particles which are in separate haloes ( see their eq .
this will be important in what follows .
notice that on smaller scales , @xmath66 . in this regime
the halo motions only account for a fraction of @xmath0 .
the remaining contribution to @xmath0 must arise from the streaming motions of pairs in which both particles are in the same halo .
this means that the fact that our model for halo motions is not accurate on small scales will not matter very much for the small scale value of @xmath40 because , on small scales , the fraction of pairs which are in separate haloes , and so are affected by this inaccuracy , is small .
we turn , therefore , to a discussion of the streaming motions of pairs in which both particles are in the same halo .
if haloes are stable , then the streaming motion within a halo exactly cancels the hubble flow : @xmath67 . in this case
, the contribution from pairs which are in the stable haloes equals unity times @xmath68 where @xmath69 denotes the number of pairs at separation @xmath3 which are in the same halo which has mass @xmath70 ; it depends on the density profiles of haloes .
for all halo shapes of interest , this expression approaches unity at very small @xmath3 , because @xmath71 , and @xmath72 on scales which are smaller than a typical halo .
so , if stable clustering is correct , then @xmath73 on small scales .
in fact , the mean pairwise velocity on small scales depends on the low - mass behaviour of @xmath74 and @xmath75in general , there is no guarantee that @xmath74 and @xmath75 will conspire to give stable clustering ( ma & fry 2000 ; sheth et al .
2000 ) .
in particular , section 4 of sheth et al .
( 2000 ) shows that the small scale term is @xmath76 } \nonumber \\ & = & { \partial{\rm ln}\,m_*(a)\over\partial{\rm ln}a } { [ \bar\xi_{\rm 1halo}(r , a ) - \xi_{\rm 1halo}(r , a)]\over 3[1 + \xi(r , a ) ] } , \label{lamseq}\end{aligned}\ ] ] where @xmath77 is proportional to the number of pairs in the same @xmath70-halo which have separation @xmath3 , and @xmath78 is the number density of virialized @xmath70-haloes at time @xmath4 .
( our notation differs slightly from that in sheth et al.they absorbed the two factors of @xmath70 into their definition of @xmath79 , whereas here we have chosen to show these factors explicitly .
hence , our @xmath80 is their @xmath79 . ) in sheth et al . , @xmath78 depended on time , and the halo profile did as well , because virialized haloes were assumed to have profiles of the form given by navarro , frenk & white ( 1997 ) , and these halo shapes depend both on @xmath81 and on the ratio of the average density within the virialized halo to the background density at the time it virialized .
( strictly speaking , the expression above assumes that all of the time dependence of the halo shape can be written as a function of @xmath82see sheth et al . for details . )
the sum of equations ( [ v12twoh ] ) and ( [ lamseq ] ) gives a complete description of the streaming motions of dark matter on all scales .
the dashed curves in fig .
[ gifhaloes ] show that this sum provides a good description of the dark matter streaming motions on all but the smallest scales ( see sheth et al . 2000 for a discussion of the discrepancy at the smallest scales ) .
now , the dashed curves correspond to independently identifying haloes at each epoch . on the other hand
, we could have followed the approach of section 2.2identify or mark the haloes today , and then follow them backward in time . with this latter approach ,
the number density of haloes is fixed to the value today , @xmath83 , but the profile changes from , say a tophat to a more centrally concentrated shape .
equation ( [ v12twoh ] ) shows that the contribution to the streaming motions from particles in separate haloes is independent of the details of how this happens
. however , recall that this contribution exactly equals the two - halo contribution to @xmath84 worked out by sheth et al .
this means that the one - halo contributions to @xmath43 must also be the same in both approaches .
in particular , this means that however the profile changes from a tophat to an nfw shape , it must change in just such a way that the final answer for the streaming motions of the dark matter particles equals equation ( [ lamseq ] ) .
indeed , we can use this requirement to constrain how the profile changes from the initial tophat to the final nfw cusp the appendix shows a worked example of how to do this .
so far , we have shown how the mean streaming motions of the dark matter and the haloes are related .
sheth et al . ( 2000 ) discuss how to do this for the second moment of the pairwise velocity distribution .
they argued that the dark matter particles receive substantial nonlinear kicks to their initial velocities ( essentially , the virial motions within haloes ) , whereas the haloes do not ( sheth & diaferio 2000 ) .
as a result the pairwise dispersion of the dark matter should be significantly larger than that of the haloes on all scales .
[ sig12h ] compares what their equation ( 31 ) predicts with the simulations ( we refer the reader to their paper for details of the model ) .
the open circles and crosses show the pairwise dispersion , @xmath85 , of the haloes and the dark matter respectively , and the dashed and solid curves show the model predictions .
( we do not show @xmath86 for the dark matter in the scdm simulations because the simulation box is sufficiently small [ 85mpc@xmath19 ] that cosmic variance affects the measurement significantly . )
the model is reasonably accurate on large scales , and not accurate on small scales .
sheth et al . ( 2000 ) discuss why this happens for the dark matter ( the inaccuracy is due to the simplifing assumptions that haloes have no substructure , the pairwise dispersion from a single halo is isotropic and independent of the pair position , and the number of pairs in the infall regime around haloes is sufficiently small that the use of virial motions to model the dispersion from infalling pairs does not not lead to a large error ) .
for the haloes , this discrepancy appears on scales which are of the order of a typical @xmath88 halo and smaller .
this suggests that the discrepancy almost surely arises from using linear theory to model the spatial distribution and velocities of haloes on scales which are smaller than the smoothing scale used to make the model prediction .
despite the quantitative discrepancies on small scales , the model is in qualitative agreement with the simulation : the pairwise dispersion of the haloes is substantially smaller than that of the dark matter .
the previous section showed that the first moment of the pairwise velocity distribution of haloes is different from that of the dark matter .
it showed that massive haloes separated by large distances are streaming together more rapidly than less massive haloes at the same separation , and that this difference scaled with one rather than two powers of the halo bias factor .
it also showed that the dark matter statistic was obtained by weighting the halo statistic by the number of dark matter particle pairs per halo .
the second moments of the pairwise velocity distributions are also different . in this case , also , the dark matter statistic is got by weighting by the number of particle pairs per halo . however , the pairwise dispersion is also sensitive to the fact that virial motions within haloes can be substantially higher than the motions of the haloes themselves . as a result ,
the pairwise velocity dispersion of dark matter particles is substantially larger than that of haloes , on all scales .
this section studies what these results imply for the pairwise motions of galaxies .
we will model galaxies as random particles in dark matter haloes .
that is , the motion of the galaxies is the same as that of the dark matter particle with which they are associated . in this sense
there is no velocity bias in our model ; the fact that velocity statistics for the dark matter and the galaxies may , nevertheless , be different , arises solely from the fact that dark matter statistics weight each halo proportional to halo mass , whereas galaxy statistics do not . such models for the difference between the statistics of galaxies and dark matter particles have received considerable attention recently .
seljak ( 2000 ) , peacock & smith ( 2000 ) and scoccimarro et al .
( 2000 ) have used them to model the spatial distribution of galaxies , sheth & diaferio ( 2000 ) describe how to model the distribution function of galaxy peculiar velocites , and sheth et al .
( 2000 ) describe how to use these models to do analytically what jing , mo & brner ( 1998 ) did numerically in their study of the pairwise velocity dispersion of galaxies . within the context of this model ,
galaxies are treated by setting @xmath89 where the two terms denote the contribution to the statistic from galaxies in the same and in different haloes , respectively
. the second term on the right hand side can be got by modifying equation ( [ v12twoh ] ) : @xmath90\over 1 + \xi_{\rm gal}(r)}\ , { v_{12}^{\rm halo}(r)\over hr } \label{v12g}\end{aligned}\ ] ] where @xmath91\nonumber \\ & \equiv & \xi^{\rm 1halo}_{\rm gal}(r ) + b_{\rm gal}^2\,\xi^{\rm lin}_{\rm dm}(r ) .
\label{xig}\end{aligned}\ ] ] here @xmath92 and @xmath93 denote the first and second moments of the distribution of the number of galaxies in @xmath70-haloes , and we set @xmath94 if @xmath95 .
there are details associated with how one treats the central galaxy in a halo , but , for the most part , these amount to a small effect ( see sheth & diaferio 2000 ) , so we have ignored them they add complications but no essential change to the logic of our argument . on scales larger than a few mpc@xmath19 , @xmath96 dominates over the one - halo contribution .
if we assume linear theory for the evolution of the two - halo term ( equation [ v12lin ] ) , then we can set @xmath97 times @xmath98 , and then equation ( [ v12 g ] ) reduces to @xmath99 . \label{v12twog}\ ] ] this shows that , on large scales , the streaming motions of galaxies can be biased relative to the dark matter
. the extent to which they are biased is related to how differently they are clustered , and this , in turn , depends on the @xmath92 relation . on smaller scales ,
the streaming motions are dominated by galaxy pairs in which both members are in the same halo .
a little thought shows that this can be computed simply by setting @xmath100 and @xmath101 in equation ( [ lamseq ] ) .
this is because the @xmath92 relation does not introduce any additional time dependence recall that the number density of haloes in the present model is fixed to the value it has today , and the galaxies are to be thought of simply as marked tracer particles within the haloes .
this has an interesting consequence .
suppose one wishes to use the pair conservation equation ( [ pairs ] ) to estimate @xmath21 .
then the model above suggests that , on small scales , simply inserting the observed galaxy correlation function into equation ( [ pairs ] ) should be reasonably accurate .
however , on larger scales , doing this leads to an estimate of @xmath102 which is incorrect , for the following reason . because there is no time dependence in @xmath92 , the result of doing this has the same time dependence as in the case of the dark matter for which @xmath103 , or the case of the haloes for which @xmath104 .
but we know that , for the haloes , doing this results in the wrong answer , because it neglects the evolution of the bias factor .
the case of galaxies is no different ; if we insert the observed galaxy correlation function into the pair conservation equation , then we are incorrectly neglecting the evolution of the bias factor of the galaxies .
this would lead one to conclude , incorrectly , that @xmath43 should scale as @xmath105 rather than as @xmath106 . to illustrate how our model works
, we will use the @xmath92 relations we obtained from the semianalytic gif @xmath51cdm model of kauffmann et al .
( 1999 ) which are now publically available .
sheth & diaferio ( 2000 ) provide a fitting formula for the @xmath92 relation of a gif galaxy catalog which was constructed by choosing all galaxies brighter than @xmath107 , after accounting for the effects of dust .
we divided that catalog up into two subsamples based on colour ( galaxies labelled as being redder or bluer than @xmath108 ) and on star - formation rate ( rates greater or less than @xmath109yr ) .
[ gifngmh ] shows these relations .
the dashed lines show the following fits : @xmath110 ^ 2}\nonumber \\ & & + ( m_{11}/30)^{0.75}{\rm e}^{-(2/m_{11})^2 } \nonumber\\ n_{\rm blue}(m ) & = & ( m_{11}/500)^{0.8 } + 0.6\,{\rm e}^{-4[\log_{10}(m_{11}/6.2)]^2}\nonumber \\
n_{\rm red}(m ) & = & n_{\rm all}(m ) - n_{\rm blue}(m ) \nonumber \\
n_{\rm hsfr}(m ) & = & 0.015 + ( m_{11}/7000)^{0.9 } + { \rm e}^{-8[\log_{10}(m_{11}/56.2)]^2}\nonumber\\ n_{\rm lsfr}(m ) & = & n_{\rm all}(m ) - n_{\rm hsfr}(m)\end{aligned}\ ] ] where @xmath111 is the halo mass in units of @xmath112 .
the solid lines are the same in both panels ; they show @xmath113 , and they equal the sum of the two dashed lines .
these relations can be used to compute the statistics of one of these galaxy samples , rather than dark matter particles , by setting @xmath92 equal to the appropriate @xmath114 relation .
in addition to these mean @xmath114 relations , our models also require the second moment of the number of galaxies per halo distribution .
we have approximated it by setting @xmath94 if @xmath95 , and @xmath115 where @xmath116^{1/2}$ ] if @xmath117 , and @xmath118 when @xmath92 is larger . because @xmath118 for a poisson distribution , this approximately accounts for the fact that the scatter in galaxy counts is sub - poisson in low mass haloes .
if we use the same galaxy sample that scoccimarro et al .
( 2000 ) did , then our model for the scatter is similar to theirs .
the correlation functions for these four subsamples are shown in fig .
[ xigals ] : crosses , open circles , triangles and squares show @xmath119 for red , blue , star - forming and quiescent galaxies in the gif simulation .
filled circles show the correlation function of the dark matter particles .
notice how similar the correlation function of the blue sample is to that of the star - forming sample , how similar the red and quiescent samples are , and how different the blue and star - forming samples are from the red and quiescent samples .
the two solid curves show the result of using our model to compute the correlation functions of the red and blue samples , because these differ the most from each other .
we did this by setting @xmath92 in equation ( [ xig ] ) equal to the appropriate @xmath114 relation . for comparison ,
the dashed curves show the result of using the second factorial moment , rather than the second moment , when computing the galaxy correlation functions ; the difference only matters on small scales .
the dotted curve shows our calculation of @xmath120 which has @xmath103 .
the bottom panel shows how @xmath121 ( the ratio of the solid and dotted curves ) depends on scale .
both panels show that our model provides a good description of the simulation results . in computing our model predictions
, we assumed that the two samples both trace their parent dark matter haloes similarly .
that is , we used the same function @xmath75 for both the red and the blue samples .
if the red galaxies were more centrally concentrated than the blue , we could have incorporated it into our analysis by adjusting @xmath75 .
in fact , fig . 2 of diaferio et al .
( 1999 ) shows that , in the semianalytic model , the red galaxies in massive clusters are concentrated more towards the centres of their parent halos than the blue ones are .
the agreement between the simulation and our model curves in which we made no such adjustment for this suggests that it must amount to only a weak effect .
we turn , therefore , to the first and second moments of the pairwise velocity distribution for these four subsamples .
[ gifgals ] shows results for the same semianalytic galaxy samples shown in fig .
[ xigals ] . as before ,
crosses , squares , triangles and open circles show galaxies classified as being red , quiescent , star - forming , or blue .
filled circles show the corresponding statistics of the dark matter particles . as for the correlation functions , the blue and star - forming samples are quite different from the red and quiescent samples .
our model shows that this arises simply from the fact that these samples have rather different @xmath114 relations there are only a few blue , star forming galaxies in clusters .
notice that blue galaxies ( circles ) have the smallest streaming motions , and red galaxies ( crosses ) have the largest @xmath43 values .
our model predicts that larger streaming motions at large separations indicate a higher amplitude of clustering on those scales .
comparison with fig .
[ xigals ] shows that the correlation functions of the red galaxies are biased high relative to the dark matter , whereas the blue galaxies are biased low .
this is in qualitative agreement with our model . the solid lines in fig . [ gifgals ]
provide a more quantitative comparison between our model predictions for the red and blue galaxies , and the values of the galaxy velocities measured in the gif simulation .
the model predictions are in reasonable agreement with the simulation , although the agreement is certainly not as good as it was for the correlation functions .
sheth et al . ( 2000 ) discuss the reason for the overestimate in @xmath6 on large scales ( e.g. , these models do not satisfy the integral constraint ) .
the bottom panels show the ratio of the galaxy velocities to those of the dark matter ; i.e. , @xmath122 , and similarly for @xmath123 .
this ratio is scale dependent on smaller scales .
our model describes the scale dependence reasonably well .
we presented a simple model of how the streaming motions of haloes and galaxies depends on separation .
we tested the model using the publically available simulations of kauffmann et al .
in the semi - analytic galaxy formation model , the mean streaming motions depend rather strongly on how the galaxy sample was selected .
for example , blue galaxies have smaller streaming motions than red galaxies .
we showed that our model was able to describe the differences between a wide range of simulated galaxy catalogues rather well ( fig .
[ gifgals ] ) .
our model predicts a very close relationship between the streaming motions of the galaxies and their spatial distribution .
optical and iras galaxies cluster differently ( e.g. marzke et al .
1995 ; fisher et al .
therefore , they must be biased differently relative to the dark matter .
if the streaming motions of optical galaxies are the same as the dark matter , then our model predicts that the streaming motions of iras galaxies must be different from that of the dark matter . in other words , whether or not the streaming motions of a given galaxy sample trace the motions of the underlying dark matter depends on how the sample was selected .
thus , in the absence of strong arguments for why a given galaxy sample is expected to be unbiased , one should be cautious when interpretting measurements of the streaming motions of galaxies .
if the correlation function and the streaming motions of two different galaxy samples have been measured , then the model described here ( equations [ v12gal ] and [ v12twog ] ) says that the square root of the ratio of the two correlation functions ( at , say , 20 mpc/@xmath124 ) should equal the ratio of the streaming motions on the same scale .
this can be used to test the validity of the model .
again , however , caution is required because this relationship is only true on large scales . for the semi - analytic galaxy samples we presented ,
this simple linear biasing was a good approximation on scales larger than about 10 mpc@xmath19 ( although our model is able to describe the scale dependence of this ratio even on smaller scales ) . in this context
, we think it worth noting that the semi - analytic galaxy samples we presented here can not explain the results of juszkiewicz et al .
( 2000 ) who found that ellipticals and spirals have the same value of @xmath43 on separations of about 10 mpc/@xmath124 , even though they estimate that the spatial distributions have bias factors which differ by a factor of two : @xmath125
. it will be interesting if future data sets confirm this .
in addition to studying how the mean streaming velocity depends on galaxy type , our fig . [ gifgals ] also shows that the second moment of the pairwise velocity distribution depends strongly on galaxy type .
for example , our model of the pairwise dispersion suggests that the dispersion of blue galaxies should be substantially smaller than that of red ones ( although this difference depends on the colour cut ) , especially on scales of 1 mpc@xmath19 or so .
thus , it is not surprising that @xmath86 for optical galaxies ( marzke et al .
1995 ) is almost a factor of two larger than for iras galaxies ( fisher et al .
the effects of redshift space distortions are larger if the pairwise dispersion is larger .
this means that , on small scales , the amplitude of the redshift space correlation function of red galaxies should be substantially smaller than the real space correlation function , but the difference between real and redshift space correlation functions of blue galaxies should not be as dramatic .
this is a generic prediction of these sorts of galaxy formation models .
this suggests that galaxy redshift samples cut by colour should provide a useful and direct test of these models .
dividing a galaxy sample by colour allows another simple test of these models . at large separations , where most pairs are in separate haloes , the model described above predicts that the cross - correlation function of the two colour samples should simply be the geometric mean of the two individual samples .
if the two galaxy samples both trace their parent dark matter haloes in the same way ( our fig .
[ xigals ] shows that this assumption describes the semianalytic model well ) , then this will be approximately true even on smaller scales .
( it will not be exactly true because the scatter in the @xmath114 relations are , typically , sub - poisson . )
data sets currently available should be able to test this prediction .
this collaboration was started at the german american young scholars institute in astroparticle physics , at the aspen center for physics in the fall of 1998 , and at ringberg castle in 1999 .
rks thanks the ias , as well as the university and observatory at torino , and lh & rs thank fermilab for hospitality where parts of this work were done .
we all thank the halo pub for inspiring nourishments .
the n - body simulations , halo and galaxy catalogues used in this paper are publically available at http://www.mpa-garching.mpg.de/numcos .
the simulations were carried out at the computer center of the max - planck society in garching and at the epcc in edinburgh , as part of the virgo consortium project .
in addition , we would like to thank the max - planck institut fr astrophysik where some of the computing for this work was done .
rks is supported by the doe and nasa grant nag 5 - 7092 at fermilab .
lh is supported by nasa grant nag5 - 7047 , nsf grant phy-9513835 and the taplin fellowship .
he also thanks the doe for an outstanding junior investigator award ( grant de - fg02 - 92-er40699 ) rs is supported by endowment funds from the ias and by nsf grant phy-0070928 at ias .
99 diaferio a. , kauffmann g. , colberg j. , white s. d. m. , 1999 , mnras , 307 , 537 fisher et al .
1994 , mnras , 267 , 927 hamilton a. j. s. , kumar p. , lu e. , matthews a. , 1991 , apj , 374 , l1 kauffmann g. , colberg j. , diaferio a. , white s. d. m. , 1999 , mnras , 303 , 188 juszkiewicz r. , ferreira p. g. , feldman h. a. , jaffe a. h. , davis m. , 2000 , science , 287 , 109 ma c. p. , fry j. , 2000 , apj , submitted marzke et al .
1995 , aj , 110 , 477 mo h. j. , white s. d. m. , 1996 , mnras , 282 , 347 mo h. j. , jing y. p. , brner g. , 1997 , mnras , 286 , 979 nityananda r. , padmanabhan t. , 1994 , mnras , 271 , 976 peacock j. , smith r. , 2000 , mnras , submitted , astro - ph/0005010 peebles p. j. e. , 1980 , the large scale structure of the universe .
princeton univ .
press , princeton press w. , schechter p. , 1974 ,
apj , 187 , 425 seljak u. , 2000 , mnras , in press , astro - ph/0001493 scoccimarro r. , sheth r. k. , hui l. , jain b. , 2000 , apj , accepted , astro - ph/0006319 sheth r. k. , saslaw w. c. , 1994 , apj , 437 , 35 sheth r. k. , diaferio a. , 2000 , mnras , submitted , astro - ph/0009166 sheth r. k. , hui l. , diaferio a. , scoccimarro r. , 2000 , mnras , submitted , astro - ph/0009167
this appendix shows how the correlation function and the mean streaming motions evolve in our model . the correlation function on comoving scale @xmath126 when the expansion factor is @xmath4 is the sum of two terms : @xmath127 where @xmath128 ^ 2 \label{xidm2h}\ ] ] ( compare equation [ xig ] ) , and @xmath129 for haloes of virial mass @xmath70 which have density profiles of the form given by navarro , frenk & white ( 1997 ) , @xmath130 where @xmath131 is given in appendix a3 of sheth et al .
( 2000 ) . here
@xmath132 is the virial radius of the halo in proper physical coordinates , @xmath133 is the comoving virial radius , and @xmath134 is a parameter which describes how centrally concentrated the halo profile is : it is the ratio of the virial radius to a central core radius , and it depends on the halo mass .
the normalization term is @xmath135^{-1}.\ ] ] the mean streaming motions from pairs in the same halo is given by equation ( [ lamseq ] ) in the main text .
we would like to study what happens if we make the profile of the region containing @xmath70 depend on time in the following way : we would like to use the same family of profiles , such as the nfw set , to describe the density run at any time , and we want to parametrize the evolution of the profile shape by changing the values of the profile s parameters . there is no physical reason why the nfw form should describe the density run around a region which has yet to virialize ; we are only using this to illustrate how our argument works .
for nfw profiles , this means that we will think of @xmath136 as the comoving boundary of the region containing @xmath70 at the time when the expansion factor is @xmath4 : initially @xmath137 , and today @xmath138 .
in addition , we will allow the concentration parameter to depend on mass as well as time : @xmath139 .
the idea is that the density run around the centre of the region of radius @xmath140 containing @xmath70 was presumably differently concentrated at @xmath36 than it is today . in particular , we would like to see if this model can produce the same streaming motions as equation ( [ lamseq ] ) in the main text . to show what is required for this to happen , set @xmath141 , and also set @xmath142 , where @xmath143 is the initial comoving radius of the halo , and @xmath144 is the comoving density of the background .
then the expression for the correlation function becomes @xmath145 the volume integral of this is @xmath146 where we have defined @xmath147 .
so @xmath148.\end{aligned}\ ] ] this will result in the same streaming motions as equation ( [ lamseq ] ) if @xmath149 where @xmath150 is the slope of the linear power spectrum on the scale on which the rms density fluctuation is 1.686 , and @xmath151 , where @xmath9 is the linear theory growth factor .
this sums up what is required of the time dependence of the concentration @xmath152 and the comoving radius @xmath140 containing @xmath70 .
note that these requirements are non - trivial , because both @xmath153 and @xmath154 depend on the mass @xmath70 , but we are requiring that the derivatives work out to be independent of @xmath70 .
we can see what this implies for the evolution of the correlation function if we insert the scalings of equation ( [ constraints ] ) in equation ( [ xixa ] ) : @xmath155 where the subscript ` 0 ' denotes the values of quantities at the present time .
this shows that the number of pairs on comoving scale @xmath126 at the present time is the same as the number of pairs which , at the earlier time when the linear growth factor was @xmath156 were on the larger comoving scale @xmath157 .
the correlation function decreases monotonically with scale , so the expression above implies that @xmath158 was smaller at early times than it is today . at very early times , therefore , the correlation function might plausibly be dominated by the two - halo term ( equation [ xidm2h ] ) .
the evolution of this term can be got from inserting the evolution of the bias factor ( equation [ biasm ] ) into it .
the integrals over @xmath70 can be done analytically , with the result that @xmath159 . at sufficiently early times , this two - halo term dominates on all scales , so our model for the correlation function reduces correctly to the linear theory expression .
our requirement that the same nfw form hold at all times means that the profile shape evolves as @xmath160 where @xmath161 is the comoving distance from the centre in units of the initial comoving scale @xmath143 , whereas @xmath162 is in units of the virial radius at @xmath163 .
the second equality follows from the scalings above for the evolution of @xmath164 and @xmath165 , and setting @xmath166 .
the profile evolves in such a way that the density on scales @xmath167 grows as @xmath4 increases , as we expect . on much smaller scales , @xmath168 , and
the profile shape is more like @xmath169 : on very small scales , the density decreases with time ! because this small scale behaviour of the density seems contrary to our intuition , we thought it worth studying the evolution of the mass as a function of comoving scale .
the scalings above imply that the fraction of the total mass @xmath170 which is in the range d@xmath161 around @xmath161 from the halo centre when the growth factor is @xmath166 is @xmath171\bigr]^2}. \label{dmds}\ ] ] fig .
[ nfwev ] shows an example of how the mass gets redistributed as the profile evolves .
it was constructed by setting @xmath172 , @xmath173 and @xmath174 in equation ( [ dmds ] ) ( the first two values approximate those of an @xmath88 halo in a @xmath51cdm simulation ) .
the solid , dashed , and dot - dashed curves show equation ( [ dmds ] ) at @xmath175 , @xmath176 and @xmath177 respectively . for most of the volume of the halo , the mass in a given comoving shell increases as @xmath178 increases .
only well within the core of the object does it decrease with time .
we only show the shape of the profile out to the radius @xmath140 which contains the mass @xmath70 . at @xmath175
, @xmath140 is the virial radius , which is at @xmath179 ; @xmath140 was larger earlier , so that the total mass contained in the profile remains constant . by @xmath180 or so
@xmath181 , indicating that the model profile must extend beyond @xmath143 if it is to enclose mass @xmath70 ; at this point the model has really broken down .
a similar analysis of the hernquist profile shows the same qualitative features : @xmath182 presumably , this apparently unphysical behaviour is a consequence of our unphysical requirement that the profile have the same functional form at all times .
|
a simple model of how objects of different masses stream towards each other as they cluster gravitationally is described .
the model shows how the mean streaming velocity of dark matter particles is related to the motions of the parent dark matter haloes .
it also provides a reasonably accurate description of how the pairwise velocity dispersion of dark matter particles differs from that of the parent haloes .
the analysis is then extended to describe the streaming motions of galaxies .
this shows explicitly that the streaming motions measured in a given galaxy sample depend on how the sample was selected , and shows how to account for this dependence on sample selection .
in addition , we show that the pairwise dispersion should also depend on sample type .
our model predicts that , on small scales , redshift space distortions should affect red galaxies more strongly than blue .
galaxies : clustering cosmology : theory dark matter .
|
the intracluster medium ( icm ) has long been proposed to play a vital role in galaxy evolution in clusters , through ram pressure and turbulent / viscous stripping of the galactic cold gas ( e.g. , gunn & gott 1972 ; nulsen 1982 ; quilis et al . 2000 ) .
as the halo gas and the cold interstellar medium ( ism ) is depleted in the stripping process , the galactic star formation will eventually be shut down and blue disk galaxies may turn into red galaxies ( e.g. , quilis et al . 2000 ) .
the removal of the cold ism also affects the accretion history of the central smbh .
stripping of the ism of the disk galaxies in clusters has been extensively studied in simulations recently , with better resolution and more physics included ( e.g. , abadi et al .
1999 ; stevens et al . 1999 ; quilis et al . 2000 ; schulz & struck 2001 ; bekki & couch 2003 ; roediger & hensler 2005 ; kapferer et al .
these simulations show that stripping has a significant impact on galaxy evolution ( e.g. , disk truncation , formation of flocculent arms , build - up of a central bulge and enhanced star formation in the inner disk at the early stage of the interaction ) . besides the impact on galaxy evolution , another significant question related to stripping is the evolution of the stripped ism .
after the cold ism is removed from the galaxy , the general wisdom is that the stripped gas is mixed with the icm through evaporation eventually .
however , it is now known that a fraction of the stripped ism turns into new stars in the galactic halo or the intracluster space , as revealed from observations ( e.g. , sun et al .
2007b , s07 hereafter ) and simulations ( e.g. , vollmer et al .
2001b ; schulz & struck 2001 ; kronberger et al . 2008 ; kapferer et al . 2009 ) .
the stripped cold ism , if it can survive long enough to reach the cluster center , can effectively heat the cluster core via ram - pressure drag , which makes gravitational heating by accretion a possible way to offset cooling ( dekel & birnboim 2008 argued for 10@xmath5 - 10@xmath6 m@xmath7 clumps ) .
current hi observations of cluster spiral galaxies still fail to detect most of the hi gas missing from hi - deficient spiral galaxies in the intracluster space ( e.g. , vollmer & huchtmeier 2007 ) , which implies that the bulk of the stripped hi gas has been heated out of the cold phase .
however , little is known about the details of mixing , as the actual strength of heat conductivity and viscosity is poorly known . how quickly is the stripped cold ism heated and mixed with the icm ? what observational signature will evaporation and mixing produce ?
what fraction of the stripped cold ism turns into new stars ?
the mixing of the stripped cold ism with the hot icm will produce multi - phase gas .
depending on the poorly known details of mixing , prominent soft x - ray emission may be produced , as well as h@xmath4 emission . in this `` unified model '' for tails of cluster
late - type galaxies , x - ray and h@xmath4 tails are simply manifestations of the cold ism tail in the mixing process .
therefore , one can better understand mixing and the relevant micro - physics from multi - wavelength data , e.g. , the amount of missing hi gas and the amount of the stripped gas in hotter phases .
obviously , a central problem is the energy transfer in the multi - phase gas , which is also a significant question for large cool cores in clusters .
moreover , the wake behind the galaxy produced by stripping provides a way to constrain the icm viscosity ( e.g. , sun et al .
2006 , s06 hereafter ; roediger & br@xmath8ggen 2008a ) .
various dynamic features , e.g. , bow shocks , tails and vortices , can be produced in stripping ( e.g. , stevens et al .
1999 ; schulz & struck 2001 ; roediger et al . 2006 ; roediger & br@xmath8ggen 2008b , rb08 hereafter ; tonnesen & bryan 2009 ) .
observational evidence of stripping of cluster late - type galaxies is present in hi and h@xmath4 observations , either through hi deficiency or tails ( e.g. , giovanelli & haynes 1985 ; gavazzi et al .
2001b ; kenney et al .
2004 ; oosterloo & van gorkom 2005 ; chung et al . 2007 ; yagi et al . 2007 ; s07 ) .
x - ray tails of late - type cluster galaxies have only been detected recently : c153 in a2125 at @xmath9=0.253 ( wang et al .
2004 ) , and ugc 6697 in a1367 at @xmath9=0.022 ( sun & vikhlinin 2005 ) . however , there are only @xmath0 60 counts from c153 so the x - ray extension is not unambiguous ( wang et al .
the data show that the x - ray tail of ugc 6697 may extend to @xmath0 90 kpc from the nucleus but most of the x - ray emission is within the galaxy .
moreover , the galaxy has a peculiar morphology and a tidal tail .
its complex kinematic behavior , revealed from the velocity map , implies the existence of a second galaxy hidden behind the main galaxy ( gavazzi et al .
thus , ugc 6697 is a very complicated system where tidal interaction is important .
there is also a wide tail ( 67 kpc@xmath1090 kpc ) of the spiral ngc 6872 in the 0.5 kev pavo group ( machacek et al .
2005 ) , but it is uncertain how it was formed as the tail terminates on the dominant early - type galaxy of the group . the brightest known x - ray tail behind a cluster late - type galaxy is in the closest rich cluster a3627 ( s06 ) , with @xmath0 80% of the x - ray emission beyond the galactic halo .
one should be aware that strong x - ray tails of late - type galaxies are rare ( e.g. , none in virgo , see section 7.1 ) .
we have done a blind search of strong x - ray tails in nearby clusters ( @xmath11 ) by extending our previous work ( sun et al .
2007a ) to the current and archives .
despite extensive cluster data in the archives , e.g. , mosaic fields of @xmath0 2 deg@xmath12 around the perseus and coma clusters with , only two more x - ray tails of cluster late - type galaxies were found .
one is also in a3627 ( ) and will be discussed in this paper .
the other one is ngc 4848 in the coma cluster ( finoguenov et al .
2004 ) that was recently observed for 29 ks by
. however , both x - ray tails are shorter ( 40 - 50 kpc ) and 2.4 - 5.5 times fainter than s tail .
thus , the proximity of , the high flux and the large length of its x - ray tail makes it the best target for detailed analysis and comparison with simulations .
one should not confuse x - ray tails of early - type galaxies with those of late - type galaxies .
early - type galaxies in clusters have abundant x - ray ism but little or no cold ism .
there are some x - ray tails of early - type galaxies reported ( e.g. , machacek et al .
2006 ; sun et al . 2007 ; randall et al . 2009 ) .
their x - ray tails are composed of the stripped hot ism and do not co - exist with cold gas .
a3627 is the closest massive cluster ( @xmath9=0.0163 , @xmath13 = 925 km / s and @xmath14=6 kev ) rivaling coma and perseus in mass and galaxy content ( kraan - korteweg et al .
1996 ; woudt et al . 2008 ) .
it is also the sixth brightest cluster in rass and the second brightest non - cool - core cluster after coma ( b@xmath15hringer et al . 1996 ) .
is a blue emission - line galaxy ( woudt et al .
2004 ) that is only @xmath0 180 kpc from the cluster s x - ray peak in projection .
its radial velocity ( 4680@xmath1671 km s@xmath3 , woudt et al .
2004 ) is close to the average velocity in a3627 ( 4871@xmath1654 km s@xmath3 , woudt et al .
2008 ) so most of its motion is probably in the plane of sky .
if is on a radial orbit , its real distance to the cluster center should be close to the projected distance .
there is no sign of a galaxy merger and no strong tidal features around .
s06 found a long x - ray tail behind , in both ( 14 ks ) and data ( mos : 18 ks , pn : 12 ks ) .
the tail extends to at least 70 kpc from the galaxy with a length - to - width ratio of @xmath0 10 .
the x - ray tail is luminous ( @xmath0 10@xmath17 erg s@xmath3 ) , with an x - ray gas mass of @xmath18 m@xmath7 .
s06 interpreted the tail as the stripped ism of mixed with the hot icm .
the data also reveal three hard x - ray point sources ( @xmath19 erg s@xmath3 ) along the tail , and the possibility of all of them being background active galactic nuclei ( agns ) is very small .
s06 suggested that some of them may be ultra - luminous x - ray sources ( ulxs ) born from active star formation in the tail .
s07 further discovered a 40 kpc h@xmath4 tail and over 30 emission - line regions downstream of , up to 40 kpc from the galaxy .
s07 concluded that they are giant hii regions in the halo downstream or in intracluster space .
sivanandam et al .
( 2010 ) observed with irac and irs on .
a warm ( @xmath0 160 k ) molecular hydrogen tail was detected to at least 20 kpc from the galaxy from the irs data , at the same position as the bright x - ray and h@xmath4 tail .
the total mass of the warm h@xmath20 gas is @xmath21 m@xmath7 .
as the irs fields only cover the front part of the x - ray tail , the above warm h@xmath20 gas mass is only a lower limit .
8 @xmath22 m pah emission is also detected from the bright hii regions identified by s07 .
the results may imply a large reservoir of colder gas than the observed warm gas downstream of , coexisting with the hot icm ( @xmath23 k ) and the h@xmath4 emitting gas ( @xmath24 k ) .
obvious follow - ups include deep exposures and optical spectroscopic observations of the candidate hii regions . in this paper
, we present the results from our deep observation of and gmos spectroscopic observations .
the plan of this paper is as follows : the and observations and the data analyses are presented in section 2 . in section 3 , we discuss the x - ray tails of .
section 4 is on the properties of the surrounding icm .
section 5 discusses the point sources and the intracluster hii regions .
we also found another 40 kpc x - ray tail in a3627 , associated with , which is discussed in section 6 .
the results are discussed in section 7 and section 8 contains the summary .
we adopt a cluster redshift of 0.0163 for a3627 ( woudt et al .
assuming h@xmath25 = 71 km s@xmath3 mpc@xmath3 , @xmath26@xmath27=0.27 , and @xmath28=0.73 , the luminosity distance is 69.6 mpc , and 1@xmath29=0.327 kpc .
the observation of was performed with the advanced ccd imaging spectrometer ( acis ) on june 13 - 15 , 2008 ( obsid : 9518 ) .
standard data analysis was performed which includes the corrections for the slow gain change and charge transfer inefficiency ( for both the fi and bi chips ) .
we examined the light curve from the part of the s3 chip with weaker icm emission .
no flares of particle background were found .
there are also no flares in the light curve of the i2+i3 chips .
the effective exposures are 138.2 ks for the bi chips and 140.0 ks for the fi chips .
this new exposure is 18 times deeper than the old one .
about 4100 net counts from the thermal gas were collected in the 0.5 - 3 kev band .
we corrected for the acis low - energy quantum efficiency ( qe ) degradation due to the contamination on acis s optical blocking filter , which increases with time and varies with position .
the dead area effect on the fi chips , caused by cosmic rays , has also been corrected .
we used ciao3.4 for the data analysis .
the calibration files used correspond to calibration database ( caldb ) 3.5.3 from the x - ray center .
we used the new hrma on - axis effective area file , `` hrmad1996-12-20axeffan0008.fits '' , which improves the accuracy of the temperature measurement at @xmath30 4 kev .
the remaining uncertainty in the source flux ( up to @xmath0 10% ) would not much affect our conclusions in this work .
in the spectral analysis , a lower energy cutoff of 0.4 kev ( for the bi data ) and 0.5 kev ( for the fi data ) is used to minimize the effects of calibration uncertainties at low energy .
the solar photospheric abundance table by anders & grevesse ( 1989 ) is used in the spectral fits .
uncertainties quoted in this paper are 1 @xmath31 .
we used cash statistics in this work .
we adopted an absorption column density of 1.73@xmath32 @xmath33 from the leiden / argentine / bonn hi survey ( kalberla et al .
if we choose to fit the absorption column from the spectra ( with the phabs model in xspec ) , the best fits are always consistent with the above value .
this absorption column is lower than the previous value from dickey & lockman ( 1990 ) , 2.0@xmath32 @xmath33 , which had been used in previous work ( e.g. , b@xmath15hringer et al . 1996 ; s06 ) .
we obtained optical spectra of the emission - line objects identified by s07 with gmos on the south ( program i d : 2008a - q50 ) . because the field is crowded with interesting sources , three masks had to be used with some sources covered in more than one mask .
observations with the first mask were done on may 3 , 2008 ( airmass : 1.18 - 1.22 , seeing : @xmath34 ) . two exposures ( 20 minutes each ) were taken , with a central wavelength difference of 60 in order to cover the inter - chip gaps .
observations with the second mask and the third mask were taken on june 26 and june 27 , 2008 respectively ( airmass : 1.17 , seeing : @xmath35 for june 26 ; airmass : 1.18 , seeing : @xmath36 for june 27 ) .
two exposures ( 10 minutes each ) were taken on each night .
all three nights were photometric .
the r400 grating + gg455 filter were used , which gives a free wavelength range of @xmath0 4500 - 8000 .
we used 1@xmath29 slits for all sources .
as queued observations , we used the calibration data of ltt4364 taken on the photometric night of may 27 , 2008 for the relative flux calibration .
the gmos data were reduced with the gemini iraf package ( version 1.9.1 ) .
we generally followed the standard procedures .
besides the accompanying cuar arc data , we also used the night sky lines for the wavelength calibration .
the latter procedure is the key to derive consistent velocities for the same source from different exposures . in this paper
, we present only the velocity information and briefly discuss the nature of these sources ( hii regions ) .
the detailed properties of these hii regions will be discussed in a future paper .
the 0.6 - 2.0 kev count image is shown in figure 1 , with point sources removed .
the image was smoothed to enhance extended low surface brightness features . compared with images from short and exposures ,
a secondary tail is clearly visible to the south of the main tail .
some substructures are also revealed , including a `` protrusion '' from the main tail and two sharp bends in the main tail . also in figure 1 , the x - ray contours from the adaptively smoothed image ( exposure corrected , not the one in the left panel ) are overlaid on the h@xmath4+continuum image from the southern observatory for astrophysical research ( soar ) .
we also show the image of the cluster and a composite color image of s tails ( figure 2 ) . in the following sections , we discuss the spatial and spectral properties of s tails .
the observed soft x - ray emission of the tails most likely comes from the interfaces between the hot icm and the cold stripped ism .
although there are no data for the cold atomic and molecular gas ( note the data on the pah emission and the warm molecular h@xmath20 , sivanandam et al .
2010 ) , the x - ray surface brightness of the tails reveals a great deal about the stripped ism .
the surface brightness of the tails is quantitatively examined .
we first derived the 0.6 - 2 kev surface brightness profiles across some parts of the tails and the leading edge .
as shown in figure 3 , both tails are narrow but highly significant features above the local background .
even the `` protrusion '' is a 4.1 @xmath31 feature .
the leading edge ( or the contact discontinuity ) is very sharp , corresponding to the h@xmath4 edge ( figure 4 ) .
the gap between two tails has a similar surface brightness as the local background ( fig . 3b and 3c ) .
for the front part of the gap , the surface brightness is ( [email protected])@xmath37 cnts s@xmath3 kpc@xmath38 , while the local background is ( [email protected])@xmath37 cnts s@xmath3 kpc@xmath38 .
the low surface brightness of the gap has implications for the 3d structure of the stripped ism tail .
one possibility for the observed double tails is the projection effect of one broad cold ism tail . according to this interpretation
, the 3d structure of the soft x - ray emission might be a cylindrical shell that produces double tails in the projected surface brightness .
as shown in figure 3 , such a configuration fails to simultaneously account for both the low surface brightness of the gap and the observed widths of two tails .
the observed emission level between tails is similar to that outside tails . basically , the width of the cylindrical shell required to make the central emission dim also makes each peak unacceptably narrow in relation to the observed profiles .
thus , the two x - ray tails are most likely detached .
this is also supported by the data ( sivanandam et al .
the 8 @xmath22 m pah emission only appears in the first 20 kpc of the main tail .
the irs data only cover a small field downstream but no enhanced warm h@xmath20 emission is present at the front part of the gap . eventually we need hi and co data to better reveal the distribution of the cold ism downstream of . in principle , the surface brightness profiles along the tails reflect the stripping history . as shown in figure 5
, both tails have wiggles and substructures along the ( time ) axis .
however , the unknown conversion from the x - ray emission to the mass of the stripped ism ( filling factor , detail of the emission spectra , etc . ) makes this difficult with only the x - ray data .
a more complete picture can be obtained with the hi and co data . from figure 5 , the total linear length of the whole tail system is at least 4.2@xmath39 or 80 kpc .
we also measured the widths of the main tail in four segments ( before sharp bends ) and the widths of the secondary tail in two segments .
the model is composed of a gaussian and a flat local background .
as shown in figure 6 , the widths of both tails do not increase with the distance from the galaxy .
this conflicts with simulations ( e.g. , rb08 ; kapferer et al .
2009 ) , as further discussed in section 7.2 .
if is moving supersonically , a bow shock may form .
the local sound speed is 1263 ( @xmath14 / 6 kev)@xmath40 km / s . for a mach number of 1.1 - 3 ,
the predicted position of the bow shock is 20@xmath41 from the leading edge ( or the contact discontinuity ) .
no such shock front is detected in the exposure
. the enhanced surface brightness would be much weaker in projection on the sky , especially if is farther from the cluster center than the projected position . with this deep exposure , for the first time
spatially - resolved spectroscopy can be done for x - ray stripping tails of a cluster late - type galaxy .
we measured gas temperatures in five regions of the main tail and two regions of the secondary tail ( figure 7 ) .
the background is from the local source - free regions .
a single apec model with galactic absorption was used .
we emphasize that the measured temperatures are only spectroscopic temperatures as the intrinsically multi - phase gas is fitted with a single-@xmath14 model .
in fact , a hard x - ray excess generally exists in the spectral fits , which implies the existence of hotter components . however with limited statistics , we elected to keep the spectral model simple . as shown in figure 7 , the variations of spectroscopic temperatures along both tails are small .
in fact , both tails have consistent temperatures .
since temperatures derived from individual regions are similar , the total spectrum of two tails ( still excluding the `` head '' region ) was also examined with several models ( figure 8 and table 1 ) .
this observation collects 3444 net counts from the thermal gas of the two tails in the 0.5 - 3 kev band .
while the fit with a single-@xmath14 model is acceptable , the addition of a second component ( either another thermal component or a power - law component ) produces better fits .
we also tried the xspec model cemekl , which allows a power - law distribution of the emission measure for multi-@xmath14 gas .
we consider the spectral results shown in table 1 as evidence for multi - phase gas , especially as the single-@xmath14 model produces a very low abundance . adding another or more thermal components can increase the abundance to the solar value .
nevertheless , the double apec or the cemekl models are no more than a phenomenological way to describe the data .
it is still unclear what emergent spectrum mixing will produce .
there are more complications , like the non - equilibrium ionization effect and the intrinsic absorption .
generally , ionization equilibrium requires the collisional ionization parameter @xmath42 @xmath43 s. for @xmath44 @xmath43 , @xmath45 3 - 32 myr , which is comparable to the age of the x - ray tails .
the x - ray gas mass of the tails can be estimated .
the main uncertainties are the geometry ( or the filling factor ) of the x - ray emitting gas and the abundance of the gas .
the filling factor of the soft x - ray emitting gas depends on the mass and configuration of the stripped cold ism gas in the tails .
the soft x - rays may originate from the interfaces between the ambient hot icm and the stripped cold ism .
a fragmented cold ism tail has larger surface area so the filling factor of the soft x - ray emitting gas would be larger .
both uncertainties can be constrained by the assumed pressure equilibrium with the ambient icm . in the following ,
we present the estimates from the one-@xmath14 fit , the two-@xmath14 fit and the cemekl fit .
we assume both the main tail and the secondary tail are cylinders with a length of 80 kpc and radii of 3.9 kpc and 3.4 kpc respectively .
the `` head '' and `` protrusion '' regions are not included . with the result from the one-@xmath14 fit ( the first row of table 1 ) ,
the total x - ray gas mass is 2.0@xmath46 @xmath47 m@xmath7 , where @xmath48 is the filling factor of the soft x - ray emitting gas .
if the estimate is done on the same individual regions as the spectral analysis ( figure 7 ) , the total gas mass only decreases by 3% .
the electron density ranges from 0.0080 @xmath49 @xmath43 to 0.017 @xmath49 @xmath43 along the tails . however , the abundance of the x - ray gas is too low for the one-@xmath14 fit , which results in gas densities that are too high .
the ambient icm electron density and temperature in the projected position of s tails is @xmath50 @xmath43 and 6.3 kev respectively .
this puts an upper limit of 0.010 @xmath43 on the electron density of the tail gas ( for a temperature of 0.8 kev ) if pressure equilibrium is assumed .
thus , the one-@xmath14 model is not sufficient .
do the two-@xmath14 fit and the cemekl fit produce more reasonable results ?
we first used the result in the fourth row of table 1 , where the abundance of the lower temperature component is fixed at the solar value and the abundance of the higher temperature component is fixed at the value of the surrounding icm .
assuming pressure equilibrium between these two components , it is found that the hotter component occupies 97% of the volume of the soft x - ray emitting gas .
the electron density of the cool and hot components is 0.018 @xmath49 @xmath43 and 0.0071 @xmath49 @xmath43 respectively .
the total x - ray gas mass is 1.4@xmath46 @xmath47 m@xmath7 .
however , the gas pressure in the tail is @xmath0 50% higher than the maximum icm pressure of the surroundings .
this problem can not be solved by increasing the abundance of the hotter component in the spectral fit , which will further increase the ism pressure .
we then tried the cemekl fit ( the eighth row of table 1 ) .
however , the x - ray tail is at least 3.2 times over - pressured to the surrounding icm , under an isobaric condition .
the total x - ray gas mass is 1.1@xmath46 m@xmath7 .
therefore , either these models are not adequate , or there is dynamic pressure ( e.g. , turbulence ) besides the thermal pressure . without better understanding of the tail spectrum and the contribution of the cold ism in the tails , it is difficult to much improve the above estimates .
we summarize s properties from the current data in table 2 , along with those of ( section 6 ) .
we also examined the spectral properties of the icm around s tails .
we followed the same approach as used in sun ( 2009 ) for eso 137 - 006 to determine the local soft x - ray background that is dominated by the galactic emission ( note the galactic latitude of -7 deg ) . with
the soft x - ray background determined , icm temperatures were determined in 12 regions around s x - ray tails ( figure 7 ) .
the projected icm temperature is about 6 kev around s tails .
region # 8 has the lowest temperature , which implies the presence of a faint ism component in that region .
indeed , the 0.6 - 2 kev surface brightness at region # 8 is a bit higher than that in regions # 11 and # 13 , ( [email protected])@xmath37 cnts s@xmath3 kpc@xmath38 vs. ( [email protected])@xmath37 cnts s@xmath3 kpc@xmath38 .
region # 7 may include some small `` protrusions '' from both tails ( figure 1 ) , which can explain its relatively lower temperature than regions # 2 - # 6 .
the icm abundance can also be determined from a simultaneous fit to the spectra of all regions , [email protected] solar .
the electron density of the surrounding icm is 0.0010 @xmath43 - 0.0014 @xmath43 in the projected positions of the tails ( b@xmath15hringer et al .
the total icm thermal pressure is 1.8@xmath51 dyn @xmath33 for @xmath52 = 10@xmath53 @xmath43 and @xmath14=6 kev .
the ram pressure is 4.4@xmath51 ( @xmath54/1500 km / s)@xmath12 dyn @xmath33 for @xmath52 = 10@xmath53 @xmath43 .
both are high enough to strip most part of disk gas , except for the very core ( e.g. , roediger & br@xmath8ggen 2005 ) .
s07 identified 29 strong candidate hii regions .
six more sources were also identified as possible hii regions .
the gmos observations were able to cover thirty - three of them ( solid blue squares in figure 9 ) .
only two sources ( dashed blue squares in figure 9 ) were not observed .
both of them are strong candidates but are faint and close to .
the gmos spectra unambiguously confirm the association of these 33 sources ( 27 strong candidates + 6 weaker candidates in s07 ) with as their velocities are all within 200 km / s of s .
the heliocentric velocity map of these 33 hii regions is shown in figure 10 . in figure 10
, we also plot their velocities vs. their distances to the major axis of the galaxy or the middle line of the bow - like front of hii regions .
we used a system velocity of 4660 km / s for , which is the weighted average from the velocities derived by woudt et al .
( 2004 ) and s07 .
the velocities of these hii regions span a small range ( @xmath55 300 km / s ) with a dispersion of 54@xmath56 km / s .
in fact , this small velocity range is comparable to the radial rotation velocity of the residual h@xmath4 nebula at the center of , @xmath0 80 km / s from our soar goodman spectra that will be presented in a future paper .
after ism is removed from the galactic disk , it is subject to turbulence in the wake .
combined with the residual rotational motion , the range of the radial velocities of stripped ism clouds should become substantially larger than the range of the rotation velocity ( e.g. , rb08 ) .
however , that is not what we observed with the velocities of hii regions as tracers .
in fact as shown in figure 10 , the imprint of the rotation velocity pattern is still present in the velocity map of these hii regions .
some example spectra of these hii regions are shown in figure 11 , including several sources that are most distant from .
the low [ oi ] @xmath576300 / h@xmath4 and [ nii ] @xmath576584 / h@xmath4 ratios are typical of giant hii regions .
the detailed properties of these hii regions ( e.g. , line equivalent width , line ratios , star - formation rate , age , metallicity and total mass ) will be presented in a future paper with the results from our data .
the deep exposure also reveals many point sources , as shown in figure 9 .
we focus on x - ray point sources in or around its x - ray tails ( within the two elliptical regions in figure 9 ) .
nineteen point sources are detected .
their positions and fluxes are listed in table 3 .
the 0.5 - 2 kev limiting flux is @xmath58 erg s@xmath3 in the region of interest ( with eight counts in the 0.5 - 2 kev band ) . the expected x - ray source density above this limiting flux is @xmath59 per deg@xmath12 ( kim et al .
for the two elliptical regions ( 9.66 arcmin@xmath12 ) , the expected number of background sources is four , while 19 point sources are detected .
the contribution of galactic sources should not be neglected as is located at a low galactic latitude ( -7 deg ) . indeed
, six point sources are positionally coincident with bright star - like objects .
however , an excess of x - ray point sources still exists , especially in the immediate downstream region of where most hii regions are .
we can divide these point sources into three groups , sources 1 - 3 that are upstream of the galaxy , sources 4 - 12 that are around the disk plane or downstream of the galaxy ( but still within the 15 kpc radius halo in projection ) and sources 13 - 19 that are unbound and near the x - ray tails .
the latter two groups are of most interest here .
the region immediately downstream of the galaxy ( but still within the galactic halo ) has nine x - ray sources , while only 0.5 are predicted from the log n - log s relation ( @xmath0 1.1 arcmin@xmath12 ) .
the unbound tail region has seven x - ray sources , while 3.3 sources are expected from the log
n - log s relation ( @xmath0 7.9 arcmin@xmath12 ) .
three sources are likely of stellar origin ( 14 , 16 and 17 ) , while source 15 is close to a confirmed hii region .
if we take 3 stars in the unbound tail region as the number density of stars , only 0.4 stars are expected in the immediate downstream region .
in fact , only source 7 coincides with a bright star .
thus , the excess of x - ray point sources in the immediate downstream region is significant .
we examined the optical counterparts of these x - ray point sources and list the results in table 3 .
the soar images ( @xmath60 , @xmath61 and h@xmath4 bands ) that are presented in s07 are used . with seven sources ( across the s3 chip )
detected in both the x - rays and the optical images , the astrometry can be determined to @xmath62 .
there are four sources ( 8 , 10 , 11 and 15 ) that are within 2@xmath29 ( or 0.65 kpc ) of the confirmed hii regions and have no other optical counterparts in the soar images .
we consider them strong candidates for intracluster x - ray binaries , while a 0.5 kpc offset can be produced by a 100 km / s kick velocity in 5 myr .
sources 9 and 12 are 0.85 - 1.3 kpc from confirmed hii regions and do not fall on any optical sources .
they are also candidates for intracluster x - ray binaries .
if associated with , their 0.3 - 10 kev luminosities are @xmath63 erg s@xmath3 ( table 3 ) , which makes them intracluster ulxs .
the x - ray spectra of the bright point sources were also examined ( table 3 ) .
it is true that the chance match between hii regions and x - ray point sources is high in the immediate downstream region .
there is a @xmath0 52% chance that an x - ray source will be within 2@xmath29 of an hii region there .
however , the excess of x - ray sources is also significant there . as ulxs
are often found around young massive star clusters , it may not be a surprise to have intracluster ulxs around intracluster massive star clusters .
we can further examine the relation between the x - ray luminosities of these putative ulxs and the total star - formation rate of the hii regions .
the total 2 - 10 kev luminosity of sources 8 , 10 , 11 and 15 is 6.5@xmath64 erg s@xmath3 .
if sources 9 and 12 are added , the total luminosity is 2.7@xmath2 erg s@xmath3 . from s07 ,
the total sfr of these 33 hii regions is @xmath0 0.7 m@xmath7/yr , assuming an intrinsic extinction of 1 mag for the h@xmath4 emission .
this sfr and the lower value of the x - ray luminosity fall well on the relation derived by grimm et al .
( 2003 ) . even for the higher value of the x - ray luminosity ( when sources 9 and 12 are included )
, agreement can be achieved if the intrinsic extinction is higher in these hii regions and the sfr in the past was higher .
[ [ section ] ] another x - ray tail of a late - type galaxy was serendipitously found in our acis - s observation of eso 137 - 006 ( the bcg of a3627 , sun 2009 ) .
it is associated with , a redder and more massive disk galaxy than ( table 2 ) .
is only @xmath0 110 kpc from the cluster s x - ray peak in projection but its radial velocity is @xmath0 900 km / s larger than a3627 s .
the tail is near the corner of the i3 chip that is 10.8@xmath39 from the optical axis of the observation ( figure 12 and 13 ) .
was also covered by one of our previous acis - i observations of a3627 ( obsid : 4958 , pi : jones ) but the exposure is too short ( 14 ks ) to reveal the x - ray tail .
we summarize the properties of s tail in this paper , mainly for the purpose of comparison with s tails .
the details of the observation of eso 137 - 006 ( obsid 8178 ) and of the data analysis were presented in section 5 of sun ( 2009 ) .
the effective exposure was 57.3 ks for the i2 chip .
a significant x - ray tail behind is revealed , which in fact can also be seen in the data ( figure 2 ) .
being aware of its x - ray tail , we observed
( @xmath9=0.0191 ) with the 4.1-m soar telescope on cerro pachon on apr . 18 , 2008 ( ut ) .
the observations were made with the soar optical imager ( soi ) , which covers a 5.2@xmath65 field of view .
the night was photometric and seeing was very good , 0.45@xmath29 - 0.65@xmath29 .
the airmass was 1.16 - 1.20 .
we took three 800-sec exposures with the ctio 6693/76 filter ( h@xmath4 band ) and two 600-sec exposures with the ctio 6606/75 filter ( h@xmath66 band ) .
both narrow - band filters
are 2 inches @xmath10 2 inches , so the non - vignetted field is only @xmath67 . dithers of 15@xmath68 were made between exposures .
each image was reduced using the standard procedures with the iraf mscred package .
the pixels were binned 2@xmath69 , for a scale of 0.154@xmath29 per pixel .
dome flats were used .
the us naval observatory ( usno ) a2.0 catalog was used for the wcs alignment , which was done with wcstools .
the spectrophotometric standards were ltt 4364 and eg 274 ( hamuy et al .
1994 ) . is a disk galaxy with a dust lane across its bulge ( figure 13 ) .
it is a more massive galaxy than and has a larger radial velocity component than ( table 2 ) .
is 120 kpc ( or 6.15@xmath39 ) from in the plane of sky .
assuming the same ratios of the halo mass to the stellar mass , the tidal radius [ @xmath70 is 0.32 -0.41 @xmath71 , where @xmath71 is the distance between the two galaxies .
this value is much larger than the tidal truncation radius of ( @xmath0 15 kpc , truncated by the cluster potential ) estimated by s07 .
even if the two galaxies are only separated by their projected distance , s gravity is over 6 times s within 15 kpc radius of s nucleus . given the large difference of their radial velocities ( @xmath0 1080 km / s ) , their actual spatial separation should be larger than the projected distance , especially at the time when stripping began .
thus , the tidal interaction between two galaxies should be small .
there are indeed no tidal features upstream of and around .
their x - ray tails are also pointing away from each other .
s x - ray tail can be traced to at least 40 kpc from the nucleus .
since it is 10.6@xmath39 off - axis and the galaxy has a large radial velocity component , the actual length of the x - ray tail may be larger .
the 3 - 8 kev image clearly reveals a bright nuclear source .
we examined s x - ray spectral properties in two regions , one within 17.4@xmath29 of the nucleus ( note the 90% encircled energy radius of the point - spread function ( psf ) at 1.49 kev is 14@xmath29 there ) and the other one excluding the nuclear region .
the first region has 460@xmath1630 net counts in the 0.5 - 10 kev band , while the second region ( the tail region ) has 253@xmath1629 net counts in the 0.5 - 2 kev band .
the spectra are shown in figure 14 and the spectral fits are shown in table 4 . hosts a seyfert2-like nucleus .
the limited statistics on the soft x - rays , combined with the poor angular resolution in the source position , prevent any definite conclusion on the gas properties .
a single-@xmath14 fit implies a surprisingly hot tail with @xmath72 2 kev .
the gas temperature can be reduced to @xmath0 1.1 kev if a power - law component is included .
however , the power - law component has to be very strong ( table 4 ) , given the small area of the tail region ( 0.81 arcmin@xmath12 , compared with the case of ) .
like s tail , the abundance of s gas can not be reliably constrained .
the x - ray gas mass of the tail can also be estimated .
we assume the tail as a cylinder with a length of 40 kpc and a radius of 2.5 kpc . using the result from the one-@xmath14 fit ( the first row of table 4 ) ,
the total gas mass is 4.2@xmath73 @xmath47 m@xmath7 , where @xmath48 is the filling factor of the soft x - ray emitting gas .
this may be an over - estimate as the gas abundance is too low .
in fact , the average electron density is 0.0188 @xmath49 @xmath43 .
the surrounding icm has a temperature of @xmath0 6.3 kev , similar to that around .
the icm electron density at the projected position is @xmath74 @xmath43 ( b@xmath15hringer et al . 1996 ) .
thus , the x - ray tail would be @xmath0 3 times over - pressured . with the addition of a power - law component and a fixed abundance at 1 solar ( the third fit in table 4 ) , the ism electron density and the total gas mass reduce to 0.0055 @xmath49 @xmath43 and 1.2@xmath73 @xmath47 m@xmath7 respectively .
the last spectral fit may largely under - estimate the contribution from the gas ( see the difference of luminosities in table 4 ) .
thus , we take an average mass estimate of @xmath75 @xmath47 m@xmath7 .
the net h@xmath4 image of also shows an h@xmath4 tail at the same position as the x - ray tail , extending to at least 20 kpc from the nucleus ( figure 13 ) .
a sharp h@xmath4 edge is located only 1.1 kpc from the nucleus .
absorption is found upstream of the edge in the disk plane .
we also searched for emission - line objects in the field from the narrow - band data , hoping to find similar intracluster hii regions as those downstream of .
however , no such emission - line regions have been found in the sources detected in the h@xmath4 band .
the upper limit on the h@xmath4 luminosity ( no intrinsic extinction ) is 10@xmath76 erg s@xmath3 and the upper limit on the equivalent width of the h@xmath4 line is 60 . assuming a metallicity of 0.4 solar and applying the starburst99 model ( leitherer et al .
1999 ) , this upper limit gives a lower limit of 7 myr for the age of any faded intracluster hii regions .
this may imply that is in a more advanced stage of stripping and evolution than , which is consistent with the likely hotter x - ray tail of and s redder optical color .
is a more massive galaxy than and may be undergoing edge - on stripping , which indeed takes longer than the face - on stripping ( e.g. , rb08 ) .
while the paucity of luminous x - ray tails of late - type galaxies in nearby clusters has been known ( e.g. , sun et al .
2007a ) , the discovery of s double x - ray tails is a surprise .
m86 also has double x - ray tails and randall et al .
( 2008 ) interpreted it as the consequence of the aspherical potential of m86 and m86 s orbit .
m86 is an early - type galaxy so its potential structure is different from s and its tail is only composed of the stripped hot ism .
m86 s tails are in fact connected to the `` plume '' on the north of the galaxy and are located at larger distance from the galaxy , instead of coming from the galaxy directly like s . moreover , a spectacular h@xmath4 complex connecting m86 and the nearby disturbed spiral ngc 4438 was recently discovered by kenney et al .
some h@xmath4 emission also extends to the position of m86 s tail .
thus , tidal interaction may also be important for m86 s x - ray tail .
there also appears to be a split feature that resembles a double tail in ngc 4438 s image ( machacek et al .
2004 ; randall et al .
. however , one part of the split feature is almost along the distorted disk plane and tidal interaction should be important in this case ( machacek et al . 2004 ; kenney et al . 2008 ) .
ngc 4438 is @xmath0 3 times more luminous than in the @xmath77 band .
its truncated halo should be larger than s ( @xmath0 15 kpc in radius , s07 ) , also because the virgo cluster is less massive than a3627 . the features in ngc 4438 only extend to @xmath0 11 kpc from the nucleus , clearly within its halo .
the other late - type galaxy in the proximity , ngc 4388 , has many h@xmath4 filaments and a spectacular hi tail ( yoshida et al .
2002 ; oosterloo & van gorkom 2005 ) .
a faint x - ray extension can be traced to @xmath0 14 kpc from the nucleus in the direction of the hi tail , from the and data .
there is also an x - ray extension to east in the disk plane , which creates the appearance of double tails ( randall et al .
ngc 4388 is @xmath0 1.5 times more luminous than in the @xmath77 band so similar to ngc 4438 , its x - ray tail is still within its halo .
thus , we do not consider the one - sided x - ray features in ngc 4438 and ngc 4388 to be x - ray double tails based upon the current data .
they are clearly much fainter than s tails , especially given s much larger distance ( @xmath0 4.3 times ) and the much higher absorption to the direction of .
we can also define strong x - ray tails as one - sided features outside of the tidal truncation radius of the galactic halo ( like the tails of and ) , in order to distinguish them from the weak x - ray tails of ngc 4438 and ngc 4388 .
the observed x - ray emission of the tails is likely from the interfaces of the hot icm and the cold stripped ism , reflecting the multi - phase nature of mixing .
although the detailed mixing process that determines the emergent x - ray spectra is poorly understood , the observed tail temperature may be related to the mass - weighted temperature that is usually adopted in simulations , @xmath78 , where @xmath79 and we assume that @xmath80 . for the x - ray tails of and
, @xmath81 if we assume @xmath82 is the observed spectroscopic temperature .
however , it is easy to find out that the mass - weighted temperature is always larger than the spectroscopic temperature derived by fitting the observed spectra with a single-@xmath14 model ( @xmath83 ) .
we examined this with the cemekl and apec models . for the cemekl model , @xmath84 .
assuming an isobaric condition and for @xmath85 -1 , @xmath86 for @xmath85 -0.5 , 1/3 @xmath87 .
it is also reasonable to assume that @xmath88 .
however , with a series of xspec simulations , we found that @xmath89 .
for example , @xmath83 = 0.5 - 1 kev when -0.5 @xmath90 0.5 , while @xmath82 = 2 - 3.6 kev for @xmath91 = 6 kev .
this is not surprising as @xmath83 is mainly determined by the centroid of the iron - l hump at low temperatures , which biases @xmath83 low .
as long as there are significant emission components at @xmath14 = 0.4 - 2 kev , the iron - l hump will be strong to make @xmath83 deviated from @xmath82 .
on the other hand , the parameter @xmath4 in the cemekl model may reflect the age of the x - ray tail .
our simulations indeed show that @xmath83 increases monotonously with @xmath4 . from the cemekl
fits to the spectra of and , s stripping event may indeed be older .
the above simple estimate implies that @xmath83 is a certain fraction of the icm temperature .
the luminosity of the x - ray tail should also increase with the magnitude of the ambient icm pressure , under an isobaric condition .
thus , the x - ray tails of late - type galaxies in groups and the outskirts of poor clusters may have temperatures ( e.g. , @xmath92 0.4 kev ) and densities too low to be bright in the and energy bands .
this helps to explain the paucity of strong x - ray tails of late - type galaxies in nearby clusters ( e.g. , sun et al .
other factors include the generally high cluster background and the small number of late - type galaxies in the field of nearby clusters as the covered cluster volume is small . with the above scenario ,
the x - ray tail of a late - type galaxy is simply the manifestation of its cold ism tail embedded in the hot icm .
optical line emission is also expected and is indeed detected , e.g. , the h@xmath4 tails of and ( s07 and section 6 ) .
hi observations of a3627 were made with australia telescope compact array ( atca ) in 1996 and no hi emission was detected from or ( vollmer et al .
the 3@xmath31 detection limit is @xmath0 3 mjy / beam in one velocity channel with a beamsize of 30@xmath29 ( vollmer et al .
adjusting to the cluster distance used in this work and assuming a linewidth of 150 km / s ( vollmer et al .
2001a ) , the corresponding hi gas mass limit is @xmath93 m@xmath7 per beam . while this limit is not weak for the remaining hi gas in the disks of and
( roughly within one beam ) , the x - ray tails of and span over 7 - 15 times the beam size so a lot of diffuse hi gas can remain undetected downstream of the galaxy .
in fact , with the sensitivity of the available atca data ( an hi column density limit of 2@xmath94 @xmath33 ) , four of the seven hi tails in the virgo cluster found by chung et al .
( 2007 ) and the long hi tail of ngc 4388 ( oosterloo & van gorkom 2005 ) would not be detected .
clearly deeper hi observations are required , although the nearby bright radio source pks 1610 - 60 ( 43 jy at the 1.4 ghz ) makes the task difficult ( 8@xmath39 - 14@xmath39 from and ) .
both and are expected to host a significant amount of atomic and molecular gas initially . from blanton & moustakas ( 2009 ) , the total neutral plus molecular gas fraction ( relative to the stellar mass ) is @xmath0 70% for and @xmath0 30% for . assuming a type of sc and a blue - band diameter of 20 kpc , the expected hi mass of is @xmath95 m@xmath7 , from the empirical relation derived by gavazzi et al .
similarly for ( a type of sa and a blue - band diameter of 30 kpc ) , the expected hi mass is @xmath96 m@xmath7 .
thus , is expected to have @xmath97 m@xmath7 of cold ism initially , while is expected to have @xmath98 m@xmath7 initially .
these numbers can be compared with the x - ray gas mass in their x - ray tails ( table 2 ) .
sivanandam et al .
( 2010 ) detected the mid - ir emission from warm h@xmath20 gas ( @xmath0 150 k ) in the galaxy and the first 20 kpc of the main tail , with a total mass of @xmath21 m@xmath7 .
the h@xmath20 lines that irs detects are rotationally excited lines that are generated very efficiently .
the warm gas they probe is likely a small fraction of the total h@xmath20 gas , if the bulk of the molecular gas is still cold .
deep hi and co observations are required to recover the bulk of the cold gas in , and their wakes .
the same question can also be asked to other cluster late - type galaxies with signs of stripping .
does a stripped ism tail show up in hi , h@xmath4 and x - rays simultaneously ?
we attempt to summarize the known examples of stripped tails for cluster late - type galaxies .
a venn diagram is plotted in figure 15 .
while a lot of stripped tails are known for cluster late - type galaxies , there is currently little known overlap between different bands , especially between x - ray tails and hi tails .
this may be due to the lack of sufficient data , either in x - rays ( e.g. , for the virgo galaxies with hi tails , chung et al .
2007 ) or in hi ( e.g. , the two galaxies in this paper ) .
we are pursuing data for the chung et al .
( 2007 ) galaxies and deeper atca data for the a3627 galaxies to better address this question . on the other hand
, there are various reasons that a one - to - one correlation may not exist in every case , even with better data . depending on its age and the surrounding icm
, an ism tail may emit predominantly in the hi band or the x - ray band .
soft x - ray emission of the tail may only be bright in high - pressure environment .
we also understand little about the details of mixing . it may be related to the mean free path of particles and the coherence length of the magnetic field , which would introduce a radial dependence on the efficiency of mixing . nevertheless , we first need better data to update the venn diagram shown in figure 15
. deep x - ray data alone , like s , can reveal some thermal history of the stripped ism .
one surprise from s x - ray data is the constancy of the spectroscopic temperatures along both tails ( figure 7 ) .
indeed extra care is required to interpret this spectroscopic temperature as it is mainly determined by the centroid of the combined iron - l humps from multi-@xmath99 gas .
nevertheless , the possible spectral difference between s tails and s tail is intriguing .
this difference , if confirmed , combined with the temperature constancy of s tails , puts a constraint on the heating timescale of the stripped cold ism . clearly the knowledge of the hi / co gas distribution in these two galaxies will help to tighten the constraint , although it is certain that this timescale depends on the icm temperature .
after the initial analytic work by gunn & gott ( 1972 ) , many simulations have been run on the ram - pressure stripping of late - type galaxies in clusters ( e.g. , abadi et al .
1999 ; schulz & struck 2001 ; vollmer et al .
2001b ; rb08 ; kapferer et al . 2009
; tonnesen & bryan 2009 and references in those papers ) .
recent efforts surpass the early ones in many respects , e.g. , resolution , time steps , realistic treatment of the galaxy orbit and some important baryon physics ( e.g. , cooling and star formation ) .
we mainly compare our results with recent simulations of three groups , the innsbruck group with the gadget-2/sph code ( kronberger et al .
2008 ; kapferer et al .
2009 ) , the bremen group with the flash / amr code ( roediger et al . 2006 ; rb08 ) and the columbia group with the _ enzo_/amr code ( tonnesen & bryan 2009 ) .
the gadget-2 simulations include recipes for cooling , star formation and stellar feedback .
the model galaxies run through a wind tunnel with a constant pressure .
the flash simulations are non - radiative and model the flight of a disk galaxy through a galaxy cluster with realistic icm and potential distributions .
the _ enzo _
simulations include cooling but not feedback ( tonnesen & bryan 2009 ) . to mimic effects of some heating processes not in simulations
, they impose a minimum temperature floor .
the model galaxy is also subject to a constant ram pressure .
the inclusion of cooling is important , as it indeed happens as revealed from our data ( the intracluster hii regions and star clusters ) and it has profound effects in simulations ( kapferer et al .
2009 ; tonnesen & bryan 2009 ) .
cooling also naturally adjusts the ism distribution on the disk plane , creating multi - phase ism with enhancement and holes ( kapferer et al .
2009 ; tonnesen & bryan 2009 ) , which is not present in the non - radiative runs .
however , cooling needs to be offsetted by stellar feedback .
unknown efficiencies of transport processes ( viscosity and heat conduction ) because of unclear strength and configuration of magnetic field , plus the other uncertainties in the prescriptions of star formation and stellar feedback , make the tasks difficult ( e.g. , tonnesen & bryan 2009 ) .
we emphasize that comparison at the current stage is not straightforward as galaxies in these simulations are always more massive than and high pressure environment was not explored in all simulations .
we also caution that it is best to compare the observational data with the mock data generated from simulations .
kapferer et al .
( 2009 ) and tonnesen & bryan ( 2009 ) attempted to create x - ray mock data but more work is required , e.g. , folding with the actual response , projection on the icm background and the use of spectroscopic temperature rather than mass - weighted temperature ( section 7.1 ) .
tails generally appear longer and wider in simulations than in observations .
the simulations by kapferer et al .
( 2009 ) show that the x - ray tails can be traced to @xmath0 300 kpc from a galaxy that is about 4 times more massive than , with an ambient pressure similar to s .
the simulated x - ray tails are also very clumpy , with widths generally increasing with the distance from the galaxy .
the flash simulations also produce longer tails than observed in this work , but are difficult to compare with the x - ray data directly .
as rb08 pointed out , the observed tail length depends on the mass - loss per orbital length , which is the highest in poor clusters as galaxies in rich clusters move too fast and the ram pressure in groups is not high enough
. the flash simulations also show tail flaring ( i.e. increasing width with distance from the galaxy downstream ) in all cases .
the velocity perpendicular to the direction of the ram pressure comes from the random motion of the icm , the disk rotation and turbulence in the wake . for a rotation velocity of 100 km / s over 50 myr ,
the offset is 5 kpc , which is not small in s case as its tails are narrow .
the tonnesen & bryan ( 2009 ) simulations produce long but narrower tails .
the tails are also very clumpy .
none of these simulations can reproduce the double tails of .
the simulated tails are generally too clumpy with strong turbulence .
this simple comparison may point to a higher icm viscosity than present in the simulations .
the reynold number is @xmath100 , where @xmath101 is the mach number , @xmath102 is the size of the remaining galactic gas disk and @xmath57 is the mean free path of particles in the icm . for unmagnetized gas , @xmath103 10 kpc around .
the radius of the remaining h@xmath4 disk and the galactic x - ray emission is only @xmath0 1.7 kpc ( s07 ) .
thus , the reynold number is on the order of unity which implies a laminar flow .
of course the presence of magnetic field would increase the reynold number . in the simulations by kapferer
( 2009 ) , a bright spot in the x - ray images is always found at distances of a few tens of kpc from the galaxy , caused by compressional heating .
this feature can be compared with the bright spot in the main tail ( figure 1 and 5 ) . however , the location of this feature is generally too far from the galaxy in simulations with high ambient pressure .
such a bright spot is also not observed in the secondary tail .
the most obvious question about s x - ray tails is why are there double tails .
the simplest explanation is to involve another dwarf galaxy forming a pair with .
this was in fact simulated by kapferer et al .
( 2008 ) , showing double gaseous tails , stellar tidal tails and a stellar bridge between two galaxies .
the x - ray gas mass of the secondary tail is 40% - 50% of the mass of the main tail , assuming the same filling factors of the x - ray gas .
if we simply take the ratio of their x - ray gas masses as the ratio of their ism gas masses , the stellar mass of this assumed dwarf that is responsible for the secondary tail should be @xmath0 30% of s stellar mass , from the ism mass - stellar mass relation summarized by blanton & moustakas ( 2009 ) .
such a galaxy should be easily detected , at least through the tidal tails and bridge it causes even if it hides behind .
however , the soar images do not show the presence of another galaxy near .
there are also no tidal features upstream of the galaxy .
there are some continuum features downstream of the galaxy as shown in s07 .
however , all of them are around confirmed hii regions and have blue colors .
our new data only reinforce the conclusion from the soar images .
the central region of the galaxy is very dusty but the galaxy appears regular without significant tidal features in the @xmath61 band ( paper iv , in preparation ) .
thus , we conclude that both x - ray tails originate from .
the discovery of two separated stripping tails from one galaxy presents a challenge to our understanding of stripping . in simulations ( e.g. , rb08 ; kapferer et al .
2009 ; tonnesen & bryan 2009 ) , cold ism is removed from the sides of the disk or through holes and is quickly mixed to form one clumpy tail because of disk rotation , momentum transferred from the icm and turbulence in the wake , especially if projected on the plane of sky .
perhaps we should begin to consider more realistic ism distributions in the galaxy .
optical images clearly show the presence of two thick spiral arms , one to the north and the other one to the south ( s07 ; better shown in the images ) .
maybe the two x - ray tails correspond to stripping of molecular gas concentrated around these two spiral arms .
co images of nearby face - on spiral galaxies indeed show that molecular gas distribution follows major spiral arms in many cases ( e.g. , kuno et al . 2007 ) .
this is also generally true for hi gas ( e.g. , adler & westpfahl 1996 ) . for
, we present a simple cartoon in figure 16 .
the north arm is apparently the trailing arm where stripping is easier so a brighter x - ray tail is produced .
the south arm is the leading arm so gas removal there is more difficult as most gas would be pushed inwards to deeper galactic potential first .
this also explains the fact that more intracluster hii regions ( at @xmath104 15 kpc from the nucleus ) are found around the main tail while the numbers of hii regions in the halo are similar at two sides .
the curvature of the secondary tail may come from the galactic rotation , which is high enough to cause the observed offset .
the main tail may also be curved in 3d but appears straight in projection .
the rather regular wake structure may imply significant viscosity . on the other hand , draping of the icm magnetic field ( e.g. , dursi & pfrommer 2008 ) can also inhibit the motion of the stripped ism perpendicular to the direction of ram pressure . in summary
, it remains to be seen whether simulations can reproduce the double x - ray tails of as observed , and the gas properties in other bands ( h@xmath4 and h@xmath20 at least ) .
we also emphasize that the comparison between data and simulations is not trivial and straightforward .
observationally we need deeper data at multiple wavelengths . for simulations ,
mock data need to be produced with sensitivities matching the actual observations .
uncertainties in the viscosity , heat conduction , cooling and stellar feedback should be explored with different runs .
eventually we need to compare the multi - wavelength data ( hi , co , ir , optical and x - rays ) with the mock data in the same band . besides the simple comparison of morphology , a lot of details of the tail properties ( e.g. , temperature distribution , filling factors of the gas in different phases and energy transfer between gas in different phases ) can be better examined . the new observations of , for the first time , unambiguously confirm the active star formation in the cold ism stripped by the icm ram pressure .
as emphasized in s07 , s immediate surroundings are devoid of bright galaxies .
there are only two galaxies that are within 4 mag of in the @xmath61 band , within 100 kpc from the end of s x - ray tail ( g1 and g2 in figure 1 of s07 ) .
their redshifts are unknown and they are 1.9 - 2.7 mag fainter than in the @xmath61 band .
there is also no such galaxy within 70 kpc from the nucleus of .
both g1 and g2 are undisturbed and there are no tidal features between and g1 ( or g2 ) .
our data do not reveal another dwarf galaxy near ( paper iv , in preparation ) .
the velocity map of the hii regions also shows no extra component , besides s rotation pattern .
it is also known that the cluster potential is not the main factor for stripping ( s07 ; sivanandam et al .
2010 ) , although it indeed affects the size of s dark matter halo .
thus , we conclude that icm ram pressure is responsible for the stripping of the ism . in our case ,
the ambient pressure is even high enough to remove some molecular gas from the galaxy .
the hi gas has typical pressures of @xmath105 k @xmath43 , while the pressure in the core of a starless giant molecular cloud can be higher than @xmath106 k @xmath43 . the ambient total thermal pressure ( electron + ions )
is 1.3@xmath107 k @xmath43 for @xmath52 = 10@xmath53 @xmath43 and @xmath14=6 kev , while the ram pressure is 3.2@xmath107 k @xmath43 for @xmath52 = 10@xmath53 @xmath43 and @xmath54 = 1500 km / s . after removing the cold ism from the galaxy ,
the key for the intracluster star formation is that the stripped ism must be able to cool , likely aided by shocks and the absence of the uv radiation field that is strong on the disk plane .
intracluster star formation in the stripped ism was first suggested in simulations ( schulz & struck 2001 ; vollmer et al .
schulz & struck ( 2001 ) even suggested formation of dwarf galaxies in this mode . in the new gadget-2/sph simulations by kronberger et al .
( 2008 ) and kapferer et al .
( 2009 ) , new stars are formed in the stripped ism , forming @xmath0 1 kpc structures with masses of up to 10@xmath108 m@xmath7 individually to hundreds of kpc downstream of the galaxy .
however , all simulations suffer from uncertainties in transport processes like viscosity and heat conduction , which are important for determining the heating rate . more evidence of intracluster star formation ,
either hii regions or blue star clusters , has been revealed recently ( e.g. , lee et al . 2000 ; gerhard et al .
2002 ; cortese et al . 2004 ; cortese et al . 2007 ; yoshida et al .
they are only at one side of the galaxy . in some cases
, tidal interaction may be important ( e.g. , lee et al . 2000 ) .
but they all lack unambiguous associations with gaseous tails .
we especially highlight the growing number of examples of one - sided blue stellar filaments and star clusters behind cluster galaxies , which includes two galaxies in a2667 and a1689 at @xmath109 0.2 ( cortese et al .
2007 ) and a dwarf galaxy rb 199 in the coma cluster ( yoshida et al . 2008 ) .
the connection between the first two galaxies at @xmath110 with was discussed in s07 . in the case of rb 199 , yoshida et al .
( 2008 ) found narrow blue filaments , knots , h@xmath4 filaments and clouds extending to 80 kpc from the galaxy in one side .
it is a rather remarkable case as the galaxy is much fainter than all previous galaxies , with a stellar mass of @xmath111 m@xmath7 .
the stellar mass of the downstream structures is @xmath112 m@xmath7 , which is @xmath0 10% of the galactic mass .
yoshida et al . ( 2008 ) also favored a mechanism similar to what we suggested for in s07 .
these three examples may imply a rather common evolution stage for cluster late - type galaxies .
as the stage with bright blue stellar trails is short ( @xmath55 200 myr ) , more similar examples are expected only in large surveys .
if we assume that intracluster star formation happens in the stripped ism of every cluster spiral , can this mode contribute a significant fraction of the intracluster light ?
the answer depends on the mass fraction of intracluster stars formed out of the stripped ism ( or the star formation efficiency ) .
s07 gave a rough estimate of @xmath0 1% , which may only be a lower estimate as blue stellar filaments and clusters ( not shining in h@xmath4 ) are not counted . while a detailed account of the total stellar mass downstream of is the aim of paper iv , here we take 1% as a lower limit .
if the efficiency of the intracluster star formation is only 1% , the intracluster light contributed from star formation in stripped ism is small , given the mass fraction of the intracluster light to the total stellar light in clusters ( @xmath0 20% , likely depending on the halo mass , e.g. , krick & bernstein 2007 ) .
however , the case of rb 199 implies a fraction of @xmath0 10% , if the original ism in rb 199 had a mass comparable to its stellar mass ( e.g. , blanton & moustakas 2009 ) . in the simulations by kronberger et al .
( 2008 ) and kapferer et al .
( 2009 ) , 7% - 12% of the initial ism turns into new stars in the wake . if an efficiency of 10% is adopted , the intracluster light contributed by ram pressure stripping is a significant fraction of the total intracluster light .
however , this efficiency may not be too high given the star formation efficiency in galaxies .
in galactic disks , @xmath0 1% of gas is converted to stars per free - fall time ( e.g. , leroy et al . 2008 ) . in 100 myr ( a typical orbital time in the disks ) , @xmath0 6% of gas is converted to stars ( e.g. , kennicutt 1998 ; leroy et al . 2008 ) .
we note that 100 myr is also about the orbital time of the residual h@xmath4 disk ( @xmath0 1.7 kpc radius with an orbital velocity of @xmath0 100 km / s ) and close to the age of the x - ray tails .
clearly , a lot of work is required to constrain the intracluster star formation efficiency .
the efficiency may also change with the mass of the host galaxy as more stars are formed inside the halo of more massive galaxies . in principle , the ratio of the intracluster snia to the other types can also inform us the significance of intracluster star formation . as active star formation is confirmed downstream of the galaxy , the discovery of ulxs is not a surprise . as we estimated in section 5.2 , the total luminosity of ulxs can be consistent with the expectation from the intracluster star formation rate .
the space density of x - ray point sources also traces the strength of the star formation , stronger in the halo and weaker around the unbound tail . a similar system may be the high - velocity system of ngc 1275 .
gonzalez - martin et al .
( 2006 ) found a concentration of eight ulxs ( @xmath113 erg s@xmath3 ) around the high - velocity system .
they may be associated with the active star formation in and downstream of the high - velocity system , similar to what is happening in .
we present the discovery of spectacular x - ray double tails behind in the closest rich cluster a3627 from our 140 ks observation .
spectroscopic data of hii regions downstream of are also discussed .
we also present the discovery of a likely heated x - ray tail associated with , also in a3627 .
the main results of this paper are : \1 ) besides the known main x - ray tail of , a fainter secondary tail is revealed with the deep observation , as well as some substructures ( section 3 and figure 1 and 2 ) .
the double tails of make it morphologically resemble radio head - tail galaxies .
the whole tail system exhibits intriguing regularity ( almost constant tail widths , constant spectroscopic temperatures and well separated tails not twisted together ) and some complexity ( substructures ) .
we consider that the soft x - ray emission of the tails come from the mixing of the stripped cold ism with the hot icm .
we also conclude that the two tails are detached and both of them originate from , stripped by the icm ram pressure . one possible scenario ( section 7.2 )
involves the two major spiral arms of that is visible from the optical data , dense ism ( especially molecular gas ) concentration around two arms , the orientation of the arms to the ram pressure and a weak level of turbulence downstream of the galaxy ( or a higher viscosity than that in the current simulations ) .
\2 ) the spectra reveal the multi - phase nature of the x - ray tails , requiring more than one thermal component to obtain good fits and reasonable abundance ( e.g. , 1 solar ) .
however , the x - ray tails are always over - pressured to the icm with the current simple models ( section 3.2 ) .
we also examined the spectroscopic temperature distribution in the tails with a single-@xmath14 model . s double tails are surprisingly isothermal ( @xmath0 0.8 kev ) , embedded in @xmath0 6 kev icm ( figure 7 ) .
\3 ) s x - ray tails also present a challenge to the current simulations of stripping of cluster spirals ( section 7.2 ) .
nevertheless , we emphasize that the comparison is not trivial and straightforward .
mock data from simulations need to be generated with sensitivities matching the actual observations .
eventually we need multi - wavelength data ( e.g. , hi , co , ir , optical and x - rays ) in order to examine the interplay and energy transfer between gas in different phases ( section 7.1 and figure 15 ) .
\4 ) spectra were obtained for 33 emission - line regions identified by s07 ( section 5.1 ) .
all of them are confirmed as hii regions downstream of ( section 5.1 and figure 9 - 11 ) .
their velocity map shows the imprint of s rotation pattern ( figure 10 ) .
for the first time , we unambiguously know that active star formation can happen in the cold ism stripped by icm ram pressure .
we also summarized the current known cases of intracluster star formation , especially ones with one - sided blue star clusters and stellar trails ( section 7.2 ) .
the intracluster star - formation efficiency is expected to be higher than 1% .
the contribution to the intracluster light from this mode can be high if the star - formation efficiency is as high as 10% ( section 7.3 ) .
\5 ) this deep observation also reveals many x - ray point sources ( section 5.2 ) .
a significant source excess is observed immediately downstream of .
four of the x - ray point sources are within 0.65 kpc from spectroscopically confirmed hii regions .
we consider them strong candidates for intracluster ulxs with @xmath114 erg s@xmath3 .
other candidates ( at least two ) have @xmath115 up to 2.5@xmath2 erg s@xmath3 .
we also conclude that the observed total x - ray luminosity of the intracluster ulxs is consistent with the expectation from the observed total star formation rate of hii regions .
\6 ) a 40 kpc x - ray tail was detected behind another late - type galaxy in a3627 , ( section 6 ) .
signatures of stripping were also found in the h@xmath4 data , with a sharp h@xmath4 edge and an h@xmath4 tail extending to at least 20 kpc from the nucleus .
no hii regions are found downstream of .
its x - ray tail is also surprisingly hot ( @xmath72 2 kev ) but the uncertainty is big .
we suggest that s stripping event is at a more advanced stage than s .
why are there two bright x - ray tails of late - type galaxies in a3627 ? besides coincidence , the high ambient pressure ( ram pressure + thermal pressure ) in a3627 should also play a role , making soft x - ray tails denser and more luminous .
this would explain the absence of bright x - ray tails of late - type galaxies in the virgo cluster ( see section 7.1 ) .
a high - pressure environment also help intracluster star formation , by moving high - density molecular gas into the intracluster space .
this paper mainly focuses on the data of . in a future paper
, we will examine the intracluster star formation downstream of in more detail , from the and data .
hi and co data will also be pursued . with a more complete set of multi - wavelength data , the nature of the x - ray tails and the intracluster star - formation could be better understood .
provides the best opportunity to simultaneously study stripping and the related important questions across the electromagnetic spectrum .
m. s. is grateful to tom matheson for his helps in the planning and the data analysis of the observations .
m. s. thanks suresh sivanandam for the preprint of the results and discussions .
we thank hans b@xmath15hringer , wolfgang kapferer and bernd vollmer for discussions .
we thank keith arnaud for the clarification on the cemekl model .
we present data obtained at the gemini observatory , which is operated by the association of universities for research in astronomy , inc .
, under a cooperative agreement with the nsf on behalf of the gemini partnership .
we also present data obtained with the southern observatory for astrophysical research , which is a joint project of the brazilian ministry of science , the national optical observatories , the university of north carolina and michigan state university .
we made use of the nasa / ipac extragalactic database and the hyperleda database .
the financial support for this work was provided by the nasa grant go7 - 8081a , the nasa grant go8 - 9083x and the grant hst - go-11683.01-a .
md and gmv acknowledge support from a nasa ltsa grant nasa nng-05gd82 g .
er acknowleges funding by the dfg priority programme 1177 ` galaxy evolution ' .
abadi , m. g. , moore , b. , bower , r. g. 1999 , mnras , 308 , 947 adler , d. s. , & westpfahl , d. j. 1996 , aj , 111 , 735 anders , e. , & grevesse n. 1989 , geochimica et cosmochimica acta , 53 , 197 bekki , k. , & couch , w. j. 2003 , apj , 596 , l13 b@xmath15hringer , h. et al
. 1996 , apj , 467 , 168 blanton , m. r. & moustakas , j. 2009 , ara&a , 47 , 159 chung , a. et al . 2007 , apj , 659 , l115 cortese , l. et al .
2004 , a&a , 416 , 119 cortese , l. et al .
2007 , mnras , 376 , 157 dekel , a. , & birnboim , y. 2008 , mnras , 383 , 119 dickey , j. m. , & lockman , f. j. 1990 , ara&a , 28 , 215 dursi , l. j. , & pfrommer , c. 2008 , apj , 677 , 993 finoguenov , a. et al .
2004 , a&a , 419 , 47 gavazzi , g. et al .
2001a , a&a , 377 , 745 gavazzi , g. et al .
2001b , apj , 563 , l23 gavazzi , g. et al .
2005 , a&a , 429 , 439 gerhard , o. et al .
2002 , apj , 580 , l121 giovanelli , r. , haynes , m. p. 1985 ,
apj , 292 , 404 grimm , h. j. , gilfanov , m. , sunyaev , r. 2003 , mnras , 339 , 793 gonzalez - martin , o. , fabian , a. c. , sanders , j. s. 2006 , mnras , 367 , 1132 gunn , j. e. , & gott , j. r. , iii 1972 , apj , 176 , 1 hamuy , m. et al .
1994 , pasp , 106 , 566 kalberla , p. m. w. et al .
2005 , a&a , 440 , 775 kapferer , w. et al .
2008 , mnras , 389 , 1405 kapferer , w. et al .
2009 , a&a , 499 , 87 kenney , j. d. p. , van gorkom , j. h. , vollmer , b. 2004 , aj , 127 , 3361 kenney , j. d. p. et al .
2008 , apj , 687 , l69 kennicutt , r. c. , jr .
1998 , apj , 498 , 541 kim , minsun et al .
2007 , apj , 659 , 29 kraan - korteweg , r. c. et al .
1996 , nature , 379 , 519 krick , j. e. , & bernstein , r. a. 2007 , aj , 134 , 466 kronberger , t. et al .
2008 , a&a , 481 , 337 kuno , n. et al .
2007 , pasj , 59 , 117 lee , h. , richer , m. g. , mccall , m. l. 2000 , apj , 530 , l17 leitherer , c. et al . 1999 ,
apjs , 123 , 3 leroy , a. k. et al .
2008 , aj , 136 , 2782 machacek , m. e. et al .
2004 , apj , 610 , 183 machacek , m. e. et al .
2005 , apj , 630 , 280 machacek , m. e. , jones , c. , forman , w. , nulsen , p. 2006
, apj , 644 , 155 nulsen , p. e. j. 1982 ,
mnras , 198 , 1007 oosterloo , t. , van gorkom , j. 2005 , a&a , 437 , 19l quilis , v. , moore , b. , & bower , r. g. 2000 , science , 288 , 1617 randall , s. et al .
2008 , apj , 688 , 208 roediger , e. , & hensler , g. 2005 , a&a , 433 , 875 ( rh05 ) roediger , e. et al .
2006 , mnras , 371 , 609 roediger , e. , & br@xmath8ggen , m. 2008a , mnras , 388 , l89 roediger , e. , & br@xmath8ggen , m. 2008b , mnras , 388 , 465 ( rb08 ) schulz , s. , & struck , c. 2001 , mnras , 328 , 185 sivanandam , s. , rieke , m. , rieke , g. 2010 , apj , in press , arxiv:0912.0075 skelton , r. e. , woudt , p. a. , kraan - korteweg , r. c. 2009 , mnras , 396 , 2367 stevens , i. r. , acreman , d. m. , ponman , t. j. 1999 , mnras , 310 , 663 sun , m. , & vikhlinin , a. 2005 , apj , 621 , 718 sun , m. et al . 2006 , apj , 637 , l81 ( s06 ) ( paper i ) sun , m. et al . 2007a , apj , 657 , 197 sun , m. , donahue , m. , voit , g. m. 2007b , apj , 671 , 190 ( s07 ) ( paper ii ) sun , m. 2009 , apj , in press , arxiv:0904.2006 tonnesen , s. , & bryan , g. l. 2009 , apj , submitted , arxiv:0909.3097 vollmer , b. et al .
2001a , a&a , 369 , 432 vollmer , b. , cayatte , v. , balkowski , c. , duschl , w. j. 2001b , apj , 561 , 708 vollmer , b. , huchtmeier , w. 2007 , a&a , 462 , 93 wang , q. d. , owen , f. , ledlow , m. 2004 , apj , 611 , 821 woudt , p. a. et al .
2004 , a&a , 415 , 9 woudt , p. a. et al .
2008 , mnras , 383 , 445 yagi , m. et al .
2007 , apj , 660 , 1209 yoshida , m. et al .
2002 , apj , 567 , 118 yoshida , m. et al .
2008 , apj , 688 , 918 ) gaussian .
this image shows the raw data in a minimally processed way .
besides two significant tails , some sub - structures like a `` protrusion '' and two sharp bends are clearly visible .
the red scale bar is 10 kpc ( or 30.6@xmath29 ) . *
right * : the 0.6 - 2.0 kev contours in red superposed on the soar h@xmath117 image ( h@xmath4 + continuum ) .
the image was background subtracted and exposure corrected .
point sources were also removed .
asmooth was used to adaptively smooth the image .
the contours are in square - root spacing and the innermost level is 3 times the outermost level ( note that this image is heavily smoothed so check figure 3 and 4 for the surface brightness ) .
the bright part of the h@xmath4 tail ( the first 20 kpc ) is also visible .
the green dashed line shows the major axis of the disk plane .
the red scale bar is 10 kpc ( or 30.6@xmath29 ) . ]
emission ( from the soar data , s07 ) is in red .
the white stellar image is the same as the one shown in fig .
1 . note that the x - ray image was adaptively smoothed ( also for other x - ray contours shown in this work ) .
the x - ray leading edge is in fact in the same place as the h@xmath4 edge ( figure 4 ) . ] ) . *
b ) * : the 0.6 - 2.0 kev surface brightness profile across the straight part of the main tail .
the width of slices is 115@xmath29 ( or 37.6 kpc ) .
the secondary tail , `` protrusion '' and point sources were masked .
a gaussian fit with a constant background is also shown .
the fwhm of the gaussian is [email protected] kpc .
* c ) * : the 0.6 - 2.0 kev surface brightness profile along the direction perpendicular to the tails at @xmath0 45 kpc from the galaxy .
point sources are masked .
the width of slices is 35@xmath29 ( or 11.4 kpc ) .
the black solid line is a simple fit with two gaussians plus a flat background .
the separation between the centers of the two tails is 14 kpc .
the fwhms are [email protected] kpc and [email protected] kpc respectively .
we attempted two fits assuming a cylindrical shell for the whole tail system ( the blue and red solid lines , only for the secondary tail , see section 3.1 ) .
as can be seen from the red and blue lines , it is impossible to fit the width of each tail and the faint emission between the tails simultaneously .
* d ) * : the 0.6 - 2.0 kev surface brightness profile across the `` protrusion '' .
the width of slices is 54.2@xmath29 ( or 17.7 kpc ) . the main tail and point sources are masked .
a gaussian fit with a constant background is also shown .
the fwhm of the gaussian is 4.05@xmath118 kpc .
the `` protrusion '' is a 4.1 @xmath31 feature . *
e ) * : the 0.6 - 2.0 kev surface brightness profile across the leading edge .
upstream of the edge , the surface brightness is nearly constant at 5.9@xmath37 cnt s@xmath3 kpc@xmath38 .
the thickness of the edge is less than 0.2 kpc ( or 0.61@xmath29 ) . ] for the main tail and 21.4@xmath29 for the secondary tail .
the end of the secondary tail is not curved as shown ( see figure 1 and 2 ) but we chose so for simplicity .
the dashed lines represent the local background .
the dotted lines in the plot of the main tail show the positions of the nucleus and the sharp bends .
we did not adjust the width of the region to include the entire bends and `` protrusion '' .
their corresponding positions in the secondary tail are also shown with the dotted lines .
the surface brightness around the nucleus is off scale ( see figure 3e ) , as we want to emphasize the faint regions . ]
range of the fwhm measured for the entire main tail ( figure 3b ) .
as the secondary tail is curved , the three parts where the width was measured have overlaps .
the result shows that the high density parts of x - ray tails are not expanding . ]
we measured temperatures in five regions of the main tail and two regions of the secondary tail ( see the left panel ) .
these temperatures should be viewed as spectroscopic temperatures for the multi - phase gas ( see section 3.2 for detail ) .
the dotted lines show the 1@xmath31 average temperature range in of the tail ( 0.773 - 0.838 kev ) , which is consistent with the temperatures of all regions .
dotted lines in the icm temperature plot are the 1@xmath31 average temperature ranges in the first six regions and the last six regions respectively . ] emission of .
the position of the leading h@xmath4 edge is consistent with that of the x - ray edge .
note that is 10.8@xmath39 off - axis in this observation so the psf is significantly blurred ( 50% - 90% encircled energy radius at 1.49 kev is 7.5@xmath29 - 14@xmath29 ) .
h@xmath4 absorption in the disk plane is visible upstream of the edge .
the h@xmath4 trail can be traced to at least 20 kpc from the nucleus , while the x - ray tail can be traced to at least 40 kpc from the nucleus .
note that the tail region is unfortunately close to a bright star ( @xmath119=10 mag ) .
the red scale bar is 10 kpc ( or 30.6@xmath29 ) . *
middle : * 0.6 - 2.0 kev contours in red superposed on the h@xmath66 image to show the galactic emission at low surface brightness . the black dashed line shows the edge of the i3 chip . the red scale bar is still 10 kpc ( or 30.6@xmath29 ) . *
right : * the h@xmath117 image of s center shows a clear dust lane .
the red scale bar is 2 kpc ( or 6.12@xmath29 ) . ] and x - rays ( see section 7.1 for detail ) .
there is little overlap now , mostly because of the lack of data .
multi - wavelength studies of tails are important to understanding mixing and other important physical processes .
the general connection of one - sided star formation behind cluster late - type galaxies with ram pressure stripping is also intriguing . ]
|
we present the discovery of spectacular double x - ray tails associated with and a possibly heated x - ray tail associated with , both late - type galaxies in the closest rich cluster abell 3627 .
a deep observation of allows us for the first time to examine the spatial and spectral properties of such x - ray tails in detail . besides the known bright tail that extends to @xmath0 80 kpc from , a fainter and narrower secondary tail with a similar length was surprisingly revealed , as well as some intriguing substructures in the main tail .
there is little temperature variation along both tails .
the widths of the secondary tail and the greater part of the main tail also remain nearly constant with the distance from the galaxy .
all these results challenge the current simulations .
the data also reveal 19 x - ray point sources around the x - ray tails .
we identified six x - ray point sources as candidates of intracluster ulxs with @xmath1 of up to 2.5@xmath2 erg s@xmath3 .
spectra of intracluster hii regions downstream of are also presented , as well as the velocity map of these hii regions that shows the imprint of s disk rotation .
for the first time , we unambiguously know that active star formation can happen in the cold ism stripped by icm ram pressure and it may contribute a significant amount of the intracluster light .
we also report the discovery of a 40 kpc x - ray tail of another late - type galaxy in a3627 , .
its x - ray tail seems hot , @xmath0 2 kev ( compared to @xmath0 0.8 kev for s tails ) .
the h@xmath4 data for are also presented .
we conclude that the high pressure environment around these two galaxies is important for their bright x - ray tails and the intracluster star formation .
the soft x - ray tails can reveal a great deal of the thermal history of the stripped cold ism in mixing with the hot icm , which is discussed along with intracluster star formation .
|
to achieve the required luminosity in a future e@xmath0e@xmath1 linear collider one has to produce two colliding beams at the interaction point ( ip ) with extremely small transverse beam dimensions . according to the linear collider design principles described in ref . 1
, the normalized beam emittance in the vertical plane ( the normalized beam emittance in the horizontal plane is larger ) at ip can be expressed as : @xmath2 where @xmath3 is the normalized beam energy , @xmath4 m is the classical electron radius , @xmath5 is the fine structure constant , @xmath6 is the maximum tolerable beamstrahlung energy spread , and @xmath7 is the mean number of beamstrahlung photons per electron at ip . taking @xmath8 and @xmath9 , one finds @xmath10mrad . to produce beams of this
small transverse emittance damping rings seem to be the only known facilities which have the potential to do this job .
the questions now are that once a beam of this small emittance is produced at the exit of the damping ring , how about the emittance growth during the long passage through the accelerating structures and the focusing channels from the damping ring at the beam energy of few gev to the ip with the beam energy of a few hundred of gev , and how to preserve it ?
many works have been dedicated to answer these questions @xcite@xcite@xcite@xcite . to start with , in sections 2 , 3 , and 4
, we consider the short range wakefield induced single bunch emittance growth and try to calculate the emittance growth by using two different methods , and show that the two methods give the same results . since the number of accelerating structures in the main linac of a linear collider is very large , the transverse random kicks on the particles can be described statistically .
firstly , we make use of the analogy between the transverse motion of a particle in a linear accelerator with the brownian motion of a molecule , which are governed by langevin equation . secondly , we solve directly fokker - planck equation . what should be noted
is that both methods are physically consistent . as a natural extension
, multibunch case is treated in section 5 .
comparisons with some numerical simulation results are made in section 6 .
the differential equation of the transverse motion of a bunch with zero transverse dimension is given as : @xmath11 @xmath12 where @xmath13 is the instantaneous betatron wave number at position @xmath14 , @xmath15 denotes the particle longitudinal position inside the bunch , and @xmath16 .
now we rewrite eq .
[ eq:1 ] as follows : @xmath17 where @xmath18 , @xmath19 , @xmath20 is the effective accelerating gradient in the linac , @xmath21 , @xmath22 , and @xmath23 is the deviation of the bunch head with respect to accelerating structures center . in this section
we consider the case where the injected bunch , quadrupoles and beam position monitors are well aligned , while the accelerating structures are misaligned . as a consequence
, @xmath24 is a random variable exactly the same as random accelerating structure misalignment errors with @xmath25 ( @xmath26 @xmath27 denotes the average over @xmath14 ) . if we take @xmath15 as a parameter and regard @xmath28 , @xmath13 , and @xmath29 as adiabatical variables with respect to @xmath14 , eq .
[ eq:2 ] can be regarded as langevin equation which governs the brownian motion of a molecule .
to make an analogy between the movement of the transverse motion of an electron and that of a molecule , we define @xmath30 , and regard @xmath31 as the particle s `` velocity '' random increment ( @xmath32 ) over the distance @xmath33 , where @xmath33 is the accelerating structure length . what we are interested is to assume that the random accelerating structure misalignment error follows gaussian distribution : @xmath34 and the velocity ( @xmath35 ) distribution of the molecule follows maxwellian distribution : @xmath36 where @xmath37 is the molecule s mass , @xmath38 is the boltzmann constant , and @xmath39 is the absolute temperature .
the fact that the molecule s velocity follows maxwellian distribution permits us to get the distribution function for @xmath40 @xcite : @xmath41 where @xmath42 by comparing eq .
[ eq:7b ] with eq .
[ eq:7a ] , one gets : @xmath43 or @xmath44 till now one can use all the analytical solutions concerning the random motion of a molecule governed by eq .
[ eq:2 ] by a simple substitution described in eq .
[ eq:7e ] . under the condition , @xmath45 ( adiabatic condition ) ,
one gets @xcite : @xmath46 @xmath47 @xmath48 @xmath49 @xmath47 @xmath50 @xmath51 @xmath52 @xmath53 where @xmath54 .
the asymptotical values for @xmath55 , @xmath56 , and @xmath57 as @xmath58 are approximately expressed as : @xmath59 @xmath60 @xmath61 inserting eqs .
[ eq:7 ] , [ eq:8 ] , and [ eq:8a ] into the definitions of the r.m.s .
emittance and the normalized r.m.s .
emittance shown in eqs .
[ eq:3 ] and [ eq:3a ] : @xmath62 @xmath63 one gets @xmath64 and @xmath65 the effects of energy dispersion within the bunch can be discussed through @xmath66 , @xmath67 , and @xmath68 , such as bns damping @xcite . from eqs .
[ eq:7 ] , [ eq:9 ] , and [ eq:9a ] one finds that there are three convenient types of variations of @xmath13 with respect to @xmath14 .
if one takes @xmath69 , one gets @xmath55 independent of @xmath14 .
if one takes , however , @xmath70 , @xmath71 is independent of @xmath14 , and finally , if @xmath72 , one has @xmath73 is independent of @xmath14 .
one takes usually the first scaling law in accordance with bns damping . to calculate the emittance growth of the whole bunch one has to make an appropriate average over the bunch , say gaussian as assumed above , as follows : @xmath74 to make a rough estimation one can replace @xmath75 by a delta function @xmath76 , and in this case
the bunch emittance can be still expressed by eq .
[ eq:9a ] with @xmath68 replaced by @xmath77 , where @xmath78 is the center of the bunch .
keeping the physical picture described above in mind , one can start directly with fokker - planck equation which governs the distribution function of the markov random variable , @xmath79 : @xmath80 with @xmath81 @xmath82 where @xmath83 is the increment of @xmath79 over @xmath33 , and @xmath84 @xmath85 denotes the average over a large number of a given type of possible structure misalignment error distributions ( in numerical simulations , this average corresponds to the average over the results obtained from a large number of different seeds , for a given type of structure misalignment error distribution function , say , gaussian distribution ) . from eq .
[ eq:2 ] one gets the increment of @xmath79 over @xmath33 : @xmath86 in consequence , one obtains : @xmath87 @xmath88 where @xmath89 has been used . inserting eqs . [ eq:16 ] and [ eq:17 ] into eq .
[ eq:13 ] , one gets : @xmath90 @xmath91 multiplying both sides with @xmath92 and integrating over @xmath79 , one has : @xmath93 assuming that @xmath94 , eq .
[ eq:19 ] is reduced to : @xmath95 solving eq .
[ eq:20 ] , one gets : @xmath96 where @xmath97 is the initial condition . apparently , when @xmath98 , one has : @xmath99 where @xmath100 .
. [ eq:22 ] is the same as what we have obtained in eq .
in fact , by solving directly fokker - planck equation we obtain the same set of asymptotical formulae derived in section 3 .
physically , the multibunch emittance growth is quite similar to that of the single bunch case , and each assumed point like bunch in the train can be regarded as a slice in the previously described single bunch .
obviously , the slice emittance expressed in eq .
[ eq:9a ] should be a good starting point for us to estimate the emittance growth of the whole bunch train . before making use of eq .
[ eq:9a ] let s first look at the differential equation which governs the transverse motions of the bunch train : @xmath101 where the subscript @xmath102 denotes the bunch number , @xmath103 is the distance between two adjacent bunches , @xmath104 is the particle number in each bunch , @xmath105 is the long range wakefield produced by each point like bunch at distance of @xmath14 .
clearly , the behaviour of the @xmath106 bunch suffers from influences coming from all the bunches before it , and we will treat one bunch after another in an iterative way .
first of all , we discuss about the long range wakefields . due to the decoherence effect in the long range wakefield only the first dipole mode will be considered . for a constant impedance structure as shown in fig . 1
, one has : @xmath107 where @xmath108 is the rms bunch length ( @xmath108 is used to calculate the transverse wake potential , and the point charge assumption is still valid ) , @xmath109 and @xmath110 are the angular frequency and the loaded quality factor of the dipole mode , respectively .
the loss factor @xmath111 in eq .
[ eq:777 ] can be calculated analytically as @xcite : @xmath112 @xmath113 @xmath114 where @xmath115 is the cavity radius , @xmath116 is the iris radius , @xmath117 is the cavity hight as shown in fig .
[ fig:1 ] , and @xmath118 is the first root of the first order bessel function . \2 .
true cm -7.5 true cm to reduce the long range wakefield one can detune and damp the concerned dipole mode .
the resultant long range wakefield of the detuned and damped structure ( dds ) can be expressed as : @xmath119 where @xmath120 is the number of the cavities in the structure .
when @xmath120 is very large one can use following formulae to describe ideal uniform and gaussian detuning structures @xcite : 1 ) uniform detuning : @xmath121 2 ) gaussian detuning : @xmath122 where @xmath123 , @xmath124 , @xmath125 is full range the synchronous frequency spread due to the detuning effect , @xmath126 is the @xmath127 width in gaussian frequency distribution .
once the long range wakefield is known one can use eq .
[ eq:7 ] to estimate @xmath128 in an iterative way , and the emittance of the whole bunch can be calculated accordingly as we will show later .
for example , if a bunch train is injected on axis ( @xmath129 ) into the main linac of a linear collider with structure _ rms _ misalignment @xmath130 , at the end of the linac one has : @xmath131 @xmath132 @xmath133 and in a general way , one has : @xmath134 finally , one can use the following formula to estimate the projected emittance of the bunch train : @xmath135 where @xmath136 ( since the bunch to bunch energy spread has been ignored ) , and @xmath137 is the average over the linac .
it is high time now for us to point out that the analytical expressions for the single and multibunch emittance growths established above give the statistical results of infinite number of machines with gaussian structure misalignment error distribution , which corresponds to using infinite seeds in numerical simulations .
to start with , we take the single bunch emittance growth in the main linac of sblc @xcite for example .
the short range wakefields in the accelerating s - band structures are obtained by using the analytical formulae @xcite and shown in fig .
[ fig1 ] . in the main linac
the beam is injected at 3 gev and accelerated to 250 gev with an accelerating gradient of 17 mv / m . the accelerating structure length @xmath1386 m , the average beta function @xmath139 is about 70 m ( @xmath140 for smooth focusing ) , the bunch population @xmath141 , the bunch length @xmath142 @xmath143 m , and and corresponding dipole mode short range wakefield @xmath144 v / pc / m@xmath145 .
inserting these parameters into eq .
[ eq:9a ] , one finds @xmath146 . if accelerating structure misalignment error @xmath147 @xmath143 m , one gets a normalized emittance growth of 8.66@xmath148 mrad , i.e. , 35@xmath149 increase compared with the designed normalized emittance of @xmath150 mrad .
the analytical result agrees quite well with that obtained from numerical simulations @xcite .
now , we apply the analytical formulae established for the multibunch emittance growth to sblc , tesla and nlc linear collider projects where enormous numerical simulations have been done .
the machine parameters are given in tables 1 to 4 which have been used in the analytical calculation in this paper .
firstly , we look at sblc . \1 . true cm \2 . true cm fig .
[ fig4 ] shows the `` kick factor '' @xmath151 defined in eqs .
[ eq:17 ] and [ eq:18 ] vs the dipole mode frequency
. + -3.8 true cm fig .
[ fig5](a ) gives the long range transverse wakefield produced by the first bunch at the locations where find the succeeding bunches , while fig .
[ fig5](b ) illustrates the square of the _ rms _ deviation of each bunch at the end of the linac with the dipole loaded quality factor @xmath152 .
the corresponding results for @xmath153 are shown in fig .
the normalized emittance growths compared with the design value at the interaction point ( @xmath154 mrad ) are 32@xmath149 and 388@xmath149 corresponding to the two cases , respectively as shown in table 4 , which agree well with the numerical results @xcite .
true cm 1.5 true cm 0.5 true cm to demonstrate the necessity of detuning cavities we show the violent bunch train blow up if constant impedance structures are used in spite of @xmath155 being loaded to 2000 as shown in fig .
[ fig5a ] . secondly , tesla ( the version appeared in ref .
10 ) is investigated . from fig .
[ fig7 ] one agrees that it is a no detuning case .
true cm \2 .
true cm \1 .
true cm from the results shown in fig .
[ fig8 ] and table 4 one finds that taking structure misalignment error @xmath156 m and @xmath157 one gets an normalized emittance growth of 24@xmath149 which is a very reasonable result compared what has been found numerically in ref .
thirdly , we look at nlc x - band main linac . to facilitate the exercise we assume
the detuning is effectuated as shown in fig .
[ fig9 ] ( in reality , nlc uses gaussian detuning ) .
[ fig10 ] shows the analytical results with @xmath158 m and @xmath159 . from table 4 one
finds a normalized emittance growth of 21@xmath149 .
then , we examine nlc s - band prelinac . assuming that the detuning of the dipole mode is shown in fig .
[ fig11 ] , one gets the multibunch transverse behaviour and the normalized emittance growth in fig .
[ fig12 ] and table 4 . finally , in figs .
[ fig13 ] and [ fig14 ] we give more information about the emittance growth vs @xmath155 in nlc x - band and s - band linacs .
1.3 true cm 1.5 true cm \1 .
true cm -4 .
true cm \1 .
true cm \2 .
true cm \1 .
true cm -4 .
true cm \1 .
true cm \2 .
true cm \1 .
true cm 0.1 true cm \1 .
true cm \0 .
true cm .the machine parameters i. [ cols="<,<,<,<,<,<,<,<",options="header " , ]
we treat the single and multibunch emittance growths in the main linac of a linear collider in analogy to the brownian motion of a molecule , and obtained the analytical expressions for the emittance growth due to accelerating structure misalignment errors by solving langevin equation .
as proved in this paper , the same set of formulae can be derived also by solving directly fokker - planck equation .
analytical results have been compared with those coming from the numerical simulations , such as sblc and tesla , and the agreement is quite well . as interesting applications ,
we give the analytical results on the estimation of the multibunch emittance growth in nlc x - band and s - band linacs .
it is a pleasure to discuss with j. le duff on the stochastic motions .
i thank f. richard for his reminding me the work of j. perez y jorba on inverse multiple touschek effect in linear colliders , and t.o .
raubenheimer for the discussion on the nlc parameters and detailed beam dynamics problems .
11 j. gao , `` parameter choices in linear collider designs '' , lal / rt 95 - 08 , and `` an s - band superconducting linear collider '' , proceedings of epac96 , barcelona , 1996 , p. 498 .
a. chao , b. richter , and c.y .
yao , `` beam emittance growth caused by transverse deflecting fields in a linear accelerator '' , _ nucl .
instr . and methods _ ,
* 178 * ( 1980 ) , p. 1 .
balakin , a.v .
novokhatsky , and v.p .
simirnov , `` vlepp : transverse beam dynamics '' , and `` stochastic beam heating '' , proc .
12th int . conf . on high energy accelerators , batavia , fermilab , usa ( 1983 ) , p. 119 , and p. 121 .
raubenheimer , `` the generation and acceleration of low emittance flat beams for future linear colliders '' , slac - report-387 , 1991 .
j. perez y jorba , `` increase of emittance in high energy e@xmath0e@xmath1 colliders by inverse multiple touschek effect in single bunches '' , _ nucl .
instr . and methods _ , * a297 * ( 1990 ) , p. 31 . s. chandrasekhar , `` stochastic problems in physics and astronomy '' , _ rev . of modern physics
15 , no . 1 ( 1943 ) , p. 1
. j. gao , `` analytical formulae and the scaling laws for the loss factors and the wakefields in disk - loaded periodic structures '' , _ nucl .
instr . and methods _ , * a381 * ( 1996 ) , p. 174 . j. gao , `` multibunch emittance growth and its corrections in s - band linear collider '' , _ particle accelerators _ , * vol .
49 * ( 1995 ) , p. 117 .
r. brinkmann , et al .
( editors ) , `` conceptual design of a 500 gev e@xmath0e@xmath1 linear collider with integrated x - ray laser facility '' , desy 1997 - 048 , vol .
ii , 1997 .
r. brinkmann , et al .
( editors ) , `` conceptual design of a 500 gev e@xmath0e@xmath1 linear collider with integrated x - ray laser facility '' , desy 1997 - 048 , vol .
|
in this paper the single and multibunch emittance growths in the main linac of a linear collider are analytically treated in analogy to the brownian motion of a molecule , and the analytical formulae for the emittance growth due to accelerating structure misalignment errors are obtained by solving langevin equation .
the same set of formulae is derived also by solving directly fokker - planck equation .
comparisons with numerical simulation results are made and the agreement is quite well .
= -1 cm 0 0 # 1#2#3_phys .
rep . _ * # 1 * ( # 2 ) # 3 # 1#2#3_phys .
rev . _
* d#1 * ( # 2 ) # 3 + 2.5 cm * analytical treatment of the emittance growth + in the main linacs of future linear colliders * * jie gao * * laboratoire de lacclrateur linaire + * in2p3-cnrs et universit de paris - sud , bp 34 , f-91898 orsay cedex
|
en el trabajo [ 1 ] se calcul la energa del estado base de sistemas de hasta 210 bosones ( tomos con espn entero ) confinados en una trampa .
el clculo se hizo por dos vas : analticamente utilizando los denominados aproximantes dobles de pad [ 11 ] y a partir del denominado mtodo variacional de monte carlo [ 12 ] .
la novedad del trabajo consisti en la extensin del tratamiento analtico empleado en [ 11 ] para calcular el espectro de energas del excitn ( un sistemas de dos partculas : un electrn y un hueco ) a sistemas de cientos de partculas .
para ello se hizo necesario calcular series para la energa en el lmite de interaccin dbil ( por teora de perturbaciones hasta segundo orden ) y en el lmite de interaccin fuerte ( oscilaciones de una denominada molcula de wigner ) . estas series son luego conectadas por medio de aproximantes dobles de pad .
por otro lado , el mtodo variacional de monte carlo combina el principio variacional de ritz con la tcnica de monte carlo para evaluar integrales multidimensionales .
siendo @xmath0 una funcin de prueba arbitraria , el principio variacional nos dice que la energa del estado base satisface la desigualdad @xmath1 donde @xmath2 es el hamiltoniano del problema y @xmath3 .
la integral , de dimension @xmath4 , es evaluada por monte carlo .
para ello generamos puntos @xmath5 con probabilidad dada por @xmath6 .
tenemos : @xmath7 el lado derecho de ( [ eq2 ] ) es el estimado para @xmath8 .
comparacin entre los resultados de monte carlo y los aproximantes dobles de pad para 20 y 210 bosones . ]
el valor de @xmath9 utilizado en nuestros clculos fue @xmath10 .
la funcin de prueba @xmath0 contenia correlaciones entre pares de partculas ( factores de jastrow ) y sus tres parmetros variacionales eran utilizados para minimizar la energa .
la generacin de puntos en el espacio multidimensional se hizo a partir del algoritmo de metropolis [ 12 ] .
este es un algoritmo secuencial en el que el punto @xmath11 se genera a partir de @xmath12 de acuerdo con lo siguiente : \(i ) se crea @xmath13 , donde @xmath14 es una perturbacin aleatoria .
( ii ) se genera un nmero aleatorio @xmath15 distribuido uniformemente en [ 0,1 ] .
( iii ) si @xmath16 , entonces @xmath17 , en caso contrario @xmath18 .
un ejemplo de los resultados de [ 1 ] se muestra en la fig .
[ fig1 ] .
el perfil de densidad , el potencial qumico y otras magnitudes en sistemas de @xmath19 - @xmath20 tomos confinados en trampas y enfriados por laser hasta temperaturas muy por debajo de 1k han sido medidas experimentalmente en fecha reciente [ 13 ] .
en el trabajo [ 2 ] se resalt e hizo uso de la analoga entre los ncleos atmicos y los sistemas de excitones en semiconductores .
un mtodo de clculo de la estructura nuclear muy empleado para ncleos pesados y que se basa en una funcin bcs [ 14 ] fue utilizado para calcular la energa de sistemas de hasta 210 pares de electrones y huecos , es decir hasta 420 partculas , con el objetivo de investigar cundo , o sea para qu densidades , el sistema se comporta como un gas de electrones y huecos y para qu densidades se comporta como un sistema de excitones descrito por la funcin bcs .
la funcin de onda variacional bcs , @xmath21 aparea huecos y electrones en estados con momento angular total igual a cero .
el estado @xmath22 del electrn tiene momento invertido respecto del hueco .
@xmath23 es la probabilidad de que no exista el par @xmath24 , mientras que @xmath25 es la probabilidad de que exista .
el principio de ritz @xmath26 nos da un estimado para @xmath8 . los coeficientes @xmath27 y @xmath28 se utilizan como parmetros variacionales sujetos a la ligadura @xmath29 .
la minimizacin conduce a la denominada ecuacin del gap @xmath30 donde @xmath31 es el potencial de interaccin entre pares , @xmath32 es la energa del estado @xmath33 renormalizada por la interaccin con el resto de los estados y @xmath34 es el potencial qumico que se determina de la ecuacin @xmath35 los coeficientes @xmath28 se han parametrizado en trminos de los @xmath36 as : @xmath37 las ecuaciones ( [ eq5 ] ) se resuelven iterativamente y muestran muy buenas propiedades de convergencia .
debido a que la funcin bcs no conserva el nmero de partculas se hace necesario un procedimiento de proyeccin .
en el trabajo se emple una proyeccin aproximada propuesta por lipkin y nogami [ 15 ] .
al resultar evidente que esta proyeccin era insuficiente se cre un mtodo estocstico basado en el algoritmo de metropolis que ofreci muy buenos resultados .
en la fig .
[ fig2 ] mostramos un ejemplo de estos resultados .
el mtodo de proyeccin estocstica mencionado en la seccin anterior no slo nos permiti hallar mejores valores para la energa en el caso de sistemas de excitones , sino que de vuelta a la fsica nuclear se calcul la energa de apareamiento en diversos modelos de estructura nuclear y en el istopo de zr con masa atmica 100 .
nuestros son comparables en exactitud con los mejores clculos cunticos por monte carlo de esta magnitud .
por su valor metodolgico , este trabajo fue publicado independientemente [ 3 ] .
la proyeccin de la funcion bcs sobre el sector de @xmath38 pares se escribe @xmath39 donde @xmath40 el estimado para la energa se obtiene de @xmath41 donde @xmath42 tiene interpretacin de energa cuando los @xmath38 estados @xmath43 estn ocupados y los coeficientes @xmath44 que son explcitamente mayores que cero , pueden interpretarse como factores de peso .
la expresin ( [ eq10 ] ) permite una evaluacin por monte carlo , donde el conjunto de estados @xmath45 se generan segn metropolis con peso @xmath46 .
en la fig .
[ fig3 ] se muestran algunos resultados en un modelo de estructura nuclear .
la ecuacin de bethe - goldstone es la base de la denominada aproximacin de pares independientes en ncleos [ 14 ] .
significa una mejora apreciable sobre los mtodos de hartree - fock o de partculas independientes en el sentido que se consideran las correlaciones de pares en la funcin de onda del ncleo .
ha sido utilizada extensivamente en fsica nuclear para el clculo de la energa de cohesin de la materia nuclear y de la energa ( masa ) de ncleos intermedios . en el trabajo [ 4 ] ,
la ecuacin de bethe - goldstone fue reformulada para sistemas de excitones confinados en puntos cunticos y empleada para el clculo de la energa de estos sistemas .
se investigaron sistemas desde el biexcitn ( 2 electrones - 2 huecos ) hasta conglomerados de 12 pares , considerando ademas sistemas cargados , 4e - 2h y 6e - 2h .
la diferencia @xmath47 entre las energas de n+1 y n excitones permite evaluar cmo depende la frecuencia del laser que crea un nuevo par con el nmero de pares presentes en el punto cuntico .
de nuestros resultados se dedujo adems la aparente inestabilidad del sistema 4e - 4h en un pozo cuntico .
la ecuacin de bethe - goldstone describe el movimiento de un par de fermiones que ocupan los estados @xmath48 y @xmath49 por debajo del nivel de fermi y se dispersan hacia estados por encima del nivel de fermi .
dicha ecuacin se escribe : @xmath50 donde @xmath51 son los operadores de energa de partculas libres en un campo externo y @xmath31 es la interaccin entre pares de partculas . @xmath52 realiza la proyeccin sobre los estados disponibles @xmath53 escribiendo : @xmath54 obtenemos para @xmath55 y @xmath56 : @xmath57 @xmath58 las ecuaciones ( [ eq15 ] ) constituyen un sistema lineal de donde se obtiene @xmath55 como funcin de @xmath56 , despus sustituyendo en ( [ eq16 ] ) obtenemos una ecuacin trascendente para @xmath56 .
la energa total se halla de : @xmath59 en la fig .
[ fig4 ] mostramos la energa de @xmath38 excitones @xmath60 calculadas por este mtodo .
en el trabajo [ 5 ] se calcul el coeficiente de absorcin , la luminiscencia y el perfil de densidades de electrones y huecos en puntos cunticos pequeos ( de 0 a 3 electrones ) en presencia de campos magnticos muy intensos , entre 8 y 50 teslas .
experimentos de luminiscencia en pozos cunticos ( donde existen infinitos electrones confinados en una regin cuasi - bidimensional ) bajo campos magnticos tan altos han sido reportados recientemente [ 16 ] . los estado finales de un proceso de absorcin e iniciales en uno de luminiscencia contienen un par electrn - hueco adicional . el sistema mas grande considerado fue , por tanto , el 4e - 1h .
las funciones de onda de onda de los estados iniciales y finales fueron obtenidas por diagonalizacin exacta de la ecuacin de schrodinger en bases de aproximadamente 5,000 funciones .
se obtuvieron resultados interesantes como la re - estructuracin de los estados de mas baja energa al variar el campo magntico , dando como resultado el corrimiento de los picos de absorcin y de luminiscencia , as como variaciones
abruptas del coeficiente de absorcin o de la luminiscencia para valores de campos correspondientes a los llenados 1/3 , 1/5 , etc del efecto hall cuntico . en la fig .
[ fig5 ] mostramos los mximos de luminiscencia correspondientes a multiexcitones con un exceso de carga @xmath61 , es decir la recombinacin desde sistemas con @xmath62 electrones y un hueco .
las resonancias dipolares gigantes ( gdr ) son excitaciones colectivas de ncleos [ 17 ] que juegan un importante papel en la foto - absorcin y en otras reacciones nucleares . en el trabajo [ 6 ] se muestra que estados muy similares existen en
sistemas de electrones y huecos en puntos cunticos semiconductores .
las energas de excitacin y los elementos de matriz del operador de dipolo son calculados a partir de la denominada aproximacin de fase aleatoria ( rpa ) en sistemas de hasta 110 pares e - h , concluyndose que en las gdr se concentra aproximadamente el 98 % de la absorcin de luz infra - roja en estos sistemas .
la rpa requiere como paso previo que el estado base sea obtenido en la aproximacin de hartree - fock ( hf ) , por lo que se implement la solucion iterativa de las ecuaciones de hf @xmath63c_{\gamma , u}^e c_{\gamma , v}^e \label{hf}\\ & - & \left .
\beta \sum_{\gamma\le\mu_f^h}\sum_{u , v } \langle s , u|1/r|t , v\rangle c_{\gamma , u}^h c_{\gamma , v}^h\right\ } c_{\alpha , t}^e = e_{e \alpha } c_{\alpha , s}^e\nonumber,\end{aligned}\ ] ] donde los orbitales electrnicos ocupados se expresan en trminos de las funciones de oscilador a travs de los coeficientes @xmath64 .
ecuaciones anlogas a ( [ hf ] ) se pueden escribir para la descomposicin de los orbitales correspondientes a los huecos .
note que los @xmath65 son las energas del oscilador en 2d @xmath66 por otro lado , en la rpa el estado base @xmath67 se supone que contiene correlaciones entre las partculas y los estados excitados se buscan en la forma @xmath68 donde el operador @xmath69 viene dado por la expresin @xmath70 el ndice @xmath71 corre sobre estados ocupados de hf , mientras que @xmath72 corre sobre estados desocupados ( por encima del nivel de fermi ) .
los coeficientes @xmath73 se determinan
por ecuaciones del tipo : @xmath74 en las cuales @xmath75 son las energ ' ias de excitacin y las matrices @xmath76 y @xmath77 vienen dadas por @xmath78 ) .
las soluciones fsicas aniquilan el estado base @xmath79 y satisfacen la condicin de normalizacin @xmath80 las ecuaciones ( [ rpaeq ] ) dan directamente las energas de excitacin de los estados excitados .
los elementos de matriz del operador dipolar , que causa la transicin del estado base a los estados excitados , se expresan en trminos de los coeficientes @xmath81 e @xmath82 .
la posicin de la gdr , es decir del estado que absorbe casi toda la fortaleza de la transicin , como funcin del nmero de pares electrn - hueco se muestra en la fig .
en fecha reciente , se ha observado experimentalmente [ 18 ] , que el 2 % de la foto - absorcin en ncleos se concentra en estados de mas baja energa de excitacin .
por contraposicin a las gdr , a estos estados se les ha llamado `` resonancias pigmeas '' .
los resultados del trabajo [ 7 ] muestran que la analoga con ncleos es casi perfecta pues los clculos de la absorcin en sistemas de electrones y huecos tambin dan como resultado una concentracin del 2 % restante en estados de baja energa de excitacin .
( ver fig .
[ fig7 ] ) .
intensidades de las resonancias gigantes y pigmeas en puntos cunticos . ]
el nmero de resonancias pigmeas , al igual que en ncleos , aumenta con el nmero total de partculas .
la posicin de la gdr y del promedio de las resonancias pigmeas tambin depende del nmero n de pares e - h en el punto cuntico .
esta dependencia es cualitativamente similar a la de ncleos , es decir la gdr se desplaza hacia energas mas altas al aumentar n , mientras que las resonancias pigmeas se desplazan hacia mas bajas energas .
experimentos recientes han mostrado la factibilidad de controlar el nmero de electrones en un punto cuntico ( desde uno hasta cuarenta ) , a la vez que reportan mediciones de la conductancia en los mismos [ 19 ] . en el trabajo [ 8 ] se tomaron parmetros que reprodujeran aproximadamente las condiciones de los experimentos en [ 19 ] y se calcul exactamente la conductancia y la densidad de niveles para un punto de 6 electrones .
hay que resaltar que el punto cuntico se halla separado de los electrodos por barreras de potencial , por lo que la conduccin a travs del mismo se realiza por medio de un proceso de tunelamiento resonante . las energas y funciones de onda de 6 y 7 electrones en el punto fueron obtenidas a partir de la diagonalizacin de la ecuacin de schrodinger en bases de hasta 40,000 funciones utilizando el mtodo de lanczos para tratar matrices tan grandes . densidad de niveles de energa para un punto cuntico con 6 electrones .
las barras representan los resultados de la diagonalizacin exacta del hamiltoniano , mientras que los smbolos son estimados a partir del coeficiente de trasmisin . ]
nuestros clculos revelan que a partir de una medicin precisa de la conductancia se puede obtener la densidad de estados excitados en sistemas de pocos electrones .
esta afirmacin se ilustra en la fig .
[ fig8 ] , donde la densidad de estados obtenida por diagonalizacin exacta es comparada con la que se obtiene a partir de la conductancia ( coeficiente de trasmisin ) calculada para un punto con 6 electrones .
el mtodo de lanczos hace uso del algoritmo de graham - schmidt para ortogonalizar una base de vectores que es creada por la propia matriz hamiltoniana ( que se supone simtrica ) .
debido a que la base es creada por la propia matriz , las propiedades de convergencia ante truncamiento son muy buenas .
quiere decir que para obtener los primeros 30 autovalores con buena aproximacin basta representar la matriz en el subespacio de los primeros 500 vectores de lanczos , aun cuando la dimensin de la matriz sea 40,000 .
sea @xmath83 un vector arbitrario .
la 1ra iteracin de lanczos consiste en lo siguiente : calculamos @xmath84 , @xmath85 , @xmath86 y definimos @xmath87 .
tenemos entonces : @xmath88 en la 2da iteracin calculamos : @xmath89 , @xmath90 , @xmath91 y definimos @xmath92 .
tenemos : @xmath93 las siguientes iteraciones se hacen de manera anloga .
notice que en la base @xmath94 el operador @xmath2 se representa por una matriz tridiagonal .
en el trabajo [ 9 ] se analizaron los efectos combinados de la energa de confinamiento del punto cuntico , la energa zeeman , la interaccin de coulomb y de campos magnticos intensos sobre la polarizacin de los espines electrnicos y la luminiscencia coherente en sistemas de hasta 40 pares e - h. posicin e intensidad de la lnea de luminiscencia como funcin del campo magntico y el nmero de excitones en el punto .
@xmath95 es la energa del excitn . ]
el esquema terico ya haba sido utilizado en [ 3 ] , es decir una funcin de onda bcs con proyeccin aproximada a lo lipkin - nogami .
la bondad de este mtodo en el caso de campos magnticos grandes consiste en que la funcin bcs reproduce el estado bsico exacto
un condensado de excitones [ 20] en el lmite de campos infinitos . mediciones de la luminiscencia en pozos cunticos como funcin de la intensidad y polarizacin de la luz que crea los excitones haban sido reportadas en [ 21 ] .
nuestros clculos arrojan slo un 10 % de polarizacin neta de los espines electrnicos , aun a campos tan altos como 20 teslas , debido a la fuerte interaccin entre electrones y huecos .
adems , se predice un corrimiento hacia el azul del pico de luminiscencia coherente con el aumento del nmero de pares e - h presentes en el punto ( o lo que es equivalente , aumento de la intensidad del laser ) .
las transiciones entre estados electrnicos internos , causadas por absorcin de fotones en el infra - rojo ( 5 - 10 mev ) , han sido estudiadas recientemente por medio de una tcnica , conocida como odr , consistente en observar sus efectos en las lneas de luminiscencia [ 22 ] . los experimentos han sido realizados en pozos cunticos y la lnea de luminiscencia tomada como referencia es la lnea fundamental del excitn . en el trabajo [ 10 ] , se calculan los niveles de energa , funciones de onda y elementos matriciales del operador dipolar para el biexcitn .
campos magneticos entre 0 y 5 teslas son considerados .
los parmetros utilizados ( masas de las partculas , etc ) corresponden a los de los experimentos [ 22 ] .
la mezcla de sub - bandas de valencia se consider aproximadamente .
el mtodo de clculo fue la diagonalizacin directa de la ecuacin de schrodinger en bases de hasta 10,000 funciones .
se hicieron predicciones para las posiciones e intensidades de las transiciones internas en el biexcitn , las cuales deben ser observadas siguiendo la linea biexcitnica en la odr .
se obtuvieron resultados interesantes sobre anti - cruzamiento de niveles y transferencia de fuerza de oscilador entre estados que colisionan . un ejemplo de estos resultados se muestra en la fig .
[ fig10 ] .
se han presentado resultados sobre propiedades fsicas de sistemas confinados del estado slido en que intervienen decenas o cientos de partculas .
se ha puesto de manifiesto la gran analoga de estos sistemas con los ncleos atmicos , lo cual fundamenta la aplicacin de mtodos de la fsica nuclear en la investigacin de los mismos .
los resultados reportados corresponden a trabajos conjuntos con r. capote , r. prez , b. rodrguez , a. delgado , e. menndez , b. partoens , a. matulis , f.m .
peeters , l. quiroga y f.j .
. 99 a. gonzalez , b. partoens , a. matulis and f. m. peeters .
b 59 ( 1999 ) 1653 - 1656 . b.a .
rodriguez , a. gonzalez , l. quiroga , f.j .
rodriguez . and r. capote .
b. 14 ( 2000 ) 71 - 83 .
r. capote and a. gonzalez .
c 59 ( 1999 ) 3477 - 3480 .
r. perez and a. gonzalez .
physica e 8 ( 2000 ) 91 - 98 .
a. gonzalez and e. menendez .
physica e 8 ( 2000 ) 333 - 341 .
a. delgado , l. lavin , r. capote and a. gonzalez .
physica e 8 ( 2000 ) 342 - 348 .
r. capote , a. delgado and a. gonzalez .
b 15 ( 2001 ) 81 - 87 . a. gonzalez and r. capote .
physica e 10 ( 2001 ) 528 - 534 .
b. rodriguez and a. gonzalez .
b 63 ( 2001 ) 205324 , 6 pages .
r. perez and a. gonzalez .
j. phys . : condens .
matter 13 ( 2001 ) l539 - l544 .
a.h . macdonald and d.s .
phys . rev .
b 33 , 8336 ( 1986 ) .
ceperley , lectures on quantum monte carlo , ictp , trieste ( 1997 ) . m.h .
anderson et .
al . , science 269 , 198 ( 1995 ) .
p. ring and p. schuck , the nuclear many - body problem , springer - verlag , new - york ( 1980 ) .
pradham et .
phys . a 201 , 357 ( 1973 ) ;
j. dobaczewski et .
c 47 , 2418 ( 1993 ) .
y. kim et .
phys . rev .
b 61 , 4492 ( 2000 ) .
j. speth ( ed . ) .
electric and magnetic giant resonances in nuclei , world scientific , singapore ( 1991 ) .
herzberg et .
c 60 , 051307(r ) ( 1999 ) ; k. govaert et .
al . phys .
c 57 , 2229 ( 1998 ) ; r.d .
herzberg e.t al .
b 390 , 49 ( 1997 ) ; j. reif et .
a 620 , 1 ( 1997 ) ; f. bauwens et .
c 62 , 024302 ( 2000 ) .
s. tarucha , d.g .
austing , and t. honda .
77 , 3613 ( 1996 ) ; t.h .
oosterkamp , j.w .
janssen , l.p .
kouwenhoven et .
82 , 2931 ( 1999 ) ; l.p .
kouwenhoven , t.h .
oosterkamp , m.w.s .
danoesastro et .
science 278 , 1788 ( 1997 ) ; s. sasaki , s.de franceschi , j.m .
elzerman et .
nature 405 , 764 ( 2000 ) ; s.de franceschi , s. sasaki , j.m .
elzerman et .
86 , 878 ( 2001 ) .
i.v . lerner and yu.e .
80 , 1488 ( 1981 ) [ sov .
phys.*jept 53 , * 763 ( 1981 ) ] .
damen , l. via , j.e .
cunningham , j. shah , and l.j .
67 , 3432 ( 1991 ) .
h.a . nickel , g. kioseoglou , t. yeo et .
b 62 , 2773 ( 2000 ) .
|
se resume un conjunto de resultados de perfil terico [ 1 - 10 ] , relativos a espectros de energas , densidad de estados excitados , polarizacin del espn , transporte a travs de barreras de potencial , y propiedades pticas ( absorcin en el infra - rojo , luminiscencia ) de sistemas tales como puntos cunticos semiconductores y trampas de tomos .
la caracterstica fundamental de los sistemas investigados consiste en que contienen un nmero de partculas entre 4 y 400 , lo cual hace muy difcil el clculo de sus propiedades . en los trabajos presentados se resalta y se hace amplio uso de la analoga entre estos sistemas y los ncleos atmicos por lo que se adaptaron mtodos de la fsica nuclear tales como : hartree - fock y la rpa para sistemas finitos , el mtodo bcs y la proyeccin de lipkin - nogami y la ecuacin de bethe - goldstone .
adems , se resolvi la ecuacin de schrodinger utilizando bases de funciones de hasta 40,000 elementos utilizando el algoritmo de lanczos y se implementaron otros mtodos como el de monte carlo variacional y los aproximantes dobles de pad .
por ltimo , se propuso un mtodo estocstico para proyectar la funcin de onda bcs .
a set of theoretical results [ 1 - 10 ] is reviewed , which concern calculations of energy spectra , density of energy levels , spin polarization , transport and optical properties ( infrared absorption , luminescence ) of semiconductor quantum dots and atomic traps .
the studied systems contain a number of particles between 4 and 400 , thus the calculation of their physical properties is a hard task .
the analogy between these systems and the atomic nuclei is stressed and used throughout the paper .
common nuclear physics methods like hartree - fock and rpa schemes for finite systems , the bcs approach and the lipkin - nogami projection , and the bethe - goldstone equation were adapted to the present context . on the other hand ,
the schrodinger equation was solved in basis with up to 40,000 functions by means of the lanczos algorithm , and other methods like variational monte carlo estimations and two - point pad approximants were also applied .
lastly , a stochastic projection of the bcs wave function was implemented .
|
the douglas rachford algorithm ( dra ) was first introduced in @xcite as an operator splitting technique to solve partial differential equations arising in heat conduction . as a result of findings by lions and mercier @xcite in the monotone operator setting
, the method has been extended to find solutions of the sum of two maximally monotone operators . when specialized to normal cone operators , the method is very useful in solving feasibility problems . to fix our setting , we assume throughout that equation i.e , a finite - dimensional real hilbert space with inner product @xmath0 and induced norm @xmath1 . given closed subsets @xmath2 and @xmath3 of @xmath4 with nonempty intersection , we consider the fundamental feasibility problem equation [ e : prob ] which frequently arises in science and engineering applications . a common approach for solving is to use projection algorithms that employ projectors onto the underlying sets ; see , e.g. , @xcite @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , and the references therein . among those algorithms , the douglas
rachford algorithm applied to has attracted much attention ; see , e.g. , @xcite and @xcite and the references therein for further information . in the convex case
, it is known , see , e.g. , lions and mercier @xcite and svaiter @xcite , that the sequence generated by the dra always converges while the `` shadow sequence '' converges to a point of the intersection .
even when the convex feasibility problem is inconsistent , i.e. , @xmath5 , it was shown in @xcite that the `` shadow sequence '' is bounded and its cluster points solve a best approximation problem ; the entire sequence converges if one of the sets is an affine subspace @xcite .
although the douglas rachford algorithm has been applied successfully to various problems involving one or more nonconvex sets , the theoretical justification is far from complete .
recently , in the case of a euclidean sphere and a line , borwein and sims @xcite have proved local convergence of the dra at points of the intersection , while aragn artacho and borwein @xcite have given a region of convergence for this model in the plane ; moreover , benoist @xcite has even shown that the dra sequence converges in norm to a point of the intersection except when the starting point belongs to the hyperplane of symmetry . in another direction , @xcite proved local convergence for finite unions of convex sets .
on the convergence rate , it has been shown by hesse , luke and neumann @xcite that the dra for two subspaces converges linearly . furthermore , the rate is then actually the cosine of the friedrichs angle between the subspaces @xcite . in the potentially nonconvex case , under transversality assumptions , hesse and luke @xcite proved local linear convergence of the dra for a superregular set and an affine subspace , while phan @xcite obtained such a rate for two super - regular sets .
specialized to the convex setting , the result in @xcite implies linear convergence of the dra for two convex sets whose the relative interiors have a nonempty intersection ; see also @xcite .
it is worth mentioning that the linear convergence of the dra may fail even for simple settings in the euclidean plane , as shown in @xcite .
based on hlder regularity properties , borwein , li , and tam @xcite established sublinear convergence for two convex basic semi - algebraic sets . for the linear convergence of the dra in the framework of optimization problems involving a sum of two functions , we refer the reader to , e.g. , giselsson s @xcite , @xcite , li and pong @xcite , liang , faili , peyr , and luke @xcite , patrinos , stella , and bemporad s @xcite , and the references therein . davis and
yin @xcite observed that the dra may converge arbitrarily slowly in infinite dimensions ; however , in finite dimensions , it often works extremely well .
very recently , the globally finite convergence of the dra has been shown in @xcite for an affine subspace and a locally polyhedral set , or for a hyperplane and an epigraph , and then by aragn artacho , borwein , and tam @xcite for a finite set and a halfspace . _
the goal of this paper is to provide various finite - convergence results . _
the sufficient conditions we present are new and complementary to existing conditions . after presenting useful results on projectors and the dra ( section [ s : aux ] ) and on locally identical sets ( section [ s : lis ] ) ,
we specifically derive results related to the following five scenarios : * @xmath2 is a halfspace and @xmath3 is an epigraph of a convex function ; @xmath2 is either a hyperplane or a halfspace , and @xmath3 is a halfspace ( see section [ s:4 ] ) . * @xmath2 and @xmath3 are supersets or modifications of other sets where the dra is better understood ( see section [ s : expand ] ) .
* @xmath2 and @xmath3 are subsets of other sets where the dra is better understood ( see section [ s : shrink ] ) .
* @xmath3 is a finite , hence nonconvex , set ( see section [ s : finiteset ] ) .
* @xmath2 is an affine subspace and @xmath3 is a polyhedron in the absence of slater s condition ( see section [ s:8 ] ) .
the paper concludes with a list of open problem in section [ s : open ] .
before we start our analysis , let us note that our notation and terminology is standard and follows , e.g. , @xcite .
the nonnegative integers are @xmath6 , and the real numbers are @xmath7 , while @xmath8 , @xmath9 , and @xmath10 .
let @xmath11 be a subset of @xmath4 .
then the closure of @xmath11 is @xmath12 , the interior of @xmath11 is @xmath13 , the boundary of @xmath11 is @xmath14 , and the smallest affine and linear subspaces containing @xmath11 are , respectively , @xmath15 and @xmath16 .
the relative interior of @xmath11 , @xmath17 , is the interior of @xmath11 relative to @xmath15 .
the smallest convex cone containing @xmath11 is @xmath18 , the orthogonal complement of @xmath11 is @xmath19 , and the dual cone of @xmath11 is @xmath20 .
the normal cone operator of @xmath11 is denoted by @xmath21 , i.e. , @xmath22 if @xmath23 , and @xmath24 otherwise . if @xmath25 and @xmath26 , then @xmath27 is the closed ball centered at @xmath28 with radius @xmath29 .
for the reader s convenience we recall in this section preliminary concepts and auxiliary results which are mostly well known and which will be useful later . let @xmath2 be a nonempty closed subset of @xmath4 .
the _ distance function _ of @xmath2 is @xmath30 the _ projector _ onto @xmath2 is the mapping @xmath31 and the _ reflector _ across @xmath2 is defined by @xmath32 where @xmath33 is the identity operator .
note that closedness of the set @xmath2 is necessary and sufficient for @xmath2 to be proximinal , i.e. , @xmath34 @xmath35 ( see , e.g. , ( * ? ? ?
* corollary 3.13 ) ) . in the following , we shall write @xmath36 if @xmath37 is a singleton .
[ f : proj ] let @xmath2 be a nonempty closed convex subset of @xmath4 , and let @xmath28 and @xmath38 be in @xmath4 . then the following hold : 1 .
[ f : proj_cvx ] @xmath39 is single - valued and @xmath40 \quad\leftrightarrow\quad x -p \in n_a(p).\ ] ] 2 .
[ f : proj_firm ] @xmath39 is _ firmly nonexpansive _ , i.e. , @xmath41 3 .
[ f : proj_nonex ] @xmath42 is _ nonexpansive _ , i.e. , @xmath43 in particular , @xmath39 and @xmath42 are continuous on @xmath4 .
[ f : proj_cvx ] : ( * ? ? ? * theorem 3.14 and proposition 6.46 ) .
[ f : proj_firm ] : ( * ? ? ?
* proposition 4.8 ) .
[ f : proj_nonex ] : ( * ? ? ?
* corollary 4.10 ) .
[ l : projab ] let @xmath2 and @xmath3 be closed subsets of @xmath4 such that @xmath44 , and let @xmath25 . then the following hold : 1 .
[ l : projab_sub ] @xmath45 .
[ l : projab_preimage ] @xmath46 @xmath47 .
[ l : projab_single ] if @xmath48 , then @xmath49 .
[ l : projab_cvx ] if @xmath3 is convex and @xmath50 , then @xmath49 .
[ l : projab_sub ] : the conclusion is obvious if @xmath51 .
assume @xmath52 , and let @xmath53 .
then @xmath54 for all @xmath55 , and so for all @xmath56 since @xmath44 .
this combined with @xmath57 gives @xmath58 .
[ l : projab_preimage ] : let @xmath57 . for all @xmath59 , we have @xmath60 , and by [ l : projab_sub ] , @xmath61 , which implies @xmath62 .
[ l : projab_single ] : assume that @xmath48 . using [ l : projab_sub ] , we have @xmath63 , and so @xmath64 it follows that @xmath65 .
l : projab_cvx ] : by [ f : proj][f : proj_cvx ] , if @xmath3 is convex , then @xmath66 is a singleton , and if additionally @xmath50 , then by [ l : projab_single ] , @xmath49 .
[ ex : proj_affine ] let @xmath67 be a real hilbert space , let @xmath68 be a linear operator from @xmath4 to @xmath67 , let @xmath69 , and set @xmath70
. then @xmath71 where @xmath72 denotes the moore penrose inverse of @xmath68 .
this follows from ( * ? ? ?
* lemma 4.1 ) , see also ( * ? ? ?
* example 28.14 ) .
[ ex : proj_h ] let @xmath73 , and let @xmath74 . then the following hold : 1 .
[ ex : proj_hyper ] if @xmath75 , then @xmath76 2 .
[ ex : proj_half ] if @xmath77 , then @xmath78 [ ex : proj_hyper ] : ( * ? ? ?
* example 28.15 ) .
[ ex : proj_half ] : ( * ? ? ?
* example 28.16 ) .
[ ex : proj_ball ] let @xmath79 with @xmath80 and @xmath26 .
then @xmath81 let @xmath25 .
we have to prove @xmath82 if @xmath83 , and @xmath84 otherwise .
indeed , if @xmath83 , then @xmath85 , and thus @xmath82 .
assume that @xmath86 .
on the one hand , for all @xmath55 , by using @xmath87 and the triangle inequality , @xmath88 on the other hand , @xmath89 , and so @xmath90 , then by combining with the convexity of @xmath3 and the above inequality , @xmath91 , which completes the formula .
[ ex : proj_epi ] let @xmath92 be convex and continuous , set @xmath93 , and let @xmath94 .
then there exists @xmath95 such that @xmath96 , @xmath97 and @xmath98 see ( * ? ? ?
* lemma 5.1 ) . in order to solve the feasibility problem , where @xmath2 and @xmath3 are closed subsets of @xmath4 with nonempty intersection
, we employ the _ douglas rachford algorithm _ ( also called _ averaged alternating reflections _ ) that generates a sequence @xmath99 by equation [ e : draseq ] ( @xmath100)x_n+1 t_a , bx_n , x_0 x , and where equation t_a , b : = ( @xmath101+r_br_a ) is the _ douglas rachford operator _ associated with the ordered pair @xmath102 .
the sequence @xmath99 in is called a _ dra sequence with respect to @xmath102 _ , with starting point @xmath103 . by
[ f : proj][f : proj_cvx ] , when @xmath2 and @xmath3 are convex , then @xmath39 , @xmath104 and hence @xmath105 are single - valued .
notice that @xmath106 and if @xmath39 is single - valued then @xmath107 in the sequel we adopt the convention that in the case where @xmath108 is not a singleton , @xmath109 .
the set of fixed points of @xmath105 is defined by @xmath110 .
it follows from @xmath111 that @xmath112 and that modified for clarity @xmath113 for the convex case , the basic convergence result of the dra sequence @xmath99 and the `` shadow sequence '' @xmath114 is as follows .
[ f : cvg ] let @xmath2 and @xmath3 be closed convex subsets of @xmath4 with @xmath115 , and let @xmath99 be a dra sequence with respect to @xmath102 . then the following hold : 1 .
[ f : cvg_fixt ] @xmath116 and @xmath117 .
[ f : cvg_anb ] if @xmath118 , then @xmath119 ; the convergence is finite provided that @xmath120 .
[ f : cvg_fixt ] : this follows from ( * ? ? ?
* theorem 1 ) and ( * ? ? ? * theorem 1 ) ; see also ( * ? ? ?
* corollary 3.9 and theorem 3.13 ) .
[ f : cvg_anb ] : clear from ( * ? ? ?
* lemma 3.2 ) .
[ d : local ] let @xmath2 and @xmath3 be subsets of @xmath4 such that @xmath121
. then @xmath2 and @xmath3 are called _ locally identical _ around @xmath122 if there exists @xmath123 such that @xmath124 .
we say that @xmath2 and @xmath3 are locally identical around a set @xmath125 if they are locally identical around every point in @xmath11 .
when @xmath2 and @xmath3 are locally identical around a point @xmath126 ( respectively , a set @xmath11 ) , we also say that @xmath102 is locally identical around @xmath126 ( respectively , @xmath11 ) .
[ l : id ] let @xmath2 and @xmath3 be subsets of @xmath4 such that @xmath121 .
then the following hold : 1 .
[ l : id_int ] @xmath2 and @xmath3 are locally identical around @xmath127 .
[ l : id_intersect ] if @xmath2 and @xmath3 are locally identical around @xmath122 , then @xmath2 , @xmath3 and @xmath128 are also locally identical around @xmath126 .
[ l : id_sup ] if @xmath44 , and @xmath126 is a point in @xmath2 such that @xmath129 , then @xmath2 and @xmath3 are locally identical around @xmath126 .
[ l : id_sub ] if @xmath2 is closed convex , and @xmath11 is a closed subset of @xmath2 such that @xmath2 and @xmath11 are locally identical around @xmath11 , then @xmath130 .
[ l : id_cvx ] if @xmath2 and @xmath3 are closed convex and locally identical around @xmath128 , then @xmath131 .
[ l : id_int ] : let @xmath132 .
then there exists @xmath123 such that @xmath133 , which implies @xmath134 , so @xmath2 and @xmath3 are locally identical around @xmath126 .
[ l : id_intersect ] : note that if @xmath124 then @xmath135 .
[ l : id_sup ] : since @xmath129 , there exists @xmath123 such that @xmath136 . combining with @xmath44 , we get @xmath137 .
[ l : id_sub ] : let @xmath138 .
it suffices to show that @xmath139 suppose to the contrary that does not hold .
since @xmath2 and @xmath11 are locally identical around @xmath11 which includes @xmath126 , @xmath140 then @xmath141\bar\varepsilon , + \infty[)$ ] @xmath142 .
now let @xmath143 and @xmath144 by the boundedness of @xmath145 and the closedness of @xmath2 , we assume without loss of generality that @xmath146 .
it follows from @xmath147 that @xmath148 . by the convexity of @xmath2 , @xmath1490 , 1[)$ ] @xmath150 , and @xmath151 , which yields @xmath152 , using the definition of @xmath153 . from @xmath154 and the closedness of @xmath11 , letting @xmath155 , we obtain @xmath156 , thus @xmath2 and @xmath11 are locally identical around @xmath157 , i.e. , @xmath158 for some @xmath26 . since @xmath159 , we find @xmath160 satisfying @xmath161 .
then @xmath162 , which contradicts the fact that @xmath163 @xmath164 .
therefore , holds .
now pick an arbitrary @xmath165 , and let @xmath166 . by combining with , @xmath167 , and so @xmath156 .
it follows that @xmath168 , which gives @xmath130 .
[ l : id_cvx ] : set @xmath169 . then @xmath11 is closed , @xmath170 , @xmath171 , and by [ l : id_intersect ] , @xmath2 , @xmath3 and @xmath11 are locally identical around @xmath11 .
now apply [ l : id_sub ] .
the following example illustrates that the assumption on convexity of @xmath2 in [ l : id][l : id_sub ] is important .
suppose that @xmath172 , that @xmath173 and that @xmath174
. then @xmath2 and @xmath11 are closed and locally identical around @xmath11 , and @xmath175 , but @xmath176 .
this does not contradict [ l : id][l : id_sub ] because @xmath2 is not convex .
[ l : proj ] let @xmath2 and @xmath3 be closed subsets of @xmath4 , and assume that @xmath2 and @xmath3 are locally identical around some @xmath122 , say there exists @xmath123 such that @xmath124 .
let @xmath177 then the following hold : 1 .
[ l : proj_subs ] if @xmath44 , then @xmath34 @xmath178 .
[ l : proj_cvx ] if @xmath2 and @xmath3 are convex , then @xmath34 @xmath179 @xmath180 @xmath181 .
equivalently , if @xmath2 and @xmath3 are convex then @xmath182 .
[ l : proj_sub ] if @xmath44 and @xmath3 is convex , then @xmath34 1 .
[ l : proj_sub.a ] @xmath183 ; 2 .
[ l : proj_sub.b ] @xmath184 ; 3 .
[ l : proj_sub.c ] @xmath185 .
[ l : proj_subs ] : observe that @xmath186 .
the conclusion follows [ l : projab][l : projab_sub ] . to prove [ l : proj_cvx ] and [ l : proj_sub ] , note that since @xmath187 , there exists @xmath26 such that @xmath188 , which yields @xmath189 [ l : proj_cvx ] : by ( * ? ? ?
* lemma 2.12 ) , it follows from @xmath190 and that @xmath191 . now using , @xmath34 @xmath179 @xmath180 @xmath192 @xmath180 @xmath181 .
hence , @xmath182 .
[ l : proj_sub.a ] : let @xmath25 .
assume that @xmath193
. then @xmath194 . by [ l : projab][l : projab_cvx ] , @xmath49 .
[ l : proj_sub.b ] : using and applying [ l : proj_cvx ] for two convex sets @xmath195 and
@xmath3 , we obtain @xmath196 . next applying [ l : projab][l : projab_preimage ] for @xmath197 and @xmath44
, we have @xmath198 , and so @xmath199 .
[ l : proj_sub.c ] : now assume @xmath63 .
then [ l : proj_sub.b ] gives @xmath48 , and [ l : projab][l : projab_cvx ] gives @xmath200 .
in this section , we assume that equation , and that equation @xmath201f : = \{(x , ) x | f(x)}. in the space @xmath202 , we set equation h : = x\{0 } b : = @xmath201f .
then the projection onto @xmath203 is given by @xmath204 the projection onto @xmath3 is described as in [ ex : proj_epi ] , and the effect of performing each step of the dra applied to @xmath203 and @xmath3 is characterized in the following result . [
f : drstep ] let @xmath205 , and set @xmath206 . then the following hold : 1 .
[ f : drstep_form ] if @xmath207 , then @xmath208 .
otherwise , there exists @xmath209 such that @xmath210 in which either ( @xmath211 and @xmath212 ) or ( @xmath213 and @xmath214 ) .
[ f : drstep_pos ] @xmath215 , or equivalently , @xmath216 @xmath217 .
[ f : drstep_form ] : ( * ? ? ?
* corollary 5.3(i)&(ii ) ) .
[ f : drstep_pos ] : clear from [ f : drstep_form ] .
we have the following result on convergence of the dra in the case of a hyperplane and an epigraph .
[ f : p - e ] suppose that @xmath218 given a starting point @xmath219 , generate the dra sequence @xmath220 by @xmath221 then @xmath220 converges _ finitely _ to a point in @xmath128 .
see ( * ? ? ?
* theorem 5.4 ) . in view of [ f : p - e ] , it is natural to ask about the convergence of the dra when @xmath2 is a halfspace instead of a hyperplane .
[ t : h - e ] suppose that either 1 .
[ t : h - e+ ] @xmath222 and @xmath223 , or 2 .
[ t : h - e- ] @xmath224 and @xmath223 with @xmath225 .
then the dra sequence converges _ finitely _ to a point in @xmath128 .
[ t : h - e+ ] : let @xmath205 . if @xmath226 , then @xmath227 , and so @xmath228 due to [ f : drstep][f : drstep_pos ] .
if @xmath229 , we are done . if @xmath230 , then @xmath231 , @xmath232 , and by [ ex : proj_epi ] , @xmath233 with @xmath234 , which implies @xmath235 .
we deduce that the dra sequence converges in at most two steps .
[ t : h - e- ] : if @xmath236 , then @xmath237 , @xmath238 , and @xmath239 , which gives @xmath240 if @xmath241 , and @xmath242 otherwise .
it is thus sufficient to consider the case @xmath243 .
then @xmath244 , and so @xmath245 due to [ f : drstep][f : drstep_pos ] .
this implies that @xmath246 now apply [ f : p - e ] .
the following example whose special cases can be found in @xcite illustrates that the slater s condition @xmath225 in [ f : p - e ] and [ t : h - e][t : h - e- ] is important .
[ ex : epi_infinite ] suppose that either @xmath247 or @xmath248 , that @xmath223 with @xmath249 , and that @xmath250 is differentiable at its minimizers ( if they exist ) .
let @xmath251 , where @xmath103 is not a minimizer of @xmath250 , and generate the dra sequence @xmath220 as in .
then @xmath252 and thus also @xmath220 do not converge finitely .
firstly , we claim that if @xmath253 , where @xmath28 is not a minimizer of @xmath250 , then @xmath254 and @xmath255 is not a minimizer of @xmath250 . indeed , by assumption , @xmath256 , so @xmath227 , and then @xmath257 .
by using [ f : drstep][f : drstep_form ] , @xmath212 and @xmath258 if @xmath255 is a minimizer of @xmath250 , then @xmath259 , and by [ e : decrease ] , @xmath260 is a minimizer , which is absurd .
hence , the claim holds . as a result , @xmath261 now assume that @xmath262 converges finitely .
then there exists @xmath263 such that @xmath264 . using again [ e : decrease ] ,
we get @xmath265 , which contradicts [ e : xn ] .
[ t : h - h ] suppose that @xmath2 is either a hyperplane or a halfspace , that @xmath3 is a halfspace of @xmath4 , and that @xmath121
. then every dra sequence @xmath99 with respect to @xmath102 converges _ finitely _ to a point @xmath28 , where @xmath266 or @xmath163 @xmath267 with @xmath268 .
if @xmath269 , i.e. , @xmath270 , then the result is trivial , so we will work in the space @xmath202 with @xmath271 , and denote by @xmath220 the dra sequence .
after rotating the sets if necessary , we can and do assume that @xmath272 , and @xmath273 , with @xmath274 and @xmath74 .
noting that @xmath275 , we distinguish the following three cases
. _ case 1 : _ @xmath276 .
then @xmath277 is the epigraph of the linear function @xmath278 if @xmath225 , we are done due to [ t : h - e][t : h - e- ] .
assume that @xmath249 .
then @xmath279 since @xmath73 implies @xmath280 .
now in turn , @xmath34 @xmath281 , and so @xmath282 , which gives @xmath283 . by the assumption that @xmath121 , we must have @xmath284 , and then @xmath285 . let @xmath205 . if @xmath286 , then @xmath287 , and @xmath288 , which gives @xmath289 , i.e. , @xmath290 , in which case @xmath291 . if @xmath292 , then @xmath293 and @xmath232 , @xmath294 , so @xmath295 _ case 2 : _
@xmath296 . then
@xmath297 after reflecting the sets across the hyperplane @xmath298 , we have @xmath299 , and @xmath3 is the epigraph of a linear function .
now apply [ t : h - e][t : h - e+ ] .
_ case 3 : _ @xmath300
. then @xmath73 and @xmath301 let @xmath205 .
if @xmath302 , we are done .
if @xmath303 , then @xmath304 , @xmath305 , and by [ ex : proj_h][ex : proj_half ] , @xmath306 which is also in @xmath272 and which yields @xmath307 now assume that @xmath308 .
we have @xmath309 and @xmath287 .
if @xmath310 , then @xmath311 , and @xmath312 . finally , if @xmath313 , then again by [ ex : proj_h][ex : proj_half ] , @xmath314 and thus , @xmath315 moreover , @xmath316 , so @xmath317 , and we get @xmath318 . the proof for the ( hyperplane , halfspace ) case is similar and uses [ f : p - e ] .
[ l : expand ] let @xmath2 and @xmath3 be closed ( not necessarily convex ) subsets of @xmath4 such that @xmath121 , and let @xmath103 be in @xmath4 .
suppose that the dra sequence @xmath99 with respect to @xmath102 , with starting point @xmath103 , converges to @xmath319 .
suppose further that there exist two closed convex sets @xmath320 and @xmath321 in @xmath4 such that @xmath322 , @xmath323 , and that both @xmath324 and @xmath325 are locally identical around some @xmath326 . then @xmath327 , @xmath328 and @xmath329 i.e. , @xmath330 . by assumption and [ l : proj][l
: proj_sub.c ] , @xmath331 , and so @xmath332 .
since @xmath319 , it follows from that @xmath333 . using again [ l : proj][l : proj_sub.c ] , @xmath334 .
we get @xmath335 , and again by , @xmath328 .
now by the definition of @xmath320 and @xmath321 , there exists @xmath123 such that @xmath336 there exists @xmath160 such that @xmath337 let @xmath338 . since @xmath339 ,
@xmath340 are ( firmly ) nonexpansive and @xmath341 is nonexpansive ( [ f : proj][f : proj_firm]&[f : proj_nonex ] ) , @xmath342 and also @xmath343 thus , @xmath344 and @xmath345 . by [ l : proj][l : proj_sub.a ] , @xmath346 and @xmath347 , which implies @xmath348 and @xmath349 .
we deduce that @xmath350 . if the assumption that @xmath322 and @xmath323 in [ l : expand ] is replaced by the assumption on convexity of @xmath2 and @xmath3 , then still holds , as shown in the following lemma .
we shall now look at situations where @xmath351 are modifications of @xmath352 that preserve local structure .
[ l : adjust ] let @xmath2 and @xmath3 be closed convex subsets of @xmath4 such that @xmath121 , and let @xmath99 be the dra sequence with respect to @xmath102 , with starting point @xmath353 .
suppose that there exist two closed convex sets @xmath320 and @xmath321 in @xmath4 such that both @xmath324 and @xmath325 are locally identical around @xmath268 , where @xmath319 is the limit of @xmath99 .
then @xmath354 i.e. , @xmath330 . recall from [ f : cvg][f : cvg_fixt ] that @xmath355 with @xmath268 . setting @xmath356 , from the assumption on @xmath320 and @xmath321 ,
there is @xmath123 such that @xmath336 furthermore , there exists @xmath160 such that @xmath337 let @xmath338 . according to [ f : proj][f : proj_firm]&[f : proj_nonex ] , @xmath39 , @xmath104 are ( firmly ) nonexpansive and @xmath42 is nonexpansive , so @xmath357 and also @xmath358 therefore , @xmath359 and @xmath360 . using [ l : proj][l : proj_cvx ] , @xmath346 and @xmath361 .
hence @xmath348 and @xmath349 .
we obtain that @xmath350 .
[ t : global ] let @xmath2 and @xmath3 be closed convex subsets of @xmath4 such that @xmath121 .
suppose that there exist two closed convex sets @xmath320 and @xmath321 in @xmath4 such that both @xmath324 and @xmath325 are locally identical around @xmath128 .
then for any dra sequence @xmath99 with respect to @xmath102 in , @xmath362 and this is still true when exchanging the roles of @xmath105 and @xmath363 in and . by [ f :
cvg][f : cvg_fixt ] , @xmath355 with @xmath268 .
now apply [ l : adjust ] .
let us exchange the roles of @xmath105 and @xmath363 in and , i.e. , @xmath163 @xmath364 , and we shall prove that @xmath365 by the assumption on @xmath320 and @xmath321 , we have @xmath366 , @xmath367 , and for all @xmath122 , there exists @xmath123 such that @xmath336 then @xmath368 therefore , @xmath128 and @xmath369 are locally identical around @xmath128 . noting that @xmath128 and @xmath369 are closed convex , and @xmath370 , [ l : id][l : id_sub ] gives @xmath371 .
next again by [ f : cvg][f : cvg_fixt ] , @xmath372 with @xmath373 . by assumption , both @xmath324 and @xmath325 are locally identical around @xmath374 , and hence the proof is completed by applying [ l : adjust ] . in the following ,
we say that the dra applied to @xmath102 _ converges finitely globally _ if the sequence @xmath375 converges finitely for all @xmath25 .
[ t : finite ] let @xmath2 and @xmath3 be nonempty closed convex subsets of @xmath4 .
then the dra applied to @xmath102 converges _ finitely globally _ provided one of the following holds : 1 .
[ t : finite_abdryb ] @xmath121 and @xmath376 ; equivalently , @xmath377 .
[ t : finite_ab ] @xmath378 and there exist two closed convex sets @xmath320 and @xmath321 in @xmath4 such that both @xmath324 and @xmath325 are locally identical around @xmath379 , and that the dra applied to @xmath380 converges finitely globally when @xmath381 .
[ t : finite_aintb ] @xmath382 , @xmath378 and there exist two closed convex sets @xmath320 and @xmath321 in @xmath4 such that both @xmath324 and @xmath325 are locally identical around @xmath379 , and that the dra applied to @xmath380 converges finitely globally when @xmath383 . let @xmath99 be a dra sequence with respect to @xmath102 .
[ t : finite_abdryb ] : it follows from @xmath121 , @xmath376 and the closedness of @xmath3 that @xmath384 , and so @xmath118 . by [ f : cvg][f : cvg_anb ] , @xmath385 finitely . now if @xmath377 , then @xmath386 , and @xmath387 , which implies @xmath376 .
conversely , assume that @xmath121 and @xmath376 .
let @xmath165 . then @xmath388
we have to show @xmath389 .
suppose to the contrary that @xmath390 .
pick @xmath391 . by convexity , @xmath392 : = { \{{\lambda a + ( 1 -\lambda)b}~\big |~{0 \leq \lambda \leq 1}\}}\subseteq a$ ] , and so @xmath392\cap { \ensuremath{\operatorname{bdry}}}b = \varnothing$ ] , which is impossible since @xmath393 and @xmath394 . hence ,
@xmath389 for all @xmath165 .
this means @xmath377 .
[ t : finite_ab ] : by assumption , @xmath395 , and so the dra applied to @xmath380 converges finitely globally .
if @xmath396 , then both @xmath324 and @xmath325 are locally identical around @xmath397 ( using the closedness of @xmath3 ) , and using [ t : global ] , @xmath398 which implies the finite convergence of @xmath99 due to the finite convergence of the dra applied to @xmath380 .
next assume that @xmath382 . then @xmath399 . by [ f
: cvg][f : cvg_anb ] , @xmath119 , and this convergence is finite when @xmath120 .
it thus suffices to consider the case when @xmath400 .
then @xmath324 and @xmath325 are locally identical around @xmath401 , and by [ l : adjust ] , holds . using again the finite convergence of the dra applied to @xmath380 ,
we are done .
[ t : finite_aintb ] : first , we show that @xmath383 .
let @xmath402 . by the assumption on @xmath320 and @xmath321
, there is @xmath123 such that @xmath336 now let @xmath403 .
then @xmath404 , and by the convexity of @xmath2 and @xmath3 , ( * ? ? ?
* proposition 3.35 ) implies @xmath405c , d\right ] : = { \{{\lambda c + ( 1 -\lambda)d}~\big |~{0 \leq \lambda < 1}\ } } \subseteq a\cap { \ensuremath{\operatorname{int}}}b$ ] .
therefore , @xmath405c , d\right ] \cap { \ensuremath{\operatorname{int}}}{\operatorname{ball}\left({c};{\varepsilon}\right ) } \subseteq a\cap { \operatorname{ball}\left({c};{\varepsilon}\right ) } = a'\cap { \operatorname{ball}\left({c};{\varepsilon}\right ) } \subseteq a'$ ] and @xmath405c , d\right ] \cap { \ensuremath{\operatorname{int}}}{\operatorname{ball}\left({c};{\varepsilon}\right ) } \subseteq { \ensuremath{\operatorname{int}}}(b\cap { \operatorname{ball}\left({c};{\varepsilon}\right ) } ) = { \ensuremath{\operatorname{int}}}(b'\cap { \operatorname{ball}\left({c};{\varepsilon}\right ) } ) \subseteq { \ensuremath{\operatorname{int}}}b'$ ] .
we deduce that @xmath383 . by assumption ,
the dra applied to @xmath380 converges finitely globally .
now argue as the case where @xmath382 in the proof of part [ t : finite_ab ] .
[ c : linear ] let @xmath2 and @xmath3 be closed convex subsets of @xmath4 such that @xmath121 .
suppose that both @xmath406 and @xmath407 are locally identical around @xmath379 when @xmath378 .
then every dra sequence @xmath99 with respect to @xmath102 converges _ linearly _ with rate @xmath408 to a point @xmath319 with @xmath268 , where @xmath409 is the _ cosine of the friedrichs angle _ between two subspaces @xmath410 and @xmath411 defined by @xmath412 if @xmath376 , then by [ t : finite][t : finite_abdryb ] , we are done .
now assume that @xmath378 . by assumption and
[ t : global ] , @xmath413 since we work with a finite - dimensional space , ( * ? ? ?
* corollary 4.5 ) completes the proof .
suppose that @xmath414 , that @xmath415 $ ] , and that @xmath416 .
then @xmath417 . by (
* theorem 4.14 ) , every dra sequence with respect to @xmath102 converges linearly .
furthermore , @xmath418 , @xmath419 , @xmath420 , @xmath421 , and both @xmath406 and @xmath407 are locally identical around @xmath397 . by applying [ c : linear ] , the linearly rate is @xmath422 .
[ p : h - b ] let @xmath2 be either a hyperplane or a halfspace , and @xmath3 be a closed ball of @xmath4 such that @xmath382 . then every dra sequence @xmath99 with respect to @xmath102 converges in _ finitely many steps _ to a point in @xmath128 .
if @xmath269 , i.e. , @xmath270 , then the result is trivial , so we will work in the space @xmath202 with @xmath271 , and denote by @xmath220 the dra sequence .
we just prove the the result for the case when @xmath2 is a hyperplane because the case when @xmath2 is a halfspace is similar .
without loss of generality , we assume that @xmath423 and that @xmath424 is the closed ball of radius @xmath425 and center @xmath426 with @xmath427 .
nothing that @xmath428 we write @xmath429 , where @xmath430 we distinguish two cases . _ case 1 : _ @xmath431 .
then the two halves @xmath432 and @xmath433 of the ball @xmath3 are symmetric with respect to the hyperplane @xmath2 . by symmetry , we can and do assume that @xmath434 .
now for any @xmath435 , we have @xmath309 , @xmath287 , and by [ ex : proj_ball ] , @xmath436 which gives @xmath437 hence , @xmath163 @xmath438 . from @xmath439
, we have @xmath440 where @xmath441 .
since @xmath442 , @xmath443 , and so @xmath444 according to [ f : p - e ] , @xmath220 converges finitely to a point in @xmath445 .
_ case 2 : _
let @xmath447 , where @xmath448 .
then @xmath449 , @xmath450 , and @xmath451 , which implies @xmath452 @xmath453 . following [ l : id][l : id_sup ] , @xmath3 and @xmath321
are locally identical around @xmath128 . by using [ t : global ] , @xmath454 and again by [ f : p - e ] , we are done .
it follows from [ ex : epi_infinite ] that the conclusion of [ p : h - b ] no longer holds without slater s condition @xmath382 .
[ p : intersect ] let @xmath455 and @xmath456 be finite intersections of closed convex sets in @xmath4 such that @xmath121 .
suppose that @xmath457 both @xmath458 and @xmath459 are locally identical around @xmath108 .
then the following holds for any dra sequence @xmath99 with respect to @xmath102 : @xmath460 since @xmath2 and @xmath3 are closed convex , [ f : cvg][f : cvg_fixt ] gives @xmath355 with @xmath268 . by assumption ,
@xmath461 both @xmath458 and @xmath459 are locally identical around @xmath108 . noting that @xmath462 , @xmath463 , the conclusion follows from [ l : expand ] .
[ c : int_intersect ] let @xmath455 and @xmath456 be finite intersections of closed convex sets in @xmath4 such that @xmath118 .
suppose that @xmath464 both @xmath458 and @xmath459 are locally identical around @xmath28
. then holds for any dra sequence @xmath99 with respect to @xmath102 . since @xmath118 ,
[ f : cvg][f : cvg_anb ] implies @xmath119 .
then @xmath465 , and [ p : intersect ] completes the proof .
[ c : h - bs ] let @xmath2 be a hyperplane or a halfspace , and @xmath456 be a finite intersection of closed balls in @xmath4 .
suppose that @xmath382 , and for all @xmath400 , there exists a unique @xmath466 such that @xmath467 .
then every dra sequence @xmath99 with respect to @xmath102 converges _ finitely _ to a point in @xmath128 . from @xmath382
, we immediately have @xmath118 .
let @xmath266 .
if @xmath468 , then @xmath469 @xmath470 , and so @xmath3 and @xmath471 are locally identical around @xmath28 , following [ l : id][l : id_int ] .
if @xmath472 , then by assumption , there exists a unique @xmath466 such that @xmath467 , which implies that @xmath3 and @xmath471 are locally identical around @xmath28 . now using [ c : int_intersect ] , @xmath473 since @xmath463 , we also have @xmath474 , and so @xmath99 converges finitely due to [ p : h - b ] . .
] [ c : p - b ] let @xmath2 be a closed convex set , and @xmath3 be a closed ball in @xmath475 such that @xmath382 .
suppose that @xmath2 is locally identical with some polyhedral set around @xmath379 , and that no vertex of @xmath2 lies in @xmath476 .
then every dra sequence @xmath99 with respect to @xmath102 converges _ finitely _ to a point in @xmath128 . by [ t : finite][t : finite_abdryb]&[t : finite_aintb ] , it is sufficient to consider the case where @xmath2 is a polyhedral set in @xmath475 satisfying @xmath382
. then @xmath118 , and using [ f : cvg][f : cvg_anb ] , @xmath119 , and this convergence is finite if @xmath120 .
it thus suffices to consider the case where @xmath400 . we can write @xmath477 , where each @xmath478 is a halfplane in @xmath475 .
since all vertices of @xmath2 are not in @xmath476 , we deduce that @xmath28 is not a vertex of @xmath2 .
hence , @xmath2 and @xmath478 are locally identical around @xmath28 for some @xmath479 . now using [ l :
expand ] , @xmath480 moreover , @xmath481 , and by [ p : h - b ] , @xmath99 converges finitely .
in this section we focus on cases where we use information of the dra for @xmath352 to understand the dra for @xmath351 where @xmath482 and @xmath483 . [
l : restrict ] let @xmath2 be a closed convex subset and @xmath3 be a closed ( not necessarily convex ) subset of @xmath4 such that @xmath121 , and let @xmath103 be in @xmath4 .
suppose that the dra sequence @xmath99 with respect to @xmath102 , with starting point @xmath103 , converges to @xmath25 .
suppose further that there exist two closed sets @xmath320 and @xmath321 in @xmath4 such that @xmath482 , @xmath483 , and that both @xmath484 and @xmath485 are locally identical around @xmath486 . then @xmath487 by assumption , there exists @xmath123 such that @xmath336 then , there is @xmath160 such that @xmath488 let @xmath338 . since @xmath39 is ( firmly ) nonexpansive ( [ f : proj][f : proj_firm ] ) , @xmath489 which implies @xmath490 . using the convexity of @xmath2 and applying [ l : proj][l : proj_sub.a ] for @xmath491
, we have @xmath492 , and also @xmath493 . noting that @xmath494 and @xmath495 we get @xmath496 , and then applying [ l : proj][l : proj_subs ] for @xmath497 yields @xmath498 . hence , @xmath499 . if @xmath320 and @xmath321 in [ l : restrict ] are convex , then @xmath363 is single - valued , and we have the conclusion that @xmath500 i.e. , @xmath330 .
[ c : restrict_j ] let @xmath2 be a closed convex subset and @xmath501 be a finite union of disjoint closed convex sets in @xmath4 such that @xmath121 , and let @xmath103 be in @xmath4 .
suppose that the dra sequence @xmath99 with respect to @xmath102 , with starting point @xmath103 , is bounded and _ asymptotically regular _
, i.e. , @xmath502 .
then @xmath99 converges to a point @xmath319 , and there exists @xmath466 such that @xmath503 according to ( * ? ? ?
* theorem 2 ) , @xmath99 converges to a point @xmath319 .
since @xmath2 is convex , @xmath108 is a singleton , and by , @xmath268
. then there exists @xmath466 such that @xmath504 . by assumption
, there exists @xmath123 such that @xmath505 @xmath506 .
this implies @xmath507 , so @xmath3 and @xmath471 are locally identical around @xmath108 .
now apply [ l : restrict ] .
let @xmath2 be a hyperplane or a halfspace , and @xmath501 be a finite union of disjoint closed balls in @xmath4 such that @xmath121 , and @xmath474 whenever @xmath508 .
let @xmath103 be in @xmath4 .
suppose that the dra sequence @xmath99 with respect to @xmath102 , with starting point @xmath103 , is bounded and _ asymptotically regular _ ,
i.e. , @xmath502 . then @xmath99 converges _ finitely _ to a point @xmath266 . using [ c : restrict_j ] , @xmath355 , and
there is @xmath466 such that @xmath503 then @xmath508 , and by assumption , @xmath474 .
now by [ p :
h - b ] , the convergence of @xmath99 to @xmath28 is finite , and @xmath509 .
if the @xmath471 in [ c : restrict_j ] are singletons and @xmath2 is either an affine subspace or a halfspace , then it is possible to obtain stronger conclusions .
let @xmath2 be an affine subspace or a halfspace , and @xmath3 be a finite subset of @xmath4 such that @xmath121 , and let @xmath103 be in @xmath4 .
suppose that the dra sequence @xmath99 with respect to @xmath102 , with starting point @xmath103 , is asymptotically regular , i.e. , @xmath502 .
then @xmath99 converges _ finitely _ to a point @xmath319 with @xmath268 .
observe that @xmath39 is single - valued as @xmath2 is convex . according to
, it suffices to show that @xmath355 finitely .
set @xmath510 let us first consider the case when @xmath2 is an affine subspace .
then we can represent @xmath70 , where @xmath68 is a linear operator from @xmath4 to a real hilbert space @xmath67 , and @xmath69 . denoting by @xmath72 the moore
penrose inverse of @xmath68 , [ ex : proj_affine ] gives @xmath511 and so @xmath512 since @xmath513 ( see ( * ? ? ?
* chapter ii , section 2 ) ) , we get [ e : llxn+ ] @xmath514 and then gives @xmath515 now in turn , @xmath516 using the asymptotic regularity of @xmath99 , and yield @xmath517 since @xmath518 lies in @xmath3 and @xmath3 is finite , there exists @xmath160 such that @xmath519 @xmath520 .
then by , @xmath521 , which together with gives @xmath522 and @xmath99 thus converges finitely .
now consider the case when @xmath2 is a halfspace . without loss of generality
, we assume that @xmath523 , where @xmath80 and @xmath524 . using [ ex : proj_h][ex : proj_half ]
, we have @xmath525 and by , @xmath526 if @xmath527 @xmath528 and @xmath529 , then gives @xmath530 , and we are done .
assume that @xmath163 @xmath531 or @xmath532 . by using , @xmath163 @xmath528 @xmath533 @xmath534 .
thus , the set @xmath535 is infinite , and denoted by @xmath536 the enumeration of that set , we have @xmath537 then @xmath538 or @xmath539 , and the asymptotic regularity of @xmath99 implies the one of @xmath540 and also of @xmath541 .
since @xmath542 , gives @xmath543 and so @xmath544 but @xmath545 is in the finite set @xmath3 , there exists @xmath546 such that @xmath547 on the other hand , implies @xmath548 and then @xmath549 which yields @xmath550 , and thus @xmath391 .
let @xmath551 .
it follows from and that @xmath552 hence @xmath553 as @xmath542 .
we obtain @xmath554 , and by combining with and , @xmath555 which completes the proof .
the following examples illustrate that without asymptotic regularity a dra sequence with respect to @xmath102 may fail to converge .
suppose that @xmath556 , @xmath557 and @xmath558 .
then @xmath121 but the dra sequence with respect to @xmath102 with starting point @xmath559 does not converge since it cycles between two points @xmath559 and @xmath560 .
suppose that @xmath556 , that @xmath561 is a halfspace , and that @xmath562 is a finite set .
then @xmath121 but when started at @xmath563 , the dra cycles between four points @xmath563 , @xmath564 , @xmath565 and @xmath566 , as shown in [ fg : h - ps ] which was created by ` geogebra ` @xcite
. notice that if @xmath2 is a halfspace and @xmath3 is a finite subset of @xmath4 such that @xmath121 , then every dra sequence with respect to _ @xmath567 _ converges finitely due to ( * ? ? ?
* theorem 4.2 ) .
recall from @xcite that if we work with an affine subspace instead of a halfspace , then the quality of convergence of the dra sequence with respect to @xmath102 is the same as the one with respect to @xmath567 .
let @xmath2 be either a hyperplane or a halfspace of @xmath4 , and @xmath3 be a finite subset of one in two halfspaces generated by @xmath2 , and let @xmath103 be in @xmath4 .
then either : ( i ) the dra sequence @xmath99 with respect to @xmath102 , with starting point @xmath103 , converges _
finitely _ to a point @xmath319 with @xmath268 , or ( ii ) @xmath5 and @xmath568 in which case @xmath114 converges _ finitely _ to a best approximation solution @xmath165 relative to @xmath2 and @xmath3 in the sense that @xmath569 .
_ case 1 : _
@xmath2 is a hyperplane . without loss of generality
, we assume that @xmath570 and that @xmath571 by [ ex : proj_h][ex : proj_hyper ] , @xmath572 therefore , @xmath573 and also @xmath574 now setting @xmath575 we have @xmath576 and so @xmath577 it follows that @xmath578 , and @xmath579 from @xmath580 and @xmath581 , we have @xmath582 , which yields [
e : bns ] @xmath583 _ case 1.1 : _ @xmath163 @xmath584 . by combining with
, the sequence @xmath585 converges , and so @xmath586 but @xmath518 lies in the finite set @xmath3 ; hence , there exists @xmath160 such that @xmath587 @xmath588 , equivalently , @xmath529 . then implies @xmath587 @xmath589 , and by , @xmath590 .
_ case 1.2 : _
@xmath591 @xmath592 . then and give @xmath593 combining with , this implies @xmath594 and the sequence @xmath595 thus converges .
since again @xmath3 is finite , there exists @xmath596 , @xmath597 such that @xmath598 @xmath599 , which yields @xmath600 due to . by combining with , @xmath601 so @xmath114 converges finitely .
furthermore , if @xmath550 , i.e. , @xmath602 , then @xmath391 , in which case @xmath121 and by , @xmath598 @xmath590 . now assume that @xmath603 .
then @xmath604 due to .
it follows from and that @xmath605 let @xmath606 , and let @xmath607 . since @xmath608
, we have @xmath609 , and so @xmath610 which implies @xmath611 noting that @xmath612 , we deduce @xmath613 .
hence @xmath614 this yields @xmath5 , and by , @xmath615 while by , @xmath598 @xmath616 is a best approximation pair relative to @xmath2 and @xmath3 .
_ case 2 : _
@xmath2 is a halfspace . by assumption ,
we assume without loss of generality that either @xmath617 or @xmath618 where @xmath80 and @xmath524 .
_ case 2.1 : _ holds .
if @xmath163 @xmath584 , i.e. @xmath619 , then @xmath620 , so @xmath621 and according to _ case 1.1 _ , we must have @xmath622 and the finite convergence of @xmath99 . if @xmath591 @xmath623 , i.e. @xmath624 , then @xmath625 , which yields @xmath626 , and we are done .
_ case 2.2 : _ holds .
if @xmath627 , i.e. @xmath628 , then @xmath629 , and thus @xmath630 .
it is therefore sufficient to consider @xmath631 , i.e. @xmath632 . then @xmath633 , @xmath634 , and by , @xmath635 .
this yields @xmath636 now apply case 1 . [
ex : h-2ps ] suppose that @xmath556 , @xmath557 and @xmath637 .
then @xmath5 , and for starting point @xmath6381 , + \infty\right[\times \{-1\}$ ] , the dra sequence @xmath99 with respect to @xmath102 satisfies @xmath639 @xmath640 and @xmath641 .
see [ fg : h-2ps ] for an illustration , created with ` geogebra ` @xcite . with the starting point @xmath642 .
in view of [ d : local ] , we recall a result on finite convergence of the douglas rachford algorithm under slater s condition . [
f : a - p ] let @xmath2 be an affine subspace and @xmath3 be a closed convex subset of @xmath4 such that _ slater s condition _
@xmath643 holds .
suppose that @xmath3 is locally identical with some polyhedral set around @xmath379 .
then every dra sequence @xmath99 with respect to @xmath102 converges _ finitely _ to a point in @xmath128 .
combine ( * ? ? ?
* theorem 3.7 and definition 2.7 ) with [ d : local ] .
a natural question is whether the conclusion of [ f : a - p ] holds when the slater s condition @xmath382 is replaced by @xmath121 and @xmath644 . in the sequel
, we shall provide a positive answer in @xmath475 ( [ t : poly2d ] ) and a negative answer in @xmath645 ( [ ex : poly3d ] ) . for the next little while
, we work with equation x = @xmath646 ^ 2 a = @xmath646\{0 } , and consider the ( counter - clockwise ) rotator defined by @xmath647 let @xmath648 $ ] , and set @xmath649 then @xmath650 is the positive @xmath28-axis , and @xmath651 is the ray starting at @xmath652 and making an angle of @xmath653 with respect to @xmath654 in counter - clockwise direction . for @xmath655 , we write @xmath656 if @xmath657 , and @xmath658 if @xmath659 .
[ f:2lines ] let @xmath648 $ ] .
then @xmath660 this follows from ( * ? ? ?
* section 5 ) .
[ l : raystep ] assume that @xmath648 $ ] , @xmath661 , @xmath662 , and @xmath663 .
let @xmath664 , and set @xmath665 .
then @xmath666 if @xmath667 , and @xmath668 otherwise . in the latter case , @xmath669 if @xmath670 , and @xmath671 furthermore , @xmath672 we have @xmath673 and @xmath674 . if @xmath675 , then @xmath676 , and so @xmath677 , which yields @xmath678 now we consider the case @xmath679
. then @xmath680 , so @xmath681 , and by applying [ f:2lines ] , @xmath682 the rest is clear .
[ l : ray ] let @xmath683 where @xmath648 $ ] .
then every dra sequence @xmath99 with respect to @xmath102 converges to a point @xmath319 , and the `` shadow sequence '' @xmath114 converges to @xmath268 in at most @xmath684 iterations , where @xmath685 set @xmath662 , and @xmath663 .
we will study the behavior of the iterations in regions @xmath686 as shown in figure [ fg : ray ] . since @xmath648 $ ] , we have @xmath687 , and so @xmath688 . set @xmath689 .
according to [ l : raystep ] , if @xmath690 , then @xmath691 ; if @xmath692 , then @xmath693 .
so it is sufficient to consider the case @xmath694 . if @xmath670 , we have immediately @xmath695 .
now we assume without loss of generality that @xmath696
. then , yields the implication @xmath697 there thus exists @xmath160 such that @xmath698 which yields @xmath699 . using again [ l : raystep ] , @xmath700 .
noting that @xmath701 we get @xmath702 .
hence , @xmath703 and @xmath704 for all @xmath705 iterations .
[ l : cone ] let either @xmath557 or @xmath706 , and let @xmath3 be the convex cone generated by the union of the rays @xmath707 with @xmath708 $ ] .
then the dra applied to @xmath102 converges _ finitely globally uniformly _ in the sense that there exists @xmath709 such that @xmath34 the sequence @xmath375 converges to a point in @xmath710 in at most @xmath684 iterations .
we shall prove this for the case @xmath557 , the other case being similar . for @xmath711 ,
set @xmath712 , @xmath713 , @xmath714 , and let @xmath715 , @xmath716 . without loss of generality , we distinguish two cases : @xmath717 or @xmath718 . _ case 1 : _ @xmath717 .
as shown in the left image in figure [ fg : ray ] , we study the behavior of the iterations in regions @xmath719 set @xmath720 .
_ case 1.1 : _ @xmath721 .
then @xmath722 , and @xmath723 , so @xmath724 _ case 1.2 : _
@xmath725 . then @xmath726 , and @xmath727 .
we also see that @xmath728 belongs to the halfspace with boundary @xmath729 and not containing @xmath730 .
thus , @xmath731 , and @xmath732 using [ l : raystep ] , this implies @xmath733 therefore , as in the proof of [ l : ray ] , there exists @xmath160 , @xmath734 such that @xmath699 .
_ case 1.3 : _ @xmath735 . by an argument similar to the above , we have @xmath699 for some @xmath160 , @xmath736 .
_ case 1.4 : _ @xmath737
. then @xmath738 and @xmath739 since @xmath740 .
therefore , @xmath741 , and @xmath742 hence , in all cases , there exists @xmath596 such that @xmath743 and @xmath744 .
this shows that @xmath745 in at most @xmath684 iterations .
_ case 2 : _
@xmath718 . partitioning @xmath746 ( see the right image in figure [ fg : cone ] ) and arguing as in the above case , we obtain that @xmath355 in at most @xmath684 iterations , where @xmath747 the proof is complete .
[ r : cone ] by the same argument , [ l : cone ] also remains true when @xmath748 $ ] .
[ t : poly2d ] let @xmath2 be either a line or a halfplane , and @xmath3 be a closed convex set in the euclidean plane @xmath475 .
suppose that @xmath121 , and that @xmath3 is locally identical with some polyhedral set around @xmath379
. then every dra sequence @xmath99 with respect to @xmath102 converges _ finitely _ to a point @xmath319 with @xmath268 . using [ t : finite ]
, it suffices to prove for the case where @xmath3 is a polyhedral set in @xmath475 satisfying @xmath121 .
then @xmath456 is a finite intersection of halfplanes @xmath471 .
now by [ f : cvg][f : cvg_fixt ] , @xmath355 with @xmath749 .
_ case 1 : _
@xmath108 is not a vertex of @xmath3
. then there exists @xmath466 such that @xmath3 and @xmath471 are locally identical around @xmath108 . applying [ l :
expand ] for @xmath750 and @xmath751 , we have @xmath752 since @xmath2 is either a line or a halfplane , and @xmath471 is a halfplane in @xmath475 , [ t : h - h ] implies that @xmath753 finitely .
_ case 2 : _
@xmath108 is a vertex of @xmath3 .
noting that there are exactly two of halfplanes @xmath471 through each vertex of @xmath3 , it can also represent @xmath754 , where each @xmath755 is a closed convex cone in @xmath475 .
we then find @xmath466 such that @xmath3 and @xmath755 are locally identical around @xmath108 . by using again [ l :
expand ] , @xmath756 here @xmath2 is either a line or a halfplane through vertex @xmath108 of the cone @xmath755 .
now apply [ l : cone ] and [ r : cone ] .
[ ex : poly3d ] suppose that @xmath414 , that @xmath757 , and that @xmath758 , where @xmath759 then for starting point @xmath760 , the dra sequence @xmath99 with respect to @xmath102 converges @xmath761 with @xmath762 , but this convergence is not finite .
it is easy to see that @xmath763 , and so @xmath764 let @xmath765 .
noting that the moore
penrose inverse of @xmath68 is given by @xmath766 we learn from [ ex : proj_affine ] that @xmath767 , and so @xmath768 by , e.g. , ( * ? ? ?
* example 6.28 ) , @xmath769 , and thus @xmath770 setting @xmath771 , we claim that if [ e : initial ] @xmath772 then @xmath773 , where @xmath774 and also holds for @xmath775 and @xmath776 .
indeed , recall that @xmath777 it follows from that @xmath778 , @xmath779 , and @xmath780 by and a direct computation , @xmath781 which means @xmath782 using again we get @xmath783 as claimed .
now let @xmath784 , the above claim implies that @xmath785 a direct argument yields @xmath786 and then @xmath787 therefore , @xmath788 linearly with rate @xmath789 , but not finitely .
we conclude with a list of specific open problems .
* do the conclusions of [ f : p - e ] and [ t : h - e ] hold when @xmath2 is any hyperplane or halfspace ? * does [ p : h - b ] remain true if @xmath2 is an affine subspace or a polyhedron ?
* does [ c : h - bs ] remain true without assumption on the uniqueness ?
* do [ c : p - b ] and [ t : poly2d ] remain true in @xmath790 with @xmath791 ?
* does [ f : a - p ] remain true if we replace `` affine subspace '' by `` halfspace '' ? * what can be said about convergence of the dra for two polyhedrons or for two balls ?
hhb was partially supported by the natural sciences and engineering research council of canada and by the canada research chair program .
mnd was partially supported by an nserc accelerator grant of hhb .
aragn artacho , j.m .
borwein , and m.k .
tam , global behavior of the douglas rachford method for a nonconvex feasibility problem , _ journal of global optimization _ , to appear .
doi : 10.1007/s10898 - 015 - 0380 - 6 h.h .
bauschke , j.y .
bello cruz , t.t.a .
nghia , h.m .
phan , and x. wang , the rate of linear convergence of the douglas rachford algorithm for subspaces is the cosine of the friedrichs angle , _ journal of approximation theory _ 185 ( 2014 ) , 6379 .
bauschke , m.n .
dao , d. noll , and h.m .
phan , on slater s condition and finite convergence of the douglas rachford algorithm for solving convex feasibility problems in euclidean spaces , _ journal of global optimization _ , to appear .
doi : 10.1007/s10898 - 015 - 0373 - 5 y. censor , w. chen , p.l .
combettes , r. davidi , and g.t .
herman , on the effectiveness of projection methods for convex feasibility problems with linear inequality constraints , _ computational optimization and applications _ 51 ( 2012 ) , 10651088 .
j. liang , j. fadili , g. peyr , and d.r .
luke , activity identification and local linear convergence of douglas rachford / admm under partial smoothness , in : _ scale space and variational methods in computer vision _ , springer , 2015 , pp 642653 p. patrinos , l. stella , and a. bemporad , douglas
rachford splitting : complexity estimates and accelerated variants , in : _ proceedings of the 53rd ieee conference on decision and control _ , 2014 , pp 4234 - 4239 .
|
solving feasibility problems is a central task in mathematics and the applied sciences .
one particularly successful method is the douglas rachford algorithm . in this paper , we provide many new conditions sufficient for _ finite _ convergence .
numerous examples illustrate our results . *
2010 mathematics subject classification : * primary 47h09 , 90c25 ; secondary 47h05 , 49m27 , 65f10 , 65k05 , 65k10 . *
keywords : * averaged alternating reflections , douglas
rachford algorithm , feasibility problem , finite convergence , projector , reflector ,
|
the interest to multiagent systems with entangled structures and nontrivial dynamics known as complex networks is stimulated by their prevalence in nature , technical applications and society , and by the common mathematical language mainly based on graph theory , statistical physics and nonlinear dynamics @xcite .
the most promising application field of the complex network view is neuroscience @xcite . revealing the architecture of structural links in neuronal structures as well as functional relations between anatomically remote brain regions allows formulating basic principles of the organization of the central nervous system in terms of complex networks and nonlinear dynamics @xcite .
new findings in neuroscience require developing novel concepts in mathematical descriptions ; for example , the hierarchical structure and activity in brain stimulates generalizations in network science , e.g. , temporal , multiplex , multilayer , multilevel networks @xcite , and hypernetworks @xcite .
the ever - changing synaptic couplings lead to studying adaptive or evolving dynamical networks @xcite , impermanent nature of most perception and cognitive processes require shifting focus in nonlinear dynamics from attractors to transients @xcite .
the transient dynamics of spatiotemporal patterns and evolving architectures of synaptic links seem to be basic properties underlying most complex brain functions @xcite .
many studies of anatomical structures and physiological processes support the idea that cognitive processes relate to interactions among distributed neuronal populations and brain areas @xcite .
it was hypothesized and experimentally shown that during each perception or cognitive act there emerges an assembly of spatially distributed neurons the dynamic core that operates in a coordinated manner @xcite .
the neurons within the assembly preferentially communicate with each other giving rise to some integration activity .
this process is highly transient which enables the neuronal network to rapidly change the activation pattern reflecting the multitasking property of brain and functional segregation @xcite . uncovering biophysical and dynamical mechanisms underlying a rich functional repertoire of the brain circuit at the micro- , meso- , and macrolevels remains a nontrivial challenge for both experimentalists and theoreticians .
understanding the principles of relating cognitive functions to coordinated activities of neuronal substrates is still in its infancy and needs to be described by an appropriate mathematical language . in this paper
we propose a descriptive model for emerging complex behaviours in a hypernetwork at the upper level with respect to a basic network with co - evolving coupling structure and nodal dynamics .
this approach is used to provide an example of how a simple structured network of coupled oscillators gives rise to the traffic in the hypernetwork .
we propose a paradigmatic model based on the concept of hypernetworks @xcite ; this model captures how network dynamics and structure evolution can lead to the emergence of a new level of description which may be considered as an emergence of functions from evolving dynamical structures .
the model gives a sketch , a mathematical image of how a small oscillatory network with co - evolving structure and dynamics generates a new functional structure at an upper level of description .
simple transient dynamics of activity clusters result in a special form of traffic in this new structure that we refer to as a hypernetwork @xcite . the outline of the model is shown in fig . [ fig_hypernetwork ] .
first we consider a neuronal population as a network of interacting nodes - oscillators coupled by directed links .
this is the bottom level presented schematically in fig .
[ fig_hypernetwork](a ) .
the nodal dynamics as well as the activity of links are governed by some deterministic evolutionary operators ( for details , see _ materials and methods _ ) .
depending on nodal dynamics the links become active or inactive thus forming different structural patterns in the network .
physiologically this can mean that some synaptic links become strong while others tend to be weak .
a structural pattern results in a specific spatiotemporal activity in the form of synchronous or asynchronous clusters . depending on the internal state as well as external conditions
the network can adapt to different stimuli and reconfigure the structure of strong and weak couplings thus changing the spatiotemporal pattern .
different perception stimuli or cognitive tasks give rise to the activation of different distributed neuronal groups or to a different order of sequential activation of various brain areas .
we relate this distributed neuronal group specified by the common task to a corresponding hypersimplex .
_ hypersimplices _ are ordered , or structured , sets of nodes with an explicit relation ; in other words , they exist at a higher level of representation than network s nodes ( see fig .
[ fig_hypernetwork](b ) ) . in our model hypersimplices
are specified by the coupling topology that leads to a certain cluster activity . by a _ cluster activity _ we mean the form of collective behaviour exhibited by oscillators .
for example , in fig .
[ fig_hypernetwork](b ) nodes with a same colour denote one cluster , i.e. , they tend to fire synchronously while other nodes are silent .
different colours relate to different clusters , i.e. , these groups of oscillators do not fire simultaneously .
the cluster activity which is a functional attribute of the network is directly specified by the current hypersimplex which is a structural attribute of the network . in process of time
the network in fig .
[ fig_hypernetwork](a ) can exhibit different configurations corresponding to particular hypersimplices in fig .
[ fig_hypernetwork](b ) , i.e. , at each moment only one hypersimplex can be found in the network .
all possible hypersimplices realized in the network form a simplicial family .
a structured set of hypercimplices determines a hypernetwork ( fig .
[ fig_hypernetwork](c ) ) where the connections between hypercimplices describe their functional relations . in our model
we assume that two hypercimplices are connected if the network can change its state from one hypersimplex to another due to the evolutionary rules governing its structure and dynamics .
internal as well as external conditions usually tend to slowly evolve or drastically change which causes the network to adapt by reconfiguring the structure of its links .
in general this leads to rewiring and a new hypersimplex appears in the system .
hence the co - evolution of structure and dynamics at the bottom - level network in fig .
[ fig_hypernetwork](a ) gives rise to a kind of traffic at the upper level in the hypernetwork ( fig .
[ fig_hypernetwork](c ) ) .
the complexity of the hypernetwork and potential transitions in it depend on the dynamical principles and evolutionary rules that govern the constituents of the basic network .
an actual path in the hypernetwork depends on the joint action of external inputs and the network s internal state .
sensory information and task requirements activate some subset in the hypernetwork thus transforming it to a simpler structure that enables to move along the hypernetwork by a specific path .
each such path is a unique pattern of transient cluster activity for a given task performance and is robust in nonstationary environments .
( a , c , e ) and the corresponding paths in the hypernetwork ( b , d , f ) for the autonomous case ( a , b ) and under constant inputs ( c - f ) . in the autonomous case one
observes a random walk in the hypernetwork ( b ) . in case of constant stimuli
there are different scenarios : for example , when the stimulus is applied to the 1st node and the network is initially in the state @xmath0 ( green circle ) one obtains a path that after 4 steps comes to the 6-cycle of cluster states ( red circles ) @xmath1 ; when the stimulus is applied to the 2nd node and the network is initially in the state @xmath2 there is a path that after 6 steps comes to the 6-cycle of cluster states @xmath3 .
, width=432 ] we illustrate this general paradigm by a model in the form of a 5-node dynamical network .
the nodes - oscillators are coupled by inhibitory links in such a way that they can generate a rather simple but nontrivial form of activity : sequential clusters that appear cyclically .
this is due to a chosen scheme of coupling where some two nodes inhibit the other two nodes , the latter inhibit the last node , and this one inhibits the former two nodes . the nodes - oscillators are tuned to have a property of post - inhibitory rebound : an oscillator receives an inhibitory input and then is released from it after which the oscillator fires a post - inhibitory burst of spikes .
inhibitory links sent by some oscillators to others allow to suppress the activity of the latter while the former are active .
thus the network can produce a 3-step cycle of clusters ( which we refer to as a cluster state ) where first two nodes synchronously fire , then the other two generate synchronous bursts , after that the remaining one becomes active , and finally it starts all over again .
for example , at the cluster state @xmath0 nodes 1 and 2 fire synchronously first while the others are silent , then nodes 3 and 4 generate synchronous bursts , and finally node 5 fires while the other four are below the threshold .
totally there are 30 cluster states @xmath4 , @xmath5 , and corresponding topologies having the described properties ( for more details , see _ materials and methods _ and ref .
@xcite ) .
the coupling structure is static for some time interval , thus enabling to produce several identical cluster cycles .
after that the structure switches to another one that depends on the dynamics exhibited by the oscillators .
the new structure has the same symmetric properties as the previous one reflecting the homeostasis inherent to biological networks . in the new state
the network also generates 3 cyclic clusters but they comprise other oscillators ; in other words , the clusters mix during the structure switching by swapping the nodes between each other .
we assume that the new topology depends on what cluster is active at the moment of switching ; since we have 3 clusters there are 3 different topologies which can be established after the switching . from time to time
the network moves from a current cluster state to one of three others .
the system is basically an adaptive network where the connection topology determines the form of nodal activity while the oscillatory dynamics influences the network structure affecting the subsequent coupling scheme after the switching .
as we talked in the previous subsection , a structured set of nodes which determines a cyclic cluster sequence can be regarded as a hypersimplex .
therefore we consider all 30 possible 3-cluster states in our network as a simplicial family .
being connected by directed links they form a hypernetwork of cluster states where some path is activated during the time course of the network co - evolution ( see fig . [ fig_hypernetwork ] ) .
we studied the co - evolution of the network in an autonomous case and under the effect of constant stimuli .
left to itself , the network generates a sequence of different cluster states and the path in the corresponding hypernetwork is a random walk shown in fig .
[ fig_path](a , b ) .
if an input is applied to the network then the path becomes stimulus - specified .
the analysis of the network dynamics showed that the path depends on an internal network state
its topology and the external stimulus , namely on what node it is applied to . in general , in the non - autonomous case the path in the hypernetwork after several steps comes to some cycle consisting of 6 cluster states and remains there until a new input comes . in the example shown in fig .
[ fig_path](c , d ) the network initially is in the state @xmath0 and the constant input is applied to node 1 .
the trajectory in the hypernetwork goes through the sequence @xmath6 and comes to the cycle @xmath1 .
when the input is applied to node 2 and the network is in the state @xmath2 at the beginning , the trajectory in the hypernetwork goes through the sequence @xmath7 and comes to the other cycle @xmath3 shown in fig .
[ fig_path](e , f ) .
we found that the application of a constant stimulus to a network node specifies the subset of active directed links in the hypernetwork which govern the co - evolution while other links become inactive . in this new hypernetwork
each hypersimplex has only one outward link so the path becomes specified and there is a limiting path a 6-cycle attracting all other trajectories .
for example , the application of a stimulus to node 1 result in a reduced hypernetwork shown in fig .
[ fig_cycle](a ) . starting from some cluster state @xmath4 ,
a trajectory moves along a well determined path towards a 6-cycle of states @xmath1 . for a better visualization
the same reduced hypernetwork is shown in fig .
[ fig_cycle](b ) where it is seen how different branches of the reduced hypernetwork converge to the same 6-cycle .
an input to node 2 leads to another reduced hypernetwork with a similar structure ( fig .
[ fig_cycle](c , d ) ) . and nodal variables @xmath8 ( @xmath9 ) which represent transitions between different cluster states realized on fpga xilinx artix-7 .
time graphs are registered from multi - channel acquisition system .
waveforms in ( a ) correspond to a random walk in the hypernetwork in an autonomous case .
waveforms in ( b ) relate to the case of a stimulus - specific traffic in the hypernetwork ( an input is applied to node 1 ) when the trajectory came to the 6-cycle @xmath10 .
, width=480 ] to create a real - time functioning 5-node dynamical network with inhibitory connections we realized its electronic model on fpga xilinx artix-7 .
the fpga electronic model allows one to process external ( sensory , informational ) signals in this network in real time . due to the real time processing and parameters scalability of the electronic model
we have the opportunity to apply it to different types of interfaces ( for example , a human - machine interface , a brain - computer interface , an artificial neural network interface , etc . ) .
the analysis shows a full agreement of theoretical and numerical predictions with experimental data . without external inputs
the experimental setup displays random transitions between different cluster states , an example is shown in fig .
[ fig_experiment](a ) . in case of an input application
, the network demonstrates transitions relating to the corresponding reduced hypernetwork where each hypersimplex has only one outward link and all paths are converged to a 6-cycle .
for example , when an input is applied to node 1 the network activity gives rise to a transition in the hypernetwork towards the cyclic sequence @xmath10 , etc .
, shown in fig .
[ fig_experiment](b ) .
in this paper , we proposed a paradigmatic model of how a network of spiking neurons can create complex responses at the higher level or representation in the corresponding hypernetwork . despite the simplicity of the coupling structure , the neuron model , and the evolutionary operator
we show the basic idea : how dynamics of the adaptive oscillatory network of spiking neurons leads to the emergence of various transient behaviours in the hypernetwork .
consecutive switchings between different hypersimplices mean that each cluster state is unstable due to the evolutionary rules governing the system but applying inputs results in a robust stimulus - specific path in the hypernetwork .
this is similar to dynamics exhibited by systems with the `` winnerless competition '' principle in refs .
@xcite , however the dynamic mechanism of the co - evolving topology and dynamics in our case is completely different than the stable heteroclinic channel in these works . besides fundamental issues concerning the dynamic origin of relational simplices in large - scale neural networks
there are several items relating to the case study model to be addressed in further research .
for example , the influence of more complicated than constant inputs , the role of noise and parameters detuning on temporal properties of sequences in the hypernetwork ( see , e.g. , @xcite ) , the appearance of topology switching from local internodal interactions , and others .
these problems remain beyond the scope of this paper and provide a baseline for future studies .
the activity of nodes - oscillators is described by the following map @xcite : @xmath11 where @xmath12 is discrete time , the variables @xmath13 and @xmath14 characterize the state of the @xmath15-th node at the moment of @xmath16 .
the nonlinear function @xmath17 , where @xmath18 is the heaviside step function : @xmath19 if @xmath20 and @xmath21 otherwise ; the parameters @xmath22 , @xmath23 , and @xmath24 control an oscillatory regime , in particular , @xmath23 and @xmath24 determine chaotic activity .
the parameter @xmath25 determines the rate for the variable @xmath26 and the parameter @xmath27 characterizes the excitatory properties of the nodes .
the term @xmath28 takes into account the additive noise and the impact on the @xmath15-th node from the other network nodes ( inhibitory links ) ; the latter is given by @xmath29 where the coefficient @xmath30 defines the coupling strength , @xmath31 are the elements of an adjacency matrix @xmath32 ( see below ) at the moment of @xmath16 , @xmath33 is the so - called reversal parameter , and @xmath34 is the threshold parameter . in our simulations
we fix the parameters @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , and @xmath41 . for these values ,
the nodes described by and display subthreshold activity in the absence of an input ( @xmath42 ) and generate chaotic bursts as a result of the inhibition by the connected nodes ( for more details , see @xcite ) . the map - based approach in modeling neuronal networks
is described in more detail in refs .
the system was used in a number of research papers for modeling the collective activity of different neuronal systems ( refs .
@xcite ) and studying rather theoretical issues on chaotic dynamics in maps ( refs .
note that the chaos itself is not necessarily needed to generate a hypernetwork and the general results in sec .
[ subsec_hypernetwork ] do not depend on whether chaotic or regular dynamics is exhibited by the network .
we use in our example chaotic oscillators to demonstrate the robustness of switching activity in adaptive networks of spiking neurons even in case of irregular dynamics .
we considered a network consisting of 5 nodes - oscillators .
the connections are directed and described by the adjacency matrix @xmath32 where @xmath43 if node @xmath44 sends a link to node @xmath15 and @xmath45 otherwise .
we consider the simplest nontrivial cluster states in this network , two clusters consist of two oscillators , and the remaining oscillator forms the third cluster . as a result ,
the network exhibits a cyclic sequence @xmath46 which means that first two nodes with numbers @xmath47 fire synchronously ( the first cluster ) , then nodes @xmath48 generate synchronous bursts , after that node @xmath49 fires and the sequence repeats . to obtain this configuration we choose the elements @xmath50 in a specific way .
we take @xmath51 , @xmath52 , @xmath53 , and the other elements of @xmath54 are equal to 0 .
the complete list of possible 3-cluster sequences of this configuration is presented in table [ tab : states ] where each column is generated by cyclic permutation of the indexes @xmath46 in the top state .
lllll @xmath55 & @xmath56 & @xmath57 + @xmath58 & @xmath59 & @xmath60 + @xmath61 & @xmath62 & @xmath63 + @xmath64 & @xmath65 & @xmath66 + @xmath67 & @xmath68 & @xmath69 + + @xmath70 & @xmath71 & @xmath72 + @xmath73 & @xmath74 & @xmath75 + @xmath76 & @xmath77 & @xmath78 + @xmath79 & @xmath80 & @xmath81 + @xmath82 & @xmath83 & @xmath84 + the coupling topology evidentally influences the nodal dynamics since it directly determines the form and order of clusters . to take into account the impact of the dynamics on the coupling topology we use the following algorithm .
we introduce an auxiliary variable @xmath85 which defines the moment @xmath86 of topology switching in the following way : @xmath87 it follows from eqs . that @xmath85 is increasing on the average starting from zero up to the threshold @xmath88 with the rate determined by the network behavior ( the mean field @xmath89 ) and a small parameter @xmath90 ( @xmath91 ) .
the dynamics of @xmath89 is chaotic due to the chaotic activity of @xmath92 , @xmath9 , which leads to small perturbations of @xmath85 . after reaching the threshold ,
the value of @xmath85 is reset to zero and begins to grow according to eq . .
the topology rewiring occurs at the moment @xmath86 of resetting of @xmath85 . at the moment of rewiring ,
two nodes are chosen , one from the current active cluster , and another from the previous active cluster .
the indexes of these nodes , @xmath93 and @xmath94 , determine how the coupling topology is changed according to the rule : @xmath95 which numbers @xmath93 and @xmath94 are chosen depends on the following algorithm .
suppose at the moment of rewiring @xmath86 the nodes with indexes @xmath48 are active , and before that the nodes @xmath47 were active .
in the clockwise ordered set starting from @xmath49 there is a pair of nodes @xmath96 and @xmath97 that have a minimum index distance determined from the clockwise ordering of the nodes .
for example , the index clockwise distance between nodes 2 and 3 is 1 while between 3 and 2 is 4 .
the indexes @xmath98 and @xmath99 specify the numbers @xmath100 and @xmath101 in eq . .
the matrix @xmath102 is obtained by swapping row @xmath93 and row @xmath94 of the identity @xmath103 matrix with 1 s on the main diagonal and 0 s elsewhere . thus @xmath102 makes a row - switching transformation , i.e. , @xmath104 is the matrix produced by exchanging row @xmath93 and row @xmath94 of @xmath54 .
the product @xmath105 is thus a matrix where firstly the rows @xmath93 and @xmath94 are exchanged and then in the obtained matrix the columns @xmath93 and @xmath94 are swapped .
the discrete - time systems describing the nodal and coupling dynamics in eqs . and as well as the evolutionary rule governing the topology switching in eqs . and
were realized on fpga xilinx artix-7 .
the signal from each oscillator of the network and the signal corresponding to average variable @xmath85 were registered from the fpga through the 12-bit digital - to - analog convertors .
one discrete time step for this network is equal 50 @xmath106 , the average spike duration is about 10 - 20 discrete steps ( about 0.5 - 1 ms ) , and the burst lasts about 25 ms . such timings are well correspond to time scales of real neurons .
the parameters are equal to @xmath107 , @xmath108 , @xmath109 , @xmath36 , @xmath35 , @xmath37 , @xmath41 , @xmath40 , and @xmath110 .
this work was supported by the russian science foundation ( project no .
16 - 42 - 01043 ) .
m. de domenico , a. sol - ribalta , e. cozzo , m. kivel , y. moreno , m. a. porter , s. gmez , and a. arenas , `` mathematical formulation of multilayer networks , '' _ physical review x _ , vol . 3 , no . 4 , p. 041022
, 2013 .
s. boccaletti , g. bianconi , r. criado , c. i. del genio , j. gmez - gardees , m. romance , i. sendia - nadal , z. wang , and m. zanin , `` the structure and dynamics of multilayer networks , '' _ physics reports _ , vol . 544 , no . 1 , pp . 1122 , 2014 .
r. m. hutchison , t. womelsdorf , e. a. allen , p. a. bandettini , v. d. calhoun , m. corbetta , s. della penna , j. h. duyn , g. h. glover , j. gonzalez - castillo , _
et al . _ , ``
dynamic functional connectivity : promise , issues , and interpretations , '' _ neuroimage _ , vol .
80 , pp . 360378 , 2013 .
m. rigotti , o. barak , m. r. warden , x .- j .
wang , n. d. daw , e. k. miller , and s. fusi , `` the importance of mixed selectivity in complex cognitive tasks , '' _ nature _ , vol .
497 , no .
7451 , pp . 585590 , 2013 .
v. afraimovich , v. zhigulin , and m. rabinovich , `` on the origin of reproducible sequential activity in neural circuits , '' _ chaos : an interdisciplinary journal of nonlinear science _ ,
14 , no .
4 , pp . 11231129 , 2004 .
v. s. afraimovich , m. i. rabinovich , and p. varona , `` heteroclinic contours in neural ensembles and the winnerless competition principle , '' _ international journal of bifurcation and chaos _ , vol .
14 , no . 04 , pp . 11951208 , 2004 .
v. afraimovich , i. tristan , r. huerta , and m. i. rabinovich , `` winnerless competition principle and prediction of the transient dynamics in a lotka volterra model , '' _ chaos : an interdisciplinary journal of nonlinear science _ ,
18 , no . 4 , p. 043103
, 2008 .
v. s. afraimovich , t. r. young , and m. i. rabinovich , `` hierarchical heteroclinics in dynamical model of cognitive processes : chunking , '' _ international journal of bifurcation and chaos _ , vol . 24 , no . 10 , p. 1450132
, 2014 .
m. courbage , o. maslennikov , and v. nekorkin , `` synchronization in time - discrete model of two electrically coupled spike - bursting neurons , '' _ chaos , solitons & fractals _ , vol .
45 , no . 5 , pp .
645659 , 2012 .
o. v. maslennikov , d. v. kasatkin , n. f. rulkov , and v. i. nekorkin , `` emergence of antiphase bursting in two populations of randomly spiking elements , '' _ physical review e _ , vol .
88 , no . 4 , p. 042907
, 2013 .
o. v. maslennikov , v. i. nekorkin , and j. kurths , `` basin stability for burst synchronization in small - world networks of chaotic slow - fast oscillators , '' _ physical review e _ , vol .
92 , no . 4 , p. 042803
, 2015 .
o. v. maslennikov and v. i. nekorkin , `` attractors of relaxation discrete - time systems with chaotic dynamics on a fast time scale , '' _ chaos : an interdisciplinary journal of nonlinear science _ , vol . 26 , no . 7 , p. 073104
|
we propose a model of an adaptive network of spiking neurons that gives rise to a hypernetwork of its dynamic states at the upper level of description . left to itself ,
the network exhibits a sequence of transient clustering which relates to a traffic in the hypernetwork in the form of a random walk .
receiving inputs the system is able to generate reproducible sequences corresponding to stimulus - specific paths in the hypernetwork .
we illustrate these basic notions by a simple network of discrete - time spiking neurons together with its fpga realization and analyze their properties .
|
the interaction of a bound electron and an atomic nucleus is characterized by the parameter @xmath4 , where @xmath5 is the nuclear charge number and @xmath6 is the fine - structure constant .
this universal `` coupling parameter '' sets the scale for calculations of the radiative corrections to various bound - state effects including the hyperfine structure ( hfs ) and the bound - electron @xmath0 factor .
traditionally , theoretical investigations of radiative corrections in light systems relied upon an expansion in powers of @xmath7 and @xmath8
. however , today it is desirable to advance theory beyond the predictive limits given by the highest available terms in the @xmath7-expansion .
this can be done by carrying out calculations with using nonperturbative ( in @xmath7 ) propagators .
such calculations demand rather sophisticated numerical techniques , which were developed relatively recently .
indeed , all - order calculations of the self - energy ( se ) correction in the presence of a magnetic field started in the 1990s @xcite , extending during past years to a wide range of reference states and nuclear charge numbers @xcite . numerical calculations of the se corrections are particularly difficult for low values of @xmath5 .
this is mainly for two reasons .
first , the goal of the calculations is the contribution beyond the known @xmath4-expansion terms . for the hfs ,
the higher - order effects are suppressed with respect to the leading correction by a factor of @xmath9 . for the @xmath0 factor
, they enter only at order of @xmath10 and thus become very small numerically in the low-@xmath5 region .
second , in actual calculations there are additional cancellations arising at intermediate stages of the numerical procedure .
these cancellations become more severe for smaller values of @xmath5 and lead to further losses of accuracy . in this article , we treat the two most important example cases of the bound - electon se corrections in external magnetic fields : the se correction to the hfs and the se correction to the bound - electron @xmath0 factor .
we evaluate both of these corrections for the ground and excited states of hydrogen and of light hydrogen - like ions .
the first attempt at an all - order evaluation of the se correction to the hfs of hydrogen was made in ref .
because of insufficient numerical accuracy , the goal was reached in an indirect way : the known terms of the @xmath4 expansion were subtracted from numerically determined all - order results for @xmath11 , and the higher - order remainder was extrapolated down toward the desired value @xmath12 .
the accuracy of the numerical evaluation of the se correction to the hfs was improved by several orders of magnitude during the past years @xcite .
however , the precision obtained was still insufficient for a direct determination of the higher - order se remainder at @xmath12 , and an extrapolation procedure had to be employed again .
the studies @xcite reported results for the higher - order contribution for the normalized difference of the @xmath13 and @xmath14 hfs intervals in @xmath15 and demonstrated a @xmath16 deviation of the theoretical prediction from the experimental result @xcite .
the accuracy of the extrapolation procedure of refs .
@xcite , however , has recently become a subject of some concern . in particular ,
an opinion was expressed in ref .
@xcite that the uncertainty of the extrapolation procedure should have been estimated as four times larger than given in refs .
@xcite , which would have brought theory and experiment into agreement . in our recent investigation @xcite
, we performed the first direct , high - precision theoretical determination of the higher - order remainder of the se correction to the hfs of @xmath13 and @xmath14 states of hydrogen and light hydrogen - like ions .
good agreement was observed with the previous extrapolated values @xcite , but the accuracy was increased by several orders of magnitude . in the present paper ,
we report the details of this calculation and extend it to the higher excited states ( @xmath17 , @xmath18 , and @xmath19 ) .
the se correction to the bound - electron @xmath0 factor is of particular importance because it is used in the determination of the electron mass value from the experimental results for the @xmath0 factor of light hydrogen - like ions @xcite .
already at the present level of experimental accuracy , calculations of the bound - electron @xmath0 factor should be performed to all orders in @xmath4 .
a number of all - order evaluations of the se correction to the @xmath0 factor have been accomplished during last years @xcite , which resulted in an improvement of the precision of the electron mass value . however , in order to match the @xmath20 level of accuracy anticipated in future experiments on the helium ion @xcite , the precision of numerical calculations of the se correction should be enhanced by several orders of magnitude .
first results of our evaluation of the se correction to the bound - electron @xmath0 factor for the @xmath13 state of light hydrogen - like ions were reported in ref .
@xcite . in the present investigation
we extend our calculation to the higher excited states ( @xmath14 , @xmath17 , @xmath18 , and @xmath19 ) and to a wider region of the nuclear charge number @xmath5 .
relativistic units ( @xmath21 ) and heaviside charge units @xmath22 are used throughout the paper .
our investigations are organized as follows . in sec .
[ general ] , we discuss general formulas pertaining to the formulation of the effect within the formalism of quantum electrodynamics .
we continue with a detailed description of the numerical approach in sec .
[ detailed ] .
numerical results are presented in sec .
[ numerical ] .
we conclude with a summary in sec .
[ conclusions ] . and entails an arbitrary number of coulomb interactions with the atomic nucleus .
the wavy line that ends with a cross denotes the interaction with the perturbing potential @xmath23 .
the latter is given by the magnetic field of the nucleus in the case of the hfs and by a constant external magnetic field in the case of the bound - electron @xmath0 factor .
[ fig : se ] ]
the se correction in the presence of a binding coulomb field and an additional perturbing potential @xmath23 is graphically represented by the feynman diagrams shown in fig . [ fig : se ] . the general expression for them
can be conveniently split into three parts @xcite , @xmath24 which are referred to as the _ irreducible _ , the _
reducible _ , and the _ vertex _ contribution , respectively .
the vertex contribution is induced by the diagram in fig .
[ fig : se](b ) .
it can be expressed as @xmath25 [ { \varepsilon}_a-\omega-{\varepsilon}_{n_2}(1-i 0)]}\,.\end{aligned}\ ] ] here , @xmath26 is the operator of the electron - electron interaction @xmath27 where @xmath28 is the photon propagator and @xmath29 are the dirac matrices .
the sums over @xmath30 and @xmath31 involve both the positive - energy discrete and continuous spectra and the negative - energy continuous spectrum .
the irreducible contribution is induced by a part of the diagrams in fig .
[ fig : se](a ) and ( c ) that can be expressed in terms of the first - order perturbation of the reference - state wave function by @xmath23 , @xmath32 { { \varepsilon}_n \ne { \varepsilon}_a } \end{array } } \frac{|n{\rangle}{\langle}n|\delta v|a{\rangle}}{{\varepsilon}_a-{\varepsilon}_n } \,.\ ] ] the expression for the irreducible contribution is @xmath33 where @xmath34 , @xmath35 is the one - loop mass counterterm , and @xmath36 is the one - loop se operator , @xmath37 in the above , @xmath38 denotes the dirac coulomb green function @xmath39^{-1}$ ] , @xmath40 is the dirac coulomb hamiltonian , and @xmath41 .
the reducible contribution is induced by a part of diagrams in fig .
[ fig : se](a ) and ( c ) that can be expressed in terms of the first - order perturbation of the reference - state energy .
it reads @xmath42 where @xmath43 . up to now
we did not specify the particular form of the perturbing potential @xmath23 , assuming only its locality . in the following ,
we consider two particular choices of @xmath23 , both representing interactions with the magnetic field : the hfs interaction and the interaction with the magnetic field ( the zeeman effect ) . in the case of the hfs interaction ,
the perturbing potential has the form of the fermi - breit interaction @xmath44}{r^3}\,,\ ] ] ( where @xmath45 denotes the nuclear magnetic moment ) and the reference - state wave function @xmath46 is the wave function of the coupled system ( electron@xmath47nucleus ) , @xmath48 where @xmath49 denotes the nuclear wave function , @xmath50 is the electron wave function , @xmath51 is the total momentum of the atom , and @xmath52 is its projection .
the nuclear variables can be separated out by using the standard technique of the racah algebra .
it can be demonstrated @xcite that the general formulas ( [ eq1])([eq8 ] ) yield contributions to the hfs if one employs an electronic perturbing potential of the form @xmath53_z}{r^3}\,,\ ] ] and takes the reference - state wave function to be the electronic wave function with the momentum projection @xmath54 , @xmath55 in the above , @xmath56 denotes the nonrelativistic limit of the expectation value of the fermi - breit operator on the reference state and the prefactor @xmath57 is given by @xmath58 where @xmath59 is the dirac quantum number of the reference state and @xmath60 is its principal quantum number . in the case of the zeeman splitting ,
the perturbing potential is @xmath61 where @xmath62 $ ] is the vector potential . in practical calculations , corrections to the zeeman splitting
are convenienly expressed in terms of corrections to the @xmath0 factor .
it can be easily shown ( see ref .
@xcite for details ) that the general formulas ( [ eq1])-([eq8 ] ) yield contributions to the electronic @xmath0 factor if one employs a perturbing potential of the form @xmath63_z\,,\ ] ] and the reference - state wave function with the momentum projection @xmath64 .
the @xmath0-factor perturbing potential ( [ eq14 ] ) differs from the hfs potential ( [ eq11 ] ) only by the power of @xmath65 and the prefactor . in the following ,
all explicit formulas for individual contributions will be presented for the case of the hfs .
when working in the coordinate representation , the corresponding formulas for the @xmath0 factor can be obtained by an obvious substitution . in momentum space ,
the formulas for the hfs and for the @xmath0 factor are different .
our present approach to the evaluation of the se correction to the @xmath0 factor closely follows the one of ref .
@xcite and is therefore not described separately .
the general formulas presented in the previous section for individual contributions are both ultraviolet ( uv ) and infrared ( ir ) divergent . in order to obtain expressions suitable for numerical evaluation , a careful re - arrangement of contributions is needed , together with a covariant regularization of divergences . the calculation of the irreducible contribution ( [ eq5 ] )
can be reduced to an evaluation of a non - diagonal matrix element of the first - order se operator ( [ eq6 ] ) .
its renormalization is well known and does not need to be discussed here .
the numerical evaluation of the irreducible contribution was performed by a generalization of the approach of refs .
@xcite , with the use of a closed - form analytic representation of the perturbed wave function @xmath66 obtained in ref .
@xcite ( see also ref .
@xcite ) .
the evaluation of the reducible and the vertex contribution is carried out after splitting them into several parts , @xmath67 where the upper index @xmath68 labels the contributions induced by the reference - state part of the electron propagators and the other indices specify the total number of interactions with the binding field in the electron propagators [ the index @xmath69 labels the terms generated by @xmath70@xmath71 such interactions ] .
the reference - state contributions @xmath72 and @xmath73 are separately ir divergent .
the divergences disappear when the contributions are regularized in the same way and evaluated together .
let us now demonstrate the cancellation of the ir divergences and obtain the finite residual . the part of the vertex and reducible contributions induced by the intermediate states degenerate in energy with the reference state is @xmath74\,,\end{aligned}\ ] ] where @xmath75 is the `` true '' reference state and @xmath76 and @xmath77 label the intermediate states that are degenerate with the reference state in energy and have momentum projections @xmath78 and @xmath79 , respectively .
( the intermediate states degenerate with the reference state in energy but of opposite parity do not induce any ir divergences because of the orthogonality of the wave functions
. however , in practical calculations we find it convenient to treat all the degenerate states on the same footing . ) for simplicity , we now consider the photon propagator in the feynman gauge .
then , the operator of the electron - electron interaction @xmath26 takes the form @xmath80 where @xmath81 with @xmath82 being the photon mass , which regularizes the ir divergences .
it can be seen that all divergences in eq .
( [ d2 ] ) originate from an integral of the form @xmath83 we now substute eq .
( [ d4 ] ) into the above expression , twice perform an integration by parts , evaluate the @xmath84 integral by cauchy s theorem , and obtain @xmath85 adding and subtracting @xmath86 in the numerator of the integrand , we separate the above expression into two parts , the first of which is convergent when @xmath87 while the other ( divergent part ) does not depend on @xmath88 . setting @xmath89 in the convergent part and evaluating the integral , we obtain @xmath90 the divergent part of @xmath91
does not depend on the radial variables and , being substituted into eq .
( [ d2 ] ) , leads to a vanishing contribution .
we thus obtain @xmath92\,.\end{aligned}\ ] ] we note that in the case when the perturbing potential @xmath23 is spherically symmetric , the reference - state contribution @xmath93 vanishes as @xmath94 . in our case , however , @xmath23 represents an interaction with the magnetic field , so that @xmath93 induces a finite contribution . in our practical calculations ,
the reference - state contribution was separated from the vertex and reducible parts by introducing point - by - point subtractions from the electron propagators in the integrands and was calculated separately according to eq .
( [ d6 ] ) .
the zero - potential parts @xmath95 and @xmath96 are separately uv divergent .
they are covariantly regularized by working in an extended number of dimensions ( @xmath97 ) and calculated in momentum space .
the elimination of uv divergences in the sum of the reducible and the vertex contributions is well documented in the literature ( see , e.g. , ref .
@xcite ) , so here we operate with the _ renormalized _ se and vertex operators , assuming that all uv divergences are already cancelled out .
the zero - potential contribution to the reducible part is simple .
it is given by @xmath98 where @xmath99 is the dirac adjoint .
the derivative of the renormalized free se operator @xmath100 can be expressed as a linear combination of 3 matrix structures , @xmath101 , @xmath102 , and the unity matrix @xmath26 , @xmath103 , \end{aligned}\ ] ] where @xmath104 and @xmath105 are scalar functions , whose explicit expression is given by eqs .
( 53)-(55 ) of ref .
@xcite . integrating over angular variables
, we immediately have @xmath106 \nonumber \\ & + a_2(\rho ) ( g_a^2+f_a^2 ) + a_3(\rho ) ( g_a^2-f_a^2 ) \bigr\ } \,,\end{aligned}\ ] ] where @xmath107 and @xmath108 and @xmath109 are the upper and the lower components of the reference - state wave function in the momentum space .
the zero - potential vertex part of the se hfs correction is induced by the hfs potential @xmath110 inserted in the free se loop .
the hfs potential ( [ eq11 ] ) in the momentum space takes the form @xmath111_z}{{\bm{q}}^2}\,.\ ] ] the zero - potential vertex part is then given by @xmath112_z}{{\bm{q}}^2}\ , \psi_a({\bm{p}}_2)\,,\end{aligned}\ ] ] where @xmath113 , @xmath114 and @xmath115 are 4-vectors with the fixed time component @xmath116 , @xmath117 , and @xmath118 is the renormalized one - loop vertex operator . for evaluating the integrals over the angular variables in eq .
( [ e6 ] ) , it is convenient to employ the following representation of the vertex operator sandwiched between the dirac wave functions @xmath119 \,,\end{aligned}\ ] ] where the scalar functions @xmath120 are given by eqs .
( a7)(a12 ) of ref .
@xcite , and @xmath121 , @xmath122 , and @xmath123 .
the dependence of the integrand of eq .
( [ e6 ] ) on the angular variables can now be parameterized in terms of the basic angular integrals @xmath124 introduced and evaluated in appendix [ app:1 ] .
the result is @xmath125 { \cal r}_1 \nonumber \\ & + [ -p_{1r}k_1(-\kappa_a)+p_{2r}k_1^{\prime}(-\kappa_a ) ] { \cal r}_2 \nonumber \\ & - p_{1r}p_{2r}k_2(\kappa_a)\,({\cal r}_3+{\cal r}_4 ) \nonumber \\ & - p_{1r}p_{2r}k_2(-\kappa_a)\,({\cal r}_5+{\cal r}_6)\bigr\}\,.\end{aligned}\ ] ] the above equation was used for the numerical evaluation .
it contains four integrations ( the fourth one , over the feynman parameter , is implicit in the definition of the functions @xmath126 ) .
all the integration were performed using gauss - legendre quadratures , after appropriate substitutions in the integration variables .
we note that the integration variables @xmath127 , @xmath128 , and @xmath129 resemble the well - known perimetric coordinates @xcite , in the sence that they weaken the ( integrable ) coulomb singularity of the integrand at @xmath130 .
the one - potential hfs vertex part @xmath131 is given by @xmath132_z } { ( { \bm{q}}-{\bm{p}}{^{\prime\prime}})^2\ , { { \bm{p}}{^{\prime\prime}}}^2}\ , \psi_a({\bm{p}}{^{\prime}})\,,\end{aligned}\ ] ] where @xmath133 is the total momentum transfer ( final minus initial ) for the electron vertex function @xmath134 , and the time component of the 4-vectors is fixed by @xmath135 and @xmath136 .
the 4-point vertex function @xmath137 is given by @xmath138 [ ( p - k - p{^{\prime\prime}})^2-m^2 ] [ ( p{^{\prime}}-k)^2-m^2]}\ , .
\nonumber \end{aligned}\ ] ] the evaluation of @xmath131 is performed by using the standard technique for the evaluation of feynman diagrams ( for a short summary of the relevant formulas , see appendix d of ref .
first , we use three feynman parameters in order to join the 4 factors in the denominator of the integrand in eq .
( [ f2 ] ) . denoting the numerator as @xmath139
, we obtain @xmath140 ^ 4}\,,\end{aligned}\ ] ] where @xmath141 , @xmath142 , and @xmath143 are the feynman parameters ( here and below it is assumed that all integrals over the feynman parameters extend from 0 to 1 ) .
we denote @xmath144 , and @xmath145 .
next , we shift the integration variable @xmath146 and perform the integration over @xmath147 .
the result is @xmath148\,,\end{aligned}\ ] ] where @xmath149 is defined so that @xmath150 is the quadratic in @xmath147 part of @xmath139 .
we note that , after shifting the integration variable , only even powers of @xmath147 yield a non - zero contribution to the integral ( i.e. , the terms proportional @xmath151 and @xmath152 vanish ) .
next , the integration over @xmath153 is carried out .
we introduce the function @xmath154 by @xmath155\,,\end{aligned}\ ] ] where @xmath156 and @xmath157 .
the integral over @xmath158 in eq .
( [ f5 ] ) can be expressed in terms of the lewis integral @xcite .
however , we prefer to perform this integration straightforwardly by merging denominators using feynman parametrization . in this way
, we end up with an additional integration to be performed numerically , but the structure of the expressions involved becomes somewhat simpler .
let us illustrate the further evaluation by considering the contribution induced by the first term in the square brackets in eq .
( [ f5 ] ) , which will be denoted by @xmath159 .
we merge the denominators by introducing two more feynman parameters , @xmath160 ^ 4}\,,\end{aligned}\ ] ] where @xmath161 , @xmath162+(1-t){\bm{q}}\,,\ ] ] and @xmath163 ^ 2 + m^2 -(1-y)p^2-y{p{^{\prime}}}^2\right\}\,.\end{aligned}\ ] ] substituting eq .
( [ f6 ] ) into eq .
( [ f5 ] ) and shifting the integration variable , we get @xmath164 where @xmath165 and @xmath166 the above equation defines the @xmath167 functions as the coefficients from the expanded form of the expression @xmath168 . performing the integration over @xmath158 in eq .
, we obtain @xmath169 because @xmath170 is linear in @xmath171 , the integral over @xmath171 is elementary and can be expressed in terms of logarithms .
the four remaining integrations over the feynman parameters remain to be evaluated numerically . to complete the evaluation of @xmath159
, one needs to obtain explicit expressions for the numerators @xmath172 and to bring them to the standard form . under `` the standard form '' we understand a linear combination of independent matrix structures , see below .
this is the most tedious part of the calculation since the expressions involved are very lenghty .
usage of symbolic computation packages is indispensable in this case .
having obtained an expression for @xmath154 , we write the correction to the hfs as @xmath173 where we used the notation @xmath174 with @xmath175 denoting the levi - civita symbol . in eq .
( [ f9 ] ) , we indicate explicitly the dependence of @xmath176 on 32 basic matrix structures @xmath177 .
the main four of these are : @xmath178_z$ ] , @xmath179_z$ ] , @xmath180_z$ ] , and @xmath181_z$ ] .
the rest is obtained by multiplying each of them by @xmath182 , @xmath183 , @xmath184 , @xmath102 , @xmath185 , @xmath186 , and @xmath187 . in order to perform the integration over all angular variables in eq .
( [ f9 ] ) except for @xmath188 , we define the angular integrals @xmath189 that correspond to the basic matrices @xmath177 by @xmath190 where @xmath51 is an arbitrary function . all @xmath189 may be expressed in terms of the elementary angular integrals listed in appendix [ app:1 ] . using the angular integrals @xmath189
, we can write the final expression for the one - potential vertex term suitable for a numerical evaluation , @xmath191 altogether , eq . ( [ f11 ] ) contains 7 integrations to be performed numerically , 3 of them being written explicitly and 4 feynman - parameter integrations contained in the definition of the function @xmath176 .
the numerical evaluation was performed using gauss - legendre quadratures for all integrations . in order to prevent losses of accuracy due to numerical cancellations
, we used quadruple - precision arithmetic ( accurate to roughly 32 decimals ) in a small part of the code , which was identified to be numerically unstable .
the evaluation was rather time - consuming ( about a month of processor time for each value of @xmath5 and each state ) and was performed with the help of the parallel computational environment at mpi heidelberg .
the one - potential vertex part has been crucial to our calculation , and so it may be appropriate to summarize once more the basic steps in its evaluation : first of all , let us recall that our `` one - potential vertex part '' actually involves two vertices inside the loop , one being a coulomb vertex and the other being a magnetic vertex ( coupling to the external field ) . therefore , there are three fermion propagators inside the loop and one photon propagator , necessitating the introduction of three feynman parameters to join denominators .
the incoming coulomb momentum and the exchanged momentum with the external field entail two further feynman parameters , one of which is integrated out analytically .
in addition to the four remaining feynman parameters , we have two radial integrations over the absolute values of the initial ( @xmath192 ) and final ( @xmath193 ) electron momenta , and an integration over the direction cosine @xmath194 ( transformed by a change of variable to an integration over @xmath195 ) .
the three additional integrations account for the resulting seven - dimensional integral . in the corresponding calculation in free qed
, one could hope to carry out the radial integrations analytically , because the incoming and outgoing fermions are on the mass shell and described by plane waves . here , however , the bound states are being off the mass shell and have a much more complicated structure , so that the radial integrations have to be evaluated numerically .
the separate calculation of the full one - potential vertex part as described in the current section leads to a numerically favourable scheme , because this part can be then subtracted from the integrand of the remaining nonperturbative vertex contribution , thereby leading to a drastic improvement in the convergence of the resulting partial - wave expansion ( see table [ tab : ver2 ] below ) .
the general expression for the many - potential vertex part @xmath196 is obtained from eq .
( [ eq2 ] ) by applying the appropriate set of subtractions in the electron propagators .
the required subtractions are given by @xmath197 where @xmath38 denotes the bound - electron propagator , @xmath198 is the free - electron proparator , @xmath199 is the electron propagator with one interaction with the binding coulomb field , and @xmath200 is the reference - state part of the bound - electron propagator .
this subtraction takes into account all terms which have been calculated separately using different approaches , as described above . in order to perform a numerical evaluation of @xmath201 ,
it is convenient to rotate the integration contour of the photon energy @xmath84 from @xmath202 to be parallel to the imaginary axis of the @xmath84 complex plane . in this work ,
we define a deformed @xmath84 integration countour @xmath203 consisting of two parts , a low - energy part @xmath204 and a high - energy part @xmath205 .
the low - energy part contains the interval @xmath206 on the lower bank of the cut of the photon propagator and the interval @xmath207 on the upper bank of the cut , with @xmath208 .
the high - energy part consists of two intervals , @xmath209 and @xmath210 .
the contour @xmath203 defined in this way differs from the one used by p. j. mohr @xcite only by the choice of the separation point @xmath211 ( the value @xmath212 instead of @xmath213 was employed in ref .
@xcite ) .
the high - energy part of @xmath201 is given by @xmath214 \,,\end{aligned}\ ] ] where the subtractions are given by eq .
( [ g1 ] ) .
the low - energy part of @xmath201 needs a careful treatment because of single and double poles situated near the contour @xmath204 , which are due to virtual bound states of lower energy than the reference state .
the single poles can be integrated via a cauchy principal value prescription , and the double poles can be converted to single poles via an integration by parts .
we thus write the low - energy part of @xmath201 as @xmath215 \nonumber \\ & -\frac{1}{\pi}\ , \sum_{0 < { \varepsilon}_n<{\varepsilon}_a } \frac{f_{nn}(\delta ) } { { \varepsilon}_a-\delta-{\varepsilon}_{n } } \,,\end{aligned}\ ] ] where @xmath216|n_1a{\rangle}\,,\end{aligned}\ ] ] the prime denotes the derivative over @xmath84 , and @xmath217 denotes the cauchy principal value of the integral . in eq .
( [ g3 ] ) , all terms that induce double poles on the interval @xmath218 ( i.e. , intermediate states with @xmath219 ) have been integrated by parts .
we recall that the term with @xmath220 is removed by the @xmath200 part of the subtraction ( [ g1 ] ) .
the need to evaluate the principal value of the integral over @xmath84 complicates the numerical calculation of the low - energy part . in the case
when there is a single pole only ( which takes place for the @xmath221 and @xmath222 reference state ) , the problem is most easily solved by employing a numerical quadrature symmeric around the position of the pole . in the general case with more than one singularity to be treated ,
this approach is not effective .
a better way is to introduce subtractions in the integrand that remove the singularities at the poles and to evaluate the principal value of the integral of the subtracted terms analytically .
we introduce the subtractions by observing that the following difference does not have any singularities on the interval @xmath218 , @xmath223|n_1a{\rangle } } { ( { \varepsilon}_a-\omega-{\varepsilon}_{n_1 } ) ( { \varepsilon}_a-\omega-{\varepsilon}_{n_2 } ) } \nonumber \\ & \qquad - \sum_{0<{\varepsilon}_{n_1}<{\varepsilon}_a } \frac{{\langle}a \delta n_1| { \rm im}\left[i({\varepsilon}_a-{\varepsilon}_{n_1})\right]|n_1 a{\rangle } } { { \varepsilon}_a-\omega-{\varepsilon}_{n_1 } } \nonumber \\ & \qquad - \sum_{0<{\varepsilon}_{n_2}<{\varepsilon}_a } \frac{{\langle}an_2| { \rm im}\left[i({\varepsilon}_a-{\varepsilon}_{n_2})\right]|\delta n_2a{\rangle } } { { \varepsilon}_a-\omega-{\varepsilon}_{n_2}}\,,\end{aligned}\ ] ] where @xmath224 we note that the terms with @xmath225 and @xmath226 present in eq .
( [ g5 ] ) do not induce any singularitites because @xmath227=0 $ ] .
the perturbed wave function @xmath228 is known analytically ( for the hfs perturbing potential , both the diagonal and the non - diagonal in @xmath229 parts ; for the @xmath0-factor perturbing potential , only the diagonal in @xmath229 part ) from the generalized virial relations for the dirac equation @xcite . in order to complete our discussion of the evaluation of the many - potential vertex part
, we present the explicit expression for it after the integration over the angular variables .
this expression reads @xmath230 \,,\end{aligned}\ ] ] where @xmath231 is a relativistic generalization of the slater radial integral , whose explicit expression is given in ref .
@xmath232 is given by @xmath233 where @xmath234 is the clebsch - gordan coefficient , @xmath235 is a normalized spherical harmonic , and @xmath236\,.\ ] ] the numerical evaluation of the many - potential vertex contribution is the most difficult part of the calculation .
the key feature that limits the accuracy achievable in a numerical calculation is the convergence of the partial - wave expansion .
we recall that the many - potential vertex contribution @xmath196 contains two and more coulomb interactions ( and a magnetic interaction ) inside the self - energy loop .
the convergence of its partial - wave expansion is much better than that for the vertex contribution with just one coulomb interaction . in order to illustrate this point , table [ tab : ver2 ]
presents a comparison of the partial - wave expansion of the vertex contribution with two and more coulomb interactions , @xmath196 , and of that with one and more coulomb interactions , @xmath237 .
it can be seen that the subtraction of the one - potential vertex contribution improves the numerical accuracy by about 5 orders of magnitude .
r .... & & + @xmath238 & & & & + + ' '' '' 3 & 1.58x0875 & 1.58x0875 & 3.36124x192571 & 3.36124x192571 + ' '' '' 7 & -0.00x2317 & 1.57x8558 & -0.00000x693015 & 3.36123x499556 + ' '' '' 15 & -0.00x0660 & 1.57x7898 & -0.00000x079070 & 3.36123x420487 + ' '' '' 30 & -0.00x0183 & 1.57x7715 & -0.00000x009848 & 3.36123x410639 + ' '' '' 60 & -0.00x0049 & 1.57x7666 & -0.00000x001246 & 3.36123x409393 + ' '' ''
120 & -0.00x0010 & 1.57x7656 & -0.00000x000122 & 3.36123x409270 + ' '' '' extrap . &
-0.00x0015(15)&1.57x7641(15 ) & -0.00000x000041(44 ) & 3.36123x409229(44 ) + the partial - wave expansion was cut off at the maximum value of @xmath239 .
quadruple - precision arithmetics was required for the dirac green function in order to control the numerical accuracy at the required level .
this computation was performed with a quadruple - precision generalization of the code for the dirac green function developed in refs .
it should be mentioned that the evaluation of the high - energy part of @xmath196 with the integration contour @xmath203 requires the dirac green function with general complex values of the energy argument .
( this is in contrast to the approach used in refs .
@xcite , where the integration contour is chosen in such a way that only the real and purely imaginary values of the energy argument are required . )
the computation of the dirac green function for general complex energies @xmath84 becomes numerically unstable when @xmath229 is large and @xmath240 is close to @xmath241 . because of this , we were not able to extend the partial - wave summation beyond @xmath242 .
the general scheme of our evaluation is as follows .
we perform the summation over @xmath229 directly in the integrand , before any integrations .
the summation is terminated when a suitable convergence criterion is fulfilled or when the cutoff value @xmath243 is reached . in order to estimate the dependence of the final result on the cutoff parameter , results for several intermediate cutoffs are stored , each consequent one being twice larger than the previous ( see table [ tab : ver2 ] for an illustration ) .
the omitted tail of the expansion was estimated by using the @xmath244 algorithm for pad approximation , and the uncertainty of the extrapolation was taken about 50%-200% of the estimated tail . according to eq .
( [ eq8 ] ) , the reducible part of the se correction involves the derivative of the se operator , `` sandwiched '' in the reference state .
the zero - potential part of the reducible contribution has already been discussed in sec .
[ zeropotredver ] ; it involves the derivative of the free electron propagator with respect to the reference - state energy .
the reference - state contribution to the reducible part has been treated in sec .
[ refstatecontrib ] , together with the reference - state contribution to the vertex term , thereby mutually cancelling the ir divergence inherent to both reference - state contributions .
the total reference - state contribution is summarized in eq .
( [ d6 ] ) . left is the many - potential reducible part , @xmath245 here , @xmath246 and @xmath247 are obtained from eq .
( [ eq6 ] ) by a replacement of the full dirac coulomb green function @xmath38 by the free green function @xmath198 and by the reference - state part of the propagator @xmath200 . for the term with @xmath200
, we have @xmath248 in coordinate space , a representation of @xmath200 reads @xmath249 { { \varepsilon}_n = { \varepsilon}_a } \end{array } } \frac{\psi_n({\bm{x}}_1 ) \ , \psi_n^+({\bm{x}}_2)}{{\varepsilon}-\omega-{\varepsilon}_a + i \ , 0 } \,,\ ] ] where we take into account all states with the same energy as the reference state , i.e. , also the state with opposite parity but the same total angular momentum as compared to the reference state ( pairs of states with the same @xmath250 are energetically degenerate according to dirac theory ) .
the evaluation of eq .
( [ ered1 + ] ) proceeds along the integration contour @xmath251 of p. j. mohr @xcite for the ( complex rather than real ) photon energy .
it is divided into a low - energy and a high - energy part .
the low - energy contour @xmath252 comprises the interval @xmath253 below the cut of the photon propagator and the interval @xmath254 on the upper bank of the cut , with @xmath255 being the reference - state energy .
the high - energy contour @xmath256 again consists of two intervals , @xmath257 and @xmath258 . because the low - energy part extends to comparatively high values of @xmath259 , the radial integrand for each single value of @xmath229 become highly oscillatory .
the behaviour of the integrand can only be improved if the full sum over intermediate angular momenta is carried out before the radial integrations .
this is already evident from the model example given in eq .
( 7.3 ) of ref .
@xcite , @xmath260 \bigr)}{r[1-\rho ] } \ ; = \ ; - \ , \sum_{|\kappa|=0}^{\infty } ( 2|\kappa|+1 ) \ , j_{|\kappa| } ( i\rho \ , r ) \ , h^{(1)}_{|\kappa| } ( i r ) \,,\ ] ] where @xmath261 is a bessel function and @xmath262 is a hankel function of the first kind ( @xmath263 ) . the right - hand side of eq .
( [ ht1 ] ) involves functions that are highly oscillatory as a function of the radial variable @xmath65 , but the left - hand side is a simple exponential .
this `` smoothing '' phenomenon after the summation over the intermediate angular momenta is crucial for the evaluation as it enhances the rate of convergence of the multi - dimensional se integrals dramatically .
the convergence of the sum over @xmath250 can be further accelerated by the so - called cnc transformation @xcite . with maximum values of @xmath229 in excess of @xmath264 being handled at ease using the cnc transformation , we are able to control the accuracy of the final evaluations .
the derivative of the green function is calculated directly using fourth - point and ( alternatively , for verification ) six - point difference schemes .
we choose suitable values of the parameters so that the green function derivative is calculated to a relative accuracy of @xmath265 .
additional modifications are necessary in the extreme infrared region of photon energies ; here the difference scheme is adjusted so that the boundaries of the integration region are not crossed and sufficient accuracy is retained . a numerical subtraction of all singular terms due to lower - lying atomic states ( e.g. , the ground state ) before doing any integrations over the photon energies and before evaluating the derivative of the dirac propagator eliminates a potential further source of numerical loss of significance for the many - potential reducible part . with reference to eq .
( [ eq5 ] ) , we recall that the irreducible part is given as @xmath266 with the renormalized se operator @xmath267 and the perturbed wave function [ see eq .
( [ eq4 ] ) ] @xmath268 { { \varepsilon}_n \ne { \varepsilon}_a } \end{array } } \frac{|n{\rangle}{\langle}n|\delta v|a{\rangle}}{{\varepsilon}_a-{\varepsilon}_n } \,.\ ] ] we only need the diagonal - in-@xmath229 component of the perturbed wave function , because the se operator is also diagonal in the total angular momentum .
the evaluation of the irreducible part is carried out along the same contour @xmath251 that is used for the many - potential reducible part . within the high - energy part
, the green function is divided into two parts .
the first is a subtraction term which involves a free propagator and an approximate one - potential term @xcite , which is obtained from the full one - potential term by commuting the coulomb potential to the left of the electron propagators .
the second is the remainder term which is the difference of the full and the approximate propagator .
the subtraction term contains all uv divergences of the irreducible part ; these are cancelled against the mass counter term @xmath35 .
the subtraction term is evaluated in momentum space , in a noncovariant integration scheme adjusted for bound - state calculations , where the spatial components of the photon momentum are integrated out before the photon energy integration . l@. ' '' '' @xmath269 & -0.0109x6549784(5 ) + ' '' '' @xmath270 & 8.2895x6864683 + ' '' ''
@xmath271 & -3.8385x4412893(5 ) + ' '' '' @xmath272 & -5.5795x8625925 + ' '' '' @xmath131 & -1.7835x813412(16 ) + ' '' '' @xmath196 & 3.3612x340923(4 ) + ' '' '' @xmath93 & -0.0000x2366906 + ' '' '' total & 0.4381x018429(16 ) + cr .. @xmath13 & @xmath5 & & + + & 1 & 0.43810x18429(16 ) & -13.83x08(43 ) + & 2 & 0.37346x7600(3 ) & -14.11x70(9 ) + & 3 & 0.30758x3838(4 ) & -14.41x21(3 ) + & 4 & 0.24100x5731(5 ) & -14.69x62(2 ) + & 5 & 0.17402x6212(7 ) & -14.96x73(2 ) + & 6 & 0.10681x5805(11 ) & -15.22x64(1 ) + & 7 & 0.03947x649(2 ) & -15.47x52(1 ) + & 8 & -0.02793x233(2 ) & -15.71x56(1 ) + & 9 & -0.09537x946(3 ) & -15.94x87(1 ) + & 10 & -0.16285x352(4 ) & -16.17x62(1 ) + & 11 & -0.23035x739(4 ) & -16.39x90(1 ) + & 12 & -0.29790x470(4 ) & -16.61x81(1 ) + + @xmath14 & @xmath5 & & + + & 1 & 0.43869x2275(3 ) & -6.12x05(85 ) + & 2 & 0.37535x2042(3 ) & -6.91x26(11 ) + & 3 & 0.31120x3194(5 ) & -7.58x33(5 ) + & 4 & 0.24666x5425(7 ) & -8.16x97(3 ) + & 5 & 0.18193x8687(9 ) & -8.70x69(2 ) + & 6 & 0.11712x3392(13 ) & -9.19x73(2 ) + & 7 & 0.05226x463(2 ) & -9.65x46(1 ) + & 8 & -0.01262x594(2 ) & -10.08x59(1 ) + & 9 & -0.07755x851(3 ) & -10.49x64(1 ) + & 10 & -0.14255x836(3 ) & -10.89x01(1 ) + & 11 & -0.20766x159(3 ) & -11.27x01(1 ) + & 12 & -0.27291x238(4 ) & -11.63x90(1 ) + + @xmath17 & @xmath5 & & + + & 1 & 0.43893x143(3 ) & -1.72x7(69 ) + & 2 & 0.37613x687(3 ) & -2.61x38(90 ) + & 3 & 0.31274x865(3 ) & -3.35x30(27 ) + & 4 & 0.24914x165(3 ) & -3.99x94(12 ) + & 5 & 0.18548x734(3 ) & -4.58x06(7 ) + & 6 & 0.12186x545(3 ) & -5.11x33(4 ) + & 7 & 0.05830x586(3 ) & -5.60x92(2 ) + & 8 & -0.00519x144(3 ) & -6.07x57(1 ) + & 9 & -0.06864x589(3 ) & -6.51x89(1 ) + & 10 & -0.13209x008(3 ) & -6.94x30(1 ) + & 11 & -0.19556x594(4 ) & -7.35x15(1 ) + & 12 & -0.25912x227(5 ) & -7.74x73(1 ) + cr .. @xmath18 & @xmath5 & & + + & 1 & 0.24939x7018(5 ) & -11.3233x21(86 ) + & & 0.2487(x5)^a & + & 2 & 0.24801x6543(5 ) & -9.3117x68(25 ) + & 3 & 0.24608x7170(7 ) & -8.1642x78(14 ) + & 4 & 0.24371x9931(7 ) & -7.3707x86(9 ) + & 5 & 0.24098x5405(8 ) & -6.7713x55(6 ) + & & 0.2397^ax & + & 6 & 0.23793x2761(8 ) & -6.2946x96(4 ) + & 7 & 0.23459x7810(8 ) & -5.9027x68(3 ) + & 8 & 0.23100x7222(8 ) & -5.5728x57(2 ) + & 9 & 0.22718x0996(8 ) & -5.2903x09(2 ) + & 10 & 0.22313x4035(8 ) & -5.0451x23(2 ) + & & 0.2202^ax & + & 11 & 0.21887x7214(8 ) & -4.8301x70(1 ) + & 12 & 0.21441x8110(9 ) & -4.6401x91(1 ) + + @xmath19 & @xmath5 & & + + & 1 & -0.12499x329(1 ) & 0.1260x9(18 ) + & & -0.1254^bx & + & 2 & -0.12498x309(1 ) & 0.0794x05(55 ) + & 3 & -0.12498x458(2 ) & 0.0321x76(39 ) + & 4 & -0.12501x321(2 ) & -0.0154x99(29 ) + & 5 & -0.12508x457(3 ) & -0.0635x28(21 ) + & & -0.1255^bx & + & 6 & -0.12521x458(3 ) & -0.1119x33(16 ) + & 7 & -0.12541x922(3 ) & -0.1606x63(12 ) + & 8 & -0.12571x467(3 ) & -0.2096x98(9 ) + & 9 & -0.12611x728(3 ) & -0.2590x28(7 ) + & 10 & -0.12664x357(3 ) & -0.3086x43(5 ) + & & -0.1271^bx & + & 11 & -0.12731x021(3 ) & -0.3585x38(5 ) + & 12 & -0.12813x408(3 ) & -0.4087x11(4 ) + @xmath273 ref . @xcite + @xmath274 ref . @xcite + for the se correction to the hfs of the @xmath13 , @xmath14 , and @xmath17 states . [ fig : hos ] ] for the se correction to the hfs of the @xmath18 and @xmath19 states .
[ fig : hop ] ] cr .. @xmath13 & @xmath5 & & + + & 1 & 2322.84x0230(2 ) & 19x.(3 . ) + & 2 & 2322.90x4037(4 ) & 23x.3(2.6 ) + & 3 & 2323.01x4295(8 ) & 22x.88(70 ) + & 4 & 2323.17x5525(14 ) & 22x.58(29 ) + & 5 & 2323.39x298(2 ) & 22x.36(15 ) + & 6 & 2323.67x243(3 ) & 22x.18(9 ) + & 7 & 2324.02x001(5 ) & 22x.02(6 ) + & 8 & 2324.44x213(7 ) & 21x.87(4 ) + & 9 & 2324.94x538(8 ) & 21x.71(3 ) + & 10 & 2325.53x651(10 ) & 21x.57(2 ) + & 11 & 2326.22x235(13 ) & 21x.42(2 ) + & 12 & 2327.00x983(12 ) & 21x.28(1 ) + + @xmath14 & @xmath5 & & + + & 1 & 2322.82x4624(2 ) & + & 2 & 2322.84x0323(8 ) & + & 3 & 2322.86x696(2 ) & 19x.(11 . ) + & 4 & 2322.90x509(3 ) & 22x.1(4.7 ) + & 5 & 2322.95x533(5 ) & 22x.5(2.4 ) + & 6 & 2323.01x836(6 ) & 22x.5(1.3 ) + & 7 & 2323.09x490(8 ) & 22x.43(74 ) + & 8 & 2323.18x571(8 ) & 22x.31(42 ) + & 9 & 2323.29x154(9 ) & 22x.18(24 ) + & 10 & 2323.41x319(9 ) & 22x.05(15 ) + & 11 & 2323.55x142(11 ) & 21x.92(11 ) + & 12 & 2323.70x704(12 ) & 21x.79(8 ) + + @xmath17 & @xmath5 & & + & 1 & 2322.82x1746(7 ) & + & 2 & 2322.82x8684(10 ) & + & 3 & 2322.84x038(2 ) & + & 4 & 2322.85x698(3 ) & 19x.(18 . ) + & 5 & 2322.87x866(5 ) & 20x.7(8.4 ) + & 6 & 2322.90x560(6 ) & 21x.5(4.6 ) + & 7 & 2322.93x798(9 ) & 21x.8(2.9 ) + & 8 & 2322.97x600(12 ) & 22x.0(2.0 ) + & 9 & 2323.01x98(2 ) & 22x.0(1.4 ) + & 10 & 2323.06x97(2 ) & 22x.0(1.0 ) + & 11 & 2323.12x58(2 ) & 21x.89(77 ) + & 12 & 2323.18x82(3 ) & 21x.81(59 ) + cr .. @xmath18 & @xmath5 & & + + & 1 & -774.25x8151(3 ) & 0.121x258(21 ) + & 2 & -774.21x2929(11 ) & 0.121x715(22 ) + & 3 & -774.13x687(2 ) & 0.122x414(19 ) + & 4 & -774.02x917(3 ) & 0.123x280(14 ) + & 5 & -773.88x876(3 ) & 0.124x305(10 ) + & 6 & -773.71x442(3 ) & 0.125x473(7 ) + & 7 & -773.50x460(3 ) & 0.126x803(5 ) + & 8 & -773.25x838(3 ) & 0.128x186(4 ) + & 9 & -772.97x356(3 ) & 0.129x711(2 ) + & 10 & -772.64x862(3 ) & 0.131x336(2 ) + & 11 & -772.28x175(4 ) & 0.133x054(3 ) + & 12 & -771.87x108(8 ) & 0.134x858(5 ) + + @xmath19 & @xmath5 & & + + & 1 & 774.29x1470 ( 3 ) & 0.148x104(21 ) + & 2 & 774.34x6522(12 ) & 0.148x294(24 ) + & 3 & 774.43x854(3 ) & 0.148x567(24 ) + & 4 & 774.56x774(4 ) & 0.148x851(22 ) + & 5 & 774.73x495(6 ) & 0.149x338(18 ) + & 6 & 774.94x028(6 ) & 0.149x816(14 ) + & 7 & 775.18x442(6 ) & 0.150x350(11 ) + & 8 & 775.46x799(6 ) & 0.150x933(7 ) + & 9 & 775.79x167(5 ) & 0.151x561(4 ) + & 10 & 776.15x615(5 ) & 0.152x231(4 ) + & 11 & 776.56x218(6 ) & 0.152x940(4 ) + & 12 & 777.01x055(6 ) & 0.153x685(4 ) + factor of the @xmath18 and @xmath19 states .
[ fig : hop : gfact ] ]
our calculation of the se correction to the hfs and the @xmath0 factor of hydrogen - like ions was performed in the feynman gauge and for a point nucleus .
the fine structure constant of @xmath275 was used in the calculation .
the small deviation of this value from the currently accepted one @xcite does not influence the numerical results for the higher - order remainder .
an example set of individual contributions to the se correction to the @xmath13 hfs of atomic hydrogen is presented in table [ tab : breakdown ] .
the se correction to the hfs of an @xmath2 state can be represented as @xmath276 \ , a_{22}+ \ln[(z\alpha)^{-2 } ] \,a_{21 } + a_{20 } \biggr\ } \nonumber \\ & + ( { { z\alpha}})^3\,\ln[(z\alpha)^{-2 } ] \,a_{31}+ ( { { z\alpha}})^3\,f_{ns}({{z\alpha } } ) \biggr]\,,\end{aligned}\ ] ] where @xmath277 is the non - relativistic hfs value , and the @xmath278 are coefficients of the @xmath7-expansion with the first index corresponding to the power of @xmath7 and the second corresponding to the power of the logarithm .
we have @xmath279 a_{22}(ns ) = & \ ; -2/3\ , , \quad a_{31}(ns ) = -13.30741592\,,\end{aligned}\ ] ] @xmath280 a_{21}(2s ) = & \ ; 0.317103926 \ , , \nonumber\\[2ex ] a_{21}(3s ) = & \ ; 0.921048823 \,,\end{aligned}\ ] ] @xmath281 a_{20}(2s ) = & \ ; 11.90110542\ , , \nonumber\\[2ex ] a_{20}(3s ) = & \ ; 10.41704775\,,\end{aligned}\ ] ] see the recent articles @xcite and references therein for earlier studies .
@xmath282 is the higher - order remainder , which we address in our numerical all - order approach .
our numerical results for the se correction to the hfs of the @xmath13 , @xmath14 , and @xmath17 states are listed in table [ tab : hfs : s ] .
the se correction to the hfs of an @xmath283 states is much less studied .
only the leading term of its @xmath4 expansion is known today .
the correction , therefore , is written as @xmath284\,,\end{aligned}\ ] ] with @xmath285 being the higher - order remainder .
the coefficient @xmath286 is given by @xmath287 and @xmath288 @xcite .
our numerical results for the se correction to the hfs of the @xmath18 and @xmath19 states are listed in table [ tab : hfs : p ] .
the results for the higher - order remainders @xmath282 and @xmath285 inferred from our numerical data are plotted in figs .
[ fig : hos ] and [ fig : hop ] , respectively . for the @xmath18 state ,
a fit of our results is consistent with the @xmath4 expansion of the form @xmath289 + a_{20}+ \ldots \,,\end{aligned}\ ] ] where the value of the logarithmic coefficient is very close to @xmath290 and the constant term is about @xmath291 . for the @xmath19 state ,
the numerical data are consistent with @xmath292 .
our results in table [ tab : hfs : p ] are in moderate agreement with those obtained previously in refs .
@xcite but significantly improve upon them in numerical accuracy
. nevertheless , we disagree with the suggestion @xcite about the possible presence of the squared logarithm in the @xmath4 expansion ( [ ee4 ] ) for the @xmath18 state .
a more careful investigation of the analytic structure of the higher - order terms is performed in the follow - up paper @xcite .
the se correction to the bound - electron @xmath0 factor of an @xmath2 state can be represented as @xmath293\,b_{41 } + b_{40}\biggr\ } \nonumber \\ & + \frac{({{z\alpha}})^5}{n^3}\,h_{ns}({{z\alpha } } ) \biggr]\,,\end{aligned}\ ] ] where the @xmath294 are known coefficients of the @xmath4 expansion : @xmath295 b_{40}(1s ) = & \ ; -10.23652432 \ , , \nonumber\\[2ex ] b_{40}(2s ) = & \ ; -10.70771560 \ , , \nonumber\\[2ex ] b_{40}(3s ) = & \ ; -11.52963397 \,,\end{aligned}\ ] ] see ref .
@xcite and references therein .
@xmath296 is the remainder incorporating all higher - order contributions .
it is remarkable that the higher - order remainder @xmath296 enters in the relative order @xmath10 rather than in the relative order @xmath9 , as in the case of the hfs .
this means that cancellations in extracting the remainder from numerical results for @xmath12 are larger for the @xmath0 factor than for the hfs by four orders of magnitude .
our numerical results for the se correction to the @xmath0 factor of the electron @xmath13 , @xmath14 , and @xmath17 states of light hydrogen - like ions are presented in table [ tab : gfact : s ] .
we observe that the higher - order remainder behaves very similarly for all @xmath2 states studied , the @xmath14 remainder being just about 2% larger than that for the @xmath13 states and the @xmath17 and @xmath14 remainders being equal within the numerical uncertainty .
the accuracy of the direct numerical determination of the @xmath13 remainder for @xmath12 and @xmath297 can easily be increased by extrapolating values obtained for higher values of @xmath5 .
an extrapolation yields the improved results @xmath298 and @xmath299 .
improved values of the higher - order remainder for the @xmath14 and @xmath17 states are most easily obtained by scaling the @xmath13 remainder . the trend of the higher - order remainder for low @xmath5 is consistent with a numerically large , @xmath300-independent coefficient @xmath301 in eq .
( [ ee30 ] ) . for the @xmath302 states
, the bound - electron @xmath0 factor is studied less thoroughly than for the @xmath303 states .
the leading term of its @xmath4 expansion is due to the electron anomalous magnetic moment ( amm ) and is immediately obtained for a general state as @xcite @xmath304 where @xmath229 is the dirac quantum number and @xmath261 is the total angular momentum of the electron state . for the @xmath217 states , the explicit results are @xmath305 and @xmath306 .
the next - order term , @xmath307 , consists of two parts , one induced by the electron amm and the other , by the emission and the absorption of virtual photons of low energy ( commensurate with the electron binding energy ) .
a simple calculation of the first part gives @xcite @xmath308 and @xmath309 .
the second part is nonvanishing for states with @xmath310 only and is more complicated .
its general expression is known @xcite but the only numerical result available for hydrogenic atoms is the estimate made in ref .
@xcite , which disagrees with our numerical values both in the sign and the magnitude .
commenting on this fact , we note that the estimate is based on a rather crude approximation .
namely , the sum over the entire discrete and continuous spectrum of virtual states was replaced by the contribution of the lowest 12 discrete bound states only .
we argue that such approximation might be inapplicable for the problem in hand .
the reason is that , e.g. , for the bethe logarithm ( which is also a contribution induced by the low - energy photons ) the dominant contribution originates from the continuum spectrum @xcite so that such approximation is clearly inadequate .
since the @xmath311 term is not presently known anyalytically , we define the higher - order remainder for the @xmath302 states as @xmath312\,.\end{aligned}\ ] ] our numerical results for the se correction to the @xmath0 factor of the electron @xmath18 and @xmath19 states of light hydrogen - like ions are listed in table [ tab : gfact : p ] . the corresponding higher - order remainder function is plotted in fig .
[ fig : hop : gfact ] .
we have discussed , in detail , a numerical evaluation , nonperturbative in the binding coulomb field , of the self - energy correction to the hyperfine splitting and of the self - energy correction to the @xmath0 factor in hydrogen - like ions with low nuclear charge number @xmath313 .
we consider the ground state , as well as the @xmath14 and @xmath17 excited states , and the @xmath18 as well as @xmath19 states .
the value of @xmath314 is employed in all calculations . at the level of precision
we are operating at , the final results for the self - energy corrections depend on the precise value employed very sensitively .
however , the main dependence on the value of @xmath6 is accounted for by the analytically known lower - order terms .
thus , the results for the remainder functions @xmath315 , @xmath316 , @xmath317 , and @xmath318 as given in tables [ tab : hfs : s ] , [ tab : hfs : p ] , [ tab : gfact : s ] and [ tab : gfact : p ] are not influenced by the value of @xmath6 employed .
even if a value of @xmath6 which differs from @xmath314 on the level @xmath319 were employed , then the values of the remainder functions would not change : their main uncertainty is due to limits of convergence of the integrals that constitute the nonperturbative self - energy corrections , as described in the preceding sections of this article .
the organization of our calculation is described in sec .
[ detailed ] .
we consider separately the reference - state contribution to the reducible and the vertex part ( sec .
[ refstatecontrib ] ) and the zero - potential contribution to the vertex and to the reducible part ( sec .
[ zeropotredver ] ) .
the one - potential and many - potential vertex parts , which represent the most challenging part of the calculation , are discussed in secs .
[ onepotver ] and [ manypotver ] .
the many - potential reducible part and the irreducible part conclude the discussion of our computational method ( secs .
[ manypotred ] and [ irreducible ] ) .
numerical calculations were carried out on the parallel computing environments of mpi heidelberg and mst rolla .
it is instructive to compare the numerical results obtained ( tables [ tab : hfs : s][tab : gfact : p ] ) to analytic results from the @xmath4 expansion .
the analytic parameterization of the self - energy correction to the hyperfine splitting according to eq .
( [ ee2 ] ) entails both logarithmic as well as nonlogarithmic corrections .
our numerical results for the scaled self - energy correction @xmath320 to the hyperfine splitting and for the nonperturbative remainder function @xmath315 are given in table [ tab : hfs : s ] for @xmath303 states .
a plot of the data ( see fig . [
fig : hos ] ) indicates that the higher - order remainders @xmath315 , for @xmath321 , may converge toward an @xmath322 coefficient which is significantly dependent on the principal quantum number .
the scaled self - energy correction @xmath323 for @xmath18 and @xmath19 states is analyzed in table [ tab : hfs : p ] .
a plot of the data ( see fig . [
fig : hop ] ) aids in a comparison to an analytic model for the correction , given in eq .
( [ ee4 ] ) .
the nonperturbative remainder @xmath324 for the hfs of @xmath217 states is seen to be well represented by an analytic model of the form @xmath325 + a_{20}+\ldots \}$ ] , where a fit to the numerical data indicates that @xmath290 and that @xmath326 .
it has been suggested in ref .
@xcite that a `` double '' ( squared ) logarithm in the @xmath4 expansion could be present for low nuclear charge ; the latter would correspond to a nonvanishing @xmath327 coefficient for @xmath217 states .
our numerical results in table [ tab : hfs : p ] do not contradict those of ref .
@xcite on the level of numerical accuracy obtained in the cited reference .
however , we can not confirm the presence of such a double - logarithmic correction ( see also @xcite ) .
a large number of analytic terms are known for the self - energy correction to the @xmath0 factor for @xmath303 states [ see eq .
( [ ee30 ] ) ] .
the higher - order remainder @xmath317 for the @xmath0 factor of @xmath303 states is thus `` separated '' from the leading - order effect by about ten orders of magnitude for @xmath12 .
thus , although our direct numerical evaluation of the self - energy correction for the ground state at @xmath12 is precise [ @xmath328 at @xmath12 ] , we can only infer the higher - order remainder @xmath330 at @xmath12 to about @xmath33110% : the result after the subtraction of lower - order terms is @xmath332 . by extrapolation of more accurate data for the remainder obtained from higher values of the nuclear charge ,
we can obtain the improved results @xmath298 and @xmath299 .
the remainder functions at very low @xmath5 appear to depend only very slightly on the principal quantum number ; they are consistent with @xmath333 approaching an @xmath300-independent coefficient @xmath301 as @xmath334 . for the @xmath0 factor of @xmath217 states ,
only the leading coefficient @xmath335 is known from the @xmath4 expansion [ see eq .
( [ ee31 ] ) ] . the self - energy remainder function @xmath318 for the @xmath0 factor of @xmath217 states
can be inferred from our numerical data in table [ tab : gfact : p ] after subtraction of the leading analytic term as given in eq .
( [ ee31 ] ) .
the numerical data for @xmath217 states are consistent with the functions @xmath318 tending toward a constant for @xmath321 . a plot of the data in fig .
[ fig : hop : gfact ] confirms this trend . to conclude
, we have performed an all - order ( in @xmath7 ) calculation of the self - energy correction to hyperfine splitting and @xmath0 factor in hydrogen - like ions with low nuclear charge numbers .
the calculation is accurate enough to infer higher - order remainder terms without any additional extrapolation , by a simple subtraction of the known terms in the @xmath7-expansion .
we improve the numerical accuracy by several orders of magnitude as compared to previous evaluations ; this leads to improved theoretical predictions for all qed effects considered in this article .
enlightening discussions with k. pachucki on the @xmath0 factor of @xmath217 states are gratefully acknowledged .
v.a.y . was supported by dfg ( grant no . 436 rus 113/853/0 - 1 ) , by rfbr ( grant no .
06 - 02 - 04007 ) , and by the foundation `` dynasty . ''
. has been supported by the national science foundation ( grant phy-8555454 ) as well as by a precision measurement grant from the national institute of standards and technology .
in this section we demonstrate how to perform the integration over the angular variables in momentum space .
the problem in hand can be formulated as follows : the general expression of the form @xmath336 where @xmath51 , @xmath337 , and @xmath338 are some arbitrary functions and @xmath339 , needs to be integrated over all angular variables exept for @xmath194 , i.e. , to be reduced to the form @xmath340 in order to write a general expression for the function @xmath341 in terms of @xmath337 and @xmath338 , we use the standard decomposition of the function @xmath51 in terms of the spherical harmonics @xmath342 , @xmath343 where @xmath344 are the legendre polynomials . from this , we immedately have @xmath345 \nonumber \\ & \times \left[\int d{\hat{\bm{p}}}_2\,b({\hat{\bm{p}}}_2)y_{lm}({\hat{\bm{p}}}_2)\right]\,.\end{aligned}\ ] ] this general formula greatly simplifies when one of the functions ( say , @xmath338 ) is unity or the identity function , i.e. , @xmath346 . when @xmath338 is unity , only the term with @xmath347 contributes , and we have : @xmath348 when @xmath349 , @xmath350 in more complicated cases with @xmath351 , formulas for @xmath352 can be in priciple obtained by using the racah algebra .
alternatively , one can observe [ from eq .
( [ app3 ] ) ] that @xmath352 is a combination of the legendre polynomials with some coefficients .
it is straightforward to find the coefficients by performing integrations in eq .
( [ app3 ] ) analytically for each particular case ( where advantage may be taken of computer algebra ) .
@xmath357_z\ , \chi_{-\kappa,\mu}({\hat{\bm{p}}}_2 ) \nonumber \\ & = \int_{-1}^1 d\xi\,f(p_{1r},p_{2r},\xi)\ , k_1(\kappa)\ , , \\
\frac{3i}{4\pi } \int d{\hat{\bm{p}}}_1\ , d{\hat{\bm{p}}}_2\ , & \
f(p_{1r},p_{2r},\xi)\ , \chi^{\dag}_{\kappa , \mu}({\hat{\bm{p}}}_1)\ , [ { \hat{\bm{p}}}_2\times { \bm{\sigma}}]_z\ , \chi_{-\kappa,\mu}({\hat{\bm{p}}}_2 ) \nonumber \\ & = \int_{-1}^1 d\xi\,f(p_{1r},p_{2r},\xi)\ , k_1^{\prime}(\kappa)\ , , \\
\frac{3i}{4\pi } \int d{\hat{\bm{p}}}_1\ , d{\hat{\bm{p}}}_2\ , & \
f(p_{1r},p_{2r},\xi)\ , \chi^{\dag}_{\kappa , \mu}({\hat{\bm{p}}}_1)\ , [ { \hat{\bm{p}}}_1\times { \hat{\bm{p}}}_2]_z\ , \chi_{\kappa,\mu}({\hat{\bm{p}}}_2 ) \nonumber \\ & = \int_{-1}^1 d\xi\,f(p_{1r},p_{2r},\xi)\ , k_2(\kappa)\ , , \\ \frac{3i}{4\pi } \int d{\hat{\bm{p}}}_1\ , d{\hat{\bm{p}}}_2\ , & \ f(p_{1r},p_{2r},\xi)\ , \chi^{\dag}_{\kappa , \mu}({\hat{\bm{p}}}_1)\ , i { \bm{\sigma}}_z\ , \chi_{\kappa,\mu}({\hat{\bm{p}}}_2 ) \nonumber \\ & = \int_{-1}^1 d\xi\,f(p_{1r},p_{2r},\xi)\ , k_3(\kappa)\,.\end{aligned}\ ] ] @xmath358 1\ , , & \kappa = 1\ , , \\[2ex ] \displaystyle \tfrac15 \ , ( 1 - 3 \,\xi^2)\ , , & \kappa = -2\,,\\[2ex ] \displaystyle \tfrac25 \ , \xi\ , , & \kappa = 2\ , , \end{array } \right .
\\[2ex ] k_1^{\prime}(\kappa ) = & \ ; \left\ { \begin{array}{cl } -1\ , , & \kappa = -1\ , , \\[2ex ] \xi\ , , & \kappa = 1\ , , \\[2ex ] \displaystyle -\tfrac25 \xi\ , , & \kappa = -2\,,\\[2ex ] \displaystyle -\tfrac15 ( 1 - 3\xi^2)\ , , & \kappa = 2\ , , \end{array } \right.\\[2ex ] \nonumber\end{aligned}\ ] ] as well as @xmath359 -\tfrac12 ( 1-\xi^2)\ , , & \kappa = 1\ , , \\[2ex ] -\tfrac14 ( 1-\xi^2)\ , , & \kappa = -2\,,\\[2ex ] -\tfrac{9}{20 } \xi(1-\xi^2)\ , , & \kappa = 2\ , , \end{array } \right .
\\[2ex ] k_3(\kappa ) = & \ ; \left\ { \begin{array}{cl } -\tfrac32\ , , & \kappa = -1\ , , \\[2ex ] \tfrac12 \xi\ , , & \kappa = 1\ , , \\[2ex ] -\tfrac12 \xi\ , , & \kappa = -2\,,\\[2ex ] -\tfrac{3}{20 } ( 1 - 3\xi^2)\ , , & \kappa = 2\ , .
\end{array } \right.\end{aligned}\ ] ] j. calmet , h. grotch and d. a. owen , phys .
a * 17 * , 1218 ( 1978 ) .
this comment estimates the low - energy contribution to the @xmath0 factor of the @xmath1 states of hydrogen to be @xmath360 . in order to convert this value to the @xmath361 factor addressed in the present investigation
, it should be multiplied by @xmath362/[2j(j+1)]$ ] . in units of the coefficient @xmath307 defined by eq .
( [ ee30 ] ) , this estimate corresponds to @xmath363 and @xmath364 , in strong disagreement with our numerical results .
|
the hyperfine structure ( hfs ) and the @xmath0 factor of a bound electron are caused by external magnetic fields .
for the hfs , the magnetic field is due to the nuclear spin .
a uniform - in - space and constant - in - time magnetic field is used to probe the bound - electron @xmath0 factor .
the self - energy corrections to these effects are more difficult to evaluate than those to the lamb shift . here , we describe a numerical approach for both effects in the notoriously problematic regime of hydrogen - like bound systems with low nuclear charge numbers . the calculation is nonperturbative in the binding coulomb field .
accurate numerical values for the remainder functions are provided for @xmath1 states and for @xmath2 states with @xmath3 .
|
gravitational lensing remains a powerful tool for studies of dark matter distribution in the universe .
one important application of gravitational lens statistics is to provide strong constraints on the cosmological models and parameters since the lensing optical depth depends on the cosmological volume element ( surdej et al .
1992 , mitchell et al .
2005 , oguri et al .
2008 , jullo et al .
2010 , rozo et al .
2011 ) .
an ideal set of observations to carry out extragalactic research and especially for searching gravitational lens systems is the cfhtls photometric survey .
previous investigators have mainly set up automatic procedures to search for gravitational lens systems .
cabanac et al .
( 2007 ) have presented 40 strong lens candidates over 28 sq .
deg . of the cfhtls field , which consist of galaxy / group deflectors producing conspicuous gravitational arc systems .
more et al . (
2012 ) have reported about 127 lens system candidates ( strong lensing legacy survey - arcs ) over 150 sq .
deg in the cfhtls . in that work , they applied semi - automatic techniques to find gravitational arcs with the help of some additional visual inspection .
maturi et al .
( 2013 ) studied the statistical color properties of gravitational arcs over 37 sq .
deg . in the cfhtls - archive - research survey and found 73 new arc candidates .
sygnet et al .
( 2010 ) have found 3 good candidates for edge - on galaxy lens applying their automatic / visual procedure to 30444 individual cfhtls relatively bright spirals 18@xmath121 .
shan et al.(2012 ) have performed a weak lensing analysis over 64 sq .
deg . of the cfhtls fields .
we propose here to search for multiply imaged qsos / agn among the counterparts of x - ray point - like sources detected with xmm .
therefore contrary to the previous approaches , we have been searching for gravitationally lensed qsos / agn at much smaller angular scales ( typically @xmath2-@xmath3 ) .
the search for gravitational lens candidates among the optical counterparts of x - ray selected qsos / agn looked a priori very promising as , first of all , it allows the selection of a sample of qsos / agn with very few contaminants .
moreover , as the x - ray qsos / agn are bright and distant sources ( observed up to a redshift @xmath4 ) , they have a high probability of being multiply imaged due to the presence of a deflector near their line - of - sight . from lensing probability calculations over a mock catalogue of x - ray sources ( reproducing the redshift distribution and differential number counts of the cosmos x - ray sources as a function of their flux ) , where we have assumed a constant comoving density of deflector galaxies modelled as singular isothermal spheres , we expect @xmath5 qsos / agn among the xmm - lss soft x - ray sources to have a deflector sufficiently close to their line - of - sight to lead to the formation of multiple lensed images of their optical counterpart ( surdej et al . 2011 , finet et al .
2012 ) .
the contiguous @xmath0 deg@xmath6 medium - deep xmm - lss field has been surveyed at high galactic latitudes with the xmm - newton satellite , centered on ra @xmath7 dec @xmath8 ( j2000 ) .
it provides the identification of a homogeneous sample of galaxy clusters and qsos / agn ( pierre et al .
2011 ) and reaches a sensitivity near @xmath9 w m@xmath10 for point - like sources in the soft band ( elyiv et al .
2012 ) . about @xmath11 of the 6000 point - like xmm - lss x - ray sources have ( an ) optical cfhtls dr6 counterpart(s ) ( chiappetti et al .
visual inspection of all counterparts led us to classify the sources as being either extended or point - like and reject the wrong associations ( melnyk et al .
2013 ) . in this paper
we describe separately the results of our gravitational lens search among the cfhtls counterparts of x - ray sources . in section 2
we detail the method used to identify multiply imaged qsos / agn in the xmm - lss field . in section 3
we describe the adopted procedure to search for multiply imaged qsos / agn , we discuss each individual candidate and report their color measurements as well as their morphology .
section 4 contains the main conclusions .
our sample of x - ray point - like sources and their optical counterparts consist of objects which have been detected in the soft and/or the hard x - ray band and coming from the multiwavelength catalogue of chiappetti et al .
the @xmath12 optical counterparts of the x - ray sources were visually inspected on @xmath13 cfhtls optical images centered on the x - ray sources .
the angular separations between the lensed images for a typical galaxy deflector and cosmological redshifts of the source ( qso / agn ) and the lens are typically of the order of 1 or several arcsec ( refsdal and surdej 1994 ) . knowing that lensed components should be point - like sources and should have rather similar color properties ( color difference smaller than @xmath14 mag . ) , this selection was also carried out using these strict criteria .
contamination by the lens , intrinsic color variation plus time delay(s ) and/or micro - lensing acting may cause color differences above typical cfhtls photometric errors .
the psf of the xmm - newton telescopes provides a resolution near @xmath15 for a point - like source .
a detected point - like x - ray source may therefore be sometimes identified with multiple ( more than one ) optical counterparts .
we have made an in - depth study of these particular cases in order to detect new gravitational lens systems ( mostly multiply imaged qsos / agn ) .
we inspected _
, _ r _ and _ i _ cfhtls direct ccd frames whenever they were available . for visual inspection we used the monochromatic images in logarithmic scale for each band as well as the color composed images . for better objectivity four independent members of the present team ( called hereafter
inspectors ) ranked each counterpart in the following way : @xmath16 - very good candidate , @xmath17 - good lens candidate , @xmath18 - possible lens candidate and @xmath19 - definitely not a candidate .
a similar approach has been followed by more et al .
( 2008 ) when searching for gravitational lenses in the extended chandra deep field south and by jackson ( 2008 ) in the cosmos field .
we finally retained nearly 300 candidates which were marked with the 1 , 2 or 3 rank by at least one inspector .
next we added the ranks from all inspectors and formed a top list of 72 probable lens candidates .
we visually inspected this list several times and removed very faint objects , objects which are obvious stars according to their spectra and so on . at the end
, we finally selected 18 multiply imaged qso / agn candidates which could be potentially strong lens systems , see fig .
[ fig1 ] and table 1 .
they are ranked there from the best to worst according to final morphological and color selection .
most of them consist of double sources . for 5 pairs with an angular separation larger than @xmath20 we considered their photometry in the @xmath21 , @xmath22 and @xmath23 bands directly from the cfhtls catalog .
we used available spectroscopic or photometric redshifts from sdss and melnyk et al .
( 2013 ) . for the remaining more compact systems we performed a multiple psf fitting analysis to precise their morphology ( point - like or extended source ) and accurate magnitude measurement of each component .
we adopted the same psf fitting technique as that used for the analysis of the gravitational lens systems he 0435 - 1223 described in ricci et al .
( 2011 ) and um673 in ricci et al .
( 2013 ) . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] candidates * j022739.7 - 050044 , j022509.7 - 050950 , j021956.2 - 052809 , j021929.6 - 040856 , j021736.4 - 041944 , j021947.4 - 041034 , j021649.2 - 054327 , j021902.8 - 054109 , j021634.8 - 043302 and j022646.2 - 045156 * consist of pairs of extended sources according to our psf fitting analysis .
j021634.8 - 043302 * consists of a wide pair and we took their magnitudes from the cfhtls data base .
the components show similar colors : @xmath24 and @xmath25 for the a and b components , @xmath26 and @xmath27 , respectively . from the sdss database ,
the estimated photo - z is @xmath28 and @xmath29 for the a and b components , respectively . in the sdss database
there is just a note that the two components are galaxies .
we performed the psf fitting of these two components and confirm that they are both extended , see fig .
[ figj021634.8 - 043302 ] . candidate * j022646.2 - 045156 * looks very different in the various bands . in the g - band
it looks like a single point - like source according to the psf fitting results . in the r - band
we see a faint secondary component .
however the bright component looks more compact than the reference star ( psf ) .
this could hint on the presence of a defect or cosmic ray in the r - band .
it was therefore not possible to correctly define the _ r _ magnitude . in the i - band
the bright component appears as an extended source according to psf fitting .
the sdss sed analysis shows that it is a qso / agn with z=1.085
. it could be that the host galaxy of this qso / agn has been detected in the i - band .
using the multiple psf fitting technique , we found 11 gravitational lens system candidates which consist of extended components : j021844.4 - 044825 , j022739.7 - 050044 , j022509.7 - 050950 , j021956.2 - 052809 , j021929.6 - 040856 , j021736.4 - 041944 , j021947.4 - 041034 , j021649.2 - 054327 , j021902.8 - 054109 , j021634.8 - 043302 and j022646.2 - 045156 . the candidate j021936.7 - 055721 consists of two components which have similar colors but one of the components is extended and the other one is point - like . using magnitudes from the cfhtls survey or from psf fitting for the case of compact systems
, we found that the following candidates have very different colors : j022055.1 - 060132 , j022500.2 - 052204 and j022324.0 - 054928 .
the two candidates j021511.4 - 034306 and j022234.3 - 031616 consist of pairs of point - like sources having very similar colors .
candidate j021511.4 - 034306 is very bright in x - ray , @xmath30 and @xmath31 w m@xmath10 in the soft and hard bands , respectively .
it corresponds to a hardness ratio @xmath32 which points to a source being likely an unobscured agn .
candidate j022234.3 - 031616 is faint in x - ray and is only seen in the soft band , @xmath33 w m@xmath10 .
the optical counterparts of the j022607.0 - 040301 candidate have slightly different colors but it could be caused by the presence of a lens , intrinsic color variation of the agn combined with a time delay and/or microlensing effects .
the system is bright in x - ray , about @xmath34 w m@xmath10 in both the soft and hard bands .
its corresponding hardness ratio @xmath35 indicates that it is a likely unobscured agn .
these last three systems ( j021511.4 - 034306 , j022234.3 - 031616 and j022607.0 - 040301 ) constitute the best gravitational lens candidates . from the xmm - lss source catalog ( chiappetti et al .
2013 ) we have constructed the color - color diagram ( @xmath36 vs. @xmath37 ; see fig .
3 ) for all spectroscopically confirmed stars and qso / agn counterparts of xmm - lss point - like sources ( see classification of sources in melnyk et al . 2013 ) . from this plot
we suggest that the j021511.4 - 034306 and j022607.0 - 040301 candidates are probably agn . for the candidate j022234.3 - 031616 we have only magnitudes in @xmath21 and @xmath22 bands and we can not classify them as qsos / agn or stars on the sole basis of the color - color diagram .
we have performed a visual inspection of some 5500 optical counterparts of x - ray point - like sources in the xmm - lss / cfhtls common field .
four independent inspectors have identified 18 compact systems which could be multiply imaged qso / agn . using the psf fitting technique we have characterized the morphology and colors of the components . as a result
, we found * three * systems ( j021511.4 - 034306 , j022234.3 - 031616 and j022607.0 - 040301 ) which met our color and morphological criteria and are the best gravitational lens candidates .
they are point - like source pairs with similar colors . with a high probability
they could be qso / agn . for a final confirmation we need spectroscopic observations of the multiple components of the selected systems .
ae ( post - doc prodex ) and js acknowledge support the esa prodex programme
xmm - lss , from the belgian federal science policy office and from the communaut franaise de belgique - actions de recherche concertes - acadmie universitaire wallonie - europe .
ae is grateful to stephen gwyn for help with the cfhtls images .
akiyama et al .
2013 in prep .
bixler j. v. , bowyer s. , & laget m. 1991 , a&a , 250 , 370 bohlin r. c. , colina l. , & finley d. s. 1995 , aj , 110 , 1316 cabanac r. a. et al .
2007 , a&a , 461 , 813 chiappetti l. , et al .
2013 , mnras , 429 , 1652 elyiv a. , et al .
2012 , a&a , 537 , a131 finet , f. , elyiv , a. , & surdej , j. 2012 , memorie della societa astronomica italiana , 83 , 944 jackson , n. 2008 , mnras , 389 , 1311 jullo e. , natarajan p. , kneib j .- p .
, daloisio a. , limousin m. , richard j. , schimd c. 2010 , science , 329 , 924 maturi m. , mizera s. , seidel g. 2013 , submitted to a&a , arxiv:1305.3608 melnyk o. et al .
2013 , accepted to a&a , arxiv:1307.0527 mitchell , j. l. , keeton , c. r. , frieman , j. a. , & sheth , r. k. 2005 , apj , 622 , 81 more a. , mckean j. p. , muxlow t. w. b. , porcas r. w. , fassnacht c. d. , koopmans l. v. e. 2008 , mnras , 384 , 1701 more a. , cabanac r. , more s. , alard c. , limousin m. , kneib j .-
, gavazzi r. , motta v. 2012 , apj , 749 , 38 oguri m. et al . 2008 , aj , 135 , 512 pickles , a. j. 1998 , pasp , 110 , 863 pierre m. , pacaud f. , juin j. b. , melin j. b. , valageas p. , clerc n. , corasaniti p. s. 2011 , mnras , 414 , 1732 refsdal , s. , & surdej , j. 1994 , reports on progress in physics , 57 , 117 ricci , d. et al .
2011 , a&a , 528 , a42 ricci , d. et al .
2013 , a&a , 551 , a104 rozo , e. , wu , h .- y . , &
schmidt , f. 2011 , apj , 735 , 118 rowan - robinson m. et al .
2008 , mnras , 386 , 697 shan h. et al .
2012 , apj , 748 , 56 sygnet , j. f. , tu , h. , fort , b. , & gavazzi , r. 2010 , a&a , 517 , a25 surdej , j. , remy , m. , & haubold , h. 1992 , european southern observatory conference and workshop proceedings , 44 , 127 surdej , j. , elyiv , a. , finet , f. , melnyk , o. , talk entitled `` agn gravitational lensing in the xxl survey '' presented during the xxl meeting http://arachnos.astro.ulg.ac.be/rpub/colloques/xxl2 , bonifacio , 5 may 2011
|
our aim was to identify gravitational lens candidates among some 5500 optical counterparts of the x - ray point - like sources in the medium - deep @xmath0 sq .
deg . xmm - lss survey .
we have visually inspected the optical counterparts of each qsos / agn using cfhtls t006 images .
we have selected compact pairs and groups of sources which could be multiply imaged qso / agn .
we have measured the colors and characterized the morphological types of the selected sources using the multiple psf fitting technique .
we found three good gravitational lens candidates : j021511.4 - 034306 , j022234.3 - 031616 and j022607.0 - 040301 which consist of pairs of point - like sources having similar colors . on the basis of a color - color diagram and x - ray properties
we could verify that all these sources are good qso / agn candidates rather than stars .
additional secondary gravitational lens candidates are also reported .
[ firstpage ] gravitational lensing : strong
|
given a cyclic group @xmath7 and a generator @xmath8 of @xmath9 , the discrete logarithm ( dl ) of @xmath10 is the element @xmath11 such that @xmath12 .
in well - chosen groups , the exponentiation @xmath13 is very fast but computing @xmath14 from @xmath15 is conjectured to be very difficult : this is the discrete logarithm problem ( dlp ) , at the heart of many asymmetric cryptosystems .
the first group proposed for dlp was the multiplicative group of a prime finite field . nowadays , the group of points of elliptic curves defined over prime fields are replacing the prime fields for dlp - based cryptosystems . in pairing - based cryptography ,
the finite fields are still used , because they are a piece in the pairing mechanism .
it is important in cryptography to know precisely the difficulty of dl computation in the considered groups , to estimate the security of the cryptosystems using them .
finite fields have a particularity : there exists a subexponential - time algorithm to compute dl in finite fields of medium to large characteristic : the number field sieve ( nfs ) . in small characteristic
, this is even better : a quasi - polynomial - time algorithm was proposed very recently @xcite . in may 2015 ,
an international team of academic researchers revealed a surprisingly efficient attack against a diffie - hellman key exchange in tls , the _ logjam _ attack @xcite .
after a seven - day - precomputation stage ( for relation collection and linear algebra of nfs - dl algorithm ) , it was made possible to compute any given individual dl in about one minute , for each of the two targeted 512-bit prime finite fields .
this was fast enough for a man - in - the - middle attack .
this experience shows how critical it can be to be able to compute individual logarithms very fast .
another interesting application for fast individual dl is _ batch - dlp _ , and _ delayed - target dlp _ : in these contexts , an attacker aims to compute several dl in the same finite field . since the costly phases of relation collection and linear algebra are only done one time for any fixed finite field , only the time for one individual dl is multiplied by the number of targets
this context usually arises in pairing - based cryptography and in particular in broadcast protocols and traitor tracing schemes , where a large number of dlp - based public / private key pairs are generated .
the time to compute one individual dl is important in this context , even if parallelization is available .
we recall that the nfs algorithm is made of four steps : polynomial selection , relation collection , linear algebra and finally , individual logarithm computation . _
this last step is mandatory to break any given instance of a discrete logarithm problem .
_ the polynomial selection outputs two irreducible polynomials @xmath16 and @xmath8 defining two number fields @xmath17 and @xmath18 .
one considers the rings @xmath19/(f(x))$ ] and @xmath20/(g(x))$ ] .
there exist two maps @xmath21 to @xmath0 , as shown in the following diagram .
moreover , the monic polynomial defining the finite field is @xmath22 , of degree @xmath23 . ( qqx ) @xmath24 $ ] ; ( middle ) ; ( kf ) @xmath19/(f(x))$ ] ; ( kg ) @xmath25/(g(y ) ) = r_g$ ] ; ( qqx ) node ( kf ) ; ( qqx ) node ( kg ) ; ( gfpn ) @xmath26/(\psi(z))$ ] ; ( kf ) ( gfpn ) ; ( rho - f ) @xmath27 ; ( kg ) ( gfpn ) ; ( rho - g ) @xmath28 ; in the remaining of this paper , we will only use @xmath29 , @xmath30 and @xmath31 .
after the relation collection and linear algebra phases , the ( virtual ) logarithm of a subset of elements in each ring @xmath32 is known .
the individual dl step computes a preimage in one of the rings @xmath32 of the target element in @xmath0 .
if one can write the target preimage as a product of elements of known ( virtual ) logarithm , then one can deduce the individual dl of the target .
the key point of individual dl computation is finding a smooth decomposition in small enough factors of the target preimage .
the asymptotic running time of nfs algorithm steps are estimated with the @xmath33-function : @xmath34 = \exp \bigl ( \bigl(c+o(1)\bigr ) ( \log q)^\alpha ( \log \log q)^{1-\alpha } \bigr ) \quad \mbox { with } \alpha \in [ 0,1 ] \mbox { and } c > 0~.\ ] ] the @xmath35 parameter measures the gap between polynomial time ( @xmath36=\log^c q$ ] ) and exponential time ( @xmath37 = q^c$ ] ) . when @xmath38 is implicit , or obvious from the context , we simply write @xmath39 $ ] .
when the complexity relates to an algorithm for a prime field @xmath40 , we write @xmath41 $ ] .
[ [ large - prime - fields . ] ] large prime fields .
+ + + + + + + + + + + + + + + + + + + many improvements for computing discrete logarithms first concerned prime fields .
the first subexponential dl algorithm in prime fields was due to adleman @xcite and had a complexity of @xmath42 $ ] . in 1986 ,
coppersmith , odlyzko and schroeppel @xcite introduced a new algorithm ( cos ) , of complexity @xmath43 $ ] .
they computed individual dl @xcite in @xmath44 $ ] in two steps ( finding a boot of medium - sized primes , then finding relations of logarithms in the database for each medium prime ) .
in these two algorithms , the factor basis was quite large ( the smoothness bound was @xmath44 $ ] in both cases ) , providing a much faster individual dl compared to relation collection and linear algebra .
this is where the common belief that individual logarithms are easy to find ( and have a negligible cost compared with the prior relation collection and linear algebra phases ) comes from . in 1993 , gordon @xcite proposed the first version of nfs dl algorithm for prime fields @xmath40 with asymptotic complexity @xmath45 $ ] .
however , with the @xmath46 $ ] algorithm there are new difficulties , among them the individual dl phase . in this @xmath46 $ ] algorithm , many fewer logarithms of small elements are known , because of a smaller smoothness bound ( in @xmath46 $ ] instead of @xmath47 $ ] ) .
the relation collection is shortened , explaining the @xmath46 $ ] running time .
but in the individual dl phase , since some non - small elements in the decomposition of the target have an unknown logarithm , a dedicated sieving and linear algebra phase is done for each of them .
gordon estimated the running - time of individual dl computation to be @xmath48 $ ] , i.e. _ the same as the first two phases_. in 1998 , weber @xcite compared the nfs
dl algorithm to the cos algorithm for a 85 decimal digit prime and made the same observation about individual dl cost . in 2003 , ten years after gordon s algorithm , joux and lercier @xcite were the first to dissociate in nfs relation collection plus linear algebra on one side and individual dl on the other side .
they used the _
special_-@xmath49 technique to find the logarithm of medium - sized elements in the target decomposition .
in 2006 , commeine and semaev @xcite analyzed the joux
lercier method .
they obtained an asymptotic complexity of @xmath50 $ ] for computing individual logarithms , independent of the relation collection and linear algebra phases . in 2013 , barbulescu ( * ? ? ? * , 7.3 ) gave a tight analysis of the individual dl computation for prime fields , decomposed in three steps : booting ( also called smoothing ) , descent , and final combination of logarithms .
the booting step has an asymptotic complexity of @xmath51 $ ] and the descent step of @xmath52 $ ] .
the final computation has a negligible cost .
[ [ non - prime - fields - of - medium - to - large - characteristic . ] ] non - prime fields of medium to large characteristic .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in 2006 , joux , lercier , smart and vercauteren @xcite computed a discrete logarithm in a cubic extension of a prime field .
they used the special-@xmath49 descent technique again .
they proposed for large characteristic fields an equivalent of the _ rational reconstruction _ technique for prime fields and the _ waterloo _ algorithm @xcite for small characteristic fields , to improve the initializing step preceding the descent .
for dls in prime fields , the target is an integer modulo @xmath1 .
the rational reconstruction method outputs two integers of half size compared to @xmath1 , such that their quotient is equal to the target element modulo @xmath1 . finding a smooth decomposition of the target modulo @xmath1 becomes equivalent to finding a ( simultaneous ) smooth decomposition of two elements , each of half the size .
we explain their method ( that we call the jlsv fraction method in the following ) for extension fields in sec .
[ sec : jlsv06 fraction ] .
[ [ link - with - polynomial - selection . ] ] link with polynomial selection .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the running - time for finding a smooth decomposition depends on the norm of the target preimage .
the norm preimage depends on the polynomial defining the number field .
in particular , the smaller the coefficients and degree of the polynomial , the smaller the preimage norm .
some polynomial selection methods output polynomials that produce much smaller norm . that may be one of the reasons why the record computation of joux _ et al . _
@xcite used another polynomial selection method , whose first polynomial has very small coefficients , and the second one has coefficients of size @xmath53 . thanks to the very small coefficients of the first polynomial , their fraction technique was very useful .
their polynomial selection technique is now superseded by their jlsv@xmath54 method @xcite for larger values of @xmath1 . as noted in @xcite ,
the fraction technique is useful in practice for small @xmath23 .
but for the jlsv@xmath54 method and @xmath3 , this is already too slow ( compared to not using it ) . in 2008 , zajac @xcite implemented the nfs - dl algorithm for computing dls in @xmath55 with @xmath1 of 40 bits ( 12 decimal digits ( dd ) , i.e. @xmath55 of 240 bits or 74 dd ) .
he used the methods described in @xcite , with a first polynomial with very small coefficients and a second one with coefficients in @xmath53 .
in this case , individual dl computation was possible ( see the well - documented @xcite ) . in 2013 ,
hayasaka , aoki , kobayashi and takagi @xcite computed a dl in @xmath56 with @xmath57 ( @xmath58 of 203 bits or 62 dd ) .
we noted that all these records used the same polynomial selection method , so that one of the polynomials has very small coefficients ( e.g. @xmath59 ) whereas the second one has coefficients in @xmath53 . in 2009 ,
joux , lercier , naccache and thom @xcite proposed an attack of dlp in a protocol context . the relation collection is sped up with queries to an oracle .
they wrote in ( * ? ? ? * b ) an extended analysis of individual dl computation . in their case , the individual logarithm phase of the nfs - dl algorithm has a running - time of @xmath60 $ ] where @xmath61 in the large characteristic case , and @xmath62 in the medium characteristic case .
in 2014 , barbulescu and pierrot @xcite presented a multiple number field sieve variant ( mnfs ) for extension fields , based on coppersmith s ideas @xcite .
the individual logarithm is studied in ( * ? ? ?
they also used a _ descent _ technique , for a global estimated running time in @xmath63 $ ] , with a constant @xmath64 . recently in 2014 , barbulescu , gaudry , guillevic and morain @xcite announced 160 and 180 decimal digit discrete logarithm records in quadratic fields .
they also used a technique derived from the jlsv fraction method and a special-@xmath49 descent technique , but did not give an asymptotic running - time .
it appears that this technique becomes inefficient as soon as @xmath23 = 3 or 4 .
[ [ overview - of - nfs - dl - asymptotic - complexities . ] ] overview of nfs - dl asymptotic complexities .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the running - time of the relation collection step and the individual dl step rely on the smoothness probability of integers .
an integer is said to be @xmath65-smooth if all its prime divisors are less than @xmath65 .
an ideal in a number field is said to be @xmath65-smooth if it factors into prime ideals whose norms are bounded by @xmath65 . usually , the relation collection and the linear algebra are balanced , so that they have both the same dominating asymptotic complexity . the nfs algorithm for dl in prime and large characteristic fields
has a dominating complexity of @xmath66 $ ] . for the individual dl in a prime field @xmath40 ,
the norm of the target preimage in the number field is bounded by @xmath1 .
this bound gives the running time of this fourth step ( much smaller than relation collection and linear algebra ) .
finding a smooth decomposition of the preimage and computing the individual logarithm ( see @xcite ) has complexity @xmath67 $ ] with @xmath68 , and @xmath69 with the improvements of @xcite .
the booting step is dominating . in large characteristic fields ,
the individual dl has a complexity of @xmath70 $ ] , dominated by the booting step again ( ( * ? ? ?
* b ) for jlsv@xmath71 , table [ tab : new booting step cpx ] for gjl ) . in generic medium characteristic fields ,
the complexity of the nfs algorithm is @xmath72 $ ] with the jlsv@xmath54 method proposed in @xcite , @xmath73 $ ] with the conjugation method @xcite , and @xmath74 $ ] with the mnfs version @xcite .
we focus on the individual dl step with the jlsv@xmath54 and conjugation methods . in these cases ,
the preimage norm bound is in fact much higher than in prime fields . without any improvements ,
the dominating booting step has a complexity of @xmath60 $ ] with @xmath62 @xcite or @xmath75 ( * ? ? ?
however , this requires to sieve over ideals of degree @xmath76 . for the conjugation method ,
this is worse : the booting step has a running - time of @xmath77 $ ] ( see our computations in table [ tab : new booting step cpx ] ) .
applying the jlsv fraction method lowers the norm bound to @xmath78 for the conjugation method .
the individual logarithm in this case has complexity @xmath79 $ ] as for prime fields ( without the improvements of @xcite ) .
however , this method is not suited for number fields generated with the jlsv@xmath54 method , for @xmath3 . in practice
, we realized that the jlsv fraction method which seems interesting and sufficient because of the @xmath78 bound , is in fact not convenient for the gjl and conjugation methods for @xmath23 greater than 3 .
the preimage norm is much too large , so finding a smooth factorization is too slow by an order of magnitude .
we propose a way to lift the target from the finite field to the number field , such that the norm is strictly smaller than @xmath78 for the gjl and conjugation methods : [ th : smaller norm bound ] let @xmath6 and @xmath80 a random element ( not in a proper subfield of @xmath0 ) .
we want to compute its discrete logarithm modulo @xmath81 , where @xmath82 , with @xmath83 the @xmath23-th cyclotomic polynomial .
let @xmath17 be the number field given by a polynomial selection method , whose defining polynomial has the smallest coefficient size , and @xmath84/(f(x))$ ] .
then there exists a preimage @xmath85 in @xmath31 of some @xmath86 , such that @xmath87 and such that the norm of @xmath85 in @xmath17 is bounded by @xmath88 , where @xmath89 is equal to 1 .
@xmath90 for the gjl and conjugation methods ; 2 .
@xmath91 for the jlsv@xmath54 method ; 3 .
@xmath92 for the conjugation method , if @xmath17 has a well - chosen quadratic subfield satisfying the conditions of lemma [ lemma : fp2 simplification ] ; 4 .
@xmath93 for the jlsv@xmath54 method , if @xmath17 has a well - chosen quadratic subfield satisfying the conditions of lemma [ lemma : fp2 simplification ] .
our method reaches the optimal bound of @xmath94 , with @xmath95 the euler totient function , for @xmath96 combined with the gjl or the conjugation method .
we show that our method provides a dramatic improvement for individual logarithm computation for small @xmath23 : the running - time of the booting step ( finding boots ) is @xmath60 $ ] with @xmath97 for @xmath98 , @xmath99 for @xmath100 and @xmath101 for @xmath102 .
it generalizes to any @xmath23 , so that the norm is always smaller than @xmath78 ( the prime field case ) , hence the booting step running - time in @xmath60 $ ] always satisfies @xmath103 for the two state - of - the - art variants of nfs for extension fields ( we have @xmath104 for prime fields ) . for the jlsv@xmath54 method ,
this bound is satisfied for @xmath105 , where we have @xmath106 ( see table [ tab : new booting step cpx ] ) .
we select three polynomial selection methods involved for nfs - dl in generic extension fields and recall their properties in sec .
[ subsec : polyselect ] .
we recall a commonly used bound on the norm of an element in a number field ( sec .
[ subsec : norm bound ] ) .
we present in sec .
[ sec : jlsv06 fraction ] a generalization of the jlsv fraction method of @xcite . in sec .
[ sec : asympt cpx ] we give a proof of the booting step complexity stated in lemma [ th : asympt cpx ] .
we sketch in sec .
[ subsec : special - q descent cpx ] the special-@xmath49 descent technique and list the asymptotic complexities found in the literature according to the polynomial selection methods .
we present in sec .
[ sec : preimage ] our main idea to reduce the norm of the preimage in the number field , by reducing the preimage coefficient size with the lll algorithm .
we improve our technique in sec .
[ sec : preimage smaller norm ] by using a quadratic subfield when available , to finally complete the proof of theorem [ th : smaller norm bound ] .
we provide practical examples in sec .
[ sec : practical examples ] , for 180 dd finite fields in sec .
[ subsec : 180dd examples ] and we give our running - time experiments for a 120 dd finite field @xmath107 in sec .
[ subsec : fp4 120dd ] .
we recall an important property of the lll algorithm @xcite that we will widely use in this paper .
given a lattice @xmath108 of @xmath109 defined by a basis given in an @xmath110 matrix @xmath33 , and parameters @xmath111 , @xmath112 , the lll algorithm outputs a @xmath113-_reduced basis _ of the lattice .
the coefficients of the first ( shortest ) vector are bounded by @xmath114 with @xmath113 close to @xmath115 ( as in ntl or magma ) , the approximation factor @xmath116 is bounded by @xmath117 ( see @xcite ) ) .
gama and nguyen experiments @xcite on numerous random lattices showed that on average , @xmath118 . in the remaining of this paper
, we will simply denote by @xmath119 this lll approximation factor .
we will study the booting step of the nfs algorithm with these three polynomial selection methods : 1 .
the joux lercier
vercauteren ( jlsv@xmath54 ) method @xcite ; 2 .
the generalized joux lercier ( gjl ) method @xcite , @xcite ; 3 .
the conjugation method @xcite . in a non - multiple nfs version ,
the jlsv@xmath71 @xcite and gjl methods have the best asymptotic running - time in the large characteristic case , while the conjugation method holds the best one in the medium characteristic case .
however for a record computation in @xmath4 , the conjugation method was used @xcite . for medium characteristic fields of record size ( between 150 and 200 dd ) ,
is seems also that the jlsv@xmath54 method could be chosen ( @xcite ) . since the use of each method is not fixed in practice , we study and compare the three above methods for the individual logarithm step of nfs .
we recall now the construction and properties of these three methods .
[ [ jouxlerciersmartvercauteren - jlsv_1-method . ] ] joux lercier smart vercauteren ( jlsv@xmath54 ) method .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this method was introduced in 2006 .
we describe it in algorithm [ alg : jlsv1 ] .
the two polynomials @xmath120 have degree @xmath23 and coefficient size @xmath121 .
we set @xmath22 monic of degree @xmath23 .
we will use @xmath122 to represent the finite field extension @xmath123/(\psi(x))$ ] . select @xmath124 , two polynomials with small integer coefficients ,
@xmath125 @xmath126 a rational reconstruction of @xmath127 modulo @xmath1 @xmath128 [ [ generalized - jouxlercier - gjl - method . ] ] generalized joux
lercier ( gjl ) method .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this method was independently proposed in @xcite and @xcite ( see also @xcite ) .
this is a generalization of the joux
lercier method @xcite for prime fields .
we sketch this method in algorithm [ alg : gjl ] .
the coefficients of @xmath8 have size @xmath129 and those of @xmath16 have size @xmath130 , with @xmath131 and @xmath132 .
choose a polynomial @xmath133 of degree @xmath134 with small integer coefficients which has a monic irreducible factor @xmath135 of degree @xmath23 modulo @xmath1 reduce the following matrix using lll @xmath136 [ [ conjugation - method . ] ] conjugation method .
+ + + + + + + + + + + + + + + + + + + this method was published in @xcite and used for the discrete logarithm record in @xmath4 , with @xmath137 .
the coefficient size of @xmath16 is in @xmath130 and the coefficient size of @xmath8 is in @xmath121 .
we describe it in algorithm [ alg : conjugation ] .
@xmath138 @xmath126 a rational reconstruction of @xmath127 @xmath139 @xmath140 in sec .
[ sec : preimage ] we will compute the norm of an element @xmath141 in a number field @xmath17 .
we will need an upper bound of this norm .
for all the polynomial selection methods chosen , @xmath16 is monic , whereas @xmath8 is not .
we remove the leading coefficient of @xmath16 from any formula involved with a monic @xmath16 .
so let @xmath16 be a monic irreducible polynomial over @xmath142 and let @xmath143/(f(x))$ ] a number field .
write @xmath144 as a polynomial in @xmath145 , i.e. @xmath146 . the norm is defined by a resultant computation : @xmath147 we use kalkbrener s bound ( * ? ? ?
* corollary 2 ) for an upper bound : @xmath148 where @xmath149 , and @xmath150 the absolute value of the greatest coefficient .
an upper bound for @xmath151 is @xmath152 .
we will use the following bound in sec .
[ sec : preimage ] : @xmath153 row and column indices . in the following
, we will define matrices of size @xmath154 , with @xmath155 .
for ease of notation , we will index the rows and columns from @xmath156 to @xmath157 instead of 1 to @xmath158 , so that the @xmath159-th row at index @xmath160 , @xmath161_{0 \leq j \leq d-1}$ ] , can be written in polynomial form @xmath162 , and the column index @xmath163 coincides with the degree @xmath163 of @xmath164 . in 2006
was proposed in @xcite a method to generalize to non - prime fields the rational reconstruction method used for prime fields . in the prime field
setting , the target is an integer modulo @xmath1 .
the rational reconstruction method outputs two integers of half size compared to @xmath1 and such that their quotient is equal to the target element modulo @xmath1 . finding a smooth decomposition of the target modulo @xmath1 becomes equivalent to finding at the same time a smooth decomposition of two integers of half size each . to generalize to extension fields ,
one writes the target preimage as a quotient of two number field elements , each with a smaller norm compared to the original preimage .
we denote by @xmath141 the target in the finite field @xmath0 and by @xmath165 a preimage ( or lift ) in @xmath166 .
here is a first very simple preimage choice .
let @xmath123/(\psi(x))$ ] and @xmath167 , with @xmath168 .
we lift the coefficients @xmath169 to @xmath170 then we set a preimage of @xmath141 in the number field @xmath166 to be @xmath171 with @xmath172 such that @xmath173/(f(x))$ ] .
( we can also write @xmath174 , with @xmath35 a root of @xmath16 in the number field : @xmath175
we have @xmath176 .
now lll is used to obtain a quotient whose numerator and denominator have smaller coefficients .
we present here the lattice used with the jlsv@xmath54 polynomial selection method .
the number field @xmath166 is of degree @xmath23 .
we define a lattice of dimension @xmath177 . for the corresponding matrix
, each column of the left half corresponds to a power of @xmath172 in the numerator ; each column of the right half corresponds to a power of @xmath172 in the denominator .
the matrix is @xmath178 the first @xmath23 coefficients of the output vector , @xmath179 give a numerator @xmath180 and the last @xmath23 coefficients give a denominator @xmath181 , so that @xmath182 with @xmath14 a scalar in @xmath142
. the coefficients @xmath183 are bounded by @xmath184 since the matrix determinant is @xmath185 and the matrix is of size @xmath186
. however the product of the norms of each @xmath187 in the number field @xmath166 will be much larger than the norm of the single element @xmath165 because of the large coefficients of @xmath16 in the norm formula .
we use formula to estimate this bound : @xmath188 and the same for @xmath189 , hence the product of the two norms is bounded by @xmath190 .
the norm of @xmath165 is bounded by @xmath191 which is much smaller whenever @xmath192 .
finding a smooth decomposition of @xmath180 and @xmath181 at the same time will be much slower than finding one for @xmath165 directly , for large @xmath1 and @xmath193 .
this is mainly because of the large coefficients of @xmath16 ( in @xmath121 ) .
the method of @xcite to improve the smoothness of the target norm in the number field @xmath17 has an advantage for the gjl and conjugation methods .
first we note that the number field degree is larger than @xmath23 : this is @xmath194 for the gjl method and @xmath177 for the conjugation method .
for ease of notation , we denote by @xmath195 the degree of @xmath16 .
we define a lattice of dimension @xmath196 .
hence there is more place to reduce the coefficient size of the target @xmath165 .
we put @xmath1 on the diagonal of the first @xmath197 rows , then @xmath198 coefficients from row @xmath23 to @xmath199 , where @xmath200 ( @xmath122 is of degree @xmath23 and has @xmath201 coefficients ) .
the rows from index @xmath195 to @xmath196 are filled with @xmath202 ( these elements have @xmath195 coefficients ) .
we obtain a triangular matrix @xmath33 .
@xmath203 since the determinant is @xmath204 and the matrix of dimension @xmath205 , the coefficients obtained with lll will be bounded by @xmath206 .
the norm of the numerator or the denominator ( with @xmath207 ) is bounded by @xmath208 the product of the two norms will be bounded by @xmath78 hence we will have the same asymptotic running time as for prime fields , for finding a smooth decomposition of the target in a number field obtained with the gjl or conjugation method .
we will show in sec .
[ sec : preimage ] that we can do even better .
in this section , we prove the following lemma on the booting step running - time to find a smooth decomposition of the norm preimage .
this was already proven especially for an initial norm bound of @xmath78 .
we state it in the general case of a norm bound of @xmath209 .
the smoothness bound @xmath210 $ ] used here is not the same as for the relation collection step , where the smoothness bound was @xmath211 $ ] .
consequently , the special-@xmath49 output in the booting step will be bounded by @xmath65 .
[ th : asympt cpx ] let @xmath212 of order @xmath81 .
take at random @xmath213 $ ] and assume that the norm @xmath214 of a preimage of @xmath215 , in the number field @xmath17 , is bounded by @xmath216 $ ] .
write @xmath217 $ ] the smoothness bound for @xmath214 .
then the lower bound of the expected running time for finding @xmath218 s.t .
the norm @xmath214 of @xmath219 is @xmath220-smooth is @xmath221 $ ] , obtained with @xmath222 and @xmath223 .
first , we need a result on smoothness probability .
we recall the definition of @xmath65-smoothness already stated in sec .
[ sec : contributions ] : an integer @xmath224 is @xmath65-smooth if and only if all its prime divisors are less than or equal to @xmath65 .
we also recall the @xmath33-notation widely used for sub - exponential asymptotic complexities : @xmath34 = \exp \bigl ( \bigl(c+o(1)\bigr ) ( \log q)^\alpha ( \log \log q)^{1-\alpha } \bigr ) \quad \mbox { with } \alpha \in [ 0,1 ] \mbox { and } c > 0~.\ ] ] the canfield
pomerance @xcite theorem provides a useful result to measure smoothness probability : [ cor : b - smooth pr ] suppose @xmath225 , @xmath226 , and @xmath227 are fixed . for a random integer @xmath224 bounded by @xmath228 $ ] and a smoothness bound @xmath229 $ ] , the probability that @xmath224 is @xmath65-smooth is @xmath230\ ] ] for @xmath231 .
we prove now the lemma [ th : asympt cpx ] that states the running - time of individual logarithm when the norm of the target in a number field is bounded by @xmath232 . from theorem
[ cor : b - smooth pr ] , the probability that @xmath224 bounded by @xmath216 $ ] is @xmath220-smooth with @xmath217 $ ] is @xmath233 $ ] .
we assume that a @xmath220-smoothness test with ecm takes time @xmath234 = l_q [ \frac{\alpha_{b}}{2 } , ( 2 \gamma \alpha_{b})^{1/2}]$ ] .
the running - time for finding a @xmath220-smooth decomposition of @xmath224 is the ratio of the time per test ( ecm cost ) to the @xmath220-smoothness probability of @xmath224 : @xmath235 l_q \bigl [ 1-\alpha_{b } , ( 1-\alpha_{b})\frac{e}{\gamma } \bigr ] ~.\ ] ] we optimize first the @xmath35 value , so that @xmath236 ( that is , not exceeding the @xmath35 of the two previous steps of the nfs algorithm ) : @xmath237 this gives the system @xmath238 so we conclude that @xmath239 .
the running - time for finding a @xmath220-smooth decomposition of @xmath224 is therefore @xmath240~.\ ] ] the minimum : this is @xmath241 and find that the minimum of @xmath242 for @xmath243 is @xmath244 . with @xmath245 and @xmath246 , we obtain the minimum : @xmath247 .
] of the function @xmath248 is @xmath249 , corresponding to @xmath223 , which yields our optimal running time , together with the special-@xmath49 bound @xmath65 : @xmath250\quad \mbox { with } q \leq b = l_q\bigl[2/3 , ( e^2/3)^{1/3}\bigr]~.\ ] ] the second step of the individual logarithm computation is the _
special-@xmath49 descent_. this consists in computing the logarithms of the medium - sized elements in the factorization of the target in the number field . the first special-@xmath49 is of order @xmath251 $ ] ( this is the boot obtained in the initialization step ) and is the norm of a degree one prime ideal in the number field where the booting step was done ( usually @xmath17 ) . the idea is to _ sieve _ over linear combinations of degree one ideals , in @xmath17 and @xmath18 at the same time , whose norms for one side will be multiples of @xmath49 by construction , in order to obtain a relation involving a degree one prime ideal of norm @xmath49 and other degree one prime ideals of norm strictly smaller than @xmath49 .
here is the common way to obtain such a relation .
let @xmath252 be a degree one prime ideal of @xmath17 , whose norm is @xmath49 .
we can write @xmath253 , with @xmath254 a root of @xmath16 modulo @xmath49 ( hence @xmath255 ) .
we need to compute two ideals @xmath256 whose respective norm is a multiple of @xmath49 , and sieve over @xmath257 .
the classical way to construct these two ideals is to reduce the two - dimensional lattice generated by @xmath49 and @xmath258 , i.e. to compute @xmath259 @xmath260 to obtain two degree - one ideals @xmath261 with shorter coefficients .
one sieves over @xmath262 and @xmath263 .
the new ideals obtained in the relations will be treated as new special-@xmath49s until a relation of ideals of norm bounded by @xmath264 is found , where @xmath264 is the bound on the factor basis , so that the individual logarithms are finally known .
the sieving is done in three stages , for the three ranges of parameters . 1 .
for @xmath265 $ ] : large special-@xmath49 ; 2 . for @xmath266
$ ] with @xmath267 : medium special-@xmath49 ; 3 . for @xmath268 $ ] : small special-@xmath49 .
the proof of the complexity is not trivial at all , and since this step is allegedly cheaper than the two main phases of sieving and linear algebra , whose complexity is @xmath269 $ ] , the proofs are not always expanded .
there is a detailed proof in @xcite and @xcite for prime fields @xmath40 .
we found another detailed proof in ( * ? ? ? * b ) for large characteristic fields @xmath0 , however this was done for the polynomial selection of @xcite ( which has the same main asymptotic complexity @xmath269 $ ] ) . in ( * ? ? ?
* , pp . 144150 ) the nfs - dl algorithm is not proposed in the same order : the booting and descent steps ( step ( 5 ) of the algorithm in @xcite ) are done as a first sieving , then the relations are added to the matrix that is solved in the linear algebra phase .
what corresponds to a booting step is proved to have a complexity bounded by @xmath79 $ ] and there is a proof that the descent phase has a smaller complexity than the booting step .
there is a proof for the jlsv@xmath54 polynomial selection in @xcite and ( * ? ? ?
* a ) for a mnfs variant .
we summarize in tab . [
tab : cpx - descent - bib ] the asymptotic complexity formulas for the booting step and the descent step that we found in the available papers .
.complexity of the booting step and the descent step for computing one individual dl , in @xmath40 and @xmath0 , in medium and large characteristic .
the complexity is given by the formula @xmath60 $ ] , only the constant @xmath38 is given in the table for ease of notation .
the descent of a medium special-@xmath49 , bounded by @xmath270 $ ] with @xmath267 , is proven to be negligible compared to the large and small special-@xmath49 descents . in ( * ? ? ?
* b , c ) , the authors used a sieving technique over ideals of degree @xmath271 for large and medium special-@xmath49 descent . [ cols="^,^,^,^,^,^,^,^ " , ]
we present an example for each of the three polynomial selection methods we decided to study .
the conjugation method provides the best timings for @xmath4 at 180 dd @xcite .
we apply the gjl method to @xmath272 according to ( * ? ? ?
* fig . 3 ) .
we decided to use the jlsv@xmath54 method for @xmath107 ( * ? ? ? * fig .
4 ) .
we take the parameters of the record in @xcite : @xmath1 is a 90 decimal digit ( 300 bit ) prime number , and @xmath274 are computed with the conjugation method .
we choose a target @xmath141 from the decimal digits of @xmath275 .
@xmath276 we first compute @xmath277 then reduce @xmath278 then @xmath279 produces @xmath85 of degree 3 and coefficient size @xmath280 .
actually lll outputs four short vectors , hence we get four small candidates for @xmath85 , each of norm @xmath281 , i.e. 90 dd . to slightly improve the smoothness search time
, we can compute linear combinations of these four reduced preimages . @xmath282
the norm of the first element is @xmath283 of 90 decimal digits , as expected . for a close to optimal running - time of @xmath284 \sim 2^{40}$ ] to find a boot , the special-@xmath49 bound would be around 64 bits .
we take @xmath1 of 60 dd ( 200 bits ) so that @xmath272 has size 180 dd ( 600 bits ) as above .
we took @xmath1 a prime made of the 60 first decimal digits of @xmath286 .
we constructed @xmath287 with the gjl method described in @xcite .
@xmath288 we set @xmath289 .
the lattice to be reduced is @xmath290\ ] ] then @xmath279 computes four short vectors @xmath85 of degree 3 , of coefficient size @xmath121 , and of norm size @xmath291 .
@xmath292 the norm of the first element is @xmath293 of 117 decimal digits ( with @xmath294 dd ) . for a close to optimal running - time of @xmath295 \sim 2^{45}$ ] to find a boot , the special-@xmath49 bound would be around 77 bits .
@xmath296 we set @xmath297 .
the subfield simplification for @xmath141 gives @xmath298 we reduce the lattice defined by @xmath299\ ] ] then @xmath279 produces these four short vectors of degree 3 , coefficient size @xmath121 , and norm @xmath300 ( smaller than @xmath78 ) .
@xmath301 the norm of the first element is @xmath302 of 155 decimal digits ( with @xmath303 ) . for a close to optimal running - time of @xmath304 \sim 2^{49}$ ] to find a boot , the special-@xmath49 bound would be approximately of 92 bits .
this is very large however .
we experimented our booting step method for @xmath107 of 120 dd ( 400 bits ) . without the quadratic subfield simplification ,
the randomized target norm is bounded by @xmath305 of 135 dd ( 450 bits ) .
the largest special-@xmath49 in the boot has size @xmath306 $ ] ( 25 dd , 82 bits ) according to lemma [ th : asympt cpx ] with @xmath307 .
the running - time to find one boot would be @xmath308 \sim 2^{44}$ ] .
we apply the quadratic subfield simplification .
the norm of the randomized target is @xmath309 of 105 dd ( @xmath310 350 bits ) .
we apply theorem [ th : asympt cpx ] with @xmath311 .
the size of the largest special-@xmath49 in the boot will be approximately @xmath312 $ ] which is 21 dd ( 69 bits ) .
the running - time needed to find one boot with the special-@xmath49 of no more than 21 dd is @xmath313 \sim 2^{40}$ ] ( to be compared with the dominating part of nfs - dl of @xmath314 \sim 2^{57}$ ] ) .
we wrote a magma program to find boots , using gmp - ecm for @xmath49-smooth tests .
we first set a special-@xmath49 bound of 70 bits and obtained boots in about two cpu hours .
we then reduced the special-@xmath49 bound to a machine word size ( 64 bits ) and also found boots in around two cpu hours .
we used an intel xeon e5 - 2609 0 at 2.40ghz with 8 cores .
we have presented a method to improve the booting step of individual logarithm computation , the final phase of the nfs algorithm .
our method is very efficient for small @xmath23 , combined with the gjl or conjugation methods ; it is also usefull for the jlsv@xmath54 method , but with a slower running - time .
for the moment , the booting step remains the dominating part of the final individual discrete logarithm .
if our method is improved , then special-@xmath49 descent might become the new bottleneck in some cases .
a lot of work remains to be done on final individual logarithm computations in order to be able to compute one individual logarithm as fast as was done in the logjam @xcite attack , especially for @xmath3 .
|
the number field sieve ( nfs ) algorithm is the best known method to compute discrete logarithms ( dl ) in finite fields @xmath0 , with @xmath1 medium to large and @xmath2 small .
this algorithm comprises four steps : polynomial selection , relation collection , linear algebra and finally , individual logarithm computation .
the first step outputs two polynomials defining two number fields , and a map from the polynomial ring over the integers modulo each of these polynomials to @xmath0 . after the relation collection and linear algebra phases ,
the ( virtual ) logarithm of a subset of elements in each number field is known .
given the target element in @xmath0 , the fourth step computes a preimage in one number field .
if one can write the target preimage as a product of elements of known ( virtual ) logarithm , then one can deduce the discrete logarithm of the target .
as recently shown by the logjam attack , this final step can be critical when it can be computed very quickly .
but we realized that computing an individual dl is much slower in medium- and large - characteristic non - prime fields @xmath0 with @xmath3 , compared to prime fields and quadratic fields @xmath4 .
we optimize the first part of individual dl : the _ booting step _ , by reducing dramatically the size of the preimage norm .
its smoothness probability is higher , hence the running - time of the booting step is much improved .
our method is very efficient for small extension fields with @xmath5 and applies to any @xmath6 , in medium and large characteristic .
|
many clues about the processes of planet formation and evolution have been discovered by studying the ensemble properties of exoplanets , such as the `` brown dwarf desert '' ( halbwachs et al .
2000 , marcy & butler 2000 ) and the tendency for metal - rich stars to have more detectable planets ( santos et al .
2003 , fischer & valenti 2005 ) . however , there are also individual exoplanets whose properties bear directly on theories of planet formation and evolution
. one of the best examples is the transiting planet hd 149026b ( sato et al .
2005 ) .
compared to saturn , hd 149026b has a similar mass but its radius is 15% smaller , despite the intense irradiation from its parent star that should _ enlarge _ the radius .
sato et al .
( 2005 ) modeled hd 149026b as a dense heavy - element core surrounded by a fluid envelope of solar composition .
they found a core mass of 70 - 80 @xmath4 , which is 65 - 75% of the total mass of the planet . this is larger than the canonical core mass of 10 - 20 @xmath4 that is expected from the core - accretion theory of planet formation ( mizuno 1980 , pollack et al .
the finding of a highly metal - enriched composition was confirmed in models by fortney et al .
( 2006 ) , ikoma et al .
( 2006 ) , broeg & wuchterl ( 2007 ) , and burrows et al .
the latter authors dubbed hd 149026b a `` super - neptune '' because the inferred mass fraction of heavy elements is similar to that of an ice giant rather than a gas giant .
interestingly , the parent star has a rather high metallicity ( [ fe / h ] @xmath5 ; sato et al .
the observation of a large core in such a metal - rich system would seem to support the core - accretion theory as opposed to coreless alternatives such as gravitational instability ( boss 1997 ) .
however , the larger - than - expected core mass raises some questions .
why did the growing protoplanet not accrete gas efficiently ? or if it did , what happened to its envelope of light elements ?
many scenarios have been proposed : a collision of two massive protoplanets ( sato et al .
2005 , ikoma et al .
2006 ) , _ in situ _ formation in a low - pressure nebula ( broeg & wuchterl 2007 ) , a viscous and evaporating gas disk ( ikoma et al .
2006 ) , and a separation of gas from planetesimals at the magnetospheric `` x point '' ( sato et al .
2005 ) .
more recently , harrington et al .
( 2007 ) found that the 8 @xmath6 m brightness temperature of hd 149026b exceeds its expected blackbody temperature , even if the planet is assumed to absorb all of the incident stellar radiation . in this sense
the planet is anomalously hot .
the high temperature may result from novel atmospheric or structural properties .
most recently , torres et al . ( 2007 ) announced the discovery of a transiting planet , hat - p-3b , whose measured mass and radius indicate that it too is highly enriched in heavy elements .
in short , hd 149026b seems to be the harbinger of an entirely new kind of planet that current models of planet formation , evolution , and structure can not accommodate without interesting and possibly exotic modifications .
because of this situation , it is desirable to improve the reliability and the precision of estimates of the system parameters , and especially a key parameter that makes this planet unusual : its small radius . one can measure the planetary radius by gathering photometry during transits , modeling the light curve , and supplementing the model with external information about the stellar radius .
previously , sato et al .
( 2005 ) analyzed 3 light curves , and charbonneau et al .
( 2006 ) added 3 light curves . in this paper
we present another 5 light curves of comparable or higher quality to the previously published data , and we simultaneously model all of the data to derive the most precise planetary , stellar , and orbital parameters that are currently available .
we present our observations and data reduction procedure in 2 and the light - curve modeling procedure in 3 .
we provide the results in 4 , along with an extended discussion about the limiting error : the uncertainty in the stellar radius .
the final section summarizes the results and speculates on future prospects for improvement .
we used three of the 0.8 m automated photometric telescopes ( apts ) at fairborn observatory to measure the transits of hd 149026b that occurred on ut 2006 april 26 , 2006 may 20 , 2007 may 3 , 2007 june 18 , and 2007 june 21 .
we observed the first three transits with the t11 apt and observed the last two transits simultaneously with the t8 , t10 , and t11 apts .
all three telescopes are equipped with two temperature - stabilized emi 9124qb photomultiplier tubes for measuring photon count rates simultaneously through strmgren @xmath7 and @xmath8 filters . on a given night , each telescope automatically acquired brightness measurements of hd 149026 ( @xmath9 , @xmath10 ) and the comparison star hd 149504 ( @xmath11 , @xmath12 ) , which was previously demonstrated to be stable in brightness at the 0.002 mag level or better ( sato et al .
we also measured the dark count rate and the sky brightness in the vicinity of each star .
we used a diaphragm of diameter @xmath13 for all the integrations .
the integration time was 20 seconds on the comparison star and 30 seconds on the ( fainter ) target star .
we computed the magnitude difference for each pair of target - comparison observations . to increase the signal - to - noise ratio of each measurement , the differential magnitudes from the @xmath7 and @xmath8 pass bands were averaged , resulting in a differential magnitude for a synthetic @xmath14 pass band .
the typical cadence of the differential magnitude measurements was 1.4 min for the first 3 transits and 1.2 min for the last 2 transits .
for each raw light curve , we fitted a linear function of time to the out - of - transit data and divided the data by this function .
this was intended to correct for differential airmass effects and other systematic errors in the photometry , and also to normalize each light curve to have unit mean flux outside of the transit .
( we also tried fitting a function of airmass rather than time , but this gave slightly poorer results . )
the final light curves are shown in fig . 1 , and
the data are given in table 1 .
the standard deviation of the out - of - transit data is approximately 0.2% in all cases , which is typical for apt observations of bright stars . for additional information on the telescopes , photometers , observing procedures , data reduction techniques , and typical photometric precision ,
see henry ( 1999 ) or eaton , henry , & fekel ( 2003 ) .
the lower right panel in fig . 1 is a composite light curve of all 8 transits observed by the apts from sato et al .
( 2005 ) and this paper .
the composite light curve was created by subtracting the mid - transit time from each of the time stamps , and then averaging into 30 second bins .
it is shown here for display purposes only ; the fitting procedure described in the next section was carried out on the unbinned data .
we fitted all of our new photometric data jointly with the three @xmath14 light curves presented by sato et al .
( 2005 ) and the @xmath15 and @xmath16 light curves presented by charbonneau et al .
( 2006).-band light curve of charbonneau et al . ( 2006 ) because of its comparatively large errors and sparse time sampling . ]
we used a parameterized model based on a two - body circular orbit .
the orbit is specified by the masses of the star and planet ( @xmath17 and @xmath18 ) , the inclination with respect to the sky plane ( @xmath19 ) , the orbital period ( @xmath20 ) and a particular midtransit time ( @xmath21 ) .
the star and planet are taken to be spheres with radii @xmath22 and @xmath23 respectively , and when their sky - projected centers are within @xmath24 of one another we use the mandel & agol ( 2002 ) formulas to compute the flux decrement due to the partial blockage of the limb - darkened stellar surface .
this is the same code that has been developed for the transit light curve project ( holman et al .
2006 , winn et al .
2007 ) . for hd 149026
, we assumed the limb darkening law to be linear , with a coefficient given by claret ( 2000 ) for a star of the appropriate temperature and surface gravity .
for the @xmath14 data we used a coefficient of 0.712 , which is the mean of the tabulated @xmath7 and @xmath8 coefficients .
not all of the parameters listed above can be determined from transit photometry alone .
one set of parameters that can be determined from an individual light curve is @xmath21 , @xmath25 , @xmath26 , and @xmath19 , where @xmath27 is the semimajor axis .
our approach was to fix @xmath17 , @xmath18 , and @xmath20 at previously determined values ( thereby fixing @xmath27 through kepler s third law ) , and then fit for @xmath23 , @xmath22 , @xmath19 , and @xmath21 .
the results for @xmath22 and @xmath23 are specific to the choice of @xmath17 , but they scale as @xmath28 because @xmath29 when the uncertainty in @xmath20 is negligible , as it is here .
we assumed @xmath30 @xmath31 , following sato et al .
( 2005 ) , a choice that was subsequently corroborated by our analysis of the observable stellar properties and the results for @xmath26 ( see 4.2 ) .
the fitting statistic was @xmath32 ^ 2 , \label{eq : chi2}\ ] ] where @xmath33 is the number of flux measurements , @xmath34(obs ) is the flux observed at time @xmath35 , @xmath36 controls the weights of the data points , and @xmath34(calc ) is the calculated flux .
experience has shown that the data weights @xmath36 should account not only for the single - measurement precision but also the time - correlated ( `` red '' ) noise that afflicts most time - series photometry ( see , e.g. , gillon et al .
the most important timescale in a transit light curve is the @xmath3710 min duration of the ingress and egress , since the resolution of ingress and egress is what permits the determination of @xmath26 and @xmath19 in addition to @xmath25 . to assess the noise on this timescale , we first calculated the standard deviation of the unbinned out - of - transit data ( @xmath38 ) for each light curve .
then we averaged the out - of - transit data into 10 min bins consisting of @xmath39 data points , where @xmath39 depended on the observing cadence , and recalculated the standard deviation ( @xmath40 ) . in the absence of red noise
, one would expect @xmath41 , but in practice @xmath40 was larger than @xmath42 by some factor @xmath43 .
therefore , we set the data weights equal to @xmath44 . the results for @xmath43 ranged from 1.05 to 1.27 .
we used a markov chain monte carlo algorithm to determine the best - fitting parameter values and confidence intervals .
this algorithm delivers an estimate of the _ a posteriori _ joint probability distribution for all of the parameters ( see holman et al . 2006 or winn et al .
2007 for more details ) . for each parameter
, we took the mode of the distribution after marginalizing over all other parameters to be the `` best value . ''
we defined the 68% confidence limits @xmath45 and @xmath46 as the values between which the integrated probability is 68% , and for which the two integrals from @xmath47 and @xmath48 were equal . at first , the preceding computations were performed using free parameters for both the orbital period and a single midtransit time . thus the transits were required to be spaced by integral multiples of a fixed period .
however , we also wanted to measure the individual transit times , in order to search for variations that might be indicative of additional bodies in the planetary system ( agol et al .
2005 , holman & murray 2005 ) . to do this , we fixed @xmath23 , @xmath22 , and @xmath19 at the best values determined in the first step , and then performed a three - parameter fit of each individual light curve .
the parameters were @xmath21 along with the zero point and slope of the linear function that was used to correct the out - of - transit data .
( fixing the values of @xmath23 , @xmath22 , and @xmath19 is justified because the errors in those parameters are not correlated with the error in the transit time . )
we did this not only for the 5 new light curves , but also for the 5 previously published light curves , to provide consistency in the treatment of errors .
we then used these transit - time measurements to refine the estimates of @xmath20 and @xmath21 ( see 4.3 ) . for our final results for the photometric parameters @xmath25 , @xmath26 , and @xmath19
, we reran the mcmc algorithm on the entire data set , using fixed values of @xmath20 and @xmath21 from our refined ephemeris .
the results of the light - curve analysis are given in table 2 and discussed in
the transit times of the individual light curves are given in table 3 , and in 4.2 we use those times to derive a new transit ephemeris .
we also use the new ephemeris along with a previously - measured midpoint of a secondary eclipse to place an upper bound on one aspect of the orbital eccentricity . in order to derive the actual planetary radius ( as opposed to the planet - to - star radius ratio ) , one must supplement the light - curve analysis with external information about the stellar radius or mass . in 4.3 we investigate four different methods to estimate the stellar radius , the results of which are given in table 4 . the planetary parameters derived from the light curves and the stellar radius estimates are given in table 5 .
the _ a posteriori _ probability distributions for @xmath25 , @xmath26 , and @xmath19 are shown in fig . 2 .
the most well - constrained of these three basic light curve parameters is the radius ratio , @xmath0 , with a precision of approximately 2.3% .
the radius ratio is determined largely by the observed transit depth , which is the smallest among all of the 20 transiting planets known to date . while there are smaller planets , such as the neptune - sized gj 436b ( gillon et al .
2007 ) , they all orbit smaller stars , making their radius ratios and transit depths larger than that of hd 149026 . less well - constrained are @xmath26 and @xmath19 , the parameters that depend on the observed durations of the ingress , egress , and total phases of the transit . table 2 gives the results for those parameters as well as the impact parameter and the durations , which can be derived in terms of @xmath25 , @xmath26 , and @xmath19 .
the data are consistent with impact parameters ranging from @xmath49 to @xmath50 . as with any eclipsing binary system
, the data can not distinguish between positive and negative impact parameters ; the probability distributions are perfectly symmetric about @xmath51 and @xmath52 . in this case
the peak probability occurs at the central values @xmath51 , @xmath52 .
the quantity @xmath26 has a highly asymmetric error bar .
the observed duration of the entire transit event enforces the upper limit on @xmath26 , while the ratio of the ingress ( or egress ) duration to the total duration enforces the lower limit on @xmath26 .
our findings are consistent with the two previous light - curve analyses , by sato et al .
( 2005 ) and charbonneau et al .
( 2006 ) , although our analysis method is different in several ways besides the use of an expanded dataset .
first , we have attempted to account for time - correlated noise in the photometry , which was neglected in the previous analyses .
second , unlike the previous authors , we have not incorporated any _ a priori _ constraints on the stellar properties into our fitting statistic .
we made this choice in order to clarify what information is derived from the light curves themselves ; for example , the previous works did not call attention to the results for @xmath25 even though that parameter is more precisely known than either @xmath23 or @xmath22 .
in addition , our analysis method provides an estimate of @xmath26 that is independent of any assumptions about the parent star , except for the very weak dependence on the chosen limb darkening parameter .
this is useful because @xmath26 can be used to determine the stellar mean density ( seager & mallen - ornelas 2003 , sozzetti et al .
2007 , holman et al . 2007 ) : @xmath53 the last term in this expression may be neglected in this case because @xmath54 and @xmath55 .
our independent estimate of @xmath56 is useful in characterizing the parent star , as described below . to determine the quantity of intrinsic interest , @xmath23
, we can multiply our result for @xmath25 by a value of @xmath22 obtained by other means .
we have investigated four different methods for determining @xmath22 : _ stefan - boltzmann law._the bolometric luminosity , effective temperature , and photospheric radius of hd 149026 are related via @xmath57 .
we use the _ hipparcos _ parallax and apparent magnitude ( @xmath58 mas , @xmath59 ; perryman et al .
1997 ) to compute the absolute @xmath60 magnitude , apply a bolometric correction of @xmath61 ( flower et al .
1996 ) , and use the spectroscopically determined @xmath62 ( sato et al . 2005 ) .
the result is @xmath63 @xmath64 .
this is essentially identical to the value quoted by sato et al .
( 2005 ) who used the same method
. _ spectral energy distribution fit._masana et al .
( 2006 ) presented an alternative means of estimating the effective temperature and bolometric correction , using @xmath65 photometry .
they also provided radius estimates for many nearby stars based on this technique .
using the _ hipparcos _ parallax and @xmath60 magnitude along with 2mass near - infrared photometry , their result for hd 149026 is @xmath66 @xmath64 . _ yonsei - yale isochrone
fit._stellar evolutionary models may be used to estimate the mass , radius , and age of a star with a given effective temperature , luminosity ( or gravity ) , and metallicity .
we used the yonsei - yale models ( yi et al .
2001 , demarque et al . 2004 ) because they are conveniently provided with tools for interpolating isochrones in both age and metallicity .
for the effective temperature , we used @xmath67 k , a weighted mean of the results of sato et al .
( 2005 ) and masana et al .
we used the photometric result for @xmath26 as our our proxy for surface gravity , and we explored the range @xmath5 with an internal uncertainty of @xmath68 . to be conservative , we adopted a somewhat larger uncertainty of @xmath69 , recognizing that different methods for determining the metallicity often produce systematic differences of this size .
] of metallicities [ fe / h ] @xmath70 . for each metallicity
we considered a range of ages from 0.1 to 14 gyr , in steps of 0.1 gyr .
we interpolated the isochrones using a fine mass grid and compared the points with the measured values of @xmath71 and @xmath72 .
we computed @xmath73 at each point based on the modeled and observed values of @xmath71 , @xmath26 , and metallicity .
then we weighted the points by @xmath74 and applied an additional weighting to take into account the density of stars on each isochrone , assuming a salpeter initial mass function .
the `` best - fitting '' stellar properties were taken to be the weighted mean of the properties of all the points . for more details and other applications of this analysis ,
see torres , winn , & holman ( in preparation ) . for hd 149026 ,
the results are @xmath75 , @xmath76 , and @xmath77 .
similar results were obtained when the spectroscopically determined value of @xmath78 was used instead of @xmath26 . the theoretical isochrones and the observational constraints
are shown in fig .
kepler s law with stellar mass prior._as mentioned earlier , the quantity @xmath26 that is determined from the transit photometry can be used to find @xmath56 ( eq . 2 ) .
with an _ a priori _ estimate of @xmath17 , one may use @xmath56 to determine @xmath22 .
taking @xmath79 @xmath31 based on the isochrone fit described above , we find @xmath80 @xmath64 .
all of the results for the stellar radius are summarized in table 4 .
they are all consistent with one another at the 1@xmath81 level , with a weighted mean of @xmath82 @xmath64 .
however , it must be emphasized that while the methods are different , they are not wholly independent .
the first two methods both rely on the _ hipparcos _ parallax , which is the largest source of error in both cases .
the latter two methods both rely on the yonsei - yale stellar evolutionary models . for this reason we can not say confidently that the uncertainty in @xmath22 is any smaller than the uncertainty in each of the individual measurements , although the mutual agreement is certainly reassuring . in what follows
we adopt the consensus value @xmath83 @xmath64 , the same value used in the previous light curve analyses . assuming a gaussian error distribution for @xmath22 , and the error distribution for @xmath25 obtained from our light curve analysis
, we find the planetary radius to be @xmath84 @xmath2 .
this can be compared to the previously published results of @xmath85 @xmath2 ( sato et al .
2005 ) and @xmath86 @xmath87 ( charbonneau et al .
2006 ) , keeping in mind our different method of analysis and treatment of observational errors .
the results are all in agreement .
indeed the differences are smaller than one would expect from gaussian statistics , given the quoted error bars , though we note that 5 of the 10 light curves we fitted were taken from those previous works .
the precision in @xmath23 is not improved because the limiting error is the uncertainty in @xmath22 , which is unchanged . for planning future observations of this system
it is important to be able to predict transit times as precisely as possible .
we used all of the transit times given in table 3 to calculate a photometric ephemeris for this system , @xmath88 where @xmath21 is the transit midpoint , @xmath89 is the integral transit epoch , and @xmath20 is the orbital period . the linear fit had @xmath90 and @xmath91 , suggesting that the errors quoted in table 3 have been somewhat overestimated .
the results are : @xmath92 } \\
p & = & 2.8758882 \pm 0.0000061~{\rm days}.\end{aligned}\ ] ] our value for the orbital period is in agreement with the previously published values and is about 25 times more precise .
figure 4 is the o@xmath93c ( observed minus calculated ) diagram for the transit times . for a circular orbit , successive transits and secondary
eclipses should be spaced by exactly half an orbital period .
recently , harrington et al .
( 2007 ) observed a secondary eclipse of hd 149026 with the _
spitzer space telescope _ , allowing the assumption of a circular orbit to be checked .
in the presence of a small but nonzero orbital eccentricity , the time difference between the midpoint of secondary eclipse , @xmath94 , and the time of transit , @xmath95 , is @xmath96 where @xmath97 is the argument of pericenter ( kallrath & milone 1999 , p. 62 ) .
harrington et al .
( 2007 ) measured the midpoint of a secondary eclipse to be hjd @xmath98 , represented by the open circle in figure 4 .
the timing offset of equation ( 6 ) is @xmath99 minutes , corresponding to @xmath100 .
the orbit does indeed appear to be nearly circular , as one would expect from the dissipative effects of stellar and planetary tidal interactions .
we have presented 5 new transit light curves of the exoplanet hd 149026b and analyzed them along with 5 previously published light curves .
the joint analysis has resulted in much more precise determinations of the orbital period and transit ephemerides , and also in a more precise value of the planet - to - star radius ratio . in some cases , this ratio is of primary interest , such as inferring the brightness temperature of the planet from the depth of a secondary eclipse ( harrington et al .
2007 ) , or testing for any wavelength dependence in the radius ratio as a means of identifying planetary atmospheric features ( see , e.g. , charbonneau et al . 2002 ) .
however , when it comes to understanding the interior structure of the planet , the quantity of primary interest is @xmath23 itself , and here we can offer no significant improvement .
the limiting error is the 7% uncertainty in the stellar radius .
this error was not reduced by acquiring more light curves , although we did find agreement between the results of 4 different ( and intertwined ) methods for estimating the stellar radius using all of the available data .
thus , we leave unchanged the interpretation of this planet as a being unexpectedly small for its mass , and likely to be highly enriched in heavy elements ( sato et al . 2005 , fortney et al .
2006 , ikoma et al .
2006 , burrows et al .
2006 ) .
further improvement will depend upon progress in measuring the stellar radius .
baines et al .
( 2007 ) recently used optical interferometry to measure the angular diameter of the planet - hosting star hd 189733 , and combined it with the _ hipparcos _ parallax to measure the stellar radius . for hd 149026 ,
similar observations are not likely to result in a more precise value of the stellar radius , at least not in the near future .
this is not only because of the 6% uncertainty in the parallax , but also because the expected angular diameter is only @xmath101180 @xmath6as , which is only 78 times larger than the measurement error that was achieved for hd 189733 .
supposing the parallax were known with 10 @xmath6as precision ( as one might hope from a space - based interferometric mission ) , the error in the stefan - boltzmann method for determining @xmath22 would be reduced to 2.7% .
the limiting errors in that case would arise from the effective temperature and bolometric correction . in the nearer term ,
a possible path forward is the continued acquistion of high - quality transit photometry , in order to improve upon our measurement of @xmath26 and thereby establish the stellar mean density with greater precision . at fixed mean density
, @xmath22 varies as @xmath28 , and our application of the yonsei - yale models to hd 149026 suggests that the stellar mass has already been pinned down to within 4.6% . if @xmath26 were known exactly , the fractional error in the stellar radius would be approximately 1.5% ( i.e. , one - third as large as the fractional error in the stellar mass ) . in effect , transit photometry measures @xmath102 , and the stellar models generally constrain a different combination of @xmath17 and @xmath22 ( see , e.g. , cody & sasselov 2002 ) .
we encourage observers to be persistent in gathering additional seasons of ground - based photometry and look forward to the results of space - based photometry for this system .
we are grateful to the anonymous referee for a thorough and helpful review of the manuscript .
g.w.h . acknowledges support from nsf grant hrd-9706268 and nasa grant nnx06ac14 g .
gt acknowledges partial support for this work from nasa grant nng04lg89 g .
@xmath103 & @xmath104 & @xmath105 & @xmath106 + @xmath26 & @xmath107 & @xmath108 & @xmath109 + @xmath19 [ deg ] & @xmath110 & @xmath111 & @xmath112 + @xmath113 & @xmath114 & @xmath115 & @xmath49 + total transit duration & @xmath116 & @xmath117 & @xmath118 + ingress or egress duration & @xmath119 & @xmath120 & @xmath121 t11 0.8 m apt & @xmath122 & @xmath123 & @xmath124 + t11 0.8 m apt & @xmath125 & @xmath126 & @xmath127 + flwo 1.2 m & @xmath125 & @xmath128 & @xmath124 + t8 , t10 , t11 0.8 m apts & @xmath129 & @xmath130 & @xmath131 + flwo 1.2 m & @xmath132 & @xmath133 & @xmath134 + flwo 1.2 m & @xmath132 & @xmath135 & @xmath136 + t11 0.8 m apt & @xmath137 & @xmath138 & @xmath139 + t11 0.8 m apt & @xmath140 & @xmath141 & @xmath142 + t11 0.8 m apt & @xmath143 & @xmath144 & @xmath145 + t8 , t10 , t11 0.8 m apts & @xmath146 & @xmath147 & @xmath127 + t8 , t10 , t11 0.8 m apts & @xmath148 & @xmath149 & @xmath150 @xmath18 [ @xmath155 & @xmath156 & spectroscopic orbit + @xmath23 [ @xmath2 ] & @xmath158 & @xmath25 from light curves and @xmath159 + @xmath160 [ cgs ] & @xmath161 & light curve and spectroscopic orbit + @xmath162 [ g @xmath163 & @xmath164 & @xmath18 , @xmath23 given above + semimajor axis , @xmath27 [ au ] & @xmath165 & kepler s law + @xmath166 & @xmath167 & timing of secondary eclipse
|
we present new photometry of hd 149026 spanning five transits of its `` super - neptune '' planet . in combination with previous data ,
we improve upon the determination of the planet - to - star radius ratio : @xmath0 .
we find the planetary radius to be @xmath1 @xmath2 , in accordance with previous theoretical models invoking a high metal abundance for the planet .
the limiting error is the uncertainty in the stellar radius .
although we find agreement among four different ways of estimating the stellar radius , the uncertainty remains at 7% .
we also present a refined transit ephemeris and a constraint on the orbital eccentricity and argument of pericenter , @xmath3 , based on the measured interval between primary and secondary transits .
|
observations of high - redshift quasars reveal that supermassive black holes ( smbhs ) with mass of @xmath8 already exist as early as redshifts @xmath9 @xcite . gas accretion and mergers of the remnant black holes ( bhs ) formed by the collapse of first - generation stars ( @xmath10 )
have been considered for producing such smbhs ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , various forms of radiative feedback can prevent efficient bh growth , making it difficult to reach @xmath8 within the age of the high - redshift universe @xcite .
one possible alternative solution is the rapid formation of supermassive stars ( smss ; @xmath0 ) and their subsequent collapse directly to massive bhs in the first galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the massive seed bhs , formed by direct collapse , shorten the total smbh growth time sufficiently , even in the presence of subsequent radiative feedback ( e.g. @xcite ) .
smss can form out of primordial gas in massive ` atomic cooling ' haloes with virial temperatures @xmath1k , if h@xmath11 formation and line cooling are prohibited through the pre - stellar collapse .
possible mechanisms to suppress h@xmath11 formation are photo dissociation by far - ultraviolet ( fuv ) radiation @xcite and collisional dissociation @xcite . in the absence of @xmath12 ,
the primordial gas remains warm ( @xmath13k ) and may collapse monolithically without strong fragmentation @xcite .
the resulting central protostar can grow via accretion at a high rate of @xmath14 .
if the accretion rate maintains such a high value , the embryonic protostar evolves to a sms with mass of @xmath0 .
the sms collapses as a whole to a single bh either directly @xcite , or via the intermediate stage of a close binary bh @xcite , through a general relativistic instability ( e.g. , * ? ? ?
* ; * ? ? ?
the massive remnant bh is then a promising seed that can grow into one of the observed smbhs at @xmath15 . in the scenario
above , a major unresolved question is whether rapid accretion will continue unabated through the protostellar evolution .
recent high - resolution simulations by @xcite suggest that the compact nuclear accretion disc , surrounding a central embryonic protostar , is gravitationally unstable .
thus , self - gravitating clumps are expected to form during the early stages of the accretion phase .
such efficient fragmentation could prevent the rapid growth of the central protostar , and preclude eventual sms formation .
this is analogous to the fragmentation of gravitationally unstable discs in the cores of lower mass ` minihaloes ' , which had been suggested to reduce the characteristic masses of population iii ( pop iii ) stars @xcite .
moreover , if the gas is slightly polluted by heavy elements ( @xmath16 ) , dust cooling can decrease the temperature and induce the fragmentation @xcite , which would be a further obstacle to sms formation .
because of this , the existence of the pristine , metal - free gas has often been considered as a necessary condition of forming a sms in semi - analytical models ( e.g. , * ? ? ?
* ; * ? ? ?
we note that even very efficient fragmentation may not prevent accretion on to a central point source ( e.g. , * ? ? ? * ) and the rapid inward migration of the fragments , on a time - scale comparable to the orbital period , may help its growth further ( e.g. , * ? ? ? * ) . in our context , however , even a slow down of the accretion rate could bring it below the critical value required for sms formation . in this paper , motivated by the above ,
we discuss the expected properties of a compact , marginally unstable nuclear protogalactic disc , and the fate of the clumps formed in the disc by gravitational instability . using analytical models
, we argue that despite fragmentation , the growth of a sms remains the most likely outcome .
the reason for this conclusion is the rapid inward migration of the fragments , and their merger with the central protostar .
the rest of this paper is organized as follows . in [ sec:2 ] , we describe the basic model of the fragmenting disc and of the clumps formed in the disc . in [ sec : fate ] , we discuss the fate of the clumps , considering various important processes : migration , accretion , contraction , and star - formation by reaching the zero - age main sequence ( zams ) .
we also consider the possibility that radiative feedback from massive stars , formed from the clumps , may halt gas accretion on to the whole system . in [ sec : metal ] , we discuss whether metal pollution and associated dust cooling could prohibit sms formation .
finally , we discuss our results and summarize our conclusions in [ sec : conc ] .
we consider the properties of a disc formed after the collapse of primordial gas inside an atomic cooling halo , when molecular hydrogen formation is suppressed .
since the parent gas cloud is hot ( @xmath17 k ) , the accretion rate on to the disc is high : @xmath18 where @xmath19 is the sound speed and @xmath20 is the gravitational constant @xcite .
the accretion disc around the central protostar can become unstable against self - gravity . to understand the stability of the disc ,
toomre s parameter @xcite defined by @xmath21 is often useful , where @xmath22 is the orbital frequency and @xmath23 is the surface density of the disc . in the marginal case of @xmath24 , strong spiral arms
are formed in the disc .
the spiral arms redistribute angular momentum and heat the disc by forming shocks ( e.g. , * ? ? ?
* ; * ? ? ?
the resulting disc is self - regulated to the marginal state .
thus , we here assume @xmath25 .
since the gas temperature in the atomic cooling halo is kept near @xmath26 k , the external accretion rate on to the disc , from larger radii , is also nearly constant .
we estimate the surface density of the self - regulated disc by assuming that it is in steady state , and that it has an effective viscosity @xmath27 ( arising from gravitational torques ) , @xmath28 the scale - height of the disc is estimated from vertical hydrostatic balance , @xmath29 we define the particle number density as @xmath30 , where @xmath31 is the proton mass .
ccc to determine the thermal state of the disc we assume that it is in equilibrium , i.e. we balance heating and radiative cooling ( @xmath32 ) .
in particular , we consider viscous heating by turbulence and spiral shocks , @xmath33 and radiative cooling @xmath34 where @xmath35 is the cooling rate in units of erg s@xmath36 @xmath37 .
for the cooling process , we consider h@xmath38 free - bound emission , which is the dominant channel in the thermal evolution at @xmath39 for a warm primordial gas in an atomic cooling halo .
the form of the cooling rate is given in
[ sec : chem ] .
finally , we adopt the standard @xmath40-prescription as a model for the viscosity by gravitational torques : @xmath41 where @xmath40 is the viscous parameter . from hydrodynamical simulations of a self - gravitating disc , the disc
is found to be susceptible to fragmentation if the viscous parameter exceeds a critical threshold value , @xmath42 ( e.g. , * ? ? ?
@xcite have studied fragmentation conditions for various specific heat ratios , and obtained @xmath43 . however , the critical value depends on both initial conditions and on numerical resolution @xcite .
recently , @xcite have considered mass loading from an infalling envelope , realistic radiative cooling , and radiative trapping of energy inside clumps , and suggested that the critical value for fragmentation is around @xmath44 .
moreover , numerical simulations of fragmentation of discs with relatively high accretion rates ( @xmath45 ) , in the context of present - day massive star formation ( e.g. , * ? ? ?
* ) and pop iii star formation @xcite have shown that the effective viscous parameter arising from gravitational torques is @xmath46 .
based on these results , we here set @xmath47 as our fiducial value , and analyze disc fragmentation using the condition of @xmath48 following previous analytical works ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . in dense ( @xmath39 ) and warm ( @xmath49 k ) primordial gas ,
the dominant cooling processes is the free - bound emission of h@xmath38 ions ( h + e@xmath38 @xmath50 h@xmath38 + @xmath51 ; @xcite ) . the cooling rate , in units of erg s@xmath36 @xmath37 , is given by @xmath52 where @xmath53 is the free - electron fraction and @xmath54 ( with @xmath55 in units of kelvin ) .
in such a warm gas , prior to protostar formation , the electron fraction is determined by the balance between recombination ( h@xmath56 + e@xmath38 @xmath50 h + @xmath51 ) and collisional ionization ( 2h @xmath50 h@xmath57 + e@xmath38 ) .
the evolution of the electron fraction follows the equation @xmath58 at the densities of interest here , the electron fraction is in equilibrium , and given simply by @xmath59 from appendix [ sec : appa ] , we obtain the specific value of @xmath60 we next obtain the properties of the marginally fragmenting disc , using the energy conservation equation ( @xmath32 ) . using eq .
( [ eq : q ] ) , the particle density can be written by @xmath61 then , we can solve the energy conservation equation with respect to @xmath22 : @xmath62 we present the profiles of the temperature , surface density , and viscous parameter as a function of @xmath22 for the case of @xmath63 in figure [ fig : profile ] .
as described above , we assume that fragmentation occurs efficiently at the radii where @xmath64 . from figure
[ fig : profile](c ) , we find that fragmentations occur in the central regions where the orbital period is shorter than @xmath65 yr . within the fragmenting region ,
the surface density is approximately given by @xmath66 , where @xmath67 g @xmath68 and @xmath69 s@xmath36 , respectively . during the earliest stages ,
i.e. prior to the formation of any central protostar embryo , or immediately following it , the disc mass dominates the central protostellar mass .
we can then obtain the radius within which the disc fragments effectively , @xmath70 moreover , during this stage , the profiles of the surface density , disc mass , and number density can be written as functions of the radial distance @xmath71 from the central protostar : @xmath72 @xmath73 and @xmath74 respectively .
the typical mass of the clumps formed at @xmath75 is estimated as @xmath76 these disc profiles and the clump mass are in good agreement with the highest resolution results in the numerical simulations by @xcite , lending credence to our simplified ` toy ' model . in the above
, we have assumed that the mass of the central protostar is negligible .
however , the central protostar grows via rapid accretion and having a central point source can then modify the disc structure .
after @xmath77 yr , the gas within @xmath78 accretes on to the central protostar , and the protostellar mass exceeds the disc mass within @xmath78 .
the fragmentation radius ( defined by @xmath79 ) remains roughly constant for the first @xmath80 yr , after which it begins to move outward slowly , according to @xmath81 where @xmath82 is the mass of the protostellar embryo at the center , and we have assumed that the central accretion rate remains constant over time . in the regions where @xmath83 , the orbital frequency is proportional to @xmath84 and thus the surface density and mass of the disc are given by @xmath85 and @xmath86 respectively . the radius where @xmath87 is given by @xmath88 the condition @xmath89 is satisfied when the protostellar mass exceeds @xmath90 , which corresponds to @xmath91 .
for convenience in the following section ( [ sec : fate ] ) , we define the transition epoch as @xmath92 yr neglecting the difference of the factor of 3 because the ratio of @xmath93 does not change significantly in the range @xmath94 yr . therefore , we use eq .
( [ eq : sig0 ] ) and ( [ eq : md0 ] ) for @xmath95 and eqs .
( [ eq : sig1 ] ) and ( [ eq : md1 ] ) for @xmath96 ( hereafter the ` late phase ' ) for the radial profile of the disc , respectively .
in the fragmenting disc , the evolution of the clumps is determined by several physical processes , including migration , accretion , and evolution to zams stars . in this section ,
we consider these processes , and discuss the fate of the clumps formed at the fragmentation radius .
the clumps formed at @xmath78 interact with the disc through gravitational torques , which results in their orbital evolution , i.e. ` migration ' .
we first discuss the time - scales for limiting cases of inward migration applicable to planets embedded in smooth discs ; this serves as a useful reference .
we then discuss the complications and uncertainties for protogalactic discs , arising from the fact that the disc is clumpy , and the nominal migration rate is very rapid . limiting cases of idealized migration
are well - known in the theory of disc - planet interactions .
one is the case when the planet mass is so small that the perturbation induced in the disc is in the linear regime .
the density waves excited in the disc by the planet exert torques on the planet through resonances , and thus promote the migration ( so - called type i migration ; e.g. @xcite ) .
the characteristic time - scale of type i migration is @xmath97 where @xmath98 , @xmath99 is the power of the surface density profile @xmath100 , @xmath101 , and @xmath102 @xcite .
a second case is when the planet mass is massive enough to clear a gap in the disc ( but still less massive than the local disc ) .
then , the planet migrates inward on approximately the discs s viscous time - scale ( so - called type ii migration ; e.g. @xcite ) : @xmath103 the theoretical basis for these time - scales is derived from idealized setups , assuming point - like planets and a smooth disc profile .
whether these results are applicable for an extended clump , in a gravitationally unstable and lumpy disc , as we discuss in this paper is poorly known .
nevertheless , recent numerical simulations have provided some useful information .
in particular , low - mass planets embedded in a turbulent disc have been found to migrate inwards , despite the stochastic kicks due to the turbulent density fluctuations , essentially on the time - scale of type i migration @xcite . if the clump grows via accretion and opens a gap during its migration in the gravitationally unstable disc , a transition between type i and type ii migration occurs and the migration speed slows down @xcite . from the disc properties discussed in
[ sec : disc ] , we expect that the characteristic migration time shifts from that of type ii to that of type i , because the mass ratio @xmath104 decreases as the central protostar ( and the mass of the corresponding steady - state disc ) grows .
we therefore estimate the transition epoch and the migration time for the late phase ( @xmath105 ) using eq .
( [ eq : typei ] ) .
the gap - opening condition has been investigated by many authors ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , but again , remains poorly known for gravitationally unstable and clumpy discs . to give a conservative estimate ,
we here do not discuss the gap - opening condition , but compare the time - scales of type i and ii migration .
the ratio of the two time - scales at @xmath78 ( eq . [ eq : rf_late ] ) is independent of the central mass , and is given by @xmath106 where we have used the condition @xmath107 . since the clumps grow via accretion at a typical rate of at least @xmath108 ( see [ sec : kh ] ) , the clump migrates towards the central protostar on a time - scale comparable or shorter than that in the type ii case , independent of the presence or absence of a gap . in the following , we therefore adopt @xmath109 as a conservative choice for the migration time - scale .
we first estimate the migration time for the case that the self - gravity of the disc within @xmath78 dominates the protostellar gravity ( i.e. @xmath95 ) . at this stage ,
the mass of the central protostar is @xmath110 . from eqs .
( [ eq : md0 ] ) and ( [ eq : m_c ] ) , the clump mass is initially smaller than the local disc mass ( @xmath111 ) . in the disc - dominated case , the orbital decay time of the clump ( @xmath112 )
is approximately given by the viscous time ( eq . [ eq : typeii ] ; see also @xcite who find a factor of few faster migration in the case they studied ) . since @xmath113 at @xmath78 , @xmath114 yr .
thus , we find that the clump orbit decays slightly faster than the orbital period .
this follows essentially from our assumption of @xmath115 , and leads to a complication : eq .
( [ eq : typeii ] ) for the type ii migration rate is strictly valid only for @xmath116 because the torques exerted on the clump are assumed to be averaged over an orbit ( at a fixed radius ) . despite this complication
, we can still safely conclude @xmath117 .
this is because assuming a slower migration ( @xmath118 ) would justify the use of orbit - averaged type ii torques , which would then yield the contradiction @xmath119 .
( @xmath120 ; @xmath95 ) , @xmath121 and @xmath122 ( @xmath83 ; @xmath123 ) , respectively . in each case , the initial position of the clump is set to the fragmentation radius @xmath124 . in the first ( @xmath125 ) case
, we show the slow - down of the migration expected if the clump grows by accretion at a rate of @xmath126 with @xmath127 ( solid ) , @xmath128 ( dashed ) , and @xmath129 ( dotted).,width=309,height=219 ] when the clump migrates to @xmath130 , its mass becomes comparable to the local disc mass ( neglecting for now any growth of either the disc or the clump during the migration ) . within @xmath131 ,
the migration speed slows down because the disc outside the clump can no longer absorb the orbital angular momentum of the clump @xcite . in the clump - dominated case ,
the migration time is somewhat modified as @xmath132 where @xmath133 and @xmath134 in figure [ fig : mig ] , we show the evolution of the orbital radius of the clump ( black lines : @xmath125 ) .
the horizontal axis is the time from the onset of the orbital decay at @xmath78 . to see the effect of the slow down
, we assume that the clump grows at the accretion rate of @xmath126 with @xmath127 ( solid curve ) , @xmath128 ( dashed ) , and @xmath129 ( dotted ) , respectively . in each case
, the clump s orbit decays within @xmath65 yr , even if we consider the slow down of the migration .
next , we consider the case that the protostellar gravity exceeds the self - gravity of the disc within @xmath78 ( i.e. @xmath96 ) . in figure
[ fig : mig ] , we show the corresponding orbital evolution for @xmath121 ( blue middle curve ) and @xmath122 ( red right most curve ) as examples of this stage . in both the cases
, we do not consider the clump growth because the initial clump mass is much smaller than the disc mass within @xmath78 and thus the slow - down effect works after the orbit has decayed by more than two orders of magnitude .
for @xmath135 , the fragmentation radius moves out to @xmath136 pc . in this case , the decay time becomes longer than the orbital period , and the clump could evolve into a normal massive star , rather than an sms .
however , this scenario is realized only by assuming that a sms with @xmath137 has already grown at the center of the disc . in the fragmenting disc ,
many clumps are formed within @xmath78 and their orbits decay through interaction with the accretion disc .
since the typical density of the clumps is @xmath138 ( eq . [ eq : numdens ] ) and since they are optically thin to the h@xmath38 free - bound emission , tiny protostars are formed in the clumps within their free - fall time @xmath139 yr , which is an order of magnitude shorter than the orbital decay time . in this section , we simply estimate the accretion rate on to the clumps ( which we assume quickly evolve to protostars ) , and then discuss the possibility of forming zams stars from the clumps . the accretion rate on to a point - like clump in a keplerian disc , where we assume a rotationally supported disc @xcite , is estimated as @xmath140 where @xmath141 is the hill radius defined by @xmath142 and @xmath143 ( e.g. , * ? ? ?
we have also assumed that the clump accretes gas orbiting in the nearby disc , within an impact parameter @xmath144 . at the fragmentation radius ,
the accretion rate is given by @xmath145 which is similar to the critical rate ( @xmath146 ) at which the evolution of a protostar changes qualitatively @xcite . below the critical rate
, the protostar grows to a usual zams star . above the critical rate ,
the protostar evolves instead to a structure with a bloated envelope resembling a giant star @xcite .
the accretion rate given by eq .
( [ eq : clump_acc ] ) is only a factor of @xmath147 above the critical value .
we therefore consider the case in which some of the clumps grow at a sub - critical rate and form massive zams stars .
the protostar embedded in the clump begins to undergo kelvin - helmholtz ( kh ) contraction , loosing energy by radiative diffusion .
after the contraction , the star reaches a zams star when the central temperature increases to @xmath148k and hydrogen burning begins @xcite . below the critical rate ,
the time - scale of the kh contraction is estimated as @xmath149 thus , it is longer than both the migration time for @xmath150 and the orbital period at the fragmentation radius .
therefore , most clumps formed promptly in the early ( disc - dominated ) phase are expected to migrate and merge with the central protostar before reaching the massive zams stars ( @xmath150 ) . on the other hand , clumps formed in the late ( central protostar - dominated ) phase ( @xmath151 )
could survive and evolve to zams stars . as we have seen in
[ sec : kh ] , the decay time of the clump orbit becomes longer than the kh time during the late stage for @xmath152 . in this case
, the clump can contract by loosing energy and reach the zams star before migrating towards the central protostar .
this , however , requires a central sms to be already present . on the other hand ,
even during the early stage for @xmath153 , some clumps could survive for their kh times , because clumps interact with each other , as well as with the disc , within the fragmentation radius . in a clumpy disc
, these gravitational interactions will make the orbital decay time longer or shorter stochastically , and can also cause clumps to be temporarily ejected from the disc @xcite .
some of these clumps can evolve to zams stars and emit strong uv radiation , which could prevent the accretion on to the protostar and on to the disc as a while .
we briefly estimate the number of the clumps which exist in the disc at a moment . the corresponding range where the fragmentation occurs is @xmath154 .
the hill radius of the clump at @xmath78 is written by @xmath155 thus , we roughly estimate the maximum number of the clumps which survives as @xmath156 if some clumps grow at rates higher than @xmath157 and their masses increase within @xmath158 , the maximum number decreases . we here consider @xmath159 as a conservative value . , the hill radius of the clump @xmath141 ( eq . [ eq : hill ] ) , the size of the circum clump disc @xmath160 , the size of the h@xmath161 region around the clump @xmath162 ( eq . [ eq : ion ] ) , the gravitational radius of the clump @xmath163 ( eq . [ eq : grav ] ) , and the size of the whole nuclear disc @xmath164 ( eq . [ eq : rd ] ) .
, width=313,height=136 ] first , let us consider radiative feedback on the accretion on to the zams star ( clump ) itself . figure [ fig : clump_disc ] illustrates the gas structure around the clumps and the characteristic radii . the gas accreting on to the zams star makes a small circum - clump disc .
the disc size is roughly estimated as @xmath165 @xcite , based on numerical simulations @xcite .
the zams star emits uv radiation and an h@xmath161 region is formed above and below the disc .
the size of the h@xmath161 region depends on the ionizing luminosity of the zams star and on the density profile of the in - falling material . for simplicity
, we estimate the size @xmath166 in the polar direction using eq .
( 36 ) of @xcite as @xmath167 where we assume that the disc mass is equal to the clump mass and we take the temperature of the h@xmath161 region to be @xmath168 k @xcite .
the sound speed in the h@xmath161 region is @xmath169 km s@xmath36 . for a zams star located at @xmath78
, we obtain @xmath170 since the sound - crossing time within @xmath162 is much shorter ( @xmath171 yr ) , the ionization front expands rapidly . when the front reaches the gravitational radius @xmath172 the ionized gas breaks out through the neutral infalling gas .
the ionizing photons subsequently heat the disc surface , and thus photo - evaporation begins to suppress the accretion rate .
the photo - evaporation rate can be expressed as @xmath173 where @xmath174 is the ionizing photon number flux @xcite . using the relation @xmath175 s@xmath176 in the mass range @xmath177
, we obtain @xmath178 which is smaller than @xmath179 by approximately an order of magnitude .
we therefore conclude that accretion on to the clump could not be suppressed by its own photoionization heating .
after the h@xmath161 region breaks out of the disc , however , the accretion proceeds only from the shadow of the disc and thus the accretion rate is reduced .
moreover , @xcite suggest that the gas behind the disc is shocked and accelerated outward by the pressure gradient after the expansion of the h@xmath161 region
. this process halts the accretion at @xmath180 .
@xcite have shown that photo - evaporation starts to suppress the accretion when @xmath181 .
note that accretion on to a clump inside @xmath78 can not be suppressed , even if the h@xmath161 region expands by a large factor .
in particular , a clump growing at a super critical rate , @xmath182 , is not affected by the radiative feedback because @xmath183 .
next , we discuss the possibility that the ionizing photons from the collection of all zams stars together suppress the accretion ( with @xmath184 ) from the parent cloud on to the disc as a whole . to make our discussion conservative , we assume that the clumps which emit strong uv radiation grow at the rate just below the critical accretion rate . for @xmath137 ,
the orbital decay time is @xmath185 yr and so each clump then grows to @xmath10 .
the corresponding total ionizing photon rate is @xmath186 s@xmath36 . from numerical simulations of the gravitational collapse of an atomic - cooling cloud @xcite ,
the rotational velocity is proportional to the keplerian velocity , @xmath187 , with @xmath188 . from this relation , the size of the whole disc around the central protostar
is given by @xmath189 where @xmath190 is the size of the quasi - spherical parent cloud fuelling the inner disc .
we approximate @xmath190 using a critical bonnor - ebert sphere with a temperature of @xmath191 k and a central density of @xmath192 , which corresponds to that of the gravitationally unstable core in the atomic cooling haloes without h@xmath11 molecules ( e.g. , * ? ? ?
therefore , we obtain @xmath193 and thus @xmath194 , which is comparable or only slightly above the critical value . therefore , even with conservative assumptions , the strong uv radiation can not deplete the gas supply on to the disc from the parent cloud , and the central protostar can grow to a sms star with @xmath195 .
the site envisioned for forming a sms is atomic cooling gas in a halo with virial temperature @xmath6 k , in which h@xmath11 cooling is prohibited prior to and throughout the protostellar collapse .
recent numerical simulations suggest that the majority of the atomic cooling haloes are polluted by heavy elements due to pop iii supernovae from prior star formation @xcite .
the resulting metallicity is @xmath16 ( @xmath196 ) if the pop iii supernovae are core - collapse type ( pair - instability type ) .
if the gas is slightly polluted with @xmath197 , the temperature decreases below @xmath198 k by dust cooling @xcite , which could promote efficient fragmentation .
the details of this fragmentation , and whether dust cooling ultimately prevents sms formation , are not yet understood .
the temperature of the gas with dust grains begins to decrease through heat exchange with cool dust grains ( @xmath199 ) when collisions between gas particles and dust are sufficiently frequent .
collisional cooling of the gas is efficient until @xmath200 k , where the gas and dust is thermally coupled ( @xmath201 ) . at the cooling phase ,
the compressional heating and collisional cooling are balanced as @xmath202 where @xmath203 is the free - fall time and @xmath204 is the energy exchange rate ( in erg s@xmath36 @xmath37 ) between the gas and dust , defined by @xmath205 for @xmath206 @xcite .
we adopt the number density and cross - section of the dust particles from , @xmath207 where the depletion factor of metals to dusts is assumed as high as the present - day galactic value @xmath208 . from eqs .
( [ eq : d1 ] ) , ( [ eq : d2 ] ) , and ( [ eq : d3 ] ) , we obtain @xmath209 above which the temperature begins to decrease compared to the zero - metallicity case . the density given by eq .
( [ eq : d4 ] ) is comparable to that at the fragmenting radius .
this implies that the discussion in [ sec : fate ] remains valid in metal - polluted gas , as long as the metallicity remains below @xmath16 .
we conclude that disc fragmentation can not prevent sms formation , as long as the metallicity is below this value . on the other hand , at higher metallicity ( @xmath210 ) , the gas temperature rapidly decreases by metal - line cooling at densities below @xmath211 @xcite .
numerical simulations of metal - enriched collapse into atomic cooling haloes also capture the character of gas fragmentation for @xmath212 @xcite .
the outcome would be a compact star - cluster which consists of many low - mass stars with @xmath213 @xcite .
in this paper , we discussed the properties of a disc around the embryo of a sms ( @xmath0 ) , expected to be present in a primordial gas without h@xmath11 molecules in massive haloes with virial temperature @xmath77 k. a high accretion rate @xmath214 , sustained for @xmath215yr , is required to form a sms .
the inner region of such a disc is gravitationally unstable , and fragments into @xmath216 clumps with characteristic mass of @xmath217 .
we discuss the possibility that this fragmentation prevents sms formation .
we argue that most of the clumps formed in the disc rapidly migrate towards the central protostar and merge with it .
the orbital decay time is shorter than or comparable to the orbital period of the clumps ( @xmath218 yr ) .
some of the clumps can grow via accretion and evolve to zams stars within their kh time , either because they survive longer due to the stochasticity of the migration process , or because they form later , further out in the disc .
our toy model , on which these conclusions are based , can be tested against simulations of pop iii star formation in lower mass minihaloes . in this case , the dominant cooling process is h@xmath11 line emission ( @xmath219 k ) and the resulting accretion rate is @xmath220 , which is smaller than that in the sms formation case . according to high - resolution numerical simulations by @xcite , clump formation occurs at @xmath221 au ( @xmath222 g @xmath68 , @xmath223 yr@xmath36 , @xmath224 km s@xmath36 and @xmath225 in their fig .
2 ) in the gravitationally unstable disc at the early stage @xmath226 yr . from eq .
( [ eq : r_f_early ] ) , we can estimate the fragmentation radius as @xmath227 au , which is consistent with the fragmentation radii in the simulations . at this stage ,
the average accretion rate is high , @xmath228 in their fig .
8 . from eq .
( [ eq : continue ] ) , ( [ eq : scaleheight ] ) , ( [ eq : viscous ] ) , and @xmath25 , @xmath229 using these equations , the sound speed at the fragmentation radius ( @xmath47 ) is estimated as @xmath230 km s@xmath36 , which agrees well with the numerical result .
the viscous time at @xmath221 au is roughly estimated as @xmath231 yr . since some clumps migrate inward within @xmath232 yr ( @xmath233 ) in simulation , other effects ( type i migration and interaction between clumps ) could accelerate the clumps migration .
therefore , our results from the toy model are expected to be conservative , and appear to capture the essential features of disc fragmentation and clump migration .
the gas in most of the massive haloes is likely to be polluted by heavy elements due to prior star formation and pop iii supernovae . with the metallicity of @xmath234 , the dust cooling decreases the temperature rapidly and could promote the fragmentation at @xmath235 @xcite . as a result ,
the existence of dusts has been considered as one of the most severe obstacles to the sms formation .
however , the fragmentation induced by the dust cooling is expected to occur only within the fragmenting radius we estimated .
this means that the clumps formed by the dust cooling can migrate towards the center on a time - scale comparable to the orbital period , in the same way as the primordial case .
we conclude that the sms formation is not prevented , unless the metal - line cooling dominates for @xmath236 .
our results therefore remove a significant obstacle for sms formation . our estimates for the orbital decay time of the clumps
are based on the well - known formulae of type i and ii planetary migration .
although some numerical simulations find that these formulae remain approximately valid in gravitationally unstable and clumpy discs , there are many uncertainties regarding the migration rate in this case . for example
, we neglect the orbital eccentricities of the clumps .
the clumps formed by the disc fragmentation are expected to have eccentric orbits and thus the angular momentum transfer between the disc and clumps could change from the case of the circular orbits . moreover , in such a situation , the interaction with eccentric clumps retard the orbital decay of the other clumps @xcite . to estimate the migration time within the clumpy disc around the sms star , more sophisticated numerical simulations are necessary .
it is worth further emphasizing the limitations and uncertainties of our disc model . in figure
[ fig : profile ] , we present the disc profile based on cooling by free - bound emission of h@xmath38 and thermal equilibrium .
hydrogen atomic cooling ( ly@xmath40 and two - photon emission ) dominates at @xmath237 k ( @xmath238 yr ) . to quantify the effect , we include ly@xmath40 cooling ( @xmath239 ; @xcite ) in the limiting optically thin case ( dashed lines in fig .
[ fig : profile ] ) .
since the gas is in fact optically thick to the ly@xmath40 photons at @xmath240 , and the cooling efficiency of the two - photon channel is smaller than the ly@xmath40 emission , the dashed lines overestimate the additional atomic cooling rate .
nevertheless , the profiles deviate from our fiducial disc model only modestly , beginning at @xmath238 yr and thus the critical period changes at most within the shaded region in figure [ fig : profile ] ( c ) .
the crucial ingredient of this paper is the critical value of the effective viscous parameter @xmath241 for fragmentation . in our model ,
the fragmentation radius depends sensitively on the value of @xmath241 . for example , if we chose @xmath242 , we obtain @xmath243 s@xmath36 ( @xmath244 yr ) , @xmath245 g @xmath68 , @xmath246 and @xmath247 pc . in this case , the orbital decay time becomes longer than the kh time of the clumps by a factor of 4 , and most of the clumps may evolve to massive stars and emit strong uv radiation .
nevertheless , the number of clumps decreases by a factor of 2 from the case of @xmath248 ( see eq . [
eq : nc ] ) .
therefore , we expect that our main conclusion also does not change . however , if @xmath249 , the fragmenting radius moves outside the size of the whole disc given by eq .
( [ eq : rd ] ) , which means that the disc can not exist , i.e. , our model is no longer self - consistent . to explore the precise value of @xmath241
is left for future investigations . in particular ,
our results motivate high - resolution numerical simulations similar to @xcite .
moreover , more realistic treatments of radiative cooling and chemical reactions at densities higher than @xmath250 @xcite should be included since the fragmentation efficiency strongly depends on the equation of state of the gas . despite these caveats
, we expect that sms formation in metal - poor gas in atomic - cooling haloes is difficult to avoid , as it requires that the high mass inflow rate , @xmath14 , is democratically distributed among @xmath251 fragments ( so that _ none _ of them accretes at a super critical rate for sms formation ) and that most of these fragments survive rapid migration and avoid coalescence with the growing central protostar for several orbital times .
the toy models presented in this paper disfavor this scenario .
we thank greg bryan , kazuyuki omukai , takashi hosokawa , kei tanaka and eli visbal for fruitful discussions .
this work is supported by the grants - in - aid by the ministry of education , culture , and science of japan ( ki ) , and by nasa grant nnx11ae05 g ( zh ) .
in this appendix , we briefly describe how to obtain the expression of eq .
( [ eq : x_e ] ) .
the reaction rate coefficient of radiative recombination @xmath252 is a function of temperature @xcite . on the other hand , that of the collisional ionization @xmath253 has a complex form
@xcite @xmath254 and @xmath255 where @xmath256 are the partition functions , @xmath257 k , and @xmath258 is defined by @xmath259 and @xmath260 where @xmath261 is the collisional transition rate , @xmath262 ( @xmath263 ) the radiative transit rate from the first excited state to the ground state ( two - photon emission , respectively ) , and @xmath264 the escape probability of photons with @xmath265 ev ( see @xcite in details ) . for a wide range of densities ( @xmath266 @xmath37 ) , the following approximation is valid : @xmath267
|
supermassive stars ( smss ; @xmath0 ) formed in the first protogalaxies with virial temperature @xmath1 k are expected to collapse into seeds of supermassive black hole in the high - redshift universe ( @xmath2 ) .
fragmentation of the primordial gas is , however , a possible obstacle to sms formation .
we discuss the expected properties of a compact , metal - free , marginally unstable nuclear protogalactic disc , and the fate of the clumps formed in the disc by gravitational instability .
interior to a characteristic radius @xmath3 pc , the disc fragments into massive clumps with @xmath4 .
the clumps grow via accretion and migrate inward rapidly on a time - scale of @xmath5 yr , which is comparable or shorter than the kelvin - helmholtz time @xmath6 yr .
some clumps may evolve to zero - age main - sequence stars and halt gas accretion by radiative feedback , but most of the clumps can migrate inward and merge with the central protostar before forming massive stars .
moreover , we found that dust - induced fragmentation in metal - enriched gas does not modify these conclusions unless @xmath7 , because clump migration below this metallicity remains as rapid as in the primordial case .
our results suggest that fragmentation of a compact , metal - poor disc can not prevent the formation of an sms .
quasars : general cosmology : theory dark ages , reionization , first stars .
|
precision test of the unitarity of the ckm matrix , allowing us to explore the sm description of the cp violation and reveal any physics beyond the sm , greatly depends on our knowledge of the nonperturbative matrix elements .
fortunately the calculation of the amplitudes of bottom meson decays is under control as the amplitudes can be expanded in terms of small ratios justified by both the large @xmath4-quark mass , and a large energy release in the decay . with this expansion ,
a number of theoretical predictions on different observables in various channels are found in global agreement with experimental measurements ( see ref .
@xcite for a review ) .
decay processes of heavy baryons consisting of a bottom quark provide complementary information with the @xmath10 meson and thereby are receiving growing attentions on both experimental and theoretical sides .
semileptonic decays , such as @xmath11 , are simplest exclusive baryonic decays and governed by heavy - to - light form factors . in this retrospect , apart from the theoretical analysis based on the heavy quark effective theory @xcite , the simplification of baryonic form factors in the large energy limit is exploited @xcite ( see ref .
@xcite for an earlier discussion ) , applying the method developed in the mesonic case @xcite . in the @xmath12 transition ,
only one form factor is nonzero after the reduction and this universal function ( soft form factor ) is also calculated within the light - cone qcd sum rules in conjunction with the effective field theory @xcite .
soft - collinear effective theory ( scet ) @xcite is a powerful tool to describe processes with particles having energy much larger than their mass .
the heavy - to - light decay of heavy baryons , for instance @xmath13 , is of this type .
scet makes use of the expansion in small ratios , in this case , @xmath14 with @xmath2 as the hadronic scale and @xmath3 as the @xmath4 quark mass .
one of the most important features of scet is that the interaction between the soft and collinear sectors is taken into account , overcoming the shortcomings in the large energy effective theory @xcite .
therefore in scet not only the reduction of the leading - power form factors is formulated on the qcd basis , but also the symmetry - breaking corrections can be systematically explored @xcite . in this work
, we will analyze the baryonic form factors in scet and follow the techniques developed in the @xmath15 form factor which takes the following factorization form at the leading power @xcite @xmath16 here @xmath17 is the energy of the final hadron and @xmath18 and @xmath19 are the short - distance coefficients obtained by matching from qcd onto the effective field theory .
the one - loop expressions for these coefficients can be found in refs .
@xcite . in what follows
we will adopt the ansatz that the final light particle is composed of collinear objects and thus hard - collinear gluon exchange is required to turn the soft spectators into energetic ones . in such picture , to the end we will show that the matrix elements parametrizing form factors , in the example of @xmath12 , are formally simple @xmath20 in which the spin indices are suppressed . for contributions dominated by soft processes which are not suppressed by @xmath21 , please see refs . @xcite .
the remainder of this work is organized as follows . in sec .
[ scet : analysis ] , we will present the form of the leading power and next - to - leading power heavy - to - light currents in scet after integrating out the hard modes , and following ref .
@xcite discuss their representations in the effective theory containing soft and collinear modes . in sec .
[ eq : qcdanalysis ] , the transition form factors are directly calculated in qcd , and we show the correspondence with the scet effective operators .
several implications from our analysis are given in sec .
[ sec : discussions ] , and a summary of our findings is presented in sec .
[ sec : conclusion ] .
we use the position - space representation of scet and closely follow the notations in refs .
we work in the b - baryon rest frame and use the light - cone coordinate , in which a momentum @xmath22 is decomposed as @xmath23 where @xmath24 are two light - like vectors : @xmath25 and @xmath26 .
the reference directions @xmath24 are chosen such that the energetic massless external lines in the recoiling system have @xmath27 of order @xmath28 , while the magnitude of @xmath29 is small .
this type of momenta is collinear : @xmath30 .
the slowly - moving degrees of freedom in the heavy baryon have soft momenta @xmath31 for the heavy @xmath4 quark , the statement of soft " refers to the residual momentum after removing the large component which becomes a label of heavy quark .
the hard - collinear mode , with @xmath32 virtuality , arises from the interaction between soft and collinear sector : @xmath33 .
power scalings of quark and gluon fields are determined by the configuration of their momenta . for the quark fields
, we have @xmath34 here @xmath35 is the velocity of the heavy quark . @xmath36 and @xmath37 are large components of the collinear , hard - collinear and heavy quark fields , respectively .
small components of the heavy quark field , @xmath38 , and collinear quarks , @xmath39 and @xmath40 , can be integrated out at tree level by solving the equation of motion .
scalings of gluon fields have a similar behavior with their momenta @xmath41 from the relativistic normalization condition , we find that the baryonic states in the effective theory , taking the @xmath42 and @xmath2 as an example , have the scaling @xmath43 where we did not specify the differences with the states in qcd .
presumably these differences may introduce more power corrections , but they are left out here , since the leading - power behavior is unlikely to change .
decay constants of baryons defined via @xcite @xmath44 scale as @xmath45 and @xmath46 with @xmath47 denoting both @xmath48 and @xmath49 . in scet , integration of the fluctuations with
large virtualities proceeds in two - steps @xcite . in the first step ,
hard scales , caused by the interaction between the collinear sector and heavy quark , and between two or more collinear sectors with different directions , are integrated out and thereby qcd is matched onto an intermediate effective theory , called scet@xmath50 . in this effective theory gauge invariant operators are built out of fields of hard - collinear quarks or soft gluons and quarks .
the leading - power and next - to - leading power terms having non - zero matrix elements between the baryonic transition are constructed as @xmath51 where the hard - collinear field with the subscript @xmath52 is evaluated at @xmath53 , while the soft field with the subscript @xmath54 is evaluated at @xmath55 , with @xmath56 being the space coordinate from the qcd current .
@xmath57 is one of the following gamma matrices @xmath58 the @xmath59 and @xmath60 are hard - collinear and soft wilson lines , respectively @xcite .
integration of the hard - collinear mode will result in the final scet , named as scet@xmath61 for convenience . in scet@xmath50 , the generic power scalings of the operators in eq .
are @xmath62 but none of them have the right quantum numbers with baryons in the initial and final state .
thus the matching of these operators from scet@xmath50 onto scet@xmath61 will induce additional power suppressions and one of our goals is to count these suppressions . to the end , we will demonstrate that the contribution from the @xmath63 operator starts at the @xmath64 , while the other types of operators have the power @xmath65 . to represent the quantum numbers of the @xmath42 and @xmath2 baryon ,
at least the fields @xmath66 and three collinear quark fields are needed . in the light - cone gauge a most general form of an operator with non - vanishing matrix elements
can be taken as @xcite @xmath67 \times ( \bar \xi_c \{1,n\!\!\!\slash_+/2\ } \gamma_k ' q_s ) ( \bar \xi_c \{1,n\!\!\!\slash_+/2\ } \gamma_l ' q_s ) ( \bar \xi_c \gamma_j ' h_v ) .
\label{eq : genericform}\end{aligned}\ ] ] where the objects in the brackets are combinations of the building blocks : @xmath68 with the integers @xmath69 being the number of occurrences of @xmath70 in an operator .
@xmath71 take one of the forms in eq .
, while @xmath72 is a basis for the remaining eight boost - invariant dirac structures .
we , following ref .
@xcite , use the power scaling , boost invariance and the matching of mass dimensions to pick up the allowed forms .
the notation for these symbols is used as : @xmath73_o = n$ ] means that @xmath74 scales with @xmath75 , the `` boost '' label corresponds to the scaling @xmath76 of @xmath74 under boosts @xmath77 , @xmath78 ; the mass dimension is denoted by @xmath79_o$ ] . using the properties of these building blocks which are discussed in detail in table 2 of ref .
@xcite , we find an operator in the final effective theory has the scalings @xmath80= 15 - 2 n_1 + 2 n_5 + 2 n_6 + 4 n_7 + 4 n_8 + 6(n_{9a}+ n_{9b}+n_{9c } ) , \nonumber\\ & & [ \alpha]=0= -n_1+n_2+n_3-n_4-n_6+n_{7 } -n_8-n_{9a}+n_{9b},\nonumber\\ & & [ { d } ] = 9-n_1-n_2+n_5+n_6+n_{7 } + 3(n_8+n_{9a}+n_{9b}+n_{9c}),\end{aligned}\ ] ] from which we have @xmath81 = 6 + [ d ] -n_3+n_4+n_5 + 2 n_6 + 2 n_{7 } + 2 n_8 + 4 n_{9a}+ 2 n_{9b}+3 n_{9c}.\end{aligned}\ ] ] in operators @xmath82 the only lorentz structure having nonzero contraction with @xmath83 is @xmath84 , and thus @xmath85 . for the @xmath63 operator , @xmath79=3 $ ] and there is only one nontrivial solution with @xmath73=9 $ ] , @xmath86 , @xmath87 and @xmath88 .
since there is no free lorentz index containing @xmath83 or @xmath89 in @xmath57 , the equality @xmath87 rules out the possibility of @xmath90 in eq . .
as for @xmath91 operator , @xmath79=4 $ ] thus @xmath73\ge 10 $ ] . since @xmath92 is an integer ,
the leading contribution from this operator has the scaling @xmath73=11 $ ] .
one solution is @xmath93 , @xmath94 , @xmath95 , @xmath88 and the other is @xmath87 , @xmath86 , @xmath96 , @xmath97 .
the latter one corresponds to the higher fock state contribution , due to the presence of an extra soft or collinear gluon .
the @xmath98 also belongs to this type . in the light - cone gauge
@xmath99 reduces to @xmath100 . in the @xmath101 operator ,
@xmath79=4 $ ] and the factor @xmath83 contracts with the gluon field @xmath102 .
after the elimination of the hard - collinear fields , @xmath83 can not be a free lorentz index and maybe it is contracted : with @xmath103 which is a constant or in the form of @xmath84 ; with a gamma matrix as @xmath104 ; with a derivative to a soft field in the form of @xmath105 ; or with a collinear gluon field as @xmath106 . in the first contraction , @xmath107 , and @xmath73\ge 10 $ ] . due to the integer constraint on @xmath92 , this operator has the scaling @xmath73=11 $ ] and its solution is similar to the one in @xmath108 .
for the rest cases , @xmath109 may be larger than @xmath110 by one unit , but @xmath111 or @xmath112 , causing more power suppressions and resulting in @xmath73 > 11 $ ] . for the operator @xmath113 , @xmath79=5 $ ] and @xmath73\ge 11 $ ] .
the solution having the power @xmath73= 11 $ ] is @xmath114 , @xmath87 and @xmath115 .
the above matching analysis indicates that the operator @xmath63 is indeed dominant and others are @xmath116 suppressed .
taking into account the power scalings of baryonic states , we obtain the scaling laws for operator matrix elements @xmath117 now we will perform a tree - level matching from scet@xmath50 to scet@xmath61 , and identify various terms to different types of operators . in this procedure , the hard - collinear quark field is first expressed as a product of soft and collinear fields and the hard - collinear gluon fields .
then the hard - collinear gluons are integrated out by solving the equation of motion for the yang - mills fields and their expressions in terms of soft and collinear quarks and gluons will be substituted back into the hard - collinear quark field . for simplicity
, we shall work in the light - cone gauge @xmath118 and the gauge invariant form can be obtained by the field redefinition technique .
the qcd currents can be matched onto the effective currents in the scet @xmath119(x ) \to
e^{-im_b v\cdot x } [ \bar \psi \gamma { \cal q}](x)\end{aligned}\ ] ] with @xmath120,\nonumber\\ & & { \cal q}= \left(1+\frac{id\!\!\!\!\slash_s}{2m_b}\right ) h_v -\frac{1}{n_-v } \frac{n\!\!\!\slash_-}{2m_b } ( ga\!\!\!\slash_{\perp c } + ga\!\!\!\slash_{\perp hc } ) h_v \nonumber\\ & & + \frac{1}{2m_b n_-v } \left[\frac{1}{in_+\partial } ( ga\!\!\!\slash_{\perp c}+ ga\!\!\!\slash_{\perp hc } ) ( ga\!\!\!\slash_{\perp c } + ga\!\!\!\slash_{\perp hc } ) \right]h_v \nonumber\\ & & -\left[\left\{\frac{1 } { m_b n_-v } \frac{n\!\!\!\slash_- n\!\!\!\slash_+}{4}- \frac { n_+v } { n_-v
i n_+ \partial } \right\ } ( n_-a_{hc } ) \right]h_v + { \cal o } ( \lambda ^4 h_v),\label{eq : treelevel}\end{aligned}\ ] ] with the derivative @xmath121 acting on the collinear field in the square bracket . in the light - cone gauge
, the collinear quark lagrangian reads as @xmath122 \frac{1}{in_+d_s } [ id\!\!\!\!\slash_\perp ] \right ) \frac{n\!\!\!\slash_+}{2 } \xi + ... , \end{aligned}\ ] ] with the ellipses standing for all other terms . here
@xmath123 and the collinear gluon in the covariant derivative denote both collinear and hard - collinear field and will be substituted as @xmath124 @xmath125 . with the use of the equation of motion ,
the @xmath126 can be integrated out and in particular , the solution ( dropping the terms not satisfying momentum conservation ) @xmath127 contributes to @xmath128 as @xmath129 in which the expressions of gluons will be specified below .
the other useful pieces are @xcite @xmath130 where the first term in @xmath131 is from the small component of the hard - collinear quark field @xmath132 and the ones in the large parentheses are from @xmath133 .
the relevant hard - collinear gluon field is expanded as @xcite @xmath134 - g\,\big[in_+\partial a^{\mu_\perp}_c , a_{\mu_\perp hc}^{(3)}\big ] \nonumber \\ & & -\,2 g t^a\bigg\{\bar\xi_ct^a\bigg(\frac { n\!\!\!\slash_+}{2}- \frac{1}{in_- \partial}\,g a\!\!\!\slash_{\perp c}\bigg ) q_s+\mbox{h.c . } \bigg\}\!\bigg\},\nonumber\\ n_-a_{hc}^{(6)}&= & -\frac{2 } { ( in_+\partial)^2 } \big [ -2 [ in_+\partial a_{\perp hc}^{(3)\mu } , a_{\mu\perp hc}^{(3 ) } ] + 2 gt^a \big\{\bar \xi_c t^a \big(\frac{1}{in_-\partial } ga\!\!\!\slash_{\perp hc}^{(3 ) } \big ) q_s + h.c . \big\}\big].\end{aligned}\ ] ] with the covariant derivative @xmath135 $ ] . before substituting the hard - collinear fields into the currents , we first count the collinear quark numbers .
the final baryonic state contains three quarks , and has collinear quark number + 3 . in order to have nonzero matrix elements ,
the effective currents in the scet must have the collinear quark number 3 as well .
let us recall that the gluon filed @xmath136 contains one collinear quark ( or antiquark depending on the interaction form in the effective theory ) , while @xmath137 may contain two collinear quarks .
for the expression of @xmath138 , we note that @xmath139 ( and also @xmath140 ) has collinear quark number @xmath141 , while @xmath142 has collinear quark number 0 .
the most nontrivial terms are : @xmath131 which has a collinear quark number @xmath143 or 0 , and @xmath128 with collinear quark number @xmath144 ( or @xmath145 ) .
the combinations having the leading and next - to - leading power scalings indeed take the forms as @xmath63 , @xmath146 and @xmath98 . substituting @xmath136 , @xmath147 and @xmath128 into the effective currents
, we have the leading term in the expansion @xmath148 the first term in the above equation contains two hard - collinear gluons emitted from the hard - collinear quark , and is depicted as the first diagram in fig .
[ fig : scetff ] . in this figure ,
the dashed lines denote the collinear quarks , while the solid lines are soft spectators .
the thick lines represent the heavy bottom quark .
the spring lines denote a collinear gluon @xmath100 while spring+solid lines denote the @xmath149 . in the @xmath150 , the trigluon term ,
corresponding to fig ( [ fig : scetff]c ) , vanishes and it can be understood as follows .
the three quarks have antisymmetric colors in both initial and final baryons , and thus the color rearrangement factor in this diagram is zero @xmath151 the current @xmath152 originates from the large component of the hard - collinear quark field @xmath153 as shown in eq .
and thereby the lorentz structure is reduced : @xmath154 as expected in the large recoil limit .
the other combinations of operators start from @xmath65 @xmath155h_v + \bar \psi^{(6)}\gamma \frac{id\!\!\!\!\slash_s}{2m_b } h_v+ ... ,\end{aligned}\ ] ] where these four pieces can be incorporated into the operators @xmath156 respectively .
it should be noted that except the second term , the other terms can have different lorentz structures with the reduced form as in eq . .
for instance , the fourth term is from the small component of the heavy bottom quark , which has the lorentz structure @xmath157 .
while the spring+solid lines denote the @xmath149 .
the hard modes have been integrated out and shrunk to the black point . ]
we also show the tree - level matching diagrams for the @xmath146 operators in fig .
[ fig : scetff ] .
however the higher fock state contributions , either from @xmath98 having the similar structure with @xmath63 except that one additional soft gluon is emitted from the hard vertex , or from the operator @xmath91 , are not depicted . graphically speaking the dominance of @xmath63
can be understood as follows . in the three diagrams ( a , d , e )
one commonality is that the two gluons interact with a soft quark from the initial state and a collinear quark in the final external state , and thereby these two vertices have the same power scaling .
however in the first diagram the quark propagator next to the weak vertex has the form @xmath158 while the rest quark propagators are of order @xmath159 , leading to the enhancement of the first diagram .
in this section , we will analyze the leading power behaviors of the baryonic transition form factors in qcd , whose feynman diagrams are depicted in fig . [
fig : qcdff ] .
we adopt the ansatz that the fast - moving baryon is composed of three collinear constituents , therefore at least two gluons are exchanged and these gluons must be far off - shell .
we will not include the contributions involving higher fock states , as at least one more gluon is needed .
as we have already shown , the trigluon diagrams give vanishing contributions and thereby will not be considered either .
there are seven diagrams shown in fig .
[ fig : qcdff ] : three of them ( a , b , c ) containing the momentum exchange by two gluons between the spectator quark system and the energetic light quark connecting the the electroweak vertex ; the same number of diagrams ( e , f , g ) having two gluons emitted from the heavy quark ; the rest diagram ( d ) in which the light spectator system receives momentum exchange from both the energetic quark and the heavy quark
. the inclusion of the flavor index will give another seven diagrams , but only leads to the exchange of momentum fractions of the light spectator quarks .
the leading twist lcda of a light baryon , such as @xmath2 , is @xcite @xmath160 with @xmath161 being the color indices and @xmath162 being the spinor indices .
for the heavy baryon , several types of lcdas emerge @xcite @xmath163 in the leading power matrix elements , only the first type of lcda contributes .
we choose the momentum fractions of the three collinear quarks in the light baryon as @xmath164 and the momentum fractions of the soft spectator quarks ( in the direction @xmath103 ) in the initial state as @xmath165 and @xmath166 .
the corresponding momenta will be denoted as @xmath167 for the collinear quarks , and @xmath168 for the soft quarks .
the calculation will be simplified by the following two observations . * if both vertices of a hard - collinear gluon are attached to collinear quarks , only the transverse component of this gluon contributes . * in the light spectator system ( usually called a diquark ) , only the diagrams with even number of gluon transverse indices are nonzero . for instance , as shown in fig [ fig : qcdff](b , e , f ) , if only one hard - collinear gluon is emitted from the light quark or the heavy quark , this gluon has to be in the form @xmath100 or @xmath169 .
the first observation can be proved by writing the amplitudes as @xmath170 \times [ \bar q_2 \gamma^{\mu } ... ] = [ \bar q_1 \gamma_{\perp\mu } ... ] \times [ \bar q_2 \gamma^{\perp\mu } ... ] , \end{aligned}\ ] ] with @xmath171 and @xmath172 being the two collinear quarks attached to the gluon .
the second one is based on the fact that the two - spectator system technically forms a trace in the spinor space .
there is no transverse index from the external wave functions , and thereby the internal ones from the exchanged gluons must be even . the leading power contributions from fig . [ fig : qcdff](a , b ) can be matched onto the @xmath63 operator . in fig . [ fig : qcdff](a ) , using the first observation , one of the two gluons ( the right one ) is connected to two collinear quarks , and only the transverse component is left . with the second observation , the other gluon must take the transverse component as well . in the numerator of the quark propagator between the two gluons ,
the collinear momentum @xmath173 does not contribute since it is next to the light spinor : @xmath174 .
this propagator is simplified as @xmath175 which scales as @xmath176 .
the other quark propagator is reduced to @xmath177 which has the scaling @xmath178 .
here we have used @xmath179 for the soft momentum fraction . combining these pieces , this diagram has the form @xmath180 where the scaling @xmath181 is from decay constants and @xmath182 comes from the two gluons propagators and the propagator in eq .
( [ eq:2propagator ] ) .
@xmath183 is the color factor @xmath184 eq
. confirms our power counting analysis given in the previous section .
furthermore as indicated in the third line of the above equation , the light spectator ( diquark ) system is proportional to @xmath185 which results in the lorentz structure @xmath186 where the large energy and heavy quark symmetries are manifestly demonstrated again . in fig .
[ fig : qcdff](b ) , the upper gluon vertex is replaced by @xmath187 and the quark propagator next to the electroweak vertex is of the form @xmath188 .
therefore this diagram has the same structure : @xmath189 in fig .
[ fig : qcdff](c ) , based on the first observation , the upper gluon is transverse ; thus there are either one or three transverse indices in the light spectator system , leading to vanishing contribution . in fig .
[ fig : qcdff](d ) , both gluons can only contain transverse components and this diagram can be matched onto the operator @xmath108 .
both the heavy quark and light quark propagators scale as @xmath176 and thus @xmath190 with again @xmath181 from decay constants and @xmath191 from the two gluon propagators . of particular interest
is that the lorentz structure in this diagram has the form @xmath192 which manifestly breaks the large recoil symmetries . in fig .
[ fig : qcdff](e ) , the gluon attaching to the two light quarks is transverse while the component @xmath193 contributes at the heavy quark propagator .
this diagram corresponds to the operator @xmath101 . using the first observation , the light quark propagator scales as @xmath176 and thus @xmath194 u_{\lambda_b } \nonumber\\ & & \times \frac { n_+p'}{2 } ( c n\!\!\!\slash_- \gamma_5)_{\alpha\beta } ( \gamma_{\perp \mu } \frac { in\!\!\!\slash_+ } { 2 ( y_2+y_3 ) n_+p ' } n\!\!\!\slash_- ) _ { \alpha\alpha ' } ( \gamma_\perp^{\mu})_{\beta\beta ' } ( n\!\!\!\slash_+ \gamma_5c)_{\beta'\alpha ' } \nonumber\\ & \propto & \lambda^{10}/\lambda^4 \sim \lambda^6 . \end{aligned}\ ] ] the first term in the square bracket obeys the large recoil symmetries , but the integral over @xmath195 in it is divergent .
it is worthwhile to point out that in the scet solution for the operator @xmath98 in the previous section , the number of the occurrence of @xmath196 is found to be @xmath94 , which means the momentum fractions for the light baryon can appear only three times .
the additional momentum fraction arises from the short - distance coefficients , for instance at tree - level shown in eq . . in fig .
( [ fig : qcdff]f ) , the gluon attaching to the @xmath4 quark can not contribute with the transverse component based on the second observation .
the @xmath169 component can be absorbed into the wilson line , one necessary piece in the gauge invariant definition of the scet operators .
thus this diagram is incorporated into the operator @xmath63 and its scaling is @xmath197 in particular this contribution cancels the one from fig .
( [ fig : qcdff]b ) . in fig .
[ fig : qcdff](g ) , the two heavy quark propagators have the offshellness of order @xmath198 and can be shrunk to one point .
suppose that the two gluons are transverse , and then it is incorporated into @xmath113 and its power scaling is @xmath199 where the momentum fraction @xmath200 in the second line comes from the wilson coefficient for the operator @xmath113 . if the two gluons take the @xmath201 component for the vertices attaching to the heavy quark , this diagram can be matched onto operator @xmath63 .
the contribution is @xmath116 suppressed compared to the leading power terms @xmath202 and the integration in this term does not converge .
as we have shown , in the dominant contribution from the @xmath63 the inverse of derivatives to both collinear fields and soft fields appear three times . in the momentum space
these factors will be converted to the inverse of momenta .
let them act on the collinear fields , we obtain the factor @xmath203 .
the energy dependence of a quark field can be read from the propagators @xmath204 the measure @xmath205 and @xmath206 are lorentz invariant , and thus @xmath207 .
then the leading power baryonic transition matrix element scales as @xmath208 where we have employed the energy independence of baryon states . using @xmath209 and the definition of the soft form factor in scet @xmath210 and restoring correct mass dimensions
, we obtain the momentum dependence @xmath211 this behavior can also be read from the qcd calculation as shown in eq . .
but it should be noticed that the above scaling law is different with the results derived in different versions of qcd light - cone sum rules @xcite in which the form factor is dominated by soft processes . to have the power counting
, we represent the form factor as an overlap integral of the wave functions in both longitudinal and transverse momentum space @xmath212 with @xmath213 and @xmath214 to be fixed by kinematics . from the normalizations of the b - baryon state
, we have @xmath215 implying that @xmath216 since @xmath217 and @xmath218 . for the light particles ,
the momentum fraction in the normalization is of order 1 , therefore for generic values of @xmath219 , @xmath220 .
however the dominance of soft processes leads to the phase suppression and in particular the scalings of the momentum fractions @xmath221 and @xmath222 result in @xmath223 . substituting the scalings for the wave - functions , we obtain @xmath224 from which we can see the contribution from the soft process is formally @xmath116-suppressed compared to the leading power contribution from the operator @xmath63 . as a comparison ,
it is also instructive to recapture the energy dependence of the @xmath15 form factor in the scet .
@xcite finds that when matching onto scet@xmath61 the leading power contribution , from the operator @xmath225 , has two powers of @xmath196 .
together with the scalings from the two collinear quark fields , the soft form factor , parametrized via @xmath226 with @xmath227 , behaves as @xmath228 when matching to scet@xmath61 , the derivatives @xmath196 and @xmath229 also contain the momentum fractions : @xmath230 or @xmath231 for the initial heavy baryon , @xmath232 or some linear combinations depending on the fields acting on .
for example , the tree - level factorization formula from fig .
[ fig : qcdff](a ) has the following integration form as shown in eq .
@xmath233 with the assumption that @xmath234 in the limit of @xmath235 @xcite , where @xmath236 denotes the lcda of @xmath42 or @xmath2 , the integration is convergent which is different with the mesonic transition form factor @xmath237 . in fig .
[ fig : qcdff](b ) and fig .
[ fig : qcdff](f ) , the involved integral @xmath238 is finite as well .
the absence of the divergences leads to the factorization of @xmath239 @xmath240 in which @xmath241 denotes the convolution over momentum fractions @xmath242 and @xmath243 , and the jet function is given as @xmath244 it should be cautious that although this formula is valid at tree - level ( order @xmath7 ) , whether it can be extended to all orders remains unknown to us and requires further analysis . on the contrary , the subleading power corrections can not be factorized , for instance the second term from the diagram shown in fig .
[ fig : qcdff](g ) , has the form @xmath245 which is divergent when @xmath246 is approaching zero .
to have some numerical estimate , we use the qcd sum rule calculation of the @xmath47(next - to - leading order in @xmath21 ) @xcite and @xmath247 @xcite @xmath248 together with the asymptotic form of @xmath249 @xcite and the parametrized model for @xmath250 @xcite @xmath251,\nonumber\\ \phi_{\lambda}(y_1,y_2,y_3)= 120 y_1y_2y_3,\end{aligned}\ ] ] where @xmath252 , @xmath253 , @xmath254 mev , @xmath255 mev and @xmath256 @xcite . with these inputs and the strong coupling constant at the scale @xmath257 gev : @xmath258 , we calculate the form factor as @xmath259 where the displayed uncertainties are from @xmath260 . for comparison ,
we quote the soft form factor @xmath239 computed in the scet sum rules @xcite @xmath261 which is larger by about one order of magnitude .
weak decays of heavy baryons provide an ideal ground for the extraction of the helicity structure of the electroweak interaction , thanks to the spin correlation and polarization embedded in decay amplitudes . in the heavy - to - light transition ,
the most important ingredients incorporating the qcd dynamics are form factors .
due to the variety in the lorentz structures , the amplitude is governed by a number of form factors .
the development of the effective field theory allows us to simplify the form factors and pick up the terms of great importance . in this work
we have analyzed the factorization properties and power scalings of heavy - to - light baryonic form factors at large recoil . using the soft - collinear effective theory
, we proved that the form factors are greatly simplified by the heavy quark and large energy symmetries at leading power in @xmath262 .
this finding indicates that only one function is necessary to parametrize the transition of @xmath263 or @xmath12 .
a general power counting analysis indicates the form factors are of the order @xmath264 .
in contrast to the mesonic case , the leading power form factor can factorize into a convolution of a hard - scattering kernel of order @xmath7 and light - cone distribution amplitudes without encountering any divergence . using the inputs mainly from qcd sum rules
, we calculate the form factor @xmath9 and find it is numerically smaller than the one governed by soft processes , although the latter is formally power - suppressed .
we have also discussed the origins for symmetry breaking effects which are suppressed by powers of @xmath0 and/or @xmath265 .
the author is very grateful to ahmed ali for valuable discussions and carefully reading this manuscript , and to thorsten feldmann for pointing out an error in the earlier version of this manuscript .
he also thanks yu - ming wang and de - shan yang for useful discussions .
this work is supported by the alexander von humboldt foundation .
m. artuso , d. m. asner , p. ball , e. baracchini , g. bell , m. beneke , j. berryhill and a. bevan _ et al .
_ , eur .
j. c * 57 * , 309 ( 2008 ) [ arxiv:0801.1833 [ hep - ph ] ] .
n. isgur and m. b. wise , nucl .
b * 348 * , 276 ( 1991 ) .
t. mannel , w. roberts , z. ryzak , nucl .
* b355 * , 38 - 53 ( 1991 ) . c. w. bauer , d. pirjol and i. w. stewart , phys .
rev . d * 65 * , 054022 ( 2002 ) [ hep - ph/0109045 ] .
m. beneke , a. p. chapovsky , m. diehl , t. feldmann , nucl .
* b643 * , 431 - 476 ( 2002 ) [ arxiv : hep - ph/0206152 [ hep - ph ] ] .
|
in the framework of the soft - collinear effective theory , we demonstrate that the leading - power heavy - to - light baryonic form factors at large recoil obey the heavy quark and large energy symmetries .
symmetry breaking effects are suppressed by @xmath0 or @xmath1 , where @xmath2 is the hadronic scale , @xmath3 is the @xmath4 quark mass and @xmath5 is the energy of light baryon in the final state . at leading order , the leading power baryonic form factor @xmath6 , in which two hard - collinear gluons are exchanged in the baryon constituents , can factorize into the soft and collinear matrix elements convoluted with a hard - kernel of order @xmath7 . including the energy release dependence
, we derive the scaling law @xmath8 .
we also find that this form factor @xmath9 is numerically smaller than the form factor governed by soft processes , although the latter is formally power - suppressed .
* keywords : * heavy quark physics , qcd , b - physics
|
in many situations , geometric objects come parametrized by the elements of an affine space @xmath2 over a field @xmath3 or @xmath4 ; the elements of @xmath2 can usually be divided into `` generic '' and `` singular '' ( what exactly we call `` singular '' depends on the particular example we are considering ) .
the subset @xmath5 of @xmath2 formed by the singular elements is called a ( generalized ) _ discriminant_. to give an example , one can take the space of polynomials of some fixed degree over @xmath6 as @xmath2 , and the subset of @xmath2 formed by the polynomials having a multiple root in @xmath6 as @xmath5 .
notice that in this example @xmath5 is the zero locus of a polynomial in the coefficients of @xmath7 ; this polynomial is called the discriminant , which maybe justifies the use of the term `` discriminant '' in a more general situation .
more examples will be considered below . due to the alexander duality
@xmath8 computing ( additively ) the cohomology of @xmath9 is equivalent to computing the borel - moore homology groups of @xmath5 .
( recall that the borel - moore homology groups @xmath10 of a locally compact topological space @xmath11 can be defined either as the homology of the one - point compactification of @xmath11 modulo the `` infinity '' or as the homology groups of the complex of locally finite singular chains ; the latter definition generalizes to arbitrary local systems in an obvious way . )
if one is interested in the borel - moore homology groups , one can often replace @xmath5 with a much more tractable space @xmath12 called a _ resolution _ of @xmath5 .
the resolution @xmath12 comes with a natural filtration such that the difference of two consecutive terms is a fiber bundle over a `` nice '' space . in some examples
this reduces the computation of @xmath10 to ( possibly lengthy ) but elementary manipulations . in the case when the elements of @xmath5 can have only discrete singular loci ,
such resolutions are called _ simplicial _ ; they were first constructed by v. a. vassiliev in @xcite , although one can trace the idea ( as well as in fact , a large part of homological algebra ) all the way back to euler s `` inclusion - exclusion '' formula ) .
conical resolutions were introduced by v. a. vassiliev in @xcite as a generalization of simplicial resolutions for the case of non - discrete singular sets ( see also @xcite ) . in @xcite v.
a. vassiliev applied conical resolutions to compute the rational cohomology of the spaces of nonsingular degree @xmath13 hypersurfaces in @xmath14 for @xmath15 and @xmath16 . in order to overcome some difficulties involved in extending the computations performed in @xcite to other examples , the author gave in @xcite a modified construction of conical resolutions based on the inclusion relations between the singular loci , rather than between the corresponding linear systems .
this modified construction was presented from the viewpoint of cubical schemes by o. tommasi @xcite .
the approach of o. tommasi enables one to recover the mixed hodge structures on the borel - moore homology groups of the discriminant . in this paper
we generalize the constructions of @xcite and @xcite .
the point of generalizing can be explained as follows .
natural filtrations on the conical resolutions of @xcite typically contain a lot of terms corresponding to `` fake '' singular loci .
this does not present a problem in @xcite , since a vast majority of these terms makes no contribution ( over @xmath17 ) to the spectral sequence corresponding to the filtration .
however , if we consider other examples ( and in particular , the spaces of nodal hypersurfaces we shall be interested in in this paper ) , the situation turns out to be different . while it is still true that most columns of term @xmath18 of the spectral sequence are zero , many of the corresponding columns of @xmath19 are not ( even rationally ) .
we address the problem by putting together some parasitic terms of the filtration that are destined to kill each other in the end .
this procedure may be seen as an adaptation of v. a. vassiliev s construction of reduced conical resolutions @xcite to the context of @xcite .
we apply our method to compute the rational cohomology ( together with the mixed hodge structures ) of the spaces of equations of nodal cubics in @xmath20 , and of the space of nodal quartics in @xmath0 .
mixed hodge structures can be encoded as follows . for a complex algebraic variety @xmath11
, we define the _ mixed hodge polynomial _ of @xmath11 to be @xmath21 by setting in this expression @xmath22 , respectively , @xmath23 , we get the poincar - serre polynomial , respectively , the poincar polynomial , of @xmath2 .
we denote by @xmath24 the space of homogeneous degree @xmath13 complex polynomials in @xmath25 variables , and we set @xmath26 to be the subvariety of @xmath24 formed by all @xmath27 such that the kernel of the hessian matrix of @xmath27 at a nonzero point @xmath28 contains a 2-plane @xmath29 ( or , equivalently , the hypersurface defined by @xmath27 has a singularity other than a simple node ) .
[ mhcubp2 ] we have @xmath30 [ mhquartp2 ] we have @xmath31 [ mhcubp3 ] we have @xmath32 we use theorems [ mhquartp2 ] and [ mhcubp3 ] to compute the mixed hodge polynomials of the corresponding moduli spaces .
[ mhqmod ] we have @xmath33 [ mhcmod ] we have @xmath34 let us note that the moduli space @xmath35 of nodal cubic surfaces was studied in @xcite from the viewpoint of complex hyperbolic geometry ; in particular , it was shown there that this space is the quotient of the complex hyperbolic 4-space by a certain discrete group @xmath36 acting with finite stabilizers .
this means that @xmath35 is a `` rational @xmath37 '' , i.e. , the rational cohomology of @xmath36 is canonically isomorphic to the rational cohomology of @xmath35 .
the paper is organized as follows : in section [ conres ] we construct conical resolutions of discriminants . in section [ nod ]
we show how the method of section [ conres ] applies to the equation spaces of nodal hypersurfaces and prove theorem [ mhcubp2 ] .
the proofs of theorems [ mhquartp2 ] and [ mhqmod ] are given in section [ quartics ] .
finally , in section [ cubics ] we prove theorems [ mhcubp3 ] and [ mhcmod ] . in the following table
we summarize some of the notation used in the paper . ' '' ''
@xmath38 & the symmetric group on @xmath39 letters + ' '' '' @xmath40 & [ cols="^ " , ] + ( here we give @xmath41 only for @xmath42 that consist of 3 elements ; two - element configurations from @xmath43 can be considered in the same way as elements of @xmath44 . )
the verification of condition [ five ] is now straightforward @xmath45 .
now we apply theorem [ thconical ] to construct a conical resolution @xmath12 of @xmath5 and a filtration @xmath46 theorem [ mhquartp2 ] follows immediately from [ spseqqvector ] let @xmath47 be the spectral sequence over @xmath17 corresponding to this filtration
. then the groups @xmath48 are given by table [ spseqquart ] .
[ spseqquart ] @xmath49 the differetials @xmath50 and @xmath51 are non - zero , and @xmath52 .
* we have @xmath53 , and @xmath54 is a complex bundle over @xmath55 of rank 10 .
hence the groups @xmath56 are as shown on table [ spseqquart ] .
for @xmath57 we have @xmath58 hence , the space @xmath59 is a fiber bundle over @xmath60 with fiber homeomorphic to the open cone over @xmath61 .
the space @xmath62 is a rank 9 complex vector bundle over @xmath59 ; this implies the statement of the theorem concerning the groups @xmath63 .
the groups @xmath64 can be computed in the same way .
for @xmath65 the space @xmath66 is equal to @xmath67 , where @xmath68 and @xmath69 are the maximal subconfigurations of @xmath70 that belong to @xmath60 and @xmath71 respectively .
notice that @xmath72 is a configuration of type [ q1 ] and @xmath73 .
since @xmath74 and @xmath75 are contractible , all groups @xmath76 are zero . for @xmath77
we have @xmath78 with @xmath79 .
hence , @xmath80 , where @xmath81 is the local system on @xmath82 induced from the alternating system @xmath83 by the inclusion @xmath84 .
the groups @xmath85 can be obtained from the fact that @xmath86 , if @xmath87 and is zero otherwise @xcite .
the space @xmath88 can be represented as @xmath89 where @xmath68 and @xmath68 are maximal subconfigurations of @xmath70 of type [ q4 ] , and @xmath90 is the union of @xmath91 for all @xmath92 of type [ q5 ] .
it is easy to see that @xmath67 has the homology of a point , hence @xmath93 the groups @xmath94 can now be computed in the same way as @xmath85 . for any @xmath95
we have a filtration on @xmath66 induced by the intersections with @xmath96 and @xmath97 .
it is easy to see that the borel - moore homology groups of the difference of any two consecutive terms of this filtration vanish , hence @xmath98 .
the eighth column of table [ spseqquart ] is obtained in exactly the same way as the sixth one . if @xmath99 , then @xmath100 is homeomorphic to the second autojoin of @xmath61 and hence , has the rational homology groups of a point @xcite .
this implies that @xmath101 for all @xmath102 .
let us now consider the tenth column of table [ spseqquart ] .
as in the case of column [ q6 ] , the groups @xmath103 are isomorphic to shifted groups @xmath104 where @xmath105 is the pullback of the local system @xmath83 on @xmath106 under the inclusion @xmath107 . the configuration space @xmath108 is fibered over @xmath109 . using this fibration and the fact that @xmath110 @xcite , it is easy to see that @xmath111 . finally , to obtain the last column of table [ spseqquart
] , we have to compute the rational borel - moore homology of the open cone over the space @xmath112 this space can be filtered as follows : @xmath113 in the first term of the spectral sequence corresponding to this filtration , the @xmath102-th column , @xmath114 , is equal to the @xmath115-th column of table [ spseqquart ] , with all weights and dimensions shifted down by @xmath116 .
the rational borel - moore homology of @xmath117 is clearly isomorphic to @xmath118 , and the space @xmath119 is fibered over @xmath109 with fiber @xmath120 since all @xmath121 s are contractible , we have @xmath122 for @xmath123 and @xmath124 otherwise , hence only the first three columns of the first term of the spectral sequence corresponding to ( [ filtr ] ) are nonzero , and they look as follows .
@xmath125 comparing this with what we already know about the spectral sequence represented on table [ spseqquart ] , we see that if any of the potentially nonzero differentials of ( [ lastcolq ] ) were zero , then @xmath126 would have a nonzero rational cohomology group in dimension @xmath127 .
this is impossible ; indeed , the group @xmath128 acts on @xmath126 with finite stabilizers , and the ( geometric ) quotient is a noncomplete variety of dimension 6 . hence , the nonzero borel - moore homology groups of the space ( [ lambdaq ] ) are @xmath129 and @xmath130 in dimensions 0 , 10 and 12 respectively .
we have thus finished computing the groups @xmath131 . to complete the proof of theorem [ spseqqvector ] ,
it remains us to compute the differentials @xmath132 .
this is done in the next few lemmas .
[ difq1 ] the differentials @xmath133 and @xmath134 are nonzero .
* proof of lemma [ difq1 ] . *
the space @xmath135 is fibered over @xmath136 ; denote the corresponding fiber by @xmath137 .
the statement of the lemma would follow , if we prove that @xmath138 if @xmath139 and @xmath140 otherwise .
take a configuration @xmath57 .
set @xmath141 , let @xmath142 be a 3-ball with boundary @xmath11 , and let @xmath143 be the complex vector bundle over @xmath11 with total space @xmath144 the bundle @xmath143 contains a constant subbundle with total space @xmath145 ; denote by @xmath146 the corresponding quotient line bundle , and let @xmath147 be the total space of @xmath146 .
finally , let @xmath148 be the result of attaching @xmath142 to @xmath147 via the zero section embedding @xmath149 . to prove the above claim about @xmath150
, it suffices to show that @xmath151 for @xmath152 , which can be done as follows .
it is easy to see that if we identify @xmath11 with @xmath61 , then @xmath146 becomes @xmath153 , hence we obtain ( recalling that @xmath154 is the total space of @xmath146 minus the zero section ) that @xmath155 and @xmath156 for @xmath157 . using the exact sequence of the couple @xmath158 , we conclude that @xmath151 for @xmath152 .
the lemma is proved.@xmath45 [ difq2 ] the differentials @xmath159 and @xmath160 are nonzero . *
proof of lemma [ difq2 ] .
* we proceed as in the proof of the previous lemma .
we shall compute the rational borel - moore homology groups of @xmath161 using a filtration a bit different from the one induced by @xmath162 .
namely , consider the filtration @xmath163 where @xmath164 ( we identify @xmath11 with a subspace of @xmath165 via the `` zero section embedding '' @xmath166 ) .
it is easy to deduce from spectral sequence ( [ lastcolq ] ) that @xmath167 and @xmath168 for @xmath169 .
the third and the fourth columns of the rational borel - moore homology spectral sequence corresponding to ( [ difq2filtr ] ) are zero , since ( as we saw above ) @xmath170 . to obtain the second column , we have to know the groups @xmath171 the space @xmath172 coincides obviously with @xmath173 minus the zero section of the bundle @xmath174 .
recalling the way we computed above the borel - moore homology of @xmath173 , we see that @xmath171 is isomorphic ( with dimension shifted by 6 and weight by 4 ) to @xmath175 , where @xmath147 is the space formed by all couples @xmath176 such that @xmath27 vanishes with multiplicity at least 2 at each point of @xmath70 , and the local system @xmath81 is the pullback of @xmath83 under the obvious projection @xmath177 .
let @xmath148 be the total space of the pullback of the fiber bundle @xmath177 to @xmath178 .
the group @xmath179 acts on @xmath148 in a natural way , and @xmath180 is isomorphic to the @xmath179-anti - invariant part of @xmath181 .
let us compute @xmath181 together with the action of @xmath179 on it . by the poincar duality
, it suffices to consider @xmath182 .
it is easy to see that @xmath183 is generated by the pullbacks @xmath184 and @xmath185 of the canonical generator of @xmath186 under the obvious projections @xmath187 .
the classes @xmath184 and @xmath185 satisfy the relations @xmath188 , and any other relation in @xmath183 is a consequence of these .
the line bundle over @xmath189 with total space @xmath190 is clearly isomorphic to @xmath191 , hence @xmath148 can be viewed as the space of nonzero vectors of the restriction of @xmath192 to @xmath178 , where @xmath193 and @xmath194 are the pullbacks of @xmath195 under the projections @xmath196 .
the first chern class of @xmath192 is @xmath197 ; therefore , the differential @xmath198 of the leray spectral sequence of the bundle @xmath199 is given by the formula @xmath200 , where @xmath201 , and @xmath202 is the canonical generator of @xmath203 .
an easy check shows now that the @xmath179-anti - invariant part of @xmath182 is @xmath17 in dimensions 2 and 7 and zero in all other dimensions .
this implies @xmath204 , and @xmath205 for @xmath206 .
hence , the space @xmath172 has nonzero borel - moore homology groups only in dimensions 9 and 14 , and these groups are equal respectively to @xmath207 and @xmath208 .
now we can write the first term of the spectral sequence corresponding to ( [ difq2filtr ] ) that converges to @xmath209 .
@xmath210 comparing this with the last four columns of table [ spseqquart ] , we see that @xmath211 , all other groups @xmath212 are zero , and the differentials @xmath159 and @xmath160 of the spectral sequence of theorem [ spseqqvector ] are nonzero .
the lemma is proved.@xmath45 lemmas [ difq1 ] and [ difq2 ] will suffice ( together with table [ spseqquart ] ) to determine the cohomology of the quotient space @xmath213 .
however , to complete the proof of theorem [ spseqqvector ] , we must compute the remaining differentials @xmath132 . in the case of @xmath214 and @xmath51 ,
the argument is analogous to the one used above to prove lemmas [ difq1 ] and [ difq2 ] .
surprisingly , the computation of @xmath215 is a lot more difficult .
[ difq3 ] the differentials @xmath214 and @xmath51 are nonzero .
* proof . * set @xmath216 and define @xmath147 as the space formed by all couples @xmath217 such that @xmath218 for some @xmath219 , and @xmath220 , where @xmath221 is the unique configuration from @xmath222 that contains @xmath70 .
the space @xmath147 is a vector bundle over @xmath222 with fiber @xmath223 , where @xmath219 and @xmath224 are the configurations from @xmath225 that are contained in @xmath70 . since @xmath74 and @xmath226 are contractible spaces , we have @xmath227 .
denote by @xmath143 and @xmath146 the fiber bundles over @xmath82 with total spaces @xmath228 and @xmath229 respectively .
the complement @xmath230 is a fiber bundle over @xmath82 with fiber over @xmath77 homeomorphic to the product @xmath231 , where @xmath232 and @xmath233 are the fibers respectively of @xmath143 and @xmath146 over @xmath70 .
set @xmath148 to be the space formed by all elements @xmath234 such that @xmath235 and @xmath236 .
( notice that @xmath148 is homotopy equivalent to @xmath237 . )
the quotient of the obvious action of the symmetric group @xmath179 on @xmath148 can be identified with @xmath82 ; let @xmath238 and @xmath239 be the pullbacks respectively of @xmath143 and @xmath146 to @xmath148 .
set @xmath240 to be the pullback of the canonical generator of @xmath241 under the projection @xmath242 given by @xmath243 notice that @xmath244/\langle a_1 ^ 3,a_2 ^ 3,a_1 ^ 2+a_2 ^ 2+a_1a_2\rangle.\ ] ] by an argument analogous to the one used in the proof of lemma [ difq2 ] , the differentials @xmath214 and @xmath51 are nontrivial if and only if @xmath168 for @xmath245 , which in turn is equivalent to @xmath246 with @xmath247 .
lemma [ difq3 ] would follow thus if we prove [ chernxieta ] we have @xmath248 * proof of proposition [ chernxieta ] .
* let us compute the total chern class of @xmath238 .
let @xmath249 be the vector bundle over @xmath250 with total space @xmath251 and set @xmath252 to be the ( restriction to @xmath148 of the ) projection @xmath253 we have an exact sequence of vector bundles over @xmath148 : @xmath254 hence , to compute @xmath255 it is enough to know @xmath256 and the cohomology mappings induced by the compositions of the inclusion @xmath257 with @xmath252 .
[ aux ] we have @xmath258 * proof of proposition [ aux ] . *
the identity @xmath259 for @xmath260 is clear . to prove the remaining two formulas ,
let us first consider the projections @xmath261 given by @xmath262 and @xmath263 let @xmath264 and @xmath265 be the canonical generators , and denote by @xmath266 is the mapping taking @xmath267 to the line spanned by @xmath268 and @xmath269 . as we saw in the proof of lemma [ chernclass ] , we have @xmath270 the space @xmath148 can be identified with the space @xmath271 of ordered triples of non - aligned points in @xmath109 via @xmath272 let @xmath273 be the pullback of @xmath274 under @xmath275 .
it follows from ( [ auxeq ] ) that @xmath276 .
the group @xmath277 acts on @xmath278 so that @xmath279 . since @xmath278 has dimension 2 and is spanned by @xmath280 we have @xmath281 , which completes the proof of the proposition.@xmath45 it follows immediately from propositions [ aux ] , lemma [ chernclass ] and the exact sequence ( [ exact ] ) that @xmath282 let us now compute the chern class of @xmath239 .
an element of the fiber of @xmath239 over @xmath283 has the form @xmath284 where @xmath285 , respectively , @xmath286 defines a conic , respectively , a line through @xmath287 and @xmath288 .
in fact , it is easy to see that the bundle @xmath239 can be represented as @xmath289 where @xmath290 and @xmath291 are the vector bundles with total spaces @xmath292 and @xmath293 respectively .
the exact sequence @xmath294 implies @xmath295 analogously , we have @xmath296 ( the second equalities in both formulas are due to @xmath297 ; in fact , this relation , which is easily checked directly , can also be deduced from @xmath298 and the fact that @xmath291 is a line bundle . )
since the rank of @xmath290 is 4 , we have @xmath299 ( and @xmath300 is obviously 0 ) . from ( [ chernxi ] ) and ( [ cherneta ] ) we obtain @xmath248 proposition [ chernxieta ]
is proved ( and hence , so is lemma [ difq3]).@xmath45 [ difq4 ] the differential @xmath301 is nonzero . *
d. allcock , j. carlson , d. toledo , `` the complex hyperbolic geometry of the moduli space of cubic surfaces '' , j. algebraic geom . 11 , 2002 , no .
4 , 659724 , arxiv : math.ag/0007048 .
a. g. gorinov , `` real cohomology groups of the space of nonsingular curves of degree 5 in @xmath109 '' , ann .
toulouse math .
, 14 , no . 3 , 2005 , 395 - 434 , arxiv : math.at/0105108 . o. tommasi , `` rational cohomology of the moduli space of genus 4 curves '' , compos
. math . 141 , 2005 , no .
2 , 359384 , arxiv : math.ag/0312055
. v. a. vassiliev , `` topology of spaces of functions without complex singularities '' ( in russian ) funktsional .
i prilozhen . 23 , 1989 , no .
4 , 2436 , 96 ; translation in funct . anal .
appl . 23 , 1989 , no .
4 , 277286 . v. a. vassiliev , `` geometric realization of the homology of classical lie groups , and complexes that are @xmath303-dual to flag manifolds '' ( in russian ) . algebra i analiz 3 , 1991 , no .
4 , 113120 , translation in st .
petersburg math .
j. 3 , 1992 , no .
4 , 809815 .
v. a. vassiliev , `` topology of discriminant complements '' , phasis , moscow , 1997 ( in russian ) .
v. a. vassiliev , `` how to calculate homology groups of spaces of nonsingular algebraic projective hypersurfaces in @xmath14 '' , proc .
steklov math .
inst . , 1999 , 225 , 121140 .
|
we present a modification of the method of conical resolutions @xcite .
we apply our construction to compute the rational cohomology of the spaces of equations of nodal cubics in @xmath0 , nodal quartics in @xmath0 and nodal cubics in @xmath1 . in the last two cases we also compute the cohomology of the corresponding moduli spaces .
|
wave interference in disordered scattering media results in speckles through the coherent addition of multiple waves , which are independent and have random amplitudes and phases @xcite . between these interfering waves ,
there exist short- , long- , and even infinite - range correlations @xcite .
these correlations have provided enriching information about mesoscopic transport , as well as a deeper understanding of fundamental phenomena such as enhanced backscattering @xcite and anderson localization @xcite . in 1990 , using speckle correlations , freund predicted that an opaque scattering medium can be used as a lens and other optical elements by designing an appropriate incident wavefront @xcite . only recently , this prediction was confirmed by the advent of innovative wave - shaping methods such as wavefront shaping @xcite , time reversal @xcite , phase conjugation @xcite , and transmission - matrix - based control @xcite . in wavefront
shaping , an optimization algorithm receives as a feedback the intensity in a target area , typically one speckle spot with an area @xmath0 .
the algorithm then modifies the spatial phase of the incident field on the scattering medium , such that the intensity in the target spot is maximized .
these wave - shaping methods have led the way for exciting applications such as non - invasive biomedical imaging @xcite , advanced optics @xcite , and cryptography and secure communication @xcite . in an earlier experiment
@xcite , it was observed that there is not only an intensity enhancement in the target speckle spot but also in an area outside the target speckle spot .
consequently , the total transmission was enhanced , even though only the intensity in a single speckle spot was monitored .
an intuitive explanation for this observation is that there is a redistribution of energy between reflection and transmission channels , since absorption is negligible in the scattering samples .
this observation was confirmed in ref .
@xcite . here
, we take a further step by investigating how the enhancement of the total transmission depends on the optimization area . in the absence of absorption
, we expect to observe a concomitant effect in the total reflected intensity .
moreover , we expect to find the effect of long - range correlations , especially of the form that exists between the reflected and transmitted speckles , as recently predicted in ref .
an optimization of the total intensity transmitted through a scattering medium , which is the extreme case of our study , has been performed in ref .
@xcite although the optimization area was not systematically varied .
the dependency that we seek will give insight to the intensity redistribution between the transmitted and reflected speckles .
such a fundamental understanding is useful for applications of wavefront shaping in efficient energy harvesting in solar cells @xcite , multi - channel quantum secure communications @xcite , imaging @xcite , and the delivery of complex beam through a scattering medium @xcite . in this paper
, we experimentally study how the optimized intensity on both interfaces of a scattering medium depends on the optimization area on the transmission side .
we imaged the transmitted intensity onto the chip of a camera , and thus , there is a freedom to control the optimization radius . we observe that as the optimization radius increases , the enhancement of the total transmitted intensity increases .
we find a concomitant decrease of the total reflected intensity , which implies that there is a redistribution of intensity from reflection to transmission .
in addition , our result reveals qualitative evidence of the long - range reflection - transmission correlation .
the experimental set - up is shown in figure [ fig : setup ] .
the light source is a helium neon ( hene ) laser , which emits at a wavelength @xmath1 nm and has an output power of 5 mw .
a combination of a half - wave plate ( @xmath2 ) and a polarizer ( p ) controls the incident power and also fixes the polarization of the beam .
a beam expander with a magnification of @xmath3 expands the beam to fill the active area of the spatial light modulator ( slm ) .
the spatial light modulator is a holoeye pluto ( 6010-nir-011 ) , which has 1920 @xmath4 1080 pixels and controls the horizontal polarization .
a beam splitter ( bs ) separates the incident and reflected light from the slm .
the reflected light is focused by a lens l1 ( focal length f = 250 mm ) .
aperture a , placed at the focal plane of the lens l1 , filters out the higher diffraction orders of the slm s pixels and transmits only the @xmath5 order . with a telescope consisting of lenses l1 and
l2 ( f = 250 mm ) , the slm is imaged onto the pupil of a microscope objective mo1 ( zeiss : infinity corrected , 63@xmath4 , na = 0.95 ) , which focuses the light onto the surface of the sample .
the sample is an ensemble of disordered zinc oxide nanoparticles that are spray - painted on top of a glass cover slide .
the sample thickness is 17 @xmath6 0.2 and the transport mean free path @xmath7 of similar samples was reported in ref .
@xcite to be @xmath8 0.2 .
the intensity transmitted through the sample is imaged onto the chip of a charged - coupled device ( ccd ) camera ( ccd1 ) using a combination of an oil - immersion objective mo2 ( olympus : infinity corrected , 60@xmath4 , na = 1.42 ) and lens l3 ( f = 500 mm ) .
the calculated magnification of imaging the back surface of the sample ( m1 ) is @xmath9 .
similarly , a combination of the focusing objective mo1 and lens l2 images the reflected intensity exiting the front surface of the sample onto the chip of a ccd camera ( ccd2 ) .
the calculated magnification on the reflection side is @xmath10 .
the cameras ccd1 and ccd2 are both of the same type ( avt dolphin 145b ) , with a pixel pitch of 6.45@xmath11 m . using the calculated magnifications , the pixel pitches on the front and back surfaces of the sample are @xmath12 m and @xmath13 m respectively .
the optical density filter odf1 ( thorlabs ne05a ) , with a measured attenuation factor af = 0.33 , attenuates the incident intensity on the sample , in order to prevent saturation of the cameras .
the reflected intensity is further attenuated by placing odf2 ( thorlabs ne10a ) , with a measured af = 0.10 in the reflection path .
the reflected intensity is in total attenuated by a factor of 0.033 . in the set - up of ref .
@xcite , it is impossible to control the optimization area since the scattering sample was directly attached to a photodetector . moreover , as a result of the refractive index contrast ( approximately a factor of 2 ) , the detected signal in ref .
@xcite is limited by significant surface reflections between the scattering sample and photodector . with our set - up
, we have the freedom to control the optimization area .
a further advantage of our study is that there is no significant surface reflections since there is an index match between the sample substrate and the immersion oil . in order to optimize multiple speckle spots , we used the partitioning algorithm , which is described in ref .
@xcite rather than the stepwise sequential algorithm , which is typically used in previous wavefront shaping experiments @xcite .
we find that the partitioning algorithm outperforms the stepwise sequential and genetic algorithms for optimizing intensity in large areas ( see appendix ) . in the optimization ,
the number of segments used is 5000 .
we systematically increased the number of transmission channels by increasing the optimization radius @xmath14 .
the number of open transmission channels @xmath15 scales linearly with the probed area @xmath16 @xmath17 where @xmath18 is the effective refractive index of the scattering medium @xcite . for a specific optimization radius
, we repeated the wavefront shaping experiment for 3 to 5 times at a fixed position on the sample .
as a reference phase pattern , we sent 100 randomly generated patterns , with the same number of segments as the optimized pattern .
compared to the optimized pattern , these randomly generated phase patterns have diffraction losses and a power incident on the sample that is larger by only 5@xmath19 , which underestimates the intensity enhancement by this amount . in order to quantify the enhancement @xmath20 of the total intensity in the target area , we define @xmath21 following refs . @xcite . @xmath22 and @xmath23 are the power in the target area with the optimized and unoptimized patterns , respectively .
@xmath24 denotes an ensemble - averaged power over the 100 different random phase patterns .
we also quantified the enhancement @xmath25 of the total transmitted intensity and the total reflected intensity as @xmath26 where @xmath27 for transmitted light , @xmath28 for reflected light , @xmath29 and @xmath30 are the total transmitted power through the sample with the optimized and unoptimized patterns respectively and @xmath31 , and @xmath32 are the total reflected power through the sample with the optimized and unoptimized patterns respectively .
the enhancement of the intensity outside the optimization area is quantified as @xmath33
= 4.72@xmath34 m , which is indicated by the red dashed circle .
, scaledwidth=45.0% ] in figs .
[ fig : ccdimages ] ( a ) and ( b ) , we show the ccd camera images of the transmitted intensity for the unoptimized and optimized incident wavefronts , respectively . in the wavefront shaping experiment shown in fig .
[ fig : ccdimages ] ( b ) , the optimization radius is 4.7@xmath11 m , which corresponds to 121 pixels .
the intensity in the optimization area increases significantly compared to the unoptimized intensity . the intensity outside the target area increases as well . for a better visualization of the intensity increase , we plot in fig .
[ fig : transmtintensity_ring ] the radial distributions of the transmitted intensity .
the radial distribution is obtained by summing the intensities within a ring with a width @xmath35 and an inner radius of @xmath36 ( see inset in fig . [
fig : transmtintensity_ring ] ) .
angular averaging helps to reduce the intensity fluctuation from the speckle pattern .
there is a significant intensity increase in the optimization area for both optimization radii @xmath14 = 15.2@xmath11 m and @xmath14 = 4.7@xmath34 m .
this intensity increase is expected since the intensity in the optimization area is the feedback to the partitioning algorithm .
the intensity outside the optimization area remarkably increases as well .
this intensity increase agrees with the observation in ref .
@xcite , where the intensity outside the optimization area was observed to increase as well .
we quantify the intensity enhancement inside and outside optimization areas , and the total transmitted intensity in the next sections .
we plot in fig .
[ fig : eta_target ] the intensity enhancement @xmath20 in the optimization area versus the optimization radius @xmath14 .
we obtained a maximum enhancement of 257@xmath4 for an optimization radius of about 0.3@xmath11 m , which corresponds to one speckle spot .
as the optimization radius increases , @xmath20 decreases drastically down to 2@xmath4 at the largest radius @xmath37 15.2@xmath38 m , which corresponds to 1716 open transmission channels . in ref .
@xcite , the enhancement for a multiple channel optimization was found to be inversely proportional to the number of speckle spots @xmath15 in the optimization area . in the model in @xcite , it is assumed that the optimized intensity is distributed equally to all optimized channels , which are considered statistically independent . from the model , the enhancement is expected to depend inversely on the square of the optimization radius @xmath14 , @xmath39 where @xmath40 is a constant that depends on the number of effectively controlled input channels on the sample . in fig .
[ fig : eta_target ] , we show a nonlinear least squares fit to the experimental data using eq . [ eq : eta ] and @xmath40 is the only adjustable parameter .
weighing all data points equally , the fit agrees well with the first two optimization radii @xmath14 = 0.3@xmath11 m and 1@xmath34 m that have 1 and 7 transmission channels respectively .
strikingly , the fit deviates from the experimental data by about a factor of 2 at large radii .
this deviation signifies that there is more intensity available in the optimization channels , than that predicted according to eq .
[ eq : eta ] , especially at large radii .
we discuss three possible reasons for the deviation of the theory from the experimental data .
first , it is known that noise has a significant effect on the single - speckle optimization @xcite .
our wavefront shaping experiments are in the regime where shot noise is much higher than the camera and laser noise that are about 1@xmath19 and 2@xmath19 respectively . according to ref .
@xcite , in this shot noise regime the enhancement of a single - speckle optimization is expected to scale linearly with the total intensity in the optimization area @xmath41 . extending this model to the optimization of multiple channels , we derive @xmath42 where @xmath43 , where @xmath44 is a pre - factor in the relationship @xmath45 .
a constant enhancement with radius obviously does not describe our experimental data , hence we reject this hypothesis .
second , the observed increased enhancement might be due to intensity redistributed from the speckles outside the optimized area to speckles inside the optimized area .
if this is the case , then the total transmitted intensity would be constant for all optimization radii .
a third hypothesis is that there is a redistribution of intensity from the reflected speckles to the transmitted speckles . in this case
, the effect of enhancing the transmitted intensity is expected to be noticeable on reflection as a reduction of the reflected intensity .
we will check these later two hypotheses in the next section .
the enhancements of the total transmitted intensity @xmath46 , the total reflected intensity @xmath47 , and the intensity outside the optimization area @xmath48 versus the optimization radius are shown in fig .
[ fig : eta_total ] .
the enhancement of the total transmitted intensity gradually increases from about 1.1 to 2 at the largest radius of 15.2@xmath11 m , which is close to the @xmath49 m large size of the detected speckle pattern .
there is also an enhancement of the intensity outside the optimization area and it is about @xmath50 at large radii . the enhancements of both the total transmitted intensity and the intensity of area outside the optimization nullifies the second hypothesis .
in contrast to the transmitted intensity , the enhancement of the total reflected intensity slowly decreases to @xmath51 as the optimization radius increases .
the decrease of the @xmath47 is not as rapid as the increase of @xmath52 because the reflected intensity is about nine times higher than the transmitted intensity .
therefore , a large intensity enhancement on transmission corresponds to a small intensity enhancement on reflection . in order to accurately compare the decrease of @xmath47 with the increase of @xmath52
, we need to know the dependence of both terms on the optimization radius .
the dependence of the enhancements on the optimization radius is unknown and we find that a power - law @xmath53 describes the experimental data well . here ,
@xmath54 and @xmath55 are adjustable parameters .
the fits to the experimental data are shown in fig .
[ fig : eta_total ] and the values of @xmath54 and @xmath55 obtained from the fits are shown in table [ table : table1 ] .
we obtained @xmath56 and @xmath57 for the enhancement of the total transmitted and reflected intensities , respectively , and these values are in remarkable mutual agreement .
.the values of the adjustable parameters b and n obtained by fitting eq .
[ eq : eta_total ] to the experimental data shown in fig .
[ fig : eta_total ] for the total transmission enhancement @xmath52 , total reflection enhancement @xmath47 and the enhancement of the intensity outside the target area @xmath48 [ cols="^,^,^",options="header " , ] .
[ table : table1 ] in fig .
[ fig : intensitychange ] , we plot the absolute changes in the transmitted and the reflected intensities after accounting for the attenuation factor of the nd filters .
the absolute changes in the transmitted and reflected intensities both collapse to the same curve , modeled with @xmath58 .
therefore , the enhancement of the transmitted intensity corresponds to a decrease in the reflected intensity .
this validates the third hypothesis that there is a redistribution of intensity from the reflection speckles to the transmission speckles of the scattering medium . at this point ,
the question arises : how does the reflected intensity change spatially ? is the change global , _
i.e. , _ does the overall reflected intensity decrease uniformly , or local , _
i.e. , _ does the intensity decrease more in the area where the transmission is enhanced ? as illustrated in fig .
[ fig : cartoon ] , a global change of the reflected intensity implies that the intensity redistributed to transmission is deducted equally from all the spatial channels .
this is expected if all input spatial channels contribute equally to all the output spatial channels . on the other hand
, a local change implies that the intensity of the spatial input channels maps one - to - one with that of output spatial channels .
the local change is expected as a result of the reflection - transmission long - range correlation predicted in ref .
@xcite . in order to observe the type of change , we plot the radial distribution of the reflected intensity in figs . [
fig : reflintensityring ] ( a ) and ( b ) for optimization radii of 15.2@xmath11 m and 8.4@xmath11 m respectively . in both optimization radii ,
the optimized ( red circles ) and unoptimized ( blue squares ) intensities matches quite well from @xmath59 to about @xmath60 5@xmath11 m .
the optimized intensity deviates asymmetrically from the unoptimized intensity between @xmath60 5@xmath11 m and 10@xmath11 m . at @xmath61 m ,
both intensities become equal and decrease in the same way . for a proper interpretation of these results ,
we describe the global and local change in the reflected intensity as follows .
first , we model the unoptimized reflected intensity @xmath62 with a 2d circular gaussian function @xmath63 here , @xmath64 is the amplitude , @xmath65 is the spatial coordinate , @xmath66 is the coordinate of the center , and @xmath67 is the standard deviation that defines the width of the function . in order to model both the global and local change , we define a second 2d circular gaussian function @xmath68 @xmath69 where @xmath70 is the displacement of the center of @xmath68 from @xmath71 and @xmath72 respectively and @xmath73 is the standard deviation of @xmath68 .
we model the optimized intensity @xmath74 , in the case of the global change , as the difference between @xmath62 and @xmath68 , when @xmath75 and @xmath76 , to get i_g ( ) & = ( - ( ) ) + & - ( - ( ) ) .
[ eq : i_global ] to model the optimized intensity @xmath77 , in case of a local change , we use the difference between @xmath62 and @xmath68 , when @xmath78 and @xmath76 , to get i_l ( ) & = ( - ( ) ) + & - ( - ( ) ) .
[ eq : i_local ] using eqs . [ eq : unoptimized ] , [ eq : i_global ] and [ eq : i_local ] , 2d gaussian functions were calculated for unoptimized light , and optimized light for either global or local changes , respectively .
a projection of the generated functions onto the x - axis is shown in the bottom left and right insets of fig .
[ fig : cartoon ] . following the same procedure for the analysis of the experimental data , we obtain the radial distribution of the calculated functions by integrating the intensity along the circumference of a ring of width of @xmath35 and inner radius of @xmath36 . in figs .
[ fig : reflintensityring ] ( c ) and ( d ) , we plot the calculated intensity versus radius for both the global and local change . to obtain the curves in figs .
[ fig : reflintensityring ] ( c ) and ( d ) , we used these parameters : @xmath79 m , as obtained by fitting a gaussian function to the unoptimized reflected intensity ; @xmath80 and @xmath81 .
for both global and local changes , the values of @xmath64 and @xmath82 are chosen such that the ratio of the total area under the calculated optimized and unoptimized functions is 0.95 , which corresponds to a comparable enhancement on reflection in fig .
[ fig : eta_total ] . both @xmath71 and @xmath72
were chosen to be 45.5@xmath34 m , which is exactly at the center of the generated 2d gaussian function .
the adjustable parameters are @xmath83 m , and @xmath84 m and we will comment on these values in the next paragraph . for the global change shown in figs .
[ fig : reflintensityring ] ( c ) , there is a symmetric deviation of the reflected intensity from the unoptimized light and the deviation is centered at the peak position near @xmath85 m . at radial positions between @xmath59 and @xmath86 m and @xmath87
m and @xmath88 m , optimized light matches with unoptimized light .
these features of the calculated global change do not correspond with the features of the experimental data shown in figs .
[ fig : reflintensityring ] ( a ) and ( b ) . for the local change shown in fig .
[ fig : reflintensityring ] ( d ) , there is interestingly an asymmetric deviation of the optimized light from the unoptimized light between @xmath89 m and @xmath90 m .
this asymmetric deviation is very similar to what is observed for the two optimization radii in figs .
[ fig : reflintensityring ] ( a ) and ( b ) .
our experiment results therefore indicate that there is a local change in the reflected intensity rather than a global change .
we now comment on the values of the adjustable parameters . in the experiment , @xmath91 m and @xmath92 m and we have used @xmath93 m in figs . [ fig : reflintensityring ] ( c ) and
( d ) , in order to have a similar asymmetric deviation in the local change .
it should be noted that the global change is almost independent of @xmath94 , since @xmath95 ( see eq .
[ eq : i_global ] ) .
the discrepancy between the experiment values of ( @xmath96 , @xmath97 ) and the adjusted values might be the result of the thermal drift of the slm , laser , mechanical stage onto which the sample is mounted and other apparatuses . from the experimental and adjusted values of ( @xmath96 , @xmath97 ) , we estimate that the beam is displaced by approximately @xmath85 m , which is conceivable since the measurements took several days .
consequently , due to the beam displacement of @xmath98 m , the optimization area mapped onto the reflection light is almost outside the reflected intensity , which has @xmath99 m .
therefore , the large optimization radii on the transmission side does not have a significant effect on the reflection side .
this explains why @xmath83 m , rather than @xmath100 m and @xmath101 m , shows in fig .
[ fig : reflintensityring ] a feature that is comparable to the feature from the experiment . despite these imperfections in the experiment
, our data qualitatively shows that there is a local change in the optimized intensity .
we have experimentally shown that as the optimization radius increases , the enhancement of the total transmitted intensity increases , while simultaneously the total reflected intensity decreases .
we also find that the enhancement of the intensity outside the optimization area increases as the optimization radius increases . from the radial intensity distribution of the reflected intensity
, we find evidence that there is a local decrease in the reflected intensity rather than a global decrease .
the local decrease confirms that the transmitted and reflected intensities are spatially correlated as recently predicted by fayard _
et al _ @xcite .
our results have prospects in extending the applications of wavefront shaping to increase the total transmitted intensity through the rough layer on top of the silicon absorber in a solar cell .
our results are also interesting for multi - channel quantum secure communication @xcite , where enhanced intensities are desired in multiple transmission channels ; for transmitting arbitrary intensity distribution through a scattering medium @xcite ; and imaging through an opaque medium .
for the optimization of intensity in large areas , we investigated three different wavefront shaping algorithms : the stepwise sequential algorithm , the partitioning , and the genetic algorithms . the details on how these algorithms work are described in refs .
firstly , the sequential algorithm modulates the phase of the segments of the slm one by one and combines them at the end of optimization .
secondly , the partitioning algorithm modulates the phase of 50@xmath19 of the segments simultaneously and keeps the optimized phases on the slm .
the modulated segments are chosen randomly at each step
. a better performance of the partitioning algorithm is expected because a larger number of segments is controlled simultaneously , and this is expected to yield a significant change in the target signal compared to the sequential algorithm .
thirdly , the genetic algorithm begins by creating a population of random phase masks , which are ranked using the measured enhancement .
the phase masks are combined using a weight proportional to the enhancement and then further mutated to create new phase masks .
the new phase masks are measured and replace the low ranking members of the population . as the whole steps are repeated
, the average enhancement of the population increases and finally converges .
we have performed wavefront shaping experiments to compare the enhancement of the three algorithms . in fig .
[ fig : ch07_comparealgorithms ] , we plot the enhancement @xmath20 in the optimization area versus the optimization radius @xmath14
. the enhancement decreases with increasing radius for all algorithms , as is expected . with the partitioning algorithm ,
the enhancement is on average approximately @xmath102 higher than for the sequential algorithm .
the partitioning algorithm outperforms the sequential algorithm because of the larger modulation signal in the optimization area .
furthermore , the partitioning algorithm has an enhancement that is @xmath103 higher than for the genetic algorithm .
we expected a similar performance of the genetic algorithm and the partitioning algorithm , since a comparable number of segments is simultaneously controlled in the two algorithms .
we attribute the lower performance to the fact that the genetic algorithm requires a large number of experimental parameters , which might differ for different optimization radii . a further detailed study of using genetic algorithm for large areas optimization is needed .
we have therefore chosen to use the partitioning algorithm since , it shows a better performance than the other two algorithms .
m. kim , y. choi , c. yoon , w. choi , j. kim , q .- h .
park , and w. choi , _ maximal energy transport through disordered media with the implementation of transmission eigenchannels _ , nature photon .
* 6 * , 581 ( 2012 ) .
s. m. popoff , g. lerosey , r. carminati , m. fink , a. c. boccara , and s. gigan , _ measuring the transmission matrix in optics : an approach to the study and control of light propagation in disordered media _ , phys .
lett . * 104 * , 100601 ( 2010 ) .
f. t. si , d. y. kim , r. santbergen , h. tan , r. a. c. m. m. van swaaij , a. h. m. smets , o. isabella , and m. zeman , _ quadruple - junction thin - film silicon - based solar cells with high open - circuit voltage _ ,
* 105 * , ( 2014 ) .
t. a. w. wolterink , r. uppu , g. ctistis , w. l. vos , k .- j .
boller , and p. pinkse , _ programmable two - photon quantum interference in @xmath104 channels in opaque scattering media _ , phys .
a * 93 * , 053817 : 1 ( 2016 ) .
d. akbulut , _ measurements of strong correlations in the transport of light through strongly scattering materials _ , ph.d .
thesis , ( university of twente ) available at www.complexphotonicsystems.com , 2013 .
|
the recent advent of wave - shaping methods has demonstrated the focusing of light through and inside even the most strongly scattering materials . typically in wavefront
shaping , light is focused in an area with the size of one speckle spot .
it has been shown that the intensity is not only increased in the target speckle spot , but also in an area outside the optimized speckle spot .
consequently , the total transmission is enhanced , even though only the intensity in a single speckle spot is controlled . here ,
we experimentally study how the intensity enhancement on both interfaces of a scattering medium depends on the optimization area on the transmission side .
we observe that as the optimization radius increases , the enhancement of the total transmitted intensity increases .
we find a concomitant decrease of the total reflected intensity , which implies an energy redistribution between transmission and reflection channels .
in addition , we find a qualitative evidence of a long - range reflection - transmission correlation .
our result is useful for efficient light harvesting in solar cells , multi - channel quantum secure communications , imaging , and complex beam delivery through a scattering medium .
|
the combined stellar wind and supernova ejecta of groups of o and b stars blow large bubbles in the interstellar medium .
the largest of these bubbles , with size scales of 100 pc to 1 kpc are commonly referred to as superbubbles .
the basic structure of a superbubble consists of a hot low - density interior , the cavity , surrounded by a cool shell of swept - up interstellar medium .
the continuous formation and dissipation of superbubbles is an important factor in the energy balance of the interstellar medium , and determines the locations of different phases of the interstellar medium on large scales @xcite .
compression of the interstellar medium in the shell may increase cooling and trigger the formation of a new generation of stars . also , the ability of large superbubbles to break out of the disk of a galaxy and initiate an outflow of chemically enriched plasma and ionizing radiation from the disk into the halo has a profound influence on the evolution of galaxies .
several examples of well - defined superbubbles have been identified in the galaxy ( e.g. heiles 1984 ; maciejewski et al .
1996 ; normandeau et al .
1996 ; heiles 1998 ; ehlerov & palous 1999 ; callaway et al .
2000 ; reynolds et al . 2001 ; mcclure - griffiths et al .
2002 , pidopryhora et al .
these have been studied in detail , thanks to their relative proximity .
observations with parsec - scale resolution of neutral and ionized gas reveal important details about the interaction between the hot ejecta and the interstellar medium .
new high - resolution surveys of galactic atomic hydrogen ( hi ) emission @xcite have provided unprecedented images with morphological and kinematic information of galactic superbubbles @xcite .
physically interesting parameters are usually derived from the observations by means of analytic models that assume spherical symmetry @xcite or axial symmetry @xcite . in this paper
we investigate the importance of mhd effects on physical quantities derived from observed superbubbles .
the effect of the galactic magnetic field is difficult to model because it introduces anisotropy in the medium , which requires three - dimensional numerical simulations .
the first three - dimensional magnetohydrodynamic simulations of superbubbles were presented by @xcite , who discussed the importance of the magnetic field in the break - out of superbubbles from the galactic disk .
@xcite also described significant departures from spherical and axial symmetry in the shape of a magnetized bubble resulting from the interaction of the expanding superbubble with the galactic magnetic field .
the shape and the size of a superbubble depend on the strength and the geometry of the galactic magnetic field as much as they depend on the density distribution of the ambient interstellar medium .
the expanding superbubble in turn redefines the geometry of the interstellar medium and the galactic magnetic field in a volume several hundred parsecs across .
@xcite and @xcite performed thee - dimensional mhd simulations that include the evolution of a superbubble in a supernova - driven turbulent multi - phase interstellar medium .
the super bubbles in these simulations show significant departures form symmetry because of inhomogeneities in the medium in which the super bubble expands . @xcite and @xcite found that a super bubble can break out into the halo in such a medium .
although these simulations provide valuable insight in the dynamics of superbubbles in a multi - phase interstellar medium , it is in general difficult to relate these simulations to specific observed super bubbles ( see however fuchs et al .
2006 for a simulation of the local bubble ) .
astrophysical parameters derived from observations , such as the age and the energy of a superbubble , or the density distribution of the ambient medium have relied on symmetric analytic models that do not include a magnetic field .
the departure from axial symmetry imposed by the magnetic field introduces systematic errors that have not been considered before . in this paper
we present three - dimensional mhd simulations of superbubbles , and we explore the errors introduced by commonly used methods to determine basic parameters from observations . in particular , we study the axial ratio of the wind - blown cavity for different times and magnetic field configurations .
the smaller simulation volume and time span used in our simulations compared to @xcite are more tailored to the latitude coverage of the galactic plane surveys .
we also calculate the first images of faraday rotation by a magnetized superbubble derived from our simulations , and emphasize the importance of such simulations to make meaningful predictions in this area .
we describe frequently used analytic models 2 and detail our numerical setup and methods in 3 .
numerical results and the analysis of magnetic effects on derived parameters are presented 4 . the specific case of the w4 supperbubble
is discussed in 5 .
faraday rotation by magnetized superbubbles is discussed in 6 , and conclusions are presented in 7 .
the exact shape and size of the bubble depends directly on the environment and the strength of the source . for weaker sources which produce bubbles
whose extent are much smaller than the scale height of the gas in the galaxy , a solution with an approximately constant density profile may be considered .
@xcite present a solution for a spherical wind - blown bubble expanding into a non - magnetized medium , in which the radius of the bubble varies according to @xmath0 where @xmath1 is the outer shock of the shell , @xmath2 is the mechanical luminosity of the source , @xmath3 is the ambient density of the undisturbed interstellar medium , and @xmath4 is the age of the bubble in myr .
the radius of the contact discontinuity was found to be @xmath5 by @xcite . for larger bubbles , associated with powerful sources ,
the situation is more complicated .
the gravitational potential of the galactic disk produces a density gradient perpendicular to the plane of the disk . at early stages ,
the @xcite solution can still be applied , but as the radius of the bubble exceeds the scale height of the surrounding medium , the density gradient of the surrounding medium affects the shape of the bubble .
@xcite ( hereafter k60 ) proposed a solution for the evolution of a point explosion in a non - magnetized exponential atmosphere that can be applied to this situation .
this solution is based on the assumptions that the thermal energy is a constant fraction of the energy deposited in the initial blast and that the energy distribution is uniform throughout the volume of the bubble except close to the shock front where the energy density can be two to three times the mean value ( see also bisnovatyi - kogan & silich 1995 ) .
@xcite extended the kompaneets solution to describe continuous energy injection , which we use in this paper .
the source is located at @xmath6 in an exponential atmosphere with scale height @xmath7 of the form @xmath8 the radius of the bubble in cylindrical coordinates takes the following form @xmath9 \label{rbasu}\ ] ] where @xmath10 is a dimensionless transformed time variable , @xmath11 is the adiabatic index , @xmath12 is the thermal energy of the bubble , and @xmath13 is the volume of the bubble .
the independent variables are the mass density at @xmath6 ( @xmath14 ) , the scale height of the atmosphere ( @xmath7 ) , and the mechanical luminosity of the source ( @xmath2 ) , which determines @xmath12 .
the unit of time in this model is then defined by @xmath15 for @xmath16 , equation ( [ rbasu ] ) reduces to two equations for the upper and lower boundary of the blast wave , @xmath17 and @xmath18 respectively @xcite @xmath19 as @xmath20 approaches 2 , the top of the bubble in the kompaneets model expands to infinite height in a finite amount of time .
physically , this means the shock acceleration in the z - direction becomes infinite because of the strong density gradient @xcite .
the bottom of the bubble ( @xmath21 , equation [ zupperlower ] ) does not penetrate downward more than @xmath22 , its location at the time of blow - out .
since the top of the kompaneets model reaches an infinite height in a finite amount of time , it can not be a valid solution at later times .
however , it can provide an adequate solution at early times , if the initial conditions are consistent with the assumptions of the kompaneets model , i.e. negligible pressure of the ambient medium and negligible inertia of the swept up medium . with the addition of a magnetic field , no three - dimensional analytic solutions exist .
in order to capture the true evolution of the these bubbles numerical simulations need to be performed .
we aim to include the physics of the galactic magnetic field in the interpretation of observed superbubbles .
the derivation of physical quantities from the data , such as the scale height of the surrounding medium or the age of a superbubble is best served by a model that takes into account the magnetic field , but not the complications of a preprocessed interstellar medium .
we do not include the effect of galactic differential rotation or a coriolis force on the superbubble in our simulations .
@xcite found that the characteristic time scale for shear from differential galactic rotation is @xmath23320 myr while the time scale for shear due to rotation of the bubble by the coriolis force is @xmath2350 myr .
these processes operate over significantly longer times scales than those considered here ( up to 20 myr ) .
the evolution of the bubble will also be affected by heating and cooling processes . in this paper
we discuss adiabatic simulations as an approximation of the situation where heating balances cooling throughout the life of the superbubble , while we use simulations with cooling to explore how the evolution changes if cooling dominates over heating .
observations of ionized gas associated with superbubble shells indicate that a significant fraction of the mass in the shell may be photo - ionized by the central star cluster and the surrounding interstellar radiation field .
the ionized mass in the shell of the orion - eridanus superbubble is @xmath24 @xcite .
the neutral mass of this shell was determined to be @xmath25 by @xcite , and @xmath26 by @xcite .
these values indicate that 13 to 28% of the mass of the orion - eridanus shell is ionized by the stars inside the bubble .
@xcite found equal amounts of neutral and ionized gas in the ophiuchus superbubble , that is possibly ionized from the outside as well as from the inside .
the shell of the w4 superbubble @xcite is clearly visible by its thermal radio continuum emission and @xmath27 emission @xcite , but not in hi .
photo - ionization of a substantial fraction of the mass of the shell indicates that heating by photoionization is a non - negligible term in the energy budget of the gas in the shell .
simulations that include cooling without photo - ionization therefore neglect a significant heating term in the energy budget of the shell . for computational reasons we can not solve radiative transport in the three - dimensional mhd simulations .
the absence of cooling in our current simulations corresponds with the approximation of equilibrium between heating and cooling in the shell . in section
[ cooling - sec ] we discuss the effect of cooling on the axial ratios of the cavity .
cccc density&@xmath3 & & @xmath28 + location&@xmath29 & & @xmath30h + velocity&@xmath31 & & @xmath32 + time&@xmath33 & & @xmath34 + luminosity & @xmath35 & & @xmath36 + pressure & @xmath37 & & @xmath38 + internal energy & @xmath39 & & @xmath40 + magnetic field & @xmath41 & & @xmath42 + cccc @xmath14 & & & @xmath2 + @xmath43 & & & @xmath44 + @xmath45:(b@xmath46,b@xmath47,b@xmath48 ) & & & @xmath49 + density profile & & & @xmath50 + & & & @xmath51 + & & & location:(@xmath52,@xmath53,@xmath54 ) + [ models ] numerical simulations of bubbles have been performed by many researchers ( see @xcite , @xcite , @xcite , @xcite , @xcite and @xcite , among others ) .
@xcite was the first to perform three - dimensional mhd simulations with radiative cooling assuming symmetry with respect to the @xmath29=0 , @xmath55=0 and @xmath56=0 planes .
in the simulations we present here , we have a complete three - dimensional bubble evolving in an unperturbed , three - dimensional environment . our simulations solve the following equations : @xmath57 - { { \bf{\nabla}}}p - \rho{{\bf{\nabla}}}\phi + \nonumber\\ + \frac{({{\bf{\nabla}}}\times { { \rm\bf{b } } } ) \times { { \rm\bf{b}}}}{4\pi } & = & 0\\ \label{a } \rho\left[\frac{\partial e}{\partial t } + ( { { \rm\bf{v}}}\cdot { { \bf{\nabla}}})e\right ] + p\ ( { { \bf{\nabla}}}\cdot { { \rm\bf{v } } } ) & = & 0\\ { { \bf{\nabla}}}\cdot { { \rm\bf{b}}}&= & 0\end{aligned}\ ] ] where @xmath3 is the density , @xmath37 is the gas pressure , @xmath41 is the magnetic field , @xmath31 is the velocity , @xmath58 is the gravitational potential , and @xmath39 is the internal energy of the gas . for our simulations we work in units of density ( @xmath14 ) , scale height ( @xmath7 ) and sound speed ( @xmath59 ) .
the conversion to dimensionless variables takes the following form @xmath60 the other parameters are converted in a similar manner ( see table [ dimlesstable ] ) .
the unit of time in our simulations is @xmath61 which is different from the unit of time in the kompaneets model @xmath62 in equation ( [ ytilde ] ) , which assumes @xmath63 . the mhd equations in dimensionless variables are solved numerically by the zeus - mp code @xcite .
the zeus - mp code used in this work is an adapted and modified version of the original 1.0b version of norman and li ( see norman 2000 ) .
one of the authors involved in this paper ( r. ouyed ) has also spent a large amount of time to modify the code .
most of these modifications were similar to those described in @xcite .
a detailed description of the original zeus code can be found in stone & norman ( 1992a&b ) which also includes basic tests of the code .
vulnerabilities in the zeus family noted by @xcite , in particular the issue of rarefaction waves and shock errors , have recently been discussed by hayes et al .
( 2006 ; and references therein ) .
while these papers acknowledge that the code possess limitations as do all numerical methods , results from zeus - mp were found to compare quite favorably with other numerical techniques .
we come to the same conclusion in [ dlcbande ] where we compare our results to that of @xcite despite differences in numerical methods used .
zeus solves the mhd equations using the operator split method .
the equations are solved in two substeps , called a source step and a transport step ( see stone & norman 1992a&b ) .
three methods for the advection of mass , momentum and internal energy in the transport step can be implemented : the first order accurate donor cell method @xcite , the second order van leer method @xcite , and the third order piecewise parabolic advection ( ppa ) method @xcite . after some numerical experiments we decided to use the van leer method because it offers the best ratio of precision to computational costs .
the basic equations of the code are written in a covariant form which allows for the use of the code in an arbitrary orthogonal coordinate system ( cartesian , cylindrical and spherical coordinates are predefined ) .
the algorithm used to guarantee that @xmath64 is the `` hsmocct '' method which combines the constrained transport ( ct ) module of @xcite , and improvements of the method of characteristics ( moc ) introduced by @xcite . in this scheme ,
if the initial b has zero divergence in the discretisation on the staggered mesh , then every time step will maintain the initial value of the divergence to the accuracy of machine round - off error .
the finite difference method is based on the discretization of each dependent variable over the spatial computational domain .
then finite difference approximations to the differential equations are solved on this discrete mesh .
the zeus code uses a staggered mesh built up of two mutually shifted grids .
the a - grid specifies positions of the zone boundaries while the b - grid specifies the zone centers .
discrete values of all dependent variables are stored for each zone .
scalars are stored at the zone centers while components of vectors are stored at the appropriate zone interfaces ( see figure 1 in stone & norman 1992a ) .
boundary conditions are implemented as two layers of ghost zones at each boundary of the computational domain ( two layers are required for higher order interpolation if the ppa method is used ) .
values of the dependent variables in the ghost zones are given by simple , explicit equations that connect these values to the values in the adjacent active zones .
for our simulations we use the so - called outflow boundary conditions where the values of all variables in the ghost zones are set equal to the values in the corresponding active zones .
the simulations are performed in a three - dimensional cartesian box with right handed coordinates @xmath52 , @xmath53 , and @xmath54 . as initial conditions ,
we specify a density distribution ( atmosphere ) and the location of one or more energy sources and their luminosities ( as described in the following sections ) . for a summary of these input parameters
refer to table [ freepar ] .
we consider two functional forms for the distribution of the ambient gas .
the first form is an exponential density distribution @xmath65 . \label{rho_exp}\ ] ] the second form is the density distribution proposed by ( hereafter dl ) from their analysis of the vertical distribution of atomic hydrogen in the galaxy .
this density distribution , in terms of dimensionless variables , is @xmath66 + \nonumber \\
+ \frac{0.107}{0.566 } \exp\left[-\frac{1}{2}\left(a_{2}\tilde{x_{2}}\right)^{2}\right ] + \nonumber \\
+ \frac{0.064}{0.566 } \exp\left[-a_{3}\tilde{|x_{2}|}\right ] \end{aligned}\ ] ] where @xmath67 , @xmath68 , and @xmath69 .
the coefficients @xmath70 express the scale heights of the three components of the dl layer in units of 100 pc .
contrary to the exponential profile , the dl layer has an equatorial plane , @xmath71 , with density decreasing in both the positive and negative @xmath53 directions .
this is the same density distribution adopted by @xcite .
the dynamical importance of the magnetic field is set through the parameter @xmath43 , defined as the ratio between the gas pressure and magnetic pressure at @xmath72 @xmath73 the magnetic field strength is @xmath74 where the numeric expression was derived using @xmath75 .
we consider two geometries of the magnetic field .
the first is a constant field , i.e. the vertical scale height of the magnetic field is infinite ( @xmath76 ) , the other takes the scale height of the magnetic pressure to be equal to the scale height of the gas pressure ( @xmath77 everywhere ) .
we refer to the constant @xmath78 case as equipartition although strictly speaking equipartition implies @xmath79 everywhere .
the scale height of the galactic magnetic field is not well known .
our initial conditions cover the possibility of a large scale height @xmath80 1 kpc corresponding to a magnetically dominated halo , and a small scale height ( equal to 2@xmath7 for an exponential atmosphere ) .
hydrostatic equilibrium is assured by imposing a gravitational potential that balances the gradient of the total pressure which is the sum of the gas pressure and the magnetic pressure .
the maintenance of hydrostatic equilibrium was tested in simulations with no source for both atmospheres .
the vertical velocity in these tests was less than @xmath81 after 10 myr simulated time .
this is consistent with numerical errors and it has no effect on the results of our simulations .
the source of the bubble is defined by its mechanical luminosity ( @xmath2 ) related to the mass loss rate ( @xmath82 ) and the outflow velocity ( @xmath44 ) according to @xmath83 the source is assigned a finite radius ( @xmath49 ) , determined by the spatial resolution of the simulation .
the mass loss rate from the source is related to the outflow velocity and the radius of the source via @xmath84 solving for @xmath85 and converting to dimensionless variables , we find @xmath86 values for @xmath87 , @xmath88 and @xmath89 are specified as initial conditions . at every time step , @xmath90 and @xmath89 are reset to their initial values to maintain constant momentum and energy input from the source . to simulate a point source ,
the radius @xmath49 should be taken as small as possible . however , in the rectangular grid , the outflow from the source will be more isotropic if @xmath49 is more than a few times the grid size .
a complication is that density is evaluated in the center of a pixel ( b - grid ) , whereas velocity is evaluated on the boundary of a pixel ( a - grid ) .
this is part of a strategy to conserve mass in the simulations , but it introduces an asymmetry in the radius of a sphere when coordinates are rounded off to integer multiples of the grid size . to maintain an isotropic source
while avoiding computationally intensive interpolations that must be performed every time step , the location of the source is oscillated in three dimensions by half a grid position every time step ( see figure [ grid - fig ] ) .
these excursions of the source along the three axes are incoherent , and average out most of the anisotropy introduced by the numerical grid for small @xmath49 .
the isotropy of the source setup was tested in hydrodynamic simulations with no density gradient .
these experiments showed that @xmath49 between two and three pixels results in good symmetry in the simulations . a small degree of anisotropy in the outflow of the source remains .
this acts as a seed for the development of instabilities , but is not considered a problem for our analysis .
cccccc + consth_zoom & 3 & 1 & no b & & @xmath91 + consth & 3 & 1 & no b & & @xmath91 + constmhd & 3.2 & 0.32 & const & & 1.16 + exph_zoom & 3 & 1 & no b & & @xmath91 + exph & 3 & 1 & no b & & @xmath91 + expcba & 3 & 1 & const & & 10 + expcbb & 3 & 1 & const & & 3 + expcbc & 3 & 1 & const & & 1 + expcbd & 3 & 1 & const & & 0.3 + expeba & 3 & 1 & equip & & 10 + expebb & 3 & 1 & equip & & 3 + expebc & 3 & 1 & equip & & 1 + expebd & 3 & 1 & equip & & 0.3 + dlh & 3&1 & no b & & @xmath91 + dlcba & 3 & 1 & const & & 10 + dlcbb & 3 & 1 & const & & 3 + dlcbc & 3 & 1 & const & & 1 + dlcbd & 3 & 1 & const & & 0.3 + dleba & 3 & 1 & equip & & 10 + dlebb & 3 & 1 & equip & & 3 + dlebc & 3 & 1 & equip & & 1 + dlebd & 3 & 1 & equip & & 0.3 + tomisaka a & 3 & 0.3 & const & & 0.7 + the simulations were performed on the capca computer cluster which consists of 64 2.4 ghz linux based processors connected by a 1 gigabit network .
the results of these simulations were analyzed using jetget @xcite and kvis , a program that is part of the karma visualization package @xcite . in the simulations that we present here , we examined the effects of magnetic field strength and geometry on the evolution of a superbubble in the two atmospheres defined in section [ atmospar ] .
the other parameters were fixed to the values in table [ simsrun ] to apply our simulations to published results on the bubble associated with the w4 region as described by @xcite and @xcite ( section [ app ] ) .
table [ simsrun ] gives a summary of the simulations and the values of the varied physical parameters .
table [ simsrunvolume ] gives details on the simulation volume . to test convergence
we ran simulations at @xmath92 and @xmath93 zones and established convergence at @xmath94 which corresponds to a resolution of 5 pc per voxel ( see section [ sec : x3x1 ] ) .
ccccc + consth_zoom & @xmath95 & @xmath961.5 to 1.5 & @xmath961 to 2 & @xmath97 + consth & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + constmhd & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + exph_zoom & @xmath95 & @xmath961.5 to 1.5 & @xmath961 to 2 & @xmath97 + exph & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expcba & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expcbb & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expcbc & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expcbd & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expeba & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expebb & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expebc & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + expebd & @xmath98 & @xmath965 to 5 & @xmath963 to 7 & @xmath97 + dlh & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlcba & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlcbb & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlcbc & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlcbd & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dleba & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlebb & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlebc & @xmath98 & @xmath965 to 5 & @xmath965 to 5 & @xmath97 + dlebd & @xmath98 & @xmath965 to 5 & @xmath965 to 5 &
@xmath97 + tomisaka a & @xmath99&0 to 10 & @xmath9610 to 10 & @xmath100 + [ comparetoanalytic ] in order to test the setup , we performed hydrodynamic simulations with a constant density profile ( consth_zoom & consth ) to compare with the @xcite model and a hydrodynamic simulation with an exponential density profile ( exph_zoom ) to compare with the k60 model .
the physical parameters for these simulations are given in table [ simsrun ] .
maintaining the same number of pixels for a smaller volume ( 300 pc on a side ) results in a finer resolution ( 1.5 pc compared to the 5 pc resolution of the other simulations listed in table [ simsrunvolume ] ) .
this allows us to follow the evolution of the bubble in detail at early times .
we have multiple reasons to consider high - resolution simulations for the comparison with the @xcite and k60 models .
the first reason is to investigate the effect of resolution on our solutions .
the second reason is to explore the effect of the finite source size . in our simulations
, the bubble begins at a finite radius at @xmath33 = 0 ( the radius of the source ) which can be made smaller in high resolution simulations .
although the @xcite model is self - similar , increasing the resolution will allow us to investigate the effect of the finite source size in the simulations at early times .
the k60 model is not self - similar , but it is only expected to agree with the simulations at early times ( section [ analytic ] ) .
both models assume that the sound speed in the ambient medium is zero by assuming the pressure in the undisturbed medium is negligible . for our simulations ,
the sound speed in the undisturbed medium is finite , but the assumption of negligible pressure is satisfied as long as the pressure inside the bubble is much larger than the pressure of the surrounding atmosphere .
the limit of negligible ambient pressure adopted in the k60 model implies @xmath101 . in our simulations ,
the time scale @xmath102 ( equation [ dimlesstime ] ) is @xmath103 years , whereas the timescale @xmath104 of the k60 model ( equation [ ytilde ] ) is @xmath105 years .
the simulations should therefore resemble the kompaneets model at times much less than 10@xmath106 years .
the size of the bubble was parameterized by the radius of the contact discontinuity , which is well defined in both the simulations and the kompaneets model .
the shell of swept - up interstellar medium is presumed to be infinitely thin in the kompaneets model , so the radius of the outer shock is not predicted . for consistency ,
the same was done for the spherical case .
the radius of the contact discontinuity for the spherical model was taken to be 86% of the outer shock radius following @xcite .
for consth_zoom , the contact discontinuity is located at 84% of the outer shock radius , which is in good agreement with the @xcite model .
figure [ radiusgraph]a compares the time evolution of the radius of consth_zoom with the spherical model .
the simulation displays a power law expansion with time @xmath107 with @xmath108 .
this is only 5.7% smaller than the slope of the @xcite model , which has @xmath109 . at an age of 1 myr
the radius of consth is 9% larger than the radius of the model .
the filled squares in figure [ radiusgraph]a show the time evolution of consth .
the radius of the bubble in this simulation is at most 10% larger than the radius in the higher resolution simulation at the same age .
this indicates that part of the difference between our simulations and the castor model may be related to the finite size of the source .
figure [ radiusgraph]b shows the time evolution of the radius of a bubble in an exponential atmosphere ( exph@xmath110zoom in tables [ simsrun ] and [ simsrunvolume ] ) , looking perpendicular to the galactic plane , for the simulations and the k60 model . also shown is the spherical model from figure [ radiusgraph]a as a reference line .
the bottom of the simulation closely follows the k60 model .
however the top of the k60 model is consistently higher than the simulation by about 16% even at early times .
this is also shown in figure [ kompaneetsfig ] . as time proceeds , the difference between the k60 model and the simulations increases significantly at the top of the bubble but not at the bottom .
@xcite also compared their 2-dimensional hydrodynamic simulations to the kompaneets model with similar results .
@xcite found consistency between their kompaneets approximation from @xcite and hydrodynamic simulations at times as late as 6.87 myr .
however , their kompaneets approximation is a solution in an atmosphere with an equatorial plane unlike the kompaneets model @xcite considered here . the axial ratio in the galactic plane ( i.e. @xmath111 ) is expected to be unity at all times for the hydrodynamic case and also at early times in the mhd case ( see section [ analysis ] ) .
the radius of the cross - section through the source of the k60 model as a function of time can be calculated to a good approximation from the spherical castor model , as shown numerically by @xcite .
@xcite derived analytic solutions of a superbubble in a uniform magnetic field in the limit of high expansion velocity , and numerical solutions for the general case of smaller expansion velocity . to compare with their solutions , we ran an mhd simulation in an atmosphere with constant density ( constmhd , see tables [ simsrun ] & [ simsrunvolume ] ) .
@xcite found that the outer shock front remains nearly spherical in a uniform magnetized medium , but the cavity is smaller in the direction perpendicular to the magnetic field , compared with the hydrodynamic solution .
we find the same in our numerical simulation , but we see no dimple in the outer shock in the direction of the magnetic field .
@xcite also did not see this dimple .
the expansion along the magnetic field lines is nearly the same as in the hydrodynamic simulation as found by @xcite .
the thickness of the shell perpendicular to the magnetic field in the analytic approximation found by @xcite , increases linearly with time according to @xmath112 where @xmath113 is the number density in @xmath114 , @xmath115 is the ambient magnetic field strength in @xmath116 g , and @xmath4 is the time in myr .
our simulation reproduces the linear increase of shell thickness perpendicular to the magnetic field with time , but the shell thickness is a factor @xmath23 2.6 larger .
the simulations by @xcite also show a very thick shell perpendicular to the magnetic field , although he did not consider a medium with constant density .
the reason for the difference is that equation ( [ ferriere ] ) was derived in the limit of high expansion velocity , defined as a small ratio of magnetic pressure to ram pressure .
this ratio increases with time as the outer shock decelerates . we find that this ratio is of order unity or larger at times @xmath80 2.5 myr in our simulation .
simulations with higher @xmath78 reproduce the shell thickness in equation ( [ ferriere ] ) , while also satisfying the assumptions made in the derivation of this equation . a simulation of a bubble expanding in an exponential atmosphere with no magnetic field was done primarily to compare with previously published results and to obtain the limiting case @xmath117 . at early times
, the bubble remains spherical ( figure [ kompaneetsfig ] ) .
the axial ratio of the cavity in the simulations _ and the axial ratio of the kompaneets model _ are indeed found to be very close to unity , even though figure [ kompaneetsfig ] shows that the top of the kompaneets model has proceeded significantly further than the simulation even at these early times . in appendix
a we show that the kompaneets model at early times can be described to third order in @xmath20 as a spherical cavity that rises in the atmosphere .
although the axial ratio remains unity to a high degree of accuracy , the geometric center of the model , defined by the point midway between the top and the bottom ( crosses in figure [ kompaneetsfig ] ) , is displaced significantly from the location of the source . at later times , the bubble grows larger than the scale height of the medium and accelerates in the vertical direction .
as the bubble expands into the upper atmosphere , the top of the bubble continues to accelerate and a rayleigh - taylor instability develops when the top reaches @xmath118 scale heights above the source , which occurs at an age of @xmath2310 myr .
the bubble breaks out of the galactic plane to form a chimney after about 16 to 17 myr .
the main difference between a constant magnetic field and equipartition magnetic field is in the evolution of the bubble at large distances from the galactic plane .
a constant magnetic field implies a high alfvn speed in the galactic halo whereas constant @xmath78 implies that the alfvn speed in the halo is the same as in the disk .
this difference has significant consequences for the structure of the shock as it expands from the disk into the halo .
figure [ expconst_b_40 ] shows a bubble in an exponential atmosphere with constant magnetic field ( @xmath119 ) oriented along the @xmath52 axis , at an age of 10 myr .
the cavity is elongated along the magnetic field as the plasma can move freely along the field lines but its motion perpendicular to the field lines is restricted .
however , the shape of the outer shock at the level of the source in the galactic plane ( @xmath111 plane ) is nearly circular because the propagation speed of the outer shock is similar in both directions . along the @xmath52 axis ,
the outer shock and the contact discontinuity are close together resulting in a thin compressed shell of swept - up interstellar medium . along the @xmath54 axis ,
perpendicular to the initial magnetic field , the distance between the outer shock and the contact discontinuity is much larger resulting in a thick shell .
looking along the field lines , the cavity is more elongated than in the hydrodynamic simulation , and the shell of swept - up interstellar medium is thicker than in the hydrodynamic case . at the top of the cavity a fast magnetosonic wave runs upwards into the halo , with very little compression of the halo gas .
in contrast to the hydrodynamic simulation , we see no evidence for a rayleigh - taylor instability developing at the top of the bubble , because the top of the bubble does not accelerate .
instead , an instability develops everywhere at the contact discontinuity as is apparent from the scalloped shape of the contact discontinuity in the @xmath120 plane ( looking along the field lines ) .
this instability also occurs in the simulations of @xcite and may be similar to the magnetically enhanced rayleigh - taylor instability proposed by @xcite . at the time shown in figure [ expconst_b_40 ] , the expansion of the cavity along the @xmath54 axis ( perpendicular to the magnetic field ) in the galactic plane has stalled as a result of magnetic tension .
the cavity still expands along the @xmath52 and @xmath53 axes .
figure [ expequip_b_40 ] shows simulation expebc also at the age of 10 myr .
in contrast with figure [ expconst_b_40 ] , figure [ expequip_b_40 ] shows a shell of compressed interstellar medium at the top of the bubble .
since the alfvn velocity does not increase with distance from the galactic plane , the outer magnetosonic shock does not accelerate into the halo but remains close to the contact discontinuity .
the shell appears approximately twice as thick looking along the field lines as looking perpendicular to the magnetic field .
the axial ratio of the cavity in the @xmath120 plane ( looking along the field lines ) is smaller than in the expcbc simulation .
this indicates the importance of magnetic tension confining the bubble , even though the total pressure mimics the pressure in the hydrodynamical exponential atmosphere . the magnetic field in the upper part of the shell
is significantly enhanced compared to the magnetic field in the undisturbed atmosphere at the same height above the plane . here too , we see an instability everywhere along the contact discontinuity in the @xmath120 plane . the simulations in the dl atmosphere show some differences compared to the evolution in the exponential atmosphere .
the equatorial plane introduces symmetry with respect to the plane @xmath72 . in the absence of a magnetic field the bubble is spherical until it expands to radius @xmath121 pc .
once the radius of the bubble grows beyond @xmath23200 pc , it becomes more elongated in the vertical direction as it balloons out into the halo ( figure [ fig : dlh ] ) .
the top and bottom of the bubble expanding into the low - density outer atmosphere , accelerate away from the galactic plane , as in the case of an exponential atmosphere .
the onset of a rayleigh - taylor instability is not seen in this simulation but it may develop at later times outside the simulation volume .
between an age of @xmath122 myr and @xmath123 myr a small decrease in the radius of the cavity in the equatorial plane is observed .
a similar stall of the expansion and subsequent contraction in the galactic plane triggered by the rapid vertical expansion of the bubble was observed by @xcite .
figure [ b40fig ] shows the simulation of a bubble in the dl atmosphere with constant magnetic field oriented along the @xmath52 axis at an age of 10 myr .
as with the exponential simulations , this simulation shows the same features as discussed by @xcite .
the cavity is elongated along the magnetic field lines .
the shape of the outer shock at the level of the source in the @xmath111 plane is nearly circular , similar to the exponential case .
this is to be expected because the density and magnetic field strengths are the same at the level of the source in both atmospheric profiles .
in contrast to the exponential atmosphere , the shape of the cavity in the @xmath124 plane is fairly circular even though the bubble expands to a size much larger than the pressure scale height .
the shape of the cavity in the @xmath120 plane is quite elongated and displays the same magnetically enhanced rayleigh - taylor instability as discussed in section [ expcbande ] . as with the expcbc case , a fast magnetosonic wave runs upwards into the halo .
figure [ dlequip_b_40 ] shows the constant @xmath78 simulation dlebc at the same age .
the discussion of the differences between the expcbc and expebc applies also to simulations in the dl atmosphere .
we performed an extra simulation ( tomisaka a ) in a larger volume to compare with model a in @xcite , with the same equatorial density .
other initial conditions were set as closely as possible to those specified by @xcite ( see tables [ simsrun ] & [ simsrunvolume ] ) , but some small differences in the setup could not be avoided .
our simulations show the same characteristics in terms of the size and shape of the cavity , and the thickness of the surrounding shell .
however we found differences of the order of 20% in the dimensions of the cavity after 10 myr which appear to be associated with small differences in initial conditions .
the fast magnetosonic shock had proceeded to almost twice the height of the shock in model a of @xcite , indicating a lower value of @xmath78 in our simulation .
the shape of the magnetosonic shock front is particularly sensitive to small differences in @xmath78 because the alfvn speed at high altitudes is very sensitive to @xmath78 .
our simulations cover a factor 30 in magnetic field strength , from relatively weak ( those labeled with a , @xmath125 ) to strong ( those labeled with d , @xmath126 ) . here
we discuss variation in the shape of a bubble with magnetic field strength at a reference age of 10 myr .
the weak magnetic field limit for the simulations converges to the hydrodynamic case as should be expected .
the differences between simulations with @xmath125 and hydrodynamic simulations are minor .
we limit the discussion to the range @xmath125 to 0.3 .
one of the most visible effects of the magnetic field is on the size of the cavity .
a decrease in @xmath78 by a factor of 30 results in a decrease in the vertical size of the cavity by a factor @xmath231.7 , independent of the density and magnetic field stratifications .
the diameter of the cavity in the @xmath127 direction ( along the field lines ) increases by a factor of @xmath23 1.4 as @xmath78 decreases .
expansion along the magnetic field is faster as the magnetic field is stronger .
the diameter of the cavity along the @xmath128 axis decreases by a factor @xmath231.6 as @xmath78 decreases from 10 to 0.3 .
the effect of a stronger magnetic field on the shape of the cavity is that the cavity becomes more elongated along the galactic plane when observed along a direction perpendicular to the galactic magnetic field .
when observed along the direction of the magnetic field , the shape of the cavity varies less , but the size of the cavity becomes significantly smaller .
the thickness of the shell in the directions perpendicular to the magnetic field also increases dramatically with increasing magnetic field strength .
however , this happens at the expense of the amount of compression of gas in the shell , which becomes nearly invisible in the simulations with @xmath126 . in practice
, the shell may not be detectable in real data confused with emission in the foreground and background , even for @xmath129 , if the magnetic field scale height is large ( see figure [ expconst_b_40 ] ) .
the increasing thickness of the shell along the @xmath128 axis is partly the result of the smaller extent of the cavity , partly because of a faster expansion of the outer shock , approximately by equal amounts . in simulations with a constant magnetic field ,
the thickness of the shell at the top of the bubble increases mainly because of the fast upward expansion of the outer shock , and to a lesser amount by the smaller extent of the cavity . in the equipartition simulations ,
the increased thickness of the top of the shell is mainly because of the smaller extent of the cavity . along the magnetic field lines ( @xmath127 ) , the thickness of the shell decreases as the magnetic field strength increases .
however , no strong increase in the density is found at the extremes of the shell along the @xmath127 axis .
figure [ fig : castorcompare ] compares the time evolution of the semi - major axis of the contact discontinuity and the outer shock in the @xmath130 plane , for a sample of our simulations ( exph , dlh , expcbc , & dlcbc ) , with that of the @xcite model .
the semi - major axis is defined as half the diameter along the @xmath127 axis at the level of the source .
the expansion along the magnetic field lines at early times ( see figure [ fig : castorcompare]a ) closely follows the hydrodynamic solution in accordance with @xcite .
the contact discontinuity of the self - similar solution was found to be at 86% of the shock radius by @xcite .
the radius of the contact discontinuity in the analytic model indeed agrees well with the radii of the contact discontinuity in our simulations at early times ( dashed curve in figure [ fig : castorcompare ] ) . at later times ( figure [ fig : castorcompare]b ) , the contact discontinuity in the expcbc and dlcbc simulations has proceeded further than the analytic solution ( dashed curve ) . on the other hand ,
the contact discontinuity in the exph and dlh simulations lags behind the analytic solution .
we interpret this behavior as the result of the hydrodynamic simulations breaking out of the disk , reducing the pressure inside the cavity , whereas the mhd simulations remain confined to the disk .
we find that the location of the outer shock in these simulations is nearly the same even after 15 myr in agreement with results presented in @xcite .
the discussion in the previous subsection shows that the magnetic field is as important as the density distribution of the medium in determining the shape of a superbubble .
observationally , the shape of the cavity is most important because structures with a large contrast in density are most easily detected in hi images of the galaxy .
figure [ fig : x3x1conv ] shows the axial ratio of the low - density cavity in the galactic plane ( @xmath131 ) versus time for different resolutions ( @xmath92 and @xmath93 ) to verify convergence ( see section [ simulations ] ) . from this , we confirm that convergence is established at @xmath94 which we use for the analysis .
figure [ x1x3ratio ] shows the relationship between the @xmath132 axial ratio and @xmath78 at an age of 5 myr and 12.5 myr .
the similarity between the four panels shows that the axial ratio of a superbubble in the galactic plane does not depend significantly on the vertical stratification of the density and the vertical stratification of the magnetic field ( see also figures [ expconst_b_40 ] , [ expequip_b_40 ] , [ b40fig ] and [ dlequip_b_40 ] ) .
the important parameters are the age of the bubble and the strength of the magnetic field in the equatorial plane . even at the relatively young age of 5 myr , a bubble in a medium with @xmath79 will be elongated along the magnetic field with axial ratio @xmath230.8 .
after 12.5 myr , the size of the bubble perpendicular to the magnetic field is only 60% of its size along the magnetic field .
variation of @xmath78 by a factor of three either way changes the predicted shape of the bubble in the galactic plane from an axial ratio @xmath133 to @xmath134 at 12.5 myr .
figure [ x2x1ratioexp ] shows the time evolution of the @xmath135 axial ratio of the low density cavity for exph ( diamonds ) , expcbc ( stars ) , expebc ( filled squares ) , and the kompaneets model ( dashed line ) . at early times , @xmath136 myr , the evolution of exph follows the kompaneets model .
the difference between the exph simulation and the kompaneets model at later times is the result of inertia of the swept - up interstellar medium .
expcbc and expebc evolve in a @xmath79 medium which resists expansion along the @xmath53 axis .
contrary to exph , at early times , the axial ratios of expcbc and expebc decrease , indicating that the expansion along the @xmath52 axis is faster than along the @xmath53 axis .
this trend continues for the expcbc simulation while it reverses for the expebc simulation where the top of the bubble accelerates at later times . in the expeb simulations , confinement in the @xmath53 direction by the magnetic field decreases rapidly as the bubble grows beyond approximately one scale height for all values of @xmath78
this allows the top of the bubble to accelaterate at later times , eventually leading to blow - out if the source is sufficiently strong @xcite .
breakout in our expebc simulation occurs at @xmath137 .
the constant magnetic field in all the expcb simulations confines the bubble in the @xmath53 direction , preventing blow - out . taking into account differences in density and luminosity of the source ,
our axial ratios agree fairly well with those of @xcite , while the results of @xcite agree better with the kompaneets solution .
differences between our results and those of @xcite appear to be related to differences in source luminosity and the density distribution in the atmosphere .
figure [ x2x1ratiodl ] shows the time evolution of the @xmath135 axial ratio of the low density cavity for dlh ( diamonds ) , dlcbc ( stars ) , and dlebc ( filled squares ) .
the evolution of the @xmath135 ratio of dlh shows the bubble accelerating in the @xmath53 direction .
the axial ratio at any time is larger than those measured for exph because the bubble is expanding in both the positive and negative @xmath53 directions .
dlcbc and dlebc evolve in a @xmath79 magnetic field .
the shape of the cavity in the @xmath124 plane remains nearly circular for both magnetic field configurations , even at later times .
a stronger magnetic field ( @xmath138 ) results in a cavity elongated along the @xmath127-axis . as for the exponential atmosphere ,
the bubble in the dleb simulations eventually breaks out , whereas the bubbles in the dlcb simulations remain confined by the magnetic field .
the elongation of a bubble along the magnetic field has important consequences for age estimates of the bubble from observations , which usually assume axial symmetry or even spherical symmetry .
the axial ratio @xmath132 can be observed for superbubbles in face - on galaxies , but not for galactic superbubbles which are observed from within the galactic plane ( @xmath111 plane ) . within the galactic plane ,
the available observations are the radius of the bubble in the plane of the sky and the expansion velocity perpendicular along the line of sight .
the age of the bubble derived from observations is @xmath139 , assuming an expansion law of the form @xmath140 .
the elongated shape of the bubble along the direction of the magnetic field implies that , on average , the expansion velocity is larger along the @xmath52-axis than along the @xmath54 axis .
_ an observer who looks at the bubble along the @xmath52 axis ( looking along the magnetic field lines ) will see the small dimension of the bubble along the @xmath54 axis but measure the larger expansion velocity along the @xmath52 axis .
an observer who looks at the bubble along the @xmath54 axis ( looking perpendicular to the magnetic field lines ) will see the large dimension of the bubble along the @xmath52 axis but measure the smaller expansion velocity along the @xmath54 axis .
_ the two observers will derive significantly different ages for the bubble from their observations . if the radius of the bubble along the @xmath52 axis is @xmath141 and the axial ratio @xmath132 is @xmath142 , the age derived by observer a ( looking along the field lines ) is @xmath143 and the age derived by observer b ( looking perpendicular to the field lines ) is @xmath144 on average , @xmath145 , therefore the ratio of the ages derived by the two observers is @xmath146 for a moderate @xmath132 axial ratio of 0.8 ( figure [ x1x3ratio ] ) , the ratio @xmath147 .
age estimates for observer a looking along the @xmath52 axis , and observer b looking along the @xmath54 axis were derived from our simulations .
the ratio of these age estimates is even somewhat smaller than @xmath148 , especially for lower values of @xmath78 , because the expansion of the shell actually stalls in the @xmath128 direction after a finite amount of time .
this illustrates the importance of mhd effects on the derivation of the age of a superbubble . .fitting the kompaneets model to simulations + in the @xmath120 plane at @xmath149 myr [ cols="^,^,^,^,^,^,^,^ " , ] [ overlaycompare ] the three - dimensional structure of a magnetized bubble not only affects age estimates , but also estimates of the scale height of the surrounding medium from the observed shape of the bubble .
@xcite outline a method for fitting the kompaneets model to observations to obtain the scale height of the ambient medium and the age of a bubble .
our simulations can be used to assess systematic errors introduced by applying an axially symmetric hydrodynamic model to a magnetized superbubble .
inspection of figures [ expconst_b_40 ] , [ expequip_b_40 ] , [ b40fig ] , and [ dlequip_b_40 ] shows that in our simulations the kompaneets model can only fit the shape of the cavity in an exponential atmosphere looking approximately along the magnetic field lines , or the circular shape of the cavity in a dl atmosphere looking approximately perpendicular to the field lines . in the latter case ,
fitting the kompaneets model to the circular shape of the cavity would imply a very large scale height . here
we focus on fitting the kompaneets model to the exph , expcbc , and expebc simulations , in the @xmath120 plane . to determine a range of kompaneets solutions that fit these bubbles , an initial estimate for @xmath20 ( equation [ ybasu ] )
was derived from the @xmath150 axial ratio of the simulation . starting with this estimate , overlays of the kompaneets solution
were visually fitted to the simulations to obtain a range of acceptable @xmath20 values , with the additional constraint that the source of the kompaneets model must coincide with the source in the simulations to within 2 pixels , i.e. the radius of the source in the simulations .
fits of the kompaneets model have also be used to estimate the age of a bubble , if the mechanical luminosity of the source and ambient density are known .
@xcite showed that at the level of the source , the radius of the kompaneets bubble is nearly equal to the radius of a spherical bubble described by @xcite . from equation ( [ rbasu ] ) , at @xmath6 , we find @xmath151 as @xmath20 and @xmath7 are determined from the fit , equation ( [ rbasuz0 ] ) can be substituted into equation ( [ rcastor ] ) with @xmath152 erg s@xmath153 and @xmath154 g @xmath114 to find the age of the bubble .
table [ overlaycompare ] shows results of visual fits of the kompaneets model to the exph , expcbc , and expebc simulations at an age of 7.3 myr , looking along the x@xmath46 axis .
scale height and age are listed as a function of @xmath20 for those models that provided an acceptable fit .
for all three simulations , the fitted scale height is smaller for larger values of @xmath20 , simply indicating a range of acceptable axial ratios that fit the shape of the bubble .
simulations exph and expcbc are fitted by the same range of @xmath20 while expebc is better fitted by higher values of @xmath20 .
this reflects the more elongated shape of the cavity in expebc seen in figure [ expequip_b_40 ] . for exph , we find that the scale height is close to 100 pc .
although exph and expcbc are fitted by the same range of @xmath20 , the scale height derived from fitting expcbc is smaller because the bubble is more confined .
this confinement of the bubble by the magnetic field while maintaining an elongated shape of the cavity is the root cause of the 30% to 50% lower scale heights resulting from fits of the kompaneets model to a magnetized bubble .
ages derived from fits to the magnetic simulations are consistently a factor @xmath155 smaller than the actual age of the simulation .
we note in passing that @xcite found that the kompaneets model blows out a factor @xmath155 earlier than their numerical hydrodynamic solutions .
as discussed earlier , the bottom of the bubble in the k60 model does not penetrate deeper than 1.4 exponential scale heights below the level of the source .
if a bubble would penetrate deeper into the atmosphere , an observer could conclude that the kompaneets model is not applicable .
however , none of our simulations penetrate as deep as the limiting value of 1.4 scale heights below the level of the source . in practise , this would not consitute a conclusive test whether or not the kompaneets model can be applied .
as a specific application of our simulations we investigated the result of @xcite that the scale height of the interstellar medium around the w4 superbubble @xcite is as small as 25 pc , approximately a factor 4 smaller than the canonical value of 100 pc .
@xcite investigated this issue with two - dimensional mhd simulations , and suggested that the elongated shape of the bubble could only be generated by a magnetic field perpendicular to the galactic plane ( see also west et al .
our simulations allow us to consider the elongated shape of the bubble seen by an observer looking along the magnetic field lines .
this perspective is not possible in two - dimensional simulations .
this section is organized as follows .
first we review some basic properties of the w4 region , because we found that the ambient density is a factor @xmath156 lower than adopted by previous authors .
second , we proceed with our analysis of the scale height near w4 .
@xcite showed that an expanding shell in the perseus arm associated with the cas ob6 association @xcite is a conical cavity with no apparent upper boundary in hi .
this structure was interpreted as a `` chimney '' , a large stellar wind bubble which has broken out of the galactic disk and expands into the galactic halo .
evidence for a vertical flow inside the cavity comes from elongated hi structures associated with a molecular cloud @xcite , and a comet - shaped molecular cloud @xcite .
the lower part of the bubble wall is bright in h@xmath157 and is also seen in radio continuum images @xcite .
the bright hii region at the bottom of the cavity is known as w4 @xcite .
the bubble appears to be confined at the bottom by a dense cloud .
the probable source of the bubble is the star cluster ocl 352 ( ic 1805 ) , which is located near the bottom of the chimney .
@xcite list 9 o stars in the cluster , with the earliest spectral type o4i for the star bd+60504 .
the presence of the o4 star in ocl 352 makes it likely that no supernova explosion has yet occured in the cluster , so that the bubble is a true stellar wind bubble .
most recent age determinations for the cluster are in the range 1 to 3 myr @xcite , but ages up to 5 myr have been suggested @xcite .
the kinetic energy released in the stellar wind of the stars in the cluster is @xmath158 @xcite , most of which is emitted by bd@xmath15960504 and two o5 stars .
@xcite reported an elongated h@xmath157 shell surrounding the hi cavity that appears closed @xmath160 ( 230 pc ) above the star cluster , outside the field of view of the original hi images of @xcite .
extended hi images that cover the top of the h@xmath157 shell do not show a cap in the neutral gas @xcite .
the h@xmath157 shell was interpreted as the inner wall of the superbubble ionized by the stars in ocl 352 .
@xcite found that the age of the superbubble exceeds the maximum possible age of ocl 352 by a factor @xmath161 , using the mechanical luminosity of the star cluster ( @xmath162 ) and the density of the ambient medium ( @xmath163 ) given by @xcite .
the large dynamical age of the bubble was taken as an indication that the star cluster ocl 352 by itself may not be the only source of the superbubble .
@xcite found that the age of the bubble is consistent with the age of the star cluster by applying the k60 model to the w4 superbubble , assuming a scale height of only 25 pc for the ambient medium .
however , @xcite noted that a lower ambient density could decrease the dynamical age derived for the w4 superbubble .
the original hi number density was derived by @xcite from an hi column density map .
this map was obtained by integration of the hi brightness temperature over a velocity range of @xmath164 .
the ambient density was then calculated by dividing the hi column density adjacent to the cavity by the diameter of the bubble , assuming that all the hi emission is local to the w4 superbubble .
this assumption leads to a significant over - estimate of the ambient density , because hi emission in the velocity range of the bubble originates from a much longer section of the line of sight .
we re - analyzed the hi data of the w4 region , now publicly available as part of the canadian galactic plane survey @xcite , to re - determine the ambient density .
the hi column density inside the cavity of the w4 superbubble , avoiding the hi features known to be inside the cavity @xcite , is @xmath165 .
the medium inside the superbubble is expected to be a highly ionized tenuous plasma at a temperature of a few million kelvin .
the ionized bubble wall seen on both sides of the cavity in radio continuum and h@xmath157 emission shows that the interior of the bubble is optically thin to lyman continuum photons . @xcite and
@xcite assume this to be true in their analysis of the ionization of the bubble wall by the star cluster .
therefore , hi observed inside the cavity must be located in the foreground and the background .
some of this hi emission could in principle belong to the bubble wall if the expansion of the superbubble had stalled .
there is no reason to believe that the expansion of the bubble has stalled because the source is still active .
although the evidence that the hi steamers identified by @xcite are located inside the cavity of the w4 bubble is compelling , the hi emission that seems to partly fill the cavity everywhere is likely unrelated gas , that was included in the density estimate of @xcite .
if the hi column density inside the cavity is subtracted , the density of the ambient medium is reduced to @xmath166 assuming the same line - of - sight dimension of the bubble
. a better estimate of the density of the ambient medium is obtained by dividing the mass of the shell of swept - up gas by the volume of the cavity .
this is impractical for the w4 bubble because there is no clear limb - brightened hi shell visible in the column density map .
@xcite found a higher density of 10 @xmath167 from their analysis of the shape of the ionization front , which depends on the details of their model .
although the lower value of the ambient density derived from the hidata may not resolve the age problem altogether , we follow the suggestion by @xcite that the location of the cluster is strong evidence for its identification as the source of the superbubble .
we now return to the issue of the small scale height found by @xcite .
@xcite found no evidence for such a small scale height in hi images of the region , but , as noted before , the hi images are contaminated to some extent by emission in the foreground and background .
the presence of molecular gas with a small scale height may have contributed to the result obtained by @xcite , but the molecular gas traced by the co line appears too fragmented to confine the shape of the bubble .
the scale height found by fitting the kompaneets model to the observed shape of the bubble may also be artificially small by a factor 2 ( see [ comparingkomp ] ) for an observer looking along the magnetic field parallel to the galactic plane .
this factor is not enough to explain the very small scale height derived by @xcite .
also , the age derived by @xcite would be too small by a factor 2 , further increasing the discrepancy between the dynamical age of the bubble and the age of the star cluster .
none of our simulations evolve into a cavity that is as narrow as the w4 super bubble .
polarimetric observations of the galactic plane at arcminute resolution @xcite allow measurements of faraday rotation of diffuse galactic synchrotron radiation and background compact extragalactic sources to probe the magneto - ionic interstellar medium down to parsec scales .
the plane of polarization of linearly polarized synchrotron emission rotates by an angle @xmath168 ( radians ) as the radiation passes through a magnetized plasma according to @xmath169 , with @xmath170 the wavelength of the radiation in meters .
the line - of - sight integral @xmath171 is the rotation measure ( radians @xmath172 ) , with @xmath173 the electron density ( @xmath167 ) , and @xmath174 the magnetic field vector @xmath175 ( @xmath176 ) projected onto the line of sight ( in parsec ) .
the rotation measure can be determined observationally by observing @xmath177 at different , but closely spaced wavelengths .
structure along the line of sight may create a measurable rotation of the plane of polarization , but it may not have a sufficiently large emission measure for any thermal emission to be detected . equation ( [ rmequation ] ) can not be inverted to obtain a unique solution to the magnetic field structure , even if independent information on the electron density @xmath173 along the line of sight is available . instead , the data can be interpreted by line - of - sight integration of a model distribution of the density and magnetic field , which depends critically on the assumed distributions of the electron density and magnetic field . if an ad hoc model for density and magnetic field is adopted to interpret the observations , careful consideration should be given to basic questions regarding self - consistency , uniqueness and dynamical properties of the proposed model .
our three - dimensional mhd simulations of magnetized superbubbles constitute a significant step forward in the modeling of faraday rotation by galactic superbubbles . in this paper
, we only consider the bubble as a faraday screen that does not emit synchrotron emission .
this situation applies in particular to compact polarized sources in the background .
the simulations provide a self - consistent set of models that can be used to calculate the rotation measure imposed on background emission by a magnetized superbubble .
the non - spherical evolution of the superbubble described in section [ simulations ] may have very significant effects on the predicted rotation measure .
the effect of nearby superbubbles on faraday rotation of background sources was first discussed by @xcite . in this paper
we discuss only the faraday rotation signature of distant bubbles , for which we can assume that all lines of sight are parallel .
we assume here that the bubble is completely ionized .
this allows a first exploration of the general effects of the topology of the density and magnetic field , without introducing more parameters .
figure [ rm - fig ] shows the rotation measure for a line of sight along the @xmath54 axis ( looking perpendicular to the galactic magnetic field ) and the @xmath52 axis ( looking along the galactic magnetic field ) of the simulation dlcbc ( table [ simsrun ] ) at an age of 10 myr ( figure [ b40fig ] ) .
the line - of - sight integration of equation ( [ rmequation ] ) was done through the entire simulated volume as shown in figure [ b40fig ] .
it includes some undisturbed medium in the foreground and the background of the bubble .
comparing the rotation measure maps with the density and magnetic field structure ( figure [ b40fig ] ) reveals some general characteristics of the faraday rotation of all bubbles in our simulations .
as the magnetic field is pushed aside , the field lines wrap around the expanding cavity .
the largest amplification and the largest change of direction of the magnetic field occur just outside the cavity , in the equatorial plane ( for the dl density distribution ) .
the density is also enhanced , compared with the surrounding undisturbed interstellar medium , but by different amounts depending on location in the shell .
simulations with a small scale height of the galactic magnetic field ( @xmath178 ) show this compression also at the top of the bubble , whereas simulations with a large scale height of the magnetic field ( constant b ) do not .
the largest faraday rotation therefore occurs in a relatively thin region around the cavity , being strongest close to the galactic plane ( see figure [ rm - fig ] ) .
the faraday rotation inside the cavity is much smaller by comparison , because of the low density and the low chaotic magnetic field there . if the line of sight is perpendicular to the direction of the undisturbed magnetic field , the near and the far side of the bubble wall have rotation measures up to @xmath23100 radians @xmath172 , with opposite signs .
these mostly cancel each other , but asymmetries between the front side and the back side of the shell create structure in the rotation measure distribution .
asymmetries arise from the instabilities in the shell mentioned in sections [ expcbande ] & [ analysis ] .
the strongest signal is expected from asymmetries that occur in locations where a perfectly symmetric bubble would have bent the galactic magnetic field sufficiently to create a significant line of sight component of the magnetic field .
as the magnetic field is tightly wrapped around the cavity , the highest values of the rotation measure resulting from front - to - back asymmetries are therefore expected for lines of sight that intersect the cavity , away from the vertical axis that intersects the source .
our three - dimensional simulations contain such asymmetries , resulting in the structures seen in figure [ rm - fig ] with amplitudes of the order of 20 radians @xmath172 . if the line of sight is parallel to the direction of the undisturbed magnetic field , rotation measures are much higher everywhere than in the previous case .
the front and back side of the shell reinforce each other , and high rotation measures are expected for lines of sight that intersect the shell .
the top of the bubble increases the rotation measure a few hundred parsecs above the galactic plane by @xmath179 compared with the undisturbed atmosphere in figure [ rm - fig ] . for the simulation dlebc ( small magnetic scale height ) ,
the increase is @xmath180 compared with the undisturbed atmosphere . within 100 pc from the galactic plane , the rotation measure varies with location because of the bubble , but the mean of the rotation measure across the bubble is almost the same as the mean rotation measure of the undisturbed atmosphere . here , the effect of the bubble is to create a significant variation in rotation measure for different lines of sight , with little effect on the mean rotation measure taken over a large area .
super bubble shells may therefore enhance the rotation measure in the disk - halo interface and affect estimates of the scale height of the galactic magnetic field from rotation measure analysis .
@xcite reported a large magnetized bubble associated with the gum nebula , based on a number of high rotation measures towards extragalactic sources .
recently , @xcite found this bubble in an all - sky image of depolarization of extragalactic sources , that was associated with excess rotation measure .
this bubble stands out in the data because of the high rotation measure in the shell .
the line of sight in this case is nearly parallel to the magnetic field in the local spiral arm .
so far we have discussed adiabatic simulations of super bubbles as an approximation that energy losses are compensated by an internal or an external radiation field . in this section
we briefly discuss simulations with cooling in the two main density stratifications discussed in this paper : the exponential atmosphere and the dickey & lockman atmosphere .
the cooling represents an extra term in the energy equation ( 10 ) and is implemented as an explicit source term in the code .
as with all physical processes , there is a maximum allowed time step for numerical stability associated with the cooling ; in our case it is reset to 10% of the radiative cooling time when it is below the normal hydrodynamic cfl condition .
the cooling function by @xcite and @xcite , using the abundances of @xcite , was implemented into zeus - mp in the form of analytic expressions adopted by @xcite .
although some abundances were revised downwards later , the difference between the cooling rates adopted in our simulations and the solar abundance cooling curve of @xcite is smaller than other sources of uncertainty introduced by finite resolution , and ionization by the central star cluster .
if an accuracy better than a factor @xmath161 is required in the cooling rate , the location of the superbubble in the galaxy also becomes important because of variation in the abundances with location in the galaxy ( e.g. * ? ? ?
the radiatively cooled simulations were done in atmospheres maintained at a temperature of 8000 k , with the same resolution as the adiabatic simulations discussed in section [ simulations ] , and with the same mechanical luminosity of the source . in these simulations , cooling begins at a time @xmath181 myr , soon after the first supernova explosion , at the beginning of the radiative phase following @xcite .
the effects of cooling on the shape of the cavity are evaluated at @xmath182 myr .
figure [ expconst_b_40_cool ] and figure [ dlconst_b_40_cool ] show simulations of a bubble with cooling in an exponential atmosphere and in a dickey & lockman atmosphere at an age og 10 myr .
these figures can be compared with the adiabatic simulations shown in figure [ expconst_b_40 ] and figure [ b40fig ] .
qualitatively the features in these simulations are very similar .
the cavity is elongated along the direction of the magnetic field .
the shell is thicker in directions perpendicular to the magnetic field in the undisturbed atmosphere ( @xmath183 and @xmath128 ) , while significant compression of the swept - up material in the direction of the magnetic field leads to a thinner , denser shell .
closer inspection shows that the cavities of both cooled simulations are more elongated in the direction of the magnetic field ( @xmath127 ) .
the shell also seems to display stronger instabilities , leading to irregularities in the shape of the cavity , and increased scatter in the derived axial ratios of the cavity . as expected , the simulations with cooling lead to shells that are thinner , with a higher density and stronger magnetic field , than the adiabatic simulations .
the effect of resolution on the cooled regions is to smear out small - scale high - density structure with the shortest cooling time and the lowest temperature .
@xcite found convergence in the maximum density and lowest temperature in their simulations for a grid size @xmath184 pc .
the density and temperature inside the swept - up shell in our simulations are probably also affected by resolution , but this paper focuses on the shape of the bubble , not the conditions inside the shell .
@xcite found that the shape of planetary nebulae in their simulations was not affected by resolution .
figure [ axial_ratio_cool - fig ] shows the evolution of the axial ratios as a function of time in these simulations .
we find that the axial ratio @xmath185 of a bubble in the dickey & lockman atmosphere is most affected by cooling ( compare figure [ axial_ratio_cool - fig]a with figures [ x2x1ratiodl ] and [ x2x1ratiodl ] ) . in the adiabatic simulation , the cavity remains nearly circular to an age of 10 myr , while in the cooled simulation the cavity is more elongated in the direction of the magnetic field .
note the change in evolution of the @xmath185 axial ratio just after the onset of cooling .
the effect of cooling on the axial ratio @xmath185 compared with the adiabatic case is less strong for the exponential atmosphere , but here too the cavity becomes more elongated in the direction of the magnetic field .
a stronger elongation is also seen in the @xmath131 axial ratio ( compare figure [ axial_ratio_cool - fig]b with figure [ fig : x3x1conv ] ) , but the effect is smaller than on the axial ratio @xmath185 . the observed change in axial ratios in the presence of cooling is most likely the result of increased magnetic tension exerted by the stronger magnetic field in the dense shell that surrounds the cavity .
this restricts the expansion perpendicular to the field ( @xmath183 and @xmath128 ) more in the cooled simulations than in the adiabatic simulations .
coupling between he magnetic field and the neutral shell is maintained through a small ionized fraction in the shell , and ion - neutral collisions .
a complete parameter study of simulations with cooling is beyond the scope of this paper .
at least some superbubbles have a substantial fraction of the mass of the shell ionized by the stars inside . in simulations with cooling , the cooling time of the shell
is very short compared to the age of the bubble , and a dense neutral shell forms .
ideally , radiative transport of the ionizing flux of the star cluster and any unrelated massive hot stars in the simulation volume should be included in the simulations . solving the radiative transport
every time step in the simulations is a computational challenge .
the mean free path of an ionizing photon is substantially smaller than the resolution of the simulations .
the ionizing flux decreases rapidly with time as the most massive stars explode as a supernova after @xmath23 3 myr , and the shell will eventually cool and become neutral .
the implicit assumption in this paper by applying adiabatic evolution to this problem is that heating by photo - ionization is balanced by cooling .
while this is a simplification , the observed ionization of superbubble walls suggests that heating by ionization of the shell is a non - negligible term in the energy equation of the shell that is not included in simuations with cooling only .
we present three - dimensional mhd simulations of a superbubble evolving in a magnetized medium . in these simulations ,
we assume two different atmospheric models , exponential and dickey & lockman ( 1990 ) , and two different magnetic field configurations , constant magnetic field and constant @xmath78 , for varying values of the magnetic field strength . with these simulations
we aim to study the importance of mhd effects on the interpretation of observed superbubbles . as noted before by @xcite , a superbubble in a magnetized medium
becomes significantly elongated along the magnetic field .
we find that the axial ratio of the bubble in the galactic plane is @xmath23 0.8 after 5 myr and @xmath186 after 12.5 myr , depending on the age of the bubble and the strength of the magnetic field but not on the vertical structure of either the magnetic field or the gas .
the elongated shape of the bubble in the galactic plane may lead to significant errors in the age determination of magnetized superbubbles from observations using symmetric hydrodynamic models .
the derived age depends on the location of the observer .
the ratio of the smallest age to the largest age derived by observers assuming axial or spherical symmetry is found to be proportional to the square of the axial ratio of the bubble in the galactic plane , which creates a discrepancy in the age of up to a factor @xmath234 .
we have analyzed systematic errors in the age of the bubble and the scale height of the ambient medium introduced by fitting the kompaneets model to a magnetized superbubble looking along the magnetic field lines .
the scale height may be underestimated by 30% to 50% and the age by 50% . in particular , we investigated the curiously small scale height of the interstellar medium near the w4 superbubble found by @xcite .
we re - analyzed hi data from the canadian galactic plane survey and found that the density of the ambient medium @xmath187 which is smaller than previously thought .
this lower density helps to diminish the apparent age discrepancy between the w4 bubble and the star cluster ocl 352 .
however , our analysis of the systematic errors introduced by fitting the kompaneets model shows that they are not big enough to explain the scale height of 25 pc found by @xcite for the w4 region .
we use the mhd simulations to predict the rotation measure distribution of superbubbles based on three - dimensional mhd simulations , and emphasize the importance of such simulations to make these predictions . as expected , the appearance of a magnetized superbubble depends on the perspective of the observer .
if the observer looks along the magnetic field lines , the largest rotation measures are seen at the intersection of the shell with the galactic plane .
the rotation measure is increased at larger distances from the galactic plane .
if an observer in the galactic plane looks perpendicular to the magnetic field lines , the rotation measures are much smaller , but most importantly most of the structure in rotation measure appears in projection on the low - density cavity , and not on the shell surrounding it . the simulations and analysis presented in this paper highlight the importance of three - dimensional mhd simulations of superbubbles evolving in the galactic magnetic field to the interpretation of new high - resolution images of the galactic plane at radio wavelengths from the international galactic plane survey .
this research has been enabled by the use of westgrid computing resources , which are funded in part by the canada foundation for innovation , alberta innovation and science , bc advanced education , and the participating research institutions .
westgrid equipment is provided by ibm , hewlett packard and sgi
we show that the kompaneets solution at early times can be approximated as an expanding sphere with a center that moves upward from the location of the source .
this analytic derivation explains quantitatively the behavior of the kompaneets solution in figure [ kompaneetsfig ] .
the center of the bubble along the vertical axis follows from the expressions for the top and bottom of the bubble ( equation [ zupperlower ] ) @xmath188 . \label{komp_center}\ ] ] the half - diameter of the kompaneets solution in the vertical direction is @xmath189 .
\label{komp_radius_z}\ ] ] the maximum radius @xmath190 of the kompaneets solution in a direction perpendicular to the density gradient was given by @xcite and @xcite , @xmath191 the taylor expansion of @xmath192 , @xmath193 , and @xmath190 in @xmath194 to third order in @xmath29 is @xmath195 where @xmath196 is the remainder that contains terms of order @xmath197 or higher .
similarly , we have @xmath198 and @xmath199 so the difference @xmath200 is of order @xmath201 @xmath202 we see that for small @xmath29 , the shift in the centre is @xmath203 , which is of second order in @xmath29 .
the difference @xmath200 is of third order in @xmath29 .
the kompaneets solution for small @xmath20 can therefore be approximated by a spherical bubble that rises in the atmosphere .
the center according to equation [ komp_center ] is shown in figure [ kompaneetsfig ] .
the offset is clear , where the kompaneets model is still visually a circle .
although both the numerical simulation and the kompaneets model have an axial ratio that is near unity , the difference between the kompaneets model and the numerical simulation is in the fact that the former is not centered on the source .
taylor , a. r. , gibson , s. j. , peracaula , m. , martin , p. g. , landecker , t. l. , brunt , c. m. , dewdney , p. e. , dougherty , s. m. , gray , a. d. , higgs , l. a. , kerton , c. r. , knee , l. b. g. , kothes , r. , purton , c. r. , uyaniker , b. , wallace , b. j. , willis , a. g. , & durand , d. 2003 , , 125 , 3145
|
we present three - dimensional magnetohydrodynamic ( mhd ) simulations of superbubbles , to study the importance of mhd effects in the interpretation of images from recent surveys of the galactic plane .
these simulations focus mainly on atmospheres defined by an exponential density distribution and the density distribution . in each case
, the magnetic field is parallel to the galactic plane and we investigate cases with either infinite scale height ( constant magnetic field ) or a constant ratio of gas pressure to magnetic pressure .
the three - dimensional structure of superbubbles in these simulations is discussed with emphasis on the axial ratio of the cavity as a function of magnetic field strength and the age of the bubble .
we investigate systematic errors in the age of the bubble and scale height of the surrounding medium that may be introduced by modeling the data with purely hydrodynamic models .
age estimates derived with symmetric hydrodynamic models fitted to an asymmetric magnetized superbubble can differ by up to a factor of four , depending on the direction of the line of sight .
the scale height of the surrounding medium based on the kompaneets model may be up to 50% lower than the actual scale height .
we also present the first ever predictions of faraday rotation by a magnetized superbubble based on three - dimensional mhd simulations .
we emphasize the importance of mhd effects in the interpretation of observations of superbubbles .
|
the quiet sun can be defined as that part of the sun far from magnetic activity .
if we exclude plages and network , the remaining quiet area covers 90% of the surface of the sun , and remains almost unchanged during the solar cycle ( e.g. , * ? ? ?
few decades ago , this internetwork ( in ) was thought to be almost devoid of magnetic fields and consequently , its relevance for the solar magnetism was not obvious @xcite .
however , with the present instrumentation , when the spatial resolution is around 1@xmath0 , the quiet sun turns out to be full of magnetic fields @xcite , with an unsigned flux density measured with techniques based on the zeeman effect of the order of 10 g ( e.g. , * ? ? ?
* ; * ? ? ?
the analysis of observations with a spatial resolution twice better yields an unsigned flux twice larger @xcite .
there are reasons to believe that these resolutions are not enough to completely resolve the magnetic structures ( see * ? ? ?
* and references therein ) .
recent measurements based on the hanle effect find an unsigned flux of , at least , 60 g @xcite , which is much higher than the magnetic flux observed with zeeman techniques .
considering the fraction of solar surface corresponding to the quiet sun , the flux measured in the in is comparable to ( or larger than ) that coming from all active regions even in the maximum of activity .
thus , the quiet sun may play a significant role in the physical processes governing the solar global magnetism ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the topology and the scale of variation of the quiet sun magnetic fields are not yet clear .
there is evidence for the existence of mixed polarities , at least when the spatial resolution is 1@xmath0 @xcite .
the small - scale magnetic structures in the quiet sun seem to be smaller than 50100 km ( see * ? ? ?
* ; * ? ? ?
this size assumes one single structure in the resolution element , however , it is to be expected that the dynamical influence of the granulation on the magnetic fields produces a distribution of many smaller structures .
this fragmentation shows up in mhd simulations of turbulent dynamos @xcite , and 3d magneto - convection ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
recently , @xcite studied various mechanisms proposed to produce intense magnetic concentrations and limited the radii of the magnetic flux tubes to less than 10 km , which is approximately the magnetic diffusion length scale @xcite .
this scenario was already proposed by @xcite with an atmosphere formed by optically thin magnetic structures with diverse properties ( see also * ? ? ?
* ) . such a scheme is called misma ( micro - structured magnetized atmosphere ) . an inversion procedure based on it
has been able to fit the whole variety of asymmetric stokes profiles for the intensity , @xmath1 and @xmath2 for the two orthonormal states of lineal polarization and @xmath3 for the circular polarization .
the stokes profiles of a spectral line give the variation with wavelength of these four parameters .
if the spectral line is formed in a homogeneous atmosphere , the stokes profiles are expected to be symmetric or antisymmetric ( e.g. , * ? ? ?
* ) . ] produced by the quiet sun magnetic fields @xcite .
observational studies based on infrared ( ir ) spectral lines find magnetic field strengths in the range below 1 kg ( sub - kg ; * ? ? ?
* ; * ? ? ?
* ) , while works using lines in the visible spectral range tend to find kg fields @xcite .
kilogauss fields are also inferred from the presence of a considerable number of g - band bright points in the quiet sun @xcite .
@xcite put forward a solution for this apparent contradiction : the co - existence in the resolution element of a distribution of field strengths going from zero to kg .
they show how to reproduce the observed bias with the vertical gradient of the field strength induced by the hydrostatic equilibrium in the atmosphere .
another possibility with the same consequences is a horizontal gradient @xcite .
the presence of a range of field strengths from sub - kg to kg fields bias the measurements based on ir stokes profiles to sub - kg fields .
such a particular behavior is due to the high sensitivity of the ir lines to magnetic fields .
the fe i @xmath415648 ( used by * ? ? ?
* ; * ? ? ?
* ) is magnetically saturated signals no longer have a linear relationship with the magnetic field strength .
the level of polarization stays constant while the splitting of the stokes v lobes grows in proportion to the field strength .
] at 400 g while the visible pair , ( e.g. , used by * ? ? ? * ; * ? ? ?
* ) saturates at 1.5 kg .
the ir polarization signals coming from the strongest fields are spread in a wide range of wavelengths and they produce little signal . on the contrary ,
weak fields saturate the ir line and a significant stokes v signal is built up in a narrow wavelength range .
weak field and strong field components appear in the same wavelengths in the visible lines so that their relative contribution is proportional to their magnetic flux .
the weak field component can dominate over the strong component in the ir stokes @xmath3 profiles , while the strong fields dominate over the weak ones in the visible profiles .
thus , an incomplete analysis of the ir profiles leads to weak fields , whereas the same analysis of the visible lines shows strong fields . in order to test this conjecture , @xcite presented simultaneous spectropolarimetric observations in the visible and ir ranges .
an independent milne - eddington inversion of the visible and ir lines suggests the co - existence of sub - kg and kg fields .
however , a more sophisticated and simultaneous inversion is desirable to provide further support to the conjecture of @xcite and @xcite .
this paper presents such simultaneous inversions under the misma approximation of the visible and the infrared stokes profiles used in @xcite .
as we discuss above , it seems to be appropriate for reproducing the complex stokes profiles produced by the quiet sun .
section [ obs ] summarizes the observations and the data reduction .
some properties of interest of the magnetograms obtained in the two spectral ranges are shown in @xmath5 [ magn ] .
the strategies for the data analysis and the inversion are explained in @xmath5 [ data ] .
we present the results in @xmath5 [ res ] whose consistency is analyzed in @xmath5 [ cons ] .
a final [ con ] discusses the main conclusions .
we re - analyze the spectra used by @xcite whose properties are only sketched in the original reference . for the sake of comprehensiveness ,
the data and reduction are detailed here .
the part of the reduction in which we bring the visible and ir spectra to equal spatial resolution is new .
we use two telescopes for the simultaneous visible and ir observation .
the observation was performed at the spanish observatorio del teide ( tenerife , spain ) .
the ir spectra were gathered with the tenerife infrared polarimeter ( tip , * ? ? ?
* ; * ? ? ?
* ) operated at the german vacuum tower telescope ( vtt ) .
spectrograms of the full stokes vector ( @xmath6 ) were obtained in the ir iron lines fe i @xmath415648 ( land factor @xmath7=3 ) and fe i @xmath415652 ( @xmath7=1.53 ) .
the visible observations were carried out with the french - italian telescope themis ( tlscope heliographique pour ltude du magntisme et des instabilits solaires ) using the spectropolarimetric mode mtr @xcite .
the lines analyzed are the pair ( @xmath7=1.66 ) and ( @xmath7=2.5 ) .
the atomic parameters of the four spectral lines are listed in table [ tbatomic ] . in order to obtain spectra in a two - dimensional field of view ( fov ) , the solar surface was scanned from east to west with a step size of 05 .
the slit width was set to 05 in both telescopes and the scan consisted in 60 positions . to improve the signal to noise ratio , the integration time
was set to 30 s per position .
lcccc 6301.499 & 3.65 & 9.766@xmath8 & @xmath9p@xmath10 @xmath9d@xmath10 & 1.67 + 6302.492 & 3.69 & 2.630@xmath8 & @xmath9p@xmath11 @xmath9d@xmath12 & 2.5 + 15648.515 & 5.426 & 8.035@xmath8&@xmath13d@xmath11 @xmath13d@xmath11 & 3 + 15652.874 & 6.246 & 3.443@xmath14&@xmath13d@xmath15 @xmath13d@xmath16 & 1.53 + [ tbatomic ] this work analyzes data obtained on august 10 2002 in a quiet sun region close to disk center ( @xmath17 ) .
we choose a region without magnetic activity as judged with the help of k and h@xmath18 slit - jaw images from video cameras at the vtt .
a video link between both telescopes providing the slit - jaw images from vtt at themis was helpful to point both telescopes to the same target .
since the absolute pointing of the telescopes was not accurate enough , and the slit - jaw images of the granulation did not allow us to recognize the region pointed by the two telescopes , we used another _ differential _
method for accurate pointing .
first , both telescopes were pointed to a characteristic small structure which could easily be identified , e.g. , a small pore .
the pointing of the vtt was moved to a quiet sun region , and then themis pointing was displaced by the same distance . as we explain below , the scans provided by the vtt and themis are post - processed for fine - tuning co - alignment .
this final step removes residual pointing errors plus some additional effects ( e.g. , differential refraction between the visible and the ir ) .
the scan under study here has an offset between telescopes of only 1(see below ) . since we scanned with a step size of 05 and with an exposure time of 30 s per step
, the time lag between ir and visible data of the same point corresponds to only 1 minute .
the data reduction of the two data sets was carried out in a similar way .
a flat - field correction was applied to the spectra . in order to get the stokes parameters , we demodulate the measurements with the calibration matrix , and finally we correct for seeing - induced crosstalk by combining the two beams of each polarimeter .
the crosstalk produced by instrumental polarization ( ip ) of the telescopes themselves was considered too .
since themis is free of ip , the visible data do not need correction .
due to the configuration of mirrors in the vtt , the ir data are affected by an important crosstalk between the different stokes parameters .
it was calibrated with a model mueller matrix set using a linear polarizer mounted at the entrance window of the telescope ( see * ? ? ?
lcc & + 50 & -100 + & -80 & -150 + fe i @xmath415648 & -370 & -400 + fe i @xmath415652 & -430 & -400 + [ tbcb ] since the purpose of this work is the simultaneous inversion of visible and ir lines , one needs a consistent scale of wavelengths in both spectral ranges .
we carried out an absolute wavelength calibration so that the lines are unshifted in this scale when the solar plasma is at rest .
the central wavelength of and was considered to be the zero - crossing of the stokes @xmath3 profiles corresponding to the signals in the network points of our fov .
such a reference gives an accuracy of 200 ms@xmath19 @xcite .
@xcite find that the zero - crossing wavelength depends on the spectral resolution .
using their simulations , we check that our resolution of 21.9m per pixel does not threaten the absolute accuracy given above .
the strongest polarization signals were chosen to trace the network ( @xmath20 with @xmath21 the intensity of the continuum near the lines ) .
weaker signals tend to be red - shifted .
since this method has not been tested for the ir lines , their absolute wavelengths need another way of calibration .
we computed the mean line profile from the data and then corrected for the expected convective blue - shift .
we average the stokes @xmath22 profile over the fov .
the minima of the lines were fitted with fourth order polynomials whose minima were taken as the central wavelengths of the spectral lines .
these wavelengths corrected for the convective blue - shifts provide the reference for an atmosphere at rest .
table [ tbcb ] presents the convective blue - shift for fe i @xmath415648 and fe i @xmath415652
obtained from a two - component inversion of many line profiles in the quiet sun @xcite and from line synthesis from numerical simulations of granulation ( * ? ? ?
* private communication ) .
we can not asses the accuracy of our absolute wavelength calibration based on them , however , the fact that the two numerical simulations provide the same convective blue - shift suggests the consistency of the approach .
one of the most important steps in the data reduction is the spatial co - alignment of the visible and ir scans .
the aim of this process is to ensure that the scans are not only taken from the same region on the sun but also that each individual pixel of the visible and ir scans come from the same point of the in region .
the correct alignment must be guaranteed for a meaningful analysis of the data , i.e. , for the simultaneous interpretation of both data sets . in order to align the scans we displace and re - scale them allowing for five degrees of freedom .
first , we correct for differences in the spatial scales along the @xmath23 and @xmath24 directions . we define @xmath23 as the direction of scanning and @xmath24 as the direction along the slit .
the step width of the scans was slightly different in both telescopes ( less than 1% ) , and the spatial scales along the slit were computed from the focal length of the telescopes . their ( relative ) agreement with respect to each other were checked using reference images of a sunspot .
then , we found the shift in the @xmath23 and @xmath24 directions by cross - correlation of the visible and ir continuum images .
it corresponds to the offset of 1@xmath0 mentioned above .
finally both slits were not exactly oriented with the same angle with respect to the sun .
there was a small rotation of 1@xmath25 which we also determine by cross - correlation .
once the correspondence between the visible and ir spatial coordinates was set , we interpolate the visible spectra to obtain the spectra at the positions of the pixels in the ir scan .
it is important to have the same spatial resolution in both data sets .
however , the angular resolution of the observations at the vtt was better than at themis due to the use of a correlation tracker @xcite . to solve this problem ,
the ir data were degraded to the resolution of the visible data by convolution with a gaussian .
the contrast of the ir continuum image depends on the angular resolution .
we set the width of the gaussian , and so the angular resolution of the ir map , by comparison with the synthetic images from numerical simulations of magneto - convection @xcite .
specifically , we tune the ir contrast so that the ratio between the visible and the ir contrasts agrees with the ratio obtained from numerical simulations ( 0.48 ) .
such theoretical ratio was found to be almost independent of the angular resolution and in excellent agreement with the value measured by @xcite using a single telescope to obtain simultaneous visible and ir data .
the central panel in figure [ img ] shows the continuum image in the ir after smearing .
the contrast of the ir image is 0.9% .
the right panel of figure [ img ] contains the continuum intensity image of the visible data ( contrast 1.9% ) .
the squares point out regions easy to identify in both images showing the goodness of the alignment . after all data reduction , including smearing of the ir data and fourier filtering of the visible data , the level of noise in polarization for stokes @xmath3 is ( 2 - 3)@xmath26 for the ir data and 5@xmath26 for the visible data .
a more detailed description of the telescopes and the data reduction can be found in @xcite .
descriptions of the instruments can be found in @xcite and @xcite for tip , and in @xcite and @xcite for themis .
figure [ virmag ] shows the magnetograms calculated from the stokes @xmath3 profiles of and fe i @xmath415648 .
the longitudinal magnetic flux density or @xmath27 was computed with the magnetograph equation under the weak field approximation using the same method as @xcite .
the magnetograms thus obtained are equivalent to the routine magnetograms ( e.g. , that obtained with the mdi instrument on board of soho ) , if we exclude the velocity fields .
figure [ virmag ] displays both magnetograms with the same scale of magnetic flux densities . at first sight , both images are similar with the exception that the signals found in visible are larger than the signals found in ir .
indeed , this is the first important difference .
since only the in fields are of interest here , the network region pointed by a circle in the lower left part of our fov was discarded from the analysis by rejecting pixels with flux densities larger than 20 g in the visible and 10 g in the ir .
the noise levels turned out to be 2 g in the visible and 1 g in the ir . the method to estimate
the noise is detailed in @xcite .
the unsigned flux density obtained from pixels in the fov with flux density below the upper limit and above noise was , @xmath28 in both cases the pixels with signals cover 45% of the total fov ( the rest of the pixels were set to zero in the calculation of @xmath29 ) .
under the same conditions of noise and spatial resolution , the measurements in the visible detect more flux than in the ir .
this suggests that visible and ir lines trace different magnetic structures .
the difference of flux is conspicuous , and one may be tempted to ascribe it to the approximations used for @xmath27 .
however , more elaborated techniques show the same behavior ( me inversions in * ? ? ?
* ; * ? ? ?
* and 5 ) . some magnetic flux concentrations are very similar in both magnetograms ( squares in the upper part of figure [ virmag ] ) but some others differ ( see the other two squares )
. there are structures having one polarity in the ir magnetogram and the opposite in the visible one .
this effect is evident even in the network area with large polarization signals ( see the circle in fig . [ virmag ] ) .
such a result can be considered as an additional evidence of the different sensitivity to magnetic fields of the visible and the ir lines . the magnetic fields in the visible magnetogram ( fig .
[ virmag ] right ) exhibit a preferred polarity while the ir magnetogram ( fig . [ virmag ] left ) looks well balanced .
the signed flux was measured to be , @xmath30 this imbalanced flux is similar to that measured in other visible observations ( e.g. , * ? ? ?
* ; * ? ? ?
yet , @xcite found almost no signed flux using the same ir lines ( they observed 0.2 g ) .
all these results indicate that visible and ir lines may trace different magnetic structures , as conjectured by @xcite .
in order to infer the magnetic field strengths from the observations , we carried out an inversion of the stokes profiles under the misma hypothesis .
micro - structured magnetized atmospheres ( mismas ) were introduced for the first time by @xcite , with the objective of reproducing the asymmetric stokes profiles observed in most solar structures .
unresolved optically thin micro - structures naturally render asymmetric stokes profiles .
the misma scenario turns out to be an appropriate scheme of analysis since the observed quiet sun stokes profiles are very asymmetric , and it has been shown to work with the asymmetries of and @xcite . in this section
we give a description of the procedure followed to invert the data .
we only analyze those stokes @xmath3 profiles clearly above noise . a threshold of three times the noise level was chosen in order to separate clear polarization signals from noisy profiles .
pixels with profiles whose peak polarization is above the threshold in both the ir and the visible data cover 40% of the fov .
this data set includes some 2280 pixels . following @xcite the stokes @xmath3 profiles in the selected data set were classified using a principal component analysis algorithm ( pca ) .
the aim is twofold . on the one hand
, it reduces the number of representative profiles and , on the other hand , it provides low noise profiles to perform inversions used as initialization for the inversion of individual stokes profiles .
the classification is carried out simultaneously for the visible and the ir stokes @xmath3 profiles .
the profiles of the four lines were scaled to the blue lobe of the stokes @xmath3 profile .
the pca method gives 38 classes .
all profiles with large polarization signals were included in one single class to avoid contaminating the other classes with network profiles .
the stokes @xmath3 threshold was chosen to be 1%@xmath21 in the visible , and this special class includes all profiles from the network patch in the bottom left of our map ( see fig .
[ virmag ] ) .
the large number of classes compared with those obtained for visible lines by ( * ? ? ?
* classes ) or for ir lines by ( * ? ? ?
* classes ) is due to the combination of the different asymmetries found in the stokes @xmath3 profiles of the two pairs of lines .
note that , in general , the classes found in this work do not correspond to those obtained by @xcite and by @xcite .
we represent each class with the average of all profiles corresponding to the class .
the individual profiles are not scaled to carry out the average .
we only multiply them by the sign of the blue lobe of .
some of these classes are shown in figures [ clas1 ] and [ clas2 ] ( the dots ) .
there is a clear improvement in the signal to noise ratio ( s / n , defined as the maximum of the unsigned stokes @xmath3 profile divided by the noise ) from the individual profiles to those representing the class .
the individual profiles have a typical s / n between 3 and 10 , whereas the two first classes have a s / n of 100 .
the ir profiles from classes 0 and 4 have tails extending far out from their extremes , as those predicted by @xcite .
these tails come from the contribution of kg field strengths .
such profiles are similar to class 0 in ( * ? ? ?
* fig . 1 ) .
the first 6 classes include 42% of the total number of profiles .
one conspicuous fact is that some classes ( e.g. , classes 11 and 20 ) show one polarity in the visible and the opposite in the ir . such a result was expected from the magnetograms in figure [ virmag ] .
approximately 20% of the profiles show the effect ( see also * ? ? ?
the misma inversion code is described by @xcite .
we will follow a strategy similar to @xcite . here
we summarize the main characteristics of the underlying model atmospheres : 1 .
there are four different components , one non - magnetic and three magnetic .
note that the inversion of visible lines performed by @xcite require two magnetic components . as @xcite discuss , it is impossible to reproduce even the mildest observed asymmetries with only one component .
here we need one additional component to characterize the weak magnetic field strength revealed by the ir lines @xcite .
the components are embedded in a background producing unpolarized stray light .
the magnetic fields are vertical , along the line of sight .
since the observed stokes @xmath1 and @xmath2 signals are rarely above the noise level , it is reasonable to assume the inclination of the magnetic field to be 0@xmath25 or 180@xmath25 .
the inversions were performed only for stokes @xmath22 and stokes @xmath3 .
this approach allows us to compute longitudinal magnetic fields , however , the observations are compatible with inclined fields producing a linear polarization hidden by noise ( see * ? ? ? * ) .
the implications of this assumption are discussed in @xmath5 [ incl ] .
the temperature of the components is forced to be the same . since the magnetic structures are optically thin the radiative exchange between components is very efficient , washing out temperature differences in seconds .
the four components are in lateral pressure balance .
motions are only allowed along magnetic field lines .
the temperatures are retrieved at four fixed heights in the atmosphere .
the parameters characterizing each component are set at the bottom of the model atmosphere .
they are the magnetic field strength , the occupation fraction , the bulk velocity , and the broadening parameter .
the stratification of these parameters is obtained under the assumption of hydrostatic equilibrium and horizontal pressure balance .
three parameters account for the stray light , and another one parametrizes both the solar macroturbulence and the instrumental broadening due to limited spectral resolution . despite the fact that the spectral resolution in the visible and the ir differ by 30%
, one can use a single macroturbulence for the two wavelength domains because the macroturbulence inferred from the inversions is completely dominated by true solar motions rather than instrumental smearing .
the total number of free parameters amounts to 25 .
they are used to fit some 400 observables ( each wavelength position of the stokes @xmath22 and @xmath3 profiles of the four lines ) .
an example of the fit to the intensity profiles is shown in figure [ clasi ] . the observed profiles ( circles ) exhibit line blends .
these are a telluric line close to the red wing of , a solar fe i line at 15647.3 , and an unidentified possibly telluric line in the red wing of fe i @xmath415648 . to avoid the contamination by these blends , the stokes @xmath22 profiles at the wavelengths of the blends
are excluded from the fit .
there are also two oh bands of solar origin close to fe i @xmath415652 , however , these bands are only visible in cool solar regions such as sunspots .
the intensity profiles in figure [ clasi ] are those from class 0 .
all models have three magnetic components . in order to distinguish between them
, we denote them as first , second and third components .
the first component is that carrying the largest mass , as was defined by @xcite , and it is followed by the second and the third .
if the three components would have the same magnetic field strength , the first would be the one producing the largest polarization signal . figure [ mod ] displays the model misma of class 0 ( see the stokes @xmath3 in fig .
[ clas1 ] ) .
the different physical parameters of the non - magnetic , the first , the second , and the third magnetic components are plotted with solid , dotted , dashed and dot - dashed lines , respectively .
the first magnetic component presents a weak magnetic field strength ( @xmath31100 g at the base of the photosphere ) , while the other two have strong field strengths ( larger than 1 kg ) .
the density of the first magnetic component is the same as for the non - magnetic component ( fig .
[ mod ] bottom - left panel ) whereas the other two components have much lower density .
the reduction is forced by the coupling between field strength and density produced by the lateral pressure equilibrium .
the higher the magnetic field the lower the gas pressure and , for the same temperature , the lower the density .
figures [ clas1 ] and [ clas2 ] show the fits produced by the misma inversion of 12 representative classes .
all kinds of observed profiles are properly reproduced with three magnetic components per resolution element . from now on , all results correspond to a single height in the atmosphere .
we select the base of the quiet sun photosphere , defined as the height where the total pressure equals to @xmath32 dyn@xmath33 . invoking lateral pressure balance among the individual model mismas ,
this definition sets a single geometrical height for all the inversions , and it corresponds to the height where the continuum optical depth is one in the 1d quiet sun model atmospheres ( e.g. , * ? ? ?
the study is focused on two particularly important physical parameters , namely , the magnetic field strength and the occupation fraction .
the latter corresponds to a volume filling factor . the magnetic field strength inferred from the misma inversion of each class is also given in figures [ clas1 ] and [ clas2 ] .
the magnetic field of the components are denoted as b@xmath11 , b@xmath10 and b@xmath34 , in order of decreasing mass .
very often the inversions require one component with a sub - kg field strength and two with strong kg fields .
such a behavior is typical of the most abundant classes ( e.g. , classes 0 , 1 , 4 ) .
the fe i @xmath415648 stokes @xmath3 profiles of these classes are similar to that expected from the synthesis by @xcite .
the lobes of the profiles have a splitting equivalent to that produced by a weak field , however , they exhibit extended wings characteristic of strong fields .
some other classes have all components with fields in the range clearly below 1 kg ( e.g. , classes 5 , 19 , 20 ) . in this case , there are no extended wings in the profiles of fe i @xmath415648 .
finally , note the presence of very asymmetric profiles .
the ir profiles tend to have a higher degree of asymmetry .
some 28% of the analyzed pixels present ir profiles with three lobes ( e.g. , class 5 ) .
we found a fraction similar to that observed by ( * ? ? ? * % ) .
there is also a small fraction of very asymmetric ir profiles with four lobes ( 3% , classes 14 and 30 ) , while there are only a few visible profiles with three lobes ( 3% , e.g. , class 30 ) .
the model atmospheres obtained from the classes were used to initialize the inversion of the individual stokes @xmath3 profiles belonging to the class . from this moment on we refer only to results obtained from the inversion of the 2280 individual profiles .
the histograms of the distribution of magnetic field strength obtained from the inversion of the individual profiles are displayed in figure [ hmag ] .
they show the percentage of pixels within bins of 50 g for the different components .
the solid , the dashed , and the dotted lines distinguish the first , the second , and the third magnetic components , respectively .
the results from the network patch were omitted ( see [ magn ] ) .
the first component tends to have much weaker fields than the other two and it has a shape similar to the distribution obtained by @xcite using only ir lines .
the mean field obtained for this component is 290 g. the second and the third components show the same behavior as the histograms obtained with visible lines by @xcite .
they have mean fields of 970 and 1100 g respectively .
despite the similarity between the magnetic field strengths of the second and the third components , they are well separated by the inversion code .
there are large differences of doppler shift between them ; 85% of the models are separated by more than 500ms@xmath19 at the base of the photosphere , and this difference increases substantially in the heights of formation of the spectral lines .
( see also @xmath5 [ three ] . )
the top panels of figure [ map ] show the maps of field strength for the three magnetic components . the magnetic field of the first , the second and the third components are displayed from left to right .
the network point ( bottom left ) is almost the only feature with strong kg fields in the first component .
most of the pixels show at least one component with strong fields . more specifically ,
92% of the analyzed pixels possess at least one field larger than 500 g while the fraction with field strengths above 1 kg is 74% .
the map with the filling factor of the magnetic fields ( the sum of occupation fractions of the three components ) is displayed in the bottom left panel of figure [ map ] .
it goes from 10% in network points to 0.5% , and the mean value is approximately 4% . if we consider the whole fov ( all values in the pixels not analyzed are set to zero ) , and we exclude the network profiles , the measured magnetic fields fill only 1.5% of the in region .
the bottom right panel of figure [ map ] gives the map of the total magnetic flux density computed as the product of the occupation fraction times the magnetic field strength .
note that the flux density is given in g ( or mx@xmath33 ) , thus it can be directly compared with other measurements with different pixel size or fov .
most parts of the analyzed surface present a flux density larger than 25 g ( sum of the three components ) . considering the whole map
, we measure a total flux density of 9.6 g which is similar to the value obtained by @xcite or @xcite with better spatial resolution .
figure [ bf ] shows the variation of the magnetic field strength of all components with the magnetic flux density .
the shaded area represents the standard deviation above and below the mean value .
a difference between the two standard deviations reveals the asymmetry between the distribution of field strength above and below the mean value . there is a strong tendency for the field strengths to weaken when the magnetic flux decreases .
@xcite obtained a less pronounced dependence , but their lowest detectable flux density was 10 g while in this work it decreases down to 1 g. 50% of the analyzed pixels can be fitted only with mixed polarities in the model misma . since we fit 40% of the total pixels ,
this fraction corresponds to 20% of the fov .
this large amount of pixels includes all profiles with three or four lobes in the ir and/or visible ( 30% ) , and also those profiles with a clear polarity in the visible stokes profiles and the opposite in the ir profiles ( 20% ) .
we find that the fields of those pixels are weaker than the rest .
figure [ hmag2 ] shows the histograms of the field strength as in figure [ hmag ] but only for models including mixed polarities .
all components show fields weaker than in figure [ hmag ] .
from the first , the second and the third magnetic components we retrieve a mean field of 250 , 700 and 880 g , respectively .
figure [ mix ] gives the percentage of mixed polarities as a function of the stokes @xmath3 signal of .
the weaker the polarization signal the larger the fraction of mixed polarities .
this behavior was already noted by @xcite and @xcite but they obtained a maximum fraction of 25% .
the fraction in figure [ mix ] flattens to 65% for stokes @xmath3 lower than 0.002 .
lccccc[!t ] total & 1.50 & 9.6 & 500 & 100 & 100 + b<500 g & 0.87 & 1.2 & 10 & 10 & 2 + b>500 g & 0.63 & 8.4 & 490 & 90 & 98 + b>1000 g&0.42 & 7.0 & 450 & 75 & 90 [ tbpdf ] we compute the probability density function ( pdf ) of the magnetic field strength from the magnetic field and occupation fractions procured by the misma inversions of the individual profiles .
the pdf gives the probability of finding a specific magnetic field strength in the observed in region .
the range of magnetic fields from 0 to 2000 g was divided into bins of 50 g. each individual intrinsic field @xmath35 is assigned to one bin . the probability to find the magnetic field of a single bin is the sum of the occupation fractions of all pixels with @xmath35 belonging to this bin .
we normalize the pdf so that its integral provides the filling factor of magnetic structures in the in region assuming that the pixels without signals have no magnetic field strength .
the pdf of the magnetic field strength condenses information on the quiet sun magnetism that can be used for comparisons with other observations and numerical simulations .
the first moment @xmath36 is connected with the magnetic flux density and the second moment provides the magnetic energy density @xmath37 : @xmath38 @xmath39 figure [ pdf ] shows the pdf of our in region . once again
, we have avoided those pixels belonging to the small network patch in the fov .
the shape of the pdf is to be expected from the histogram of figure [ hmag ] .
most of the filling factor is concentrated in the weak fields , while most of the magnetic flux comes from the strong fields .
table [ tbpdf ] summarizes the properties of the pdf .
it shows the filling factor , the flux density , the fraction of flux , and the fraction of energy in four different ranges of magnetic field strengths , i.e. , the full range , the weak fields ( @xmath40 g ) , the strong fields ( @xmath41 g ) , and the kg fields ( @xmath42 g ) .
three quarters of the total observed flux and 90% of the magnetic energy is in the kg fields , despite they represent only 30% of the total filling factor .
there is an accumulation of field strength between 1600 and 1800 g. it may be due to the combined effect of a magnetic intensification mechanism plus the existence of an upper limit that the magnetic field strength can not exceed .
this limit of 1800 g corresponds to a magnetic pressure equal to the gas pressure of the non - magnetic plasma at the base of the photosphere .
the magnetic intensification mechanism would try to pile up magnetic fields right before the upper limit .
when we average over the fov , the _ signed _ magnetic flux does not cancel completely .
the flux density , taking into account the direction of the magnetic field , turns out to be @xmath43 g. this net flux was expected from the results in [ magn ] ( equation [ [ virm2 ] ] ) .
@xcite already observed the presence of an imbalanced flux in almost all the observation he studied .
he measured an imbalance ( defined as the signed flux divided by the unsigned flux ) between @xmath44 and @xmath45 , and found a direct correlation between the imbalance of the in fields and the imbalance of the network fields around them .
our imbalance is @xmath46 and our network region is negative almost everywhere .
figure [ pdf2 ] shows the pdf of the signed magnetic field strength , i.e. , the magnetic field strength times the sign of the vertical component of the magnetic field . there is a large contribution of strong fields in the negative flux part of the pdf , while the positive part has a continuous fall off from weak to strong fields .
the weak fields are well balanced while the strong fields have a preference toward negative polarity , i.e. , the same polarity as the network patch . for fields weaker than 500
g the imbalance is only @xmath47 , and for field strengths stronger than 1 kg the imbalance is @xmath48 .
@xcite already observed the preference of the strong fields towards one polarity , and figure [ pdf2 ] shows a clear observational evidence of such an effect .
some of the properties of the observational material used in the work , and the assumptions of the inversion , are not ideal but dictated by feasibility . in order to strengthen the validity of the results
, we devote this section to analyze the assumptions and results that may be more critical . our visible and ir spectra average the in properties in spatial scales of 15 arcsec and temporal scales of 30 sec ( [ obs ] ) .
the time lag between visible and ir spectra would be of concern only if this spatio - temporal average changes significantly during 1 min .
fortunately , large changes of the average properties are not expected for a number of reasons .
first , the integration time of the individual measurements smears out the short time - scale variations of the in fields , and only variations in the range between 30 sec and 1 min are of concern .
second , low spatial resolution observations of in magnetic fields show how many structures remain stable for tens of minutes , and longer ( e.g. , * ? ? ?
* ; * ? ? ?
obviously , our time lag does not represent a problem for these structures . in view of the results coming from high resolution observations and numerical simulations of magneto - convection ,
this time stability can not not be interpreted as the stability of individual magnetic structures but the stability of the average properties .
for example , a strong granular downdraft lives longer than the granules and , during its lifetime , it continuously advects magnetized plasma towards a specific point of the solar surface @xcite
. observed with low resolution , one would detect a magnetic patch that remains in place for a long period of time , despite the fact that it is formed by many small - scale short - lived magnetic structures that are continuously advected and engulfed by the downdraft .
finally , for the average magnetic properties to present significant variations during the time lag , the typical velocities of the magnetized plasma would have to be unphysically high .
consider the horizontal velocities required to sweep out during the time lag all magnetic elements existing in a resolution element . in this case
the magnetic structures in the same resolution element of the visible and ir maps will be completely different , questioning any interpretation based on their simultaneity .
the required velocities would have to be of the order of the speed needed to cross a resolution element during the time lag , i.e. , 18 km s@xmath19 .
such speed is too high to be common .
it is supersonic , and much larger than the observed vertical velocities .
obviously , the lack of simultaneity becomes more problematic as the spatio - temporal resolution of the observation improves . from our point of view
, the spatial and temporal integration of the measurements represents a problem more serious than the time lag and the small spatial resolution differences that may remain between the ir and the visible data sets , .
the polarization detected in each resolution element results from adding up contributions from many magnetic structures having very different properties .
our inversions provide only a crude representation of what may be happening in the real sun .
this problem , however , is not specific of our work but affects all previous observational works on the in magnetic fields . in the case of the misma inversions
the problem is ameliorated because a sketchy model misma actually represents many different underlying atmospheres , including very complex ones .
all those atmospheres having the same average properties produce the same spectrum , and these average properties are the ones that the inversions retrieve ( see the discussion in * ? ? ?
the inversions were carried out under the assumption of longitudinal magnetic fields , which minimizes the number of free parameters and simplifies the inversions .
this assumption , however , biases the inferred pdf(@xmath35 ) .
fortunately , the importance of the bias can be estimated and it turns out to be small . if a magnetic field has an inclination @xmath49 with respect to our vertical line - of - sight , the stokes @xmath3 spectrum remains approximately as in the case of vertical field but scaled with the factor @xmath50 .
this is a general property of the radiative transfer equation , and is not restricted to weak fields on plane - parallel atmospheres ( see * ? ? ?
* ) . consequently ,
if an observed stokes @xmath3 profile is reproduced with the wrong inclination , all the parameters of the inversion would remain the same except for the occupation fraction @xmath51 .
the occupation fraction provides a global scaling factor for the stokes @xmath3 profiles , and so it absorbs the full bias .
our pdf(@xmath35 ) is biased because it is based on occupation fractions underestimated by a factor @xmath52 , @xmath53 with the subscript @xmath54 denoting one of the @xmath55 pixels with a field strength @xmath35 , and the symbol @xmath56 standing for a normalization constant .
obviously , @xmath57 sets a lower limit to the true @xmath58 , @xmath59 since @xmath60 . on the other hand , it is possible to set an upper limit by estimating the maximum inclination allowed by the noise in stokes @xmath1 and @xmath2 .
following , @xcite , we synthesize the linear polarization produced by our model atmospheres when the magnetic field of the three components is inclined . for each model misma
we determine the largest inclination @xmath61 producing undetectable linear polarization , i.e. , below the observational noise .
if we use this inclination to correct the occupation fractions , one can define a new @xmath62 which represents an upper limit to the true @xmath63 , @xmath64 where the inequality follows from the fact that @xmath65 for all @xmath54 .
we find that the unsigned flux corresponding to @xmath62 is only 25% larger than the flux of @xmath57 , indicating that the assumption of vertical fields have a very limited effect on the pdf estimated in the paper . due to the similarity between the magnetic field strengths of the second and the third magnetic components , one may wonder whether they are really needed to reproduce the line shapes .
moreover , if they are not needed , the existence of mixed polarities may be an artifact since the signals of the two components cancel our producing no observable residual . as we stress in [ minv ]
, the three components are needed . here
we elaborate on the reasons . reproducing the stokes @xmath3 asymmetries of the visible lines
requires two magnetic components , even for the mildest asymmetries characteristic of the network .
one single magnetic component produces stokes @xmath3 area and the peak asymmetries of the same magnitude , in contradiction with the observations @xcite .
somehow unexpectedly , @xcite find that these two components are enough to explain not only the mildest asymmetries , but the whole range of observations including extreme profiles with an even number of lobes .
these two components have similar kg magnetic field strengths , but very different velocities ( * ? ? ?
* figs . 12 and 14 ) .
the large difference of velocities is demanded by the observed stokes @xmath3 asymmetries , and it is the reason why the inversion code is able to separate them . here
we analyze profiles similar to those reproduced by @xcite and , consequently , a minimum of two magnetic components is needed .
in addition , we also want to reproduce the ir lines which , according to the literature , indicate the existence of weak fields in the quiet sun .
a third weak magnetic field component was needed too .
this argument does not imply that all the observed profiles necessarily require three components .
it implies that the typical cases need three components ( regular and , with fe i @xmath415648 and fe i @xmath415652 showing weak fields ) .
it also implies that that typical stokes @xmath3 profiles have enough information to constrain the three components .
finally , it implies that if a single misma scenario is going to be used to reproduce all the asymmetries , it must contain three components .
one of the main conclusions of this work is the co - existence of kg and sub - kg fields in resolution elements of 15 as inferred from the simultaneous inversion of ir and visible stokes profiles .
the use of visible lines to measure kg magnetic field strength has been criticized .
the analysis of extremely noisy stokes @xmath3 profiles of and may induce a bias towards kg ( see * ? ? ?
* ) , and the stokes @xmath3 profiles from the visible lines are the major tracers of the kg magnetic fields in our inversion .
however , the kg that we infer are not produced by such a bias for a number of reasons .
the classes have much less noise than the individual profiles , and yet the inversion yields as much kg of the individual profiles .
the classes that yield at least one kg magnetic component represent 75% of the total pixels .
they include the two main classes with a s / n larger than 100 .
another 10 classes have s / n larger than 30 , which ensures inversions without the bias pointed out by @xcite .
another way to test the consistency of our results is to invert only the ir lines to see whether it is possible to infer kg fields in the quiet sun without visible lines .
this inversion was performed in the same way as in [ minv ] but only for fe i @xmath415648 and fe i @xmath415652 with misma models including only 2 magnetic components ( see more details in * ? ? ?
the result of such analysis was a major magnetic component similar to the first component in figure [ hmag ] and a second with weak and strong fields .
a field stronger than 1 kg was found in 50% of the inverted pixels ( those with the strongest ir signals ) .
the strong fields were retrieved from the extended wing of the stokes @xmath3 profiles .
figure [ pdf3 ] gives the pdf obtained from these inversions .
the strong field contribution is not as large as in figure [ pdf ] , but it still carries most of the magnetic energy and flux .
the presence of fewer strong fields as compared to the full inversions including visible lines is due to the incompleteness of the information given by the ir lines , since most of the contributions from strong fields remains hidden by noise .
the signed flux has a behavior similar to that of the full inversion described in [ sflux ] .
it is more important for the fields stronger than 1 kg while the weak fields are balanced .
figure [ pdf3 ] also shows the pdf obtained from the milne - eddington inversion of the ir lines ( the dashed line ; see * ? ? ?
this inversion was performed with a single magnetic component and shows only the weak field component .
a new observational technique was used to study the quiet sun magnetism .
simultaneous visible ( and ) and infrared ( fe i @xmath415648 and fe i @xmath415652 ) polarimetric observations were analyzed under the assumption of micro - structured magnetized atmosphere ( misma ) .
we generate model mismas able to reproduce all kinds of asymmetries presented in the observed stokes @xmath3 profiles , even when there are clear differences in the asymmetries of the visible and the ir stokes profiles .
the model atmospheres have three magnetic components .
the simultaneous inversion of the visible and ir pairs reveals the co - existence of kg and sub - kg fields in the low photosphere .
we have observed this combination of weak and strong fields in 74% of the pixels under study , which means 30% of the pixels of the fov ( see [ mags ] ) .
this was already observed by @xcite from the separate inversion of the ir and visible lines , however , this work presents the first evidence of the co - existence of kg and sub - kg fields from a simultaneous inversion of lines in the two spectral ranges . from the three magnetic components , the one having most of the mass includes most of the weak fields .
the two other components tend to show strong kg fields .
the preference of the first component to have the weakest fields is due to the coupling between field strength and plasma density .
the weaker the field the larger the density so that the component of weakest field tends to have the largest mass .
the histogram of field strengths of these minor components ( fig .
[ hmag ] ) looks similar to that obtained by @xcite , while the first component has a distribution similar to that obtained with the analysis of ir lines by @xcite .
this supports the hypothesis of @xcite and @xcite : the study of the quiet sun magnetism with only visible lines or only ir lines biases the resulting field strengths ; the in fields must be studied combining stokes profiles from both spectral ranges .
the inferred magnetic fields occupy 1.5% of the total fov . other observations with better spatial resolution ( 1@xmath0 ;
* ; * ? ? ?
* ) yield a total filling factor of 1% .
we obtain a larger value thanks to the combined use of the information from visible and ir lines , while all previous works used lines from a single spectral range .
the filling factor inferred for the same observations used in this paper by @xcite was only 0.85% .
the difference is due to the different model atmospheres used to reproduce the observations .
while here we use a realistic multicomponent model , @xcite carried out milne - eddington inversions with a single magnetic component .
thus , the filling factors in pixels with mixed polarities ( see [ mixp ] ) are underestimated .
we have measured an average flux density over the fov of 9.6 g ( those pixels without polarization signals contribute to this average with zero field strength ; [ mags ] ) .
this flux is similar to that obtained by analysis of visible lines with 1@xmath0 spatial resolution ( see * ? ? ?
* ; * ? ? ?
* ) while it is two times larger than the magnetic flux derived from ir lines @xcite .
albeit our observations have lower resolution ( 15 ) , the use of visible and ir lines reveals more magnetic structures than the use of lines from a single spectral range .
the flux density is also larger than the magnetic flux obtained in [ magn ] using the magnetograph equation applied to a single line .
a combination of factors explain such bias .
first , we obtained magnetic field strengths larger than the saturation limit of the lines , thus the circular polarization signals are no longer proportional to the magnetic flux and equation ( [ virm1 ] ) is a lower limit of the real flux .
second , the presence of mixed polarities ( see [ mixp ] ) produces cancellation of polarization signals and the magnetic flux is underestimated .
third , the lines weaken in strong magnetic concentrations , leading to a well known diminishing of polarization signals ( e.g. , * ? ? ?
finally , the visible and ir signals often trace different concentrations coexisting in the resolution element .
then , combined inversions allow us to detect more flux than the inversion of one of the spectral ranges .
the inversion indicates the presence of mixed polarities in at least 20% of the fov ( [ mixp ] ) .
this is the largest fraction of mixed polarities obtained from direct measurements .
the fraction of pixels with mixed magnetic polarities grows as the polarization signals weaken ( fig .
[ mix ] ) .
more than half of the profiles with stokes @xmath3 signal lower than 0.2% of the continuum intensity have mixed polarities .
we have estimated the probability of finding a given magnetic field strength in the quiet sun , i.e. , the probability density function of the magnetic field strength ( pdf ;
[ spdf ] ) .
it is largest for the weakest fields , however , an extended kg tail goes up to a limit around 1800 g. despite the fact that the kg fields have only 30% of the total filling factor , they carry 75% of the total flux and 90% of the total magnetic energy .
there is a clear imbalance of the magnetic flux towards one polarity .
we measure a signed flux density of @xmath43 g which is mostly due to the strong kg fields ( [ sflux ] ) .
the signed pdf ( fig . [ pdf2 ] ) shows how the weak fields are well balanced whereas the strong fields have a clear preference for the negative polarity .
this imbalanced flux has the same sign as the surrounding network , as it was already observed by @xcite .
this imbalance is qualitatively similar to those imposed in some mhd simulations of granular convection ( see * ? ? ?
* ; * ? ? ?
these simulations have an initial unipolar field .
the convection reprocesses this field to yield a continuous pdf with a clear imbalance in the strongest fields .
the misma inversion of the ir profiles alone also gives rise to a pdf with extended kg tail ( [ cons2 ] ) .
such tail does not appear in milne - eddington inversions of the same profiles , which tend to select the component of weakest field in the resolution element .
the magnetic fields studied in this paper fill only 1.5% of the total fov . the unsigned flux density that we measure is only a lower limit of the real flux .
larger fluxes are obtained in observations with better spatial resolution @xcite and , furthermore , measurements based on the hanle effect are compatible with a flux density of 60 g or more @xcite . a large fraction of the magnetic structures remains undetected in zeeman based observations .
based on our biased observations , we have produced a pdf with the distribution of magnetic field strengths . obviously , this is not the real solar pdf .
a natural step forward would be working out an unbiased pdf with the help of different measurements based on the zeeman effect , the hanle effect , and also mhd simulations .
this work of synthesis is carried out in a separate paper @xcite . despite these and other efforts
, we are still far from having a complete picture of the quiet sun magnetism .
there is a reduced number of observations and each one is gathered with different instruments under different conditions .
thus , we need a survey of polarimetric data across many in regions of the sun , if we want to study the quiet sun with a systematic and global point of view
. such data would allow us to start exploring the connection of the in with active regions , the chromosphere and corona and , eventually , to follow the evolution of the in with the solar cycle .
observations with better spatial and temporal resolution are mandatory .
they are coming up thanks to the advent of new generation ground based and space - borne solar telescope @xcite .
these improvements , together with the development of more realistic numerical simulations , guarantee a rapid evolution of the field .
thanks are due to the support astronomers and operators of themis and vtt for help during the observation .
the vtt is operated by the kiepenheuer - institut fr sonnenphysik , freiburg , and the french - italian telescope themis is operated by cnrs - cnr , both at the spanish observatorio del teide of the instituto de astrofsica de canarias .
the work has been partly funded by the spanish ministry of science and technology , project aya2004 - 05792 and by the deutsche forschungsgemeinschaft through grant 418 spa-112/14/01 . ,
v. , collados , m. , snchez almeida , j. , gonzlez , v. , cruz - lopez , a. , manescau , a. , joven , e. , paez , e. , diaz , j. j. , feeney , o. , snchez , v. , scharmer , g. , & soltau , d. 1999 , in asp conf .
183 , high angular resolution solar physics : theory , observations , and techniques , ed .
t. rimmele , k. s. balasubramaniam , & r. radick ( san francisco : asp ) , 264
|
we study the quiet sun magnetic fields using spectropolarimetric observations of the infrared and visible fe i lines at 6301.5 , 6302.5 , 15648 and 15653 .
magnetic field strengths and filling factors are inferred by the simultaneous fit of the observed stokes profiles under the misma hypothesis .
the observations cover an intra - network region at the solar disk center .
we analyze 2280 stokes profiles whose polarization signals are above noise in the two spectral ranges , which correspond to 40% of the field of view . most of these profiles can be reproduced only with a model atmosphere including 3 magnetic components with very different field strengths , which indicates the co - existence of kg and sub - kg fields in our 15 resolution elements .
we measure an unsigned magnetic flux density of 9.6 g considering the full field of view .
half of the pixels present magnetic fields with mixed polarities in the resolution element .
the fraction of mixed polarities increases as the polarization weakens .
we compute the probability density function of finding each magnetic field strength .
it has a significant contribution of kg field strengths , which concentrates most of the observed magnetic flux and energy .
this kg contribution has a preferred magnetic polarity , while the polarity of the weak fields is balanced .
|
the standard model ( sm ) , by meeting all confrontations with experiments , stands as a remarkably simple parameterisation of known physics .
yet it has many unsatisfactory aspects which leads to a belief that there must exist a simpler underlying structure of which the sm is the low energy piece .
this structure is believed to make its appearance at much higher energies , where we can approach this with renormalisation group techniques to extrapolate the sm parameters to the unexplored scales @xcite .
recalling that in the sm , the runnings of the gauge , yukawa and quartic scalar couplings are logarithmic with the energy scale , the gauge couplings do not all meet at a point , but do tend to unify near @xmath0 gev .
extensions to the sm such as extra - dimensional scenarios accessible to sm fields have the virtue , thanks to the couplings now having a power law running , of bringing the unification scale down to an explorable range @xcite . note that many other extensions to the sm exist which alter the runnings in different ways , such as supersymmetry ( susy ) , where a range of new particles ensure the gauge couplings do meet at a point , but runnings remain logarithmic @xcite .
the story of extra - dimensional physics can be thought to begin in the 1920s with kaluza ( 1919 ) and klein ( 1926 ) @xcite who had the idea to add a fifth dimension to unify the only two forces known at that time .
later in the 1970s and 1980s the birth of supergravity and superstring theories renewed the interest in extra - dimensional models .
however these dimensions are expected to be very small ( @xmath1 ) and will not be probed by experiments any time soon . however , beginning in the 1990s new extra - dimensional scenarios which could be larger than the planck length appeared .
antoniadis @xcite proposed @xmath2 scale extra - dimensions to explain susy breaking , and in order to solve the hierarchy problem the large extra dimensions approach was introduced by arkani - hamed , dimopoulos and dvali ( add ) @xcite . in these models , the metric is flat and the strength of the gravitational interaction is diluted , which leads to interesting consequences for low - energy phenomenology .
another approach was introduced by randall and sundrum @xcite with only one curved extra - dimension ( warped extra - dimensions ) .
the new dimension is compactified on a finite interval @xmath3 , with the endpoints of the interval being 3-branes .
the metric of this space is not flat , where the gravity fields propagating in the fifth dimension suffer exponential suppression and live on a different brane to the sm particles .
extra - dimensional models lead to many phenomenological implications which can be tested at colliders and also can be used as a tool to answer many issues in the sm , such as the hierarchy problem @xcite ; tev scale extra - dimensional scenarios giving rise to new susy breaking mechanisms @xcite ; the generation of neutrino mass and new sources of cp violation @xcite ; the unification without susy with suppression of proton decay @xcite ; triggering electroweak symmetry breaking without a higgs boson @xcite ; providing cosmologically viable dark matter candidates @xcite and many other applications related to black holes and gravity @xcite . with the large hadron collider ( lhc ) now up and running , exploration of the realm of new physics that may operate at the tev scale has begun @xcite . among these models with extra spatial dimensions the universal extra - dimension ( ued ) model makes for an interesting tev scale physics scenario @xcite ; as it features a tower of kaluza - klein ( kk ) states for each of the sm fields , all of which have full access to the extended space - time manifold @xcite .
this particular scenario has recently been extensively studied in the literature @xcite and has been a fruitful playground for addressing a variety of puzzles in the sm .
we therefore collect in a comprehensive manner and in one place the necessary tools for making renormalisation group analyses of the sm and minimal supersymmetric sm ( mssm ) ued extensions .
note that we review only these particular ued scenarios here , as alternative warped or higher - dimensional ued models are largely unexplored in the literature , or require different toolsets , as shall be briefly discussed later .
the observable parameters of the sm are : 6 quark masses , 3 lepton masses , 4 parameters of the cabibbo - kobayashi - maskawa ( ckm ) matrix @xcite and 3 gauge couplings .
the renormalisation group equations ( rges ) for the ckm matrix being obtained from the rges for the yukawa couplings .
this can also be extended to include neutrino masses and mixings possible in the leptonic sector . in this review
we first introduce the various models and their varieties we shall consider ( section [ sec:2 ] and section [ sec:3 ] for the supersymmetric extensions to this ) , next reviewing the rges for the gauge couplings constants ( section [ sec:4 ] ) and yukawa couplings for the sm and ued scenarios and 5-dimensional minimal supersymmetric sms ( 5d mssm ) ( section [ sec:5 ] ) .
this shall be followed by a review of the ckm parameters evolution ( sections [ sec:7 ] ) .
extensions to massive neutrino scenarios and their mixings evolution will follow in section [ sec:9 ] . with a summary and the prospects for future research directions in section [ sec:10 ] .
the ued model places particles of the sm in the bulk of one or more compactified extra dimensions @xcite . in our case we have a single flat extra dimension of size @xmath4 , compactified on an @xmath5 orbifold .
as such we will have an infinite tower of kk modes with the zero modes corresponding to the sm states .
these kk modes are in the tev scale and modify the running of the rges at relatively low energy scales @xcite .
the ued model , like any higher dimensional theory , is an effective field theory which is valid up to some scale @xmath6 , at which a new physics theory emerges . as a result , once the kk states are excited , these couplings exhibit power law dependencies on @xmath6 .
this can be illustrated if @xmath7 , to a very good accuracy , the generic sm beta function is shown to have the power law evolution behaviour @xcite : @xmath8 where @xmath9 is a generic contribution from a single kk level , and where its coefficient is not a constant but instead @xmath10 , with @xmath11 , reflecting the power law running behaviour . as a result of faster running ,
the gauge couplings tend to lower the unification scale down to a relatively low order , which might be accessible to collider experiments . the first version of this model we shall consider , the bulk ued model has the 5-dimensional kk expansions of the weak doublet ( @xmath12 ) and singlet ( @xmath13 ) as well as the higgs and gauge fields ( @xmath14 ) as shown ( the corresponding coupling constants among the kk modes are simply equal to the sm couplings up to normalisation factors , e.g. @xmath15 ) below : @xmath16\right\ } \ ; , \nonumber \\
f(x , y ) & = & \frac{1}{{\sqrt { \pi r } } } \left\ { { f_l}(x ) + \sqrt 2 \sum\limits_{n = 1}^\infty \left[f_l^n ( x)\cos \left(\frac{{ny}}{r}\right ) + f_r^n(x ) \sin \left(\frac{{ny}}{r}\right)\right]\right\ } \ ; \nonumber .
\label{ued7}\end{aligned}\ ] ] the zero modes in the above equations are identified with the 4-dimensional sm fields , whilst the complex scalar field @xmath17 and the gauge field @xmath18 are @xmath19 even fields , and there is a left - handed and a right - handed kk mode for each sm chiral fermion .
note that in models with ued momentum conservation in the extra dimensions , we are led to the conservation of kk number at each vertex in the interactions of the 4-dimensional effective theory ( or strictly speaking , the kk parity @xmath20 is what remains conserved , where @xmath21 is the kk number ) . in the bulk we have the fermion and gauge field interactions as follows : @xmath22 where @xmath23 , and @xmath24 , for further details see @xcite . after integrating out the compactified dimension , the 4-dimensional effective lagrangian has interactions involving the zero mode and the kk modes .
however , these kk modes can not affect electroweak processes at tree level , and only contribute to higher order electroweak processes .
the one - loop feynman diagram contributions to the yukawa couplings in the sm and ued model have been explicitly illustrated in @xcite . in the ued model ,
where for each energy level @xmath25 , we effectively have a heavier duplicate copy of the entire sm particle content
. however , new contributions from the @xmath26 , @xmath27 interactions ( that of the fifth component of the vector fields ) also contribute .
in contrast , the fifth component of the gauge bosons @xmath28 is a real scalar and does not have any zero mode , transforming in the adjoint representation of the gauge group .
a simple alternative to this model is that of the brane localised ued model , where we have the same fields but where the fermion matter fields can not propagate in the bulk and they are restricted to the brane . for the case of brane
localised matter fields , only the boson fields ( the gauge fields and the scalar fields ) can propagate in the bulk space . however , if the compactification radius @xmath4 is sufficiently large , due to the power law running of the gauge couplings , it will enable us to bring the unification scale down to an exportable range at the lhc scale .
another useful model we shall consider is the 5d mssm defined in @xcite .
the 5d mssm is a five dimensional @xmath29 supersymmetric model compactified on the @xmath30 orbifold which breaks the 5d lorentz invariance to the usual 4d one .
this breaking gives a momentum conservation along the fifth dimension which conserves the kk number at tree level and kk parity at loop level @xcite .
one of the main implications of kk - parity invariance is that the lightest kk mode is stable and can be a cold dark matter candidate . in this compactification
we can recover the mssm as zero modes since we obtain chiral fermions .
the gauge sector is then described by a 5d @xmath29 vector supermultiplet which consists ( on - shell ) of a 5d vector field @xmath31 , a real scalar @xmath32 and two gauginos , @xmath33 and @xmath34 .
the action for which can be given by : @xmath35\right ] \ ; , \label{eqn:0}\end{aligned}\ ] ] with @xmath36 and @xmath37 . @xmath38 $ ] and @xmath39 normalises the trace over the generators of the gauge groups . from the decomposition of the 5d supercharge ( which is a dirac spinor ) into two majorana - type supercharges , which constitute a @xmath40 superalgebra in 4d
, one can rearrange these fields in terms of a @xmath40 , 4d vector supermultiplet , @xmath41 : @xmath42 : @xmath29 vector supermultiplet containing @xmath43 and @xmath33 , @xmath44 : @xmath29 chiral supermultiplet containing @xmath34 and @xmath45 . both @xmath42 and @xmath44 ( and their component fields )
are in the adjoint representation of the gauge group @xmath46 . using the supermultiplets one can write the original 5d @xmath29 supersymmetric action eq.([eqn:0 ] ) in terms of @xmath29 4d superfields and the covariant derivative in the @xmath47 direction@xcite : @xmath48 \ ; , \label{eqn:0a}\ ] ] with @xmath49 .
@xmath50 is the covariant derivative in the 4d @xmath29 superspace ( see refs@xcite . ) and @xmath51 . to find the feynman rules to a given order in the gauge coupling @xmath52
, one can expand and quantise the action@xcite .
the beta functions for the couplings of the operators in the superpotential are governed by the wave function renormalisation constants @xmath53 due to the non - renormalisation theorem @xcite .
the higgs superfields and gauge superfields will always propagate into the fifth dimension .
different possibilities for the matter superfields will be discussed , where superfields containing sm fermions can propagate in the bulk or are restricted to the brane . for the case where all fields can propagate in the bulk , the action for the matter fields would be@xcite : @xmath54 again , this action can be expanded and quantised .
the @xmath44-field should be odd under @xmath19 symmetry because it appears together with a derivative @xmath55 , whereas v is even .
for the two matter superfields , we choose @xmath56 to be even and the conjugate @xmath57 to be odd such that @xmath57 vanishes on the brane .
only the even fields have zero modes .
we can write the action for the second case where all superfields containing sm fermions are restricted to the brane . in which case the part of the action involving only gauge and higgs fields is not modified , whereas the action for the superfields containing the sm fermions becomes : @xmath58 due to the 5d @xmath29 susy , yukawa couplings are forbidden in the bulk .
however , they can be introduced on the branes , which are 4d subspaces with reduced susy .
one can also add the effective neutrino mass operator ( also called lepton number violating weinberg operator ) , with dimensional coupling @xmath59 in which we are interested to show its evolution and therefore the majorana mass term for neutrinos .
we will write the following interaction terms , called brane interactions , containing yukawa - type couplings : @xmath60 where @xmath61 and @xmath62 are the lepton and up - type higgs doublet chiral superfields respectively .
this operator is used to study neutrino masses and mixings , where rges for this effective operator have been derived in the context of the four - dimensional sm @xcite and mssm@xcite and shall be discussed further in section [ sec:9 ] .
the evolution of the gauge couplings in four dimension at one loop are given by : @xmath63 where @xmath64 and @xmath65@xcite , using a @xmath66 normalisation . if we consider our 5d theory eq.([beta4d])can be written in terms of the scale parameter @xmath67 : @xmath68{g_i}^3 \ ; , \label{gaugecoupling3}\end{aligned}\ ] ] where @xmath69 take the following form in the case of the model ued sm@xcite : @xmath70 with @xmath71 being the number of generations of matter fields in the bulk .
next we consider the beta functions of the gauge couplings in the 5d mssm .
in fact , after compactification of the 5d mssm , where the master beta functions of the gauge couplings in the 5d mssm as follows @xcite : @xmath72 where @xmath71 again represents the number of generations of fermions which propagate in the bulk . in figs.[gauge_uedsm ] and [ gauge_5d ]
we have plotted the running of the gauge couplings for the ued sm case and the 5d mssm respectively for the brane localised and bulk field cases , and for several choices of compactification scales for the extra - dimension ( @xmath4 ) . from these plots , and the discussion given in refs .
@xcite , we find that for the three gauge coupling constants to approach a small region at some value of @xmath67 requires an extremely large value of @xmath73 , which is of no phenomenological interest at present .
for the case of our fields being brane localised in the ued model , we see a similar behaviour : the extra - dimensions naturally lead to gauge coupling unification at an intermediate mass scale for the compactification radii considered here .
( red ) , @xmath74 ( blue ) , @xmath75 ( green ) with : in the left panel , all matter fields in the bulk ( ued bulk ) ; and the right panel for all matter fields on the brane ( ued brane ) ; for three different values of the compactification scales ( 2 tev ( solid line ) , 8 tev ( dot - dashed line ) , 15 tev ( dashed line ) ) as a function of the scale parameter @xmath67 in the ued sm.,title="fig:",scaledwidth=45.0% ] ( red ) , @xmath74 ( blue ) , @xmath75 ( green ) with : in the left panel , all matter fields in the bulk ( ued bulk ) ; and the right panel for all matter fields on the brane ( ued brane ) ; for three different values of the compactification scales ( 2 tev ( solid line ) , 8 tev ( dot - dashed line ) , 15 tev ( dashed line ) ) as a function of the scale parameter @xmath67 in the ued sm.,title="fig:",scaledwidth=45.0% ] ( red ) , @xmath74 ( blue ) , @xmath75 ( green ) ) with : in the left panel , all matter fields in the bulk ; and the right panel for all matter fields on the brane ; for three different values of the compactification scales ( 2 tev ( solid line ) , 8 tev ( dot - dashed line ) , 15 tev ( dashed line ) ) as a function of the scale parameter @xmath67 in the 5d mssm.,title="fig:",scaledwidth=45.0% ] ( red ) , @xmath74 ( blue ) , @xmath75 ( green ) ) with : in the left panel , all matter fields in the bulk ; and the right panel for all matter fields on the brane ; for three different values of the compactification scales ( 2 tev ( solid line ) , 8 tev ( dot - dashed line ) , 15 tev ( dashed line ) ) as a function of the scale parameter @xmath67 in the 5d mssm.,title="fig:",scaledwidth=45.0% ]
in the quark sector of the sm , we have ten experimentally measurable parameters , i.e. six quark masses , three mixing angles , and one phase ( these angles and phase being encoded in the ckm matrix which we shall discuss in section [ sec:7 ] ) . at present
there has been considerable effort to understand the hierarchies of these mixing angles and fermion masses in terms of the rges@xcite .
note though that when using the rges as a probe , the initial values we shall adopt are very important , where we shall scale for the gauge couplings and the fermion masses at the @xmath76 scale are shown in table [ ta1 ] .
furthermore , we shall also attempt , in section [ sec:9 ] , to develop the rges of the lepton sector ( including possible mixing angles and phases ) , which will require knowledge of the evolution of a parameter @xmath39 , where the lowest order operator which generates majorana neutrino masses after electroweak symmetry breaking ( ewsb ) , is the lepton - number violating weinberg operator @xcite .
this lowest order operator ( appearing with dimension @xmath77 in four space - time dimensions ) can be written as : @xmath78 where @xmath61 and @xmath79 are the lepton and the higgs doublet fields .
@xmath80 is the typical heavy energy scale for the range of validity of the low - energy effective theory , where renormalisation group equations for this effective operator have been derived in the context of the four - dimensional sm and mssm@xcite . in the present case
we consider the effective neutrino mass operator with dimensional coupling @xmath59 ; after spontaneous symmetry breaking , the majorana neutrino masses can be written as @xmath81 ( @xmath82 being the vev of the higgs field and @xmath83 , the ratio of the vevs of our two higgs doublets ) and @xmath84 for bulk propagating , and @xmath85 for brane localised matter superfield scenarios respectively .
as such , we have set @xmath86 as the renormalisation point , and use @xmath87 and @xmath88 .
the general form of evolution equations for yukawa couplings and neutrino @xmath39 coupling at the one loop can be written in the following form refs@xcite : @xmath89k+k[{y_e}^\dag y_e ] \right)c_3 \ ; . \nonumber\end{aligned}\ ] ] where @xmath90 $ ] .
the sm is a limiting case for the ued , where the kk states decouple . the coefficients in the evolution equation are defined by : @xmath91 , @xmath92 , @xmath93 , @xmath94 , @xmath95 , @xmath96 , @xmath97 .
the ued contribution is obtained when kk states enter , where due to the orbifolding the zero mode for fermions are chiral , which are replaced by dirac fermions at each kk level .
this lead to the factor 2 appearing in @xmath98 and @xmath99 since the kk left and right - handed chiral states contribute to the closed fermion one loop diagrams .
that is , @xmath100 , @xmath101 , @xmath102 , @xmath103 , @xmath104 , @xmath105 , @xmath106 .
for the case where the fermions are restricted to the brane , we obtain the coefficients from ref@xcite . and ref @xcite .
@xmath107 , @xmath108 , @xmath109 , @xmath110 , @xmath111 , @xmath112 , @xmath113 .
note that the coefficient @xmath114 since we do not have a trace of fermionic loops as the fermions are restricted to the brane .
similarly , the general form of the evolution equations for the various mssms , where we shall use a notation similar to the ones of refs@xcite .
@xmath115k+k[{y_e}^\dag y_e ] \right ) c \ ; . \nonumber\end{aligned}\ ] ] where @xmath116 , @xmath117 , @xmath118 .
where for the mssm , as a limiting case of the 5d models we shall consider in the following , and also when @xmath119 the coefficients in the evolution equations are : @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 .
the coefficients in the 5d mssm , for all three generations propagating in the bulk , can be expressed as : @xmath126 , @xmath127 , @xmath128 , @xmath129 , @xmath130 , @xmath131 .
however , when all matter superfields are constrained to live on the 4d brane , the coefficients of the evolution equations are given by : @xmath132 , @xmath133 , @xmath134 , @xmath135 , @xmath136 , @xmath137 .
it is well known that in the sm , the quark sector s flavor mixing is parameterised by the ckm matrix , which makes it possible to explain all flavor changing weak decay processes and cp - violating phenomena to date . in particular , for the standard parameterisation of the ckm matrix , which has the form : @xmath138 where @xmath139 , @xmath140 etc .
are the sines and cosines of the three mixing angles @xmath141 , @xmath142 and @xmath143 , and @xmath144 is the cp violating phase .
the ckm matrix arises from a consideration of the square of the quark yukawa coupling matrices being diagonalised by using two unitary matrices @xmath145 and @xmath42 , @xmath146 in which @xmath147 , @xmath148 , @xmath149 and @xmath150 , @xmath151 , @xmath152 are the eigenvalues of @xmath153 and @xmath154 respectively , with @xmath155 . from the full
set of one - loop coupled rge for the yukawa couplings and the ckm matrix , together with those for the gauge coupling equations , one can obtain the renormalisation group flow of all observables related to up- and down - quark masses and the ckm matrix elements .
we write down the general form for the evolution of @xmath156 , @xmath157 and the variation of each element of the ckm matrix @xmath158 @xcite in the sm , the ued sm , the mssm and the 5d mssm .
@xmath159\ ; , \nonumber\\ 16{\pi ^2}\frac{{dh_j ^2}}{{dt } } & = & h_j ^2[2({t_d}a - { g_d } ) + 3{b}h_j ^2 - 2{b}\sum\limits_i { f_i^2 } { \left| { { v_{ij } } } \right|^2 } ] \ ; , \label{eqn:343 } \\ 16{\pi ^2}\frac{{dy_e^2}}{{dt } } & = & y_e^2[2({t_e}a - { g_e } ) + 3{b}y_e^2 ] \ ; , \nonumber \\ 16{\pi ^2}\frac{{d{v_{ik}}}}{{dt } } & = & -\frac{3}{2}{b}\left[\sum\limits_{m , j \ne i } { \frac{{f_i^2 + f_j^2}}{{f_i^2 - f_j^2 } } } h_m^2{v_{im } } v_{jm}^*{v_{jk } } + \sum\limits_{j , m \ne k } { \frac{{h_k^2 + h_m^2}}{{h_k^2 - h_m^2 } } } f_j^2v_{jm}^*{v_{jk}}{v_{im}}\right ] \ ; , \nonumber\end{aligned}\ ] ] where @xmath160 in the sm , @xmath161 , @xmath162 in the ued bulk sm and @xmath163 , @xmath164 in the ued brane sm .
the gauge couplings @xmath14 for the sm , the ued bulk sm and the ued brane sm are written in sec.([smcoef ] ) , sec.([uedbulkcoef ] ) and sec .
( [ uedbranecoef ] ) respectively .
@xmath165\ ; \nonumber\\ 16{\pi ^2}\frac{{dh_j ^2}}{{dt } } & = & h_j ^2[2({t_d}\tilde{c } - { g_d } ) + 6{c}h_j ^2 + 2{c}\sum\limits_i { f_i^2 } { \left| { { v_{ij } } } \right|^2 } ] \ ; , \label{eqn:344 } \\ 16{\pi ^2}\frac{{dy_e^2}}{{dt } } & = & y_e^2[2({t_e}\tilde{c } - { g_e } ) + 6{c}y_e^2 ] \ ; , \nonumber \\ 16{\pi ^2}\frac{{d{v_{ik}}}}{{dt } } & = & { c}\left[\sum\limits_{m , j \ne i } { \frac{{f_i^2 + f_j^2}}{{f_i^2 - f_j^2 } } } h_m^2{v_{im}}v_{jm}^*{v_{jk } } + \sum\limits_{j , m \ne k } { \frac{{h_k^2 + h_m^2}}{{h_k^2 - h_m^2 } } } f_j^2v_{jm}^*{v_{jk}}{v_{im}}\right ] \ ; , \nonumber\end{aligned}\ ] ] where we use the same forms as in secs.([mssmcoef ] , [ 5dbulkcoef ] , [ 5dbranecoef ] ) to fix the coefficients @xmath166 , @xmath167 and gauge couplings @xmath14 to describe each model . in fig.[fig : ft_uedsm ] the initial yukawa couplings are given by the ratios of the fermion masses to the higgs vacuum expectation value .
the yukawa couplings evolve in the usual logarithmic fashion when the energy is below 2 tev , 8 tev , and 15 tev for the three different cases .
however , once the first kk threshold is reached , the contributions from the kk states become more and more significant .
the evolution of @xmath168 ( see eq.([eqn:343 ] ) ) depends explicitly on the cutoff @xmath6 , which have finite one - loop corrections to the beta functions at each massive kk excitation level .
therefore , the running of the yukawa couplings , or more precisely , the one - loop kk corrected effective four dimensional yukawa couplings , begins to deviate from their normal orbits and start to evolve faster and faster .
note also observe that the yukawa couplings are quickly evolving to zero , however , a satisfactory unification of these seems to still be lacking . as such , we have so far observed the yukawa couplings all decrease with increasing energy , which agrees with what is observed in the sm , however , the yukawa couplings are driven dramatically towards extremely weak values at a much faster rate .
this is an interesting feature that distinguishes the ued model from that of the sm .
for the top quark in the ued sm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the sm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] for the top quark in the ued sm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the sm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] the 4d mssm contains the particle spectrum of a two - higgs doublet model extension of the sm and the corresponding supersymmetric partners .
the two higgs doublets @xmath170 and @xmath171 , with opposite hypercharges , are responsible for the generation of the up - type and down - type quarks respectively .
the vacuum expectation values of the neutral components of the two higgs fields satisfy the relation @xmath172 . as a result
, the initial yukawa couplings are given by the ratios of the fermion masses to the appropriate higgs vacuum expectation values as follows : @xmath173 where we define @xmath174 , which is the ratio of vacuum expectation values of the two higgs fields @xmath170 and @xmath171 .
furthermore , below the supersymmetric breaking scale the yukawa and gauge couplings run in the usual logarithmic fashion , giving a rather slow change for their values .
therefore , for supersymmetric breaking theories around tev scales , for simplicity , we take the supersymmetric breaking scale @xmath175 in the present numerical study , and run the rges from @xmath76 up to the high energy scales for our three different compactification scales . once again , once the first kk threshold is crossed , the power law running of the various beta functions causes the yukawa coupling to rapidly increase following the rapid increase in the gauge coupling constants .
this behaviour can be observed for both small and large @xmath176 cases .
however , as illustrated in the first graph of fig .
[ fig : ft_bulk ] , for small @xmath176 , the yukawa coupling has a large initial value , therefore it blows up at a relatively low energy as compared with the case for large @xmath176 . as a result , as one evolves upward in the scale , the top yukawa coupling is rising with a fast rate and is pushed up against the landau pole where it becomes divergent and blows up . in the vicinity of this singular point
the perturbative calculation becomes invalid , and the higher order corrections become significant . for the top quark in the bulk case of 5d mssm as a function of the scale parameter @xmath67 , for ( left panel ) @xmath177 and ( right panel ) @xmath178 where the solid line is the mssm evolution and for different compactification scales : @xmath169
= 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] for the top quark in the bulk case of 5d mssm as a function of the scale parameter @xmath67 , for ( left panel ) @xmath177 and ( right panel ) @xmath178 where the solid line is the mssm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] in the brane localised matter field scenario , the beta function has only linear terms in @xmath179 , which is comparable with the @xmath179 term in the beta function for the gauge couplings . as depicted in fig .
[ fig : ft_brane ] , for a small value of @xmath176 , we have a large initial value of @xmath168 and the gauge coupling contribution to the yukawa beta function is sub - dominant only .
therefore the yukawa coupling @xmath168 increases rapidly as one crosses the kk threshold , resulting in a rapid approach of the singularity before the unification scale is reached . however ,
for an intermediate value of @xmath176 , we have a relative smaller initial condition for the top yukawa coupling and the yukawa terms in the beta function become less important .
the contributions from the gauge couplings may then become significant , which leads to a net negative contribution to the beta functions .
therefore , the curvature of the trajectory of the top yukawa evolution might change direction , becoming more obvious for a large value of @xmath176 .
this behaviour provides a very clear phenomenological signature , especially for scenarios with a larger @xmath176 and that are valid up to the unification scale . for the top quark in the brane case of 5d mssm as a function of the scale parameter @xmath67 , for ( left panel ) @xmath177 and ( right panel ) @xmath178 where the solid line is the mssm evolution and
for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,scaledwidth=45.0% ] in fig.[fig : vub_uedsm ] we plot the evolution of @xmath180 for the ued bulk and brane cases . for the evolution of @xmath181 and
@xmath182 we can observe similar behaviours , i.e. , they all increase with the energy scale ; as can be seen from eq.([eqn:343 ] ) , the evolution of the ckm matrix is governed by the yukawa couplings and the factor @xmath179 .
they evolve faster in the region where the power law scaling of the yukawa couplings becomes substantial .
therefore , the renormalisation effect is explicit for mixings involving the third family . in the ued sm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the sm , for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] in the ued sm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the sm , for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] in fig.[fig : vub_bulk_brane_30 ] we plot the energy dependence of @xmath180 from the weak scale all the way up to the high energy scales for different values of compactification radii @xmath169 for the bulk and brane cases in 5d mssm for @xmath178 .
the running of the ckm matrix is governed by the terms related to the yukawa couplings , where @xmath183 can be used to observe the mixing angle , @xmath184 .
it decreases with the energy scale in a similar manner regardless of whether @xmath176 is small or large .
however , for a large initial value of @xmath168 ( small @xmath176 ) , the mixing angles have a more rapid evolution and end in the regime where the top yukawa diverges and develops a singularity . in the 5d mssm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and brane case ( right panel ) for @xmath178 where the solid line is the mssm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] in the 5d mssm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and brane case ( right panel ) for @xmath178 where the solid line is the mssm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] we next turn our attention to the quark flavor mixing matrix , especially the complex phase of the ckm matrix which characterises cp - violating phenomena . from fig.[fig :
j_uedsm ] the variation in the jarlskog parameter ( @xmath185 ) becomes very significant .
the larger the value of the compactification radius @xmath4 , the faster @xmath185 evolves to reach its maximum . in the ued sm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the sm , for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] in the ued sm as a function of the scale parameter @xmath67 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the sm , for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] from fig.[fig : j_bulk_brane_30 ] , in contrast , the jarlskog parameter decreases quite rapidly once the initial kk threshold is passed .
however , when @xmath176 is large , we have a relatively longer distance between the initial and terminating energy track , the evolution of @xmath185 can be driven towards zero or even further . for the matter fields constrained to the brane , in figs.[fig : vub_brane ] and [ fig : j_bulk_brane_30 ]
we observe that the evolutions of these mixing angles and cp violation parameter are decreasing irrespective of whether the top yukawa coupling grows or not . for small @xmath83
we see similar evolution behaviours for these parameters as in the bulk case .
however , as @xmath176 becomes larger , the top yukawa coupling evolves downward instead of upward .
the decreases in these ckm parameters then becomes much milder towards the unification scale ; though the reduction to effectively zero in the jarlskog parameter persists . as a result ,
for the brane localised matter field scenario , it is more desirable to have a large @xmath176 for theories that are valid up to the gauge coupling unification scale . in the 5d mssm as a function of the scale parameter @xmath67 for @xmath178 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the mssm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ] in the 5d mssm as a function of the scale parameter @xmath67 for @xmath178 , for the bulk case ( left panel ) and the brane case ( right panel ) where the solid line is the mssm evolution and for different compactification scales : @xmath169 = 2 tev ( red , dotted line ) , 8 tev ( blue , dot - dashed line ) , and 15 tev ( green , dashed line).,title="fig:",scaledwidth=45.0% ]
in a similar way to what was done for quark parameters , we can study the evolution of the masses , mixing and phases in the neutrino sector .
indeed the values of the measured mixing angles and the expected sensitivity of future experiments will allow to test at least partially the predicted evolutions of the neutrino parameters .
the mixing matrix which relates gauge and mass eigenstates is defined to diagonalise the neutrino mass matrix in the basis where the charged lepton mass matrix is diagonal .
it is usually parameterised as follows@xcite : @xmath186 with @xmath187 and @xmath188 ( @xmath189 ) .
we follow the conventions of ref .
@xcite to extract mixing parameters from the pmns matrix .
experimental information on neutrino mixing parameters and masses is obtained mainly from oscillation experiments . in general @xmath190
is assigned to a mass squared difference between @xmath191 and @xmath192 , whereas @xmath193 to a mass squared difference between @xmath192 and @xmath194 .
the current observational values are summarised in table [ ta2 ] .
data indicates that @xmath195 , but the masses themselves are not determined . in this work
we have adopted the masses of the neutrinos at the @xmath76 scale as @xmath196 ev , @xmath197 ev , and @xmath198 ev , as the _ normal _ hierarchy ( whilst any reference to an _ inverted _ hierarchy would refer to @xmath199 ev , with @xmath200 and satisfying the above bounds ) . for the purpose of illustration , we choose values for the angles and phases as the @xmath76 scale as : @xmath201 , @xmath202 , @xmath203 , @xmath204 , @xmath205 and @xmath206 .
the evolution equation for the observables in our 5d mssm are taken from@xcite . as expected @xmath176 plays an important role as all the mixing angles and phases depend on @xmath207 ( see appendix c in@xcite ) .
however , the new degrees of freedom ( the extra - dimensional fields giving rise to kk excitations of the zero modes ) become important at energies corresponding to their masses . for the ued sm , we see different behaviour for the brane case fig.[ued_brane_deltamass ] and bulk case fig .
[ ued_bulk_deltamass ] .
once the kk threshold is reached , both @xmath208 and @xmath209 decrease with increasing energy in the brane case , but they increase with the energy in the bulk case for the different radii of compactification .
the evolution of masses depends on the evolution of @xmath210 and @xmath39 coupling and the rg runnings in the ued sm bulk model are generally larger than those in ued sm brane model .
this is due to the fact that the coefficient @xmath211 in the brane model ( sec.([uedbranecoef ] ) ) and @xmath212 in the bulk model ( sec.([uedbulkcoef ] ) ) and also there is difference in @xmath213 in the two equations due to the trace of charged - fermion yukawa couplings in bulk model whereas such a contribution does not exist in brane model due to the absence of fermion kk excitations ( see the @xmath214 term in sec.([smcoef ] ) ) .
this lead to the increasing of observables in the bulk case and the decreasing in the brane case . for the 5d mssm , in general
, in the brane case , the evolution has the same form for the three masses @xmath215 , @xmath216 , @xmath217 .
this leads to a reduction of up to a factor of two for the masses at @xmath218 ( for a large radius , @xmath219 tev ) with respect to the mssm values at low energies .
the situation is more involved when analysing the mass squared differences .
we plot in figs.[brane_deltamass ] and [ bulk_deltamass ] the evolution of @xmath208 and @xmath209 both for the matter fields on the brane and for all fields in the bulk for tan@xmath220=30 and different radii of compactification . in the brane case
different behaviours as a function of the energy scale are possible as a relatively large interval in energy range is allowed for the effective theory . as explicitly illustrated in fig .
[ brane_deltamass ] , the relevant radiative corrections controlled by the gauge fields in secs.([mssmcoef ] , [ 5dbranecoef ] ) become dominant as energy goes up , which tends to reduce mass splitting , and an approximately degenerate neutrino masses spectrum at the high energy scale @xmath221 becomes favourable .
this is in contrast with the mssm , where the neutrino mass splitting becomes large at an ultraviolet cut - off .
therefore , it is very appealing that the neutrino mass splitting at low energy could be attributed to radiative corrections resulting from a degenerate pattern at a high energy scale . in fig.[bulk_deltamass ] , the bulk case tends to a non - perturbative regime , where the unitarity bounds of the effective theory are reached much faster and only a much shorter running can be followed using the effective theory . concerning the evolution of the mixing angles , as can be seen in figs.[theta13_bulk][theta12_tanb30 ] , in the ued bulk and brane cases , we have very small variation from the sm case because there is no dependence on tan@xmath220 and there is no quadratic term of @xmath179 in the rges
. however the mixing angles variation is more significant in the 5d mssm in which the largest effect is for @xmath141 .
as observed , due to the large quadratic term of @xmath179 in the beta function , the @xmath141 has a rapid and steep variation in the bulk case .
however , for the brane case , it has a relatively longer evolution track with the @xmath141 then being pulled further down until the termination point ( where the effective theory becomes invalid ) . in contrast , the running of @xmath143 and @xmath222 is much milder .
however , a running to @xmath223 can not be observed in any situation .
the running of the dirac phase @xmath144 in the ued sm case is very small .
the variation is stable and similar for the bulk and brane cases , there is very small deviation from the sm because all other mixing angles vary only by small quantities and the coefficient @xmath166 which appear in the variation of @xmath144 ( see appendix d of @xcite ) are linear in @xmath179 and there is no dependence on @xmath83 . noting that the dirac phase @xmath144 determines the strength of cp violation in neutrino oscillations . in the 5d mssm , the runnings we include
follow the general features presented in fig.[delta_bulk_brane_50 ] , with large increases possible once the first kk threshold is crossed .
the recent results from the daya bay and reno reactor experiments have established non zero values of @xmath143 .
therefore , the leptonic cp violation characterised by the jarlskog invariant @xmath224 becomes promising to be measured in the future long baseline neutrino oscillation experiments .
for leptogenesis related to the matter - antimatter asymmetry , we should note that the parameters entering the leptogenesis mechanism can not be completely expressed in terms of low - energy neutrino mass parameters .
note that in some specific models the parameters of the pmns matrix ( which contains cp asymmetry effects ) can be used @xcite . here
, the cp - violating effects induced by the renormalisation group corrections could lead to values of the cp asymmetries large enough for a successful leptogenesis , and the models predicting maximum leptonic cp violation , or where the cp - violating phase @xmath144 is not strongly suppressed , become especially appealing .
specific models with large extra dimensions in which leptogenesis is relevant at low scale can also be found in ref .
@xcite .
the running of the mixing angles are entangled with the cp - violating phases@xcite .
the phases @xmath225 and @xmath226 do not affect directly the running of the masses , while the phase @xmath144 has a direct effect on the size of @xmath227 , although its importance is somewhat reduced by the magnitude of @xmath143 . for further discussions of the correlation between these phases and mixing angles , refer to refs @xcite for details . finally , whilst the above results and analysis were for the normal hierarchy of neutrino masses , we did also review the inverted hierarchy , where from an analysis of the equations presented in the appendices of @xcite we obtain the same features and results for neutrino mass runnings ( though with different initial values at the @xmath76 scale ) . as
such , the figures for @xmath208 and @xmath209 remain unchanged .
possible changes in the angles and phases arise from the different signs for the @xmath228 terms present in each evolution equation , where the @xmath141 results remain approximately the same , and the small runnings of @xmath143 and @xmath222 would be up rather than down .
the present review of the renormalisation group evolution of the masses , mixing angles and phases of the ued models in the quark and lepton sectors brings together the results obtained in the recent years in this subject using a common notation .
the important physical points are discussed and the equations are written in compact way to show the unified approach to the different sectors of these models . for more technical details
we refer to the existing literature .
we plot their running up to the gauge unification scale only when relevant , since the introduction of new ultraviolet cutoff becomes imperative due to the scalar potential stability condition , and beyond this scale new physics should appear .
in contrast , in the ued brane model , the physics parameters have a full running till the gauge unification scale , since the higgs self coupling evolution has a finite value which thus excludes the vacuum stability concern and validates the theory up its full scale @xcite .
the ued model has substantial effects on the hierarchy between the quark and lepton sectors and provides a very desirable scenario for grand unification .
the scale deviation of renormalisation curves from the usual sm one depends closely on the value of the compactified radius r. the smaller the radius is , the higher the energy scale we need to differentiate the ued curve from the sm one .
a comparison between theoretical predictions and experimental measurements will be available once the lhc will be running at its full centre of mass energy .
this will set limits on the parameters of the ued model , and a precise determination of @xmath185 , @xmath180 or @xmath181 at high energy may lead to a discrimination between the sm and extra dimensional models .
in the case of the 5d mssm , we have reviewed the behaviour of the evolution equations for the quark and neutrino sector in a minimal supersymmetric model with one extra - dimension . for quarks ,
the 5d mssm scenarios with matter fields in the bulk or on the brane , give both results with small or no quark flavor mixings at high energies , especially for the mixings with the heavy generation .
the evolution of these ckm parameters have a rapid variation prior to reaching a cut - off scale where the top yukawa coupling develops a singularity point and the model breaks down . for the brane localised matter fields model , we can only observe similar behaviour for small values of @xmath176 , while for large @xmath176 , the initial top yukawa coupling becomes smaller , the gauge couplings then play a dominant role during the evolution of the yukawa couplings , and therefore the yukawa couplings decrease instead of increasing . concerning the neutrino sector , the evolution equations for the mixing angles , phases , @xmath208 and @xmath209 , within the two distinct scenarios , is also considered .
a larger @xmath83 typically leads to larger renormalisation group corrections .
neutrino masses evolve differently in the two models due to the sign of the ( different ) dominant contributions in the bulk and in the brane cases . to conclude this review we would like to note some of the remaining incomplete areas of investigation in the study of quark and lepton sector runnings in ued models .
whilst we have reviewed the simplest sm and mssm ued models , other alternative extra - dimensional geometries exist .
note that our two scenarios of all matter fields freely propagating in the bulk or brane localised represent the only possibilities for calculating unitary ckm or pmns matrices , where extensions to the runnings of yukawas with different numbers of matter fields in the bulk or brane are trivial extensions of the equations already reviewed here .
alternative extra - dimensional geometries are still to be investigated , such as 2ued models ( preliminary work @xcite contains errors and are incomplete studies of these sectors ) or situations with warped randall - sundrum style extra - dimensions ; though warp factors provide an additional problem of vertex factors now depending on the kk numbers , and so the equations would become of a completely different form to the ones provided here ( excepting extreme limiting cases ) .
asc would like to thank ad , at and the ipnl for their hospitality during his stay in lyon , where the first stage of this work was performed .
000 h. georgi , h. r. quinn and s. weinberg , phys . rev .
lett . * 33 * , 451 ( 1974 ) .
k. r. dienes , e. dudas and t. gherghetta , nucl .
b * 537 * ( 1999 ) 47 [ arxiv : hep - ph/9806292 ] .
k. r. dienes , e. dudas and t. gherghetta , phys .
b * 436 * ( 1998 ) 55 [ arxiv : hep - ph/9803466 ] .
j. c. pati and a. salam , phys .
d * 10 * , 275 ( 1974 ) [ erratum - ibid .
d * 11 * , 703 ( 1975 ) ] ; h. georgi and s. l. glashow , phys .
* 32 * , 438 ( 1974 ) ; h. fritzsch and p. minkowski , annals phys .
* 93 * , 193 ( 1975 ) ; f. gursey , p. ramond and p. sikivie , phys .
b * 60 * , 177 ( 1976 ) .
o. klein , z. phys .
* 37 * ( 1926 ) 895 [ surveys high energ .
* 5 * ( 1986 ) 241 ] .
i. antoniadis , phys .
b * 246 * ( 1990 ) 377 .
n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
b * 429 * ( 1998 ) 263 [ hep - ph/9803315 ] .
i. antoniadis , n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
b * 436 * ( 1998 ) 257 [ hep - ph/9804398 ] .
n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
d * 59 * ( 1999 ) 086004 [ hep - ph/9807344 ] .
n. arkani - hamed , l. j. hall , d. tucker - smith and n. weiner , phys .
d * 62 * ( 2000 ) 105002 [ hep - ph/9912453 ] .
l. randall and r. sundrum , phys .
* 83 * ( 1999 ) 3370 [ hep - ph/9905221 ] .
n. arkani - hamed , s. dimopoulos , g. r. dvali and j. march - russell , phys .
d * 65 * ( 2002 ) 024032 [ hep - ph/9811448 ] .
n. arkani - hamed and m. schmaltz , phys .
d * 61 * ( 2000 ) 033005 [ hep - ph/9903417 ] .
b. lillie , jhep * 0312 * ( 2003 ) 030 [ hep - ph/0308091 ] .
k. r. dienes , e. dudas and t. gherghetta , phys .
b * 436 * ( 1998 ) 55 [ hep - ph/9803466 ] .
k. r. dienes , e. dudas and t. gherghetta , nucl .
b * 537 * ( 1999 ) 47 [ hep - ph/9806292 ] .
m. s. carena , a. delgado , e. ponton , t. m. p. tait and c. e. m. wagner , phys .
d * 68 * ( 2003 ) 035010 [ hep - ph/0305188 ] .
l. randall and m. d. schwartz , jhep * 0111 * ( 2001 ) 003 [ hep - th/0108114 ] . g. cacciapaglia , c. csaki , c. grojean and j. terning , phys .
d * 71 * ( 2005 ) 035015 [ hep - ph/0409126 ] . c. csaki , c. grojean , j. hubisz , y. shirman and j. terning , phys .
d * 70 * ( 2004 ) 015012 [ hep - ph/0310355 ] . c. csaki , c. grojean , l. pilo and j. terning , phys .
* 92 * ( 2004 ) 101802 [ hep - ph/0308038 ] . c. csaki , c. grojean , h. murayama , l. pilo and j. terning , phys .
d * 69 * ( 2004 ) 055006 [ hep - ph/0305237 ] .
y. nomura , jhep * 0311 * ( 2003 ) 050 [ hep - ph/0309189 ] .
h. davoudiasl , j. l. hewett , b. lillie and t. g. rizzo , jhep * 0405 * ( 2004 ) 015 [ hep - ph/0403300 ] .
h. davoudiasl , j. l. hewett , b. lillie and t. g. rizzo , phys .
d * 70 * ( 2004 ) 015006 [ hep - ph/0312193 ] .
r. barbieri , a. pomarol and r. rattazzi , phys .
b * 591 * ( 2004 ) 141 [ hep - ph/0310285 ] .
g. servant and t. m. p. tait , nucl .
b * 650 * ( 2003 ) 391 [ hep - ph/0206071 ] .
h. -c .
cheng , k. t. matchev and m. schmaltz , phys . rev .
d * 66 * ( 2002 ) 036005 [ hep - ph/0204342 ] .
t. appelquist , h. c. cheng and b. a. dobrescu , phys .
d * 64 * , 035002 ( 2001 ) [ arxiv : hep - ph/0012100 ] .
s. dimopoulos and g. l. landsberg , phys .
* 87 * ( 2001 ) 161602 [ hep - ph/0106295 ] .
s. b. giddings and s. d. thomas , phys .
d * 65 * ( 2002 ) 056010 [ hep - ph/0106219 ] .
i. antoniadis , k. benakli and m. quiros , phys
. lett .
b * 331 * ( 1994 ) 313 [ hep - ph/9403290 ] , e. accomando , i. antoniadis and k. benakli , nucl . phys .
b * 579 * ( 2000 ) 3 [ hep - ph/9912287 ] .
a. delgado , a. pomarol and m. quiros , phys .
d * 60 * ( 1999 ) 095008 [ hep - ph/9812489 ] .
n. arkani - hamed , h. c. cheng , b. a. dobrescu and l. j. hall , , phys .
d * 62 * , 096006 ( 2000 ) [ arxiv : hep - ph/0006238 ] .
t. appelquist , b. a. dobrescu , e. ponton and h. u. yee , phys .
lett . * 87 * , 181802 ( 2001 ) .
g. bhattacharyya , a. datta , s. k. majee and a. raychaudhuri , arxiv : hep - ph/0608208 .
a. j. buras , a. poschenrieder , m. spranger and a. weiler , nucl .
b * 678 * , 455 ( 2004 ) [ arxiv : hep - ph/0306158 ] ; a. j. buras , m. spranger and a. weiler , nucl .
b * 660 * , 225 ( 2003 ) [ arxiv : hep - ph/0212143 ] .
d. hooper and s. profumo , phys . rept .
* 453 * , 29 ( 2007 ) [ arxiv : hep - ph/0701197 ] ; p. colangelo , f. de fazio , r. ferrandes and t. n. pham , phys .
d * 77 * , 055019 ( 2008 ) [ arxiv:0709.2817 [ hep - ph ] ] and references therein .
a. datta , k. kong and k. t. matchev , new j. phys .
* 12 * , 075017 ( 2010 ) [ arxiv:1002.4624 [ hep - ph ] ] .
m. kobayashi and t. maskawa , prog .
* 49 * , 652 ( 1973 ) : n. cabibbo , phys .
* 10 * , 531 ( 1963 ) .
a. pomarol and m. quiros , phys .
b * 438 * ( 1998 ) 255 [ hep - ph/9806263 ] .
a. delgado , a. pomarol and m. quiros , jhep * 0001 * ( 2000 ) 030 [ hep - ph/9911252 ] .
a. s. cornell and l. -x .
liu , phys .
d * 83 * ( 2011 ) 033005 [ arxiv:1010.5522 [ hep - ph ] ] .
l. -x .
liu and a. s. cornell , pos kruger * 2010 * , 045 ( 2010 ) [ arxiv:1103.1527 [ hep - ph ] ] .
t. p. cheng , e. eichten and l. f. li , phys .
d * 9 * , 2259 ( 1974 ) .
a. deandrea , j. welzel , p. hosteins and m. oertel , phys .
d * 75 * ( 2007 ) 113005 [ hep - ph/0611172 ] . c. bouchart , a. knochel and g. moreau , phys .
d * 84 * , 015016 ( 2011 ) [ arxiv:1101.0634 [ hep - ph ] ] .
t. flacke , desy - thesis-2003 - 047 .
a. hebecker , nucl .
b * 632 * ( 2002 ) 101 [ hep - ph/0112230 ] .
n. arkani - hamed , t. gregoire and j. g. wacker , jhep * 0203 * , 055 ( 2002 ) [ arxiv : hep - th/0101233 ] .
n. marcus , a. sagnotti and w. siegel , nucl .
b * 224 * , 159 ( 1983 ) .
e. a. mirabelli and m. e. peskin , phys .
d * 58 * , 065002 ( 1998 ) [ arxiv : hep - th/9712214 ] .
i. l. buchbinder , s. j. j. gates , h. s. j. goh , w. d. i. linch , m. a. luty , s. p. ng and j. phillips , phys .
d * 70 * , 025008 ( 2004 ) [ arxiv : hep - th/0305169 ] .
a. s. cornell , a. deandrea , l. -x .
liu and a. tarhini , phys .
d * 85 * ( 2012 ) 056001 [ arxiv:1110.1942 [ hep - ph ] ] .
a. s. cornell , a. deandrea , l. -x .
liu and a. tarhini , arxiv:1206.5988 [ hep - ph ] . c. csaki , arxiv : hep - ph/0404096 . y. a. kubyshin , arxiv : hep - ph/0111027 .
v. a. rubakov , phys .
usp . * 44 * , 871 ( 2001 ) [ usp .
nauk * 171 * , 913 ( 2001 ) ] [ arxiv : hep - ph/0104152 ] .
a. perez - lorenzana , j. phys .
* 18 * , 224 ( 2005 ) [ arxiv : hep - ph/0503177 ] .
m. quiros , arxiv : hep - ph/0302189 ; n. arkani - hamed , t. gregoire and j. g. wacker , jhep * 0203 * ( 2002 ) 055 [ hep - th/0101233 ] . i. antoniadis , c. munoz and m. quiros , nucl . phys .
b * 397 * ( 1993 ) 515 [ hep - ph/9211309 ] .
p. c. west , `` introduction to supersymmetry and supergravity , '' j. wess and j. bagger , `` supersymmetry and supergravity , '' j. iliopoulos and b. zumino , nucl .
b * 76 * ( 1974 ) 310 ; j. wess and b. zumino , phys .
b * 49 * ( 1974 ) 52 .
p. h. chankowski and z. pluciennik , phys .
b * 316 * ( 1993 ) 312 [ hep - ph/9306333 ] .
s. antusch , m. drees , j. kersten , m. lindner and m. ratz , phys .
b * 519 * ( 2001 ) 238 [ hep - ph/0108005 ] .
k. s. babu , z. phys .
c * 35 * , 69 ( 1987 ) .
l. -x .
liu and a. s. cornell , arxiv:1204.0532 [ hep - ph ] .
m. bando , t. kobayashi , t. noguchi and k. yoshioka , phys .
d * 63 * , 113017 ( 2001 ) [ arxiv : hep - ph/0008120 ] .
m. bando , t. kobayashi , t. noguchi and k. yoshioka , phys .
b * 480 * , 187 ( 2000 ) [ arxiv : hep - ph/0002102 ] .
k. sasaki , z. phys .
c * 32 * , 149 ( 1986 ) .
m. e. machacek and m. t. vaughn , nucl .
b * 236 * , 221 ( 1984 ) .
l. x. liu , int .
j. mod .
a * 25 * , 4975 ( 2010 ) , arxiv:0910.1326 [ hep - ph ] . c. balzereit , t. mannel and b. plumper , eur .
j. c * 9 * , 197 ( 1999 ) [ arxiv : hep - ph/9810350 ] .
t. k. kuo and l. x. liu , arxiv : hep - ph/0511037 . z. -z .
xing , h. zhang and s. zhou , phys .
d * 77 * , 113016 ( 2008 ) [ arxiv:0712.1419 [ hep - ph ] ] .
k. nakamura _ et al .
_ [ particle data group collaboration ] , j. phys .
g g * 37 * ( 2010 ) 075021 and 2011 partial update for the 2012 edition .
s. weinberg , phys .
* 43 * ( 1979 ) 1566 .
m. blennow , h. melbeus , t. ohlsson and h. zhang , jhep * 1104 * ( 2011 ) 052 [ arxiv:1101.2585 [ hep - ph ] ] .
p. h. chankowski and z. pluciennik , phys .
b * 316 * ( 1993 ) 312 [ hep - ph/9306333 ] . k. s. babu , c. n. leung and j. t. pantaleone , phys . lett .
b * 319 * ( 1993 ) 191 [ hep - ph/9309223 ] .
s. antusch , j. kersten , m. lindner and m. ratz , nucl .
b * 674 * ( 2003 ) 401 [ hep - ph/0305273 ] .
z. maki , m. nakagawa and s. sakata , prog .
theor . phys .
* 28 * ( 1962 ) 870 .
f. p. an _ et al . _
[ daya - bay collaboration ] , phys .
* 108 * ( 2012 ) 171803 [ arxiv:1203.1669 [ hep - ex ] ] .
m. raidal , a. van der schaaf , i. bigi , m. l. mangano , y. k. semertzidis , s. abel , s. albino and s. antusch _ et al .
_ , eur .
j. c * 57 * ( 2008 ) 13 [ arxiv:0801.1826 [ hep - ph ] ] .
w. buchmuller , p. di bari and m. plumacher , nucl .
b * 665 * ( 2003 ) 445 [ hep - ph/0302092 ] .
p. -h .
gu , phys .
d * 81 * ( 2010 ) 073002 [ arxiv:1001.1340 [ hep - ph ] ] .
s. luo and z. -z .
xing , arxiv:1203.3118 [ hep - ph ] .
t. ohlsson and s. riad , phys .
b * 718 * , 1002 ( 2013 ) [ arxiv:1208.6297 [ hep - ph ] ] .
|
we review the renormalisation group evolution of quark and lepton masses , mixing angles and phases both in the ued extension of the standard model and of the minimal supersymmetric standard model .
we consider two typical scenarios : all matter fields propagating in the bulk , and matter fields constrained to the brane .
the resulting renormalisation group evolution equations in these scenarios are compared with the existing results in the literature , together with their implications .
|
quantum turbulence(qt ) , which consists of a tangle of quantized vortices , has been investigated since the thermal counterflow experiments of vinen [ 1 - 4 ] half a century ago , while the underlying physics is far from being fully understood [ 5,6 ] .
the numerical simulations are the useful source of knowledge about qt , because the whole dynamics of this system is too complicated to be described analytically .
one of the powerful schemes of the simulation is the vortex filament model based on pioneering works by schwarz [ 7,8 ] .
schwarz performed the numerical simulation of counterflow turbulence under the periodic boundary condition by using the localized induction approximation ( lia ) which neglects a non - local term of the biot - savart integral [ 8 ] .
however he could not obtain the statistically steady state ( sss ) because the vortices lie in planes normal to @xmath5 to prevent them from reconnecting .
therefore the unphysical mixing procedure , in which half of the vortices are randomly selected to be rotated by 90@xmath6 around the axis defined by the flow velocity , was used , and only this method enabled him to obtain the sss .
this failure reminds us that the lia is unsuitable through the absence of the interaction between vortices .
recently , thermal counterflow in superfluid @xmath7he was visualized by using solid hydrogen particles , and the velocity distribution of particles was observed [ 9 ] . in this experiment , two types of particles appeared .
some particles move straight along the normalflow with the approximately same velocity as normalflow .
other particles move zigzag along the direction of superfluid and has different velocity from superfluid . in the latter case , particles seem to be trapped in the core of the vortices .
we try to apply our numerical results for understanding these observations .
section 2 describes the equation of motion of vortices and the method of numerical calculation . in sec.3
we show the typical numerical results of vortex tangle .
section 4 studies the validity of the lia by comparing the lia calculation with the full biot - savart calculation . in the section 5
we present the velocity distribution of vortices in the sss of vortex tangle and compare it with that of particles in the experimental observation .
[ sec:2 ] at 0k the velocity @xmath8 of the filament at the point * s*(@xmath9,t ) is given by @xmath10 where the prime denotes the derivatives with respect to the arc length @xmath9 , and @xmath11 is the quantized circulation , @xmath12 is a cutoff parameter corresponding to a vortex core radius , @xmath13 and @xmath14 are the length of the two adjacent line elements that hold the point*s * between .
the first term shows the localized induction field arising from a curved line element acting on itself .
the second term represents the non - local field obtained by carrying out the integral of the biot - savart integral along the rest of the filament .
the third term @xmath15 is an applied field .
the lia used in some works ( e.g.,[7,8,10,11 ] ) means neglecting the second non - local term .
in contrast the calculation without the lia is called the full biot - savart calculation . at finite temperatures ,
the velocity @xmath16 is given by @xmath17,\ ] ] where @xmath18 and @xmath19 are the temperature - dependent friction coefficients , @xmath20 is the normalfluid velocity , and @xmath21 is calculated from eq.(1 ) . in the work [ 8 ]
the author neglected the third term of eq.(2 ) .
we performed the full biot - savart calculation in this work .
the concrete reconnection procedure used in this work is the following .
every vortex is represented by a string of points at intervals of almost @xmath22 .
when a point on a vortex approaches another point on another vortex more closely than the fixed space resolution @xmath23 , we join these two points and reconnect the vortices [ 12 ] . for the integration of the motion of eq.(2 ) in time we used the classical 4th order runnge - kutta method . we usually start with an initial vortex configuration of six vortex rings as shown in fig.1(upper left ) .
in this section we present the numerical simulations of counterflow turbulence at the temperature @xmath24=1.9k , in the computing box 0.1 @xmath25 0.1 @xmath25 0.1 @xmath26 , at applied conterflow velocity @xmath3=0.286 , 0.381 , 0.572 cm / s .
a typical result is shown in fig.1 . [ cols="^,^ " , ] [ test4 ] the vortices in our simulations have lower vertical velocities than the superfluid velocity just like the experimental observation . for the quantitative agreement with observation , we need to introduce other effects such as a distortion of vortices by particles .
the obtained numerical results with the full biot - savart law demonstrate that the initially smooth vortex rings develop to a vortex tangle of the statistically steady state .
we compared the numerical results by the full biot - savart calculation to that by the lia calculation .
the lia calculation could not sustain the statistically steady state in contrast to the full biot - savart calculation .
we compared the velocity distributions of vortices obtained by our numerical simulations to the observations by the visualization experiment of couterflow by using solid hydrogen particles .
our results agree with the experimental observation qualitatively in that most vortices have lower velocities than the superfluid velocity .
|
we performed the numerical simulation of quantum turbulence produced by thermal counterflow in superfluid @xmath0he by using the vortex filament model .
the pioneering work was made by schwarz , which has two defects .
one is neglecting non - local terms of the biot - savart integral ( localized induction approximation , lia ) , and the other is the unphysical mixing procedure in order to sustain the statistically steady state of turbulence .
we succeeded in making the statistically steady state without the lia and the mixing .
this state shows the characteristic relation @xmath1 between the line - length - density @xmath2 and the counterflow relative velocity @xmath3 with the quantitative agreement of the coefficient @xmath4 with some typical observations .
we compare our numerical results to the observation of experiment by paoletti _
et al _ , where thermal couterflow was visualized by solid hydrogen particles .
|
some important physical quantities displaying quantization properties , may sometimes be represented by means of _ momentum _ space integrals which exhibit their topologically invariant character .
considerable effort has been devoted to find these representations , since they are often very useful to prove their quantization , as well as their stability under perturbations .
this is the case , for instance , of the transverse conductivity @xmath3 in @xmath4 , which can be represented as a momentum space integral given by @xcite @xmath5 \ , \ ] ] where @xmath6 is the full fermion propagator .
as shown in @xcite , due to the ultraviolet behavior of the fermion propagator , this integral reduces to the kronecker topological invariant which labels the homotopy classes in @xmath7 , that is , @xmath3 is quantized .
a similar representation for the induced chern - simons coefficient has been obtained in @xcite by using a cubic lattice regularization for the euclidean fermionic action . in this case , the possible values for the chern - simons coefficient turn out to be labeled by the winding number characterizing the mapping between the three - dimensional torus in momentum space and the normalized quaternion corresponding to the fermion propagator : @xmath8 .
topology in momentum space has also been advocated in @xcite , in order to discuss the stability of neutrino masses in the standard model and the spectrum of excitations in effective two dimensional models such as helium-3 . on the other hand , in the context of yang - mills theories ,
one of the main open problems is that of understanding the non - perturbative generation of a mass gap for the gauge fields as a consequence of the dynamics in the infrared regime .
this problem is of crucial importance for confinement in qcd , and for analyzing finite temperature effects .
for instance , in @xcite , a gap equation , based on the introduction of gauge invariant mass terms , has been proposed and applied to the yang - mills theory in @xmath2 dimensions .
the aim of this work is to obtain a useful momentum space representation for detecting the existence of an induced gauge invariant mass , @xmath0 , for an abelian gauge field . in particular
, we shall be able to derive a general expression for @xmath0 , valid for different space - time dimensions and geometries in momentum space .
the paper is organized as follows .
section 1 is devoted to a short review of the well - known relationship between the induced gauge invariant mass and the chiral anomaly in @xmath1 dimensions . in section 2 , a general momentum space representation for the induced gauge invariant mass
is provided .
section 3 is devoted to the applications of the aforementioned representation to the case of square lattice geometry and to the case of finite temperature in @xmath2 dimensions .
section [ sec : conc ] presents our conclusions .
by gauge invariant mass , @xmath0 , we understand the value of the constant appearing in a gauge invariant mass term for an abelian gauge field . in @xmath9 dimensional euclidean spacetime ,
a gauge invariant mass term lagrangian @xmath10 is given explicitly by @xmath11 where @xmath12 is the transverse kronecker @xmath13 in @xmath14 dimensions .
this mass term for @xmath15 is explicitly gauge invariant , although , of course , at the price of introducing a non - locality in the action .
that is often the reason for not including this term in the classical lagrangian of a standard local quantum field theory . in spite of this fact ,
terms like this naturally arise when evaluating radiative corrections to the effective action . in _
@xmath16 in @xmath1 dimensions , i.e. , the schwinger model @xcite , it is induced by the one - loop vacuum polarization graph .
moreover , for such a model , the one - loop result is exact , since it does not suffer from higher order corrections .
these results can also be shown to be related to the chiral anomaly in @xmath1 dimensions , whence the non - renormalization of @xmath0 is inherited .
for this model , the non - locality can be made good since , in the bosonization approach , the gauge field may be written in terms of two scalar fields : @xmath17 . in terms of these scalars ,
the gauge invariant mass term becomes a standard , local , mass term for the @xmath3 field . in the @xmath1
dimensional case , the value of @xmath0 for a single fermionic flavor has the well - known value @xmath18 . in this article , we are interested in deriving general expressions for @xmath0 , as a first step to extending the useful @xmath1 dimensional result into several non trivial directions . in order to define the problem more
precisely , we write @xmath0 in terms of the exact vacuum polarization tensor @xmath19 .
the ward identity for @xmath19 , @xmath20 , allows us to write : @xmath21 where @xmath22 is a scalar function .
thus @xmath0 may also be obtained as : @xmath23 this definition implicitly assumes gauge and lorentz invariance , two conditions that , except explicit indication on the contrary , shall be maintained in everything that follows .
we shall be first concerned with models that can be described by an euclidean action with the following structure : @xmath24 where @xmath25 and @xmath26 denote the fermion and gauge field lagrangians , respectively .
let us begin with a consideration of the @xmath1 dimensional case , in the simplest non - trivial situation of massless @xmath27 .
this will amount to a re - derivation of known results , although we shall present them here from a different perspective .
the gauge field lagrangian is @xmath28 while fermionic matter is described by a massless dirac field @xmath29 , with a lagrangian : @xmath30 where we adopted the conventions : @xmath31 @xmath32 the exact vacuum polarization tensor @xmath33 is , as usual , defined in terms of the connected gauge field two - point function @xmath34 by : @xmath35 it is easy to see that there is , indeed , a relation between @xmath0 and the chiral anomaly @xmath36 , since the former may actually be derived from the latter , at least for massless @xmath16 in @xmath1 dimensions .
to that end , we define @xmath36 by @xmath37 where @xmath38 denotes the quantum average of the axial current in the presence of an external gauge field @xmath39 ( we use the cursive @xmath40 notation to distinguish it from the _ dynamical _ gauge field @xmath15 ) .
hence , the integrated form of the anomaly @xcite is @xmath41 where we used the relation : @xmath42 , and @xmath43 is the vector current . then , from our knowledge of the chiral anomaly , we may use the fact that @xmath44 is linear in @xmath45 to adopt the linear approximation for @xmath46 in ( [ eq : defanom1 ] ) , namely @xmath47 with @xmath48 denoting the coordinate space vacuum polarization function .
it should be noted that the @xmath48 appearing in ( [ eq : linear ] ) is exact , since only an expansion in the external field has been performed . inserting then ( [ eq : linear ] ) into ( [ eq : defanom ] ) and fourier transforming
, we find that : @xmath49 assuming that both limits exist . taking ( [ eq : pi0 ] ) into account ,
this amounts to : @xmath50 where @xmath51 is the total flux of @xmath52 in euclidean spacetime .
equation ( [ eq : anom2 ] ) yields @xmath0 in terms of the anomaly .
of course , the left hand side in equation ( [ eq : anom2 ] ) being the spacetime integral of the anomaly , will have the form of a coefficient times @xmath53 , namely : @xmath54 thus the content of ( [ eq : anom2 ] ) is that the coefficient @xmath55 gives precisely the value of @xmath56 .
now , only configurations with @xmath57 may have a finite action , and this implies that @xmath58 could have only been @xmath59 or an integer multiple of this quantity , i.e. , @xmath60 it must of course be possible to prove this quantization , starting from the fermionic side of the problem .
indeed , this is the case for the one - loop approximation , since we also know that , to that order , the integral of the anomaly is in fact the index of the dirac operator , namely : @xmath61 where @xmath62 denotes the number of zero modes of positive and negative chirality in the given background . in our conventions , it is easy to realize that . in the case in which a current - current interaction is introduced ,
the corresponding effect can be evaluated by means of the bosonization rules , which give @xmath63 where @xmath64 denotes the current - current coupling constant .
this relationship would imply , through equation ( [ eq : anom2 ] ) , a renormalized gauge invariant mass , in agreement with the result of ref.@xcite . also , when the fermions are massive , the right hand side of equation ( [ eq : defanom ] ) has to be supplemented with the additional term @xmath65 coming from the explicit chiral symmetry breaking .
accordingly , equation ( [ eq : anom1 ] ) reads @xmath66 due to the short ranged behavior of the massive dirac fields , the left hand side identically vanishes , which together with equation ( [ eq : pi0 ] ) gives a zero gauge invariant mass @xmath0 .
in particular , this implies that the contribution of the chiral anomaly cancels against that of the explicit chiral symmetry breaking term @xcite .
we try , in what follows , to use the fermionic point of view exclusively , in an attempt to derive the previous results therefrom , without assuming that a loop expansion has been performed . to that end , let us derive an exact expression for @xmath0 in terms of the fermion propagator .
we start from the ( @xmath67-dimensional ) general expression for @xmath33 , as given by the schwinger - dyson equations @xcite , which amount to the diagram of figure [ fig : fig1 ] , ( 0,0 ) # 1#2#3#4#5 @font ( 6703,3580)(1425,-4337 ) ( 2251,-2311)(0,0)[lb ] ( 7276,-2311)(0,0)[lb ] ( 4651,-961)(0,0)[lb ] ( 6376,-3286)(0,0)[lb ] ( 2851,-3286)(0,0)[lb ] ( 4726,-3436)(0,0)[lb ] where the white blobs are included to mean that the lines are full fermion propagators , while the black blob represents the full vertex functions .
the external legs are of course to be truncated , but we have drawn them for the sake of clarity . the analytic expression corresponding to figure [ fig : fig1 ] is @xmath68\ ] ] where @xmath69 is the momentum propagator and @xmath70 the exact vertex function .
noting that @xmath71 may be expressed as in ( [ eq : pimn ] ) , we see that @xmath72 or @xmath73\;.\ ] ] since the theory is gauge invariant , we may relate @xmath74 to the fermion propagator , by using the exact ward identity : @xmath75 this identity may be used in ( [ eq : mg1 ] ) to obtain @xmath76\;\ ] ] @xmath77\;,\ ] ] which is an expression we shall consider in detail below .
it is still formal , in the sense that we have not yet made explicit any regularization method . before exploring its divergences in the general case ,
we note that , to one loop order , they are the well - known divergences in the vacuum polarization graph of figure [ fig : fig2 ] .
( 0,0 ) # 1#2#3#4#5 @font ( 6703,3395)(1425,-4152 ) ( 2251,-2311)(0,0)[lb ] ( 7276,-2311)(0,0)[lb ] ( 4726,-3736)(0,0)[lb ] ( 4651,-961)(0,0)[lb ] ( 6376,-3286)(0,0)[lb ] ( 2851,-3286)(0,0)[lb ] in ( 1 + 1 ) the expression for @xmath0 suffers from both uv and ir divergences , as it is clear from ( [ eq : mg2 ] ) , since , when applied to the free dirac propagator , it yields : @xmath78 in order to make sense of this expression , we exclude a circle of radius @xmath79 around the origin of the momentum plane to avoid the ir singularity , and also use a pauli - villars regulator to tame the uv divergences .
this amounts to defining @xmath80 where @xmath81 is the region illustrated in figure [ fig : fig3 ] , and @xmath82 denotes the mass of the regulator field .
the momentum @xmath83 denotes the radius of the integration region , and of course @xmath84 since the theory is already regularized .
( 0,0)).,title="fig : " ] # 1#2#3#4#5 @font ( 5626,5624)(3038,-5933 ) ( 6706,-4651)(0,0)[lb ] ( 6931,-2311)(0,0)[lb ] ( 5131,-3526)(0,0)[lb ] by a straightforward application of the two dimensional gauss theorem , we may convert these integrals of divergences into fluxes of radial vector fields .
this procedure yields @xmath85\ ] ] which in the @xmath86 limit becomes : @xmath87 as it should be .
the procedure we have followed has a simple electrostatic analogy : @xmath0 is given by the integral of the divergence of an ` electric field ' , so by following that analogy it is proportional to the total ` electric charge ' . because of the regulator , though , only the charges at the origin are relevant ; namely , the subtraction due to the uv regulator leads to : @xmath88\ ] ] where the integral is taken along a small curve of radius @xmath79 enclosing the origin .
@xmath89 denotes the outer normal to @xmath90 . in what follows
we shall argue that this kind of expression can be generalized to the full theory .
indeed , we can always say that the role of _ any _ uv regularization will be to modify the large momentum behavior of the propagator , in such a way that all the points with infinite momentum can be identified . for the pauli - villars case ,
this can be shown to hold simply by combining the contributions of both fermion propagators into one integral , to define a regularized propagator .
namely , @xmath91 \;,\ ] ] which contains the divergence ( in momentum space ) of a vector field which decreases as @xmath92 at infinity
. this will be one of the properties we shall demand of a regularization at any finite order of the loop expansion ; namely , the points at infinity in the momentum space may , from the point of view of this calculation , be identified .
the other condition is that , in the small momentum region , i.e. momenta much smaller than the cutoff , the behavior of the propagator should be the same as for the unregularized propagator .
thus only the small region around zero may contribute .
this , in turn , will produce a non - vanishing answer only when the field is massless , as we shall see now . assuming a standard dirac action for the fermions , the general form of the fermion propagator in momentum space is , of course , @xmath93 where @xmath94 and @xmath95 are real functions that depend only on the scalar @xmath96 . then an application of ( [ eq : mg2 ] ) yields : @xmath97\;,\ ] ] @xmath98\ ] ] where @xmath99 counts the dimension of the dirac algebra representation . evaluating the integral along @xmath90 , @xmath100 since @xmath86
, we may conclude from here that , if the fermion is massive , @xmath101 , then @xmath102 .
this is indeed the case for the massive schwinger model .
regarding the massless case , we obtain : @xmath103 where we have used the normalization condition @xmath104 , which holds at any finite order of the loop expansion .
this condition fixes the residue of the perturbative electron propagator at the pole @xmath105 . strictly speaking
, this normalization will change in an interacting theory .
indeed , the fermion propagator can in general , be rewritten by using the spectral representation @xcite : @xmath106 where @xmath107 and @xmath108 are real functions .
it is then clear , by the linearity of this expression , that there will be a finite gauge invariant mass if there is an isolated pole at zero momentum , namely , if we can write : @xmath109 where @xmath110 is the multiparticle threshold . in this situation , we would obtain , @xmath111 where @xmath112 is a constant , smaller that @xmath113 because of the spectral condition : @xmath114 equation ( [ eq : mg13 ] ) may seem to contradict our remarks on the relation between @xmath115 and the anomaly , for example ( [ eq : quant ] ) .
the resolution of this apparent paradox is that the anomaly , as we understood it in ( [ eq : defanom ] ) , is defined in terms of the divergence of the current .
the matrix elements of this current , when evaluated through the reduction formulae , will require the introduction of a @xmath112 factor ( @xmath116 for each fermionic field ) , thus the proper relation that generalizes ( [ eq : quant ] ) is : @xmath117 the original expression for @xmath0 required a regularization procedure in order to be well defined .
however , being determined by the low momentum behavior of the propagator , @xmath0 should be independent of the regularization .
for example , for a generalized pauli - villars regularization in the one - loop case , one might include a set of @xmath118 regulator fields , and define : @xmath119 where @xmath120 denote the regularizing masses , and the index @xmath121 is reserved for the unregularized propagator , which has @xmath122 and @xmath123 .
the masses @xmath120 and the coefficients @xmath124 for @xmath125 are chosen in order to verify the desired behavior in the uv . however , since only the behavior around zero momentum is relevant , and all the regulators are massive , then just the unregularized massless propagator contributes .
it should be clear from what we have said above that the geometry of the momentum space is crucial in the problem of evaluating @xmath0 .
an interesting example of this is the case of a lattice regularized theory . indeed , assuming that the coordinate space has been discretized to an ( infinite ) square lattice with lattice spacing @xmath126 ,
the momentum space becomes a ( continuous ) torus .
the lattice points are defined then as the set of points @xmath127 with @xmath128 for @xmath129 .
the naive ( i.e. , no wilson term ) lattice fermion propagator for a massive dirac fermion then becomes : @xmath130 where @xmath131 , and the @xmath132 are continuous variables ( because the lattice is infinite ) , in the first brillouin zone , namely : @xmath133 then , the expression for @xmath0 is : @xmath134\;,\ ] ] where @xmath135 denotes the first brillouin zone .
the fact that we are dealing with a momentum torus is obvious because of the periodicity of the functions @xmath136 , a property which is preserved of course for more general propagators .
but the functions @xmath136 will simultaneously vanish for the @xmath137 values in the set @xmath138 which contain not just the origin but also three other unwanted points . the behavior of the lattice propagator is identical for each of these points , and equal to the one in the continuum propagator .
indeed , @xmath139 where @xmath140 is any of the points where both @xmath136 s vanish .
then , when @xmath141 , an application of the gauss law on the torus , with the circles around the four poles of the propagator excluded yields four times the contribution of the physical pole at zero .
this is because of the contribution of the ` charges ' at the doublers .
this is just a manifestation of the ` doubling ' problem of lattice fermions , related to the nielsen - ninomiya theorem @xcite .
thus , @xmath142_{lattice } \;=\ ; 4 \frac{e^2}{\pi } \;.\ ] ] another circumstance where the structure of momentum space allows for the emergence of a non - trivial gauge invariant mass , is the case of @xmath16 in @xmath2 dimensions at finite temperature in the presence of ` large ' @xmath143 field configurations @xcite .
parity conserving versions of @xmath144 , like the model introduced by dorey and mavromatos @xcite have been extensively studied as quantum field theory models at finite temperature @xcite .
most of what we shall say about this here , however , holds true both for the parity conserving and the parity breaking cases .
we now discuss @xmath144 with regards only to one of its aspects , namely , its gauge invariant mass .
we recall that in finite temperature quantum field theory lorentz invariance is lost . indeed ,
in the matsubara formalism , which we shall adopt , the time coordinate runs from @xmath145 to @xmath146 .
this lack of lorentz invariance implies that one shall , in principle , have different masses for the spatial and temporal components of the gauge field .
namely , the natural extension of ( [ eq : defmt ] ) to this case would be @xmath147 where @xmath148 is the transverse kronecker @xmath13 for the @xmath149 spatial dimensions , and @xmath150 is the imaginary time .
the two components @xmath151 and @xmath152 are the electric and magnetic masses , which can of course be different for @xmath153 .
these two masses play the role of components of a ` mass tensor ' for @xmath15 .
we shall deal exclusively with the magnetic mass .
it is not difficult to apply a similar derivation to the one used for the @xmath154 case , to obtain @xmath155\ ] ] where the sum runs over the matsubara frequencies @xmath156 , and @xmath69 is the finite temperature fermion propagator . for the free , massless fermion propagator in the presence of a large @xmath143
, we have @xcite : @xmath157 where @xmath158 .
@xmath143 is assumed to be constant , a fact what can always be achieved by a small gauge transformation .
the value of this constant is of course determined by the quantity @xmath159 , where @xmath160 is an arbitrary time dependent configuration .
then , @xmath161 which looks like a series of @xmath1 dimensional contributions . of course , whenever @xmath162 , we have a non - vanishing contribution , of the same kind as those appearing in @xmath1 dimensions .
that condition on @xmath143 amounts to @xmath163 thus , we may write the result for @xmath164 as : @xmath165 \;.\ ] ] the @xmath166 dependence should have been expected by dimensional analysis , since in @xmath2 dimensions @xmath167 has the units of a mass .
regarding the existence of particular points where the magnetic mass is generated , the physical reason for that is that when the condition for @xmath143 is met , dimensional reduction occurs because there is a massless mode , and the reduced model is then tantamount to a schwinger model .
in this work a useful momentum space representation for the gauge invariant mass has been obtained and applied to different situations , namely : massless and massive @xmath16 in @xmath1 dimensions , the @xmath1 dimensional square lattice and @xmath4 at finite temperature .
the key ingredient for this representation is the validity of gauge invariance expressed by the ward identity , and of course its consistency with the dynamics defined by the schwinger - dyson equations
. it should be clear from ( [ eq : mg2 ] ) that the value of @xmath56 depends entirely on the infrared behavior of the fermion propagator , resulting indeed from the infrared dynamics .
in particular , the momentum representation for the gauge invariant mass in @xmath1 dimensions turns out to be related to the chiral anomaly .
this relationship suggests that @xmath0 should possess some kind of stability against perturbations .
for instance , the introduction of a current - current interaction amounts to a smooth change @xcite of the coefficient @xmath168 in equation ( [ eq : mg12 ] ) , while preserving the vanishing of @xmath169 .
this means that , apart from a normalization factor , @xmath56 is still finite and non - vanishing .
a further comment we would like to add , and which is related to the previous discussion is one concerning the meaning of formula ( [ eq : mg2 ] ) .
the main difference between our momentum space representation for @xmath0 and similar representations @xcite lies in the presence of the factor @xmath168 .
this is related to the fact that in our case the formula can not be written just in terms of the su(2 ) projection of the fermion propagator as in the other cases .
therefore , when including interactions @xmath0 is not protected from being renormalized . in this sense , the quantized gauge invariant mass of the @xmath1 dimensional schwinger model is not stable against interactions . in the present case ,
what is stable against perturbations is the ratio between @xmath0 and @xmath112 .
if perturbations are included in the massive schwinger model or in the massless case , @xmath0 will however still be @xmath102 or @xmath170 , respectively .
we finish with a comment and outlook on the possible implications of our results to higher dimensional systems at zero temperature , where dimensional reduction does not occur .
it should be noted that the main difficulty is the fact that @xmath115 , as given by ( [ eq : mg2 ] ) , seems to be not interesting for the higher dimensional case , because of the ir behavior of the fermion propagators .
one possible way out of this , could be to consider non - standard fields like , for instance , dipole fields .
this , however , should require a re - derivation of the main results , since the structure of the model will , in principle , be different from the standard minimal coupling case .
results on this will be reported elsewhere .
the conselho nacional de desenvolvimento cientfico e tecnolgico cnpq - brazil , the fundao de amparo pesquisa do estado do rio de janeiro ( faperj ) , fundacin antorchas ( argentina ) , and the sr2-uerj are acknowledged for their financial support .
bib t. matsuyama , progress theoret.phys .
* 77 * , 711 ( 1987 ) .
k. ishikawa and t. matsuyama , nucl .
b280 * , 523 ( 1987 ) .
a. coste and m. lscher , nucl
b323 * , 631 ( 1989 ) .
g. e. volovik , _ interplay of real space and momentum space topologies in strongly correlated fermionic systems _ , proccedings of xviii cfif autumn school `` topology of strongly correlated systems '' , lisbon 8 - 13 october 2000 , * cond - mat/0012195*. r. jackiw and so - young pi , phys . lett . * 368b * , 131 ( 1996 ) .
j. schwinger , phys .
rev . * 128 * 2425 ( 1962 ) . for a general reference on anomalies
see , for example : + s. b. treiman , r. jackiw , b. zumino and e. witten , _ current algebra and anomalies _ , world scientific , singapore 1985 .
see , for example , c. itzykson and j - b zuber , _ quantum field theory _
, mc graw hill 1985 , chapter x. h. grosse , e. langmann and e. raschhofer , ann . of phys . *
253 * , 310 ( 1997 ) ; hep - th/9609206 .
r. rajaraman , _ solitons and instantons _ , north - holland 1987 .
l. h. karsten , phys . lett . *
104b * , 315 ( 1981 ) ; + l. h. karsten and j. smit , nucl .
phys * b183 * , 103 ( 1981 ) ; + h. b. nielsen and m. ninomiya , nucl .
phys * b185 * , 20 ( 1981 ) ; nucl .
phys * b193 * , 173 ( 1981 ) ; phys
* 105b * , 219 ( 1981 ) . for a recent review
see : y. shamir , hep - lat/9509023 .
see , for example , i. j. r. aitchison and c. d. fosco , nucl .
* b578 * , 199 ( 2000 ) . n. dorey and n. e. mavromatos , nucl . phys . *
b386 * 614 ( 1996 ) , + n. dorey and n. e. mavromatos , phys . lett . *
b266 * 163 ( 1991 ) .
|
we derive a general expression for the gauge invariant mass ( @xmath0 ) for an abelian gauge field , as induced by vacuum polarization , in @xmath1 dimensions . from its relation to the chiral anomaly , we show that @xmath0 has to satisfy a certain quantization condition .
this quantization can be , on the other hand , explicitly verified by using the exact general expression for the gauge invariant mass in terms of the fermion propagator .
this result is applied to some explicit examples , exploring the possibility of having interesting physical situations where the value of @xmath0 departs from its canonical value .
we also study the possibility of generalizing the results to the @xmath2 dimensional case at finite temperature , showing that there are indeed situations where a finite and non - vanishing gauge invariant mass is induced .
|
an iron - based superconductor was first discovered in the lafeaso system with a so - called zrcusias - type structure whose superconductivity is induced by partial substitution of flourine for oxygen.@xcite after this discovery , the development of new materials with an fe layer in which the superconductivity is considered to appear , have been performed , resulting in the discoveries of other iron - based superconductors , such as the so - called 122 system,@xcite fese@xmath9te@xmath10,@xcite lifeas,@xcite and so on . on the other hand , new superconductors with a zrcusias - type structure which consist of elements other than fe are also explored , for example , la@xmath9sr@xmath10niaso@xcite .
recently , mizoguchi and his collaborators have discovered the superconductivity in ceni@xmath0bi@xmath1 with a superconducting transition temperature @xmath2 4.2 k.@xcite series compounds cembi@xmath1 and cemsb@xmath1 have zrcusias - type structures with space group of @xmath11 corresponding to the so - called 1111 system , where m is a transition metal element such as mn , fe , ni , cu , ag , and so on.@xcite for cembi@xmath1 , bi1 , m , ce and bi2 sites correspond to as , fe , rare earth element and o sites , respectively .
the `` parent '' compound of ceni@xmath0bi@xmath1 , cenibi@xmath1 is a moderately heavy fermion antiferromagnet ; magnetic ordering appears at about 5 k and the electronic specific heat coefficient @xmath12 is relatively large ( 470 mj / k@xmath13 mol).@xcite this parent compound does not exhibit a superconductivity although the electric conductivity is metallic down to the lowest temperature.@xcite the superconductivity is induced by the deficiency of ni atom,@xcite as the superconductivity in the iron - based superconductor of the so - called 1111 system is induced by oxygen - deficiency.@xcite the resistivity and superconducting shielding signal in the magnetic field show the superconducting transition at about 4.2 k. at about 5 k , specific heat @xmath14 exhibits a jump of @xmath154 j / k mol . this jump may be attributed to the magnetic ordering of the ce 4@xmath8 moment because the entropy estimated by an integration of @xmath16 below 5 k corresponds with @xmath17ln2 , which is expected from the twofold degenerated ground state of the ce 4@xmath8 electron .
the clear jump of the specific heat is not observed at around @xmath6 because the jump caused by the superconducting transition is much smaller than the jump caused by the magnetic ordering , suggesting that the charge carrier which causes the superconductivity is not coupled with the ce 4@xmath8 electron and the mass is not enhanced.@xcite if the ce 4@xmath8 electron is coupled with the superconducting carrier in the present compound , the magnetic fluctuation of 4@xmath8 electron can induce the unconventional superconducting order parameter . in such a case , temperature dependence of the magnetic bragg peak should have some anomaly at @xmath6 .
then we have performed powder neutron diffraction measurements on ceni@xmath0bi@xmath1 in order to investigate the existence of the coupling between the ce 4@xmath8 electron and the superconducting carrier .
first , we present the structural parameters determined by rietveld analysis in sec .
. the elecrtonic state of this compound discussed in ref .
6 is based on the present result . in sec .
iiib , we report the magnetic structure and the temperature dependence of the ordered moment of the ce 4@xmath8 moment .
our results suggest that the ce 4@xmath8 electron does not contribute to the superconductivity in this compound , consistent with the suggestion in ref .
for the compound of ceni@xmath10bi@xmath1 , we found through a series of synthesis experiments that samples synthesized by conventional solid state reaction have a constant ni - deficiency of x=0.8.@xcite a powder sample of ceni@xmath0bi@xmath1 for the present neutron measurements was prepared by following method .
ce , ni and bi powders with a nominal composition of ceni@xmath0bi@xmath1 were used as starting materials .
the mixed powder was evacuated in a silica tube and heated at 500 for 10 h and 750 20 h. the obtained powder was ground and pressed into pellets , and it was heated 800 for 10 h in the evacuated silica tube . inductively coupled plasma ( icp ) spectroscopy was used to confirm the chemical composition , ceni@xmath0bi@xmath1 , which was consistent with the nominal composition .
the sample appears stable thermally in a dry box or an evacuated glass tube , but decomposes in an hour when exposed to an ambient atmosphere .
this decomposition would be induced by the reaction of an unusual valence state of bi ions in the material with water vapor in the atmosphere .
thus , we paid attention so as to avoid decomposition or degradation of the sample . the sample of about 14.3 g was used in neutron diffraction measurements .
electric resistivity and superconducting shielding measurements confirm that @xmath6 value and a volume fraction of shielding signal are the same as the values reported in ref .
the powder neutron diffraction pattern for the analysis on a crystal structure was collected by using the high - resolution powder diffractometer ( hrpd ) installed in the reactor jrr-3 of japan atomic energy agency ( jaea ) .
the neutron wavelength was 1.8234 @xmath18 and the collimation was open ( effective value of 35)-40-s-6 ( s denotes sample ) .
the pattern was collected at room temperature .
the powder sample was set in a vanadium holder enclosed in al cans filled with he gas in order to avoid the decomposition in the atmosphere .
diffraction data for the analysis on the magnetic structure were collected by using the triple - axes spectrometer tas-2 in reactor jrr-3 of jaea .
the neutron wavelength used was 2.3590 @xmath18 and the collimation was 14-40-s-40-80. the sample enclosed in al can was mounted in a closed - cycle refrigerator .
figure 1 shows the neutron powder diffraction pattern of ceni@xmath0bi@xmath1 obtained from the hrpd .
the observed pattern is collected by using the hrpd at room temperature .
vertical bars show the calculated position of bragg reflections including the impurities .
the solid line at the bottom of the figure is the difference between observed and calculated intesities , width=321 ] the observed data are shown by crosses .
structural analysis on the neutron powder diffraction pattern is performed by using the program rietan2000.@xcite the space group of @xmath11 is used because we know that the series compounds of cembi@xmath1 and cemsb@xmath1 have same structure as , shown in sec . i.
the diffraction pattern is analyzed including nibi@xmath19 , nibi and ni as impurities . because a small peak which can not be assigned by above impuries is observed at 2@xmath20 61.2 ,
the intensity data in the 2@xmath21 range of 60.9 - 61.5 are excluded in the analysis .
the occupation factor of ni is fixed at 0.8 which is determined by icp spectroscopy .
the obtained structural parameters of ceni@xmath0bi@xmath1 are shown in table i. .atomic positions of ceni@xmath0bi@xmath1 ( space group @xmath11 ) determined by rietveld analysis of neutron powder diffraction data at room temperature . obtained lattice parameters are @xmath22=4.5439(1 ) and @xmath4=9.6414(2 ) .
the @xmath17-factor , @xmath23 , is 6.04 @xmath24 , [ cols="<,<,<,<,<,<,<",options="header " , ] the errors of the parameters shown in the table are mathematical standard deviations obtained by the analysis . the diffraction pattern calculated by using refined parameters is shown in fig . 1 by solid lines .
the calculated line can reproduce the observed data .
mass fractions of the impurities are 9.1@xmath25 for nibi@xmath19 , 1.4@xmath25 for nibi and 0.8@xmath25 for ni , respectively .
the larger thermal parameter of ni relative to other sites may be due to the impurities .
results of 2@xmath21 scan at 11.2 k and 2.9 k , obtained from tas-2 are shown in fig .
2(a ) by gray and black lines , respectively .
bi@xmath1 obtained at 11.2 k ( gray line ) and 2.9 k ( black line ) .
the data are collected by using tas-2 .
miller indices are shown for nuclear bragg peaks .
the peaks marked by asterisks are contributed from impurities , for example , nibi@xmath19 .
( b ) magnetic bragg intensity obtained by subtracting the intensity at 11.2 k from the intensity at 2.9 k.,width=7 ] at 2.9 k a clear bragg peak appears at right - hand side of 002 reflection , corresponding to 100 reflection .
the intensities of 102 and 111 reflections are slightly enhanced at 2.9 k. these results show the existence of magnetic ordering at lower temperatures and the coexistence of magnetic ordering and the superconductivity . figure 2(b ) shows the intensity obtained by subtracting the intensity at 11.2 k from the intensity at 2.9 k. clear magnetic bragg peaks are observed at lower temperatures at the reciprocal lattice points corresponding to 100 , 102 , 111 , and 201 .
the data are scattered around 2@xmath20 53 and 63 , due to large nuclear bragg intensities of 112/103 and 113/200 reflections whose intensities are about 70000 and 50000 counts at their peak positions , respectively .
magnetic 113 reflection does not have clear peak profile due to the large experimental error mentioned above , although it should have almost the same intensity as the magnetic 201 reflection in the magnetic structure presented below . from these results
, we can consider the magnetic structure at low temperatures . for 10@xmath27 reflections ,
the clear magentic bragg peaks are observed at @xmath28 and they are nearly absent at @xmath29 , indicating that the magnetic moments of two ce atoms in a unit cell at ( 1/4 1/4 0.2691 ) and ( 3/4 3/4 0.7309 ) are antiparallel .
this intesity modulation of magnetic 10@xmath27 reflections against @xmath27 also shows that ni moments are almost independent of the observed magnetic bragg peaks even if the ni atoms have magnetic moments .
the scattering intensity from the ni moments should not depend on @xmath27 because two ni atoms are located on a flat layer at @xmath30=1/2 . in magnetic susceptibility measurements ,
the contribution of ni atom to the curie - weiss - like susceptibility is nearly absent.@xcite these results suggest that ni atom is nonmagnetic . the clear peaks are not observed at 00@xmath27 positions in fig .
if the ce moments are perpendicular to @xmath4 axis , magnetic 00@xmath27 reflections with @xmath29 must have large intensities . especially , the intensity of 001 reflection is expected to be about four times as large as the intensity of 100 reflection .
the absence of magnetic 00@xmath27 reflections indicates that the two ce moments are oriented along @xmath4 axis .
it is consistent with the anisotropy of the magnetic susceptibility , indicating that the antiferromagnetic easy - axis is the @xmath4 axis.@xcite then we can get the magnetic structure shown in fig .
3 . bi@xmath1 .
the ce atom at ( 1/4 1/4 0.2691 ) is the origin .
gray arrows show the magnetic moments of ce atoms.,width=5 ] in the figure , ce atom at ( 1/4 1/4 0.2691 ) is located at the origin .
the integrated intensities of observed nuclear and magnetic bragg reflections , @xmath31 , are shown in table ii .
the former intensities are estimated by using the diffraction data at 11.2 k and the latter ones are estimated by using the data shown in fig .
llllll + @xmath32 & @xmath31 & @xmath33 & @xmath32 & @xmath31 & @xmath33 + 0 0 1 & 8354(428 ) & 10316 & 1 0 0 & 1579(128 ) & 1594 + 0 0 2 & 8354(152 ) & 8607 & & & + 1 0 1 & 759(93 ) & 652 & & & + 1 0 2 & 7148(301 ) & 7734 & 1 0 2 & 827(141 ) & 715 + 003/110 & 3643(200 ) & 3843 & & & + 1 1 1 & 7490(304 ) & 5111 & 1 1 1 & 905(149 ) & 1008 + 1 1 2 & 51380(951 ) & 52171 & & & + 1 0 3 & 10002(325 ) & 7569 & & & + 0 0 4 & 3626(108 ) & 3520 & & & + 113/200 & 38757(1001 ) & 40004 & & & + 2 0 1 & 2842(231 ) & 2418 & 2 0 1 & 363(123 ) & 200 + the intensities of nuclear bragg reflections calculated by using the structural parameters shown in table i , @xmath33 , are also shown in right - hand side of @xmath31 in the table .
the calculated intensities almost correspond with the observed intensities . to estimate the ordered moment of ce sites at 2.9 k , the intesities of magnetic reflections obtained from the magnetic structure shown in fig .
3 are fitted to the observed intensities by using least - squares fitting program .
the estimated value of the ordered moment is 1.43(5 ) @xmath34 . here , we use the isotropic magnetic form factor of ce@xmath35.@xcite the calculated intensities of magnetic bragg reflections are shown in the right - hand side of the table .
the calculated intensities agree with the observed intensities , indicating that present magnetic structure and the obtained ordered moment are reasonable .
although 113 reflection should have the intensity of @xmath15200 in the calculation , the observed intensities can not be estimated due to the experimental error , as mentioned above .
the inset of fig .
4(a ) is the temperature ( @xmath26 ) dependence of the peak profile of 100 reflection . and the extrapolation of the fitting function to zero kelvin . in the inset : temperature dependence of the peak profile of 100 reflection.,width=7 ] as shown in the inset , 100 reflection appears below 5.45 k and the peak gradually develops with decreasing @xmath26 . figures 4(a ) and 4(b ) show the @xmath26 dependences of the integrated intensities of 100 and 102 reflections , respectively .
the intensity of 102 reflection has a large error bar in this scale because this reflection includes the nuclear scattering .
the intensities of both reflections increase with decreasing @xmath26 below about 5 k , consistent with the temperature at which the jump of the specific heat is observed.@xcite as shown in the main panel of fig .
4(a ) , the temperature development of the magnetic bragg intensity is monotonous down to 2.9 k and an anomaly is not observed at the superconducting transtion temperature , 4.2 k. it is quite different from the behaviors observed in heavy fermion superconductors in which the magnetic ordering and the superconductivity coexist . for example ,
cd - doped cecoin@xmath7 , ceco(in@xmath9cd@xmath10)@xmath7 with @xmath36=0.1 and 0.075 , exhibit the antiferromagnetic orderings of the ce moment at neel temperatures , @xmath37 3.0 k and 2.5 k , respectively .
superconductivities appear below @xmath15 1.2 k and 1.7 k , and the antiferromagnetic ordering and the superconductivity coexist below these temperatures.@xcite in these compounds , the magnetic bragg peaks appear below @xmath5 and their intensities increase with decreasing @xmath26 .
however , the developments of the magnetic bragg intensities are suppressed and they become almost flat to the temperature below @xmath6.@xcite the present @xmath26 dependence of the magnetic bragg intensity suggests that the ce 4@xmath8 electron is almost independent of the superconductivity and this compound is not a heavy fermion superconductor . the solid line in fig .
4(a ) shows the fitting of data below 5.0 k to the function that the intensity is proportional to @xmath38 and the extrapolation of the fitting function to 0 k. the amplitude of the momtent at 0 k obtained by the fitting is 1.71 @xmath34 . in the experimental accuracy ,
any structural change is not observed and the tetragonal symmery remains down to 2.9 k. in the tetragonal symmetry , energy levels of the ce 4@xmath39 state split into three doublets , @xmath40 , @xmath41 , and @xmath42 , by the crytalline electric field , where the two former doublets consist of linear combination of @xmath43 and @xmath44 states , and the latter consists of @xmath45 state .
in the case of ceagsb@xmath1 which also has a zrcusias - type structure , the ground state is the @xmath42 state , revealed by the measurement of the crytalline electric - field excitation .
the amplitude of the antiferromagnetic ordered moment is in good agreement with @xmath46 0.43 @xmath34.@xcite on the other hand , the ordered moment in this compound is much larger than the moment of ceagsb@xmath1 , indicating that the ground state of this compound is not the @xmath42 state .
such differences of the amplitude of the ordered moment and/or the ground state are caused by the differences of the structural parameters and constituent elements .
actually , other compounds with a zrcusias - type structure , cecubi@xmath1 and ceagbi@xmath1 , exhibit antiferromagnetic ordering with easy axes parallel to the @xmath4 direction .
their saturated ordered moments which are estimated by magnetization curves are 1.76 and 2.1 @xmath34 , respectively .
these behaviors are similar to the present compound while in cecubi@xmath1 and ceagbi@xmath1 , the superconductivity is not observed and their magnetization curves show metamagnetic behavior which is not observed in the present compound .
ceni@xmath0bi@xmath1 has the same crystal structure as the so - called 1111 system which has a maximum @xmath6 of about 55 k among iron - based high-@xmath6 superconductors .
theoretical studies on an iron - based high-@xmath6 superconductor point out that magnetic fluctuation arising from a nesting between fermi surfaces at @xmath47 and m points in an unfolded brillioun zone is an origin of the superconductivity@xcite .
experimentally , strong corelations between the high-@xmath6 superconductivity and the spin fluctuation are observed in an iron - based high-@xmath6 superconductor . the inelastic neutron scattering measurements on lafeaso@xmath9f@xmath10 with @xmath36=0.057 , 0.082 and 0.157 , show that the spin fluctuation observed in the samples with @xmath36=0.057 and 0.082 , whose @xmath6 are about 30 k , almost disappears in the sample with @xmath36=0.157 in which the high-@xmath6 superconductivity is almost suppressed.@xcite on the other hand , the superconductivity may appear without magnetic fluctuation in ceni@xmath0bi@xmath1 , because ni is nonmagnetic and the ce moment is almost decoupled with the superconducting carrier .
theaabsence of the magnetic fluctuation may indicate that the @xmath6 of ceni@xmath0bi@xmath1 is much lower than the values of iron - based superconductors although they have same crystal structures .
we have performed neutron powder diffraction measurements on a new superconductor , ceni@xmath0bi@xmath1 , with @xmath2 4.2 k. the structural parameters at room temperature are determined by the rietveld method .
the clear magnetic bragg peaks are observed at @xmath3=(0 0 0 ) below about 5 k. in the magnetic ordering phase , the two ce moments in the unit cell are antiparallel along the @xmath4 axis and the ni atom does not contribute to the magnetic bragg reflections . below @xmath6 , the superconductivity and antiferromagnetic ordering coexist .
the magnetic bragg intensity monotonously increases with decreasing @xmath26 below about 5 k and does not exbihit an apparent anomaly at @xmath2 4.2 k , suggesting that the 4@xmath8 electron of the ce atom is almost independent of the superconductivity .
the saturated ordered moment is about 1.7 @xmath48 .
the authors would like to thank k. kaneko for his fruitful discussion .
this work was supported by the funding promgram for world - leading innovative r&d on science and technology ( first ) , japan , and also supported by a grant - in - aid for specially promoted research 17001001 from the ministry of education , culture , sports , science and technology , japan .
y. kamihara , t. watanabe , m. hirano , and h. hosono , j. am .
chem . soc . *
130 * , 3296 ( 2008 ) .
m. rotter , m. tegel , d. johrendt , phys .
* 101 * , 107006 ( 2008 ) .
hsu , j .- y .
luo , k .- w .
yeh , t .- k .
chen , t .- w .
huang , p. m. wu , y .- c .
lee , y .- l .
huang , y .- y .
yan , and m .- k .
wu , natl .
105 * , 14262 ( 2008 ) .
x. c. wang , q. q. liu , y. x. lv , w. b. gao , l. x. yang , r. c. yu , f. y. li , and c. q. jin , solid state commun . * 148 * , 538 ( 2008 ) .
l. fang , h. yang , p. cheng , x. zhu , g. mu and h .- h .
wen , phys .
b * 78 * , 104528 ( 2008 ) .
h. mizoguchi , s. matsuishi , m. hirano , m. tachibana , e. takayama - muromachi , h. kawaji , and h. hosono , phys .
lett . * 106 * , 057002 ( 2011 ) .
h. flandorfer , o. sologub , c. godart , k. hiebl , a. leithe - jasper , p. rogl , and h. nol , solid state commun .
* 97 * , 561 ( 1996 ) .
a. thamizhavel , a. galatanu , e. yamamoto , t. okubo , m. yamada , k. tabata , t. c. kobayashi , n. nakamura , k. sugiyama , k. kindo , t. takeuchi , r. settai , and y. nuki , j. phys .
. jpn . * 72 * , 2632 ( 2003 ) .
m. h. jung , a. h. lacerda , and t. takabatake , phys .
b * 65 * , 132405 ( 2002 ) .
h. kito , h. eisaki , and a. iyo , j. phys .
* 77 * , 063707 ( 2008 ) .
f. izumi and t. ikeda , mater .
forum , * 321 - 324 * , 198 ( 2000 )
. magnetic susceptibility of ceni@xmath0bi@xmath1 can be fitted by the curie - weiss law in the temperature range of 30 @xmath49 300 k. from the curie constant , the effective moment is 2.80 @xmath34 , which corresponds to the effective moment of single crystal cenibi@xmath1 , 2.88 @xmath34 and 2.60 @xmath34 for @xmath50 and @xmath51 , reported in ref .
these values are larger than 2.54 @xmath34 of a free ce@xmath35 ion . however
, if the ni atom has a local moment and also contributes to the curie - weiss - type susceptibility , the curie constant and/or the effective moment have to be much larger than the observed values .
p. j. brown , _ international tables for crystallography _ , edited by a. j. c. wilson ( kluwer academic , dordrecht , 1992 ) , vol .
c , chap . 4 .
l. d. pham , tuson park , s. maquilon , j. d. thompson , and z. fisk , phys .
* 97 * , 056404 ( 2006 ) .
m. nicklas , o. stockert , tuson park , k. habicht , k. kiefer , l. d. pham , j. d. thompson , z. fisk , and f. steglich , phys .
b * 76 * , 052401 ( 2007 ) .
sunil nair , o. stockert , u. witte , m. nicklas , r. schedler , k. kiefer , j. d. thompson , a. d. bianchi , z. fisk , s. wirth , and f. steglich , proc .
* 107 * , 9537 ( 2010 ) .
s. araki , n. metoki , a. galatanu , e. yamamoto , a. thamizhavel , and y. nuki , phys .
b * 68 * , 024408 ( 2003 )
. i. i. mazin , d. j. singh , m. d. johannes , and m. h. du , phys .
lett . * 101 * 057003 ( 2008 ) .
v. cvetkovic and z. tesanovic : europhys . lett . *
85 * 37002 ( 2009 ) .
k. kuroki , s. onari , r. arita , h. usui , y. tanaka , h. kontani , and h. aoki : phys . rev
* 087004 ( 2008 ) . f. ma and z .- y .
lu , phys .
b * 78 * 033111 ( 2008 ) .
s. wakimoto , k. kodama , m. ishikado , m. matsuda , r. kajimoto , m. arai , k. kakurai , f. esaka , a. iyo , h. kito , h. eisaki , and s. shamoto , j. phys .
. jpn . * 79 * 074715 ( 2010 ) .
|
we have performed powder neutron diffraction on the new superconductor , ceni@xmath0bi@xmath1 with a superconducting transtion temperature @xmath2 4.2 k. the structural parameters of this compound at room temperature are determined by rietveld analysis .
below about 5 k , the clear magnetic bragg peaks with propagation vector @xmath3=(0 0 0 ) are observed .
the observed intensities of magnetic bragg peaks can be explained by the magnetic structure that the two ce moments in the unit cell are antiparallel along @xmath4 axis .
the magnetic bragg peaks are observed in the superconducting state , indicating the coexistence of the antiferromagnetic ordering and the superconductivity in this compound .
the intensity of magnetic bragg peak monotonously increases with decreasing temperature below @xmath5 and does not exhibit apparent anomaly at @xmath6 , obviously different from cases of heavy fermion superconductors in which the magnetic ordering and the superconductivity coexist , for example , cd - doped cecoin@xmath7 .
these results suggest that the 4@xmath8 electron of the ce atom is not coupled with the superconducting carrier , and the magnetic ordering is almost independent of the superconductivity in ceni@xmath0bi@xmath1 .
|
in superstring theory the elementary particles are not point - like objects but extended , string - like objects .
it is surprising that this apparently small change allows us to answer fundamental questions that in the context of the quantum field theory of point - like particles can not even be posed .
for example : why is the standard model gauge group @xmath3 ? why are there three families of particles ?
why is the mass of the electron @xmath4 mev ?
why is the fine structure constant @xmath5 ?
in addition , only superstring theory has the potential to unify all gauge interactions with gravity in a consistent way . in this sense
, it is a crucial step in the construction of the fundamental theory of particle physics to find a consistent superstring model in four dimensions accommodating the observed standard model ( sm ) , i.e. we need to find the superstring standard model ( sssm ) .
actually , this is the main task of what we call string phenomenology . in the late eighties , the compactification of the @xmath0 heterotic string on six - dimensional orbifolds proved to be an interesting method to carry out this task ( for a brief historical account see the introduction in ref . and references therein ) .
for example , it was shown that the use of two wilson lines on the torus defining the symmetric @xmath6 orbifold can give rise to four - dimensional supersymmetric models with gauge group @xmath7 and , automatically , three generations of chiral particles@xcite .
in addition , it was also shown that the fayet
iliopoulos ( fi ) d - term , which appears because of the presence of an anomalous @xmath8 , can give rise to the breaking of the extra @xmath8 s . in this way it was possible to construct@xcite supersymmetric models with gauge group @xmath9 , three generations of particles in the observable sector , and absence of dangerous baryon- and lepton - number - violating operators orbifold@xcite , as well as in the context of the @xmath10 orbifold@xcite , and @xmath11 orbifold@xcite , have been analysed . ] .
unfortunately , we can not claim that one of these @xmath6 orbifold models is the sssm , since several problems are always present .
for example , the initially large number of extra particles , which are generically present in these constructions , is highly reduced through the fi mechanism , since many of them get a high mass ( @xmath12 gev ) . however ,
in general , some extra @xmath13 triplets , @xmath14 doublets and @xmath15 singlets still remain at low energy . on the other hand ,
given the predicted value for the unification scale in the heterotic string@xcite , @xmath16 gev , the values of the gauge couplings deduced from lep experiments cannot@xcite be obtained gev . ] .
it was also not possible to obtain in these models the necessary yukawa couplings reproducing the observed fermion masses@xcite .
at this point , it is fair to say that almost 20 years have gone by since string phenomenology started , and the sssm has not been found yet . ] as acquittal on the charge we should remark that there are thousands of models ( vacua ) that can be built .
some of them have the gauge group of the sm or gut groups , three families of particles , and other interesting properties , but many others have a number of families different from three , no appropriate gauge groups , no appropriate matter , etc . a perfect way of solving this problem would be to use a dynamical mechanism to select the correct model ( vacuum ) .
such a mechanism should be able to determine a point in the parameter space of the heterotic string determining the correct compactification producing the gauge group @xmath17 , three families of the known particles , the correct yukawa couplings , etc .
the problem is that such a mechanism has not been discovered yet .
so , for the moment , the best we can do is keep trying , i.e. to use the experimental results available ( such as the sm gauge group , three families , fermion masses , mixing angles , etc . ) , to discard models .
although the model space is in principle huge , a detailed analysis can reduce this to a reasonable size .
for example , within the @xmath6 orbifold with two wilson lines , one can construct in principle a number of order 50000 of three - generation models with the @xmath18 gauge group associated to the first @xmath19 of the heterotic string . however , a study implied that most of them are equivalent@xcite , and in fact , at the end of the day , only 192 different models were found@xcite .
this reduction is remarkable , but we should keep in mind that the analysis of each one of these models is really complicated .
nevertheless , a certain degree of optimism is important when working in string phenomenology , and one can argue that if the sm arises from superstring theory there must exist one model with the right properties . in the present review we will adopt this viewpoint , and will assume that the sm arises from orbifolds constructions . instead of the painful work of searching for the correct orbifold model , we will try to deduce the phenomenological properties that such a model must have in order to solve the crucial problems mentioned above , with the hope that this analysis will allow us to make predictions that can be tested at the lhc .
in fact , all those problems , extra matter , gauge coupling unification , and correct yukawa couplings , are closely related .
the first two because the evolution of the gauge couplings from high to low energy through the renormalization group equations ( rges ) , depends on the existing matter@xcite . in section 2
we will discuss a solution to the gauge coupling unification problem implying the prediction of three generations of supersymmetric higgses and vector - like colour triplets at low energies@xcite . in this solution
the fi scale plays an important role .
concerning the third problem , how to obtain the observed structure of fermion masses and mixing angles , this is in our opinion the most difficult task in string phenomenology .
for example , the right model must reproduce also the correct mass hierarchy for quarks and leptons , @xmath20 , @xmath21 , etc .
, and this is not a trivial task , although it is true that one can find interesting results in the literature .
in particular , orbifold spaces have a beautiful mechanism to generate a mass hierarchy at the renormalizable level .
namely , yukawa couplings between twisted matter can be explicitly computed and they get suppression factors , which depend on the distance between the fixed points to which the relevant fields are attached@xcite@xmath22@xcite . the couplings can be schematically written as @xmath23 , with @xmath24 , and the @xmath25 are the moduli fields associated to the size and shape of the orbifold .
the distances can be varied by giving different vacuum expectation values ( vevs ) to these moduli , implying that one can span in principle five orders of magnitude the yukawa couplings@xcite@xmath22@xcite .
unfortunately , this is not the end of the story , since nature tells us that a weak coupling matrix exists with weird magnitudes for the entries , and that therefore we must arrange our up - and down - quark yukawa couplings in order to have specific off diagonal elements . ] . in section 3
we will see that to obtain this _ at the renormalizable level _ is possible if three higgs families and the fi breaking are present@xcite .
thus we have a common solution for the three problems mentioned above .
on the other hand , it is well known that dangerous flavour - changing neutral currents ( fcncs ) may appear when fermions of a given charge receive their mass through couplings with several higgs doublets@xcite .
this situation might be present here since we have three generations of supersymmetric higgses .
in section 4 we will address this potential problem , finding that viable scenarios can be obtained@xcite .
since we are interested in the analysis of gauge couplings , we need to first clarify which is the relevant scale for the running between the supersymmetric scale @xmath26 and the unification point .
let us recall that in heterotic compactifications some scalars singlets @xmath27 develop vacuum expectation values ( vevs ) in order to cancel the fi @xmath28-term , without breaking the sm gauge group .
an estimate about their vevs can be done with the average result @xmath29 gev ( see e.g. ref . ) .
after the breaking , many particles , say @xmath30 , acquire a high mass because of the generation of effective mass terms .
these come for example from operators of the type @xmath31 . in this way extra vector - like triplets and doublets and
also singlets become very heavy .
we will use the above value as our relevant scale , the so - called fi scale @xmath32 gev . as discussed in the introduction , we are interested in the unification of the gauge couplings at @xmath16 gev .
this is not a simple issue , and various approaches towards understanding it have been proposed in the literature@xcite .
some of these proposals consist of using string gut models , extra matter at intermediate scales , heavy string threshold corrections , non - standard hypercharge normalizations , etc . in our case
, we will try to obtain this value by using first the existence of extra matter at the scale @xmath26 .
we will see that this is not sufficient and , as a consequence , the fi scale must be included .
let us concentrate for the moment on @xmath33 and @xmath34 .
recalling that three generations appear automatically for all the matter in @xmath6 orbifold scenarios with two wilson lines , the most natural possibility is to assume the presence of three light generations of supersymmetric higgses .
this implies that we have four extra higgs doublets , @xmath35 , with respect to the case of the mssm .
unfortunately , this goes wrong . whereas @xmath36 remains unchanged ,
since the number of extra triplets is @xmath37 , the line for @xmath38 is pushed down with respect to the case of the mssm . as a consequence ,
the two couplings cross at a very low scale ( @xmath39 gev ) .
we could try to improve this situation by assuming the presence of extra triplets in addition to the four extra doublets .
then the line for @xmath36 is also pushed down and therefore the crossing might be obtained for larger scales . however , even for the minimum number of extra triplets that can be naturally obtained in our scenario , @xmath40 , i.e. @xmath41 , the `` unification '' scale turns out to be too large ( @xmath42 gev ) .
one can check that other possibilities including more extra doublets and/or triplets do not work@xcite .
thus , using extra matter at @xmath26 we are not able to obtain the heterotic string unification scale , since @xmath33 never crosses @xmath34 at @xmath16 gev .
fortunately , this is not the end of the story .
as we will show now , the fi scale @xmath43 is going to play an important role in the analysis . in order to determine whether or not the heterotic string unification scale can be obtained ,
we need to know the number of doublets @xmath44 and triplets @xmath45 in our construction with masses of order the fi scale @xmath43 .
it is possible to show that within the @xmath6 orbifold with two wilson lines , three - generation standard - like models must fulfil the following relation for the extra matter : @xmath46 .
then , it is now straightforward to check that only models with @xmath47 and therefore @xmath48 , may give rise to the heterotic string unification scale@xcite ( other possibilities for @xmath49 , @xmath50 do not even produce the crossing of @xmath33 and @xmath34 ) .
this is shown in fig . 1 for an example with @xmath51 , and assuming @xmath52 gev .
there we are using @xmath53 gev as will be discussed below . = 2.5 in note that at low energy we then have ( excluding singlets ) 3\{(3,2)+2(|3,1)+(1,2 ) } + 3\{(3,1)+(|3,1)+ 2(1,2 ) } , [ sm2 ] i.e. the matter content of the supersymmetric sm with three generations of higgses and vector - like colour triplets .
let us remark that in these constructions only the following patterns of matter with masses of order @xmath43 are allowed : a ) n_3^fi=0 , n_2^fi=12 & & 3\{4(1,2 ) } , + b ) n_3^fi=6 , n_2^fi=18 & & 3\{(3,1)+(|3,1)+6(1,2 ) } , + c ) n_3^fi=12 , n_2^fi=24 & & 3\{2[(3,1)+(|3,1)]+8(1,2 ) } , + d ) n_3^fi=18 , n_2^fi=30 & & 3\{3[(3,1)+(|3,1)]+10(1,2 ) } [ extra5 ] .
thus for a given fi scale , @xmath43 , each one of the four patterns in eq .
( [ extra5 ] ) will give rise to a different value for @xmath54 .
adjusting @xmath43 appropriately , we can always get @xmath16 gev .
in particular this is so for @xmath55 gev as shown in fig .
it is remarkable that this number is within the allowed range for the fi breaking scale as discussed above . for the pattern in fig . 1 corresponding to case _ a ) _ we have @xmath56 , and therefore @xmath57 gev . of course
, we can not claim to have obtained the heterotic string unification scale until we have shown that the coupling @xmath58 joins the other two couplings at @xmath59 .
the analysis becomes more involved now and a detailed account of this issue can be found in ref . .
let us just mention that the fact that the normalization constant , @xmath60 , of the @xmath61 hypercharge generator is not fixed in these constructions as in the case of guts ( e.g. , for @xmath62 , @xmath63 ) is crucial in order to obtain the unification with the other couplings . summarizing , the main characteristic of this scenario is the presence at low energy of extra matter .
in particular , we have obtained that three generations of higgses and vector - like colour triplets are necessary .
since more higgs particles than in the mssm are present , there will be of course a much richer phenomenology .
note for instance that the presence of six higgs doublets implies the existence of sixteen physical higgs bosons , where eleven of them are neutral and five charged . concerning the three generations of vector - like colour triplets , say @xmath28 and @xmath64
, they should acquire masses above the experimental limit @xmath65(200 gev ) .
this is possible , in principle , through couplings with some of the extra singlets with vanishing hypercharge , say @xmath66 , which are usually left at low energies , even after the fi breaking .
for example , in the model of ref . , there are 13 of these singlets
. thus couplings @xmath67 might be present . from the electroweak symmetry breaking ,
the fields @xmath66 a vev might develop . note in this sense that the giudice masiero mechanism to generate a @xmath68 term through the khler potential is not available in prime orbifolds as @xmath6 .
thus an interesting possibility to generate vevs , given the large number of singlets present in orbifold models , is to consider couplings of the type @xmath69 , similarly to the next - to - minimal supersymmetric standard model ( nmssm ) .
it is also worth noticing that some of these singlets might not have the necessary couplings to develop vevs and then their fermionic partners might be candidates for right - handed neutrinos have been proposed recently@xcite to solve the @xmath68 problem . ] . for the models studied in refs .
the extra colour triplets have non - standard fractional electric charge , @xmath70 and @xmath71 respectively .
in fact , the existence of this kind of matter is a generic property of the massless spectrum of supersymmetric models .
this means that they have necessarily colour - neutral fractionally charged states , since the triplets bind with the ordinary quarks .
for example , the model with triplets with electric charge @xmath71 will have mesons and baryons with charges @xmath72 and @xmath73 . on the other hand , the model studied in ref .
has ` standard ' extra triplets , i.e. with electric charges @xmath74 and @xmath75 ; these will therefore give rise to colour - neutral integrally charged states .
for example , a @xmath76-like quark @xmath28 forms states of the type @xmath77 , @xmath78 , etc .
let us finally mention that a detailed discussion about the stability of these charged states , how to solve possible conflicts with cosmological bounds , and their production modes can be found in ref . .
crucial ingredients in the above analysis were that all three generations of supersymmetric higgses remain light ( @xmath79 , @xmath80 ) , @xmath81 , and the fi breaking . and , precisely , both ingredients favour to obtain the correct yukawa couplings at the renormalizable level .
namely , having three families of higgses introduces more yukawa couplings , and after the fi breaking some physical particles appear combined with other states , and the yukawa couplings are modified in a well controlled way .
this , of course introduces more flexibility in the computation of the mass matrices .
let us recall that the @xmath6 orbifold is constructed by dividing @xmath82 by the @xmath83 ^ 3 $ ] root lattice modded by the point group ( @xmath84 ) with generator @xmath85 , where the action of @xmath85 on the lattice basis is @xmath86 , @xmath87 , with @xmath88 .
the two - dimensional sublattices associated to @xmath83 ^ 3 $ ] are shown in fig .
[ lattice ] . in orbifold constructions ,
twisted strings appear attached to fixed points under the point group . in the case of the @xmath6 orbifold there are 27 fixed points under @xmath84 , and therefore there are 27 twisted sectors .
we will denote the three fixed points of each two - dimensional sublattice as shown in fig .
[ lattice ] .
thus the three generations arise because in addition to the overall factor of 3 coming from the right - moving part of the untwisted matter , the twisted matter come in 9 sets with 3 equivalent sectors on each one .
let us suppose that the two wilson lines correspond to the first and second sublattices .
the three generations correspond to move the third sublattice component ( x @xmath89 o ) of the fixed point keeping the other two fixed .
as mentioned in the introduction , we must arrange our up - and down - quark yukawa couplings in order to have specific off diagonal elements , @xmath90 in principle this property arises naturally in the @xmath6 orbifold with two wilson lines@xcite@xmath22@xcite . for example , if the su(2 ) doublet @xmath91 corresponds to ( o o o ) , the three generations of ( 3,2 ) quarks to ( o o ( o , x , @xmath89 ) ) and the three generations of ( @xmath92 , 1 ) up - quarks to ( o o ( o , x , @xmath89 ) ) , then there are three couplings allowed from the space group selection rule ( the components of the three fixed points in each sublattice must be either equal or different ) : @xmath93 associated to ( o o o)(o o o)(o o o ) with @xmath94 1 , @xmath95 associated to ( o o o)(o o x)(o o @xmath89 ) with @xmath96 , and @xmath97 associated to ( o o o)(o o @xmath89)(o o x ) with @xmath98 . in this simple example
one gets one diagonal yukawa coupling without suppression factor and two off diagonal degenerate ones @xmath99 , but other more realistic examples producing the observed structure of quark and lepton masses and mixing angles can be obtained using three generations of higgses@xcite .
let us first study the situation before taking into account the effect of the fi breaking .
consider for example the following assignments of observable matter to fixed point components in the first two sublattices ; q o o u^c o o d^c x o + h^u o o h^d o [ assignments ] in this case the up- and down - quark mass matrices , assuming three different radii , are given by m^u = gna^u , m^d = gn_1 a^d , [ quarkmasses ] where @xmath100 is the gauge coupling constant , @xmath101 is proportional to the square root of volume of the unit cell for the @xmath6 lattice , and a^u= ( ccc v_1^u & v_3^u_5 & v_2^u_5 + v_3^u_5 & v_2^u & v_1^u_5 + v_2^u_5 & v_1^u_5 & v_3^u ) , a^d= ( ccc v_1^d & v_3^d_5 & v_2^d_5 + v_3^d_5 & v_2^d & v_1^d_5 + v_2^d_5 & v_1^d_5 & v_3^d ) .
[ quarkmasses2 ] here @xmath102 , @xmath103 denote the vevs of the higgses @xmath79 , @xmath80 respectively , and @xmath104 .
for example , for @xmath105 one has @xmath106 .
the elements in the above matrices can be obtained straightforwardly .
for example , if the higgs @xmath107 corresponds to ( o , o , o ) , then since the three generations of ( 3,2 ) quarks @xmath108 correspond to ( o , o,(o , x,@xmath89 ) ) and the three generations of ( @xmath109,1 ) quarks @xmath110 to ( o , o,(o , x,@xmath89 ) ) , there are only three allowed couplings , @xmath111 the corresponding suppression factors are given by 1 , @xmath112 , @xmath112 respectively , and are associated with the elements 11 , 23 , 32 in the matrix @xmath113 .
these matrices clearly improve the result obtained with only one higgs family . however , it is possible to show that although the observed quark mass ratios and cabbibo angle can be reproduced correctly , the 13 and 23 elements of the ckm matrix can not be obtained@xcite .
fortunately , this is not the end of the story because the previous result is modified when one takes into account the fi breaking .
in particular , it will be possible to get the right spectrum and a ckm with the right form@xcite . as discussed in the introduction and section 2 , some scalars
@xmath27 develop large vevs in order to cancel the fi @xmath28-term generated by the anomalous @xmath8 .
thus many particles @xmath30 are expected to acquire a high mass because of the generation of effective mass terms , and in this way vector - like triplets and doublets and also singlets become heavy and disappear from the low - energy spectrum .
this is the type of extra matter that typically appears in orbifold constructions .
the remarkable point is that the sm matter remain massless , surviving through certain combinations with other states@xcite .
let us consider the simplest example , a model with the yukawa couplings c_1 _ 1 f , c_2 _ 1 _ 2 , [ massive ] where @xmath114 denotes a sm field , @xmath115 denote two extra matter fields ( triplets , doublets or singlets ) , and @xmath116 are the fields developing large vevs denoted by @xmath117 .
it is worth noting here that @xmath114 can be an @xmath110 , @xmath118 , @xmath119 , @xmath120 or @xmath121 field , but not a @xmath108 field .
this is because in these orbifold models no extra ( 3,2 ) representations are present , and therefore the standard model field @xmath108 can not mix with other representations through yukawas .
clearly the ` old ' physical particle @xmath114 will combine with @xmath115 .
it is now straightforward to diagonalise the mass matrix arising from the mass terms in eq .
( [ massive ] ) to find two very massive and one massless combination .
the latter is given by f ( c_2^ * f - c_1^ * _ 2 ) .
[ massless ] notice for example that the mass terms ( [ massive ] ) can be rewritten as @xmath122 , where @xmath123 .
indeed the unitary combination is the massless field in eq .
( [ massless ] ) .
the yukawa couplings and hence mass matrices of the effective low energy theory are modified accordingly .
for example , consider a model where we begin with a yukawa coupling @xmath124 .
since we have f= ( c_2 f + c_1^ * _2 ) , [ oldfield ] then the ` new ' coupling ( involving the light state ) will be , which would induce another contribution to @xmath125 , is not in fact allowed .
for this to be the case the fields @xmath126 and @xmath114 would have had to have exactly the same @xmath127 charges .
this is not possible since different particles all have different gauge quantum numbers . ]
@xmath128 the situation in realistic models is more involved since the fields appear in three copies .
all these effects modify the mass matrices of the low - energy effective theory ( see eq .
( [ quarkmasses ] ) ) , which , for the example studied in ref .
, are now given by ^u = gna^u^c a^u b^u^c , m^d = gn_1a^d^c a^d b^d^c , [ finalud ] where a b = ( ccc v_1 _ 5 & v_3_5 & v_2 + v_3_5 ^ 2 & v_2 & v_1 + v_2_5 ^ 2 & v_1_5 & v_3/_5 ) , [ define ] and the parameters @xmath129 , @xmath130 , and @xmath131 depend on are in general complex vevs , and therefore they can give rise to a contribution to the cp phase .
this mechanism to generate the cp phase through the vevs of the fields cancelling the fi @xmath28-term was used first , in the context of non - renormalisable couplings , in ref . .
for a recent analysis , see ref . .
] @xmath132 @xmath133 , and their possible values are discussed in ref . .
as shown in ref . , for natural values of those parameters and the vevs , one can find configurations that obey the electroweak symmetry breaking conditions , and can account for the correct quark masses and mixings .
in addition to the magnitudes of the ckm matrix elements we also require a cp violating phase .
although it has been shown that observable cp violation can not be obtained at the renormalizable level in odd order orbifolds@xcite for a minimal higgs sector , the above matrices having in addition to the ` mixing ' of states three families of higgses avoid this problem@xcite .
thus one possibility here ( in addition to the one already mentioned in footnote _ i _ ) is to assume that the vevs of the moduli have an imaginary phase , which can occur when the flat moduli directions are lifted by supersymmetry breaking and find their minimum where the phases are non - zero@xcite .
such a phase feeds directly into @xmath134 .
it is easy to check that this phase is physically observable , and leads to a non - zero @xmath135 phase for the ckm matrix which is of order one .
let us finally mention that the correct masses for charged leptons can be obtained following a similar approach , as discussed in ref .
[ ] . for neutrinos
this turns out to be not sufficient , but a see - saw mechanism arising in a natural way in orbifolds might solve the problem [ ] .
the most challenging implication of an extended higgs sector is perhaps the occurrence of tree - level fcncs mediated by the exchange of neutral higgs states . clearly , having six higgs doublets ( and thus six quark yukawa couplings ) the transformations diagonalising the fermion mass matrices do not diagonalise the yukawa interactions .
since experimental data is in good agreement with the sm predictions , where such an effect is not present , the potentially large contributions arising from the tree - level interactions must be suppressed in order to have a model which is experimentally viable . in general , the most stringent limit on the flavour - changing processes
emerges from the small value of the @xmath136 mass difference@xcite . [ cols="^,^ " , ] a detailed discussion of fcncs in multi - higgs doublet models was presented in ref .
( see also the references therein ) .
we summarise here some relevant points and apply the method to the orbifold case@xcite , focusing on the neutral kaon sector and investigating the tree - level contributions to @xmath137 .
the latter is simply defined as the mass difference between the long- and short - lived kaon masses , @xmath138 where @xmath139 is the effective hamiltonian for the diagrams in fig .
[ fig : effh ] .
once all the contributions to @xmath140 have been taken into account , the prediction of this orbifold model regarding @xmath137 should be compared with the experimental value , @xmath141 mev . in ref .
the numerical approach was divided in two steps .
firstly , one focus on the string sector of the model , and for each point in the space generated by the free parameters of the orbifold ( @xmath142 ) , one derives the up- and down - quark mass matrices and computes the ckm matrix . further imposing the conditions associated with electroweak symmetry breaking , and fixing a value for @xmath143 , one
can then determine the values of @xmath144 and @xmath145 .
a secondary step requires specifying the several higgs parameters , which must obey the minimum criteria finally , the last step comprehends the analysis of how each of the yukawa patterns constrains the higgs parameters in order to have compatibility with the fcnc data . in particular , we want to investigate how heavy the scalar and pseudoscalar eigenstates are required to be in order to accommodate the observed value of @xmath137 .
let us summarize the analysis of the orbifold parameter space by commenting on the relative number of input parameters and number of observables fitted .
working with the six higgs vevs ( @xmath146 ) , and the orbifold parameters @xmath145 , @xmath112 , @xmath147 and @xmath148 , one can obtain the correct electroweak symmetry breaking ( @xmath149 ) , as well as the correct quark masses and mixings ( six masses and three mixing angles ) . in order to discuss now the tree - level fcncs ,
let us remark that the present orbifold model does not include a specific prediction regarding the higgs sector .
for instance , we have no hint regarding the value of the several bilinear terms , nor towards their origin .
concerning the soft breaking terms , the situation is similar . in the absence of further information
, we merely assume that the structure of the soft breaking terms is the usual one ( see ref . for further details ) , taking the higgs soft breaking masses and the @xmath150-terms as free parameters ( provided that the electroweak symmetry breaking and minimisation conditions are verified ) . in the absence of orbifold predictions for the higgs sector parameters , and motivated by an argument of simplicity , we begin our analysis by considering textures for the soft parameters as simple as possible . in particular , we arrive to four representative cases with the following associated scalar and pseudoscalar higgs spectra * @xmath151 gev ; + @xmath152 gev .
* @xmath153 gev ; + @xmath154 gev . * @xmath155 gev ; + @xmath156 gev . * @xmath157 gev ; + @xmath158 gev . in fig .
[ fig : dmk ] we plot the ratio @xmath159 versus @xmath112 , for cases ( a)-(d ) , and @xmath160 .
all the points displayed comply with the bounds from the ckm matrix . from fig .
[ fig : dmk ] it is clear that it is quite easy for the orbifold model to accommodate the current experimental values for @xmath137 . even though the model presents the possibility of important tree - level contributions to the kaon mass difference , all the textures considered give rise to contributions very close to the experimental value .
although ( a ) and ( b ) are not in agreement with the measured value of @xmath137 , their contribution is within order of magnitude of @xmath161 . as seen from fig .
[ fig : dmk ] , with a considerably light higgs spectrum ( i.e. @xmath162 tev ) , one is safely below the experimental bound , as exhibited by cases ( c ) and ( d ) .
this is not entirely unexpected given the strongly hierarchical structure of the yukawa couplings ( notice from eq .
( [ define ] ) that @xmath163 is suppressed by @xmath164 ) .
let us finally mention that the analysis for other neutral meson systems , @xmath165 , @xmath166 and @xmath167 , can be carried out in an analogous way@xcite
. additionally , and given the existence of flavour violating neutral higgs couplings , and the possibility of having complex yukawa couplings , it is natural to have tree - level contributions to cp violation . in the kaon sector
, indirect cp violation is parameterised by @xmath168 . from experiment one
has @xmath169 .
a comparison of this quantity with the theoretical result in orbifold models can be found in ref . .
we have attacked the problem of the unification of gauge couplings in heterotic string constructions . in particular , we have obtained that due to the fayet - iliopoulos scale , @xmath33 and @xmath34 cross at the right scale when a certain type of extra matter is present . in this sense three families of supersymmetric higgses and vector - like colour triplets might be observed in forthcoming experiments . the unification with @xmath58 is obtained if the model has the appropriate normalization factor of the hypercharge .
let us recall that although we have been working with explicit orbifold examples , our arguments are quite general and can be used for other schemes where the standard model gauge group with three generations of particles is obtained , since extra matter and anomalous @xmath8 s are generically present in string compactifications .
another advantage of these models is that they naturally predict three generations , and also that the three generations of higgs fields give enough freedom to allow an entirely geometric explanation of masses and mixings .
the fayet - iliopoulos mechanism plays also an important role here .
namely , after the gauge breaking some physical particles appear combined with other states , and the yukawa couplings are modified in a well controlled way . on the other hand , the presence of six higgs doublets poses the potential problem of having tree - level fcncs . by assuming simple textures for the higgs free parameters , we have verified for example that the experimental data on the neutral kaon mass difference can be easily accommodated for a quite light higgs spectra , namely @xmath170 tev .
the presence of a fairly light higgs spectrum , composed by a total of 21 physical states , may provide abundant experimental signatures at future colliders , like the tevatron or the lhc .
in fact , flavour violating decays of the form @xmath171 , or @xmath172 may provide the first clear evidence of this class of models .
this work was supported in part by the spanish dgi of the mec under proyectos nacionales fpa2006 - 05423 and fpa2006 - 01105 ; by the comunidad de madrid under proyecto hephacos , ayudas de i+d s-0505/esp-0346 ; and also by the european union under the rtn program mrtn - ct-2004 - 503369 .
j. giedt , ` spectra in standard - like @xmath6 orbifold models ' , _ annals phys . _
* 297 * ( 2002 ) 67 [ arxiv : hep - th/0108244 ] .
t. kobayashi , s. raby , r .- j .
zhang ` searching for realistic 4d string models with a pati - salam symmetry : orbifold grand unified theories from heterotic string compactification on a @xmath10 orbifold ' _ nucl .
* b704 * ( 2005 ) 3 [ arxiv : hep - ph/0409098 ] .
w. buchmuller , k. hamaguchi , o. lebedev , m. ratz , ` supersymmetric standard model from the heterotic string ' , _ phys .
* 96 * ( 2006 ) 121602 [ arxiv : hep - ph/0511035 ] ; ` supersymmetric standard model from the heterotic string ( ii ) ' , arxiv : hep - th/0606187 .
o. lebedev , h.p .
nilles , s. raby , s. ramos - snchez , m. ratz , p.k .
vaudrevange and a. wingerter , ` a mini - landscape of exact mssm spectra in heterotic orbifolds ' , arxiv : hep - th/0611095 .
kim and b. kyae , ` string mssm through flipped su(5 ) from @xmath11 orbifold ' , arxiv : hep - th/0608085 ; ` flipped su(5 ) from @xmath175 orbifold with wilson line ' , arxiv : hep - th/0608086 .
kim , j .- h .
kim and b. kyae , ` superstring standard model from @xmath175 orbifold compactification with and without exotics , and effective r - parity ' , arxiv : hep - ph/0702278 .
p. ko , t. kobayashi and j.h .
park , ` quark masses and mixing angles in heterotic orbifold models ' , _ phys .
* b598 * ( 2004 ) 263 [ hep - ph/0406041 ] ; ` lepton masses and mixing angles from heterotic orbifold models ' , _ phys . rev . _
* d71 * ( 2005 ) 095010 [ hep - ph/0503029 ] .
j. giedt , `` the km phase in semi - realistic heterotic orbifold models '' , _ nucl . phys . _
* b595 * ( 2001 ) 3 [ erratum - ibid .
* b632 * ( 2002 ) 397 ] [ arxiv : hep - ph/0007193 ] ; `` cp violation and moduli stabilization in heterotic models '' , _ mod .
* a17 * ( 2002 ) 1465 [ arxiv : hep - ph/0204017 ] .
b. acharya , d. bailin , a. love , w.a .
sabra and s. thomas , ` spontaneous breaking of cp symmetry by orbifold moduli ' , _ phys .
_ * b357 * ( 1995 ) 387 [ hep - th/9506143 ] ; d. bailin , g.v .
kraniotis and a. love , ` cp violation by soft supersymmetry breaking terms in orbifold compactifications ' , _ phys .
_ * b414 * ( 1997 ) 269 [ hep - th/970524 ] ; ` cp violating phases in the ckm matrix in orbifold compactifications ' , _ phys .
lett . _ * b435 * ( 1998 ) 323 [ hep - th/9805111 ] .
|
we review the possibility that the supersymmetric standard model arises from orbifold constructions of the @xmath0 heterotic superstring , and the phenomenological properties that such a model should have .
in particular , trying to solve the discrepancy between the unification scale predicted by the heterotic superstring ( @xmath1 gev ) and the value deduced from lep experiments ( @xmath2 gev ) , we will predict the presence at low energies of three families of higgses and vector - like colour triplets .
our approach relies on the fayet - iliopoulos breaking , and this is also a crucial ingredient , together with having three higgs families , to obtain in these models an interesting pattern of fermion masses and mixing angles at the renormalizable lebel .
namely , after the gauge breaking some physical particles appear combined with other states , and the yukawa couplings are modified in a well controlled way . on the other hand
, dangerous flavour - changing neutral currents may appear when fermions of a given charge receive their mass through couplings with several higgs doublets .
we will address this potential problem , finding that viable scenarios can be obtained for a reasonable light higgs spectrum .
|
this paper is about denotational semantics of ( a fragment of ) linear logic .
the focus is on ( denotational ) static and dynamic semantics , i.e. respectively _ semantics of proofs _ and _ semantics of proofs and their cut - elimination_. denotational semantics were introduced by scott @xcite as mathematical models for programming languages . through the proofs - as - programs correspondence , it is equivalent to study semantics of programs and semantics of proofs , and the current paper will focus on a fragment of linear logic .
denotational semantics are data - driven semantics , i.e. their focus is on data and data types .
a program then corresponds to a function that maps some input data to some output data .
this view of _ programs as functions _
is however arguably too coarse , as it models only the _ what _ ( a program computes ) and not the _ how _ ( the program computes ) .
consider a program @xmath0 and an input data @xmath1 , and the result @xmath2 that @xmath0 outputs when given @xmath1 as input .
denotational semantics will represent both @xmath3 @xmath0 given input @xmath1 and @xmath2 the result of @xmath0 s computation given input @xmath1 by the same object , i.e. @xmath4 .
dynamic semantics try to provide mathematical models that capture also _ how _ programs compute .
formally a dynamic semantic will be such that @xmath5 , but will provide an interpretation to the computation as well , i.e. there exists an operation @xmath6 in the model such that @xmath7 .
this paper presents a functor mapping the interpretations of multiplicative linear logic ( ) proofs in dynamic semantics to the interpretations of proofs in denotational semantics .
intuitively , one can understand this functor as taking a dynamic semantics and collapsing the interpretation of execution to an equality .
more precisely , we show how this collapse can be performed in a way that preserves _ quantitative information_. formally , the functor is defined as the antisymmetric tensor algebra construction .
since the paper is concerned with semantics of a fragment of linear logic , we will now speak of proofs rather than programs and of cut - elimination rather than program execution .
however , the reader should make a note to herself that , through the curry - howard correspondence , proof - theoretic and computer science terminologies are interchangeable .
[ [ static - semantics . ] ] static semantics .
static semantics are denotational models that _ do not account for _ the dynamics , i.e. the interpretation of a proof in a static model should be an invariant of the cut - elimination .
a well - studied static semantics for linear logic is the so - called _ relational model _ where formulas @xmath8 are interpreted as sets @xmath9 and proofs of the linear implication @xmath10 are interpreted as binary relations on @xmath11 .
another example of static semantics considered in this paper is girard s coherence spaces , which can be defined categorically as a tight double - glueing construction @xcite applied to the relational model .
_ quantitative _ denotational semantics are those which capture quantifiable properties of the programs it represents , such as time , space , and resource consumption .
that is , quantitative semantics are denotational models of computation that mirror more information than the so - called _ qualitative _ models , e.g. the relational model , coherence spaces .
origins of this approach can be traced back to girard s work on functor models for lambda - calculus @xcite that inspired linear logic @xcite which exhibited for the first time a decomposition of the semantic interpretation of lambda - terms as taylor series .
quantitative semantics has also provided static semantics for various algebraic extensions of lambda calculus such as probabilistic lambda calculi @xcite . by considering a refinement of the relational model , laird _
@xcite provided a uniform account of several static models accounting for quantitative notions .
finally , some quantitative generalisations of coherence space semantics have been considered , namely probabilistic @xcite and quantum coherence spaces @xcite . however , no uniform account of those has yet appeared in the literature .
[ [ dynamic - semantics . ] ] dynamic semantics .
we regroup under this terminology those semantics that do not solely account for proofs but also for their _ dynamics _ , e.g. game semantics and various _ geometry of interaction _ ( goi ) constructions . formally
, the interpretation of a proof of @xmath10 cut with a proof of @xmath1 will differ from the interpretation of the cut - free proof of @xmath2 obtained by cut - elimination .
although different models and techniques regrouped under this name may differ on certain aspects , their interpretation of programs / proofs is quite similar @xcite . on the technical level
, we will work in this paper with the author s _ interaction graphs _ construction .
these models include all of girard s geometry of interaction ( goi ) models @xcite , and provide a far - reaching generalisation of those .
indeed , interaction graphs should be understood as a quantitative variant of geometry of interaction . moreover , since goi models are strongly related to game semantics ( even more so when we restrict to as it done in this paper ) , interaction graphs can be thought of as a quantitative generalisation of those as well .
interaction graphs thus provide a natural framework to deal with quantitative dynamic interpretations with great generality .
one should note that goi models differ from those of game semantics by how the formulas are interpreted .
game semantics start with interpretations of formulas ( games ) and then define the interpretations of proofs . on the contrary , and though proof interpretations are the same , goi constructions define the interpretation of formulas from the `` untyped '' interpretation of proofs
. categorically , the latter construction is realised by a ( tight ) double - glueing construction @xcite defined from a notion of orthogonality between morphisms . [
[ contributions . ] ] contributions .
the first contribution of this paper is the definition of the categorical foundations of the interaction graphs constructions through the introduction of _
triskells_. this categorical apparatus is of interest in itself as it generalises both the bicategory of spans and the categories of matrices and arrays over a semiring .
furthermore , unlike in the work of laird _ et al .
_ , our definition is performed internally to the underlying category and does not refer to external objects .
this opens the way to defining even wider generalisations based on the replacement of the underlying category of sets with e.g. any topos . secondly , we shed light onto a new relationship between dynamic and denotational semantics .
this relationship is not only shown to hold for _ qualitative semantics _ , but also for _ quantitative semantics_. this provides formal grounds to the claim that the author s interaction graphs construction should be understood as a quantitative version of dynamic semantics
goi and game semantics .
though not the first work bridging dynamic and static semantics @xcite , the technique used here is novel and is the first to be applied to quantitative models . finally , we describe how this functor not only preserves the interpretation of proofs but also maps _
orthogonalities_. i.e. we show how this functor lifts to a map from the double - glueing constructions considered in interaction graphs and the double - glueing construction of coherence spaces .
in particular , the specific model detailed in the first paper by the author @xcite is shown to map to `` strict '' probabilistic coherence spaces .
this correspondence , although confined to specific cases when considering categories of relations , can be lifted to a much more general correspondence when considering the generalisation of weighted relations provided by triskells .
we end the paper by sketching a very general construction of _ quantitative coherence spaces _ suggested by the previous sections .
[ [ organisation - of - the - paper . ] ] organisation of the paper .
the first section introduces the notion of _ triskell _ in a category .
triskells are generalisations of spans that will account for the quantitative aspects available in the interaction graphs constructions .
the category of triskells will then ( section ) be shown to have the structure of a _ traced monoidal category _ , i.e. it has the structure to define a goi interpretation of proofs .
we also show how triskells can be mapped to weighted relations in the sense of laird _ et al .
these two results show that the category of triskells can accomodate both quantitative goi and quantitative static semantics .
the monoidal products interpreting the linear logic tensor in these two kinds of models are not the same however .
we thus introduce ( ) an endofunctor which maps one monoidal product to the other .
this functor thus maps the interpretation of a proof of in ( quantitative ) dynamic models to the interpretation of the same proof in ( quantitative ) denotational models .
finally , we study in the double - glueing construction of goi and show that it relates to the coherence space double - glueing through the functor just exhibited .
this result is however quite specific as it is based upon the notion of determinant and trace of matrices .
we thus end this paper by defining a notion of determinant and trace of triskells , allowing for the definition of both quantitative goi models and quantitative coherence space semantics in a very general setting [ [ acknowledgements . ] ] acknowledgements .
this work was partly supported by the european commision s marie skodowska - curie individual fellowship ( h2020-msca - if-2014 ) 659920 - react and the anr 12 js02 006 01 project http://lipn.univ-paris13.fr/~pagani/pmwiki/pmwiki.php/coquas/coquas[coquas ] .
the author also thanks hugh steele for his interest in this work , and his numerous comments about earlier versions of this paper .
the idea of triskells comes from the author s work on interaction graphs @xcite .
triskells are a generalisation of spans @xcite .
while the latter can be used to represent directed graphs , triskells possess the extra structure needed to interpret weighted edges .
a _ triskell _ @xmath12 in a category @xmath13 is defined as three morphisms sharing the same source : it is a tuple @xmath14 with @xmath15 .
we call @xmath16 ( resp .
@xmath17 , resp .
@xmath18 ) and @xmath19 ( resp .
@xmath20 , resp .
@xmath21 ) the _ source object _ ( resp .
_ target object _ ,
. _ weight object _ ) and _ source map _
( resp . _ target map _ , resp . _ weight map _ ) .
we denote by @xmath22 the set of all triskells with source object @xmath16 , target object @xmath17 and weight object @xmath18 . for composition to be defined ,
the set of weights must be fixed once and for all and endowed with an associative binary product , i.e. @xmath18 should have the structure of a monoid .
let @xmath23 be a monoidal category with @xmath24 the associativity , right and left unit maps of the monoidal product . a monoid object @xmath25 in @xmath13 is an object @xmath26 together with a multiplication @xmath27 and a unit @xmath28 , which satisfy associativity , i.e. @xmath29\otimes { { \tt m}})={{\tt m}}\circ({{\tt m}}\otimes\identity[{\mathrm{m}}])\circ\alpha$ ] , and identity , i.e. @xmath30\otimes { { \tt u}_{{\mathrm{}}}})=\rho_{{\mathrm{m}}}$ ] and @xmath31)=\lambda_{{\mathrm{m}}}$ ] . for our purposes
, we will suppose that @xmath13 is a cartesian category , i.e. the monoidal product is the cartesian product and the unit is a terminal object .
we will write @xmath32 the product , and @xmath33 $ ] the associated operation of _ pairing _ of morphisms .
this hypothesis is enough to ensure the existence , for each object @xmath34 of a unit map @xmath35 , i.e. a map such that @xmath36=f={{\tt m}}\circ [ { { \tt u}_{{\mathrm{e}}}},f]$ ] for all @xmath37 . from now on ,
all monoid objects considered will be monoid objects for the cartesian monoidal structure over @xmath13 .
we now fix a cartesian category @xmath13 with pullbacks , i.e. @xmath13 is finitely complete .
we also pick a monoid object @xmath18 , and two triskells @xmath38 with weight object @xmath18 and such that the source object of @xmath39 coincides with the target object of @xmath12 . we can define ( ) their _ composition _ @xmath40 as the tuple @xmath41)$ ] , where @xmath42 is a designated pullback of @xmath20 and @xmath43 . as
when working with spans , associativity is only satisfied up to isomorphism and the natural structure we obtain is not that of a category but rather a bicategory .
however , identifying isomorphic objects leads to a category .
cc [ cols="^,^ " , ] & + let @xmath13 be a complete category and @xmath18 a monoid in @xmath13 .
the bicategory @xmath44 is : @xmath45 the category @xmath46 is then defined as the quotient of @xmath44 w.r.t .
2-isormorphisms .
when they exist , coproducts will be denoted by @xmath47 , while the _ copairing _ of morphisms will be denoted by @xmath48 .
given @xmath49 and @xmath50 , we also define the morphism @xmath51 as the copairing @xmath52 where @xmath53 and @xmath54 are the natural inclusions .
the product of the underlying category @xmath13 naturally induces a monoidal product on triskells written @xmath55 ( ) . if @xmath13 has coproducts , then @xmath46 has two additional operations : the sum ( ) another monoidal product written @xmath56 , and the union , written @xmath57 ( ) , which is defined between two triskells with same source and target objects
. we will denote by @xmath58 where @xmath59 is the unique morphism @xmath60 the unit of the monoidal product @xmath55 .
whenever the underlying category has coproducts , we will denote by @xmath61 where @xmath62 denotes the unique map @xmath63 the unit of the monoidal product @xmath56 .
the category of triskells is of interest since it is the ambient category in which the author s interaction graphs construction takes place @xcite ( taking @xmath13 as the category of countable sets and restricting to finite objects ) .
although later works @xcite make use of a larger category ( in particular , one considers a sort of abelian enrichment to deal with additives @xcite ) , we restrict our discussion to the category of triskells in this paper .
if @xmath13 is the category of countable sets @xmath64 , then a triskell over @xmath18 whose source and target objects are finite sets is exactly a @xmath18-weighted directed graph from a set of edges @xmath34 to a set of vertices @xmath65.[remgraphs ] ] as defined in earlier work by the author @xcite .
composition corresponds to the composition of graphs by considering length 2 paths , while the trace defined below coincides with the _ execution _ @xcite .
the previous remarks explain the following terminology .
when considering a triskell @xmath14 over the category of sets , we will refer to the set @xmath34 as the _ set of edges _ , and we will refer to any element of @xmath34 as an _
we now fix once and for all the underlying @xmath13 to be a category of sets , i.e. @xmath13 is the category @xmath67 of all sets of cardinality less than @xmath68 where @xmath68 is any fixed cardinal .
we suppose moreover that @xmath67 has countable coproducts , i.e. @xmath69 .
we now explain how one can define a categorical trace @xcite on @xmath46 by constructing the _ triskell of paths_. we start from a triskell @xmath12 whose source and target objects are of the form @xmath70 and @xmath71 respectively
. one can define from it a triskell with same source and target @xmath72 which we denote @xmath73 .
the composition of @xmath74 copies of this triskell @xmath73 with the same source and target objects @xmath72 defines the `` triskell of paths of length @xmath74 '' .
pre- and post - composition with `` projections '' onto @xmath75 and @xmath76 respectively gives rise to the triskell of paths of length @xmath74 from @xmath76 to @xmath75 ( ) . finally , countable coproducts allows to build the triskell @xmath77 of all ( finite ) paths from @xmath76 to @xmath75 in @xmath12 , through the countable version of the union of triskells ( ) .
[ thm_tracedmonoidal ] for any monoid @xmath18 in @xmath67 , @xmath78 is traced monoidal .
the above theorem should hold in a more general setting .
first , the construction of the trace can be performed as long as the underlying category @xmath13 is finitely complete and has countable coproducts .
then , the proof that @xmath46 is indeed traced monoidal relies mainly on two additional properties of the category of sets , namely that coproducts are disjoint and stable under pullbacks
. in particular , one should be able to replace @xmath67 by any topos .
when @xmath18 is a complete semiring , one can define a functor from the category of triskells over @xmath18 to the category of weighted relations over @xmath18 through the following _ contraction of triskells_. given a triskell @xmath79 , the map @xmath80 $ ] factors uniquely as @xmath81 , where @xmath82 is a monomorphism ( and @xmath83 is an epimorphism ) .
we can additionally define a map @xmath84 as follows : @xmath85 .
we can therefore defined a triskell @xmath86 the contraction of @xmath12 as @xmath87 .
this triskell has a particular property : the pairing of the source and target maps is a monomorphism ; such triskells will be called _
simple_. this mimics a natural contracting operation @xcite that maps a ( weighted ) graph @xmath88 to a simple ( weighted ) graph @xmath89 with the same set of vertices @xmath90 .
this operation on triskells actually defines a functor from the category of triskells to the category of weighted relations defined by laird _
@xcite or equivalently the category @xmath91 of matrices with coefficients in a ( complete ) semiring @xmath18 .
let @xmath18 be a complete semiring .
we define the category @xmath92 as follows .
objects are sets and morphisms from @xmath93 to @xmath94 are matrices in @xmath95 .
composition , written @xmath96 , is defined as the matrix product .
the identity morphism is the identity matrix .
now , a matrix @xmath97 in @xmath95 can be written as a triskell @xmath98 : writing @xmath99 , we define this triskell by @xmath100 .
this correspondence is not one - to - one however , as zero coefficients can be represented both by edges with weight zero ( as in the triskell defined above ) or by a lack of edge .
we therefore define also the minimal triskell associated to @xmath97 as @xmath101 where @xmath102 and @xmath103 is the restriction of @xmath21 to @xmath104 .
conversely , every simple triskell @xmath105 defines a matrix @xmath106 in @xmath107 .
since @xmath80 $ ] is a monomorphism , we can consider @xmath34 a subset of @xmath108 .
then the coefficient @xmath109 is equal to @xmath110 if @xmath111 and null otherwise . to obtain an exact correspondence
, one should consider relations weighted in a complete semiring @xmath18 ( with a zero ) and simple triskells over @xmath112 @xmath18 bereft of its zero element .
it should be noted that the composition of relations is _ not _ the composition of ( simple ) triskells : the composition @xmath113 of two simple triskells
@xmath38 need not be simple in general and @xmath114 , with @xmath115 the natural projections . ] .
however , a simple computation shows that @xmath116 for all matrices @xmath117 in @xmath118 .
two monoidal products are considered by laird _
et al . _ on the category @xmath92 , namely the _ sum _ @xmath119 and the _ tensor _ @xmath120 .
it is not difficult to check that , through the identification of relations with simple triskells , those correspond to the monoidal products @xmath56 and @xmath55 on triskells .
furthermore , the operation of union of triskells corresponds to matrix addition .
let @xmath18 be a complete semiring .
we define the contraction functor @xmath121 from @xmath46 to @xmath122 as follows .
it acts as the identity on objects , and maps a triskell @xmath12 to the matrix @xmath123 of the contracted triskell @xmath86 .
[ thm_contract_monoidal ] the functor @xmath121 is a strict monoidal functor for the two monoidal structures , i.e. @xmath124 and @xmath125 .
this suggests the category @xmath126 as a natural generalisation of the category of weighted relations .
we will try to follow this intuition in the next sections .
we have now shown that triskells can be used to define both quantitative denotational and dynamic semantics for . however , these interpretations are based upon the interpretation of the tensor product as two distinct monoidal products . in the next section ,
we explain how one can define a strict monoidal functor on the category of triskells which maps the monoidal product @xmath56 onto the monoidal product @xmath55 .
we now define a functor from @xmath92 to itself which we call the _ fock functor _ , by analogy with the _ antisymmetric fock functor_. we recall that this functor is induced by the construction of the _ grassman algebra _ or _ exterior algebra _
* xix , 1 , page 733 ) . for the definition to make sense in our setting
, we will need now to suppose that the monoid @xmath18 has a structure of a continuous commutative ring , i.e. we need additive inverses in @xmath18 .
it is important to note that such algebraic structures do exist ( for instance , taking the ring of formal power series over a group ) , though they are not as usual as the well - known semiring @xmath127 . to consider simpler examples
, one can restrict to the subcategory of relations ( resp .
triskells ) between _ finite _ sets .
the continuity requirement is then useless and one could consider simple rings , such as the real numbers .
consider a triskell @xmath12 in @xmath128 , and choose @xmath129 and @xmath130 .
we define the set @xmath131 $ ] as the set of all pairs @xmath132 where @xmath133 is a permutation and @xmath134 is a sequence of edges such that @xmath135 and @xmath136 .
then for all such pairs @xmath132 we define the weight @xmath137(\vec{e})=\prod_{e\in\vec{e } } w(e)$ ] .
we then define @xmath137=\sum_{(\sigma,\vec{e})\in m_{{\mathcal{t}}}[\bar{a},\bar{b } ] } \epsilon(\sigma)\omega_{{\mathcal{t}}}[\bar{a},\bar{b}](\vec{e})$ ] where @xmath138 is the signature of the permutation @xmath133 .
notice that @xmath137 $ ] is the leibniz formula for the determinant of the submatrix of @xmath97 defined as @xmath139=(m_{a , b})_{a\in\bar{a } , b\in\bar{b}}$ ] . for any simple triskell @xmath12 with source object @xmath93 and target object @xmath94 , we denote by @xmath140 the existence of @xmath141 with @xmath142 , @xmath143 and @xmath144 . since @xmath12 is supposed simple , at most one such @xmath145 exists given a pair @xmath146 .
we define the _ fock _ functor @xmath147 from the category @xmath92 to itself as follows ( we consider morphisms defined as triskells ) . on objects @xmath147 acts as the finite powerset functor : @xmath148 . on morphisms
, we have @xmath149 in @xmath150 if and only if @xmath151 , @xmath152\neq\emptyset$ ] and @xmath153 $ ] .
let us consider the relation @xmath154 corresponding to the matrix @xmath155 between @xmath156 and @xmath157 .
then @xmath158 is a relation between @xmath159 and @xmath160 defined as follows .
first , @xmath161 is in relation with @xmath161 with weight @xmath162 .
then @xmath163 and @xmath164 are both in relation with @xmath157 with respective weights @xmath165 and @xmath166
. illustrates two more involved examples .
+ [ fock_monoidal ] the functor @xmath147 is strict monoidal from @xmath167 to @xmath168 .
this shows that the image of @xmath169 through the composition @xmath170 is the category @xmath171 .
i.e. the interaction graphs interpretation @xmath172 of a proof @xmath173 is mapped to its weigthed relational model s interpretation @xmath174 .
this result , however , can be improved by lifting the fock functor to the category of triskells .
we will now show how to lift the fock functor from simple triskells to general triskells , i.e. we will show there exists a functor @xmath175 such that the following diagram commutes .
we will define @xmath176 later on ; let us point out for now that it is a quotient of @xmath46 w.r.t . a simple congruence . ( td ) at ( 0,2 ) @xmath177 ; ( tp ) at ( 4,2 ) @xmath178 ; ( rd ) at ( 0,0 ) @xmath179 ; ( rp ) at ( 4,0 ) @xmath180 ; ( td ) ( rd ) node [ midway , left ] @xmath121 ; ( tp ) ( rp ) node [ midway , left ] @xmath121 ; ( rd ) ( rp ) node [ midway , above ] @xmath147 ; ( td ) ( tp ) node [ midway , above ] @xmath175 ; although it is necessary that @xmath18 have the structure of a complete ring ( or simply a ring if we restrict to the subcategories where objects are finite sets ) for the above diagram to commute , we will in fact not need much more than a monoid structure on @xmath18 for defining @xmath175 .
this is yet another argument in favour of considering triskells instead of relations .
this lifted functor is defined in a way which is similar to the definition @xmath147 , but simpler as one does not perform any summations .
i.e. every element of @xmath131 $ ] will give rise to a new edge .
the only thing we will need are signs , i.e. ways to interpret @xmath138 for a permutation @xmath133 .
this is done by picking a monoid @xmath18 that can be written as the product of a monoid with the monoid @xmath181 ( with multiplication ) . in the following
, we will call such structure a _ signed monoid_. we chose here to fix @xmath18 as a signed monoid , but we could have as well defined @xmath175 as a functor from @xmath126 to @xmath182 for any monoid @xmath18 .
we define @xmath175 as follows . on objects , @xmath175 acts as the finite powerset functor . on morphisms , it maps the triskell @xmath183 to the triskell @xmath184 defined as follows : for each @xmath132 in @xmath131 $ ] , there is exactly one edge of source @xmath185 , target @xmath186 and weight @xmath187(\vec{e})$ ] .
it turns out that @xmath175 is _ not _ a functor from @xmath46 to itself . to understand this , and understand why the same situations are not a problem with relations , let us consider the triskells ( actually these are relations ) @xmath12 and @xmath73 shown in and . then the fact that @xmath175 is not a functor can be seen by computing @xmath188 ( ) and @xmath189 ( ) . as the reader can see for herself , the contraction operation will cancel the edges from @xmath156 to @xmath190 in the triskell @xmath188 .
this is why @xmath147 is indeed a functor while @xmath175 is not .
however , one can show that @xmath175 is a functor _ up to some equivalence _ on morphisms .
for this , we define a congruence on the set of triskells over @xmath191 as follows : two triskells @xmath12 and @xmath73 are equivalent if and only if there exists two `` zero triskells '' @xmath192 and @xmath193 and a triskell @xmath194 such that @xmath195 and @xmath196 . here
zero triskell _ is a triskell whose set of edges @xmath197 can be decomposed as @xmath198 so that there is a bijection @xmath199 satisfying @xmath200 , @xmath201 and @xmath202 . then @xmath175 can be shown to be a functor from @xmath126 to the quotient @xmath203 of @xmath126 by this congruence , and this congruence is the smallest that makes @xmath175 a functor .
notice moreover that this congruence is contained into the equivalence defined from the contracting functor @xmath121 , i.e. the equivalence defined by @xmath204 if and only if @xmath205 .
this implies that the functor @xmath121 is well - defined on the quotient category @xmath203 , as two equivalent triskells in @xmath126 have the same image through @xmath121 .
lastly , one easily checks that this quotient is compatible with the monoidal structure , i.e. @xmath206 is a monoidal category .
[ fock_trik_monoidal ] the functor @xmath175 is strict monoidal from @xmath207 to @xmath206 , and @xmath208 .
a consequence of is the fact that a proof s interpretation in the dynamic models is mapped , through @xmath170 , to the interpretation of the same proof in the relational model , modulo the fact that we are working with a continuous semiring ( or a ring when considering only finite sets as objects ) .
more importantly , this last theorem [ fock_trik_monoidal ] extends this result when one considers triskells instead relations .
i.e. a proof s interpretation in the interaction graphs model for a monoid @xmath18 is mapped through @xmath175 to the interpretation of the same proof in the triskells denotational model for the monoid @xmath209 ( modulo the simple quotient considered above ) .
we will now show how this result actually extends to the double - glueing constructions .
in this section , we start discussing the image of the double - glueing construction considered in interaction graphs .
first , we limit our discussion to the subcategory of @xmath210 where all objects are finite sets .
a further restriction will be that @xmath18 is chosen as a subring of the complex numbers @xmath211 .
these constraints are quite strong but they were introduced by the author @xcite to obtain a combinatorial version of girard s hyperfinite geometry of interaction @xcite . in this particular case , we can define in the category of relations both the ( normalised ) trace i.e. the mean of diagonal coefficients and the determinant of matrices .
we now recall the following well - known theorem from linear algebra .
[ thm_matdettrace ] let @xmath1 be a matrix with complex coefficients , then @xmath212 .
we now fix the subring @xmath213 $ ] of the complex numbers as our weight object @xmath18 , and we restrict to matrices with operator norm at most @xmath162 .
this second restriction implies that the operations interpreting multiplicative linear logic will not create weights greater than @xmath162 @xcite , i.e. the fact that @xmath213 $ ] is not a subring ( i.e. not closed under addition ) will not induce incoherences in the model.$ ] .
this is clear for the disjoint union used to interpret the tensor product , it is proved for the trace in earlier works @xcite . ]
these restrictions still provide an adequate framework for defining the combinatorial variant of girard s hyperfinite model @xcite and probabilistic coherent spaces .
girard s hyperfinite goi uses the fuglede - kadison determinant @xmath214 @xcite , a ( positive-)real - valued determinant which corresponds , in finite dimensions , to the absolute value of the usual determinant ( to the power @xmath215 ) .
this determinant is used to define the orthogonality considered in girard s goi models , i.e. two matrices @xmath8 are ( det-)orthogonal , noted @xmath216 , when @xmath217 , i.e. when @xmath218 .
the trace , on the other hand , can be used to define ( a variant of ) the orthogonality considered in probabilistic / quantum coherence spaces , i.e. two matrices @xmath8 are ( trace-)orthogonal , denoted @xmath219 , when @xmath2200,1[$ ] . and
@xmath162 which are here forbidden .
this slight change however do not impact the result , i.e. we still define from this orthogonality a model of multiplicative linear logic . ]
the theorem above shows that @xmath221 , i.e. when we restrict to norm 1 matrices with coefficients in @xmath213 $ ] : @xmath2220,1 [ ~\longleftrightarrow~ \abs{\tr({\mathfrak{f}}(a){\mathfrak{f}}(b))}\in]0,1 [ ~\longleftrightarrow~ { \mathfrak{f}}(a){\perp_\mathrm{coh}}{\mathfrak{f}}(b)\ ] ] this implies that the interaction graph model detailed in the author s first work @xcite is mapped to the model of _ strict _ coherent spaces through the functor @xmath147 .
this result , while interesting on its own , can be generalised if one works with triskells instead of relations .
the reason triskells provide a better framework is that we will be able to get rid of the restrictions we imposed in this section .
indeed , we asked for objects to be restricted to finite sets : the reason for this is that dealing with infinite sets implies dealing with infinite sums .
thus infinite sets force us out of the usual domain of definition of matrix determinants .
the second restriction , i.e. that the weight object is a subring of the complex numbers , is also linked to the definition of the determinant .
indeed , the definition of a determinant for matrices over arbitrary semirings is quite involved and has no clear solution @xcite .
lastly , the condition on the norm comes from the very definition of girard s model : execution can not be defined in general when the norm is greater than @xmath162 @xcite .
we will now show how one can define a much more general and satisfying correspondence when dealing with triskells .
such a generalisation will provide many more models since there are fewer restrictions on the weight object @xmath18 .
this is done by introducing notions of determinant and trace of a triskell that generalise the usual definition for complex matrices .
we ask but one thing about these extensions , namely that they map the determinant to the trace through @xmath175 .
that is , we ensure that a generalisation of holds .
the remark above motivates a definition of determinant and trace in the categories of triskells .
this can be done in a coherent way , namely such that the equivalent of holds .
note also that the trace satisfies the property that @xmath223 , and will be used to define a notion of quantitative coherent spaces in @xmath126 .
let @xmath224 be a complete monoid
i.e. a monoid with a notion of infinite sums
, let @xmath18 be a signed monoid , and @xmath225 be any map , i.e. not necessarily a monoid ( or equivalently group ) homomorphism .
we define the @xmath82-trace and the @xmath82-determinant of any triskell with same source and target objects the finite set @xmath1 as @xmath226 } m(\epsilon(\sigma)\prod w(e_{i}))\ ] ] in the last expression , @xmath227 should be understood as the set of permutation over @xmath1 , and for all @xmath228 we denote by @xmath229 $ ] the subset of edges @xmath83 such that @xmath230 and @xmath231 .
if one keeps to the subcategory where objects are finite sets , then completeness of @xmath232 is not necessary and any monoid would do . as a particular case
, the identity map on complex numbers allows one to recover the usual determinant and trace of matrices .
although the trace defined as it is above can be used to deal with matrices over infinite sets , the determinant considered there can not , as one needs to define infinite products for the expression to make sense . given a monoid @xmath233 , we define the extended monoid @xmath234 by adding an absorbing element @xmath235 , i.e. @xmath236 for all @xmath237 .
one can then define , in this extended monoid , a satisfying notion of infinite products , namely @xmath238 is defined equal to @xmath239 if for all @xmath240 we have @xmath241 and equal to @xmath235 otherwise . by implicitely considering elements of @xmath233 as their image in @xmath234 and extending @xmath82 to @xmath234 by defining @xmath242 as being equal to the unit of the sum in @xmath243
, the expression above for determinant makes sense for arbitrary triskells . with this definition of determinant and trace ,
we obtain the following theorem .
[ thm_triskdettrace ] for any triskell @xmath194 with same source and target sets , @xmath244 .
we will describe here a notion of _ quantitative coherence spaces_. although coherence spaces and their probabilistic and quantum variations have been defined as a double - glueing construction on the category of ( weighted ) relations , we here describe them as a double - glueing construction applied to the category of triskells .
this change in the underlying category allows for a wider array of definable models , as the category of weighted relations uses a complete semiring to represent quantitative features , while triskells need only the simpler structure of a monoid .
we stress that this double - glueing on triskells gives rise to a double - glueing construction on weighted relations through the contraction functor @xmath121 . given a complete monoid @xmath224 , a map @xmath245 and a subset @xmath246 of @xmath232
, we define a binary relation the orthogonality on triskells by @xmath247 .
we define the orthogonal of a set of triskells @xmath1 as the set @xmath248 .
a quantitative coherence space ( qcs ) @xmath1 is a pair @xmath249 of a set @xmath250 and a bi - orthogonally closed set @xmath251 of triskells over @xmath250 ( i.e. of source and target @xmath250 ) , i.e. @xmath252 .
let @xmath253 and @xmath254 be qcs .
their tensor product is defined as @xmath255 as it is usual with this way of representing coherence spaces , no nice representation of the dual connective @xmath256 can be obtained .
the linear implication , however , is nicely characterised .
let @xmath257 be a triskell over @xmath258 , and @xmath194 be a triskell over @xmath259 .
we define the application @xmath260 $ ] as the triskell @xmath261 , where : @xmath262 , a\in e^{a}[x , x']\}\ ] ] @xmath263 if @xmath253 and @xmath254 are qcs , the linear arrow is defined as follows : @xmath264\in\cond{b}\})\ ] ] [ modelmll ] for any @xmath82 and @xmath246 , this defines a model of .
these models generalise previously introduced coherence spaces models . to see this
, we will need some additional definitions .
say a triskell @xmath265 is _ diagonal _ when @xmath266 and for all @xmath145 , @xmath267 . restricting morphisms to diagonal ( resp .
hermitian ) triskells defines a subcategory of the category of triskells . restricting the definitions of qcsjust considered to
this subcategory leads to the usual notions of coherence spaces and probabilistic coherence spaces ( in the style of girard @xcite ) .
restricting in the same manner to the straightforward notion of _ hermitian _ triskells is hermitian if @xmath268 , @xmath269 , and @xmath270 bijective s.t . @xmath271 . ]
leads to girard s quantum coherence spaces model .
[ girardqcoh ] the restriction of qcsto _ simple diagonal triskells _
( and finite objects ) yields usual coherence spaces when @xmath272 and probabilistic coherence spaces when @xmath273 ( for ) .
restriction to _ simple hermitian triskells _
( and finite objects ) leads to quantum coherence spaces ( for ) .
now , probabilistic coherence spaces were generalised to deal with infinite objects by danos and ehrhard ; this allowed them to define exponentials , something unavailable in girard s construction .
it turns out that their generalisation is also a special case of our construction . to formalise this , we consider the subcategory of diagonal triskells and the qcsone can define in this category . to simplify things , first notice that it is not necessary to consider @xmath273 :
since we are allowing multiple edges with same source and target , all @xmath274-weighted ( simple ) triskells can be obtained from @xmath213$]-weighted triskells through the collapse functor .
secondly , danos and ehrhard require two additional conditions that can be translated in our setting by considering the following definition .
we will use here _ ( weighted ) partial identities _ to express it , i.e. if @xmath275 , the partial identity on @xmath2 weighted by @xmath276 is the triskell @xmath277 where @xmath278 denotes the inclusion of @xmath2 in @xmath250 and @xmath279 is the constant function equal to @xmath280 ( definable through the `` unit '' map @xmath281 ) .
we say that a qcs@xmath1 is _ everywhere bounded below _ if for every @xmath282 there is a @xmath283 such that the partial identity of @xmath284 weighted by @xmath280 lies in @xmath251 .
if both @xmath1 and @xmath285 are everywhere bounded below , we say that @xmath1 is _
bounded_. this restriction to specific bi - orthogonally closed sets is reminiscent of a similar restriction in the igmodels @xcite .
it is not clear if the two restrictions are related , but they are strikingly similar at a first glance .
a more detailed study of these aspects will have to be relegated to later work , as it is intimately related to the interpretation of additives .
the constructions for multiplicative connectives in danos and ehrhard paper are then the same as the one considered here .
the essential point is that the tensor product of bounded qcsis again a bounded coherence space .
this leads to the following proposition .
[ deqcoh ] the restriction to _ bounded qcs _ in the subcategory of _ diagonal triskells _
( when @xmath286 $ ] ) yields danos and ehrhard probabilistic coherence spaces model ( for ) .
we have shown how qcscan be related to previous coherence space semantics .
we will now explain how it relates to the author s interaction graphs ( ig ) models .
we first prove a number of properties for the introduced notions of determinant and trace , which will then lead us to stating a formal link between the two families of models .
[ lineartrace ] if @xmath287 is a semiring in which infinite sums are invariant w.r.t .
every permutation of the index set ( not only finite ones ) and @xmath82 is a monoid morphism from @xmath18 to @xmath288 , i.e. @xmath289 , then @xmath290 is linear , i.e. @xmath291 and @xmath292 .
we can now relate these notions of trace and determinant to the measurement introduced in earlier work on interaction graphs @xcite , and denoted by @xmath293 .
we consider here triskells over finite sets , and the real numbers as the semiring @xmath232 .
notice the latter is not complete , but completeness is needed only for dealing with triskells over infinite sets .
[ measurementdet ] suppose @xmath294 and that @xmath82 is a monoid morphism such that @xmath295 .
suppose moreover that there is a family @xmath296 with @xmath297 and that @xmath8 are triskells over finite sets .
writing @xmath298 , we have : @xmath299{a , b}\ ] ] in other words , the orthogonality defined in interaction graphs by @xmath300{a , b}\in\bot$ ] and the orthogonality of qcsare mapped one to the other when we restrict to the hypotheses of the theorem .
this yields the following theorem .
[ quantfunctormodel ] the lifted fock functor maps the igmodel ( for ) to qcsmodels ( for ) defined on its image .
it is natural to wonder about the generalisation of qcsto larger fragments of linear logic .
first , the model extends quite straightforwardly to additives .
given qcs@xmath253 and @xmath254 , we define @xmath301 as usual , the dual @xmath302 is denoted @xmath303 . with this definition , and are easily extended to the fragment . however , the most interesting question concerns the extension to exponential connectives .
it turns out that one can define exponentials using the _ symmetric fock functor_. we define @xmath304 as follows . on objects , @xmath304 acts as the finite multisets @xmath305 functor . on morphisms , the functor maps the triskell @xmath183 to the triskell @xmath306 :
for each @xmath132 in @xmath131 $ ] , there is exactly one edge in @xmath306 of source @xmath185 , target @xmath186 and weight @xmath307(\vec{e})$ ] .
[ symmmonoidal ] the functor @xmath304 is strict monoidal from @xmath207 to @xmath308 .
given a qcs@xmath253 we define the following qcs .
@xmath309 [ modelll ] for any @xmath82 and @xmath246 , this defines a model of .
let us notice that this definition of exponentials is similar to that of exponentials in weighted relations models of laird _ et al .
more formally , given a triskell @xmath12 contracting to the weighted relation @xmath310 , the triskell obtained by applying the symmetric fock functor @xmath311 is mapped , through the contraction functor , to the relation denoted by @xmath312 in laird _
et al . _
we finally explain how this differs from the exponentials considered by danos and ehrhard @xcite .
first , we remark that if @xmath12 is a diagonal triskell , then @xmath311 will be diagonal as well .
then starting from a simple diagonal triskell @xmath12 corresponding to the relation @xmath154 over a set @xmath76 with weights in @xmath213 $ ] , we can compute the triskell @xmath313 and compare it to danos and ehrhard exponential @xmath314 . for any multiset @xmath82
, we can then show that @xmath315.r^{!}_{m}$ ] , where @xmath316 $ ] denotes the multinomial coefficient @xmath316=(\sum_{x\in{\mathrm{x } } } m(x))!/ \prod_{x\in{\mathrm{x } } } m(i)!$ ] ( identifying @xmath316 $ ] as a function @xmath317 ) .
although not equal , the two interpretations of exponentials are therefore related . of course , the exact definition of danos and ehrhard can be considered in our setting by defining a variant of the exponential in which weights are scaled by a @xmath318}$ ] factor .
this is however a less general construction as it can only be performed when the latter scalar can be defined .
it appears that the category of triskells provides an interesting generalisation to the category of weighted relations when it comes to defining quantitative denotational models . in this paper , we have presented the structure only for multiplicative linear logic , and it thus makes sense to study extensions to interpret proofs of larger fragments , e.g. additives and exponential connectives of linear logic , fixed points . as such extensions were considered by the author for the interaction graphs models , a related question is what these extensions are mapped to through the `` functor '' @xmath175 .
furthermore , these extensions can be studied not only at the level of proof interpretations , but also with respect to the double - glueing constructions .
indeed , quantitative coherence spaces models were defined on the category of triskells for larger fragments in the last section .
it should be noted , however , that to define satisfying exponential connectives , one has to consider generalisations of graphs @xcite .
lastly , the notion of triskell seems of interest in itself , especially considering the fact that its definition is internal to the underlying category @xmath13 .
it then appears that both dynamic and denotational models can be defined in categories of triskells over any topos instead of a category of sets .
|
we exhibit a new relationship between dynamic and static semantics . we define the categorical outlay needed to define interaction graphs models , a generalisation of girard s geometry of interaction models , which strongly relate to game semantics .
we then show how this category is mapped to weighted relational models of linear logic .
this brings into vision a new bridge between the dynamic and static approaches , and provides formal grounds for considering interaction graphs models as quantitative versions of goi and game semantics models .
we finally proceed to show how the interaction graphs models relate to a very general notion of quantitative coherence spaces .
|
the intermetallic compound ucoge belongs to the select group of superconducting ferromagnets @xcite . in this intriguing group of materials superconductivity
develops in the ferromagnetic state at a temperature @xmath5 well below the curie temperature @xmath6 for ferromagnetic ordering @xcite .
moreover , below @xmath5 , superconductivity and ferromagnetic order coexist on the microscopic scale .
the superconducting ferromagnets discovered so far are uge@xmath7 ( under pressure , ref . ) , urhge ( ref . ) , uir ( under pressure , ref . ) and ucoge ( ref . ) .
the co - occurrence of ferromagnetism and superconductivity is at odds with the standard bcs ( bardeen - cooper - schrieffer ) scenario for phonon - mediated spin - singlet superconductivity , since the ferromagnetic exchange field impedes spin - singlet cooper pairing @xcite . instead
, alternative models have been proposed that exploit the itinerant nature of the ferromagnetic order where critical spin fluctuations , connected to a magnetic instability , mediate an unconventional , spin - triplet type of pairing @xcite . indeed , these uranium intermetallics , where the @xmath8-electrons are delocalized , are all close to a magnetic instability that can be induced by mechanical pressure , chemical doping or an applied magnetic field @xcite . unraveling the properties of superconducting ferromagnets might help to understand how spin fluctuations can stimulate superconductivity , which is a central theme for materials families as diverse as heavy - fermion , high-@xmath5 cuprate and iron - oxypnictide superconductors .
ucoge crystallizes , just like urhge , in the orthorhombic tinisi structure with space group @xmath9 ( ref . ) .
the coexistence of superconductivity and ferromagnetism was first reported by huy _
_ @xcite .
high - quality single crystals with a typical residual resistance ratio , @xmath10 = @xmath11(300k)/@xmath11(1k ) , of 30 have a curie temperature @xmath12 k and show superconductivity with @xmath13 k. ucoge is a uniaxial ferromagnet .
the spontaneous magnetic moment , @xmath1 , points along the @xmath14-axis and attains the small value of @xmath15 per u - atom in the limit @xmath16 .
proof for the microscopic coexistence of superconductivity and ferromagnetic order is provided by @xmath17sr ( muon spin relaxation and rotation ) @xcite and @xmath18co - nqr ( nuclear quadrupole resonance ) @xcite experiments .
evidence for spin - triplet cooper pairing has been extracted from the magnitude of the upper critical field @xmath19 ( measured with the external magnetic field directed perpendicular to @xmath1 ) , which greatly exceeds the pauli limit for spin - singlet superconductivity @xcite .
the important role of spin - fluctuations in promoting superconductivity is established by the large anisotropy of the upper critical field , @xmath20 ( refs . ) .
for @xmath21 the magnetic transition becomes a cross - over , spin fluctuations are rapidly quenched and , accordingly , superconductivity is suppressed , while for @xmath22 spin fluctuations become more pronounced and superconductivity is enhanced . at the microscopic level ,
the close link between anisotropic critical magnetic fluctuations and superconductivity was recently put on a firm footing by @xmath18co - nmr ( nuclear magnetic resonance ) @xcite and inelastic neutron scattering @xcite .
yet another salient property of ucoge is the unusual s - shaped curvature of the upper critical field for a field direction along the @xmath4-axis , @xmath23 , which yields the large value of @xmath24 t when @xmath16 ( ref . ) .
this field - reinforced superconductivity seems to be closely connected to a field - induced quantum critical point as a result of the progressive depression of the curie temperature @xcite .
the peculiar response of the magnetic and superconducting phases to a magnetic field calls for a detailed investigation of the anisotropy in the magnetic , thermal and transport properties . here
we present an extensive angle dependent magnetotransport study on high - quality single crystals of ucoge for fields directed in the @xmath25- and @xmath26-planes of the orthorhombic unit cell .
we identify a pronounced maximum in the magnetoresistance for a sample with @xmath27 when the component of the field along the @xmath14-axis reaches a value @xmath0 t. measurements of @xmath2 as a function of pressure show a rapid increase of @xmath2 to 12.8 t at 1.0 gpa .
the uniaxial nature of @xmath2 and its large pressure variation provide strong indications for a close connection to an unusual polarizability of the u and co moments .
transport measurements around the superconducting transition in fixed magnetic fields @xmath28 reveal our samples exhibit the s - shaped @xmath29-curve when properly oriented in the magnetic field . at temperatures of @xmath30 and @xmath31 k , as indicated .
the current was applied along the @xmath4-axis .
left inset : high - field magnetoresistance ( @xmath32 ) up to @xmath33 t at @xmath34 k. right inset : @xmath2 as a function of temperature determined from the maximum in the magnetoresistance.,width=302 ]
single cystals of ucoge were prepared in a tri - arc furnace by the czochralski technique as described in ref . .
bar - shaped samples with typical dimensions @xmath35
mm@xmath36 were cut from the crystals by means of spark erosion .
magnetotransport measurements were carried out on three samples with the current , @xmath37 , along the @xmath4- and @xmath14-axis , with @xmath10-values of 30 ( sample # 1 ) and 8 ( sample # 2 , sample # 3 ) , respectively .
experiments on sample # 1 were performed in a @xmath36he refrigerator ( heliox , oxford instruments ) in the temperature range 0.24 - 15 k and fields up to 14 t , and in a dilution refrigerator ( kelvinox m100 , oxford instruments ) in the temperature range 0.04 - 1.0 k and fields up to 16 t. in the kelvinox the sample was mounted on a swedish rotator for angle dependent measurements .
in addition , experiments were carried out at the high field magnet laboratory in nijmegen in a dilution refrigerator and in fields up to 33 t. sample # 2 was measured under pressure using a hybrid clamp cell attached to the cold plate of the @xmath36he refrigerator .
details of the pressure cell and calibration data are given in ref . .
sample # 3 was measured in the @xmath36he refrigerator at ambient pressure and @xmath38 k. the ac - resistivity data were acquired using a low - frequency ( @xmath39 hz ) resistance bridge or a phase sensitive detection technique using a lock - in amplifier .
care was taken to prevent heating of the sample by selecting a low excitation current ( @xmath40a ) .
k. upper panel : field rotation in the @xmath25-plane ; @xmath41 = -58 , -53 , -48 , -43 , -38 , -33 , -28 , -18 , -8 , 2 , 12 , 22 and 27 degrees , where @xmath42 corresponds to @xmath43 .
lower panel : field rotation in the @xmath26-plane ; @xmath41 = -53 , -48 , -43 , -38 , -28 , -18 , -8 , 2 , 12 , 22 and 27 degrees .
the current is always applied along the @xmath4-axis .
inset : @xmath2 as a function of @xmath41 .
the solid line represents @xmath44 .
, width=302 ]
in fig . 1 we show the resistivity of ucoge ( sample # 1 ) as a function of the magnetic field applied along the @xmath14-axis . at the lowest temperature @xmath45
k ( @xmath46 the initial steep rise signals the suppression of superconductivity at @xmath47 t. next , @xmath48 steadily increases and passes through a pronounced maximum at @xmath0 t. increasing the temperature shows that the maximum at @xmath2 is a robust property and can be identified up to at least 10 k. the temperature variation of @xmath2 is relatively weak as shown in the right inset of fig . 1
. we remark that the overall resistivity rapidly increases with temperature and in the normal phase the initial low - field magnetoresistance is negative . in the left inset
we show data taken at @xmath34 k in strong magnetic fields up to 33 t. the maximum at @xmath2 is most pronounced . for fields exceeding 12 t
the magnetoresistance displays a steady increase that leads to the large value of 40 @xmath49 cm at the maximum field . in order to investigate the magnetocrystalline anisotropy of @xmath2
, we have measured the angle dependent magnetoresistance .
the data taken in a dilution refrigerator at @xmath50 k are shown for a field rotation in the @xmath25-plane and in the @xmath26-plane in the upper and lower panels of fig . 2 , respectively .
the major experimental observation is the steady upward shift of the maximum in @xmath51 when the field is rotated away from the @xmath14-axis .
the value @xmath52 is proportional to @xmath53 , where @xmath41 is the angle at which the field is tilted from the @xmath14-axis .
this functional behavior is illustrated in the inset in fig . 2 and holds for the @xmath25- as well as for the @xmath26-plane .
for @xmath54 the maximum in @xmath48 falls outside the magnetic field range probed in the dilution refrigerator .
we remark that the value of the maximum magnetoresistance @xmath55 at @xmath2 is quasi field - angle independent .
this tells us the angle dependent magnetoresistance data may be collapsed onto a single reduced curve @xmath56 _ versus _ @xmath57 .
we conclude the maximum in @xmath51 takes place when the component of the magnetic field along the @xmath14-axis reaches @xmath0 t. this confirms its uniaxial nature , just as for the ferromagnetic order .
the suppression of superconductivity in the field - angle interval probed in fig .
2 still takes place at a low value of @xmath29 @xcite .
the pressure variation of @xmath2 was investigated for sample # 2 for @xmath58 in the pressure range 0.26 - 1.29 gpa in the @xmath36he refrigerator . in this longitudinal configuration
the field variation @xmath51 is very different as shown in fig .
this is confirmed by the longitudinal magnetoresistance of sample # 3 , measured at @xmath59 ( see right inset fig .
3 ) . after the initial steep rise , due to the suppression of superconductivity
, @xmath51 steadily decreases and shows a kink near 9.2 t rather than a maximum at ambient pressure .
the field at which the kink appears identifies @xmath2 . under pressure @xmath2 increases
rapidly ( @xmath60 ) up to 12.8 t at 1.0 gpa ( see left inset fig . 3 for @xmath2-values at @xmath61 k and 1.0 k ) .
the temperature variation is weak . at our highest pressure
( 1.29 gpa ) @xmath2 falls outside the available field range in the @xmath36he refrigerator . at pressures of 0.26 , 0.52 , 0.77 , 1.03 and 1.29 gpa as indicated
the temperature is @xmath38 k. right inset : magnetoresistance of sample # 3 for @xmath58 at ambient pressure at @xmath38 k. left inset : @xmath2 as a function of pressure at @xmath38 k ( circles ) and @xmath62 k ( squares ) .
the value @xmath63 at ambient pressure ( symbol : star ) is taken from sample # 3 .
the solid line is a fit to the data at @xmath38 k with @xmath64 where @xmath65 t / gpa@xmath66 .
, width=302 ] in fixed magnetic fields from 0 to 16 t with steps of 1 t ( from right to left).,width=302 ] the upper critical field @xmath67 for a field direction perpendicular ( @xmath68 or @xmath28 ) to the ordered moment ( @xmath69 ) is extremely sensitive to the precise orientation of the magnetic field @xcite . in order to substantiate the unusual @xmath70-behavior of our single crystals
, we measured sample # 1 as a function of field orientation in the dilution refrigerator .
special care was taken to enable field rotation in the @xmath25-plane .
after fine tuning to @xmath28 we measured the resistivity in fixed magnetic fields .
the results are shown in fig .
4 . for @xmath71
the superconducting transition sets - in at 0.6 k and has a width @xmath72 k. on applying a magnetic field , the superconducting transition progressively shifts to lower temperatures , and is still visible up to the highest field ( 16 t ) .
striking features are ( _ i _ ) the different depression rate of @xmath5 in the field range 5 - 9 t , and ( _ ii _ ) the narrowing of @xmath73 across the same field range . the upper critical field , determined by taking the midpoints of the transitions , is shown in fig . 5 .
@xmath74 has an unusual curvature for @xmath75 t and extrapolates to the large value of 17 t in the limit @xmath16 , in good agreement with the results reported in ref . .
in the inset we show the large angular variation of @xmath29 around @xmath28 measured at @xmath50 k. for a tilt - angle of typically @xmath76 the upper critical field has diminished by a factor of 3 ( ref . ) . of ucoge ( sample # 1 ) measured for @xmath28 .
inset : angular variation of @xmath3 in the @xmath25-plane at @xmath77 k.,width=302 ]
the major result from the angle dependent magnetoresistance measurements is the pronounced maximum at a field @xmath2 , which occurs when the component of the magnetic field along the @xmath14-axis reaches a value of 8.5 t. the characteristic field @xmath2 is a robust property of our samples , but the shape of the magnetoresistance @xmath78 is different for sample # 1 ( with a maximum at @xmath2 , fig . 1 ) and samples # 2,#3 ( with a kink at @xmath2 , fig . 3 ) .
we remark that there are two obvious differences between the experiments .
firstly , the sample quality is very different as quantified by the residual resistivity value @xmath79 of 10 and @xmath80 cm , respectively .
possibly , for samples # 2,#3 magnetic disorder makes a large contribution to @xmath81 , which can be reduced by the magnetic field resulting in a negative @xmath82 .
the second difference is the measurement geometry , _ i.e. _ transversal ( @xmath83 for sample # 1 ) _ versus _ longitudinal ( @xmath84 for samples # 2,#3 ) magnetoresistance , since the lorentz force on the current in general leads to more scattering and a positive @xmath82 .
moreover , in the transverse configuration the charge carriers will scatter more effectively from the magnetic moments and the predominantly longitudinal spin fluctuations , compared to the longitudinal configuration where the current and @xmath1 are aligned . these differences in sample quality and measurement geometry
might also explain the slightly different @xmath2-values extracted from fig . 1 ( @xmath85 t ) and fig . 3 ( @xmath86 t ) at ambient pressure
. the large variation of @xmath82 with the @xmath10-value and geometry is uncommon and its understanding is highly relevant in view of the strongly anisotropic magnetic properties of ucoge .
measurements in the transverse geometry with @xmath43 have not appeared in the literature so far , while longitudinal ( @xmath14-axis ) magnetoresistance data have been reported on two samples of different quality : ( _ i _ ) a magnetoresistance trace taken on a sample with @xmath27 at @xmath87 k shows a weak initially positive @xmath82 with a small structure near @xmath88 t and three additional kink - like features in the field range 17 - 30 t ( ref . ) , and ( _ ii _ ) the magnetoresistance of a sample with @xmath89 has an overall negative @xmath82 with a kink at @xmath90 or @xmath88 t ( ref . ) , as in our fig . 3 . in the latter study
the angular variation of @xmath2 , measured at @xmath87 k by tilting the field from the @xmath14-axis towards an arbitrary direction in the @xmath91-plane , was also found to follow the @xmath92-law .
an appealing scenario that has been put forward to explain the change in magnetoresistance at @xmath2 is a ferro - to - ferrimagnetic transition @xcite .
this proposal is largely based on a recent polarized neutron diffraction experiment on ucoge carried out for @xmath43 @xcite . in low magnetic field
( 3 t ) the small ordered moment @xmath1 is predominantly located at the u atom , but in a large field of 12 t a substantial moment , antiparallel to the u moment , is induced on the co site .
this unusual polarizability of the co @xmath93 orbitals may give rise to a field - induced ferri - magnetic - like spin arrangement .
support for this scenario was obtained by field - dependent ac - susceptibility data @xcite which exhibit a maximum near @xmath2 .
recently , the dc - magnetization @xmath94 was measured at @xmath95 k in pulsed magnetic fields up to 52 t @xcite .
for @xmath43 the data do not show a clear sign of a ( meta)magnetic transition , however , a weak structure appears near @xmath2 in the derivative @xmath96 , and a second change of slope occurs near 23.5 t. sensitive torque cantilever experiments might be helpful to resolve the possibly anomalous behavior of the magnetization around @xmath2 .
further arguments in favor of a magnetic transition are : ( _ i _ ) the uniaxial ( ising - type ) behavior of the ferromagnetic order is reflected in @xmath2 , and ( _ ii _ ) the pressure variation of @xmath2 ( see fig .
3 ) is large and has a magnitude comparable to the pressure dependence of @xmath6 @xcite assuming 1 k @xmath97 t per @xmath98 ( the critical pressure for the suppression of ferromagnetic order is 1.4 gpa ) . in this scenario
the pressure increase of @xmath2 may be related to the reduced polarizability of the co moment under pressure .
another possible origin of the structure in @xmath82 near @xmath2 is a lifshitz transition , _
i.e. _ a field - induced topological change of the fermi surface .
notably it has been suggested that the multitude of small kink - like features observed in @xmath82 for @xmath84 at @xmath87 k could hint at a fermi surface reconstruction @xcite .
quantum oscillations have been reported for ucoge for @xmath28 but could not be detected for a field direction along or close to the @xmath14-axis .
a second indication for the possibility of a field - induced fermi surface modification comes from thermoelectric power data @xcite , which show two pronounced peaks at 11.1 and 14.6 t for @xmath28 .
while the former peak is associated with field - reinforced superconductivity , the latter peak and the ensuing sign change of the thermopower provide evidence for a topological change of the fermi surface . in the related material
urhge the field - induced disappearance of a small fermi - surface pocket was recently demonstrated by quantum oscillations measurements @xcite .
overall , the ising - like nature of the ferromagnetic ground state results in a complex magnetotransport behavior . moreover , the magnetization @xcite , thermal expansion @xcite , thermoelectric power @xcite and thermal conductivity @xcite all have a strong magnetocrystalline anisotropy , which makes it difficult to unravel the behavior of ucoge . however , at the positive side , it is the strong anisotropy that results in longitudinal ferromagnetic fluctuations that play a major role in inducing spin - triplet superconductivity @xcite
. a greater understanding of the anisotropy is therefore likely to be important for our understanding of the superconductivity .
the unusual superconducting behavior is demonstrated by the @xmath3 curve reported in fig . 5 .
we recall the upward curvature for @xmath75 t and the large value of 17 t in the limit @xmath16 .
the sample ( # 1 ) used here comes from the same single - crystalline batch as used in our first measurements of the upper critical field ( ref . ) , where @xmath99 was found to reach a value of 5 t for @xmath28 .
this discrepancy can now be attributed to a small misorientation of @xmath100 ( see the inset in fig . 5 ) .
the precise orientation of the sample with respect to the magnetic field direction remains an absolutely crucial feature for the behavior of this material .
the field - reinforced superconductivity appears to be connected to critical spin fluctuations associated with a field - induced quantum critical point , where the latter is reached by the suppression of the curie temperature in strong magnetic fields for @xmath22 @xcite .
a second , more recently , proposed cause for the field - reinforced superconductivity is a lifshitz transition @xcite
. finally , we mention the progress made in modeling the intricate and anisotropic @xmath3 of ucoge by a strong - coupling eliasbergh model exploiting the ising - type spin fluctuations @xcite and by the completely broken symmetry scenario for parallel - spin @xmath101-wave superconductors @xcite .
in summary , we have presented an extensive angle dependent magnetoresistance study on single crystals of ucoge for fields directed in the @xmath25- and @xmath26-planes of the orthorhombic unit cell .
we pinpoint a pronounced structure in the magnetoresistance , which takes place when the component of the field along the @xmath14-axis reaches a value @xmath0 t. this behavior is very pronounced for transverse measurement geometry and rather weak for longitudinal geometry .
measurements of @xmath2 as a function of pressure show a rapid increase of @xmath2 to 12.8 t at @xmath102 gpa .
the uniaxial nature of @xmath2 and its large pressure variation are consistent with the interpretation that the change in the magnetoresistance regime at @xmath2 is related to an unusual polarizability of the u and co moments .
transport measurements in fixed magnetic fields confirm the unusual s - shaped @xmath3-behavior after carefully aligning the sample along the field @xmath28 . in order to further unravel the intriguing properties of ucoge , notably with respect to the close connection between field - induced phenomena , such as a quantum critical point or lifshitz transition , and superconductivity
, it requires an unremitted research effort to probe the strongly anisotropic thermal , magnetic and transport properties with help of high - quality single crystals .
acknowledgements @xmath103 t.v .
bay acknowledges support of the vietnamese ministry of education and training .
this work was part of the research program of fom ( dutch foundation for fundamental research of matter ) .
part of this work was carried out at the hfml - ru / fom , member of the european magnetic field laboratory ( emfl ) .
s. s. saxena , p. agarwal , k. ahilan , f. m. grosche , r. k. w. haselwimmer , m. j. steiner , e. pugh , i. r. walker , s. r. julian , p. monthoux , g. g. lonzarich , a. huxley , i. sheikin , d. braithwaite , and j. flouquet , nature ( london ) * 406 * , 587 ( 2000 ) .
|
we report a magnetoresistance study of the superconducting ferromagnet ucoge .
the data , taken on single - crystalline samples , show a pronounced structure at @xmath0 t for a field applied along the ordered moment @xmath1 .
angle dependent measurements reveal this field - induced phenomenon has an uniaxial anisotropy .
magnetoresistance measurements under pressure show a rapid increase of @xmath2 to 12.8 t at 1.0 gpa .
we discuss @xmath2 in terms of a field induced polarization change .
upper critical field measurements corroborate the unusual s - shaped @xmath3-curve for a field along the @xmath4-axis of the orthorhombic unit cell .
|
the issue of central singularity of a black hole ( bh ) is an open problem in bh physics .
although it is commonly accepted that only a not yet accessible quantum theory of gravity would be competent to solve the problem appropriately , several phenomenological scenarios have been considered in the literature in order to study bhs with regular centers ( for a review , see @xcite ) . in 1968
, bardeen @xcite introduced a compact object with an event horizon and without an intrinsic singularity , namely bardeen bh ; it is the first regular bh model in general relativity .
the bardeen spacetime is spherically symmetric without violating the weak energy condition and the inside of the horizon is desitter - like wherein the matter has a high pressure . in 2000 , ayn - beato and garca @xcite reinterpreted the bardeen model as the gravitational field of a nonlinear magnetic monopole .
a few years later , hayward @xcite investigated the formation and evaporation of a new kind of the regular solution , i.e. hayward bh , in which its static region is bardeen - like and the dynamic regions are vaidya - like .
recently , a family of rotating regular solutions have been obtained by applying the newman - janis algorithm to the hayward and to the bardeen spacetimes @xcite .
afterwards , a general class of regular solutions using a general mass term described by a function which generalizes the bardeen and hayward mass terms have been constructed @xcite .
the regular bhs have extensively been studied in the recent literature ( see for instance , @xcite ) . besides , the authors of refs .
@xcite have utilized a spherically symmetric matter distribution leading to no curvature singularity .
in fact , they have presented a physically inspired type of noncommutativity corrections to bh solutions . in this method ,
the point - like structure of mass , instead of being totally localized at a point , is described by a smeared structure throughout a region of linear size @xmath3 . in other words , the mass density of a static
, spherically symmetric , particle - like gravitational source can not be a delta function distribution , but will be given by a gaussian distribution @xmath4 .
it has been demonstrated that the modified metric does not allow the bh to decay below the planckian relic .
the evaporation process ends when the size of the bh reaches a planck - sized remnant , explaining the bh released from the curvature singularity at the origin .
here , the regularity of the metrics emerges from the appearance of a minimal length preparing a natural cut - off at small scales .
the idea of a minimal length is confirmed by many outcomes of various approaches to quantum gravity @xcite .
this universal cut - off is entered into the energy - momentum tensor of the einstein equations , and stands for the degree of delocalization of the matter distribution @xcite . from the other viewpoint , all the various arguments concerning the so - called hawking information loss paradox rely on semi - classical methods and guess about the behavior of systems in the quantum gravitational regime , but there are substantial struggles over the success of the arguments .
the basis of the information loss problem turns back to hawking s discovery that the theory of quantum fields in a curved spacetime indicates that bhs will emit thermal radiation at a temperature inversely proportional to their mass @xcite .
conservation of energy points out that the bh will lose mass through this procedure , and if nothing stops the evaporation of the bh will ultimately cease to exist .
this proposition of total evaporation is necessary for hawking s argument , and may be refused by a remnant proposal that we will be considering in this paper ( for reviews on resolving the paradox , see @xcite ) . in other words ,
planck scale physics may terminate the hawking radiation and prohibit the appearance of a singularity in the center of a bh , i.e. the appearance of a bh remnant ; for example , in a recent result by paul and majhi @xcite , the nature of the cascade of hawking emission spectrum in the presence of a back reaction was studied .
they observed that under a physical background , below a particular value of the mass , which is of the plank mass order , the hawking radiation must stop wherein a remnant is formed .
in this work we shall include the influences of inspired noncommutativity to the one of the most popular models of non - singular bhs , i.e. the bardeen bh , and analyze the remnants of noncommutative bardeen ( nb ) bhs .
afterwards , we consider the thermodynamic properties of the nb solution , providing its hawking temperature . throughout this paper natural units are used with the following definitions : @xmath5 .
in this section we first include the noncommutative effects in the line element of bardeen and then analyze the consequences of the resulting metric .
now , let us start from the schwarzschild - like class of metrics which describe the spacetimes in the so - called kerr - schild form and in the presence of matter @xmath6 where @xmath7 is the minkowski line element expressed in a spherical basis and @xmath8 is a null vector in minkowski coordinates .
the expression @xmath9 is found to be @xmath10 according to the kerr - schild decomposition , the above equation has a general validity , so its generic structure is kept and it is not sensitive to different forms of the mass term @xmath11 . for the bardeen metric we have @xmath12 where @xmath0 is the magnetic charge of the bh .
now , using the noncommutativity approach @xcite , the metric describing the noncommutative geometry inspired bardeen bh is given by @xmath13 where @xmath11 can now be written in terms of the smeared mass distribution @xmath14 as follows : @xmath15 and @xmath14 can implicity be given in terms of the lower incomplete gamma function , @xmath16 the radiating behavior of such a noncommutative regular bh can now be easily investigated by plotting the temporal component of the metric , @xmath17 , versus the radius @xmath18 for an extremal bh with different values of @xmath0 ( see fig .
( [ fig:1 ] ) ) . the plot presented in fig .
[ fig:1 ] shows , for several values of minimal nonzero mass @xmath1 , the possibility of having an extremal configuration with one degenerate event horizon as the parameter @xmath0 grows .
as this figure shows , the coordinate noncommutativity leads to the existence of a remnant mass in which the nbbh can shrink to .
the line element ( [ mat:4 ] ) has a coordinate singularity at the event horizon radius , @xmath19 , that can be obtained from the equation @xmath20 as follows @xmath21 with @xmath22 the analytical solution of eq .
( [ mat:7 ] ) for @xmath23 in a closed form is impossible , but it is possible to solve it to find @xmath24 , which provides the mass of the nbbh as a function of the horizon radius @xmath19 .
this leads to @xmath25},\ ] ] where the gauss error function @xmath26 is defined by @xmath27 .
the results of the numerical solution of the mass as a function of the horizon radius are displayed in fig .
[ fig:2 ] .
as expected , from the mass equation ( [ mat:9 ] ) , the noncommutativity indicates a minimal nonzero mass in order to have an event horizon .
so , in the noncommutative case , for @xmath28 there is no event horizon . for more details ,
the numerical results for the remnant size of the bh and for different values of @xmath29 are presented in table [ tab:1 ] which are comparable to fig .
[ fig:2 ] .
according to table [ tab:1 ] , as @xmath0 increases both the minimal nonzero mass and the minimal nonzero horizon radius are enlarged which subsequently lead us to the final result : @xmath30 this means in the limit @xmath31 , the magnetic charge is proportional to the remnant mass and to the remnant radius . c|c|c + magnetic charge & remnant mass & remnant radius + @xmath32 & @xmath33 & @xmath34 + @xmath35 & @xmath36 & @xmath37 + @xmath38 & @xmath39 & @xmath40 + @xmath41 & @xmath42 & @xmath43 + @xmath44 & @xmath45 & @xmath46 + @xmath47 & @xmath48 & @xmath49 + @xmath50 & @xmath51 & @xmath52 + @xmath53 & @xmath54 & @xmath55 + & & + + [ tab:1 ] here , it should be emphasized that the physical interpretation of the noncommutative parameter @xmath56 is the smallest fundamental cell of an observable area in noncommutative geometry , in the same way that the planck constant @xmath57 explains the smallest fundamental cell of an observable phase space in quantum mechanics .
the scale of @xmath3 is , possibly and most reasonably , of the order of an inverse characteristic energy of the planck scale .
most of the phenomenological examinations of the noncommutativity models have assumed that the noncommutative energy scale can not lie far above the tev regime @xcite .
given that the fundamental planck scale in models with large extra dimensions is as small as a tev , for solving the hierarchy problem @xcite , therefore it could be possible to set the noncommutative effects in a tev energy scale . in this process , the minimum value of the nbbh mass increases to a value more than its value for the noncommutative schwarzschild one ( see table [ tab:1 ] ) .
thus , in the theory of regular bhs , if the parameter @xmath0 becomes sufficiently large , i.e. @xmath58 with a sufficiently small noncommutative inverse length parameter ( @xmath59 tev ) , then the noncommutativity effect can concretely decrease the possible formation and detection of bhs in tev - scale collisions at particle colliders , such as the large hadron collider ( lhc ) , where @xmath60 is the parton - parton center - of - mass energy which is proportional to the tev energy scale . ] and the ultra - high energy cosmic ray ( uhecr ) .
however , if the fundamental planck scale is of the order of a few tev , then the lhc may produce bhs .
these bhs may have masses on the order of tev . in this sense , on the other hand , the complete decay of bhs is impossible and the final planck - sized remnant can be thought of as the order of tev .
based on our computations , the total evaporation of the bh is not possible in principle .
therefore , the idea of a stable bh remnant as a candidate to conserve information has fixed .
note that , currently there are some proposals about what happens to the information that falls into a bh .
one of the main proposals is that the bh never disappears completely , and the information is not lost , but would be stored in a stable remnant .
a remnant proposal reflects the fact that the semi - classical approaches used to derive the hawking effect are obviously inapplicable when the bh reaches the planck mass .
possibly , planckian physics will propose some way of conserving the information held in the bh .
the remnant proposals are generally divided into two categories .
the first is stable remnants and the second is long - lived remnants .
if a remnant is stable , then planck scale quantum gravitational effects shut down the hawking radiation , and the bh remnant continues to exist for all future time .
inasmuch as the bh , and its information , are prohibited from vanishing totally , the troubling result of hawking s argument , i.e. the non - unitary evolution is prevented .
for example , in ref .
@xcite , we have shown that , as a well - known result of the spacetime noncommutativity , a part of information may be preserved in a stable bh remnant . on the other hand ,
the process for long - lived remnants is completely different .
they eventually disappear . in spite of the fact that the semi - classical models imply that no information can escape in the hawking radiation
, these models will fail in planckian processes .
this scenario is inspired by this cognition that the physics at the planck scale might return the information to the external universe , and once the coherence of the external universe is made safe , the remnant could safely disappear . as an example of long - lived kinds of remnants
, one can point to ref .
@xcite in which an exact @xmath61 dependent case of a noncommutative schwarzschild - like metric for a vaidya solution was calculated . as an important result of ref .
@xcite , the idea of a stable bh remnant as a candidate to conserve information has been failed which means that if we pick up a time - dependent gaussian distribution of mass / energy , then it will be possible to find a zero remnant mass , albeit in a long - time limit
. one of the serious problems with these remnants is the probability of their detection .
given that the interactions of bh remnants are purely gravitational , the cross - section is highly small , and a direct observation of these remnants seems impossible .
a possible indirect evidence might be related to the cosmic gravitational wave background .
in contrast to photons , the gravitons radiated during evaporation would be instantly frozen .
since , the bh evaporation finishes when it reduces to a remnant , hence , the graviton spectrum should have a cut - off at the planck mass .
in general , such a cut - off is expected to have a redshift on the order of @xmath62 gev .
moreover , we know that the nature of dark matter is hitherto remained an open problem . there exist many dark matter candidates in which most of them are non - baryonic weakly interacting massive particles .
a candidate which is not closely connected to particle physics is the relics of primordial bhs @xcite .
some specific inflation models naturally induce a great number of such bhs , e.g. the hybrid inflation model can generally produce a required abundance of primordial bh remnants for them to be the main source of the dark matter @xcite .
while we do not yet have any credible candidate for a full quantum gravity theory , more phenomenological procedures have tried to investigate micro bhs .
the recent anticipations imply the conceivable results of lhc experiments , containing the creation of these objects .
the possible experimental production of bhs at particle colliders is one of the most significant subfields in extra dimension models .
these newly formed miniature bhs first lose their hairs associated with the multipole and the angular momenta , then classically reach the stable schwarzschild solutions , and finally evaporate via hawking radiation up to promising planck - sized remnants .
the hawking temperature is generally subjected to corrections from many sources , particularly , those related to a bh with the mass of the order of the planck mass .
hence , the study of tev - scale bhs in the uhecr and particle colliders requires a perfect examination of how temperature corrections affect bh thermodynamics . in the following , with the above motivation , we would like to find the temperature corrections of the nb solution .
when the nbbh radiates , its hawking temperature can be calculated to find @xmath63.\end{aligned}\ ] ] for large bhs , @xmath64 , and @xmath32 , the gauss error function tends to unity and the exponential term is reduced to zero .
thus , one recovers the standard result for the temperature of a schwarzschild bh , i.e. @xmath65 . at this stage , the numerical result of the hawking temperature versus the horizon radius is presented in fig .
[ fig:3 ] .
this figure shows that the temperature peak drops with increasing the parameter @xmath0 .
therefore , we expect that the nbbhs to be colder than the noncommutative schwarzschild bhs . in addition , the size and the mass of the nbbh remnant at the final stage of the evaporation increase with increasing the magnetic charge . hence , the remnant of a nbbh may be big compared to the noncommutative schwarzschild case . as an important note ,
if we had chosen the hayward solution , as another popular example of regular bhs , solely the mass term would have altered , however the general properties would have directed to entirely comparable consequences to those above .
now , let us consider the metric describing the noncommutative hayward ( nh ) bh , in schwarzschild coordinates , which is immediately given by eq .
( [ mat:4 ] ) with this new mass term @xmath66 where @xmath67 is a positive constant measuring the deviations from the classical kerr metric .
the lack of responsiveness of the results to the kind of the regular bh can be easily exhibited by plotting the hawking temperature as a function of radius for nhbhs ( see fig .
[ fig:4 ] ) . comparing these results with the results of fig .
[ fig:3 ] shows the close similarity of outcomes in these two types of regular bhs . at this point
it is worth pointing out that most of the results in the noncommutative framework are confirmed by the so - called generalized uncertainty principle ( gup ) context @xcite .
it is widely accepted that the heisenberg uncertainty principle should be reformulated owing to the noncommutative nature of spacetime at the planck scale .
the application of the gup to bh thermodynamics has attracted considerable attention in the literature which leads to significant modifications to the emission process , particularly at the final stages of the bh evaporation ( there is a large body of literature on this subject ; see for example , @xcite ) . as a result of gup effects on this issue
, we have shown that a modification of the de broglie relation and corresponding commutation relations in the quantum tunneling framework of the bh evaporation lead to correlations between emitted modes of evaporation @xcite . in this setup , information leaks out of the bh in the form of non - thermal gup correlations and , on the other hand , the inclusion of quantum gravity effects as the gup expression can halt the evaporation process , so that a stable bh remnant is left behind , including a part of the bh information content ( see also @xcite ) .
in addition , recently , the authors of ref .
@xcite investigated the gup effect on the thermodynamics of a schwarzschild - tangherlini bh and found that the gup corrected hawking temperature is smaller than the original case ; it goes to zero when the mass of the bh reaches a minimal value , which is supported by the results obtained in the framework of the inspired noncommutativity .
also , the results of their study concerning the possibilities to observe a micro bh in the lhc have shown that the minimum energy for the production of the bh is larger than the current energy scales of lhc . as a final remark ,
our results are supported by the results obtained in the framework of gravity s rainbow @xcite .
ali and his coworkers @xcite have argued that since a remnant depends critically on the structure of the rainbow functions , therefore a remnant is formed for all black objects in the context of gravity s rainbow and this is a model - independent phenomenon .
their calculations have shown that the behavior of hawking s radiation changes considerably near the planck scale in gravity s rainbow such that black objects do not evaporate completely and a remnant is left . moreover
, they have found that the mass of their remnant is greater than the energy scale at which experiments were carried out at the lhc @xcite .
in summary , we have applied the noncommutativity effects to bardeen bhs .
the noncommutative effects become susceptible when the mass of the bh reaches the order of the planck scale , it stops radiating and leads to a bh remnant .
it is concluded that , for an adequately large magnetic charge of a nbbh there is a linear relationship between the remnant mass and the remnant radius that is just the same as appeared in the relation between the horizon radius and the bh mass for the standard schwarzschild case .
we have found that the temperature peak of the nbbh decreases as the parameter @xmath0 increases .
thus , a nbbh is colder than a noncommutative schwarzschild bh . in this setup ,
the final stage of the evaporation of a noncommutative regular bh is a remnant in which it has an increasing size with raising its own characteristic parameter . as a consequence , in the theory of noncommutative regular bhs ,
the minimum value of energy for the production of such a bh at the current energy scales of lhc is larger , so the possibility for its detection is less .
financial support by lahijan branch , islamic azad university grant no . 17.20.5.3517 is gratefully acknowledged .
the authors thank to c. bambi for valuable suggestions .
+ ansoldi , s. 2008 , arxiv:0802.0330 bardeen , j. m. , 1968 , in conference proceedings of gr5 , tbilisi , ussr , p.174 ayn - beato , e. , & garca , a. 2000 , phys .
b , 493 , 149 hayward , s. a. 2006 , , 96 , 031103 bambi , c. , & modesto , l. 2013 , phys .
b , 721 , 329 neves , j. c. s. , & saa , a. 2014 , phys .
b , 734 , 44 azreg - anou , m. 2014 , , 90 , 064041 frolov , v. p. 2014 , jhep , 2014 , 49 amir , m. , & ghosh , s. g. 2015 , arxiv:1503.08553 lorenzo , t. d. , giusti , a. , & speziale , s. 2016 , gen .
grav . , 48 , 31 ghosh , s. g. , & amir , m. 2015 , eur
j. c , 75 , 553 garcia , a. , et al .
2015 , j. math .
phys . , 56 , 032501 toshmatov , b. , et al .
2014 , , 89 , 104017 ghosh , s. g. 2015 , eur . phys . j. c , 75 , 532 stuchlik , z. , & schee , j. 2015 , int . j. mod .
d , 24 , 1550020 nicolini , p. 2005
, j. phys .
a , 38 , l631 nicolini , p. , smailagic , a. , & spallucci , e. 2006 , phys .
b , 632 , 547 ansoldi , s. , et al .
2007 , phys .
b , 645 , 261 nozari , k. , & mehdipour , s. h. 2008 , class .
grav . , 25 , 175015 nicolini , p. 2009 , int . j. mod . phys . a , 24 , 1229 snyder , h. s. 1947 , phys .
, 71 , 38 dewitt , b. s. 1964 , , 13 , 114 yoneya , t. 1976 , prog .
phys . , 56 , 1310 rovelli , c. 1993 , nucl .
b , 405 , 797 fontanini , m. , spallucci , e. , & padmanabhan , t. 2006 , phys .
b , 633 , 627 smailagic , a. , & spallucci , e. 2003 , j. phys . a , 36 , l467 smailagic , a. , & spallucci , e. 2003 , j. phys .
a , 36 , l517 smailagic , a. , & spallucci , e. 2004 , j. phys .
a , 37 , 7169 hawking , s. w. 1975 , comm .
43 , 199 preskill , j. 1992 , arxiv : hep - th/9209058 page , d. n. 1993 , , 71 , 3743 stephens , c. r. , t hooft , g. , & whiting , b. f. 1994 , class .
grav . , 11 , 621 russo , j. g. 2005 , arxiv : hep - th/0501132 chen , p. , ong , y. c. , & yeom , d .- h .
2015 , phys .
, 603 , 1 dymnikova , i. , & khlopov , m. 2015 , int . j. mod . phys .
d , 24 , 1545002 banks , t. , oloughlin , m. , & strominger , a. 1993 , , 47 , 4476 paul , a. , & majhi , b. r. 2016 , arxiv:1601.07310 hinchliffe , i. , kersting , n. , & ma , y. l. 2004 , int . j. mod .
a , 19 , 179 rizzo , t. g. 2006 , jhep , 0609 , 021 casadio , r. , & nicolini , p. 2008 , jhep , 0811 , 072 spallucci , e. , smailagic , a. , & nicolini , p. 2009 , phys . lett .
b , 670 , 449 antoniadis , i. 1990 , phys .
b , 246 , 377 witten , e. 1996 , nucl .
b , 471 , 135 arkani - hamed , n. , dimopoulos , s. , & dvali , g. r. 1998 , phys .
b , 429 , 263 randall , l. , & sundrum , r. 1999 , , 83 , 3370 nozari , k. , & mehdipour , s. h. 2009 , jhep , 03 , 061 mehdipour , s. h. 2010 , , 81 , 124049 zeldovich , y. b. , & novikov , i. d. 1966 , sov .
, 10 , 602 linde , a. 2000 , phys .
, 333 , 575 amati , d. , ciafaloni , m. , & veneziano , g. 1989 , phys .
b , 216 , 41 konishi , k. , paffuti , g. , & provero , p. 1990 , phys .
b , 234 , 276 maggiore , m. 1994 , , 49 , 5182 kempf , a. , mangano , g. , & mann , r. b. 1995 , , 52 , 1108 scardigli , f. 1999 , phys .
b , 452 , 39 adler , r. j. , chen , p. , & santiago , d. i. 2001 , gen .
, 33 , 2101 medved , a. j. m. , & vagenas , e. c. 2004 , , 70 , 124021 bolen , b. , & cavaglia , m. 2005 , gen .
grav . , 37 , 1255 nozari , k. , & mehdipour , s. h. 2006 , int . j. mod
a , 21 , 4979 nozari , k. , & mehdipour , s. h. 2008 , europhys .
, 84 , 20008 mehdipour , s. h. 2009 , int
a , 24 , 5669 mehdipour , s. h. 2012 , can .
j. phys . ,
90 , 425 feng , z. w. , et al .
2015 , arxiv:1512.09219 ali , a. f. 2014 , , 89 , 104040 ali , a. f. , faizal , m. , & khalil , m. m. 2014 , jhep , 12 , 159 ali , a. f. , faizal , m. , & khalil , m. m. 2015 , nucl . phys .
b , 894 , 341 ali , a. f. , faizal , m. , & khalil , m. m. 2015 , phys .
b , 743 , 295
|
we derive the mass term of the bardeen metric in the presence of a noncommutative geometry induced minimal length . in this setup , the proposal of a stable black hole remnant as a candidate to store information is confirmed .
we consider the possibility of having an extremal configuration with one degenerate event horizon and compare different sizes of black hole remnants . as a result , once the magnetic charge @xmath0 of the noncommutative bardeen solution becomes larger , both the minimal nonzero mass @xmath1 and the minimal nonzero horizon radius @xmath2 get larger .
this means , subsequently , under the condition of an adequate amount of @xmath0 , the three parameters @xmath0 , @xmath1 , and @xmath2 are in a connection with each other linearly . according to our results ,
a noncommutative bardeen black hole is colder than the noncommutative schwarzschild black hole and its remnant is bigger , so the minimum required energy for the formation of such a black hole at particle colliders will be larger .
we also find a closely similar result for the hayward solution .
|
generating initial conditions for numerical experiments of a system with a certain density profile is a well - defined but difficult procedure .
there are two steps for constructing the initial conditions of one desired model : 1 ) calculate the steady - state distribution function of the desired model ; 2 ) use monte carlo method to generate the initial conditions .
the main difficulty comes from the first step .
one possibility is to make some simplifying approximations about the nature of the distribution function ( @xcite ; @xcite ) .
hernquist ( 1993 ) described a prescription for constructing n - body realization assuming that velocity distribution at any given point is maxwellian .
it is called the local maxwellian approximation ( @xcite ) , of which the advantage is easy to implement .
kazantzidis , magorrian & moore ( 2004 , kmm hereafter ) show that the halo initialized using the local maxwellian approximation ( the maxwellian halo hereafter for simplicity ) can be significantly far from equilibrium , and that for high resolution simulations it is more reasonable to use a stable system generated with a certain steady - state distribution function ( the df halo hereafter ) .
in particular , they demonstrated that an unstable cuspy halo constructed using the local maxwellian approximation will soon lose its cusp structure due to its own evolution , which is previously claimed to be due to minor merger between a host halo and the falling sub - system . on the contrary ,
a stable cuspy halo survives the minor merger .
this investigation indicates that the non - equilibrium maxwellian halo will surely lead to spurious evolution of the host halo when dealing with the response , e.g. the responding of a host halo to sinking sscs will otherwise be mixed with the artificial evolution of the host halo , induced by its unstability .
this effect has been investigated in our recent work on the bulge formation in late - type spirals ( @xcite ) . the halos on which kmm studied are those with cuspy structure .
they have not explored the halos with flat inner profile , e.g. the burkert profile .
the progress in observations ( e.g. @xcite ; and the references therein ) shows , however , that the cuspy dm profiles , such as the nfw ( @xcite ) profile , are not adequate for a large fraction of dwarf galaxies , which are dominated by dm halos , and suggests that the core density profiles , the burkert ( 1995 ) profile for example , are more suitable to these galaxies .
it would then be very much instructive to extend kmm s study to the burkert halo .
it is the halo that contains less mass in the central region than those in the cuspy halos , e.g. the nfw profile .
it follows that the technical difficulties will arise for high - resolution simulations of the core halos , especially for those studies focused on the central regions ( @xcite ; @xcite ) . for a given mass of dm halo ,
our test shows that in order to reach the same resolution in the central 10@xmath0 region the number of particle for the burkert halo should be a hundred times more than that for the nfw halo .
it would need much more computing time with the n - body simulations . in this paper , we aim at describing the way to generate the initial conditions for an isotropic , spherical burkert halo , and perform simulations using the monte carlo method ( henon 1971a , b ) to study the stable situation of the maxwellian and the df burkert halos .
this may have important implications for the dynamical evolution of sscs in a responding dark matter halo , resulting in different bulge formation history in late - type spirals .
the organization of the paper is as follows .
the procedure to constructing the initial conditions for our model is described in section 2 .
section 3 gives the results of the simulations .
we discuss the results and illustrate the important implications of generating the df halo in section 4 .
finally , section 5 notes our conclusion .
in this paper , we restrict our models to isotropic , non - rotating halos . according to the jeans@xmath18 theorem ( @xcite ; @xcite )
, the distribution function of any steady - state spherical system can be expressed as @xmath19 , where e is the binding energy and @xmath20 is the angular momentum vector . as usual , the relative potential @xmath21 and the relative energy @xmath22 ( a dimensionless energy in units of @xmath23 ) are defined as : @xmath24 where @xmath25 is the potential and @xmath26 is a constant fulfilling @xmath27 for @xmath28 and @xmath29 for @xmath30 . in an isotropic spherical model the distribution function depends on @xmath22 only , that is , @xmath31 . integrating @xmath32 over all velocities
, we can get the density profile : @xmath33 the inversion of the above equation gives the distribution function ( @xcite ; @xcite ) , @xmath34\ ] ] the @xmath35 factor can be evaluated from eq.(1 ) and eq.(2 ) .
the second term of the right - hand side vanishes in large distance in our model .
here we describe two approaches to initialize the numerical simulations for the burkert profile using the local maxwellian approximation and the steady - state distribution function , respectively .
having the density profile @xmath36 , we can calculate the model s cumulative mass distribution m(r ) and the gravitational potential @xmath37
. then we can calculate the distribution function from eq.(6 ) .
the integrand in eq.(6 ) diverges at one or both of the limits , but this can be solved using standard techniques for improper integral ( @xcite ) .
the distribution function is obtained on a grid of @xmath22 equally spaced in @xmath38 .
the accuracy of the numerical integration is checked by comparing the density profile derived from eq.(5 ) and that given by eq.(1 ) .
the two profiles are very well accordant .
we find that the distribution function is non - negative everywhere , which proves that the assumption of the isotropic velocity distribution for the spherical burkert model is reasonable and physical .
the procedure to obtain the physical quantities of the sampling particles is the following .
we randomly sample the initial positions and velocities of the particles . for our model ,
the probability density function ( windrow 2000 ) of the particle having the relative energy @xmath22 at the radius r is : @xmath39 we use the rejection method ( @xcite ) to sample the particles in two steps : 1 ) generate the position of the particle .
the probability density function with radius r is : @xmath40 2 ) generate the energy of the particle . once given r , we can get the relative potential @xmath41 . then the probability density function having
the relative energy @xmath22 and potential @xmath41 is : @xmath42 once the position r and the relative energy @xmath22 of a particle are given , the speed v can be easily determined by @xmath22=@xmath43+@xmath21 .
finally , the tangential and radial velocities of each particle can be randomly sampled considering that our model is isotropic .
the way to sample the position of a particle is the same as that depicted above .
the difference is how to generate the velocity of the particle .
the collisionless boltzmann equation for a spherical system can be written as : @xmath44=-\rho\frac{d\phi}{dr}\ ] ] where @xmath45 , @xmath46 , and @xmath47 are the velocity dispersions in spherical coordinates ( @xcite ) . assuming the model is isotropic , it can be integrated to : @xmath48 using the known density profile , the radial dispersion can be computed as a function of radius . in the local maxwellian approximation ,
the 1d velocity distribution at a given point is approximated by a gaussian ( @xcite ) : @xmath49 the speed of a halo particle is initialized from the eq.(12 ) .
then the tangential and radial velocities are randomly sampled in the same way as that described in the above subsection .
many approaches have been used in simulations to study the galactic dynamics .
one is to give the statistical description of the system , e.g. distribution function @xmath50 .
rosenbluth et al . ( 1957 ) discussed a way to calculate @xmath32 by solving a much complicated equation , fokker - planck equation .
but it needs to make a number of arbitrary simplifications .
a commonly used alternative way is n - body simulations .
unfortunately , this approach requires much computing time , which will greatly increase with the number of particles . as we mentioned before , in the case of a core halo like the burkert one ( 1995 ) , a large number of particles are in demand to keep high resolution in the central region . in view of this
, we adopt a time - saving method , the monte carlo procedure ( henon 1971a , b ) , rather than the n - body simulations .
this procedure has been proved to reproduce the behavior of the system given by the fokker - planck equation and to be much faster than the n - body simulations .
the basic ideas of the monte carlo procedure is the following .
we can divide the gravitation filed of the system into two parts : a main smoothed - out field and a small fluctuating one . the particle moves during every time step @xmath51 t which satisfies : @xmath52 where @xmath53 and @xmath54 are the crossing time and the relaxation time , respectively .
the halo evolves accordingly .
the relaxation time is related to the crossing time ( @xcite ) : @xmath55 and the crossing time can be calculated as a function of the total mass m and the total energy @xmath22 of the system ( @xcite ) : @xmath56 where c is a numerical constant . neglecting the fluctuating field in a first approximation ,
the motion of the particle is governed by the main spherical symmetry field which changes with the density or mass distribution of the halo .
the particle will then take a plane rosette motion ( @xcite ) . in any given state
the particle is characterized by its position , energy and its angular momentum .
one could derive the statistic of the system by randomly choosing the new positions and velocities of the particles on their respective orbits . taking them as the new conditions
, we can calculate a new density ( or mass ) distribution and accordingly a new potential of the system , which will in turn determine the positions of particles at the next step .
the equilibrium state of the halo can be tested by comparing the new density ( or mass ) distribution with the initial one . due to this character ,
if we find some deviation from the initial state of a system at one step , the deviation should grow in the simulations until the system relaxes to a new stable equilibrium ( as figure [ denprovsd]b shows ) .
in other words , what we could illustrate with the monte carlo method at the present time is that the system is practically stable or unstable . for our simulations ,
the relaxation time of the system is much long and we need not consider the two - body relaxation in our study . if the system is in stable equilibrium , the statistic of the system such as the potential field ( or the density and mass distribution ) will keep the same though the positions and velocities of the particles are randomly chosen in each step . on the contrary , in an unstable system , from the density or mass profile we can know that the statistic of the system have changed after one step .
the parameters of our simulations are given in table 1 .
clclclclcl model & @xmath57 & @xmath2 & @xmath4 & @xmath8 & @xmath15 & n & initializing method + & @xmath58 & @xmath3 & kpc & kpc & kpc & + a & @xmath59 & 0.05 & 4.0 & 91.24 & 118.2 & 2 @xmath60 & exact distribution function + b & @xmath59 & 0.05 & 4.0 & 91.24 & 118.2 & 2 @xmath60 & local maxwellian approximation +
( 10.8cm,9.3cm)fig1.eps we run our models for 20 steps and generate 20 individual snapshots .
the radial density profiles shown in this section are obtained by binning the particles from individual snapshot .
figure [ denprovsd]a ) illustrate the radial density profiles for model a , where the density distributions at three different steps , denoted with dotted , dashed and dot - dashed lines , are plotted . for comparison , the ( initial ) burkert profile given by eq.(1 )
is also shown in the figure , denoted with thin solid line .
obviously the radial density profiles for model a have no evolution over the simulations , all of them are indeed very well matched to the ( initial ) burkert profile . in view of this , we conclude that the df burkert halo , i.e. the burkert halo initialized with the exact distribution function , is in stable equilibrium .
the opposite situation occurs for model b. we present the results in figure [ denprovsd]b ) , where the density distributions derived at three randomly choosing steps are indicated .
the profiles denoted with dotted , dashed and dot - dashed lines illustrate significant deviation from the ( initial ) burkert profile at the central and outer regions .
the central distribution tends to steepen as clearly shown in the insert diagram of figure [ denprovsd]b ) and the outer densities decline .
the disparity from the initial one simply shows that the maxwellian burkert halo , i.e. the burkert halo generated with the local maxwellian approximation is not in equilibrium .
following kmm , we analyze the actual velocity structure at various distances for model a , compared with the corresponding gaussian distributions used in the maxwellian approximation ( model b ) .
the solid and dotted lines in figure [ veldisvsdiscen ] correspond to the true and gaussian velocity distributions , respectively , at four different distances from the center .
it is evident that only at about the scale radius the gaussian distribution is a close approximation to the reality .
the disparity in the velocity structure from the gaussian distribution is very strong near the center .
the true velocity distributions there are strongly peaked than a gaussian .
it is still apparent at far from the center , where the true velocity structure is shallower than a gaussian on the contrary .
the feature in velocity distributions described above is indeed reflected in the maxwellian burkert halo , where the deviation from the ( initial ) burkert halo is obvious at the central and outer regions .
the calculations show that the total energy and angular momentum of particles in the maxwellian burkert halo are less than those in the df burkert halo , i.e. the ( initial ) burkert halo .
that is , the maxwellian burkert halo is colder than the df burkert halo .
this explains the hoist of the central density profile in the maxwellian halo .
in other words , the unstable situation of the maxwellian halo is closely related to taking a gaussian as the velocity distribution at any given point . in some existing n - body simulations , the considered system with a predetermined mass density profile
are initialized using the maxwellian approximation . in order to start simulations with a system in stable or quasi - stable equilibrium ,
the authors allow the constructed system to relax for some time so that a ceratin quasi - stable equilibrium is attained . according to kmm s investigation and the study presented in this work , however
, the simulations should start with a system not having the predetermined mass density profile .
that is to say , spurious `` evolution '' of the system initialized with the maxwellian approximation still can not be avoided in their way .
( 7.8cm,7.3cm)fig2.eps here we would like to further demonstrate the significance of initializing a system with the distribution function by exploring the dynamical evolution of a compact super star cluster ( ssc hereafter ) in a dm dominated galaxy , depicted with the burkert profile .
the term `` compact '' here means no tidal stripping is effected . in this case , the most important process involved is the dynamical friction the compact ssc experiences , which reads as ( @xcite ) @xmath61 where @xmath62 m and @xmath63 ( with @xmath64 ) are , respectively , the mass and velocity of the ssc experiencing the dynamical friction .
the mass of the dm particle , @xmath65 , is much smaller than m. the quantity @xmath66 is the maximum impact parameter and @xmath67 a typical speed .
neither @xmath66 nor @xmath67 is precisely defined . following binney & tremaine ( 1987 ) , we take @xmath68 kpc and @xmath69 .
if the dm has a maxwellian velocity distribution with dispersion @xmath70 , the dynamical friction can be written as : @xmath71\mathbf{v_{m}}\ ] ] where erf is the error function , and @xmath72 the velocity dispersion @xmath70 can be evaluated from the jeans equation .
eq.(16 ) and eq.(18 ) are , respectively , the basis for considering the dynamical friction in the df burkert halo and the maxwellian burkert one . the initial ssc is assumed to move at a radius r=1kpc with the local circular velocity .
the sinking history of the ssc is presented in figure [ trajto ] a ) and b ) for ssc with different masses .
the thick solid lines in both panels express the case in model a , i.e. in the burkert halo initialized using the distribution function , while the thick dotted lines the case in model b , i.e. in the maxwellian burkert halo .
we can see from figure [ trajto]a ) that the ssc with typical mass of @xmath73 ( @xcite ) sinks quickly , in about @xmath74 , to the central region within a hundred @xmath0 in model a , while takes much longer time in model b. it means that the ssc in model a experiences much stronger dynamical friction and loses angular momentum quickly at first .
this situation is consistent with the velocity distributions that we discussed above .
in fact , in an isotropic halo , the effect of the dynamical friction comes from the dm particles whose velocities are less than that of the ssc ( @xcite ) . as we illustrated in figure [ veldisvsdiscen ] , the true velocity structure in the inner part of the halo is more peaked than a gaussian
. thus more particles with low velocities exist , resulting in stronger dynamical friction on ssc . inside
200 @xmath0 from the center , the ssc in model a sinks slowly than that in model b , mainly due to its low velocity and low deceleration accordingly .
( 10.8cm,8.8cm)fig3.eps a similar sinking story is presented in figure [ trajto]b ) for a ssc with mass of @xmath75 , where we can find same trend of faster sinking in model a than in model b. the interesting thing is that the ssc with smaller mass reaches the center , less than 10 @xmath0 , in shorter period of time , as compared with the corresponding situation in model a shown in figure [ trajto]a ) .
the compact ssc with larger mass experiences stronger dynamical friction , resulting in stronger deceleration but longer sinking time . as we know
, when the dynamical friction exists the periodic motion of a test particle in a spherical potential is damped like that of a damped simple harmonic oscillator . if the dynamical friction is strong enough , it is possible for the motion to lose completely its oscillation behavior , the so - called over - damping . for an over - damped ssc ,
the energy of ssc is in the form of potential energy in most of time , which makes the energy loss by dynamical friction much less effective than otherwise . for an incompact ssc
, the tidal stripping must not be ignored . in that case , the peeling process will make a heavy ssc lighter , and the over - damping may not appear .
the shorter sinking time for sscs in a dm halo initialized with the steady - state distribution function should lead to higher probability of forming massive bulges in late - type galaxies ( @xcite ) , as well as of forming nuclear star clusters at young ages ( @xcite ) .
it will also bring the intermediate mass black holes , the seed black holes , within sscs to the center of a dm halo at early stage ( @xcite ) .
all of these matter are hot topics in astrophysics , which emphasizes the important implications of using the distribution function to initialize the system .
constructing appropriate and reasonable initial conditions for an isolated equilibrium system is one of the most important points to various numerical simulations related to the formation and evolution of galaxies . in this paper , we extend kmm s study on cuspy density profiles to the burkert profile . using a time - saving monte carlo method , we have shown clearly that the burkert halo initialized with the local maxwellian approximation tends to steepen , which is closely related to adopting a gaussian as the velocity distribution at any given points .
this spurious evolution leads to underestimating the dynamical friction on sscs moving in dark matter dominated galaxies .
this important demonstration gives us a valuable clue to clarify the embarrassed situation in our previous investigation ( @xcite ; @xcite ) , i.e. the adopted maxwellian approximation in these works causes longer sinking time for sscs so as not to form massive bulges in a few gyrs and not to form a young nuclear star cluster in the burkert halo . to initialize the dark matter halos with the steady - state distribution function
should be critical to what we are pursuing , as what we have shown in this paper .
the authors would like to thank the referee , andi burkert , for his valuable comments which improve the description of the paper .
this work is supported by nkbrsf g19990754 and nsfc 10373008 .
fu is partly supported by the nsfc 10233020 .
binney , j. , & tremaine , s. 1987 , galactic dynamics ( princeton : princeton university press ) burkert , a. 1995 , apj , 447 , l25 eddington , a.s .
1916 , mnras , 76 , 572 fu , y.n . , huang , j.h . ,
deng , z.g .
2003a , mnras , 339 , 442 fu , y. n. , huang , j. h. , & deng , z. g. 2003b , carnegie obs .
, vol . 1 : coevolution of black holes and galaxies , ed
. l. c. ho ( pasadena : carnegie observatories ) , fu , y.n . , liu , w.h . , deng , z.g .
, & huang , j.h .
2005 , to be submitted henon , m. 1971a , apss , 13 , 284 henon , m. 1971b , apss , 14 , 151 hernquist , l. 1993 , apjs , 86 , 389 hernquist , l. , & quinn , p.j .
1988 , apj , 331,682 huang , j.h . ,
deng , z.g .
2003 , pasj , 55 , l89 kazantzidis , s. , magorrian , j. , & moore , b. 2004 , apj , 601,37 ( kmm ) lynden - bell , d. 1962 , mnras , 214 , 1 marchesini , d. , donghia , e. , chincarini , g. , firmani , c. , conconi , p. , molinari , e. , & zacchei , a. 2002 , , 575 , 801 navarro , j.f . ,
frenk , c.s . & white , s.d.m .
1997 , , 490 , 493 press , w.h . , flannery , b.p .
, teukolsky , s.a . , & vetterling , w.t .
1986 , numerical recipes , cambridge university press .
rosenbluth , m.n . ,
macdonald , w.m . , & judd , d.l .
1957 , phys .
, 107 , 1 toomre , a. , & toomre , j. 1972 , apj , 178,623 windrow , l.m .
2000 , apjs , 131 , 39 van albada , t.s . 1968 , bain , 19,479
|
in this paper a prescription for generating an equilibrium spherical system depicted with cuspy density profiles is extended to a core density profile , the burkert profile , which is observationally more suitable to dwarfs . by using a time - saving monte carlo method instead of n - body simulations ,
we show that the burkert halo initialized with the distribution function that depends only on energy is in a practically stable equilibrium . the one generated using
the local maxwellian approximation is unstable , where the flat core density structure tends to steepen .
this is fundamentally different from the previous study on a halo with cuspy density profile .
the deviation of the unstable `` burkert '' from the initial burkert profile is found to be closely related to taking a gaussian as the velocity distribution at any given point .
the significance of not using the local maxwellian approximation is further demonstrated by exploring the dynamical evolution of compact super star clusters in the burkert halo . in particular
, the local maxwellian approximation will result in underestimating the dynamical friction and overestimating sinking time scale .
accordingly , it will lead to lower probability of forming massive bulges and young nuclear star clusters in bulgeless galaxies .
|
the study of small lumps of strange quark matter , called strangelets , plays an important role for research the quark gluon plasma ( qgp ) in recent relativistic heavy ion collision ( rhic ) experiments .
the reason is that although many signatures of qgp such as @xmath0 suppression , strangeness enhancement , thermal dilepton electro magnetic radiation , etc .
, have been found @xcite , but it is still ambiguous because these signatures can also be explained by hadron gas @xcite . to search
an unambiguous signature of qgp is the key for rhic experiments .
the strangelet , as was argued by greiner et .
al @xcite , is a good candidate which could serve as an unambiguous signature for the qgp .
the essential problem for detectability of strangelet in rhic experiments is to study its stability during the formation of qgp . employing mit bag model
, many authors discussed this problem @xcite .
they came to the same conclusion that the electric charge and the strangeness fraction are of vital important to the experimental searches of strangelets because these two elements affect a change of stability remarkably .
but the effect of charge on stability is different for different authors .
jaffe and his co - workers @xcite argued that the strangelet is slightly positive charged .
they considered strong and weak decay by nucleon and hyperon emission together and concluded that the stable strangelets will have a low but positive change to mass ratio .
contrary to jaffe et .
al , after considering the initial condition of possible strangelet production in rhic carefully , greiner and his co - workers @xcite argued that strangelets are most likely highly negative charged .
here we hope to emphasize that above discussions are limited in the framework of mit bag model and at zero temperature .
since the quark deconfinement phase transition can occur at high temperature or / and high density only , it is of interest to extend the investigation for the stability of strangelets to finite temperature .
this is the objective of this paper .
however , mit bag model is a permanent quark confinement model because the confined boundary condition does not change with temperature . in principle
, one can not use this model to study the phase transition of qcd directly .
the best that we can do is to employ a model which can almost reproduce the properties of strange quark matter obtained by mit bag model , but in which the quark confinement is not permanent .
the quark mass density- and temperature - dependent model ( qmdtd ) @xcite suggested by us is one of such candidates .
many years ago , a quark mass density - dependent model ( qmdd ) was suggested by fowler , raha and weiner @xcite and then it was employed by many authors to discuss the properties of strange quark matter @xcite . according to the qmdd model ,
the masses of @xmath1 quarks and strange quarks ( and the corresponding anti - quarks ) are given by @xmath2 where @xmath3 is the baryon number density , @xmath4 is the current mass of the strange quark and @xmath5 is the vacuum energy density . as was proved by ref.@xcite ,
the properties of strange matter in the qmdd model are nearly the same as those obtained in the mit bag model .
in fact , it is not surprised if one notices that the confinement mechanism of mit bag model is almost the same as that of qmdd model @xcite .
but when we employ the qmdd model to discuss the properties of strangelet , a lot of difficulties emerge @xcite .
at first , the radius of strangelet decreases as the temperature increases ; secondly , it can not mimic the correct phase diagram of qcd because the temperature @xmath6 tends to infinite when @xmath7 . to overcome these difficulties
, we suggest a qmdtdmodel @xcite . instead of a constant @xmath5 in qmdd model , we argue that @xmath8 is a function of temperature and introduced an ansatz @xcite @xmath9 , 0\leq t\leq t_c , \label{15 } \\
b(t ) & = & 0,t > t_c , \label{15 - 1}\end{aligned}\ ] ] where @xmath10 is the vacuum energy density inside the bag ( bag constant ) at zero temperature , @xmath11 is the critical temperature of quark deconfinement phase transition , and @xmath12 , @xmath13 are two adjust parameters .
since @xmath5 is zero when @xmath14 , a condition @xmath15 is imposed and only one parameter @xmath12 can be adjusted . as pointed out by ref.@xcite , in order to satisfy two physical conditions of strangelet at finite temperature , namely , ( 1 ) , the radius of the strangelet must increases when the temperature rises , ( 2 ) , the energy of the strangelet must increases when the temperature rises , the parameter @xmath12 is restricted in a small range : @xmath16 in this paper , we fix the value of two parameters @xmath12 , @xmath13 in this suitable range @xmath17 with these parameters set , we use the qmdtd model to discuss the thermodynamical properties of strangelet , especially , to investigate the stability of strangelet via strong hadron emission and weak hadronic decay .
we will study the effects of temperature , charge and strangeness fraction on the stability and hope that our study can have impact to the detectability of the strangelet .
the organization of this paper is as follows . in the following section ,
we give the formulae of thermodynamical calculations for qmdtd model .
the results of thermodynamical properties of strangelet are presented in section 3 . in section 4
, we will discuss the stability of strangelet via possible strong and weak decays .
the last section is a summary .
to calculate the dynamical and thermodynamical quantities of the strangelet , we must look for the density of states first . the density of states of a spherical cavity in which the free particles be contained can be expressed as @xmath18 where @xmath19 is the total number of particle states and can be written in terms of dimensionless variable @xmath20 as @xmath21 where @xmath22 is the radius of the bag .
the three terms of the right hand side of eq.([2 ] ) refer to the contributions of the volume , surface and curvature , respectively .
the coefficients @xmath23 , @xmath5 and @xmath24 are expected to be very slow varying functions of @xmath20 , and their expressions are model dependent . for mit bag model ,
these coefficients have been obtained by numerical calculations in our previous paper @xcite .
the volume term @xmath23 is a constant , and it has the value of @xmath25 where @xmath26 is the total degeneracies .
for example , it is the total number of spin and color degrees of freedom for a quark with flavor treated separately .
the surface term @xmath5 is @xmath27 \tan ^{-1}\left ( \frac km\right ) -\left ( \frac mk\right ) -%
\frac \pi 2\right\ } , \label{5}\ ] ] where @xmath28 is the mass of quark . equations ( [ 4 ] ) and ( [ 5 ] ) are in good agreement with the ones given by multi - reflection theory @xcite .
the curvature term @xmath24 can not be evaluated by this theory except for the two limiting case : @xmath29 and @xmath30 .
madsen proposed that @xcite @xmath31 but as was pointed out by ref.@xcite , the beat fit of numerical data for curvature term is @xmath32 now we are in the position to calculate the thermodynamical quantities of strangelets for qmdtd model .
the thermodynamical potential @xmath33 is @xmath34 where @xmath35 stands for @xmath36 ( or @xmath37 ) quarks , @xmath38 for quarks and antiquarks .
@xmath39 is the density of states for various flavor quarks , it is given by eqs.([a])-([our - c ] ) .
@xmath40 is the corresponding chemical potential ( for antiparticle @xmath41 ) .
@xmath42 is the single particle energy and @xmath43is mass for quarks and antiquarks . according to the qmdtd model @xcite , the masses of quarks are @xmath44 , 0\leq t\leq t_c , \nonumber \\ m_{u,\bar{u},d,\bar{d } } & = & 0,t\geq t_c , \label{mq } \\
m_{s,\bar{s } } & = & m_{s0}+{\frac{b_0}{3n_b}}\left [ 1-a\left ( \frac t{t_c}% \right ) + b\left ( \frac t{t_c}\right ) ^2\right ] , 0\leq t\leq t_c , \nonumber \\
m_{s,\bar{s } } & = & 0,t\geq t_c , \label{ms}\end{aligned}\ ] ] where @xmath4 is the current mass of the strange quark matter , @xmath3 is the baryon number density @xmath45 and @xmath23 is the baryon number of the strangelet , @xmath46 is the volume of the strangelet . using the standard statistical treatment , and noticing that @xmath33 is not only a function of temperature , volume and chemical potential , but also of density
, it can be proved that the total pressure @xmath47 and the total energy density @xmath48 are given by @xcite @xmath49 @xmath50 the number density of each particle can be obtained by means of @xmath51 at finite temperature , we must include the contributions of the anti - particles , therefore , the baryon number for @xmath35 quark is given by @xmath52 + 1}-\frac 1{\exp [ \beta ( \varepsilon _ i+\mu _ i)]+1}\right ) .
\label{20}\ ] ] the strangeness number @xmath53 of the strangelet reads @xmath54 the baryon number @xmath23 of the strangelet satisfies @xmath55 the electric charge @xmath56 of the strangelet is @xmath57 at finite temperature , the stability condition of strangelets for the radius reads @xmath58 where the free energy @xmath59 of strangelet is , @xmath60 @xmath61 is the total energy , and @xmath62 is the entropy . given strangeness number , baryon number and electric charge , for any strangelet , we can calculate chemical potentials for @xmath36 quarks self - consistently by the equations ( [ 21]),([22]),([23 ] ) and ( [ 13 ] ) .
then the thermodynamic potential , free energy and the stable radius of the strangelet at finite temperature can be obtained self - consistently .
the numerical calculations have been done with the parameters set : @xmath63 and the possible area for strangelet in our calculations is chosen as @xmath64 we calculate the free energy of the strangelet first .
figure 1 shows the free energy per baryon @xmath65 in a @xmath66 plane for all possible strangelets with @xmath67 and @xmath68 .
this is a set of downward protruding curves .
the corresponding curves but for @xmath69 are shown in figure 2 .
comparing these two figures , we find that the positions of dots appeared in figure 2 are lower than the corresponding positions ( with the same strangeness number @xmath53 and charge @xmath56 ) in figure 1 .
for example , for @xmath70 and @xmath71 , the free energy per baryon of the dot is @xmath72 in figure 1 , but reads @xmath73 in figure 2 . for fixed @xmath53 and @xmath56 , the free energy of the strangelet decreases when temperature increases .
this result is reasonable because the entropy increases with the temperature , the term @xmath74 in eq.([13 - 1 ] ) increases considerably and @xmath59 decreases .
we will see in the next section this result affects on the decay of the strangelet remarkably . to illustrate our result transparently , we draw the @xmath65 vs @xmath53 curves for fixed @xmath75 but different temperatures @xmath6 @xmath76 and @xmath77 in figure 3 , respectively .
we find that the minimum of these two curves are located at the same strangeness @xmath78 but with different free energies .. the whole curve for @xmath68 is located on the upper position of the curve for @xmath69 .
we also draw the @xmath65 vs @xmath53 curve for @xmath79 and @xmath69 in figure 4 . compared with figure 3 , the shape of the curve in figure 4 almost does not changes except the position of the minimum is changed to @xmath80 .
these results are similar to that one given by @xcite .
now we turn to study the charge and strangeness dependences of the radius of strangelet .
we draw @xmath65 vs @xmath22 curves with fixed @xmath67 , @xmath81 and @xmath82 but different @xmath83 and @xmath84 in figure 5 , respectively .
the stable radiuses given by eq.([13 ] ) for different strangeness are different .
figure 5 shows that the stable radius changes from @xmath85 to @xmath86 when strangeness number changes from @xmath87 to @xmath84 .
the same curves for @xmath67 , @xmath88 and @xmath69 but different @xmath89 are shown in figure 6 .
we see that the stable radius become @xmath90 for @xmath91 and @xmath92 for @xmath93 .
therefore , we come to a conclusion that the stable radius of strangelet decreases when @xmath53 increases . to illustrate the charge dependence of the stable radius , we draw @xmath65 vs @xmath94 curves with fixed @xmath67 , @xmath91 and @xmath69 but different electric charge @xmath95 and @xmath96 in figure 7 , respectively . the stable radius increases from @xmath85 to @xmath97 when the electric charge rises from @xmath98 to @xmath96 .
the same curves for @xmath67 , @xmath99 and @xmath69 but different charge @xmath100 are shown in figure 8 .
we see that the stable radius changes from @xmath86 to @xmath101 when @xmath56 decreases from @xmath98 to @xmath102 .
we find that the stable radius of strangelet increases with electric charge .
in this section , we follow the line of ref.@xcite to investigate the stability of the strangelet and extend their study to finite temperature by using the qmdtd model . as was pointed out by refs.@xcite , small clusters of strange matter are most favoured for detection . as two examples ,
hereafter we study two cases with @xmath67 and @xmath103 , respectively . instead of the binding energy at zero temperature , we calculate the free energy per baryon of the possible strangelet at finite temperature first .
figure 9 and 10 show the free energy per baryon of all possible strangelets as a function of the strangeness @xmath53 at the temperature @xmath69 but for @xmath67 and @xmath103 , respectively .
the solid lines in these two figures connect the masses of the nucleon , @xmath104 and @xmath33 . as a first cut for potential candidates of stable strangelets , these ones lying above this line can ( or probably will ) completely decay to the pure hadron state via strong processes , and only these ones beneath the line will be possible for metastable or stable strangelets @xcite .
we consider the influence of temperature now . as shown in last section , the free energy per baryon increases when temperature decreases .
the positions of the dots will raise and many dots will across the line and become unstable when temperature decreases .
figure 11 shows this result clearly .
the figure 11 is the same as figure 9 except for temperature @xmath68 .
comparing these two figures , we find many dots across the line when temperature decreases from @xmath105 to @xmath106 . at low temperature
, more strangelets can decay to the pure hadronic state by strong process . here
we must emphasize that although any strangelet initially formed under this line can not decay to a pure hadron state , it is still possibly unstable because it can decay to a hadron and another strangelet changing baryon number , strangeness number and charge @xcite .
we will look for possible strong decays , i.e. single baryon \{n , p , @xmath107 and @xmath33 } emission and mesonic decays at the finite size configuration at finite temperature .
the baryon number @xmath23 , strangeness number @xmath53 and electric charge @xmath56 are conserved in the strong process .
a general expression of a strong baryon decay for a strangelet @xmath108 can be written as @xmath109 and this process is allowed if the free energy balance of the corresponding reaction is @xmath110 where @xmath59 stands for the total free energy of the strangelet and @xmath111 stands for a baryon with strangeness number @xmath112 and electric charge @xmath113 . our results of numerical calculation are shown in figures 12 , 13 and 14 .
for @xmath114 and @xmath115 all strangelets including the ones lying above the lines of figure 9 and the ones which satisfy the inequality ( [ 29 ] ) are drawn by the circles in figure 12 .
these strangelets can strong decay and are unstable .
the unstable strangelets for @xmath103 and @xmath69 are drawn by circle in figure 13 .
figure 12 and figure 13 show that the strangelets with small strangeness number @xmath53 and situated in the left side of the figure are unstable . to study the temperature effect , the same figure for @xmath67 and @xmath116
are shown in figure 14 . comparing figure 12 and
figure 14 , we find that many strangelets become unstable when temperature decreases .
lower temperature is in favour of the strong decay of the strangelet . according to the definition of greiner et .
al @xcite , a strangelet is called metastable in the following if its energy lies under the correspond ( free ) hadronic matter of the same baryon number , charge and strangeness , and if it can not emit a single hadron or multiple hadrons by strong processes . at finite temperature , instead of energy , we use free energy .
a metastable strangelet can then only decay via weak process like the nonleptonic ( hadronic ) decays . in this sub - section ,
we study all possible weak hadronic decay for metastable strangelets . in the weak process
, the baryon number @xmath23 and the electric charge @xmath56 are conserved , but the strangeness number @xmath53 is not conserved .
for the weak decay , @xmath117 .
therefore , a general expression of a weak baryon decay for a strangelet @xmath108 can be written as @xmath118 this process is allowed if the free energy balance of the corresponding reaction is @xmath119 the metastable strangelets which satisfy the inequality ( [ 31 ] ) are shown in figure 12 , 13 and 14 with filled circle .
we find that they are almost situated in the area with negative charge ( or slightly positive charge ) and high strangeness number .
in particular , high strangeness is in favour of the stability of strangelet . as shown in figures 12 , 13 and 14 , a minimal strangeness @xmath120 above which ( @xmath121 ) strangelets are metastable or stable exists .
the values of the minimal strangeness @xmath120 are @xmath122 in figure 12 ( @xmath123 ) and @xmath124 in figure 13 ( @xmath125 ) . defining strangeness fraction @xmath126 as @xmath127 greiner and his co
- workers had predicted that there exists a critical @xmath128at zero temperature and pointed that the metastable or stable strangelets could only be found above this value @xmath129our results strongly support their prediction and extend it to the finite temperature .
we obtained @xmath130 at @xmath69 .
this value does not depend on the baryon numbers .
but as shown in figure 14 , @xmath131 depends on the temperature .
it becomes @xmath132 at @xmath68 .
now we turn to discuss the charge dependence of the stability of strangelet .
as shown in figures 12 , 13 and 14 , higher negative charge is in favour of the metastable and/or stable strangelets . a maximal charge , which is @xmath133 in figure 12 ( @xmath134 ) , @xmath135 in figure 13 ( @xmath136 ) and @xmath137 in figure 13 ( @xmath138 ) , respectively , exists .
the strangelets would be stable or metastable when @xmath139 .
defining charge fraction @xmath140 as @xmath141 we find that the maximal charge fraction depends on not only the temperature but also the baryon number @xmath23 .
for example , @xmath142 equals to @xmath143 for @xmath67 and @xmath144 for @xmath103 when @xmath69 .
the strangelet which is stable against both strong and weak decay is called stable strangelet .
the stable strangelets are shown in figure 12 , 13 and 14 by filled squares .
we find that only a few strangelets can not decay via strong and weak reactions and be completely stable .
for example , as shown in figure 12 , for @xmath67 and @xmath69 , in total @xmath145 strangelets , only @xmath146 strangelets are stable ( @xmath147 ) , ( @xmath148 ) , ( @xmath149 ) and ( @xmath150 ) .
finally , we hope to emphasize that the high negative charge and the high strangeness are in favour of the stable strangelets obviously at finite temperature .
the stable strangelets in figures 12 , 13 and 14 are all highly negative charged .
the conclusion given by greiner et .
al at zero temperature @xcite is still correct at finite temperature .
in summary , by using the qmdtd model , we have studied the thermodynamical properties and the stability of strangelets at finite temperature .
we obtain : ( 1 ) , for fixed strangeness and charge , free energy per baryon decreases as the temperature increases .
the stable radius of the strangelet decreases when the strangeness increases or the charge decreases .
( 2 ) , the higher temperature is in favour of the stability of the strangelets . comparing figure 12 with figure 14 , we see clearly that the total area of metastable and stable strangelets expands when the temperature increases .
( 3 ) , the higher negative charge and higher strangeness number are in favour of the stability of the strangelets .
the stable strangelets are highly negative charged and have high strangeness .
this work is supported in part by the nnsf of china under contract nos . 19975010 , 10047005 and 19947001 .
figure 1 . the dots stand for the free energy per baryon @xmath65 in a @xmath66 plane for all possible strangelets with @xmath67 and @xmath68 .
the free energy per baryon @xmath65 as functions of radius @xmath22 with @xmath154 , @xmath81 and @xmath69 , the three lines represent the strangeness number @xmath83 and @xmath84 , respectively .
the free energy per baryon @xmath65 as functions of radius @xmath22 with @xmath154 , @xmath88 and @xmath69 , the two lines represent the strangeness number @xmath91 and @xmath155 , respectively .
the dots stand for the free energy per baryon @xmath65 with various strangeness numbers @xmath53 for all possible strangelets with baryon number @xmath67 at temperature @xmath69 .
the masses of nucleon , @xmath104 and @xmath156 are represented by the filled squares , respectively .
figure 12 . the electric charge @xmath56 as a function of the strangeness number @xmath157 for unstable strangelets ( open circles ) , metastable strangelets ( filled circles ) and stable strangelets ( filled squares ) with baryon number @xmath67 at temperature @xmath69 .
j. barrette et .
al , phys .
b 252 , 550 ( 1990 ) ; m. aoki et .
al , phys .
69 , 2345 ( 1992 ) ; k. borer et .
al , phys .
72 , 1415 ( 1994 ) ; d. beavis et .
al , phys .
75 , 3078 ( 1995 ) ; t. armstong et .
al , phys .
79 , 3612 ( 1997 ) ; g. appleqvist et .
al , phys .
76 , 3907 ( 1996 ) .
|
using the quark mass density- and temperature dependent model , we have studied the thermodynamical properties and the stability of strangelet at finite temperature .
the temperature , charge and strangeness dependences on the stability of strangelet are investigated .
we find that the stable strangelets are only occured in the high strangeness and high negative charge region .
|
optical amplification in a strongly driven sodium atoms system without population inversions has been observed in the experiment @xcite where the atomic recoil effect and doppler shift were purposely avoided by a specific systematic configuration .
when the pumping field is off - detuned from the atomic transition frequency @xmath0 , the experimental pump - probe gain spectrum exhibits a normal absorption - gain profile similar to that of fig.1(a ) which is drawn from a direct simulation and has been obtained by bonifacio _
et al _ in their original work @xcite . of the probe field versus pump - probe detuning @xmath1 with parameters @xmath2 , @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 . ( a ) the atomic motion is totally removed from the model ; ( b ) the motion effect is included ( traditional carl).,title="fig : " ] of the probe field versus pump - probe detuning @xmath1 with parameters @xmath2 , @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 .
( a ) the atomic motion is totally removed from the model ; ( b ) the motion effect is included ( traditional carl).,title="fig : " ] the vertical ordinate of fig.1 stands for a relative gain of the probe field and the horizontal ordinate @xmath8 denotes the detuning between the pump and the probe field
. however , this anti - symmetry gain profile around @xmath9 can be drastically changed ( fig.1(b ) ) by the atomic recoil motion which plays an important role in the amplification mechanism of the collective atomic recoil laser ( carl ) @xcite . as shown in fig.1(b ) , the gain peak within the traditional madey region @xcite is greatly enhanced , and which has been a signature of light amplification induced by cooperative atomic recoil effect .
being modified by the recoil motion and exchanging energies with one common pump field , the atoms will exhibit some kinds of self - organization behavior @xcite or dynamic synchronization @xcite which induces phase transitions in the relevant system @xcite@xcite .
indeed , in the traditional carl system , the atomic density grating as well as the population inversion grating on the scale of the electromagnetic wavelength emerges spontaneously during the light amplification process @xcite .
however , some theoretical simplicities of the traditional carl model , as have been pointed by some authors @xcite , impede a cw gain output of the system .
therefore , the steady - state carls were proposed and have been experimentally demonstrated @xcite@xcite , but , with a relative weak gain . in this paper , we report a result in order to give a more detailed theoretical analysis to a modified carl system @xcite which produces a cw laser output with a higher gain . in our previous scheme ,
the active cold atoms are trapped in a harmonic potential , which can be realized by loading the atoms into an optical lattice or an overall magnetic trap , instead of introducing a light molasses @xcite@xcite or adding a thermal collision / momentum damping mechanism to the atomic motion @xcite .
we suppose the trap to be strong in radial direction and loose in axial direction ( cigar - shaped trap ) , which reducing our system effectively to one - dimension .
our modification by this configuration produces a very different laser gain profile illustrated by fig.2 @xcite . of the probe field in a trap with frequency of @xmath10 and the time @xmath11 .
the other parameters are the same as fig.1 for a convenient comparison . ]
fig.2 displays a discrete spectral structure which is very similar to that of a nonlinear jaynes - cummings model of one atom in a high - q cavity @xcite .
when the pump - probe detuning satisfies a raman condition @xmath12 , where @xmath13 is an arbitrary integer and @xmath14 is the effective axial trapping frequency , a distinct light amplification process occurs for certain @xmath13 as displayed by fig.2 . in this case , a self - organization of all the atomic oscillators along the trap axis is demonstrated by atomic distribution pattern and atomic orbit bunching in the phase space in this paper .
the following contains our model with a more rigorous analysis of all the above interesting results in a framework of semiclassical method .
the relevant system is supposed to be enclosed in a ring cavity @xcite and the operating medium is composed of a collection of @xmath15 two - level atoms of mass @xmath16 , moving along the axial direction @xmath17 of a trap described by @xmath18 $ ] .
the atoms are driven by a strong far off - detuned pump field propagating along @xmath17 and probed by a counterpropagating weak probe field .
if we set @xmath19 , the radial trapping mode will be hardly excited in the recoil process and a one - dimensional trap , @xmath20 , should be a promising approximation . with above simplification ,
the explicit form of the hamiltonian of this system is @xmath21 where the operators @xmath22 and @xmath23 are the canonically conjugate momentum and center - of - mass position of the @xmath24th atom , @xmath25 and @xmath26 denote the usual spin operators of the atomic internal degrees of freedom with a transition frequency of @xmath0 .
the probe field and the pump field with the carrier frequencies @xmath27 and @xmath28 are denoted by the photon creation operators @xmath29 and @xmath30 , respectively .
the coupling constant between the atom and the @xmath31th optical field is given by @xmath32 ( @xmath33 ) .
following the conventional commutation relations of all above operators , the heisenberg equations for the expectations of the relevant operators under the semiclassical approximation are @xmath34 -\gamma \left ( \sigma _ { zj}-1\right ) , \\ \frac{d\sigma
_ { j}}{d\tau } & = & i\left ( \delta _ { 20}+\frac{p_{j}}{2}\right ) \sigma _ { j}-\rho \sigma _ { zj}\left ( a_{1}e^{i\theta _ { j}}+a_{2}\right ) -\gamma \sigma _ { j } , \notag \\ \frac{da_{1}}{d\tau } & = & i\delta _ { 21}a_{1}+\frac{1}{n}\sum_{j=1}^{n}\sigma _ { j}e^{-i\theta _ { j}}-\kappa a_1 , \notag\end{aligned}\]]where the atomic polarization and population damping @xmath35 as well as the probe field damping @xmath36 have been phenomenologically added .
the strong pump field @xmath37 is treated as a real undepleted field and removed from eq.([eq2 ] ) as an external condition @xcite .
some dimensionless variables or transformations have been introduced in above derivations , i.e. , @xmath38 , @xmath39 , @xmath40 ( @xmath33 ) , @xmath41 , @xmath42 , @xmath43 , @xmath44 , @xmath45 and @xmath46 under the assumption @xmath47 and @xmath48 , for simplicity . the scaling parameters @xmath49 and @xmath50 used above are two - photon recoil frequency @xmath51 and the dimensionless carl parameter @xmath52 , respectively .
it should be noted that eq.([eq2 ] ) are identical to those derived in the ref.@xcite@xcite@xcite except for one single term @xmath53 for the harmonic trapping force , and all the simulations throughout this paper will be based on these equations . in order to analyze the unique discrete gain profile in fig.2 ,
we focus on the long - time evolution of @xmath54 when the probe damping is weak so as to be enclosed in a high - finesse ring cavity @xcite .
the gain quantity depicted in fig.2 is defined by @xmath55 with @xmath56 being the initial value of the weak probe field . by formally integrating eq.([eq2 ] )
, the probe field can be expressed as @xmath57and the probe gain @xmath58 + \left ( e^{-2\kappa \tau } -1\right ) , \label{gain}\]]where @xmath59 is defined by @xmath60with @xmath61 being defined as a collective atomic coherence parameter by ref.@xcite . for an enough long time of evolution , provided @xmath62 with @xmath63 @xcite , all the atoms in the trap will approximately perform a harmonic oscillation with different amplitudes @xmath64 and phases @xmath65 , described by @xmath66 $ ] .
then @xmath67 in eq.([laplace ] ) will be expanded as @xmath68 e^{in[\phi _ { j0}(\tau ) -\frac{\pi } { 2}]}\int_{0}^{\tau } d\tau ^{\prime } \sigma _ { j}\left ( \tau ^{\prime } \right ) e^{\left [ \kappa -i\left ( \delta _ { 21}-n\cdot \nu \right ) \right ] \tau ^{\prime } } \right ] \text { , } \label{eq4}\]]where the function @xmath69 $ ] is the @xmath13-th order bessel function of the first kind . in above derivations , we have assumed that the amplitude @xmath70 and phase @xmath71 are slowly varying quantities after a long time evolution and can be moved out from the integral of eq.([eq4 ] ) . if we extend the upper integral limit of eq.([eq4 ] ) to infinity ( @xmath72 ) ,
the integral defines a laplace transformation of @xmath73 , i.e. , @xmath74 = \int_{o}^{\infty } d\tau ^{\prime } \sigma _ { j}\left ( \tau ^{\prime } \right ) \cdot $ ] @xmath75
\tau ^{\prime } } $ ] , on condition that @xmath76 does nt increase much more than @xmath77 . generally @xmath78 is limited by bloch sphere of two - level atom and can be approximately expanded by a fourier series with @xmath79 as a fundamental frequency @xmath80where @xmath81 is the fourier amplitude of the @xmath13-th order harmonic components .
furthermore , if the pump - atom detuning satisfies @xmath82 , the long - time average polarization @xmath83 will be @xmath84where @xmath85 is the rabi frequency of the pump field and @xmath86 can be determined by taking an adiabatic approximation of atomic internal freedom as done in @xcite . if we only take the dc term of @xmath78 in eq.([fourier ] ) for a qualitative spectrum analysis , eq.([eq4 ] ) reads @xmath87 e^{in[\phi _ { j0}(\tau ) -\frac{\pi } { 2}]}\int_{0}^{\tau } d\tau ^{\prime } e^{\left [ \kappa -i\left ( \delta _ { 21}-n\cdot \nu \right ) \right ] \tau ^{\prime } } \right ] \notag \\ & = & s_{0}\sum_{n=-\infty } ^{\infty } \frac{1}{n}\sum_{j=1}^{n}j_{n}\left [ \theta _ { j0}(\tau ) \right ] e^{in[\phi _ { j0}(\tau ) -\frac{\pi } { 2}]}\frac { e^{\kappa
\tau -i\left ( \delta _ { 21}-n\cdot \nu \right ) \tau } -1}{\kappa -i\left ( \delta _ { 21}-n\nu \right ) } .
\label{eq6}\end{aligned}\]]as the probe - field damping rate @xmath88 , we have @xmath89 , then the gain under resonant condition is @xmath90where @xmath91 } $ ] is the asymptotic coherence parameter which is determined by the secular average atomic polarization @xmath92 and the instantaneous atomic distribution at @xmath93 .
the coefficient @xmath94 of eq.([gainlimit ] ) shows here that when the damping rate @xmath95 is small enough and the evolution time @xmath96 , the intensity gain of the probe field will be extremely high in the neighborhood of @xmath97 , and thus results in a uniform discrete gain structure as shown in fig.2 simulated directly on eq.([eq2 ] ) .
although the raman gain condition will not be changed if the higher harmonic components of eq.([fourier ] ) is included , except for a more complicated sum formula for eq.([eq6 ] ) , the height of the gain peak for a specific @xmath13 in fig.2 ca nt be interpreted by above analytical method . as shown by fig.2 , for example
, the light amplification mostly for positive multiple of the trap frequency and approximately up to @xmath98 is still a problem .
the further simulations only indicate that the light amplification for positive detuning is a feature owing to the red - detuned of the pump field from the atom frequency .
however , this regular gain spectrum still can be developed as an efficient detection method to analyze the micro - trap potential formed by a local interference field .
under the raman resonant condition @xmath97 , another ingredient that enhances the maximum gain of probe field is the synchronization of the atomic motion in the harmonic trap , which is indicated by the term of @xmath99 in eq.([gainlimit ] ) and can be demonstrated through a bunching of the atomic trajectories in the phase space . to some extent ,
the present problem described by eq.([eq2 ] ) under the adiabatic approximation is similar to the famous kuramoto model of ensembles of coupled oscillators @xcite , in which the complex order parameter @xmath100 plays an equivalent role as that of the bunching parameter @xmath101 @xcite in the traditional carl system @xcite . for the complex order parameter
, @xmath102 measures the phase coherence of all the oscillators and @xmath103 describes their average phase .
a main result of kuramoto model is : when the coupling intensity ( in present model the coupling intensity is related to the rabi frequency , @xmath104 , of the pump field ) exceeds a certain threshold , the incoherent oscillators are mutually synchronized and @xmath102 grows exponentially . in present case , when the resonant conditions @xmath105 and @xmath82 are both satisfied , the long - time output gain of @xmath54 reduces to@xmath106 where @xmath107 } $ ] is the secular complex order parameter determined only by the atomic distribution in the phase space .
eq.([gain0 ] ) indicates that the field gain is explicitly dependent on the increase of collective atomic coherence , @xmath108 .
atoms in phase space ( a ) at @xmath109 ( initial distribution ) , ( b ) at @xmath110 , ( c ) at @xmath111 and ( d ) the corresponding time evolution of the probe intensity @xmath112 under the resonant condition with parameters @xmath113 , @xmath114 , @xmath3 , @xmath115 , @xmath5 , and @xmath6.,title="fig:",width=302,height=226 ] atoms in phase space ( a ) at @xmath109 ( initial distribution ) , ( b ) at @xmath110 , ( c ) at @xmath111 and ( d ) the corresponding time evolution of the probe intensity @xmath112 under the resonant condition with parameters @xmath113 , @xmath114 , @xmath3 , @xmath115 , @xmath5 , and @xmath6.,title="fig:",width=302,height=226 ] atoms in phase space ( a ) at @xmath109 ( initial distribution ) , ( b ) at @xmath110 , ( c ) at @xmath111 and ( d ) the corresponding time evolution of the probe intensity @xmath112 under the resonant condition with parameters @xmath113 , @xmath114 , @xmath3 , @xmath115 , @xmath5 , and @xmath6.,title="fig:",width=302,height=226 ] atoms in phase space ( a ) at @xmath109 ( initial distribution ) , ( b ) at @xmath110 , ( c ) at @xmath111 and ( d ) the corresponding time evolution of the probe intensity @xmath112 under the resonant condition with parameters @xmath113 , @xmath114 , @xmath3 , @xmath115 , @xmath5 , and @xmath6.,title="fig:",width=302,height=226 ] at an early stage of evolution , as the atoms incoherently scatter the pumping photon into probe mode and their spatial distribution in the trap almost keeps uniform with @xmath116 ( fig.3(a ) ) , the probe field fluctuates around a small value . after a period of lethargy , a self - organization structure is established through the recoil motion modified by the trap and the atoms in phase space concentrate into several clusters ( fig.3(b ) ) ,
resulting in an exponential growth of the probe field .
when this process is saturated ( fig.3(c ) ) by a counteraction of the spontaneous damping processes @xcite , the coherent parameter reaches its maximum value , or extremely @xmath117 ( all the atoms moving synchronously as one big atom ) , and a stable output of @xmath54 will be set up .
fig.3(d ) , which is directly simulated from eq.([eq2 ] ) , displays a typical amplification process of the probe intensity under this resonant condition and , finally , establishes a saturated output with an efficiency nearly to @xmath118 ( here the efficiency is defined by the intensity rate between pumping field and the output probe field ) . according to a further fourier analysis
, the fast oscillation observed in fig.3(d ) at saturation is induced by the harmonic motion of the atoms in the trap . and
( b ) @xmath119 with parameters @xmath10 , @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath120.,title="fig : " ] and ( b ) @xmath119 with parameters @xmath10 , @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath120.,title="fig : " ] during the lasing process , the self - organization structure related to the light amplification corresponds to a spontaneous bunching of atomic trajectories in the phase space ( fig.4 ) , which , actually , is similar to the atomic density grating in the free space carl system when @xmath101 is large . in order to demonstrate the spatial features of this atomic self - organization behavior , the initial spatial distribution of the atoms
are set uniform in a region of @xmath121 $ ] and the initial atomic momentum follows a gaussian distribution ( showed by fig.3(a ) ) with a thermal spread of width above recoil temperature ( about @xmath122 ) @xcite . before the increment of @xmath123 ,
the atomic trajectories are almost ubiquitous in the phase space and , after a stable output of @xmath54 is established , most atomic trajectories converge into the antinode area ( around @xmath124 ) and the other ones into the node area ( around @xmath125 ) of a standing wave induced by an interference between the amplified probe field and the strong pump field , with its wavelength being defined by @xmath126 . the bunching positions of the most atomic trajectories into the antinode area are a clear result due to the red - detuned of the pump field @xmath127 in fig.4 . according to more simulations ,
we find that the amplification and the pattern formation of the atoms in phase space under the raman resonant condition are robust to other parametric values provided that the gas medium is operated on a far - off red - detuned region and the trap is strong enough @xcite . following above conditions the change of collective coupling constant @xmath128 in the resonant gain region will not qualitatively change the self - organization effect and the long time high cw gain output of this model .
the experimental realization of above cw carl system can be carried out by loading the laser - cooled atoms into an overall magnetic trap or into a far - off detuned optical lattice .
thereafter , the trap is displaced along its axis for about one light wavelength @xmath129 to trigger the atomic motion , and then the atoms are pumped by a strong red - detuned field .
a gain of a seed probe field in the backward direction under the resonant conditions @xmath130 will be detected then .
this process will be enhanced further by a collective behavior of the atoms due to the synchronization of their recoil motion under raman resonant condition .
the strong pump field feeds the energy to the system ( excite the trapping oscillating mode by atomic recoil motion ) and the dipole force , generated by the interference of the probe and pump light , bunches the atoms into several groups to adjust them synchronically extracting kinetic energy from the trap mode into the probe field @xcite .
this scheme will also present a clue to introduce a new configuration of free electron lasers ( fel ) where the high speed electrons are injected into a ( or an array of ) strong quantum trap ( e.g. quantum dots ) . by stimulating the electrons motion through electromagnetic field on the electrode with appropriate frequency
, the high gain field with certain frequency will be generated when the electrons oscillate collectively in ( or through ) these quantum traps . in conclusion , we have investigated the light amplification in a carl with the atomic center - of - mass motion influenced by an external harmonic potential . under appropriate parametric condition that the pump field is red - detuned from the atomic transition frequency , our results show that carl confined by a robust trap can give rise to a cw laser with a higher lasing gain than that of the carl in the free space . the optical gain profile as a function of the pump - probe detuning displays a discrete structure and another kind of atomic self - organization related to the light gain is demonstrated in the phase space .
this work suggests a new way to promote a cw carl device with a higher gain and , perhaps , analogically implies a new configuration of fel .
we would like to thank dr .
o.zobay for helpful discussions . the work is supported by the science foundation of the education bureau of shaanxi province under grant no.05jk140 and by the key project no.zk2406 of baoji university of arts & science .
r. bonifacio and l. de salvo , _ nucl .
methods phys .
sec . a _ * 341 * , 360 ( 1994 ) ; r. bonifacio , l. de salvo , l. m. narducci , e. j. dangelo , _ phys . rev
. a _ * 50 * , 1716 ( 1994 ) ; l. de salvo , r. cannerozzi , r. bonifacio , e. j. dangelo , l. m. narducci , _ phys . rev . a _ * 52 * , 2342 ( 1995 ) ; r. bonifacio and p. verkerk , _ opt .
commun . _ * 124 * , 469 ( 1996 ) .
g. l. lippi , g. p. barozzi , s. barbay , j. r. tredicce , _ phys .
lett . _ * 76 * , 2452 ( 1996 ) ; p.r.hemmer , n. p. bigelow , d. p. katz , m. s. shahriar , l. de salvo , r. bonifacio , _ phys . rev
. lett . _ * 77 * , 1468 ( 1996 ) .
|
self - organization effects related to light amplification in the collective atomic recoil laser system with the driven atoms confined in a harmonic trap are investigated further . in the dispersive parametric region , our study reveals that the spontaneously formed structures in the phase space contributes an important role to the light amplification of the probe field under the atomic motion being modified by the trap .
|
if colloidal particles get trapped at fluid - fluid interfaces they interact effectively via deformations of the interface .
these so - called capillary interactions can easily be tuned by changing external fields and they depend sensitively on the shape of the particles .
this makes them good candidates for designing self - assembling systems and provides a convenient experimental playground for studying basic issues of statistical mechanics in two dimensions in the presence of long - ranged interactions @xcite . on the other hand ,
recent experiments @xcite show that capillary forces between elongated particles floating on spherical interfaces can have important consequences for stabilizing so - called pickering emulsions , which are formed by particle - covered droplets ( e.g. , oil ) in a solvent ( e.g. , water ) . whereas considerable theoretical progress has been made in understanding capillary interactions at flat interfaces @xcite ,
basic issues such as the balance of forces acting on the interface and the influence of the incompressibility of the liquid enclosed by spherical interfaces have not yet been fully resolved .
curved interfaces pose the additional difficulty @xcite that in the presence of external forces the condition of mechanical equilibrium demands to either fix the center of mass of the droplet by an external body force or to pin the droplet surface , e.g. , to a solid plate .
the experimentally relevant issues of the boundary conditions at the plate and of their influence on the pair potential between capillary _ monopoles _ have been studied previously @xcite . here , we study _ higher _ capillary multipoles and derive the corresponding effective pair potentials for pointlike particles in terms of an expansion in spherical harmonics . by using a finite element method @xcite we compare these analytical results with those obtained by the numerical minimization of the surface free energy of drops in the presence of ellipsoidal particles .
we consider a spherical , full , and spatially fixed droplet of radius @xmath0 with colloidal particles trapped at its surface with surface tension @xmath1 .
we model the effect of particles in terms of an external surface pressure field @xmath2 ( see below ) parameterized by spherical coordinates @xmath3 on the unit sphere .
the equilibrium shape of the droplet subjected to the pressure field @xmath4 follows from minimizing the corresponding free energy functional @xmath5 $ ] expressed in terms of the dimensionless radial displacement of the interface @xmath6 . in the limit of small interfacial gradients @xmath7
one has @xcite @xmath8= \int\ !
d\omega\ , \left[\frac{1}{2}(\nabla_a v)^2 - v^2 - \big(\pi(\omega)+\mu\big)v\right ] , \label{free_en_o2}\ ] ] where @xmath9 is the dimensionless
_ _ a__ngular gradient on the unit sphere @xcite .
the first two terms in eq .
( [ free_en_o2 ] ) represent the surface free energy , the third term corresponds to the work done by the dimensionless external pressure @xmath10 in displacing the interface and the fourth term serves to implement the volume conservation .
the value of the lagrange multiplier @xmath11 follows from imposing the volume constraint @xmath12 @xmath13 can be identified with the shift in the internal pressure of the droplet with respect to the laplace pressure @xmath14 of an unperturbed , perfectly spherical droplet .
the stationary condition @xmath15 leads to the euler - lagrange equation @xmath16 we expand both the deformation @xmath17 and the pressure field @xmath18 in spherical harmonics @xmath19 so that eq . ( [ helmholtz_pi ] ) turns into an infinite set of algebraic equations @xcite : @xmath20v_{lm}=\pi_{lm}+\mu\delta_{l0 } , \label{helmholtz_multipole}\ ] ] where @xmath21 and @xmath22 , and the spherical multipoles are defined as @xmath23 for @xmath24 or @xmath25 . the volume constraint in eq .
( [ volume_constr2 ] ) implies @xmath26 and thus @xmath27 , which means that the internal pressure shift counterbalances the external pressure .
according to eq .
( [ helmholtz_multipole ] ) the @xmath28 components of the deformation @xmath29 are undefined .
this is consistent with the fact that those components describe translations of the whole droplet without any change in shape which do not change the free energy . on the other hand all @xmath28 components of the external pressure
@xmath30 must cancel , reflecting the condition of balance of forces acting on the droplet .
indeed , the multipoles @xmath31 are proportional to the cartesian components @xmath32 of the total force @xmath33 acting on the droplet .
hence the condition @xmath34 is equivalent to @xmath35 with @xmath36 .
as mentioned above we consider a pressure field in the form of the superposition of @xmath37 pointlike forces @xmath38 pointing into directions @xmath39 : @xmath40 where @xmath41 and @xmath42 is the dirac delta distribution expressed in terms of spherical coordinates . the pressure field as a whole must obey the condition of balance of forces which can be expressed as @xmath43 for @xmath36 . the corresponding interface deformation can be written as the superposition of single particle contributions : @xmath44 where @xmath45 is green s function describing the response of the interface in direction @xmath46 to a point - force applied in direction @xmath47 and which , according to the constraints of constant liquid volume and fixed center of mass obeys the equation @xcite @xmath48 where the right hand side is a modified dirac delta function @xmath49 with the @xmath50 and @xmath28 components projected out . from integrating both sides of eq .
( [ greens_equation ] ) over the unit sphere it follows that the @xmath50 component of @xmath51 vanishes , which reflects the condition of constant volume @xmath52 .
the @xmath28 component of @xmath51 also vanishes , which can be seen by multiplying both sides of eq .
( [ greens_equation ] ) by a radial vector @xmath53 , integrating over the unit sphere and using the fact that @xmath53 can be expressed in terms of spherical harmonics with @xmath28 only .
this reflects the condition that the center of mass of the droplet is fixed in space , which can be formulated as @xmath54 .
[ l][l][1]@xmath55 [ c][c][1]@xmath56 [ l][l][1]@xmath57 [ r][l][1]@xmath58 [ c][c][1]@xmath59 [ c][c][1]@xmath60 [ c][c][1]@xmath61 [ c][c][1]@xmath62 [ c][c][1]@xmath63 [ c][c][1]@xmath64 [ c][c][1]@xmath65 [ c][c][1]@xmath66 [ c][c][1]@xmath67 [ c][c][1]@xmath68 [ c][c][1]@xmath69 ( north pole ) and @xmath59 on the unit sphere .
the angles @xmath55 and @xmath57 represent orientations of the local coordinate frames @xmath70 and @xmath71 associated with the pressure distributions ( see main text).,title="fig:",scaledwidth=45.0% ] according to the above considerations the pressure field due to a _
single _ point - force violates the balance of forces acting on the droplet .
therefore , if one wants to cast the solution still into in the form given by eq .
( [ v1 ] ) with @xmath72 , there must be an additional pressure field counterbalancing the single point - force . inserting @xmath73 into eq .
( [ helmholtz_pi ] ) and using eq .
( [ greens_equation ] ) , one obtains the corresponding pressure field @xmath30 in the form @xmath74 with the effective pressure @xmath75 , where @xmath56 is the angle between the directions @xmath46 and @xmath47 . the latter pressure must emerge in order to cancel the @xmath28 component of the external pressure @xmath76 . therefore @xmath77 corresponds to a body force fixing the center of mass ( @xmath78 ) of the droplet without contributing to the deformation field @xmath29 .
the expression in eq .
( [ v1 ] ) for the deformation under the action of point - forces can be generalized to the case of a pressure field @xmath79 due to an arbitrary continuous force distribution : @xmath80 the total pressure @xmath81 acting on the interface again has an additional component @xmath77 accomplishing the fixing of the center of mass , i.e. , @xmath82 .
one can express the free energy in terms of green s function by integrating by parts eq .
( [ free_en_o2 ] ) , by using eq .
( [ helmholtz_pi ] ) and the divergence theorem , and finally by inserting @xmath17 as given by eq .
( [ upi ] ) @xcite : @xmath83
we study the pressure @xmath84 localized around two different directions @xmath85 and @xmath59 . to this end
we introduce the following decomposition : @xmath86 where @xmath87 ( @xmath88 ) denotes the rotation transforming the original coordinate frame @xmath89 into the coordinate frame @xmath70 ( @xmath71 ) , referred to as @xmath90 ( @xmath91 ) , associated with the pressure distribution @xmath92 ( @xmath93 ) ( see fig . [ sphere ] ) .
we use the parameterization in terms of euler angles , in which an arbitrary rotation @xmath94 can be composed of a rotation around the @xmath63-axis by the angle @xmath95 , around the ( rotated ) @xmath62-axis by the angle @xmath96 , and finally around the ( rotated ) @xmath63-axis by the angle @xmath1 . under the rotation @xmath97 of the coordinate frame
the coordinates transform according to @xmath98 , as indicated in eq .
( [ decomposition ] ) . the total free energy can be written as @xmath99 , where @xmath100\ln(r_0/a_i)+o(1)$ ] is the self - energy of particle @xmath101 @xcite , which does not depend on the relative position of the particles on the droplet but depends on the particle size @xmath102 . for a particle located at the north pole this is defined as the smallest distance for which @xmath103 for all @xmath104 and @xmath105 ; this corresponds to the solid angle circumscribing the particle - liquid interface .
the interaction free energy @xmath106 is given by the cross - terms in eq .
( [ free_en_sph ] ) with @xmath30 from eq .
( [ decomposition ] ) : @xmath107 where @xmath108 are the spherical harmonics coefficients of @xmath45 . without loss of generality
we can take @xmath109 such that the reference frame @xmath90 coincides with the original reference frame @xmath89 ( see fig .
[ sphere ] ) , which implies that @xmath110 .
spherical harmonics transform under the representation of the group of rotations according to @xmath111 , where @xmath112 is the wigner @xmath113-matrix and reads ( adopting the convention used in ref .
@xcite ) @xmath114 where @xmath115 is known as the wigner ( small ) @xmath116-matrix .
we use the parameterization in terms of the orientations @xmath55 and @xmath57 of the coordinate frames @xmath90 and @xmath91 , respectively , relative to the great circle connecting the points @xmath109 and @xmath59 on the unit sphere ( see fig . [ sphere ] ) .
the rotation @xmath88 is then parameterized by the triad of the euler angles @xmath117 . by using the orthogonality @xmath118 of the spherical harmonics , and due to the identities @xmath119 and @xmath120 we finally obtain @xmath121 having in mind colloidal particles of radius @xmath122 as sources of the effective surface pressure @xmath30 , we would like to relate the spherical multipoles @xmath123 , @xmath124 , @xmath22 to prescribed forces , torques , and higher capillary multipoles defined for the case of a flat interface . if @xmath125 the interface can be treated as to be locally flat in the close neighborhood @xmath126 of angular extent @xmath127 around each particle separately .
assuming for reasons of simplicity that @xmath126 is centered at the @xmath63-axis and approximating it by a circular disk @xmath128 of radius @xmath122 in the tangent plane we can use the asymptotic form of spherical harmonics @xcite @xmath129{}i^{|m|+m}a_{l|m|}\theta^{|m| } e^{im\phi},\ ] ] where @xmath130 in order to obtain @xmath131{}\\ i^{|m|+m}a_{l|m|}\left(\dfrac{a}{r_0}\right)^{|m|+1}q_{|m|}e^{im\tilde{\phi}_{|m| } } , \label{qlm - qm } % \end{equation}\end{gathered}\ ] ] where we have introduced the dimensionless moduli @xmath132 and the phase @xmath133 where @xmath134 , @xmath135 denotes a complex number , and @xmath136 , with @xmath137 being a projection onto the plane tangent to the reference sphere at @xmath39 . from eq .
( [ qnsph ] ) it follows that the moduli @xmath138 are the usual capillary multipoles defined at a locally flat interface ( see refs .
accordingly , due to the conditions of force and torque balance on a flat interface @xcite , @xmath139 represents a `` capillary monopole '' , i.e. , the total external force acting on the particle , @xmath140 a `` capillary dipole '' , i.e. , the total external torque . in the case of a free particle , @xmath141 represents the lowest non - vanishing multipole .
furthermore , we note that for each particle @xmath101 one can always choose the orientation @xmath142 of the coordinate frame @xmath143 such that the phase of the order @xmath144 vanishes , i.e. , @xmath145 .
we can write the interaction energy @xmath146 for a pair of capillary multipoles @xmath138 and @xmath147 of arbitrary orders @xmath134 and @xmath148 as a sum of terms such that in eq .
( [ multipole_expansion ] ) the sum over @xmath149 is taken under the constraint @xmath150 with @xmath151 and @xmath152 fixed to @xmath153 and @xmath154 ; this implies @xmath155 .
accordingly , the interaction free energy @xmath146 for a pair of capillary multipoles @xmath156 and @xmath157 scales with the droplet radius @xmath0 as @xmath158 .
note , that in the case of two monopoles ( @xmath159 and @xmath160 ) the interaction does not depend on @xmath0 but only on the moduli @xmath139 and @xmath161 .
the wigner d - matrix in eq .
( [ multipole_expansion ] ) can be expressed in terms of the jacobi polynomials @xmath162 @xcite : @xmath163^{1/2 } \left(\sin\dfrac{\bar{\theta}}{2}\right)^{m'-m}\times\\ \times \left(\cos\dfrac{\bar{\theta}}{2}\right)^{m'+m}p^{(m'-m , m'+m)}_{l - m'}(\cos\bar{\theta } ) .
% \end{equation}\end{gathered}\ ] ] by using certain properties of the jacobi polynomials ( see , e.g. , ref .
@xcite ) we finally obtain @xmath164 { } \dfrac{q_{n}q'_{n'}}{(-2)^{n+n'+1}n!n'!\pi}\dfrac{a^{n}a'^{\,n'}}{r_0^{n+n ' } } \sum_{l\geq \max\{2,n , n'\}}\dfrac{(2l+1)}{(l+2)(l-1)}\times\\ % \theta(l - n)\theta(l - n')\\ \vspace*{-0.5 cm } \times\left\ { \begin{array}{l } \dfrac{(l+n')!}{(l - n)!}\bigg[(-1)^{n}\cos(n\phi_{1}+n'\phi_{2})\left(\cos\dfrac{\bar{\theta}}{2}\right)^{n'-n}\left(\sin\dfrac{\bar{\theta}}{2}\right)^{n'+n } p_{l - n'}^{(n'+n , n'-n)}(\cos\bar{\theta } ) \\
\hspace*{-0.2 cm } \begin{array}{lll }
\ + \cos(n\phi_{1}-n'\phi_{2})\left(\cos\dfrac{\bar{\theta}}{2}\right)^{n'+n}\left(\sin\dfrac{\bar{\theta}}{2}\right)^{n'-n}p_{l - n'}^{(n'-n , n'+n)}(\cos\bar{\theta})\bigg ] , & n>0 , & n'>0,\\ ( -1)^n\cos(n\phi_{1})p_l^n(\cos\bar{\theta } ) , & n>0 , & n'=0,\\ 2^{-1}p_l(\cos\bar{\theta } ) , & n=0 , & n'=0 , \end{array } \end{array}\right . \label{fn1n2 } % \end{equation}\end{gathered}\ ] ] where @xmath165 and @xmath166 are the legendre and the associated legendre polynomials , respectively . for a first discussion
, we evaluate the interaction energy in eq .
( [ fn1n2 ] ) for @xmath167 and @xmath168 . in the case of two identical pointlike _ monopoles _ , @xmath169 , one obtains @xcite @xmath170{}\\ \frac{q_0 ^ 2}{4\pi}\left[\frac{1}{2 } + \frac{4}{3}\cos\bar{\theta } + 2\cos\bar{\theta}\ln\left(\sin\dfrac{\bar{\theta}}{2}\right ) \right ] , \label{greens_function } % \end{equation}\end{gathered}\ ] ] which gives back green s function @xmath171 as given in ref .
we note that according to eq .
( [ greens_function ] ) one has @xmath172 , which renders the deformation @xmath173 due to a pointlike force , known for a flat interface , where @xmath174 is the arc length ( see the dashed line in fig .
[ sphere ] ) and @xmath0 plays the role of the capillary length . in the case of pointlike _ dipoles _ , @xmath175 , one can evaluate the series involving the jacobi polynomials by using their generating function ( see the appendix ) : @xmath176{}\dfrac{q_1 ^ 2}{8\pi}\left(\dfrac{a}{r_0}\right)^2\times \\ \times \big [ \cos(\phi_1+\phi_2)f_+(\bar{\theta } ) + \cos(\phi_1-\phi_2)f_-(\bar{\theta } ) \big ] \label{df11-final } % \end{equation}\end{gathered}\ ] ] where @xmath55 and @xmath57 are the orientations of the particles as indicated in fig .
[ sphere ] and chosen such that the phases @xmath177 defined in eq .
( [ phin ] ) vanish for @xmath178 , and where @xmath179 we note that the dependences on the angular separation @xmath56 and on the orientations @xmath180 do not factorize .
therefore , in order to minimize the free energy , the particles adopt orientations which depend on @xmath56 .
according to eq .
( [ df11-potential ] ) below one can distinguish three branches of the free energy ( see the dotted green lines in fig .
[ fig : potentials ] ) corresponding to three different minimal orientations , out of which the one with the lowest free energy is the equilibrium one . assuming that the rotational relaxation of the particles is much faster than the translational one
, an effective interaction potential can be obtained by minimizing the free energy with respect to @xmath55 and @xmath57 : @xmath181{}\\ \dfrac{q_1 ^ 2}{8\pi } \left\{\begin{array}{llll}\vspace*{0.1 cm } -f_+(\bar{\theta})+f_-(\bar{\theta } ) , & \text{for } & \bar{\theta}<\bar{\theta}_0 , & \,\,\uparrow\,\,\,\uparrow \\ \vspace*{0.1 cm } -f_+(\bar{\theta})-f_-(\bar{\theta } ) , & \text{for } & \bar{\theta}_0<\bar{\theta}<\bar{\theta}_1 , & \rightarrow\leftarrow \\ \vspace*{0.1 cm } \quad f_+(\bar{\theta})-f_-(\bar{\theta } ) , & \text{for } & \bar{\theta}>\bar{\theta}_1 , & \,\,\uparrow\,\,\,\downarrow \end{array}\right .
\label{df11-potential } % \end{equation}\end{gathered}\ ] ] where @xmath182\simeq0.38 $ ] and @xmath183 are the zeros of @xmath184 and @xmath185 , respectively ; the arrows indicate the minimal orientational configuration ( e.g. , @xmath186 corresponds to @xmath187 etc ) . in the case of pointlike _ quadrupoles _ , @xmath188 , one obtains @xmath189{}\\ -\dfrac{3q_2 ^ 2}{64\pi } \left(\dfrac{a}{r_0}\right)^4 \cos(2\phi_1 + 2\phi_2 ) \dfrac{1}{\sin^4(\bar{\theta}/2)}. \label{df22-final } % \end{equation}\end{gathered}\ ] ] distinct from the case of dipoles , the free energy has a minimum for @xmath190 ( so that @xmath191 ) , independent of @xmath56 .
this corresponds to a monotonic attraction ( see the solid blue line in fig .
[ fig : potentials ] ) .
we discuss the influence of the curvature of the interface on the interaction potentials by comparing the expressions for the curved interfaces ( eqs .
( [ df11-final])-([df22-final ] ) , solid lines in fig .
[ fig : potentials ] ) with the well established results for the case of a flat interface ( see , e.g. , ref .
@xcite , dashed lines in fig .
[ fig : potentials ] ) , with the corresponding spatial separation @xmath116 between the particles given by @xmath192 .
a striking difference is the non - monotonicity for curved interfaces of the free energy for monopole - monopole and dipole - dipole interactions , which in both cases leads to short - ranged attraction and long - ranged repulsion .
these peculiarities can be traced back to the presence of an external body force necessary for fixing the center of mass .
for @xmath193 the expected reduction of the results for curved interfaces to those for a flat interface occurs only for very small angular separations .
this is not the case for quadrupoles ( @xmath194 ) , for which at all separations there is no qualitative discrepancy and only a minor quantitative difference from the flat limit .
[ l][l][1]@xmath56 [ l][l][1]@xmath195 [ l][l][1]@xmath196 [ l][l][1]@xmath197 flat [ c][c][1]@xmath198 [ l][l][1]@xmath159 [ l][l][1]@xmath178 [ l][l][1]@xmath194 fig02 ( 65,8.5))-([df22-final ] ) , solid lines ) for two capillary multipoles of the same order @xmath199 , each shifted as a whole such that @xmath200 .
the dashed lines correspond to the following expressions : @xmath201/(2\pi)$ ] for @xmath159 , @xmath202/(2\pi)$ ] for @xmath178 , and @xmath203/(4\pi)$ ] for @xmath194 , representing for comparison the case of a flat interface with the equivalent spatial separation @xmath192 of the particles .
as expected , for @xmath204 each full curve approaches its corresponding dashed curve .
the green dotted lines correspond to three metastable branches of the free energy for @xmath178 ( see eq .
( [ df11-potential ] ) ) with two relevant intersection points ( at @xmath183 , where @xmath185 , and @xmath205 , where @xmath184 ) indicated by filled squares . in each interval
@xmath206 $ ] , @xmath207 $ ] , and @xmath208 $ ] the full green curve corresponds to the lowest of the three branches and as such describes thermal equilibrium.,title="fig:",scaledwidth=15.0% ] [ l][l][1]@xmath56 [ c][c][1]@xmath209 [ l][l][1]@xmath210 [ l][l][1]@xmath211 [ l][l][1]@xmath212 [ l][l][0.9]@xmath213 [ l][l][0.9]@xmath214 [ l][l][0.9]@xmath215 [ l][l][0.9]@xmath39 [ l][l][0.9]@xmath216 [ c][c][0.9]@xmath0 [ l][l][0.9]@xmath217 [ c][c][1]@xmath218 [ l][l][1 ] [ l][l][1]point - quadrupoles [ l][l][1]@xmath219 [ l][l][1]@xmath220 fig04 ( 53,48 ) for two ellipsoidal particles of aspect ratio @xmath221 , contact angle @xmath222 , various droplet radii @xmath0 , and two different orientational configurations : tip - to - tip and side - to - side ( as visualized by the snapshots from the finite - element free energy minimization ) . in view of large relative numerical errors for small values of @xmath223
the zero of the free energy is taken as the mean value of the data points with @xmath224 ( the error bars are comparable or smaller than the size of the symbols ) .
the thick solid line corresponds to eq .
( [ df22-final ] ) with @xmath225 or @xmath226 ( in either case eq .
( [ df22-final ] ) yields the same free energy ) .
the capillary quadrupole of each of the particles is approximated by the expression @xmath227 ; this relation is valid for a flat interface @xcite at large interparticle separations , with the undulation @xmath228 of the interface taken to be the one at the surface of a single ellipsoidal particle . as an approximation for a droplet
we adopt the same relation , @xmath229 , with the undulation @xmath230 of the interface at the bottom part of a particle centered at the north pole of the droplet ( see the green part of the upper inset ) .
this undulation has been calculated numerically for three droplet radii : @xmath231 , @xmath232 , @xmath233 .
@xmath0 is the radius of the droplet _ without _ the particles .
the immersion @xmath216 and the tilt angle @xmath215 of the long axis of each particle @xmath101 with respect to the plane perpendicular to the direction @xmath39 are subjected to minimization .
the colored lines end at small values of @xmath56 because the particle surfaces touch each other for nonzero @xmath56.,title="fig:",scaledwidth=14.2% ] ( 70,33 ) for two ellipsoidal particles of aspect ratio @xmath221 , contact angle @xmath222 , various droplet radii @xmath0 , and two different orientational configurations : tip - to - tip and side - to - side ( as visualized by the snapshots from the finite - element free energy minimization ) .
in view of large relative numerical errors for small values of @xmath223 the zero of the free energy is taken as the mean value of the data points with @xmath224 ( the error bars are comparable or smaller than the size of the symbols ) .
the thick solid line corresponds to eq .
( [ df22-final ] ) with @xmath225 or @xmath226 ( in either case eq .
( [ df22-final ] ) yields the same free energy ) .
the capillary quadrupole of each of the particles is approximated by the expression @xmath227 ; this relation is valid for a flat interface @xcite at large interparticle separations , with the undulation @xmath228 of the interface taken to be the one at the surface of a single ellipsoidal particle . as an approximation for a droplet
we adopt the same relation , @xmath229 , with the undulation @xmath230 of the interface at the bottom part of a particle centered at the north pole of the droplet ( see the green part of the upper inset ) .
this undulation has been calculated numerically for three droplet radii : @xmath231 , @xmath232 , @xmath233 .
@xmath0 is the radius of the droplet _ without _ the particles .
the immersion @xmath216 and the tilt angle @xmath215 of the long axis of each particle @xmath101 with respect to the plane perpendicular to the direction @xmath39 are subjected to minimization .
the colored lines end at small values of @xmath56 because the particle surfaces touch each other for nonzero @xmath56.,title="fig:",scaledwidth=6.75% ] ( 70,15 ) for two ellipsoidal particles of aspect ratio @xmath221 , contact angle @xmath222 , various droplet radii @xmath0 , and two different orientational configurations : tip - to - tip and side - to - side ( as visualized by the snapshots from the finite - element free energy minimization ) . in view of large relative numerical errors for small values of @xmath223
the zero of the free energy is taken as the mean value of the data points with @xmath224 ( the error bars are comparable or smaller than the size of the symbols ) .
the thick solid line corresponds to eq .
( [ df22-final ] ) with @xmath225 or @xmath226 ( in either case eq .
( [ df22-final ] ) yields the same free energy ) .
the capillary quadrupole of each of the particles is approximated by the expression @xmath227 ; this relation is valid for a flat interface @xcite at large interparticle separations , with the undulation @xmath228 of the interface taken to be the one at the surface of a single ellipsoidal particle . as an approximation for a droplet
we adopt the same relation , @xmath229 , with the undulation @xmath230 of the interface at the bottom part of a particle centered at the north pole of the droplet ( see the green part of the upper inset ) .
this undulation has been calculated numerically for three droplet radii : @xmath231 , @xmath232 , @xmath233 .
@xmath0 is the radius of the droplet _ without _ the particles .
the immersion @xmath216 and the tilt angle @xmath215 of the long axis of each particle @xmath101 with respect to the plane perpendicular to the direction @xmath39 are subjected to minimization .
the colored lines end at small values of @xmath56 because the particle surfaces touch each other for nonzero @xmath56.,title="fig:",scaledwidth=6.75% ] ( 86,33 ) for two ellipsoidal particles of aspect ratio @xmath221 , contact angle @xmath222 , various droplet radii @xmath0 , and two different orientational configurations : tip - to - tip and side - to - side ( as visualized by the snapshots from the finite - element free energy minimization ) . in view of large relative numerical errors for small values of @xmath223
the zero of the free energy is taken as the mean value of the data points with @xmath224 ( the error bars are comparable or smaller than the size of the symbols ) .
the thick solid line corresponds to eq .
( [ df22-final ] ) with @xmath225 or @xmath226 ( in either case eq .
( [ df22-final ] ) yields the same free energy ) .
the capillary quadrupole of each of the particles is approximated by the expression @xmath227 ; this relation is valid for a flat interface @xcite at large interparticle separations , with the undulation @xmath228 of the interface taken to be the one at the surface of a single ellipsoidal particle . as an approximation for a droplet
we adopt the same relation , @xmath229 , with the undulation @xmath230 of the interface at the bottom part of a particle centered at the north pole of the droplet ( see the green part of the upper inset ) .
this undulation has been calculated numerically for three droplet radii : @xmath231 , @xmath232 , @xmath233 .
@xmath0 is the radius of the droplet _ without _ the particles .
the immersion @xmath216 and the tilt angle @xmath215 of the long axis of each particle @xmath101 with respect to the plane perpendicular to the direction @xmath39 are subjected to minimization .
the colored lines end at small values of @xmath56 because the particle surfaces touch each other for nonzero @xmath56.,title="fig:",scaledwidth=3.36% ] next , we compare our approximate analytic theory with the numerical minimization of the free energy for floating ellipsoidal particles with semi - axes @xmath122 and @xmath234 , which we performed by using a finite element method @xcite . the numerical procedure consists of minimizing the free energy functional @xmath235=\\ = \gamma s_{lg } -\gamma\sum_{i=1,2}\cos\theta_{p , i}s_{pl , i } - \lambda(v - v_l ) , \label{functional } % \end{equation } \end{gathered}\ ] ] with respect to the shape of the interface @xmath236 , immersions of the particles @xmath237 , and their tilts @xmath238 ( relative to planes perpendicular to the directions @xmath85 and @xmath59 , respectively , see the inset in fig .
[ fig : numerical ] ) , while the angular separation @xmath56 and the tangential orientations @xmath180 are kept fixed ; @xmath239 is the area of the liquid - gas interface , and @xmath240 and @xmath241 are the area of the particle - liquid interface and the contact angle , respectively , of particle @xmath101 .
the last term in eq .
( [ functional ] ) ensures conservation of the liquid volume @xmath242 with @xmath243 as the corresponding lagrange multiplier . the effective potentials ( fig . [ fig : numerical ] ) obtained numerically for prolate spheroids ( i.e. , elongated spheres ) with aspect ratio @xmath221 and droplet radii @xmath244 differ for particles oriented tip - to - tip ( @xmath226 ) and side - to - side ( @xmath225 ) .
this signals the importance of capillary multipoles higher than only a quadrupole , because the point - quadrupole approximation yields the same free energy for both orientations ( see eq .
( [ df22-final ] ) ) .
this discrepancy is not surprising because the longer particle axes ( which set the length scale @xmath122 ) are actually not very small compared with the droplet radius @xmath0 and thus in this case the point - particle limit is a rather crude approximation .
our numerical analysis is restricted to such small droplets due to the limited effectiveness of the finite - element method for larger drops caused by too large numerical errors .
for the same reason we consider only particles with a large aspect ratio ( @xmath221 ) for which the deformation of the interface is sufficiently large so that the numerical noise does not smear out the angular dependence of the free energy . nonetheless , even in this regime our analytic theory captures the main qualitative features of the effective interaction potential between floating ellipsoids .
presumably the reason for this is that even for large particles the droplet shape deviates only slightly from a sphere .
( due to a fixed volume of the liquid larger particles cause a larger displaced liquid volume @xmath245 ; primarily this leads to an increased effective radius @xmath246{v_l+v_{displ}}$ ] of the droplet and thus generates only a constant in the deformation field @xmath29 . ) accordingly , away from the particles our perturbation theory remains applicable .
thus even for large particles the deformation of the droplet can be approximated by the superposition of contributions from pointlike quadrupoles substituting the actual particles .
the additional dependence on the orientations revealed by our numerical calculations , which goes beyond the quadrupole approximation , resembles the well established case of a flat interface , which has been analyzed experimentally @xcite and theoretically @xcite in terms of a multipole expansion .
those studies showed that the point - quadrupole approximation , within which tip - to - tip and side - to - side configurations have the same free energy , is reliable only either for particles with a very small eccentricity ( @xmath247 ) or at large spatial separations @xmath248 .
otherwise , one should rather consider an expansion in terms of elliptic coordinates , from which it follows that the tip - to - tip configuration is energetically more favourable .
this prediction agrees with the experimental observations @xcite .
our calculations indicate that the same configuration is the preferred one also in the case of spherical interfaces .
however , we have not analyzed all the possible orientations .
it is noteworthy that optimal orientations other then tip - to - tip and side - to - side have been observed at a flat interface for a pair of ellipsoidal particles of _ different _ sizes @xcite .
we can also discuss the effective capillary potential @xmath249 for a single capillary quadrupole on a _ sessile _ , spherical caplike droplet forming a contact angle @xmath250 with the supporting substrate , where @xmath95 is the angular position of the particle ( @xmath251 ) measured from the droplet symmetry axis and relative to the center of the sphere circumscribing the droplet , and @xmath55 the orientation of the particle .
as before the tilt angle @xmath252 is taken to be the optimal one . in the case
@xmath253 the method of images can be applied , analogous to the case of monopoles @xcite .
the substrate acts like a mirror and the free energy of a _ single _ particle equals one half of the energy of _ two _ particles on a _ full _ droplet , i.e. , @xmath254 where the @xmath255 sign corresponds to a free contact line at the substrate and the @xmath256 sign to a pinned one .
( since the particle is force - free , there are no additional images , distinct from the case of monopoles , see ref .
@xcite . ) for a free contact line eq .
( [ eq : dfell ] ) holds in general , i.e. , for arbitrary particles ( not necessarily ellipsoidal ) .
in such a case @xmath257 is replaced by the actual interaction energy @xmath258 between the particles ( possibly containing also higher multipoles ) . on the other hand , for a pinned contact line eq .
( [ eq : dfell ] ) holds only for pointlike particles in conjunction with eq .
( [ df22-final ] ) . for a pointlike quadrupole , according to the expression for @xmath257 in eq .
( [ df22-final ] ) , the dependence of @xmath259 on the orientation @xmath55 drops out so that within this approximation the particle is always attracted to a free contact line and repelled from a pinned one .
thus the effective potential is given by the black line in fig .
[ fig : numerical ] with @xmath260 ( with an additional minus sign in the case of a pinned contact line ) .
in the case of an actual ellipsoidal particle the free energy is no longer given by the rhs of eq .
( [ df22-final ] ) so that the dependence on @xmath55 does no longer drop out but becomes important in the vicinity of the contact line at the substrate . as stated above , if this contact line is _ free _ ( still assuming @xmath253 ) , due to symmetry arguments the free energy of a single particle on a sessile drop equals one half of the energy for two particles on a full droplet ( as predicted by the method of images for the more restrictive case of pointlike quadrupoles ) , for which the full numerical results are plotted in fig .
[ fig : numerical ] .
accordingly , for a fixed angular position @xmath261 of the particle its orientation perpendicular to the contact line ( solid lines in fig .
[ fig : numerical ] ) is energetically more favorable .
however , the global minimum of the free energy corresponds to the particle touching the substrate ( i.e. , @xmath262 so that @xmath204 ) and being oriented parallel to the contact line ( dashed lines in fig . [
fig : numerical ] for @xmath263 ) .
however , we can not judge if the latter configuration remains the optimal one for arbitrary aspect ratios and contact angles at the particle .
( this and analyzing the influence of the contact angle at the substrate actually call for a separate numerical study . )
if the contact line at the substrate is _ pinned _ , the global free energy minimum corresponds to the particle being positioned at the drop apex . in such a case , due to a large distance from the contact line with the substrate , the pointlike quadrupole approximation remains reliable .
one can estimate the depth of this effective confining potential as follows . according to eqs .
( [ eq : dfell ] ) and ( [ df22-final ] ) with @xmath227 and assuming @xmath264 ( which applies to an ellipsoidal particle of aspect ratio @xmath221 and contact angle @xmath222 , see the caption of fig . [
fig : numerical ] ) , the displacement of the particle , e.g. , from @xmath265 to @xmath266 , leads to a free energy increase of approximately @xmath267 for a droplet as large as @xmath268 ( assuming a surface tension @xmath269 which corresponds to a water droplet in air ) .
hence , one could experimentally measure the capillary confining potential , e.g. , by observing the brownian motion of micron - sized particles on sessile drops . finally , we emphasize that the effects of confinement should be also taken into account in determining the effective _ pair _ potentials between particles on _ sessile _ drops ( for the case of monopoles see ref .
in particular , the confinement leads to closer packing of particles on a droplet , which should not be mistaken for an additional attractive pair - interaction .
on of the authors ( j.g . ) acknowledges financial support from the polish ministry of science ( 2010/2011 ) through grant nr ip2010 012270 .
in eq . ( [ fn1n2 ] ) for @xmath175
the following series appears : @xmath270 where @xmath271 or @xmath272 , and for @xmath188 : @xmath273 where @xmath274 or @xmath275 .
those series can be evaluated using the generating function @xcite : @xmath276 where @xmath277 and @xmath278 .
one finds the relationships @xmath279\end{gathered}\ ] ] and @xmath280 which can be evaluated with the results @xmath281 and @xmath282 taking @xmath283 and inserting these results into eq .
( [ fn1n2 ] ) finally yields eqs .
( [ df11-final])-([df22-final ] ) .
|
the effective capillary interaction potentials for small colloidal particles trapped at the surface of liquid droplets are calculated analytically .
pair potentials between capillary monopoles and dipoles , corresponding to particles floating on a droplet with a fixed center of mass and subjected to external forces and torques , respectively , exhibit a repulsion at large angular separations and an attraction at smaller separations , with the latter resembling the typical behavior for flat interfaces .
this change of character is not observed for quadrupoles , corresponding to free particles on a mechanically isolated droplet .
the analytical results for quadrupoles are compared with the numerical minimization of the surface free energy of the droplet in the presence of ellipsoidal particles .
|
[ intro ] the movement of droplets in external gradients fascinates scientists and layman alike at least since newton s description @xcite of hauksbee s experiment with drops of orange oil that move between two non - parallel glass plates towards the point of smallest plate distance @xcite . in another example
a drop of liquid freely immersed in another liquid subject to a temperature gradient will move towards the higher temperature region due to marangoni forces caused by surface tension gradients @xcite .
a drop sitting on a solid substrate also moves in a temperature @xcite or wettability @xcite gradient .
especially the marangoni force is already used to manipulate droplets , for example in light induced drop movement @xcite .
similar concepts of a directed movement of small amounts of soft matter in a given gradient are also realized in models of cell motility @xcite
. however , even more intricate are situations where matter spontaneously starts a directed movement in initially homogeneous settings .
small pieces of camphor that move on a liquid surface by emitting a surfactant have also been studied for centuries @xcite . more recently
oil droplets containing volatile additives and interacting droplets of different volatile oils have been reported to move on solid surfaces due to the solutal marangoni effect caused by evaporation / condensation @xcite . also intricate
is the spontaneous movement of a juxtaposed pair of droplets with different wetting properties , so called bi - slugs , along a capillary tube @xcite .
this effect was already mentioned by marangoni @xcite .
another example are drops immersed in a second liquid .
if the drops contain a soluble surfactant undergoing an isothermal chemical reaction at their surface the drops may start to move @xcite .
apart from chemical reactions , drop movement can also be driven by surface phase transitions @xcite .
the movement is possible because such _ active _ drops change their surrounding and produce a gradient that drives their motion .
[ sketch ] recent experiments also found chemically driven running droplets on solid substrates @xcite . in these cases small droplets of solution are put on partially wettable substrates .
a droplet changes the substrate by adsorption or desorption of a solute rendering the substrate underneath the droplet less wettable than the bare substrate .
the radial symmetry still assures an equilibrium position that is , however , unstable . as a result
, fluctuations break the symmetry and the drop starts to move in a self - sustained manner .
in the course of its movement it changes the substrate and leaves a less wettable trail behind ( see fig.[sketch ] ) .
we distinguish two types of experiments : in case i the substrate is changed in an irreversible way whereas in case ii it recovers its initial state after the droplet has passed .
the droplet does not rest again and follows a random trajectory until trapping itself between self - produced less wettable patches ( type i ) @xcite .
alternatively , a running drop may perform a periodic movement ( type ii ) @xcite .
the droplet often moves with nearly constant velocity for a rather long time ( tens of minutes ) .
however , in all cases the motion ceases when the overall system reaches thermal equilibrium .
next , we shortly recall the specific results of different sets of experiments , yielding running droplets . in ref.@xcite droplets of @xmath0-alkanes ( @xmath0-octane and @xmath0-dodecane ) are used that contain 1,1,2,2-tetrahydroperfluorodecyltrichlorosilane ( cf@xmath1(cf@xmath2)@xmath3(ch@xmath2)@xmath2sicl@xmath1 ) .
the silane molecules form dense crafted monolayers on silicon or glass and render these surfaces hydrophobic .
the droplet motion is stalled when no further hydrophilic surface is available .
no limitation due to silane depletion inside the droplet is observed . for millimeter
sized drops ( smaller than the capillary length ) droplet velocities between 1mm / s and 10cm / s are observed , depending on liquid viscosity , silane concentration and droplet size .
the velocity increases with the silane concentration and the droplet size .
the reaction rate is estimated to be 1100s@xmath4mol@xmath4 .
typical reaction times are given as 0.010.2s .
another experiment is performed in a chemically different system @xcite using millimeter - sized droplets of a nonpolar solution of @xmath0-alkylamine ( 1 mm solution of c@xmath5nh@xmath2 or c@xmath6nh@xmath2 ) in decahydro - naphthalene ( @xmath7mj / m@xmath8 , @xmath9ns / m@xmath8 ) .
therein , silicon substrates with microprinted high - energy surfaces are employed , that expose a dense packing of carboxylic acid functionalities ( co@xmath2h ) .
the amines dissolved in the droplet adsorb at the substrate and produce a surface of lower energy exposing methyl groups . the effect of different adsorbates @xcite and reaction kinetics @xcite is investigated . the velocity decreases with increasing droplet size .
the only example of a type ii experiment @xcite features droplets of oil ( 5 mm iodine solution of nitrobenzene saturated with potassium iodide ) on glass substrates immersed in aqueous solution of a cationic surfactant ( 1 mm stearyl trimethyl ammonium chloride ) . in this system the stearyl trimethyl ammonium ion ( sta ) absorbs at the glass substrate outside the oil droplet and renders the glass lyophilic .
when the oil droplet moves on top of the coating the sta desorbs into the oil . in this way
a wettability contrast between front and back of the moving droplet is created and sustained .
in contrast to the type i case the substrate recovers its lyophilic state soon after the droplet has passed , because new sta absorbs at the glass . the aqueous phase can be seen as an infinite reservoir of adsorbent .
movement only finally stops when the oil droplet is saturated with sta .
similar phenomena can be seen in metallurgic systems where droplets of liquid metals or alloys react with the metallic substrate , for instance , by alloying .
the layer between the droplet and the substrate may be less wettable than the bare substrate resulting in the migration of reactive islands .
this was studied for tin islands on copper surfaces that move and leave tracks of bronze behind @xcite .
contrariwise , the layer can be more wettable than the bare substrate .
this is typical for the related process of reactive spreading ( also called reactive wetting ) where a droplet of liquid on a ( nearly ) non - wettable surface starts to spread after forming a more wettable layer underneath @xcite .
variants are possible in which during spreading the substrate becomes less wettable in the center of the drop @xcite . however , reactive spreading processes do normally not result in running droplets ( but see the suddenly displaced droplets in ref.@xcite ) . for type
i experiments an implicit equation for the velocity @xmath10 of the droplet was derived @xcite from a simple theoretical argument .
based on a balance of friction force and driving capillary forces one obtains @xmath11 , where @xmath12 is the reaction rate , @xmath13 the size of the droplet and @xmath14 a constant .
the dynamic contact angles at the advancing and receding ends of the droplet are then assumed to be identical ( @xmath15 ) , i.e.the droplet profile is approximated by a spherical cap with @xmath15 given by @xmath16 .
the static contact angles at the advancing edge @xmath17 and at the receding edge @xmath18 are different due to the chemical gradient .
the expression for the velocity is found from a first order reaction on the substrate that yields chemical concentrations @xmath19 and @xmath20 at the respective ends of the droplet .
the expression for the velocity predicts a monotone increase of the droplet velocity with the droplet length @xmath13 and the reaction rate @xmath12 in line with experimental observations @xcite .
the droplet velocity in the limiting case of a saturated chemical reaction was also given in @xcite .
however , the experiments of ref.@xcite show the opposite trend ; the velocity decreases with increasing drop sizes and effective reaction rate . in the framework of ref.@xcite the decrease
is explained as a result of the flattening of the drops by gravity .
related works discuss running droplets in a random medium @xcite and the forced wetting of a plate immersed into a reactive fluid @xcite . in this paper we propose and analyze dynamical models for self - propelled running droplets for both , type i and type ii experiments .
our models consist of coupled evolution equations for the film thickness profile and the substrate coverage .
thereby the wettability of the substrate is modeled by a coverage dependent disjoining pressure .
the different types of experiments are taken into account by incorporating different reaction terms .
the models are capable of reproducing the different experimentally found regimes .
a short account of a variant of the model for the type i experiments , i.e. for the case of a irreversibly changed substrate was recently presented @xcite .
there it was found that the dynamic contact angles at the advancing and receding ends of the droplet are not identical but rather resemble the corresponding static ones . even for a strong driving force
the deviation between dynamic and static angles is only about 10 percent .
the bifurcation leading from sitting to running droplets for a finite reaction rate was identified as a drift - pitchfork bifurcation .
here we analyze the type i model in more detail and extend our analysis to the type ii model .
in section [ model ] the type i and type ii dynamical models are presented .
results for the two models are discussed in section [ ad ] and [ addes ] , respectively .
stationary running droplets and sitting droplets are characterized in dependence of the control parameters reaction rate , droplet volume , diffusion constant of the adsorbate field and desorption rate of the adsorbate ( for type ii model only ) .
the resulting families of solutions and their linear stability are used to derive phase diagrams that describe the existence regions for running and sitting droplets . to further elucidate the drift - pitchfork bifurcation that mediates the transition from sitting to running droplets also in the type ii model we analyze the linear modes that destabilize sitting droplets close to the bifurcation . before we conclude in section [ conc ] , we use numerical simulations to illustrate and analyze periodic droplet motion that is possible in the type ii model .
the simulations are compared with our continuation results and the type ii experiment of ref.@xcite .
[ model ] [ evequ ] the evolution of a thin liquid film on a horizontal smooth solid substrate is described by an equation for the thickness profile @xmath21 derived from the navier - stokes equations using long - wave or lubrication approximation @xcite @xmath22 the parameters @xmath23 and @xmath24 are the surface tension and viscosity of the liquid , respectively .
they define the viscous time scale @xmath25 , where @xmath13 is a typical length for the system .
the change in time of the film thickness profile equals the gradient of a flow that results as the product of a mobility and a pressure gradient .
the mobility @xmath26 corresponds to a parabolic velocity profile @xmath27 within the film .
the velocity component orthogonal to the substrate @xmath28 can be obtained using the continuity equation @xmath29 the pressure @xmath30 contains the curvature ( or laplace ) pressure @xmath31 and the disjoining pressure @xmath32 .
the latter comprises effective molecular interactions between the film surface and the substrate and accounts for the wetting properties of the substrate @xcite . as discussed in detail below , here a mathematically simple function @xmath32
is used that is common in the literature . in many situations
the qualitative outcome only depends on very general characteristics of the disjoining pressure @xcite .
the used form allows for solutions of eq.([film ] ) that represent static ( i.e. sitting ) droplets with a finite mesoscopic equilibrium contact angle .
the disjoining pressure is chosen such that the droplets coexist with an ultrathin precursor film .
the evolution of the density of the adsorbed layer on the substrate determines the wettability and is modeled by a reaction - diffusion equation for the dimensionless field @xmath33 @xmath34 where the function @xmath35 describes adsorption or desorption on the substrate .
the second term allows for a diffusion of the chemical species along the substrate . for simplicity
we assume that the adsorbate has the same diffusion constant on the bare and the droplet covered substrate .
however , assuming different constants would not change the presented results qualitatively .
note that for type i experiments @xmath36 directly corresponds to the substrate coverage .
however , in a type ii experiment the droplet dissolves a more wettable coating . in this case
the coverage corresponds to @xmath37 . also here
one can use eq.([rd ] ) , however , the signs of adsorption and desorption term are switched . with this convention , in both cases
the wettability decreases with increasing @xmath36 .
one can neglect the dynamics of the concentration field of the chemical in the bulk of the droplet by assuming a fast equilibration of the solute concentration within the moving droplet as compared to the reaction at the substrate .
the fast equilibration is caused by diffusion and convective motion within the droplet whereas the latter is driven by the lateral movement of the droplet along the substrate .
it corresponds to the limit of a small damkhler number da@xmath38 giving the ratio of reaction velocity at the substrate and diffusion velocity in the droplet @xcite .
the parameter @xmath12 is a typical reaction rate , @xmath39 stands for a typical concentration of the chemical species in the droplet and @xmath40 is the diffusion constant in the droplet .
[ reaction ] corresponding to the two different sets of experiments that the model shall describe we use two different reaction terms . for type
i the initial coverage of the substrate is zero ( i.e. @xmath41 ) and we only allow for adsorption underneath the droplet @xmath42 the reaction saturates at @xmath43 because this is the maximal possible coverage .
the function @xmath44 represents a ( smooth ) step function that approaches one and zero inside and outside the droplet , respectively . for type ii the rest state of the substrate without any droplet is the fully covered substrate ( again corresponding to @xmath41 , see above section [ evequ ] ) and we allow for desorption underneath the droplet and adsorption outside the droplet @xmath45\,\phi .
{ \label{reac2}}\ ] ] the time scales of the reactions at the substrate inside and outside the droplet are defined by the effective rate constants @xmath46 and @xmath47 , respectively . note that they have the dimension s@xmath4 .
the function @xmath44 may be the step function @xmath48 or the smooth function @xmath49 + 1\}/2 .
{ \label{xi2}}\ ] ] the value of @xmath50 is chosen slightly larger than the thickness of the precursor film and the maximal drop height is always @xmath51 .
a small value of @xmath52 ensures that the switch between a predominant adsorption reaction and a predominant desorption reaction occurs over a small film thickness range .
note , that @xmath53 for @xmath54 .
changes in the details of the reaction term do not affect the results qualitatively .
[ disjpress ] for the disjoining pressure we use throughout the present work @xmath55 where @xmath56 nm is the born repulsion length that defines a lower cut - off for the film thickness .
the parameter @xmath57 and @xmath58 are the long- and short - range components of the total spreading coefficient @xmath59 ( for @xmath41 ) .
we use @xmath60 and @xmath61 corresponding to a destabilizing long - range van der waals and a stabilizing short - range interaction @xcite . for @xmath41
the pressure allows for drops sitting on a stable precursor film as can be seen studying the corresponding densities of the excess surface energy @xmath62 , related to the disjoining pressures by @xmath63 .
the short - range interaction contains the influence of the coating and crucially influences the static contact angle @xcite . to account for the varying wettability caused by the different substrate coverage we let the short - range part of the spreading coefficient @xmath58 depend linearly on the field @xmath64 .
the signs are chosen in a way that for both types of experiments @xmath65 assures that an increase in @xmath36 corresponds to a lower wettability , i.e. to a larger equilibrium contact angle @xmath66 given by @xmath67 @xcite .
the constant @xmath68 relates the coverage to the wettability and therefore defines the magnitude of the possible wettability gradient .
note , that our eq.([pi2 ] ) corresponds to the linear relation between @xmath69 and @xmath36 assumed in ref.@xcite .
[ dimless ] we rewrite eqs.([film ] ) to ( [ pi2 ] ) by introducing scales @xmath70 , @xmath71 , and @xmath72 for @xmath73 , @xmath74 , and @xmath75 , respectively
. then one obtains @xmath76 for the scaled spreading coefficient @xmath77 . as length scale
@xmath72 we use the value of the film thickness where the local free energy @xmath78 has its minimum , i.e. , where @xmath79 .
this gives @xmath80 and implies that the ratio of the strength of the two antagonistic effective molecular interactions is only an implicit parameter of the system .
the length @xmath72 also corresponds to the thickness of the precursor film for a sitting drop on the bare substrate . defining the dimensionless overall reaction rate @xmath81 , diffusion constant @xmath82 , and ratio of reaction rates @xmath83
, we obtain from eqs.([film ] ) to ( [ pi2 ] ) the dimensionless coupled evolution equations for the thickness profile @xmath75 and the field @xmath36 @xmath84\right\ } { \label{sys1}}\\ \partial_t\,\phi\,&=&\,r\,r(h,\phi)\,+\,d\,\delta\phi { \label{sys2}}\end{aligned}\ ] ] with the disjoining pressure @xmath85 and the options for the reaction term @xmath86\,\phi .
{ \label{r2}}\end{aligned}\ ] ] to give an impression of the influence of the coverage we give in fig.[sitting ] droplet profiles for different values of the coverage @xmath36 .
thereby , @xmath36 is assumed to be independent of time and constant along the substrate .
increasing the ratio @xmath87 clearly leads to increasing contact angles and decreasing droplet length . as given in the legend .
the coverage is assumed to be independent of time and constant along the substrate .
the system size is @xmath88 and all droplets have the volume @xmath89 . ]
[ sitting ] in the following we explore the full non - linear behavior of the dynamical models presented above .
the main interest lies in an exploration of the existence regions of qualitatively different solutions depending on control parameters like reaction rate @xmath12 , ratio of desorption and adsorption rates @xmath90 , and droplet volume @xmath91 .
it was already shown that a type i model allows for three - dimensional running droplets that move with constant velocity @xmath10 and droplet shape @xcite . here ,
we restrict our attention to two - dimensional drops to be able to explore a large part of the parameter space . to determine such droplets we therefore use eqs.([sys1],[sys2 ] ) replacing @xmath92 by @xmath93
. then continuation techniques @xcite are employed to calculate two - dimensional running droplets moving with constant speed .
this is achieved by switching to the comoving frame @xmath94 and imposing appropriate boundary conditions . in the comoving frame
running droplets correspond to steady solutions .
integration in time of the full system is also used in some cases .
details on all numerical techniques used can be found in the appendix .
[ ad ] we start with model type i , where adsorption takes place only underneath the droplet , i.e. we combine the evolution equations for the film thickness profile ( [ sys1 ] ) and the field of adsorbate coverage ( [ sys2 ] ) with the reaction term ( [ r1 ] ) and the disjoining pressure ( [ pi2b ] ) .
first we focus on the characterization of the changing solution behavior in dependence of the overall reaction rate @xmath12 . without diffusion along the substrate ( @xmath95 ) one finds unstable sitting droplets and stable running droplets for all @xmath12 .
the droplets and the coverage profile move with constant velocity @xmath10 and constant shape .
we emphasize that this state corresponds to a subtle dynamical equilibrium given that coverage profile and droplet velocity depend on each other .
fig.[m1prof ] shows for two different reaction rates profiles of moving droplets [ ( a ) and ( b ) ] where the streamlines correspond to contour lines of the stream function @xmath96 and indicate the flow in the comoving coordinate system .
note that the velocity fields are obtained by @xmath97 . also shown
are the corresponding profiles for the coverage @xmath36 [ ( c ) and ( d ) ] .
the two sets of profiles belong to two qualitatively different regimes that are prominently visible in the profiles of the coverage ( corresponding to the results obtained for a different disjoining pressure in @xcite ) . and
( c , d ) the substrate coverage @xmath36 for two different reaction rates @xmath12 .
droplets on the left and right correspond to the non - saturated ( @xmath98 ) and saturated ( @xmath99 ) regime and move with velocities @xmath100 and @xmath101 , respectively .
the streamlines plotted in ( a ) and ( b ) indicate the convective motion inside the droplets in the comoving frame and have the spacing @xmath102 and @xmath103 , respectively .
the direction of the movement is indicated by arrows .
the remaining parameters are @xmath104 , @xmath105 , @xmath88 , @xmath106 , @xmath107 , @xmath108 and the droplet volume is @xmath89 . ]
[ m1prof ] for low reaction rates [ fig.[m1prof](a ) and ( c ) ] the coverage starts to increase at the advancing contact zone and continues to increase up to the receding contact zone where it is still well below the saturation value of @xmath43 .
we call this the _ non - saturated regime_. for high reaction rates [ fig.[m1prof](b ) and ( d ) ] the coverage starts to increase at the advancing contact zone as in the non - saturated regime
. however , it increases much faster and reaches the saturation value @xmath43 already underneath the droplet . at the receding contact zone
it is always at the saturation value .
we call this the _ saturated regime_. the droplets in the two regimes behave qualitatively different when the reaction rate is changed as shown in fig.[m1vrcr ] .
( logarithmic scale ) for different strength of the diffusion of the coverage @xmath36 along the substrate @xmath109 . shown
are ( a ) the velocity @xmath10 for different diffusion constants @xmath109 ( see legend ) ; and ( b ) the advancing ( @xmath110 , dashed lines ) and the receding ( @xmath111 , solid lines ) dynamic contact angles for @xmath105 ( heavy lines ) and @xmath112 ( thin lines ) .
the arrows with the labels @xmath113 , @xmath114 and @xmath115 in ( a ) indicate the locations for the maximum velocity , drift - pitchfork bifurcation and the saddle - node bifurcation , respectively , for @xmath112 .
the remaining parameters are as in fig.[m1prof ] . ]
[ m1vrcr ] for low reaction rates in the non - saturated regime the coverage does not reach the saturation value @xmath43 at the rear of the droplet .
this implies that the driving wettability gradient between front and rear of the droplet has not yet reached its maximally possible value .
therefore an increase in the reaction rate leads to a steeper increase in the spatial profile of the coverage underneath the droplet . in consequence a larger wettability gradient results implying a larger velocity
however , again it is subtle to determine the dynamic equilibrium .
first , the higher velocity reduces the contact time of the droplet with a given point of the substrate .
second , it also reduces the length of the droplet ( see below ) .
both effects restrict the growth of the wettability gradient .
the resulting increase of the velocity with increasing reaction rate can be seen in fig.[m1vrcr](a ) for @xmath116 .
note that it does only weakly depend on the diffusion along the substrate .
fig.[m1vrcr](b ) shows the corresponding dynamical contact angles at the advancing and receding edges of the moving droplets .
below @xmath117 both angles increase , implying a decrease in the length of the droplet ( that has constant volume ) . because in the non - saturated regime
the coverage in the contact zone at the advancing edge does effectively not depend on @xmath12 , the increase in the advancing contact angle is caused by the larger velocity only .
seeing the full dynamic contact angle at a moving contact line as a superposition of a static ( or equilibrium ) part and a dynamic part , one can say that only the dynamic part of the advancing contact angle increases with @xmath12 .
the increase at the front is in line with the increase with velocity found for sliding drops on an incline ( see for example ref.@xcite ) .
however , at the receding edge the coverage in the contact zone depends strongly on @xmath12 .
therefore the rather strong increase in the receding contact angle with increasing @xmath12 is caused by changes in both , the static and the dynamic part .
for a larger @xmath12 the wettability at the back is smaller , i.e. the local equilibrium contact angle increases .
however , the dynamic receding contact angle normally decreases with increasing velocity @xcite .
this implies that the increase in the receding angle found here results from an increase of the static part only .
it is even counteracted by a decrease of the dynamic part .
also in the contact angles , a relatively small @xmath109 has no visible influence in the non - saturated regime .
the maximum wettability gradient , i.e. driving force , and therefore the maximum droplet velocity is reached for the bare substrate ( @xmath41 ) at the advancing edge and maximum coverage ( @xmath43 ) at the receding edge .
the point of largest velocity [ fig.[m1vrcr](a ) ] corresponds to the point of the largest difference between the receding and advancing contact angles [ fig.[m1vrcr](b ) ] .
we call the corresponding reaction rate @xmath113 .
next we discuss the behavior in the saturated regime ( @xmath118 in fig.[m1vrcr ] ) . for larger reaction rates
the coverage at the rear of the droplet remains at its saturation value .
however , the driving force and in consequence the velocity decrease slightly with increasing @xmath12 ( cf . also ref.@xcite ) .
note , that for small @xmath109 in fig.[m1vrcr](a ) this is barely visible .
the decrease from @xmath113 to @xmath119 corresponds to about 2% of the maximal velocity .
as detailed next this slight decay is caused by the dynamics in the advancing contact zone . in the saturated regime
the time scale of the reaction is short compared to the one of the droplet movement .
therefore the coverage increases very steeply underneath the droplet leading to an elevated coverage already in the advancing contact zone ( see fig.[m1prof]b ) .
the resulting decrease of the overall wettability gradient felt by the droplet causes the slight decrease of the velocity . in this regime
the slight increase of the advancing contact angle results from an increase of the static part ( because of the decreasing wettability ) that overcomes a decrease of the dynamic part ( decreasing velocity ) .
the slight increase of the receding angle comes from the dynamic part only .
we emphasize here that fig.[m1vrcr](b ) clearly shows that the advancing contact angles are always smaller than the receding ones . as was already shown in ref.@xcite also here the differences between the static and the dynamic contact angles at the front and the rear , respectively , are an order of magnitude smaller than the difference between the two static ( or the two dynamic ) contact angles . and the diffusion constant @xmath109 .
the solid line indicates a drift - pitchfork bifurcation where a running droplet solution branches off a stationary droplet solution .
the dotted line indicates a saddle - node bifurcation where two running droplet solutions merge .
shown is also the transition between non - saturated and saturated regime ( dashed line ) .
the remaining parameters are as in fig.[m1prof ] . ]
[ m1phase ] let us finally discuss the influence of diffusion along the substrate on the dependencies shown in fig.[m1vrcr ] .
the result without diffusion ( not shown ) coincides nearly perfectly with the shown curves for @xmath105 .
only the decay in the saturated regime is slightly slower .
we mention here that the decay also depends slightly on the used cutoff @xmath50 ( cf .
eq.([xi1 ] ) and ref.@xcite ) . increasing the diffusion has a rather small influence on the non - saturated regime but changes the saturated regime even qualitatively .
the velocity decreases with the reaction rate because the adsorbate produced underneath the droplet close to the advancing edge diffuses onto the bare substrate ahead of the moving droplet .
this effectively reduces the wettability there , implying a slower movement .
in fact , running droplets cease to exist above a reaction rate @xmath114 where the velocity drops to zero again , corresponding to a supercritical bifurcation for large diffusion ( see curve for @xmath120 ) .
the bifurcation structure can , however , be more involved . for a smaller @xmath112
one finds that the branch of running droplets joins the branch of sitting drops in a subcritical bifurcation at @xmath114 , i.e. decreasing @xmath10 the running droplet branch first turns back in a saddle - node bifurcation at @xmath115 and then joins the sitting droplet branch at @xmath114 .
we indicate the definition of @xmath115 and @xmath114 in fig.[m1vrcr](a ) using the curve for @xmath112 as example . however , the subcritical part of the branch is difficult to see .
we refer the reader to the discussion of model ii in section [ addes ] for more details . between @xmath114 and @xmath115 both the large velocity running droplet and the sitting droplet
are linearly stable and correspond to metastable states .
the branch emerging from the subcritical bifurcation is linearly unstable and corresponds to threshold solutions that separate the two stable solutions .
the prominent features ( @xmath113 , @xmath114 and @xmath115 ) of the solution branches shown in fig.[m1vrcr ] can be used to determine phase diagrams by continuation techniques @xcite .
the phase diagrams show existence regions for the different types of solutions in the parameter space spanned , for instance , by the reaction rate and the diffusion constant ( see fig.[m1phase ] ) . to this end
we followed the maximum of the @xmath121 dependence at @xmath113 , which marks the transition between the non - saturated and saturated regime for running droplets ( dashed line in fig.[m1phase ] ) .
continuation of the loci of the saddle - node bifurcation at @xmath115 and the drift - pitchfork bifurcation at @xmath114 gives the border of the existence region for stable running droplets ( dotted line ) and stable sitting droplets ( solid line ) , respectively .
the region between the two latter borders corresponds to a coexistence region where running and sitting droplets are metastable .
note , that unstable sitting drops do also exist everywhere in the existence region of the running droplets .
such sitting droplets are steady solutions of the governing equations .
however , because they are unstable , even infinitely small perturbations will grow with a characteristic rate . in consequence
the droplets start to move and adopt the shape and speed of the stable running droplet solution . practically , it is difficult to change the reaction rate over orders of magnitude .
experiments normally only cover a much smaller range @xcite .
nevertheless , the analysis of the @xmath121 dependence ( fig.[m1vrcr ] ) for droplets of fixed volume gives a very good characterization of the general system behavior .
an experimentally important parameter is the size or volume of the droplet .
the velocity may increase @xcite or decrease @xcite with increasing volume depending on the parameter regime @xcite .
we show in fig.[vli ] the dependency of velocity on droplet volume ( logarithmic scale ) for a variety of reaction rates .
also there one can well distinguish a saturated and a non - saturated regime . in the non - saturated regime
the droplet velocity increases with increasing size , whereas in the saturated regime it decreases . for small droplets
one always finds a non - saturated regime whereas droplets of a very large size are always in a saturated regime ( in fig.[vli ] the latter is not yet reached for the curve with @xmath98 ) .
the explanation for this behavior is closely related to the one given for the @xmath121 dependence . for small droplets
the reaction does not reach saturation until the rear of the droplet has passed .
this implies that an increase in droplet size gives more time for the reaction leading to a larger value of @xmath36 at the back and , in consequence , to a larger velocity .
for very large droplets the reaction has enough time to reach saturation ( @xmath43 ) at the rear , i.e. an increase in size does not increase the driving wettability gradient .
it may even decrease slightly due to an increase of @xmath36 in the contact zone at the advancing edge ( see above ) .
however , a larger droplet has a larger viscous dissipation leading to a decreasing velocity with increasing size . inspecting fig.[vli ]
shows that for @xmath122 and @xmath123 ( i.e. the volume used in fig.[m1vrcr ] ) one is well in the saturated regime . on drop size , i.e. droplet volume @xmath91 ( logarithmic scale ) .
curves are plotted for different reaction rates @xmath12 ( see legend ) .
remaining parameters are as in fig.[m1prof ] . ]
[ vli ] and the droplet volume @xmath91 ( in units of @xmath124 ) for two diffusion constants in the @xmath36-field @xmath105 ( a ) and @xmath112 ( b ) .
the solid and dotted lines indicate a drift - pitchfork and a saddle - node bifurcation , respectively .
the dashed line indicates the transition between non - saturated and saturated regime .
remaining parameters are as in fig.[m1prof ] . ]
[ m1existvol ] the volume @xmath125 where the maximal velocity is obtained can be followed in the space spanned by droplet volume and reaction rate .
the obtained existence region for non - saturated and saturated running droplets are shown in fig.[m1existvol ] .
having studied the type i model describing experiments where the passing droplet irreversibly changes the substrate we next report on results for our type ii model where the substrate recovers .
in contrast to type i this allows for the description of a periodic droplet movement as experimentally observed in ref.@xcite .
[ addes ] this section is devoted to the type ii model that extends and generalizes the model of type i discussed above . model ii accounts for experimental situations where both , the droplet and its surrounding medium are able to change the substrate by adsorption or desorption . in this way
a moving droplet makes the substrate less wettable , but after the droplet has passed the substrate may relax to its initial state .
such systems are modeled by extending the reaction kinetics for the @xmath36-field ( [ r1 ] ) by an additional desorption term , i.e. we combine the evolution equations for the film thickness profile ( [ sys1 ] ) and the adsorbate field ( [ sys2 ] ) with the reaction term ( [ r2 ] ) and the disjoining pressure ( [ pi2b ] ) .
we will refer to this set of equations as type ii model .
the reaction term ( [ r2 ] ) is chosen such that adsorption and desorption of @xmath36 take predominantly place underneath and outside the droplet , respectively .
( see legend ) for ( a ) the profiles of moving droplets and ( b ) the corresponding profiles in the @xmath36-field .
the drop profiles at different @xmath12 differ only slightly by visual inspection .
the differences are more prominent in the @xmath36-profiles .
parameters are @xmath126 , @xmath127 , @xmath108 , @xmath104 , @xmath105 , @xmath128 ( @xmath88 , @xmath106 ) . ]
[ m2profr ] as above for model i , we use continuation techniques to calculate running droplet solutions of model ii as solutions stationary in a comoving frame .
examples of the resulting droplet profiles along with the profiles of the @xmath36-field are shown for a very small diffusion @xmath109 in figs.[m2profr ] and [ m2profs ] .
thereby , fig.[m2profr ] displays the results for several values of the reaction rate @xmath12 for a fixed ratio @xmath90 of the desorption to adsorption rates .
the drop profiles change only slightly but the @xmath36-profile undergoes prominent changes . as in model i there is practically no @xmath36-field in front of the droplet ( @xmath41 ) .
the field increases ( i.e. the wettability decreases ) underneath the droplet as in model i , but in contrast the @xmath36-field decreases behind the droplet due to the desorption .
the concentration of the @xmath36-field at the rear of the droplet is increasing with @xmath12 until it reaches saturation .
because @xmath12 is the overall reaction rate also the desorption of @xmath36 behind the droplet becomes faster , i.e. the tail of the @xmath36-field becomes shorter . of desorption and reaction ( see legend ) on ( a ) the profiles of moving droplets and ( b ) the corresponding profiles in the @xmath36-field for model ii .
the drop profiles at different @xmath90 differ only slightly by visual inspection .
the differences are prominent in the tail of the @xmath36-profile behind the droplet . @xmath129 and the remaining parameters are as in fig.[m2profr ] . ]
[ m2profs ] fig.[m2profs ] displays droplet and @xmath36 profiles for a fixed reaction rate @xmath12 but different values for the desorption to adsorption ratio @xmath90 .
the droplet shape and the form of the @xmath36-field underneath the droplet are almost independent of @xmath90 as one would expect .
however the decay of the elevated @xmath36 value occurring behind the droplet is strongly influenced .
for a small @xmath90 , i.e. slow desorption as compared to adsorption , @xmath36 decays slowly approaching qualitatively the behavior of model i. increasing @xmath90 reduces drastically the length of the @xmath36 tail behind the droplet .
the length of the tail is very important when studying the periodic movement of a droplet on a finite stripe - like substrate ( see below ) . in the following the existence regions of running and sitting droplets
are determined in their dependence on the control parameters reaction rate , desorption to adsorption ratio , diffusion constant and droplet volume .
this is done along the lines explained in section [ ad ] .
continuation gives branches of stationary solutions depending on one control parameter . on these branches
special points that separate qualitatively different behavior are identified and followed in the space of parameters .
the linear stability of the stationary solutions is also determined .
details on the used techniques can be found in the appendix .
-field for running and sitting droplets ( see legend ) for model ii .
the drop profiles differ only slightly by visual inspection .
the differences are more prominent in the @xmath36-profiles . @xmath129 and the remaining parameters are as in fig.[m2profr ] . ]
[ sitdrop ] .
the branch of running droplets emerges from zero reaction rate and undergoes a subcritical drift pitchfork bifurcation at finite @xmath130 .
the reaction rate - velocity curve is shown for three different values of the ratio between desorption and adsorption @xmath90 .
the labels @xmath113 , @xmath114 and @xmath115 indicate the locations for the maximum velocity , drift - pitchfork bifurcation and the saddle - node bifurcation , respectively , for @xmath131 .
( b ) existence of running and sitting droplets depending on the reaction rate @xmath12 and the ratio between desorption and adsorption @xmath90 .
shown is also the boundary between non - saturated and saturated regime ( dashed line ) .
the remaining parameters are as in fig.[m2profr ] .
[ m2vrs ] , title="fig : " ] + .
the branch of running droplets emerges from zero reaction rate and undergoes a subcritical drift pitchfork bifurcation at finite @xmath130 .
the reaction rate - velocity curve is shown for three different values of the ratio between desorption and adsorption @xmath90 .
the labels @xmath113 , @xmath114 and @xmath115 indicate the locations for the maximum velocity , drift - pitchfork bifurcation and the saddle - node bifurcation , respectively , for @xmath131 .
( b ) existence of running and sitting droplets depending on the reaction rate @xmath12 and the ratio between desorption and adsorption @xmath90 .
shown is also the boundary between non - saturated and saturated regime ( dashed line ) .
the remaining parameters are as in fig.[m2profr ] .
[ m2vrs ] , title="fig : " ] studying first the influence of @xmath12 and @xmath90 we present in fig.[m2vrs](a ) and ( b ) branches of stationary solutions in dependence of @xmath12 for different values of @xmath90 characterized by their velocity and the resulting existence regions of running and sitting droplets in the @xmath132 parameter plane , respectively . beside the moving droplets there
exist sitting droplets ( steady states ) for all values of @xmath12 and @xmath90 .
they are symmetrical with respect to the droplet maximum and may be stable or unstable .
an example is shown in fig.[sitdrop ] . the curve for @xmath126 in fig.[m2vrs](a )
is used to better illustrate special values of the reaction rate introduced above in section [ ad ] .
they mark the loci of the maximum of the @xmath121 dependence at @xmath113 , of the saddle - node bifurcation at @xmath115 and of the drift - pitchfork bifurcation at @xmath114 . as before ,
continuation in @xmath12 of these loci when changing @xmath90 gives the phase diagram fig.[m2vrs](b ) .
in particular one obtains the border between the non - saturated and saturated regime for running droplets ( @xmath113 ) , the border of the existence region for stable running droplets ( @xmath115 ) , and the border of the existence region for stable sitting droplets ( @xmath114 ) . in the small region
where both , running and sitting droplets are stable , one finds metastability .
there noise , for instance , in the form of substrate inhomogeneities may lead to intermittent droplet movement . the sitting droplets existing in the gray shaded area of fig.[m2vrs](b ) are all unstable .
inspection of fig.[m2vrs ] reveals the rather strong influence of the desorption / adsorption ratio @xmath90 especially on the transition from the non - saturated to the saturated regime and the slowing down in the saturated regime .
the extension in @xmath12 of the saturated regime is drastically reduced with increasing @xmath90 . for decreasing @xmath90
the maximum of the @xmath121 curve slowly transforms into a plateau and reaches for @xmath133 the form typical for model i. comparing the strong decrease of the velocity with increasing @xmath12 in the saturated regime [ fig.[m2vrs](a ) ] to the slight decrease found for @xmath133 ( model i , fig.[m1vrcr](a ) , curve for @xmath105 ) poses the question why @xmath90 has such a strong influence . the influence of @xmath90 parallels the influence of the diffusion @xmath109 in model i. there for large @xmath109 the adsorbate diffuses in front of the advancing droplet rendering the substrate there less wettable .
thereby it decreases the overall wettability gradient between front and rear of the droplet . here , for large @xmath90 the @xmath36-field
is removed behind the rear of the droplet rendering the substrate there more wettable .
that also implies a reduction of the overall wettability gradient . to elucidate the mechanism of the transition between moving and sitting drops we focus for a moment on the drift - pitchfork bifurcation at @xmath114 that separates the metastable and the running droplet region
this type of bifurcation is well known from reaction - diffusion ( see , for instance , @xcite and references therein ) and hydrodynamic systems ( see , for instance , @xcite and references therein ) , where it mediates the transition between steady and travelling structures . in our system it breaks the reflectional symmetry of the sitting droplets leading to moving asymmetric droplets and acompanying travelling asymmetric adsorbate profiles . at the bifurcation a real eigenvalue switches sign , i.e.at
@xmath114 the velocity of the moving drops is zero . beyond the bifurcation sitting droplets become unstable and start to move slowly .
that makes it unlike a hopf bifurcation ( associated with the zero crossing of the real part of a pair of complex eigenvalues ) where the travelling structure has a finite velocity at the bifurcation ( for waves on flowing thin liquid films see the discussion in ref.@xcite ) .
the bifurcation may be subcritical or supercritical ( see figs.[m1vrcr ] , [ m2vrs ] and [ m2vrd ] ) and in consequence the bifurcating branch corresponds to unstable or stable moving drops . approaching the bifurcation their respective velocity goes to zero as @xmath134 providing an unambiguous signature of the drift - pitchfork bifurcation .
the basic mechanism of the drift - pitchfork bifurcation is connected to the behaviour of the neutral ( or goldstone ) mode related to the translational symmetry of the system .
this mode with eigenvalue zero is obtained by analyzing the linear stability of the stationary solutions . in general , each continuous symmetry is related to such a neutral mode . in a sense ,
the neutral modes are the modes that are closest to zero .
this implies that a perturbation or modulation of these modes in an additional degree of freedom ( if existing ) gives modes that are probable candidates to cross zero and become instability modes .
for instance , the transversal ( fingering ) instability of a liquid front results from a transversal modulation of the longitudinal translational neutral mode @xcite .
also one of the two coarsening modes of two liquid droplets is the combination of translational neutral modes of the individual droplets directed in opposite directions @xcite . here
, the drift instability is associated with a mode representing a relative shift between the translational modes of the height and the @xmath36 profiles .
right at the bifurcation this mode corresponds exactly to the translational neutral mode .
although elsewhere one can still identify the translational mode when only looking at the sub - mode for the height profile _ or _ the one for the @xmath36 profile , looking at the complete mode one realizes that the relative weight of the two sub - modes is shifted in favor of one of them .
this corresponds to the introduction of a relative shift between the two fields .
the relative shift breaks the overall reflection symmetry and leads to the movement of the drops . .
the branch of running droplets emerges from zero reaction rate and undergoes a subcritical drift pitchfork bifurcation at finite @xmath130.the reaction rate - velocity curve is shown for three different values of the diffusion constant in the @xmath36-field @xmath109 .
( b ) existence of running and stationary droplets depending on the reaction rate @xmath12 and the diffusion constant of the @xmath36-field @xmath109 .
shown is also the transition between non - saturated and saturated regime ( dashed line ) .
remaining parameters are as in fig.[m2profr ] .
[ m2vrd ] , title="fig : " ] + .
the branch of running droplets emerges from zero reaction rate and undergoes a subcritical drift pitchfork bifurcation at finite @xmath130.the reaction rate - velocity curve is shown for three different values of the diffusion constant in the @xmath36-field @xmath109 .
( b ) existence of running and stationary droplets depending on the reaction rate @xmath12 and the diffusion constant of the @xmath36-field @xmath109 .
shown is also the transition between non - saturated and saturated regime ( dashed line ) .
remaining parameters are as in fig.[m2profr ] .
[ m2vrd ] , title="fig : " ] next we discuss the influence of @xmath12 and @xmath109 presenting in fig.[m2vrd ] dependencies of droplet velocities on reaction rate for different diffusion constants and the resulting phase diagram in the @xmath135 plane for a fixed desorption / adsorption ratio @xmath90 .
one notes first that the strength of diffusion has nearly no influence on the non - saturated regime and the transition value @xmath113 [ the dashed line in fig.[m2vrd](b ) is practically vertical ] .
note that fig.[m2vrd ] focuses on relatively small diffusion . for large diffusion the droplets
will also stop as shown for model i in fig.[m1phase ] .
however , the saturated regime does depend on @xmath109 quantitatively as well as qualitatively . for fixed @xmath12 , with increasing @xmath109
the stable running droplets become slower . in parallel
its existence range in @xmath12 shrinks slowly . as in model i this behavior
is mainly caused by the diffusion of the @xmath36-field in front of the advancing droplet .
there it increases the coverage thereby reducing the overall wettability gradient , i.e. slowing down the droplets .
the qualitative change concerns the character of the drift - pitchfork bifurcation .
with increasing @xmath109 it becomes less subcritical and at a critical @xmath136 it becomes supercritical , i.e. there exists no coexistence region for sitting and running drops any more . for small diffusion @xmath137
the existence range of stable sitting drops shrinks with increasing @xmath109 whereas for larger diffusion @xmath138 it grows . and the droplet volume @xmath91 ( in units of @xmath124 ) .
shown is also the transition between non - saturated and saturated regime ( dashed line ) .
remaining parameters are as in fig.[m2profr ] . ]
[ m2vrh ] finally , in fig.[m2vrh ] we present the phase diagram for the dependence on reaction rate and droplet volume for fixed desorption / adsorption ratio @xmath90 and diffusion constant @xmath109 .
the locations of the saddle - node , the drift - pitchfork bifurcation and the velocity maximum are all shifted slightly towards smaller @xmath12 when increasing the droplet volume .
the range in @xmath12 of the non - saturated regime shrinks slightly , but the range of the saturated regime and the metastable region remain practically constant , they are only shifted towards smaller @xmath12 .
obviously the two contact regions and the substrate outside the droplet are not affected by a change in the droplet volume
. however , an increase in droplet size increases the viscous forces and therefore stalls the droplet movement at smaller reaction rates .
finally , we employ numerical simulations of running droplets and show that the type ii model is able to describe different experimentally found modi of periodic droplet movement @xcite . to perform the simulations we use the routine d02cjc provided by the nag library @xcite .
it is based on a variable order , variable step size adam s method .
the simulations either utilize periodic boundary conditions ( to model droplets on a ring - like track @xcite ) or boundary conditions that mimick non - wettable borders of an otherwise wettable channel ( to model droplets on finite stripe - like tracks ) @xcite .
simulations are started from steady droplet solutions which develop in the absence of a chemical field , i.e. imposing @xmath41 .
then the droplet movement is initiated by breaking the symmetry of the @xmath36-field by imposing a small gradient and starting the adsorption / desorption reaction .
after a short initial transient the running droplets follow periodic trajectories that do not depend on details of the initial symmetry breaking .
for the periodic boundary condititions the initial solution was one stationary droplet , whereas in the case of non - wettable boundaries the initial solution were two stationary droplets .
in the case of periodic boundary conditions the droplets move with constant speed and shape after an initial phase .
fig.[m2persim](a ) shows space - time plots of the evolution of the film thickness for different reaction rates @xmath12 .
the droplet velocity increases with the reaction rate as expected from our continuation results .
fig.[m2persim](b ) shows a comparison of the droplet velocities obtained in the simulations and by continuation .
the values for the simulations are estimated after the droplets have reached a constant speed and shape .
one finds that both velocities match fairly well .
however , for higher reaction rates the simulations slightly overestimate the velocities compared to the continuation results .
this results from the lower and equidistant discretization used in the simulation in time .
( indicated on the right ) and ( b ) a comparison of drop velocities obtained by numerical simulation ( @xmath139 ) and continuation ( solid line ) for different values of the reaction rate @xmath12 .
remaining parameters are @xmath104 , @xmath126 , @xmath105 , @xmath140 , @xmath141 , @xmath127 and @xmath108 .
dark ( light ) colors correspond to small ( large ) film thickness .
[ m2persim ] ] ( indicated on the right ) and ( b ) the velocity of the center of mass of the film thickness depending on time examplary for @xmath99 ( top ) and @xmath142 ( bottom ) . the thin dotted line in ( b ) at @xmath143
is meant to guide the eye .
remaining parameters are as in fig.[m2persim ] .
dark ( light ) colors correspond to small ( large ) film thickness .
[ m2refsim ] ] -field ( bottom ) .
@xmath144 and remaining parameters are as in fig.[m2persim ] .
[ wf ] , title="fig : " ] + -field ( bottom ) .
@xmath144 and remaining parameters are as in fig.[m2persim ] .
[ wf ] , title="fig : " ] experiments with droplets in a finite wettable channel found regular rhythmic motion @xcite or different types of shuttling motion along with slowing and stopping behavior @xcite .
the here performed simulations show a smooth transition between the different types of periodic movements depending on our control parameters .
in general , the droplets in a wettable channel with non - wettable walls move periodically between the two walls as shown for different reaction rates @xmath12 in the space - time plots in fig.[m2refsim](a ) .
the initial solution of two stationary droplets quickly coarses upon starting the chemical reactions .
the prevailing droplet oscillates between the channel walls with a frequency that depends on @xmath12 .
the droplet movement can be classified into two identical but antisymmetric halfcycles ( i.e. with different signs in the velocity , and profiles that are related by reflection ) .
fig.[m2refsim](b ) shows the droplet velocity depending on time during one halfcycle exemplary for two reaction rates @xmath12 .
we find that each half cycle typically contains three phases , distinguishable by their different velocities .
after meeting the non - wettable boundary ( phase i ) the droplet velocity is very low or even zero in the case of very small reaction rates @xmath12 . in this phase ,
the @xmath36-field , which has been produced by thof e passing droplet has first to decrease , until the droplet can return on its own path .
this phase of very small velocity is followed by a short phase of a very high velocity ( phase ii ) until the droplet returns to a medium velocity ( phase iii ) , that is kept until it meets the opposite wall and the next half cycle begins .
the subphases can also be very well distinguished in fig.[wf ] , were we show hidden line plots of the film thickness profiles ( top ) and coverage profiles ( bottom ) for one period of droplet movement .
one can clearly see that in phase i ( velocity close to zero ) , after the droplet has encountered the non - wettable wall , the concentration in the @xmath36-field is very high and drops steeply in phase ii when the droplet starts moving again .
similar subphases of droplet motion have also been observed in ref.@xcite . in the context of that work
the continuous transition from fig.[m2refsim](a ) between @xmath145 ( top ) and @xmath142 ( bottom ) can be seen as their transition between shuttling motion and intermittent shuttling motion. a simple bead - spring model based on a mechanical analogy is used in ref.@xcite .
it models the different experimentally observed regimes varying the wettability of the walls .
although it well captures the overall behaviour it can not resolve the more hydrodynamic aspects of the motion like the flow field inside the running droplets or the dynamical contact angles .
[ conc ] in the present work we have developed and analysed two models for chemically - driven self - propelled running droplets on solid substrates .
such moving droplets were described for several experimental systems using different liquids , substrates and reacting substances @xcite .
the movment of the droplets is driven by a self - produced wettability gradient that is perpetuated with the droplet itself by means of a desorption or adsorption reaction underneath the droplet .
two types of experimental systems were reported on : ( i ) adsorption underneath the droplet decreases the wettability of the substrate irreversibly @xcite ; and ( ii ) desorption underneath the droplet removes a more wettable coating that is recovered behind the droplet through an adsorption from a surrounding medium @xcite .
we have described the droplet dynamics for both types of experimental systems using coupled evolution equations for the profiles of the film thickness and the adsorbate coverage .
the equations can be derived from the navier - stokes equations using a long - wave or lubrication approximation @xcite and assuming a small damkhler number .
the two types of experiments have been mapped onto two types of models . in the type i model
a wettability - decreasing reaction takes place underneath the droplet . in the type ii model
this mechanism is extended by additionally introducing a wettability - increasing reaction that takes place at the substrate outside the droplet .
the wettability of the substrate enters both models through a disjoining pressure supplementing the laplace pressure in the thin film equation .
we have chosen here a disjoining pressure consisting of a long - range destabilizing part @xmath146 and a short - range stabilizing part @xmath147 used , for instance , in ref.@xcite to study coarsening in dewetting .
the long - range part corresponds to van der waals interaction and is not influenced by the adsorbate .
all the influence of the adsorbate goes into the short - range part that in the simplest case selected here depends linearly on the adsorbate coverage . using continuation techniques and numerical simulations we have analyzed the solution behaviour of both models .
thereby we have focused on stationary running and sitting droplets in two dimensions .
both models display a transition from a non - saturated to a saturated regime with increasing reaction rate .
the transition is also obtained when increasing the droplet volume . in the non - saturated regime
an increase in the reaction rate leads to a larger wettability gradient implying a larger droplet velocity . in the saturated regime
an increase in the reaction rate does not increase the coverage at the rear of the droplet , i.e. it does neither lead to a larger wettability gradient nor to a larger droplet velocity .
however , it has turned out that the driving force and in consequence the velocity are decreasing slightly with increasing reaction rate .
this effect is due to a rise in the adsorbat concentration in the advancing contact zone .
a similar behavior occurs when the droplet volume increases .
there , however , in the saturated regime the velocity clearly decreases with increasing volume because the constant driving force ( wettability gradient ) is counteracted by an increasing viscous friction .
the latter dependencies found for the non - saturated and the saturated regime correspond very well to experimental results of ref.@xcite ( fig.5 ) and ref.@xcite [ fig.7(a ) ] , respectively . to our knowledge there exist , however , no experimental results for a physico - chemical system that show the qualitative change between the two regimes in dependence of the droplet velocity on its volume . with the combination of materials used in ref.@xcite
this should be possible because their fig.6(a ) shows the transition from the non - saturated to saturated regime for increasing solute concentration within the droplet .
this corresponds directly to the dependence on the reaction rate shown here in fig.[m1vrcr ] . allowing for diffusion of the adsorbate along the substrate does not affect the results in the non - saturated regime too much .
however , it leads to a stronger decrease of the velocity in the saturated regime because adsorbate is transported to the substrate in front of the running droplet thereby decreasing the overall wettability gradient .
this may even lead to a transition towards sitting droplets .
this transition occurs either continuously through a supercritical drift - pitchfork bifurcation or discontinuously through a saddle - node bifurcation . in the latter case a metastable parameter region
exist where running and sitting drops can coexist .
we have determined the existence regions of running droplets depending on the reaction rate , the diffusion constant in the surface coating , the droplet volume and the desorption / adsorption ratio ( only for model ii ) .
both models show a strong dependence of the existence regions of sitting and running droplets on the diffusion constant and only a week dependence on the droplet volume . the strong dependence on the diffusion constant results mainly from the diffusion of the @xmath36-field in front of the advancing droplet , thereby greatly affecting the driving force and the droplet velocity . the type ii model also shows a strong dependence on the desorption / adsorption ratio , which is due to a decreased substrate coverage in close proximity behind the droplet , effectively decreasing the driving force .
the primary effect of the desorption reaction is the possibility for the droplet to return on its own path , which we have illustrated by numerical simulations of droplets moving on a wettable stripe with non - wettable borders in an oscillatory manner , which has also been observed experimentally @xcite .
finally , we have shown that different modes of periodic movement called shuttling motion and intermittent shuttling motion in ref.@xcite are covered by the presented model .
the analysis of both models has shown that for the chemically - driven running droplets the advancing contact angle is always smaller than the receding one . as was already shown in ref.@xcite also here
the differences between the static and the dynamic contact angles at the front and the rear are one order of magnitude smaller than the difference between the two dynamic contact angles .
this challenges the assumption of equal dynamic contact angles at the front and the rear that was used in ref.@xcite to develop a simple description of self - propelled running droplets .
a simple quantitative theory should instead be based on the assumption that the respective dynamic contact angles equal the different static contact angles at the front and rear .
the proposed type i and type ii model based on thin film or lubrication theory can very well reproduce the main features of droplet motion that have been observed experimentally .
however , they fail to reproduce the reported damped oscillations in the droplet shape overlaying the continuous droplet movement @xcite
. these oscillations could be on the one hand the result of a weakly inhomogeneous surface , since the authors of ref.@xcite themselves suggested that they have no full control of the experimental surface properties .
on the other hand the oscillations could be a sign of a hopf bifurcation along the solution branch of running droplets .
we are well aware that the models presented in this paper are minimal models that , however , reproduce qualitatively most aspects of the behaviour of chemically self - propelled droplets . refinements of the presented theory could include the viscous motion of the ambient medium as present in the type ii experiments .
this extension can still be based on the lubrication approximation , for instance , along the lines of the two - layer systems studied in refs.@xcite .
a second important extension could cover the case of a higher damkhler number .
such a model has to include a description of the transport of the solute within the droplet .
[ auto ] sitting and running droplets are steady solutions in the laboratory and comoving frame , respectively .
they can be followed in parameter space using numerical continuation techniques @xcite , for instance , employing the continuation software auto97 @xcite .
the following section highlights some technical details of the employed techniques .
the basic idea behind continuation is that unknown solutions of an algebraic system for a certain set of control parameters are obtained by iterative techniques from known solutions nearby in parameter space .
differential equations of the form @xmath148 subject to initial , boundary and integral constraints are discretized in space and then the resulting algebraic system is solved iterativly . here
the dash indicates the first derivative with respect to @xmath149 and @xmath150 denotes the set of control parameters .
the presence of boundary conditions and/or integral conditions requires the presence of free parameters which are determined simultaneously and are part of the solution to the differential equation .
the package auto97 is limited to the continuation of ordinary differential equations ( ode s ) , thus it can only be used to compute droplet solutions in two dimensions . as an example we consider the continuation of stationary running droplets , steady in a comoving frame . after transforming eqs.([sys1 ] ) and ( [ sys2 ] ) into the comoving frame with velocity @xmath10 and integrating the resulting time - independent thin - film equation we have the system of ode s @xmath151 where @xmath152 is an integration constant .
it has the physical meaning of a mean flow in the comoving frame . @xmath153 ,
@xmath154 , @xmath155 , @xmath156 and @xmath157 denote @xmath75 , @xmath158 , @xmath159 , @xmath36 and @xmath160 , respectively .
the dashed quantities denote first derivatives with respect to @xmath149 .
the system is flux conservative , thus we need to specify the integral condition @xmath161 where @xmath13 is the system length and @xmath162 is the mean film thickness . furthermore , for a complete description of the system we introduce periodic boundary conditions for the film thickness @xmath163 and the mixed boundary conditions @xmath164\right\}\times\nonumber\\ & & \left(\phi_2(0)-\phi_1(0)\lambda_1\right)\\ 0 & = & \xi(h_1(0))\lambda_2+\left\{\xi(h_1(l))+s\left[1-\xi(h_1(l))\right]\right\}\times\nonumber\\ & & \left(\phi_2(l)-\phi_1(l)\lambda_2\right)\end{aligned}\ ] ] for the @xmath36-field ( see appendix [ boundary ] ) . since translation of a running droplet solution in the @xmath149-direction also yields a valid solution one needs to introduce a pinning condition in the form of an additional boundary or integral condition , which will not be specified further in here .
as mentioned earlier , the number of boundary conditions ( nbcd ) and integral conditions ( nint ) imposes a constraint on the number of parameters ( npar ) , that have to be varied during the continuation process .
specifically npar = nbcd+nint - ndim+1 , where ndim is the dimensionality of the system of ode s .
this condition leaves us here with three free parameters , which are the principal continuation parameter ( e.g.the reaction rate @xmath12 ) , the mean flow @xmath152 and the droplet velocity @xmath10 .
auto97 uses the method of orthogonal collocation for discretizing solutions , where the solution is approximated by piecewise polynomials with 27 collocation points per mesh interval .
the mesh is adaptive as to equidistribute the discretization error .
having specified the ode system in standard form with boundary and integral conditions auto97 then tries to find stationary solutions to the discretized system , by using a combination of newton and chord iterative methods .
once the solution has converged auto97 proceeds along the solution branch by a small step in the parameter space defined by the free continuation parameters and restarts the iteration . + the challenge usually is , to provide auto97 with a nonuniform starting solution for the continuation . for our purpose
it is sufficient to start the continuation close to the point of the primary bifurcation point , where the stable uniform solution of the ode system ( [ as1])([as2 ] ) or a system similar to ( [ as1])([as2 ] ) with periodic boundary conditions undergoes a hopf - bifurcation . in the vicinity of the bifurcation
one can determine analytically small - amplitude sinusoidal stationary traveling waves .
by selectively using reaction , boundary or integral conditions as primary continuation parameter one finally computes fully nonlinear solutions for the film thickness and the coverage .
+ auto97 is not only able to follow solution branches but can also detect bifurcations , like saddle - node bifurcation or branching points , and can then follow these bifurcations in parameter space .
[ boundary ] the use of periodic boundary conditions does not rule out interactions between droplets in consecutive periods either via the film thickness or the @xmath36-field .
interactions through the @xmath36-field arise if the desorption of the coating is slow compared to the drop movement .
one way to avoid this problem is to use very large periods , such that the @xmath36-field has enough time to recover .
this approach is successful in suppressing interactions but is very costly from a computational point of view .
an alternative approach is to use other than periodic boundary conditions for the @xmath36-field .
the following paragraphs illustrate this approach .
we generally consider a running droplet steady in a comoving frame , that has its maximum approximately in the center of the computational domain of length @xmath13 .
we assume that the film thickness profile obeys periodic boundary conditions @xmath165 and @xmath166 .
the latter restriction ensures that the minimal film thickness is close to equilibrium and the period is large compared to the droplet . in model
i we assume that there is no chemical reaction taking place outside the droplet .
therefore , outside the computational domain for @xmath167 and @xmath168 the following ordinary differential equation holds @xmath169 which has the general solution @xmath170 with @xmath171 and @xmath172 being yet undetermined constants . for simplicity
we assume that @xmath173 , i.e.the droplet is moving to the right . in front of the droplet it is assumed that @xmath174 as @xmath175 .
furthermore we assume that the @xmath36-field adopts a finite value behind the droplet such that @xmath176 as @xmath177 .
first we consider the boundary at @xmath178 with the boundary value for the @xmath36-field @xmath179 and the first derivative of the @xmath36-field @xmath180 . since @xmath181 is a finite value we find @xmath182 , which leads to the neumann boundary condition @xmath183 next we consider the boundary at @xmath184 with the boundary value for the @xmath36-field @xmath185 and the first derivative of the @xmath36-field @xmath186 .
integrating ( [ ode1 ] ) we find @xmath187 where @xmath188 is an integration constant . since @xmath189 as @xmath175 and also @xmath190 as @xmath191
we find @xmath192 , which leads us to the mixed boundary condition @xmath193 we assume for the @xmath75-field outside the computational domain with @xmath167 or @xmath168 @xmath194 .
then the concentration profile for the @xmath36-field for @xmath167 or @xmath168 can be computed analytically by solving a linear 2@xmath195 order ordinary differential equation in @xmath149 of the form @xmath196 which has the general solution @xmath197 where @xmath198}\ ] ] @xmath199\right\}}\ ] ] and @xmath172 and @xmath188 are yet undetermined constants . for
@xmath200 @xmath36 adopts a small but finite value @xmath181 .
first we consider the boundary at @xmath178 .
this leaves us with @xmath192 . using the boundary values for the @xmath36-field @xmath179 and the first derivative of the @xmath36-field @xmath180 at the boundary
@xmath178 the following system of equations holds @xmath201 solving for @xmath172 one finds the following mixed boundary condition @xmath202\right\}\left(\phi_{x0}-\phi_0\lambda_1\right)\,.\ ] ] the same procedure can be performed at the boundary @xmath184 with @xmath185 and @xmath186 yielding the condition @xmath203\right\}\left(\phi_{xl}-\phi_l\lambda_2\right)\ ] ] in the continuation algorithm
the two periodic boundary conditions for the @xmath36-field were substituted by the two mixed boundary conditions obtained above .
we thank e. knobloch for his comments on the drift - pitchfork bifurcation , and acknowledge support through the eu rtn unifying principles in non - equilibrium pattern formation
( contract mrtn - ct-2004005728 ) . 10
[ 2]#2 [ 1]#1 [ 1]#1 , ( , london , ) , ( reprinted 4th ed . 1730 ) , book iii , part 1 , querie 31 .
, , ( ) . and , , ( ) . , , and , , ( )
, , and , , ( ) . , , ( ) . , , ( )
, , , , , and , , ( ) . , and , , ( ) . and , , ( ) .
, , ( ) , observation16 . , , and , , ( ) . ,
( ) . and , , ( ) . , , and , , ( )
. and , , ( ) . and , , ( ) . , , and , , ( ) . , , , and , ( 6 ) , , ( pages ) ( ) . , , , , , , and , , ( ) . ,
( ) . , , and , , ( ) . and , , ( )
, , ( ) . , , and , , ( ) .
, , , and , , ( ) . , , and , , ( ) . and , , ( )
, , and , , ( ) . , , and , , ( ) . and , , ( )
, , ( ) . and , in _ lecture notes in physics _
( , ) , , . , , ( ) . , , ( )
, , and , , ( ) . , , and , , ( ) .
, , ( , oxford , ) . , ( , london , ) . , , , and , , ( ) . , ( , new york , ) , 2nd ed . note , that the disjoining pressure used in ref.@xcite was @xmath204 $ ] where for @xmath41 , @xmath205 and @xmath206 are the apolar and polar components of the total spreading coefficient @xmath207 , respectively , and @xmath72 is a correlation length @xcite .
one usually describes the choice @xmath208 and @xmath209 as a combination of a stabilizing long - range van der waals and a destabilizing short - range polar interaction .
the apparent contradiction of qualitative similar results for model i for different verbal descriptions and combinations of signs used here and in ref.@xcite results from a subtle feature of the combination of exponential and power law . combining a term @xmath210 and
one @xmath211 leads for a proper choice of parameters to a dominance of @xmath212 for large and very small @xmath75 .
the exponential only dominates for intermediate thicknesses ( see ref.@xcite for a related phase diagram ) .
this implies that above verbal description only covers part of the feature of the disjoining pressure . on the contrary
, the combination of two power laws used here clearly attributes the long range and short range forces to the terms @xmath213 and @xmath214 , respectively .
we therefore believe , that the chosen disjoining pressure more accurately represents the physical situation .
, , ( ) . , , and , , ( ) . , , and , , ( ) . , , , , , and , ( , montreal , ) . , , , , and , , ( ) . , , and , , ( ) . and , , ( ) . and , , ( ) . , , , and , , ( ) . , , , , , and , , ( ) . and , , ( ) . , , and , , ( ) . and , , ( ) . and , , ( ) . and , , ( ) . and , , ( )
. , , , and , , ( ) .
( ) , www.nag.co.uk . and , , ( ) . , , , and , , ( )
, , and , , ( ) .
|
we study chemically driven running droplets on a partially wetting solid substrate by means of coupled evolution equations for the thickness profile of the droplets and the density profile of an adsorbate layer .
two models are introduced corresponding to two qualitatively different types of experiments described in the literature . in both cases
an adsorption or desorption reaction underneath the droplets induces a wettability gradient on the substrate and provides the driving force for droplet motion .
the difference lies in the behavior of the substrate behind the droplet . in case i the substrate
is irreversibly changed whereas in case ii it recovers allowing for a periodic droplet movement ( as long as the overall system stays far away from equilibrium ) .
both models allow for a non - saturated and a saturated regime of droplet movement depending on the ratio of the viscous and reactive time scales .
in contrast to model i , model ii allows for sitting drops at high reaction rate and zero diffusion along the substrate .
the transition from running to sitting drops in model ii occurs via a super- or subcritical drift - pitchfork bifurcation and may be strongly hysteretic implying a coexistence region of running and sitting drops .
|
let @xmath6 be standard one - dimensional brownian motion and @xmath7 a continuous function defined on some interval @xmath8 .
a standard result is that the zero set of @xmath6 has no isolated points almost surely , see theorem 2.28 in @xcite . by the cameron - martin theorem ( see theorem 1.38 in @xcite or theorem 2.2 in chapter 8 in @xcite ) the zero set of the process @xmath4 has no isolated points almost surely if @xmath3 is in the cameron - martin space @xmath9 ( integrals of functions in @xmath10 ) .
we will prove that the same is true for any function @xmath3 which is @xmath5-hlder continuous .
since all functions in @xmath9 are @xmath5-hlder continuous , this is a stronger statement than the one implied by the cameron - martin theorem . for any function @xmath11 defined on some subset ( or the whole ) of @xmath12 denote by @xmath13 the set of zeros of @xmath11 in @xmath14 .
we remove the origin from consideration since the origin is an isolated zero of the process @xmath4 for any @xmath3 growing fast enough in the neighborhood of the origin , say @xmath15 .
[ thm : holder-1/2 ] for @xmath16 which is @xmath5-hlder continuous on compact intervals , the set @xmath17 has no isolated points almost surely .
the condition that @xmath3 is @xmath5-hlder continuous is sharp in the following sense .
[ thm : isolated_zeros ] for every @xmath1 there is an @xmath2-hlder continuous function @xmath16 such that the set @xmath17 has isolated points with positive probability .
theorem [ thm : isolated_zeros ] will follow directly from proposition [ prop : isolated_zeros_first_example ] .
an example of function @xmath3 satisfying theorem [ thm : isolated_zeros ] is given in section [ isolated zeros - first example ] .
for @xmath18 , let @xmath19 denote the middle @xmath20-cantor set and let @xmath21 be the corresponding cantor function , shifted to an interval away from the origin ; see figure [ fig : cantor ] and section [ isolated zeros - first example ] for a precise definition . for @xmath22 , the set @xmath23 has isolated zeros with positive probability , and all such zeros
are contained in the cantor set @xmath19 .
the proof of this claim consists of constructing a subset of @xmath19 which contains zeros of @xmath24 with positive probability , and in which any zero is isolated .
en route we obtain the following result of independent interest .
[ prop : hitting characterisation ] let @xmath21 be a cantor function
. then @xmath25 if and only if @xmath26 .
the case @xmath27 of the above theorem has already been resolved by taylor and watson ( see example 3 in @xcite ) .
their interest in the graph of the restriction @xmath28 stemmed from the fact that , although the projection of this set on the vertical axis is an interval , the graph of brownian motion does not intersect this set almost surely .
$ ] ( functions @xmath29 from the construction in section [ isolated zeros - first example ] ) for @xmath30 and @xmath31
. approximations of the cantor set ( sets @xmath32 from the construction in section [ isolated zeros - first example ] ) are drawn in bold . ]
part ( ii ) of proposition [ prop : basic_facts ] shows that isolated zeros of the process @xmath4 can occur only where the function @xmath3 increases or decreases very quickly . in the following theorem we bound the hausdorff dimension of such sets
the hausdorff dimension of a set @xmath33 will be denoted @xmath34 .
[ set a_f has dim less than 1/2 ] for any continuous function @xmath16 there exists a set @xmath35 , such that @xmath36 and such that , almost surely , all isolated points of @xmath17 are contained in @xmath35 .
it is a classical result that the zero set of brownian motion has hausdorff dimension @xmath5 almost surely , see theorem 4.24 in @xcite . of course
, for any compact interval @xmath37 not containing @xmath38 and any continuous function @xmath7 the event @xmath39 will have a non - zero probability , and it is easy to construct a function @xmath16 with the same property .
however , we prove that adding a continuous drift can not decrease the hausdorff dimension of the zero set almost surely .
this is the content of the following theorem .
[ thm : hausdorff_lower_bound ] for continuous function @xmath16 , the set @xmath17 has hausdorff dimension greater than or equal to @xmath5 with positive probability . as the following example shows , upper bounds on the hausdorff dimension of the zero set
can not be obtained without additional assumptions on the drift @xmath3 .
recall that fractional brownian motion @xmath40 with hurst index @xmath41 is a continuous , centered gaussian process , such that @xmath42 . taking the drift @xmath3 to be an independent sample of fractional brownian motion with hurst index @xmath43 ,
one gets that the hausdorff dimension @xmath44 is bounded from below by @xmath45 , almost surely .
this follows from the proof of the same fact for unperturbed fractional brownian motion in theorem 4 in chapter 18 of @xcite ( see also proposition 5.1 and section 7.2 in @xcite ) .
[ thm : hausdorff_upper_bound ] let @xmath16 be either @xmath5-hlder continuous on compact intervals or of bounded variation on compact intervals .
then the hausdorff dimension of @xmath17 is at most @xmath5 , almost surely .
arguments from the proof of theorem [ thm : hausdorff_lower_bound ] and theorem [ thm : hausdorff_upper_bound ] imply the following corollary .
[ cor : sharp_haudorff_for_holder ] if @xmath46 \to \mathbb{r}$ ] is a @xmath5-hlder continuous function , such that @xmath47 , then the hausdorff dimension of @xmath17 is equal to @xmath5 , almost surely .
[ rem : no_isolated_extremes ] for a function @xmath48 , define @xmath49 and denote the set of its record times by @xmath50 . for standard brownian motion @xmath6 , a result of lvy says that the processes @xmath51 and @xmath52 have the same distribution ( see e.g. theorem 2.34 in @xcite ) , which implies that sets @xmath53 and @xmath54 have the same distribution ( on the borel sigma algebra of families of closed subsets of @xmath12 , generated by the hausdorff metric ) . in general , for brownian motion with drift
there is no such correspondence .
actually , one can see that there are no isolated points in the set of record times of the process @xmath4 almost surely .
this is proven in part ( ii ) of proposition [ prop : zeros extremes and record times ] .
connection between hlder continuity of drift @xmath3 and path properties of @xmath4 has already been observed . for @xmath55 , a function @xmath56
is called polar if , for @xmath57-dimensional brownian motion @xmath6 started at the origin and any point @xmath58 , the probability that there is a @xmath59 such that @xmath60 is positive . in @xcite graversen
constructed @xmath2-hlder continuous functions which are polar for two dimensional brownian motion , for @xmath61 . in @xcite
le gall showed that @xmath5-hlder continuous functions are not polar for two dimensional brownian motion and that the same conclusion holds in higher dimensions when @xmath3 satisfies a slightly stronger condition than @xmath62-hlder continuity . in a recent paper @xcite it was shown that for any @xmath63 there are @xmath2-hlder continuous functions @xmath3 such that the image of @xmath4 covers an open set almost surely .
we will briefly review some related results on intersections of brownian trajectories with non - smooth paths . let @xmath64 be the family of all functions @xmath65\to \mathbb{r}$ ] such that @xmath66 and @xmath67 . define the local time of brownian motion @xmath6 on @xmath3 by the formula @xmath68}(b(s))ds$ ] .
it was proved in @xcite that the supremum over @xmath69 of @xmath70 is finite for @xmath71 and infinite for @xmath72 .
the function @xmath73 is continuous over @xmath74 $ ] for @xmath71 .
suppose that @xmath75 is a continuous function and let @xmath76 be brownian motion reflected on @xmath11 ; see @xcite for a precise definition .
let @xmath77 be the set of all @xmath59 such that @xmath78 .
it was proved in @xcite that for every continuous function @xmath11 we have @xmath79 and for some continuous functions
@xmath11 we have @xmath80 .
for an interval @xmath37 we denote its length by @xmath81 and say it is _ dyadic _ if it is of the form @xmath82 $ ] for integers @xmath83 and @xmath84 . for intervals @xmath37 and @xmath85 , we will write @xmath86 if @xmath85 is located to the right of @xmath37 . [ 6.30.1 ] we will repeatedly use the following simple observations .
\(i ) suppose that @xmath87 , @xmath88 , are events and for some @xmath89 and all @xmath90 we have @xmath91 .
recall that @xmath92 is the event that infinitely many @xmath87 s occur . then @xmath93 .
\(ii ) as an easy consequence of the cauchy - schwarz inequality we have @xmath94 , for any nonegative variable @xmath95 . see lemma 3.23 in @xcite .
[ prop : basic_facts ] let @xmath16 be a continuous function .
* let @xmath33 be a set such that for any @xmath96 there is an @xmath72 such that @xmath97 .
then , almost surely any point in @xmath98 is isolated in @xmath17 .
* almost surely all isolated points of @xmath17 are located inside the set @xmath99 , where @xmath100 and @xmath101 .
\(i ) first assume that the set @xmath102 is contained in @xmath103 $ ] for some large @xmath104 .
if a zero @xmath105 is not isolated , then we can find a sequence @xmath106 , converging to @xmath107 , which , for some @xmath61 , necessarily satisfies @xmath108 however , this is impossible since , by levy s modulus of continuity , almost surely , there exists an @xmath109 such that for all @xmath110 and all @xmath111 we have @xmath112 , see e. g. theorem 1.14 in @xcite .
if @xmath102 is unbounded , apply the above reasoning to @xmath113 $ ] and let @xmath104 go to infinity .
\(ii ) first define @xmath114 and notice that any isolated zero of the process @xmath4 must equal @xmath115 , for some @xmath116 .
this is because , for any zero @xmath117 , not of the form @xmath115 , and a sequence of rational numbers @xmath118 converging to @xmath107 from below , we have @xmath119 .
therefore , it is enough to prove that for each @xmath120 , the event that @xmath121 and that @xmath115 is isolated in the set @xmath17 , has probability zero .
fix a positive integer @xmath122 and define sequences of functions @xmath123 and @xmath124 since @xmath3 is continuous , it is easy to see that for each @xmath125 , the functions @xmath126 and @xmath127 are measurable . also define @xmath128 for all @xmath129 it holds that @xmath130 and @xmath131 , for all @xmath125 . since @xmath115 is a stopping time , the process @xmath132 , is , by the strong markov property , a brownian motion independent of the sigma algebra @xmath133 .
let @xmath134 denote the event that @xmath135 happens for infinitely many @xmath125 s .
since the random variables @xmath136 are measurable with respect to @xmath133 , blumenthal s 0 - 1 law implies that @xmath137 is equal to 0 or 1 on the event @xmath138 . on the event @xmath138 , for every @xmath125 , we have @xmath139 since the right hand side does not depend on @xmath125 , by remark [ 6.30.1 ] ( i ) , @xmath140 on the event @xmath138 . similarly , if @xmath141 denotes the event that @xmath142 happens for infinitely many @xmath125 s then @xmath143 on the event @xmath138 . by the definition of the sequences @xmath144 and @xmath145 , if @xmath146 holds then @xmath115 is not an isolated zero from the right .
therefore , the probability that @xmath147 and that @xmath115 is an isolated point of @xmath17 is equal to zero . taking the union over all rational @xmath148 s and observing
that @xmath149 proves the claim .
[ rem : basic_facts_extended ] by proposition [ prop : basic_facts ] ( ii ) and the time reversal property of brownian motion it follows that , almost surely all isolated points of @xmath150 are contained in the set @xmath151 every point @xmath152 for which the limits @xmath153 are equal to @xmath154 or @xmath155 is a strict local minimum or maximum .
since a function can have only countable many strict local extrema , almost surely all isolated points of @xmath150 are contained in the set @xmath156 in particular , all isolated points of @xmath150 are contained in the set of points of increase or decrease of @xmath3 ( recall that @xmath152 is a point of increase of @xmath3 if for some @xmath157 we have @xmath158 for @xmath159 and @xmath160 for @xmath161 and points of decrease are defined analogously ) .
therefore , if @xmath3 is a function with at most countable many points of increase or decrease , then @xmath150 has no isolated points almost surely .
examples of such functions include functions constructed by loud in @xcite which satisfy a certain local reverse hlder property at each point ( see also the construction in @xcite ) .
these functions are defined as @xmath162 where @xmath163 , for @xmath164 , a positive integer @xmath102 such that @xmath165 , and a continuous function @xmath166 which has value @xmath38 at even integers , value @xmath167 at odd integers and is linear at all other points . to prove that these functions have at most countably many points of increase or decrease , we proceed analogously as in the proof of the lower bound in @xcite .
we will take @xmath168 and show that @xmath152 is not a point of increase .
first observe that @xmath169 is odd for infinitely many integers @xmath84 .
for such an integer @xmath84 assume that @xmath170 and denote @xmath171 .
then by construction @xmath172 for all @xmath173 , @xmath174 and @xmath175 , for @xmath176 . therefore @xmath177 using the fact that @xmath178 , it is easy to check that the right hand side above is negative , and since @xmath179 can be arbitrarily close to @xmath152 the claim follows . if @xmath180 then define @xmath181 , which now satisfies @xmath182 and proceed analogously to prove that @xmath183 .
this is straightforward from part ( ii ) of proposition [ prop : basic_facts ] . for @xmath184
, we will define the middle @xmath20-cantor set and denote it by @xmath19 .
take a closed interval @xmath37 of length @xmath81 .
define @xmath185 as the set consisting of two disjoint closed subintervals of @xmath37 of length @xmath186 , the left one ( for which the left endpoint coincides with the left endpoint of @xmath37 ) and the right one ( for which the right endpoint coincides with the right endpoint of @xmath37 ) .
continue recursively , if @xmath187 , then include in the set @xmath188 its left and right closed subintervals of length @xmath189 .
define the set @xmath32 as the union of all the intervals from @xmath190 . for any @xmath125 ,
the family @xmath190 is the set of all connected components of the set @xmath32 .
the cantor set is a compact set defined as @xmath191 .
it is easy to show that @xmath192 .
now we recall the construction of the standard cantor function .
define the function @xmath193 so that it has values @xmath38 and @xmath167 at the left and the right endpoint of the interval @xmath37 , respectively , value @xmath5 on @xmath194 and interpolate linearly on the intervals in @xmath185 .
recursively , construct the function @xmath195 so that for every interval @xmath196\in \mathfrak{c}_{\gamma , n}$ ] , the function @xmath195 agrees with @xmath29 at @xmath107 and @xmath152 , it has value @xmath197 on @xmath198 and interpolate linearly on the intervals in @xmath188
. see figure [ fig : cantor ] .
it is easy to see that the sequence of functions @xmath199 converges uniformly on @xmath37 .
we define the cantor function @xmath21 as the limit @xmath200 .
note that for any @xmath125 and all @xmath201 the functions @xmath21 and @xmath29 agree at the endpoints of intervals @xmath202 . another way to characterize the cantor set @xmath19 and
the cantor function @xmath21 is by representing it as fixed points of certain transformations .
define linear bijections @xmath203 \to [ 0,1]$ ] and @xmath204 \to [ 0,1]$ ] by @xmath205 and @xmath206 . the cantor set @xmath19 defined on @xmath207 $ ] is the unique nonempty compact set that satisfies @xmath208 , and the corresponding cantor function is the unique continuous function that satisfies @xmath209 it is not hard to see that @xmath21 is @xmath210-hlder continuous .
the value @xmath27 is the threshold at which the functions @xmath21 become @xmath5-hlder continuous . for @xmath211 the function @xmath212 will give an example for theorem [ thm : isolated_zeros ] .
this threshold is sharp , since by proposition [ thm : holder-1/2 ] , for @xmath213 there are no isolated points in the zero set @xmath214 almost surely . for simplicity
, we will assume the initial interval @xmath37 to be @xmath215 $ ] , but we note that the analysis works for all compact intervals . to satisfy the assumptions of theorem [ thm : isolated_zeros ] , the function @xmath216 should of course
be extended to @xmath217 $ ] , by , say , value @xmath38 on @xmath218 and value @xmath167 on @xmath219 . for an interval @xmath220 \in \mathfrak{c}_{\gamma.n}$ ] ,
define @xmath221 as the event @xmath222 $ ] , and the random variable @xmath223 , where @xmath224 is the indicator function of the event @xmath221 .
note that there is a constant @xmath225 , such that for any @xmath226 we have @xmath227 if @xmath228 happens for infinitely many @xmath125 s , then we can find a sequence of intervals @xmath229 \in \mathfrak{c}_{\gamma , n_k}$ ] , such that @xmath230 , thus @xmath231 . since @xmath232 , the sequence @xmath233 will have a subsequence converging to some @xmath234 , which obviously satisfies @xmath235 .
therefore @xmath236 to estimate the probabilities @xmath237 from below we will use the inequality from remark [ 6.30.1 ] ( ii ) , for which we need to bound the second moment @xmath238 from above .
first we express it as @xmath239 now fix @xmath125 and intervals @xmath240 $ ] and @xmath241 $ ] from @xmath190 , so that @xmath242 , and denote @xmath243 and @xmath244 , for @xmath245 .
let @xmath246 be the process @xmath247 which is , by the markov property and brownian scaling , again a brownian motion , independent of @xmath248 , and thus independent of the event @xmath249 .
the event @xmath250 happens when @xmath251 for the interval @xmath252 $ ] of length @xmath253 .
* case @xmath211 : * fix intervals @xmath254 and @xmath255 in @xmath256 , which are contained in a single interval in @xmath257 .
label the intervals from @xmath190 contained in @xmath254 by @xmath258 , and those contained in @xmath255 by @xmath259 , so that @xmath260 and @xmath261 .
set @xmath262 and @xmath263 for some @xmath264 , and define @xmath265 , @xmath266 , @xmath246 and @xmath267 as before .
conditional on @xmath221 , the left endpoint of the interval @xmath267 is at least @xmath268 , and since @xmath269 we have @xmath270 . because @xmath267 has length at most @xmath271
we obtain @xmath272 which , by summing over @xmath264 gives @xmath273 here we used the trivial bound for @xmath274 .
the sum on the right hand side can be written as @xmath275 for @xmath276 . since @xmath277
, we see that @xmath278 since @xmath279 is a bounded function on @xmath12 , ( [ eq : conditional probabilities estimates ] ) implies that for any fixed @xmath254 and @xmath255 as above @xmath280 for some @xmath281 .
therefore , summing the inequality in ( [ eq : second_moment_estimates_below_1/4_3 ] ) over all @xmath254 and @xmath255 and @xmath282 , and using it together with ( [ eq : second moment ] ) and ( [ eq : hitting probabilities ] ) , we have @xmath283 which is bounded since @xmath284 . thus we have shown that , for a fixed @xmath211 , the second moments @xmath238 are bounded from above . now the lower bound in the second inequality in ( [ eq : hitting probabilities ] ) and the inequality from remark [ 6.30.1 ] ( ii ) imply that @xmath285 is bounded from below and the claim follows from ( [ eq : approximation of probabilities_1 ] ) and remark [ 6.30.1 ] ( i ) . and @xmath250 ( graph of brownian motion intersects two bold vertical intervals ) . ] .
[ fig : intersection_with_cantor ] * case @xmath286 : * again pick @xmath287 such that @xmath288=i
< j = [ s_2,t_2]$ ] and define @xmath265 , @xmath266 , @xmath246 and @xmath267 as before
. by @xmath289 denote the largest integer such that both @xmath37 and @xmath85 are contained in a single interval from @xmath257 .
assume that @xmath221 happens .
clearly the endpoints of the interval @xmath267 satisfy @xmath290 the sequence @xmath291 is bounded , and therefore the interval @xmath267 is contained in a compact interval , which does not depend on the choice of @xmath125 , @xmath289 , @xmath37 or @xmath85 . using this and the fact that the length of @xmath267 is bounded from above by @xmath271 and from below by @xmath292 , we get that for some positive constants @xmath293 and @xmath294 we have @xmath295 note that , since the sequence @xmath291 is bounded for @xmath27 , estimates ( [ eq : easy conditional probabilities ] ) also hold for @xmath27 . substituting ( [ eq : easy conditional probabilities ] ) and the upper bounds from ( [ eq : hitting probabilities ] ) into ( [ eq : second moment ] ) , and summing over all intervals @xmath37 and @xmath85 , we obtain @xmath296 since @xmath297 , we have bounded @xmath238 from above by a constant not depending on @xmath125 , and the claim follows as in the case @xmath211 .
* case @xmath27 : * assume that @xmath298 and define @xmath299 as the first zero of @xmath24 in the cantor set @xmath19 ( @xmath299 exists since @xmath300 is a closed set ) .
for an interval @xmath301 \in \mathfrak{c}_{\gamma , n}$ ] assume that @xmath302 . since @xmath299 is a stopping time , and by brownian scaling , the conditional probability @xmath303 is equal to the probability that brownian motion at time @xmath167 is between @xmath304 and @xmath305 . since @xmath306
we see that @xmath307 and @xmath308 .
moreover , the assumption @xmath309 implies that @xmath310 which leads to @xmath311 .
thus we can bound the probability @xmath312 for some @xmath313 , see also figure [ fig : approximation from above ] .
, conditional on the event that there is a zero of @xmath24 in an interval @xmath314 ( @xmath299 is the first such zero ) the probability of the event @xmath221 ( brownian motion intersecting the right hand side of the rectangle ) is bounded from below . ]
therefore , @xmath315 and , since the events @xmath316 are disjoint for different @xmath317 , we have @xmath318 for every @xmath125 .
thus we obtain @xmath319 the arguments leading to ( [ eq : approximation of probabilities_1 - 5 ] ) are true for all @xmath320 and therefore ( [ eq : approximation of probabilities_2 ] ) holds for @xmath320 . for @xmath321 \in \mathfrak{c}_{\gamma , n}$ ] and @xmath322 ,
let @xmath323 denote the interval from @xmath257 that contains @xmath37 , and let @xmath324 be the left and right subintervals of @xmath323 , respectively .
let @xmath325 denote the @xmath326-ary expansion of the left endpoint of @xmath37 ( note that the @xmath326-ary expansion of @xmath107 contains only the digits @xmath38 and @xmath327 and has length at most @xmath125 , here we add zeros at the end , if necessary , to make it of length @xmath125 ) .
it is easy to see that @xmath328 if @xmath329 and @xmath330 if @xmath331 . call an interval @xmath37 _ balanced _ if the sequence @xmath332 contains at least @xmath333 zeros and otherwise unbalanced . for a balanced interval @xmath321 \in \mathfrak{c}_{\gamma , n}$ ]
let @xmath334 denote the event that @xmath37 is the leftmost balanced interval for which @xmath221 happens , and , as before , let @xmath335 denote the @xmath326-ary expansion of @xmath107 .
assume that for some @xmath322 we have @xmath329 and pick an interval @xmath187 such that @xmath336 . since @xmath337 we can use the same arguments that lead to the lower bound in ( [ eq : easy conditional probabilities ] ) to conclude that the probability @xmath338 , for some constant @xmath339 .
summing over all @xmath336 and over all @xmath289 such that @xmath329 gives @xmath340 by ( [ eq : hitting probabilities ] ) and since events @xmath334 are disjoint , we have @xmath341 to estimate the probability that @xmath221 happens for some unbalanced interval @xmath37 notice that the number of such intervals is bounded from above by @xmath342 for some @xmath343 . by ( [ eq : hitting probabilities ] )
this gives @xmath344 now ( [ eq : 1/4 case balanced part ] ) and ( [ eq : 1/4 case unbalanced part ] ) yield @xmath345 and the claim follows from ( [ eq : approximation of probabilities_2 ] ) .
since the cantor function @xmath21 is @xmath210-hlder continuous , the following proposition proves theorem [ thm : isolated_zeros ] .
[ prop : isolated_zeros_first_example ] for @xmath22 consider the cantor function @xmath21 .
then the set @xmath23 has isolated points with positive probability . for @xmath33
define @xmath346 as the event @xmath347 .
we claim that there exists a constant @xmath348 , such that for any interval @xmath349 $ ] of length @xmath350 , we have @xmath351 to prove this , fix an interval @xmath85 and take the largest integer @xmath125 such that @xmath352 .
notice that @xmath85 can be covered by no more than two consecutive binary intervals @xmath353 and @xmath354 of length @xmath355 .
moreover , there are consecutive @xmath356 such that @xmath357 for @xmath245 , and @xmath358 , see figure [ fig : cantor probabilities ] . now using the notation from the proof of theorem [ prop : hitting characterisation ] and the arguments that lead to ( [ eq : approximation of probabilities_1 - 5 ] ) we obtain @xmath359 which yields @xmath360 but by the first inequality in ( [ eq : hitting probabilities ] ) , the probability on the right hand side is bounded from above by @xmath355 . using this fact in ( [ eq : approximation_by_diagonals_for_small_gamma ] ) and summing the expression for @xmath245 , we obtain ( [ eq : cuts_probabilities ] ) . by theorem [ prop : hitting characterisation ] the set @xmath361 is non - empty with some probability @xmath89 .
take an arbitrary @xmath362 and @xmath363 such that @xmath364 . for @xmath365
consider the interval @xmath366 $ ] and define the set @xmath367 . by ( [ eq : cuts_probabilities ] ) and the choice of @xmath363 , we have that @xmath368 . therefore , the probability that there is a zero of @xmath369 in the set @xmath370 is at least @xmath371 ( here @xmath372 is the interior of the set @xmath373 ) .
now the claim will be proven if we show that any such zero is isolated .
take @xmath374 and any @xmath375 in the same connected component of @xmath372 .
the largest integer @xmath289 such that both @xmath107 and @xmath152 are contained in the same interval of @xmath257 satisfies @xmath376 .
moreover , @xmath377 and @xmath378 .
now it is clear that @xmath152 satisfies the condition in part ( i ) of proposition [ prop : basic_facts ] with @xmath379 .
therefore , the statement follows from part ( i ) of proposition [ prop : basic_facts ] .
$ ] can be covered by two dyadic intervals @xmath353 and @xmath354 of comparable size .
intervals @xmath380 are such that @xmath381 , for @xmath245 . ]
[ rem : isolated zeros almost surely ] it is not difficult to construct a continuous function @xmath3 such that the set @xmath17 has isolated points almost surely . let @xmath21 be the cantor function with @xmath211 defined on the interval @xmath207 $ ] .
construct the function @xmath16 such that for every @xmath382 and @xmath383 we have @xmath384 and define @xmath3 on @xmath385 arbitrarily , see figure [ eq : repeat cantor ] . by proposition [ prop : isolated_zeros_first_example ] and the cameron - martin theorem ,
the probability that @xmath17 has an isolated point in the interval @xmath386 $ ] is positive , denote it by @xmath387 . by brownian scaling the probability that there is an isolated point of @xmath17 in the interval @xmath388 $ ]
is also equal to @xmath387 , for any @xmath382 . therefore , in view of remark [ 6.30.1 ] ( i ) , the probability that there is an isolated point of @xmath17 in the interval @xmath388 $ ] , for infinitely many @xmath125 s is bounded from below by @xmath387 . by blumenthal s zero - one law
this event has probability one , which proves the claim . from remark
[ rem : isolated zeros almost surely ] . on each interval
@xmath388 $ ] function @xmath3 is a scaled and shifted copy of the cantor function ( on the interval @xmath385 the function is defined arbitrarily ) . ]
the following proposition shows that , with positive probability , the set @xmath17 can have only isolated points , or even only one point .
[ prop : only one zero ] there exists a continuous function @xmath3 such that , with positive probability , @xmath17 is a singleton .
take @xmath21 to be the cantor function with @xmath211 defined on the interval @xmath215 $ ] .
since @xmath23 has isolated points with positive probability , there are two rational numbers @xmath389 such that , with positive probability , there will be only one zero of @xmath24 in the interval @xmath390 .
denote this event by @xmath391 and on this event the unique zero by @xmath299 .
note that on the event @xmath392 the unique zero @xmath299 is necessarily a local extremum of the process @xmath24 and by part ( i ) of the upcoming proposition [ prop : zeros extremes and record times ] this event has probability zero .
furthermore , on the event @xmath393 the unique zero @xmath299 is necessarily a local point of increase of the brownian motion . by a result of dvoretzky , erds and kakutani in @xcite , almost surely , brownian motion has no points of increase and thus @xmath394 see also theorem 5.14 in @xcite .
next define @xmath395 and notice that by the markov property and the discussion above @xmath396 .
this implies that the set @xmath397 is of positive lebesgue measure and so is @xmath398 , for @xmath399 small enough .
now the claim will follow if we prove that there is a modification @xmath3 of the function @xmath21 on the intervals @xmath400 and @xmath401 , such that * with positive probability there are no zeros of @xmath4 in @xmath400 and @xmath402 , * for any @xmath403 , conditional on @xmath404 , with positive probability there are no zeros of @xmath4 in @xmath401 . for 1 )
define @xmath3 to be linear on @xmath405 $ ] with @xmath406 and @xmath407 , if we do not require @xmath47 . to prove that the probability that both @xmath408 and @xmath409 happen is positive , by the cameron - martin theorem , it is enough to prove that @xmath410}b(t ) > -\epsilon , b(q_1 ) \in s_2=s_1-f_\gamma(q_1)-\epsilon\big)>0.\ ] ]
while this is intuitively obvious it can be proven by using the reflection principle at the first time the brownian motion hits the level @xmath411 to conclude that the probability that both @xmath412 and @xmath413 happen is equal to the probability that @xmath414 , where @xmath415 is obtained by reflecting the set @xmath416 around @xmath411 . since @xmath417
we have @xmath418 and therefore @xmath410}b(t ) > -\epsilon , b(q_1 ) \in s_2\big ) = \mathbb{p}(b(q_1 ) \in s_2 ) - \mathbb{p}(b(q_1 ) \in s_3 ) > 0.\ ] ] now the probability that there is a unique zero in the interval @xmath419 is equal to some @xmath89 .
if we do require @xmath47 , redefine @xmath3 on the interval @xmath420 as @xmath421 . here
@xmath422 is chosen large enough so that for @xmath423 , for which @xmath3 is continuous , we have @xmath424 . to satisfy the condition 2 ) , replace @xmath21 on @xmath401 by a linear function of slope @xmath167 , that is @xmath425 .
to prove that this @xmath3 satisfies the required condition on @xmath401 it is enough to prove that , for standard brownian motion @xmath6 , with positive probability there are no zeros of the process @xmath426 .
this probability is equal to @xmath427 by ( 5.13 ) , sect .
3.5 in @xcite
. see figure [ fig : only one zero ] . from proposition [ prop :
only one zero ] .
it is a continuous function defined as @xmath428 on @xmath429 $ ] , linear on both @xmath430 $ ] and @xmath431 and is a part of the cantor function on @xmath432 $ ] . for this trajectory of @xmath6 , there is only one zero of the process @xmath4 , labelled as @xmath433 . ]
the next proposition justifies remark [ rem : no_isolated_extremes ] .
[ prop : zeros extremes and record times ] let @xmath16 be a continuous function and define the process @xmath434 .
* almost surely there are no points in @xmath435 which are local extrema . *
define @xmath436 and the set of record times of the process @xmath76 as @xmath437 .
almost surely there are no isolated points in the set @xmath438 .
\(i ) take an interval @xmath432 $ ] with @xmath439 and let @xmath122 be the maximum of the process @xmath76 on this interval . then , since the process @xmath76 has independent increments , @xmath440 and @xmath441 are independent .
since @xmath440 has a continuous distribution , so has @xmath442 , and therefore @xmath443 for any @xmath444 .
taking @xmath445 and a union over all rational @xmath389 proves the claim for local maxima .
similarly the statement holds for local minima .
see figure [ fig : zeros and extrema ] .
having a zero and a local extremum at some @xmath59 ; this is an event of probability zero . ]
\(ii ) for any continuous function @xmath11 , any record time @xmath446 is a maximum of @xmath11 on the interval @xmath447 $ ] , for every @xmath448 .
let @xmath449 be an isolated point in @xmath450 .
then @xmath107 is a local maximum , because otherwise we would have record times to the right of @xmath107 , arbitrarily close to @xmath107 .
since @xmath11 is continuous and there are no record times in the interval @xmath451 for some @xmath452 , there has to be an @xmath453 , which is also a local maximum and such that @xmath454 and @xmath455 . applying these observations to @xmath456
, we see that , in order to prove the claim , it is enough to show that the process @xmath457 does not have two equal local maxima almost surely .
see figure [ fig : isolated record times ] .
this is well known for standard brownian motion and can be proven in the same way for the process @xmath76 .
namely , for two intervals @xmath458 $ ] and @xmath459 $ ] , with @xmath460 , define the random variables @xmath461 and @xmath462 , and let @xmath463 .
clearly these three random variables are independent . since @xmath464 is a continuous random variable ,
so is @xmath465 and @xmath466 .
therefore , almost surely the maxima on @xmath458 $ ] and @xmath459 $ ] are different . taking the union over all possible rational @xmath467 , @xmath468 , @xmath469 and @xmath470 as above proves the claim . has an isolated record time at time @xmath107 then there exists a record time @xmath454 for which @xmath471 . in this case
both record times @xmath472 and @xmath107 have to be local maxima .
this is an event of probability zero . ]
in this section we prove theorems [ set a_f has dim less than 1/2 ] , [ thm : hausdorff_lower_bound ] and [ thm : hausdorff_upper_bound ] .
recall that an interval @xmath37 is called _ dyadic _ if it is of the form @xmath82 $ ] for some integers @xmath83 and @xmath84 . for a dyadic interval @xmath37 , the set of its subintervals which are dyadic and of length @xmath473 , will be denoted by @xmath474 . for intervals
@xmath37 and @xmath85 we will write @xmath86 if @xmath85 is located to the right of @xmath37 .
[ proof of theorem [ set a_f has dim less than 1/2 ] ] recall the definition of sets @xmath475 and @xmath476 from part ( ii ) of proposition [ prop : basic_facts ] .
it is enough to prove that both of these sets have hausdorff dimension at most @xmath5 .
this follows from the case @xmath477 of part ( i ) of the following lemma , which estimates the hausdorff dimension of larger sets .
[ lemma : set_with_isolated_zeros ] ( i ) for a locally bounded function @xmath16 and @xmath164 the sets @xmath478 and @xmath479 have hausdorff dimension at most @xmath2 .
( ii ) assume @xmath16 has bounded variation on compact sets . for any @xmath164 the set @xmath480 has hausdorff dimension at most @xmath2 .
\(i ) since @xmath481 it is enough to prove the claim for @xmath482 .
first cover the set @xmath482 by sets @xmath483 , that is @xmath484 .
note that for any positive integer @xmath90 , and any @xmath90 points @xmath485 from @xmath486 such that @xmath487 , we have @xmath488 therefore , for any dyadic interval @xmath8 of length @xmath355 there is a constant @xmath348 such that , for any @xmath90 points @xmath489 from @xmath490 , we have @xmath491 now for any positive integer @xmath84 and any dyadic interval @xmath492 such that @xmath493 define @xmath494 and @xmath495 . the family @xmath496 : j \in { \mathcal{g}}_m(i ) , b_{f , n}^+ \cap j \neq \emptyset\}$ ] is a cover of the set @xmath497 with intervals of diameter at most @xmath498 . from ( [ eq : hausdorff estimates ] )
it is easy to see that @xmath499 since @xmath84 was arbitrary , by the definition of hausdorff dimension we obtain @xmath500 , for any dyadic interval @xmath37 of length @xmath355 .
therefore @xmath501 and taking the union over all @xmath125 gives @xmath502 .
\(ii ) it is enough to prove that for any @xmath422 the set @xmath503 has hausdorff dimension bounded by @xmath2 from above . representing @xmath504 as the union @xmath505 will prove the claim .
fix a dyadic interval @xmath37 and an @xmath448 . for
every @xmath506 we can find @xmath507 such that @xmath508 . denote the interval @xmath509 $ ] by @xmath510 .
clearly the family of intervals @xmath511 is a cover of the set @xmath512 . by besicovitch s covering theorem ( see theorem 2.6 in @xcite ) we can find an integer @xmath90 , not depending on @xmath399 , and a subcover which can be represented as a union of @xmath90 at most countable disjoint subfamilies @xmath513 . more precisely , there are sets @xmath514 , @xmath515 such that @xmath516 clearly for any @xmath517 we have @xmath518 where @xmath122 is the total variation of @xmath3 on the interval @xmath37 .
the last inequality follows form the fact that the intervals @xmath510 , @xmath519 are disjoint . now summing the above inequality over @xmath515 we obtain @xmath520 since @xmath521 is a cover of @xmath512 of diameter at most @xmath522 and neither @xmath90 nor @xmath122 depend on @xmath399 , the @xmath2-dimensional hausdorff measure of @xmath512 is finite .
therefore @xmath523 and , since the dyadic interval @xmath37 is arbitrary , the claim follows .
the proof of theorem [ thm : hausdorff_lower_bound ] will be an application of the percolation method due to hawkes @xcite , which we now describe .
see chapter 9 in @xcite for more on this method . fix a dyadic interval @xmath37 , a real number @xmath524 and set @xmath525 ( the construction and the results can be stated for any interval @xmath37 ) .
construct the set @xmath526 by including in it ( as subsets ) each of the two dyadic intervals from @xmath527 with probability @xmath387 , independently of each other . to construct @xmath528 , for each dyadic interval @xmath492 such that @xmath529 , include each of its dyadic subintervals from @xmath530 in @xmath528 with probability @xmath387 , independently of each other .
we obtain a decreasing sequence of compact sets @xmath531 whose intersection we denote by @xmath532 $ ] . comparing this to the galton - watson process with binomial offspring distribution @xmath533
, we see that @xmath534 \neq \emptyset ) > 0 $ ] .
the following theorem is due to hawkes ( theorem 6 in @xcite ) . for any set @xmath535 ,
if @xmath536 \neq \emptyset ) > 0 $ ] then @xmath537 . in the above construction of the percolation
set we can change the retention probabilities of intervals at each level .
if @xmath538 is a retention probability at level @xmath125 , we denote the union of intervals kept at level @xmath84 by @xmath539 , and the limiting percolation set by @xmath540 $ ] .
we will use the following result which is an easy corollary of hawkes theorem .
[ cor : hawkes ] let @xmath541 and let @xmath542 be a sequence converging to @xmath543 .
for any set @xmath535 , if @xmath544 \neq \emptyset)>0 $ ] then @xmath537 .
let @xmath545 and @xmath546 be such that @xmath547 for all @xmath548 .
there is a realization @xmath6 of @xmath549 for which @xmath544\neq \emptyset \mid s_{(\beta_n)}(m_0 ) = b)>0 $ ] .
we have @xmath550 \neq \emptyset \mid s_{(\beta_n)}(m_0 ) = b ) \leq \mathbb{p}(a \cap \gamma[\alpha ] \neq \emptyset \mid s_\alpha(m_0 ) = b),\ ] ] since we can couple two percolation processes so that for @xmath551 , if an interval @xmath492 is retained in the percolation process with retention probabilities @xmath552 , it is also retained in the process with retention probability @xmath553 . since @xmath554 we get @xmath555 \neq \emptyset)>0 $ ] and , by hawkes theorem , @xmath556 . since @xmath545 was arbitrary the claim follows .
it is enough to prove that the hausdorff dimension of @xmath557 $ ] is greater than or equal to @xmath5 with positive probability .
let @xmath542 be a sequence converging to @xmath5 from below , to be chosen later .
consider the percolation process on the interval @xmath215 $ ] with retention probabilities @xmath552 , independent of brownian motion .
fix a positive integer @xmath84 . for a dyadic interval @xmath558)$ ]
denote by @xmath559 its center and , for a fixed @xmath157 , consider the event @xmath560 define @xmath561)}\mathbf{1}({f}_{m,{\epsilon}}(i))$ ] , where @xmath562 is the indicator function of the event @xmath563 .
using trivial bounds on the transition density of brownian motion , the first moment of @xmath564 can be estimated simply by @xmath565 for some constant @xmath348 depending only on @xmath566}|f(t)|$ ] , and where @xmath567 .
in the same way , for @xmath242 we can estimate the conditional probability @xmath568 for @xmath569)$ ] such that @xmath86 let @xmath289 be the largest integer so that both @xmath37 and @xmath85 are contained in a single interval from @xmath570)$ ] . in other words
there are consecutive intervals @xmath571)$ ] , contained in a single interval from @xmath570)$ ] , such that @xmath572 and @xmath573 .
then @xmath574 using independence , ( [ eq : conditional hawkes ] ) and ( [ eq : conditional hawkes 2 ] ) @xmath575 summing ( [ eq : second_moment_help_any_drift ] ) over all @xmath572 and @xmath573 for a fixed @xmath254 and @xmath255 as above @xmath576 for some universal constant @xmath281 . summing this over all @xmath254 and @xmath255 and all @xmath577 , and using ( [ eq : first_moment_any_drift ] )
, we can estimate the second moment @xmath578 ) , i < j } \mathbb{p}({f}_{m,{\epsilon}}(i ) \cap { f}_{m,{\epsilon}}(j ) ) + \mathbb{e}(y_{m,\epsilon } ) \nonumber \\ & \leq 8c_2\epsilon^2\sum_{\ell=1}^{m-1 } \big(2^\ell 2^{2(m-\gamma_m ) + \gamma_\ell-3\ell/2 } \big ) + 2\epsilon 2^{m-\gamma_m } \nonumber \\ & \leq 8c_2 \epsilon^2 2^{2(m-\gamma_m)}\sum_{\ell=0}^{m-1 } 2^{\gamma_\ell-\ell/2 } + 2\epsilon 2^{m-\gamma_m}.\end{aligned}\ ] ] now choose a sequence @xmath542 which converges to @xmath5 from below , and so that the series @xmath579 converges ( for example take @xmath580 , then @xmath581 is up to an additive constant equal to @xmath582 ) . with such @xmath542 and for the sequence @xmath583 that converges to zero ,
we define @xmath584 .
by ( [ eq : first_moment_any_drift ] ) we have @xmath585 and by ( [ eq : second_moment_any_drift ] ) we have @xmath586 , where @xmath587 is a universal constant . remark [ 6.30.1 ] ( ii ) yields @xmath588 .
thus the probability of the event @xmath589 is also bounded from below by @xmath590 . on this event
there is a sequence @xmath591 , such that each @xmath592 is the center of a dyadic interval from @xmath593)$ ] contained in @xmath594 , and such that @xmath595 .
the sequence @xmath591 contains a subsequence that converges to some @xmath596 $ ] such that @xmath597 .
therefore we have @xmath598 \neq \emptyset ) \geq c_1 ^ 2/c_3 > 0 $ ] . by the independence of brownian motion and the percolation process and corollary [ cor : hawkes ]
it follows that @xmath599 [ rem : uniform bound for holder ] note that , for a continuous function @xmath3 defined on @xmath215 $ ] , the lower bound from depends only on @xmath566}|f(t)|$ ] .
it is enough to prove that @xmath600 , for any dyadic interval @xmath37 .
first we will prove the claim for functions which are @xmath5-hlder on compact intervals .
assume @xmath3 is a @xmath5-hlder continuous function on @xmath37 , that is @xmath601 for some @xmath602 and all @xmath603 .
for an interval @xmath604 \subset i$ ] set @xmath605 , if there exists @xmath606 such that @xmath607 , and @xmath608 otherwise , and define the interval @xmath609 $ ] . on the event @xmath610 define the stopping time @xmath611 . since @xmath612 , by the strong markov property , conditional on the sigma algebra @xmath613 and on the event @xmath614 , the probability @xmath615 that @xmath616 is equal to the probability that @xmath617 , where @xmath618 is the interval @xmath267 shifted by @xmath619 and scaled by @xmath620 .
since the interval @xmath618 has length at least @xmath621 and contains the origin , @xmath615 is bounded from below by a constant not depending on the choice of the interval @xmath85 ; see also arguments in the proof of theorem [ prop : hitting characterisation ] leading to ( [ eq : approximation of probabilities_1 - 5 ] ) . therefore , for some @xmath622 we obtain @xmath623 .
this implies @xmath624 for some @xmath281 .
now consider the covering @xmath625 of the set @xmath626 , consisting of the dyadic intervals from @xmath627 which intersect the set @xmath626 .
since every interval in @xmath625 has length @xmath628 , by ( [ eq : hausdorff conditional ] ) we have @xmath629 fatou s lemma implies @xmath630 therefore , almost surely , we can find a sequence of coverings @xmath631 of @xmath626 , with @xmath632 and @xmath633 .
this implies that @xmath634 , a.s .
now let @xmath3 be of bounded variation on compact intervals and define the set @xmath635 by part ( ii ) of lemma [ lemma : set_with_isolated_zeros ] we have @xmath636 .
therefore , it is enough to bound the dimension of the zero set in the complement , @xmath637 , where @xmath638 .
we can cover @xmath639 by a countable union of the sets @xmath640 for a fixed @xmath125 , all @xmath641 with @xmath642 satisfy @xmath643 .
therefore , the restriction @xmath644 is @xmath5-hlder continuous , that is , for some positive constant @xmath293 and all @xmath645 we have @xmath646 .
define the function @xmath647 to be @xmath648 for all @xmath649 .
we will define @xmath650 for @xmath651 using linear interpolation , in a sense . if @xmath652 then set @xmath650 to be any constant function .
assume that @xmath653 .
note that the set @xmath654 is closed from the left , that is if @xmath233 is an increasing sequence of points in @xmath654 converging to some @xmath107 , then @xmath655 . since @xmath3 has bounded variation , by assumption , the right limit @xmath656 exists for all @xmath657 . thus if @xmath658 define @xmath659 , @xmath660 as well as @xmath661 and @xmath662 and notice that @xmath663 .
now define @xmath650 on @xmath664 $ ] to be linear with @xmath665 and @xmath666 .
if @xmath667 does not exist , define @xmath667 as the left endpoint of @xmath37 and @xmath668 , and do similarly if @xmath669 does not exist .
clearly @xmath650 is @xmath5-hlder continuous with the same constant @xmath293 .
we can apply the first part of the proof to @xmath670 , to conclude that the set @xmath671 has hausdorff dimension at most @xmath5 almost surely . since @xmath672
, we obtain @xmath673 and @xmath674 . since @xmath675 the claim follows .
we finish the paper with the proof of corollary [ cor : sharp_haudorff_for_holder ] .
assume that @xmath676 for all @xmath677 .
for every positive integer @xmath125 define a continuous function @xmath678 \to \mathbb{r}$ ] as @xmath679 , for all @xmath680 $ ] .
clearly @xmath681}|f_n(t)| \leq c\sqrt{2}$ ] and , by remark [ rem : uniform bound for holder ] and brownian scaling , there is a constant @xmath422 such that @xmath682 ) \geq 1/2 ) = \mathbb{p}(\dim(\operatorname{\mathcal{z}}(b - f_n ) ) \geq 1/2 ) \geq c.\ ] ] thus by blumenthal s zero - one law , almost surely , there are infinitely many integers @xmath125 such that the hausdorff dimension of the set @xmath683 $ ] is greater or equal than @xmath5 .
the upper bound follows from theorem [ thm : hausdorff_upper_bound ] .
authors would like to thank david aldous , zhen - qing chen , serguei denissov and steven evans for helpful discussions and references .
toni antunovi and julia ruscher would like to thank microsoft research where this work was completed .
krzysztof burdzy was supported in part by nsf grant dms-0906743 and by grant n n201 397137 , mnisw , poland .
a. dvoretzky , p. erds , and s. kakutani .
nonincrease everywhere of the brownian motion process . in _ proc .
4th berkeley sympos .
math . statist . and prob .
, pages 103116 .
california press , berkeley , calif . ,
|
it is well known that standard one - dimensional brownian motion @xmath0 has no isolated zeros almost surely .
we show that for any @xmath1 there are @xmath2-hlder continuous functions @xmath3 for which the process @xmath4 has isolated zeros with positive probability .
we also prove that for any continuous function @xmath3 , the zero set of @xmath4 has hausdorff dimension at least @xmath5 with positive probability , and @xmath5 is an upper bound on the hausdorff dimension if @xmath3 is @xmath5-hlder continuous or of bounded variation .
|
the study of kinetic growth models such as directed percolation , eden growth , ballistic deposition , diffusion limited aggregation , random deposition with and without relaxation , cluster - cluster aggregation , etc .
, is motivated by their interest in many areas of scientific research and technology such as polymer science , crystal and polycrystalline growth , gelation , fracture propagation , epidemic spreading , bacterial and fungi growth colonies , colloids , etc . @xcite . within this context the eden model @xcite
has become an archetype growth model .
eden clusters are compact but the self - affinity that characterizes the behavior of the growing interface is of much interest ( see e.g. @xcite ) .
few years ago ausloos et al .
@xcite have introduced an additional degree of freedom to the eden model , namely the spin of the added particles .
more recently , the eden growth of clusters of charged particles has also been studied @xcite . in the magnetic eden model ( mem )
@xcite with spins having two orientations ( up and down ) the growth of the cluster starts from a single seed , e.g. a spin up seed , placed at the center of the two - dimensional square lattice , whose sites are labelled by their rectangular coordinates @xmath5 . then
, the growth process of the resulting magnetic cluster consists in adding further spins to the growing cluster taking into account the corresponding interaction energies . by analogy to the ising model @xcite one takes @xmath6 as the coupling constant between nearest - neighbor ( nn ) spins @xmath7 and the energy @xmath8 is then given by @xmath9 where @xmath10 means that the summation is taken over occupied nn sites .
the spins can assume two values , namely @xmath11 . throughout this work we set the boltzmann constant equal to unity ( @xmath12 )
, we consider @xmath13 ( i.e. , the ferromagnetic case ) and we take the absolute temperature @xmath14 measured in units of @xmath6 . in the mem a spin is added to the cluster with a probability proportional to the boltzmann factor exp@xmath15 , where @xmath16 is the total energy change involved . it should be noted that at each step all sites of perimeter are considered and the probabilities of adding up and down spins have to be evaluated . after proper normalization of the probabilities the growing site and the orientation of the spin
are determined through a pseudo - random number generator .
it is worth mentioning that the mem has originally been motivated by the study of the structural properties of magnetically textured materials @xcite . while these previous studies of the mem were mainly devoted to determine the lacunarity exponent and the fractal dimension of the set of parallel oriented spins @xcite , the aim of the present work is to complement these previous investigations by studying the critical behavior of the mem using extensive monte carlo simulations and applying a finite - size scaling theory .
also , our study is performed in confined ( stripped ) geometries which resemble recent experiments where the growth of quasi - one - dimensional strips of fe on a cu(111 ) vicinal surface @xcite and fe on a w(110 ) stepped substratum @xcite have been performed .
in fact , the preparation and characterization of magnetized nanowires is of great interest for the development of advanced microelectronic devices @xcite .
furthermore , the growth of metallic multilayers of ni and co separated by a cu spacer layer has recently been also studied @xcite .
another goal of the present work is to compare the results obtained for the mem with the well known behavior of the classical ising model @xcite , an archetypical model in the study of thermally driven ( reversible ) phase transitions in equilibrium systems .
the ising hamiltonian @xmath17 is given by @xmath18 where @xmath10 means that the summation runs over all nn sites , @xmath19 is the state of the spin at the site of coordinates @xmath5 and @xmath6 is the coupling constant ( @xmath13 ) .
the mem is also similar to a family of models for the stochastic growth of crystals generically known as crystal growth models ( cgm ) @xcite , for a review see e.g. @xcite . as in the mem , in the case of cgm each atom
is adsorbed with a given probability conditional to the actual configuration of neighboring atoms on the previous layer(s ) .
however , in contrast to the mem , the crystal is supposed to grow layer after layer .
it should also be noticed that relationships established between cgm and a special class of ising models @xcite have allowed to derive exact results .
therefore , useful comparisons with the mem will be also discussed in the presentation of our results .
this paper is organized as follows : in section 2 we give details on the simulation method , section 3 is devoted to the presentation and discussion of the results obtained for the mem in @xmath20-dimensions , while section 4 refers to results corresponding to @xmath21-dimensions . in sections 3 and 4 ,
detailed discussions comparing our results with the behavior of the ising magnet are outlined .
finally our conclusions are stated in section 5 .
the mem in @xmath22dimensions is studied in the square lattice using a rectangular geometry @xmath23 with @xmath24 and imposing periodic boundary conditions along the @xmath25direction .
the location of each site on the lattice is specified through its rectangular coordinates @xmath5 , ( @xmath26 , @xmath27 .
the starting seed for the growing cluster is a column of parallel oriented spins placed at @xmath28 .
it should be noticed that previous simulations of the mem were restricted to rather modest cluster sizes , i.e. containing up to 8000 spins @xcite , while in the present work clusters having up to @xmath29 spins have been typically grown .
we have also studied the mem in @xmath30dimensions employing a @xmath31 geometry ( @xmath24 ) with periodic boundary conditions along both @xmath25directions .
magnetic eden clusters grown on a stripped geometry of finite linear dimension @xmath33 at low temperatures show an interesting behavior that we call magnetization reversal .
in fact , we have observed that long clusters are constituted by a sequence of well ordered magnetic domains .
spins belonging to each domain , of average length @xmath34 , have mostly the same orientation and consecutive domains have opposite orientation .
let @xmath35 be the characteristic length for the occurrence of the magnetization reversal .
since @xmath36 , we then conclude that the problem has two characteristic length scales , namely @xmath37 and @xmath35 , such that @xmath38 . in ordinary thermally driven phase transitions , the system changes from a disordered state at high temperatures to a spontaneously ordered state at temperatures below some critical value @xmath39 where a second - order phase transition takes place .
regarding the ising model , one has that , in the absence of an external magnetic field ( @xmath40 ) , the low temperature ordered phase is a state with non - vanishing spontaneous magnetization ( @xmath41 ) .
this spontaneous symmetry breaking is possible in the thermodynamic limit only .
in fact , it is found that the magnetization @xmath42 of a finite sample can pass with a finite probability from a value near @xmath43 to another near @xmath44 , as well as in the opposite direction .
consequently , the magnetization of a finite system , averaged over a sufficiently large observation time , vanishes at every positive temperature .
the equation @xmath45 holds if the observation time ( @xmath46 ) becomes larger than the ergodic time ( @xmath47 ) , which is defined as the time needed to observe the system passing from @xmath48 to @xmath49 .
increasing the size of the sample the ergodic time increases too , such that in the thermodynamic limit ergodicity is broken due to the divergence of the ergodic time , yielding broken symmetry . since monte carlo simulations are restricted to finite samples , the standard procedure to avoid the problems treated in the foregoing discussion is to consider the absolute magnetization as an order parameter @xcite . turning back to the mem , we find that the phenomenon of magnetization reversal also causes the magnetization of the whole cluster to vanish at every non - zero temperature , provided that the length of the cluster @xmath50 ( which plays the role of @xmath46 ) is much larger than @xmath37 ( which plays the role of @xmath47 ) .
therefore , we have measured the mean absolute column magnetization , given by @xmath51 in the stripped geometry used in this work the bias introduced by the lineal seed ( a starting column made up entirely of up spins ) can be avoided by calculating relevant properties after disregarding spins within a distance approximately equal to few times @xmath33 from the seed .
the procedure of column averaging out from the transient region represents a significant advantage of the stripped geometry used for the simulation of the mem .
in fact , when a single seed at the center of the sample is used , the definition of the average magnetization of the whole cluster is strongly biased by the cluster s kernel orientation at the early stages of the growing process .
in addition , using several randomly generated seeds we could also establish that the system evolves into a given stationary state independently of the seed employed .
= 4.0in=3.2 in the mean column magnetization is a fluctuating quantity that can assume @xmath52 values .
then , for given values of both @xmath33 and @xmath14 , the probability distribution of the mean column magnetization @xmath53 can be evaluated , since it represents the normalized histogram of @xmath54 taken over a sufficiently large number of columns in the stationary region @xcite . in the thermodynamic limit
the probability distribution @xmath55 of the order parameter of an equilibrium system at criticality is universal ( up to rescaling of the order parameter ) and thus it contains very useful and interesting information on the universality class of the system @xcite .
figure 1 shows the thermal dependence of @xmath56 for a fixed lattice size ( @xmath57 ) as obtained for the mem . at high temperatures
@xmath56 is a gaussian centered at @xmath58 but when the temperature gets lowered , the distribution broadens and develops two peaks at @xmath59 and @xmath60 .
further decreasing the temperature causes these peaks to become dominant while the distribution turns distinctly non - gaussian , exhibiting a minimum just at @xmath58 .
the emergence of the maxima at @xmath61 is quite abrupt .
this behavior reminds us the order parameter probability distribution characteristic of the one dimensional ising model .
in fact , for the well studied @xmath62dimensional ising model @xcite , we know that for @xmath63 , @xmath64 is a gaussian centered at @xmath65 , given by @xmath66 where the susceptibility @xmath67 is related to order parameter fluctuations by @xmath68 = 4.0in=3.0 in = 4.0in=3.0 in decreasing temperature the order parameter probability distribution broadens , it becomes non - gaussian , and near @xmath39 it splits into two peaks that get the more separated the lower the temperature . for @xmath69 and linear dimensions @xmath33 much larger than the correlation length @xmath70 of order parameter fluctuations ,
one may approximate @xmath64 near the peaks by a double - gaussian distribution , i.e. @xmath71 where @xmath72 is the spontaneous magnetization , while the susceptibility @xmath67 is now given by @xmath73 from equation ( 4 ) it turns out that the gaussian squared width @xmath74 associated with high temperature distributions is very close to the 2nd moment of the order parameter , i.e. @xmath75 equation ( 8) is a consequence of the gaussian shape of the order parameter probability distribution and , thus , it holds for the mem as well . from the known one - dimensional exact solution for a chain of @xmath33
spins @xcite one can obtain @xmath76 then , equations ( 5 ) and ( 9 ) lead us to @xmath77 ( where it has been taken into account that @xmath78 due to finite - size effects , irrespective of temperature ) . from equations ( 8) and ( 10 ) we can see that the high temperature gaussian probability distribution broadens exponentially as @xmath14 gets lowered , until it develops delta - like peaks at @xmath79 as a consequence of a boundary effect on the widely extended distribution . it should be noticed that for @xmath80 this phenomenon is prevented by the finite critical temperature which splits the gaussian , as implied by equation ( 6 ) . turning back to the mem , figure 1
strongly suggests that an analogous mechanism should be responsible for the thermal dependence exhibited by the mem s order parameter distribution function .
so , by analogy to equation ( 9 ) , we assume the relation @xmath81 to hold for the mem , where we have introduced a phenomenological parameter @xmath82 , and the susceptibility @xmath67 is given by equation ( 5 ) .
we find an excellent agreement to the data by choosing the value @xmath83 as observed in figure 2(a ) , where ln - linear plots of @xmath84 versus @xmath85 are shown for strip widths varying in the range @xmath86 .
figure 2(b ) shows plots of @xmath87 versus @xmath14 for the same lattices .
this figure shows that increasing @xmath33 the order parameter curves approach the one that corresponds to the thermodynamic limit ( i.e. , @xmath88 , where @xmath89 is the heaviside function ) .
= 4.0in=3.0 in = 4.0in=3.0 in however , it should be pointed out that the results obtained for the @xmath90-mem and the @xmath91-dimensional ising model do not exactly coincide for finite lattices , as figure 3(a ) shows for the case of the magnetization . anyway , this fact should not alarm us , since it can be seen that differences in the results obtained for both models are a direct consequence of the finite - size nature of the lattices used in the simulations and consequently they tend to vanish in the thermodynamic limit .
this is actually shown by figure 3(b ) , where log - linear plots of @xmath92 versus @xmath93 for two different fixed values of temperature are presented .
thus , we conclude that in view of the full qualitative and quantitative agreement between both models we can safely establish that , as in the 1d ising model , the @xmath90-mem is not critical ( i.e. it also undergoes a phase transition at @xmath94 ) .
= 4.0in=3.0 in we have also computed the number of already occupied nn sites every time that a new particle was added to the spin system , and thus we have obtained the normalized probability @xmath95 of having @xmath96 occupied nn sites .
figure 4 shows the behavior of @xmath95 as a function of temperature . using this probability
we have evaluated @xmath97 irrespective of the temperature .
this result can be understood considering that the growing process that leads to the assignment of a spin @xmath19 to each lattice site of coordinates @xmath5 can be studied by means of a bond model .
in fact , we can assign a bond to each pair of neighboring sites , pointing from the earlier occupied site to the later occupied one .
so , the process that leads to a given spin configuration can be specified by the fields @xmath98 and @xmath99 , where sub - indexes @xmath100 and @xmath101 refer to the upper bond of @xmath5 ( i.e. , the bond that connects the site of coordinates @xmath5 with that of coordinates @xmath102 ) , and to the right bond of the site of coordinates @xmath5 , respectively @xcite .
we take @xmath103 if the bond points outwards and @xmath104 if it is directed inwards . therefore
, the net bond flux at a given lattice site @xmath5 is given by : @xmath105 and the possible values that @xmath106 can take are @xmath107 .
after some algebra , it follows that @xmath108 holds for every site on the lattice .
moreover , it can be seen that , for an arbitrary @xmath109-dimensional lattice of coordination number @xmath110 , @xmath111 . then , @xmath112 is the mean number of occupied nn , since @xmath113 .
for the two - dimensional square lattice , @xmath114 and equation ( 13 ) yields @xmath115 , in agreement with the result we have already obtained by means of monte carlo simulations .
= 4.0in=3.0 in further insights into the mem s growing process can be gained by studying the mean energy change involved in the addition of a new particle to the system . the process of adding a new spin involves an energy change @xmath16 and from the definition of the @xmath32mem , the possible values that @xmath16 can take are @xmath116 ( in units of @xmath6 ) .
figure 5 shows plots of the normalized probability @xmath117 versus @xmath14 for each of these values .
the non - equilibrium nature of the mem manifests itself through much more complex probability distributions @xmath117 ( see figure 5 ) than the corresponding to the equilibrium ising model where @xmath16 can take only five different values , namely @xmath118 ( in units of @xmath6 ) .
the results shown in figures 4 and 5 confirm the non - trivial nature of the link established between the mem at stationarity and the ising model in equilibrium .
it should be noticed that for the case of crystal growth models ( cgm ) @xcite the growing conditions are quite different than those of the mem .
in fact , in cgm the crystal grows layer by layer in a given direction @xcite .
furthermore , the probability distribution of the predecessor spin layer is sampled from the * equilibrium * distribution , so will be the probability of spins in subsequent layers .
this particular growth mechanism allow to establish dual transformations with the kinetic ising model @xcite and to extract some exact results .
in contrast , the growing interface of the mem is self - affine and the system is far from equilibrium .
so , the link between the 1d ising model and the @xmath22dimensional mem is quite challenging .
= 4.0in=3.1 in in order to compare the @xmath119dimensional mem and the 2d ising model , we have first studied the order parameter probability distribution @xmath120 , where @xmath121 takes now @xmath122 possible values ( see figure 6 ) . for high temperatures , the probability distribution corresponds to a gaussian centered at @xmath123 . at lower temperatures we observe the onset of two maxima located at @xmath124 @xmath125 , which become sharper and
approach @xmath126 as @xmath14 is gradually decreased . = 4.0in=3.0 in figure 7 shows the location of the maximum of the probability distribution as a function of temperature for both @xmath127-dimensional mem models , with @xmath128 . while for the @xmath129 case we observe a smooth transition from the @xmath130 value characteristic of high temperatures to nonzero @xmath131 values that correspond to lower temperatures , the curve obtained for @xmath132 shows , in contrast , a heaviside - like jump . as already discussed , the behavior exhibited by the @xmath133-dimensional mem ( e.g. , as displayed by figures 6 and 7 ) is the signature of a thermal continuous phase transition that takes place at a finite critical temperature .
the broken symmetry at a finite critical temperature @xmath39 implied by the thermal continuous phase transition can be explained in terms of the broken ergodicity that occurs in the system when we tend to the thermodynamic limit ( @xmath134 ) making use of the temperature dependence exhibited by the order parameter distribution function .
in fact , if we set the characteristic length of mem s domains @xmath37 equal to an ergodic length @xmath135 , we can carry out a complete analogy with the ising model by associating @xmath135 to @xmath47 ( the ising model ergodic time ) and the cluster s total length @xmath50 ( already defined in section 3 ) to the ising model observation time @xmath46 . in this way
, we encounter that excursions of @xmath121 from @xmath136 to @xmath137 and _ vice versa _ occur at length scales of the order of @xmath135 . when the cluster s total length becomes larger and larger ( @xmath138 ) the whole cluster s magnetization is averaged to zero .
furthermore , @xmath135 diverges as the strip s width becomes larger and larger , and again broken symmetry arises as the consequence of broken ergodicity . as already anticipated and as it follows from figures 6 and 7 , the @xmath133-dimensional mem exhibits a thermally driven order - disorder transition at a finite temperature . in the thermodynamic limit ( @xmath134 ) we expect to determine a critical temperature @xmath39 such that @xmath139 for @xmath63 while @xmath140 remain non - vanishing at temperatures below @xmath39 .
from the finite - size scaling theory , developed for the treatment of finite - size effects at criticality and under equilibrium conditions @xcite , it is well known that if a thermally driven phase transition occurs at a temperature @xmath141 in the thermodynamic limit , then in a confined geometry this transition becomes smeared out over the temperature region @xmath142 around a shifted effective transition temperature @xmath143 , and the following relationships hold : @xmath144 and @xmath145 where the rounding and shift exponents are given by @xmath146 , respectively ( recalling that @xmath147 is the exponent that characterizes the divergence of the correlation length at criticality ) . furthermore , from well established finite - size scaling relations , the following anstze hold just at criticality @xmath148 : @xmath149 and @xmath150 where @xmath151 and @xmath152 are the order parameter and the susceptibility critical exponents , respectively .
note that @xmath153 , as given by equation ( 17 ) , refers to the maximum of @xmath154 as a function of @xmath14 for fixed lattice size @xmath33 .
= 4.0in=3.0 in in view of the encountered analogies between the mem and the ising model , it is natural to test the validity of equations ( 14 - 17 ) for the case of the mem in @xmath21-dimensions .
it should be noted that as in the case of equilibrium systems , in the present case various effective " l - dependent critical temperatures can also be defined . in particular
, we will define @xmath155 as the value that corresponds to @xmath156 for fixed @xmath33 , and @xmath157 as the one corresponding to the maximum of the susceptibility for a given @xmath33 , assuming that the susceptibility is related to order parameter fluctuations in the same manner as for equilibrium systems ( as given by equations ( 5 ) and ( 7 ) ) .
then , we should be able to obtain @xmath39 from plots of @xmath158 versus @xmath93 ( for @xmath159 ) , as it is shown in figure 8 . following this procedure we find that both @xmath155 and @xmath157 extrapolate ( approximately ) to the same value , allowing us to evaluate the critical temperature @xmath160 in the thermodynamic limit .
after determining @xmath39 , the correlation length exponent @xmath147 can be evaluated by means of equation ( 15 ) , making the replacement @xmath161 .
in fact , taking @xmath162 at the mean , maximum and minimum values allowed by the error bars , we obtain six log - log plots of @xmath163 versus @xmath33 for @xmath159 .
the slope of each of these plots , not shown here for the sake of space , yields a value for @xmath147 .
the obtained values are : @xmath164 @xmath165 and @xmath166 @xmath167 thus our estimate is given by @xmath168 , where the error bars reflect the error derived from the evaluation of @xmath39 , as well as the statistical error .
= 4.0in=3.0 in figure 9 shows plots of the susceptibility versus @xmath14 as obtained using lattices of different side .
it is found that the susceptibility exhibits a peak which becomes sharper and shifts towards lower temperatures when @xmath33 is increased .
this behavior is , in fact , already anticipated by equation ( 17 ) , and it allows us to evaluate @xmath169 from the slope of a log - log plot of @xmath170 versus @xmath33 , as figure 10 shows . the linear fit yields @xmath171 . using this value and the value formerly obtained for @xmath147 we thus determine @xmath172 .
= 4.0in=3.0 in = 4.0in=3.0 in figure 11 shows log - log plots of @xmath173 versus @xmath33 for the mean , maximum and minimum allowed values of @xmath39 .
considering only the larger lattices , the linear fits to the data according to equation ( 16 ) yield the following estimates : @xmath174 , @xmath175 and @xmath176 .
we then assume the value @xmath177 , where the error bars reflect the error derived from the evaluation of @xmath39 , as well as the statistical error . from this value and the value formerly obtained for @xmath147
we thus determine @xmath178 .
the critical exponents of the mem in @xmath119dimensions as obtained using a finite - size scaling analysis are , so far : @xmath168 , @xmath179 and @xmath180 .
if we recall the exactly known critical exponents of the @xmath4 ising model , i.e. @xmath181 , @xmath182 and @xmath183 , we find that the @xmath30mem has the same critical exponents within error bars . these results
further support our conjecture on the connection between the mem in @xmath21-dimensions and the ising model in @xmath184 dimensions . as in the case of the mem in @xmath185 ,
we have also computed the number of already occupied nn sites every time that a new particle was added to the spin system .
we found that the value @xmath186 holds for all temperatures , which is indeed the result given by equation ( 13 ) , since @xmath187 for the three - dimensional square lattice . at this stage , we may recall that for the @xmath32mem @xmath188 equals the coordination number of the @xmath132 ising model , and that we found that both models have the same critical temperature and exhibit the same critical behavior .
reasoning by analogy , we may expect a coincidence between the critical temperature for the @xmath30mem and the corresponding one for a @xmath129 ising model defined on a lattice of coordination number @xmath189 . however , this comparison can not be carried out , since the critical temperature of an ising model depends on both the coordination number @xmath110 and the topological structure of the lattice , but for @xmath80 and a given value of @xmath110 the topological structure is not unique .
for instance , for @xmath129 and @xmath189 , we can pass from the honeycomb lattice ( hl ) to the expanded kagom lattice ( ekl ) through the application of a star - triangle transformation and obtain the exact values of their critical points , which turn out to be @xcite @xmath190 ( hl ) and @xmath191 ( ekl ) .
in the present work we have studied the growth of magnetic eden clusters with ferromagnetic interactions between nearest neighbor spins in a @xmath0dimensional rectangular geometry ( for @xmath128 ) , using monte carlo simulations and applying a finite - size scaling theory .
the results obtained allow us to conjecture a nontrivial correspondence between the mem for the irreversible growth of magnetic materials and the classical ising model under equilibrium conditions .
in fact , we have found that the @xmath0dimensional mem and the @xmath62dimensional ising model behave identically ( unless finite - size differences that vanish in the thermodynamic limit ) at criticality , i.e. that both models belong to the same universality class .
we also conjecture that this correspondence would remain at higher dimensions @xmath192 .
the results obtained strongly suggest a link between the temporal evolution of equilibrium systems and the stationary growth of nonequilibrium systems .
we thus believe that this work will stimulate further developments in the field of nonequilibrium kinetic growth models . a more precise numerical test of the posed conjecture will certainly require a considerable computational effort but it will be of great interest .
furthermore , analytical developments aimed to establish a theoretical framework for the understanding of far from equilibrium growth phenomena will become stimulated by the reported findings .
this work was supported financially by conicet , unlp , cic ( bs . as . ) , anpcyt and fundacin antorchas ( argentina ) and the volkswagen foundation ( germany ) .
the authors thank m. muoz for helpful discussions and his proposal of the bond representation for the mem .
99 f. family and t. vicsek , dynamics of fractal surfaces .
world scientific , singapore ( 1991 ) .
a. l. barabasi and h. e. stanley , fractal concepts in surface growth .
cambridge university press , new york ( 1995 ) .
fractals and disordered media , eds .
a. bunde and s. havlin , springer - verlag , heidelberg ( 1991 ) .
fractals in science , eds .
a. bunde and s. havlin , springer - verlag , heidelberg ( 1995 ) .
m. marsili , a. maritan , f. toigo and j. r. banava .
phys . * 68 * , 963 ( 1996 ) .
m. eden , in : symp . on information
theory in biology , edited by h. p. yockey , pergamon press , new york ( 1958 ) and proceedings of the fourth berkeley symposium on mathematics , statistics and probability , edited by f. neyman university of california press , berkeley ( 1961 ) .
iv , p. 223 .
d. e. wolf and j. kertsz .
. lett . * 4 * , 651 ( 1987 ) .
m. cieplak , a. maritan and j. r. banavar .
lett . * 76 * , 3754 ( 1996 ) . c. s. ryu and in - mook kim , phys .
e. * 53 * , 5643 ( 1996 ) .
s. s. manna and d. dhar , phys .
e , * 54 * , r3063 ( 1996 ) .
e. v. albano , phys .
e. * 56 * , 7301 ( 1997 ) .
m. marsili and m. vendruscolo .
. lett . * 37 * , 505 ( 1997 ) .
m. ausloos , n. vandewalle and r. cloots .
lett . , * 24 * , 629 ( 1993 ) , n. vandewalle and m. ausloos .
e. * 50 * , r635 ( 1994 ) .
y. v. ivanenko , n. i. lebovka and n. v. vygornitskii , eur .
j. b. * 11 * , 469 ( 1999 ) .
e. ising , z. phys .
* 31 * , 253 ( 1925 ) .
j. shen et al .
b. * 56 * , 2340 ( 1997 ) .
o. pietzsch , a. kubetzka , m. bode and r. wiesendanger .
lett . * 84 * , 5212 ( 2000 ) .
j. hauschild , u. gradmann and h. j. elmers .
. lett . * 72 * , 3211 ( 1998 ) . h.
j. elmers , j. hauschild and u. gradmann .
b. * 59 * , 3688 ( 1999 ) .
u. bovensiepen et al .
lett . * 81 * , 2368 ( 1998 ) .
b. m. mccoy and t. t. wu . the two dimensional ising model .
harvard university press , cambridge , ma .
t. r. welberry and r. galbraith .
* 6 * , 87 ( 1973 ) .
a. m. verhagen .
. phys . * 15 * , 219 ( 1976 ) . i. g. enting .
j. phys . c. ( solid state physics ) . *
10 * , 1379 ( 1977 ) .
* 29 * , 231 ( 1982 ) .
. phys . * 34 * , 615 ( 1984 ) .
phys . * 49 * , 139 ( 1987 ) .
j. stephenson , phys .
b. * 1 * , 4405 ( 1970 ) and j. math . phys . * 11 * , 420 ( 1970 ) .
k. binder and d. w. heermann , in monte carlo simulation in statistical physics , springer series in solid state sciences 80 , springer - verlag , berlin ( 1992 ) .
k. binder and d. stauffer , a simple introduction to monte carlo simulation and some specialized topics , in applications of the monte carlo method in statistical physics , ed .
k. binder , 2nd edition .
springer - verlag , berlin ( 1987 ) .
k. binder , monte carlo methods , in encyclopedia of appl .
vch publishers inc .
, heidelberg ( 1994 ) .
k. binder , introduction to monte carlo methods , chapter 5 , p.124 , in conference proceedings vol .
49 , monte carlo and molecular dynamics of condensed matter systems , eds .
k. binder and g. ciccotti , sif , bologna ( 1996 ) .
k. binder .
b * 43 * , 119 ( 1981 ) .
a. d. bruce , j. phys . c.
* 14 * , 3667 ( 1981 ) . m. m. tsypin and h. w. j. blte , phys .
e. * 62 * , 73 ( 2000 ) .
d. p. landau and e. m. lifshitz , statistical physics , 3rd edition , part 1 .
pergamon press , oxford .
r. j. baxter , exactly solved models in statistical mechanics , academic press ( 1982 ) .
k. huang , statistical mechanics , 2nd edition , john wiley & sons , ( 1987 ) , page 363 .
since we are dealing with periodic boundary conditions in the @xmath193-direction , we consider @xmath194 if @xmath195 and analogously @xmath196 if @xmath197 throughout . m. n. barber , in `` phase transitions and critical phenomena '' , edited by c. domb and j. l. lebowitz .
( academic , new york ) .
( 1983 ) , vol 8 , page 146 .
v. privman(ed . ) .
`` finite size scaling and numerical simulations of statistical systems '' .
world scientific , singapore .
m. fisher .
* 113 * , 969 ( 1959 ) .
|
the magnetic eden model ( mem ) [ n. vandewalle et al . , phys .
rev .
e. * 50 * , r635 ( 1994 ) ] with ferromagnetic interactions between nearest - neighbor spins is studied in @xmath0dimensional rectangular geometries for @xmath1 . in the mem ,
magnetic clusters are grown by adding spins at the boundaries of the clusters .
the orientation of the added spins depends on both the energetic interaction with already deposited spins and the temperature , through a boltzmann factor .
a numerical monte carlo investigation of the mem has been performed and the results of the simulations have been analyzed using finite - size scaling arguments . as in the case of the ising model , the mem in @xmath2 is non - critical ( only exhibits an ordered phase at @xmath3 ) . in @xmath4
the mem exhibits an order - disorder transition of second - order at a finite temperature .
such transition has been characterized in detail and the relevant critical exponents have been determined .
these exponents are in agreement ( within error bars ) with those of the ising model in 2 dimensions . further similarities between both models have been found by evaluating the probability distribution of the order parameter , the magnetization and the susceptibility .
results obtained by means of extensive computer simulations allow us to put forward a conjecture which establishes a nontrivial correspondence between the mem for the irreversible growth of spins and the equilibrium ising model .
this conjecture is certainly a theoretical challenge and its confirmation will contribute to the development of a framework for the study of irreversible growth processes .
.5 cm .5 cm 21truecm 15truecm
|
ising spin - glass models yield phase diagrams with distinctively complex ordered phases in three - dimensions ( @xmath3 ) .
the global phase diagram of such models includes ferromagnetic , paramagnetic , and glassy phases , and the transitions among these are of second order , belonging to different universality classes .
an important part of the corresponding theoretic and computational studies is based on the edwards - anderson ( ea ) model @xcite .
the ea model is defined via the hamiltonian @xmath4 where the summation is over nearest - neighbors , @xmath5 are ising spins , and @xmath6 denotes the quenched uncorrelated exchange interaction , obtained in the current work from the following and most popular random bimodal distribution @xmath7 in the present paper , we consider a bimodal spin - glass model with a spatially longitudinal anisotropic bond randomness on the simple cubic lattice .
the random exchange is applied only in the @xmath1 direction , whereas in the @xmath2 - planes , the exchange is taken to be ferromagnetic . thus the disorder is longitudinal and the interactions in the @xmath2 directions are ferromagnetic .
this study follows the more general study of a spatially uniaxially anisotropy spin - glass system considered on an hierarchical lattice by gven _
et al_. @xcite , and the transverse anisotropic model on the simple cubic lattice studied recently by the present authors , where the random exchange was applied in the interactions on the @xmath2 - planes , while the interactions in the @xmath1 direction were ferromagnetic @xcite . the generalized anisotropic case studied in ref .
@xcite may be defined by the hamiltonian @xmath8 studied also in the current paper .
exchange interactions are uncorrelated quenched random variables , taking values @xmath9 on the @xmath2 - planes and @xmath10 on the z axis . accordingly , the bimodal distribution of @xmath11 has the general form @xmath12 where @xmath13 denotes the @xmath1 axis ( @xmath14 ) or the @xmath2 - planes ( @xmath15 ) , @xmath16 denotes the corresponding exchange interaction strength , and @xmath17 are the probabilities of two neighboring spins ( @xmath18 ) having antiferromagnetic interaction .
the above hamiltonian includes the standard isotropic model [ eqs .
( [ eq:1 ] ) and ( [ eq:2 ] ) ] , which corresponds to @xmath19 and @xmath20 . in this case , several accurate studies have been carried out , and the critical behavior of the corresponding ferromagnetic - paramagnetic ( f - p ) and spin glass - paramagnetic ( sg - p ) phase transitions has been well estimated @xcite . the ferromagnetic - spin glass ( f - sg ) transition line has also been studied @xcite , and the multicritical point , where the transition lines meet , located along the nishimory line , has been accurately defined @xcite .
as mentioned above , our investigations concern particular cases of the spatially anisotropic @xmath3 spin - glass system described by eqs .
( [ eq:3 ] ) and ( [ eq:4 ] ) . in the present study
we continue with the longitudinal anisotropic model @xmath21 , whereas the previous studies @xcite concerned the transverse anisotropic model @xmath22 , both with @xmath23 .
the main motivation is the identification of possible effects caused by the introduced anisotropy on the global phase diagrams and the investigation of the universality aspects of these models .
this is carried out by estimating in each case the corresponding critical exponents along the different transition lines .
our findings are compared to both the isotropic model and our previously studied transverse anisotropic case @xcite .
similarities and differences are pointed out and the importance of anisotropy , as well as that of the relevant frustration features of the models , are critically discussed .
the rest of the paper is laid out as follows : in the following section we briefly review the monte carlo ( mc ) scheme implemented and some details of our simulations .
section [ sec:3 ] starts with an outline of the finite - size scaling ( fss ) schemes used in the paper ( see section [ sec:3a ] ) , then presents our numerical data and the estimation of the corresponding critical behavior .
in particular , in sec .
[ sec:3b ] the critical behavior at the f - p transition line is estimated .
for the f - p case @xmath24 , a detailed fss analysis is presented that shows clearly that the system , in this regime , belongs to the universality class of the random ising model , as expected .
further cases on the f - p line are considered related to the universality class of the random ising model , the details of the corresponding estimations are omitted , but the critical estimates are given in table [ tab:1 ] at the end of the manuscript . in sec .
[ sec:3c ] the sg - p transition is discussed with an emphasis on the symmetric case for which a zero - temperature study is also carried out .
this is followed by sec .
[ sec:3d ] , where the f - sg transition is presented .
section [ sec:4 ] summarizes all our critical estimates and the global phase diagrams of the isotropic ea model and the two anisotropic variants @xmath25 and @xmath26 are sketched and discussed , pointing out similarities , as well as differences with previous work .
finally , we conclude with our findings and comment on the results of the current manuscript in sec .
[ sec:5 ] .
our approach to frustrated systems @xcite is based on the parallel tempering ( pt ) method @xcite .
a pt protocol uses an adequate number of lattice sweeps ( usually metropolis mc steps @xcite ) followed by pt exchange moves between neighboring temperatures .
simulations extend to a temperature range , which depends on the linear lattice size @xmath27 , and is appropriate for the estimation of the critical properties of the system . in other words , we generate mc data that cover several finite - size anomalies of the system .
the temperatures are selected using a constant acceptance exchange method @xcite and the acceptance rate is chosen within the range @xmath28 .
this pt practice is a quite general approach used , also by several other authors @xcite , and has been carefully implemented and compared to other alternative pt methods in our recent papers @xcite . to obtain the temperature sequences for the production runs , we have used relatively short preliminary runs to generate adequate mc data in the range of interest .
a simple histogram method @xcite was then used to determine from the energy probability density functions the temperatures , satisfying the constant acceptance exchange condition @xcite .
since these temperature sequences depend weakly on the disorder realization , we average over several realizations to find an averaged sequence , which is subsequently used in the extended runs . for each case
, corresponding to a certain set of @xmath29 $ ] values , the pt scheme was carefully tested for equilibration .
several independent runs of a large number of disorder realizations were carried out in order to obtain , in the temperature range of interest , the relevant disorder averaged parameters @xmath30 $ ] , where @xmath31 denotes a thermal average of some thermodynamic quantity .
we ensured that the temperature sequence went well deep into the paramagnetic phase in order to avoid entrapment and ensure equilibration .
for moderately small lattice sizes , the number of temperatures used were approximately @xmath32 , but for larger sizes we had to use more than @xmath33 temperatures , even though we had adjusted the acceptance rate in a rather low value , of the order of @xmath34 .
as expected , equilibration was much easier in the paramagnetic and ferromagnetic phases , than in the glassy phase .
furthermore , as sample - to - sample fluctuations are substantially larger for glassy systems , not only the averaging time , but also the number of random samples simulated had to be accordingly adjusted . in more detail ,
for the lower part of the phase diagram , @xmath35 , the signature of the f - p transition was very clear .
this can be observed from the illustrations of section [ sec:3b ] , where the finite - size anomalies of the system are illustrated . in these cases
we have used @xmath36 to @xmath37 different lattice sizes varying from @xmath38 to @xmath39 and @xmath40 to @xmath41 disorder realizations . in particular , for the case @xmath24 we simulated systems up to @xmath42 spins .
this specific case was used as the representative of the f - p line and a detailed fss analysis was carried out and is presented in sec .
[ sec:3b ] . in this particular case , the corresponding ensembles of realizations varied from @xmath41 samples for @xmath38 to @xmath40 samples for the largest lattice @xmath43 . near the multicritical point ,
a much larger set of realizations was needed . for @xmath44 , the number of realizations varied from @xmath45 for @xmath38 to @xmath46 for @xmath47 . for the case
@xmath48 we started with @xmath38 and @xmath49 realizations going up to @xmath50 with @xmath51 samples .
deep in the sg - p transition line , near the symmetric point @xmath52 , smaller system sizes were simulated .
for @xmath53 we used @xmath54 samples for @xmath55 and @xmath56 samples for @xmath52 .
we reached the size of @xmath47 with @xmath57 samples for @xmath55 and the size of @xmath58 with @xmath59 samples for @xmath52 .
further details on our simulations will be be given in the sections .
we start this section by outlining the fss analysis ( sec .
[ sec:3a ] ) used for the estimation of critical parameters , and we also present , in subsections , estimations for the transition lines and the corresponding critical behavior of the present anisotropic model .
all the results shown below refer to the disorder - averaged thermodynamic quantities , as explained in more detail below . for instance in sec .
[ sec:3b ] a detailed study is presented for the case @xmath60 .
similar analysis has also been performed for several other values of @xmath61 along the global phase diagram , but the details of some cases are omitted for reasons of brevity . however , a full summary of results for the critical temperatures and the correlation length s exponent for all the values of @xmath61 simulated is given in table [ tab:1 ] at sec .
[ sec:3d ] .
among other alternatives , we attempt here to estimate the critical properties applying the traditional fss route , in which one observes the scaling laws of several finite - size anomalies of the system .
the corresponding peaks of specific heat @xmath62 , magnetic susceptibility @xmath63 , and the peaks of logarithmic derivatives of the order parameter with respect to the inverse temperature @xmath64 will be used . for a disordered system , we assume that the averaged over disorder @xmath30 $ ] parameter of the system is the relevant observable , where @xmath31 denote the thermal average of some thermodynamic quantity , such as the ones mentioned above . furthermore , the standard scaling power laws , commonly used for the thermal averages of a pure system , are assumed to apply to the corresponding @xmath30 $ ] observable of the disordered system .
finite - size anomalies are now defined with the help of @xmath30 $ ] . in this scaling scheme , the maxima of the specific heat
are assumed to follow the scaling law @xmath65^{\ast}\sim l^{\alpha/\nu}$ ] , whereas the peaks of the magnetic susceptibility are expected to scale as @xmath66^{\ast}\sim l^{\gamma/\nu}.\ ] ] additionally , the peaks of the logarithmic derivative of power @xmath67 of the order parameter with respect to the inverse temperature @xmath64 , defined as @xcite , @xmath68 are assumed to follow a power - law behavior with the correlation length exponent of the form @xmath69^{\ast } \sim l^{1/\nu}.\ ] ] ( color online ) ( a ) specific - heat curves as a function of the temperature .
( b ) shift of the corresponding pseudocritical temperatures obtained from panel ( a ) where the specific heat attains its maximum.,width=8 ] the shift behavior of these finite - size anomalies @xmath30^{\ast}$ ] is used for the estimation of critical temperatures and the correlation length exponent by applying the standard fitting formula for second - order phase transitions @xmath70 for the present study we have used two different order parameters : ( i ) the usual magnetization @xmath71 , which is appropriate for describing transitions that involve the ferromagnetic phase , defined as @xmath72 where @xmath73 is the spin variable and @xmath74 the number of lattice sites and ( ii ) the spin - glass overlap order parameter @xmath75 , appropriate for the sg - p transition ( also for the f - p transition ) , defined respectively as @xmath76 where @xmath77 denotes the spin of the site @xmath78 and @xmath79 represent two replicas of the same disorder realization . for the disordered system the corresponding , averaged over disorder , order parameter is obtained from the thermal averages of eqs .
( [ eq : m ] ) and ( [ eq : q ] ) .
( color online ) temperature dependence of the second moment of the logarithmic derivative of the order parameter with respect to the inverse temperature ( main panel ) and the corresponding fss behavior of the peaks ( inset).,width=8 ] following common practice @xcite , we use a definition for the binder cumulant @xcite of @xmath31 ( @xmath71 or @xmath75 ) , based on the disorder averaged moments @xmath80}{3[\langle z^2\rangle]^{2}}.\ ] ] the crossings of the order - parameter s fourth - order binder cumulant @xmath81 or @xmath82 have been used as an alternative method in estimating the phase - diagram points . furthermore ,
another route that provide us with complementary estimates of both the critical temperature and the exponent @xmath83 is that of data collapse @xcite .
the application of this method is carried out assuming a scaling hypothesis of the form @xmath84 with @xmath31 the above described order parameters .
for the fitting procedure we have used _ autoscale _ , a program that performs a fss analysis for given sets of simulated data @xcite .
the program implements the above scaling assumption which is expected to hold best close to the critical point and optimizes an initial set of scaling parameters that enforce a data collapse of the different sets .
the optimum data collapse , achieved by the minimization procedure of the scaling parameters via the downhill simplex algorithm , is carried out by finding a fair compromise between a ( rather small ) value of a @xmath85/dof - like quantity and a reasonably large interval on the re - scaled abscissa .
( color online ) illustration of the magnetic susceptibility as a function of the temperature ( main panel ) and the fss of the peaks ( inset).,width=8 ] our main interest is the identification of the universality class of the model and this is , in general , characterized mainly from the critical exponent @xmath83 of the correlation length .
thus , we shall focus our analysis on the estimation of this exponent .
additional tests will include the estimation of the critical temperature via various fss schemes , reflecting also the accuracy of our numerical data , as well as the estimation of the magnetic exponent ratio @xmath86 .
we start our analysis with fig .
[ fig : spec_heat_scaling ] , illustrating the scaling behavior of the specific heat .
in particular , the upper panel ( a ) of this figure illustrates the specific - heat curves @xmath65 $ ] as a function of the temperature for the complete set of system sizes studied .
the observed shift behavior of the specific - heat peaks is quantified in the corresponding lower panel ( b ) , where we study the fss of the relevant pseudocritical temperatures @xmath87 , i.e. , the temperatures where the specific heat attains its maximum . applying the standard fitting formula for second - order phase transitions ( [ eq : shift ] ) on the numerical data we get the estimates @xmath88 and @xmath89 for the critical temperature and the correlation length s exponent , respectively , the latter indicating that the model shares the random ising universality class , for which accurate estimates are @xmath90 @xcite , @xmath91 @xcite , and @xmath92 @xcite
. a further verification of the value of the critical exponent @xmath83 is provided via the fss analysis of the peaks of the logarithmic derivative of the power @xmath67 of the order parameter with respect to the inverse temperature @xmath64 , as shown in fig .
[ fig : ropes ] . in particular ,
in the main panel of this figure we illustrate the temperature dependence of the second moment [ eq .
( [ eq : moment ] ) ] for the whole spectrum of lattice sizes studied and in the corresponding inset the fss of the peaks .
these are expected to scale as eq .
( [ eq : moment_scaling ] ) and the power - law fitting , shown by the solid line , provides an estimate @xmath93 for the exponent , in agreement with the values given above .
( color online ) crossings of the order - parameter s fourth - order binder cumulant ( main panel ) and infinite limit size extrapolation of the crossing points ( inset).,width=8 ] we conclude our standard fss analysis with the magnetic properties of the model , and in particular with the estimation of the magnetic exponent ratio @xmath86 through the fss of the magnetic susceptibility . in the main panel of fig .
[ fig : magn_susceptibility ] we present the temperature dependence of the magnetic susceptibility and the fss of the corresponding peaks in the inset .
the solid line shows a power - law fitting of the form ( [ eq : chi ] ) using the complete lattice - range spectrum .
the outcome of the fitting provides an estimate @xmath94 for the exponent ratio @xmath86 , in good agreement to the best - known literature estimates for the random ising model which slightly vary around the value @xmath95 @xcite .
the above estimates are now further verified from a different approach , see fig .
[ fig : crossings ] , where the crossings of the order - parameter s fourth - order binder cumulant @xmath96 are illustrated .
although it is clear that the curves cross around the value @xmath97 which marks the location of the critical temperature of the system , a more refined analysis ( see the inset ) using the infinite - limit size extrapolation @xmath98 , where @xmath99 , of the crossing points @xmath100 of the pairs @xmath101 gives an estimate of @xmath102 for the critical temperature .
last but not least , we provide complementary estimates of both the critical temperature and the exponent @xmath83 , via the method of data collapse @xcite , as already discussed above in sec.[sec:3a ] . in the current case ,
the optimum data collapse for the order - parameter s fourth - order binder cumulant @xmath96 is shown in fig . [
fig : collapse_v_m ] and the resulting value for the critical temperature @xmath103 is in agreement with the previous estimates of fig . [ fig : crossings ] from the crossings of the same thermodynamic quantity . additionally , the estimate @xmath104 for the critical exponent of the correlation length is also in agreement with the previously obtained estimates .
( color online ) collapse of the order - parameter s fourth - order binder cumulant of @xmath71.,width=8 ] similar collapse attempt was performed for the binder cumulant of the overlap order parameter @xmath82 and both critical estimates are presented in table [ tab:1 ] .
further cases for the lower part of the phase diagram , @xmath105 , were considered and the estimates are also in table [ tab:1 ] . for the values @xmath106 and @xmath107 we observe a considerable deviation in the estimated critical temperatures obtained from the corresponding collapse of @xmath96 and @xmath82 , indicating that we are close to the multicritical point , in which the magnetization ceases to operate as the appropriate order parameter of the system . in particular , at @xmath108
we observe a significant deviation also in the estimate of the correlation length exponent . in this case
the exponent produced from the collapse of the magnetization data is @xmath109 , whereas that produced from the overlap data is @xmath110 .
the last value is close to @xmath111 given by hasenbusch _
et al . _
@xcite , confirming that @xmath108 is close to the multicritical point .
the f - sg is the least investigated of the transition lines . a recent comprehensive fss analysis for the isotropic ea model
was performed by ceccarelli _
et al_. @xcite .
these authors found a new universality class by determining two points of the f - sg line corresponding to the temperatures @xmath164 and @xmath165 .
these two phase - diagram points , when compared to previous literature estimates for the multicritical and the @xmath113 f - sg critical point , clearly supported the earlier proposed reentrant behavior of the line .
the extensive simulations of ref .
@xcite involved lattice sizes within the range @xmath166 , using many thousands of disorder realizations , even for @xmath50 .
illustration of the crossing behavior of the magnetization s fourth - order binder s cumulant versus @xmath61 for the linear sizes @xmath167 , @xmath168 , and @xmath58 for three relevant temperatures outlined
. the resulting f - sg transition line appears to be forward , according to the finite - size approximation shown.,width=8 ] our attempt here to estimate the f - sg line for the present anisotropic ea model is not as extensive , we considered lattice sizes @xmath167 , @xmath168 , and @xmath58 , with @xmath169 , @xmath170 , and @xmath171 samples , respectively .
the simulations were carried out for several values of @xmath61 , close to the multicritical value @xmath172 .
figure [ fig : multi ] illustrates the behavior of the fourth - order binder s cumulant of the magnetization @xmath96 .
the temperature sequence for each @xmath61 case went well deep into the paramagnetic phase ( @xmath173 ) in order to avoid entrapment and ensure equilibration . while in most cases the minimum temperature was @xmath174 , a very low temperature , namely @xmath175 ,
was used in some production runs in order to record ground - state properties . at @xmath176 in the range @xmath177 for linear sizes @xmath167 and @xmath168 , @xmath96 does not show crossing and goes to zero as the size increases , indicating that we are in the paramagnetic region .
below the phase - diagram boundary we find crossing points as illustrated in fig .
[ fig : multi ] for three temperatures , namely @xmath174 , @xmath178 , and @xmath179 , corresponding to crossing probabilities @xmath180 , @xmath181 , and @xmath182 , respectively .
the above were determined by the mean value of the crossings of the lines produced by second order polynomial fittings on the simulated points .
note that these values are rough estimates of the asymptotic results , since no extrapolation has been attempted .
the f - sg transition line appears , at least according to the present finite - size approximation , to be forward and not reentrant , as shown for the isotropic model by ceccarelli _
et al_. @xcite . global phase diagrams .
transition lines separate ferromagnetic ( f ) , spin glass ( sg ) , and paramagnetic ( p ) phases , and the multicritical points are denoted by .
solid lines and full symbols illustrate the phase diagram of the isotropic ea model .
f - p line points are taken from ref .
@xcite , the multicritical point from ref .
@xcite , the zero - temperature f - sg point from ref .
@xcite , and the symmetric sg - p point from ref .
the dotted lines and open symbols correspond to the transverse anisotropic model @xmath183 described in detail in ref .
finally the dashed lines and open symbols correspond to the present longitudinal anisotropic model @xmath184.,width=8 ] from the above discussion , the f - sg transition line for the present longitudinal anisotropic model appears to be forward , although an uncertainty remains because of the rather limited finite - size data at hand .
however , the shifts of the crossings introduced by adding @xmath58 reinforces our claim .
this behavior , if verified , could indicate to a link between frustration and the slope of the transition lines , since the present model has quite different frustration features close to the multicritical point . thus , a need for a more extensive fss analysis in this regime remains . by using two estimates for the phase diagram , close to the multicritical point for the f - p line and the above three estimates of the f - sg transition line we find , using linear fittings , @xmath185 and @xmath186 for the multicritical point .
the @xmath113 f - sg critical point found by the linear fittings based on the above three estimates is @xmath187 , which is not far from the value obtained by a collapse of @xmath188 performed on our restricted ground - state data ( @xmath189 ) .
these estimates indicate a clear forward behavior . concluding the above discussion
, it is also useful to recall that a forward ferromagnetic spin - glass line has been illustrated for a transverse anisotropic model with ratio of interactions @xmath190 ( see upper right panel of fig.3 in ref .
@xcite ) . for the same ratio of interactions , the authors of re .
@xcite also reported a very narrow spin - glass phase for the longitudinal anisotropic version of the model without commenting on the forward or reentrant character of this line ( see the lower left panel of fig.3 in ref .
the origin of these findings , including those of the present paper , is not diaphanous and this makes the problem even more interesting .
from the closing remarks of the previous section it is expected that for the present model the multicritical point is close to @xmath108 .
the data for the f - sg transition line in sec .
[ sec:3d ] are also in agreement with this estimate .
thus , two values @xmath55 and @xmath52 of our study belong to the sg - p transition line . for brevity
we omit the details of our analysis for the first value , and we concentrate here only in the most interesting symmetric , @xmath52 , case including a relevant ground - state study . in the vicinity of the symmetric case @xmath112 at low enough temperatures ,
the presence of the spin - glass phase was clearly detected .
the first indication was that the magnetization in this regime was not an appropriate order parameter ( see also below the discussion for the @xmath113 magnetization order ) .
furthermore , at @xmath114 and @xmath115 ( see fig . [
fig : collapse05 ] ) , the data collapse analysis of @xmath82 produced critical exponent values @xmath116 and @xmath117 respectively , which are compatible to the estimated values of the sg - p transition found in the literature , namely the values @xmath118 @xcite , @xmath119 @xcite , @xmath120 @xcite , and the recent estimates @xmath121 @xcite and @xmath122 @xcite .
similar results were also found in our recent study of the anisotropic case @xmath123 , where @xmath83 was estimated @xmath124 @xcite .
the collapse of the binder s cumulant of the overlap order parameter , illustrated in the main panel of fig .
[ fig : collapse05 ] , gives an estimate of the order of @xmath125 for the critical temperature .
we have also performed a linear extrapolation of the crossings points of @xmath82 , as illustrated in the inset of this figure .
the fit is performed in @xmath126 , with @xmath127 .
the resulting estimate is now @xmath128 in accordance with the value produced by the data collapse method .
therefore , the critical temperature of the longitudinal anisotropic model ( denoted by @xmath129 ) for the symmetric case is estimated to be considerably higher from the corresponding critical temperature of the isotropic model .
let us recall that the critical temperature of the isotropic model , denoted here by @xmath130 , is of the order of @xmath131 @xcite , and the corresponding temperature for the transverse anisotropic model @xmath132 , denoted here by @xmath133 , was estimated in the same range @xmath134 @xcite .
the striking coincidence of points and @xmath135 , was further supported by the detailed ground - state study of ref .
@xcite . in relevance to the above discussion
, we now present in fig .
[ fig : gs ] the finite - size behavior of the ground - state energy per site(@xmath136 ) for both anisotropic versions of our studies together with the known behavior of the isotropic @xmath3 ea model .
evidently , the clear difference in the limiting value of @xmath137 reflects the considerably higher critical temperature of the longitudinal anisotropic model .
the known asymptotic estimations for the isotropic model are @xmath138 @xcite , and @xmath139 @xcite , and the second value is indicated by the dashed line in fig .
[ fig : gs ] . for the present longitudinal anisotropic model
the dashed line indicates the estimate @xmath140 . in the second part of this figure
, we also show the size - decline of the @xmath113 magnetization order in the bulk @xmath141 $ ] . in the same figure
an alternative order parameter is used based on the magnetization of the @xmath2 planes , defined as @xmath142 $ ] . apparently ,
this pseudo - ordering ( more pronounced for @xmath143 ) is only a finite - size effect that keeps also the ground - state energy in much lower values .
the data shown in fig .
[ fig : gs ] were obtained via the pt algorithm using a practise analogous to that detailed in ref .
the ensemble of disorder realizations varied from @xmath144 for @xmath53 to @xmath145 for @xmath50 . critical temperature at the symmetry axis @xmath112 .
the collapse of the overlaps s fourth - order binder cumulant with the method outlined in sec .
[ sec:3 ] gives @xmath125 , while the infinite - limit size extrapolation of @xmath146 crossings , illustrated in the inset , gives @xmath128 .
details are provided in sec .
[ sec:4],width=8 ] .
we will now try to give an intuitive explanation of the above observations concerning the critical temperatures at the symmetric cases of the above spin - glass models .
we suggest that , at these symmetry points - @xmath147 - the critical temperatures @xmath130 , @xmath133 , and @xmath129 , are determined by some simple global frustration features of the models .
first , let us assume that the coincidence of @xmath130 and @xmath148 observed in our earlier papers @xcite is due to the fact that the density of the elementary squares of the lattice , denoted by @xmath149 , which can not simultaneously satisfy all their bonds ( frustrated elementary squares or plaquette ) is equal to @xmath150 for both models .
this is also a plausible explanation of the apparent match in their ground - state energy found in ref .
@xcite(see also fig .
[ fig : gs ] ) . for the current model , at the point @xmath115 ,
the corresponding density is @xmath151 , giving a lesser amount of frustration , thus producing a much higher @xmath152 , according to the above hypothesis .
further global frustration features of the models can be invoked in order to support the above argument . as an example , let us consider the class of completely unfrustrated bonds .
these type of bonds ( @xmath153 ) on the simple cubic lattice have zero frustrated plaquette attached to them and one can define the corresponding densities @xmath154 as fractions of the total number of bonds .
then , it is again easily seen that for both the isotropic model at @xmath155 and also for the transverse anisotropic model @xmath156 , @xmath157 , whereas for the longitudinal anisotropic model this fraction is @xmath158/3=1/6=0.166\cdots$ ] .
thus , the longitudinal anisotropic model carries weaker frustration features , leading to higher critical temperature .
similar arguments can be applied for all five classes of bonds , i.e , @xmath159 , @xmath160 , @xmath161 , @xmath162 , and @xmath163 on the simple cubic lattice or their combinations , strengthening the hypothesis that @xmath152 is mainly determined by simple global frustration features of the models .
( a ) behavior of ground - state energies per site @xmath137 for the @xmath3 ea model and the two anisotropic versions .
( b ) size - decline of the @xmath113 magnetization order in the bulk and @xmath2-planes of the present longitudinal anisotropic model . in both panels
error bars denote sample - to - sample fluctuations.,width=8 ]
in this section we present the global phase diagrams of the following spin models : the well - known isotropic ea spin - glass model @xcite , the transverse anisotropic case @xmath183 presented in our previous work @xcite , and finally the present longitudinal anisotropic model @xmath191 .
the three models are symmetric around the @xmath192- , @xmath193- , and @xmath112-axis respectively , thus the @xmath194 parts of the diagrams are omitted .
the phase diagrams are illustrated in fig .
[ fig : phd ] .
point , is the well studied pure ising model with @xmath195 @xcite .
the full star at @xmath196 in the f - p line denotes the improved model proposed in ref .
@xcite , where scaling corrections are minimum .
open and half full stars represent respectively points of the transverse anisotropic model @xmath197
@xcite and @xmath24 of the present longitudinal anisotropic model .
note that , the @xmath198-axis labelled as @xmath192 , represents the probability @xmath192 for the isotropic model , @xmath193 for the transverse anisotropic model , and @xmath61 for the present longitudinal anisotropic model .
further details are given in the caption of fig .
[ fig : phd ] .
the phase diagrams of the isotropic and transverse anisotropic models were discussed in refs .
@xcite , where the most striking feature of the coincidence of points and @xmath135 , was supported by the detailed ground - state study of ref . @xcite . as seen now from fig .
[ fig : phd ] , the present model yields a sg - p point @xmath199 at a significantly higher temperature and the origin of this phenomenon we believe stems from the different frustration features , already discussed in sec .
[ sec:3c ] .
.[tab:1]summary of estimates for the critical temperature and the critical exponent @xmath83 obtained for all the values of @xmath61 studied . for @xmath200 two sets of results
are given , obtained from the data collapse method as discussed in the main text , corresponding to the order - parameter s @xmath96 ( second and third column ) and overlap s @xmath82 ( fourth and fifth ) fourth - order binder cumulant .
for @xmath201 , estimates only via the collapse of @xmath82 are given . [ cols="^,^,^,^,^",options="header " , ]
in the present manuscript , we presented the phase diagram and critical behavior of a further anisotropic case of the @xmath0 three - dimensional ising model on the simple cubic lattice .
the current spatially anisotropic @xmath0 bond randomness was applied only in the @xmath1 direction , whereas in the other two directions , @xmath2 - planes , a ferromagnetic exchange was implemented .
the phase diagram of the model under study was compared to that of the corresponding isotropic one , as well as to the phase diagram of the anisotropic case in which the @xmath0 bond randomness was applied only in the @xmath2 - planes .
the observed differences in the global phase diagrams were critically discussed , assuming that some global frustration features of the models determine the critical temperature , especially at the symmetry points of the models .
as we have shown , the differences in frustration features , at the symmetry point , give rise in the present model to a considerable increase of the transition temperature .
furthermore , our data for the ferromagnetic - spin glass transition line are supporting a forward behavior in contrast to the reentrant behavior of the isotropic model , and this is also proposed as an effect of the frustration features of the model .
our scaling analysis verified , once more , the interesting feature of the irrelevance for the critical behavior of the spatial anisotropy .
this aspect of universality was also found in the renormalization - group study of ref .
the signs of this universality can be directly identified by simply comparing our critical estimates shown in table [ tab:1 ] to those of the relevant literature .
it is natural to expect that the general universality , observed among these models , is a reflection of the spatially stochastic character of the quenched disorder and the frustration , although the differences in some global characteristics may produce very interesting changes in the corresponding phase diagrams .
it will be useful , as a feature task , to further understand the effects of the global frustration features on both phase diagrams and also on the critical behavior of these simple spin - glass models .
research in this direction is currently under way .
edwards and p.w .
anderson , j. phys .
f : metal physics * 5 * , 965 ( 1975 ) . k. binder and a.p .
young , rev .
* 58 * , 801 ( 1986 ) . c. gven , a.n .
berker , m. hinczewski , and h. nishimori , phys .
e * 77 * , 061110 ( 2008 ) .
t. papakonstantinou and a. malakis , phys .
e * 87 * , 012132 ( 2013 ) .
g. ceccarelli , a. pelissetto , and e. vicari , phys .
b * 84 * , 134202 ( 2011 ) .
m. hasenbusch , f. parisen toldin , a. pelissetto , and e. vicari , phys .
b * 76 * , 094402 ( 2007 ) .
m. hasenbusch , a. pelissetto , and e. vicari , phys .
b * 78 * , 214205 ( 2008 ) .
katzgraber , m. krner , and a.p .
young , phys .
b * 73 * , 224432 ( 2006 ) .
t. jrg , phys . rev .
b * 73 * , 224431 ( 2006 ) .
campbell , k. hukushima , and h. takayama , phys .
b * 76 * , 134421 ( 2007 ) .
ballesteros , a. cruz , l.a .
fernndez , v. martn - mayor , j. pech , j.j .
ruiz - lorenzo , a. tarancn , p. tllez , c.l .
ullod , and c. ungil , phys .
b * 62 * , 14237 ( 2000 ) .
m. palassini and s. caracciolo , phys .
. lett . * 82 * , 5128 ( 1999 ) .
n. kawashima and h. rieger , arxiv : cond - mat/0312432 .
a. billoire , l.a .
fernndez , a. maiorano , e. marinari , v. martn - mayor , and d. yllanes , j. stat .
mech . : theory exp .
( 2011 ) p10019 .
m. hasenbusch , f. parisen toldin , a. pelissetto , and e. vicari , phys .
b * 76 * , 184202 ( 2007 ) . y. ozeki and h. nishimori , j. phy .
. jpn . * 56 * , 3265 ( 1987 ) .
y. ozeki and n. ito , j. phys . a : math
. gen . * 31 * , 5451 ( 1998 ) .
singh , phys .
lett . * 67 * , 899 ( 1991 ) .
doussal and a.b .
harris , phys .
* 61 * , 625 ( 1988 ) .
doussal and a.b .
harris , phys .
b * 40 * , 9249 ( 1989 ) .
h. nishimori , _ statistical physics of spin glasses and information processing : an introduction _ ( oxford university press , usa , 2001 ) .
h. nishimori , j. phys .
c : solid state phys . * 13 * , 4071 ( 1980 ) .
h. nishimori , prog .
66 * , 1169 ( 1981 ) .
h. nishimori , j. phys .
* 55 * , 3305 ( 1986 ) .
a. malakis and t. papakonstantinou , phys .
e * 88 * , 013312 ( 2013 ) .
t. papakonstantinou and a. malakis , j. phys .
: conf . ser . *
487 * , 012010 ( 2014 ) .
swendsen and j .- s .
wang , phys .
lett . * 57 * , 2607 ( 1986 ) .
k. hukushima and k. nemoto , j. phys .
. jpn . * 65 * , 1604 ( 1996 ) .
e. marinari , g. parisi , and j.j .
ruiz - lorenzo , _ in spin glasses and random fields , edited by a.p .
( world scientific , singapore,1998 ) , p.59 .
m.e.j newman and g.t .
barkema , _ monte carlo methods in statistical physics _
( clarendon , oxford , 1999 ) . n. metropolis , a.w .
rosenbluth , m.n .
rosenbluth , a.h .
teller , and e. teller , j. chem .
* 21 * , 1087 ( 1953 ) .
e. bittner , andreas nubaumer , and w. janke , phys .
* 101 * , 130603 ( 2008 ) .
y. sugita and y. okamoto , chem .
* 314 * , 141 ( 1999 ) .
e. bittner and w. janke , phys .
e * 84 * , 036701 ( 2011 ) .
d.j . earl and m.w .
deem , phys .
* 7 * , 3910 ( 2005 ) .
swendsen and j.s .
wang , phys .
lett . * 58 * , 86 ( 1987 ) ; u. wolff , _ ibid . _ * 62 * , 361 ( 1989 ) .
landau and k. binder , _ monte carlo simulations in statistical physics _
( cambridge university press , cambridge , 2000 ) .
a. malakis , a.n .
berker , n.g .
fytas , and t. papakonstantinou , phys .
e * 85 * , 061106 ( 2012 ) .
a.m. ferrenberg and d.p .
landau , phys .
b * 44 * , 5081 ( 1991 ) .
o. melchert , arxiv:0910.5403 . h.g .
ballesteros , l.a .
fernndez , v. martn - mayor , a. muoz sudupe , g. parisi , and j.j .
ruiz - lorenzo , phys .
b * 58 * , 2740 ( 1998 ) .
m. hasenbusch , f. parisen toldin , a. pelissetto , and e. vicari , j. stat .
mech . : theory exp .
( 2007 ) p02016 .
a. pelissetto and e. vicari , phys
b * 62 * , 6393 ( 2000 ) . p.e .
berche , c. chatelain , b. berche , and w. janke , eur .
j. b * 38 * , 463 ( 2004 ) .
n.g . fytas and p.e .
theodorakis , phys .
e * 82 * , 062101 ( 2010 ) ; p.e .
theodorakis and n.g .
fytas , eur .
j. b * 81 * , 245 ( 2011 ) .
k.f . pal , physica a * 223 * , 283 ( 1996 ) .
hartmann , europhys .
lett . * 40 * , 429 ( 1997 ) .
m. hasenbusch , phys .
b * 82 * , 174433 ( 2010 ) .
m. campostrini , a. pelissetto , p. rossi , and e. vicari , phys .
e * 65 * , 066127 ( 2002 ) .
r. guida and j. zinn - justin , j. phys .
a : math . gen . * 31 * , 8103 ( 1998 ) .
p. butera and m. comi , phys .
b * 65 * , 144431 ( 2002 ) . y. deng and h.w.j .
blte , phys .
e * 68 * , 036125 ( 2003 ) .
blte , e. luijten , and j.r .
heringa , j. phys . a : math .
* 28 * , 6289 ( 1995 ) .
hartmann , phys .
b * 59 * , 3617 ( 1999 ) . w. janke .
_ monte carlo methods in classical statistical physics _
( springer berlin heidelberg , 2008 )
|
we study the @xmath0 three - dimensional ising model with a longitudinal anisotropic bond randomness on the simple cubic lattice . the random exchange interaction is applied only in the @xmath1 direction , whereas in the other two directions , @xmath2 - planes
, we consider ferromagnetic exchange . by implementing an effective parallel tempering scheme
, we outline the phase diagram of the model and compare it to the corresponding isotropic one , as well as to a previously studied anisotropic ( transverse ) case .
we present a detailed finite - size scaling analysis of the ferromagnetic - paramagnetic and spin glass - paramagnetic transition lines , and we also discuss the ferromagnetic - spin glass transition regime .
we conclude that the present model shares the same universality classes with the isotropic model , but at the symmetric point has a considerably higher transition temperature from the spin - glass state to the paramagnetic phase . our data for
the ferromagnetic - spin glass transition line are supporting a forward behavior in contrast to the reentrant behavior of the isotropic model .
|
in spite of decades of intensive investigations , the broad line region ( blr ) of active galactic nuclei ( agn ) is yet not fully understood .
the direct detection of the blr remains a challenge for modern instruments , since the angular size of the blr is less than @xmath0 even for the closest agns .
the only information from the blr comes in the form of the broad emission lines ( bel ) , that are a very prominent feature in the spectra of the so - called type 1 agn @xcite .
nevertheless , the investigations of the bel s properties ( flux and shape ) , and especially of their variability , indicates that the blr is linked with the accretion process onto the supermassive black hole in the center of an agn .
first of all , the blr gas is photoionized by the continuum radiation from the accretion disk , and secondly , at least part of the blr follows its geometry , i.e. even though the blr geometry is not ubiquitously described it is believed that one part of the blr is a part of the accretion disk or has a disk - like geometry @xcite .
an important method to indirectly map the geometry and kinematics of the the blr is the `` reverberation mapping '' @xcite , a method that is based on multiple spectroscopic observation and the variability of spectral properties . in particular
, the bel s flux varies with respect to the ionizing continuum flux with a certain time delay ( of the order of days to weeks ) , due to the light - travel time from the source of the continuum photon to the source of the bel photon , that is the blr .
therefore , by applying the cross - correlation function ( ccf ) between the continuum and bel light curves an estimate of the time - lag @xmath1 between the two signals can be obtained , which is basically the photometric blr radius @xmath2 , where c is the speed of light .
this is a powerful tool to observe the unobservable , as one can estimate the size of the blr and consequently the mass of the supermassive black hole , since the blr gas is assumed to be virialized , and the method has been exploited in many papers ( see e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* etc . ) .
the reverberation - mapped agn are of particular importance because they anchor the scaling relationships ( blr radius - luminosity relationship , see e.g. * ? ? ?
* ; * ? ? ?
* ) that allow the estimate of the blr radius , and the supermassive black hole mass , from a single - epoch spectrum . however , this method is based on several assumptions , which makes it still pretty uncertain , especially for the supermassive black hole mass estimate .
it is assumed that the blr gas is photoionized by the continuum coming from the central source , and that is gravitationally bound to the supermassive black hole so that the virial theorem can be used @xcite .
moreover , it strongly depends on the geometry of the blr , and the kinematics and structure of the blr is not ubiquitously described ( see e.g. * ? ? ?
* ; * ? ? ?
* and reference therein ) .
the blr is probably very complex , often with evidence for multiple components ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and can have systematic motions such as infalls , outflows , circular motions , which should be detected in the bel profiles
. if the line profile is varying , the change of the line shape should be according to the change in the blr geometry and kinematics . in principle , the reverberation assumption can be tested , e.g. one test is if the bel fluxes correlate with the continuum flux . therefore , detailed studies of different broad line profiles is required , and this can be done with long - term ( of the order of decade and longer ) monitoring of type 1 agn with different bel properties . in this paper
we give a comparative overview of our investigation based on the international long - term optical monitoring campaign , coordinated by the special astrophysical observatory of the russian academy of science ( sao ras ) . within this campaign , several type 1 agn
have been monitored for decades , out of which data were published for 5 objects @xcite .
the objects have been selected so that representatives of type 1 agn with different bel profiles are in the sample : arp 102b and 3c 390.3 have bels with double - peaked line profiles , ark 564 is a narrow - line seyfert 1 ( nlsy1 ) , and ngc 4151 and ngc 5548 are well - studied seyfert 1.5 with strongly variable bel profiles , with bumps and asymmetries present in the line profiles ( see figure 1 ) .
all 5 agn have been discussed individually in referenced papers , but here we aim to give for a first time a comparative analysis of their variabilty properties , that is to : i ) summarize already obtained most important results of their bel variability , ii ) give new comparative analysis of the bel properties , iii ) discuss their complex blr physics and geometry , all in the framework of the estimations of the mass of the supermassive black hole in agn , and suggestions for future exploitation of our large data sets .
our data sets are observed , reduced and analyzed using the same procedures , thus the total analyzed data form an uniform set of data .
the paper is organized as follows , in section 2 we briefly report on observation and data reduction , in section 3 we present the main result , which are discussed in section 4 , and in section 5 we outline our conclusions and future investigations . and
h@xmath3 line are marked.,width=321 ] line light curves for all 5 objects ( name denoted in bottom left ) . the line flux and corresponding error - bars are normalized to the maximal flux for better comparison and x - axis shows time in units of modified julian date.,width=340 ]
with the international long - term optical monitoring campaign , coordinated by sao ras several type 1 agn have been monitored for decades , out of which data were published for 5 objects in 20012013 ( see table 1 for details ) .
the high quality spectra were taken with 6 different telescopes : ( i ) 6-m and 1-m telescopes of sao ras ; ( ii ) 2.1-m telescope of guillermo haro astrophysical observatory , mexico ; ( iii ) 2.1-m telescope of the observatorio astronomico nacional at san pedro martir , mexico , and ( iv ) the 3.5-m and 2.2-m telescopes of calar alto observatory , spain .
the data acquisition , reduction and different calibration ( e.g. corrections for different position angle , seeing and aperture effects ) , and flux measurements were described in details in shapovalova et al .
( 2001 , 2004 , 2008 , 2010a , 2012 , 2013 ) .
table 1 gives a summary of the data ( object name , monitored period , redshift and type , line properties , and references ) .
for all objects we measured different line parameters , such as total line fluxes , broad line fluxes , line widths , and line - flux time - lags , that will be investigated here ( see references listed in table 1 for details on other line parameters ) .
some of the measured h@xmath4 and h@xmath3 line parameters are given in table 1 : the full - width half - maximum ( fwhm ) of the mean spectrum , the photometric radius of the blr @xmath5 , where @xmath6 is the line - flux time - lag obtained from the ccf analysis , the maximal - to - minimal flux ratio @xmath7 , the variability parameter @xmath8 , and the continuum luminosity at 5100 .
the variability parameter @xmath8 is used to estimate the amount of the line - variability and is calculated according to @xcite . for the estimates of time - lags from the ccf analysis a z - transformed discrete correlation function ( zdcf )
was used @xcite .
luminosities are calculated using the mean continuum flux at 5100 and the online calculator for luminosity distance @xcite , for which we adopted the cosmological parameters @xmath9 , @xmath10 , and @xmath11 , and a hubble constant , @xmath12 km / s mpc@xmath13 .
all flux measurements from this campaign are publicly available in publications listed in table 1 and in the corresponding vizier online data catalog .
the same procedures were performed for the observations , data reduction , measurements and analysis , so that the resulting data sets of these 5 objects basically are uniform .
this makes them valid for further analysis and comparison .
( upper ) and the variability indicator , the ratio of maximal - to - minimal flux @xmath7 ( bottom ) versus the fwhm of h@xmath4 ( circles ) and h@xmath3 ( diamonds ) lines for all 5 objects .
the vertical dashed lines correspond to the line fwhm of @xmath14 and @xmath15.,width=332 ]
our sample consists of 5 type 1 agn with different bel profiles ( figure 1 ) : arp 102b and 3c 390.3 have bels with double - peaked line profiles , ark 564 is a nlsy1 , and ngc 4151 and ngc 5548 are strongly variable seyfert 1.5 .
the results of our long - term monitoring campaign are summarized in table 1 , and figure 1 and 2 , that show the comparison of optical spectra for all 5 objects , given in arbitrary flux unites and shifted for better display ( figure 1 ) , and the h@xmath4 line flux light curves for all 5 objects ( figure 2 ) , where the line flux and corresponding errors are normalized to the maximal flux for better comparison .
similar as in @xcite , in figure 3 we plot the ratio of the maximal - to - minimal line flux against the fwhm of h@xmath4 ( circles ) and h@xmath3 line ( diamonds ) .
we divided the @xmath7 vs. fwhm plane according to the line fwhm in three groups of objects : 1 ) nlsy1 having fwhm @xmath162000 km / s @xcite , 2 ) agn with broad emission lines ( seyfert 1 ) with 2000 @xmath16 fwhm @xmath16 10000 km / s ) , and 3 ) double - peaked line agn ( dpl agn ) with fwhm @xmath17 10000 km / s .
figure 3 gives an indication that there are different types of objects ( with different geometry and type of variability ) that should be considered simultaneously . in order to test the photoionization assumption , we plot in figure 4 the correlation between the h@xmath4 line and continuum flux for all objects ( name indicated on each plot )
the correlation coefficient and the corresponding p - value are listed in table 2 .
the two objects , arp 102b and ark 564 have very low correlation coefficients ( 0.37 and 0.59 , respectively ) , indicating a weak response of the line to the continuum variability .
.line and continuum flux correlations for all 5 objects . the correlation coefficient and the corresponding @xmath18 value ( in brackets )
are given for each pair of data . [ cols="^,^,^",options="header " , ] in principle , the geometry of the blr should be seen in the line shape , e.g. in the most obvious case if the rotational motion is present in the blr , then the line profile should have two peaks , the red corresponding to the gas moving away , and the blue one moving towards the observer .
moreover if there is a change in the blr geometry and kinematics , the line shape should vary accordingly . in order to show how the line profiles vary in two dpl objects , we plot the mean and rms profile for the h@xmath4 line of 3c 390.3 ( figure 5 , upper panel ) and arp 102b ( figure 5 , bottom panel ) . from the shape and intensity of the rms profiles
, it is obvious that in case of 3c 390.3 the line profiles vary significantly , while in arp 102b the line profiles remain almost unchanged . finally , in figure 6
, the radius of the blr for h@xmath4 line versus the continuum luminosity at 5100 , is given .
the solid line represents the scaling relation from @xcite , @xmath19 $ ] , where @xmath3 is the slope of the blr radius - luminosity relationship and @xmath20 is the scaling factor , and in figure 6 @xmath21 and @xmath22 from @xcite are plotted .
the dashed line is a simple linear fit of the above equation ( not considering the error - bars ) through all objects except from ark 564 , obtaining the fitting parameters of @xmath23 and @xmath24 .
line for all 5 objects ( name indicated on each plot ) .
the dashed line represents the linear best fit through all data.,width=332 ] line of 3c 390.3 ( upper panel ) and arp 102b ( bottom panel).,width=332 ] line .
the solid line represents the blr radius - luminosity scaling relation from @xcite where @xmath25 , while the dashed line is our fit through all objects except ark 564 , where @xmath26.,width=332 ]
we have analyzed the variability of the optical spectra for five type 1 agn with different properties of the broad emission lines .
the data sets were processed in a uniform way , which makes them valid for further analysis and comparison . during the long - term monitoring , that is by rule longer than a decade , the line and continuum flux for all objects are varying .
most of the objects are strongly varying , i.e. ngc 5548 , ngc 4151 , and 3c 390.3 , that is best seen from the normalized light curves ( figure 2 ) , but also from the ratio of the maximal - to - minimal flux ( figure 3 , table 1 ) and the variability parameter f@xmath27 that is @xmath2840% ( table 1 ) .
this is supported with the correlation analysis , as these strongly varying objects are showing significant correlation between the line and continuum fluxes ( figure 4 , table 2 ) . on the other hand , the line fluxes of the ark 564 , that is a nlsy1 ,
stay almost constant during the monitored period ( only changing by @xmath287 - 8% ) , that was also noticed before for this object @xcite .
it is also interesting that in case of arp 102b , that is the other extreme object with the largest line widths ( fwhm @xmath28 15000 km / s ) , the variability is also low ( @xmath2820% ) .
the lack of variability in both of these cases is supported by the weak correlation between the continuum and line fluxes ( figure 4 , table 2 ) , e.g. for h@xmath4 of arp 102b it is @xmath29 ( @xmath18=0.5e-04 ) , and even worse for h@xmath3 of the same object ( @xmath30 , @xmath18=0.22 ) .
the lack of variability can be due to poor sampling of the data , e.g. in case of arp 102b the mean sampling is 40 days @xcite .
however , the lack of correlation between the line and continuum flux also points to additional sources of ionization in the blr apart from the central agn continuum source .
we should note here that our result for observed low variability in nls1 ark 564 is in agreement with a comparative study of nls1 and broad line seyfert 1 galaxies ( bls1 ) variability given by @xcite .
they found that the majority of nls1-type agn show variability on timescales from several days to a few years , but with the variabilty amplitudes smaller than in the case of bls1-type agn .
moreover , there are some results in particular study cases that indicate peculiar blr in nls1 agn , as e.g. in nls1 mrk 493 ( see * ? ? ?
* ) where the broad lines have not changed in a period of couple of years ( between two observations ) , and it is suspected that the weak broad h@xmath3 and h@xmath4 may come partly from the unresolved central blr , but also may be partly produced by the violent starburst in the circumnuclear ring ( clearly resolved on hst image ) .
it seems that nls1-type agn have some specific physical properties that have to be considered in monitoring campaigns .
another peculiarity to note is the fact that the line flux appears to flatten at high continuum flux , that is especially seen in the case of ngc 4151 for which the line flux saturates above the continuum flux of @xmath31 ( see thorough analysis and discussion in * ? ? ?
this effect can be also noticed in the case of ngc 5548 and only weakly in 3c 390.3 .
this implys that either there is more ionizing to optical flux than expected for a typical agn spectrum , or that , again , lines are not produced purely by the photoionization from the central continuum source . the photometric radius of the blr c@xmath6 increases with the width of the bel s profiles ( figure 3 ) , except for the case of arp 102b .
this is in accordance with the general picture of the central engine , as it appears that in objects with lower accretion rate ( @xmath32 ) , those with larger line widths @xcite the blr is larger .
arp 102b though remains a mystery , as here some other geometry , apart from the obvious accretion disk geometry that is usually used to explain the double - peaked line profiles , may be considered ( see * ? ? ?
in favor to this is the fact that during more than two decades , the line profiles of this object have not changed significantly ( figure 5 , bottom panel ) .
this is not expected in case of the relatively compact accretion disk ( @xmath281000 gravitational radii ) that was used to modeled these very broad double - peaked line profiles ( see models in * ? ? ?
* ; * ? ? ?
. additionally , there is an indication that an outflow in the blr of arp 102b is detected as shown in figure 15 of @xcite .
finally , the radius - luminosity relationship plotted for these five objects shows that a linear trend is present ( figure 6 ) .
obviously the nlsy1 ark 564 is clearly an outliner .
this is expected due to the low flux variability of this object ( @xmath287% for h@xmath4 line ) and poor significance of the obtained blr radius .
therefore , a simple fit was done through all objects except ark 564 , and we obtained @xmath23 ( figure 6 ) .
this result is different from bentz et al .
( 2013 ) , who obtained @xmath33 , that is also plotted as solid line in figure 6 .
this illustrates that considering only this very low - luminosity end ( @xmath34l@xmath35(5100)@xmath36erg / s ) of the empirical radius - luminosity scaling relation is very sensitive to the measured continuum luminosity and blr radius .
one reason could be the contribution of the host - galaxy continuum flux , that was not considered here , apart from subtracting the extended source correction factor @xmath37 , that is an aperture - dependent correction factor standardly used in our campaign to account for the host - galaxy contribution .
moreover , the sample is far from being statistically significant .
however , this still illustrates that new results of the monitoring campaigns may help in obtaining more accurate results and all agn properties should be taken into account ( e.g. level of line variability or line production mechanism ) , as well as obtaining direct distances need for calculations of the luminosity as discussed in @xcite .
it is obvious that for some objects , the blr geometry is changing during decades of observation , e.g. in case of 3c 390.3 the rms profile of h@xmath4 shows strong variability especially in the blue peak @xcite .
one possible scenario to model this variability is with the orbiting bright spots in accretion disk @xcite .
other proposed scenario , e.g. for ngc 4151 that also has strongly variable bel s profiles @xcite is that this well - studied object is hosting a binary supermassive black hole with two blr orbiting around them @xcite .
@xcite found by analyzing the 20-year long ngc 4151 h@xmath3 light curve and radial velocity curves of the line components the evidence for a sub - parsec scale supermassive binary black hole system with an orbital period of @xmath285800 days .
the comparative study of the main results of sao ras long - term monitoring campaign of five type 1 agn with bels having different properties are given .
the variations of the bels parameters of the h@xmath4 and h@xmath3 line are investigated with the aim to constrain the properties of the blr and test the reverberation mapping as a method to estimate the size of the blr , and consequently the mass of the supermassive black hole in the center of an agn .
the main conclusions are : * the five selected objects are varying during the monitored period , where some are strongly variable ( @xmath2840% ) in the spectral line fluxes ( e.g. ngc 4151 and 3c 390.3 ) ; * the blr is believed to be dominantly photoionized by the central continuum source in the majority of agn , but in some objects ( in this analysis 3 out of 5 , i.e. arp 102b , ark 564 , and ngc 4151 ) , other heating mechanisms such as a shock - ionization due to outflows , may contribute to the formation of the bel ; * the empirical radius - luminosity scaling relation is very sensitive , especially in this low - luminosity end .
it depends on the long - term reverberation monitoring results , so that these investigations continue to be important even though spectroscopy is a very telescope and time - consuming observational technique ; * during decade(s ) of monitored period , the size and geometry of the blr can change , and this should be considered when bels are used to estimate the size of the blr and the mass of the supermassive black hole .
the monitoring results presented here and in several other papers pay attention to the importance of the spectroscopic monitoring campaigns in the investigation of the blr properties and mass estimates of the supermassive black holes in agn .
this work was supported by the ministry of education , science and technological development of republic of serbia through the project astrophysical spectroscopy of extragalactic objects ( 176001 ) , intas ( grant n96 - 0328 ) , rfbr ( grants n97 - 02 - 17625 n00 - 02 - 16272 , n03 - 02 - 17123 , 06 - 02 - 16843 , n09 - 02 - 01136 , 12 - 02 - 00857a , 12 - 02 - 01237a , n15 - 02 - 02101 ) , conacyt research grants 39560 , 54480 , and 151494 .
d.i . has been awarded loral - unesco `` for women in science '' national fellowship for 2014 .
we would like to thank the anonymous referees for very useful comments .
|
the spectral variability of active galactic nuclei ( agn ) is one of their key features that enables us to study in more details the structure of agn emitting regions .
especially , the broad line profiles , that vary both in flux and shape , give us invaluable information about the kinematics and geometry of the broad line region ( blr ) where these lines are originating from .
we give here a comparative review of the line shape variability in a sample of five type 1 agn , those with broad emission lines in their spectra , of the data obtained from the international long - term optical monitoring campaign coordinated by the special astrophysical observatory of the russian academy of science .
the main aim of this campaign is to study the physics and kinematics of the blr on a uniform data set , focusing on the problems of the photoionization heating of the blr and its geometry , where in this paper we give for a first time a comparative analysis of the variabilty of five type 1 agn , discussing their complex blr physics and geometry in the framework of the estimates of the supermassive black hole mass in agn .
galaxies : active - galaxies quasar : individual ( arp 102b , 3c 390.3 , ngc 5548 , ngc 4151 , ark 564 ) line : profiles
|
j. liebig and f. whler , ann .
phys . * 96 * , 369 ( 1830 ) .
n. marcelino , s. brnken , j. cernicharo , d. quan , e. roueff , e. herbst , and p. thaddeus , a&a * 516 * , a105 ( 2010 ) .
r. a. perry and d. l. siebers , nature * 324 * , 657 ( 1986 ) .
j. a. miller and c. t. bowman , int . j.
kinet . * 23 * , 289 ( 1991 ) .
l. e. snyder and d. buhl , astrophys .
j. * 177 * , 619 ( 1972 ) .
d. buhl , l. e. snyder , and j. edrich , astrophys .
j. * 177 * , 625 ( 1972 ) .
i. zinchenko , c. henkel , and r. q. mao , a&a * 361 * , 1079 ( 2000 ) .
nguyen - q - rieu , c. henkel , j. m. jackson , and r. mauersberger , a&a * 241 * , l33 ( 1991 ) .
j. m. jackson , j. t. armstrong , and a. h. barrett , apj * 280 * , 608 ( 1984 ) .
j. m. roberts , p. r. veres , a. k. cochran , c. warneke , i. r. burling , r. j. yokelson , b. lerner , j. b. gilman , w. c. kuster , r. fall , and j. de gouw , pnas * 108 * , 8966 ( 2011 ) . c. t. bowman , 24th symp .
( int . ) combust . , 859 ( 1992 ) .
p. j. young , l. k. emmons , j. m. roberts , j .- f .
lamarque , c. wiedinmyer , p. veres , and t. c. vandenboer , j. geophys
117 * , d10308 ( 2012 ) .
j .- j . klossika and r. schinke , j. chem . phys . * 111 * , 5882 ( 1999 ) .
h. laine berghout , shizuka hsieh , and f. fleming crim , j. chem . phys . * 114 * , 10835 ( 2001 ) .
t. a. spiglanin , r. a. perry , and d. w. chandler , j. chem . phys . *
87 * , 1568 ( 1987 ) .
t. a. spiglanin and d. w. chandler , j. chem .
phys . * 87 * , 1577 ( 1987 ) .
w. k. yi and r. bersohn , chem .
. lett . * 206 * , 365 ( 1993 ) .
r. a. brownsword , t. laurent , r. k. vatsa , h. r. volpp , and j. wolfrum , chem .
lett . * 249 * , 162 ( 1996 ) .
r. a. brownsword , t. laurent , r. k. vatsa , h. r. volpp , and j. wolfrum , chem .
lett . * 258 * , 164 ( 1996 ) .
r. a. brownsword , m. hillenkamp , t. laurent , r. k. vatsa , and h. r. volpp , j. chem . phys . * 106 * , 4436 ( 1997 ) .
r. a. brownsword , m. hillenkamp , h. r. volpp , and r. k. vatsa , res
. intermed . *
25 * , 339 ( 1999 ) .
d. r. yarkony , j. chem . phys . * 114 * , 2614 ( 2016 ) .
d. r. yarkony , mol .
* 99 * , 1463 ( 2001 ) .
h. wang , s .- l .
liu , j. liu , f .- y .
wang , b. jiang , and x .- m .
yang , chinese j. chem .
phys . * 20 * , 388 ( 2007 ) .
z. zhang , z. chen , c. huang , y. chen , d. dai , d. h. parker , and x. yang , j. phys .
a * 118 * , 2413 ( 2014 ) .
|
full - dimensional semiclassical dynamical calculations are reported for the photofragmentation of isocyanic acid in the s@xmath0 state .
these calculations , performed for the first time , allow to closely reproduce the key features of high - resolution imaging measurements at photolysis wavelengths of 201 and 210 nm while providing insight into the underlying dissociation mechanism .
isocyanic acid , hnco , is the simplest molecule that contains h , c , n and o atoms , which are essential for life and major constituents of chemical species in chemistry and biology . since its discovery in 1830 @xcite
, hnco has been identified in diverse media .
for instance , this molecule is found in over sixty galactic sources and in nine external galaxies @xcite .
it is relatively highly abundant and it is suggested to be strongly involved in prebiotic chemistry occurring there .
especially , it is a major component of the photochemistry of building blocks of life ( dna bases and amino acids ) .
isocyanic acid has been identified in the atmosphere @xcite where hnco sources are from both anthropogenic and biomass burning , and also released from combustion ( e.g. fuels , cigarettes @xcite ) .
it is known to be toxic and easily dissolves in water ( in lungs ) posing obvious health risks such as the development of atherosclerosis , cataracts and rheumatoid arthritis @xcite and inflammations that can lead to cardiovascular disease . to - date
, the health effects of isocyanic acid , its role in the atmosphere and its photochemistry are not fully known nor understood .
the photochemistry of hnco under uv radiations or in combustion sources is complex . besides the importance of isocyanic acid in diverse fields ( see above ) , its unimolecular decomposition processes represent benchmarks for tetratomic molecules evolution upon electronic excitation .
adiabatically , the lowest hnco singlet electronic states can not lead to the formation of nh ( x@xmath1 ) + co(x@xmath2 ) products . instead , the observed formation results from electronically excited nh(a@xmath3 ) photofragment ( with co(x@xmath2 ) ) .
previous studies showed that around 200 nm wavelengths , hnco mainly photodissociates in the first excited singlet state s@xmath0 @xcite .
internal conversion to the s@xmath4 ground state ( @xmath5a@xmath6 ( 1 @xmath5a ) ) through seams of conical intersections with the s@xmath0 state ( @xmath5a@xmath7 ( 2 @xmath5a ) ) leads to the photofragments , h + nco @xcite , by spin - allowed unimolecular decay , to nh(x@xmath1 ) + co(x@xmath2 ) , after intersystem crossing and decomposition on the lowest triplet state potential .
up - to - date experimental investigations of hnco photodissociation have been performed by yang and co - workers @xcite . in 2007
, they @xcite the exact quantum treatment of the photodissociation dynamics of tetratomic molecular systems is at the capacity limit of today s theoretical and numerical techniques @xcite .
this is particularly true for the title process , as four different atoms are involved and no symmetry can be used to converge calculations by means of reasonable basis sizes ( not to mention the significant mass of the three second raw atoms that is a complicating factor ) .
therefore , we used the newly developed classical trajectory method ( ctm ) in a quantum spirit @xcite .
this semiclassical approach proved to lead to nearly quantitative predictions for several bimolecular , unimolecular and vibrational predissociation processes @xcite . as for photodissociations , ctm in a quantum spirit involves the three following steps : ( i ) selecting the set of initial conditions of trajectories from the wigner distribution associated with the internal state of hnco before the optical excitation @xcite ; ( ii ) running trajectories from these initial conditions up to the separated products @xcite ; and ( iii ) deducing from the final conditions the pair - correlated keds , vsps and aps by applying bohr quantization rule @xcite to product vibrational motions @xcite .
this basic rule of quantum mechanics states in a language suited to the present problem that one should only take into account those classical trajectories reaching the products with integer vibrational actions .
this has a significant influence on the final results when only a small number of vibrational states are available to the products , as is the case for the fragmentation of hnco at 201 and 210 nm . on the other hand , between ten and several tenths of rotational states are available to the products for this process , so rotational motions need not be pseudo - quantized .
strict application of bohr s condition of quantization is , however , too restrictive from a numerical point of view and , instead , trajectories are usually assigned gaussian statistical weights such that the closer the vibrational actions to integer values , the larger the weights .
this procedure is known as gaussian binning ( gb ) @xcite and ctm in a quantum spirit will thus be called gb - ctm . within the framework of bimolecular collisions , gb has been shown to be consistent @xcite with the semiclassical theory of molecular collisions pioneered by miller and marcus @xcite .
more technical details on the gb - ctm calculations presented further below can be found in the supplemental material .
the coordinate system used in this work is represented in fig .
[ figure1 ] .
-1 cm is the dihedral angle around the nc bond ( @xmath8 for the @xmath9 configuration )
. we also give the body - fixed frame convention.,title="fig:",width=188 ] none of the previous theoretical treatments of the title reaction were performed in full - dimensionality @xcite .
instead , hnco was kept planar throughout its fragmentation , being thus reduced to a five - dimensional ( 5d ) system . for the sake of realism , however , out - of - plane motions should be taken into account .
about two decades ago , klossika and schinke @xcite built an @xmath10 planar potential energy surface ( pes ) fitted to high - level _ ab - initio _ calculations ( ks - pes ) .
we thus took advantage of the availability of this high - quality ks - pes to which we added a contribution depending mainly on @xmath11 and the dihedral angle @xmath12 , fitted to equation of motion coupled - cluster calculations .
this modification is detailed in the supplemental material .
panel a in fig .
[ figure2 ] displays the bi - dimensional cut of the resulting 6d - pes along @xmath11 and the @xmath13 angle , together with two typical trajectories discussed later below .
panel b in fig .
[ figure2 ] shows a similar cut along @xmath11 and the @xmath12 torsional angle ( for both figures , the remaining coordinates were kept at the values corresponding to the trans minimum in the @xmath10 state , given in table v of ref .
@xcite ) .
it is worth noting that r@xmath14 and @xmath12 are strongly coupled to each other , which necessarily modifies the dynamics as compared to those obtained in a planar description . -0.5
cm potential along @xmath11 and the nco @xmath13 angle .
the red and blue lines are two typical trajectories ( see text ) .
b ) similar cut along @xmath11 and the dihedral @xmath12 angle.,title="fig:",width=302 ] the gb - ctm keds at 210 nm are compared with their experimental counterpart in fig . [ figure3 ] ( solid lines ) .
these distributions are deduced from slicing images of co in the vibrotational states ( @xmath15 ) = ( 0,30 ) and ( 0,35 ) @xcite .
the energy available to the products is such that @xmath5nh can only be in the vibrational ground state .
the contributions assigned to the @xmath16 rotational levels are also shown ( dashed curves ) . as a matter of fact
, the main features of the experimental keds ( threshold and cut - off kinetic energies , positions of the rotationally - resolved peaks , overall shape ) are well reproduced by our theoretical treatment .
in particular , both sets of data show the production of rotationally cold @xmath5nh molecules , with @xmath16 smaller than @xmath17 10 . -0.5 cm and @xmath18 ( assuming a perfect energy - resolved experiment).,title="fig:",width=340 ] fig . [ figure4 ] shows similar comparison at 201 nm .
the displayed experimental keds were obtained from slicing images of @xmath5nh in the vibrotational states ( @xmath19 ) = ( 0,5 ) and ( 1,6 ) @xcite ( see figs .
s1-s3 for more @xmath5nh states ) .
again , the overall agreement between experiment and theory ( solid curves ) is very satisfying ( the best agreement being achieved for ( 0,9 ) , ( 1,6 ) , ( 1,8 ) ) .
the bimodality of the rotational state distributions of co in the vibrational ground state , seen in the right - lower panel of fig .
[ figure4 ] , is clearly reproduced in the left - lower panel .
the production of rotationally hot co , discussed further below , is striking in both the experimental and theoretical results .
this is also the case for the ( 1,4 ) and ( 1,8 ) states ( see fig .
s3 ) .
we wish to emphasize that the quality of the accord for both wave lenghts would be lost without respecting bohr quantization , as recently illustrated in ref .
@xcite .
the dashed curves in the left panels are the co vibrational state contributions to the total ked .
interestingly , most of them ( see also figs .
s1-s3 ) are bimodal to some extent .
the areas below the dashed curves are proportional to the populations of their corresponding @xmath20 values , thus leading to the @xmath5nh state - resolved vsd of co after _ ad - hoc _ renormalization to unity ( see fig .
s4 ) .
one thus sees from figs .
[ figure4 ] and s1-s3 that theory does not predict any vibrational inversion at 201 nm , whatever the @xmath5nh state probed .
this is in contrast with the conclusions of yang and co - workers @xcite who found that the vsds of co peak at 1 for @xmath21=0 and nearly all the values of @xmath16 .
however , these authors recognize that one can not avoid any ambiguity in the co vibrational assignment of a given pair - correlated ked when its @xmath20-resolved components overlap @xcite . -0.5 cm and @xmath22 .
the correspondence between combs and @xmath20 is shown in the right panels.,title="fig:",width=340 ] the theoretical aps ( between -0.95 and -0.8 at both 210 and 201 nm ; see fig .
s4 ) are slightly lower than the measured ones ( between -0.8 and -0.5 ) .
nevertheless , our approach corroborates the experimental finding that the fragmentation proceeds through a fast direct impulsive mechanism .
this is also consistent with the short theoretical lifetime of hnco between its optical excitation and its fragmentation ( 305 and 175 fs at 210 and 201 nm , respectively ) .
we discuss in the supplemental material how relaxing two constraints ( @xmath23 ; internal conversion to s@xmath4 ignored ) in our calculations should improve the predictions . finally , in order to check the validity of the supposed link between ( i ) bimodality of rotational state distributions and ( ii ) involvement of two @xmath24 and @xmath9 hnco isomers in the fragmentation dynamics @xcite , we have calculated the distribution of the dihedral angle @xmath12 at @xmath25 bohr , distance nearly defining both transition states ( tss ) ( see fig .
s5 ) . for the vast majority of the trajectories ( @xmath26 ) , @xmath12 is larger than 90 , which separates the basins of attraction of the @xmath24 and @xmath9 hnco isomers and their associated tss .
@xmath12 being equal to 180at the @xmath9 ts , the @xmath24 ts crossing is clearly a minor event .
note that at the franck - condon ( fc ) geometry ( @xmath27 ) , the system lies within the @xmath9 basin .
though selecting initial conditions according to the wigner distribution necessarily generates many trajectories with an initial motion directed towards the @xmath24 basin , the forces towards the @xmath9 basin result sufficiently strong to redirect the trajectories towards it . consequently ,
the bimodality observed theoretically does not result from the involvement of two @xmath24 and @xmath9 hnco isomers in the dissociation dynamics .
rather , we have found that the bimodality is the consequence of an impulsive - deflective mechanism suggested by the two trajectories displayed in fig .
[ figure2 ] ( representing more than 80 % of the trajectories ) .
the red path undergoes a strong impulsion from the wigner region down to the bottom of the well , followed by a deceleration insufficient to prevent the system from crossing the @xmath9 ts with a large angular velocity @xmath28 and thus , a large rotational excitation of co. the blue trajectory performs one loop before leaving the well , but its path beyond the inner ( last ) turning point is fairly similar to the red path . if all the trajectories were like the two previous ones , the rotational distribution would be a relatively narrow unimodal peak centred at @xmath29 . however , part of the trajectories appear to be deflected by the potential wall lying on the right - side of the mountain pass en route to the product valley ( see fig .
[ figure2 ] ; this region is approximately defined by @xmath11 within [ 3.8 , 4.8 ] and @xmath13 below 90 ) .
this is , for instance , slightly the case of the red path .
this deflection , if strong enough , may make the final direction of the trajectories nearly parallel to @xmath11 , leading thereby to a rotationally cold co. consequently , the strong initial impulsion and the subsequent deflection create the large peak and the smaller bump , respectively , as is clearly visible in the lower panels of fig . [ figure4 ] . in summary ,
the theoretical dynamics of the photofragmentation hnco @xmath30 nh(a@xmath3 ) + co(x@xmath31 ) at 201 and 210 nm have been examined , assuming this process entirely takes place in the s@xmath0 state .
specifically , accurate semiclassical trajectory calculations on a new 6d - pes have been carried out .
the pair - correlated kinetic energy distributions ( keds ) and anisotropy parameters ( aps ) measured by yang and co - workers @xcite , following state - selective detection of either co or @xmath5nh via rempi , have been calculated .
the positions , widths and shapes of the keds are in good agreement with the experimental ones for most co or @xmath5nh states probed .
the aps are also in satisfying agreement with experiment , though the perpendicular polarization of the angular distributions is slightly overestimated by theory . on the other hand , we found that the vibrational state distributions ( vsds ) at 201 nm do not involve any inversion , in contrast with the vsds deduced from the experimental keds @xcite .
however , experimentally deduced vsds are necessarily equivocal when the co - product state - resolved components of the keds overlap , as is the case at 201 nm , whereas the theoretical determination of the vsds is unambiguous .
we thus see no contradiction between our results and the measurements of yang and co - workers @xcite .
last but not least , theory precludes the possibility of any link between the existence of two @xmath24 and @xmath9 hnco isomers and the bimodality observed in the keds , as speculated in ref.@xcite .
instead , our treatment shows that bimodality should be the consequence of an impulsive - deflective mechanism due to two repulsive walls of the 6d - pes , the first one producing rotationally hot co products , the second one cooling part of them .
these findings provide new insights into the dissociation dynamics of the title tetratomic prototype fragmentation . at the moment ,
pair - correlated keds , vsps and aps for dissociations involving more than three atoms have been much more often measured than theoretically predicted . in this regard
, the present work fills an important gap ensuring the synergy between experiment and theory for these complex processes .
this work can be straightforwardly extended to the description of state - to - state photodynamics through imaging techniques for polyatomics playing crucial roles in atmospheric and environmental chemistry .
we would like to thank dr .
gustavo a. garcia ( synchrotron soleil , france ) for fruitful discussions .
|
in recent years , fibre - optic connections in telecommunication networks have proven to be suitable for frequency comparisons and frequency distribution with high stability over long distances . in general the frequency of either an ultra - stable continuous - wave ( cw ) laser or a stable microwave reference
is transmitted over an optical fibre and received at the remote site .
several experiments @xcite have shown that frequency comparisons at or below the current accuracy level of the best atomic frequency references , a few times 10@xmath1 @xcite , are feasible over long - haul fibre connections .
for example , the optical frequency of the 1s-2s transition in atomic hydrogen was recently measured with respect to a remote cs frequency standard with 4.5@xmath2 relative uncertainty , employing a 920 km long fibre - optic link performing at the @xmath3 uncertainty level @xcite .
fibre - optic methods for remote frequency comparison can provide significantly better stability than current satellite - based methods . these include two - way satellite time and frequency transfer ( twstft ) @xcite and ( carrier - phase ) common - view global positioning system ( ( cp/)cv - gps ) comparisons @xcite .
the accuracy limits of satellite based methods are on the order of @xmath4 at one day @xcite , which is already insufficient to compare state - of - the - art cesium fountain clocks operating at less than @xmath5 uncertainty level @xcite
. plans for high - resolution laser spectroscopy experiments at vu university amsterdam laserlab and at van swinderen institute , university of groningen , would be greatly facilitated by direct frequency comparisons over an optical fibre - link at stability levels better than @xmath6 .
for this purpose , a @xmath0 km fibre - optic connection between both laboratories has been established using a @xmath7 km dwdm channel provided by surfnet . the top part of fig .
[ fig : surfnet_fibre_network ] gives an overview of the surfnet fibre network in the netherlands .
the optical path between vu university amsterdam and van swinderen institute groningen is marked in green .
the bottom part shows the details of the optical path .
the link consists of two unidirectional fibres .
interferometric detection and active compensation of fibre length changes @xcite is therefore not possible .
however , it is possible to create a bidirectional path in the optical fibre , and implement a compensation system @xcite .
it must be noted that , in contrast to some of the other frequency comparison links @xcite , our link is part of a carrier - grade dwdm optical network in which the fibre is shared with other users , and several other wavelength channels are simultaneously used for data transfer .
the er@xmath8-amplifiers are used by all active dwdm channels . in amsterdam @xmath9 m of intra - office fibre bridges the distance between the surfnet node and the laboratory . in groningen
@xmath10 km of additional underground fibre is needed to bridge the distance from the surfnet node at the computing centre of the university of groningen to the laboratory , adding to a total link length of @xmath0 km . ]
the use of a standard dwdm channel enables us to characterise the performance of unidirectional transmission of optical frequency references in public transport networks carrying live network traffic .
it also allows a systematic study of environmental conditions on the link stability , such as soil temperature variations , which is the main aim of this work .
unidirectional frequency transfer will be useful for institutions and industries which require accurate and reliable time and frequency references , e.g. for calibration of equipment .
for example , optical frequency distribution may be used for accurate length and , in the future , mass measurements ( through the josephson effect and the watt balance ) by referencing to the si second .
this article is structured as follows . in sec .
[ sec : analysis ] a model for the influence of temperature on the stability of fibre - optic frequency transfer is presented , along with a soil temperature model . in sec .
[ sec : method ] , we present our approach to determine the stability of the @xmath0 km optical frequency link , as well as details of the setups used in amsterdam and groningen .
results are discussed in sec .
[ sec : measurement ] , followed by conclusions and an outlook presented in sec .
[ sec : conclusions ] .
the phase @xmath11 accumulated by a monochromatic light wave guided along a certain path of length @xmath12 and effective refractive index @xmath13 can be written as @xmath14 where @xmath15 is the frequency of the light ( in radians per second ) .
several physical processes can lead to phase ( and thus frequency ) variations in fibre - optic links . at short time
scales ( @xmath16 sec ) environmental vibrations couple to the fibre and can therefore cause path length variations , e.g. via stress - induced refractive index variation , which may occur at frequencies up to tens of khz . at time scales longer than 1 s , significant phase variations also occur because of thermal expansion of the fibre and thermally induced changes in the refractive index .
these variations are typically slow and , for a fibre which is installed mainly underground in an outdoor environment , they are mostly affected by the diurnal and seasonal soil temperature cycles @xcite .
our work is focused on the understanding of the long - term stability of fibre links in relation to such temperature variations .
the phase variations due to a time - varying temperature @xmath17 are @xmath18 we can express the relative length variations of the fibre as a function of temperature as @xmath19 where @xmath20 is the thermal expansion coefficient of the fibre . furthermore it is customary to write @xmath21 with @xmath22 the thermo - optic coefficient . note that both @xmath20 and @xmath22 are weakly dependent on temperature , which we ignore here .
a typical ( room - temperature ) value of the thermal expansion coefficient of the fibre glass is @xmath23/ @xmath24c @xcite .
the thermo - optic coefficient has a typical value of @xmath25/@xmath24c @xcite , and is therefore the main cause of phase variations due to temperature changes .
similar effects due to varying air pressure are approximately two orders of magnitude smaller @xcite and have therefore not been included in the model .
considering the heat flux in isotropic media ( soil ) for a vertical temperature gradient and varying temperature , and modelling the temperature variation as a sinusoidal periodic signal , van der and @xcite derive the equation for the temperature of the soil at a certain depth @xmath26 and time @xmath27 as @xmath28 where @xmath29 is the average temperature at the surface ( @xmath30 ) , @xmath31 is the amplitude of the temperature variation at the surface with period @xmath32 , and @xmath33 is an arbitrary time offset .
the phase constant @xmath34 includes the soil constant @xmath35 which is determined by the thermal conductivity @xmath36 , the specific heat capacity @xmath37 , and the mass density @xmath38 of the soil according to @xmath39 equation can be applied to both diurnal and annual variations in temperature .
a remark must be made that accurate modelling of the soil temperature is delicate and involves , among others , the groundwater levels and groundwater freezing rates in winter @xcite . in this model
the amplitude of the temperature wave decreases exponentially with depth , while it undergoes a phase shift linear with depth .
the equation is universal and can also be applied to other materials as long as the physical properties @xmath36 , @xmath38 and @xmath37 are known . with the help of eq .
, average frequency deviations @xmath40 from the nominal frequency @xmath41 can be calculated for any depth , using eq . for temperature differences
@xmath42 over a time interval @xmath43 as @xmath44
in order to characterise the frequency stability of the fibre link , two different methods are used . the first method
consists of the one - way transmission of a c - band - wavelength cw laser , locked to a mode of an er@xmath8-doped fibre frequency comb laser in amsterdam . in groningen , the transmitted laser frequency
is measured using a similar frequency comb laser .
both frequency combs are locked to gps - disciplined atomic clocks .
therefore , the stability of the optical frequency measurement in amsterdam and in groningen depends on the fibre link as well as the atomic clock stabilities .
the second method employs a closed fibre loop amsterdam
amsterdam . in this case
the laser frequency can be compared with itself after its roundtrip through the fibre loop .
the second method takes advantage of the fact that the laser frequency instability on the time scale of the measurement is much smaller than the instabilities introduced in the fibre loop .
figure [ fig : fibre_network_experimental_setup ] gives an overview of the measurement setup that is used to characterise the frequency stability of the fibre link .
a narrow - linewidth cw laser ( redfern integrated optics inc . ( rio ) planex ,
3 khz linewidth , 20 mw output power , and wavelength 1559.79 nm [ itu channel 22 ] ) is phase - locked to the fibre frequency comb laser ( menlo systems fc1500 ) and used as an absolute optical frequency reference . to verify the lock quality ,
the in - loop beat signal is counted .
the optical reference frequency is sent to groningen via the fibre link ( injected power 0 dbm ) .
the nominal length @xmath12 of the fibre - link is @xmath0 km between the laboratories in amsterdam and groningen .
the optical fibres of the pair are located in a fibre - bundle and thus follow nearly the same physical path . in groningen
the light is split .
part of the light is used in a frequency comparison against the local optical frequency standard ( menlo systems fc1500 ) , while the other part of the light is sent back to amsterdam ( received power 06 dbm ) . as pointed out above , the comparison of the frequency of the light after the roundtrip with the cw laser source reveals the noise contribution from the fibre link .
both optical frequency combs are locked to rubidium ( rb ) frequency standards ( srs fs725 ) , which are disciplined to the 1 pps output of gps receivers ( amsterdam : trimble acutime 2000 , groningen : navsync cw46 ) .
the combined instability of the gps - receiver output and rb clocks is transferred to the frequency comb lasers via the various rf locks used to stabilise the frequency comb laser repetition rate frequency @xmath45and carrier envelope offset frequency @xmath46 .
figure [ fig : fibre_network_experimental_setup_laser_locking ] shows the details of the cw laser stabilisation setup .
part of the light of the diode is split off and fed to a fibre - based beat - note unit consisting of a dwdm filter to reject a large part of the frequency comb spectrum , a fused coupler , and a fibre - coupled photodiode ( @xmath10 ghz bandwidth , 50/125 @xmath47 m multi - mode fibre coupled ) to detect the rf beat signal .
the rf beat signal ( @xmath48 mhz ) is bandpass filtered ( filter bandwidth @xmath49 mhz ) and amplified before comparison with a signal generator ( agilent 33250a , referenced to the rb clock ) , via counting phase detector electronics .
feedback on the diode laser phase is achieved via a fast pid controller acting directly onto the diode - laser injection current ( bandwidth @xmath50 mhz ) . to verify proper cw - laser locking conditions during the experiment , the in - loop rf beat - frequency
is recorded with a rb - referenced counter .
for dead - time - free counting of rf frequencies , either a quasi - continuous double agilent 53132a counter setup is used , or an agilent 53230a counter in a continuous reciprocal frequency counting mode .
finally , a regulated variable optical attenuator ensures that the optical power injected into the fibre link remains constant .
-fibre frequency comb laser .
the photo - diode signal of the fibre - coupled beat unit is amplified , filtered and split by a 3 db power splitter ( s ) for input to the phase detector and the counter .
the stabilisation setup is fully referenced to the gps - disciplined rb frequency standard .
the monitor ports are used to observe optical power variations of the planex laser before and after the variable optical attenuator ( voa ) , which regulates the laser power to a constant level before injection into the telecommunication network . ] to characterise the stability of the fibre link , the setup of fig .
[ fig : fibre_network_experimental_setup_roundtrip_measurement ] is used .
light of the diode laser is split by a fused coupler .
part of the light is sent via the roundtrip amsterdam
amsterdam , while another portion of the light is frequency shifted by an acousto - optic modulator ( aom ) .
the output of the aom is combined with the roundtrip optical signal after transmission by the fibre link ( power @xmath51 dbm ) and the beat frequency is detected with a fibre - coupled photo - diode .
any frequency variations introduced by the fibre link can be measured as frequency deviations from the aom frequency , which is generated by a rb - referenced dds ( analog devices ad9912 ) .
mhz ) . in both cases
the aom was driven by a rb - referenced dds unit whith a set accuracy of @xmath52 @xmath47hz .
frequency deviations of the link are recorded with a rb - referenced counter . ] to measure the roundtrip stability two aom + counter setups were used in succession during the experiments .
initially an agilent 53230a counter in reciprocal continuous mode was used to count the beat frequency of a free - space double - pass aom unit ( 300 mhz , results of sec .
[ sec : clock_transfer_results ] ) .
after that , this setup was replaced by a zero - dead - time k+k fmx-50 counter to count the beat frequency of a fibre - coupled aom ( -42 mhz , results of sec .
[ sec : roundtrip_instability_results ] ) .
the remote ( groningen ) optical frequency characterisation setup is depicted in fig .
[ fig : fibre_network_experimental_setup_clock_comparison ] .
the link laser is amplified using a semiconductor optical amplifier , and guided to a free - space beat unit .
the beat frequency is counted using a k+k fmx-50 counter .
the frequency comb laser and counters are frequency referenced to the rb - standard .
two series of measurements have been performed on the fibre link , namely an optical - versus - gps measurement for which the optical frequency is measured simultaneously at amsterdam and groningen and compared , and a roundtrip measurement at the amsterdam site ( fig .
[ fig : fibre_network_experimental_setup ] ) .
it is known that soil temperature variations of the fibre may lead to significant frequency instability ( see , for example , @xcite ) . in this section ,
we present a model to describe and predict the influence of soil temperature variations on the frequency stability of underground fibre links .
the results of the temperature model are compared with actual soil temperature and link stability measurements .
apart from the instability contributed by the fibre link , the stability of the frequency transfer is limited by three sources .
first , the stability of the frequency comb laser is limited by the rb / gps rf reference oscillator used to control the parameters of the comb .
this reference has a specified relative instability of @xmath53 at 1 second and below @xmath54 between @xmath55 and @xmath56 seconds , with a minimum around @xmath57 .
this long - term frequency instability is transferred to the cw link laser through the various rf locks in the setup .
the rf locks themselves also contribute to the frequency instability , which we assess as follows . using a second , similar , cw laser ( rio orion )
locked to the frequency comb , and employing the virtual beat note technique @xcite , the combined instability of the locking electronics is determined to be @xmath58 at 1 second .
thus , on time scales longer than 1s , the rf noise is effectively averaged out so that its influence on the link measurements may be neglected .
a second source of instability plays a role in the roundtrip measurements , for which a frequency drift of the laser source may lead to an apparent frequency shift of the ( delayed ) light transmitted by the fibre loop . considering the 3.2 ms roundtrip propagation delay of the light over the fibre link ,
the ( linear ) drift of the frequency standard ( @xmath59 at 1 s ) only plays a role at the level of @xmath60 at 1 s , decreasing as @xmath61 , the inverse of the counter gate time , due to the fractional time overlap @xmath62 between reference and roundtrip light in the frequency comparison .
this is one order of magnitude smaller than the typically measured link stability at 1 s , and averages down more rapidly with increasing @xmath63 . for the optical - versus - gps comparison ,
a third source of instability is due to the fact that the rb clocks in amsterdam and groningen are independently locked to gps time .
small intrinsic phase differences between gps signals , received at geographically separated locations , thus propagate through the frequency locks , and may manifest itself as additional noise in the link frequency measurement .
the intrinsic frequency transfer stability of a fibre link is best measured on a closed loop , so that the frequency of the input signal can be directly compared with that of the roundtrip signal .
figure [ fig : roundtrip_stability-120706_vs_130930 ] shows the results of two roundtrip stability measurements .
the first measurement is based on a 13-hour time series of frequency measurements acquired on 2012 - 07 - 06 , using the free - space aom unit ( fig .
[ fig : fibre_network_experimental_setup_roundtrip_measurement ] ) while recording with the agilent 53230a . for this dataset the absolute frequency of the roundtrip optical signal was calculated and used to determine the overlapping allan deviation ( odev ) . to compare the measurement with the soil temperature model , statistics on the scale of several days are needed .
therefore a longer roundtrip stability measurement took place from 2013 - 09 - 30 to 2013 - 10 - 09 , using the fibre - coupled aom while recording with the k+k fmx50 counter .
this measurement contains a few periods during which the beat signal was too low for the fmx50 input circuits to record properly , leading to frequency outliers .
outliers were removed according to chauvenets criterion ( @xmath64 , 4102 datapoints ( 0.52% of the total set ) removed ) and were replaced with the median of the dataset .
we have verified that the odev statistics are not influenced significantly by this operation . in the previous section we established that for averaging periods of more than 10 s
the frequency transfer instability of the link is @xmath65 .
given the stability of the laboratory frequency standards ( rb clocks ) it is therefore to be expected that a direct measurement of the link laser frequency at the remote site ( groningen ) yields the mutual clock stability . during the 2012 measurement session ,
the absolute laser frequency at the remote site was recorded .
the result of this measurement , together with the roundtrip stability and the stability of the link laser lock frequency are presented in fig .
[ fig : link_measurement_120706 ] . the link laser frequency was determined with respect to the frequency comb , taking into account variations of @xmath45and @xmath46 .
the determination of the absolute laser frequency is limited by the resolution of the @xmath45counter to @xmath66/s .
the maximum observable roundtrip instability is therefore about 1@xmath2/s .
the observed instability is significantly higher , showing that the fibre may already influence the roundtrip stability at the second time scale .
the frequency of the laser used for the characterisation of the remote link was determined with respect to the nearest frequency comb mode .
together with the mode number determination , this directly delivers the remote absolute optical frequency . as can be seen in fig .
7 , the link laser frequency instability at the remote site is much larger than the total roundtrip instability .
therefore , the local frequency references are the limiting factor .
this means that we have effectively performed a frequency comparison between the gps referenced rb atomic clocks over the fibre link .
the measured performance is slightly worse than expected based on the allan variance graph in the datasheet ( @xmath67 at @xmath68 s ) of the clocks .
this is probably due to imperfect gps reception , and uncorrelated frequency noise of the fibre combs used in this experiment .
based on coarse estimates we expected ( soil ) temperature fluctuations to have a major influence on the passive stability of the fibre link .
the royal netherlands meteorological institute ( knmi ) provided us with soil temperature measurements taken at the cabauw site ( see fig .
[ fig : surfnet_fibre_network ] ) .
temperatures are measured at depths of 0 , 2 , 4 , 6 , 8 , 12 , 20 , 30 and 50 cm every 12 seconds and averaged over 10 minute intervals .
out of five knmi locations the site at cabauw is the only location for which soil temperature data with such high temporal resolution is available in the netherlands .
the provided datasets consist of one set covering the 9 day link measurement with a 12 second time interval in october 2013 @xcite , and a 2 year dataset ( 2011 , 2012 ) with a 10 minute interval @xcite .
equation is used to convert these temperature series to frequency deviations expected on a @xmath0 km fibre link with the given thermal expansion and thermo - optic coefficients . in the next sections we use these reference data to compare the link stability with the soil temperature model , eq . . the raw frequency data from the 9-day link stability measurement
is compared to the frequency deviations as derived from the knmi soil temperature dataset over the same time period .
the first question that needs to be addressed is to what extent such a comparison , correlating temperature effects measured in locations separated by tens of kilometres , is meaningful .
the cabauw measurement site ( fig .
1 ) is located more than 40 km south of the ( geographically ) 200 km long trajectory of the fibre . therefore the exact trends in soil temperature at the cabauw site and along the fibre link can differ , due to local variations in solar irradiation and precipitation .
nevertheless , significant temperature correlations are expected as the different locations are relatively close , as seen from a meteorological and climatological point of view .
slope , which indicates that on shorter time scales temperature noise is significantly more prominent in the knmi measurements than in the temperature - dependent link stability . ]
figure [ fig : roundtrip_comparison_measurement_knmi ] shows the roundtrip stability in comparison to the fibre - link stability calculated from the knmi soil temperature data .
knmi aims at a relative accuracy among sensors of 0.01 k and an absolute accuracy of 0.1 k. the noise in the soil temperature measurements amounts to a few mk .
more accurate soil temperature observations will be difficult to make , let alone capturing all relevant variations along the optical path .
all stabilities derived from soil temperature show a @xmath69 slope for shorter time scales , which indicates the presence of white noise in the temperature measurements that does not appear in the fibre link instability .
the noise levels at these time scales obscure frequency drift due to the daily temperature cycle .
this cycle leads to the rising slope at longer time scales , with a local maximum at an averaging time of half a day . on even longer time scales
, this instability cycle continues with a one - day period , and with local maxima decreasing in height over time .
this is most clearly seen in the surface temperature curve .
the measured roundtrip stability curve shows averaging at short time scales less than 100 s ,
while at time scales larger than 1000 s frequency drifts due to the diurnal temperature cycle dominate the instability .
the origin of the level of instability at time scales less than 4@xmath70 s remains unclear .
possible causes are the several hundreds of meters of the fibre link located inside buildings , which are subject to significant and relatively fast temperature variations ( due to e.g. airconditioning systems ) , while other factors like stress induced frequency fluctuations can not be entirely ruled out . for averaging times larger than 4@xmath70 s , the link measurement shows qualitative agreement with the knmi data at depths of 2030 cm .
the correlation between soil temperature and link frequency drift can be inferred more directly from the temperature and frequency measurement time series . to this end ,
the roundtrip frequency deviations from the aom frequency , @xmath40 , are compared to the frequency deviations @xmath71 due to soil temperature @xmath72 at depth @xmath13 , estimated using eq . .
both datasets are averaged over two - hour windows to reduce noise levels .
figure [ fig : roundtrip_time_series ] shows the raw data of the nine - day link measurement in comparison with @xmath71 at 20 cm and 30 cm depth ( 2 hour averages ) .
the depth at which the fibre link is buried is not precisely known , and it furthermore may vary along the path of the link .
an estimate of the effective depth of the link may be obtained by assigning a weighting factor @xmath73 to each time series @xmath71 , and minimising ( by least squares fitting ) the difference @xmath74 between the measured and calculated frequency deviations @xmath75 the frequency offset parameter @xmath76 is needed to include a systematic offset between the link measurement data and the knmi data .
such an offset may be caused by an overall relative temperature change between the fibre link path and the cabauw site .
figure [ fig : roundtrip_time_fit_parameters ] shows the variation of the fit parameters @xmath77 and @xmath76 over time .
these parameters and their time dependence are obtained as follows .
first , all input data ( i.e. the measured frequency deviations and the frequency deviations estimated from soil temperature measurements ) are averaged with a one - hour window .
of the averaged data sets , a 24-hour subset is taken ( labeled by the median of the time stamps in the set ) , for which the coefficients @xmath77 and @xmath76 are found by least - squares fitting .
this last step is repeated for a 24-hour subset which is offset by six hours with respect to the previous subset , until the entire data set is covered . from fig .
[ fig : roundtrip_time_fit_parameters ] it follows that best agreement is found for an average fibre depth of about 30 cm .
the least - squares fit method yields solutions which are generally well aligned in phase with the measured data ( fig .
[ fig : roundtrip_time_series ] ) . however , fig .
[ fig : roundtrip_time_series ] also shows that the agreement between the amplitude of the frequency deviations at 30 cm depth and the measured data is poor . for the most important depths ; the @xmath77 found for the other depths are negligibly small .
averages and standard deviations over this dataset are @xmath78 @xmath79 , @xmath80 , and @xmath81 . ] in the time series of the link measurement ( fig .
[ fig : roundtrip_time_series ] ) the frequency deviations before 3 october and after 6 october appear to follow the predictions based on knmi data , while between 3 october and 6 october the curves seem to be substantially less correlated .
this discrepancy might be linked to the fact that during this period , the fibre - link path received considerably more precipitation on 3 october 2013 than the measurement site at cabauw @xcite . moreover ,
solar irradiation differed substantially from day to day between cabauw and the fibre link in this period @xcite .
it is conceivable that this led to local soil temperature variations along the fibre link and , thus , to the observed discrepancy .
this behaviour illustrates the limited power of the temperature model for predicting or estimating instantaneous frequency variations based on soil temperature measurements .
soil temperature data can be used to predict the long - term stability of fibre links , but it can be burdensome to obtain or construct long historical records .
for example in 1961 soil temperature was measured 3 times a day @xcite . also , soil temperature is less often measured at meteorological measurement sites than other quantities . in the netherlands knmi
has such data available for only four sites . in case of scarce soil temperature data , the sinusoidal soil temperature model ( eq . ) can be used to construct an artificial temperature cycle by superposition of a diurnal and an annual temperature cycle , which can be used to estimate the frequency transfer stability of fibre links .
such models can also be used to predict frequency transfer stability for various types of soil .
the amplitude of the diurnal temperature variation itself varies approximately sinusoidally during the year , and is given by @xmath82 where @xmath83 is the average diurnal temperature variation , @xmath84 is the amplitude of the annual variation of the diurnal amplitude , @xmath85 is the annual period , and @xmath86 is used to shift the temperature cycle to fit the model to the long term measurement data of knmi .
the total annual temperature cycle now becomes @xmath87 where @xmath88 and @xmath89 are given by eq . , but now with the amplitude @xmath31 of the diurnal variation being a function of time .
we obtain a set of model parameters for eq . by fitting eq . to the data from van der and @xcite as an estimate of the annual variations , and to the data from @xcite to estimate the cycle of diurnal temperature variations .
the soil constant was taken @xmath90 , which is representative for sand ( being in a state between wet and dry ) , relatively dry loam , and clay @xcite .
the obtained average , amplitude and phase values are given in table [ tab : soil_model_fitted_parameters ] .
.the parameters of the soil temperature model of eq . retrieved by a least - squares fit to data obtained from @xcite and @xcite . [ cols="<,^,^,^",options="header " , ] the soil temperatures resulting from the model are plotted in fig .
[ fig : soil_temperature_yearly_fluctuation ] and compared with the measured knmi data set covering the year 2011 .
the plot compares the data at the surface and at 50 cm depth and shows that the model is indeed in reasonable agreement with direct soil temperature measurements .
a long time series of soil temperatures was generated from the model , and estimates of the long term frequency stability of a hypothetical fibre link of the amsterdam groningen type were calculated using eq . .
figure [ fig : soil_temperature_allan_deviations ] shows the calculated stabilities at the day - to - year time scale , and for various depths . from the graph it is clear that more deeply buried fibres offer a substantial stability improvement at the timescale of days , while the stability at the year scale is much closer to that of fibre buried closer to the surface . ) at several depths .
frequency stability at ( solid ) the surface , ( long dashed ) 50 cm depth , ( short dashed ) 100 cm depth and ( dotted ) 200 cm depth .
the dash - dotted lines indicate the @xmath69 behaviour with a maximum instability of @xmath91 at half a day for the diurnal variation and @xmath92 at half a year for the annual variation . ]
figure [ fig : soil_temperature_model_measurement ] compares the frequency stability , computed using eq . and the soil temperature measurement series of 2011 and 2012 , with that obtained from eq . and the sinusoidal model , eq . .
the most prominent feature is the discrepancy for 50 cm depth , likely due to fluctuations in weather conditions on long ( days / weeks / months ) time scales , which cause a higher instability at time scales between a day and half a year .
these fluctuations are clearly visible in fig .
[ fig : soil_temperature_yearly_fluctuation ] .
this noise also leads to a smoothing of the strong minima in the frequency instability curves , which are a consequence of the sinusoidal temperature behaviour .
we investigated the passive frequency stability of a @xmath0 km carrier - grade , unidirectional fibre link between vu university amsterdam laserlab and van swinderen institute , university of groningen , using a single dwdm channel , and with live optical data traffic present in other dwdm channels .
the observed frequency instability of the link lies in the range @xmath93 to @xmath94 for averaging times 10
8@xmath95 s. this result implies that such fibre links are well suited to distribute the frequency of commercial rb atomic clocks with negligible loss of accuracy .
a model for thermo - optical fibre length variation was developed which relates frequency variations to soil temperatures as a function of depth .
we employed this model taking actual long - term soil temperature measurements as input , as well as predictions obtained from existing soil temperature models .
we combined link frequency measurements with soil temperature measurements at the knmi cabauw site to show that the observed link stability corresponds to an average fibre depth of about 30 cm .
qualitative agreement is found between the soil temperature model and knmi measurements , while predictions of frequency stability based upon this model agree with actual roundtrip frequency measurements to within an order of magnitude .
although the predictive power of the soil - temperature model is limited , it does provide insight into the relation between soil temperature , fibre - optic path length variations and frequency - transfer stability .
our model thus allows estimating the passive frequency stability of fibre links for averaging times ranging from days to years , and allows to estimate upper bounds on the passive link instability .
such information will be useful for the design of future one - way frequency distribution systems based on underground fibre - optic infrastructure .
the presented results show that soil temperature fluctuations have a large impact on the passive frequency stability of optical carriers over underground fibre links for time scales longer than approximately 1000 seconds .
the results of our study confirm the conclusions of previous work that fibre - optic infrastructure is sufficiently stable for one - way atomic clock frequency distribution over hundreds of kilometres distance , and with @xmath96 relative instability @xcite .
this figure compares favourably to the stability of commercial gps - disciplined rb clocks .
the residual frequency variations are sufficiently small to back up the local oscillator of gps - referenced clocks with indefinite holdover ( provided a non - gps - referenced master clock is used ) .
this opens up the perspective of a terrestrial `` flywheel '' oscillator , embedded in the fibre - optic telecommunications network , which can be used to back - up gps - referenced oscillators during periods of gps outage .
this work was supported by the netherlands foundation for fundamental research of matter ( fom ) through the program `` broken mirrors & drifting constants '' .
j.c.j.k . thanks the dutch organisation for scientific research ( nwo ) and the dutch technology foundation ( stw ) for support .
o. lopez , a. amy - klein , c. daussy , c. chardonnet , f. narbonneau , m. lours , and g. santarelli , `` 86-km optical link with a resolution of @xmath97 for rf frequency transfer , '' eur .
j. d * 48 * , 3541 ( 2008 ) .
k. jaldehag , s .- c .
ebenhag , p. o. hedekvist , and c. rieck , `` time and frequency transfer using asynchronous fiber - optical networks : progress report , '' in `` proceedings of the 41@xmath98 annual precise time and time interval ( ptti ) meeting , '' ( 2009 ) , pp .
383396 .
g. grosche , o. terra , k. predehl , r. holzwarth , b. lipphardt , f. vogt , u. sterr , and h. schnatz , `` optical frequency transfer via 146 km fiber link with @xmath99 relative accuracy , '' opt . lett . * 34 * , 22702272 ( 2009 ) . k. predehl , g. grosche , s. m. f. raupach , s. droste , o. terra , j. alnis , t. legero , t. w. hnsch , t. udem , r. holzwarth , and h. schnatz , `` a 920-kilometer optical fiber link for frequency metrology at the 19@xmath100 decimal place , '' science * 336 * , 441444 ( 2012 ) .
o. lopez , a. haboucha , b. chanteau , c. chardonnet , a. amy - klein , and g. santarelli , `` ultra - stable long distance optical frequency distribution using the internet fiber network , '' opt .
expr . * 20 * , 2351823526 ( 2012 ) .
t. rosenband , d. b. hume , p. o. schmidt , c. w. chou , a. brusch , l. lorini , w. h. oskay , r. e. drullinger , t. m. fortier , j. e. stalnaker , s. a. diddams , w. c. swann , n. r. newbury , w. m. itano , d. j. wineland , and j. c. bergquist , `` frequency ratio of al@xmath101 and hg@xmath101 single - ion optical clocks ; metrology at the 17@xmath100 decimal place , '' science * 319 * , 18081812 ( 2008 ) .
n. hinkley , j. a. sherman , n. b. phillips , m. schioppo , n. d. lemke , k. beloy , m. pizzocaro , c. w. oates , and a. d. ludlow , `` an atomic clock with @xmath102 instability , '' science * 341 * , 12151218 ( 2013 ) .
b. j. bloom , t. l. nicholson , j. r. williams , s. l. campbell , m. bishof , x. zhang , w. zhang , s. l. bromley , and j. ye , `` an optical lattice clock with accuracy and stability at the @xmath102 level , '' nature * 506 * , 7175 ( 2014 ) .
a. matveev , c. g. parthey , k. predehl , j. alnis , a. beyer , r. holzwarth , t. udem , t. wilken , n. kolachevsky , m. abgrall , d. rovera , c. salomon , p. laurent , g. grosche , o. terra , t. legero , h. schnatz , s. weyers , b. altschul , and t. w. hnsch , `` precision measurement of the hydrogen 1@xmath1032@xmath103 frequency via a 920-km fiber link , '' phys . rev . lett . * 110 * , 230801 ( 2013 ) . v. zhang , t. e. parker , j. achkar , a. bauch , l. lorini , d. matsakis , d. piester , and d. g. rovera , `` two - way satellite time and frequency transfer using 1 mchip / s codes , '' in `` proceedings of the 41@xmath105 annual precise time and time interval ( ptti ) meeting , '' ( 2009 ) , pp .
371382 .
d. piester , a. bauch , l. breakiron , d. matsakis , b. blanzano , and o. koudelka , `` time transfer with nanosecond accuracy for the realization of international atomic time , '' metrologia * 45 * , 185198 ( 2008 ) .
d. w. allan and m. a. weiss , `` accurate time and frequency transfer during common - view of a gps satellite , '' in `` proceedings of the 34@xmath104 annual frequency control symposium , '' ( 1980 ) , pp .
334346 .
v. s. zhang , t. e. parker , and m. a. weiss , `` multi - channel gps / glonass common - view between nist and usno , '' in `` proceedings of the 2000 ieee / eia international frequency control symposium and exhibition , '' ( 2000 ) , pp .
598606 .
ma , p. jungner , j. ye , and j. l. hall , `` delivering the same optical frequency at two places : accurate cancellation of phase noise introduced by an optical fiber or other time - varying path , '' opt .
* 19 * , 17771779 ( 1994 ) . e. a. elias , r. cichota , h. h. torriani , and q. de jong van lier , `` analytical soil - temperature model : correction for temporal variation of daily amplitude , '' soil sci .
j. * 68 * , 784788 ( 2004 ) .
r. emardson , p. o. hedekvist , m. nilsson , s .- c .
ebenhag , k. jaldehag , p. jarlemark , c. rieck , j. johansson , l. r. pendrill , p. lthberg , and h. nilsson , `` time transfer by passive listening over a 10-gb / s optical fiber , '' ieee trans .
57 * , 24952501 ( 2008 ) .
|
results of optical frequency transfer over a carrier - grade dense - wavelength - division - multiplexing ( dwdm ) optical fibre network are presented .
the relation between soil temperature changes on a buried optical fibre and frequency changes of an optical carrier through the fibre is modelled .
soil temperatures , measured at various depths by the royal netherlands meteorology institute ( knmi ) are compared with observed frequency variations through this model .
a comparison of a nine - day record of optical frequency measurements through the @xmath0 km fibre link with soil temperature data shows qualitative agreement .
a soil temperature model is used to predict the link stability over longer periods ( days - months - years ) .
we show that one - way optical frequency dissemination is sufficiently stable to distribute and compare e.g. rubidium frequency standards over standard dwdm optical fibre networks .
|
recently , there have been several studies of the milky way s magnetic field by the means of observing the rotation measure of compact polarized objects like extra - galactic point sources ( e.g. @xcite ) or pulsars ( see talks by a. noutsos and j .- l .
however , we still do not know the large - scale magnetic field configuration or even the number of field reversals within our galaxy .
one problem is that through faraday rotation studies of extra - galactic point sources we only derive the average magnetic field parallel to the line of sight @xmath0 through our galaxy weighted by the electron density @xmath1 , because the rotation measure @xmath2 is given by : @xmath3 here @xmath2 is given in rad / m@xmath4 , @xmath5 in @xmath6 g , @xmath1 in @xmath7 , and the pathlength @xmath8 in pc .
in addition extragalactic sources may suffer from intrinsic faraday rotation of unknown magnitude .
faraday rotation studies of pulsars average @xmath0 between us and the pulsar weighted by @xmath1 .
in addition pulsars distances are usually quite uncertain and pulsars probe only a very small area in space since they are only about 15 km in diameter .
one big problem of the averaging procedure is that there could be numerous field reversals along the line of sight , which would be averaged out .
this ambiguity could be solved if we had anchor points for the magnetic field within our galaxy .
we propose to determine these anchor points with polarization and faraday rotation studies of supernova remnants ( snrs ) , since these can be used as magnifying glasses of their ambient magnetic field .
we assume that the supernova remnants are adiabatically expanding into an ambient medium of constant density @xmath9 and constant homogenous magnetic field @xmath10 . in our model
the snr is circular and has a shell with a width of about 10% of the snr s radius in which the ambient medium and its magnetic field are compressed by a factor of 4 . while the electron density in the shell is assumed to be constant the magnetic field changes with the angle between the snr s expansion direction and the ambient magnetic field ( see fig .
[ sketch ] ) .
it is proportional to the `` number '' of magnetic field lines , which are swept up by the expanding shock wave , hence the magnetic field inside the shell is strongest where the snr is expanding perpendicular to the ambient magnetic field and 0 , where it is parallel .
the synchrotron emission is integrated along the line of sight from the back of the snr to the front and appropriately faraday rotated ( see equation [ rm ] ) .
the synchrotron flux density @xmath11 produced by a spectrum of relativistic electrons @xmath12 at frequency @xmath13 is given by : @xmath14 here @xmath15 is the energy and the values for @xmath16 and @xmath17 are defined by @xmath12 .
( from top to bottom : @xmath18 , @xmath19 , @xmath20 , and @xmath21).,width=491 ] in fig .
[ model ] we display the simulated emission structure in total intensity and polarized intensity and the internal rotation measure that would be observed from a typical supernova remnant at different viewing angles @xmath22 . in this simulation
the ambient magnetic field is pointing away from us from the front left to the back right and @xmath22 is the angle between the plane of the sky and the magnetic field lines the snr is expanding into . for negative @xmath22 here , the magnetic field would point towards us from the back left to the front right . in the simulation shown in fig .
[ model ] the ambient magnetic field is @xmath23 g , the ambient density is 1@xmath7 , the distance of the snr is 2kpc , and its diameter is 20pc .
the simulation was convolved to a resolution of 2.5. in emission we find the typical bilateral structure of snrs in total power and polarized intensity with a tangential magnetic field up to a @xmath22 of about @xmath20 , beyond which the snr becomes circular and thick - shelled with a radial magnetic field .
the surface brightness of the snr is decreasing form @xmath18 to @xmath24 by a factor of almost 10 .
the rotation measure structure reveals a few very interesting characteristics .
the internal rotation measure in the centre of the snr is always 0 , which indicates that from observations we can determine the foreground @xmath2 there .
the entire snr except for @xmath18 is dominated by rm of one sign ; in fig . [ model ] it is mostly negative .
this indicates whether the ambient magnetic field is pointing away from us or towards us .
there is also a smaller area of opposite sign .
its location indicates the orientation of the ambient magnetic field and its distance from the centre gives an approximate value for @xmath22 .
in addition we find that the rm on the two arcs shows a linear behaviour up to about @xmath25 ( see fig .
[ grad ] ) .
this can be described by its gradient , which depends entirely on @xmath26 and the size of the snr and is almost independent of @xmath22 and the @xmath2 in the middle , which is determined from @xmath22 and the foreground @xmath2 . to show how we can exploit these characteristics
we will give two examples .
in fig . [ da530 ] we display a polarized intensity and rotation measure map of the snr da530 , which is located high above the galactic plane at a latitude of about @xmath27 .
the overlaid vectors in the polarized intensity image indicate a tangential magnetic field , which we expected from the simulations for a snr with a @xmath22 of less than @xmath20 .
the rotation measure in the centre is about 0rad / m@xmath4 , hence we can neglect foreground effects .
the rm map is dominated by negative rotation measures , but there is a small area of positive rm to the left of the centre .
this indicates that the snr is expanding into a magnetic field , which is pointing away from the front left to the back right . as can be seen in fig .
[ da530 ] the rm values on both shells differ significantly , however , the gradient seems to be the same .
this implies that the ambient density and magnetic field strength are the same for both shells
. the difference could be either in the foreground rm or the ambient magnetic fields for the two shells have a different @xmath22 . to find such a large difference in rotation measure on such a small scale in the foreground is very unlikely , since it would require either another snr or an hii region in the foreground to produce such a large effect
both of which would be easily detectable by some other means .
the only possibility left is twisted ambient magnetic field .
a simulation for both shells indicates that the top shell is expanding into a magnetic field with @xmath28 and the bottom one with @xmath29 .
the lower surface brightness of the top shell supports this finding .
the radio surface brightness goes down with @xmath22 , because the magnetic field inside the snr is more and more along the line of sight ( see also equation [ flux ] ) .
radio observations of other galaxies show twisted magnetic spurs emerging from star forming regions ( e.g. review by @xcite ) .
da530 is located above an area of the milky way , which is rich in star forming regions , hii regions , and snrs .
is da530 expanding inside these twisted magnetic spurs ?
in terms of the large - scale galactic magnetic field , the community generally agrees that the field is directed clockwise in our local arm , as viewed from the north galactic pole , and that there is at least one large - scale magnetic field reversal ( a region of magnetic shear where the field is seen to reverse directions by roughly @xmath30 ) in the inner galaxy between the local and sagittarius - carina arms ( @xcite,@xcite ) , as indicated by the filled arrows in quadrant 1 ( q1 ) in fig .
in addition to the existence and location of magnetic field reversals is the question of field alignment with the optical spiral arms .
it is often assumed that the field is closely aligned with the spiral arms .
@xcite demonstrated conclusively that the field in the fourth quadrant ( q4 ) of the sagittarius - carina arm is directed clockwise , the same as the local field .
this presents a continuity problem if the field does indeed follow the optical arms and is counter - clockwise in q1 of the sagittarius - carina arm , as shown in fig .
[ mw ] ( left ) . however , if the field is much less inclined than the optical spiral arms , perhaps even purely azimuthal as suggested by @xcite , the continuity problem is resolved , as shown in fig .
[ mw ] ( right ) . to find additional evidence a good place to look is towards the anti - centre region of the milky way .
if the large - scale magnetic field of our galaxy is azimuthal the magnetic field component parallel to the line of sight @xmath0 would be 0 towards a galactic longitude of @xmath30 and if the magnetic field follows the spiral arms we would expect @xmath31 towards galactic longitude between @xmath32 and @xmath33 .
the snr g182.4 + 4.3 ( fig .
[ g182 ] ) is conveniently located close to the anti - centre at a galactic longitude of @xmath34 .
unfortunately only one shell of this snr is visible , likely because the bottom shell is expanding towards the galactic plane into higher density and the top shell away from the plane into very low density .
however , without emission from the centre of the snr we can not easily determine the foreground @xmath2 . on the bottom shell we find a rm gradient which goes from about 0 in the right to about 50rad / m@xmath4 on the left ( fig .
[ g182 ] ) .
this implies that the ambient magnetic field is pointing towards us from back left to front right .
this is not surprising , since the magnetic field in the outer galaxy should be directed from higher to lower longitudes and we are here at a longitude which is higher than the @xmath31 point for both possible magnetic field configurations . neglecting the foreground contribution for the rm we can determine an upper limit for @xmath22 by simulating a snr for the rm map of g182.4 + 4.3 .
the result is independent of the distance to g182.4 + 4.3 .
we derive @xmath35 , hence the galactic longitude at which @xmath31 must be larger than @xmath36 . if we now assume that the foreground magnetic field has the same @xmath22 as the snr s ambient field and the foreground field is constant at @xmath23 g , which would be a reasonable assumption , we would find @xmath37 .
this supports the assumption that the milky way has an azimuthal magnetic field structure .
the dominion radio astrophysical observatory is a national facility operated by the national research council canada .
the canadian galactic plane survey is a canadian project with international partners , and is supported by the natural sciences and engineering research council ( nserc ) .
this research is based on observations with the 100-m telescope of the mpifr at effelsberg .
|
as supernova remnants expand , their shock waves are freezing in and compressing the magnetic field lines they encounter ; consequently we can use supernova remnants as magnifying glasses for their ambient magnetic fields .
we will describe a simple model to determine emission , polarization , and rotation measure characteristics of adiabatically expanding supernova remnants and how we can exploit this model to gain information about the large scale magnetic field in our galaxy .
we will give two examples : the snr da530 , which is located high above the galactic plane , reveals information about the magnetic field in the halo of our galaxy .
the snr g182.4 + 4.3 is located close to the anti - centre of our galaxy and reveals the most probable direction where the large - scale magnetic field is perpendicular to the line of sight .
this may help to decide on the large - scale magnetic field configuration of our galaxy .
|
in recent years , authors such as @xcite , hereafter b07 ; @xcite ; @xcite ; @xcite ; @xcite ; and others , have made use of archival data ( observations of calibration fields , or in some cases , entire telescope archives ) to search for radio transients .
other studies have targeted particular fields with observations dedicated to the search for transients ( * ? ? ? * and references therein ) .
such studies are preparing the groundwork for a new generation of high - throughput radio telescopes : both dedicated new observatories for example , the allen telescope array @xcite , lofar @xcite , and askap @xcite and upgrades to older instruments for example , apertif @xcite and evla @xcite .
a major goal for these new facilities is to survey large areas of sky at high sensitivity , both to build up deep , wide - field images , and to search for transient and variable sources from a range of progenitors @xcite .
currently , large areas of rate versus sensitivity parameter space remain to be explored @xcite , and the radio transient population is not well understood . in observations where radio transients are detected , sometimes counterparts are seen in existing surveys at other wavelengths .
for example , transients may coincide with the nuclear regions of galaxies , in which case ( depending on the galaxy redshift and hence the inferred radio luminosity of the transients ) they are most likely due to active galactic nuclei ( agns ) , or perhaps supernovae ( sne ) or gamma ray bursts ( grbs ) . or they may be offset from the peak of the optical emission , as is the case for two of the transients seen by b07 , leading to their interpretation as sne or grbs . for some radio transients ,
however , no counterparts are seen in archival images at other wavelengths , and no follow - up observations taken close in time to when the transients were seen are available .
follow - up observations will become more common as radio observatories move increasingly into a regime of issuing transient alerts for rapid follow - up by other facilities , but the field is still in its infancy compared to optical transient facilities which are now issuing transient alerts as a matter of routine ( e.g. , * ? ? ?
* ) . however , the lack of galaxy counterparts in deep optical images can help exclude some models for these sources ( such as agns , sne and grbs , unless these sources are at very high redshift or extremely obscured ) , and the lack of stellar counterparts can also constrain models with galactic progenitors for these events .
b07 consider radio sne , grbs , late - type stars , soft gamma - ray repeaters , x - ray binaries , pulsars , microlensing , and reflected solar flares as possible progenitors .
they conclude that late - type stars are the most likely progenitors for transients with mjy radio fluxes and no optical counterparts .
@xcite , hereafter o10 , suggest old isolated galactic neutron stars as a possible progenitor for the sources without optical identifications , and @xcite suggest that the sources with optical counterparts may be due to neutron star
neutron star or neutron star black hole mergers .
the latter are of particular interest because they are expected to also be producers of gravitational waves .
x - ray observations can be a powerful discriminant among progenitor models , but information on the x - ray properties of the transients and transient hosts is limited . with this in mind , we obtained _ chandra _ observations of the b07 field .
we assume an @xmath7 cosmology @xcite .
magnitudes are given in the vega system unless stated otherwise .
b07 analyzed archival data at 5 and 8.4ghz of a very large array ( vla ) calibration field centered at @xmath8 , @xmath9 .
the data spanned @xmath10yr with 944 epochs , approximately one per week , each with an integration time of @xmath10min , all on the same field .
the images had a full - width at half - maximum ( fwhm ) of 86 and 51 at 5 and 8.4ghz respectively , and typical root - mean - square ( rms ) fluctuations at the center of the images were @xmath11@xmath12jy per epoch .
b07 compared the images from epoch to epoch , and detected 7 single - epoch transient sources at 5ghz , and 1 at 8.4ghz .
the transients are listed in table [ tab : radio ] .
b07 denoted these sources with identifiers beginning `` rt '' and the epoch in which they were observed ; for convenience and clarity here we denote the 5ghz and 8ghz transients with identifiers beginning `` 5 t '' and `` 8 t '' respectively , although their formal identifiers are those beginning `` rt '' with an epoch of observation as cross - referenced in table [ tab : radio ] .
for each of these transients , there was no simultaneous coverage at the other of the two observing frequencies , and no detection of the sources in longer - timescale average images .
lllllllll 5s1 & @xmath13 @xmath14 0.17 & @xmath15 @xmath14 0.2 & x29 & @xmath16 & @xmath17 & @xmath18 & @xmath19 & galaxy group + 5s2 & @xmath20 @xmath14 0.17 & @xmath21 @xmath14 0.5 & nc & & & & & none + 5s3/8s1 & @xmath22 @xmath14 0.17 & @xmath23 @xmath14 0.2 & nc & & & & & galaxy + 5s4 & @xmath24 @xmath14 0.17 & @xmath25 @xmath14 0.2 & & @xmath26 & @xmath27 & & & galaxy + 5s5/8s2 & @xmath28 @xmath14 0.17 & @xmath29 @xmath14 0.2 & x58 & @xmath30 & @xmath31 & @xmath32 & @xmath33 & none + 5s6/8s3 & @xmath34 @xmath14 0.17 & @xmath35 @xmath14 0.2 & & @xmath36 & @xmath37 & & & none + 5s7/8s4 & @xmath38 @xmath14 0.23 & @xmath39 @xmath14 0.8 & & @xmath40 & @xmath41 & & & none + 5s8 & @xmath42 @xmath14 0.49 & @xmath43 @xmath14 1.7 & nc & & & & & galaxy ? + 5t1 ( rt19840502 ) & @xmath44 @xmath14 0.35 & @xmath45 @xmath14 0.8 & & @xmath46 & @xmath47 & & & none + 5t2 ( rt19840613 ) & @xmath48 @xmath14 0.35 & @xmath49 @xmath14 0.6 & & @xmath50 & @xmath51 & & & @xmath52 galaxy + 5t3 ( rt19860115 ) & @xmath53 @xmath14 0.73 & @xmath54 @xmath14 2.6 & & @xmath55 & @xmath56 & & & none + 5t4 ( rt19860122 ) & @xmath57 @xmath14 0.56 & @xmath58 @xmath14 2.3 & & @xmath46 & @xmath47 & & & none + 5t5 ( rt19920826 ) & @xmath59 @xmath14 0.58 & @xmath60 @xmath14 4.0 & & @xmath61 & @xmath62 & & & galaxy ? + 5t6 ( rt19970528 ) & @xmath63 @xmath14 0.17 & @xmath64 @xmath14 0.3 & & @xmath65 & @xmath66 & & & @xmath67 galaxy ? + 5t7 ( rt19990504 ) & @xmath68 @xmath14 0.92 & @xmath69 @xmath14 1.2 & x19 & @xmath70 & @xmath71 & @xmath72 & @xmath73 & none + 8t1 ( rt19970205 ) & @xmath74 @xmath14 0.17 & @xmath75 @xmath14 0.2 & & @xmath76 & @xmath77 & & & none + 5l1 ( rt19870422 ) & @xmath78 @xmath14 0.58 & @xmath79 @xmath14 1.6 & & @xmath80 & @xmath81 & & & @xmath67 galaxy + 8l1 ( rt20010331 ) & @xmath82 @xmath14 0.17 & @xmath83 @xmath14 0.2 & x124 & @xmath84 & @xmath85 & @xmath86 & @xmath87 & none + b07 also constructed a deep image of the field using all data at each frequency .
the resulting images at 5 and 8.4ghz had an rms of 2.6 and 2.8@xmath12jy , respectively .
sources which were seen in many of the individual epochs and also in the deep field image were termed `` steady '' by b07 .
they saw 8 steady sources at 5ghz , shown in table [ tab : radio ] with identifiers beginning `` 5s '' , and 4 of these correspond to steady sources in the deep image at 8.4ghz , shown with labels beginning `` 8s '' in table [ tab : radio ] .
b07 also compared images made by making images from two months worth of data at a time , and found two additional `` long - timescale '' transients , one at each frequency , shown with identifiers beginning `` 5l '' and `` 8l '' in table [ tab : radio ] . optical and infra - red ( oir ) observations of the field were also reported by b07 , both with lris on keck i , to limiting magnitudes of @xmath88mag and @xmath89mag , and with pairitel to @xmath90mag , @xmath91mag and @xmath92mag .
of the two long - duration transients , one has a match with a @xmath67 galaxy .
of the eight short - duration transients , one has a likely match ( with a @xmath52 galaxy ) .
o10 also present observations in @xmath93-band using the hale 5.08-m telescope to a limiting magnitude of @xmath93 ranging from @xmath94 to @xmath95 for those sources undetected in oir by b07 .
they found no additional identifications .
two transients are referred to by b07 as `` possible '' matches , one with a galaxy at @xmath67 , and the other with a galaxy with unknown redshift ( see table [ tab : radio ] ) , but the images of o10 suggest that both of these associations are unlikely .
we consider these two , like the remaining six b07 radio transients , to be unidentified in all the oir images from b07 and o10 . as discussed above , b07
suggest a range of possible progenitors for these transients , and o10 expand on this discussion .
o10 also suggest that the transients seen at the nasu pulsar observatory by @xcite and other papers from the same group may be members of the same class .
recently , @xcite , @xcite , and @xcite suggested that the nasu transients are probably not astronomical events , and that their quoted rate is at odds with other measurements , although the parameters of the nasu survey are not clear from the literature ( see * ? ? ?
however , the b07 transient rate is consistent with measurements from other surveys ( see , for example , @xcite , fig .
12 ) , and two of the b07 transients have clear oir counterparts .
o10 show that for most classes of object , in particular extragalactic events including supernovae and gamma ray bursts , one would expect a large fraction of the transients to have associated host galaxies .
they argue that the most attractive explanation is that the transients are associated with isolated old galactic neutron stars . in order to test this hypothesis , and the flare star scenario discussed by b07
, we obtained x - ray observations of the field .
x - ray observations were obtained with _
chandra _ during cycle 11 , using the acis - i instrument with no grating or filter .
the exposure time was 49,441s and the observation start date was ut 2010 july 3 01:14:32 .
the observation was centered at @xmath96 , @xmath97 so as to encompass the majority of the vla field . for our analysis
, we started with the `` level 2 '' data products , and these were processed at the _ chandra _ x - ray center with pipeline ( `` ascds '' ) version 8.3 .
we performed further processing with the _ chandra _ interactive analysis of observations ( ciao ) version 4.3 software and version 4.4.1 of the calibration data base ( caldb ) . for the work described below , we used an image from the four acis - i ccds , including photons in the @xmath98kev bandpass .
a catalog was created from the full - band image , by running the ciao source detection algorithm `` wavdetect '' .
we used 4 different pixel binnings ( 1 , 2 , 4 , and 8) in order to detect sources that are either truly extended or sources that appear extended because the psf broadens off - axis .
we combined the catalogs from these 4 runs .
a total of 128 sources were detected , and their properties are shown in table [ tab : xray ] . in each run , the threshold was set such that 1 spurious source would be expected , so one would expect up to 4 of the sources in table [ tab : xray ] to be spurious . in most cases ,
the parameters shown in table [ tab : xray ] correspond to the lowest pixel binning [ sec : binning]in which a source was detected . in some cases ( such as the extended source x29 , or off - axis sources )
, the lowest binning clearly did not enclose all of the counts and a higher binning was used .
the binning corresponding to the reported source parameters is shown in table [ tab : xray ] .
lllllllll x1 & @xmath99 & @xmath100 & 1.46 & 4 & @xmath101 & @xmath102 & + x2 & @xmath103 & @xmath104 & 1.63 & 4 & @xmath105 & @xmath106 & + x3 & @xmath107 & @xmath108 & 1.25 & 2 & @xmath109 & @xmath110 & + x4 & @xmath111 & @xmath112 & 1.29 & 2 & @xmath113 & @xmath114 & + x5 & @xmath115 & @xmath116 & 0.95 & 1 & @xmath117 & @xmath118 & + x6 & @xmath119 & @xmath120 & 1.98 & 4 & @xmath121 & & + x7 & @xmath122 & @xmath123 & 0.86 & 1 & @xmath124 & @xmath125 & + x8 & @xmath126 & @xmath127 & 1.22 & 3 & @xmath128 & @xmath129 & + x9 & @xmath130 & @xmath131 & 1.11 & 3 & @xmath132 & @xmath133 & + x10 & @xmath134 & @xmath135 & 0.85 & 1 & @xmath136 & @xmath137 & + x11 & @xmath138 & @xmath139 & 1.02 & 2 & @xmath140 & @xmath141 & + x12 & @xmath142 & @xmath143 & 1.40 & 3 & @xmath144 & @xmath145 & + x13 & @xmath146 & @xmath147 & 1.37 & 3 & @xmath148 & @xmath149 & + x14 & @xmath150 & @xmath151 & 0.93 & 1 & @xmath152 & @xmath153 & + x15 & @xmath154 & @xmath155 & 1.28 & 3 & @xmath156 & @xmath157 & + x16 & @xmath158 & @xmath159 & 1.69 & 4 & @xmath160 & & + x17 & @xmath161 & @xmath162 & 0.93 & 2 & @xmath163 & @xmath164 & + x18 & @xmath165 & @xmath166 & 1.01 & 2 & @xmath167 & @xmath168 & + x19 & @xmath169 & @xmath170 & 0.83 & 2 & @xmath70 & @xmath72 & 5t7 + x20 & @xmath171 & @xmath172 & 1.43 & 3 & @xmath173 & @xmath174 & + x21 & @xmath175 & @xmath176 & 1.16 & 3 & @xmath177 & @xmath178 & + x22 & @xmath179 & @xmath180 & 0.72 & 1 & @xmath181 & @xmath182 & + x23 & @xmath183 & @xmath184 & 0.83 & 1 & @xmath185 & @xmath186 & + x24 & @xmath187 & @xmath188 & 0.81 & 1 & @xmath189 & @xmath141 & + x25 & @xmath190 & @xmath191 & 1.02 & 2 & @xmath192 & @xmath193 & + x26 & @xmath194 & @xmath195 & 0.70 & 1 & @xmath196 & @xmath197 & + x27 & @xmath198 & @xmath199 & 0.74 & 1 & @xmath200 & @xmath201 & + x28 & @xmath202 & @xmath203 & 0.69 & 1 & @xmath204 & @xmath205 & + x29 & @xmath206 & @xmath207 & 1.16 & 4 & @xmath208 & @xmath18 & 5s1 + x30 & @xmath209 & @xmath210 & 0.82 & 1 & @xmath211 & @xmath212 & + x31 & @xmath213 & @xmath214 & 1.11 & 1 & @xmath215 & @xmath216 & + x32 & @xmath217 & @xmath218 & 1.04 & 2 & @xmath219 & @xmath220 & + x33 & @xmath221 & @xmath222 & 1.38 & 2 & @xmath223 & @xmath224 & + x34 & @xmath225 & @xmath226 & 1.00 & 2 & @xmath227 & @xmath228 & + x35 & @xmath229 & @xmath230 & 0.71 & 1 & @xmath231 & @xmath232 & + x36 & @xmath233 & @xmath234 & 0.77 & 1 & @xmath235 & @xmath236 & + x37 & @xmath237 & @xmath238 & 0.73 & 1 & @xmath239 & @xmath240 & + x38 & @xmath241 & @xmath242 & 1.43 & 3 & @xmath243 & @xmath244 & + x39 & @xmath245 & @xmath246 & 0.76 & 1 & @xmath247 & @xmath248 & + x40 & @xmath249 & @xmath250 & 0.90 & 2 & @xmath251 & @xmath252 & + x41 & @xmath253 & @xmath254 & 0.76 & 1 & @xmath255 & @xmath256 & + x42 & @xmath257 & @xmath258 & 0.66 & 1 & @xmath259 & @xmath260 & + x43 & @xmath261 & @xmath262 & 1.06 & 2 & @xmath263 & @xmath264 & + x44 & @xmath265 & @xmath266 & 1.32 & 4 & @xmath267 & @xmath268 & + x45 & @xmath269 & @xmath270 & 0.71 & 1 & @xmath271 & & + x46 & @xmath272 & @xmath273 & 0.86 & 2 & @xmath274 & @xmath275 & + x47 & @xmath276 & @xmath277 & 1.12 & 3 & @xmath278 & @xmath279 & + x48 & @xmath280 & @xmath281 & 0.85 & 3 & @xmath282 & @xmath283 & + x49 & @xmath284 & @xmath285 & 0.97 & 2 & @xmath286 & & + x50 & @xmath287 & @xmath288 & 1.31 & 3 & @xmath289 & @xmath290 & + x51 & @xmath291 & @xmath292 & 0.72 & 1 & @xmath293 & @xmath294 & + x52 & @xmath295 & @xmath296 & 0.67 & 1 & @xmath297 & @xmath125 & + x53 & @xmath298 & @xmath299 & 1.55 & 4 & @xmath300 & @xmath301 & + x54 & @xmath302 & @xmath303 & 0.72 & 1 & @xmath304 & @xmath305 & + x55 & @xmath306 & @xmath307 & 0.76 & 1 & @xmath308 & @xmath309 & + x56 & @xmath310 & @xmath311 & 0.81 & 1 & @xmath312 & @xmath313 & + x57 & @xmath314 & @xmath315 & 0.80 & 1 & @xmath316 & @xmath317 & + x58 & @xmath28 & @xmath318 & 0.68 & 1 & @xmath30 & @xmath32 & 5s5/8s2 + x59 & @xmath319 & @xmath320 & 0.91 & 2 & @xmath321 & @xmath322 & + x60 & @xmath323 & @xmath324 & 0.73 & 2 & @xmath325 & @xmath326 & + x61 & @xmath327 & @xmath328 & 0.90 & 1 & @xmath329 & @xmath330 & + x62 & @xmath331 & @xmath332 & 1.37 & 3 & @xmath333 & @xmath334 & + x63 & @xmath335 & @xmath336 & 0.72 & 1 & @xmath337 & @xmath338 & + x64 & @xmath335 & @xmath339 & 0.73 & 1 & @xmath340 & @xmath341 & + x65 & @xmath342 & @xmath343 & 0.78 & 2 & @xmath344 & @xmath345 & + x66 & @xmath346 & @xmath347 & 0.75 & 1 & @xmath348 & @xmath349 & + x67 & @xmath350 & @xmath351 & 1.00 & 2 & @xmath352 & @xmath353 & + x68 & @xmath354 & @xmath355 & 1.04 & 3 & @xmath356 & @xmath357 & + x69 & @xmath358 & @xmath359 & 0.76 & 1 & @xmath360 & & + x70 & @xmath361 & @xmath362 & 1.28 & 3 & @xmath363 & @xmath364 & + x71 & @xmath365 & @xmath366 & 0.64 & 1 & @xmath367 & @xmath368 & + x72 & @xmath369 & @xmath370 & 0.77 & 1 & @xmath371 & @xmath372 & + x73 & @xmath373 & @xmath374 & 0.73 & 1 & @xmath375 & @xmath376 & + x74 & @xmath377 & @xmath378 & 1.04 & 2 & @xmath379 & & + x75 & @xmath380 & @xmath381 & 1.64 & 4 & @xmath382 & @xmath383 & + x76 & @xmath384 & @xmath385 & 1.06 & 1 & @xmath386 & @xmath387 & + x77 & @xmath388 & @xmath389 & 1.83 & 4 & @xmath390 & @xmath256 & + x78 & @xmath391 & @xmath392 & 1.72 & 4 & @xmath393 & @xmath394 & + x79 & @xmath395 & @xmath396 & 0.68 & 1 & @xmath397 & & + x80 & @xmath398 & @xmath399 & 0.66 & 1 & @xmath400 & @xmath401 & + x81 & @xmath402 & @xmath403 & 0.78 & 2 & @xmath404 & & + x82 & @xmath405 & @xmath406 & 2.28 & 4 & @xmath407 & @xmath408 & + x83 & @xmath409 & @xmath410 & 0.66 & 1 & @xmath411 & @xmath412 & + x84 & @xmath413 & @xmath414 & 0.68 & 1 & @xmath415 & @xmath416 & + x85 & @xmath417 & @xmath418 & 0.72 & 1 & @xmath419 & @xmath420 & + x86 & @xmath421 & @xmath422 & 1.37 & 3 & @xmath423 & & + x87 & @xmath424 & @xmath425 & 0.68 & 1 & @xmath426 & @xmath256 & + x88 & @xmath427 & @xmath428 & 0.73 & 1 & @xmath429 & @xmath430 & + x89 & @xmath431 & @xmath432 & 0.89 & 1 & @xmath433 & @xmath434 & + x90 & @xmath435 & @xmath436 & 0.72 & 1 & @xmath437 & @xmath438 & + x91 & @xmath439 & @xmath440 & 1.53 & 4 & @xmath441 & @xmath442 & + x92 & @xmath443 & @xmath444 & 0.72 & 1 & @xmath445 & @xmath446 & + x93 & @xmath447 & @xmath448 & 1.34 & 2 & @xmath449 & @xmath279 & + x94 & @xmath450 & @xmath451 & 0.92 & 2 & @xmath452 & @xmath453 & + x95 & @xmath454 & @xmath455 & 0.89 & 1 & @xmath456 & @xmath457 & + x96 & @xmath458 & @xmath459 & 1.04 & 2 & @xmath460 & @xmath461 & + x97 & @xmath462 & @xmath463 & 0.67 & 1 & @xmath464 & & + x98 & @xmath465 & @xmath466 & 0.96 & 1 & @xmath467 & @xmath468 & + x99 & @xmath469 & @xmath470 & 1.02 & 3 & @xmath471 & @xmath472 & + x100 & @xmath473 & @xmath474 & 0.68 & 1 & @xmath426 & @xmath125 & + x101 & @xmath475 & @xmath476 & 0.69 & 1 & @xmath477 & @xmath478 & + x102 & @xmath479 & @xmath480 & 1.02 & 2 & @xmath481 & @xmath482 & + x103 & @xmath483 & @xmath484 & 1.58 & 4 & @xmath485 & @xmath486 & + x104 & @xmath487 & @xmath488 & 0.70 & 1 & @xmath489 & @xmath490 & + x105 & @xmath491 & @xmath492 & 0.81 & 1 & @xmath493 & @xmath494 & + x106 & @xmath495 & @xmath496 & 0.96 & 2 & @xmath497 & @xmath498 & + x107 & @xmath499 & @xmath500 & 1.25 & 3 & @xmath501 & @xmath502 & + x108 & @xmath503 & @xmath504 & 0.70 & 1 & @xmath505 & @xmath506 & + x109 & @xmath507 & @xmath508 & 0.76 & 1 & @xmath509 & @xmath510 & + x110 & @xmath511 & @xmath512 & 1.55 & 4 & @xmath513 & @xmath514 & + x111 & @xmath515 & @xmath516 & 0.71 & 1 & @xmath517 & @xmath518 & + x112 & @xmath519 & @xmath520 & 1.88 & 4 & @xmath521 & @xmath522 & + x113 & @xmath523 & @xmath524 & 0.74 & 1 & @xmath525 & @xmath526 & + x114 & @xmath527 & @xmath528 & 1.22 & 2 & @xmath529 & @xmath530 & + x115 & @xmath531 & @xmath532 & 1.44 & 3 & @xmath533 & @xmath534 & + x116 & @xmath535 & @xmath536 & 1.86 & 4 & @xmath537 & @xmath538 & + x117 & @xmath539 & @xmath540 & 0.78 & 1 & @xmath541 & @xmath542 & + x118 & @xmath543 & @xmath544 & 0.68 & 1 & @xmath545 & @xmath546 & + x119 & @xmath547 & @xmath548 & 1.34 & 4 & @xmath549 & @xmath550 & + x120 & @xmath551 & @xmath552 & 0.75 & 1 & @xmath553 & @xmath554 & + x121 & @xmath555 & @xmath556 & 0.81 & 1 & @xmath557 & @xmath526 & + x122 & @xmath558 & @xmath559 & 0.95 & 2 & @xmath386 & @xmath86 & + x123 & @xmath560 & @xmath561 & 0.72 & 1 & @xmath562 & @xmath563 & + x124 & @xmath564 & @xmath565 & 0.95 & 2 & @xmath84 & @xmath86 & 8l1 + x125 & @xmath566 & @xmath567 & 1.50 & 4 & @xmath568 & @xmath486 & + x126 & @xmath569 & @xmath570 & 1.96 & 4 & @xmath571 & & + x127 & @xmath572 & @xmath573 & 1.53 & 4 & @xmath574 & @xmath575 & + x128 & @xmath576 & @xmath577 & 1.53 & 4 & @xmath578 & @xmath579 & + event filtering was performed using @xmath580 pixel islands .
this approach , compared to the use of @xmath581 pixel islands , may result in a slightly larger fraction of cosmic rays being misidentified as real sources , but also has the benefit of being less likely to reject real astronomical sources.[sec : islands ] hardness ratios were computed as @xmath582 where @xmath583 is the counts ( after background subtraction ) in the @xmath584kev band and @xmath585 the counts in @xmath586kev .
we tabulate @xmath587 in table [ tab : xray ] . in some cases the formal errors on @xmath587 resulted in values outside the range @xmath588
; here we truncate the values at these limits and report upper or lower limits on @xmath587 as appropriate . in fig .
[ fig : hr ] , we plot @xmath587 as a function of number of counts for the sources in table [ tab : xray ] . hardness ratio as a function of number of @xmath98kev counts for the sources in table [ tab : xray ] .
the four cases where there are possible matches to our vla sources ( table [ tab : radio ] ) are marked . ] in fig .
[ fig : optxr ] , we show a multi - wavelength view of the field , created by combining the @xmath589- , @xmath590- , @xmath591- , @xmath592- and @xmath93-band data from b07 , and overlaying the b07 radio catalog ( table [ tab : radio ] ) , and the x - ray catalog from table [ tab : xray ] . optical / ir image of the field with radio and x - ray positions overlaid .
the color image was made from the data from b07 : keck @xmath589- and @xmath590-band images ( in the blue and green channels respectively ) and the mean of the pairitel @xmath591- , @xmath592- , and @xmath93-band images ( in the red channel ) .
green circles are the radio sources from b07 , as labelled in table [ tab : radio ] , and white circles are the _ chandra _ sources reported here , as labelled in table [ tab : xray ] .
the sizes of the white ellipses correspond to the size of the _ chandra _ sources as determined by the wavdetect software , except for cases where the size was 5 or smaller , in which case we show a circle with a radius of 5 for clarity .
the green circles are plotted with a radius of 5 for clarity ; typical radio position uncertainties are @xmath593 ( table [ tab : radio ] ) .
the yellow dashed box shows the edge of the _ chandra _ field . ]
we searched around the radio positions ( table [ tab : radio ] ) for matches in the x - ray catalog ( table [ tab : xray ] ) . in order to determine the maximum radius where a match is believable , we ran monte carlo simulations .
the areal density of _ chandra _ sources in our image is @xmath594 .
we ran 1000 iterations of a routine where we created mock x - ray catalogs with the same density ( over an area twice as large as our vla field ) , and then searched for matches in this simulated catalog around the vla positions .
we calculated @xmath595 , the fraction of vla sources with a match in the catalog closer than radius @xmath596 , for both the mock and real _ chandra _ catalogs . at small @xmath596
, we expect few matches in the mock catalog , but as @xmath596 increases , the fraction of sources with matches also increases due to chance associations with neighboring sources . in fig .
[ fig : matchfrac ] , we plot @xmath595 for both the mock and real catalogs . for the mock catalog ,
the match fraction is very closely approximated by the expected analytic form , @xmath597 where @xmath598arcsec@xmath599 is the areal density of _ chandra _ sources . for a match at a given radius @xmath596 , @xmath600 gives an estimate of the probability that the match is due to chance .
this assumes a uniform surface density of x - ray sources ; if sources in the mock catalogs were clustered ( as they are in the real catalog ) , @xmath600 would tend to be lower over areas of the x - ray image away from concentrations of x - ray sources , leading to a tendency to overestimate the false match probability . proper motion is also not accounted for ; we discuss this further in section [ sec : proper1 ] .
_ left : _ fraction of sources , @xmath595 , with a match at radius @xmath601 , shown as a function of @xmath596 , where we matched the 18 sources in the b07 vla catalog with our _ chandra _ catalog ( dashed line ) and with 1000 monte carlo iterations of a mock catalog with the same areal density as our _ chandra _ catalog ( solid line ) .
the curve from the monte carlo simulations is a very good match to the expected analytic form , @xmath602 , where @xmath598arcsec@xmath599 is the areal density of _ chandra _ sources .
the real _ chandra _ catalog shows an excess above random of matches to the vla sources , showing that at least some of the radio sources are really associated with x - ray sources .
_ right : _ the same plot , but removing the 8 steady vla sources and plotting the curves only for the 10 transient sources . an excess above random
is still seen .
, title="fig : " ] _ left : _ fraction of sources , @xmath595 , with a match at radius @xmath601 , shown as a function of @xmath596 , where we matched the 18 sources in the b07 vla catalog with our _ chandra _ catalog ( dashed line ) and with 1000 monte carlo iterations of a mock catalog with the same areal density as our _ chandra _ catalog ( solid line ) .
the curve from the monte carlo simulations is a very good match to the expected analytic form , @xmath602 , where @xmath598arcsec@xmath599 is the areal density of _ chandra _ sources .
the real _ chandra _ catalog shows an excess above random of matches to the vla sources , showing that at least some of the radio sources are really associated with x - ray sources .
_ right : _ the same plot , but removing the 8 steady vla sources and plotting the curves only for the 10 transient sources . an excess above random
is still seen .
, title="fig : " ] for @xmath603 , @xmath604 , i.e. , 5% of vla sources have a match in the mock catalogs within this radius .
if sources in the x - ray catalog are physically associated with sources in the radio catalog , we expect an excess of matches at small @xmath596 over the expectation from the monte carlo simulations , and indeed this is seen when we compare @xmath600 and @xmath605 in fig .
[ fig : matchfrac ] .
the 5% false match probability leads us to expect @xmath606 false match out of our 18 radio sources at a radius of 117 , but for matches that are closer than this , the probability increases that matched sources are physically associated rather than superposed by chance .
we used equation [ eqn : thetamock ] to calculate values of @xmath600 ( i.e. , the probability that the association is due to chance ) for radio sources with a match in the real _ chandra _ catalog closer than @xmath607 .
two of the steady sources , and two of the radio transients ( one short , and one long ) have possible _ chandra
_ counterparts , and we discuss each of these cases below ( sections [ sec : matchstead ] and [ sec : matchtran ] ) .
bandwidth smearing in the radio image is not a significant issue .
at worst it increases the size of unresolved sources close to the edge of the field in the 5ghz vla a - array observations of b07 by around 70% , but in most cases is significantly less than this . even in the worst case
it should not make a large difference to the statistical significance of matches between the x - ray and radio catalogs . due to the small number of matches between x - ray and radio sources , it is not possible to look for systematic offsets between the x - ray and radio astrometric frames . however , vla astrometric accuracy is typically @xmath608 or better , so systematic errors are likely dominated by the @xmath609 _ chandra _ pointing uncertainty , which is already included in the positional uncertainties reported in table [ tab : xray ] .
we show in [ sec : icrs ] that the _ chandra _ data are well - matched to the optical data .
sub - arcsecond systematic offsets would have only a small effect on the confidence levels for the matches discussed below .
steady radio source 5s1 was matched to x - ray source x29 at a distance of @xmath610 .
the 90% positional uncertainty of x29 is 116 .
the positional uncertainties reported by b07 for source 5s1 evaluate to a 90% uncertainty of 053 . adding these in quadrature
gives a total 90% positional uncertainty of @xmath611 .
source x29 is extended , and 5s1 lies within the boundaries of the _ chandra _ source ( fig .
[ fig : optxr ] ) . from equation [ eqn : thetamock ] , @xmath612 ,
so such an association has a @xmath613% probability of happening by chance .
there are 18 sources altogether ( eight steady and ten transient ) , and the binomial probability of a false association for one or more of them is 54.6% , so the formal confidence for this association is just 45.4% . however , source x29 is bright and extended , and is clearly associated with a group of galaxies in the oir image ( fig . [
fig : optxr ] ) , one of which is clearly the host of the radio source , so the sources both do appear to have a physical association with the galaxy group ( [ sec:5s1 ] ) . steady source 5s5/8s2 was matched to source x58 at a distance of @xmath614 .
the 90% positional uncertainty of x58 is @xmath615 .
the 90% positional uncertainty of 5s5 is @xmath616 . adding these in quadrature gives a total 90% positional uncertainty of @xmath617 , so the match is well within the uncertainties .
the false match probability is @xmath618% , or @xmath619% for at least one out of 18 sources .
so we can be @xmath620% confident that the association is real . in fig .
[ fig : postages ] we show x - ray images with radio contours overlaid of the two fields with x - ray counterparts to our steady sources .
images of the two fields with x - ray counterparts to our steady radio sources 5s1 ( left ) and 5s5 ( right ) .
the colors in the image represent the number of _ chandra _ @xmath98kev counts per pixel .
the image is binned in blocks of @xmath621 pixels .
the radio data are from the deep 5ghz image , with contours from 15
35@xmath12jy beam@xmath1 in steps of 5@xmath12jy beam@xmath1 for 5s1 , and from 50 200@xmath12jy beam@xmath1 in steps of 50@xmath12jy beam@xmath1 for 5s5 .
the size of the blue ellipse for source x29 corresponds to the size of the _ chandra _ source as determined by the wavdetect software .
for source x58 the fitted size was @xmath622 and the circle is shown with a radius of 5 for clarity .
, title="fig : " ] images of the two fields with x - ray counterparts to our steady radio sources 5s1 ( left ) and 5s5 ( right ) .
the colors in the image represent the number of _ chandra _
@xmath98kev counts per pixel .
the image is binned in blocks of @xmath621 pixels .
the radio data are from the deep 5ghz image , with contours from 15
35@xmath12jy beam@xmath1 in steps of 5@xmath12jy beam@xmath1 for 5s1 , and from 50 200@xmath12jy beam@xmath1 in steps of 50@xmath12jy beam@xmath1 for 5s5 .
the size of the blue ellipse for source x29 corresponds to the size of the _ chandra _ source as determined by the wavdetect software . for source x58
the fitted size was @xmath622 and the circle is shown with a radius of 5 for clarity .
, title="fig : " ] long ( duration @xmath623 months ) transient source 8l1 was matched to source x124 at a distance of @xmath624 .
the 90% position uncertainty for x124 is 095 .
the radio position uncertainties reported by b07 for source 8l1 evaluate to a 90% uncertainty of @xmath625 .
adding these in quadrature gives a total 90% positional uncertainty of 128 .
so the measured positional offset is just a little larger than the formal positional uncertainty . from equation [ eqn : thetamock ] , @xmath626 ,
so such an association has a @xmath627% probability of happening by chance .
there are ten radio transients , and the binomial probability of one or more of them being associated with a _
chandra _
source by chance is @xmath628% .
hence , we can be @xmath629% confident that the association between 8l1 and x124 is real .
we note , however , that x124 was detected with @xmath84 counts , so the x - ray source is not a very strong detection . additionally , this source is not present in a catalog filtered using @xmath581 pixel islands ( section [ sec : islands ] ) .
examination of the event list before filtering shows that four out of 21 events occurred at the same time in the same pixel .
these events are likely not due to photons from the source , but may be due to a cosmic ray or hot pixel .
the absence of x124 from the @xmath581 pixel filtered catalog means that we can not rule out that it may be a spurious source due to cosmic rays or particle background .
additional x - ray data would be useful to confirm or reject this x - ray detection . short duration transient source 5t7 was matched to source x19 at a distance of @xmath630 .
the positional uncertainty of source x19 is 083 .
the positional uncertainty of source 5t7 is 306 .
adding these in quadrature for a total uncertainty of 317 , we see that the match radius is almost twice the formal uncertainties .
however , the x - ray source is towards the edge of the x - ray field and appears somewhat extended , and the synthesized beam of the radio observations in which this transient was detected is relatively large ( fwhm @xmath631 ) ; bandwidth smearing for this source is not a significant problem , though ( broadening the radio source by around 10% in the radial direction ) .
equation [ eqn : thetamock ] shows that the false match probability is @xmath632% .
the binomial probability of one or more out of ten transients being associated with a _
chandra _
source this close is 11.0% , so we can be 89.0% confident that this association is real .
we discuss this source further in section [ sec : x19 ] . in fig .
[ fig : postage ] , we show x - ray images with radio contours overlaid of the two fields with possible x - ray counterparts to our radio transients .
images of the two fields with possible x - ray counterparts to our radio transients 8l1 ( left ) and 5t7 ( right ) .
the colors in the image represent the number of _ chandra _ @xmath98kev counts per pixel .
the image is binned in blocks of @xmath621 pixels .
the radio contours are from 15
35@xmath12jy beam@xmath1 in steps of 5@xmath12jy beam@xmath1 for the two - month average vla b - array data for 8l1 , and from 150 250@xmath12jy beam@xmath1 in steps of 50@xmath12jy beam@xmath1 for the single - epoch lower resolution vla d - array data for 5t7 .
the size of the blue ellipse for source x19 corresponds to the size of the _ chandra _ source as determined by the wavdetect software .
for source x124 the fitted size was @xmath622 and the circle is shown with a radius of 5 for clarity .
, title="fig : " ] images of the two fields with possible x - ray counterparts to our radio transients 8l1 ( left ) and 5t7 ( right ) .
the colors in the image represent the number of _ chandra _
@xmath98kev counts per pixel .
the image is binned in blocks of @xmath621 pixels .
the radio contours are from 15
35@xmath12jy beam@xmath1 in steps of 5@xmath12jy beam@xmath1 for the two - month average vla b - array data for 8l1 , and from 150
250@xmath12jy beam@xmath1 in steps of 50@xmath12jy beam@xmath1 for the single - epoch lower resolution vla d - array data for 5t7 .
the size of the blue ellipse for source x19 corresponds to the size of the _ chandra _ source as determined by the wavdetect software .
for source x124 the fitted size was @xmath622 and the circle is shown with a radius of 5 for clarity .
, title="fig : " ] for the eight transients and three steady sources without a counterpart from our _ chandra _ catalog within @xmath607 ( not including the three steady sources which are outside the area of our _ chandra _ coverage ) , we measured counts from the _ chandra _ @xmath98kev image .
we used apertures with radii ranging from 5 10 ( depending on how far away from the center of the _ chandra _ image the positions were , to account for distortion of the psf ) centered at the radio positions .
the counts are reported in table [ tab : radio ] .
we also tabulate 90% poisson upper limits for the count rates following @xcite .
we generated a catalog from the @xmath590-band image of b07 , who report an astrometric uncertainty relative to the international celestial reference system ( icrs ) of 250 mas in each axis for this image.[sec : icrs ] we matched this @xmath590-band catalog to our _ chandra _ catalog , to check for systematic astrometric shifts in our _ chandra _ data .
for the 40 _
chandra _
sources with a match in the @xmath590-band catalog within a radius of 10 , we found a median offset of @xmath633 , @xmath634 .
we conclude that our _ chandra _ data are astrometrically well - matched to the icrs .
we neglect these small offsets in our analysis of the matches , because they are smaller than the other errors . in table
[ tab : radio ] , we note the oir matches as reported by b07 for the transient and long transient radio sources .
we also report where there are oir counterparts for the steady sources as seen in the lris and pairitel images .
two sources ( 5s3 and 5s4 ) have clear counterparts which appear as faint , small angular size sources in the oir images .
one source ( 5s8 ) has a probable association with an ir source which is outside the footprint of the optical coverage .
these three sources are likely agns or perhaps starburst galaxies at intermediate redshifts .
none has an x - ray counterpart .
[ sec:5s1]source 5s1 is associated with a galaxy in a group or cluster which appears to be at intermediate redshift .
the cluster is coincident with an extended x - ray source ( x29 ) which is presumably due to bremsstrahlung from the cluster halo ( [ sec : cluster ] ) .
transient source 5t6 is also quite nearby , and may be associated with a galaxy at @xmath67 , as reported by b07 , although as noted above the offset between the galaxy and the radio position is rather large to be plausible .
it is possible that the potential host galaxy is a member ( on the outskirts ) of this cluster it is 42 from the cluster radio source 5s1 , and 35 from the centroid of the x - ray halo x29 , which if the cluster is also at @xmath67 , correspond to 163kpc and 136kpc respectively , within the virial radius for even a poor cluster . for the sources with oir identifications , we searched the nasa extragalactic database for redshifts or other information about the counterparts , but no further information was found . for nearby sources such as brown dwarfs
, high proper motions may result in positional mismatches between the same source in the radio , oir , and x - ray data , which can complicate the interpretation of the data .
for the two radio transients with nearby x - ray matches , the time between the x - ray and radio observations was @xmath635yr .
from the offsets in table [ tab : radio ] , this implies that if the x - ray and radio sources were exactly coincident at the time of the radio transient ( and ignoring any systematic position errors ) , the progenitors of 8l1 and 5t7 would have proper motions ( with 90% confidence uncertainties ) of @xmath636 arcsec yr@xmath1 and @xmath637 arcsec yr@xmath1 , respectively .
source x19 was detected with 40.5 counts in our x - ray data , and we are able to fit models to the x - ray spectrum ( fig . [
fig : x19 ] ) , as opposed to simply predicting fluxes from basic models for the fainter sources . using the xspec software
, we fit the fluxes using three models : a power law , where the photon index and absorbing column were allowed to vary , and blackbody and bremsstrahlung models where the temperature and absorbing column were allowed to vary .
formal errors were determined using the @xcite statistic .
best fit parameters with 90% confidence limits , along with the reduced @xmath638 values for the best - fitting models , and associated fluxes are shown in table [ tab : x19 ] .
[ sec : specfit ] _ top : _ x - ray spectrum of source x19 .
the model shown is the best - fitting absorbed power - law with @xmath639@xmath3 , and @xmath640 ( table [ tab : x19 ] ) . _
bottom : _ the @xmath641 residuals ( i.e. residuals divided by uncertainties ) , denoted @xmath642 , for the fit to the spectrum shown in the top panel . ]
lllll power law & @xmath643 & @xmath640 & 20 & 2.8 + blackbody & @xmath644 & @xmath645kev & 12 & 5.5 + bremsstrahlung & @xmath646 & @xmath647kev & 18 & 2.9 + the x19 spectrum is about equally well fit by an absorbed power law with @xmath648 as by a bremsstrahlung model with @xmath649kev .
a single - temperature blackbody is a worse fit to the data than these two models .
there are insufficient data to perform two - component fits to the spectrum . in fig .
[ fig : x19lc ] , we show the x - ray light curve for x19 . by eye
there appears to be some variability but the error bars are large , and most of them are consistent with the weighted average of the _ chandra _ counts .
a kolmogorov - smirnov test shows a 32% chance that the source is constant , i.e. , variability is detected at only the @xmath650 level .
_ left : _ light curve for source x19 during our _ chandra _ observation .
the dashed line shows the weighted average of the _ chandra _ counts .
the dotted line shows the background count rate .
[ fig : x19pow ] _ right : _ power spectrum for source x19 , using the @xcite normalization , where a power of 2 indicates consistency with poisson noise with no additional variability from the source . , title="fig : " ] _ left : _ light curve for source x19 during our _ chandra _ observation .
the dashed line shows the weighted average of the _ chandra _ counts .
the dotted line shows the background count rate .
[ fig : x19pow ] _ right : _ power spectrum for source x19 , using the @xcite normalization , where a power of 2 indicates consistency with poisson noise with no additional variability from the source .
, title="fig : " ] the power spectrum for x19 is shown in fig .
[ fig : x19pow ] . for most frequencies ,
the variability is consistent with the poisson noise level with no additional variability from the source , although there is tentative evidence of variability at low frequencies .
additional epochs of x - ray data could help to determine if the source is truly variable on longer timescales .
however , with only a small number of counts per second , it will remain difficult to detect variability in this source on short timescales using _ chandra _ ( unless it brightens significantly in the future ) .
although 5t7 , the source apparently matched to x19 , has no counterpart in the deep oir imaging of b07 or o10 , it is offset from the x19 position by 557 . on examination of the @xmath93-band images of the field presented in o10 ( their fig . 1 , bottom right panel )
, we noticed a faint source offset from the radio position , but which appears to be consistent with the x - ray position .
the source is not present in the @xmath93-band data of b07 , so it must be fainter than the detection limit in b07 ( @xmath651 ) but brighter than the detection limit in o10 ( @xmath652 ) . from its appearance in the o10 data , we estimate @xmath653 .
the source is also not visible to @xmath654 and @xmath655 in the b07 data .
we searched the wide - field infrared survey explorer ( wise ) catalog @xcite for counterparts to x19 , and found that it was detected in the two shortest wavelength bands ( 3.4 and 4.6 , with magnitudes @xmath656 and @xmath657 respectively ) , and undetected at the two longer wavelengths ( 12 and 22 , to 12.9 and 9.1mag respectively ) .
on examination of the second palomar observatory sky survey ( poss - ii ) digitized images @xcite we also found faint sources at the same position in @xmath658- and @xmath590-band , and no detection in @xmath659-band . quoted
limiting magnitudes are @xmath660 , @xmath661 , and @xmath662 . from the poss - ii images , we estimate @xmath663 , @xmath664 , and @xmath665 .
we also searched the galaxy evolution explorer ( galex ; * ? ? ? * ) release 6 catalog for ultraviolet ( uv ) counterparts to x19 , and discovered a source with a near - uv ( 1750 2800 ) magnitude in the ab system of @xmath666 , corresponding to a vega magnitude of @xmath667 . in the far - uv ( 1350 1750 )
, there is no galex detection to ab @xmath668mag , corresponding to a vega magnitude fainter than 19.9 .
postage stamp images at the position of x19 are shown in fig .
[ fig : x19post ] .
postage stamp images of x19 . from top left to bottom right , in order of increasing wavelength : _ chandra _ ; galex far - uv and near - uv ; poss - ii @xmath669 , @xmath670 , and @xmath671 ; @xmath591 , @xmath592 , and @xmath93 from b07 ; wise channels 1 4 . overlaid on each image
are radio contours are from 150
250@xmath12jy beam@xmath1 in steps of 50@xmath12jy beam@xmath1 for the single - epoch vla d - array data for 5t7 .
, title="fig : " ] postage stamp images of x19 . from top
left to bottom right , in order of increasing wavelength : _ chandra _ ; galex far - uv and near - uv ; poss - ii @xmath669 , @xmath670 , and @xmath671 ; @xmath591 , @xmath592 , and @xmath93 from b07 ; wise channels 1 4 .
overlaid on each image are radio contours are from 150 250@xmath12jy beam@xmath1 in steps of 50@xmath12jy beam@xmath1 for the single - epoch vla d - array data for 5t7 .
, title="fig : " ] as noted in section [ sec : proper1 ] , if the offset between the positions of 5t7 ( seen at ut 1999 may 4 ) and x19 ( observed ut 2010 july 3 ) is due to movement of the source , a proper motion of @xmath672 arcsec yr@xmath1 is required . however , the position of the oir counterpart to x19 as seen in the @xmath669 , @xmath670 , and @xmath671-band poss - ii data ( taken at ut 1994 april 6 , 1994 june 09 , and 1997 june 28 , respectively ) is coincident with the wise position ( data taken in early 2010 ) , as well as the galex position ( ut 2006 november 5 ) .
therefore , proper motion can not explain the 557 offset between the 5t7 and x19 positions , which are bracketed by epochs where the uv / oir counterpart to x19 is not seen to move .
we performed spectral energy distribution ( sed ) fitting using the uv and oir data for x19 .
a @xmath638 minimization using a library of 726 galaxy and qso models , and 254 stellar models ( including 100 brown dwarf models ) , was performed using the le phare software @xcite .
the best - fitting galaxy or qso template , a qso1 from the swire template library @xcite redshifted to @xmath673 , had @xmath674 ; the best - fitting stellar model , a brown dwarf from @xcite , was a poor match to the data ( @xmath675 ) . in fig .
[ fig : x19sed ] , we show the sed of x19 along with the multiwavelength uv / oir photometry . spectral energy distribution of x19 . from left to right , in order of increasing wavelength , the data points are : galex far - uv and near - uv ; poss - ii @xmath669 , @xmath670 , and @xmath671 ; @xmath591 , @xmath592 , and @xmath93 from b07 and o10 ; wise channels 1 4 . the best - fitting qso model , a qso1 from @xcite at @xmath5 , is shown . ] the x - ray light - curve and spectrum , and the positional coincidence of the x - ray and uv / oir sources , make for a convincing argument that x19 is a qso at @xmath5 .
the offset between x19 and 5t7 must then be due to one of two possibilities : either 5t7 is not at @xmath5 , in which case x19 and its multiwavelength counterpart are unassociated with the radio transient , which remains unidentified in x - rays ; or 5t7 is at @xmath5 , in which case it appears to be offset by @xmath676kpc from the x - ray position .
we discuss this source further in section [ sec : x19prog ] .
it is important to note here that it is unlikely that all the radio sources belong to the same population .
three , or maybe four of the steady sources appear to be associated with galaxies in the oir images .
these are likely agns or starburst galaxies
. the remainder may be galactic or extragalactic , but are likely distant or highly obscured .
the transient sources with no counterparts may be very distant , highly obscured , or intrinsically faint in quiescence .
again , this is true whether they are galactic or extragalactic , although in the latter case , as argued by o10 , we would still expect to see more counterparts .
the two transient sources with clear oir counterparts , and perhaps the two with possible counterparts , may , as noted by b07 , be supernovae , although if so , they are unusually radio - luminous . for these four sources ,
no x - rays were seen in our _ chandra _ image .
this would not be particularly surprising if these sources are supernovae , particularly since the x - ray observations were taken between 11 and 27 years after the radio transients were seen .
it has also been recently suggested @xcite that the radio emission in the two cases with clear counterparts may be from neutron star
neutron star or neutron star black hole mergers . of the six transient sources with no oir identification , two have possible x - ray counterparts in our _ chandra _ image (
although one is a marginal detection and the other is offset too much for the association to be certain ) , and four have only upper limits .
the two sources with `` possible '' oir ids as noted by b07 ( which we argue are also unidentified ) also have only upper limits in our _ chandra _ data .
the count rates of the upper limits are similar to those for the detections , however , so it is conceivable that all the transients have similar count rates , of order @xmath677 counts s@xmath1 .
they may or may not be due to a single class of progenitor .
it is worth noting that the only short - duration transient with a possible x - ray counterpart ( 5t7/x19 ) is by far the brightest of the transients in the radio ( table [ tab : xflux ] ) , which could indicate that it is a member of a different class of objects to the others .
llllll 5t1 ( rt19840502 ) & & @xmath47 & @xmath678 & @xmath679 & 0.45 + 5t2 ( rt19840613 ) & & @xmath51 & @xmath680 & @xmath681 & 0.57 + 5t3 ( rt19860115 ) & & @xmath56 & @xmath682 & @xmath683 & 0.37 + 5t4 ( rt19860122 ) & & @xmath47 & @xmath678 & @xmath679 & 1.59 + 5t5 ( rt19920826 ) & & @xmath62 & @xmath684 & @xmath685 & 0.64 + 5t6 ( rt19970528 ) & & @xmath66 & @xmath686 & @xmath687 & 1.73 + 5t7 ( rt19990504 ) & x19 & @xmath688 & @xmath689 & @xmath690 & 7.04 + 8t1 ( rt19970205 ) & & @xmath77 & @xmath691 & @xmath692 & 2.23 + 5l1 ( rt19870422 ) & & @xmath81 & @xmath693 & @xmath694 & 0.51 + 8l1 ( rt20010331 ) & x124 & @xmath695 & @xmath696 & @xmath697 & 0.70 + we can achieve an improvement in sensitivity by `` stacking '' the x - ray images at the radio positions of the six sources with no oir identification and only x - ray upper limits ( 5t1 , 5t3 , 5t4 , 5t5 , 5t6 , and 8t1 ; see table [ tab : radio ] ) .
stacking is a well - established ( e.g. , * ? ? ?
* ; * ? ? ?
* ) method for combining data from individual sources in order to study the statistical properties of objects which are undetected in single images .
prior information about positions at a frequency where sources are individually detected ( in this case , the radio positions ) is used to create `` cut - out '' images centered on those positions at a second frequency ( here , from our x - ray data ) where sources may not be individually detected .
if @xmath698 images each with rms @xmath699 are combined , then the noise level of the stacked image will decrease approximately as @xmath700 .
this allows sources below the original detection threshold to be studied .
stacks can also be created at `` blank sky '' positions offset from those of the known sources , in order to determine the background noise level in the absence of any additional emission present at the source positions .
we created a stack from the mean of six x - ray images aligned at the radio positions , and eight blank sky stacks ; four at positions offset 30 in the cardinal directions , and four offset 42 in the ordinal directions from the radio positions .
in other words , we created stacked images at positions on a @xmath580 grid , with spacing 30 , centered on the radio positions .
we measured _ chandra _ counts in a 5-radius aperture at the center of each stacked image .
the stacks offset from the radio positions will be dominated by background photons , whereas the stack centered at the radio positions will contain any additional flux from the radio sources , if present ( assuming proper motions smaller than a few arcseconds over the time between the vla and _ chandra _ observations ) .
the mean and rms of the eight background values was @xmath701 counts .
there were 7.93 counts in the aperture in the stack made at the radio positions .
so there is less than @xmath650 excess of x - ray counts at the radio positions of the non x - ray detected sources .
we used the webpimms v4.2 software to convert the count rates into flux densities ( for the two _ chandra _ sources near to b07 transients ) or upper limits ( for the eight transients with no x - ray counterpart ) .
the calculations are somewhat dependent on the model used for the x - ray emission .
we chose to model two cases : a neutron star , and a flare star .
the neutron star case was modelled as a blackbody with energy @xmath702kev , since isolated neutron stars older than @xmath703y have temperatures @xmath704kev @xcite .
we caution , however , that the only isolated neutron stars which have been observed to date are relatively young , and may not be representative of older populations .
slow accretion onto magnetic neutron stars with fields of @xmath705 g can cause @xmath706% of the x - ray luminosity to be emitted from broadened cyclotron emission features in the hard x - ray ( @xmath635kev ) regime @xcite .
other authors , however ( e.g. * ? ? ?
* ) note that although the spectrum for such objects is slightly harder than thermal , it is not markedly different from a blackbody .
[ sec : flaremodel]for the flare star we used a bremsstrahlung model with @xmath707kev . both flare and post - flare spectra are often modelled in the literature using two - temperature models ( e.g. * ? ? ? * ) .
since we have no reason to believe that the progenitors were in outburst when observed with _
chandra _
( in contrast to when they were seen to flare in our vla observations ) , the post - flare model is appropriate here .
we choose a 1kev single temperature model , which is a valid approximation for both flare stars @xcite and brown dwarfs @xcite .
blackbody , raymond - smith , and thermal bremsstrahlung models give similar fluxes ( to within 30% ) in the @xmath98kev energy range .
we choose a thermal bremsstrahlung model .
more detailed spectral information could be used to constrain two - temperature fits or non - thermal components for both the neutron star and flare star cases , but since our sources are mostly undetected or barely detected , a simpler approach is warranted . in both the neutron star and flare star cases
, we assumed a galactic column density of @xmath708@xmath3 @xcite .
we show the resulting flux densities and limits in table [ tab : xflux ] .
for source x124/8l1 , which was detected with only 8.8 counts , we have insufficient data to perform spectral modeling .
however , we can use the @xmath587 constraints ( @xmath86 ) to constrain progenitor models .
we used webpimms to compute hardness ratios for blackbodies with a range of energies , and we find @xmath709kev at 90% confidence .
the best fit @xmath710kev.[sec : hr ] the x - ray luminosities of the transients ( ignoring @xmath711-corrections if they are extragalactic ) are @xmath712 where @xmath713 is the model - dependent x - ray flux or upper limit from table [ tab : xflux ] or table [ tab : x19 ] , and @xmath714 is the luminosity distance to the progenitor in kpc .
the radio luminosities of the transients are @xmath715 where @xmath716 is the transient radio flux density at 5ghz ( table [ tab : xflux ] ) , or the upper limit ( from b07 ) in quiescence ( typically a few tens of @xmath12jy ) , and @xmath714 is the luminosity distance to the progenitor in kpc .
o10 suggest that the b07 transients may be due to isolated old galactic neutron stars at distances between @xmath593 5kpc .
this implies x - ray luminosities around @xmath717
@xmath718erg s@xmath1 for x124 , and similar values as upper limits for the x - ray non - detections .
this is consistent with the typical x - ray luminosities , @xmath719erg s@xmath1 , of old neutron stars accreting from the interstellar medium ( ism ; * ? ? ?
* ; * ? ? ?
this luminosity depends on the accretion rate from the ism , which in turn depends on the ism density and the neutron star space velocity , and also on the field strength and spin rate of the neutron star .
it also depends on the details of bondi accretion in the presence of magnetic fields , which are not well understood @xcite .
however , although the measured x - ray fluxes and upper limits are consistent with a population of isolated neutron stars at distances between @xmath593 5kpc , the hardness ratio measurements for x124 ( section [ sec : hr ] ) exclude a simple one - component model with @xmath702kev , the expected energy for neutron stars . since x19 is clearly not a neutron star , and x124 is the only other source for which we can compute hardness ratios , we can not extrapolate this conclusion to the remainder of the x - ray sources .
additionally , as previously noted , accreting neutron stars may also have non - thermal components , and the spectra of old isolated neutron stars may not be identical to that of younger cooling neutron stars . also , as previously noted , x124 may in fact be a spurious detection .
@xcite note that optical luminosities of isolated neutron stars are around 5 orders of magnitude fainter in the optical than in the x - ray .
our x - ray fluxes of @xmath720erg s@xmath1 @xmath3 would correspond to @xmath721erg s@xmath1 @xmath3 optical fluxes for this model , or @xmath722 out of reach even for the planned james webb space telescope .
accretion can cause brightening by a factor of a few in the optical @xcite , but not likely enough to make our objects detectable .
another class of potential neutron star progenitors discussed by b07 and o10 are soft gamma - ray repeaters ( sgrs ) or magnetars .
these are attractive as potential progenitors because they are known to exhibit dramatic flares @xcite , and would have no oir counterparts in quiescence in our data . the x - ray emission from magnetars
almost always has a blackbody component with @xmath723kev , and sometimes also has a power - law component @xcite .
our x - ray data for x124 are not sensitive enough to constrain multi - component fits .
typical magnetar x - ray luminosities are @xmath724erg s@xmath1 , so our sources would have to be at @xmath725kpc , or underluminous if closer , to explain their non - detection or marginal detection in x - rays .
as noted by b07 and o10 , the surface density of magnetars is too low to explain all the b07 transients as being due to this class of object .
low mass stars and brown dwarfs are numerous , optically faint , and active at radio wavelengths in 5 to 30% of cases @xcite .
m - dwarfs at distances as large as 1kpc have been identified as the dominant contribution to optical transient rates @xcite . in an extreme case ,
the dme star ev lac was observed to undergo a factor of @xmath726 flare to a radio luminosity @xmath727@xmath728 , with a rise time of minutes and a decay time of hours @xcite .
flare star ( i.e. , dme star ) radio luminosities in quiescence are generally no brighter than @xmath727@xmath728 @xcite , implying even extreme flares ought to be no brighter than @xmath729@xmath728 .
the radio luminosities of our transients ( equation [ eqn : rlum ] ; table [ tab : xray ] ) are @xmath730@xmath728 .
so if the progenitors are flare stars , they must be closer than 1kpc , or else the transients must be due to unusually dramatic events . from the x - ray
radio relation @xcite , our quiescent x - ray flux limits imply quiescent radio flux densities of @xmath606@xmath12jy ( consistent with the non - detection of the transients in the deep radio image of b07 to depths of a few tens of @xmath12jy ) .
the inferred x - ray luminosities ( equation [ eqn : xlum ] ) are @xmath731 for progenitors at 1kpc , plausible for flare stars @xcite . predicting x - ray luminosities during outburst from the peak radio luminosities of the transients ( assuming the x - ray
radio relation holds during outbursts ) , we obtain x - ray luminosities of @xmath732 .
however , the non - detection of the transients to @xmath733mag by o10 implies that if they are due to m5 stars , they must be at @xmath734pc @xcite . so unless the radio flares are unusually luminous , this implies that if flare stars are the progenitors of the transients , they must be of later spectral type than m5 .
o10 suggest that brown dwarfs at typical distances of around 200pc are plausible progenitors for the b07 transients , although they note that we would expect the radio emission to be strongly circularly polarized , in contrast to the low levels of polarization observed by b07
. if they have spectral types as late as t8 , they would be too faint to be detected by o10 if at distances @xmath735pc @xcite .
our x - ray fluxes imply luminosities of @xmath736erg s@xmath1 at 30pc .
this is a typical luminosity for quiescent emission from brown dwarfs @xcite ; sources with non - detections in our x - ray data could be intrinsically less luminous , more obscured , or further away . as noted in section [ sec : proper2 ] , the offset between x19 and 5t7 can not be explained by proper motion .
there is no uv / oir counterpart to x124 , however , so it is plausible that a proper motion of @xmath636 arcsec yr@xmath1 is responsible for the @xmath737 offset between x124 and 8l1 .
we can use the 90% confidence limits for the proper motion , along with plausible ranges of tangential velocity , @xmath738 km s@xmath1 , for different classes of progenitors , to infer plausible ranges of distance , @xmath714 ( in parsecs ) to x124/8l1 .
a typical low mass star with @xmath739@xmath740 would have to be closer than @xmath741pc .
some dwarf stars have tangential velocities @xmath742@xmath740 , however @xcite , so x124/8l1 could plausibly be a high proper motion object at @xmath743pc .
this would appear to marginally rule out an m star progenitor for this source ( since , as discussed above , m stars should have been seen by o10 unless at @xmath6kpc ) , but the proper motion is consistent with a brown dwarf progenitor .
the measured proper motion is also not unreasonable for a neutron star at @xmath744pc @xcite , particularly if the source is in the high - velocity tail of the distribution .
a neutron star with tangential velocity @xmath745@xmath740 @xcite could explain the proper motion of x124/8l1 if at @xmath746kpc .
the lower limit for distances of old isolated neutron star progenitors from the oir non - detections ( @xmath6kpc ) , as noted above , implies that the tangential velocity must be @xmath747@xmath740 . for both the flare star and neutron star cases , relatively small systematic position offsets could reduce the size of the measured proper motion , bringing upper limits on the distance from transverse velocities into better agreement with lower limits from to the @xmath748-band non - detections by o10 .
if the faint x - ray detection is not in fact due to an astronomical source , we can not constrain the proper motion of the progenitor . the x - ray luminosity of x19
is @xmath749@xmath750 , typical for x - ray selected agns @xcite .
the non - detection of x19 to @xmath751mjy in the deep radio image of b07 implies that the quiescent radio luminosity of the quasar is @xmath752@xmath728 .
radio loud agns are defined as those having radio to optical luminosity ratio @xmath753 , where @xmath754 @xcite .
our estimate of the b - band flux density , @xmath755mjy , implies @xmath756 from the deep radio data , so the agn is not radio loud in quiescence . as noted above , the facts that x19 appears
somewhat extended due to off - axis broadening of the _ chandra _ psf , and that the vla d - array radio beam is rather large for 5t7 , suggest that the uncertainty in the position may be somewhat higher than the formal uncertainties .
if , in fact , 5t7 is associated with the agn , this would imply that the agn brightened by a factor of @xmath706 ( see b07 , figure 4 ) to @xmath757@xmath728 , and then faded back by a factor of @xmath706 into undetectability on a timescale of days ( and been brighter when detected as a transient by a factor @xmath758 relative to its mean luminosity on a timescale of 22 years ; see b07 , table 2 ) . during the flare , @xmath759 ,
so the agn would have changed from radio - quiet to radio - loud in this scenario .
it is more likely , however , due to the positional offset , that the radio source is unassociated with the nuclear x - ray source .
if the radio transient is also at @xmath5 , the positional offset implies it is @xmath676kpc from the nucleus . in this case , the @xmath757@xmath728 radio transient could plausibly be due to a grb afterglow in the outer regions or halo of the agn host galaxy , although it would be among the brightest yet seen @xcite .
if in fact 5t7 is a chance association with x19 , then the radio transient remains undetected in x - rays and should be considered along with the other non - detections as a possible brown dwarf or neutron star .
for source 5s1 , apparently associated with a galaxy cluster , we used webpimms to model the x - ray source x29 as a power law with @xmath760 ( optically thin bremsstrahlung ) at @xmath67 , and a galactic column density of @xmath708@xmath3 , which results in a luminosity ( extrapolated to rest frame 0.05 40kev ) of @xmath761ergs@xmath1 .
following @xcite , we infer a mass of @xmath762@xmath763 for the cluster
. the cluster could be part of a structure associated with abell2047 ( see b07 ) , which is centered 106 away , or @xmath623mpc if both are at @xmath764 , perhaps explaining the fact that several of the b07 sources seem to be at this redshift .
of the ten b07 transients , eight are undetected in x - rays ; one is associated with an x - ray source of marginal ( @xmath4 ) significance , which may be due to cosmic rays or particle background ; and one may be associated with a qso or its host galaxy at @xmath5 , or may be a spurious positional coincidence .
because of the uncertainty regarding these two associations , it is plausible that all of the b07 transients remain undetected in our x - ray images . the measured x - ray flux for the transient with a marginal x - ray detection ( x124/8l1 ) , and the upper limits for the undetected transients , are consistent with the x - ray luminosity expected for brown dwarfs at distances @xmath744pc , one of the possible progenitors discussed by o10 .
the small offset between the x124 and 8l1 positions could plausibly be due to the proper motion of such an object . if instead the transients are due to flaring m stars , most of the radio transients ought to have been detected in quiescence in both the x - ray and oir images , unless the flares were extreme events ( not out of the question , since we know the radio flux densities increased by a factor @xmath765 ) from progenitors at distances @xmath6kpc .
the x - ray fluxes are also consistent with the suggestion of o10 that the transients could be isolated neutron stars .
the radio and oir fluxes are consistent with this picture as well . if x124 is truly a counterpart to 8l1 , hardness ratio constraints appear to rule out a simple one - component spectral model , but there are insufficient data to attempt a more sophisticated fit with additional thermal or non - thermal components .
if the progenitors have x - ray fluxes slightly fainter than our sensitivity limit , deeper observations would be expected to detect more of the transients in x - rays , as well as improving the significance of the fainter detection ( or rejecting it as a real source ) .
x - ray or radio monitoring campaigns lasting for several months would have a good chance of catching these sources in outburst since for the most likely classes of progenitor , flares are likely to repeat . near- or mid - infrared data would also be useful to confirm or refute the brown dwarf hypothesis .
support for this work was provided by the national aeronautics and space administration through _ chandra _ award number go0 - 11152x issued by the _ chandra _ x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 .
this research made use of montage , funded by the national aeronautics and space administration s earth science technology office , computation technologies project , under cooperative agreement number ncc5 - 626 between nasa and the california institute of technology .
montage is maintained by the nasa / ipac infrared science archive .
arons , j. , & lea , s. m. 1976 , , 207 , 914 arons , j. , & lea , s. m. 1980 , , 235 , 1016 bannister , k. w. , murphy , t. , gaensler , b. m. , hunstead , r. w. , & chatterjee , s. 2011 , , 31 becker , a. c. , et al .
2004 , , 611 , 418 becker , r. h. , helfand , d. j. , white , r. l. , & proctor , d. d. 2010 , , 140 , 157 bell , m. e. , et al .
2011 , arxiv:1103.0511 , , in press .
berger , e. 2006 , , 648 , 629 bildsten , l. , & rutledge , r. e. 2000 , , 541 , 908 bower , g. c. , saul , d. , bloom , j. s. , bolatto , a. , filippenko , a. v. , foley , r. j. , & perley , d. 2007 , , 666 , 346 bower , g. c. , et al .
2010 , , 725 , 1792 bower , g. c. , & saul , d. 2011 , , 728 , l14 cash , w. 1979 , , 228 , 939 chabrier , g. , baraffe , i. , allard , f. , & hauschildt , p. 2000 , , 542 , 464 cordes , j. m. , & chernoff , d. f. 1998 , , 505 , 315 croft , s. , et al .
2010 , , 719 , 45 croft , s. , et al .
2011 , , 731 , 34 . de vries ,
w. h. , hodge , j. a. , becker , r. h. , white , r. l. , & helfand , d. j. 2007 , , 134 , 457 faherty , j. k. , burgasser , a. j. , cruz , k. l. , shara , m. m. , walter , f. m. , & gelino , c. r. 2009 , , 137 , 1 faucher - gigure , c .- a . , & kaspi , v. m. 2006 , , 643 , 332 frail , d. 2011 , http://fermi.gsfc.nasa.gov/science/grb2010/tue/dale_frail.pdf gaensler , b. m. , et al .
2005 , , 434 , 1104 gehrels , n. 1986 , , 303 , 336 gdel , m. 2002 , , 40 , 217 ilbert , o. , et al .
2006 , , 457 , 84 imanishi , k. , tsujimoto , m. , & koyama , k. 2001 , , 563 , 361 jarosik , n. , et al .
2011 , , 192 , 14 johnston , s. , et al .
2008 , experimental astronomy , 22 , 151 kalberla , p. m. w. , burton , w. b. , hartmann , d. , arnal , e. m. , bajaja , e. , morras , r. , pppel , w. g. l. 2005 , , 440 , 775 kellermann , k. i. , sramek , r. , schmidt , m. , shaffer , d. b. , & green , r. 1989 , , 98 , 1195 kulkarni , s. r. , & rau , a. 2006 , , 644 , l63 law , n. m. , et al .
2009 , , 121 , 1395 lazio , t. j. w. , et al .
2008 , , 672 , l115 leahy , d. a. , darbro , w. , elsner , r. f. , weisskopf , m. c. , kahn , s. , sutherland , p. g. , & grindlay , j. e. 1983 , , 266 , 160 lusso , e. , et al .
2010 , , 512 , a3 martin , d. c. , et al .
2005 , , 619 , l matsumura , n. , et al .
2009 , , 138 , 787 mereghetti , s. 2008 , , 15 , 225 nakar , e. , & piran , t. 2011 , arxiv:1102.1020 nelson , r. w. , salpeter , e. e. , & wasserman , i. 1993 , , 418 , 874 ofek , e. o. 2009 , , 121 , 814 ofek , e. o. , breslauer , b. , gal - yam , a. , frail , d.,kasliwal , m. m. , kulkarni , s. r. , & waxman , e. 2010 , , 711 , 517 ofek , e. o. , frail , d. a. , breslauer , b. , kulkarni , s. r. , chandra , p. , gal - yam , a. , kasliwal , m. m. , & gehrels , n. 2011 , arxiv:1103.3010 , apj , submitted .
osten , r. a. , hawley , s. l. , allred , j. c. , johns - krull , c. m. , & roark , c. 2005 , , 621 , 398 osten , r. a. , et al .
2010 , , 721 , 785 ostriker , j. p. , rees , m. j. , & silk , j. 1970 , , 6 , 179 patten , b. m. , et al . 2006 , , 651 , 502 perley , r. , et al .
2009 , ieee proceedings , 97 , 1448 polletta , m. , et al .
2007 , , 663 , 81 preibisch , t. , et al .
2005 , , 160 , 582 reid , i. n. , et al .
1991 , , 103 , 661 reiprich , t. h. , & bhringer , h. 2002 , , 567 , 716 rttgering , h. 2003 , , 47 , 405 rutledge , r. e. , fox , d. w. , bogosavljevic , m. , & mahabal , a. 2003 , , 598 , 458 toropina , o. d. , romanova , m. m. , toropin , y. m. , & lovelace , r. v. e. 2003 , , 593 , 472 toropina , o. d. , romanova , m. m. , toropin , y. m. , & lovelace , r. v. e. 2005 , , 76 , 508 treves , a. , turolla , r. , zane , s. , & colpi , m. 2000 , , 112 , 297 verheijen , m. a. w. , oosterloo , t. a. , van cappellen , w. a. , bakker , l. , ivashina , m. v. , & van der hulst , j. m. 2008 , the evolution of galaxies through the neutral hydrogen window , 1035 , 265 welch , j. , et al . 2009 , ieee proceedings , 97 , 1438 wright , e. l. , et al .
2010 , , 140 , 1868 yakovlev , d. g. , gnedin , o. y. , gusakov , m. e. , kaminker , a. d. , levenfish , k. p. , & potekhin , a. y. 2005 , nuclear physics a , 752 , 590 zane , s. , turolla , r. , & treves , a. 2000 , , 537 , 387 zibetti , s. , white , s. d. m. , schneider , d. p. , & brinkmann , j. 2005 , , 358 , 949
|
we present a 50ks _ chandra _ acis - i x - ray observation of the @xcite vla archival field .
the observations reach a limiting sensitivity of @xmath0 counts s@xmath1 , corresponding to a flux of a few times @xmath2erg s@xmath1 @xmath3 for the models we explore
. the _ chandra _
observations were undertaken to search for x - ray counterparts to the eight transient sources without optical counterparts , and the two transient sources with optical counterparts seen by @xcite neither of the sources with optical counterparts was detected in x - rays .
one of the eight optical non - detections is associated with a marginal ( @xmath4 ) x - ray detection in our _ chandra _ image . a second optically - undetected @xcite transient
may be associated with a @xmath5 x - ray detected quasar or its host galaxy , or alternatively is undetected in x - rays and is a chance association with the nearby x - ray source .
the x - ray flux upper limits , and the one marginal detection , are consistent with the interpretation of @xcite that the optically - undetected radio transients are flares from isolated old galactic neutron stars .
the marginal x - ray detection has a hardness ratio which implies a temperature too high for a simple one - temperature neutron star model , but plausible multi - component fits are not excluded , and in any case the marginal x - ray detection may be due to cosmic rays or particle background .
the x - ray flux upper limits are also consistent with flare star progenitors at @xmath6kpc ( which would require the radio luminosity of the transient to be unusually high for such an object ) or less extreme flares from brown dwarfs at distances of around 100pc .
|
the @xmath0 state has been one of the interesting problems of qcd both in theory [ 1,2,3 ] and in experiment [ 4 ] .
qcd predictions for production cross section of @xmath0 fails to agree with the experimental results .
color octet senario is introduced to bring about the agreement [ 5,6 ] . in this work ,
we introduce the fermi motion into the @xmath0 production .
we only include the charm fragmentation contributions .
the tevatron run i , the tevatron run ii , the rhic and the cern lhc energies are considered to demonstrate the effect .
with the choice of a light cone the wave function , we introduce the fermi motion effect in fragmentation production of the @xmath0 bound state . in the leading order
perturbative regime , the fragmentation functions for @xmath0 production without and with the fermi motion are obtained [ 7 ] . in the absence of fermi motion , where the confinement parameter is @xmath1 , the fragmentation function has the following form @xmath2\biggr\},\end{aligned}\ ] ] where @xmath3 is the strong interaction coupling constant and @xmath4 is the color factor .
the quantity @xmath5 is the average transverse momentum squared of the initial state heavy quark , the parameter @xmath6 is defined as @xmath7 and finally @xmath8 is given by @xmath9 ^ 2.\end{aligned}\ ] ] in the presence of fermi motion the fragmentation function is obtained as @xmath10 + \eta\bigl [ 2+(-6 + 4x)z\nonumber\\ & & + ( 9 - 8x+2x^2)z^2 - 2(2-x+x^2)z^3+(1+x^2)z^4\bigr]\nonumber\\ & & + \xi z^2\bigl[1 + 2x^3(2 - 3z)z+z^2 + 2x^4z^2 + 2x(-1+z-2z^2)\nonumber\\ & & + x^2(2 - 8z+9z^2)\bigr]+\eta^2(1-z)^2+\xi^2(1-x)^2x^2z^4\bigr\}.\end{aligned}\ ] ] the function @xmath11 reads as @xmath12\bigl[\eta(-1+z)+\xi ( -1+x)xz^2 - 1+(1-x+x^2)z\bigr]\bigr\}^2.\end{aligned}\ ] ] moreover here we have defined @xmath13 .
the two cases of the fermi motion off and on may be compared using the above fragmentation functions .
note that @xmath1 and @xmath14 correspond to these cases .
for @xmath15 value , we have chosen the range of @xmath160.0 - 0.6 gev .
the behavior of the fragmentation function ( 3 ) is shown in fig . 1 for @xmath17 gev .
we take the maximum value of @xmath15=0.6 gev and use it for our further considerations . production .
while the dashed curve represents the function when the fermi motion is off , the solid ones show the behavior of this function when the fermi motion is on .
the @xmath15 values are indicated .
the function picks up as @xmath15 increases .
this gives raise to increase in the fragmentation probability .
, width=321 ]
we have employed the idea of factorization to evaluate the @xmath0 production cross section at hadron colliders . for @xmath18 collisions we may write @xmath19.\end{aligned}\ ] ] where @xmath20 are parton distribution functions with momentum fractions of @xmath21 and @xmath22
, @xmath23 is the charm quark hard production cross section and @xmath24 represents the fragmentation of the produced heavy quark into @xmath25 state with confinement parameter @xmath15 at the scale @xmath26 .
we have used the parameterization due to martin - roberts - stiriling ( mrs ) [ 8 ] for parton distribution functions and have included the heavy quark production cross section up to the order of @xmath27 [ 9 ] .
the dependence on @xmath26 is estimated by choosing the transverse mass of the heavy quark as our central choice of scale defined by @xmath28 and vary it appropriate to the fragmentation scale of the particle state to be considered .
this choice of scale , which is of the order of @xmath29 ( parton ) , avoids the large logarithms in the process of the form @xmath30 or @xmath31 .
however , we have to sum up the logarithms of order of @xmath32 in the fragmentation functions . but this can be implemented by evolving them by the altarelli - parisi equation [ 10 ] .
the following form of this equation is used here @xmath33 here @xmath34 is the altarelli - parisi splitting function .
the boundary condition on the evolution equation ( 7 ) is the initial fragmentation function @xmath35 at some scale @xmath36 . in principle
this function may be calculated perturbatively as a series in @xmath3 at this scale .
detection of final state requires kinematical cuts of the transverse momentum , @xmath29 , and the rapidity , @xmath37 .
we have imposed the required @xmath38 and @xmath39 in our simulations for different colliders as required and have used the following definition of rapidity @xmath40 $ ] .
and its subsequent decay @xmath41 at the tevatron run i energies .
while the dashed curve is obtained using ( 1 ) or equally ( 3 ) with @xmath1 , the solid one is due to ( 3 ) with @xmath42 gev .
the result is compared with the cdf run i data .
other contributions are not included .
the scale is chosen to be @xmath43 .
, width=321 ] at tevatron run ii .
the two curves are obtained using ( 3 ) with @xmath44 and 0.6 gev respectively .
the scale has been set to @xmath43 .
, width=321 ] at the rhic and the cern lhc .
the shift due to the fermi motion is significantly increased in the case of lhc but shows to be less important at the rhic . in both cases
the fragmentation function ( 3 ) is employed with @xmath44 and 0.6 gev respectively.,width=321 ]
in this work we have employed a light - cone wave function to introduce the fermi motion in @xmath0 production in direct fragmentation channel and obtained its fragmentation function in leading order perturbative regime .
the behavior of @xmath0 fragmentation function with such a wave function is illustrated in fig .
we have studied the production rates at which the @xmath0 state is produced with and without the effect of fermi motion .
first we present the @xmath29 distribution of @xmath45 for the cases of the fermi motion off and on along with the cdf data at run i in the fig .
the branching ratio br(@xmath46)=0.0597 is taken from [ 11 ] .
the poor agreement with data is due to the fact that here we have only considered the contribution of @xmath47 .
similar behavior at tevatron run ii energies is shown in fig .
we have also extended our study to the cases of the rhic and the cern lhc @xmath48 colliders .
here we provide the @xmath29 distributions of the differential cross sections for @xmath49 and compare the two cases of the fermi motion off and on in the fig .
4 . in all cases
we have used @xmath42 gev for the confinement parameter in the fragmentation functions .
naturally , the results for @xmath15 in the range of 0 - 0.6 gev fall between the above results . calculation of the total integrated cross sections for each case shows that the total cross sections for with and without fermi motion essentially remain unchanged within the uncertainties of mote carlo simulations .
the reason is first due to the momentum sum rule which the fragmentation functions should satisfy . in other words
although the modification of fragmentation functions by the fermi motion redistributes the final states , the integrated cross sections are left unchanged . alternatively
although the fermi motion increases the fragmentation probability for the state , i.e. , introduces a state with overall higher mass , the cross section is lowered by just the same amount when we introduce the effect in calculation of the total integrated cross section .
it is evident from the figures 2,3 and 4 that the effect increases with increasing @xmath50 .
the kinematical cuts play important role apart from @xmath51 .
the large cross section at the rhic compared with the lhc in the fig .
4 is an example .
there are two main sources of uncertainities.the first is about the simulation of @xmath0 production at hadron colliders such as the uncertainties along with the fragmentation functions and parton distribution functions .
these kind of uncertainties are well discussed in the literature .
the second source of uncertainty is due to the choice of the confinement parameter in the fragmentation function .
relying on our discussion in section 2 , our choice of @xmath42 gev seems to be justified .
future determination of this parameter will shed more light on this situation .
it is worth mentioning that our choice of charm quark mass , i.e. , 1.25 gev , have put our results in their upper side and that the change of the charm quark mass in its acceptable range does not have significant impact on the fermi motion effect in @xmath52 production .
m. cacciari and m. greco , phys .
( 1994 ) 1586 ; k. sridhar , a.d .
martin and w.j .
stirling , phys .
b 438 ( 1998 ) 211 ; a. petrell , nucl . phys .
86 ( 2000 ) 533 .
fred cooper , et al .
, arxiv : hep - ph/0402219 v2 ( 2004 ) c. albajar , et al .
b256 ( 1991 ) 112 ; f. abe , et al .
, phys . rev .
69 ( 1992 ) 3704 ; f. abe , et al . ,
71 ( 1993 ) 2537 ; raphael granier de cassagnac , the phenix collaboration , j. phys .
g : nucl . and part .
( 2004 ) s1341 ; d. acosta , et al .
d71 ( 2005 ) 032001 .
bodwin , e. braaten and g. p. lepage , phys . rev . d51 ( 1995 ) 1125 ; e. braaten , m.a .
doncheski , s. fleming and m. mangano , phys .
b333 ( 1994 ) 548 ; g.t .
bodwin , e. braaten and g.p .
lepage , phys .
d46 ( 1992 ) r1914 ; m. cacciari and m. greco , phys .
73 ( 1994 ) 1586 .
p. nason , s. dawson and r.k .
ellis , nucl .
b303 ( 1988 ) 607 ; w. beenakker , h. kuijf , w.l .
van neerven and j. smith , phys .
d40 ( 1989 ) 54 ; w. beenakker , w.l .
van neerven , r. meng , g. schuler and j. smith , nucl . phys .
b351 ( 1991 ) 507 .
|
the fermi motion effect in @xmath0 production in various hadron colliders is studied .
we deduce that in agreement with sum rules which the fragmentation functions should satisfy , while the effect has considerable influence on the fragmentation probabilities and the differential cross sections , the total cross sections are essentially left unchanged .
|
] let @xmath0 be a slim , planar , semimodular lattice , an _ sps lattice_. as in g. czdli and e.t .
schmidt @xcite , _ inserting a fork _ to @xmath0 at the covering square @xmath1 , firstly , replaces @xmath1 by a copy of @xmath2 ( see the lattice @xmath2 in figure [ f : s7 ] ) . secondly ,
if there is a chain @xmath3 such that the element @xmath4 has just been added and @xmath5 is a covering square in the lattice @xmath0 ( and so @xmath6 is not on the boundary of @xmath0 ) but @xmath7 at the present stage of the construction , then we insert a new element @xmath8 such that @xmath9 and @xmath10 .
let @xmath11 $ ] denote the lattice we obtain when the procedure terminates .
we say that @xmath11 $ ] is obtained from @xmath0 by _ inserting a fork to @xmath0 _ at the covering square @xmath1 .
see figure [ f : forks ] for an illustration . ]
let @xmath0 be an sps lattice and @xmath1 a covering square of @xmath0 .
in this series of research notes we examine the connections between the congruence lattice of @xmath0 and the congruence lattice of @xmath11 $ ] . in part
i , g. grtzer @xcite , we proved the following result : [ t : extension ] let @xmath0 be a slim , semimodular , planar lattice .
let @xmath1 be a covering square of @xmath0 .
then every congruence of @xmath0 extends to @xmath11 $ ] .
let @xmath0 be an sps lattice and let @xmath12 be a covering square of @xmath0 .
we call @xmath1 a _ tight square _ if @xmath13 covers exactly two elements , namely , @xmath14 and @xmath15 , in @xmath0 ; otherwise , @xmath1 is a _ wide square_. [ t : wide ] let @xmath0 be an sps lattice .
let @xmath1 be a wide square .
then @xmath11 $ ] is a congruence - preserving extension of @xmath0 .
let @xmath16 be a congruence of @xmath0 .
then there is a smallest congruence @xmath17 of @xmath11 $ ] generated by @xmath16 . by theorem [ t : extension ] , @xmath16 is the restriction of @xmath17 to @xmath0 . moreover , if @xmath16 is a join - irreducible congruence of @xmath0 , then @xmath17 is a join - irreducible congruence of @xmath11 $ ] .
[ t : tight ] let @xmath0 be an sps lattice .
let @xmath1 be a _
tight square_. then @xmath11 $ ] has exactly one join - irreducible congruence , @xmath18 , that is not the minimal extension of a join - irreducible congruence of @xmath0 . in this note
we will examine the congruence @xmath18 .
we will use the notations and concepts of lattice theory , as in @xcite .
as illustrated in figure [ f : cong2 ] , we say that @xmath19 $ ] is _ up congruence - perspective _ to @xmath20 $ ] and write @xmath19 \cperspup [ c , d]$][gon : cperspup ] if @xmath21 and @xmath22 ; similarly , @xmath19 $ ] is _ down congruence - perspective _ to @xmath20 $ ] and write @xmath19 \cperspdn [ c , d]$ ] if @xmath23 and @xmath24 .
if @xmath19 \cperspup [ c , d]$ ] _ or _ @xmath19 \cperspdn [ c , d]$ ] , then @xmath19 $ ] is _ congruence - perspective _ to @xmath20 $ ] and we write @xmath19 \cpersp [ c , d]$ ] .
\cperspup [ c , d]$ ] and @xmath19 \cperspdn [ c , d]$ ] ] if for some natural number @xmath25 and intervals @xmath26 $ ] , for @xmath27 , @xmath28 then we call @xmath19 $ ] _ congruence - projective _ to @xmath20 $ ] , and we write @xmath29 \cproj [ c , d]$ ] .
we now state a classic result in a special case .
[ l : cproj ] let l be a lattice , @xmath30 in @xmath0 , and let @xmath31 be a prime interval . then @xmath31 is collapsed by @xmath32 iff @xmath29 \cproj { \f q}$ ] .
in fact , there is a prime interval @xmath33 in @xmath19 $ ] such that @xmath34 .
there are two technical lemmas we need .
[ l : technical ] let @xmath0 be a finite lattice .
let @xmath35 be an equivalence relation on @xmath0 with intervals as equivalence classes .
then @xmath35 is a congruence relation iff the following condition and its dual hold : @xmath36 we want to prove that if @xmath37 and @xmath38 , then @xmath39 .
the proof is a trivial induction first on @xmath40 $ ] and then on @xmath41 $ ] .
let ( c@xmath42 ) denote the dual of ( c@xmath43 ) .
the following statement is a variant of lemmas 245 and 246 of @xcite .
[ l : technical2 ] let @xmath0 be a finite lattice .
let @xmath44 be an extension of @xmath0 .
let us assume that every congruence of @xmath0 extends to @xmath44 .
then @xmath44 is a congruence - preserving extension iff the following condition is satisfied : 1 . for every prime interval @xmath33 of @xmath44 , @xmath33 not in @xmath0
, there exists a prime interval @xmath31 of @xmath0 such that @xmath45 in @xmath44 .
let @xmath0 be an sps lattice .
let @xmath46 be a covering square of @xmath0 , let @xmath14 be to the left of @xmath15 we need some notation for the @xmath11 $ ] construction , see figure [ f : forksdetails ] .
we start the construction by adding the elements @xmath47 , @xmath48 , and @xmath49 so that the set @xmath50 forms a sublattice @xmath2 .
we say that this @xmath2 sublattice and the covering square @xmath1 are _ associated _ with each other .
let @xmath51 , @xmath52 .
if @xmath53 is the largest number so that @xmath54 , @xmath55 , and @xmath56 have already been defined , and @xmath57 is a covering square in @xmath0 , then we add the element @xmath58 , so we get two new covering squares @xmath59 and @xmath60 .
we proceed similarly on the right .
so @xmath11 $ ] is constructed by inserting the elements in the set @xmath61 = \set{t , z_{l,1 } \succ \dots \succ z_{l , n } , z_{r,1 } \succ \dots \succ z_{r , m}}\ ] ] so that @xmath50 is a sublattice @xmath2 , moreover , @xmath62 for @xmath63 , and @xmath64 for @xmath65 .
the new elements are black filled in figure [ f : forksdetails ] .
we name a few join - irreducible congruences of @xmath0 and @xmath11 $ ] that will plays an important role : @xmath66 and @xmath67 in @xmath0 and @xmath68}{a_l , i}$ ] , @xmath69}{a_r , i}$ ] and @xmath70}{t , i}$ ] .
the following results are well known .
[ l : known ] let @xmath0 be an sps lattice .
an element of @xmath0 has at most two covers .
let @xmath71 .
let @xmath72 cover the three elements @xmath73 , @xmath74 , and @xmath75 .
then the set @xmath76 generates an @xmath2 sublattice . if the elements @xmath73 , @xmath74 , and @xmath75 are adjacent , then the @xmath2 sublattice of is a cover - preserving sublattice . finally , we state the structure theorem for sps lattices of g. czdli and e.t .
schmidt @xcite : [ t : structure theorem ] let @xmath0 be an sps lattice .
there exists a planar distributive lattice @xmath77 such that @xmath0 can be obtained from @xmath77 by a series of fork insertions .
[ l : moretwo ] let @xmath0 be an sps lattice . let @xmath1 be a _
wide square_. then @xmath78 or @xmath79 in @xmath11 $ ] . since @xmath1 is wide , the element @xmath13 covers an element @xmath72 in @xmath0 , with @xmath80 .
either @xmath72 is to the left of @xmath14 or to the right of @xmath15 , let us assume the latter . by lemma [ l : known ]
, the set @xmath81 generates an @xmath2 sublattice in @xmath0 .
then @xmath82 , computed in the @xmath2 sublattice generated by @xmath83 and @xmath84 , computed in the @xmath2 sublattice generated by @xmath85 , yielding @xmath79 .
the symmetric case yields @xmath78 .
now we prove theorem [ t : wide ] . by theorem [ t :
extension ] , the assumptions of lemma [ l : technical2 ] hold for @xmath0 and @xmath86 $ ] .
so by ( p ) of lemma [ l : technical2 ] , we have to show that for every prime interval @xmath33 of @xmath11 $ ] , if @xmath33 is not in @xmath0 , then there exists a prime interval @xmath31 of @xmath0 such that @xmath45 in @xmath11 $ ] . the prime intervals of @xmath11
$ ] that are not in @xmath0 can be listed as follows : @xmath87 , & \dots , [ z_{l , n } , x_{l , n}];\label{e : set1}\\ & [ t , i ] ; \label{e : set2}\\ [ z_{r,1 } , x_{r,1 } ] , & \dots , [ z_{r , m } , x_{r , m}];\label{e : set3}\\ [ y_{l,1 } , z_{l,1 } ] , & \dots , [ y_{l , n } , z_{l , n } ] ; \label{e : set4}\\ [ y_{r,1 } , z_{r,1 } ] , & \dots , [ y_{r , m } , z_{r , m}].\label{e : set5}\end{aligned}\ ] ] we choose the following prime intervals @xmath31 of @xmath0 to satisfy ( p ) of lemma [ l : technical2 ] : for the prime intervals in the lists and , choose @xmath88 $ ] ; for the prime intervals in the list , choose @xmath89 $ ] .
let @xmath0 be an sps lattice .
let @xmath46 be a covering square of @xmath0 .
we call @xmath1 _ distributive _ if the ideal generated by @xmath1 , that is , @xmath90 , is distributive .
let @xmath0 be an sps lattice .
let @xmath91 and @xmath92 be covering @xmath2 sublattices of @xmath0 in the covering squares @xmath93 and @xmath94 , with unit elements @xmath95 and @xmath96 , respectively .
let @xmath97 mean that @xmath98 .
let us call @xmath91 _ minimal _ if @xmath97 fails for any covering @xmath2 sublattice @xmath94 of @xmath0 .
the following statement see g. czdli and e.t .
schmidt @xcite .
[ l : minimal ] let @xmath0 be an sps lattice and let @xmath1 be be a covering square of @xmath0 .
let @xmath99 be the @xmath2 sublattice of @xmath11 $ ] associated with @xmath1 .
then @xmath99 is minimal iff @xmath1 is distributive .
let @xmath0 be an sps lattice and let @xmath1 be a distributive covering square .
we define in @xmath11 $ ] an equivalence relation @xmath18 as follows .
all equivalence classes of @xmath18 are singletons except for the following : @xmath100 [ l : distr ] let @xmath0 be an sps lattice and let @xmath1 be a distributive covering square .
then @xmath18 is a congruence relation of @xmath11 $ ] . by lemma [ l : technical ]
, we have to verify ( c@xmath43 ) and ( c@xmath42 ) . to verify ( c@xmath43 ) , let @xmath101 be covered by @xmath102 in @xmath11 $ ] and @xmath103 .
if @xmath104 and @xmath105 , where @xmath106 , then @xmath107 and @xmath108 , because @xmath109 is in the list .
if @xmath110 , we proceed the same way with @xmath111 and the list .
we proceed `` on the right '' with the lists and .
finally , @xmath112 can not happen because @xmath47 is covered only by one element .
the verification of ( c@xmath42 ) is very similar .
note that we use implicitly that an element of @xmath113 can not be covered by three elements , see lemma [ l : known](i ) .
also an element of @xmath114 can not cover three elements , because @xmath114 is a planar distributive lattice .
lemma [ l : distr ] is closely related to the construction of the lattice @xmath115 from @xmath116 in g. czdli @xcite , where a fork is inserted into a distributive covering cell .
let @xmath0 be an sps lattice and let @xmath1 be a covering square @xmath12 of @xmath0 .
we call @xmath1 a _ tight square _ if @xmath13 covers exactly two elements , namely , @xmath14 and @xmath15 , in @xmath0 ; otherwise , @xmath1 is a _
wide square_. for a tight square @xmath1 , a _
protrusion _ is an element @xmath117 , @xmath63 ( or symmetrically , @xmath118 , @xmath65 ) , such that @xmath117 covers three or more element in @xmath0 ( equivalently , three or more element in @xmath11 $ ] ) ; see figure [ f : protrusion ] , especially for the notation @xmath119 , shown for @xmath120 , for the element of @xmath0 covered by @xmath117 that is immediately left of @xmath121 . if @xmath117 is a protrusion and @xmath122 has two covers , we denote by @xmath123 the cover different from @xmath124 ; we define @xmath125 , and so on , similarly .
let @xmath126 be the largest integer for which @xmath127 is defined .
the elements @xmath124 ,
, @xmath128 are the _ extensions _ of the protrusion .
there is always one extension , namely , @xmath124 . in figure
[ f : protrusion ] , @xmath120 and @xmath129 , there are three extensions .
let @xmath130 denote the set of all left protrusions , that is @xmath131 iff @xmath117 is a left - protrusion .
we define @xmath132 symmetrically . and
extensions @xmath133 , @xmath134 , @xmath135 with two protrusions ] for every @xmath131 , define @xmath136 and @xmath137 .
we define , symmetrically , @xmath138 and @xmath139 .
finally , let @xmath140 , the _ protrusion congruence_. we will also need the congruence @xmath141 and @xmath142 , the congruences of @xmath11 $ ] generated by @xmath143 and @xmath142 , respectively . [ l : pi ] let @xmath0 be an sps lattice and let @xmath1 be a tight square .
let @xmath131 . in @xmath11
$ ] , all nontrivial congruence classes of @xmath141 except for @xmath144 and @xmath145 are in @xmath146 . by the structure theorem ( theorem [ t : structure theorem ] ) , the lattice @xmath11 $ ] can be obtained from a planar distributive lattice @xmath77 by inserting @xmath25 forks .
we induct on @xmath25 .
if @xmath147 , there is no protrusion , so there is no @xmath143
. there is nothing to prove .
let us assume that the statement holds for @xmath148 .
let @xmath149 be a covering square of @xmath11 $ ] such that @xmath150 $ ] is minimal @xmath2 sublattice of @xmath11 $ ] .
let @xmath151 - f[a]$ ] .
then @xmath152 is a sublattice of @xmath11 $ ] and @xmath153 = l[s]$ ] .
we can obtain @xmath152 from the planar distributive lattice @xmath77 by inserting @xmath148 forks .
so the statement holds for @xmath152 . by inserting the fork at @xmath99 ,
may make the congruence @xmath141 larger in @xmath154 in @xmath153 = l[s]$ ] , but on the upper left and upper right boundaries of @xmath155 there is no change , so there is no effect outside of the filter @xmath146 .
[ l : protrusion ] let @xmath0 be an sps lattice , @xmath1 a tight square of @xmath0 . let @xmath131 .
then @xmath156 case : @xmath117 is a protrusion and @xmath181 . if @xmath177 , @xmath178 or if @xmath182 , @xmath183 , we proceed as in case .
@xmath184 can not happen because @xmath121 has only one cover .
so we are left with @xmath182 , @xmath185 .
then @xmath186 and @xmath187 case : @xmath117 is a protrusion and @xmath192 . in this case , either @xmath193 , @xmath194 or @xmath195 , @xmath196 . if @xmath193 , @xmath194 , then @xmath197 and @xmath198 by . if @xmath195 , @xmath196 , then @xmath186 and @xmath199 by .
case : @xmath117 is a protrusion and @xmath181 . if @xmath177 , @xmath178 or if @xmath182 , @xmath183 , we proceed as in case .
if @xmath208 , @xmath183 , then @xmath211 and @xmath212 by .
finally , let @xmath177 , @xmath213
. then @xmath214 and @xmath215 by .
[ l : ] let @xmath0 be an sps lattice , @xmath1 a tight square of @xmath0 .
let @xmath131 .
then the congruence classes of @xmath141 in @xmath11 $ ] are the congruence classes of @xmath143 in @xmath0 and @xmath222 moreover , neither @xmath143 nor @xmath141 has any nontrivial congruence classes outside of @xmath114 . in particular , @xmath223 .
[ t : tight ] let @xmath0 be an sps lattice .
let @xmath1 be a _
tight square_. let @xmath227 $ ] be a prime interval of @xmath0 .
let us assume that @xmath228 in @xmath11 $ ] . then either @xmath229 or @xmath230 in @xmath0 .
so let @xmath228 in @xmath11 $ ] for a prime interval @xmath33 of @xmath11 $ ] . by lemmas [ l : cproj ] and [ l : g ] , there is a sequence of intervals in @xmath11 $ ] : @xmath231 \in g.\ ] ] ( or symmetrically ) , using the notation @xmath232 , @xmath233 .
we can assume that ( s ) was chosen to minimize @xmath25 .
in particular , @xmath234 and @xmath235 alternate in ( s ) . the interval @xmath236 $ ] has a prime subinterval @xmath237 $ ] perspective to @xmath31 . since @xmath224 is a set of prime interval closed under prime perspectivity , it follows that @xmath236 $ ] has @xmath31 as a subinterval by minimality . if @xmath241 = [ z_{l , i } , x_{l , i}]$ ] ( or symmetrically ) , with @xmath245 , then @xmath246 and so @xmath247 or @xmath248 .
the first possibility contradicts the minimality of @xmath13 , while the second yields that @xmath249 , as claimed .
second , let @xmath250 and @xmath239 \in g$ ] .
there are two possibilities : @xmath251\label{e : up}\\ \intertext{and } \inv{e_{n-2}}{f_{n-2 } } & \cperspdn [ { e_{n-1}}{f_{n-1}}]\label{e : dn}\end{aligned}\ ] ] if holds , then @xmath252 and @xmath253 ( which implies that @xmath117 precedes the first protrusion ) , contradicting the minimality of @xmath25 .
[ c : tight ] let @xmath1 be a _
tight square _ in an sps lattice @xmath0
. then the congruence @xmath18 of @xmath11 $ ] is covered by one or two congruences in @xmath257)$ ] , namely , by @xmath258 and @xmath259 . , _ notes on planar semimodular lattices .
the size of a minimal congruence lattice representation with rectangular lattices .
( szeged ) ( 2010 ) , 326 .
g. grtzer and e.t .
schmidt , _ congruences of planar semimodular lattices .
i. a short proof of the representation theorem .
_ arxiv : 1303.4464 .
algebra universalis ( 2013 ) .
|
g. czdli and e.t . schmidt introduced in 2012 the fork extension . continuing from part i
, we investigate the congruences of a fork extension .
|
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.