ebook img

Gravitational lensing by point masses on regular grid points PDF

0.83 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Gravitational lensing by point masses on regular grid points

J.A:Gravitationallensingbypointmassesonregulargridpoints Gravitational lensing by point masses on regular grid points Jin H. An⋆ MITKavliInstituteforAstrophysics&SpaceResearch,MassachusettsInstituteofTechnology,77MassachusettsAvenue,CambridgeMA02139,USA MonthlyNoticesoftheRoyalAstronomicalSociety ABSTRACT 7 Itisshownthatgravitationallensingbypointmassesarrangedinaninfinitelyextendedregular 0 latticecanbestudiedanalyticallyusingtheWeierstrassfunctions.Inparticular,wedrawthe 0 criticalcurvesandthecausticnetworksforthelensesarrangedinregular-polygonal–square, 2 equilateraltriangle,regular hexagon– grids. From this, the mean numberof positive parity n images as a function of the average optical depth is derived and compared to the case of a the infinitely extendedfield of randomlydistributed lenses. We find that the high degree of J the symmetryin the lattice arrangementleads to a significantbias towards cancelingof the 1 shearcausedbytheneighboringlensesonagivenlenspositionandlensingbehaviourthatis 3 qualitativelydistinctfromtherandomstar field. We also discusssomepossible connections tomorerealisticlensingscenarios. 2 v 2 9 1 1 INTRODUCTION icalcurvesandthecausticnetworks.InSect.4,westudythemean 2 number of positive parity images, which is compared to the case 1 Beyond the Galactic and Local-Group (in particular, Magellanic oflensingbyaninfiniterandomstarfield.InSect.5,weconsider 6 Clouds and M31) microlensing experiments (Alcock et al. 2000; 0 theeffectoftheexternalpotentialbymeanofaddingexternalshear Afonsoetal.2003;CalchiNovatietal.2005;deJongetal.2006), / (andconvergence),anddiscusssomepossibleconnectionstomore h the main interest on the point-mass lenses lies in the effect of realisticlensingscenarios. p the individual stars in a lensing galaxy (Chang & Refsdal 1979, Wearguethat,whilethesystemconsideredheremaybesome- - 1984; Nityananda & Ostriker 1984; Irwin et al. 1989; Witt et al. o whatartificialinitsconstructandratherabstractinitsnature,the 1995; Schechter & Wambsganss 2002; Keeton et al. 2006). For r study presented in this paper can lead us to some insight on mi- t these cases, each point-mass lens cannot be treated individually s and their collective effects usually differ drastically from that of crolensingathighopticaldepth. a : the simple linear superposition of them. The usual approach to v this“high-optical-depthmicrolensing”problemis“inverseraytrac- i X ing”(Kayseretal.1986;Schneider&Weiss1987;Wambsgansset 2 LENSEQUATION al.1990,1992),thatis,examiningstatisticalpropertiesoflensing r Inmostastronomicalsituations,gravitationallensingisdescribed a observablesusingMonte-Carlorealizationsofthelensingsystem. However,asthenumberoftheindividualpoint-masslensestore- by the lens equation y = x ∇ψ, which is derived from the − lowest-order nontrivial approximation to the path of light propa- producetherealisticscenariowellapproachesthenumberofstars inthegalaxy,thisbecomesveryexpensiveratherquicklyinterms gation(Schneideretal.1992; Pettersetal.2001;Kochanek etal. 2005).Here, yand xarethe(2-dimensional)vectorsrepresenting oftherequiredresources.Anothercomplementaryapproachtothe problem is from the random walk and the probability theory to- thelinesofsighttoward,respectively,theangularpositionsofthe getherwiththethermodynamicapproximation(thatis,effectively sourceundertheabsenceoflensingandthelensedimage.Thelens- consideringthelimitattheinfinitenumberofthestars),whichhas ingpotentialψisthegravitationalpotential(equivalently,thegrav- itational‘Shapiro’timedelay)integratedalongthelineofsightup beenquitesuccessfulincertainwell-definedproblems(Nityananda &Ostriker1984;Schneider1987;Schneideretal.1992). toascaleconstant(Blandford&Narayan1986). Inthefollowing,weconsideradifferentapproachtotheprob- Inparticular,gravitationallensingbyapointmassisdescribed bythepotentialψ(x)=θ2ln x l ifthepoint-masslensislocated lem: seeking complete analytic solutions. In many complex sys- E | − | alongthelineofsightindicatedbythevector l(Paczyn´ski1986). tems,imposinghighlevelofsymmetrycanreducetheproblemto asimplerone,whichmaybeotherwisetoocomplicatedtoanalyze. Here,thescaleconstantθEisknownasthe‘Einstein(ring)radius’ anddeterminedbythemassofthelens(θ2 M)andthedistances Ina similar spirit,in thispaper we consider gravitational lensing E ∝ by equal point masses that are arranged in an infinitely extended among thesource, thelens, and theobserver. Withthispotential, wefindthelensequationforpoint-masslensing, regular lattice. After we derive the appropriate lens equation in Sect.2,theyareexaminedinSect.3tofindthecorrespondingcrit- x l y= x − (1) − x l2 | − | ⋆ E-mail:[email protected] wheretheunitofangularmeasurementshasbeenchosensuchthat c 2007RAS (cid:13) 2 J. An θ = 1.Theequationisgeneralizedtothecaseofmultiplepoint- 1972,orseealsoAppendix).Here,Γ Γ(x)isthegammafunc- E x ≡ masslenseswithasmoothly-varying‘largescale’potential(Young tion,℘(z)istheWeierstrassellipticfunctionand℘(z)isitscomplex ′ 1981)suchthat derivative.Wealsouseanabbreviatednotationfor(thepowerto) thegammafunctionsuchthat[Γ(x)]n = Γn inordertoavoidpro- y=x−Xi=N1 mMi|xx−−llii|2 −∇ψext. (2) ltpihefeenrdraeitxaio)lnhaanoldff-pκpaerrei=ondtπhoe/fs(e2tshωea)2nledimsbntrhiasecckoaetpittcsi.ccaIanlsedaxedopdftih℘tio-(fVnu,ineωctrt0iio&≈n(O1s.es8et5rAi4kpeisr- ⋆ Here,theunitofangularmeasurementsisnowgivenbyθ corre- E 1983; Paczyn´ski 1986) of the lenses. If we consider the square spondingtoafiducialmassscaleMandsubsequentlythe‘external’ cell given by the centres of each lensing lattice, it is straightfor- lensingpotentialψ shouldberescaledappropriately.Inaddition, ext wardtoseethatthereexistsalenscorrespondingtoeachcellwith the masses of individual lenses m enter into the equation as the i sidelengthof2ω.Hence,themeansurfacedensityofpointmasses ratiotothefiducialscale. is in fact given by σ = (2ω) 2, and therefore, the optical depth − Next,letusthinkofthecasethattheequalpoint-masslenses by κ = πσ = π/(2ω)2. Moreover, the 90 -rotational symme- ⋆ ◦ are located at the lattice points on an infinitely extended square try of the system implies that the total shear due to all the other grid.Inaddition,weignoretheeffectoftheexternalpotentialfor lenses on any given lens cancels out to be zero [mathematically themoment. Utilizingthecomplex-number notation(Bourassa& this indicates the constant term in the Laurent-series expansion Kantowski 1975; Subramanian et al. 1985; Witt1990; An2005), of f (z) is null]. Consequently, we find that f (z) = ℘(z;g ,0) thelensequation(2)thencanbewrittenas ′ ′ − 2 and f(z) = ζ(z;g ,0)+C.Here,ζ(z)istheWeierstrasszetafunc- 2 η=z f(z); tion(Abramowitz&Stegun1972) andC isanarbitrarycomplex − constant.Ifwechoosethe‘centre’ofthesystemtocoincidewith 1 ∞ 1 (3) the lens at the coordinate origin, then we have C = 0. We note f(z)= = , Xi z−λi m,Xn= z−2(m+ni)ω that,whiletheconstantCcannotbeindependentlydetermined,its −∞ choiceisphysicallyinconsequential.Thesituationisanalogousto whereη,z,andλarethecomplexifiedvariablescorrespondingtoy, lensing by an infinitely extended mass screen. While a na¨ıve ex- x,and l,respectively,andtheover-barnotationisusedtoindicate pectation from the symmetry appears to indicate null deflections complexconjugation.Wefurtherassumethatthelenslocationsare along every lineof sight, detailed physical consideration leads to given byλ = 2(m+ni)ωwherem andnareanypair of integers theuniformfocusingwithrespecttoanarbitrarychoiceofthecen- (bothrunningfromthenegativeinfinitytothepositiveinfinity)and tre.Inthecurrent situation,thediscretetranslationsymmetryim- ω is the half distance between adjacent grid points. Without any pliesthearbitrarinessinthechoiceofthecentreandconsequently lossofgenerality,ωcanberestrictedtobeapositivereal.Strictly thatofC,whichresultsintheinfinitesuminequation(3)notbeing speaking,the‘deflectionfunction’ f(z)givenformallyintermsof well-determined.However,theconstant-deflectionterminthelens theinfinitesuminequation(3)isnotwell-definedasitisnotcon- equation againonlyleads toaconstant offset betweenthesource vergent.Thisisinfactanaturalconsequenceofhavinganinfinite and the image/lens plane, which has no observable consequence totallensingmass, whichisunphysical. However, whatisimpor- andthereforecanbeignored. tantinourunderstandingofthelensingisnottheabsoluteamount Someresultsareimmediatefromtheresultinglensequation ofthedeflectionbutthedifferenceofthedeflectionsbetweenneigh- bouringlinesofsights.Thatistosay,ifweaddanyconstanttothe η=z−ζ(z;g2,0). (6) deflectionfunctioninthelensequation,theresultinglensequation Forinstance,thepropertyofζ(z)suchthat isreducedtotheequivalentonewithouttheconstantbyintroduc- π ingthe“offset”tothesourcepositiontocanceltheconstant.Two ζ z+2(m+ni)ω;g ,0 =ζ(z;g ,0)+ (m ni) (7) equationsthenareobservationallyequivalentbecausethereisnoa 2 2 2ω − (cid:2) (cid:3) prioriinformationregardingthesourcepositionintheabsenceof forarbitraryintegersmandnindicatesthattheverticesz,z+2ω, thelens. Similarly,despitetheseemingglobal failureofequation z+2(1+i)ω,andz+2iωofasquareinthelensplanemaptothe (3),wecanstillproceedbyfocusingonthelocalpropertiesofthe pointsonthesourceplane:η = z ζ(z),η+ ,η+(1+i) ,and lensingdescribedbyequation(3). η+i where = 2ω π/(2ω),−and therefoℵre the mean iℵnverse Forthis,wefirstwritedownthe(formal)second-order com- magnℵificationℵisgivenb−y[ /(2ω)]2 = [1 π/(2ω)2]2.Theresult plexderivativeofthedeflectionfunction; confirms the expectation thℵat the sufficien−tly large beam of light wouldseethesystemasifitwereauniformscreenofmasswitha f′′(z)=m,Xn∞= [z−2(m2+ni)ω]3 (4) rceoanlvleirngee,necqeu(aotriotnhe(6o)ptmicaaplsdeepacthh)sκeg=mπenσt =betπw/e(e2nω)t2w.oAalodnjagcethnet −∞ ≡−℘′(z|ω,iω)=−℘′(z;g2,0) ltehnesreesaroentionfitnhietewnhuomlebereraolfliinmea,gaensd. Hthouwseivteirs, uenasleilsysaκrgu=ed1,thaaltl ⋆ where butafinitenumberofimageshavenegativeparity. ω 4 Γ8 g = 0 = 1/4 κ2. (5) 2 (cid:18)ω(cid:19) 16π4 ⋆ 2.1 lensesontriangularandhexagonalgrid Wefindthatthisisrelatedtothedefinitionofthespecialfunction Analogous to the preceding case, we can also set up the lens knownastheWeierstrassellipticfunction1(Abramowitz&Stegun equation for the equal point-mass lenses located at the lattice points on an infinitely extended equilateral-triangular grid. The 60 -rotational symmetry of the system again implies that the to- ◦ 1 Throughoutthispaper,theargumentsoftheWeierstrassfunctionsfollowedbyaverticalbar– – talshear on anygiven lenscausedby theremainingset of lenses | denotethehalf-periodswhereastheellipticinvariantsareindicatedbythosefollowedbyasemicolon also cancels out. This fact, combined with the general definition (;).Fordetails,seethelistedreferencesorAppendix.Wheneveritisappropriate,thesearguments maybesuppressed,providedthatthereislittledangerofconfusion. of℘-functionindicatesthatthelensequationcanagainbewritten c 2007RAS,MNRAS000,1–19 (cid:13) Lensingbyregularpoint-masslattices 3 Figure1.Contourplotsoftheequipotentiallinesofthelensingpotentialforthelatticelenses.TopLeft:squarelattice.TopRight:triangularlattice.Bottom Left:hexagonallatticewithequation(14).BottomRight:hexagonallatticewithalensbeingatthecenter. downusingtheWeierstrassfunctions.Ifthelensesareplacedover convergenceisκ = π/(2√3ω2)(andthemeansurfacedensityσ ⋆ thegridgivenbyλ = 2ω(m+neπi/3)wheremandnareintegers, ofthepointmassesonthetriangulargridisgivenbyσ 1 =2√3ω2 − thecorrespondinglensequationisgivenby where2ωisthesidelengthoftheunittriangularcell). Wenotethatthetheprecedingtwocasesofthelensarrange- η=z f(z); f(z)=ζ(zω,eπi/3ω)=ζ(z;0,g ) (8) − | 3 mentcorrespondtotwoofthethreepossibleregulartessellationsof thetwo-dimensionalplane.Itmaybeofsomeinteresttoconsider where the regular lens placement corresponding to the remaining regu- ω 6 √3Γ6 3 lar tessellation – the hexagonal or honeycomb tiling. It turns out g = 2 = 1/3 κ3 (9) thattheregularhexagonal gridcaseiscloselyrelatedtotheequi- an3dω(cid:18)ω (cid:19)1.530is8tπh3erealh⋆alf-periodsoftheequiharmoniccaseof lfaotremraslatr(i3a0n◦g-urolatratgedri)dl.a2rgBeyrtrreimanogvuilnagrgerviedryofthsiidrdeloefngththeolefn2s√th3aωt ℘-func2ti≈on(seeAppendix)whereasκ = π/(2√3ω2)isthecorre- fromthebasetriangulargridwithsidelengthof2ω,theremaining ⋆ lensesformahexagonalgrid(ortheverticesofhoneycombcells). spondingopticaldepth.Likewise,thequasiperiodicityofζ(z;0,g ), 3 π ζ z+2(m+neπi/3)ω;0,g3 =ζ(z;0,g3)+ m+ne−πi/3 (10) √3ω (cid:2) (cid:3) (cid:0) (cid:1) wheremandnarearbitraryintegers,indicatesthatthevertexpoints z,z+2ω,andz+2eπi/3ωoftheequilateraltriangleinthelensplane mapontothesourceplanepoints:η=z ζ(z),η+iandη+ieπi/3 wherei = 2ω π/(√3ω)sothatthem−eaninversemagnification 2 Notethattheyareinfactso-calleddualtilingofeachother.Thatis,ifweconsideragridwith − vertexpointsatthecentresofeachtriangularcelloftheequilateraltriangulargrid,thentheresulting is again given by (1−κ⋆)2 where the optical depth or the mean latticeformsaregular-hexagonal(honeycomb)gridandalsoviceversa. c 2007RAS,MNRAS000,1–19 (cid:13) 4 J. An Figure2.Critical curves(dottedlines)andcausticnetworks (solidlines) Figure3.SameasFig.2exceptforω=1.2[κ⋆=π/(2ω)2 0.545]. forlensesonasquarelatticewithω=1.4[κ⋆=π/(2ω)2 0.401]. ≈ ≈ for the lensing potential for some purposes such as the time de- Consequently,thecorrespondinglensequationmaybewrittenas lay.Althoughthedirectinfinitesumoftheindividuallogarithmic potential of thepoint-mass lensisdivergent everywhere, one can η=z d(z;g ); − 3 get around this difficulty through the known antiderivative of the d(z;g3)=ζ(z|ω,eπi/3ω)−ζ(z|√3eπi/6ω,√3iω) (11) Weierstrasszeta function. First, wenote therelation between the g realpotentialandthecomplexdeflectionfunction(An2005;An& =ζ(z;0,g3)−ζ(cid:16)z;0,−273(cid:17) tEhvearnesa2l0p0o6te);nttihaelcisomgipvleenxbleynψs.eHqueareti,otnheisogpievreantobry∂η=indzi−ca2t∂esz¯ψthief whereg isrelatedtoωthroughequation (9)providedthat 2ωis z¯ 3 ‘Wirtingerderivative’(e.g.,Schramm&Kayser1995)withrespect stillthesidelengthofthebasetriangulargrid(andalsothatofthe toz¯,thatis,ifψ=ψ(x,y)andz=x+yi,then unithoneycombcell).However,sinceeverythirdoflenshasbeen removedfromthebasetriangularcell,theopticaldepthisreduced ∂ψ∂x ∂ψ∂y 1∂ψ i ∂ψ ∂ψ= + = + (13) toκ⋆ =π/(3√3ω2),andthus z¯ ∂x ∂z¯ ∂y ∂z¯ 2∂x 2 ∂y 33/2Γ6 3 becausex=(z+z¯)/2andy=i(z¯ z)/2.Forthecurrentscenario,we g3 = 24π13/3 κ⋆3. (12) have f(z)=2∂z¯ψ=2∂zψ¯ =2∂zψ−(notethatψ=ψ¯ sinceψisreal). Icnomadpdleitxiodni,ffde′r(ezn;tgi3a)tio=n,−ih.e(.z,;tgh3e)d(tehreivpartiivmeedwistyhmrebsopleicntdtiocazt)e.sTthhee ∂Mcoψomrep=olevx0e.ra,Tnfha(lezyn)t,iic∂sfa(uψcnoc+mtiψo¯pnl)eψx=ca(∂nz)aψlsyut+cich∂ftuψh¯nact=tfio(∂zn)ψa=n+dψ∂s′co(ψzt)h=e=re∂fze(ψzx)ci,saatsnndda z¯ c z c c z c z c z c z¯ c functionh(z;g )isstudiedinAppendixA2.1.Itisagainstraightfor- thus,2ψ=ψ +ψ¯ ,thatis,ψ(x,y)= [ψ (x+yi)]uptoanadditive wardtoestabl3ishthatthemeaninversemagnificationis(1 κ )2. constant(Anc2005c).Here, [f]istheℜrealcpartofacomplex-valued − ⋆ ℜ Unliketheprecedingtwocases,thesystemdescribedbylensequa- function f.Consequently, wefindthat thereal potential up toan tion (11) does not have the lens at the centre, but the centre of additiveconstantfortheregularlensinglatticeisgivenby thesystemcorrespondstothelocationonthelensplanesuchthat d′(0)=h(0)=0.However,wenotethatthisisdeliberatelychosen ln|σ(z;g2,0)| (cid:3) feonrcmeoatfhtehmeahtoicnaelyscimomplbicgitryid,afnrdomdotehsensoqtuiamreployratnryiainngtruilnasricgrdiidff.eIrn- ψ(x,y)=llnn|σ(zσ;0(z,;g03,)g|3) 7△ (14) proeansrpttihececuthlatoorn,ewaynecyonlmoetnbestghlroeidc1a2eti0xo◦pn-er,roiwetanhtciiocehnsazalecsrtyoumaslmhlyeeatirrmyfproolfimetshtehthesayrtsetamenmayinwleiinntghs wherez=x+(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)σy(iza;n0d,−σg(3z/;2g72),g(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)3)istheWeierstrasssigmafunction (Abramowitz&Stegun1972),whichcanbedefinedastheanti-log- pointmasses.Inotherwords,forallthreecasesofthepointmasses derivativeoftheWeierstrasszetafunction,thatis,(d/dz)lnσ(z)= ontheregulargrid,thetotalshearfromalltheadjacentlensesex- ζ(z).Here,wehavealsousedthepropertyofthecomplexlogarithm actlycancelsoutatanylensposition. functionsuchthat [lng] = ln g wheregisanycomplex-valued ℜ | | function. Note that an additional linear term may be required to beaddedintheexpressionforthepotentialifthecentreofthesys- 2.2 lensingpotential temischosendifferentlyfromthoseofthelensequationsdiscussed Whilemostofthelensingpropertiesofthelatticelenscanbestud- earlier.Fig.1showscontourplotsfortheequipotentiallinesforthe iedusingthelensequations,itisstillusefultohaveanexpression potentialgiveninequations(14). c 2007RAS,MNRAS000,1–19 (cid:13) Lensingbyregularpoint-masslattices 5 Figure4.Causticnetworksforasquarelatticelenswithω=1(κ⋆ 0.785;topleft),ω=0.8(κ⋆ 1.227;topright),ω=0.77(κ⋆ 1.325;bottomleft),and ≈ ≈ ≈ ω=0.73(κ⋆ 1.474;bottomright). ≈ 3 CRITICALCURVESANDCAUSTICS Hence, the topography of the critical curves may be studied from the lines of constant values of ℘(z;1,0) (square grid) and Asusual,theJacobiandeterminantofequation(6),(8)or(11) | | ℘(z;0,1) (triangular and hexagonal grid). In particular, the criti- | | 1 ℘(z;g ,0)2 (cid:3) calcurvesfor thesquare (ortriangular) lenslatticewiththehalf- −| 2 | periodωarebasically thecurves of constant ℘(z;1,0) = ω2/ω2 Jisth=e1in−ve|rfs′e(zm)|2ag=nifi11ca−−ti||o℘hn((zzo;;gf03;t)hg|2e3)i|n2div7△idualimageatthelimi(t15o)f tc[houerrv|o℘ets(hzeo;rf0h,c1aon)n|ds=,tatωnhto2s|/℘eω(f22zo];r0ret,hs1ce)a|hl=eodnωeby2y/c(ao√mfa3bcωtlo)artstocicfaelω|ea/drωeb0yfo(ou√rn3|ωdω/f/ωrωo2m2).aOtnhdne0 rotatedby 30 .Since℘-functionsarewell-definedellipticfunc- thepointsource.Inaddition,thecriticalcurvescanbefoundfrom ◦ − f (z) = 1 or equivalently solving f (z) = e 2iφ for z(φ) withφ tions, for all three cases the critical curves are an infinite set of ′ ′ − | | − nonintersectingclosedcurvesexceptwhenωassumesaparticular beingarealparameter(Witt1990;Wambsganssetal.1992;An& valueatwhichthecurvesaregivenbyinfinitelyextendedboundary Evans2006),andthecausticsfromtheirimagesunderthemapping linesthatdividethelensplane. given by the lens equation. Since f (z) is elliptic (hence, biperi- ′ odic),thereexistsaninfinitesetofcriticalcurvesandcausticsal- thoughtheycanallberecoveredfromthediscretegridtranslation 3.1 squarelattice ofa‘unit’curve. Thehomogeneityrelationof℘-functionindicatesthat Fπo/r(2tωhe2)squ0a.4re57g]r,iedacchasoef,tihfeωcrit<ica(lωcu0/rv√e2s)is≈cen1t.r3e1d1at[tih.ee,cκe⋆ntr>e ω 2 ω 0 ≈ ℘(z;g ,0)= 0 ℘ 0 z;1,0 ; ofasquarecelldefinedbythefouradjacentlenses(i.e.,kω+ipω 2 (cid:18)ω(cid:19) (cid:16)ω (cid:17) (16) wherekand pareoddintegers).Anyimageintheregion‘within’ ℘(z;0,g )= ω2 2℘ ω2 z;0,1 thecriticalcurveshaspositiveparityandviceversa.Ontheother 3 (cid:18)ω(cid:19) (cid:16)ω (cid:17) hand,ifω > 2−1/2ω0 [κ⋆ < π/(2ω20)],thecriticalcurvesarecen- andfromequation(A2)that tredaround eachlens location. For thiscase, the images‘within’ thecriticalcurvesnowhavenegativeparity.Finally,ifω=2 1/2ω − 0 h(z;g )= e−2πi/3 ω2 2 ℘ eπi/6ω2z;0,1 −2. (17) [κ⋆ =π/(2ω20)],thecriticalcurvesaregivenbytwoinfinitesetsof 3 3 (cid:18)ω(cid:19) " (cid:18) √3 ω (cid:19)# paralleldiagonallines,andthewholelensplaneisevenlydivided c 2007RAS,MNRAS000,1–19 (cid:13) 6 J. An Figure5.Criticalcurves(dottedlines)andcausticnetworks(solidlines)for Figure6.SameasFig.5exceptforω=1.15[κ⋆=π/(2√3ω2) 0.685]. lensesonaequilateraltriangularlatticewithω=1.25[κ⋆=π/(2√3ω2) ≈ ≈ 0.580]. curves2ω(κ 1)increasewhiletheirsizescontinuallydecrease. ⋆− Consequently,thenetworkeventuallyreducesanarrayofseparate intochecker-liketiledregionsaccordingtotheparityoftheimages tilesoftheunitcaustics,whichexhibitnooverlapbetweenneigh- inthem. boring caustics. The critical value at which the sizes of the unit Forω , 2−1/2ω0,thelensequationmapseachconnectedand causticscoincide withtheseparations between them isfound ap- closed ‘unit’criticalcurve toacausticthatisalsoconnected and proximatelytobeω 0.748(κ 1.405). closed. Fig. 2 shows the critical curves and the caustic network One physical in≈terpretatio⋆n≈of the caustics in gravitational for ω = 1.4. Like the critical curves, the whole caustic network lensing is that they are boundaries between regions inthe source is composed of the lattice arrangement of ‘unit’ caustic curves, plane that produce different number of images. In addition, it is but the unit curve exhibits self-crossing in contrast to the critical alsoknownthatapairofimagesofoppositeparityappearsordis- curves.Theshapeoftheunitcausticcurve,whichisroughlysim- appearswheneverthesourcecrossesthecaustics,andthatthenum- ilartoavariant of (irregular)octagram (i.e., 8/3-star-polygon3), ber of positive parityimages, if the source liesoutside any caus- { } isactuallygeneric for thecasethat ω > 2−1/2ω0. The‘centre’ of tics, is one or nil depending on the characteristics of the system. eachunitcausticcurveformsasimilarsquaregridtothelenslat- Hence,themaximumnumberofpositiveparityimagescanbede- ticebutthesidelengthofthegridonthesourceplaneisgivenby terminedfromexaminationofthecausticsnetwork.Forexample, 2ω(1−κ⋆).Asωgetslarger(orequivalentlyκ⋆ getssmaller),the the quasi-octagram caustics for ω > 2−1/2ω0 allow at most four sizeof thecausticsshrinksandtheyasymptoticallyreduce tode- positiveparityimagesintheregionaroundthecentresofcaustics. generatepoints.Forω < 2−1/2ω0,thecausticnetworkcanstillbe It is easy to convince oneself that the minimum number of posi- understoodasthelatticearrangement of‘unit’causticcurves,the tiveparityimagesforthecorresponding scenarioisonewhenthe shapeofwhichissimplydescribedasbeingadiagonallystretched source liesoutside any caustic. As for the case ω 6 2 1/2ω , the − 0 square.Again,theseparationbetweenthecentresoftheunitcaus- examinationofthecausticnetworkleadsustoconcludethat,while ticsaregivenby2ω|1−κ⋆|.However,thisissmallerthanthe‘size’ themaximumnumberofthepositiveimagesapproachesinfinityat loafpepaicnhguanmitocnagusthtiecnifeκig⋆h∼bo1rinsogtchaautstthicesn,ewtwhoicrhklceaandsexthoibaitroavtheerr- 0κ⋆.80=31.(ωi.e..,ω1.0=39π.1I/n2/o2th≈er0w.8o8r6d)s,,ittheisngurmeabteerrothfapnosfoituivreopnalyritiyf complexnetwork(albeitregularthankstothesymmetryofthesys- imagesisboundedbyfourprovidedthatω&1.039(κ .0.727)or ⋆ tem).Thecriticalcurvesandcausticnetworkforω=1.2areshown thatω.0.803(κ &1.217).Thesourcepositionswithnopositive ⋆ inFig.3.Thetransitionofthecausticnetworkoverω = 2−1/2ω0 parityimagestarttoexistifω.0.776orequivalentlyκ⋆ &1.039. canbeunderstoodasthecontactbetween(theverticesof)adjacent Finally, if ω . 0.748 (κ & 1.405), there is no overlap between ⋆ quasi-octagram caustics leading to their connection. The further neighboringcaustics,andthusthenumberofpositiveparityimages evolution of thecaustic topography for ω < 2−1/2ω0 ispresented isoneifthesourceliesinthecaustic,andnilifotherwise. inFig.4.Sincetheseparationbetweentheunitcausticsreducesto nilasκ 1, thenetwork isthedensest around κ 1.Onthe ⋆ → ⋆ ≈ otherhand,asκ⋆increasespastunity,theseparationsbetweenunit 3.2 equilateraltriangularlattice Thelenses onanequilateral triangular latticeproduce thecritical curvesandthecausticnetworkinabasicallyconsistentpatternas 3 A{p/q}-star-polygon,withp,qpositiveintegers,isafigureformedbyconnectingwithstraight thesquarelatticelensalthoughmanydetailsarequitedifferent.For lineseveryq-thpointoutofregularlyspaced ppointslyingonacircumference.Seehttp: //mathworld.wolfram.com/StarPolygon.html. triangularlattices,theunitcriticalcurvesforω<(ω2/√32)≈1.214 c 2007RAS,MNRAS000,1–19 (cid:13) Lensingbyregularpoint-masslattices 7 Figure7.Criticalcurves(dottedlines)andcausticnetworks(solidlines)for Figure8.SameasFig.7exceptforω=1.3[κ⋆=π/(3√3ω2) 0.358]. lensesonahoneycomblatticewithω=1.45[κ⋆=π/(3√3ω2) 0.286]. ≈ ≈ 3.3 regularhexagonallattice [κ >π/(21/331/2ω2) 0.615]arelocatedineachoftheequilateral Inqualitativeterms,thecriticalcurvesofthehoneycomblatticeare ⋆ 2 ≈ triangularcelldefinedbythreeadjacentlenses.Ontheotherhand, basicallythe‘dual’ofthatofthetriangularcase.Thecriticalcurves those for ω > 2 1/3ω [κ > π/(21/331/2ω2)] are centred at each forω < (22/33 1/2ω ) 1.402[κ > π/(31/224/3ω2) 0.308]are − 2 ⋆ 2 − 2 ≈ ⋆ 2 ≈ lensposition.Notethat,incontrasttothesquarelattice,thenumber analogoustothoseforthetriangularlatticewithω > 2 1/3ω and − 2 ofpositionswithzeroshearistwicethatofthepoles(i.e.,thelens locatedwithineachhexagonalcellcentredatthepositionsofnull positions)sothatthereistwo-to-onecorrespondence betweenthe shear. On the other hand, if ω > 22/33 1/2ω , the critical curves − 2 unitcriticalcurvesoftheformerandthatofthelatter.Atthecrit- arelikethoseforthetriangularlatticewithω < 2 1/3ω andcen- − 2 icalvaluethatω = 2 1/3ω [κ = π/(21/331/2ω2)],thelensplane tredaroundeachlenspositions.Thedualcharacteristicfurtherin- − 2 ⋆ 2 isdividedintotwoclassesbythecriticalcurves:triangularregions dicates that there are twice more poles than the locations of null aroundthenullshearpositionswithinwhichtheimageshaveposi- shear(whichareinfactfourth-orderzeros),andthusthereisone- tiveparityandhexagonalregionsaroundthelensescorresponding to-two correspondence between unit critical curves of the former totheimagesofnegativeparity. tothelatter(formally,the‘windingnumber’oftheformeristwice For ω > 2 1/3ω , the shape of the unit caustics may be de- thatofthelatter). − 2 scribedassnowflake-likeoranirregularquasi-dodecagram( 12/5- On the other hand, the caustics of a honeycomb lattice are { } star-polygon). Each unit caustic is arranged in the same equilat- qualitativelydistinctfromthoseofatriangularlattice.Fig.7shows eral triangular grid with side length of 2ω(1 κ ) to form the anexampleforasparsehoneycomb latticewithω = 1.45,which − ⋆ wholecausticnetwork.Fig.5showsanexampleofthis,thecaus- is in fact archetypal for ω > 22/33 1/2ω . The unit caustics are − 2 ticnetworkforω = 1.25. Themaximumnumber ofpositivepar- givenbyasix-sidedclosedself-intersectingcurve,whichmaybe ity images associated with these unit caustics is found to be six. labeledabi-triangle,andtheyarearrangedinasimilarhexagonal If ω < 2 1/3ω , the unit caustics are given by stretched triangu- pattern as the latticeof thelens with side length 2ω(1 κ ). We − 2 − ⋆ larcells,andthewholenetworkisgivenbytilingthesourceplane alsonotethatthemaximumnumber of positiveparityimagesas- with these cells, which may overlap with one another. Here, the sociatedwiththesecausticsisthree.Foradensehoneycomb lens centres of each cell actually form a hexagonal (honeycomb) grid lattice(ω < 22/33 1/2ω ), theunit causticsare found tobe inthe − 2 with side length of 2ω 1 κ /√3, which is in fact the mapped shapeofastretchedhexagon(anexampleofwhichforω = 1.3is | − ⋆| image of the dual grid to the lens lattice in the lens plane, and giveninFig.8)andarrangedinatriangulargridwithsidelengthof the adjacent cells alternate their orientation with their symmetry 2√3ω 1 κ .Theyalsoexhibittrendsofvaryingoverlapsimilar | − ⋆| maintained.Anexampleofthesecausticsnetworksispresentedin tothesquareortriangularlattice,asκ getslargerpasttheunity.We ⋆ Fig. 6 for ω = 1.15. Analogous tothe case of the square lattice, findthatthecriticalvaluefornooverlappingcausticsforthehon- denseoverlappingofunitcausticsleadstoverycomplexnetwork eycomblatticeisapproximatelygivenbyω 0.622(κ 1.565). ≈ ⋆≈ around κ 1. It is found that that the critical value that corre- Despitetheirvariety,wealsonoticesomecommoncharacter- ⋆ ∼ spondstonooverlapbetweenadjacentcausticsisaboutω 0.791 isticsofthecausticnetworksofthreekindsofregularlensinglat- ≈ (κ 1.449).Thetriangularlenslatticewithsidelengthlessthan tices.Forexample,withanopticaldepththatisgreaterthanacer- ⋆ ≈ thisvalue(henceadenserlattice)producestriangularunitcaustics taincriticalvalue,theunitcurvesaremoreorlessinasimilarshape arrangedonahoneycombgridwithnooverlap,andthenumberof asthelensinglatticeexceptforthefactthattheyaredistortedina positiveparityimagesiseitheroneornildependingontheposition waystretchingtheveticesradiallyoutward.Theydenselyoverlap ofthesourcerelativetothecaustics. withoneanotheraroundκ 1.However,asκ getslargerpastthe ⋆∼ ⋆ c 2007RAS,MNRAS000,1–19 (cid:13) 8 J. An Figure9.Top:meannumberofpositiveparityimageforlensesonasquare lattice(solidline),atriangularlattice(short-dashedline),ahexagonallattice (long-dashedline),andfortherandomstarfield(dottedline).Alsoplotted indot-dashedlinesisthemeantotalmagnificationhµi=(1−κ⋆)−2.Bottom: meannumbernormalizedbythemeantotalmagnification.Thelinetypes arethesame. Figure10.Distributionoflensingshear,γ.Thelinetypesarethesameas unity,theirsizeshrinkwhereastheseparationsamongthemgrow, Fig.9. and so the networks eventually reduce to a simple array of tiles. Likewise,towardstheloweropticaldepthsbelowthecriticalvalue, thenetworksarecomprisedofstar-shapedunitcurveswhoseexact itive images reside and thus there are no positive images formed characteristicsarerelatedtomeetingpatternofthelensinglattice ‘outside’ of any caustics. On the other hand, for ‘sparse’ lattices atthecommonvertexpoints. thatproducethenetworkcomposedofagridsetofself-intersecting ‘star-shaped’curves,theself-crossingoftheunitstructuresrequires someconsiderationonthemeaningof‘area’thatproperlyaccounts fortheimagemultiplicity.Wearguethatthesolutionisrathersim- 4 MEANNUMBEROFPOSITIVEPARITYIMAGES ple.Oncetheareacalculationisreducedtothelineintegralalong The area in the caustic accounting for its multiplicitycan be un- itsboundary(seee.g.,An2005;An&Evans2006)usingthefun- derstoodasthecoveringfractionofthesourceplanebytheregion damentaltheoremofmultivariatecalculus4,thecorrespondingline with‘extraimage’pairsonceitisproperlynormalized.Ingeneral, integralalongthecausticactuallyresultsintheareacountedwith however,thenormalizationisill-definedforlocalizedlenses.This multiplicity. That is to say, they can be considered as the ‘area- difficultyisamelioratedwiththeintroductionofthelatticeofpoint excess’ or the ‘over-covering factor’ of the source positions that massessincethewholeplaneisnowevenlydividedintoidentical producepositiveparityimages.Unlikethepreviouscase,thenor- cells.Withtheregularlensinglattice,theareaundertheunitcaustic malizationcellinthelensplaneistheunitcellof thedual ofthe canbeproperlynormalizedwithrespecttotheareaofthemapped lensinglattice,i.e.,thecelldefinedbytheadjacentnullshearloca- image (on the source plane) of the unit cell comprising the lens- tions(whichisinfactthecentreofeachcellofthelensinglattice). inglattice.Theresult,ifthemultiplicityisproperlyaccountedfor, The corresponding critical curves lie completely within this unit isactuallythemeannumberofextraimagepairs,whichhasbeen normalizationcellandfurthermoreenclosetheregionsoftheim- calculated for the case of the random star field using techniques ageplanewherenegativeparityimagesreside.Subsequently,there fromtheprobabilitytheory(Wambsganssetal.1992;Granotetal. isatleastonepositiveparityimageforanygivensourceposition 2003). – that is, the source ‘outside’ caustics forms one positive parity For‘dense’latticesthatproducenonintersecting(butpossibly image. Asa result, the mean number of extra image pairs isless overlapping) unit caustics, the relevant area under the unit caus- thanthenumberofpositiveparityimagesbyone.Wenotethatthe ticsisstraightforwardtocalculatesince,despiteoverlapping,they meannumberofpositiveparityimagesactuallyvariescontinuously aresimplyreduced toaset of closedloops. Then,theareaunder overthecriticalvaluedividingthedenselatticefromthesparseone one such loop normalized to the area of a unit cell in the source whilethemeannumberofextraimagepairsjumpsbyone.Infact, plane, Aω(1−κ⋆)2 where Aω is the area under the unit cell de- finedbytheadjacentlenspositionsinthelensplane,isthemean numberbothofextraimagepairsandofpositiveparityimagesfor 4 ThetheoremisusuallyknownasGreen’stheoreminelementarymultivariatecalculuswhenrefer- a given source position because the corresponding critical curves ringtotherelationbetweentwo-dimensionalintegraloveradomainintwo-dimensionalspaceand one-dimensionalintegralalongtheboundaryofthedomain.Thisisalsoaspecialcaseofso-called intheimageplanecompletely enclosetheregionwhereanypos- (generalized)Stoke’stheorem,thenameunderwhichthetheoremisusuallyknownas. c 2007RAS,MNRAS000,1–19 (cid:13) Lensingbyregularpoint-masslattices 9 designatingacertainimagepairasanextrainvolvessomedegrees valueofκ insteadofκ canbechosenasthenormalizationandthe ⋆ 0 ofarbitrarinesswhereascountingthenumberofpositiveparityim- resultingsimilarityrelationisvalidforallthreeregularlattices. agesiswell-defined.Forthesereasons,henceforth,weshalldiscuss The resulting distributions for all three regular lattices are themeannumberofpositiveparityimagesexclusively. showninFig.10togetherwiththesamedistributionfortherandom Fig.9showstheresultingmeannumberofpositiveparityim- starfield(Nityananda&Ostriker1984; Schneider 1987; Wambs- agesasafunctionofκ .Comparisontothecaseoftherandomstar ganssetal.1992),whichisgivenby ⋆ field(withzero external shear) (Wambsganss et al. 1992) reveals κ γ 1 γ some significant differences between the regular lensing lattices Pran(γ|κ⋆)= (κ2 +⋆γ2)3/2 = κ2 1+(γ/κ )2 3/2. (22) andtherandomfield,particularlyinthebehaviourtowardκ . ⋆ ⋆ ⋆ ⋆ →∞ Whilethemeannumberfallsoffasκ⋆−4fortherandomstarfieldand Unliketherandomfieldcase,w(cid:2)efindthatth(cid:3)ereisnonzeroproba- triangularlensinglattices(albeitthenormalizationfortherandom bilitythatthelineofsightexperiencesnonetshearforthesquare fieldbeinglargerbyafactorof 4),itsbehavioursforthelenseson latticelens.Bycontrast,thesamedistributionlinearlytendstozero ∼ squarelatticesandhoneycomblatticesarecharacterizedbyslower asγ 0forthetriangularlattice(andalsotherandomfield)andit fsaolml-oewffshagtivceonunbtyer∼intuκ⋆−it3ivaen.dT∼hatκ⋆−i2s.,5,inretshpeecrtaivnedloym. Asttafirrsfite,ltdh,isthies rdeivlaetre→gdestoatsh∼efγa−c1t/t2hfaotrtthheezheoronseyocfo℘m(zb;lgat,ti0c)e,.℘T(hz;e0se,gbe)haanvdiohu(rzs;gare) 2 3 3 distance stotheclosestneighboring lensisdistributedaccording aresecond,first,andfourth-order,respectively;thatistosay,their to Taylor-seriesexpansionsattheirzeroesaregivenby P(s)ds=2πσse−πσs2ds (18) ℘(z;g2,0)≃−g42 (z−z0)2+O(|z−z0|6) Twhheermeeσan=anκd⋆/thπeivsatrhiaenmceeaarnesguirvfeancebynuωm=be1r/d(2en√sσity)aonfdth[(e4/sπta)rs. ℘(z;0,g3)≃ig13/2(z−z0)+O(|z−z0|4) (23) g 1]ω2 =[1−(π/4)]κ⋆−1.Inotherwords,asthedensityofstellarfiel−d h(z;g3)≃ 273 z4+O(z10) increases, the random fluctuations around the mean decrease and wherez isazeroofthecorrespondingfunction(notethatzeroofh onewouldexpect thattherandomstellarfieldmayapproach toa 0 regulTarhfieealcdt.ualresult canbeunderstood usingthedistributionof iisndaitcza0te=s0th).aFtoArω|γ|0≪γP1(,γth′)eddγi′sc≃ussπio|zn−inzt0h|2ewprheecreedizn0gipsatrhaegrnauplhl the(internal)lensingshearγduetostarsexperiencedbythegiven shearpositionandRztracesthelocationsatwhichtheshearisgiven byγ.Bynoticingtheshearisgivenby ℘(z;g ,0), ℘(z;0,g ) and lineof sight towards alensed image. For afieldof point masses, | 2 | | 3 | h(z;g ) forsquare,triangular,andhoneycomblattices,wecanthen oncethisdistributionisknown,themeannumberofpositiveparity | 3 | findtheleading-orderbehavioursforthesheardistributionsasγ imagesisrecoveredthrough(Wambsganssetal.1992) → 0fromtheaboveTaylorexpansions5; 1 hN+i=hµiZ0 (1−γ2)P(γ)dγ. (19) P(cid:3)(γ|κ⋆)≃ κ1⋆ 2Γ681π/44 +O(cid:20)(cid:16)κγ⋆(cid:17)2(cid:21)≈ 0.2κ0⋆88 Withlensesonaregularlattice,wehaveshownthattheshearatthe 1 211π9 γ γ 4 0.2326 gtiiovne,nainmdahgeenpclea,ntehelodciasttiroinbuistiogniveonnbthyethwehWoleeieimrstargaesspelallnipetricedfuucnecs- P△(γ|κ⋆)≃ κ⋆ 33/2Γ118/3 κ⋆ +O(cid:20)(cid:16)κ⋆(cid:17) (cid:21)≈ κ⋆2 γ (24) to that on the unit cell in the image plane. Then, the cumulative 1 24π9/2 κ 1/2 γ 0.1705 dwihsterriebuthtieonshoefasrhiesasrmisalfloeurntdhabnytthheegfirvaecntiovnaloufe,araenadincotnhseeuqnuietnctelyl,l P7(γ|κ⋆)≃ κ⋆ 33/4Γ91/3 γ⋆! +O(cid:16)κ⋆(cid:17)≈ κ⋆1/2γ1/2, itsderivativegivesthedifferentialdistributionoftheshear.Further- whichagreewiththenumericalresultsshowninFig.10(notethat more,thehomogeneityrelationof℘-functionimpliesthesimilarity κ = π for Fig. 10). In other words, the higher-than-linear-order ⋆ relationbetweenthedistributioncorrespondingtodifferentω.For (i.e.,degenerate) zerosoftheshear leadtononzerofiniteoreven example,withasquarelensinglattice,wehavethat divergent probability for the zero net shear. Physically speaking, accidental canceling of the net shear in the random star field is γ 1 Z0 P(γ′|κ⋆)dγ′= AωAωh|℘(z;g2,0)|6γi mmeunchtolfesthselilkeenlsyest,ofoorcwcuhricthhatnheirnemthaeyhbigehsliygnsiyfimcamnettcrhicanacrerasntghea-t 1 ω ω2 π γ theimageislocatedatanexactlybalancedpositionwherethenet = Aω0Aω0"(cid:12)(cid:12)℘(cid:16)ω0 z;1,0(cid:17)(cid:12)(cid:12)6 ω20 γ= 4ω20κ⋆# shearisnull.Thisisbasicallythereasonforslowerfall-offsofthe κ0γ/κ⋆ (cid:12)(cid:12)(cid:12) (cid:12)(cid:12)(cid:12) meannumber ofpositiveparityimagesindensesquareorhoney- =Z0 P(γ′|κ0)dγ′ (20) cliokmebPla(tγti|κc⋆e)s.≃FroSmκ⋆−e1(qγu/aκt⋆io)nα(a2s1γ),/iκf⋆th→esh0ewarhdeirsetrSibuistioancboenhstaavnets, α=(2/n) 1> 1istheasymptoticpowerindexforthesheardis- − − P(γ|κ⋆)= A1ω0 ddγ′Aω0(cid:20)(cid:12)(cid:12)(cid:12)℘(cid:16)ωω0 z;1,0(cid:17)(cid:12)(cid:12)(cid:12)6 κκ⋆0 γ′(cid:21)(cid:12)(cid:12)(cid:12)(cid:12)γ′=γ t(r1i9b)utaisonκ⋆an→dn∞isretdhuecoersdteorofthezerosoftheshear,thenequation = A1ω0 ddγgddgAω(cid:12)0(cid:20)(cid:12)(cid:12)℘(cid:16)ωω0 z;1,0(cid:12)(cid:17)(cid:12)(cid:12)6g(cid:21)(cid:12)(cid:12)(cid:12)g=(cid:12)(cid:12)κ0γ/κ⋆ (21) hN+i= (1−1κ⋆)2 Z01dγ(1−γ2)P(γ|κ⋆) (25) κ κ (cid:12)(cid:12) (cid:12)(cid:12) (cid:12)(cid:12) S 1 = κ⋆0 P(cid:18)κ⋆0 γ κ0(cid:19) (cid:12) ≃ (κ⋆−1)2κ⋆α+1 Z0 dγ(1−γ2)γα. where κ0 = π/(2ω0)2 and Aω[|℘(z;g2,0)| 6 γ] is thearea of the unitcellwhere|℘(z;g2,0)|6γforasquarelensinglatticewithside 5 Itisactuallyeasiertoworkoutfromtheseriesfortheinversefunctions,whichareinfacthyper- lengthofω(seeeq.[5]forg2asafunctionofωandκ⋆).Infact,any geometricfunction(seealsoAppendix). c 2007RAS,MNRAS000,1–19 (cid:13) 10 J. An Figure11.Locationsoflensedimageswithasquarelensinglattice withω = 1.35(κ⋆ 0.431).Thelocations ofthesourcewithrespecttothecaustic ≈ networksareindicatedinthetoprightpanel.Intheremainingthreepanels,thelocationsoftheindividuallensedimagesaremarkedbyopencircles(positive parityimages)orcloseddots(negativeparityimages).Alsodrawnindottedlinesarethecriticalcurvesandthelocationsofthelensesaremarkedbycrosses. ShNin+cie∼theκ⋆−la(αs+t3i)natesgκra⋆li→sa∞fin,iwtehcicohnsitsanctofnosrisαten>t−w1it,hwtehefinrdestuhlatst κb⋆e/hγav2i[oiu.er.,oPf(tth)e≃sht−e2arasdits→trib∞ut]i.oSnuapspγo/seκ⋆th→atp∞owtoerb-seePries(γe|xκ⋆p)an≃- foundearlier(Fig.9). sionofP(t)att=∞isgivenbyP(t)≃t−2(1+ ∞n=1cnt−pn)(andit Infact,thesheardistributionsforregularlensinglatticesare isuniformlyconvergentinanintervalcontainingPt= ).Then,for quitedistinctfromthatoftherandomfield,anditissomewhatre- κ 0,wefindthat(assumingp,1/nforanypositi∞veintegern); ⋆ → markablethatthebehaviours of N aregenerally similarfor all + csiacsaelslyatalneainsttefgorrastimonal(loκr⋆c.oTnhviosliusthiroenla)iteodftthoethsheefaarctdtihstartibhuNt+ioini,sabnad- Z01/κ⋆dtP(t)≃1−κ⋆−Xn∞=1 cpnnκ⋆p+n+11 therefore any ‘roughness’ in the shear distribution is ‘smoothed’ outin N+ .Inparticular,wenotethatthesimilarityrelationofthe 1/κ⋆dtt2 (t) 1 P ∞ cnκ⋆pn−1 (27) sheardhistriibutionisessentiallytheresultofthescaleinvarianceand Z0 P ≃ κ⋆ − 0−Xn=1 pn−1 sisoashfuonucldtiobneogfenγe/rκic.,aTnhdenc,oPnse(qγu|κe⋆n)tl=yκw⋆−e1Pha(γv/eκ⋆)whereP(γ/κ⋆) hN+i 1 2κ +P κ2 + ∞ 2cnκ⋆pn+1 ⋆ µ ≃ − ⋆ 0 ⋆ (pn 1)(pn+1) h i Xn=1 − 1 1/κ⋆ hN+i= (1−κ⋆)2 Z0 dt(1−κ⋆2t2)P(t). (26) w(thhaetreis,Pf0arisfraomfintite=con)stbaenhtatvhiaoturwoefakly(t)d,eapnedndthsuosnNthe/glµobal + ∞ P h i h i ≃ Furthermore, since theshear for thelinesof sight that passsuffi- 1−2κ⋆+O(κ⋆2)andhN+i=1+O(κ⋆2),regardlessofthehigher-order cientlyclosetoanyofthelensesissimplygivenbyγ d 2where behaviourof (t),providedthat p > 1.Infact,wehave (t) = ≃ − P Pran d istheseparationbetweenthegivenlineof sightandthattothe t(1+t2) 3/2 = t 2(1+t 2) 3/2 sothat p = 2for (t),and also − − − − Pran lens, it isstraightforward to establish that the generic asymptotic thepower-seriesexpansionsof℘-functionsattheirpolesandzeros c 2007RAS,MNRAS000,1–19 (cid:13)

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.