ACCEPTEDFORPUBLICATIONINASTROPHYSICALJOURNALLETTERS PreprinttypesetusingLATEXstyleemulateapjv.6/22/04 THEFUNDAMENTALPLANEFORz=0.8- 0.9CLUSTERGALAXIES INGERJØRGENSEN1,KRISTINCHIBOUCAS1,KATHLEENFLINT1,2,MARCELBERGMANN3,JORDIBARR4,ROGERDAVIES4 AcceptedforpublicationinAstrophysicalJournalLetters ABSTRACT WepresenttheFundamentalPlane(FP)for38early-typegalaxiesinthetworichgalaxyclustersRXJ0152.7– 1357(z=0.83)andRXJ1226.9+3332(z=0.89),reachingalimitingmagnitudeofM =- 19.8magintherest B frameoftheclusters. WhilethezeropointoffsetoftheFPforthesehighredshiftclustersrelativetoourlow redshift sample is consistent with passive evolution with a formation redshift of z ≈3.2, the FP for the form high redshift clusters is not only shifted as expected for a mass-independentz , but rotated relative to the 6 form 0 low redshift sample. Expressed as a relation between the galaxy masses and the mass-to-light ratios the FP 0 is significantly steeper for the high redshift clusters than found at low redshift. We interpret this as a mass 2 dependencyofthestarformationhistory,ashasbeensuggestedbyotherrecentstudies. Thelowmassgalaxies n (1010.3M⊙) have experienced star formation as recently as z≈1.35 (1.5 Gyr prior to their look back time), a whilegalaxieswithmasseslargerthan1011.3M⊙ hadtheirlastmajorstarformationepisodeatz>4.5. J Subject headings: galaxies: clusters: individual: RXJ0152.7–1357 – galaxies: clusters: individual: 8 RXJ1226.9+3332–galaxies:evolution–galaxies:stellarcontent. 1 1 1. INTRODUCTION Our study of these two clusters is part of the Gemini/HST v TheFundamentalPlane (FP)forelliptical(E)andlenticu- Galaxy Cluster Project, which is described in detail in Jør- 3 lar (S0) galaxies is a key scaling relation, which relates the gensen et al. (2005). We adopt a ΛCDM cosmology with 0 effective radii, the mean surface brightnesses and the veloc- H0=70kms- 1Mpc- 1,ΩM =0.3,andΩΛ=0.7. 4 itydispersionsinarelationlinearinlog-space(e.g.,Dressler 1 etal.1987;Djorgovski&Davis1987;Jørgensenetal.1996, 2. OBSERVATIONALDATA 0 hereafterJFK1996). The FP can be interpretedas a relation Spectroscopy for RXJ0152.7–1357 and RXJ1226.9+3332 6 betweenthegalaxymassesandtheirmass-to-light(M/L)ra- were obtained with the Gemini Multi-Object Spectrograph 0 tios. For low redshift cluster galaxies the FP has very low (GMOS-N, Hook et al. 2004) at Gemini North. The data / h internalscatter, e.g.JFK1996. Itisthereforea powerfultool forRXJ0152.7–1357arepublishedinJørgensenetal.(2005). p for studying the evolution of the M/L ratio as a function of ThereductionoftheRXJ1226.9+3332spectroscopywasdone - redshift(e.g., Jørgensen et al. 1999; Kelson et al. 2000; van using similar techniques, with suitable changes to take into o r de Ven et al. 2003; Gebhardtet al. 2003; Wuytset al. 2004; account the use of the nod-and-shuffle mode of GMOS-N t Treu et al. 2005; Ziegler et al. 2005). These authorsall find (Jørgensenet al. in prep.). We use Hubble Space Telescope s a that the FP at z=0.2–1.0 is consistent with passive evolution (HST)archivedataofthetwoclustersobtainedwiththeAd- : ofthestellarpopulationsofthegalaxies,generallywithafor- vanced Camera for Surveys (ACS). In this paper we use ef- v mation redshift z >2. Most previous studies of the FP fectiveradii,r , andmeansurfacebrightnesses,hIi ,derived i form e e X at z=0.2–1.0 cover fairly small samples of galaxies in each fromeitherF775WorF814Wobservations,calibratedtorest- r clusterandarelimitedto a narrowrangeinluminositiesand frame B-band, see Chiboucas et al. (in prep.) for details. a thereforeinmasses,makingitverydifficulttodetectpossible The GALFIT program (Peng et al. 2002) was used to de- differences in the FP slope. A few recent studies indicated termine r and hIi . We fit the cluster members with Sérsic e e a steepening of the FP slope for z∼ 1 galaxies (di Serego (1968) and r1/4 profiles. The combination which enters the Alighieri et al. 2005; van der Wel et al. 2005; Holden et al. FP,logr +βloghIi (β=0.7-0.8),differsverylittleforthetwo 2005). These studies and studies of the K-band luminosity choiceseof profiles.e In the following we use the parameters function(Toftetal. 2004)andtheredsequence(deLuciaet fromr1/4-fitsfor consistencywith ourlow redshiftcompari- al. 2004) at z≈0.8- 1.2 suggest a mass dependency of the son data. Noneof the mainconclusionsof this paperwould formationepoch. change had we chosen to use the Sérsic fits. Masses of the WepresenttheFPfortwogalaxyclustersRXJ0152.7–1357 galaxiesarederivedasMass=5σ2r G- 1. e at z=0.83 and RXJ1226.9+3332 at z=0.89. Our samples Our Coma cluster sample serves as the low redshiftrefer- reachapparenti′-bandmagnitudesof 22.5–22.8mag, equiv- encesample(Jørgensen1999).WehaveobtainednewB-band alenttoanabsolutemagnitudeofM =- 19.8magintherest B photometry of this sample with the McDonald Observatory frame of the clusters. No other published samples suitable 0.8-meter telescope and the Primary Focus Camera (Claver for studies of the cluster galaxy FP at z>0.8 go this deep. 1995).Thedatawerereducedinastandardfashionandeffec- tiveparameterswerederivedasdescribedin Jørgensenetal. 1 Gemini Observatory, 670 N. A‘ohoku Pl., Hilo, HI 96720; ijor- (1995). Table 1 summarizes the sample sizes and some key [email protected],[email protected] 2CurrentlyatStateUniversityofNewYorkatStonyBrook,TheReinven- clusterproperties. tionCenter,StonyBrook,NY11794;kathleen.fl[email protected] 3GeminiObservatory,LaSerena,Chile;[email protected] 3. THEFUNDAMENTALPLANEATZ=0.8-0.9 4DepartmentofAstrophysics,UniversityofOxford,KebleRoad,Oxford We first establish the FP for the Coma cluster data. In or- OX13RH,UK;[email protected],[email protected] der to limit the effect of differences in sample selection for 2 FIG.1.—TheFPforRXJ0152.7-0152(orange),RXJ1226.9+3332(red),andComa(blue).Smallersymbols–galaxieswithMass<1010.3M⊙,excludedfrom theanalysis. RXJ1226.9+3332id=711andid=966(withSérsicindexn<1.5)arelabeledandexcludedfromtheanalysis. (a)&(b): Edge-onviewoftheFP. (c):TheFPface-on,fortheComaclustercoefficients.(d):TheFPasMassvs.M/Lratio.Solidbluelineon(a),(b)&(d):FittotheComaclustersample.Solid greenlineon(a)&(d): TheComaclusterfitoffsettothemedianzeropointofthehighredshiftsample. Orange-redlineon(b)&(d): Fittothehighredshift sample.Thefitshownon(b)isnottheoptimalFPforthehighredshiftsample,sinceithasthecoefficientforloghIi fixedat0.82.Dashedlineson(c)and(d): e LuminositylimitsfortheComacluster(blue),andbothredshiftclusters(orange). On(c)thesolidblueandgreenlinesmarkthe“exclusionzones”(Benderet al.1992)fortheComaclusterandhighredshiftsample,respectively,assumingtheslopeandzeropointsasshownon(a). Dashedgreenlineson(d): Models fromThomasetal.(2005),seetextfordiscussion. Internaluncertaintiesareshownasrepresentativeerrorbars. On(c)theinternaluncertaintiesarethesizeof thepoints. minedusingabootstrapmethod,seeJFK1996fordetails.The TABLE1 rmsof the fit is 0.08in logr . The coefficientsare in agree- e GALAXYCLUSTERSANDSAMPLES mentwithotherdeterminationsavailableintheliterature(e.g., JFK1996;Collessetal.2001;Blakesleeetal.2002;Bernardi Cluster Redshift σclustera Ngalaxiesb Nanalysisc Ref.d etal.2003). Coma=Abell1656 0.024 1010kms- 1 116 105 (1) Figure1showstheComaclusterFPface-onaswellastwo RXJ0152.7–1357 0.835 1110kms- 1 29 20 (2) edge-onviews of the relation, with the high redshift sample RXJ1226.7+3332 0.892 1270kms- 1 25 18 (3) overplotted. The FP forthe highredshiftsample is notonly offset from the Coma cluster FP, but appears “steeper”. As aClustervelocitydispersion there is no significant FP zero point difference between the bNumberofgalaxiesobserved two high redshiftclusters we treat the high redshiftgalaxies cNumberofgalaxiesincludedintheanalysis,seetext. d(1)Jørgensen1999;(2)Jørgensenetal.2005;(3)Thispaper asone sample. Derivingthe FP forthe highredshiftsample usingthesametechniqueandsamplecriteriaasfortheComa cluster,wefind theComaclustersampleandthehighredshiftsample,weex- logr =(0.60±0.22)logσ- (0.70±0.06)loghIi +1.13 (2) cludegalaxieswithMass<1010.3M⊙aswellasemissionline e e withanrmsof0.09inlogr . Thedifferenceinthecoefficient galaxies. Thesumoftheabsoluteresidualsperpendicularto e forlogσ betweenEq.1andEq.2is∆α=0.70±0.23,a3σ therelationwasminimized.Wefind detectionofadifferenceintheFPslope. Theinternalscatter logre=(1.30±0.08)logσ- (0.82±0.03)loghIie- 0.443 (1) of the two relations is similar. Figure 1d shows the FP as a relationbetweenthe galaxymassesand the M/L ratios. The wherer isthetheeffectiveradiusinkpc,σ thevelocitydis- persionein kms- 1, and hIi is the surface brightness within fittotheComasample,excludingthelowmassgalaxies,gives e re inL⊙pc- 2. Theuncertaintiesonthecoefficientsaredeter- logM/L=(0.24±0.03)logMass- 1.75 (3) 3 withanrmsof0.09inlogM/L. Fittingthehighredshiftsam- ple,usingthesamemasslimit,gives logM/L=(0.54±0.08)logMass- 5.47 (4) with an rms of 0.14 in logM/L. The internal scatter in logM/L of the two relations are not significantly different. Wefind0.07and0.08fortheComasampleandthehighred- shift sample, respectively. Even with the same mass limit enforced on both samples one might argue that the fits are FIG. 2.—Distributions ofFPcoefficients and the slope, a, ofthe M/L ratio–massrelationfor1000sub-samplesoftheComaclustersample(black) stillaffectedbythedifferenceintheluminositylimit. There- andfor1000bootstrapsamplesofthehighredshiftsample(red,dashed).See fore,wealsofitasub-sampleoftheComasamplelimitedat textfordiscussion. M =- 19.8mag. The coefficientforlogMass is in this case B 0.28±0.06.Thus,thedifferencebetweenthecoefficientsfor the high redshift and the low redshift samples is at the 3σ ∆logM/L = 0.935∆logage (Jørgensen et al. 2005), this level. gives an epoch for the last major star formation episode of z ≈ 3.2. However, the steeper M/L ratio–mass relation 4. POSSIBLESYSTEMATICEFFECTS form foundfor high redshiftclusters comparedto the Coma clus- To test how well we recover input re, hIie and logre+ ter may be due to a difference in the epoch of the last ma- βloghIi (β =0.7-0.8), we simulate HST/ACS observations jorstarformationepisodeasafunctionofgalaxymass. The e of galaxies with Sérsic profiles with n = 0.8- 4.6 and ef- low mass galaxies have experienced the last major star for- fective parameters matching our Coma sample. For n>2, mation episode much more recently than is the case for the the r1/4-fits recover logr with an rms of 0.15. However, highmassgalaxies. Thedifferencebetweenthehighandlow e logr +βloghIi is recovered with an rms scatter of only redshiftsamplesis∆logM/L=- 0.30logMass+3.72,equiv- e e ≈0.02 for β between 0.7 and 0.8. There are no systematic alentto ∆logage=- 0.32logMass+4.0. Thus,forthe low- effects as a function of effective radii or luminosities, see estmassgalaxies(1010.3M⊙)thelastepochofstarformation Chiboucas et al. (in prep.) for details. Simulations of spec- may have been as recent as z ≈1.35. This is only ≈1.5 form tra matching the instrumental resolution, signal-to-noise ra- Gyrpriortowhenthelightthatwenowobservewasemitted tiosandspectralpropertiesofourobservationaldatashowed fromthegalaxiesinthehighredshiftsample. Thereappears that velocity dispersions below the instrumental resolution tobejustenoughtimeforthegalaxiestonolongerhavede- (logσ = 2.06) may be subject to systematic errors as large tectableemissionlinesduetothemassivestarsformedatthat as ±0.15 in logσ (Jørgensen et al. 2005). Excluding from time. Very shortly after the end of the last major star for- theanalysisthefourgalaxiesinthehighredshiftsamplewith mationepisodethesegalaxiesfollowatightFP.Forgalaxies logσ<2.06,wefindaslopefortheM/Lratio–massrelation withMass≈1010.8M⊙ wefindzform≈1.9,whilezform>4.5 of0.47±0.06,whiletheFPcoefficientsarenotsignificantly forgalaxieswithMass>1011.3M⊙. differentfromthosegiveninEq.2. Thomas et al. (2005) used absorption line index data for Finally, we address whether selection effects can be the nearbyE/S0galaxiestoestablishroughstarformationhisto- causeof the differencesin the relationsforthe two samples. ries of the galaxies as a function of their masses. They find Wechoose1000randomsub-samplesof38galaxiesfromthe thatthemostmassivegalaxiesformthemajorityoftheirstars Comasample, roughlymatchingthemassdistributionofthe athighredshift,whilelowermassgalaxiescontinueforming high redshift sample. We confirm the match in mass distri- starsatmuchlaterepochs.Thomasetal.convertvelocitydis- butionsbyusing a Kolmogorov-Smirnovtest. The probabil- persions to galaxy masses using a model dependentrelation itythatthesub-samplesandtherealhighredshiftsampleare thatis inconsistentwith ourdata. We thereforecorrecttheir drawn from the same parent distribution is above 90% for masses to consistency with our data by using the empirical more than 90% of the realizations. For the remainder the relationbetweenourmassestimatesandthemeasuredveloc- probabilityis above70%. We then comparethe fits to these ity dispersions. The lower of the two dashed green lines on sub-samples to the results from bootstrapping the high red- Figure1dshowstheresultbasedonthestarformationhistory shiftsample. FortheFPcoefficientsthesub-samplefitsover- inhighdensityenvironmentsasestablishedbyThomasetal. lapthebootstrapfitsinonly1.6%ofthecases(Fig.2a),while andtheM/LmodelingfromMaraston(2005).Ourdatashow fortheM/Lratio–massrelationtheslopeforthesub-samples slightly less evolution in the M/L ratios between z≈0.8-0.9 overlapwiththebootstrapfitsin3.7%ofthecases(Fig.2b). andthepresentthanpredictedbyThomasetal. However,itis This shows that the FP and the M/L ratio–mass relation for strikingthattheslopeofthepredictedrelationisinagreement thehighredshiftsamplearedifferentfromtherelationsfound with our data. As an experiment we shifted the predictions fortheComasampleatthe96–98%confidencelevel. fromThomasetal.tothebestagreementwithourdata. The Based on the simulations of the data and the selection ef- upperofthetwodashedgreenlinesshowthisfortheforma- fects, we conclude that the differences in relations we find tionlookbacktimesshifted2.5Gyrearlierforallmassessuch between the Coma sample and the high redshift sample are thattheearliestformationlookbacktimeis14Gyr(roughly unlikelyto be dueto systematic effectsin the data or dueto theageoftheUniverseinthiscosmology). Theabsolutefor- differencesinselectioneffects. mationepochsfromThomasetal. may notbe correct, since their analysis depends on stellar population models. How- 5. THESTARFORMATIONHISTORYOFE/S0CLUSTER ever,theirresultsontherelativetimingofthestarformation GALAXIES episodes as a functionof galaxy mass closely match our re- The median offset of logM/L for the high redshift sam- sultsforthishighredshiftsample. ple relative to the Coma sample is –0.38. Using stellar Thomaset al. predictthat star formationis on-goingfor a population models from Maraston (2005), which show that longerperiodinlowmassgalaxiesthaninhighmassgalaxies. 4 Basedonthis,weestimatethattheinternalscatterintheM/L– relation than found for the Coma cluster. We interpret this massrelation,inlogM/L,shouldbe≈0.06at1010.3M⊙ but as due to a mass dependency of the epoch of the last ma- only≈0.01at1011.3M⊙. Wecannotconfirmsuchadecrease jor star formation episode. The lowest mass galaxies in oftheinternalscatter. However,itwouldmostlikelyrequire thesample (1010.3M⊙) haveexperiencedsignificantstar for- alargersampleand/orsignificantlysmallermeasurementun- mation as recent as zform ≈1.35, while high mass galaxies certaintiestotestthisprediction. (Mass>1011.3M⊙)havezform>4.5.Thisisingeneralagree- Factors other than the mean ages of the stellar popula- ment with the predictions for the star formation histories of tions could be affecting the M/L ratios of the galaxies. For E/S0galaxiesfromThomasetal.(2005)basedontheiranal- RXJ0152.7–1357 we found based on absorption line index ysis of line index data for nearby galaxies. The scatter of datathatalargefractionofthegalaxiesmayhaveα-element FP for these two z=0.8- 0.9 clustersis as low as foundfor abundance ratios, [α/Fe], about 0.2 dex higher than found theComacluster,andwefindnosignificantdifferenceinthe in nearby clusters (Jørgensen et al. 2005). This could af- scatterforlowandhighmassgalaxies. Thisindicatesthatat fect the M/L ratios in a systematic way. Maraston (private a given galaxy mass the star formation history for the E/S0 comm.) finds from modeling that stellar populations with galaxies is quite similar. In a future paper we will discuss [α/Fe]=0.3,solarmetallicitiesandagesof2-7Gyrmayhave theseresultsinconnectionwithourabsorptionlineindexdata M/L ratiosin the bluethatareabout20percenthigherthan forthegalaxiesinbothhighredshiftclusters. thosewith[α/Fe]=0.0.Whileitisstilltooearlytousethese modelsfordetailedanalysisofhighredshiftdata,itindicates Based on observations obtained at the Gemini Observa- thatforfuturedetailedanalysisoftheFPwemayhavetoin- tory(GN-2002B-Q-29,GN-2004A-Q-45),whichisoperated cludeinformationabout[α/Fe]ofthegalaxies. byAURA, Inc.,underacooperativeagreementwithNSFon behalf of the Gemini partnership: NSF (US), PPARC (UK), 6. CONCLUSIONS NRC (Canada), CONICYT (Chile), ARC (Australia), CNPq We find that the FP for E/S0 galaxies in the clusters (Brazil) and CONICET (Argentina). Based on observations RXJ0152.7–1357(z=0.83)andRXJ1226.9+3332(z=0.89) madewiththeNASA/ESAHubbleSpaceTelescope. IJ,KC, isoffsetandrotatedrelativetotheFPofourlowredshiftcom- and KF acknowledgesupportfromgrantHST-GO-09770.01 parisonsampleofComaclustergalaxies. Expressedasa re- fromSTScI.STScIisoperatedbyAURA, Inc. underNASA lation between the M/L ratios and the masses of the galax- contractNAS5-26555. ies, the high redshift galaxies follow a significantly steeper REFERENCES Bender,R.,Burstein,D.,&Faber,S.M.1992,ApJ,399,462 Jørgensen, I., Bergmann, M., Davies, R., Jordi, B., Takamiya, M., & Bernardi,M.,etal.2003,AJ,125,1866 Crampton,D.2005,AJ,129,1249 Blakeslee,J.P.,Lucey,J.R.,Tonry,J.L.,Hudson,M.J.,Narayanan,V.K., Kelson,D.D.,Illingworth, G.D.,vanDokkum,P.G.,&Franx,M.2000, &Barris,B.J.2002,MNRAS,330,443 ApJ,531,184 Claver,C.F.1995,Ph.D.thesis,Univ.Texas Maraston,C.2005,MNRAS,362,799 Colless,M.,Saglia,R.P.,BursteinD.,Davies,R.L.,McMahanJr.,R.K.,& Peng,C.Y.,Ho,L.C.,Impey,C.D.,&Rix,H.-W.2002,AJ,124,266 Wegner,G.2001,MNRAS,321,277 Sérsic,J.L.1968,AtlasdeGalaxiasAustrales(Córdoba:Obs.Astron.Univ. deLucia,G.,etal.2004,ApJ,610,L77 Nac.Córdoba) diSeregoAlighieri,S.,etal.2005,A&A,442,125 Thomas,D.,Maraston,C.,Bender,R.,&deOliveira,C.M.2005,ApJ,621, Djorgovski,S.,&Davis,M.1987,ApJ,313,59 673 Dressler, A., Lynden-Bell, D., Burstein, D., Davies, R. L., Faber, S. M., Toft,S.,Mainieri, V.,Rosati,P.,Lidman,C.,Demarco,R.,Nonino,M.,& Terlevich,R.,WegnerG.1987,ApJ,313,42 Stanford,S.A.2004,A&A,422,29 Gebhardt,K.,etal.2003,ApJ,597,239 Treu,T.,etal.2005,ApJ,633,174 Holden,B.P.,etal.2005,ApJ,620,L83 vanderWel,A.,Franx,M.,vanDokkum,P.G.,Rix,H.-W.,Illingworth,G. Hook,I.M.,Jørgensen,I.,Allington-Smith, J.R.,Davies,R.L.,Metcalfe, D.,&Rosati,P.2005,ApJ,631,145 N.,Murowinski,R.G.,&Crampton,D.2004,PASP,116,425 vandeVen,G.,vanDokkum,P.G.,&Franx,M.2003,MNRAS,344,924 Jørgensen,I.1999,MNRAS,306,607 Wuyts,S.,vanDokkum,P.G.,Kelson,D.D.,Franx,M.,&Illingworth,G. JørgensenI.,FranxM.,&Kjærgaard,P.1995,MNRAS,273,1097 D.2004,ApJ,605,677 Jørgensen, I., Franx, M., & Kjærgaard, P. 1996, MNRAS, 280, 167 Ziegler,B.L.,Thomas,D.,Böhm,A.,Bender,B.,Fritz,A.,&Maraston,C. (JFK1996) 2005,A&A,433,519 Jørgensen,I.,Franx,M.,Hjorth,J.,&vanDokkum,P.G.1999,MNRAS, 308,833