UV to Mid-IR Observations of Star-forming Galaxies at z~2 Stellar Masses and Stellar Popula

更新时间：2023-07-23 00:46:01 阅读量：实用文档文档下载

说明：文章内容仅供预览，部分内容可能不全。下载后的文档，内容与下面显示的完全一致。下载之前请确认下面内容是否您想要的，是否完整无缺。

uv是什么意思推荐度：
相关推荐

We present the broad-band UV through mid-infrared spectral energy distributions (SEDs) of a sample of 72 spectroscopically-confirmed star-forming galaxies at z=2.30+/-0.3. Located in a 72 arcmin-squared field centered on the bright background QSO, HS1700+6

002 raM 92 2v5843050/hp-ortsa:viXraDRAFT:February2,2008

PreprinttypesetusingLATEXstyleemulateapjv.6/22/04

UVTOMID-IROBSERVATIONSOFSTAR-FORMINGGALAXIESATZ～21:STELLARMASSESAND

STELLARPOPULATIONS

AliceE.Shapley2

UniversityofCalifornia,Berkeley,DepartmentofAstronomy,601CampbellHall,Berkeley,CA94720

CharlesC.Steidel,DawnK.Erb,andNaveenA.Reddy

CalforniaInstituteofTechnology,MS105-24,Pasadena,CA91125

KurtL.Adelberger3

ObservatoriesoftheCarnegieInstitutionofWashington,813SantaBarbaraStreet,Pasadena,CA91101

MaxPettini

InstituteofAstronomy,MadingleyRoad,CambridgeCB3OHAUK

PaulineBarmbyandJiashengHuang

Harvard-SmithsonianCenterforAstrophysics,60GardenStreet,Cambridge,MA02138

DRAFT:February2,2008

ABSTRACT

Wepresentthebroad-bandUVthroughmid-infraredspectralenergydistributions(SEDs)ofasam-pleof72spectroscopically-con rmedstar-forminggalaxiesatz=2.30±0.3.Locatedina72arcmin2 eldcenteredonthebrightbackgroundQSO,HS1700+643,thesegalaxieswerepre-selectedtolieatz～2basedsolelyontheirrest-frameUVcolorsandluminosities,andshouldberepresentativeofUV-selectedsamplesathighredshift.Inadditiontodeepground-basedphotometryspanningfrom0.35 2.15µm,wemakeuseofSpitzer/IRACdata,whichprobestherest-framenear-IRatz～2.Therangeofstellarpopulationspresentinthesampleisinvestigatedwithsimple,single-componentstellarpopulationsynthesismodels.Theinabilitytoconstraintheformofthestar-formationhistorylimitsourabilitytodeterminetheparametersofextinction,age,andstar-formationratewithoutusingexternalmulti-wavelengthinformation.Emphasizingstellarmassestimates,whicharemuchlessa ectedbytheseuncertainties,we nd logM

inthez～2erahasoccurredonanumberoffronts:newsurveyshavenowsuccessfullyidenti edsamplesofz～2galaxiesbasedupontheirverybrightfar-IRandradioemission(e.g,Chapmanetal.2004),theirsigna-turerest-framefar–UVcolors(Adelbergeretal.2004b;Steideletal.2004),theirhighluminosityintheobservedKband(Daddietal.2004b),redrest-frameopticalcol-ors(vanDokkumetal.2003;Franxetal.2003),orsomecombinationoftheabove(e.g.,Abrahametal.2004;Shapleyetal.2004).Theresultsofmostofthesesur-veyssuggestthatitwasnearz～2thatgalaxiesasmassiveasanyweseeinthelocaluniversebeganreach-ingmaturity,andwhereatransitionfromrapidlystar-formingtoasymptoticallyquiescentmaybeoccurring“beforeoureyes”forrelativelymassivegalaxies(e.g.,McCarthyetal.2004;Daddietal.2004a;Shapleyetal.2004).Fortuitously,galaxiesintheredshiftrange2<z<2.5areparticularlywell-suitedstudies～～

tospectroscopicbothintheobservedoptical(Steideletal.2004;Daddietal.2004b)andespeciallyintheobservednear-IR,wherenebularlinesvaluableformeasurementofkine-matics,star-formationrates,andchemicalabundancesareobservationallyaccessible(e.g.,Erbetal.2003,2004,2005;Shapleyetal.2004;vanDokkumetal.2004).Oneofthefundamentalphysicalmeasurementsthatcanbemadeforhighredshiftgalaxies,andoneofthemoreimportantforplacingthemintothetheoreticalcon-text,isthestellarmass.Anumberofauthorsinthelastseveralyearshaveinterpretedobserved-frameUVtonear-IRphotometrywithstellarpopulationsynthesismodelsinordertoinferthestellarmassforsamplesofz>2galaxies(e.g.,Papovichetal.2001;Shapley2001;～

etal.Daddietal.2004b).Othershaveusedsimilarmethodsandthetechniqueofphotometricredshiftstoobtainmeasuresofthetotalstellarmassasafunctionofredshift(e.g.,Dickinsonetal.2003;Fontanaetal.2003;Glazebrooketal.2004).Stellarmassmeasuresbecomeincreasinglyuncertainasonegoestohigherwhereground-basedobservations,limitedtoλ<redshift,

2.3µm,probeincreasinglyshorterrest-framewavelengths.～

Atz>2,theKsbandcorrespondstotherest-frameoptical,wherecurrentorrecentstarformationmaydominatetheobservedluminosityratherthanthestellarmassthathasaccumulatedoveragalaxy’slifetime.Forthisreason,thearrivalofdatafromtheIRACinstrument(Fazioetal.2004)onboardtheSpitzerSpaceTelescope(hereafterSpitzer)hasbeenwidelyanticipatedasameanstosig-ni cantlyimprovethemeasurementofstellarmassandstarformationhistorysinceitiscapableofobtainingverysensitiveobservationsoutto8µm,allowingphotometricaccesstotherest–frame1 2µmlightfromgalaxiestoredshiftsz～4,andprovidingparticularlysensitiveob-servationsofthisregionforgalaxieswithmoremodestredshiftsnearz～2.

Inthispaper,wediscussthebroad-bandpropertiesofareasonablylargesampleofspectroscopicallycon- rmed,UV-selectedgalaxiesatz～2ina eldthatwasobservedbytheIRAC“In-OrbitCheckout”(IOC)pro-gramintheearlydaysofSpitzer.Theprincipalaimhasbeentoexplorethebroad-bandspectralenergydis-tributions(SEDs)forarobustlyde nedsampleofhighredshiftgalaxiesthatnaturallyspansalargerangeinstellarmass,star-formationrate,andstar-formationhis-tory.ThemainadvantagesofaUV-selectedsampleatz～2,describedindetailelsewhere(Adelbergeretal.2004b;Steideletal.2004)arethatcon rmingopticalspectracanbeeasilyobtained,photometricselectionishighlye cientandrequiresusingonlyrelatively“inex-pensive”opticalground-basedphotometry,anditispos-sibletotunecolorcriteriatoyieldaroughlyvolume-limitedsampleoverthedesiredredshiftrange.Inthisparticularcase,anadditionaladvantagewasthatthespectroscopicsampleofgalaxiesalreadyexisted,asdiddeepphotometryfromobserved0.35to2.15µmandtar-getednear-IRspectroscopicobservations.Detailsofthenear-IRphotometryandspectroscopywillbepresentedinErbetal.(2005).TheHS1700+6416 eldisoneofseveralwehaveusedinasurveydesignedtocombineob-servationsofz～2galaxieswithsensitiveobservationsoftheintergalacticmedium(IGM)inthesamevolume.Inthisrespect,HS1700+6416(zem=2.72)isoneoftheverybest-studiedhighredshiftQSOsintheentiresky,duetoitsunusualbrightness(V～16).Italsohappenstolieinthe“continuousviewingzone”forbothSpitzerandHST;thisfact,plustheavailabilityoftheancillarydeepgroundbaseddata,isthemainreasonthe eldwaschosenforthedeepIOCobservationsbytheIRACteam.Incombinationwiththeground-basedphotometricandspectroscopicdata,theIRACobservationsofthis eldprovideameanstoestablishtherangeofproper-tiesrepresentedamonggalaxiesselectedusingUVcolors;thesemaybeusefulforcomparisontogalaxysamplesatsimilarredshiftsbutselectedusingothertechniques.Ofparticularinteresttousisaquantitativeassessmentofthe“valueadded”bylongerwavelengthphotometryforassessingthegalaxypropertiesathighredshiftsascomparedtowhatonewouldinferfromground-basedmeasurementsalone(sincerealisticallyonlyafractionofground-based eldswillbeobservedbySpitzerduringitslifetime).Asitturnsout,theHS1700 eldisalsointer-estingbecauseitallowsforapreliminarycomparisonofthedependenceofz～2galaxypropertiesonlarge-scaleenvironment(Steideletal.2005).

Thepaperisorganizedasfollows:§2presentstheground-basedandSpitzerobservations;§3describesthephotometryandtheselectionofthesamplewhichisdis-cussedintherestofthepaper;§4describestheuseofstellarpopulationsynthesismodelstoinferphysicalpropertiesofthegalaxiesfromtheirbroad-bandspectralenergydistributions(SEDs);§5summarizestheresultsofthemodelingfocusingonthegalaxystellarmassesandmass-to-lightratios(M/L);§6discussestheresultsandtheirgeneralimplications;§7summarizesthemainresults.

Weassumeacosmologywith .7throughout.

m=0.3, Λ=0.7,andh=02.OBSERVATIONS2.1.OpticalImaging

ImagesoftheHS1700+64 eldwereobtainedin2001MayusingtheWilliamHerschel4.2mtelescopeonLaPalmaandthePrimeFocusImager.TheWHTsystemusesa2-chipmosaicofMarconi2kx4kCCDswithespe-ciallygoodUVquantum′e ciency,andprovidesa eldofviewof16′by16sampledat0′′.236perpixel.Theconditionsduringtheobservingrunwereexcellent,withsub-arcsecondseeinginallbands.The eldwasroughly

centeredonthepositionoftheQSO,and15arcsecondditherswereexecutedbetweensubsequentexposuresinordertoallowforfringeremovalintheRbandandthecreationofdarkskysuper ats(particularlynecessaryintheUnband).ThedatawerereducedusingproceduresdescribedindetailbySteideletal.(2003).

AdditionalimagesintheUregionoftheWHTnbandcoveringthecen-tral5′by7′ eldwereobtainedin2001MayusingthebluesideoftheLowResolu-tionImagingSpectrometerontheKeckI10mtelescope(McCarthyetal.1998;Steideletal.2004).Thesewerescaled,varianceweightedandcombinedwiththeWHTdatatoformthe nalUofthereducednimage.

Propertiesandstackedimagesaresum-marizedinTable1.

2.2.Near-IRImaging

WeobservedtheHS1700+64 eldduringtwoobserv-ingruns,2003Mayand2003October,usingthePalo-mar5.1mHaletelescopeandtheWideFieldInfraredCamera(WIRC;Wilsonetal.2003).TheimagesweretakenintheK-short(Ks) lter,obtainedinasequenceof30-secondintegrations,movingthetelescopeinapre-scribedditherpatternaftereachsequenceof4expo-sures.Theconditionsduringbothobservingrunsweregenerallygood,withseeingrangingbetween0′′.55and0′′

.99,buttheambienttemperaturewaswarmresult-inginrelativelyhighKmagarcsec 2).TheWIRCsbandskybackgrounds(～12.4cameraemploysaRockwellHgCdTe“Hawaii-2”2kx2karray,providinga eldofviewof8′.5x8′.5withaspatialsamplingof0.249arcsecperpixel.

TheWIRCdatafromeach(typically54minute)dithersequencewerereduced,registered,andstackedusingasetofIDLscriptswrittenbyK.Bundy(privatecommu-nication).TheresultingimageswerethenregisteredandcombinedusingIRAFtasks.The nalKof11.03hoursofintegrationsimagerepre-sentsatotalinthepartsoftheimagethatreceivedthefullintegrationtime.The nalmosaicwastrimmedtoincludeonlythepartsoftheimagereceiving50%ormoreofthetotalintegrationtime(～90%ofthetrimmedimagereceivedthefullexposuretime).Theimagewascalibratedwithreferencetothenear-IRstandardsofPerssonetal.(1998).

2.3.Spitzer/IRACImaging

ObservationswiththeInfraredArrayCamera(IRAC)wereobtainedin2003Octoberduringthe“In-OrbitCheckout”(IOC)ofSpitzer.TheobservingstrategyanddatareductionsaredescribedindetailbyBarmbyetal.(2004).TheIRACobservationsweredesignedtoover-lapasmuchaspossiblewiththeground-basedKcompromiseswereadoptedsandopticalimages,butsometoavoidknownforegroundgalaxyclustersinthe eld,sothattheIRACpointingswerebiasedtowardtheSWrel-ativetotheground-baseddata,whicharecenteredonthebrightQSO.Asdetailedbelow,theIRACimagesinthe4IRACbands(seeTable1)completelyoverlapoveronlyarelativelysmallfractionofthe eldobservedintheKinthesbanddescribedabove,withthesmallestoverlapIRACchannel1(3.6µm)andchannel3(5.8µm)images,butmuchlargeroverlapinchannels2(4.5µm)and4(8.0µm).Figure1illustratestheregionscovered

intheIRACimagescomparedtothosecoveredinthePalomarKsimage.

ThephotometriccalibrationsfortheIRACimageswereasdeterminedbytheIRACteam(Barmbyetal.2004);however,weusedasigni cantlydi erentmethodforperformingthephotometryandestimatingthepho-tometricuncertainty(see3.2),resultinginsigni cantlymoreconservative(rger)estimatederrorsatagiven uxdensity.

3.PHOTOMETRY3.1.OpticalandNear-IR

Theobservablesfromourground-basedimagesareRmagnitudesandisophotalR Kcolors,withisophotesde nedins,G R,andUtheRdetectionn Gcat-alog.Inordertomimictheactualobservingprocess,wethereforeestimateduncertaintiesinisophotalcolorratherthanmagnitudeforalloptical/IRbandsexceptR.TheuncertaintieswereestimatedseparatelyfortheopticalandIRphotometry,byaddinglargenumbersofsimulatedgalaxiesofknownmagnitudestotheimagesandcomparingtheirrecoveredphotometrywiththeirinputparameters.Thesimulatedgalaxiesweredrawnfromtemplates′′ofexponential′′diskswithscalelengthsrangingfrom0.05to0.4,smoothedtomatchtheseeingoftheindividualimages.FortheR Kweselectedgalaxiesatrandomwithinsuncertain-ties,therange21.0<R<26.0and1.0<R Ks<5.5andplacedthematrandompositionsontheRandKatatime(toavoidovercrowding)foratotalsimages,200of250,000fakegalaxies.Wethenperformedphotometryasusualontheimages,andcomparedthecatalogofdetectionswiththeinputlistofsimulatedgalaxies.Weconsideredthequantity [R Kstepsofs]=(R K0.5magins)meas (R KandbinnedinRs)true,Themeanvalueofmeasand0.2magin(R Ks)meas. [R Ktypicals]isanestimateofthebiasinthephotometry;forob-jects, [R KR Ks]<0.05mag.Thestandarddeviationσ( [s])providesanestimateoftheR Ktaintyforobjectsthatfallineachbin;thiswastypicallysuncer-σ( [R Kares])～0.25mag.R KgiveninTable2.UncertaintiessuncertaintiesforeachobjectinR,G RandUn Gweredeterminedinasimilarfashion,byaddinglargenumbersoffakegalaxiestotheUn,GandRimagesandcomparingtheirrecoveredandinputprop-erties,withtheadditionalrestrictionthatonlythoseob-jectswhoserecoveredphotometrymetourselectioncrite-riawereconsidered.WebinnedinRandG Rtodeter-minetheRandG Runcertainties,whichweretypicallyσ( R)～0.14magandσ( [G R])～0.07mag,andinGandUn GtodetermineU( [Umag.n Guncertainties,typi-callyσn G])～0.13AgainuncertaintiesforeachobjectaregiveninTable2.Similartechniquesforestimatingphotometricuncertaintieshavebeenusedin,e.g.,Adelberger&Steidel(2000);Steideletal.(2003);Shapleyetal.(2001).

3.2.IRAC

WemeasuredIRAC uxesby rstcomputinganem-piricalpointspreadfunction(PSF)foreachchannelbymedianaveragingthesignalfromseveralisolatedpointsourceslocatedthroughoutthemosaicedIRACdataintheHS1700 eldandnormalizingtheresultingPSFto

Fig.1.—Aschematicmapofthesurvey eldconsidered.ThegrayscaleimageisthePalomar5mWIRCKsimage,withthepositionsofobjectswith1.4≤z≤2.8markedwithblackcircles.Thecoloredcontours(blue=IRAC/Ch1;green=IRAC/Ch2;yellow=IRAC/Ch3;red=IRAC/Ch4)indicateregionsoftheimagethatreceived>10%ofthetotalIRACintegrationstimeineachIRACband,i.e.,>1.1hours.Forfullresolution gure,gotohttp://astron.berkeley.edu/～aes.

haveunit ux.Weobtainedaccurateobjectpositions<0.2′′rmsuncertainty)for>95%ofIRACsourcesin(～

the eldfromtheirKs-bandcounterpartsintheWIRCimagingofthis eld,orfromtheirR-bandcounterpartsforthesmallfractionofobjectsdetectedinRbutnotinKs.Wethenusedtheempirically-determinedIRACPSFtosimultaneouslymodeltheemissionofallobjectswithinasub-imageofsize24′′surroundingeachobjectofinterest,similartomethodsgenerallyusedforphotome-tryofstellarobjectsincrowded elds.Thisproceduretosomeextentmitigatesthee ectsofconfusionnoise,sincetheobjectpositionsareknownapriorifromthemuchhigherspatialresolutionKsbandimage,andallowsforsomede-blendingofobjectsthatareonlypartiallyre-solvedintheIRACimages.ThemethodisalsosimilartothatemployedbyFern´andez-Sotoetal.(1999)andPapovichetal.(2001)forcomparingground-basednear-IRandHST/WFPC-2photometry.Photometricerrorswerecomputedbyaddinginquadraturetheerrorinthesource ux(Poissonnoise)andthedispersioninmea-sured uxvaluesfoundby ttingPSFsto100randompositionscontainingnoobvioussourcesneartheobjectofinterest.Thislatterestimateshouldaccounttoalargeextentfornoiseduetosourceconfusion,whichispartic-ularlyrelevantinthe3.6and4.5µmbands.Exceptincasesofverybrightsources,thebackgrounddispersiondominatestheerrors.ThePSFfull-widthathalfmaxi-mum(FWHM)andapproximate5σsensitivitylimitsforeachchannelarepresentedinTable2.Weomittedanymeasured uxeswithuncertaintieslargerthan0.5magni-tudesfromconsiderationin ttingtheSEDsasdescribedin§4.1.Inaddition,wesettheerrorsto0.10magnitudesformeasurementswhoseformaluncertaintiescalculatedusingtheaboveprescriptionwere<0.10magnitudes,in

Fig.2.—Redshifthistogramfor72spectroscopicallyiden-ti eddetections.galaxies2.30±0.015,ThisinthediscussedhistogramHS1700+643bySteidelcontains eldetal.awith(2005).

signi cantnear-IRspikeandatIRACz=ordertoaccountforsystematicuncertaintiesinthepho-tometrythatarenototherwiseaccountedfor(e.g.,sourceconfusiononscalessmallerthanthePSF,estimationoflocalbackground,etc.).

3.3.SampleSelection

BecausetheredshiftofHS1700+643iszthepropertiesoftheem=2.72andwewereinterestedingalaxiesinthe“foreground,”thespectroscopicobservationstar-getedphotometricallyselectedcovertheredshiftrange1.9<galaxiesthatwouldneatly

z<2.7.Toaccomplishthisweselectedacombination～of～

objectssatisfyingei-therthe“BX”photometriccriteriaofAdelbergeretal.(2004b)(z=2.20±0.32)orthe“MD”photometriccri-teriaofSteideletal.(2003)(z=2.79±0.27).Atotalof100objects(81“BX”and19“MD”)hasmeasuredspec-troscopicredshiftsz>1.4fromobservationswiththeLowResolutionImagingSpectrometer,describedinde-tailinSteideletal.(2004).BecausethesurfacedensityofBXcandidatesis～6timeshigherthanthatofMDcandidatestothesameR=25.5apparentmagnitudelimit,theexpectedredshiftdistributionforaBX/MDphotometricsample,basedonthetotalof1133spec-troscopicredshiftsforBX/MDcandidatesinallsurvey elds,is z =2.24±0.36.

Thesampleofgalaxiesdiscussedintherestofthepaperisculledfromtheregionofoverlapbetweentheground-basedUnGRimages,theground-basedKage,the4IRACbands(see gure1)andthespectro-sim-scopicobservationsofUV-selectedgalaxies.Withinthe8′.3by7′.8regioncoveredbythePalomarKisatotalof508BXandMD-typegalaxysimage,therecandi-datesidenti edfromtheopticalphotometryusingthese-lectioncriteriade nedbyAdelbergeretal.(2004b)andSteideletal.(2003),respectively;ofthose,389(77%)oftheobjectsaresigni cantlydetectedintheKWithinthe64.7squarearcminregioncovereds-bandim-age.bytheKsimage,weobtainedspectrafor162ofthe508pho-tometriccandidatesandspectroscopicallyidenti ed98

objects.Theidenti edfractionrepresentsa60.5%spec-troscopicsuccessrate,similartothatofourz～2surveyoverall(Steideletal.2004).AsdiscussedinSteideletal.(2004),weexpecttheunidentifedobjectshavethesameredshiftdistributionasthespectroscopicsuccesses.Ofthe98objectswithspectroscopicidenti cations,92aredetectedintheKsimagetoKs(Vega)～22;ofthe92ob-jectswithspectroscopicIDsandK.4andhaveameanredshiftsdetections,79(86%)areatz>1anddispersionof2.26±0.30,entirelyconsistentwiththeexpectedred-shiftdistributionforallBX/MDcandidatesasdiscussedabove.Ofthe79spectroscopicallyidenti edz～2galax-ies,72aredetectedinoneormoreoftheIRACpassbandsandincludedinthestellarpopulationanalysis.Invari-ably,objectsdetectedsigni cantlyat2.15µmarealsodetectedintheIRAC3.6and4.5µmbands;the7galax-iesdetectedat2.15µmbutnotwithIRACsimplyfalloutsideofthewell-exposedregionsoftheIRACpoint-ings.

Insummary,the72z～2galaxiesdiscussedinthispaperarerepresentativeofaUV-selectedsampletoR=25.5,butareobjectswhicharealsodetectedinboththenearandmid-IR.TheonlypossiblebiasofthissamplerelativetoanysampleselectedusingthesameUVcriteriaisthatitexcludesthe～23%ofBX/MDcandidatesinthe eldthatarenotdetectedtoKdistributionforthisspectroscopics～22.Theredshiftsamplewithnearandmid-IRdetectionsisshownin gure2.Verystrikingly,thereisasigni cantgalaxyoverdensityinthisdistributionatz=2.30,whichallowsforanexam-inationofgalaxypropertiesasafunctionoflarge-scaleenvironment(Steideletal.2005).

Itisofinterest,forsomepurposes,toaskwhatfractionofamid-IRselectedgalaxysampleatthesameredshiftisrepresentedbyobjectsselectedusingtheUVcolorcri-teria.Thisisadi cultquestiontoanswer,asthereisasyetnolargespectroscopicsampleofmid-IRselectedhighredshiftgalaxies(and,aswearguebelow,photo-metricredshiftsmaybeaninadequateapproachtotheproblem).However,wecanmakearoughestimateasfollows.Recently,Barmbyetal.(2004)suggestedare-gioninIRAC3.6/4.5/5.8µmcolorspacethatmaymakeitpossibletoselecthighredshiftgalaxiesfromIRACcol-orsalone,basedonmodelsandtheobservedlocationsofz～3UV-selectedgalaxiesinIRACcolorspace.In g-ure3weshowthelocationsintheUoftheobjectsdetectedintheKnGRcolorplaneofallsbandthatarewithintheregionoftheIRACimagescoveredbyIRACat3.6,4.5,and5.8µmandwhichhaveIRACuncertaintiesof<0.3magineachband.4.Outofatotalof618galaxieswithKs,3.6µm,4.5µm,and5.8µmmeasurements,443arein-cludedintheopticalphotometriccatalogtoR=25.5,99ofwhichsatisfythecriteriaproposedbyBarmbyetal.(2004)inthe3.6-5.8µmcolorspace.Asshownin g-ure3,46ofthe99sourcessatsifyingtheBarmbyetal.colorcriteriadoindeedlieintheregionsofUfromwhichz～2BM(z=1.70±nGRcolorspace0.34),BX(z=2.20±0.32),MD(z=2.72±0.3)andC/D/M(z=3.0±0.3)(Steideletal.2003)galaxiesaredrawn.

Thesizeofthisregionisapproximately30arcmin2onlyregion～40%oftheobjectsina,although5.8µmareimage.

included,primarilybecauseKs-selectedofthelimitedcatalogdepthwithinofthisthe

Fig.3.—PlotshowingthelocationinopticalUnGRcolorspaceofallobjectsinaK-selectedcatalogtoK 22.Theregionswithinthesolidblacklinesdenotethoseusedtoselectthegalaxiesdiscussedinthispaper(“BX”and“MD”);thedottedcurvesdenoteregionscontaining“BM”(z=1.7±0.3,lowerleft)and“C/D”LBGs(z=3.0±0.3,upperleft).Objectssatisfyingthe“highz”galaxycriteriasuggestedbyBarmbyetal.(2004)basedonIRACcolorsaremarkedinred.Asdiscussedinthetext,asigni cantfraction(～40%)oftheIRAC“highz”objectsintheopticalcatalogappeartolieatz<1.4.

However,thereisasigni cantnumberofgalaxies(～40)whichsatisfytheBarmbyetal.criteriabutarelocatedinregionsinhabitedprimarilybyz～0.7 1.5starform-ingorpassivegalaxiesinopticalcolorspace(see,e.g.,Adelbergeretal.2004b).Another～10haveopticalcol-orsthatareoutsideoftheUVselectionwindows,butareconsistentwithmorereddenedstarforminggalaxiesatz～1.5 2.5(cf.Daddietal.2004a).Ofthe175galax-iesthatarewelldetectedintheKs,3.6,4.5,and5.8µmbandsbutabsentfromtheopticalcatalog,81sat-isfytheBarmbyetal.colorcriteria.Assumingthatthesamepercentage(～40%)ofthesearelikelytobeintheforegroundasfortheopticallydetectedgalaxies,～50

galaxiestoKs～22areplausiblyatz～1.55.Mostof

29ofthe81objectsareactuallydetectedintheRbandtoR～26butarenotincludedin gure3becausethecolorsarenot

thegalaxiesinthisredsamplearerelativelyfaintintheKsband,withonly12of81havingKs<20,ofwhichperhaps6-8arehighredshiftobjects.

Insummary,ofthe180galaxieswelldetectedintheintheKs,3.6,4.5,and5.8µmbandsthatalsosatisfytheBarmbyetal.criteria,46wouldbeselectedbased

onUnGRcolorstobeatz～1.5,～70(～40%)are

likelytobeatz～1.5,whiletheremaining～60galaxies

>areplausiblez～1.5candidatesthatdonotsatisfythe

UnGRselectioncriteria.Thus,weconcludethat～50%ofgalaxiesatz～2.3±0.4toanapparentmagnitudelimitofK～22(K～23.8ontheABsystem)wouldbeselectedusingtherest-UVcolorcriteriatoR=25.5.ThisstatisticisbroadlyconsistentwiththeresultsoftheFIRESsurvey(Franxetal.2003)whichtargetedgalax-reliablymeasured.

ieswithcolorsJ K>2.3,mostofwhicharesimilarlyopticallyfaint,aswellaswiththe“IRACEROs”dis-cussedbyYanetal.(2004).Wedeferfurtherdiscussionoftheoverlapanddistinctivenessofsamplesselectedinvariouswaystofuturework.

4.POPULATIONSYNTHESISMODELING

4.1.Sample

InordertoexplorethestellarpopulationsofUV-selectedz～2galaxiesintheHS1700+64IRACpointing,weconsideredonlythesubsampleof72galaxieswithmeasuredredshifts,Ktionsinatleastonechannel.smagnitudes,andIRACdetec-Asdiscussedbelow,thesetofcolorsandmagnitudesforBXandMDgalaxiesisnotsu cienttosimultaneouslydetermineaccuratepho-tometricredshiftsandstellarmassesforgalaxiesthatdonothavespectroscopicredshifts.

Thephotometricmeasurementsusedtomodelthestel-larpopulationsaregiveninTable2.NotethatallofthemagnitudesandcolorsareontheABsystem,withtheexceptionoftheR KVegasystemforscolor,forwhichtheKeasiercomparisontoothersmagisontheworkonsimilarhighredshiftgalaxies.6Forthe13galaxiesatz>2.5,wedidnotincludetheUn GcolorintheSED t,sinceabsorptionbyHIintheintergalacticmediumbeginstohaveastronge ectonthemeasuredUnmag-nitudeattheseredshifts.

Beforemodelingthegalaxymagnitudesandcolors,wecorrectedtheUticLyαemissionnorGmagnitudeforthee ectsofgalac-andabsorption.LyαequivalentwidthsweremeasuredfromtheLRISspectrausedtoidentifyredshifts.Thesecorrectionswereappliedtogalaxiesatz=1.7 1.9(Un)orz>2.47(G).WealsohaveHαand[NII]emissionlinemeasurementsforfourgalaxiesintheHS1700sample(BX691,BX717,MD103,andMD109,allfromErbetal.(2003)).WecorrectedtheobservedKby0.10,0.17,0.08,and0.10magnitudes,respectivelytosaccountforthecontributionofnebularemissiontotheobserved ux.Ksmagnitudesfortheremainderofthegalaxiesintheredshiftrange2.016<z<2.504remainuncorrected.BasedonamuchlargersampleofHαspec-traofgalaxiesinthesameredshiftrange(whichwillbediscussedindetailbyErbetal.(2005)),thetypicalcor-rectiontothebroad-bandKs uxis+0.09±0.07magni-tudes,althoughsomeofthemostextremecaseshaveHαequivalentwidthslargeenoughtochangetheKmag.Thesmagni-tudebyasmuchasseveraltenthsofapossibleimpactoftheKscorrectionsonthemodelingresultsisdiscussedbelow.

Only22%oftheBX/MDphotometriccandidateswithbothKsphotometryandmeasurementsinatleastoneIRACbandarespectroscopicallyidenti ed,andwere-strictthestellarpopulationmodelingtothisspectro-scopicsample.Totestthepotentialsuccessofmodel-ingphotometriccandidateswithoutredshifts,wetriedto tsimultaneouslytheredshiftandstellarpopula-tionparametersforthesampleofgalaxieswithspec-troscopicredshifts,pretendingthattheredshiftwasanundeterminedfreeparameter.WithonlyUplusIRACphotometry,weobtainedanaccuracynGRKofs |zphot zspec|/(1+zspec) =0.09.Atthemeanred-6

systemKsbymagnitudesadding1.82orRto theKstabulatedcolorsmayvalues.

beconvertedtotheAB7

shiftofthesample,thiserrorcorrespondstoatypical|zphot zspec|of0.3,whileinseveralcasestheredshiftestimatesfailcatastrophically,with|zphot zspec|>1.WenotethatthisaccuracyisverysimilartothatfoundforphotometricredshiftestimatesintheFIRESsur-vey,basedon39galaxieswithspectroscopicredshiftsintheHubbleDeepFieldSouth(HDF-S)(Rudnicketal.2003;Labb´eetal.2003).Perhapsmoresigni cantly,thestellarpopulationparametersinferredwhensimul-taneously ttingfortheredshiftdonotuniformlyagreewiththosefoundwhentheredshiftisknown.Speci -cally,we ndatypicalfractionalof |M phot M spec|/M spec

stellarmassdi erence

=0.56±1.23,whichislargerthananyofthefractionalstellarmassuncertaintieswe ndduetophotometricuncertainties,oruncertaintiesinourchoiceofstar-formationhistoryforsingle-componentmodels(seediscussioninsection5.2).Notonlyisthescatterinthisfractionalo setquitelarge,butthedistri-butionisskewedtowardoverestimatingthethattheaverage (M phot M spec)/M spec

mass,such

=0.37±1.30.7Thisexercisedemonstratesthat,withonlyUIRACphotometry,evenforobjectsknownnGRKtohavesplusred-shiftsinthetargetedrange(whichpresumablyreducestheprobabilityofcatastrophicfailures),considerableun-certaintieswouldbeintroducedifobjectswithoutspec-troscopicredshiftswereincluded.

4.2.ModelingProcedure

FollowingaproceduresimilartothatofShapleyetal.(2001)andShapleyetal.(2004),weapplyBruzual&Charlot(2003)modelsto ttheSEDsoftheUV-selectedz～2galaxiesintheHS1700Spitzerpointing.WeusedmodelswithsolarmetallicityandaSalpeterinitialmassfunction(IMF)extend-ingfrom0.1 100M⊙.Also,asrecommendedbyBruzual&Charlot(2003),weadoptedthePadova1994stellarevolutiontracks.Papovichetal.(2001)haveinvestigatedhowthebest- tparametersdependsystematicallyonthechoiceofmetallicityandIMF,thoughwewillnotincludesuchadiscussionhere.Wedonotethat,atleastforgalaxiesbrightintherest-frameoptical(Ks≤20),solarmetallicityappearstobeagoodapproximation,andtypicalz～2galaxiesappeartohavemetallicitiesthatareonlyslightlylowerthansolar(Shapleyetal.2004).WealsonotethatitiswellestablishedthataSalpeterIMFover-predictsthenum-berofstarslessmassivethan1M⊙(or,alternatively,thatitover-predictsthestellarM/L)comparedtoobservationsinthelocaluniverse(e.g.,Belletal.2003).WeusetheSalpeterIMFtofacilitatecomparisonwithotherwork(e.g.,Coleetal.2001;Shapleyetal.2001);useoftheChabrier(2003)IMF,theIMFproposedbyBaldry&Glazebrook(2003),orthe“diet”SalpeterIMFusedbyBelletal.(2003),wouldreducetheinferredstellarmassesbyafactorof～1.5 1.8.DustextinctionistakenintoaccountwithaCalzettietal.(2000)starburstattenuationlaw.Reddy&Steidel(2004)haveshownthatusingtheCalzettilawandcontinuousstar-formationmodelstoinferunobscuredstar-formation

WhentoMweexclude vegalaxieswithwildlydiscrepantM

phot

relative spec

(allareoverestimates),we ndarelativelysym-metric0.07±distribution0.44.aroundzero,with (M phot M spec)/M spec =

ratesfromUVcolorsandmagnitudesaccuratelypre-dictstheaverageX-rayandradiocontinuum uxesofz～2star-forminggalaxies,thoughtheobjecttoobjectscattermaybelarge.Inordertodeterminehowwellthestellarpopulationparameterscanbeconstrained,weinvestigatedarangeofstar-formationhistoriesoftheformSFR(t)∝exp( t/τ),withe-foldingtimesofτ=0.01,0.05,0.10,0.2,0.5,1,2,and5Gyr,aswellascontinuousstar-formation(CSF)models.Forthemainanalysis,wedidnotconsidermorecomplexstar-formationhistories,suchasshort-durationburstsoccurringatrandompasttimessuperimposedonasmoothlydecliningstar-formationhistory.Thedataformostgalaxiesdonotplacestrongconstraintsamongtherangeofsimple,single-componentstar-formationhistoriesconsidered,andthereforetheuseofamorecomplexmodeldoesnotseemjusti ed.Weonlybrie yemployextremetwo-componentmodelsinsection5.4,inordertodeterminehowmuchstellarmasscouldbecontainedinthesegalaxiesinmaximallyoldunderlyingbursts.

Ourgoalwastoconstrainthestellarpopulationpa-rametersofeachgalaxy.The ttedparametersincluded:dustextinction(parameterizedbyE(B V)),age(tstar-formationrate(SFR),stellarmass(Msf), ),andstar-formationhistory(τ).Foreachτ,weusedagridofBruzual&Charlot(2003)modelswithagesrangingbe-tween1MyrandtheageoftheUniverseattheredshiftofthegalaxybeingmodeledandextinctionsrangingbe-tweenE(B V)=0.0andE(B V)=0.7(thoughnoneofthegalaxieshasabest- tE(B V)>0.5).ForeachcombinationofE(B V),tsf,andτ,wecomputedthefullUnGRKsplusIRAC3.6 8.0µmSEDofthemodelplacedattheredshiftoftheobservedgalaxy.Incom-putingthisSED,wereddenedtheintrinsicmodelgalaxyspectrumbydust,andfurtherattenuateditinaman-nersimulatingabsorptionbytheintergalacticmediumofneutralhydrogen(Madau1995),whicha ectsthepre-dictedUnandGmagnitudes.Bruzual&Charlot(2003)CSFmodelsarebydefaultnormalizedtohavestar-formationratesof1M⊙yr 1,whiledecliningmodelsarenormalizedtohaveatotalmassof1M⊙att=∞.Wesearchedforthemodelnormalizationthatminimizedthevalueofχ2withrespecttothemeasuredSED,whichyieldedanestimateofthestar-formationrateandstellarmass.ThecombinationofE(B V),tsf,τandnor-malizationyieldingthelowestχ2wascalledtheoverall“best- t”.Figure5showstheoverallbest- tmodelto-getherwiththeobserveddatapointsforeachofthe72galaxiesinthesample.Inmostcasesthe“best- t”τmodelisnotsigni cantlybetterthanthatofthebest- ttingmodelforeveryothervalueofτconsidered,asdiscussedbelow.

4.3.SystematicUncertainties

Evenneglectingthee ectsofphotometricuncertain-ties,parametersderivedfromSED ttingaresub-jecttosigni cantdegeneraciesandsystematics,pri-marilybecauseoftheinabilitytoconstrainthestar-formationhistory(Shapleyetal.2001;Papovichetal.2001).Theseuncertaintiesapplyevenusingthesimplestsingle-componentmodels,aswedohere,sincethebest- tage,extinction,andstar-formationratealldependsys-tematicallyontheτusedtodescribethestar-formation

history.Stellarmassestimatesaremorerobustinthefaceofuncertaintiesinthestar-formationhistory.Herewereviewthesystematicuncertaintiesinage,extinc-tion,star-formationrate,andstellarmassthatresultfrompopulationsynthesismodelingofgalaxySEDsathighredshift.

Theinferredage,tsf,foragivenvalueofτiscon-strainedmainlybythemagnitudeoftheage-sensitiveBalmerbreak.AgivenBalmerbreakstrengthindicatestherelativenumberofAandOstarsandcorrespondstoolderageswhenlargervaluesareassumedforτ(i.e.agivenrelativenumberofA/Ostarsisreachedatlatertimesformoregraduallydecliningstar-formationhisto-ries).

Theinferredextinction,E(B V),alsodependssys-tematicallyonthestar-formationhistory:decliningstar-formationhistorieswithtsf/τ>1withlessdustextinc-tioncanproducesimilarG RandR Kamodelwithalargerτ,smallertscolorstothoseofthereddeningsf/τandmoredustextinction.Inthiscase,oftherest-frameUVslopeiscausedbyamixtureofstarswithlaterspec-traltypesonaverageformodelswithsmallerτ,whereasdustextinctionaccountsfortheUVreddeningincontin-uousstar-formationmodelsorthosewithlargervaluesofτ.

ThesystematicdependenceoftheinferredSFRonthevalueofτthatparameterizesthestar-formationhistorystemsfromtwoe ects.First,thestar-formationrateisderivedfromtheextinction-correctedUV-luminosity.Itthereforedependsonthebest- tvalueofE(B V),whichaswediscussedabove,dependsonthestar-formationhistoryusedtomodeltheobservedSED.Second,aconversionfromextinction-correctedUV-luminositytoSFRisrequired.Thisconversiondependsonbothtincreasingsfandτ,andforagivengalaxy,isgener-allyanfunctionofτ.TheconversionfromUV-luminositytoSFRequilibratesafterabout100Myrformodelswithτ≥1Gyr(Kennicutt1998).How-ever,theratioofSFRtoUVluminositydeclinesasafunctionoftimeformodelswithsmallerτ.Intherangeofparameterspacespannedbythegalaxiesinoursample,τ=100MyrmodelstypicallyhaveUV-to-SFRconversions1.2 1.8timeslowerthancontinuousSFRmodelsthat tthesamecolors.Whilenotlikely,basedonexternal(spectroscopic)information,modelswithτ=10MyrhavemuchmoreextremeconversionsfromUV-luminositytoSFR,asmuchas25 1000timeslowerthanthelimitapproachedbythemodelswithlargeτ.Thisiseasilyunderstoodsincethelargestτ=10Myragethatcanreasonably ttheobservedcolorsofthegalaxiesinoursampleisroughlytsf～100Myr.Atthispoint,thestar-formationrateisroughly20000timeslowerthanattsf=0.However,theUV-luminosityhasonlydeclinedbyafactorof～20,becausewhiletherearenoO-starsleft,thelifetimeoflateB-starsisroughly100Myr,sotheyhavenotyetdisappeared.

Oursingle-component-modelestimatesofstellarmassesarenotplaguedbythesameamountofsystem-aticuncertainty,especiallywhenconsideringmorereal-isticexponentialtimescales,τ≥100Myr.Therehasbeenmuchdiscussionintheliteratureaboutthevalueofrest-framenear-IRphotometryinmeasuringstellarmasses(Glazebrooketal.2004;Belletal.2003).Infact,asshownbythemodel tstoBX490andBX505,

Fig.4.—Examplesofcon denceintervalsobtainedfromMonteCarlomodelingforthreedi erentgalaxiesselectedtospanarangeofinferredproperties,asdescribedinthetext(BX879,BX691,andBX563).Thecon denceintervalshavebeenprojectedontothespacesof(E(B-V),Age)and(Age,Mass),andthedarkandlightcontourscorrespondto68%and95%con denceranges,respectively.Whilethetimeconstant,τ,istreatedasafreeparameterrangingfrom10Myrto∞,thegridinτcontainsonlytendistinctvalues.Thediscretenatureofτtranslatesintodiscreteregionsofparameterspace,thatwouldbeconnectedsmoothlyifτwereallowedtovarycontinuously.Thise ectisconveyedmostclearlyinthe(E(B-V),Age)plots,sincebothparameterssystematicallydependonτ.Forfullresolution gure,gotohttp://astron.berkeley.edu/～aes.

whichhaveverysimilarrest-framenear-IRmagnitudes,rest-framenear-IRphotometryaloneisinsu cienttoconstrainthestellarmass-to-lightratioofstar-forminggalaxiestobetterthanafactorof10(seealso§5.4).However,withrest-frameUV,optical,andnear-IRpho-tometry,thenear-IRM/Lratio(andthereforethestel-larmass)isreasonablywell-determined,almostindepen-dentofthevalueofτusedto ttheSED.Withmea-surementsoftherest-frameUVslope(observedG Rcolorathighredshift),therest-frameUV/opticalcolor(observedR Ks),andrest-frameUV/near-IRcolor(observedR IRAC),thevarianceininferrednear-IRM/Lratioduetouncertaintiesintheformofthestar-formationhistory(forτ≥100Myr)istypically10%.Therelationshipbetween(G R,R Ks,R IRAC)andM/Lratiostartstobreakdownforstar-formationhistorieswithτ<100Myr.ThesystematicuncertaintyintheM/Lratiothereforeincreases,exceptinthecaseswheretsf≤10Myr(and≤τforallstar-formationhisto-riesunderconsideration),whenallstar-formationhisto-riesyieldthesamebest- tstellarpopulationparameters.However,formostgalaxiesinthesample,thereisnoindi-cationthatmodelswithτ≤100Myrwithtsf/τ>1arepreferred,andinsomecasessuchstar-formationhistoriesareactuallyruledout(seediscussioninsection6.2).Externalconstraintscanbeusedtoevaluatethelikeli-hoodthattheseextreme“decayingburst”star-formationhistoriesarerealistic.Forexample,Reddy&Steidel(2004)haveshownthatthemeanextinction-correctedstar-formationrateforz～2galaxiesintheChandraDeepFieldNorth,inferredfromrest-frameUVluminosi-tiesandcolors,assuming>100MyrofcontinuousstarformationandtheCalzettiextinctionlaw,agreesverywellwiththeaveragestar-formationratederivedfromstackedX-rayandradioimagesofthesamegalaxies.Ifτ<100hadbeenassumed,theaverageUV-derivedstar-formationratewouldhavebeenseveraltimessmaller,andnolongerinagreementwiththestar-formationrate

inferredfromtheaverageX-rayorradiocontinuum ux.Inprinciple,thedegeneraciesthata ectprimarilythein-terpretationoftheUVcontinuumslopecanberemovedifthereisexternalinformationonthenumberofmassivestarsstillpresentinthegalaxies.Forexample,usingobservedHα uxesand/orhighqualityfar-UVspectra,itisrelativelyeasytoestablishwhetherOstarsarestillpresentinthegalaxyspectra.Ingeneral,thehighsuccessrateindetectingHαemissionfromUV-selectedgalax-ies,andinparticularthefactthatthestar-formationrateinferredfromHα uxesisneversmallerthanthatinferredfromtheUVcontinuum(e.g.,Erbetal.2003),suggeststhatitmustbequiteraretocatchastarburstwithtSF/τ>>1andτ≤100Myr.Additionally,intheindividualLRISrest-frameUVspectrawithsu -cientS/N,oneseesevidencefortheexistenceofmassivestarsfromthepresenceofCIVandSiIVP-Cygnistel-larwindfeatures,and,insomecasesLyαemission.Ifτ<100Myrandtwouldnotsf/τweremuchgreaterthan1inmostcases,weexpectto ndsuchstrongevidenceformassiveOstarsintherest-frameUVspectraoftheseobjects.

Thesystematice ectsdiscussedabovedonotincludetheuncertaintiesthatresultfromincludingmodelswithunderlyingshortburstsofdi erentstrengthsoccurringatrandomtimesinthepast(Kau mannetal.2003).Wedidnotaddthatlayerofcomplexitytoourmod-elingprocedure,giventhedi cultywefoundindis-tinguishingamongevensimpleexponentiallydecliningandcontinuousstar-formationhistories.However,asPapovichetal.(2001)demonstrated,underlyingmaxi-mallyoldburstswithtsf>>τcanincreasetheM/Lbyafactorofaseveralwithouthavinganappreciablee ectontherest-frameUV-to-opticalcolors.Glazebrooketal.(2004) ndthatincludingrandomburstsintotherangeofstar-formationhistoriesusedto ttheVIz′Kcolorsof1.0≤z<2.0galaxiescausesthederivedmassestoin-creasebylessthanafactoroftwo,althoughtheamountbywhichonemightbeunderestimatingthetruestellarmasswilldependontheobservedgalaxycolors.Were-turntothisquestioninsection5.4.

4.4.MonteCarloCon denceIntervalsbasedon

PhotometricErrors

Inadditiontothebest- tstellarpopulationparame-tersforeachgalaxy,wealsofoundtheparametercon -denceintervalsallowedbythephotometricuncertainties.Alargeset(n=1000)offakeSEDswasgeneratedforeachgalaxy,byperturbingtheobservedSEDinaman-nerconsistentwiththeestimatedphotometricerrorbars(sections3.1and3.2).EachperturbedSEDwas tinexactlythesamemannerastheobservedSED,andasetofbest- tstellarpopulationparameterswaschosentominimizeχ2withrespecttotheperturbedSED.

Wedeterminedthecon denceintervalsoverE(B V),tsf,SFR,M ,andτ,suchthat68.3%and95.4%con- denceintervalscontainthecorrespondingfractionsoftheMonteCarlorealizations.Morethanhalfofthegalaxiesinoursamplehave95.4%con denceintervalsthatspanallτvalues,indicatingthedi cultyofdis-tinguishingamongthedi erentstar-formationhistorieswithbroad-bandphotometryalone.Inpractice,themostusefulnumberwederivefromtheMonteCarlocon denceintervalsforeachgalaxyistheuncertaintyonthestellarmass.Forthisquantity,itisimportanttoknownotonlytherandome ectsofphotometricerrors,butalsothesystematice ectofallowingτtovary.The1σMonteCarlostellarmassuncertaintiesincolumn7ofTable3,derivedfromsimulationsinwhichτwasallowedtovary,re ecttheuncertaintyinτ.WealsolisttheMonteCarlostellarmassuncertaintiesforCSFstar-formationhisto-riesinTable3.

Todemonstratethecovarianceofdi erentparame-ters,andatleastsomeoftherangeofstellarpopula-tionspresentinoursample,weshowin gure4thetsfvs.E(B V)andM vs.tderivedfromMonteCarlosimulations.sfcon denceinter-valsThesecon- denceintervalsdonotassumea xedvalueforτ,butaretheresultsofsimulationsinwhichτwasallowedtovary.Thecon denceintervalsdisplayedincludethoseofBX879(z=2.308),whichappearstobeaveryyoung(tstantaneoussf≤10Myr),dustystellarpopulationwithhighin-star-formationrateandlowstellarmass,allofwhichparametersareindependentofτbecauseoftheextremelyyoungage;BX691(z=2.190),anevolvedstellarpopulationwithtsf/τ>1,τconstrainedtobebetween0.5and2Gyr,moderatecurrentstar-formationrateandlargestellarmass;andBX563(z=2.296),forwhichtherearenoconstraintsonthestar-formationhis-tory,butthestellarmassofwhichisclosetothesamplemedian.ThelargeextentoftheBX563con denceinter-valintsfvs.E(B V)spacedemonstratessomeofthefundamentallimitationsofpopulationsynthesismodel-ing.

4.5.CaveatstotheModelResults

Thegoalofpopulationsynthesismodelingistoin-ferphysicalquantitiesfromobservedgalaxymagnitudesandcolors,inparticularthestellarmass.Inthebest-casescenariowewouldalsoliketoplaceconstraintsonthetimescaleofthecurrentstar-formationepisode,andthereforethelevelofdustextinctionandthedust-correctedstar-formationrate.Therearecertainlimita-tionsofthemodelingprocedure,however,thatareworthpointingout,inadditiontothesystematicuncertainties

Fig.5.—ObservedandmodelSEDsforall72ofthegalaxieswithspectroscopicredshiftsandmeasurementsintheKleastexponentiallyoneIRACchannel,sortedbyredshiftfromlowesttohighest.Thenumbersfortheinferredsbandandatdatapointsaredecliningexcluded.

model,whichisshownshownwithasolidcurve.Thedottedcurverepresentsstellarmassthebest- taretakenmodelfromwhenthebest- ttingtheIRACdiscussedinsection4.3.

anobscuredAGN(see§5.5).MD157showsrest-frameFirst,itisnotalwayspossibleto ndastatisticallyac-UVspectroscopicevidenceforhostingabroad-lineAGN,ceptable ttotheobservedgalaxycolors,especiallyusingwhichprobablycontaminatestheKrelativelysimplemodels.Sevengalaxiesinthe2sampleIRAC8.0µmband.TheSEDss ux(strongHα)andoftheremaininghavebest- tmodelswithsigni cantlyhigherχvaluesthreegalaxies(BX581(z=2.406),BX756(z=1.738),thanthoseoftherestofthesample.Twoofthesegalax-andBX681(z=1.740))areverydi culttounderstand,ies(BX490(z=2.403)andBX794(z=2.253))havewithbothlargeKlargeR Ksresidualsthatarelikelytobeexplainedpositiveandnegative,sandIRACresidualsthatarebothanddon’tstemfromobviousir-intermsofcontaminationfromHαemissionintheKregularitiesinanyoftheimagingdata.

band.Anothergalaxy,BX561(z=2.426),hasagoodsAsecondissueistheapparentdiscrepancybetween texceptfortheIRAC8.0µmpoint,whichyieldsasig-modelparametersbasedonthebest ttobroad-bandni cantpositiveresidual,possiblyduetothepresenceofphotometryandthoseinferredfromrest-frameUVspec-

Fig.5.—(continued)

tra,whichappliesforasmallnumberofgalaxiesinthecheckonthestellarpopulationparametersderivedfromsample:BX1032(z=1.883),BX1087(z=1.871),andtheSED- ttingprocedure.ThespectraofBX1032andBX536(z=1.977).Notonlydidweobtainstatisti-BX1087bothshowprominentCIVP-Cygnifeatures,callyacceptable tstotheobservedphotometryofthesewhichareproducedinthewindsofmassivestars,associ-galaxies,butalsotheresultsoftheMonteCarlosimu-atedwithactiveongoingstarformation.Therest-framelationsappeartoruleoutstar-formationhistorieswithUVspectraofBX1087andBX536bothcontainbroadτ>200Myr,andtofavorbest- ttHeIIemission,whichisproducedbyWolf-Rayetstars,quitemodestsf/τsigni cantlygreaterthanunity,withstar-formationanotherindicationofayoungstellarpopulation.Finally,ratesofafewM⊙yr 1.Thebroad-bandphotometryofthespectrumofBX1032exhibitsnebularemissionlinesthesegalaxiesimpliesstar-formationhistoriesinwhichsuchasCIII]andHeII,furtherindicationsofactivecur-thebulkofstellarmassformedinthepast,andthecur-rentstarformation.Inshort,therest-frameUVspectrarentstar-formationrateismuchlowerthanthepastav-oftheseterage.Wealsohaverest-frameUVKeckI/LRISspectrasf/τ>>1galaxiescontainfeaturesthatarenotconsistentwiththebest- tmodelsusedtodescribeforthesegalaxies,whichcanbeusedasaconsistencytherest-frameUVtonear-IRSED,whichisallthemore

Fig.5.—(continued)

strikingsincethemodelsappearfairlywell-constrained.thereforeweincludetheminallfurtherdiscussionofstel-Clearly,therearecasesinwhichthesimple,single-larmasses.

componentSED- ttingprocedurefailstocapturethetruenatureofthegalaxystellarpopulation.Rest-frame5.RESULTS

UVandopticalspectra,andX-ray,submillimeter,and5.1.TheStellarpopulationsofz～2Galaxiesinthe

radio uxes,providevaluablecomplementaryinforma-tionthatmustbeincludedtobetter-constrainallofHS1700Field

thestellarpopulationparameters.Thelimitationsin-InFigure5,weshowtheobservedoptical,near-IR,herentinSED tting,withoutexternalconstraints(andandSpitzerphotometryandbest- tmodelsfortheen-especiallywithoutspectroscopicidenti cations),shouldtiresampleofUV-selectedz～2galaxiesintheHS1700beborneinmindwhenconsideringtheresultingstel- eldwithspectroscopicredshifts,Ks,andSpitzer/IRAClarpopulationparameters.Ontheotherhand,thestel-photometry.Thesolidlinesindicatebest- texponentiallarmassesinferredforthesethreegalaxiesarelargelyBruzual&Charlot(2003)modelsincludingtheIRACindependentofthestar-formationhistoryassumed,and

measurements,whereasthedottedlinesindicate tstoonlyUnGRKsphotometry.Thestellarmassinferred

Fig.5.—(continued)

fromthebest- ttotheentireSED(includingIRACpho-tometry)isalsoshown,alongwiththeredshiftofeachobject.InTable3,welistthebest- tstellarpopula-tionparametersforeachobject,includingtheentireop-ticalthroughIRACSED.Parametersarelistedforthebest- tmodelassumingτ=∞(CSF),andalsofortheexponentiallydecliningmodelthatformallyyieldedthelowestvalueofχ2whenτwasallowedtovaryasafreeparameter.Thereisasubsetofgalaxieswithextremelyyoungbest- tages(tsf≤10Myr),forwhichallval-uesofτyieldnearlyidenticalstellarpopulationparam-eters.Thesegalaxiesinclude:BX928,BX1075,BX794,MD128,BX879,BX490,MD155,MD152,BX588,andMD126.Insuchcaseswheretsfissmallerthananyoftheτvaluesconsidered,thestar-formationhistory(asparameterizedbyτ)becomesirrelevant.

FrombothFigure5andTable3,itisclearthatalargerangeofrest-frameUVtorest-framenear-IRSEDsisincludedinoursample.Inparticular,wehighlightthevarietyofrest-frameUV/near-IRcolors,whichtranslatesdirectlyintoalargerangeofinferrednear-IRmass-to-lightratios.

5.2.DistributionofStellarMasses

Figure6showsthedistributionofstellarmassesin-ferredfromtheoptical–IRACSEDs.Themassforeachobjectistakenfromcolumn7ofTable3.Thenearlylog-normaldistributionofstellarmasscanbecharac-terizedbyameanandstandarddeviationlogM =10.32±0.51M⊙.FortheCSFmodels(theresultsofwhicharealsocompiledintable3),we ndadistribu-

tionlogM (CSF)=10.30±0.53.Figure7comparesthestellarmassobtainedassumingthebest- tτmodeltothestellarmassimpliedbythebest- tCSFmodels.Theagreement,bothintermsofthemassdistributionandthemassesofindividualobjects,isexcellent,withnoobvioussystematictrendoro set.

Forsomepurposes,itisofinteresttobreakdowntheinferredstellarmassdistributionasafunctionofob-servedKs-bandmagnitude,sinceanumberofothersur-veysofgalaxiesatsimilarredshiftsareselectedinthatband.TheresultsaresummarizedinTable4,wheretheKslimitsarechosentocorrespondtothelimitsimposedbysomeoftheseothersurveys.Asexpected,thereisacleartrendofincreasinginferredstellarmasswithin-creasingrest-frameopticalluminosity,although,asdis-cussedbelow,thereissigni cantscatterinrest-opticalM/Lata xedmassorluminosity.

Thetypical(median)MonteCarlouncertaintiesinin-dividualstellarmassestimatesare～30 35%,derivedfromsimulationsinwhichτwasallowedtovaryasafreeparameter.IfwemodelallgalaxiesinthesamplewithCSFstar-formationhistories,we ndthatthemedianMonteCarlostellarmassuncertaintyis～35%.Thesevaluesindicatethat,forthesamephotometryandun-certainties,CSFmodelsactuallyyieldstellarmassun-certaintiesatleastaslargeastheonesderivedwhenτisallowedtovaryasafreeparameter.

5.3.ResultswithandwithoutSpitzer/IRACPhotometry

OneofthefundamentallimitationsoftheanalysisinShapleyetal.(2001)andPapovichetal.(2001)was

Fig.6.—Histogramsofbest- tstellarmasses,with(shaded)andinstellarthewithouttext,(unshaded,theoveralldashed)massdistributionIRACphotometry.Asdiscussedinfactorthemassofmeasurementaresimilar～1.5 2whenoftoindividualwithin～15%,andthetotalinferredtheIRACstellaralthoughdataaremassestheincluded.

areuncertaintiessmallerbyathetheinabilitytodiscriminateamongdi erentstar-formationhistorieswithonlyrest-frameUVandopticalphotometry.Wehaveshownthat,evenwiththeaddi-tionofSpitzer/IRACphotometry,thisinabilitytocon-strainstar-formationhistoriespersistsformostgalaxiesinourz～2sample.Thebroad-bandSEDsarepar-ticularlydegeneratewithrespecttoage,dustcontent,andstar-formationrates,whiletheinferredstellarmasssu ersfrommuchlesssystematicuncertainty.Anotherquestionraisedinthepre-Spitzererawaswhetherornotrest-framenear-IRphotometryofz≥2galaxieswoulduncoverconsiderablymorestellarmassfromoldburststhanhadbeeninferredfromSEDswhoselongestwave-lengthmeasurementwastheground-basedKtherest-frameopticalatthesband,cor-respondingtoredshiftsofinteresthere.SuchstellarcompositepopulationsmighthavebeenrevealedthroughIRAC uxesinexcessoftherest-framenear-IRextensionofmodel tstotherest-frameUV

throughopticalSEDs.Figure8showstherelationshipbetweenM (NoIRAC),thestellarmassinferredwhentheIRACphotometryisexcludedfromthe t,andM (IRAC),thestellarmassderivedusingIRACphotometry.Figure6displaysthisrelationshipinonedimension,withthedistributionofM (NoIRAC)indicatedasanunshadedhistograminfrontoftheshadedhistogramofM (IRAC).Whatboth guresdemonstrateisthat,onaverage,thestellarmassinferredwithandwithoutIRACdataagreewithoutalargesystematico set,thoughthescatterinthisrelationshipissigni cantandlargerthanthetypical(formal)uncertaintiesintheindividualstellarmassestimatesusingthefullSED.Table4comparestheseinferredstellarmassdistributionsasafunctionoflimitingobservedKTheaverageoflogMlogMs ux. (NoIRAC)- (IRAC)=0.06,withastandarddeviationof0.35dex;theme-diandi erenceisonly0.01dex.We ndonly vegalax-

Fig.7.—Acomparisonoftheinferredstellarmassesforconstantstarmationformationments.models(CSF)(bestmodelsτ)andforthebest- tdecliningstarfor-onthereMonteTheCarloerrormodelingbarsareforgivenasthepresented72galaxieswithIRACmeasure-thephotometricinTableerrors.3,andNotearebasedmodels,arebetteragreementalthoughnoobviousinsystematicdi erencesbetweentheCSFandthatτwithsomethecasesobservedtheτdata.

modelsresultinsigni cantlyieswhoseoptical–IRACSEDsimplystellarmassesmorethanafactoroftwolargerthanthatinferredfromtheoptical–KBX898,whosesSEDs.Themostdramaticsuchexampleisobserved4.5µm uxis～3.5timeslargerthanthatpredictedbythe ttotheoptical–Kalone.Theinferredstellarmassis～12timeslargersSEDiftheIRAC4.5µmpointisincludedintheSED t.Un-fortunately,BX898hasasigni cantIRACmeasurementonlyat4.5µm,andcontaminationofthismeasurement(or,astatisticaloutlierintheobservedKout.

smeasurement)cannotberuledItismorecommonforthe ttotheoptical–KtooverpredictthestellarmassrelativetothatinferredsSEDfromtheoptical–IRACSED.Webelievethatthisdis-crepancyarisesinmostcasesbecauseofthecontamina-tionoftheK0≤z≤scontinuum uxbyHα+[NII]lineemissionat2.2.5.TheadditionoftheIRACphotome-trytotheSEDsofgalaxieswithstrongHαemissionsometimesrevealstheKsdatapointasapositiveoutlierrelativetotheoverallSED t(e.g.BX794andBX490),andthebest- tBalmerbreakissmallerthantheob-servedR Kscoloralonewouldimply.IntheabsenceofIRACphotometry,theKstellarsmeasurementstronglyin- uencestheinferredmassbecauseitistheonlyproxyforthestrengthoftheBalmerbreak,resultinginabiastowardsomewhathighervalues.Insupportofthisinterpretation,forthefourgalaxiesintheHS1700Ks/IRACpointingswithpublishedHαmeasurements,we ndanaverageratioof1.55betweenM (NoIRAC)andM (IRAC)whennocorrectionforHαemissionisapplied.WhenHαemissioniscorrectedfor,we ndanaverageM (NoIRAC)/M (IRAC)of1.11,i.e.closertounity.MorecompleteHαsurveysarenecessarytore-movetheKsemissionlinebias,especiallyinordertomodelobjectsforwhichKsisthereddestavailableband

Fig.8.—Acomparisonoftheinferredstellarmassesforthebest- twithtometry.IRACdecliningstarformationmodels(bestτ)forthe72galaxiesthestellarphotometricThemeasurements,errorerrors.barsarederivedTherebasedwithandwithoutIRACpho-areonveryMontefewgalaxiesCarlomodelinggiventrytowardsisexcluded.massissigni cantlyunderpredictedwhenIRACforwhichphotome-theextrahigherItismorecommonfortheno-IRAC tstobiasedTheKredcontributionstellartomasses,theKinatleastsomecasesbecauseofthes-bandlightbyHα+[NII]emission.forband,pointswhilearethegalaxiesbluepointswhoseareredshiftsobjectsplacewith2H.016αoutside<zofthesemissionwhichissigni cantpossible.

contaminationoftheK<by2.504smeasurementHα(Erbetal.2005).

WithoutIRACphotometry,we ndtypicalMonteCarlostellarmassuncertaintiesof50 60%forCSFmod-els,whichis～1.5timeslargerthanthetypicalstellarmassuncertaintiesestimatedwhenincludingtheIRACphotometry.Whenτisallowedtovary,thetypicalstel-larmassuncertaintyis60 70%,～2timeslargerthanthetypicaluncertaintyinferredwithIRACphotometry.Thus,whilethelackofIRACphotometrydoesnotap-peartocausealargebiasinthestellarmassestimates,theinclusionofIRACphotometrysigni cantlyreducestheformaluncertainties.

WhilewehavenotedthepossibilityofHαcontamina-tionoftheKcurrentsamples-bandmeasurementsforthe～60%ofthehavingredshiftswherethisisapoten-tialproblem,thetotalinferredstellarmassforthefullsampleof72galaxiesdi ersbyonly～15%whenIRACmeasurementsareincludedorexcluded(inthesensethattheinferredtotalmassishigherwithouttheIRACdataincluded).Thus,whiletheadditionoflonger-wavelengthdataleadstosigni cantlysmalleruncertaintiesinthees-timatesofthestellarmassesofindividualgalaxies,theIRACdatadonotappeartobeessentialforobtainingreasonablestellarmassestimatesforgalaxiesatz～2,particularlywhenstatisticsoflargesamplesaremostrel-evant.

5.4.Mass-to-LightRatiosversusWavelength

Itisoftenassertedthatestimatesofstellarmassfromphotometricmeasurementsbecomeincreasinglyreliableasoneobtainslongerwavelengthdata.Wehavecertainlyfoundthistobethecase,asdiscussedabove;however,it

isnottruethatlong-wavelengthobservationscanbeusedtomeasurestellarmasseswithoutunderstandingthere-centstar-formationhistoryofthegalaxy,evenwhenob-servationsextendintotherest-framenear-IRasfortheIRACobservationsofz～2galaxies.Thispointisillus-tratedin gures9and10;wehavechosentheobservedKsbandandtheobservedIRACchannel2(4.5µm)bandsincetheformeristhelongestwavelengthmeasurementpossiblefromtheground,andthelatterliescloseto～1.5µmintherestframeatthemeanredshiftofthecurrentsample(68/72ofthegalaxiesinthesamplehavebeenmeasuredinIRACchannel2),wherethepeakintheSEDproducedbyolderstellarpopulationsisexpected.Figure9showstheabsolutemagnitudeofthegalaxiesbasedontheirobservedKthemedianredshiftof z s-bandmagnitudes,whichat=2.29correspondstorest-frame0.65µm(i.e.,Rband).NotethatwhilethereisclearlyacorrelationbetweenabsoluteR-bandmagni-tudeandstellarmass,thereisalargeenoughrangeinstar-formationhistoryamongthesamplethatameasure-mentofaparticularrest-frameluminositywouldmapquitepoorlyontostellarmass.ThedottedanddashedcurvesshowthatthereisarangeininferredM/Lofafactorof～70amongthesample.Thislargefac-torappliesevenforobjectsthatarebrightintheKband(Kindicatedwithopencircles.Theveryslargeststellars≤20.0),M/Lvaluesapproachthatofpresent-daygalaxies(Belletal.2003),buttheaveragevalueis～5timessmaller, (M/L)R =0.53(wherebothMandLareinsolarunits,andLisevaluatedintherest-frameRband).Thesituationisimprovedat4.5µm(～rest-frame1.4µm),asshownin gure10;thescatterinthestellarM/Lisafactorof～15,withindicationsthatthescatterdecreasesforthemostluminousgalaxiesintherest-framenear-IR,whichtendtohaveM/Lveryclosetothatofpresent-daygalaxies.ThescatterinM/Lisalsogreatlyreducedwhenthesampleislimitedtoob-jectswithR Ks>3.5(thereddestquartile),indicatedwithopensquares.Still,amongthetotalsamplestud-iedhere,theaveragevalueis(M/L)the1.local4µm=0.38,roughly2.5timessmallerthantypicalinuniverseatthesamerestwavelengths.ThelocalM/Lratioswerede-rivedwithanIMFcontainingfewerlow-massstarsthanarefoundinthestandard0.1 100M⊙SalpeterIMF,usedforthecurrentwork.Adoptingthe“diet”SalpeterIMFusedinBelletal.(2003)wouldlowertheaveragez～2M/Lratiosbyafactorof～1.5.

Asimpleinterpretationoftheseresultsisthatgaug-ingtheimportanceofcurrentorrecentstarformationonthegalaxyluminosities,evenintherest-framenear-IR,dependsontheobservationsofthefullUV-to-near-IRSEDofthegalaxy.Thatis,onecannotgodirectlyfromnear-IRluminositytostellarmasswithoutknowledgeofthelikelycontributionofyoungstarstothatluminosity,whichcanonlycomefrommeasurementsatshorterwave-lengths.TheregionofthespectrumthatismostsensitivetotheratiobetweenthecurrentstarformationandtheintegralofpaststarformationisbetweentheUVandthevisual,convenientlymeasuredbytheobservedR Kcolorinthecurrentsample.Thispointisillustratedins gure11,whichshowsthattheinferredstellarmassisremarkablywell-correlatedwithR Ks.Infact,thecor-relationbetweenR Kscolorandinferredstellarmassisalmostassigni cantasthecorrelationbetweenstellar

Fig.9.—Inferredstellarmassforthebest- tmodel,versustheinbandstellarabsolutemass/lightmagnituderatiosatatKs(2.155µm)Notethatthescatterofare70.atThethe9meanobjectsredshiftintheofsampletheagivensample)Kscanluminositybeaslarge(rest-frameasafactorRreddestindicatedwithquartilewithofopenthesamplecircleswithsurroundingwithKs≤20R Kthe(Vegapoints,normalized)whilethes>3.stellopensquares.Dashedanddottedlinesbracket5aretheindicatedrangeofatinrest-framearM/LinRtheband.sample,andaregiveninsolarunits,evaluateduniverse,z～2whilegalaxiestheapproachesTheveryaveragevaluethatlargestisofopticalstellarM/Lseen～5galaxiestimessmaller.

inthepresent-daymassandMofthe1correlations.4µm–usingaSpearmantest,thesigni -canceare5.9σand6.4σ,respectively–andmoresigni cantthanthecorrelationbetweenstel-larmassandrest-frame0.65µmabsolutemagnitude(5σ).Oneimplicationofthisresultisthat,intheabsenceofIRACdata,itmaybemoree ectivetouseR KthenKscolorrathers uxtoselectz～2galaxieswithlargestellarmasses.Anotherimplicationisthattherearefewgalaxiesthathavelargeinferredstellarmassesatz～2without

showingevidenceforlongstar-formationhisto-ries(cf.,Shapleyetal.2004)thataresuggestedbyredrest-frame2000–6500 Acolors.

Ofcourse,cautionisrequiredinassessingthetruesig-ni canceofcorrelationsthatarebasedonthemodelresultsandnotondirectlymeasuredquantities;suchisthecaseforallstellarmassestimates.Tosomeex-tenttheresultsinthissectionsimplyre ecthowtheobservedSEDisusedtoinferstellarmassbythemod-elingsoftware,asasinglesmoothlydecliningorcontin-uousepisodeofstarformationthatdominatestheemis-sionatallwavelengthsfromrest-frameUVthroughnear-IR.Insuchaframework,theratioofrest-frameUVtoopticalornear-IRlightindicatesthestellarM/Linafairlyrobustmanner(seesection4.3).Wecanobtainadi erentresultforthestellarmassifweinsteadun-derstandthegalaxyUV-to-near-IRSEDasthesuper-positionofayoung,roughlycontinuousepisodeofstarformationandanoldburstthatpeakedsometimeinthepast.Inthemostextremecaseofamaximally-oldburstandaveryyoung(tsf≤10Myr)continuousepisode,theoldburstwilldominatetherest-framenear-IRlight,theyoungepisodewilldominatetheemissionatrest-frameUVwavelengths,andthemaximumpossiblestel-

Fig.10.—Inferredstellarmassforthebest- tmodel,versustheabsoluteluminositymagnitudeobjectsatthemeanat4.5redshiftµm,equivalenttotherest-frame1.4µmindicatedwithrangewithK0arecircledofandthethosesample.withAsR forK gure9,thes≤20.s>3.5areevaluatedofstellaropenisatrest-frameM/Lsquares.in1the.4µm.sample,DashedanddottedlinesbrackettheWhileandaregiveninsolarunits,itverynosity.canstillmuchbereducedaslargecomparedasafactortothattheof10measurescatterataininthestellarKM/Lsband,theparablemostThemassivez～2galaxiessamplehavingspansaattainedrangeofgiven4.5µmlumi-aanear-IRfactorofM/L15,com-with(scatterM/L)topresent-daygalaxies,butwithtypicalgalaxieshaving1in.4µmnear-IR 0.4,M/Lafactorforobjectsof～2.with5smaller.R KThere5.

ismuchlesss>rmasswillbeinferredfromtherest-framenear-IRlu-minosities.Whilethisscenarioisratherunrealistic,itdoesprovideanextremeupperlimitonthestellarmassimpliedbytherest-framenear-IRlight.Lessextrememultiple-componentmodelsprovideanaturalrepresen-tationoftheepisodicstar-formationhistoriesthatariseinahierarchicalmodelofgalaxyformation,andwillal-waysresultinlargerstellarmassestimatesthanthosederivedfromthesingle-componentmodelingpresentedthusfarinthiswork.

Inordertosetaroughupperlimitonthestellarmassesimpliedbythebroad-bandSEDs,weusedatechniqueverysimilartothatofDaddietal.(2004b),whichisde-signedtoinferthemaximumM/Lratiofromtherest-framenear-IRdatapoints.Thismethodaccountsformostofthelightatthosewavelengthswithamaximally-oldburst,ontopofwhichissuperimposedayoungstellarpopulationthataccountsformostoftherest-frameUVemission.Theagesofthemaximallyoldmodelsrangefrom2.2-4.3Gyrforthegalaxiesinoursample.We rst tayoung(tsf≤10Myr)CSFmodeltotheobservedphotometry,normalizingthismodeltotheRmagnitude.Wethensubtractedthisyoungmodelfromtheobservedphotometry,normalizedthemaximallyoldτ=100MyrbursttotheresidualIRACmagnitudes,andsummedthestellarmassfromthetwocomponents.Thissumisdominatedbythemassoftheoldcomponent,andistyp-ically～3timeslargerthanthesingle-componentmass.ThecorrelationbetweenMsamelevel1.4asµmandstellarmassremainssigni cantattheforthesingle-componentmodels.WhileR Kscolorisstillsigni cantlycorrelated

Fig.11.—Therelationshipbetweeninferredstellarmassandthe2.26observed±R Kcolorforthesampleof72galaxieswith z withR0K .30.≤KThereisaverystrongcorrelationbetweeninferredM= ands20color,.0areparticularlycircled.

forR Ks>3.5.Again,objectsswithstellarmass(atthe～4σlevel),thecorrelationislesssigni cantthanforthecaseofthesingle-componentmodelsdiscussedabove.Speci cally,thereismuchmorescatteramongobjectswithblueR KtheM/Lcanchangethemostusingthisscolors,forwhichtypeofmodel-ing.ForobjectswithR Ks>3.5,theinferredM/LandthecorrelationbetweenstellarmassandR Kcolorarefairlyinsensitivetothetypeofmodelused.sItiseasytoseethattheamountofadditionalstellarmassthatcouldbe“hidden”bythecurrentstarforma-tionepisodeisstronglydependentontheobservedratioofUVtooptical/near-IRlight.Forthebluestobjects,theratiooftruestellarmasstoinferredstellarmasscouldbe≥5whilereddergalaxies(exceptinextremecasesofdustobscuration)couldnothidemuchadditionalstellarmass.Forthegalaxiesinthesamplewiththelargeststel-larmassesderivedfromsingle-component ts,thetwocomponentmodelingprovidesstellarmassupperlimitsonly10 40%higherthantheoriginalestimates.Thus,ingeneral,objectswhichalreadyhavelargeinferredstellarmassesandoldinferredagescannotbemuchalteredbymoreextremeassumptions,whileobjectsthatarecom-pletelydominatedbythecurrentepisodeofstarforma-tion(thesewouldtendtobegalaxieswithveryyounginferredages)couldbemoremassivebyfactorsaslargeas～5 10.ThetypicalBX/MDgalaxy,whichisbe-tweentheseextremes,probablyhasatruestellarmassthatisafactorof≤3largerthanthevalueinferredfrompopulationsynthesismodeling.Finally,thisdecomposi-tionoftheSEDintoyoungandmaximally-oldcompo-nentsprovidesafairlyrobust(butextreme)upperlimittothestellarmassconsistentwiththeobservedphotom-etry,sincemorerealistic,longercurrentstar-formationepisodes,wouldmakeanincreasingcontributiontotherest-framenear-IRlight,andmorerealisticpastburstswouldoccuratz<∞andhavesmallerM/Lratios.

5.5.

AGN

Fig.12.—TherelationshipbetweentheabsolutemagnitudeintheredshiftobservedindicatesoftheRsample)band(equivalentandinferredtorest-framestellar2100 Aatthemeanlacknocorrelationofthesetwoquantities.mass.ThereASpearmanisatestframeofthosewith～correlation1500 betweenMandtheabsolutemagnitudeatsimilar rest-R AK.Again,objectswithKs≤open20.0squares.

arecircledands>3.5areindicatedwithAsdescribedabove,thereisoneobject,MD157,whosespectrumshowsclearsignaturesofAGNactivityinthefar-UV(strongCIVandHeIIemissionaswellasabroadLymanαline).Inaddition,MD157isextremelybrightintheIRAC-4band(ABmagof18.51;seeTable2),suchthatobtainingareasonable ttoitsSEDusingstellarpopulationsrequiredexcludingthe8µmdatapoint.Twootherobjects,MD174andMD94,haverest-frameopti-calspectraindicativeofAGN(bothhaveHα/[NII] 1,andbroadHαwithvFWHM≥1000kms 1).Allthreeobjectsexhibitsigni cantlyredderthanaverageR Kcolors,rangingfrom4.46 5.15andbest- tstellarmassessrangingfrom8×1010M⊙to4.5×1011M⊙.

AnimportantquestionishowmuchthepresenceoftheAGNa ectsthestellarmassestimates.Thelargestellarmassesareinferredbecauseofthebrightrest-framenear-IRphotometryandtheveryredrest-frameUV/opticalandUV/near-IRcolors,butifadustyAGNweredom-inatingtherest-IRpartoftheSED,thesestellarmassinferencescouldbeerroneous.Alargefractionofsimilarlybrightandredz>other

2galaxieshavingrelevantspectroscopicmeasurements～

alsoindicatethepresenceofAGN.Forexample,4ofthe7galaxieswithspectroscopicredshiftsfromtheFIRESsurvey(Franxetal.2003)pre-sentedbyvanDokkumetal.(2003,2004)exhibitUVand/oropticalsignaturesofAGNactivity(highioniza-tionUVlines,broadHαandHα/[NII]lineratiosnearunity),similartoMD157,MD174,andMD94.Itisdi -culttoknowwhetherAGNactivityleadstoerroneouslylargestellarmassestimates,orwhetheritispreferen-tiallyobjectsthathavealreadyaccumulatedlargestel-larmassesthattendtoharborsigni cantAGNactivity.WhentheAGNisnotobviouslyenergeticallydominantintheobserved-framemid-IR,wesuspectthatthestellarmassestimatesareprobablyreasonable;forobjectslikeMD157,whichisobviousasanAGNinboththerest-framefar-UVandnear-IR,itisprobablysafesttoassume

thatthestellarmasshasbeenover-estimatedusingSED tting.

Thereareotherobjectsthatappeartohave uxesat8µmthataresigni cantlyinexcessofthatexpectedfromtheSEDmodeling(e.g.,BX756,BX535,MD69,BX505,BX581,BX561,BX523)butwherewedonothaveanyobviousspectroscopicevidenceforthepresenceofanAGN.Itisconceivablethatthesemid-IRexcessescouldbeduetoobscuredAGNactivity;deepChandraimagesofthis eldcouldhelpresolvethisquestion.

6.DISCUSSION

Wehaveseenabovethatmodelingofthebroad-bandSEDsofz～2starforminggalaxieshassigni cantlimitations,evenwithspectroscopicredshiftsandwhenlonger-wavelengthdataareavailablefordirectmeasure-mentoftherest-framenear-IRlight.Wehaveshownthatreasonablyconsistentstellarmassestimatesareob-tainedindependentoftheexactformofthestarforma-tionhistory,butthatthisconsistencyshouldbeviewedwithcautionbecauseofthefactthatarobustepisodeofstarformationcanchangetheinferredM/Lbyafactorofseveral,evenatrest-frame～1.5µm.Oneobtainsin-creasinglyreliableestimatesofthetotalstellarmassofagalaxyastheintegralofstarformationoverthelifetimeofthegalaxyexceedsthatinthepastseveralhundredMyrbyafactorofafew.Inthemostreliablecases,wheretheUVlightisfarsurpassedbytheemissionatrest-frameopticalandnear-IRwavelengths,theinferredstellarmassdoesnotdependverystronglyonwhetherasmoothorepisodicprescriptionisusedtomodelthewayinwhichtheintegratedstellarmassaccumulatedasafunctionoftime.Whentherest-frameUVluminos-ityismorecomparabletorest-frameopticalandnear-IRluminosities,however,itisinevitablethatthestellarmassesinferredfromphotometrictechniqueswillre ectthetypicalstellarmassproducedinanepisodeofstarfor-mation,perhapscausedbyamergeroraccretioneventoverthecourseofthelastfewhundredMyr,andnotthetotalstellarmass.Acorollarytotheaboveargumentsisthatitismostlikelyonewouldmeasurelargestel-larmasses(usingthetypeofmodelingdescribedinthispaper)whenthecurrentrateofstarformationissigni -cantlysmallerthanthepastaverage.Otherwise,mostoftherest-framenear-IRandopticallightwillbeascribedtothecurrentstar-formationepisode,whichhasmuchsmallerM/LevenintheIRACbands.Thesearenotnewarguments(see,e.g.,Papovichetal.2001;Shapleyetal.2001),thoughtheinclusionofIRACwavelengthstocon-straintherest-framenear-IRemissionplacesthemonamorerobustfooting,andthelargerrangeofgalaxypropertiesobservedinthecurrentsamplehighlightstheimportanttrendsmoreclearly.

Theconceptofepisodicstarformationinformstheabovediscussion.Whilesuchvariablestar-formationhis-toriesappearareasonablerepresentationofthebuild-upofstellarmassingalaxiesathighredshift,sim-plybasedongeneralconsiderationofthehierarchicalmodelofgalaxyformationinwhichmerging(andmerger-inducedstarformation)playsalargerole,wealso nddirectevidencefromthegalaxiesinoursample.Specif-ically,wecanobtaininterestingconstraintsonthered-shiftatwhichtheonsetofstarformationoccurredforthegalaxiesinoursample,despitelargesystematicun-

certaintiesinthestellarpopulationages(section4.3).Themaximumstellarpopulationageisobtainedforagivensetofrest-frameUVtonear-IRcolorswhenacon-tinuousstar-formationhistoryisused,sowecanusetheCSFresultstoinfertheearliestredshiftsatwhichthecurrentepisodesofstarformationcouldbegin(cf.,Fergusonetal.2002).ThemedianCSFageforthesam-pleis～ingtheredshiftandCSFagedistri-butionswe ndthatthecurrentepisodeofstarformationhadnotbegunyetatz=3foratleast～65%(46/72)ofthegalaxiesinthesample.Thetruefractionisprobablyhigher,sinceCSFmodelsprovideonlyanupperlimitontheageofthestellarpopulationdominatingthecur-rentstar-formationepisode.Therefore,we ndthatthemajorityofgalaxiesinoursamplehaveyoungagesforthecurrentepisodeofstarformationcomparedtotheageoftheUniverseatz～2,allowingthepossibilityofmucholderpreviousburstsofstarformation.Relativelyyounginferredageswouldnaturallyariseinapictureofepisodicstar-formation.

Arelatedpointisthatselectingonthebasisoffar-UVproperties,aswehavedonefortheBX/MDsample,connectsonlyobliquelytostellarmass.Onemightex-pectthatgalaxieswouldhavetoattainaparticulartotalmassthresholdbeforetheycouldsupportstarformationdetectableattheapparentmagnitudelimitofthesur-vey,butoncethatthresholdisexceeded,acombinationofextinction(whichisprobablystar-formationratede-pendent;e.g.,Adelberger&Steidel(2000))andscatterinSFRperunittotal(orstellar)massmight“random-ize”themixofgalaxypropertiesincludedinthesample.Indeed,asshownin gure12,we ndthatthereisnohintofacorrelationbetweenMUVandinferredstellarmass,aswasfoundpreviouslyinthez～3LBGsampleofShapleyetal.(2001).

Theseissuesmakeitparticularlydi culttomeasurethetotalstellarmassthathasformedbyaparticularredshift(Dickinsonetal.2003;Glazebrooketal.2004;Fontanaetal.2003),ortomakeadetailedcomparisonofthestellarmassofevensimilarly-selectedgalaxiesatdi erentredshifts.Forexample,analysisofthespa-tialclusteringofz～3LBGsandz～2BXgalax-ies(Adelbergeretal.2004a)suggeststhatcorrelationlengthofLBGdescendantsatz～2wouldbesimilartotheobservedBXcorrelationlength,withtheimpli-cationthatz～2LBGdescendantsresideindarkmat-terhalosinthesamemassrangeasthosehostingBXgalaxies.Itisthereforeofinterestalsotocomparethestellarmassdistributionofspectroscopically-con rmedBX/MDobjectswiththatofz～3LBGsfromthesam-pleofShapleyetal.(2001).Adirectevolutionarycom-parisonbetweenthestellarmassdistributionsforz～2andz～3UV-selectedsamplesisnotstraightforwardformanyreasons.Mostimportantly,whilethereislittledoubtthattheobjectsatz～2andz～3withlargeinferredmassesandagesreallyaremassiveandold(seesection6.2),themodelingresultslikelyrepresentlowerlimitsonthestellarmassesformoretypicalobjectsinbothsamples,wherethecurrentstarformationispro-ducingenoughopticalandnear-IRlighttopotentiallymaskpreviousgenerationsofstars.Ifstarformationisepisodic,asseemsquitelikelyonmanygrounds,thenitseemspossibleinprinciplethatthesamegalaxycouldbemeasuredatdi erenttimes,buttheinferredstellarmass

mightnotevenbeamonotonicallyincreasingfunctionoftime.Similarly,itmightbedi culttotracetheevolu-tionofstellarmassduringtheepochwhenmostgalaxiesarestillveryactivelystar-forming.

Bearingalloftheseconcernsinmind,wenowsumma-rizetheinferredpropertiesoftypicalUV-selectedgalax-iesatz～2.Wethenturninsomedetailtothemostmassiveandoldestgalaxiesinthesample,forwhichweobtainrobusttotalstellarmassmeasurements,andshowhowthesemassivegalaxiesprovideplausibleprogenitorsforpassivegalaxiesidenti edatz～1.5 2.0.Finally,wemakeageneralcommentabouttherelationshipsamonggalaxypopulationsselectedwithdi erentphotometriccriteria.

6.1.ThePropertiesofthe“Typical”UV-Selected

Galaxyatz～2

Theresultsofthispapersuggestthatthetypicalz～2.2“BX”galaxyhasastellarmassof～3×1010M⊙,anumberthatcouldpossiblybehigherbyafactorof～2 3withouta ectingtheobservedSEDs.Al-thoughthelengthoftimestarshavebeenforminginin-dividualgalaxiesisnotwell-determinedbythemodeling,Steideletal.(2005)showthatgalaxieswithinthesignif-icantover-densityatz=2.30areolderandcontainmorestellarmassonaveragethanthoseinlower-densityen-vironments.Itisveryclearfromtheseresultsthatevenatz～2thestar-formationhistoryisstronglydepen-dentonlargescaleenvironment–moresothanontheparticularredshiftatwhichthegalaxiesareobserved.Thetypicalcurrentstar-formationrateis～50M⊙yr 1(Reddy&Steidel2004),althoughunobscuredstarfor-mationdownto～3M⊙yr 1couldmakeitintothesample,andweknowthatamongtheBXsamplesub-mmsourcesthathaveimpliedSFRofupto>lurk

1000M(Chapmanetal.2004).UV-selectedgalaxies

～

⊙yr 1thatarealsoopticallybright(K<20)havetypicalinferredSFRhigherbyafactorofs～

～3(Reddyetal.2005a).

Thespacedensityofgalaxies similar3tothosedis-cussedinthispaperis～3×10Mpc 3atz 2.2(Reddyetal.2005b).Theco-movingcorrelationlengthofthe1UV-selectedgalaxiesatz～2isr0=4.2±0.4h Mpc(Adelbergeretal.2005),implyingthatthetyp-icalgalaxylivesinadarkmatterhaloofmass～2×1012M⊙.Assumingthatthegalaxieshaveabaryonfrac-tionthatisthesameastheuniversalvalue,thees-timated11massinbaryonsassociatedwiththehalois～3×10M⊙,suggestingthatonaverageroughly10%ofthebaryonshavebeenconvertedintostars.Con-siderationoftheexpectedevolutionoftheclusteringpropertiesoftheUV-selectedgalaxiessuggeststhatbyz～1theywillbefoundmainlyamongearly-typegalax-ies,andwillhavecompletedmostoftheirstarforma-tion.Themostmassiveexampleswillhavecompletedtheirstarformationevenearlier,andexhibitsigni cantlystrongerclusteringthantheaverageUV-selectedgalax-iesatz～2(Adelbergeretal.2005).Asdiscussedfur-therinsection6.2andotherrecentwork,the～2.5Gyrbetweenz～2andz～1marksthetransitionofalargefractionofmassivegalaxiesfromstronglystarformingtolargelyquiescent(Daddietal.2004a;Shapleyetal.2004;Cimattietal.2004;McCarthyetal.2004;Fontanaetal.2003).

6.2.MassiveGalaxiesinRest-UV-SelectedSamples

Themostmassivegalaxiesinthecurrentsample,withstellarmasses>1011M⊙,cannotcontainsigni -cantadditionalstellarmassthatismaskedbythecur-rentepisodeofstarformation.Weobtainroughlythesamestellarmassestimateregardlessofwhetherasingle-componentmodelisusedto tthecolors,orayoungcurrentstar-formationepisodesuper-imposedonanoldpastburst.Inthecontextofsimple,single-componentmodels,thesegalaxiesalsohaveSEDsthatprovidesig-ni cantconstraintsonthestar-formationhistory.UnlikethemajorityofgalaxiesintheHS1700sample,forwhichwe ndthatallormostτvaluesareconsistentatthe95.4%levelwiththedata,wecanruleoutmodelswithτ<200Myrandτ=∞(i.e.,CSF)forallbutoneofthegalaxieswithM >1011M⊙.Morediscriminationamongτvaluesleadstointerestingconstraintsonthedurationofthestar-formationepisodesinthesegalaxies.Theminimumτ=200Myrforthese vegalaxiesim-pliesminimumagesoftvalueofτhoweversf≥600 900Myr.Themostlikelyis1 2Gyr,basedonMonteCarlosimulationsinwhichτwasallowedtovary.Forτ=1 2Gyr,theagesinferredforthesegalaxiesrangefrom2 3Gyr,comparabletotheageoftheUniverseatz～2.3.Thestellarmassesforthissubsampleofgalaxiesareallwell-constrained(≤20%uncertainties),includingthesystematicuncertaintyinstar-formationhistory8.Thesuperiorconstraintonstar-formationhistoryfortheseobjectsresultsfromtwofactors: rst,themostmassivegalaxiesareamongthebrightestintheKbands9,resultinginsmallerthanaverageRsandIRAC KandIRACphotometricuncertainties.Fourofthe vesmassivegalaxieswithsigni cantconstraintsonτarealsodetectedinallfourIRACbands,providingbettercon-straintsontheshapeoftheSEDintheregiondominatedbylonger-lived11stars.Secondly,allbutoneoftheobjectswithM >10M⊙havesigni cantlyredderthanaver-ageR Ks(andR IRAC)colors10,butaverageorbluerthanaverageG Rcolors.ThecombinationofbluetoaverageG RandredR Knear-IRscolorsindicatesthattherest-frameopticalandlightcannotbeascribedtothemassivestarstracingtheinstantaneousstar-formationrateanddominatingtherest-frameUVemission,andthereforemustbeproducedbylessmas-sivestarsformedprimarilyinthepast.SincetheinferredextinctioncorrectionwillbesmallwhentheUVcolorisblue,eitherthestarformationwasmuchhigherinthepast,orthecurrentSFRhasbeensustainedforaverylongtime.IftheimpliedCSFtimeisolderthantheageoftheuniverse,thenτmodelswillberequiredto tthedatawell.

ThereareotherobjectsinthesamplewithcomparablysmallR KsandIRACuncertainties,butwithweaker

quiteWenotethatthemassivegalaxiesinthecurrentsampleareweresimilartothosepresentedbyShapleyetal.(2004),which9alldrawnfromanotherofoursurvey elds,Q1623+26.

2002)We11thatnote,onlyforhalfcomparisonofthegalaxieswiththeinK20oursurveysample(CimattiwithMetal. >10thanM⊙alsohaveKs≤20.0.Adeeperapparentmagnitudelimitthe10mostKs=massive20.0isgalaxiesevidentlyatnecessaryz～2(Glazebrooktogainacompleteetal.2004).

censusofMostofthegalaxieswithM >1011M⊙

withhavebest- tmodelsFIRESJ survey.

Ks>2.3,implyingthattheywouldbeselectedbythe

constraintsonτ.AnexampleisMD94,whichhasveryredR KsandG RcolorsandisinferredtohaveastellarmassM>1011M⊙.Inthiscase,thecolorscanbeadequately tbymodelsspanningthewholerangeofτ,includingCSF,mainlybecausetheinferredcurrentSFR(derivedfromtheextremelyredG Rcolor)ishighenoughtohaveproducedthenear-IRlightinmuchlessthantheageoftheuniverse,evenforCSF.Thebest- tτmodel,though,hasanagethatisafactorof～3shorterthantheCSFmodel,andacurrentSFR3timessmaller(however,theimpliedinitialSFRgiventheageandτvalues1wouldhavebeen25timeslarger,or2500M⊙yr !).

Wereturnnowtothemassivegalaxieswithsuperiorconstraintsonthesingle-componentstar-formationhis-tories.Ofcourse,asmoothly-decliningstar-formationhistoryisnotaverylikelyscenarioforthebuild-upofstellarmass,sincethestar-formationrateisexpectedtobemodulatedbyepisodicmergerandaccretionevents.Itisnotreallypossibletoestablishwhetherornotstarfor-mationhasbeensustainedcontinuouslyatadetectableleveloveralongperiodoftime,evenforgalaxieswhosestarformationmusthavebegunatveryhighredshift.Forexample,ingeneral,galaxiesforwhichtheτmodelsarerequiredforagood ttosingle-componentmodelscanalsobeadequately tbyassumingasingle,earlyburstofstarformationtoproducetheoptical/near-IRlightandasmallamountofrecentstarformation(withanassociatedstellarmassofonly～1 2%oftheearlyburst)thataccountsfortherest-framefar-UVlight.Forsuchtwo-componentmodels,thereisagainnouniqueso-lution,butreasonable tstothedatacanbefoundformodelswherealargeburstwithτ=100Myroccurredatz>4andtherecent(smaller)burstislessthan100Myrold11.Whilewecannotdistinguishbetweensmoothorepisodicstar-formationhistories,theestimatesoftotalstellarmassderivedfromtwo-componentmodelsagreeverywellwiththoseobtainedfromthesimplesingle-component ts.Forthesemassiveandoldgalaxies,theinferredtotalstellarmassisthereforearobustquantity.InananalysisofthestellarpopulationsofUV-selected2.0≤z≤3.5galaxiesintheHubbleDeepField(HDF-N)Papovichetal.(2001)foundnocomparablymassiveandoldobjects.Thesmallratiobetweenthetypicalinferredstar-formationtimescalefortheseHDFLBGsandthetimeintervalbetween2.0≤z≤3.5wasusedtomotivateascenarioofmultiple,intermittentstar-formationepisodeswithadutycycleof≤1Gyr.Asdescribedinthebeginningofsection6,mostofthecur-rentsamplecaninfactbedescribedinthismanner.Inretrospect,itisnotsurprisingthatthesampleofPapovichetal.(2001)didnotcontainanyM >1011M⊙galaxiesinthe5.0arcmin2HDF-N eld,despiteasimilaropticalmagnitudelimitintheHDFspectro-scopicsub-sample.Wefoundonly26withspectroscopicredshiftsintheentire65arcminWIRC eld.Sincethespectroscopiccompletenessofoursampleis～22%,theapproximatesurfacedensityofobjectsintheredshiftintervalz 2.3±0.4is～0.5arcmin 2,sothatonewouldexpect<2suchmassiveandoldgalaxiesinan

～between1500WeMnotethatsuchmodelswouldimplySFRsthataveraged⊙z=yr 15andoverz=the4.4.

rst200Myrofthegalaxy’slife,e.g.21

areawiththesizeoftheHDF-Ngivenspectroscopicin-completeness.Infact,Labb´eetal.(2003)havepointedoutthediscrepancyinthenumberofgalaxieswithrednear-IRcolorsintheHDF-NversustheirFIRES eldintheHDF-S,withtheHDF-Nhavingonly10-20%ofthenumbersintheHDF-S.Giventhatthegalaxieswithlargestellarmassarestronglyclustered(Daddietal.(2004b),Adelbergeretal.(2005))thevariancebetweensmallsur-veyareasisexpectedtobelargeforsuchrareobjects.

6.3.UV-SelectedProgenitorsofMassiveandPassive

Galaxies?

WhilemassiveandmaturegalaxiesrepresentasmallfractionbynumberoftheBX/MDpopulation(6/72withM≥1011M⊙,5ofwhichalsohavetSF～theageoftheUniverse),theirestimatedspacedensity,～10 4Mpc 3,issimilartothespacedensityofthemassivestar-forminggalaxiesfromtheK20surveypresentedbyDaddietal.(2004b)at1.7≤z≤2.3,andalsothemassivegalax-ieswitholdandpassivestellarpopulationspresentedbyCimattietal.(2004)andMcCarthyetal.(2004)at1.3≤z≤2.0.Inbothofthelatterpapers,theauthorssuggestthattheprogenitorsoftheoldandpassivegalax-iesaretobefoundamongsubmillimeter-andnear-IR-selectedsamplesatz>2.Hereweshowthatthemostmassiveandmatureobjectsinthe2.0≤z≤2.6UV-selectedsamplecouldquiteplausiblybetheprogenitorsofmassiveandpassivesystemsat1.3≤z≤2.0.Tomakethiscomparison,weadoptstarformationhistoriesthatareconsistentbothwiththoseusedto tthenear-IRselectedmassiveandpassivegalaxiesbyCimattietal.(2004)andMcCarthyetal.(2004)andwiththeSEDsofall5oftheBX/MDmassiveandoldgalaxiesatz～2.3.WethenevolvethesemodelsforwardintimetothemeanredshiftsoftheCimattietal.(2004)andMcCarthyetal.(2004)samples,computetheaverageKmagnitudes,R K KssandIscolors,stellarmasses,andstar-formationrates,andcomparethemwiththepropertiesofthepassivegalaxies.

Cimattietal.(2004)presentfourgalaxiesat z =1.7,with Ks =18.7, R Ks =6.3,stellarmassesrang-ingfrom13 3×1011M⊙,andaspacedensityof～10 4Mpc .Whenadecliningmodelwithτ=0.3Gyrisusedto tthecolors,theassociatedagesandforma-tionredshiftsare～2Gyr,andz～4,respectively.Ifweapplyaτ=0.3GyrmodeltothemassiveBX/MDSEDsat z =2.3,andthenevolvethemforwardtoz=1.70,theywouldhave Ks =19.4, R K M =2.2×1011Ms =6.2,⊙,agesrangingfrom1.9-2.3Gyr,andSFRsof≤1M⊙yr 1,makingthemfainterthanR=25.0.We ndasimilarlevelofagreementbetweentheBX/MDmodelsevolvedtoz=1.5andtheparame-tersofthe5objectsinMcCarthyetal.(2004)(theirta-ble1)forwhichaτ=0.5Gyrmodelwasusedto tthespectraandcolors.Thus,extrapolatingstarformationhistoriesforthemassiveBX/MDsforwardintimewouldmakethemplausiblyconsistentwiththepassivegalax-iesinthenear-IRselectedsamplesobservedatsomewhatlowerredshift.Equivalently,iftheBX/MDexamplesareinterpretedasmassiveoldburstswithasmallamountofrecentstarformationsuperposed,byz～1.7theywouldhaveessentiallythesamecharacteristicsasthepassivegalaxiesprovidedthatthecurrentepisodeofstarforma-

本文来源：https://www.bwwdw.com/article/o6zm.html

相关文章：

管家婆-辉煌版完整培训教程06-28

温泉旅游综合体独特的规划设计——天成国际景观08-28

党员个人《大国粮仓》观后感优秀例文07-30

《高等学校研究生英语综合教程-上》原文及翻译09-29

重庆中考英语词汇05-20

上一篇：初二人教版物理欧姆定律单元测试下一篇：2014年考研数学一真题与解析