DRAFTVERSIONMARCH24,2017 PreprinttypesetusingLATEXstyleAASTeX6v.1.0 RELATIONSBETWEENTHESIZESOFGALAXIESANDTHEIRDARKMATTERHALOSATREDSHIFTS0<z <3 KUANG-HANHUANG1,S.MICHAELFALL2,HENRYC.FERGUSON2,ARJENVANDERWEL3,NORMANGROGIN2,ANTON KOEKEMOER2,SEONG-KOOKLEE4,PABLOG.PE´REZ-GONZA´LEZ5,ANDSTIJNWUYTS6 1UniversityofCaliforniaDavis,1ShieldsAvenue,Davis,CA95616,USA;[email protected] 2SpaceTelescopeScienceInstitute,3700SanMartinDrive,Baltimore,MD21218,USA 7 3MaxPlanckInstituteforAstronomy,Koenigstuhl17,D-69117Heidelberg,Germany 1 4CenterfortheExplorationoftheOriginoftheUniverse,DepartmentofPhysicsandAstronomy,SeoulNationalUniversity,Seoul,Korea 0 2 5DepartamentodeAstrof´ısica,FacultaddeCC.F´ısica,UniversidadComplutensedeMadrid,E-28040,Madrid,Spain r 6DepartmentofPhysics,UniversityofBath,ClavertonDown,Bath,BA27AY,UK a M ABSTRACT WederiverelationsbetweentheeffectiveradiiR ofgalaxiesandthevirialradiiR oftheirdarkmatter 2 eff 200c 2 halos over the redshift range 0 < z < 3. For galaxies, we use the measured sizes from deep images taken withHubbleSpaceTelescopefortheCosmicAssemblyNear-infraredDeepExtragalacticLegacySurvey;for ] halos,weusetheinferredsizesfromabundancematchingtocosmologicaldarkmattersimulationsviaastellar A mass–halo mass (SMHM) relation. For this purpose, we derive a new SMHM relation based on the same G selectioncriteriaandotherassumptionsasforoursampleofgalaxieswithsizemeasurements. Asacheckon . h therobustnessofourresults,wealsoderiveR –R relationsforthreeindependentSMHMrelationsfrom eff 200c p theliterature. WefindthatgalaxyR isproportionalonaveragetohaloR ,confirmingandextendingto eff 200c - o highredshiftsthez = 0resultsofKravtsov. Late-typegalaxies(withlowSe´rsicindexandhighspecificstar r formationrate[sSFR])followalinearR –R relation,witheffectiveradiiat0.5 < z < 3closetothose t eff 200c s predictedbysimplemodelsofdiskformation;atz < 0.5,thesizesoflate-typegalaxiesappeartobeslightly a below this prediction. Early-type galaxies (with high Se´rsic index and low sSFR) follow a roughly parallel [ Reff–R200c relation,∼0.2–0.3dexbelowtheoneforlate-typegalaxies. Ourobservationalresults,reinforced 3 byrecenthydrodynamicalsimulations,indicatethatgalaxiesgrowquasi-homologouslywiththeirdarkmatter v 1 halos. 0 Keywords:galaxies: evolution—galaxies: high-redshift—galaxies: structure—methods: dataanalysis 0 4 0 1. INTRODUCTION galaxiesare,onaverage,proportionaltothesizesoftheirsur- . rounding dark matter halos. For galactic disks, this propor- 1 The size of a galaxy, as measured by its half-mass radius 0 R, for example, is among the most basic of its properties. tionality in sizes follows directly from the assumed propor- 7 tionality of the specific angular momentum of baryons and Together with the mass M, the size R determines the bind- 1 : ing energy, −E ≈ GM2/(4R), and hence the energy radi- darkmatterresultingfromtidaltorquesintheearlystagesof v galaxy formation (Fall & Efstathiou 1980; Mo et al. 1998). ated away during the formation of the galaxy. For galactic i X disks, with stars and gas on nearly circular orbits with ro- Thisassumptionunderliespracticallyallofthesemianalyti- cal models of galaxy formation in current use (e.g., Cole et r tation velocity V , the size R is determined by the angu- rot a al.2000;Crotonetal.2016). Recenthydrodynamicalsimu- lar momentum J ≈ MRV , which in turn determines the rot energy E = −1/2MVr2ot ≈ −G2M5/(8J2). The basic de- lationsofgalaxyformationconfirmtheapproximatepropor- tionalitybetweenthespecificangularmomentumofgalaxies scriptionofgalaxiesingeneralconsistsofM,R,andV ,or rot and their dark matter halos (Genel et al. 2015; Pedrosa & equivalently M, E, and J, while for disk-dominated galax- Tissera2015;Tekluetal.2015;Zavalaetal.2016). ies,anytwoofthesequantitiessuffice. There have been numerous searches for the expected de- As a result of the hierarchical growth of galaxies, we ex- creaseingalacticsizeswithredshiftbasedonmeasurements pecttheirmassesandradiitoincreasewithcosmictimeand ofdeepimagestakenwiththeHubbleSpaceTelescope(HST) thus to decrease with redshift. In the simplest models of over the past dozen years (e.g., Ferguson et al. 2004; Hathi galaxy formation, the sizes of the baryonic components of etal.2008;Moslehetal.2012). Thesesearchesallfindthat galaxies were smaller in the past, by roughly the predicted amount,althoughtherearesignificantdifferencesinthepre- E-mail:[email protected] 2 cisedeclineofgalacticsizeswithredshiftamongthesestud- logicalparameters: h=0.7,Ω =0.27,andΩ =0.73. m Λ ies (compare, e.g. Shibuya et al. 2015 and Curtis-Lake et al.2016). Partofthediscrepancyamongtheseresultsstems 2. OBSERVATIONS fromthefactthattheapparentevolutioninsizesdependson Forthisstudy,weneedagalaxysamplewithhomogeneous howgalaxiesatdifferentredshiftsarecompared,whetherat data quality that enables accurate size measurements. HST fixedstellarmassorluminosityoratvariablestellarmassor images are required because galaxies at z > 1 are gener- luminosity. ally smaller than 1(cid:48)(cid:48). We also need a galaxy sample with Kravtsov(2013)usedstellarmass–halomass(SMHM)re- good constraints on redshifts, stellar masses, and star for- lations derived via the technique of abundance matching to mationrates,sothatwecanconnectgalaxiestodarkmatter comparetheobservedsizesofpresent-daygalaxieswiththe halos and distinguish star forming galaxies from quiescent sizesoftheirmatcheddarkmatterhalosincosmologicalN- galaxies. The Cosmic Assembly Near-infrared Deep Extra- bodysimulations.Hefoundthatthesizesofgalaxiesatz =0 galacticLegacySurvey(CANDELS)isthebestdatasetcur- are proportional on average to the sizes of their halos. Fur- rentlyavailableforthisstudy:allfiveCANDELSfields,cov- thermore,thecoefficientofproportionalityisconsistentwith ering ≈ 800 arcmin2 in total, have HST images at optical a simple model in which galactic disks grow with approxi- andnear-IRwavelengthswithuniformquality(Groginetal. mately the same specific angular momentum as their halos 2011; Koekemoer et al. 2011). The high angular resolution until z ∼ 2 and then stop growing after that. The question ofHST((cid:46)0(cid:48).(cid:48)15inthenear-IR)isabletoresolvemostgalax- immediately arises whether the same or a different relation iesatz ≤ 3. Inaddition, ancillaryspectroscopicandimag- holds between the sizes of galaxies and their halos at high ingdatacombinewithHST datatoprovidetightconstraints redshifts. The purpose of this paper is to answer this ques- ongalaxyredshifts, stellarmasses, andstarformationrates. tion. CANDELShasthreetiersofdepth. TheWideregioncovers Theadvantageofcomparingthesizesofgalaxiesatmulti- ∼675arcmin2toa5σlimitingmagnitudeH ∼27.3mag 160 pleredshiftswiththesizesoftheirmatchedhalosatthesame ina0(cid:48).(cid:48)17aperture. TheDeepregioncovers∼ 125arcmin2 redshifts,aswedohere,isthattheresultsarethenexpressed toH ∼28.1mag. ThesurveyalsoencompassestheHub- 160 directlyinsimple,physicallymeaningfulterms. Thisframe- ble Ultra-Deep Field (HUDF)—the HUDF09 (Bouwens et workalsohelpstoclarifytheresultsofprevioussearchesfor al. 2010) and HUDF12 (Ellis et al. 2013; Koekemoer et al. theevolutionofgalacticsizes. 2013;seealsoIllingworthetal.2013)—covers∼ 5arcmin2 Therearealreadyacoupleofindicationsthatthesizesof toH ∼29.7mag. 160 galaxies and their halos evolve in lockstep. First, semiem- We take the photometry, spectroscopic and photometric pirical models of galaxy formation that make this assump- redshifts, and stellar-mass estimates from the CANDELS- tionagreebetterwithdeepHST imagesthanthesamemod- team catalogs (Guo et al. 2013; Galametz et al. 2013; San- elswithdifferentassumptionsabouttheevolutionofgalactic tinietal.2015;Nayyerietal.2016;G.Barroetal. 2017,in sizes (Taghizadeh-Popp et al. 2015). Second, recent mea- preparation; M. Stefanon et al. 2017, in preparation). The surementsofthesizesandrotationvelocitiesofgalacticdisks sizeestimatesaretakenfromvanderWeletal.(2012). at1 < z < 3and0.2 < z < 1.4indicatethattheyhaveap- WeselectgalaxiesintheCANDELSsurveyat0 < z < 3 proximatelythesamespecificangularmomentaastheirdark for this study. We cap our galaxy redshifts at z = 3 be- matterhalos(Burkertetal.2016;Continietal.2016). While cause this is the highest redshift that HST still samples red- these results are suggestive, it is still important to make a ward of rest-frame 4000A˚, and because selection biases in- direct, independent comparison of the sizes of high-redshift duced by cosmological surface brightness dimming are ex- galaxieswiththesizesoftheirmatchedhalos,theinvestiga- pected to be relatively mild for z ≤ 3 (Taghizadeh-Popp et tionwedescribehere. al. 2015). Sources are detected using SExtractor (Bertin & The plan for the remainder of this paper is the following. Arnouts1996)inH . Roughly10%ofthesesourceshave 160 In Section 2, we describe our sample of galaxies and mea- high-quality spectroscopic redshifts, which are used in cali- surements of their sizes and other properties. In Section 3, bratingthephotometricredshiftsfortheremainingsources. we discuss the abundance-matching method and its imple- Galaxy sizes are measured in H and J by fitting a 160 125 mentationwithfourdifferentSMHMrelations. InSection4, single Se´rsic profile to each galaxy using GALFIT (Peng et wepresenttheresultsofourcomparisonofgalaxyandhalo al. 2010). We define galaxy sizes as effective radii (R ) eff sizes,andinSection5,wediscusstheuncertaintiesinthese along the major axis, the radii within which Se´rsic profiles results. WediscusssomeimplicationsofourresultsinSec- containhalfofthetotalintegratedlight. Wediscussthede- tion6.Weshowtheconnectionbetweenthegalaxysize–halo projectionfrom2Dto3Dlaterwhencomparingwiththeoret- size relation and the more familiar galaxy size–stellar mass ical expectations. Our overall sample is dominated by late- relationinanappendix. Allmagnitudesquotedinthispaper type galaxies at all redshifts, whose disk components have areintheABsystem, andweassumethefollowingcosmo- thesame2Dand3Dhalf-lightradii. Usingsimulationswithartificialgalaxiesandcomparisons 3 Table1.GalaxySampleSizes Redshift Wide Deep HUDF Total z M a med ∗,low (M ) (cid:12) 0.0<z<0.5 4388 923 50 5361 0.34 1.0×107 0.5<z<1.0 9706 2435 116 12257 0.73 5.0×107 1.0<z<1.5 6666 1395 113 8174 1.23 8.2×107 1.5<z<2.0 5152 1224 90 6466 1.70 1.7×108 2.0<z<2.5 2580 727 47 3354 2.23 2.1×108 2.5<z<3.0 1483 497 54 2034 2.69 3.8×108 AllRedshifts 29975 7201 470 37646 ··· ··· a TypicalstellarmassofthegalaxiesfromHUDFwith26.6mag<H < 160 26.8magandnearthemedianofeachredshiftbin.Inthelowestredshift bin,weimposeahardcutinstellarmassat107M . (cid:12) of measurements in different imaging depths, van der Wel upmanymorelowsurfacebrightnessobjects,wewouldhave etal.(2012)concludedthatbrighterthanH = 24.5mag expected to see them show up in the tail of the distribution. 160 in the Wide region, the systematic (random) errors of R Instead,weseemorelarge-radiusobjectsintheWidesample, eff measurementsarebelow∼20%(30%). Meanwhile,thesys- mostofwhichareprunedawayasbadfits,butwithouthaving tematic(random)errorsofSe´rsicindexnmeasurementsare much impact on the median R . A Kolmogorov-Smirnov eff below∼50%(60%). Thequotederrorshereareforgalaxies testyieldspvaluesconsistentwiththesamplesbeingdrawn with n > 3, which tend to have larger errors than galax- from the same underlying distribution. The bottom panels ies with n < 3. Therefore, we select all galaxies brighter of Figure 1 show the same comparison for the Deep region than H = 24.5 mag in the Wide region, H = 25.2 inthemagnituderange24.2mag< H < 25.2mag. We 160 160 160 magintheDeepregion,andH =26.7magintheHUDF madeasimilarcomparisonforthestellarmassdistributions, 160 (SExtractor-measured magnitudes). These magnitude limits alsofindingnostatisticallysignificantdifferencebetweenthe correspondtosimilarsignal-to-noiselimits. HUDFandtheDeepandWidesamples. Inadditiontomagnitudecuts,weprunethesampleasfol- We have also estimated the completeness of our sample lows. We reject all sources that have problematic photom- fromthedetectionefficienciesfortheCANDELSsurveyde- etry (generally those at the borders of the image or falling rived by Guo et al. (2013). They inserted artificial galaxies on stellar diffraction spikes). We eliminate sources that are into images from the Wide, Deep, and HUDF regions and identified as active galactic nuclei (AGNs) via X-ray or IR analyzed them with SExtractor in the same way as the real spectral energy distributions (SEDs). We discard as point surveytodeterminethedetectionefficiencyasafunctionof sources all objects that have half-light radii (measured by apparent magnitude H , effective radius R , and Se´rsic 160 eff SExtractor) smaller than 2.6 pixels. We enforce the follow- index n (see their Fig. 5). From these results, we estimate ingcriteriatoeliminategalaxieswithpoorGALFITfits: (1) thatoursampleasawholeismorethan85%complete. This theGALFITmeasurementisflaggedaspoorinthecatalogs highlevelofcompletenesshelpstoensurethatselectionbi- from van der Wel et al. (2012); (2) the error in the mea- aseshaverelativelylittleimpactonourgalaxysize–halosize sured R exceeds 0.3R ; (3) the measured n lies outside relations(estimatedinSection5). eff eff the range 0.1 < n < 8, which usually signals problematic Studyinggalaxysizeevolutiondemandsthatwecompare fits. TheGALFIT,AGN,andpoint-sourcecriteriacombined R valuesatasimilarrest-framewavelengthacrossredshift eff rejectroughlyone-fourthofthesourcesthatsatisfythemag- bins,sothatwecaneliminatethecontributionsfromdustor nitude cuts. The numbers of sources that pass all the cuts stellar age gradient to the observed size evolution. We fol- abovearelistedinTable1. lowtheprocedureinvanderWeletal.(2014)tocorrectfor TheexistenceoftheverydeepHUDFdataallowsustotest galaxy color gradients and place galaxy sizes on the same whetherselectioneffects,measurementbiases,orthepruning rest-framewavelength. Todothis,weusegalaxysizesmea- procedurearebiasingoursamplesneartheirfaintlimits. In suredinH forgalaxiesatz > 1.5andusethesizesmea- 160 thetoppanelsofFigure1,wecomparethesizedistributions suredinJ atz < 1.5. Colorgradientsthatleadtodiffer- 125 in the Wide region and the HUDF for the magnitude range ent galaxy sizes at different wavelengths are accounted for 23.5 mag < H < 24.5 mag before and after pruning, by a correction factor that is a function of galaxy redshift, 160 findingnosignificantdifference. IftheHUDFwerepicking stellar mass, and galaxy type (late-type or early-type). As 4 Raw sample HUDF Pruned sample 2.0 2.0 WIDE s s nt nt u1.5 = 0.29 u1.5 = 0.27 o o c = 0.28 c = 0.28 d d e = 0.83 e = 0.35 z z ali1.0 ali1.0 m m or or n n 0.5 0.5 0.0 0.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 1.0 0.5 0.0 0.5 1.0 log(R /arcsec) log(R /arcsec) Raw sample HUDF Pruned sample 2.0 2.0 DEEP s s nt nt u1.5 = 0.25 u1.5 = 0.23 o o c = 0.26 c = 0.24 d d e = 0.59 e = 0.53 z z ali1.0 ali1.0 m m or or n n 0.5 0.5 0.0 0.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 1.0 0.5 0.0 0.5 1.0 log(R /arcsec) log(R /arcsec) Figure1. HistogramsofeffectiveradiusReff forgalaxiesinnarrowmagnituderangesintheWide,Deep,andHUDFregionsofoursample. ThetoppanelscomparethedistributionsofR intheWideandHUDFregionsinthemagnituderange23.5mag<H <24.5mag,while eff 160 thebottompanelscomparethedistributionsofR intheDeepandHUDFregionsinthemagnituderange24.2mag<H <25.2mag.For eff 160 reference,theselectionlimitsofoursampleintheseregionsareH =24.5(Wide),25.2(Deep),and26.7mag(HUDF).Theleftandright 160 panelscomparethedistributionsbeforeandafterthesamplepruningdescribedinSection2.Thelegendsinthepanelslistthemedianvaluesof R inthefourhistograms,andtheKolmogorov-Smirnovprobabilitiesthatthehistogramsaredrawnfromthesameunderlyingdistribution. eff Theconsistencyofthehistogramsinregionswithdifferentdepths,beforeandafterpruning,indicatesthatthedistributionofgalacticsizesin oursampleisunbiasedevenneartheselectionlimits. theresultofthiscolorgradientcorrection,themeasurements than the magnitude limits mentioned above, where we are are converted into the R near rest-frame 5000A˚. The size confidentthatourmeasurementsarerobustandunaffectedby eff correction is typically only a few percent, but it does reach size-dependentbiases.Foreachredshiftinterval,weestimate ∼60% in some cases. For more details about the color gra- thetypicalstellarmassofthefaintestgalaxiesM bytak- ∗,low dient correction, we refer the readers to van der Wel et al. ing the median SED-fitted stellar mass estimate of galaxies (2014),Section2.2,andtheirequations(1)and(2). within0.1magoftheHUDFmagnitudelimit. Thevaluesof Stellar masses and star formation rates are estimated by M arelistedinTable1andshownasthicktickmarksat ∗,low comparing our photometry with model SEDs, adopting a thebottomsofFigures5–9. SED-basedstarformationrates Chabrier (2003) initial mass function (IMF). Here the stel- can be uncertain by ∼ 0.4 dex (Salmon et al. 2015); there- lar masses of galaxies include all luminous stars and dark fore,theuncertaintiesinthespecificstarformationrates(sS- remnants at the time of observation (but not stellar ejecta). FRs) are roughly (cid:46) 0.6 dex for our galaxy sample. In this This method of estimating stellar masses has been exten- paper,weselectsubsamplesintheupperandlower20%tails sively tested in Mobasher et al. (2015), and they found that of the sSFR distribution. Because we are making a differ- typicalstellarmassuncertaintiesare∼0.25dexforthemag- entialcomparisonbetweentherelativelylargepopulationsin nitude limits adopted here. The primary sources of system- thesetails,ourresultsarenotsensitivetothesSFRuncertain- atic uncertainties are IMF and stellar evolution models; for ties. galaxies with strong nebular emission lines, systematic un- certaintiesforstellarmasscanbeupto∼0.4dex. 3. ABUNDANCEMATCHING WerestrictthisstudytogalaxieswithstellarmassesM∗ > In this study, we employ the technique of abundance 107M(cid:12). Above this limit, we include all galaxies brighter matchingtoestimatethemassandhencethesizeofthedark 5 matter halo associated with each galaxy in our sample. In results for the four bins of width ∆z = 0.25 covering the essence, this technique compares the measured sizes of ob- range 0.5 < z < 1.5 into two bins of width ∆z = 0.5. served galaxies with the inferred sizes of matched halos in In this step, we weight the observed comoving densities of cosmologicaldarkmattersimulations. Thebasicassumption galaxiesbythecomovingvolumeineach∆z =0.25binand isthattherankorderingofgalaxy(stellar)massesM reflects thenfitadoubleSchechterfunctiontothecombinedcomov- ∗ on average the rank ordering of halo (virial) masses M , ingdensitiesineach∆z = 0.5bin. Forourlowestredshift 200c i.e., that the cumulative number densities of galaxy masses bin, 0 < z < 0.5, weadopttheTomczaketal. stellarmass andhalomassesareequal:n (>M )=n (>M ).This function in their lowest redshift bin, 0.2 < z < 0.5, be- g ∗ h 200c ansatz leads directly to a correspondence between M and causeitagreeswellwiththeoneat< z >= 0.1derivedby ∗ M knownasthestellarmass–halomassrelation. While Moustakas et al. (2013). Finally, we have derived the halo 200c the assumption that galaxy masses and halo masses follow massfunctionn (>M )fromtheMillennium-IIsimula- h 200c thesamerankorderingisareasonableapproximationforsta- tion (Boylan-Kolchin et al. 2009) at the snapshot closest to tistical studies based on large samples such as ours, it can- themiddleofeachredshiftbinandthenmatchedthistothe notbeexactlytrueforindividualgalaxies,whichexperience stellarmassfunctionasdescribedabovetoobtaintheSMHM stochastic events such as mergers and starbursts throughout relation. theirhistories. As a check on this procedure, we have independently de- Given an SMHM relation, we compute the halo mass rived our own stellar mass function from scratch by the M ofeachgalaxyinoursamplefromitsstellarmassM . 1/V methodforthegalaxiesinallfiveCANDELSfields 200c ∗ max WethencomputethevirialhaloradiusR usingthestan- in the six ∆z = 0.5 bins (albeit with approximate K- 200c dardformula corrections in our estimates of V ). The resulting stellar max mass function is nearly identical to the rebinned one from (cid:20) 3M (cid:21)1/3 R = 200c , (1) Tomczaketal.(2014). Thisaddstoourconfidenceintheva- 200c 4π·200ρ (z) crit lidity of SMHM relation 1, which we regard as the primary SMHMrelationinthisstudy. where ρ (z) is the critical density of the universe at red- crit Becauseourgalaxysamplecoversawiderrangeinstellar shiftz. Inordertoassesshowsensitiveourresultsaretothe mass than the Tomczak et al. sample, we linearly extrapo- choiceofSMHMrelation,weperformallofourcalculations late the SMHM relation in log–log space to both lower and withfourdifferentSMHMrelations. AlloftheseSMHMre- highermasses. ThesolidlinesinFigure2showtheSMHM lationsarebasedontheChabrier(2003)stellarIMFandthe relationderiveddirectlyfromtheTomczaketal. data,while samehalomassdefinitionM .TheyareplottedinFigures 200c the dashed lines show the extrapolated parts of the SMHM 2,3,and4anddiscussedbelow. relation. SMHM relation 1. We have derived this new SMHM re- SMHM relation 2. Behroozi et al. (2013) derived this lation specifically for this study so that it is as consistent as SMHM relation from published stellar mass and halo mass possiblewiththeCANDELSdataset,selectioncriteria,and functionsoverawiderangeofredshifts(0<z <8). Thisis SED fitting procedure for our sample of galaxies with size probablythemostprevalentSMHMrelationintheliterature. measurements. In particular, we combine the stellar mass However, sinceitisbasedonstellarmassfunctionsthatare functionn (>M )fromTomczaketal.(2014)withourde- g ∗ quitedifferentfromthosederivedusingCANDELSdata,itis terminationofthehalomassfunctionn (>M )fromthe h 200c notidealforthepresentstudy. Weuseitmainlytogaugethe Millennium-IIsimulation(Boylan-Kolchinetal.2009). sensitivity of our results to different SMHM relations. For Tomczaketal.(2014)derivedthestellarmassfunctionof consistency, we convert their halo mass M , defined us- galaxiesat0.2<z <3inthreeofthefiveCANDELSfields, vir ing a redshift-dependent overdensity factor ∆ (z) (Bryan usingselectioncriteriaandproceduresforestimatingstellar vir & Norman 1998), to our halo mass definition M . The massessimilartothoseforoursample,asdescribedinSec- 200c conversionassumesanNFWhalomassprofileandthehalo tion2. Wehavecomparedourstellarmasseswiththosede- rivedbyTomczaketal.(2014)1andfindnosystematicoffset mass–concentrationmodelcalibratedinDiemer&Kravtsov (2015). The corrections are very small in general (< 0.1 andonlyasmallscatter(∼ 0.1dex). Tomczaketal. fitteda dex). doubleSchechterfunctiontotheobservedstellarmassfunc- SMHM relation 3. This is the same SMHM relation tion in differential form dn (> M )/dM in each of eight g ∗ ∗ adopted by Kravtsov (2013). He derived his own SMHM redshift bins. We adopt the Tomczak et al. results directly relation out of concerns that previous relations used stellar for the three bins of width ∆z = 0.5 covering the range massfunctionsthatarebiasedatboththehigh-massandlow- 1.5 < z < 3.0. However, for simplicity, we combine their mass ends. By using the same SMHM relation as Kravtsov (2013), we can directly compare our galaxy size–halo size 1ThesestellarmassesarepublishedbytheZFOURGEteam(Straatman relationwithhisatz =0. etal.2016)andcanbedownloadedfromhttp://zfourge.tamu.edu. SMHMrelation4. ThereareseveralSMHMrelationssep- 6 aratedbygalaxytypeatz < 0.5intheliterature, whichwe 0.5 plotinFigure3. Theserelationsusedifferentapproachesto 1.0 deriving the ratio between stellar masses and halo masses, 1.5 ranging from abundance matching (Rodr´ıguez-Puebla et al. ) 2015) to weak lensing (Hudson et al. 2015; Mandelbaum 2.0 / et al. 2016) to a mixture of the two methods (Dutton et * 2.5 al. 2010). We adopt the SMHM relation from Rodr´ıguez- ( 3.0 Pueblaetal.(2015)becauseithasthelargestdynamicrange Dutton+10 (late-type) Hudson+15 (blue, z=0.5) in halo mass and is in the middle of the range spanned by 3.5 Dutton+10 (early-type) Hudson+15 (red, z=0.5) Rodriguez+15 (blue) Mandelbaum+16 (blue) the other type-dependent relations from the literature. We 4.0 Rodriguez+15 (red) Mandelbaum+16 (red) use the Rodr´ıguez-Puebla et al. SMHM relations for blue 10 11 12 13 14 15 ( / ) and red central galaxies at z = 0 for galaxies in our sam- ple with Se´rsic index n below and above 2.5, respectively. Figure3.RatioofgalaxystellarmassM∗tohalovirialmassM200c SinceRodr´ıguez-Pueblaetal. definedtheirhalomassusing plottedagainstM200c forfourlow-redshiftSMHMrelationsfrom theliteraturethatdependongalaxycolorortype. Thesewerede- ∆ (z), we have applied the same conversion to M as vir 200c rivedbyabundancematching(Rodr´ıguez-Pueblaetal.2015),weak wedidforSMHMrelation2. lensing(Hudsonetal.2015;Mandelbaumetal.2016),oracombi- We compare the four SMHM relations in Figure 4. Ev- nationofbothtechniques(Duttonetal.2010).ThreeoftheSMHM relationspertaintoz =0andonetoz =0.5(Hudsonetal.2015). idently, there are significant discrepancies among these Notethelargediscrepanciesamongthesecolor-andtype-dependent SMHM relations, especially the first and second, for which SMHMrelations. thedifferencescanbeupto∼0.5dexatz ∼3. OurSMHM relation1, derivedspecificallyfortheCANDELSsampleat 1.0 0 < z < 3, shows stronger redshift evolution than SMHM relation 2 from Behroozi et al. (2013). As already noted, 1.5 thisdifferencecomesmainlyfromthedifferentstellarmass ) 2.0 functions used as input to these SMHM relations. Fortu- 2.5 nately, as we show in Sections 4 and 5, our main scientific / * resultsarerelativelyinsensitivetotheadoptedSMHMrela- 3.0 ( tion,largelyduetotheweakdependenceofhalosizeonhalo 3.5 SMHM 1 (. < < . ) SMHM 3 (= ) mass(R200c ∝M210/03c). 4.0 SSMMHHMM 12 ((.=<. )< . ) SSMMHHMM 44 ((brelude, , ==)) SMHM 2 (= . ) 4.5 10 11 12 13 14 15 SMHM Relation 1 1.0 ( / ) 1.5 Figure4.RatioofgalaxystellarmassM∗tohalovirialmassM200c 2.0 plottedagainstM200cforthefourSMHMrelationsadoptedinthis ) work. SMHMrelation1: derivedasdescribedinSection3forall 2.5 galaxies at 0 < z < 3 and displayed here at 0 < z < 0.5 and /* 2.5<z<3.0,whichbrackettherelationatintermediateredshifts. g( 3.0 SMHMrelation2: derivedbyBehroozietal.(2013)forallgalax- o l 3.5 ies at 0 < z < 8 and displayed here at z = 0.1 and z = 3.0. 0.0 < z < 0.5 1.5 < z < 2.0 SMHMrelation3:derivedbyKravtsov(2013)forallgalaxiesonly 4.0 0.5 < z < 1.0 2.0 < z < 2.5 at z = 0. SMHM relation 4: derived by Rodr´ıguez-Puebla et al. 1.0 < z < 1.5 2.5 < z < 3.0 (2015)separatelyforblueandredgalaxiesonlyatz=0.Notethat 4.5 10 11 12 13 14 15 therearesignificantdifferencesamongtheseSMHMrelations,but log( / ) becausehalosizedependsweaklyonhalomass(R ∝M1/3), 200c 200c ourmainresultsarenotsensitivetothesedifferences. Figure2.RatioofgalaxystellarmassM∗tohalovirialmassM200c plottedagainstM forourprimarySMHMrelationinsixred- 200c shiftbinscoveringtherange0 < z < 3. WederivedthisSMHM ThemainresultsofthispaperaredisplayedinFigures5– relationbyabundancematchingfromanevolvingstellarmassfunc- 9 and described in this section. The uncertainties in these tionappropriatefortheCANDELSsample(Tomczaketal.2014) results,mostlystemmingfromtheSMHMrelationandmor- and the evolving halo mass function in the Millennium-II simula- tion(Boylan-Kolchinetal.2009)asdescribedinSection3. Solid phologicalclassification,arediscussedinSection5. linesarebaseddirectlyonthestellarmassfunctionfromTomczak Ourfirstmainresultisthatgalaxysizesareproportionalto etal.(2014);welinearlyextrapolatetheSMHMrelationinlog–log halosizesoverawiderangeofsizeandmass.Figure5shows spacetocoverthestellarmassrangeofoursample(dashedlines). galaxyR plottedagainsthaloR at0<z <0.5forthe eff 200c fourdifferentSMHMrelations.Ineachpanel,themediansof logR inbinsofwidth∆logR = 0.15dexareplotted eff 200c 4. RESULTS aspentagons,andthe16th–84thpercentilerangesasvertical 7 bars; only the bins with more than five galaxies are shown. This helps explain why our overall relation in Figure 5 is The halo radius limit corresponding to the reference stellar higher than Kravtsov’s at intermediate masses. Our sample mass M from Table 1 is shown as a thick tick mark at is dominated by late-type galaxies (∼90% have n < 2.5), ∗,low thebottomofeachpanel. Thecoefficientofproportionality whileKravtsov’ssampleisdominatedbyearly-typegalaxies αintherelationR =αR isnearlythesameinallfour (∼ 80% by number). He noted that late-type galaxies are eff 200c cases; the median values of α are 0.021, 0.025, 0.023, and systematically larger in R than early-type galaxies at in- 1/2 0.024 for SMHM relations 1–4, respectively. These R – termediatestellarmasses,whichiswhereweseethelargest eff R relationsareapproximatelylinear,butwithsomesub- offset between these sequences in Figure 5. The changing 200c tledifferencesdependingontheadoptedSMHMrelation. morphological mix as a function of mass also helps explain Kravtsov (2013) also found a linear relation, using com- theapparentcurvatureoftheoverallrelationinFigure5,be- pletely independent samples of galaxies at z = 0 and de- cause early-type galaxies dominate the high- and low-mass projected 3D half-mass radii R rather than the projected endsoftherelation. 1/2 2Dhalf-lightradiiR . ThesolidlineinFigure5showshis Our third main result is that the R –R relation for eff eff 200c derivedrelationR =α(cid:48)R withα(cid:48) =0.015,assuming late-typegalaxiesisclosetothepredictionsofthesimplean- 1/2 200c R = R for pure-disk galaxies. The bulk of our sam- alyticmodelofdiskformation.Thescaleradiusandeffective eff 1/2 plebynumberliesabovethisrelationby∼0.2dex,agreeing radiusofanexponentialdiskembeddedinadarkmatterhalo better at the high- and low-mass ends. There are a number withavirial(outer)radiusR andaspinparameterλare 200c of possible explanations for this offset, one of them being givenby thedifferencebetween2Dhalf-light(effective)and2Dhalf- λ R = √ R (2) d 200c massradii. Szomoruetal.(2013)notedthatforthegalaxies 2 moremassivethan5×1010M at0<z <2.5,rest-frameg- (cid:12) and band2Dhalf-lightradiiareonaverage∼25%largerthan2D R =1.68R , (3) half-massradii(presumablyduetotheinfluenceofbulges), eff d whichcouldaccountfor∼0.1dexoftheoffset. Wewillad- when the disk and halo have the same specific angular mo- dress other explanations below in connection with morpho- mentum (J/M). Equation (2) is exact for isothermal halos logicaltypes,deprojectioneffects,andtheredshiftevolution. (Fall&Efstathiou1980;seetheirFigure3andequation42; Our second main result is that the R –R relations Fall 1983, see his equation 4) and is approximate for NFW eff 200c are offset for late-type and early-type galaxies. To sepa- haloswithtypicalconcentrations(Moetal.1998;Burkertet rate morphological types, we split our sample in two dif- al. 2016). This prediction is shown as the dashed lines in ferent ways: (1) high-n (early-type) and low-n (late-type) Figures6to9forλ = 0.035, thepeakoftheuniversalspin subsamples, and (2) low-sSFR (early-type) and high-sSFR parameterdistribution(Bullocketal.2001;Bettetal.2007). (late-type)subsamples.Weonlyincludethehighestandlow- We find that late-type galaxies at 0 < z < 0.5 lie ∼ 0.2 est 20% of the sample in either n or sSFR in the hope that dex below the J/M equality line; in other words, our late- this procedure will isolate disk-dominated from spheroid- type galaxies have slightly less specific angular momentum dominated galaxies. The resulting R –R relations for thantheirdarkmatterhalos. Thisoffsetisconsistentwithdi- eff 200c late-andearly-typegalaxiesusingallfourSMHMrelations rectmeasurementsofspecificangularmomentumatz = 0, areshowninFigures6and7. whichindicateJ/M retentionfactorsη ∼ 80%±20%for j We see in both Figures 6 and 7 that galaxies of different galacticdisks(Fall&Romanowsky2013). typesfollowsequencesroughlyparalleltotheR ∝ R Ourfourthmainresultisthatthereisremarkablylittleevo- eff 200c linewithanoffsetof∼ 0.2dexat0 < z < 0.5. Thisresult lution in the R –R relation from z = 3 to z = 0. eff 200c isrelativelyrobustagainstSMHMrelationandmorphologi- This is shown in Figures 8 and 9. As in the previous dia- cal classification method: early-type (high-n or low-sSFR) grams, we select the highest and lowest 20% tails of the n galaxies have smaller R than late-type (low-n or high- andsSFRdistributions. WeonlyshowresultsforSMHMre- eff sSFR)galaxiesatthesamehalomasses. Theeffectpersists lation 1, but we have checked that they are similar for the even if we compare 3D half-light radii rather than 2D half- other SMHM relations. Figures 8 and 9 show again that in light radii R , although with a smaller separation between allredshiftbins,late-typegalaxiesfollowanearlylinearre- eff thesequences. Theparallelsequencesofearly-andlate-type lation: R = αR . At 0.5 < z < 3, late-type galaxies eff 200c galaxies in the R –R diagram are reminiscent of the haveα≈0.034inFigure8(α≈0.029inFigure9)andlie eff 200c parallelsequencesofspheroid-anddisk-dominatedgalaxies close to the J/M equality line (within (cid:46) 0.1–0.2 dex) with intheJ/M vs. M diagram(Fall1983;Romanowsky&Fall no discernible evolution. (There is a slight offset to smaller 2012;Fall&Romanowsky2013). Thelatterisduetoacom- sizes in the late-type sample when selected by sSFR rather bination of different sizes (by a factor of ∼2) and different thanSe´rsicindex.) Thisresultagreeswithrecentdirectmea- rotation velocities (also by a factor of ∼2–3) of spheroid- surements of specific angular momentum at 0.2 < z < 1.4 anddisk-dominatedgalaxiesofthesamestellarmass. (Continietal.2016)andat1 < z < 3(Burkertetal.2016), 8 SMHM Relation 1 SMHM Relation 2 SMHM Relation 3 SMHM Relation 4 . < < . . < < . . < < . . < < . c p k / Kravtsov13 / kpc / kpc / kpc / kpc Figure5.GalaxyeffectiveradiusReffplottedagainsthalovirialradiusR200cinthelowestredshiftinterval(0<z<0.5)forthefullsampleofgalaxies.The fourpanelsshowresultsforSMHMrelations1,2,3,and4asindicated.Thefaintgraydotsrepresentindividualgalaxies,whilethefilledpentagonsandvertical barsindicatethemedianvaluesand16th–84thpercentilerangesofReff inbinsofwidth0.15inlogR200c.ThediagonallinesshowtheR1/2–R200crelation atz =0fromKravtsov(2013)assumingReff =R1/2. Thethicktickmarkatthebottomofeachpanelindicatesthehalosizecorrespondingtothereference stellarmassM∗,low listedinTable1. NotethattheReff–R200c relationsaresimilarforthefourdifferentSMHMrelationsandareroughlyconsistentwith Kravtsov’sresults. TheReff–R200crelationsarelinearinafirstapproximationbutexhibitsomecurvatureathighandlowmassesasaresultofthechanging mixofgalaxymorphologies.ComparewithFigures6and7. SMHM relation 1 SMHM relation 2 SMHM relation 3 SMHM relation 4 . < < . . < < . . < < . . < < . c p k / J/M equality Kravtsov13 Lowest 20% in Highest 20% in / kpc / kpc / kpc / kpc Figure6.GalaxyeffectiveradiusReffplottedagainsthalovirialradiusR200cinthelowestredshiftinterval(0<z<0.5)forsubsamplesofgalaxieswiththe lowestandhighest20%ofthemeasuredSe´rsicindexnasproxiesforlate-andearly-typegalaxies,respectively.ThefourpanelsshowresultsforSMHMrelations 1,2,3,and4asindicated.Thefaintblueandreddotsrepresentindividuallow-nandhigh-ngalaxies,respectively,whilethefilledbluesquares,openredcircles, andverticalbarsindicatethecorrespondingmedianvaluesand16th–84thpercentilerangesofReff inbinsofwidth0.15inlogR200c.Thediagonalsolidlines showtheR1/2–R200crelationatz =0fromKravtsov(2013)assumingReff =R1/2,whilethediagonaldashedlinesshowthepredictionforgalacticdisks withthesameJ/Mastheirsurroundinghalos.Thethicktickmarkatthebottomofeachpanelindicatesthehalosizecorrespondingtothereferencestellarmass M∗,lowlistedinTable1.NotethattheReff–R200crelationforlow-ngalaxiesissystematicallyabove,androughlyparallelto,therelationforhigh-ngalaxies. TheReff–R200crelationsforbothsubsamplesofgalaxiesaremorelinearthantherelationsforthefullsample.ComparewithFigures5and7. whichshowthatJ/M ingalacticdisksisnearlythesameas could be due to either size-measurement biases (due to dif- intheirdarkmatterhalos. fuse outer halos surrounding central galaxies in groups and Kravtsov(2013)speculatedthatthesizesofgalaxiesgrew clusters) or the breakdown of abundance matching for the inproportiontothesizesoftheirhalosuntilz ∼ 2andthen group-orcluster-masshalos. stopped,whiletheirhaloscontinuedtogrowinmassandsize. 5. UNCERTAINTIES We find instead that the R –R relations at z < 2 are eff 200c very similar to those at z > 2. Our R –R relations How robust are these results? The uncertainties in this eff 200c forthelate-typegalaxiesatz <0.5havesmalleramplitudes study potentially include measurement and statistical errors thanthoseatz > 0.5,indicatingapossibleslowdowninthe internal to the CANDELS data set, as well as external sys- growthofdisks,butthisdeviationismild(∼0.2dex)andnot tematicerrorsfromtheadoptedSMHMrelationsandstellar establishedbeyondalldoubt(seebelow). population models. Here we provide a brief assessment of The R –R relation for early-type galaxies is also theseuncertainties. eff 200c nearly constant. We see in Figures 8 and 9 that the trend As noted in Section 2, errors in the measurements of ef- for early-type galaxies at all redshifts roughly parallels that fective radii Reff (from fits to Se´rsic profiles) are relatively for late-type galaxies, but shifted down by ∼ 0.2 dex at small: < 20%(systematic)to30%(random). Evenifthese 0 < z < 0.5 and by ∼ 0.2–0.3 dex at 0.5 < z < 3. There errorswereattheupperendofthisrangeforallgalaxiesand is a slight hint of a “turnover” at the most massive end at varied systematically with galactic masses and sizes, they 0 < z < 0.5 (see Figures 8 and 9). This turnover, if real, wouldhaveanegligibleinfluenceonthecoefficientandex- ponentofthegalaxysize–halosizerelation: R = αRβ eff 200c 9 SMHM relation 1 SMHM relation 2 SMHM relation 3 SMHM relation 4 . < < . . < < . . < < . . < < . c p k / J/M equality Kravtsov13 Highest 20% in sSFR Lowest 20% in sSFR / kpc / kpc / kpc / kpc Figure7.GalaxyeffectiveradiusReffplottedagainsthalovirialradiusR200cinthelowestredshiftinterval(0<z<0.5)forsubsamplesofgalaxieswiththe highestandlowest20%ofthemeasuredsSFRasproxiesforlate-andearly-typegalaxies,respectively.ThefourpanelsshowresultsforSMHMrelations1,2,3, and4asindicated.Thefaintblueandreddotsrepresentindividualhigh-sSFRandlow-sSFRgalaxies,respectively,whilethefilledbluesquares,openredcircles, andverticalbarsindicatethecorrespondingmedianvaluesand16th–84thpercentilerangesofReff inbinsofwidth0.15inlogR200c.Thediagonalsolidlines showtheR1/2–R200crelationatz =0fromKravtsov(2013)assumingReff =R1/2,whilethediagonaldashedlinesshowthepredictionforgalacticdisks withthesameJ/Mastheirsurroundinghalos.Thethicktickmarkatthebottomofeachpanelindicatesthehalosizecorrespondingtothereferencestellarmass M∗,lowlistedinTable1. NotethattheReff–R200crelationforhigh-sSFRgalaxiesissystematicallyabove,androughlyparallelto,therelationforlow-sSFR galaxies.TheReff–R200crelationsforbothsubsamplesofgalaxiesaremorelinearthantherelationsforthefullsample.ComparewithFigures5and6. with|∆α/α| (cid:46) 0.02and|∆β| (cid:46) 0.08(assuminga∼ 20% size–halo size relations from the detection efficiencies for or smaller systematic deviation in R for a factor of 10 or the CANDELS survey derived by Guo et al. (2013) as fol- eff more variation in R ). Because the sample size in this lows. TheydividetheR –H planeintoregionsthatare 200c eff 160 studyissolarge(N ∼ 38000),theeffectsofrandomerrors 0–50%,50–90%,and90–100%complete. Mostofoursam- in the size measurements on the mean R –R relations ple(88%)liesintheregionof90–100%completeness,while eff 200c areevensmaller.Inasituationlikethis,withnegligibleinter- theremainder(12%)liesintheregionof50–90%complete- nalerrors,formaltestsofgoodnessoffitarenotinformative, ness. To place an upper limit on the impact of selection bi- andwedonotattemptthem. ases,weadoptthelowerlimitsof90%and50%onthecom- Thedominantuncertaintiesinourgalaxysize–halosizere- pletenessinthesetworegionsoftheR –H plane,assign eff 160 lationsaremostlikelycausedbypossiblesystematicerrorsin weights 2.0 (i.e., 1/0.5) and 1.1 (i.e., 1/0.9) to the galax- our adopted SMHM relations. We can judge the magnitude ies in our sample in these regions, and then recompute the of these errors by comparing the R –R relations plot- R –R relations. ForR (cid:38)100kpc,wefindnegligi- eff 200c eff 200c 200c tedinFigures5to7forthefourdifferentSMHMrelations. blecorrectionstothemedianR –R relations,whilefor eff 200c This comparison indicates that the SMHM relation may be R (cid:46) 100kpc, wefindcorrectionsbelow0.1dexforall 200c responsible for systematic errors at the level of ∼ 0.1–0.2 galaxy types and redshifts 0 < z < 3. We conclude from dex,perhapsalittlelessforthecombinedsampleofgalaxies, this exercise that selection biases are likely to be subdomi- perhapsalittlemoreforthesubsamplessplitbymorpholog- nantsourcesofuncertaintyinourR –R relations. eff 200c icaltype. Quantitativemeasuresofthedeviationsamongthe Basedonthisassessmentofuncertainties, mostofthere- R –R relations at 0 < z < 0.5 confirm these impres- sultsofthispaperappeartoberobust. Inparticular,thereis eff 200c sions. a strong, approximately linear correlation between the sizes Thecontributionstotheerrorbudgetfromtheadoptedstel- of galaxies and their dark matter halos over the full range lar population models, which determine the stellar masses of redshifts examined here, 0 < z < 3. The coefficient of andspecificstarformationrates,aresmallerthanthosefrom proportionalityislargerforlate-typegalaxiesthanforearly- the adopted SMHM relations. Systematic errors in stellar type galaxies, which follow roughly parallel sequences, ex- masses could affect the R –R relations at about the ceptpossiblyatthehighestredshifts. Forlate-typegalaxies, eff 200c samelevelassystematicerrorsinR . Theclassificationof theobservedR –R relationisgenerallyconsistentwith eff eff 200c the3Dshapesofgalaxies(i.e.,flatdisksvs.roundspheroids) simple models in which galactic disks grow with the same bySe´rsicindexisanothersourceofuncertainty,becauseitis specificangularmomentumastheirdarkmatterhalos. There basedonlyontheradialdeclineoftheprojected2Dsurface issomeevidenceforaslowdownindiskgrowthatz < 0.5, brightnessprofiles. FittingasingleSe´rsicprofileinsteadofa buttheapparentdeviationfromtheJ/M equalitylineisonly detaileddisk/bulgedecompositionpossiblyaddsfurtherun- ∼0.2dex. certainty. Nevertheless, the R –R relations we obtain eff 200c fromsubsamplessplitbySe´rsicindexagreeatthe(cid:46)0.1dex levelwiththosefromsubsamplessplitbyspecificstarforma- tionrate. We estimate the impact of selection biases on our galaxy 10 SMHM Relation 1 . < < . . < < . . < < . c p k / . < < . . < < . . < < . c p k / J/M equality Kravtsov13 Lowest 20% in Highest 20% in / kpc / kpc / kpc Figure8. GalaxyeffectiveradiusReff plottedagainsthalovirialradiusR200c atdifferentredshiftsforsubsamplesofgalaxieswiththelowestandhighest 20%ofthemeasuredSe´rsicindexnasproxiesforlate-andearly-typegalaxies,respectively. ThesixpanelsshowresultscomputedfromSMHMrelation1in redshiftintervalsof∆z =0.5coveringtherange0<z <3. Thefaintblueandreddotsrepresentindividuallow-nandhigh-ngalaxies,respectively,while thefilledbluesquares,openredcircles,andverticalbarsindicatethecorrespondingmedianvaluesand16th–84thpercentilerangesofReffinbinsofwidth0.15 inlogR200c.ThediagonalsolidlinesshowtheR1/2–R200crelationatz=0fromKravtsov(2013)assumingReff =R1/2,whilethediagonaldashedlines showthepredictionforgalacticdiskswiththesameJ/M astheirsurroundinghalos. Thethicktickmarkatthebottomofeachpanelindicatesthehalosize correspondingtothereferencestellarmassM∗,lowlistedinTable1.NotethattheReff–R200crelationsforbothlow-nandhigh-ngalaxiesarenearlyconstant withredshift,andthattheoneforlow-ngalaxiesisclosetothepredictedrelationforequalityofJ/Mindisksandhalos.ComparewithFigure9. Table2.VerificationofMainResults SMHM1 SMHM2 SMHM3 SMHM4 1.TheR –R relationsareroughlylinearinallredshiftbins. T T T T eff 200c 2.TheR –R relationsareoffsetforearly-andlate-typegalaxies. T T T T eff 200c 3.TheR –R relationforlate-typegalaxiesareclosetotheJ/M equalityline. T T T T eff 200c 4.TheR –R relationshowslittleevolutionbetweenz=0andz=3. T T T T eff 200c WehaveplottedandexaminedtheR –R relationsat scientific results of this study are robust relative to discrep- eff 200c all redshifts (0 < z < 3) for all four SMHM relations to anciesamongtheSMHMrelations(becauseoftheweakde- determinewhetherornottheysupportthefourmainresults pendenceofR onM ). 200c 200c discussedinSection4. Theoutcomeofthistestisrecorded in Table 2 by a T (for true) or F (for false) for each combi- 6. DISCUSSION nation of SMHM relation and result. All of the entries are We have found that the sizes of galaxies are proportional Ts. Table2thereforereinforcesourconclusionthatthemain on average to the sizes of their dark matter halos over a