Galaxies in X-ray Groups. III. Satellite Color and Morphology Transformations
Matthew R. George, Chung-Pei Ma, Kevin Bundy, Alexie Leauthaud, Jeremy Tinker, Risa H. Wechsler, Alexis Finoguenov, Benedetta Vulcani
DDraft version August 28, 2018
Preprint typeset using L A TEX style emulateapj v. 11/10/09
GALAXIES IN X-RAY GROUPS. III. SATELLITE COLOR AND MORPHOLOGY TRANSFORMATIONS
Matthew R. George , Chung-Pei Ma , Kevin Bundy , Alexie Leauthaud , Jeremy Tinker , Risa H.Wechsler , Alexis Finoguenov , Benedetta Vulcani Draft version August 28, 2018
ABSTRACTWhile the star formation rates and morphologies of galaxies have long been known to correlatewith their local environment, the process by which these correlations are generated is not well un-derstood. Galaxy groups are thought to play an important role in shaping the physical propertiesof galaxies before entering massive clusters at low redshift, and transformations of satellite galaxieslikely dominate the buildup of local environmental correlations. To illuminate the physical processesthat shape galaxy evolution in dense environments, we study a sample of 116 X-ray selected galaxygroups at z = 0 . − − M (cid:12) and centroids determined with weak lensing.We analyze morphologies based on HST imaging and colors determined from 31 photometric bandsfor a stellar mass-limited population of 923 satellite galaxies and a comparison sample of 16644 fieldgalaxies. Controlling for variations in stellar mass across environments, we find significant trendsin the colors and morphologies of satellite galaxies with group-centric distance and across cosmictime. Specifically at low stellar mass (log( M (cid:63) /M (cid:12) ) = 9 . − . >
50% among field galaxies to <
20% among satellites near the centersof groups. This decline is accompanied by a rise in quenched galaxies with intermediate bulge+diskmorphologies, and only a weak increase in red bulge-dominated systems. These results show thatboth color and morphology are influenced by a galaxy’s location within a group halo. We suggest thatstrangulation and disk fading alone are insufficient to explain the observed morphological dependenceon environment, and that galaxy mergers or close tidal encounters must play a role in building up thepopulation of quenched galaxies with bulges seen in dense environments at low redshift.
Subject headings: galaxies: bulges – galaxies: clusters: general – galaxies: evolution – galaxies: halos– galaxies: statistics – X-rays: galaxies: clusters INTRODUCTION
Galaxy properties are correlated with their environ-ment. Dense regions host galaxies with greater masses,lower star-formation rates, and morphologies that aremore bulge-dominated than in low density regions. Sincethe early work of Dressler (1980), studies of environmen-tal correlations have expanded to a variety of indicatorsof star formation, morphology, and environment. SeeBlanton & Moustakas (2009) for a recent review of thesecorrelations in the local universe.Numerous processes may be culpable for the depen-dence of galaxy properties on environment (see Boselli &Gavazzi 2006 for a review). Galaxy interactions throughmergers or tides can disrupt stellar kinematics and re-move gas. These interactions can also apply torques that [email protected] Department of Astronomy, University of California, Berke-ley, CA 94720, USA Lawrence Berkeley National Laboratory, 1 Cyclotron Road,Berkeley, CA 94720, USA Kavli Institute for the Physics and Mathematics of the Uni-verse (Kavli IPMU, WPI), Todai Institutes for Advanced Study,University of Tokyo, Kashiwa 277-8583, Japan Center for Cosmology and Particle Physics, Department ofPhysics, New York University, 4 Washington Place, New York,NY 10003, USA Kavli Institute for Particle Astrophysics and Cosmology,SLAC National Accelerator Laboratory, 2575 Sand Hill Rd.,Menlo Park, CA 94025, USA Physics Department, Stanford University, Stanford, CA94305, USA Department of Physics, University of Helsinki, GustafH¨allstr¨omin katu 2a, FI-00014 Helsinki, Finland drive gas inward, perhaps feeding star formation or acentral black hole. Halos can play a role through tidalforces and dynamical friction. Pressure from hot densegas inside halos may strip gas from infalling galaxies, andshock heating in massive halos can prevent accretion ofcold gas that would feed star formation.Despite extensive observational and theoretical work,the dominant physical mechanisms responsible for theenvironmental correlations remain unclear. An impor-tant clue is that the scale on which these correlations ap-pear is similar to the size of the dark matter halos hostinggalaxies (e.g., Kauffmann et al. 2004; Blanton & Berlind2007). Relatedly, satellite galaxies have a primary rolein the buildup of the red sequence in dense environments(e.g., Weinmann et al. 2006; van den Bosch et al. 2008;Wetzel et al. 2012a; Peng et al. 2012).Since many galaxy characteristics are interrelated, itis important to study the correlations for each prop-erty independently while fixing other variables. Whenconstraining environmental effects, one ideally controlsfor differences in redshift, stellar mass, halo mass, andlocation within a group. The colors and morphologiesof galaxies are also correlated, so it is advantageousto split galaxies simultaneously by color and morpho-logical classes to distinguish between processes that af-fect star formation rates and structural properties differ-ently. Several studies have suggested that environmentalprocesses affect star formation more significantly thanmorphology (e.g., Kauffmann et al. 2004; Blanton et al.2005; Christlein & Zabludoff 2005; Weinmann et al. 2009;Kovaˇc et al. 2010), suggesting that the gas that feeds star a r X i v : . [ a s t r o - ph . C O ] M a y George et al.formation is depleted or removed without significantly al-tering galactic structure or stellar kinematics.While many of the observational studies have focusedon large surveys at low redshift or on massive clustersat higher redshifts, less is known about the more com-mon group-scale environments and their evolution overtime. In this paper, we study the colors and morpholo-gies of galaxies in groups spanning the redshift range z = 0 . −
1. We focus in particular on the propertiesof satellite galaxies as a function of stellar mass, group-centric distance, and redshift. This sample of groupshas been identified from the COSMOS field (Scovilleet al. 2007) based on their extended X-ray emission andhave masses determined with weak lensing in the range10 − M (cid:12) (Leauthaud et al. 2010). Galaxy mem-bership has been assigned using precise photometric red-shifts with a Bayesian procedure tested extensively withmock catalogs and a spectroscopic subsample (Georgeet al. 2011). Furthermore, by finding galaxies that maxi-mize the lensing signal on small scales we have identifiedcentral galaxies that accurately trace the center of massof their halos (George et al. 2012). Additional studiesof these groups with magnification lensing (Ford et al.2012; Schmidt et al. 2012) and clustering (Allevato et al.2012) have confirmed the halo mass estimates. These in-gredients make for a robust characterization of group en-vironments with which to study satellite properties andtheir evolution over time.We describe the data used in this analysis in Section 2along with mock catalogs used to estimate the purityof the satellite population analyzed. Section 3 presentsthe results including our primary finding of a decliningfraction of blue late disk galaxies and a rise in red earlytypes toward group centers and over cosmic time. InSection 4, we discuss the implications of these findingsfor the physical mechanisms that shape galaxy evolution,and put this work in the context of previous studies andpossibilities for the future. DATA AND MOCKS
The COSMOS field (Scoville et al. 2007) has been thesubject of a broad array of multi-wavelength imagingand spectroscopy campaigns. In this study we make useof several pieces of this data: galaxy colors, photomet-ric redshifts, and stellar masses derived from over thirtybands of ultraviolet, optical, and infrared data (Ilbertet al. 2009, 2010; Bundy et al. 2010); galaxy morphologiesdetermined by Scarlata et al. (2007) from high-resolutionimaging taken by the Advanced Camera for Surveys(ACS) aboard the
Hubble Space Telescope (Koekemoeret al. 2007); and a catalog of massive galaxy groups se-lected based on their X-ray emission seen with
XMM-Newton and
Chandra and for which halo mass profileshave been characterized using weak lensing with
Hub-ble imaging (Leauthaud et al. 2010; George et al. 2011,2012). Most of these data have been compiled in thegroup member catalog published in George et al. (2011)and we refer the reader to that paper and referencestherein for further details including the flux and masslimits for galaxies and groups. Here we briefly explainthe stellar masses, colors, and morphologies used in thispaper, as well as the identification of group centers andthe mock catalogs used for validating group membershipassignment and estimating the effects of contamination from field galaxies.
Group Membership and Mocks
As a measure of galaxy environment, we use a cata-log of groups described in George et al. (2011). Thesegroups have been identified through a wavelet detectionof extended X-ray emission following Finoguenov et al.(2010) and the sample is an expanded version of an earlierCOSMOS catalog (Finoguenov et al. 2007) with deeperX-ray data. A Bayesian membership algorithm is usedto associate galaxies with groups based on precise pho-tometric redshifts and proximity to the X-ray position,with a prior accounting for the relative field population asa function of magnitude and redshift. Following Georgeet al. (2011), we select galaxies with membership prob-ability P mem > . flag include =1 in the catalog). Halo masses for thesegroups have been measured with stacked weak lensing tobe in the range M ≈ − M (cid:12) and are correlatedwith X-ray luminosity (Leauthaud et al. 2010; Georgeet al. 2012). Over this limited range in halo masses, wehave not detected significant trends in galaxy propertieswith X-ray luminosity, but to avoid the influence of out-liers we eliminate a small number of groups with halomasses estimated from their X-ray luminosities follow-ing Leauthaud et al. (2010) that are above or below the10 − M (cid:12) range.We split galaxies into central, satellite, and field pop-ulations across several stellar mass and redshift bins toisolate dependences among different properties. For eachgroup, the central galaxy is defined as the most mas-sive group member within a projected distance of theX-ray position equal to the scale radius of a Navarro-Frenk-White (Navarro et al. 1996) profile. George et al.(2012) showed with weak lensing tests that this definitiontraces the halo center to within roughly 75 kpc, thoughin roughly 30% of groups the center is still ambiguous. Inthis paper we do not specifically study the central galax-ies in these X-ray selected halos because their abundancelimits statistical constraints on the properties of this pop-ulation. The remaining group members are called satel-lites, and are the primary focus of this work. Field galax-ies are taken to be those not assigned to any extendedX-ray source and are thus expected to reside in halos lessmassive than the X-ray detection limit ( P mem = 0 fromGeorge et al. 2011; see Figure 1 of that paper for the X-ray flux limit as a function of redshift). Because we donot detect faint X-ray groups at high redshift, the halomass range occupied by the field population can evolvedue to this selection effect. However, at low redshift only ∼
10% of galaxies in our sample live in X-ray groups, sowe expect the evolving halo mass limit to have only asmall effect on the statistics of the field population. Ta-ble 1 gives the size of each of these samples used in ouranalysis.Extensive tests of the accuracy of the photometric red-shifts and membership assignment algorithm have beencarried out using mock catalogs from simulations as wellas a subsample of galaxies with spectroscopic redshifts(George et al. 2011). The overall purity and completenessof the member selection within 0 . R is 84% and 92%,respectively, down to our flux limit of F814W = 24 . TABLE 1Environmental Census
Stellar Mass [log( M (cid:63) /M (cid:12) )]Type [9 . , .
3) [10 . , . z = 0 . − . N groups = 47Satellites 237 218Field 2455 1993 z = 0 . − . N groups = 40Satellites 122 189Field 4497 3529 z = 0 . − . N groups = 29Satellites ... 157Field ... 4170 Note . — No contamination correc-tions have been applied. Ellipses denotea bin below our stellar mass completenesslimit.
The accuracy of member selection depends most signifi-cantly on distance from the group center and on the fluxof a galaxy. The first effect is due to the decreasing den-sity of true members with group-centric distance, whichimplies that selecting galaxies uniformly with radius (outto R ) will result in higher contamination from fieldgalaxies in the outskirts. The second effect from galaxyflux arises from the decreasing precision of photomet-ric redshifts for fainter galaxies, ranging from σ z (cid:46) . < . σ z = 0 .
03 at F814W = 24. It isnotable that with the many filters used in constructingthese photometric redshifts, precision does not dependsignificantly on spectral type, so red and blue galaxiesof a given magnitude are selected with similar complete-ness.In our analyses of member populations, we correct forcontamination from field galaxies using the mock cata-logs described in George et al. (2011). A halo catalog isconstructed from a LasDamas simulation (C. K. McBrideet al., in preparation) and populated with group mem-bers following the halo occupation model of Leauthaudet al. (2012). Mock galaxies are matched to real COS-MOS galaxies in narrow bins of stellar mass and redshiftto assign magnitudes in the F814W band, which in turnare used to assign mock photometric redshift errors. Weestimate the purity of the member selection algorithmin bins of group-centric distance and magnitude, and as-sume that the colors and morphologies of contaminatingfield galaxies are representative of the field population atthat magnitude. Contamination corrections are appliedto the satellite populations only, since the field popula-tion is much larger. These corrections only account forcontamination of the member list and not for incomplete-ness; the latter effect is smaller, does not significantlydepend on group-centric distance, and should be fairlyuniform across galaxy types since the precision of photo-metric redshifts is effectively achromatic (see Figures 2and 5 of George et al. 2011). Stellar Masses, Colors, and Morphologies Details regarding the LasDamas “Consuelo” simulationcan be found at http://lss.phy.vanderbilt.edu/lasdamas/simulations.html
Stellar masses and spectral classes have been deter-mined by fitting stellar population synthesis models tothe spectral energy distributions of galaxies, varying theage, amount of dust extinction, and metallicity in themodels. We use the stellar masses from Bundy et al.(2010) and color classifications from Ilbert et al. (2010),both of which are based on Bruzual & Charlot (2003)models fit with a Chabrier (2003) initial mass functionbut take somewhat different approaches to fitting thedata. Using the unextincted rest-frame color (NUV − r + )of the best-fitting template for each galaxy, we split theminto two color classes: • red : (NUV − r + ) > . • blue : (NUV − r + ) < . − yr − .It is worth emphasizing that these colors are correctedfor dust extinction, so that the red sample selects a pop-ulation of quiescent galaxies similar to those identifiedusing two color cuts based on near-ultraviolet, optical,and near-infrared measurements, minimizing contamina-tion from dusty star-forming galaxies (e.g., Bundy et al.2010).Galaxies are classified morphologically based on au-tomated measurements of several structural parametersfrom ACS imaging in the F814W band. We use theZurich Estimator of Structural Types (ZEST; Scarlataet al. 2007) catalog, which is derived from measurementsof the asymmetry, concentration, Gini coefficient, secondmoment of the brightest pixels, and ellipticity. Scarlataet al. (2007) performed a principal component analysis ofthese measured quantities to identify the most importantstructural parameters, and divided galaxies into classes(early type, disk, and irregular) based on their locationin this principal component space. A further subdivi-sion of the disk category was derived by correlating mea-surements of the S´ersic index for a bright sample with I AB < . • early type spheroidals and bulge-dominated disksincluding even relatively inclined S0 galaxies(ZEST types 1 and 2.0) • intermediate bulge+disk galaxies (ZEST type 2.1) • late disks with little or no bulge component (ZESTtypes 2.2 and 2.3).Irregulars and unclassified galaxies make up a small frac-tion of the sample, typically not more than a few percentin any stellar mass bin. We note that morphologicalK-corrections, which have not been applied, should nothave a large effect on our sample because the F814Wband probes rest-frame optical wavelengths over our en-tire redshift range (e.g., Lotz et al. 2004; Cassata et al.2005; Bundy et al. 2010). George et al. < z <
30 kpc ≈ < z <
30 kpc ≈ < z <
30 kpc ≈ < z <
30 kpc ≈ Spheroidal Bulge+Disk Late Disk < z <
30 kpc ≈ < z <
30 kpc ≈ < z <
30 kpc ≈ < z <
30 kpc ≈ NUV − R µ Distance from Group Center [R/R ] S t e ll a r M a ss [ l og ( M ? / M (cid:12) ) ] Fig. 1.—
Group members as a function of stellar mass and distance from the group center. Colors for each galaxy represent the averageunextincted rest-frame template (NUV − r + ) color, with shading proportional to the logarithmic surface brightness µ . The gray band at thebottom of the high-redshift panels shows the stellar mass completeness limit for a passive population calculated with our flux limits F814W= 24 . K s = 24 following the approach of Bundy et al. (2010) (see also Figure 1 of George et al. 2011). Central galaxies in these groupsare shown in the light gray band on the left side of each of the outer panels. The middle frame shows morphological classifications for arandom sample of galaxies chosen to span the range of colors observed; objects in this panel are sorted vertically by color and horizontaloffsets within each classification are arbitrary. ransformers 5 RESULTS
We begin our analysis with a visualization of a fewproperties of the galaxies in our sample. Figure 1 showsthe distributions of stellar masses and group-centric dis-tances in different redshift bins for all galaxies selected asgroup members. Each point represents one object and isdisplayed using the ACS image of the galaxy (Koekemoeret al. 2007), colored according to the average unextinctedrest-frame template (NUV − r + ) color. Centrals are po-sitioned on the left of each panel with small horizontaloffsets, and satellites are plotted at their distance fromthe corresponding central. Each image is basically theset of adjacent pixels with flux above a noise thresholdshown with a logarithmic surface brightness scale thathas its maximum set to the peak value for each galaxy.We note that the apparent size of a galaxy on the plot isquite sensitive to its flux since images of brighter galax-ies have more pixels above the noise threshold, and thisis not a perfect indicator of the physical effective radiusof a galaxy. A small fraction of objects have deblend-ing issues or edge effects from the cutout image size, butwe note that the images are processed independently forvisualization and analysis purposes.Several trends are evident when visualizing galaxies inthis manner. Stellar mass is a strong determinant ofgalaxy properties; massive galaxies are more likely to belarge, red, and spheroidal. Physical properties also de-pend on the location within a group. Central galaxiesare massive (by definition), but also typically red andelliptical. Blue centrals tend to be less massive than redcentrals at high redshifts even within the fairly narrowrange of halo masses studied here, as discovered by Tin-ker et al. (2012). Satellites closer to group centers aremore likely to be red and show fewer spiral features thanin the outskirts, particularly at low stellar masses. Thereis a relative dearth of low mass satellites near group cen-ters; this could be a hint of satellite depletion due tomergers, but challenges in measuring photometry of faintobjects near massive extended central galaxies could bea contributing factor so further investigation is needed.Similar evidence of mass segregation and the influence ofmass and environment on star formation has been seenin groups and clusters out to z ∼ z = 1.Sample variance due to the finite survey volume does af-fect our ability to measure absolute redshift trends asthe number densities vary significantly due to large-scalestructures (particularly evident at z = 0 . − . z = 0 . − . M (cid:63) /M (cid:12) ) >
10. Whilethese classifications do not always agree with one’s vi-sual impression, there are clear differences in structuralparameters among the classes, and correlations betweenthese automated measurements and a galaxy’s positionin a group can provide an interesting test of the depen-dence of morphology on environment. We emphasize that the ZEST morphologies are correlated with, but notidentical to, traditional visual classification of ellipticalsand spirals. We compare results from multiple morpho-logical indicators in Section 3.3.
Radial Trends: Blue Late Disks into RedBulge+disks
We can quantify some of the trends from Figure 1 bymeasuring the fraction of galaxies of a given color andmorphology as a function of stellar mass, group-centricdistance, and redshift. Figure 2 shows the fraction ofgalaxies in each of the six combinations of color and mor-phology categories described in Section 2.2. For exam-ple, the cyan triangles represent the fraction of galaxiesin the stellar mass and redshift range shown that areboth blue and have late disk morphology. This popula-tion makes up the majority among field galaxies (shownat
R > R ) but its fraction declines among satel-lites, contributing less than 20% of the satellite popu-lation at
R < R /
3. Meanwhile, the proportion ofred bulge+disk galaxies rises from 7% in the field to 40%among satellites in the inner radial bin. . . . . . . . . . . . . . log(M ? /M (cid:12) )=[9.8, 10.3)z=[0.2, 0.5)Red SpheroidalRed Bulge+DiskRed Late DiskBlue SpheroidalBlue Bulge+DiskBlue Late Disk Distance from Group Center [R/R ] F r a c t i on | M ? , z Fig. 2.—
Color and morphological fractions as a function ofgroup-centric distance. These fractions are calculated for satellitesin three equally-spaced bins out to R after applying contam-ination corrections. The field population is plotted to the right.Points are assigned small horizontal offsets for clarity. Error barsare the 1 σ standard deviation of 500 bootstrap samples. There is aclear transition from blue late disks dominating among field galax-ies and outer satellites to red bulge+disks among inner satellites. Figure 2 highlights the most significant environmentaltrends in our sample by focusing on low mass, low red-shift galaxies. We repeat the exercise in Figure 3 withhigher mass and redshift bins to study how these trendsvary. The broad picture is similar; blue late disks dom-inate in the field while the red bulge+disk populationbecomes more prominent toward group centers. Amonglow mass galaxies at z > .
5, the blue late disks dom-inate everywhere, while more massive galaxies at lower George et al.redshift have a substantial population of red spheroidals.Red late disks and blue spheroidals do not make up alarge portion of galaxies at any mass or redshift studied.The red fraction can be determined from these plots bysumming the three red lines, and similarly the spheroidalfraction is the sum of the solid lines with circular mark-ers. The red fraction rises toward group centers for lowmass galaxies, but is relatively flat among massive galax-ies. We note that the abscissa for these plots is the pro-jected group-centric distance, and that the true radialtrends measured in spherical shells are likely more sig-nificant than observed. . . . . . . . log(M ? /M (cid:12) )=[9.8, 10.3) z = [ . , . ) log(M ? /M (cid:12) )=[10.3, 10.8) . . . . . . . . . . . . . z = [ . , . ) . . . . . . . . . . . . . z = [ . , . ) Red SpheroidalRed Bulge+DiskRed Late DiskBlue SpheroidalBlue Bulge+DiskBlue Late Disk
Distance from Group Center [R/R ] F r a c t i on | M ? , z Fig. 3.—
Color and morphological fractions of group members asa function of distance from the group center, for different bins ofstellar mass (columns) and redshift (rows). Line styles and errorbars are as defined in Figure 2. Figure 2 is repeated in the topleft panel, for comparison with weakening environmental trends athigher mass and redshift.
Redshift Trends
Since satellites tend to fall toward halo centers, group-centric distance is related to the timescale that galaxieshave been inside the group. The range of redshifts sam-pled with this data set provides another measure of timeto study evolution. Figure 4 is a transpose of Figure 3 toshow the redshift trends in the distribution of colors andmorphologies. Again, there is a decline in blue late disksamong low-mass satellites near the centers of groups, nowcompensated by a rise in both red spheroidals and redbulge+disks. This trend weakens away from group cen-ters and at higher masses.
Checks for Systematics
Environment . . . . . . . log(M ? /M (cid:12) )=[9.8, 10.3) I nne r S a t e lli t e s log(M ? /M (cid:12) )=[10.3, 10.8) . . . . . . . O u t e r S a t e lli t e s . . . . . . . . . . . . . . . . . F i e l d Redshift F r a c t i on | M ? , R / R c Fig. 4.—
Color and morphological fractions as a function of red-shift, for different bins of stellar mass (columns) and environment(rows). Inner and outer satellites are separated at a projectedgroup-centric distance of 0 . R . Line styles and error bars areas defined in Figure 2. There are several possible biases or other effects in thedata to consider in order to ensure the robustness of theseresults. First, we revisit the contamination of our satel-lite sample due to interloping field galaxies, discussedin Section 2.1. In Figures 2, 3, and 4 we have plottedvalues of population fractions for satellites corrected forcontamination estimated from mock catalogs. Contami-nation corrections are always smaller than the statisticalerror bars estimated via bootstrapping, except for lowmass blue disk galaxies where it is 10% greater than theerror because the field population is so large. Though thesample of satellites near R is significantly contam-inated by field galaxies, the corrections to the relativefractions of each galaxy type are small because the fieldpopulations are not markedly different from the outersatellites. Color Distribution
When classifying galaxies by their spectral energy dis-tributions (SEDs), our primary aim is to distinguish star-forming galaxies from those that are quenched. Thoughtraditional indicators from emission lines or spectralbreaks are not directly available from photometry, the31-band SEDs used here provide a wealth of informationabout spectral types, including an estimate of dust ex-tinction that separates star-forming galaxies that appearred due to dust from those that are truly passive. Theransformers 7 log(M ? /M (cid:12) )=[9.8, 10.3) z = [ . , . ) log(M ? /M (cid:12) )=[10.3, 10.8) − z = [ . , . ) − z = [ . , . ) Late DiskBulge+DiskSpheroidal
NUV-r + N Fig. 5.—
Distribution of rest-frame, extinction-corrected, tem-plate colors by morphological type. Galaxies from all environmentsare included. Vertical dotted lines show the cut used to segregatered and blue galaxies. template-based, extinction-corrected (NUV − r + ) colorsused in this paper generally have a bimodal distributionwith the color cut from Section 2.2 falling on the red endof the “green valley.” The color distributions for differ-ent morphological types are shown in Figure 5. Shiftingthe cut slightly in either direction shifts the amplitudeof the red fraction in Figures 2, 3, and 4 up or down,but the trends with group-centric distance and redshiftdo not vary significantly. We have tested an alternative“red sequence” selection suggested by Ilbert et al. (2010),NUV − r + > . M (cid:63) /M (cid:12) ) − . z − .
5, using rest-frame absolute magnitudes but applying no extinctioncorrection. The transformation of blue late disks intored bulge+disks among low mass satellites is still evident.Similarly, applying the two-color cut in the NUV − r + ,r + − J plane used by Bundy et al. (2010) does not qual-itatively change our results.
Morphologies
Morphological classification is a challenging problemand significant scatter exists between the types assignedto galaxies in both visual and automated analyses. Ingeneral, visual analyses emphasize the presence or ab-sence of spiral features, categorizing objects as spiralsor ellipticals, often with an intermediate class of S0sgrouped with ellipticals. Automated analyses measurestructural parameters such as concentration and asym-metry which are then generally tied to a training set ofvisual classifications. The correlation between the prop-erties measured is imperfect, so we test the impact onour results of using a variety of automated morpholog-ical classifications, all based on the ACS F814W imag-ing. The alternative classifications come from Tasca et al.(2009), which presented three separate techniques.The differences between the results of each catalog andthose from the ZEST classification used in this paper are driven by how bulge+disk galaxies are classified, sincethe other catalogs do not split the spiral/disk categoryinto multiple bins as ZEST does. For instance, amongthe 237 satellites used for Figure 2, the ZEST classifi-cations are 22% spheroidal, 36% bulge+disk, 37% latedisk, and 5% irregular or unclassified. The three sep-arate classifications for these galaxies from Tasca et al.(2009) vary between 39 −
59% E/S0s and 59 −
40% spi-rals. When bulge+disk galaxies are mostly classified asE/S0s, the spheroidal populations (both blue and red) inFigures 3 and 4 are significantly elevated but show sim-ilar trends with group-centric distance and redshift. Inthe opposite case where bulge+disk galaxies are mostlytreated as disks, the spheroidal fraction is essentially un-changed. While the dominant population in each panelof Figure 3 and 4 can change depending on which cate-gory the intermediate bulge+disk galaxies fall into, theradial trend in low mass satellites is unchanged; the bluelate type fraction declines toward group centers and iscompensated with a rise in red early types. The redshifttrend for this transition is strongest for low mass satel-lites near group centers and weakens toward larger radiiand stellar masses. Though the scatter between mor-phological classifications signals that our results shouldbe interpreted and compared to others with caution, thesignificant trends with group-centric distance and red-shift for the intermediate bulge+disk galaxies suggeststhat the ZEST classification has identified a populationin transition.
A Population of Blue Spheroidals
While the tests described above suggest that our mea-sures of environment, color, and morphology are ro-bust, we do note a puzzling population of massive bluespheroidals that can be seen most clearly in the bot-tom right panel of Figure 4 as well as the right col-umn of Figure 5. Those plots suggests that half of allspheroidals in that stellar mass range are blue, exceed-ing measurements in other studies (e.g. Kaviraj et al.2007, 2008; Bamford et al. 2009; Schawinski et al. 2009;Ilbert et al. 2010), but see also Cross et al. (2004) whofound a large fraction of blue ellipticals in a luminosity-selected sample at moderate redshift. We see a similarlylarge blue fraction among spheroidals at higher masses(log( M (cid:63) /M (cid:12) ) > .
8, not plotted) where spheroidalsmake up a higher proportion of all galaxies, and mostprominently at low redshift ( z < . − − r + ) colors (see e.g. Smith et al. 2012,for a discussion of galaxy properties contributing to the“UV upturn”). While this population is interesting in George et al.its own right, it is most prominent at high stellar massesand does not show a strong dependence on environment,so we do not consider it further here. DISCUSSION AND CONCLUSIONS
Our results indicate a complex relationship betweencolor, morphology, stellar mass, and group-centric dis-tance. Trends in color and morphology are distinct, andthe use of a single property as proxy for galaxy typedoes not capture the whole picture. The most inter-esting trend seen is the shift in dominance among thelow mass population from blue late disk galaxies in theoutskirts to red bulge+disk types in group interiors. Ifpart of one population was merely disappearing from thesample, either due to changes in mass or merging withother galaxies, then the other populations would all beexpected to grow by an equal factor. The fact that thedecline in one population is roughly balanced by the riseof a single other population suggests a transformationprocess. We now discuss these results in the context ofpast analyses, the physical implications of the presentwork, and future avenues to clarify the role of environ-ment in galaxy evolution.
Connection to Previous Observations
There is a long history of research into the covariancebetween the stellar masses, colors, morphologies, and en-vironments of galaxies and its evolution with time. Af-ter controlling for differences in stellar mass or luminos-ity, numerous studies at low redshift have found thatcolor is more strongly correlated with environment thanmorphology (e.g., Kauffmann et al. 2004; Blanton et al.2005; Christlein & Zabludoff 2005; van den Bosch et al.2008; Bamford et al. 2009; Skibba et al. 2009; Weinmannet al. 2009). The implication of these studies is thatthe well-known correlation between morphology and en-vironment is secondary to correlations between morphol-ogy and stellar mass or color, with the latter propertiesmore physically linked to environment. Still, some ofthese studies find residual correlations between morphol-ogy and environment after controlling for color and stel-lar mass or luminosity, particularly at low masses andamong late to intermediate morphologies (Blanton et al.2005; Weinmann et al. 2009; Skibba et al. 2012).Moving from large low-redshift studies to moderateredshifts ( z ∼ . z ∼ z = 0 . −
1. Our X-ray group catalog gives a large, clean,and fairly representative selection of 10 − M (cid:12) ha-los (Finoguenov et al. 2010), and has been well-calibratedbased on its weak lensing signal (Leauthaud et al. 2010;George et al. 2012). We detect significant trends in bothcolor and morphology with group-centric distance. Someprevious studies measured weak or insignificant gradientsin morphology by using a simple dichotomy of spiralsand ellipticals, and we can reproduce these results whenusing a coarse morphological binning (see discussion inSection 3.3.3). But in contrast to those results, we seeclear morphological gradients at fixed stellar mass andcolor once the morphological classification considers dif-ferences in the bulge content of disk galaxies.The importance of this intermediate morphology be-tween pure disks and spheroidals has been noted beforein the context of S0 galaxies which have been known todominate the evolution in the morphology-density rela-tion (Dressler et al. 1997; Postman et al. 2005; Smithet al. 2005; Boselli et al. 2006; Moran et al. 2007; Oeschet al. 2010; Lackner & Gunn 2013). Smith et al. (2005)also showed that morphological evolution occurs laterin less dense regions than in the densest regions asso-ciated with clusters, with little evolution at intermedi-ate densities from z = 1 to 0 . Implications for Physical Mechanisms of GalaxyTransformation
The radial gradients and redshift trends measured inSection 3 suggest a transformation among low masssatellites from blue late disks into red bulge+disks andspheroidals. The mechanism for this transformationmust affect both color and morphology, or more phys-ically, the star formation rate and stellar kinematics.Some processes (see e.g. Boselli & Gavazzi 2006, for areview) halt star formation without significantly alter-ing stellar structure, such as gas removal via ram pres-sure stripping or weak tidal interactions, suppression ofgas accretion within dense shock-heated environments,or quasar feedback. On the other hand, galaxy mergersand strong tidal interactions can affect both the distribu-tion of gas needed to form stars as well as the stellar mor-phology. A third possibility is that color and morphologyransformers 9changes are physically coupled; bulge growth could sta-bilize a galaxy against disk fragmentation and suppressstar formation (Martig et al. 2009), or gas loss may leavea disk unable to dissipate energy from tidal interactionsor may drive instabilities leading to bulge growth. Yetanother scenario is that a bulge only appears more promi-nent after a galaxy is quenched because the previouslystar-forming disk has faded.We can test these models by studying the morpholog-ical dependence on environment among quenched galax-ies. Mechanisms like gas stripping or disk fading that donot directly affect morphology should produce a higherfraction of quenched galaxies in dense environments, butamong quenched galaxies the morphological distributionshould be constant across environments. On the otherhand, if quenched galaxies in dense environments havemore bulge-dominated morphologies than quenched fieldgalaxies, it would suggest that mergers or strong tidalinteractions altering the structure of galaxies occur inaddition to, or in conjunction with, the suppression ofstar formation. The results of this test are plotted inFigure 6, where we show the fractions of red galaxiesthat are early and late type as a function of environment.We have used a broad redshift range and combined thespheroidal and bulge+disk categories to reduce statis-tical errors for this smaller population of red galaxies.Figure 6 demonstrates that quenched satellites, particu-larly those in the inner regions of groups, are more likelyto be bulge-dominated than their field counterparts.We interpret this to mean that some physical processin dense environments is driving bulge growth. Thoughsome field studies (Bundy et al. 2010; Masters et al.2010) have already noted a tendency of passive disks tobe more concentrated than star-forming disks, we findthat the morphological evolution of quenched galaxies iseven stronger in groups. Our result is consistent withprevious arguments favoring bulge growth over disk fad-ing based on the higher typical luminosities of galaxieswith intermediate morphologies compared to late types(Christlein & Zabludoff 2004; Burstein et al. 2005). Thedata presented here show that even at fixed mass theremust be bulge enhancement that accompanies quenchingin groups.The results of Figure 3 suggest that the timescale forthe morphological transition from late disk to bulge+diskmust be comparable to the timescale for quenching in or-der to turn blue late disks directly into red bulge+diskgalaxies. For example, if quenching occurred via gasstripping or removal at a rate faster than any structuralchanges, we would expect to see blue late disks turn-ing into red late disks. Instead we observe a growth inred galaxies with earlier morphologies whenever blue latedisks decline. The fraction of blue late disks plummetsby more than a factor of two in ∼ . . . . . . . . . . . . log(M ? /M (cid:12) )=[9.8, 10.3)z=[0.2, 0.8)Spheroidal & Bulge+DiskLate Disk Distance from Group Center [R/R ] F r a c t i on | M ? , z , r e d Fig. 6.—
Morphological fractions among red galaxies as a func-tion of distance from the group center. The small but significantexcess in early-type morphologies among inner satellites relativeto field galaxies suggests that mergers or tidal interactions causemore significant bulge growth among satellites. categorized as disks in the top level ZEST classification.This echoes earlier results finding a buildup of S0s inclusters but weaker evolution in the relative abundance ofellipticals (e.g., Dressler et al. 1997). Though the fractionof red spheroidals does not correlate significantly withgroup-centric distance (Figure 3), it does grow globallywith time (Figure 4). This may indicate that differentprocesses are responsible for producing bulge+disks andspheroidals.The discreteness of morphological classification makesdetailed comparison of transformations in morphologyand color somewhat difficult, but the structural evolutionappears to contrast with that in color, where a clear bi-modality separates red and blue states and the relativelylow intermediate population suggests a fast quenchingtimescale once it begins (e.g., Wetzel et al. 2012a, thoughsee Balogh et al. 2011). One could test the hypothesisthat galaxies pass through an intermediate state betweenblue late disk and red bulge+disk by measuring the timesince quenching based on the SEDs of galaxies in these in-termediate states. Is there a difference in the mean colorof blue bulge+disks and blue late disks, or between redbulge+disks and red late disks? Figure 5 does not showconclusive differences in the colors of these populations,and the satellite sample size is too small to constraindifferences in the distribution of colors within the redand blue populations. More detailed analysis with spec-troscopic data could better constrain such evolutionarymodels.The group environment can play an important rolein building up the well-studied color-morphology-densityrelations in nearby clusters. A group of mass 10 . M (cid:12) at z = 1 should grow through mergers by an average of0 . z = 0 (Fakhouri et al. 2010), so some of ourhigh redshift groups will be the progenitors of massive0 George et al. R c R c G r oup s Late Disk Bulge+Disk SpheroidalBlueRed disk instabilitiestidal interactionsminor mergers m a j o r m e r ge r s qua s a r f eedba ck r a m p r e ss u r e s t a r v a t i on MorphologyColor
Fig. 7.—
Illustrations of our main results and interpretation. Left panel shows the projected positions of satellites in an ensemble groupwith the same stellar mass and redshift range as Figure 2 with only the blue late disks (which dominate the outskirts) and red bulge+disks(which dominate the interior) displayed. The schematic diagram at right shows the effects of various physical mechanisms on color andmorphology; the large arrow indicates the observed transformation from blue late disks to red bulge+disks, suggesting a combination ofprocesses. clusters and others may accrete onto them. For compar-ison, a typical star-forming galaxy should grow by 0 . . z = 1 to 0 (Elbaz et al. 2011).Wetzel et al. (2012b) show that the dominant popula-tion of satellites in the most massive halos at z = 0 werealready satellites in groups when accreted. Our resultsdemonstrate that both the color and morphology of thesesatellites are influenced in the group environment, show-ing the importance of “preprocessing” of galaxies priorto entering massive clusters. The bulge+disk populationwe see here is not quite as morphologically evolved asthe growing S0 population seen at low redshift and inmassive clusters, but likely precedes it.We summarize our main results and interpretation inFigure 7. The left panel shows the positions of satel-lites for a stacked group and highlights the trend in Fig-ure 2 with blue late disks dominating the outskirts andred bulge+disk galaxies making up most of the innersatellite population. This combination of color and mor-phology transformations suggests some combination ofgas removal and bulge growth, shown in the right panel.While the diagram is a simplification of a wide variety ofmodel predictions, the primary point is that both colorand morphology are affected in group satellites, so phys-ical mechanisms that explain environmental correlationsshould reflect this. Future Prospects
While these measurements add insight to the transfor-mation of galaxy properties, much work remains to bedone to constrain the variety of physical mechanisms thatmay be responsible. A cleaner bulge-disk decompositionwould be useful to track the growth of bulges and theimportance of disk fading (e.g., Lackner & Gunn 2012,2013). Incorporating the physical sizes of each compo- nent would also help to link these transformations withthe significant growth seen in early types since z ∼ HST , the CANDELS survey (Grogin et al. 2011;Koekemoer et al. 2011) is pushing these studies to higherredshift and should also allow bulge-disk decompositionto be studied at multiple wavelengths, identifying whenand where star formation is happening or stopping. Al-ternatively, visual classifications from an ongoing GalaxyZoo project in the CANDELS fields will track the evo-lution of spiral features and bars which are challengingfor automated techniques. Impending wide field imagingsurveys will add greatly to the statistics of these analyses.Different measures of environment may also help dis-entangle the relevant physical mechanisms. While wehave argued in favor of halo-based indicators (see Georgeet al. 2011 for a discussion), the local galaxy density within a halo , in addition to group-centric distance, mayshed light on galaxy-galaxy interactions and infallingsubstructure (e.g., Blanton & Berlind 2007; Cibinel et al.2012; Woo et al. 2012). However, this indicator is a noisyquantity affected by shot noise, redshift errors, and pecu-liar velocities, likely requiring deep and complete spectro-scopic data for clean results. Larger surveys will also en-able a wider range of halo masses to be probed at higherredshift. Comparison of different halo mass proxies canprovide a test of environmental mechanisms, for exampleif X-ray bright groups are more efficient at ram pressurestripping satellites than groups with less hot gas.These studies can also be extended to the interplay ofenvironmental mechanisms with non-stellar componentsof galaxies, namely the gas content and central blackholes in galaxies. Current and planned radio arrays willextend the study of neutral and molecular gas beyondthe local Universe allowing a clearer picture of how star-formation is fed and quenched. And the growth of bulgesransformers 11seen in this paper should be connected with accretiononto the central black hole in order to explain the tightcorrelation seen locally between these components.Finally, while we have simply presented an empiricaldescription of the data here, we can extend this work bymodeling the evolution of color and morphology with en-vironment. Peng et al. (2010) and Wetzel et al. (2012b)present simple empirical descriptions of the fraction ofquenched galaxies as a function of stellar mass and envi-ronment. Incorporating morphologies into this frame-work will further illuminate the physical processes atwork that build up the environmental correlations weobserve.We thank Andrew Wetzel, Frank van den Bosch, andNic Ross for helpful discussions, and Claire Lackner for constructive comments on a draft. MRG acknowl-edges support from the US Department of Energy’sOffice of High Energy Physics (DE-AC02-05CH11231)and a Graduate Research Fellowship from the US Na-tional Science Foundation. CPM is partially supportedby a grant from the Simons Foundation ( http://cosmos.astro.caltech.edu . This research has madeuse of the NASA/IPAC Infrared Science Archive, whichis operated by the Jet Propulsion Laboratory, CaliforniaInstitute of Technology, under contract with the NationalAeronautics and Space Administration.. This research has madeuse of the NASA/IPAC Infrared Science Archive, whichis operated by the Jet Propulsion Laboratory, CaliforniaInstitute of Technology, under contract with the NationalAeronautics and Space Administration.