Morphological evolution in situ: Disk-dominated cluster red sequences at z ~ 1.25
MMon. Not. R. Astron. Soc. , 1–13 (2002) Printed 6 October 2018 (MN L A TEX style file v2.2)
Morphological evolution in situ: Disk-dominated clusterred sequences at z ∼ . Roberto De Propris (cid:63) Malcolm N. Bremer and Steven Phillipps Finnish Centre for Astronomy with ESO, University of Turku, Finland H.H. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol, BS8 1TL, UK, United Kingdom
ABSTRACT
We have carried out a joint photometric and structural analysis of red sequence galax-ies in four clusters at a mean redshift of < z > ∼ .
25 using optical and near-IR HSTimaging reaching to at least 3 magnitudes fainter than M ∗ . As expected, the photome-try and overall galaxy sizes imply purely passive evolution of stellar populations in redsequence cluster galaxies. However, the morphologies of red sequence cluster galaxiesat these redshifts show significant differences to those of local counterparts. Apartfrom the most massive galaxies, the high redshift red sequence galaxies are signifi-cantly diskier than their low redshift analogues. These galaxies also show significantcolour gradients, again not present in their low redshift equivalents, most straight-forwardly explained by radial age gradients. A clear implication of these findings isthat red sequence cluster galaxies originally arrive on the sequence as disk-dominatedgalaxies whose disks subsequently fade or evolve secularly to end up as high S´ersicindex early-type galaxies (classical S0s or possibly ellipticals) at lower redshift. Theapparent lack of growth seen in a comparison of high and low redshift red sequencegalaxies implies that any evolution is internal and is unlikely to involve significantmergers. While significant star formation may have ended at high redshift, the clus-ter red sequence population continues to evolve (morphologically) for several Gyrsthereafter. Key words: galaxies: evolution — galaxies: clusters — galaxies: luminosity function,mass function — galaxies: clusters: general — galaxies: elliptical and lenticular, cD
The common picture of galaxy populations in clusters isthat they are remarkably homogeneous in nearby as well asdistant objects, out to z ∼ . z ∼ .
5, other than pure passive evolution oftheir stellar populations (De Propris et al. 1999, 2007; An-dreon 2006, 2008; Muzzin et al. 2008; Mancone et al. 2010,2012) at ∼ L ∗ and brighter, implying no significant growthof cluster galaxies over the past 2/3 of the Hubble time.Similarly, distant clusters studied thus far are usually domi-nated by red sequence galaxies with colours consistent withpassively evolved local samples, having high formation red-shifts and short ( ∼ (cid:63) E-mail:email@address (AVR); otheremail@otheraddress(ANO) dama & Arimoto 1997; Blakeslee et al. 2003; Mei et al.2006a,b, 2009, 2012). This argues for a model where clus-ter galaxies assembled most of their stellar content at earlytimes and with their stellar populations formed over shortperiods (e.g. Pipino & Matteucci 2004).Unlike old early-type field galaxies at similar redshifts(Longhetti et al. 2007; Cimatti et al. 2008; Damjanov et al.2009; Whitaker et al. 2012; Cassata et al. 2013; Williamset al. 2014) which are more compact than similarly massivelow redshift counterparts (Shen et al. 2003), the sizes of redcluster galaxies also do not appear to have evolved appre-ciably (Cerulo et al. 2014; Delaye et al. 2014; Jørgensen &Chiboucas 2013; Jørgensen et al. 2014) since high redshift.Consequently, the red sequence cluster galaxies have beenpresumed to be ”red and dead”, at least as far as their starformation activity and growth is concerned.More recent work has shown that at least for some clus-ters at higher redshifts there is evidence for a population ofmassive blue star-forming galaxies in their centres (Brodwinet al. 2013; Zeimann et al. 2013; Alberts et al. 2014; Mei et al.2014) with a possible ’reversal of fortune’ in the morphology-colour-density relation at z > . c (cid:13) a r X i v : . [ a s t r o - ph . GA ] M a r R. De Propris et al. unlike the classical Butcher-Oemler effect (Butcher & Oem-ler 1978, 1984; Dressler 1984) where the excess blue systemsin high redshift clusters tend to be low mass galaxies (DePropris et al. 2003b; Pimbblet & Jensen 2012), but insteadinvolves star-forming galaxies with masses comparable tothe spheroids that dominate the red sequence populationin the local Universe. At these redshifts, there appears tobe diversity in star formation and growth between clusters,e.g. some show evidence of on-going mass assembly amongluminous galaxies (Rudnick et al. 2012; Fassbender et al.2014) while others appear to resemble passive local systems(Andreon et al. 2014; Koyama et al. 2014).This diversity is the likely signature of the end of theepoch of significant star formation and growth in the coresof massive clusters, any later evolution in this galaxy popu-lation is necessarily more subtle. Morphologically, it is usu-ally assumed that passive red colours correlate with spheroiddominated structures, i.e. the red sequence galaxies are earlytypes, but we should note that passive or red spirals (disks)also exist at low redshifts (Koopmann & Kenney 1998; Wolfet al. 2009) which indicates that stellar population andstructural evolution do not have to be synchronous.One key advantage of studying cluster galaxy popula-tions as opposed to field or group systems is that once agalaxy resides within a massive cluster it remains in such anenvironment thereafter. We can therefore identify the pro-genitors of local red sequence galaxies in distant clustersprecisely because of the lack of significant evolution in lu-minosity functions and colours, whereas in the field galaxypopulation evolution can only be measured through the sta-tistical changes in the properties of the sample at differentredshifts. The aim of this paper is to measure the evolutionof luminosity, colour, size and shape for galaxies in clustersat 1 . < z < . z < . In order to probe both the photometry and morphologyof red sequence galaxies down to low near-IR luminosities(and consequently stellar masses) at 1 < z < . z > z = 1 .
27 as only optical data (but no in-frared) are available in the archive.For this work we aim to be able to reach at leastM ∗ + 2 . ( M stellar /M (cid:12) ) ∼ .
5) assuming only passive evolutionin luminosity. We then require optical data (probing therest frame blue/near-UV) of a depth sufficient to identifyand characterise most or all of the red sequence galaxies incombination with the near-IR. In practice, the reddest bandavailable in the archival data is the WFC3 H F W , whichcorresponds to the rest-frame R band at the redshifts of in-terest here. Given the current state of the archive we choseto study four clusters at 1 < z < . z = 0 .
98, RDCS1252+2927 at z = 1 .
24, ISCS1434 at z = 1 .
24 and XMM2235 at z = 1 .
40, with near-IR data ob-tained as part of Proposal 12051, PI Perlmutter. Exposuretimes were of around 4ks each. Optical imaging in the i and z bands, to sample the rest-frame U and B were derived froma variety of sources (XMM1229 from program 10496, PI:Perlmutter; RDCS1252 from PID 9290, PI: Ford; ISCS1434also from PID 10496 and XMM2235 from PID 10531, PI:Mullis). Exposure times vary between 1.5 and 7.2 ks in i and between 3 and 10 ks in z .Initial photometry was carried out exactly as in ourearlier work (De Propris et al. 2013). We used Sextractor (Bertin & Arnouts 1996) with the same parameters as usedpreviously to measure both total and aperture magnitudeson each separate image and bandpass. In addition to thecluster fields, we used similarly-deep reference fields whichare treated in exactly the same way in order to understandthe expected contribution from foreground/background in-terlopers when determining luminosity functions. For thereference fields we used the same data on the ExtendedGroth Strip as that used in De Propris et al. (2013), aswell as H band images of the Early Release Data 2 field,which covers part of the southern GOODS field (Giavaliscoet al. 2004). The latter provides us with i and z photometry,allowing us to derive the red sequence luminosity functionsin addition to the overall cluster luminosity functions. Noteall magnitudes are on the AB system.For star-galaxy separation, we plot the central surfacebrightness (measured in the Sextractor detection aperture)vs. total magnitude in Fig. 1. Stars are identified as a highsurface brightness ‘edge’ at all luminosities. All objects re-maining after star-galaxy separation were visually inspectedto remove possible contaminants such as cosmic ray streaks,diffraction spikes, brighter galaxies segmented by the algo-rithm and (for cluster fields) large gravitational arcs. Notethat we performed a similar procedure on our reference fieldsas well. These will be discussed in more detail in a futurepaper.In addition to the standard flux cuts typically used inphotometric selection of galaxy samples, we also considerthe effect of surface brightness on the completeness of anysample of red sequence objects. While luminous galaxies aredetected within a broad wedge-shaped region in Fig. 1, thereis a horizontal ‘cut’ at low surface brightnesses, which repre-sents the detectability limit set by the sky brightness. Whilethis significantly affects the bulk of the galaxy distribution at c (cid:13) , 1–13 isky cluster red sequence galaxies
20 22 24 26 H F160W [mag.] µ ( H ) [ m a g a r c s ec - ] Blue galaxiesRed galaxies
Figure 1.
Central surface brightness vs. total magnitude in the H band for galaxies in our clusters. Red sequence galaxies (de-fined below) are plotted as filled grey circles, while blue galaxies(including non cluster members) are shown as plus signs. The di-agonal cutoff at bright surface brightnesses represents the locus ofstars. There is a region below µ ( H ) ∼ . − wherethe density of galaxies falls at H > .
5, giving the impression of a’cut’ in the plot. This (shown as a dashed line) is the explicit sur-face brightness limit of our data. Galaxies may be brighter thanthe actual completeness limit but still be undetected if their cen-tral surface brightness is below the sky (cf. Disney 1976, Phillipps& Driver 1995). H ∼
25, comparatively low surface brightness outliers startto be lost by
H > .
5, so we limit our sample selectionto brighter than this. The surface brightness limit inducesa significant selection effect (e.g., Disney 1976; Phillipps &Driver 1995) which we will discuss later in this paper inrelation to the red sequence luminosity functions.
We first reconfirm several results from published studies ofthe luminosity functions and colour magnitude relations ofdistant clusters.The data at hand reach to as faint (e.g. Strazzullo etal. 2010) or fainter luminosities than has been previouslyachieved at this redshift, enabling us to better measure thefaint-end slope of the galaxy luminosity function and studythe evolution of the red sequence, which allows us to relateour further analysis of sizes and morphologies to galaxies more representative of the descendant population in lowerredshift clusters rather than purely the most massive galax-ies. As in our previous paper on a similar HST dataset at0 . < z < . H band from the background-subtracted counts in each cluster field, following the proce-dure described by Colless (1989). We have shifted all clustersto a mean redshift of z = 1 .
25, following the procedure inDe Propris et al. (1999), assuming a distance modulus fromthe conventional cosmological model and differential k + e corrections (between the cluster z and z = 1 .
25) from aBruzual & Charlot (2003) model forming its stars at z = 3with solar metallicity and an e-folding time of 1 Gyr, whichgenerally yields a good fit to the colour-magnitude relationsof the red galaxies in distant clusters (e.g., Mei et al. 2009;Rudnick et al. 2012).Fig. 2 shows the composite H band luminosity functionderived from the combination of the four clusters, havingshifted the photometry of each to z = 1 .
25, the best fit-ting Schechter function and the derived error ellipses. Er-rors include a contribution from non-Poissonian clusteringcomputed as in Huang et al. (1997) and Driver et al. (2003).The best fit to this composite function has H ∗ = 21 . α = − . H ∗ is the apparent magnitude at z = 1 . ∗ . The uncertainties on both are shownin the inset of Fig 2 as 1 , σ error ellipses. In the figurewe also show the composite infrared luminosity function of10 local clusters measured from a pure spectroscopic sam-ple in De Propris & Christlein (2009), shifted to z = 1 . H = 22, thefaint end slope appears well-determined (by eye it is consis-tent with α ∼ − ∗ andthe resulting uncertainty in positioning the exponential cut-off affects the determination of the faint end slope. The un-certainty in the bright end is dominated by the backgroundsubtraction. As we show later, the effect of this is decreasedsignificantly in the determination of the luminosity functionof the red sequence alone, purely because the bright red se-quence galaxies fall in an otherwise less populated region ofcolour-magnitude space. A larger cluster sample would beneeded to pinpoint M ∗ more accurately.For our chosen cosmology our luminosity function at z = 1 .
25 has M ∗ H = − .
6. Given the local luminosityfunction from De Propris & Christlein (2009) and assum-ing H − K = 0 . M ∗ values within 5%, given the e + k corrections cal-culated from the passively evolving model. The implied evo-lution of M ∗ , in comparison to low redshift clusters, is con-sistent with a pure passive evolution scenario where brightgalaxies are assembled at high redshift and suggests that nosignificant growth has occurred in the galaxy luminosities(or rather, in stellar masses) since at least the redshifts ofthe present sample of clusters.This has already been established by several previousstudies (e.g., De Propris et al. 1999; Andreon 2006; De Pro- Throughout this paper we assume the concordance values forcosmological parameters: Ω M = 0 .
27, Ω Λ = 0 .
73 and H = 73km s − Mpc − c (cid:13)000
73 and H = 73km s − Mpc − c (cid:13)000 , 1–13 R. De Propris et al.
19 20 21 22 23 24 H F160W [mag.] N u m b e r o f G a l a x i e s Figure 2.
The composite background-subtracted H band lumi-nosity function for all galaxies in the cluster fields. The pointswith error bars show the data, the thick dark line the best fit-ting Schechter function (with H ∗ = 21 . α = − .
78) and thethick grey line the local cluster K -band luminosity function fromDe Propris & Christlein (2009), (having M K = − . h and α = − . H − K = 0 . z = 1 .
25. The inset shows the 1 , σ error ellipses on thevalues of M ∗ and α , the parameters of the Schechter function.There is no evidence of evolution (other than purely passive) inthe shape of the luminosity function since z = 1 .
25 over a rangeof ∼
100 in luminosity. pris et al. 2007; Muzzin et al. 2008; Mancone et al. 2010),some reaching to even higher redshifts (Andreon et al. 2014;Wylezalek et al. 2014). Similarly, the value for α is consis-tent with the zero redshift values in clusters as measured byBarkhouse et al. (2007) for 57 Abell clusters in R . Andreon(2008) and Mancone et al. (2012) also found a similar lackof evolution for α since z ∼ α ≈ − I for several clusters at 0 . < z < . z = 1 .
25 downto at least 2.5 magnitudes below the M ∗ point. Consistentwith the results of Andreon et al. (2014) and Wylezalek etal. (2014), it appears that any evolution must essentiallypreserve the shape of the galaxy luminosity function acrossmore than a decade in stellar mass. Any mergers and accre- tion from the field would have to be very finely compensatedby galaxy growth and destruction to produce remarkablyself-similar objects across two thirds of the age of the Uni-verse. In order to obtain the red sequence luminosity function, wemust first consider the colour-magnitude diagrams of theclusters. We generate colour-magnitude relations for galax-ies in the cluster fields in i − z vs. H as shown in Fig. 3,where colours were measured in 0 . (cid:48)(cid:48) apertures and mag-nitudes are total magnitudes. The existence of well-definedred sequences is apparent, though of varying strength be-tween clusters reflecting their overall richness. As expected,their colour-magnitude relations are consistent with passiveevolution of the local relations to the appropriate redshift(cf. Kodama & Arimoto 1997; Mei et al. 2012; Snyder et al.2012). The relations can be followed to the magnitude limitsof our data, with no apparent evidence for a weakening atthe faint end in the best defined cases (e.g. RDCS1252).We select out red sequence galaxies in the followingmanner: we derive the slope and intercept of the red se-quence by minimum absolute deviation (Armstrong & Kung1978) and then take all objects within ± .
25 mag. of theridge line as belonging to the red sequence. We make thesame cuts on the colour-magnitude distributions of fieldgalaxies in the ERS2 field to estimate the contribution fromforeground and background galaxies, normalising to the ar-eas of the observed cluster fields. As with the total luminos-ity functions in Fig. 2, we have then shifted all clusters tothe mean redshift of z = 1 .
25 and we have estimated thenon-Poissonian contributions to the errors.The resulting luminosity function in H for red sequencegalaxies is shown in Fig. 4. This has H ∗ = 21 .
14 and α = − .
71, with error ellipses as shown in the inset. We also showthe local luminosity function (as in Fig. 2), shifted to theredshift of the clusters as described for the previous figure.Even though the local luminosity function is for all galaxies,red sequence galaxies dominate local clusters. Our measured H ∗ translates to M ∗ H = − .
6, again in excellent agreementwith the local value assuming a passively evolving model,implying no significant luminosity (mass) growth. Note thatin the Coma cluster De Propris et al. (1998) measured M ∗ H = − .
58 and α = − .
78 for a sample of 111 members, all ofwhom but one lie on the red sequence. The expected k + e correction of − . M ∗ and α measured for the whole cluster (above) andthe red sequence, implies that these high redshift systemsare already dominated by red, quiescent galaxies at theseearly times. The error bars on individual points are smallerthan those in the full luminosity function (Fig. 2) becausethe background/foreground contamination is lower in thatpart of the colour-magnitude space.The consistency of the faint end of the LF for the redsequence down to H (cid:39)
24, with that seen in nearby clustersargues that the faint end of the red sequence in massive clus-ters, at least down to ∼ M ∗ +3, is already in place at the red-shifts we study. Crawford et al. (2009) and Rembold & Pas-toriza (2012) also find no evidence that α at high redshiftsis significantly different from the present day value, while c (cid:13) , 1–13 isky cluster red sequence galaxies -0.500.51 i F W - z F L P [ m a g . ]
19 20 21 22 23 24 25H
F160W [mag.]-0.500.51 i F W - z F L P [ m a g . ]
19 20 21 22 23 24 25H
F160W [mag.]XMM1229 RDCS1252ISCS1434 XMM2235
Figure 3.
Colour-magnitude diagrams in i − z vs. H for galaxies in all our cluster fields, as identified in the figure legend. Red sequencescan clearly be seen for all clusters. Galaxies for which we derive morphological parameters are marked with filled red circles. Failuresinclude galaxies which are on the edges of the fields, or for which GALFIT only returns highly uncertain parameters. Cerulo et al. (2014) find no weakening or drop-off at fainterluminosities in the red sequence for XMM 1229+0151, unlikeprevious claims (e.g., De Lucia et al. 2007).
There may be several reasons for the discrepancy betweenour results and those from other studies that have claimed tosee evolution in the faint-end slope of red sequence galaxies.One possibility is surface brightness selection effects, as wehave discussed in De Propris et al. (2013). We plot the cen-tral surface brightness vs. total magnitude for red and bluegalaxies in the cluster fields for the H band data in Fig. 1. Itis clear that there is a quite abrupt surface brightness cut-offat µ ( H ) ∼ . − which then generates an everincreasing magnitude incompleteness even if the detectionlimit of the data is much fainter. The onset of this effect liesbetween H = 23 and 24.In addition, we note that red galaxies are more likely tobe affected by this surface brightness limit than blue galax-ies. They generally have lower central surface brightnesses, probably because they lack bright star-forming regions. Inparticular, red galaxies seem to avoid the high concentra-tion regions of this plot at low luminosities, suggesting thatfainter red galaxies are more diffuse and therefore less likelyto be detected. This is a well known characteristic of dwarfsin nearby clusters and is often used to distinguish themfrom background, higher surface brightness, galaxies (e.g.Sandage et al. 1985).To at least the M ∗ +2 . c (cid:13)000
There may be several reasons for the discrepancy betweenour results and those from other studies that have claimed tosee evolution in the faint-end slope of red sequence galaxies.One possibility is surface brightness selection effects, as wehave discussed in De Propris et al. (2013). We plot the cen-tral surface brightness vs. total magnitude for red and bluegalaxies in the cluster fields for the H band data in Fig. 1. Itis clear that there is a quite abrupt surface brightness cut-offat µ ( H ) ∼ . − which then generates an everincreasing magnitude incompleteness even if the detectionlimit of the data is much fainter. The onset of this effect liesbetween H = 23 and 24.In addition, we note that red galaxies are more likely tobe affected by this surface brightness limit than blue galax-ies. They generally have lower central surface brightnesses, probably because they lack bright star-forming regions. Inparticular, red galaxies seem to avoid the high concentra-tion regions of this plot at low luminosities, suggesting thatfainter red galaxies are more diffuse and therefore less likelyto be detected. This is a well known characteristic of dwarfsin nearby clusters and is often used to distinguish themfrom background, higher surface brightness, galaxies (e.g.Sandage et al. 1985).To at least the M ∗ +2 . c (cid:13)000 , 1–13 R. De Propris et al.
19 20 21 22 23 24
H [mag.] N u m b e r o f G a l a x i e s Figure 4.
The composite background-subtracted H band lumi-nosity function for red sequence galaxies in the cluster fields. Thepoints with error bars show the data, the thick dark line the bestfitting Schechter function (with parameters as in the text) andthe thick grey line the local cluster K -band luminosity functionfrom De Propris & Christlein (2009), with H − K = 0 . z = 1 .
25 assuming the passivelyevolving model described in the text. The inset shows the 1 , σ error ellipses on the values of M ∗ and α , the parameters ofthe Schechter function. There is no evidence of evolution (otherthan purely passive) in the shape of the luminosity function since z = 1 .
25 over a range of ∼
100 in luminosity.
Having selected a sample of galaxies representative of thered sequence population at z = 1 .
25 over a wide range ofluminosities, we can now explore their morphological prop-erties and compare these to equivalent properties in a sam-ple of zero redshift red sequence members, in this case takenfrom the Virgo Cluster (Ferrarese et al. 2006). As the sam-ple of high redshift galaxies is chosen based on a simplecolour cut and without spectroscopic confirmation, a frac-tion of this sample will be foreground or background inter-lopers, either with the same intrinsic colour, or scattered intothe sample through photometric uncertainties. Although thecontamination rate will be low (or zero) at bright magni-tudes it increases at fainter levels. We assess this as partof the background subtraction procedure for producing thered sequence luminosity function. The interloper fraction ata given magnitude bin is simply the difference between the total number count and the count used in the luminosityfunction divided by the total number count. Brighter than H = 22 contamination is negligible and does not influencesubsequent results. Between 22 < H <
23 the interloperfraction rises to up to 40% and therefore is taken into ac-count when interpreting results in this magnitude range.In order to quantify the morphology of the red sequencegalaxies we use GALFIT (Peng et al. 2002, 2010) to simul-taneously measure their sizes (effective radii) and intensityprofile shapes (S´ersic indices) by assuming a single S´ersicprofile fit. We interactively carry out fitting on all galaxiesin the H and z band images, using bright stars in the clus-ter fields as a point spread function reference, inspecting theresidual images as an indicator of the quality of the fit. Welimit this process to galaxies brighter than H = 23 .
0. Fainterthan this the radial fits become insufficiently reliable and,as shown earlier, surface brightness incompleteness selectsagainst larger and more diffuse galaxies even though morecompact galaxies are detected to significantly fainter levels.
Fig. 5 (top panel) shows the effective radii of galaxies inour clusters as a function of H -band absolute magnitudefor each galaxy when evolved to z = 0 as described earlier.The equivalent data in Ferrarese et al. (2006) from the ACSVirgo Cluster Survey (VCS) is also plotted as the zero red-shift comparator. At z = 1 .
25, the observed H band moreclosely matches the restframe R band, while the VCS dataare measured in the z band. However, as shown by previousstudies (e.g., ? Tamura & Ohta 2000) and the plotted Fer-rarese et al. data itself, colour gradients for local ellipticalsare small, especially in the redder bands, so we expect thatthere will be little difference between the effective radii in z and R in Virgo.We see little evidence of size evolution in cluster galaxiesacross this redshift range. The high redshift galaxies at M ∗ and fainter have R eff ( H ) ∼ − ∼ z are more compact than those seen lo-cally (e.g.,Whitaker et al. 2012; Cassata et al. 2013; Williamset al. 2014 and references therein), though this work usuallyconcentrates on the more luminous objects. Our above re-sult on the surface brightness vs. luminosity plot (Fig. 1)already suggests that we are not seeing particularly com-pact luminous galaxies in our clusters.However, in other measured parameters there are cleardifferences between the values for the high redshift sam-ple and those for the low redshift comparators. Red se-quence galaxies in our high redshift clusters have, on av-erage, lower S´ersic indices, n , than those in Virgo at the c (cid:13) , 1–13 isky cluster red sequence galaxies l og R e ff ( kp c ) z=1.25 clustersVirgo 246810 R e ff ( kp c ) z=1.25 clustersVirgo0246810 S e r s i c n S e r s i c n -26 -25 -24 -23 -22 -21M H (mag)00.20.40.60.81 A x i s R a ti o ( b / a ) A x i s R a ti o ( b / a ) Figure 5.
Top Panel:
Sizes ( H -band effective radii) of galaxies in our clusters (see legend); Middle Panel:
S´ersic indices (from singleS´ersic fits);
Bottom Panel:
Ellipticities (axis ratios), all plotted vs. M H (evolved to zero redshift) and compared to data in the Virgocluster from Ferrarese et al. (2006). Compare with Fig.6 from Gutierrez et al. (2004): there is no strong evidence for size evolution.However, there is a population of objects with n < M ∗ point. In the equivalent panels on the right hand side of the figure, we show histograms of the distributions in R eff , n and b/a for galaxies in the cluster sample (grey stippled lines) compared to the counterparts (in luminosity) for Virgo (red lines), demonstratingan excess of galaxies with low n and more elliptical axis ratios in the high redshift systems we study. equivalent H − band magnitude (Fig. 5, middle panel) as-suming passive evolution. We plot the distribution of Sersicindices for all cluster galaxies and Virgo in the histogram(middle panel) in Fig. 5. This shows than n is systemati-cally lower by about 1.5 compared to Virgo for our sample.This difference is confirmed by a K-S test that rejects thehypothesis that the two samples are drawn from the sameparent distribution at more than 99 . H − band magnitude of M H = −
21 have n > n < n < < H <
23 ( − . < M H < − . n for the true red sequence galaxies to typically have theS´ersic indices as high as their Virgo counterparts. The num-bers of objects in the plots can be directly compared, asthere are ∼
50 Virgo galaxies to the luminosities we con-sider, and about 140 galaxies in total for the high redshiftclusters.Similarly, we find that the high redshift sample containsmore objects with flattened axis ratios (i.e., more disk-like)than in Virgo (bottom panels of Fig. 5). Again the histogramshows that the high redshift clusters have a long ’tail’ at low b/a and a K-S test also confirms that the two samples havenot been drawn from the same distribution at more thanthe 5 σ level.The lack of objects in Virgo with n < b/a < . c (cid:13)000
50 Virgo galaxies to the luminosities we con-sider, and about 140 galaxies in total for the high redshiftclusters.Similarly, we find that the high redshift sample containsmore objects with flattened axis ratios (i.e., more disk-like)than in Virgo (bottom panels of Fig. 5). Again the histogramshows that the high redshift clusters have a long ’tail’ at low b/a and a K-S test also confirms that the two samples havenot been drawn from the same distribution at more thanthe 5 σ level.The lack of objects in Virgo with n < b/a < . c (cid:13)000 , 1–13 R. De Propris et al. all clusters at z ∼ .
25 compared to Virgo; no cluster isresponsible for the observed results on its own and they allshare the above features (i.e., a more disk-like and flattenedpopulation compared to the local data).An obvious interpretation of this result is that most ofthe high redshift red sequence galaxies are more disk-like orcontain more significant disk components than their low red-shift counterparts (in agreement with similar observationsby Rembold & Pastoriza 2012). Only the brightest galaxieshave S´ersic indices consistent with those of genuine ellip-ticals, and even then they are usually lower than those ofequivalently bright and massive Virgo counterparts. Otherparameters are consistent with this interpretation. Clearly,many of the high redshift galaxies show very significant elon-gation, the entire distribution of axial ratios (Fig, 5 bottompanel) is consistent with expectations from a sample of ran-domly oriented disks (e.g. Lambas et al. 1992).Compared to the Virgo cluster, a population of morehighly elongated objects is present at all luminosities, whilesuch systems can only be seen for the fainter Virgo galaxies(and recall that even Virgo dwarf ellipticals, with n ∼ z > n than their Virgo counterparts.The more highly elliptical galaxies in Fig. 5 have lower n and correspond to the red sequence objects in the lowerleft region of Fig. 6. All this points to a different (disky)morphology for most red sequence galaxies in our high z clusters. Indeed, the range of projected b/a suggests thatthese disks are more like the thin ones seen in modern dayspiral galaxies, than the typically ’thicker’ ones seen in SOs(Laurikainen et al. 2010), the most common disk galaxies inlow z clusters. That is, cluster galaxies at this epoch havebecome ‘red’ without becoming ‘early type’ yet, though theyalready have roughly the same stellar mass as their localcounterparts. The images of the red sequence galaxies bearthis out: Fig 7 shows postage stamps of red sequence galaxies(in the z band) within ∼ M ∗ and with S´ersicindices n <
2, demonstrating the disky nature of the bulkof the population.
Local red sequence galaxies show weak negative (bluer out-wards) colour gradients (Tamura & Ohta 2000); these weakgradients are observed to have the same size to z ∼ .
6, im-plying that they are due to metal abundance gradients (LaBarbera et al. 2003). We measure the colour gradients in z − H (approximately rest-frame B − R ) for our < z > = 1 . z and H bands,∆ log R eff = log( R eff ( z ) /R eff ( H )). The reason we do thisis that, for these faint and small objects, we cannot easily Sersic Index n A x i s r a ti o ( b / a )
S´ersic indices of galaxies in our clusters vs. axis ratio,compared to data in the Virgo cluster from Ferrarese et al. (2006).We see that at z ∼ .
25 clusters contain a population of low n galaxies with flattened shapes, thus resembling disks. derive a radial luminosity profile which is accurate enoughto derive colour gradients directly as in the lower redshiftstudies of Vader et al. (1988) or Tamura & Ohta (2000). Forthese high redshift systems, sky subtraction will be uncer-tain, the differences in the point spread functions betweenACS and WFC3 are difficult to model and any small mis-alignment between the blue and red image may result inspurious results for directly determined colour profiles (e.g.,from ellipse ). Our approach minimises these issues, as itsimply compares the effective radii derived from each imageand measures the colour gradient via their ratio. The cen-tering of the profile, the effects of the PSF and sky level aremodelled for each galaxy via GALFIT and are, in principle,accounted for.As shown in Fig. 8, the high redshift galaxies are foundto have significantly positive ∆ log R eff (i.e., a larger R eff in the bluer band), hence negative colour gradients in theusual sense in rest-frame B − R (observed z − H ). These aremuch larger than those observed for Virgo galaxies (typi-cally consistent with little or no gradient when measured c (cid:13) , 1–13 isky cluster red sequence galaxies Figure 7.
HST/ACS 2 × z − band images for 50 red sequence galaxies with n < ∼ M ∗ drawn from all four clusters. Each image greyscale stretches between the mean sky level to the square-root of the central surfacebrightness (so more diffuse, lower surface brightness galaxies appear to have noisier images). in g − z ). Here we find that the high redshift n < R eff than foundlocally (0.1 to 0.4 as opposed to < .
1) while the high red-shift spheroids (large n ) have ∆ log R eff of 0 . .
5. Usingequation (4) from La Barbera et al. (2002), we find that amedian ∆ log R eff of 0.13 corresponds to a colour gradientof − .
25 mag. in B − R per decade in radius. Local systemsin La Barbera et al. (2002) have typical gradients of –0.1to –0.3 mag., for comparison. Similarly significant negativecolour gradients in a substantial fraction of early type galax-ies at 1 . < z < .
9, albeit among field galaxies, have beenmeasured by Gargiulo et al. (2012).This argues that these galaxies have significantly bluerexteriors than their local counterparts, while still being over-all red in colour - star formation has likely ended through-out, but more recently in the outer parts. Colour gradientsin local red sequence galaxies are commonly attributed to agradient in mean metal abundance with radius (e.g., Fosteret al. 2009). The sizes of the colour gradients we measure aremore consistent with age gradients, as they are too large tobe produced by metallicity alone (Saglia et al. 2000; La Bar-bera et al. 2003), with higher redshift galaxies being youngerat larger radii. Star formation may therefore have continuedfor longer periods in the outer parts of these galaxies, beforethey were quenched on to red sequence. (Of course, their de-scendants will also be younger in the outer parts, but thecolour differential decays rapidly with time). The large scat-ter in the colour gradients in Fig 8 suggests that cessation ofstar formation may have taken place over a range of timesfor each galaxy. This is similar to the findings of Jørgensenet al. (2005), where fainter cluster galaxies in a z = 0 . z ∼ . ∼ z = 0 . z = 1 . z = 1 . z = 3 (the earliest that star formationis likely to end in any disk), it is unsurprising that gradientsof the observed strength are only seen at the higher redshiftsstudied in this work.Given that the galaxies we are studying are red andhave ceased significant star formation at an earlier epoch,the clear implication is that colour (stellar population) evo-lution precedes morphological evolution (Skibba et al. 2009;Kovac et al. 2010) and that most of these red sequence galax-ies are galaxies with significant disk components that fade orevolve secularly into bulge-dominated red (quiescent) galax-ies in the present universe (e.g., Jaffe et al. 2011). This is notthe first time such an evolutionary scenario has been sug-gested to explain the properties of high redshift populations.Bundy et al. (2010) proposed a similar scheme for the evolu-tion of red sequence galaxies at 1 < z < after star formation is sup-pressed, as disks fade and bulge components become moresignificant (in a relative if not absolute sense). Although it isa field sample, several red disks are also observed by Bruceet al. (2012) in the CANDELS survey, albeit at somewhathigher redshifts.Given that only a minority of the bright galaxies evenin Virgo are true ellipticals, the majority being S0s (Fer-rarese et al. 2006), it is likely that most of these “red disks” c (cid:13)000
9, albeit among field galaxies, have beenmeasured by Gargiulo et al. (2012).This argues that these galaxies have significantly bluerexteriors than their local counterparts, while still being over-all red in colour - star formation has likely ended through-out, but more recently in the outer parts. Colour gradientsin local red sequence galaxies are commonly attributed to agradient in mean metal abundance with radius (e.g., Fosteret al. 2009). The sizes of the colour gradients we measure aremore consistent with age gradients, as they are too large tobe produced by metallicity alone (Saglia et al. 2000; La Bar-bera et al. 2003), with higher redshift galaxies being youngerat larger radii. Star formation may therefore have continuedfor longer periods in the outer parts of these galaxies, beforethey were quenched on to red sequence. (Of course, their de-scendants will also be younger in the outer parts, but thecolour differential decays rapidly with time). The large scat-ter in the colour gradients in Fig 8 suggests that cessation ofstar formation may have taken place over a range of timesfor each galaxy. This is similar to the findings of Jørgensenet al. (2005), where fainter cluster galaxies in a z = 0 . z ∼ . ∼ z = 0 . z = 1 . z = 1 . z = 3 (the earliest that star formationis likely to end in any disk), it is unsurprising that gradientsof the observed strength are only seen at the higher redshiftsstudied in this work.Given that the galaxies we are studying are red andhave ceased significant star formation at an earlier epoch,the clear implication is that colour (stellar population) evo-lution precedes morphological evolution (Skibba et al. 2009;Kovac et al. 2010) and that most of these red sequence galax-ies are galaxies with significant disk components that fade orevolve secularly into bulge-dominated red (quiescent) galax-ies in the present universe (e.g., Jaffe et al. 2011). This is notthe first time such an evolutionary scenario has been sug-gested to explain the properties of high redshift populations.Bundy et al. (2010) proposed a similar scheme for the evolu-tion of red sequence galaxies at 1 < z < after star formation is sup-pressed, as disks fade and bulge components become moresignificant (in a relative if not absolute sense). Although it isa field sample, several red disks are also observed by Bruceet al. (2012) in the CANDELS survey, albeit at somewhathigher redshifts.Given that only a minority of the bright galaxies evenin Virgo are true ellipticals, the majority being S0s (Fer-rarese et al. 2006), it is likely that most of these “red disks” c (cid:13)000 , 1–13 R. De Propris et al. -26 -25 -24 -23 -22 -21 -20M H [mag.]-0.200.20.40.6 ∆ l og R e ff
Colour gradients of galaxies in our clusters (see legend) vs. absolute H magnitude (left panel) and S´ersic index (right panel)compared to data in the Virgo cluster from Ferrarese et al. (2006). While nearly all Virgo galaxies have small colour gradients, irrespectiveof luminosity or Sersic index, our galaxies have both larger gradients and a wider spread, indicating the existence of age gradients witha large spread in ages for the outer regions of galaxies. will evolve into classical lenticulars rather than true ellip-ticals. While these high redshift galaxies could themselvesbe labelled as lenticulars (as we see no evidence of spiralarms), they are significantly diskier (in terms of their S´ersicindices, range of axis ratios and colour gradients) than clas-sical zero-redshift cluster S0s. Early on, Michard (1994) pro-posed that all but the brightest ellipticals can be observedto host disks and should be classified as S0s. With the ex-ception of very luminous ellipticals, most early-type galaxiesare found to contain a rotating stellar disk (Emsellem et al.2011), with several S0s found to be fast rotators (Emsellemet al. 2007), and, by implication, therefore disky. In terms ofthe classification used by Bundy et al. (2010), based on theirS´ersic indices, our sample of cluster red sequence galaxies aremainly “early disks” with true ellipticals only dominating atthe high mass end. The stellar mass at which the dominantpopulation changes is similar in the clusters studied hereand in the highest redshift subsample of Bundy et al. (2010)(e.g. see their figure 2).While the red disky systems discussed here appear rel- atively extreme in their properties (S´ersic indices, range ofaxis ratios and colour gradients), comparable red disk galax-ies have been identified at lower redshifts. Passive red disks,with little or no spiral structure, have been found to be a no-table sub-population in the red sequence of the Abell 901/2supercluster at z = 0 .
165 (Wolf et al. 2009) Balogh et al.(2009) finds dust-reddened spirals with early-type morphol-ogy (S0/a) on the red sequence in groups, although theseobjects tend to be star-forming rather than passive. About10% of red sequence galaxies are observed to have infraredcolours indicative of active or recent star formation in thePRIMUS sample (Zhu et al. 2011). On the other hand, Mas-ters et al. (2010) identify truly passive spirals in their sampleand show that they tend to have large masses, likely simi-lar to our targets. These red disks have similar, or slightlyyounger, ages than ellipticals (Robaina et al. 2012; Tojeiroet al. 2013), which is likely the case for our sample as well,given our interpretation of the observed colour gradients asage gradients.Passive spirals are known in nearby clusters as well (van c (cid:13) , 1–13 isky cluster red sequence galaxies den Bergh 1976; Koopmann & Kenney 1998). Their fre-quency may even increase in the higher redshift MORPHSsample (Poggianti et al. 1999), where they are believed togradually replace S0s. These passive red spirals also tend tooccur more commonly at intermediate densities (Bamfordet al. 2009; Masters et al. 2010) and although the numbersare small, many of our objects seem to lie somewhat out-side of the cluster core. Similarly, Ferr´e-Mateu et al. (2014)show that in CL0152-13 at z = 0 .
83 a small population ofcluster members lying outside of the two main subclustershas younger ages than the apparently old ellipticals in thecluster core.These red disks may originate from quenched diskgalaxies, which are abundant in local groups (Feldmann etal. 2011). Carollo et al. (2013, 2014) find that quenchingand fading of disks may be responsible for the apparentevolution of the typical size of the early type galaxy pop-ulation as a function of time. At higher redshift, Fontana etal. (2004) and Abraham et al. (2007) claim that at z > z = 1 in the COSMOSfield, but vanish by z = 0 . ∼ − z ∼ . z = 1 .
25 appear to dis-appear on similar timescales. Whether this occurs by diskfading, removal or secular evolution is an issue that can beaddressed with better and more data on clusters above z = 1than present in the archive currently. Such a dataset wouldalso allow us to considerably improve our estimate for theevolution of the luminosity function parameters. In this work we have demonstrated that while the stellarpopulations of typical z ∼ .
25 cluster red sequence galax-ies appear to have already reached a state of only passiveevolution, their morphologies must clearly evolve over thesubsequent ∼ . • Red sequence cluster galaxies in four < z > = 1 .
25 clus-ters appear to have formed their stars and assembled theirmass completely by this redshift, at least down to 3 magni-tudes below the M ∗ point, evolving passively afterwards. • Unlike field galaxies, these red sequence galaxies do notshow size evolution when compared to zero redshift clustergalaxies. • However, there is clear evidence that these objects mustcontinue to evolve morphologically at these redshifts. Apart from the most massive of the high redshift galaxies, thesesystems appear to have lower S´ersic indices than those of asimilar mass in low redshift clusters, projected axis ratiosextending to lower (b/a) than local counterparts, and clearnegative colour gradients much larger than those encoun-tered locally. These colour gradients are so large that theycan only really be attributed to age gradients in the stellarpopulations, with the galaxies being younger outwards. • Taken together all of these observations imply that thebulk of the red sequence galaxies in massive z ∼ .
25 clustersare galaxies with significant disk components (diskier thanpresent-day lenticulars) unlike those identified in the denseregions of low redshift clusters which have the characteristicsof classical ellipticals and S0s. • Given the clear difference in the S´ersic indices andcolour gradients in the two epochs, a clear prediction is thatthe disks must reduce in prominence over time through age-ing of their stellar populations and/or their secular evolutioninto bulges. The present results do not suffice to determinewhether this is due to disk fading (implied by the observa-tion that the effective radii do not change), bulge luminos-ity growth (suggested by the increase in the S´ersic index) orboth.
ACKNOWLEDGMENTS
We would like to thank Inger Jørgensen and Anna Cibinelfor useful discussions. This work is based on observationsmade with the NASA/ESA Hubble Space Telescope, andobtained from the Hubble Legacy Archive, which is a col-laboration between the Space Telescope Science Institute(STScI/NASA), the Space Telescope European Coordinat-ing Facility (ST-ECF/ESA) and the Canadian AstronomyData Centre (CADC/NRC/CSA). Some of the data pre-sented in this paper were obtained from the Mikulski Archivefor Space Telescopes (MAST). STScI is operated by the As-sociation of Universities for Research in Astronomy, Inc.,under NASA contract NAS5-26555. The PIs of the originalprojects which produced these data are thanked for provid-ing excellent deep archival images with diverse uses beyondthe original programmes. We also thank the anonymous ref-eree for a helpful report that helped to clarify a number ofissues.
REFERENCES
Abraham R. G. et al. 2007, ApJ, 669, 184Alberts S. et al. 2014, MNRAS, 437, 437Andreon S. 2006, A&A, 448, 447Andreon S. 2008, MNRAS, 386, 1045Andreon S., Newman A. B., Trinchieri G., Raichoor A.,Ellis R. S., Treu T. 2014, A&A, 565, 120Armstrong R. D., Kung M. T. 1978, Applied Statistics, 27,363Balogh M. et al. 2009, MNRAS, 398, 754Bamford S. et al. 2009, MNRAS, 393, 1324Barkhouse W. A., Yee H. K. C., Lopez-Cruz O. 2007, ApJ,671, 1471Bertin E., Arnouts S. 1996, A&AS, 117, 393Binggeli B., Popescu C.C. 1995, A&A, 298, 63 c (cid:13)000
Abraham R. G. et al. 2007, ApJ, 669, 184Alberts S. et al. 2014, MNRAS, 437, 437Andreon S. 2006, A&A, 448, 447Andreon S. 2008, MNRAS, 386, 1045Andreon S., Newman A. B., Trinchieri G., Raichoor A.,Ellis R. S., Treu T. 2014, A&A, 565, 120Armstrong R. D., Kung M. T. 1978, Applied Statistics, 27,363Balogh M. et al. 2009, MNRAS, 398, 754Bamford S. et al. 2009, MNRAS, 393, 1324Barkhouse W. A., Yee H. K. C., Lopez-Cruz O. 2007, ApJ,671, 1471Bertin E., Arnouts S. 1996, A&AS, 117, 393Binggeli B., Popescu C.C. 1995, A&A, 298, 63 c (cid:13)000 , 1–13 R. De Propris et al.
Blakeslee J. et al. 2003, ApJ, 596, L143Brodwin M. et al. 2013, ApJ, 779, 138Bruce V. A. et al. 2012, MNRAS, 427, 1666Bruzual G., Charlot S. 2003, MNRAS, 344, 1000Bundy K. et al. 2010, ApJ, 719, 1969Butcher H. & Oemler A. 1978, ApJ, 226, 559Butcher H. & Oemler A. 1984, ApJ, 285, 426Carollo C. M. et al. 2013, ApJ, 773, 112Carollo C. M. et al. 2014, arXiv 1402.1172Cassata P. et al. 2013, ApJ, 775, 106Cerulo P., Couch W. J., Lidman C., Delaye L., Demarco R.,Huertas-Company M., Mei S., Sanchez-Janssen R. 2014,MNRAS, 439, 2790Chang Y.-Y., van der Wel A., Rix H.-W., Wuyts S., ZibettiS., Ramkumar B., Holden B. 2013, ApJ, 762, 83Cibinel A. et al. 2013, ApJ, 777, 116Cimatti A. et al. 2008, A&A, 482, 21Colless M. 1989, MNRAS, 237, 799Crawford S. M., Bershady M. A., Hoessel J. G. 2009, ApJ,690, 1158Damjanov I. et al. 2009, ApJ, 695, 101Delaye L. et al. 2014, MNRAS, 441, 203De Lucia G. et al. 2007, MNRAS, 374, 809den Brok M. et al. 2011, MNRAS, 414, 3052De Propris R., Eisenhardt P. R., Stanford S. A., DickinsonM. 1998, ApJ, 503, L45De Propris R., Stanford S. A., Eisenhardt P. R., DickinsonM., Elston R. 1999, AJ, 118, 719De Propris R. et al. 2003a, MNRAS, 342, 725De Propris R., Stanford S. A., Eisenhardt P. R., DickinsonM. 2003b, ApJ, 598, 20De Propris R., Stanford S. A., Eisenhardt P. R., Holden B.P., Rosati P. 2007, AJ, 133, 2209De Propris R., Christlein D. 2009, Astr. Nach. 330, 943De Propris R., Phillipps S., Bremer M. N. 2013, MNRAS,434, 3469Disney M. 1976, Nature, 263, 573Dressler A., 1984, ARA&A, 22, 185Driver S. P., Odewahn S. C., Echevarria L., Cohen S. H.,Windhorst R. A., Phillipps, S., Couch W. J. 2003, AJ,126, 2662Emsellem E. et al. 2007, MNRAS, 379, 401Emsellem E. et al. 2011, MNRAS, 414, 888Fasano G., Poggianti B. M., Couch W. J., Bettoni D., Kjær-gaard P., Moles M. 2000, ApJ, 542, 673Fassbender R. et al. 2014, A&A, 568, 5Feldmann R., Carollo C. M., Mayer L. 2011, ApJ, 736, 88Ferrarese L. et al. 2006, ApJS, 164, 334Ferr´e-Mateu A., S´anchez-Blasquez P., Vazdekis A., de laRosa I. G. 2014, ApJ, 797, 136Fontana A. et al. 2004, A&A, 424, 23Gargiulo A., Saracco P., Longhetti M., La Barbera F., Tan-burri S. 2012, MNRAS, 425, 2698Giavalisco M. et al. 2004, ApJ, 600, L93´Gutierrez C. M., Trujillo I., Aguerri J. A. L., Graham A.W., Caon N. 2004, ApJ, 662, 604Huang J.-S., Cowie L. L., Gardner J. P., Hu E. M., SongailaA., Wainscoat R. J. 1997, ApJ, 476, 12Jaffe Y. et al. 2011, MNRAS, 417, 1996Jørgensen I., Chiboucas K., Toft S., Bergmann M., ZirmA., Schivon R.P., Gr¨utzbauch R. ArXiv:1408.2906Jørgensen I. & Chiboucas K., 2013, AJ, 145, 77 Jørgensen I., Bergmann M., Davies R., Barr J., TakamiyaM., Crampton D. 2005, AJ, 129, 1249Kodama T., Arimoto N. 1997, A&A, 320, 41Koopmann R. A., Kenney J. D. P. 1998, ApJ, 497, L75Kovac K. et al. 2010, ApJ, 708, 505Koyama Y., Kodama T., Tadaki K.-I., Hayashi M., TanakaI., Shimakawa R. 2014, ApJ, 789, 18La Barbera F., Busarello G., Merluzzi P., Massarotti M.,Capaccioli M.. 2002, ApJ. 571, 790La Barbera F., Busarello G., Massarotti M., Merluzzi P.,Mercurio A. 2003, A&A, 409, 21Lambas D. G., Maddox S. J., Loveday J. 1992, MNRAS,258, 404Laurikainen E., Salo H., Buta R., Knapen J.H., ComeronS., 2010, MNRAS, 405, 1089Longhetti M. et al. 2007, MNRAS, 374, 614Lopez-Cruz O., Barkhouse W., Yee H. K. C. 2004, ApJ,614. 679Mancone C. L., Gonzalez A. H., Brodwin M., Stanford S.A., Eisenhardt P. R., Stern D., Jones C. 2010, ApJ, 720,284Mancone C. L. et al. 2012, ApJ, 761, 141Martinet N. et al. 2014, arXiv, 1412.5821Masters K. L. et al. 2010, MNRAS, 405, 783Mei S. et al. 2006a, ApJ, 639, 81Mei S. et al. 2006b, ApJ, 644, 759Mei S. et al. 2009, ApJ, 690, 42Mei S. et al. 2012, ApJ, 754, 141Mei S. et al. 2014, astro-ph, 1403.7524Michard R. 1994, A&A, 288, 401Muzzin A., WIlson G., Lacy M., Yee H. K. C., Stanford S.A. 2008, ApJ, 686, 966Oesch P. A. et al. 2010, ApJ, 714, L47Pipino A., Matteucci F. 2004, MNRAS, 347, 968Peng C. Y., Ho L. C., Impey C. D., Rix H.-W. 2002, AJ,124, 266Peng C. Y., Ho L. C., Impey C. D., Rix H.-W. 2010, AJ,139, 2097Phillipps S., Driver S. 1995, MNRAS, 274, 832Pimbblet K. A., Jensen P. C. 2012, MNRAS, 426, 1632Poggianti B. M. et al. 1999, ApJ, 518, 576Poggianti B. M. et al. 2013, ApJ, 762, 77Rembold S. B., Pastoriza M. G. 2012, MNRAS, 422, 719Robaina A. R., Hoyle B., Gallazzi A., Jimenez R., van derWel A., Verde L. 2012, MNRAS, 427, 3006Rudnick G. H., Tran K.-V., Papovich C., Momcheva I.,Willmer C. 2012, ApJ, 755, 14Saglia R. P., Maraston C., Greggio L., Bender R., ZieglerB. 2000, A&A, 360, 911Sandage A., Binggeli B., Tammann G. A. 1985, AJ, 90,1681Shen S., Mo H. J., White S. D. M., Blanton M. R., Kauff-mann G., Voges W., Brinkmann J., Csabai I. 2003, MN-RAS, 343, 978Skibba R. et al. 2009, MNRAS, 399, 966Snyder G. F. et al. 2012, ApJ, 756, 114Strazzullo V. et al. 2010, A&A, 524, 17Tamura N., Ohta K. 2000, AJ, 120, 533Tamura N., Ohta K. 2003, AJ, 126, 596Tojeiro R. et al. 2013, MNRAS, 432, 359Tran K.-V. H. et al. 2010, ApJ, 719, L126 c (cid:13) , 1–13 isky cluster red sequence galaxies Vader J. P., Vigroux L., Lachieze-Rey M., Souviron J. 1988,A&A, 203, 217van den Bergh S. 1976, ApJ, 206, 883Vulcani B. et al. 2013, A&A, 550, 58Vulcani B., Poggianti B. M., Fritz J., Fasano G., MorettiA., Calvi R., Paccagnella A. 2014, arXiv, 1410.6481Whitaker K. E. et al. 2012, ApJ, 745, 179Williams C. et al. 2014, ApJ, 780, 1Wolf C. et al. 2009, MNRAS, 393, 1302Wylezalek D. et al. 2014, ApJ, 786, 17Zeimann G. R. et al. 2013, ApJ, 779, 137Zhu G. et al. 2011, ApJ, 726, 110This paper has been typeset from a TEX/ L A TEX file preparedby the author. c (cid:13)000