ZENS. IV. SIMILAR MORPHOLOGICAL CHANGES ASSOCIATED WITH MASS QUENCHING AND ENVIRONMENT QUENCHING AND THE RELATIVE IMPORTANCE OF BULGE GROWTH VERSUS THE FADING OF DISKS*

In computing different environmental parameters, which in principle should probe the surroundings of galaxies on different scales, several factors can introduce cross talk among such parameters. This complicates attempts to identify the physical mechanisms behind environmentally driven galaxy evolution. We discuss below some of the most important of these factors that are relevant for our analysis, and we summarize our approaches to minimizing their impact on the results.

2.1.1. Identification of Centrals and Satellites and Definition of Relaxed and Unrelaxed Groups

2.1.2. An LSS Density Estimate Unaffected by Group Richness

2.1.3. Disentangling Halo Mass and LSS Density Effects

To separate galaxies into quenched and (moderately or strongly) star-forming systems, we adopted a combination of several indicators. Quenched galaxies are required to satisfy all of the following spectral and color criteria: (1) no detected emission in Hα and Hβ and (2) $(\mathrm{NUV}-I)\gt 4.8$, $(\mathrm{NUV}-B)\gt 3.5$, and $(B-I)\gt 1.2$.

The final assignment of a galaxy to the quenched population was based on the sSFR derived from an SED fit to its photometric data (see Section 2.3), validated, however, through the spectral and color requirements listed above. This multiple-constraint approach to the identification of quenched galaxies leads to a higher purity of the sample compared to the use of a single (often optical) color criterion; samples based on the latter may suffer from a nonnegligible contamination from, for example, dust-reddened star-forming galaxies, especially at masses $\lesssim {10}^{11}\;{M}_{\odot }$ (see Paper III and Woo et al. 2013).

In discussing our results in Section 5, we will compare the bulge and disk properties of the quenched satellites with those of star-forming satellites of similar stellar mass in the ZENS sample. In our work we define galaxies to be strongly star-forming if they show (1) strong [O ii], [N ii], Hα, and Hβ line emission and (2) $(\mathrm{NUV}-I)\lt 3.2$, $(\mathrm{NUV}-B)\lt 2.3$, and $(B-I)\lt 1.0$ colors. Galaxies with color and spectral properties between strongly star-forming and quenched galaxies are classified as moderately star-forming objects. Further details are given in Paper III.

Galaxy stellar masses and SFRs were derived in Paper III by fitting model SEDs to the photometric data of the ZENS galaxies using the Zurich Extragalactic Bayesian Redshift Analyzer+ (ZEBRA+; Oesch et al. 2010). ZEBRA+ is an upgraded version of our publicly released ZEBRA code (Feldmann et al. 2006). Stellar population models were adopted from the Bruzual & Charlot (2003) library with a Chabrier (2003a) initial mass function. We used two types of star-formation histories: exponentially decaying models with a broad range of characteristic τ timescales (from very short to very long e-folding τ values) and constant star-formation models. Details on the grid of SED models are given in Table 2 of Paper III.

All SED fits were visually inspected to ensure consistency between the spectral+color analysis and the SFR estimates derived from the fits. For 2% of ZENS galaxies, a second iteration in the fits was required to resolve initial discrepancies between the SED-based SFRs and the spectral+color assessments. Specifically, the original SED fits to this small fraction of galaxies with clearly star-forming spectra and colors returned very low SFRs. In these few cases, we rerun the SED fits after restricting the sample of templates to star-forming SED models only.

We use as our definition of stellar mass the integral of the past SFR. Our stellar masses are thus about 0.2 dex higher than the "actual" stellar mass that is sometimes used, that is, the stellar mass that remains after mass return to the interstellar medium from the evolving stellar population.

With our definition, the stellar mass completeness of the ZENS galaxy sample for quenched systems is ${M}_{{\rm{galaxy}}}\sim {10}^{10}\;{M}_{\odot }$.

A strength of ZENS is its correction for measurement biases in the structural parameters. These were calibrated to eliminate spurious trends with the size of the seeing-driven point-spread function (PSF) and with galaxy magnitude, size, concentration, and ellipticity. A quantitative structural–morphological classification was then implemented on the basis of these corrected structural parameters. The classification was mainly based on the light-defined bulge-to-total (B/T) ratio, which was measured via two-dimensional bulge+disk fits to the galaxy surface brightness distributions, reinforced by imposing quantitative boundaries for the different morphological types in standard nonparametric quantities such as concentration, Gini, M₂₀, and smoothness parameters (see Table 2 and Figure 18 of Paper II for details). Bulges were modeled with a Sérsic profile and disks with a single exponential profile (see Paper II for details).

We note that in the literature the word "morphology" is typically associated with visual (or visual-like) classifications reflecting the combination of galaxy structure and star-formation properties and involving no quantitative estimate of the contribution of the disk and bulge components (see, e.g., Conselice 2003; Lotz et al. 2004; van der Wel 2008; Nair & Abraham 2010; Kartaltepe et al. 2015 or the Galaxy Zoo project, Lintott et al. 2011). From a physical perspective, however, it is the B/T ratio that carries most of the meaning of a "morphological class." In our ZENS work, we have therefore chosen to avoid the subjectivity of a visual morphological classification and adopted a physically meaningful definition of "morphology" that is primarily based on the B/T ratio. We furthermore make use of nonparametric quantities to facilitate the separation between neighboring morphological classes (e.g., S0 and early-type spiral galaxies). Consistent with our previous work and adopted philosophy, in the following we will call our classification "morphological". The reader should keep in mind that other authors, including the quoted work above, would, however, define our classification as a "structural" one. Furthermore, in our discussion below, we will distinguish between "morphology" and "structure" to indicate respectively light-weighted and mass-weighted B/T values.

For the present analysis, we divide galaxies into three morphological bins, based on the basic ZENS morphological classes described in Paper II. These three bins are here referred to as the early-type, disk-dominated, and bulgeless disk galaxies. With reference to the morphological classes defined in Paper II, the early-type bin includes both elliptical galaxies, that is, galaxies described by one single component with Sérsic index $n\gt 3$, and bulge-dominated disks with B/T $\geqslant \;0.5$. The disk-dominated bin is composed of intermediate-type disks with $0.2\;\leqslant$ B/T < 0.5, that is, of disk galaxies with a substantial but not dominant bulge component, and the bulgeless bin consists of late-type disks with B/T $\lt \;0.2$.

We note here that, at the galaxy mass scales of this study, bulgeless disks contribute negligibly to the morphological mix of the quenched galaxy population. Therefore, in the following we will simplify the presentation of our results by reporting only the fractional weight ${f}_{{Q}_{\mathrm{ETG}}}$ of the quenched early-type satellite population, relative to the whole quenched satellite population of similar mass. Given the paucity of bulgeless disks within the ${M}_{{\rm{galaxy}}}\geqslant {10}^{10}\;{M}_{\odot }$ quenched satellite population, a value of $(1-{f}_{{Q}_{\mathrm{ETG}}})$ gives, in practice, the contribution of the aforementioned disk-dominated galaxies to the total quenched satellite population. Ellipticals represent instead between 5% and 15% of the whole early-type population in the two mass bins considered in the following.

The completeness mass limit of ${10}^{10}\;{M}_{\odot }$ for quenched galaxies motivates our adoption of this value as the lower galaxy mass limit in this study. We thus conduct our analyses in two well-defined bins of galaxy stellar mass: a low galaxy stellar mass bin defined by ${10}^{10}{M}_{\odot }\leqslant {M}_{{\rm{galaxy}}}\leqslant {10}^{10.7}\;{M}_{\odot }$ and a high-mass bin, ${10}^{10.7}{M}_{\odot }\leqslant {M}_{{\rm{galaxy}}}\leqslant {10}^{11.5}\;{M}_{\odot }$. Where necessary, we will refer to these mass bins with the abbreviated notation $\mathrm{log}M$ ∈ $[10,10.7[$ and $\mathrm{log}M$ ∈$[10.7,11.5]$. The median galaxy mass values of our samples in these two mass bins are respectively ${10}^{10.32}$ ${M}_{\odot }$ and ${10}^{10.85}$ ${M}_{\odot }$. The explored range of galaxy stellar mass straddles the characteristic (and seemingly invariant) value of $M\ast$ of the Schechter (1976) function that describes both the quenched and star-forming galaxy populations. As discussed in Peng et al. (2010), $M\ast$ emerges as the characteristic mass in the quenching of galaxies. With our convention for defining stellar masses, $M*\;\sim \;{10}^{10.85}\;{M}_{\odot }$, which matches the median value of our galaxy sample in the high galaxy stellar mass bin.

As shown in the Appendix, the fraction of quenched satellites and the morphological mix of these quenched satellites remain practically unchanged, in any environment, with the inclusion or exclusion of the unrelaxed groups in the computations. To maximize the size of the different galaxy samples, we will therefore use the entire ZENS group sample, independent of the group dynamical state, in studying the dependence on the different environments of the quenched satellite fraction and its morphological composition (Sections 3 and 4).

Our analysis below includes estimates for the fraction f_Q of quenched satellites and the fraction ${f}_{Q,\mathrm{ETG}}$ of quenched satellites that have an early-type morphology. In order to include the impact of interloper galaxies on these fractions, we used our estimated average ∼20%–40% contamination by interlopers in the selection of group galaxy members (Paper I). Specifically, we generated 100 sample realizations for each mass and environmental bin in our study, in each of which we rejected a random 20% up to 40% of the galaxies before recalculating the quenched (and quenched early-type) fractions. We then measured the dispersion around the median fractions of all these realizations as an estimate for the impact of interlopers on these quantities. In the following, the quoted uncertainties on the measured fractions include both binomial statistics errors (Gehrels 1986) and these systematic contributions to the errors, which have been added in quadrature.

Figures 1(a) and (b) show the results of Table 1 for f_Q in a graphic form. Specifically, the upper panels of Figure 1(a) show the quenched fraction f_Q of our lower-mass $\mathrm{log}M$ ∈ $[10,10.7[$ satellite galaxies as a function of group halo mass M_Halo (upper right) and normalized halo-centric radius $R/{R}_{\mathrm{vir}}$ (upper left). The lower panels show the same results, this time split into the three bins of halo-centric radius (lower left plot) and the three bins of halo mass (lower right plot) that are listed above and in Table 1 (and illustrated by the dotted lines in Figure 10). Figure 1(b) shows the same for the higher mass satellite galaxies ($\mathrm{log}M$ ∈$[10.7,11.5]$), but now, as discussed above and in the Appendix, we split the sample in only two bins of halo mass in the lower left panel. In both figures, we plot the quenched fractions in the corresponding environmental bins, that is, the values reported in Table 1, at the median values of the environmental parameters within the bins in question. The black dashed and dotted lines in all panels of both figures, and the red horizontal lines and hatched areas in their top-right panels, are discussed in Sections 5.1 and 5.4, respectively, where we compare the observed trends for f_Q with classical SAMs and with the Peng et al. (2010) predictions for mass and environment quenching (see caption of Figure 1(a)).

Zoom In Zoom Out Reset image size

The significance of the differences in f_Q and ${f}_{{Q}_{\mathrm{ETG}}}$ measured across the M_Halo–R parameter space is given in Table 3. Specifically, we provide the p value (probability of obtaining the observed difference from the same parent sample) for a z test performed on the fractions measured in the lowest and highest environmental bins, for any given environmental indicator and galaxy sample considered in Figures 1(a)–2(b).

At the galaxy mass scales that we are probing, we find no significant dependence on halo mass of the quenched satellite fraction f_Q when this is integrated over all halo-centric distances: about 50% and 70% of $\mathrm{log}M$ ∈ $[10,10.7[$ and $\mathrm{log}M$ ∈$[10.7,11.5]$ satellite galaxies, respectively, are quenched systems at all halo masses. A quantitative z test indicates consistency between the quenched fraction in small and massive groups for both low- and high-mass galaxies (p value greater than 60%); bootstrapping simulations give a $1\sigma$ upper-limit slope $\frac{{{df}}_{Q}}{d\mathrm{log}M}\leqslant 0.1$ for the f_Q versus M_Halo relation across the range covered by our data. In Paper I, we estimated a typical uncertainty of about 0.3 dex for the halo masses, which in general would tend to weaken any actual dependence of parameters on halo mass, even though the width of the halo mass bins in our analysis is larger than this typical uncertainty. Mock simulations, discussed in Paper I, indicate that the observed flat relation between f_Q and M_Halo does imply a negligible intrinsic slope for this relationship. We note that, in their SDSS analysis, Peng et al. (2012) find that the fraction of quenched satellites is invariant with halo mass at fixed local overdensity, while Woo et al. (2013) find a weak positive correlation between their satellite red fraction and halo mass.

Consistent with other independent analyses, our sample shows instead a significant environmental dependence of f_Q on halo-centric distance R: the quenched fraction, averaged over all halo masses, decreases with increasing R (see upper right plots of Figures 1(a) and (b)). This is also seen in the SDSS analysis of Woo et al. (2013), and, indirectly, of Peng et al. (2012) if their local projected density ρ is interpreted as a proxy for the halo-centric radius; other studies have also found a similar trend (e.g., Rasmussen et al. 2012; George et al. 2013). In our data, this radial trend is seen mostly, and most strongly, at the highest halo masses. When averaged over all halo mass scales, the fraction f_Q of low-mass quenched satellites in our sample declines from $\gtrsim 60\%$ in the cores of groups down to $\lesssim 50\%$ at the virial radius, whereas for the high-mass satellites it decreases from $\gtrsim 80\%$ down to $\lesssim 60\%$. The total quenched fraction keeps decreasing at radii larger than R_vir in the high galaxy mass bin where our $R\gt {R}_{\mathrm{vir}}$ measurement gives ${f}_{Q}\sim 50\%$. In the low-mass bin, the quenched satellite fraction flattens outside the virial radius; for both galaxy mass bins, however, a two-sided z test shows a $\lt 1\%$ probability that the inner ($R/{R}_{\mathrm{vir}}\leqslant 0.5$) and outer ($R/{R}_{\mathrm{vir}}\gt 1$) measured quenched fractions could be consistent with one another.

In the ZENS sample, the results that are obtained when considering only galaxies located at large halo-centric distances $R\gt {R}_{\mathrm{vir}}$ or in low-mass halos differ from the global trends that we have discussed so far when integrating over all R_vir or M_Halo (see bottom panels of Figures 1(a) and (b)). Low-mass groups, in fact, do not display a clear radial variation of f_Q in both galaxy mass bins here studied. The flattening of f_Q versus $R/{R}_{\mathrm{vir}}$ for ${10}^{10}\;{M}_{\odot }\lt {M}_{{\rm{galaxy}}}\lt {10}^{10.7}\;{M}_{\odot }$ galaxies is driven by groups with ${M}_{\mathrm{Halo}}\lt {10}^{13.2}\;{M}_{\odot }$ (a stronger radial dependence is observed in more massive groups). The lack of strong radial variations of f_Q at low halo masses may be genuinely due to the inability of these environments (e.g., due to a low intragroup medium density) to produce environmental effects. The possibility remains, however, that this may simply reflect the difficulty of identifying galaxy members, including the central galaxy, in these small, loose groups (see also above and the discussion in Paper I regarding the uncertainties affecting the definition of the group centers). Mixing of the inner and outer satellite populations could also be causing the mild decrease in f_Q with increasing group mass for galaxies located at large halo-centric distances in low-mass groups (see lower left panels of Figures 1(a) and (b)).

Figures 2(a) and (b) show in graphic form the results of Table 1 for the fraction of quenched satellites that have early-type morphology, ${f}_{{Q}_{\mathrm{ETG}}}$. Specifically, they show the morphological composition of the quenched satellite population for the two bins of satellite stellar mass, as functions of halo mass and halo-centric radius. The black (dashed, dot-dashed, and dotted) lines in both figures are again the predictions of classical SAMS, which we discuss in Section 5.4.

Zoom In Zoom Out Reset image size

As commented in Section 2.4, in ZENS there are virtually no bulgeless quenched galaxies at the galaxy mass scale that we are studying, and the quenched satellites consist therefore largely of intermediate-type disks with a dominant disk component but a nonnegligible bulge and of a few elliptical galaxies.

Within the errors, we find no trends in the morphological mix of the quenched satellite population with either halo mass or halo-centric radius, despite the significant radial trends seen in the total fraction of quenched satellites in Figures 1(a) and (b). All of the data points in Figures 2(a) and (b) are consistent with a constant early-type fraction of about 50% for $\mathrm{log}M$ ∈ $[10,10.7[$ satellite galaxies, and about 65% for the more massive $\mathrm{log}M$ ∈$[10.7,11.5]$ ones. Thus, while it is clear from Figures 1(a) and (b) that more satellites have been quenched in the cores of groups than in their outskirts, the relative fractions of early-type and disk-dominated morphologies in the quenched satellite population do not change with group-centric distance. Similarly, no significant variation is observed in the fraction of elliptical galaxies among all quenched galaxies: we measure constant values of about 5%–10% and 20% in the two galaxy mass bins across most environments. The only exception is for $\mathrm{log}M$ ∈ $[10.7,11.5]$ galaxies at $R\gt {R}_{\mathrm{vir}}$, in which ellipticals represent 50% of the quenched population and cause the slight upturn in ${f}_{{Q}_{\mathrm{ETG}}}$ observed in the upper right panel of Figure 2(b). This difference is, however, only marginally significant (p value = 0.08), and it could be affected by contamination in the group outskirts.

The constancy of the ${f}_{{Q}_{\mathrm{ETG}}}$ with halo-centric radius in Figures 2(a) and (b) does not of course violate the well-known morphology–density relation of Dressler (1980). Our results indicate that this is due to the changing fraction of quenched galaxies with radius f_Q(R), rather than a change in the morphologies of the quenched galaxies themselves—although, as we discuss in Sections 5.2 and 5.3, there is a systematic difference in the morphologies (but, as we argue, not in the structure) of quenched satellites and their star-forming counterparts of similar mass.

Mass quenching is the process that controls the overall mass functions of star-forming and passive galaxies (Peng et al. 2010); it is clearly associated with the mass of the galaxy, although it is hard to distinguish whether it is the stellar mass, halo mass, or even black hole mass that is relevant (see discussion and references in Section 1). Environment quenching appears to act only on satellites (e.g., van den Bosch et al. 2008; Peng et al. 2012; Wetzel et al. 2012; Kovač et al. 2014), and such satellite quenching has been shown to be independent of stellar mass (when expressed as a differential quenching efficiency, as in Peng et al. 2010).

Given the well-known correlation of early-type morphology with mass, a simple hypothesis would have been to suppose that mass quenching produces early-type morphologies, while satellite quenching is not associated with any morphological transformation and adds quenched disk systems to the total population of passive galaxies. Based on our results, we can rule out this simple idea.

At a given stellar mass, some (constant) fraction of satellites should have been quenched by mass quenching, independent of environment (Peng et al. 2010). We would expect this to also be the f_Q of satellites at large radii, $R\gt {R}_{\mathrm{vir}}$. This is tantamount to saying that there is no satellite quenching per se at these large radii. This is not inconsistent with reports of alleged environmental-quenching effects on galaxies at very large distances from the centers of massive clusters (e.g., an enhanced red fraction and reduced H i emission; Wetzel et al. 2014 and references therein) since these effects are largely contributed by central galaxies of lower-mass halos and disappear when satellites only are considered (Wetzel et al. 2012). Note also that, as shown in the Appendix, most of the massive satellites that are found at very large halo-centric distances are members of unrelaxed groups without a well-defined center, which casts some doubts on their identification as $R\gt {R}_{\mathrm{vir}}$ satellites of the given halo.

Using the quenching efficiencies for mass quenching of Peng et al. (2010) and taking into account a ∼0.2 dex offset between their stellar masses (actual, including return to the interstellar medium) and ours (integral of SFR), we would estimate that about 40% and 70% of satellites, respectively, in our low and high stellar mass bins should have been mass quenched. This is consistent with what we observe at large radii in the low-mass bin, but, in the high-mass bin, the ZENS f_Q at $R\gt {R}_{\mathrm{vir}}$ is lower than this. However, at these high masses, the number of galaxies at $R\gt {R}_{\mathrm{vir}}$ is small (see Figure 11) and the corresponding statistical uncertainty quite large. We therefore set the mass-quenched f_Q value in the high-mass bin to the Peng et al. (2010) expectation of 0.7, which is less than 2σ discrepant from our observed value at $R\gt {R}_{\mathrm{vir}}$. We stress that this choice for the mass-quenched f_Q value is not critical to our discussion below. The adopted mass-quenched f_Q values are shown as horizontal red lines in the top-right panels of Figures 1(a) and (b).

Several physical mechanisms have been suggested to be the origin of mass quenching and of environment quenching. Independent of the physics behind each of these quenching channels, however, the observed increase in f_Q with decreasing halo-centric radius in these figures must reflect the increased importance of satellite quenching toward the centers of the halos (we hatch the relevant areas in the top-right panels of Figures 1(a) and (b) to guide the eyes). If mass quenching and satellite quenching had a different ability to affect galaxy structure, then this would produce a clear trend with halo-centric distance in our B/T-based morphological composition of the quenched satellite fractions shown in Figures 2(a) and (b): the morphology associated with mass quenching would dominate at large group radii, and that associated with satellite quenching would become increasingly more important (up to about a 50:50 split) toward the central group regions where the f_Q has almost doubled. Put another way, since the radial variation of f_Q is due to the variation in relative importance of the two quenching channels, we would expect, if these channels had different morphological signatures, to then see a variation of ${f}_{{Q}_{\mathrm{ETG}}}$ with f_Q. We see no such variation in ${f}_{{Q}_{\mathrm{ETG}}}$, and, as a result, we can reject this simple hypothesis. We conclude either that neither of the two quenching channels induces a morphological transformation, or that both of the two result in the same morphological changes.

Insights about the latter possibility can be obtained by comparing the morphological mix of the quenched satellites with that of the star-forming satellites at large radii, $R\gt {R}_{\mathrm{vir}}$, since these are presumably representative of the progenitors of the quenched population. The fraction of $R\gt {R}_{\mathrm{vir}}$ star-forming satellites that are morphological early types⁸ is substantially lower, in each mass bin, than the ${f}_{{Q}_{\mathrm{ETG}}}$ that we have found for the quenched satellites, i.e., $11{\%}_{-0.03}^{+0.04}$ and 35%${}_{-0.13}^{+0.18}$ for the low and high stellar mass bins as compared with 50% and 65%, respectively.

The difference in the early-type fractions of star-forming and quenched satellites may be taken to indicate that quenching is associated with structural changes in the galaxies, that is, changes affecting their underlying mass distribution, such as substantial growth of stellar mass in the bulge (or inner disk) components. It is well established that quenched galaxies have radial surface brightness profiles that are steeper than those of star-forming galaxies and are well described by high Sérsic indices (e.g., Kormendy et al. 2009 and references therein), reflecting high central mass densities (e.g., Fang et al. 2013 and references therein). Also, since the pioneering work of Mihos & Hernquist (1994), inward flows of gaseous material have long been recognized to take place during galaxy mergers and through disk instabilities. Such inward gas flows are likely to contribute to the growth of stellar mass in galactic bulges (see, e.g., Courteau et al. 1996; MacArthur et al. 2003; Immeli et al. 2004; our own work, i.e., Carollo et al. 1997, 1998, 2001, 2007 and Carollo 1999, 2004; the comprehensive review by Kormendy & Kennicutt 2004 and references therein; Bournaud et al. 2011, and many others). This idea has been further developed in more recent theoretical studies (e.g., Dekel & Burkert 2014) and is suggested to be observed at high redshifts (e.g., Genzel et al. 2006; Cameron et al. 2011) and, at least for low-mass bulges, certainly observed in the local universe, where small bulges show signs of "rejuvenation" of their stellar populations, with up to 10%–30% of their total mass consistent with having formed in the past few gigayears (Thomas & Davies 2006; Carollo et al. 2007). If mergers and disk instabilities were directly or indirectly connected with galactic quenching, it would not be implausible to draw a connection between quenching and the growth of bulges. Establishing the direction of causality in any such change would remain, however, very difficult. Quenching could be directly linked to the growth of the bulge (e.g., Martig et al. 2013; Genzel et al. 2014), or the quenching process itself could be associated with bulge growth, or the prominence of the bulge could be linked to a third property (e.g., galactic mass or the presence of an AGN) that itself controls the quenching of star formation.

A key but often neglected consideration is, however, that changes in the disk surface brightness profiles following quenching are almost certain to occur once star formation ceases, and this will inevitably increase the prominence of the bulges even if the bulge itself remains unchanged. The different early-type fractions in star-forming and quenched galaxies could therefore arise from changes in only the light distributions of galaxies, that is, in their light-defined $B/T$ values and thus morphologies, due in particular to differential surface brightness fading after star formation ceases. The effects of washing out the presence of massive bulges under highly star-forming disks at $z\sim 2$ has been, for example, highlighted by Tacchella et al. (2015).

There are theoretical suggestions whereby neither mass nor environment quenching mechanisms would produce a structural change. For example, radio-mode AGN feedback can provide a quenching mechanism that leaves unaffected the mass distributions of bulges and disks (e.g., Gabor & Davé 2012). Gas removal once the galaxies become satellites will also preserve disks. Since gas removal as a satellite-quenching mechanism is implemented in several current traditional SAMs, this gives us the opportunity to quantitatively compare our results on ${f}_{Q}$ and ${f}_{{Q}_{\mathrm{ETG}}}$ with a couple of such SAMs in Section 5.4.

In the next two sections we explore instead the global and the bulge and disk properties in the quenched and star-forming satellite populations. We study the effects of the fading of their disks (with no growth of bulges) to investigate whether structural as opposed to morphological changes are mainly responsible for the observed differences between these populations and how this can be connected with our results on the morphological fractions of mass- and environment-quenched galaxies.

We have argued in the previous section that the morphological outcome of the different quenching channels appears to be the same, even though these quenched satellites have different morphologies from the star-forming satellites, presumed to be representative of the progenitors of the quenched galaxies. Figure 5 shows in blue (upper panels) the distribution of the measured light-defined $B/T$ ratios for all star-forming satellites and in magenta (central panels) that in our original sample of quenched satellites⁹ (we defer to the next section the description of the lower panels of Figures 5 and 6). To reduce statistical errors, star-forming satellites at all halo-centric distances are included, but the following results do not change when using only the $R\gt {R}_{\mathrm{vir}}$ star-forming sample. The median $B/T$ values for the star-forming samples are ${0.23}_{-0.01}^{+0.02}$ and ${0.37}_{-0.05}^{+0.04}$ in the low and high stellar mass bins; the corresponding values for the quenched satellites are ${0.47}_{-0.02}^{+0.01}$ and ${0.60}_{-0.02}^{+0.05}$, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The quenched satellites also have smaller half-light radii than star-forming satellites. Figure 6 plots the histograms of total galaxy half-light radii ${r}_{1/2}$ for the star-forming (blue, upper panels) and the quenched (magenta, central panels) satellite samples: the average sizes of the quenched satellites are ${2.7}_{-0.1}^{+0.1}$ kpc and ${3.7}_{-0.2}^{+0.2}$ kpc at low and high stellar masses, compared with ${4.8}_{-0.2}^{+0.1}$ kpc and ${6.5}_{-0.5}^{+1.0}$ kpc for the star-forming samples. These values reflect the global mass–radius relations for quenched and star-forming nearby galaxies (see, e.g., the extensive discussion in Paper II and references therein).

The increased $B/T$ and the smaller half-light radii in the observed quenched satellites relative to the star-forming ones could be taken to indicate some growth of stellar mass in the bulges of star-forming satellites when they are moved from the star-forming "main sequence" (on the galaxy SFR versus ${M}_{{\rm{galaxy}}}$ plane, e.g., Daddi et al. 2007) to the quenched population (either before quenching, i.e., in the gas flow regime, or after quenching, i.e., in the secular stellar disk bar-like instability regime; see Kormendy & Kennicutt 2004 and references therein). In order to study this further, Figure 7 shows the surface brightness profiles for the bulge and disk components of the quenched and star-forming satellites that we derived from analytic, PSF-deconvolved bulge+disk decompositions in Paper II. In particular, we plot the typical profile of each component as a shaded area that encompasses the 25th and 75th percentiles of all surface brightness profiles. The average star-forming disk profile is shown in blue, and the average bulge in the star-forming satellites is shown in black. In magenta and gray we plot the observed disk and bulge profiles in the quenched satellites.

Zoom In Zoom Out Reset image size

It is noticeable how similar the bulge profiles of star-forming and quenched satellites are at constant galaxy stellar mass, despite the substantially lower average light-based $B/T$ value measured for the former relative to the latter (but we discuss below a small effect in the inner profiles within 1 kpc; see insets in Figure 7). In contrast, the surface brightness profiles of the disks in the quenched satellites appear to be overall fainter and steeper (with smaller exponential scale lengths) than those of the presumed-progenitor star-forming disks of the same integrated galactic mass.

The fact that the overall bulge surface brightness profiles are so similar in the quenched and in the star-forming satellites argues in principle against substantial structural changes in the bulge components that are due to quenching. The two most stable parameters for the bulges are the total luminosities and the central surface brightnesses (see Paper II). This is also confirmed by the simulations that we describe in Section 5.3.

We therefore plot in Figure 8 the $I$-band bulge surface brightness within 1 kpc, $\langle \mu {\rangle }_{1\mathrm{kpc}}$, versus the total bulge $I$-band absolute magnitude for the quenched and star-forming satellites with a bulge component in the two bins of galaxy mass. The median of the total $I$-band bulge luminosities is similar in the star-forming and quenched satellites, that is, $I$-band absolute magnitudes of $-{19.50}_{-0.08}^{+0.11}$ (or $-{19.82}_{-0.06}^{+0.08}$ if we exclude galaxies with very small bulges, $B/T\lt 0.1$, from the star-forming sample) versus $-{19.92}_{-0.05}^{+0.05}$ mag respectively in the low-mass bin, and $-{20.94}_{-0.15}^{+0.25}$ ($-{20.98}_{-0.15}^{+0.15}$) versus $-{21.10}_{-0.08}^{+0.08}$ at the high masses. There is a small shift in the average bulge surface brightness within 1 kpc at constant $I$-band magnitude between the two satellite samples. This is better shown in the bottom panels of Figure 8, where we plot the residuals ${\rm{\Delta }}\langle {\mu }_{1\mathrm{kpc}}\rangle$ for the bulges in the quenched (magenta) and star-forming (blue) samples, relative to the average relation between bulge $\langle \mu {\rangle }_{1\mathrm{kpc}}$ and bulge $I$-band magnitude (indicated by the dashed black lines in the top panels of Figure 8). The linear parameters of these average fits in the low and high galaxy mass bins are respectively given by $a=0.95\pm 0.03,\;$ $b=37.2\pm 0.6$ and $a=0.81\pm 0.08,\;$ $b=34.74\pm 1.7$.

Zoom In Zoom Out Reset image size

The small shift of order 0.3 mag in bulge ${\rm{\Delta }}\langle {\mu }_{1\mathrm{kpc}}\rangle$ is consistent with a small amount of stellar mass being added to the centers of the bulges of quenched satellites. It is clear, however, that any such increase in the overall mass of the bulge cannot be substantial and in particular cannot significantly change the overall bulge ($I$-band) luminosity.

The comparison between the disk profiles of star-forming and quenched satellites, on the other hand, indicates that the change in the global light-defined $B/T$ and the reduction in the galaxy half-light radii between star-forming and quenched satellites are primarily associated with differences in their disks. Figure 9 shows the measured disk parameters for individual star-forming (blue) and quenched (magenta) satellites. These indicate that the central disk surface brightnesses averaged within the innermost 1 kpc do not change very much ($\langle \mu {\rangle }_{\mathrm{SF},1\mathrm{kpc}}={19.72}_{-0.04}^{+0.05}$ $[{19.75}_{-0.10}^{+0.16}]$ mag versus $\langle \mu {\rangle }_{Q,1\mathrm{kpc}}={19.74}_{-0.05}^{+0.05}$ [${19.59}_{-0.09}^{+0.18}$] mag at low (high) galaxy masses), whereas the distribution of scale lengths of the disks in the quenched satellites is shifted to smaller values as compared with the scale lengths of star-forming satellites (${h}_{\mathrm{SF}}={3.3}_{-0.1}^{+0.1}[{4.7}_{-0.3}^{+0.7}]$ kpc versus ${h}_{Q}={2.2}_{-0.1}^{+0.1}[{3.3}_{-0.2}^{+0.2}]$ kpc again at low (high) masses).

Zoom In Zoom Out Reset image size

We note that, because at least at late epochs (and possibly always) star formation is mostly associated with disks, a morphological change in the direction of an increase of the light-defined $B/T$ would be expected as an aftermath of quenching, even in the absence of structural (i.e., mass-based B/T) changes, simply due to the fading in surface brightness of a star-forming disk once star formation ceases. A fading of the disks would increase the apparent $B/T$ and shift galaxies toward earlier morphological types, potentially explaining the observed $B/T$ of the quenched satellites. Fading of a star-forming disk in the aftermath of quenching clearly also reduces the overall half-light radius of a galaxy. This would also explain the smaller sizes of quenched satellites and has previously been invoked to help reconcile the differences between predicted and observed size functions of $\sim M\ast$ quenched galaxies at high redshifts in the context of the apparent change in size of the passive galaxy population (Carollo et al. 2013a). We therefore explore below in more detail the quantitative impact of disk fading on the global structure of galaxies.

To quantify the impact of postquenching disk fading on the $B/T$ and sizes of galaxies, a first approach is simply to take the bulge-to-total ratio $B/T$ of the star-forming disks, ${(B/T)}_{\mathrm{SF}}$, and recompute the new "quenched" ${(B/T)}_{Q}$ that would be obtained with a reduced disk contribution. The resulting quenched ${(B/T)}_{Q}$ will be a nonlinear function of the initial ${(B/T)}_{\mathrm{SF}}$ and the reduced surface brightness of the quenched disk, which we represent by $\xi ={D}_{Q}/{D}_{\mathrm{SF}}\lt 1$. Specifically, with

$$\begin{eqnarray}&&{(B/T)}_{\mathrm{SF}}=\displaystyle \frac{B}{B+{D}_{\mathrm{SF}}},\end{eqnarray} \tag{ 1 }$$

and under the assumption that disk fading is not accompanied by any change in the bulge component, $B$, we get

$$\begin{eqnarray}&&{(B/T)}_{Q}=\displaystyle \frac{B}{B+{D}_{Q}}=\displaystyle \frac{{(B/T)}_{\mathrm{SF}}}{\xi +{(1-\xi )(B/T)}_{\mathrm{SF}}}.\end{eqnarray} \tag{ 2 }$$

For a disk fading of about 1 mag, $\xi \sim 0.4$, a typical star-forming galaxy with an original ${(B/T)}_{\mathrm{SF}}=0.4$ will have ${(B/T)}_{Q}=0.63$ after quenching, sufficient to make it an ETG.

We can look at this from a different perspective by looking at the actual surface brightness profiles of individual star-forming galaxies in ZENS as decomposed into their bulge and disk components. Specifically, we use the light-defined $B/T$ decomposition for each star-forming satellite galaxy in ZENS (see Paper II and Figure 7), and we generate a simulated $I$-band image with the same bulge and disk structural parameters, except that the disk surface brightness is reduced, uniformly at all radii, by a certain amount: ${\rm{\Delta }}{\mu }_{\mathrm{fade}}=-2.5\mathrm{log}\xi$. We consider fading of 1.0–1.5 mag, corresponding to the fading of a disk that has been forming stars at a constant rate prior to being quenched 1–3 Gyr before the current epoch of observation (using for this estimate the Bruzual & Charlot 2003 stellar population synthesis models with a Chabrier 2003b initial mass function and a solar metallicity). These artificially quenched images are then convolved with the typical ZENS $I$-band PSF, and an appropriate level of noise is added. We then carry out the same bulge+disk decomposition fits as done for the real ZENS galaxies (see Section 2.4 and Paper II for details). The results¹⁰ are shown in red in Figures 5–7 and 9.

It is clear from Figures 5 and 7 that fading of the disk enhances the prominence of the central bulge, leading to a higher light-defined $B/T$. The light-defined $B/T$ values for the average galaxy profiles change at ${10}^{10}\lt {M}_{\odot }\lt {10}^{10.7}{M}_{\odot }$ from 0.21 for the star-forming average profile to 0.43 or 0.50 for the two fading models being examined; the corresponding change at ${10}^{10}\lt {M}_{\odot }\lt {10}^{10.7}{M}_{\odot }$ is from 0.35 to 0.57 or 0.64. The median $B/T$ values of the distributions of individual $B/T$ shown in Figure 5 change from ${0.23}_{-0.01}^{+0.02}$ and ${0.37}_{-0.05}^{+0.04}$ for the original star-forming galaxies in the low and high stellar mass bins, to an average of about 0.50 $\pm \;0.02$ and 0.55 $\pm \;0.06$ for both low and high masses for the 1 Gyr (dark red solid) and 3 Gyr (light red dashed) disk-fading models, respectively. These values are quite close to the ${0.47}_{-0.02}^{+0.01}$ and ${0.60}_{-0.02}^{+0.05}$ medians of the $B/T$ distributions of real quenched satellites in the two mass bins. A Mann–Whitney–Wilcoxon U test (or a Kolmogorov–Smirnov test) indicates that the probability that the measured $B/T$ distributions of the quenched satellites and the 1-Gyr-old artificially faded star-forming satellites are drawn from the same parent samples are substantial (precisely, $p$ values >55% or >6% and >10% or >30% in the low- and high-mass bins for the two tests, respectively; the corresponding values for a t test are >25% and $\gt 10\%$).

Figure 6 shows, however, that the galaxy half-light radii ${r}_{1/2}$ obtained from these radially uniform fading models (e.g., ${4.1}_{-0.1}^{+0.1}$ kpc and ${5.4}_{-0.4}^{+1.0}$ kpc in the two mass bins using the representative 1 Gyr postquenching model) remain still significantly larger than those of the actual quenched satellites, 2.7 kpc and 3.7 kpc, respectively. The disk scale lengths of the artificially faded disks are also too large (3.2${}_{-0.1}^{+0.1}$ kpc and ${4.3}_{-0.3}^{+0.7}$ kpc at low and high masses, respectively), and their central surface brightnesses are too faint (20.68${}_{-0.08}^{+0.08}$ and ${20.95}_{-0.25}^{+0.14}$ mag; see Figure 9).

A reduction in scale length would, however, be expected, due to a differential fading of the disks with galactic radius, e.g., due to a stellar age gradient within the disks. Smaller sizes could also arise if disks were smaller, at given mass, at earlier epochs, when the quenched satellites were actually quenched. Both would be a natural consequence of an inside-out growth of individual disks. We can quantify the first effect (differential fading) as follows. We assume that both the original (star-forming) and the subsequently quenched disks can be represented by exponential light profiles with respective scale lengths h_SF and h_Q and central surface brightnesses ${\mu }_{{SF},0}$ and ${\mu }_{Q,0}$. The radial change in surface brightness will itself be an exponential with scale length:

$$\begin{eqnarray}&&{h}_{\mathrm{fade}}=\displaystyle \frac{{h}_{\mathrm{SF}}{h}_{Q}}{{h}_{\mathrm{SF}}-{h}_{Q}}={({h}_{Q}^{-1}-{h}_{\mathrm{SF}}^{-1})}^{-1}.\end{eqnarray} \tag{ 3 }$$

With

$$\begin{eqnarray}&&{\mu }_{\mathrm{SF}}(r)={\mu }_{{SF},0}+1.085{{rh}}_{\mathrm{SF}}^{-1}\end{eqnarray} \tag{ 4 }$$

and

$$\begin{eqnarray}&&{\mu }_{Q}(r)={\mu }_{Q,0}+1.085{{rh}}_{Q}^{-1},\end{eqnarray} \tag{ 5 }$$

the fading ${\rm{\Delta }}\mu {(r)}_{\mathrm{fade}}$ at different radii will be given by

$$\begin{eqnarray}&&{\rm{\Delta }}\mu {(r)}_{\mathrm{fade}}=({\mu }_{Q,0}-{\mu }_{{SF},0})+1.085r({h}_{Q}^{-1}-{h}_{\mathrm{SF}}^{-1}).\end{eqnarray} \tag{ 6 }$$

An apparent halving of the scale length, i.e., ${h}_{Q}\sim 0.5{h}_{\mathrm{SF}}$ (roughly what we see in our data) corresponds therefore to the application of a differential fading of the same scale length as the original disk, ${h}_{\mathrm{fade}}\sim {h}_{\mathrm{SF}}$, and to ${\rm{\Delta }}\mu {(r)}_{\mathrm{fade}}=1.085{{rh}}_{\mathrm{SF}}^{-1}$. Such little fading at the center, i.e., ${\mu }_{Q,0}-{\mu }_{{SF},0}\sim 0$ and 1 mag of fading at the original scale length of the star-forming disk (consistent with the 1 Gyr passive evolution model above), can thus completely explain the smaller sizes of quenched satellites relative to disk-faded star-forming satellites.

The overall conclusion from these analyses is that a simple fading of the disks after star formation ceases can, in principle, explain most of the change in the observed (light-defined) B/T and in the mean half-light radii that is seen relative to a plausible set of star-forming progenitors, without the need for any substantial mass growth or other changes in the bulge components. This therefore also supports the idea that neither mass quenching nor satellite quenching produces a significant structural change in the stellar mass distribution of satellite galaxies. Clearly, this conclusion is based on analysis of galaxies at the present epoch, albeit reflecting quenching that happened at earlier epochs. It therefore may or may not hold for quenching occurring at very much earlier times, such as at $z\sim 2$ when passive galaxies first appear in substantial numbers (e.g., Ilbert et al. 2013).

Generally speaking, a long-standing problem of SAMs remains the overprediction of the total number of quenched satellites (Bower et al. 2006; De Lucia et al. 2006; Font et al. 2008; Kimm et al. 2009). Even recent models that implement more accurate prescriptions for the removal of the satellites' gas reservoirs still predict too many galaxies in the green valley between the red sequence and the blue cloud (Font et al. 2008). This is evident especially in Figure 1(a), in the low galaxy mass bin of our ZENS analysis and, although in a much reduced form, also in Figure 1(b) in our high galaxy mass bin. In both figures, the dotted and dashed black lines show respectively, for the SAMs of De Lucia et al. 2006 (DL06) and Guo et al. 2011 (G11),¹¹ the dependence of f_Q on the group halo mass (left panels) and halo-centric distance (right panels).

The main improvements in the G11 model relative to the earlier DL06 prescription are a more accurate calculation of the ram pressure from the intragroup medium and a more gradual resulting stripping of gas from satellites as these move toward the inner group regions. While the gradual gas stripping narrows the gap between the data and theoretical predictions by lowering the quenched satellite fractions in the G11 relative to the DL06 SAMs, the G11 model still overpredicts the absolute number of quenched satellites, in particular in our low galaxy mass bin. Most likely the star-formation histories of low-mass galaxies at ${M}_{{\rm{galaxy}}}\lt M\ast$ must be improved, and, at the same time, these galaxies should be prevented from losing their gas too easily.

At face value, the recipes implemented in both the DL06 and G11 models do reasonably well in predicting almost no dependence of the total (i.e., integrated over R) quenched satellite fraction f_Q on group halo mass and the dependence of f_Q on ${\delta }_{\mathrm{LSS}}$. Also, both models predict some decrease in the quenched satellite fraction with increasing halo-centric distance. In the low galaxy mass bin, although the instantaneous gas stripping algorithm at work in the DL06 models leads to a somewhat milder decrease in f_Q with mass than the gradual gas stripping algorithm in G11 (which gives satellites the time to establish a more significant halo-centric radial trend in the total quenched fraction), the amplitude of the radial decrease in f_Q in our data is roughly consistent with both models. In G11, the timescales for establishing the radial trend of f_Q with $R/{R}_{\mathrm{vir}}$ are typically ∼4 Gyrs, or a factor of ∼2 longer than the estimates for ${\tau }_{\mathrm{quench}}$ that are given by our analysis of the color properties of satellites in our Paper III and by several independent studies (Balogh et al. 2000; Wang et al. 2007; Weinmann et al. 2009; Feldmann et al. 2010, 2011; Rasmussen et al. 2012). Together with the large overprediction of quenched galaxies with a disk morphology that we discuss below, this suggests that gas removal from disk satellites orbiting within a common group halo is likely a faster but less universal process than what is currently implemented in the SAMs. This is also suggested by some numerical simulations of satellite evolution in group potentials (e.g., Feldmann et al. 2010). In the high galaxy mass bin, the G11 model reproduces the observed trend of f_Q with $R/{R}_{\mathrm{vir}}$ at distances $R\lt {R}_{\mathrm{vir}}$, although it shows a rather flat relation at halo-centric distances $R\gt {R}_{\mathrm{vir}};$ the DL06 model fails instead to predict the observed decrease of f_Q with $R/{R}_{\mathrm{vir}}$ even within the virial radius. Given this somewhat better performance of G11 in reproducing all observed global trends of f_Q in both stellar mass bins, we adopt this model as the fiducial one for the comparison of the observed trends of ${f}_{{Q}_{\mathrm{ETG}}}$ with M_Halo and R (and ${\delta }_{\mathrm{LSS}}$).

G11 define as morphological the early-type galaxies with a bulge-to-total ratio $B/T\gt 0.7$ in stellar mass; in ZENS we adopted a similar B/T threshold to select our early-type sample, but measured in I-band light. In Figures 2(a), (b), and 4(a), (b) we overplot the model predictions for the $B/T\gt 0.7$ sample as dash-dotted black lines. Given the difference in the definitions between data and models, we also plot equivalent model lines for a morphological mass selection equal to $B/T\gt 0.5$ to explore the effects of the precise B/T threshold value. The models are then treated as were the observations: for example, in the low galaxy mass bin (Figures 2(a) and 4(a)), where the number of galaxies enables a detailed investigation, the models are shown both integrated over the second environmental parameter (halo-centric distance for the plots with halo mass as the abscissa, and halo mass for the plots with halo-centric distance as the abscissa; top panels of the figures) and also split in the same bins of the second environmental parameter as done for the data (bottom panels).

The G11 model underpredicts the observed ${f}_{{Q}_{\mathrm{ETG}}}$. This discrepancy is larger in the low galaxy mass bin, where the amplitude of ${f}_{{Q}_{\mathrm{ETG}}}$ is underpredicted by a factor of ∼2–3 (see Figures 2(a) and 7). At these masses, the dominant mechanism to form spheroids in the G11 model is via disk instabilities. These instabilities depend on internal galaxy properties only and not on environment (see Equation (34) in G11). Furthermore, the gradual environmental "strangulation" that is implemented in the G11 model leads to satellite quenching by the stripping of the hot gas halos while leaving their prequenching morphologies unaffected.¹² This leaves the morphological mix of the quenched satellite fraction equal to that of the progenitor star-forming satellite population.

The main conclusions that we draw by comparing our data with the above SAMs, and in particular with the G11 model, is that in the regime of galaxy masses where environmentally driven satellite quenching starts being relevant, at galaxy masses below $M\ast$, the model (1) overpredicts the total quenched satellite fraction in all environments and (2) underpredicts the fraction of satellites with an early-type morphology in all environments: effectively it overpredicts the fraction of quenched satellites that retain a substantial disk component. This likely indicates that the model subjects too many star-forming disk satellites to environmental quenching, thereby leading, at least at these galaxy mass scales, to too many quenched satellites with a disk-dominated morphology.