Effect of Local Environment and Stellar Mass on Galaxy Quenching and Morphology at 0.5 < z < 2.0^*

The ZFOURGE catalogs include photometric redshifts and rest-frame colors calculated using EAZY (Brammer et al. 2008) from the 0.3 to 8 μm photometry for each galaxy. Of import here is that ZFOURGE uses templates and photometric zeropoints that are iteratively adjusted during the fitting process to improve the accuracy of the photometric redshifts.

The precision of photometric redshifts has the ability to potentially introduce spurious signals or wash out structure (Cooper et al. 2005; Quadri et al. 2012). However, our estimates of the quality of the ZFOURGE photometric redshifts show them to be very accurate, and as we demonstrate below, sufficient to recover galaxy environmental densities. By comparing photometric redshifts of galaxy pairs and to spectroscopic subsamples, Straatman et al. (2016) show that the typical photometric-redshift uncertainties are ${\sigma }_{z}/(1+z)=0.01\mbox{--}0.02$ to the K_s-band magnitude limit for galaxies between z = 0.5 and z = 2.0, with negligible dependence on galaxy color, but there is dependence on magnitude and redshift (Straatman et al. 2016, see their Section 5.4). Quantitatively, at ${K}_{s}\lt 25.0$ AB mag, the photometric-redshift uncertainties of quiescent galaxies at z = 2 are better than those of star-forming galaxies at the same redshift—only about 5%. Other studies with ZFOURGE have shown that these redshifts are sufficient to identify protoclusters out to $z\sim 2$ (e.g., Spitler et al. 2012; Yuan et al. 2014; Forrest et al. 2017).

In addition to the photometric redshifts, we also make use of stellar masses for galaxies provided in the ZFOURGE catalogs. The stellar masses were derived by fitting stellar population models to the photometry using FAST (Kriek et al. 2009), assuming exponentially declining star formation histories, solar metallicity, and a Chabrier (2003) initial mass function.

Because we are concerned with the galaxy quiescent fractions, it is important that we use a data set that is complete in stellar mass for both star-forming and quiescent galaxies. Quiescent galaxies have higher mass-to-light ratios and therefore will have a higher mass-completeness limit than star-forming galaxies at fixed magnitude. Here, we adopt 90% mass-completeness limits for galaxies with a quiescent stellar population using the technique described by Quadri et al. (2012). In a given narrow redshift bin, we select all quiescent galaxies and scale their fluxes (and therefore their stellar masses) downward until they have the same magnitude as the measured magnitude limit, K_s = 25.5 AB mag, for all three ZFOURGE fields. Then, we define the mass-completeness limit as the stellar mass at which we recover 90% of the dimmed galaxies at each redshift. We provide the adopted completeness limits for ZFOURGE at $0.5\lt z\lt 2.5$ in Table 1.

Table 1.  Stellar Mass-completeness Limits for ZFOURGE Galaxies at $0.5\lt z\lt 2.5$

Redshift	$\mathrm{log}(M/{M}_{\odot })$	Redshift	$\mathrm{log}(M/{M}_{\odot })$
0.5	8.09	1.6	9.24
0.6	8.25	1.7	9.31
0.7	8.40	1.8	9.38
0.8	8.52	1.9	9.44
0.9	8.64	2.0	9.51
1.0	8.74	2.1	9.57
1.1	8.84	2.2	9.62
1.2	8.93	2.3	9.68
1.3	9.01	2.4	9.73
1.4	9.09	2.5	9.79
1.5	9.17	⋯	⋯

Download table as:  ASCIITypeset image

Our goal is to measure the fraction of quiescent galaxies as a function of stellar mass, environment, and redshift. From the parent sample of all galaxies in the ZFOURGE catalog, we first select all well-detected galaxies (USE flag = 1) and group them into three bins of redshift, $0.5\lt z\lt 1.0$, $1.0\lt z\lt 1.5$, and $1.5\lt z\lt 2.0$. We then further subdivide the samples into bins of galaxy stellar mass, $8.8\lt \mathrm{log}(M/{M}_{\odot })\lt 9.8$, $9.8\lt \mathrm{log}(M/{M}_{\odot })\lt 10.5$, and $10.5\lt \mathrm{log}(M/{M}_{\odot })\lt 11.5$. In each of these bins, we classify galaxies as either star forming or quiescent using their rest-frame U − V and V − J colors, denoted by ${(U-V)}_{0}$ and ${(V-J)}_{0}$, respectively. This UVJ color–color space is useful to separate galaxies with colors of quiescent and star-forming stellar populations (including the effects of dust attenuation; Williams et al. 2009; Patel et al. 2012). Due to the small systematic variations in the rest-frame colors of galaxies at fixed stellar mass and redshift in different surveys, we follow our previous method (Kawinwanichakij et al. 2016, see Section 2.2 in that paper) to self-calibrate the region in the UVJ color–color space to delineate star-forming from quiescent galaxies. We then select quiescent galaxies whose rest-frame colors satisfy

$$\begin{eqnarray}{(U-V)}_{0} & \gt & 1.2\times {(V-J)}_{0}+0.2\\ {(U-V)}_{0} & \gt & 1.3\\ {(V-J)}_{0} & \lt & 1.6.\end{eqnarray} \tag{ 1 }$$

A summary of the number of galaxies from each galaxy mass, redshift, and density subsample is given in Table 2.

Table 2.  Number of Quiescent and Star-forming Galaxies in Different Stellar Mass and Density Quartiles in ZFOURGE at $0.5\lt z\lt 2.0$

Stellar Mass Range	Redshift Range	Lowest-density Quartile			Highest-density Quartile
		${N}_{\mathrm{Quiescent}}$	${N}_{\mathrm{Star} \mbox{-} \mathrm{forming}}$	${N}_{\mathrm{Total}}$	${N}_{\mathrm{Quiescent}}$	${N}_{\mathrm{Star} \mbox{-} \mathrm{forming}}$	${N}_{\mathrm{Total}}$
$8.8\lt \mathrm{log}(M/{M}_{\odot })\lt 9.8$	$0.5\lt z\lt 1.0$	24	662	686	113	427	540
	$1.0\lt z\lt 1.5$	10	491	501	22	406	428
	$1.5\lt z\lt 2.0$	8	259	267	5	212	217
$9.8\lt \mathrm{log}(M/{M}_{\odot })\lt 10.5$	$0.5\lt z\lt 1.0$	39	129	168	96	138	234
	$1.0\lt z\lt 1.5$	32	146	178	42	144	186
	$1.5\lt z\lt 2.0$	24	196	220	40	165	205
$10.5\lt \mathrm{log}(M/{M}_{\odot })\lt 11.5$	$0.5\lt z\lt 1.0$	30	33	63	87	50	137
	$1.0\lt z\lt 1.5$	15	37	52	51	51	102
	$1.5\lt z\lt 2.0$	19	64	83	54	74	128

Download table as:  ASCIITypeset image

Figure 1 shows the UVJ color–color diagram for galaxies in our ZFOURGE samples at $0.5\lt z\lt 2.0$ split into low stellar mass ($8.8\lt \mathrm{log}(M/{M}_{\odot })\lt 10.2$) and high stellar mass ($10.2\lt \mathrm{log}(M/{M}_{\odot })\lt 11.5$) subsamples, and as a function of environment as defined by local overdensity (see Section 3 below). Although all panels show a similar range of galaxy colors, differences in the distributions are clearly evident with mass (and also environment). The distribution of lower-mass galaxies is weighted more toward bluer star-forming objects (by number), while the distribution of higher-mass star-forming galaxies show higher dust attenuation, consistent with other studies (e.g., Wuyts et al. 2011). All panels also show the existence of a "red sequence" of quiescent galaxies, but this is much more pronounced in the higher-density regions (denser environments), which we discuss more below.

Zoom In Zoom Out Reset image size

In our analysis, we study the morphological differences between quiescent and star-forming galaxies in different environments and as a function of stellar mass. The majority of galaxies in our sample fall within the CANDELS coverage from HST/WFC3, with the effective semimajor axis, a_eff, and Sérsic index, n, measured by van der Wel et al. (2012) using the HST/WFC3 F160W (H₁₆₀)-band imaging. We refer the reader to van der Wel et al. for the measurement of the random and systematic uncertainties of the estimated morphological parameters using simulated galaxy images. We cross-matched the sources in our catalog with those of van der Wel et al. The fractions of our galaxy sample with available morphological parameters from van der Wel et al. are 80%, 90%, and 82% for CDFS, COSMOS, UDS, respectively. We note that 10%–20% of our galaxy sample have no morphological information from HST/WFC3 because those galaxies are in the regions around the edges of ZFOURGE fields where there is no HST/WFC3 coverage. We further define the circularized effective radius as ${r}_{\mathrm{eff}}={a}_{\mathrm{eff}}\sqrt{q}$, where a_eff is the effective semimajor axis and $q=b/a$ is the ratio of the semiminor to semimajor axis. In addition to the morphological parameters from van der Wel et al., we calculate the inner 1 kpc of the stellar mass surface density, the procedure for which we describe in Appendix B.

We show the quiescent fraction of galaxies as a function of the overdensity in Figure 3. We apply a mass limit of $\mathrm{log}(M/{M}_{\odot })\gt 9.5$ in all three redshift bins, which corresponds to our completeness limit at z = 2.0. We also compare our quiescent fraction with those from Quadri et al. (2012), who used the galaxy sample from the UKIDSS Ultra-Deep Survey, so we apply a mass limit of $\mathrm{log}(M/{M}_{\odot })\gt 10.2$, corresponding to the completeness limit at z = 2.0 used by Quadri et al. (2012). The error bars indicate the 1σ uncertainty based on Poisson statistics for the number of quiescent galaxies in a bin. At all redshift ranges, we see evidence for a higher quiescent fraction of galaxies at higher densities. This effect is very strong at $z\lt 1$ (left panel of Figure 3), but decreases at higher redshift, $z\gt 1$ (middle and right panels of Figure 3). This is in agreement with previous studies of the star formation–density relation (e.g., Quadri et al. 2012), where with the ZFOURGE data we have extended the result to lower masses and higher redshifts (for a recent study of a galaxy sample with stellar mass and redshift range comparable to our sample, see Guo et al. 2017).

Zoom In Zoom Out Reset image size

In this section, we showed that the quiescent fraction of galaxies is higher in denser environments over all redshifts probed in this study. In principle, it is possible that this result is caused by differences in the stellar mass distribution and/or the redshift distribution of quiescent and star-forming galaxies. To check for this, we create samples of quiescent and star-forming galaxies such that their stellar mass and redshift distributions are matched following the method described by Kawinwanichakij et al. (2016, see their Section 3.2). The left panel of Figure 4 shows the differences in overdensity between the quiescent and star-forming galaxies before this matching at $1.5\lt z\lt 2.0$ (a p-value of the differences as measured by a K–S test of $\ll {10}^{-3}$), and the right panel shows that the difference persists even after matching the stellar mass and redshift distributions (a p-value of $\ll {10}^{-3}$). We obtain even more significant results in our other (lower) redshift bins. We conclude that, at all redshifts studied here, quiescent galaxies are more common in overdense regions compared to star-forming galaxies even taking into account differences in redshift and stellar mass.

Zoom In Zoom Out Reset image size

Figure 5 shows the quiescent fraction as a function of stellar mass in bins of redshift, separating galaxies in the highest (${\delta }_{75}$) and lowest (${\delta }_{25}$) density quartiles. Qualitatively, at $0.5\lt z\lt 1.5$, galaxies in the highest-density quartile show higher quiescent fractions than galaxies with the same mass in the lowest-density quartile in all stellar mass bins. This is in agreement with Allen et al. (2016), who show that the fraction of quiescent galaxies with $\mathrm{log}(M/{M}_{\odot })\gtrsim 9.5$ at $z\sim 0.95$ increases with decreasing distance to the cluster core. At higher redshift, $1.5\lt z\lt 2.0$, this trend persists for more massive galaxies ($\mathrm{log}(M/{M}_{\odot })\gtrsim 10$).

Zoom In Zoom Out Reset image size

Figure 6 shows the quiescent fraction as a function of redshift in bins of stellar mass, separating out the lowest and highest-density quartiles. Qualitatively, galaxies have higher quiescent fractions in high-density environments compared to galaxies in low-density environments out to $z\sim 2$ and for galaxies with stellar mass $\mathrm{log}(M/{M}_{\odot })\gtrsim 9.8$. For lower stellar mass galaxies ($8.8\lt \mathrm{log}(M/{M}_{\odot })\lt 9.8$), the quiescent fractions are higher in the higher-density environment at least to $z\lesssim 1.5$, but at higher redshift, $z\gtrsim 1.5$, the quiescent fraction shows less dependence on environment.

Zoom In Zoom Out Reset image size

The range of quiescent fraction as a function of overdensity and stellar mass is large, ranging from nearly 100% to less than 1%. To illustrate this, we show the quiescent fraction in log scale instead of linear scale in both Figures 5 and 6. This better presents the separation of quiescent fractions of galaxies in different environments and also the relative quiescent fraction in each stellar mass bin. Because more massive galaxies tend to exist in higher-density environments, it is logical to ask to what extent this drives the trend among mass, redshift, and environment. This is a reasonable question, as the number density of quiescent galaxies, even at high redshift, increases with both stellar mass and environment (e.g., C. Papovich et al. 2017, in preparation).

To answer this, we perform a similar procedure to Quadri et al. (2012). We computed the mean overdensity of quiescent and star-forming galaxies in narrow 0.25 dex bins in stellar mass (where we expect such narrow bins to have a negligible change in overdensity), which we show in Figure 7. This figure shows evidence that quiescent galaxies have a higher overdensity than mass-matched star-forming galaxies (down to the stellar mass limit of each redshift), and this trend exists to $z\sim 2$. These trends extend correlations found by Quadri et al. (2012) to lower stellar masses. In particular, the density contrast between star-forming and quiescent galaxies is largest for lower-mass galaxies, $\mathrm{log}(M/{M}_{\odot })\simeq 9\mbox{--}10$. As argued by Quadri et al. (2012), this suggests that the environment plays a dominant role in galaxy quenching, and here we show this extends to the lowest stellar masses. Our analysis supports this assertion, which we discuss more below.

Zoom In Zoom Out Reset image size

To quantify the environment and stellar mass in quenching the star formation activity in galaxies, we follow the approach of Peng et al. (2010, which is similar to the methods of van den Bosch et al. 2008; Quadri et al. 2012; Kovač et al. 2014; Lin et al. 2014). We define the environmental quenching efficiency, ${\varepsilon }_{\mathrm{env}}$, as the fraction of galaxies at a given stellar mass, M, that are quenched in excess of those in the lowest-density environment (presumably these are galaxies that would be forming stars in the lowest-density environments, but have had their star formation truncated due to some physical process related to the environment). The environmental quenching efficiency is then

$$\begin{eqnarray}&&{\varepsilon }_{\mathrm{env}}(\delta ,{\delta }_{0},M)=\displaystyle \frac{{f}_{q}(\delta ,M)-{f}_{q}({\delta }_{0},M)}{1-{f}_{q}({\delta }_{0},M)},\end{eqnarray} \tag{ 4 }$$

where f_q is the quiescent fraction for galaxies with stellar mass M and overdensity, δ. ${\delta }_{0}$ is the overdensity of the low-density reference environment, where we choose ${\delta }_{0}={\delta }_{25}$, i.e., the overdensity demarcating the lowest 25th percentile of the overdensity distribution (see Figure 2). We note, however, that we are parameterizing non-environmental quenching with stellar mass because it correlates with the other quantities associated with (stellar) mass quenching such as the central stellar mass and black hole mass (e.g., Woo et al. 2013, 2015, 2017b; Zolotov et al. 2015; Tacchella et al. 2016a, 2016b; Terrazas et al. 2016)

For galaxies with $\delta \leqslant {\delta }_{25}$, our definition of ${\varepsilon }_{\mathrm{env}}$ explicitly assumes that environment quenching is negligible (i.e., ${\varepsilon }_{\mathrm{env}}(\delta \lt {\delta }_{0})\approx 0$) for all stellar masses. This is a reasonable assumption as there is no apparent evolution in the shape of the quiescent galaxy stellar mass function in low-density environments over the redshift and stellar mass range considered here (C. Papovich et al. 2017, in preparation), as would be expected if galaxy quenching correlates only with stellar mass. For the remainder of this paper, we will also denote the (stellar mass-dependent) quiescent fraction of galaxies in the lowest- and highest-density quartiles as ${f}_{q}({\delta }_{25},M)$ and ${f}_{q}({\delta }_{75},M)$, respectively.

Similarly, we define the (stellar) mass quenching efficiency ${\varepsilon }_{\mathrm{mass}}$ as the fraction of galaxies at a fixed overdensity, $\mathrm{log}(1+\delta )$, that are quenched compared to the star-forming fraction at low masses. Specifically, we define the mass quenching efficiency to be,

$$\begin{eqnarray}&&{\varepsilon }_{\mathrm{mass}}(M,{M}_{0},\delta )=\displaystyle \frac{{f}_{q}(\delta ,M)-{f}_{q}(\delta ,{M}_{0})}{1-{f}_{q}(\delta ,{M}_{0})},\end{eqnarray} \tag{ 5 }$$

where in practice we take the reference mass M₀ to be the stellar mass at the completeness limit for a given redshift, and we compute the mass quenching efficiency for galaxies with $\delta \lt {\delta }_{75}$.

Figure 8 compares the strength of the environmental quenching and (stellar) mass quenching efficiencies as a function of stellar mass for galaxies in the highest-density environments ($\delta \geqslant {\delta }_{75}$). At all redshifts, the (stellar) mass quenching efficiency increases with stellar mass. At $1\lt z\lt 1.5$ and $1.5\lt z\lt 2$, the magnitude of environmental quenching efficiency is on par with the (stellar) mass quenching efficiency: in the highest-density environments, roughly half of all galaxies are quenched by the environment.

Zoom In Zoom Out Reset image size

At lower redshifts, $0.5\lt z\lt 1$, the evolution of the environmental quenching efficiency is strongest for lower-mass galaxies. For example, in the mass range $8.8\lt \mathrm{log}(M/{M}_{\odot })\,\lt 10$, ${\varepsilon }_{\mathrm{env}}$ increases from <10% at $z\gt 1$ to ∼30% at $z\lt 1$. Moreover, at these redshifts, the environmental quenching efficiency dominates over (stellar) mass quenching efficiency for these lower-mass galaxies (in the highest-density environments). Therefore, in the highest-density environments, the majority of quiescent lower-mass galaxies have been quenched by environmental processes rather than by other processes (see also Hogg et al. 2003; van den Bosch et al. 2008; Quadri et al. 2012). Comparing the magnitudes of the environmental and (stellar) mass quenching efficiencies gives an estimate of the effect, which is of order ${\varepsilon }_{\mathrm{env}}/{\varepsilon }_{\mathrm{mass}}\gt 5$ for galaxies with $\mathrm{log}(M/{M}_{\odot })\,=8.8\mbox{--}9.8$, i.e., the environment accounts for the quenching of five out of six galaxies in this mass range.

Figure 8 also shows that at $0.5\lt z\lt 1.0$, the environmental quenching efficiency appears to be nearly independent of stellar mass. At $z\gt 1$, the environmental quenching efficiency shows a clearer dependence on stellar mass: more massive galaxies experience stronger environmental quenching. This persists at least to $z\simeq 2$ for galaxies with $\mathrm{log}(M/{M}_{\odot })\gt 9.8$.

We explore the evolution with redshift of both the environmental quenching efficiency and the mass quenching efficiency for galaxies in three stellar mass bins: $9.5\lt \mathrm{log}(M/{M}_{\odot })\lt 9.8$, $9.8\lt \mathrm{log}(M/{M}_{\odot })\lt 10.2$, and $\mathrm{log}(M/{M}_{\odot })\,\gt 10.2$ in Figure 9. At lower redshifts, $0.5\lt z\lt 1$, the environmental quenching efficiency of galaxies at all masses is ≈0.3. This is generally consistent with that measured with the same (relatively higher) stellar mass at $0.3\lt z\lt 0.6$ from zCOSMOS (Peng et al. 2010). However, this "constant" quenching efficiency is a coincidence of epoch. At higher redshifts, the environmental quenching efficiency of low-mass galaxies decreases and is very low ($\lesssim 5 \%$) at $z\gt 1.5$, while for more massive galaxies it remains roughly constant (or possibly slightly declining) out to $z\sim 2$. The evolution of the environmental quenching efficiency depends both on redshift and stellar mass, and its effects are not separable from stellar mass at higher redshift.

Zoom In Zoom Out Reset image size

Comparing to the literature, the environmental quenching efficiency we derive at $1\lt z\lt 1.5$ is modestly lower than that derived for a sample of spectroscopically confirmed galaxy clusters at $0.87\lt z\lt 1.63$ with $\mathrm{log}(M/{M}_{\odot })\gt 10.3$ from the Spitzer Adaptation of the Red-Sequence Cluster Survey (SpARCS; Nantais et al. 2016). However, this may be expected as the SpARCS sample includes very rich clusters at these redshifts, which include galaxies in even higher overdensities than the galaxies in our highest-density environments in ZFOURGE. Our result of increasing quiescent fractions in high-density environments implies that the strength of the environment quenching efficiency increases with overdensity, and it is very reasonable that this efficiency is even higher in the rich environments of galaxy clusters.

Figure 9 also shows that the strength of the (stellar) mass quenching efficiency increases with increasing stellar mass and decreasing redshift. This is consistent with the overall decrease in star formation activity in galaxies at later cosmic times (e.g., Madau & Dickinson 2014).

Our main result is that there is strong evidence for both (stellar) mass quenching and environmental quenching for galaxies to high redshift. For massive galaxies, $\mathrm{log}(M/{M}_{\odot })\,\gt 10.2$, environmental quenching is evident, and nearly unchanging (or slowly declining), over the redshift range of our sample, $0.5\lt z\lt 2$. For lower-mass galaxies, the environmental quenching efficiency evolves strongly with redshift, at least to $\mathrm{log}(M/{M}_{\odot })=9.5$, where our data are complete. For such lower-mass galaxies, the environmental quenching declines by roughly an order of magnitude from z = 0.5 to 2. At our lower redshift range, $0.5\lt z\lt 1.0$, the environmental quenching efficiency dominates over the (stellar) mass quenching efficiency by a factor of >5:1 for galaxies with $\mathrm{log}(M/{M}_{\odot })=8.8\mbox{--}9.8$ (Figure 8). Therefore, the majority of low-mass quiescent galaxies are quenched by their environment (Hogg et al. 2003; Quadri et al. 2012). Our result here is consistent with that of Geha et al. (2012), who found that the number of quiescent low-mass galaxies with $7\lt \mathrm{log}(M/{M}_{\odot })\lt 9$ in the field is very low ($\lt 0.06 \%$), demonstrating that star formation in low-mass galaxies ($\mathrm{log}(M/{M}_{\odot })\lt 9$) are suppressed by being near more massive galaxies (although see Geha et al. 2017). In addition, our results are in excellent agreement with the recent study by Guo et al. (2017), who used CANDELS data to measure the distance from low-mass galaxies to the nearest massive neighbor galaxies. They found that environmental quenching is the dominant quenching mechanism for galaxies with $\mathrm{log}(M/{M}_{\odot })\,\lt 9.5$ out to $z\sim 1$. At higher redshift, Guo et al. (2017) observed minimal environmental quenching for low-mass galaxies, which is consistent with our finding here, but our observation with the ZFOURGE survey provides us with sufficient statistics and accurate environment measurements (due to the precise photometric redshifts) to strengthen this result.

We note that at $0.5\lt z\lt 1.0$, even in low-density environments, the fraction of massive quiescent galaxies with stellar mass $\mathrm{log}(M/{M}_{\odot })\gtrsim 10.8$ is comparable to those in the high-density environment. The observation of quiescent galaxies in voids (low-density environment) has been reported by Croton et al. (2005). Croton & Farrar (2008) further compared the luminosity function of void galaxies in the 2dF Galaxy Redshift Survey to that from a galaxy formation model built on the Millennium simulation. These authors demonstrated that a population of quiescent galaxies in low-density environments will arise naturally due to a combination of a shift in the halo mass function in low-density environments and an environment independent of a star formation suppression mechanism efficient above a critical halo mass of ${M}_{\mathrm{vir}}\sim {10}^{12.5}{M}_{\odot }$ (radio-mode AGN).

Some hint of the quenching mechanism comes from the timescales and the evolution in the quenching efficiency. The lack of significant environmental quenching in low-mass galaxies at $z\gt 1$ suggests that the quenching timescale is at least 3–5 Gyr (corresponding to the lookback time from z = 1 to an infall epoch of $z=3$–6). This is consistent with quenching times from other studies of environmental processes (e.g., Peng et al. 2010, 2015; Tinker & Wetzel 2010; Quadri et al. 2012; Slater & Bell 2014; Wetzel et al. 2015; Darvish et al. 2016; Fossati et al. 2017; Guo et al. 2017). Several recent studies (e.g., Fillingham et al. 2015; Peng et al. 2015; Davies et al. 2016) argue that environmental quenching for galaxies with $\mathrm{log}(M/{M}_{\odot })=8.0\mbox{--}10.0$ are primarily driven by starvation because quenching timescales and cold gas-depletion timescales are comparable. For more massive galaxies, $\mathrm{log}(M/{M}_{\odot })\gtrsim 10$, the quenching timescale could be shorter, given that we see higher environmental quenching efficiency for these galaxies even in our highest redshift bin, $1.5\lt z\lt 2.5$. This suggests a mass-dependent quenching mechanism, such as "overconsumption" (McGee et al. 2014), which arises as more massive (star-forming) galaxies have shorter gas-depletion times (which we discuss more below). Our results support these findings, but with the additional requirement (also discussed below) that the quenching process also transform the morphologies of the quenching galaxies.

The environmental processes driving the quenching must occur in environments with overdensities comparable to those of our high-density quartile. The ZFOURGE fields contain some massive groups (Fossati et al. 2017) but no massive, virialized clusters given the cosmological volume contained in the ZFOURGE/CANDELS fields. Furthermore, our overdensity estimator based on the third nearest neighbor distance measurements are primarily sensitive to group-sized scales (Muldrew et al. 2012). Therefore, the environmental quenching efficiency we measure pertains to physical mechanisms within such environments and not necessarily to more massive clusters, which may have even stronger environmental quenching efficiency. In addition, even though we do not separate our galaxy sample into central and satellite galaxies in this study, we note that if environmental effects are specific to satellite galaxies, the observed trend here would be even stronger (see Fossati et al. 2017).

One way to constrain the cause or causes of environmental quenching is to test if they also affect the morphological structures of the quenched galaxies. Previous studies have demonstrated a relation between galaxy morphology and star formation activity, quenched fractions, and implied gas fractions (e.g., Franx et al. 2008; Wuyts et al. 2011; Bell et al. 2012; Papovich et al. 2015), and quenching is driven by the processes that change morphology and grow black holes (non-environmental effects which we refer to as "(stellar) mass quenching" in this study; e.g., Dekel & Burkert 2014; Woo et al. 2015; Zolotov et al. 2015; Terrazas et al. 2016).

In contrast, quenching from environmental processes manifests in different ways. One null hypothesis is that environmental quenching has no effect on galaxy morphology. If this is true, then it would suggest that quiescent galaxies in high-density environments (which are affected by quenching processes that correlate with stellar mass and environment; Figure 8) would have different morphologies from quiescent galaxies in low-density environments (which are affected only by mass quenching); instead, their morphologies would be more similar to the star-forming population in dense environments. We test this hypothesis here by comparing the morphological distributions of quiescent and star-forming galaxies in our different environments.

We begin by showing the number of galaxies as functions of stellar mass and projected local density (environment) at $0.5\lt z\lt 1.0$, $1.0\lt z\lt 1.5$, and $1.5\lt z\lt 2.0$ as a function of both $\mathrm{log}{(1+{\delta }^{{\prime} })}_{3}$ and $\mathrm{log}(M/{M}_{\odot })$ in Figure 10 (top panels). We then show the quiescent fraction of galaxies at the same redshift ranges as a function of both $\mathrm{log}{(1+{\delta }^{{\prime} })}_{3}$ and $\mathrm{log}(M/{M}_{\odot })$ in Figure 10 (middle panels). We observe both environmental quenching and mass quenching out to $z\sim 2.0$—the quiescent fraction of galaxies increases with both stellar mass and overdensity as we have shown earlier.

Zoom In Zoom Out Reset image size

The middle panel of Figure 10 also shows that for massive galaxies with $\mathrm{log}(M/{M}_{\odot })\gt 10$, the contour of "constant color" (quiescent fraction) goes from nearly vertical at low redshift to nearly horizontal (the high-density regions) at high redshift, demonstrating that we get 50% of quenching from mass quenching and 50% from environmental quenching, even at $z\sim 2$ (in the high-density regions). This finding is consistent with what we have shown in Figure 8.

The bottom panels of Figure 10 show the median Sérsic index of galaxies at $0.5\lt z\lt 1.0$, $1.0\lt z\lt 1.5$, and $1.5\lt z\lt 2.0$ as a function of stellar mass ($\mathrm{log}(M/{M}_{\odot })$) and environment ($\mathrm{log}{(1+{\delta }^{{\prime} })}_{3}$). As shown in the middle panels, we find that the distribution of the median Sérsic index closely resembles that of quiescent fractions of galaxies as in previous works (see references above). In this work, we see that these quantities closely track each other remarkably well across the all masses, environments, and redshifts we probe (this can be seen visually in Figure 10). The similarity of the quiescent fraction and Sérsic index distributions is consistent with the picture in which the (stellar) mass quenching is reflecting the quenching processes that are more directly correlated with morphology, such as bulge-building/compacting mechanisms (e.g., Lang et al. 2014; Woo et al. 2015, 2017b; Zolotov et al. 2015; Tacchella et al. 2016a, 2016b; Terrazas et al. 2016).

In addition, it is interesting that quiescence and concentrated morphology (Sérsic index) go together even at high redshift ($z\gt 1.5$) and for the environmental quenching, indicating that only particular galaxies are susceptible to environmental quenching, and those are the galaxies that already have a high Sérsic index. These galaxies have time to make it into dense environments, or they collapsed early and have concentrated morphologies. At low redshifts, it is different, as the environment affects galaxy star formation independent of the galaxy properties.

There are some indications of deviations from this. As a function of environmental density, for lower-mass galaxies with $\mathrm{log}(M/{M}_{\odot })\lt 10.2$, there is some indication that the change in quiescent fraction is faster than the change in galaxy Sérsic index (at fixed stellar mass), and this exists at all redshifts. Figure 11 shows the ratio of quiescent fraction to the median Sérsic index of galaxies (${f}_{q}/n$) in 0.2 dex bins of projected local density. At low masses ($\mathrm{log}(M/{M}_{\odot })\lt 10.2;$ top panels), the plots show that the ratio of ${f}_{q}/n$ is roughly constant for densities $\mathrm{log}{(1+{\delta }^{{\prime} })}_{3}\lesssim 0$ at all redshifts, but that the ratio increases at higher overdensity. This is caused by a faster increase in the quiescent fraction while the median Sérsic index remains roughly constant (or increases slower) as a function of projected local density ($\mathrm{log}{(1+{\delta }^{{\prime} })}_{3}$) out to $z\sim 2$, at least for these low-mass galaxies (this is evident from a close inspection of Figure 10). Our finding here is consistent with that of Weinmann et al. (2009), who demonstrated that satellite-specific processes mildly enhance the concentration of galaxies once they become satellites. This may be taken as some evidence that these low-mass galaxies retain some memory of the morphology of their star-forming progenitors. However, as we discuss below, once galaxies quench (even as a result of their environment), some process also transforms galaxy morphologies on fast timescales as the distributions of the morphological and structural parameters of quiescent galaxies in high and low densities appear highly similar (see Figure 12).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In contrast, for more massive galaxies ($\mathrm{log}(M/{M}_{\odot })\gt 10.2;$ bottom panels of Figure 11), there is no evidence that the ratio of quiescent fraction to the median Sérsic index of these galaxies increases with environmental density. It is also interesting that there are more quiescent high-mass galaxies in dense environments. They already came in with high Sérsic index, providing tentative evidence for a morphological version of the pre-processing—the processes that make galaxies concentrated have happened already.

We compared the morphological properties of the galaxy populations using the cumulative distributions of the Sérsic index (n), effective radius (r_e), axis ratio (b/a) (using values from van der Wel et al. 2012), and stellar mass surface density in the inner 1 kpc (${{\rm{\Sigma }}}_{1\mathrm{kpc}}$) for quiescent and star-forming galaxies in the highest- and lowest-density environments, using the measurements described in Appendix B. At all redshifts and stellar masses in our sample, the quiescent galaxies have higher Sérsic indices, smaller effective radii, higher axis ratios, and higher mass surface densities than star-forming galaxies (Figure 15), in agreement with previous studies both in the local universe (e.g., Bell 2008; Fang et al. 2013; Omand et al. 2014) and at high redshift (e.g., van Dokkum et al. 2011; Wuyts et al. 2011; Barro et al. 2012, 2017; Bell et al. 2012; Cheung et al. 2012; Szomoru et al. 2012; Lang et al. 2014).

Turning to the environmental dependence, we also find that quiescent galaxies in high-density environments have different structural parameters from star-forming galaxies in the same environments. This would not be expected if a significant portion of quiescent galaxies in high-density environments were recently quenched and retained the morphologies of star-forming galaxies. This is true for all subsamples in stellar mass and redshift that we probe, $8.5\lt \mathrm{log}(M/{M}_{\odot })\lt 11.0$ and $0.5\lt z\lt 2.0$. Figure 12 shows a summary of the p-values from K–S tests comparing the distributions of quiescent galaxies in the highest -density quartile to those of star-forming galaxies in the highest-density quartile at $0.5\lt z\lt 1.0$, and Appendix B shows the cumulative distributions of the morphological parameters that we tested (Sérsic indexes, effective radii, axis ratios, stellar mass surface densities) for galaxies at the same redshift. We do not show the higher redshift bins, but we find the same results in all bins of mass and redshift. In all cases, the p-values are $\ll {10}^{-2}$. In other words, we can reject the hypothesis that their morphologies are drawn from the same parent distribution.

Similarly, there is no evidence (or at best, weak evidence) that quiescent galaxies in high-density environments have different morphologies from quiescent galaxies in low-density environments at any stellar mass or redshift, except at $0.5\lt z\lt 1.0$. At this low redshift, we find tentative evidence that low-mass quiescent galaxies ($8.8\lt \mathrm{log}(M/{M}_{\odot })\lt 9.8$) in high-density environments have larger effective radius than their counterparts in low-density environments, and high-mass quiescent galaxies ($\mathrm{log}(M/{M}_{\odot })\gt 10.5$) have higher Sérsic indices than their counterparts in low-density environments. However, the p-values of these are only 0.02 (equivalent to $\sim 2\sigma$ significance under the assumption of a Gaussian distribution). Figure 12 shows a summary of the p-values from K–S tests comparing the distributions of quiescent galaxies in the highest-density quartile to those in the lowest density quartile at $0.5\lt z\lt 1.0$. In all cases, we are unable to reject the hypothesis that the morphological distributions are the same (p-values > ${10}^{-1.3}$).

Our results that the morphologies of neither quiescent nor star-forming galaxies depend on environment are generally consistent with previous studies, with some notable exceptions. Some studies have found differences in the sizes and Sérsic indices of galaxies in the (lower-density) field and (higher-density) environments at high redshift, but this has mostly been restricted to comparisons between the field and clusters. Papovich et al. (2012) and Bassett et al. (2013) study the structural and morphological properties of galaxies in a z = 1.62 protocluster and compare those with the field galaxies at the same stellar mass and redshift. Both studies show that the cluster quiescent galaxies have larger average effective sizes compared to field galaxies at fixed mass (see also Cooper et al. 2012; Zirm et al. 2012; Lani et al. 2013; Delaye et al. 2014). In addition, Bassett et al. (2013) found that quiescent cluster galaxies have smaller Sérsic indices compared to the field galaxies (but this was driven by several quiescent galaxies on the edge of the cluster that may be a rare population of recently quenched star-forming galaxies), whereas the star-forming galaxies in both cluster and field show no difference in their morphologies.

On the other hand, Newman et al. (2014) do not detect a significant difference between the mass–radius relation of the quiescent galaxies in the cores of clusters (within ${R}_{\mathrm{proj}}\,\lt 1$ pMpc) and quiescent field galaxies at $z\sim 1.8$. Recently, Allen et al. (2016) studied the dependence of the mass–size relation on environment using field and cluster galaxies at $z\sim 1$. The cluster halos of their sample are on the order of ${10}^{13}{M}_{\odot }$, which are comparable to the group-sized environments we probe here. Allen et al. ruled out a size difference between quiescent field galaxies and quiescent cluster galaxies. Similarly, they also showed that the Sérsic indices of field quiescent galaxies and cluster quiescent galaxies are consistent. Our results are also consistent with Woo et al. (2017a), who compared the specific SFR–${{\rm{\Sigma }}}_{1\mathrm{kpc}}$ relation for field and satellite galaxies from SDSS with stellar mass $\mathrm{log}(M/{M}_{\odot })=9.75\mbox{--}11.0$, and find that, in a given stellar mass bin, quiescent galaxies have higher ${{\rm{\Sigma }}}_{1\mathrm{kpc}}$ relative to star-forming galaxies by ∼0.2–0.3 dex regardless of whether they are field or satellite galaxies. Therefore, our results add to the growing body of literature that the environment at most weakly affects the morphologies of galaxies when matched in mass, star formation activity, and redshift.

The lack of evidence for any environmental dependence of the morphological parameters of star-forming or quiescent galaxies has important consequences for the physical effects that drives environmental quenching. Over the mass and redshift ranges considered here, quiescent galaxies even in high-density environments have very different morphological properties from star-forming galaxies at that epoch, and thus do not appear simply as recently quenched star-forming galaxies. The implication is that the environmental quenching process transforms galaxy morphologies, and it must do so on timescales comparable to the quenching process.

However, it is important to keep in mind the following caveats. We perform the analysis using morphologies as traced by light ("light"-weighted morphology), which may lead to different morphologies as traced by stellar mass ("stellar mass"-weighted morphology Fang et al. 2013). Second, galaxies grow in size with time, but once galaxies are quenched, it stops growing in size at some earlier time. As a result, quiescent galaxies necessarily need to be smaller than star-forming galaxies at a given stellar mass and epoch. Based on this argument, Lilly & Carollo (2016) demonstrated that a high degree of environmental transformation might not be needed if one keeps track of the morphologies of the progenitors of quiescent galaxies and links mild environmental processes, such as stripping of the galaxy's outer part, disk fading, or removal of dust, to reproduce the observed quiescent morphology.

The fact that the environmental quenching efficiency evolves with redshift and (at higher redshift) stellar mass implies that the quenching mechanism itself is correlated with those quantities (namely, time and galaxy stellar mass). This observation is consistent with the "overconsumption" model (McGee et al. 2014), where the environmental quenching time depends on the stellar mass of the satellite. In this model, the cosmological accretion of gas is halted once a galaxy becomes a satellite of a larger halo, and the decline in star formation in satellites is then due to the exhaustion of a gas reservoir through star formation and outflows ("starvation"). The timescale on which the galaxy quenches is equal to the total gas available at the point of accretion, divided by the gas consumption rate.

Given the strong correlation between SFR and stellar mass (e.g., Tomczak et al. 2016), more massive star-forming galaxies have shorter gas-depletion timescales. In the overconsumption model, McGee et al. (2014) predict that delay times should depend both on galaxy stellar mass and redshift. Using this model, Balogh et al. (2016) showed that the high SFRs of massive galaxies ($\mathrm{log}(M/{M}_{\odot })\sim 10.5$) in their sample of groups and clusters at high redshift, $0.8\lt z\lt 1.2$, lead to short delay times at $z\sim 1$, consistent with the quenching timescale of at least 3–5 Gyr. This is consistent with the lack of significant environmental quenching of low-mass galaxies at $z\gt 1$, which we observed here. In addition, McGee et al. argue that, given the strong redshift evolution of the star formation rate, the quenching timescales should be shorter at $z\gt 1.5$ and is possible even with moderate outflow rates.

Nevertheless, the overconsumption model by itself does not account for the differences in the morphological distributions of the galaxies in our study. Qualitatively, in its simplest form, the overconsumption model predicts no morphological evolution of star-forming galaxies (galaxies simply exhaust their gas supply and retain their morphological appearance at infall modulo the effects of disk fading; see the discussion in Section 5.4 below). Therefore, while overconsumption can mainly account for environmental quenching of a galaxy at high redshift ($z\gtrsim 1$) after it becomes a satellite, it alone probably cannot account for the lack of observed morphological differences in quiescent galaxies in the high- and low-density environments discussed above.

At lower redshift, as galaxy-specific SFRs decline and associated outflow rates decrease, the quenching time predicted from the McGee et al. overconsumption becomes long ($\gt 10$ Gyr), and other environmental effects that are more closely aligned with dynamical processes in the halo may become more important and ultimately dominate (Balogh et al. 2016). This may also drive the environmental quenching efficiency to be more constant with stellar mass at later times ($z\lesssim 0.5\mbox{--}1$), as is observed here and in previous studies (e.g., Peng et al. 2010; Quadri et al. 2012; Kovač et al. 2014).

At low redshift ($z\lt 1$), there is still clear evidence that the act of becoming quiescent is accompanied by a change in galaxy structure, and this is true even when environmental processes are responsible for quenching star formation in galaxies. As we discussed above, the environmental quenching processes at low redshift are more likely driven by dynamical processes. Strangulation—the removal of the gas reservoir—is not expected to significantly affect galaxy morphology (see also van den Bosch et al. 2008; Woo et al. 2015). Ram-pressure stripping can remove cold gas from galaxies (e.g., Kenney & Koopmann 1999; Oosterloo & van Gorkom 2005; Chung et al. 2007; Sun et al. 2007; Abramson et al. 2011; Vollmer et al. 2012; Kenney et al. 2015), but again the morphology of a galaxy is not expected to be significantly modified (Weinmann et al. 2006; van den Bosch et al. 2008). Moreover, a hot gaseous halo is a requirement for ram-pressure stripping to be effective in satellites (e.g., Larson et al. 1980; Balogh et al. 2000; Kawata & Mulchaey 2008; McCarthy et al. 2008), so it is not clear that this can be a dominant mechanism in the lower-mass systems that dominate our study.

Although several studies argue that ram-pressure stripping is likely a rapid quenching mechanism, in groups and interacting pairs, it is primarily effective in quenching lower -mass galaxies ($\mathrm{log}(M/{M}_{\odot })\lt 8.0$; Slater & Bell 2014; Davies et al. 2015; Weisz et al. 2015; Fillingham et al. 2016) and is a less relevant quenching mechanism for more moderate-mass galaxies (Woo et al. 2017a, $9.75\lt \mathrm{log}M/{M}_{\odot }\lt 10$). Given the stellar mass range of galaxies in our samples, their environmental quenching efficiencies, and the expected range of group halo masses (e.g., Fossati et al. 2017), these processes (by themselves) seem unlikely to dominate the overall environmental trends that we observe.

It may be that multiple environmental processes are at work to quench star-forming galaxies, while others transform their morphologies. One candidate for environmental processes that would affect galaxy morphologies is mergers and interactions (similar to "merging quenching;" Peng et al. 2010), which are expected to be more frequent in denser environments at both z = 0 and higher redshifts (Fakhouri & Ma 2009). Each merger and interaction can build up the density in the inner kiloparsec of a galaxy (Lake et al. 1998). Such interactions are also shown to be more frequent for higher-mass galaxies out to z = 2.5 (Xu et al. 2012; Man et al. 2016), where the interactions could increase gas consumption, redistribute angular momentum, and form spheroids.

Frequent galaxy–galaxy encounters also lead to strong tidal torques, which could drive material to galaxy centers, fuel starbursts, build bulges, (e.g., Sobral et al. 2011), and can also lead to disk stripping. These could be combined with disk fading, which enhances the relative importance of the bulge and shifts the galaxy morphology toward being more bulge dominated (higher Sérsic index and higher ${{\rm{\Sigma }}}_{1\mathrm{kpc}}$; e.g., Carollo et al. 2013, 2016). Testing for disk fading requires morphological measurements weighted by stellar mass (rather than weighted by luminosity, which they are at present). For this, we require spatially resolved optical–near-IR structures in galaxies, which will be possible through forthcoming observations with JWST. A higher rate of (minor) mergers has been invoked to explain the accelerated morphological evolution in galaxies in higher-density environments (e.g., at $z\sim 1.6$, Papovich et al. 2012; Rudnick et al. 2012; Lotz et al. 2013), and this could explain the weak evidence that massive quiescent galaxies in the highest-density environments have increased Sérsic indices compared to massive quiescent galaxies in low-density environments (based on the p-values; Figure 12 and Appendix B). Although these processes act on galaxies over a range of redshift, they also have the ability to transform galaxy morphologies once they become satellites and would help to explain the structural differences in quiescent and star-forming galaxies in different environments.

To summarize, dense galaxy regions are complex environments, and our results suggest that there are multiple processes at work. The redshift and stellar mass evolution of the environmental quenching efficiency favors models where the gas supply is truncated as galaxies become satellites (e.g., starvation), combined with stellar mass-dependent star formation and outflows (e.g., overconsumption). This must be combined with processes such as more frequent interactions and mergers that are prevalent in denser environments and are capable of transforming galaxy morphologies. These processes would naturally connect the quenching timescale with the morphological transformation timescale, which is required to explain the data. This leads to the prediction that massive galaxies in denser environments have more tidal features than those in less dense environments. Again, forthcoming observations with JWST will provide deeper imaging to test our prediction.