On the Gas Content, Star Formation Efficiency, and Environmental Quenching of Massive Galaxies in Protoclusters at z ≈ 2.0–2.5

Previous studies probing the gas content of galaxies in high density environments suffer from small samples of extreme sources like DSFGs or AGN. In contrast, our study benefits from a relatively large sample of massive galaxies which are spectroscopically confirmed to be protocluster members (drawn from Casey et al. 2015; Casey 2016; and Hung et al. 2016). As described in detail in the aforementioned references, the targets were selected from several redshift surveys, which followed up cluster candidates from large near-infrared (NIR)-selected catalogs (mostly K-band selected galaxies) in the COSMOS field (e.g., Lilly et al. 2007, 2009; Muzzin et al. 2013; Yuan et al. 2014; Kriek et al. 2015; Tasca et al. 2017), with the exception of a few galaxies found serendipitously through other means.

Although some protocluster members were probably missed in those follow-up spectroscopic campaigns, there are no obvious biases related to the sample selection that might affect our results. At high stellar masses our sample comprises ∼70% of the known members with a stellar mass of M_⋆ ≳ 10¹⁰ M_⊙. The SFRs of the targets lie on (and below) the estimated star-forming main sequence (Figure 1), probing the typical star-forming galaxy population at this epoch. The fact that some of the most massive galaxies lie below the main sequence is attributed to environmental effects, as discussed in Section 4.3. Note that if any bias exists such that we systematically undersample galaxies with low SFRs, it would only reinforce our conclusion about the higher fraction of passive galaxies in the protocluster structures.

Zoom In Zoom Out Reset image size

In summary, our final sample includes a total of 68 sources within the two structures, 27 in the z = 2.47 overdensity and 41 in the z = 2.10 structure. Although our work focuses on galaxies with M_⋆ ≳ 10¹⁰ M_⊙, we include several members with 10⁹ M_⊙ < M_⋆ < 10¹⁰ M_⊙ which lie within the ALMA primary beam (see Tables 1 and 2 and Figure 1). This is one of the largest samples of z ≳ 2 galaxies in dense environments for which gas masses have been derived, comprising sources representative of the "normal" star-forming galaxy population, and hence, ideal to test our understanding of gas fueling in protocluster environments.

This subsample of massive sources has a good analogous control sample, which allows for a direct comparison between the properties of the field and the protocluster galaxies and, consequently, a better understanding of the environmental effects. The adopted field scaling relations come from Scoville et al. (2016), who reported ALMA Band 7 (ν = 345 GHz) observations toward ∼55 targets at the same redshifts ($\langle z\rangle =2.2$), UV-luminosities, and masses (10¹⁰ M_⊙ ≲ M_⋆ < 4 × 10¹¹ M_⊙; see Figure 1), yet residing in less dense environments (Darvish et al. 2018). These sources were drawn from the NIR-selected catalogs of Ilbert et al. (2013) and Laigle et al. (2016), similar to those used for the identification of the protocluster galaxies targeted in this work. Additionally, their gas masses were derived employing exactly the same methodology adopted here (see details in Section 3.1), and stellar masses and star formation rates were also estimated following similar procedures (see Section 3.2), allowing us to make a balanced comparison.

Given that the completeness of our sample drops at M_⋆ ≲ 10¹⁰ M_⊙ where, additionally, we lack a control sample, the conclusions from this work focus only on the most massive objects with M_⋆ ≳ 10¹⁰ M_⊙.

ALMA Band 6 observations were conducted on 2017 April 4 as part of the Cycle 4 program 2016.1.00646.S (PI: C. Casey), using the 12 m antennae in a relatively compact configuration (with the longest baselines at 0.46 km). These observations comprise a total of 46 pointings with an average on-source integration time of ∼5 minutes, encompassing a total of 68 spectroscopically confirmed galaxies within the two studied protocluster structures. The correlators were configured to maximize the bandwidth in order to increase the continuum sensitivity, providing a total bandwidth of 7.5 GHz centered at 233 GHz (≈1.3 mm).

Data reduction was individually performed following the ALMA reduction pipeline scripts in CASA (version 4.7.2). A few noisy channels in two different spectral windows were flagged for all the pointings before imaging. The continuum maps were obtained using a natural weighting of the visibilities in order to obtain the highest sensitivity, and using a pixel size of 012 (roughly seven times smaller than the beam size), yielding central noises between σ_{1.3 mm} = 40–50 μJy/beam and a typical synthesized beam size of θ_FWHM ≈ 09.

These continuum images are used to search for dust emission for each individual galaxy within 1'' of their respective optical positions. This search radius accounts for any astrometry offset between the ALMA and the optical images (e.g., Dunlop et al. 2017) and for real misalignments between the dusty regions and the bulk of the stellar emission (e.g., Swinbank et al. 2010; Hodge et al. 2016). To derive the flux densities a simple peak-finding algorithm is implemented on the primary-beam corrected images. A source is considered detected only if it satisfies a detection threshold of ≥3σ, for which less than one spurious detection is expected given our positional priors. In this process, the local noise is computed to be the 68th percentile of the distribution of pixel values, which corresponds to ±1σ for a Gaussian distribution. The errors calculated with this method are consistent with those derived by measuring the standard deviation of small off-source apertures. From all the 68 protocluster galaxies which lie within our surveyed area, 19 sources were detected above our adopted threshold, implying a detection rate of ≈25%–30% in each structure. The derived flux density of the detected targets, and the 3σ upper limits of the nondetections, are reported in Tables 1 and 2 (Appendix B), while Hubble Space Telescope (HST)/ALMA cutouts are shown in Appendix C.

To derive better constraints on the average flux densities of the nondetections, a stacking analysis was performed. We extract small cutouts in the image plane centered at the position of each galaxy and then combine them in a weighted average. Weights are estimated as the squared inverse of the noise measured around each source. Finally, the adopted stacked flux density (or upper limit if <3σ) corresponds to the maximum value within 1'' of the center of the stacked image. This stacking procedure was performed for different subsamples defined by SFR, stellar mass, or redshift, which are reported in Table 3.

Scoville et al. (2014, 2016) developed the physical and empirical bases for using the long wavelength Rayleigh–Jeans dust emission as a probe of the ISM gas content of galaxies, calibrating a ratio between the specific luminosity at rest-frame 850 μm to the total ISM mass. Our ALMA observations trace the rest-frame ∼410 and 350 μm emission for the z ∼ 2.10 and 2.47 protocluster galaxies, respectively, probing hence the Rayleigh–Jeans regime. Following Scoville et al. (2016) we derive the ISM masses using:

$$\begin{eqnarray}\begin{array}{rcl}{M}_{\mathrm{mol}} & = & 1.78\,{S}_{{\nu }_{\mathrm{obs}}}[\mathrm{mJy}]\,{(1+z)}^{-4.8}{\left(\displaystyle \frac{{\nu }_{850\mu {\rm{m}}}}{{\nu }_{\mathrm{obs}}}\right)}^{3.8}\\ & & \times \ {({\text{}}{D}_{L}[\mathrm{Gpc}])}^{2}\left\{\displaystyle \frac{6.7\times {10}^{19}}{{\alpha }_{850\mu {\rm{m}}}}\right\}\displaystyle \frac{{{\rm{\Gamma }}}_{0}}{{{\rm{\Gamma }}}_{\mathrm{RJ}}}\,{10}^{10}\,{M}_{\odot },\end{array}\end{eqnarray} \tag{ 1 }$$

where α_{850 μm} is the empirically calibrated ratio between long-wavelength dust luminosity and molecular gas (i.e., ${\alpha }_{850\mu {\rm{m}}}\,\equiv \langle {L}_{{\nu }_{850\mu {\rm{m}}}}/{M}_{\mathrm{mol}}\rangle$), ν_obs = 233 GHz is the observed frequency, D_L is the luminosity distance at the redshift z, and Γ_RJ is the correction for the departure in the rest frame of the Planck function from Rayleigh–Jeans (i.e., ${B}_{{\nu }_{\mathrm{rest}}}/{\mathrm{RJ}}_{{\nu }_{\mathrm{rest}}}$) given by

$$\begin{eqnarray}&&{{\rm{\Gamma }}}_{\mathrm{RJ}}({T}_{{\rm{d}}},{\nu }_{\mathrm{obs}},z)=\displaystyle \frac{h{\nu }_{\mathrm{obs}}(1+z)/{{kT}}_{{\rm{d}}}}{{e}^{h{\nu }_{\mathrm{obs}}(1+z)/{{kT}}_{{\rm{d}}}}-1},\end{eqnarray} \tag{ 2 }$$

with ${{\rm{\Gamma }}}_{0}={\rm{\Gamma }}({T}_{{\rm{d}}}=25\,{\rm{K}},\lambda =850\,\mu {\rm{m}},z=0)$. The uncertainties in the derived masses from this calibration are expected to be less than 25% (Scoville et al. 2016).

ISM masses were calculated for all the ALMA-detected galaxies and their associated errors were obtained by propagating the uncertainties on the flux densities. For those galaxies not detected in continuum emission, 3σ upper limits on the ISM were derived using the flux density upper limits. Additionally, the stacked fluxes described above were also used to derive stacked ISM masses (or upper limits in case of nondetections) for different subsamples divided by SFR or stellar mass. These results are summarized in Tables 1–3 (B).

The large ancillary data available in the COSMOS field provide an exquisite set of photometric measurements in multiple bands for our sample of galaxies, which can be used to estimate stellar masses and star formation rates through spectral energy distribution (SED) modeling. We use the COSMOS2015 catalog (Laigle et al. 2016; the same used by Scoville et al. 2016, from which our control sample is extracted) to obtain the photometry of each of our targets by matching counterparts within a 1'' radius. The catalog includes YJHK_s observations from the UltraVISTA survey, Y-band images from the Subaru telescope, infrared photometry from Spitzer, in addition to other data spanning from Galaxy Evolution Explorer near-ultraviolet to SPIRE/Herschel far-infrared, comprising more than 30 bands.

We perform our own SED fitting using all the available photometry, including our new ALMA data, using the MAGPHYS code (da Cunha et al. 2008, 2015), which adopts an energy balance technique between the stellar and dust emission. This approach implies that the stellar component is coupled to the dust-emitting region. While this does not necessarily apply for extreme DSFGs (e.g., Hodge et al. 2016), it usually does for less extreme star-forming galaxies like those studied in this work. The fitting was done with a fixed redshift according to the spectroscopic redshift of each source, using the spectral population synthesis models of Bruzual & Charlot (2003), with a Chabrier (2003) IMF, and a continuous delayed exponential star formation history, similar to the parameters adopted in the work by Laigle et al. (2016). The stellar masses derived for the galaxies presented here are in good agreement with those reported in the COSMOS2015 catalog with a mean ratio of $\mathrm{log}({M}_{{}_{{\rm{COSMOS}}}^{* }})/\mathrm{log}({M}_{{}_{{\rm{MAGPHYS}}}^{* }})=1.01\pm 0.06$. All the estimated SFRs and M_⋆ are also reported in Tables 1 and 2 and shown in Figure 1.

The star formation efficiency, typically measured as the SFR per unit gas mass ($\mathrm{SFE}\equiv \mathrm{SFR}/{M}_{\mathrm{mol}}$), is an important quantity in the understanding of the star formation activity and stellar mass growth, which is directly linked to the gas depletion timescale ($\tau =1/\mathrm{SFE}$). Testing its universality or dependence on other parameters, such as stellar mass, redshift, or environment, is required in order to have a complete view of the structure formation in the universe. Our observations and the measurements of the molecular gas content and SFR described above (see Section 3) allow us to study the SFE in one of the largest and most complete samples of protocluster galaxies at z > 2, shedding light on the influence of the environment on this quantity.

In Figure 2 we explore the SFE of these protocluster galaxies via the SFR–M_mol plane. Thirteen out of the 18 detected galaxies shown in this plot (∼70%) lie, within the error bars, on the field scaling relation. To quantitatively compare the SFEs of the protocluster galaxies with those of our control sample, we perform a Kolmogorov–Smirnov test between the two distributions, from which we derive a probability of p = 0.81 that both of them are drawn from the same parent distribution (when including only those detected galaxies with ${M}_{* }\,\gt {10}^{10}\,{M}_{\odot }$, if we include all the detections the probability is p = 0.27).

Zoom In Zoom Out Reset image size

At the probed evolutionary stage of these protoclusters, the SFEs of the most massive galaxies show consistent values to those expected for coeval field galaxies. Although a nonnegligible fraction of the detected galaxies seems to lie below the field relation while some upper limits suggest galaxies lying above it. From the four detected sources that lie significantly below the field relation (or to the right of it), two are part of the z = 2.10 protocluster and two are from the z = 2.47 structure. The two sources from the z = 2.47 protocluster also lie below the star-forming main sequence (Figure 1), which implies that their low SFE is driven by the reduced SFR rather than by an excess of molecular gas. Indeed, these sources lie on the expected relation if their offset from the main sequence is taken into account, as shown in Appendix A. The two sources from the z = 2.10 structure show SFRs in agreement with main-sequence galaxies and their high SFEs seem to be caused by their enhanced molecular gas masses (see Appendix A). Interestingly, most of the results derived from the stacking of the nondetections, which probe—on average—galaxies with M_* ∼ 1 × 10¹⁰ ${M}_{\odot }$ (see Table 3), seem also to be in agreement with the extrapolation of the field scaling relation with a few sources lying above it (see Appendix A).

Heterogeneous results regarding the SFEs of protocluster galaxies have been presented in the literature, including some in agreement with our results. For example, Lee et al. (2017) detected CO(3 − 2) line emission in seven star-forming galaxies associated with the protocluster 4C23.56 at z = 2.49 and derived a median SFE consistent with the reference sample (although different results have been reported in the same structure, as discussed below; Tadaki et al. 2019). Dannerbauer et al. (2017) detected an extended CO(1 − 0) emitting disk in a protocluster member galaxy at z ≈ 2.15, whose gas properties and SFR follow the same relation as normal field galaxies. Similarly, Darvish et al. (2018) used ALMA dust continuum observations to investigate the role of environmental density on the molecular gas content in a large sample of ∼400 massive (M_⋆ ≳ 10¹⁰ M_⊙) galaxies within z ≈ 0.5–3.5, where the density was estimated by the projected surface density of galaxies over different redshift slices, and concluded that the SFE is independent of galaxy overdensity. Wang et al. (2018) obtained CO(1 − 0) observations toward a concentrated group of 14 galaxies at z ≈ 2.51 which might be associated with the z ≈ 2.47 protocluster structure presented here (see also Cucciati et al. 2018; J. B. Champagne et al. 2019, in preparation). Their estimated SFEs show a large variety of values, although an interesting tentative trend suggests a high SFE toward the center of the small surveyed core. Gómez-Guijarro et al. (2019) reanalyzed the Wang et al. sample with deeper observations and present new CO observations in two new protoclusters at z = 2.13 and 2.60, finding rather similar SFEs to the field. Additionally, studies of more evolved (collapsed) clusters at z ∼ 1.6–2.1 targeting CO emission lines have found similar SFEs or depletion timescales (τ = 1/SFE) to the ones measured in field galaxies at similar redshifts (Rudnick et al. 2017).

On the other hand, Tadaki et al. (2019) presented CO(3 − 2) observations toward 66 Hα-selected galaxies in three protoclusters at z = 2.16, 2.49, and 2.53. Interestingly, in the stellar mass range of $10.5\lt \mathrm{log}({M}_{\star }/{M}_{\odot })\lt 11.0$ they reported SFEs lower (i.e., longer depletion timescales) than expected from the scaling relation, although these galaxies only represent 30% of all the sources within this stellar mass range. The rest of the galaxies (including all those with ${M}_{\star }\gt {10}^{11}\,{M}_{\odot }$) are in agreement with the field relations. In line with these results, Noble et al. (2017) found depletion timescales systematically higher than the scaling relation for nine CO(3 − 2)-detected galaxies in a cluster at z ∼ 1.6, although most are within one standard deviation of the relation (plus several nondetections whose upper limits are in agreement with the field). Similarly, Hayashi et al. (2018) presented CO(3 − 2) observations in several cluster galaxies at z ∼ 1.5 finding again systematically longer depletion timescales, although ∼50% of their sample might be in agreement the expected relation for the field (when nondetections are taken into account).

Although our results suggest that most of the massive (${M}_{\star }\gt {10}^{10}\,{M}_{\odot }$) galaxies have SFEs similar to coeval field galaxies, it is clear that other works in the literature have revealed (proto)cluster galaxies with a large variety of properties, with some of them having SFEs similar to coeval field galaxies and others lying off the expected relations. Larger samples of high-redshift clusters and protoclusters, as well as deeper observations to mitigate nondetections, are hence required to fully understand the star formation activity and gas fueling in these overdense structures.

Figure 3 shows the combined gas mass function of the two protoclusters, which is formally a lower limit given the incompleteness of our follow-up survey (only a fraction of the known protocluster members were observed; see Section 2.1) and the possible existence of more unidentified member galaxies. This estimation assumes a volume of 15,000 cMpc³ for each structure, as estimated by Casey (2016). Note that this volume encompasses the z ≈ 2.47 overdensity studied here, but not the larger structure reported in the literature with a redshift range of z = 2.42–2.51 (Cucciati et al. 2018; although the volume would only increase by a factor of ∼2). The first bin of our gas mass function comes from the stacked detection of 20 galaxies (14 in the z = 2.10 structure and 6 in the 2.47; see Table 3), while the rest come directly from the individual detections.

Zoom In Zoom Out Reset image size

The derived lower limit of the protoclusters' gas mass function is of the same order of magnitude as the field gas mass function at these redshifts,¹² as shown in Figure 3). Therefore, it is likely that these protoclusters have an enhanced gas mass function or a higher gas volume density than the one measured in the field, after taking into account all the incompleteness described above. Indeed, enhanced gas volume densities have been measured for other similar overdense structures (e.g., Lee et al. 2017).

In Figure 4 we plot the derived gas mass fraction, ${f}_{\mathrm{gas}}={M}_{\mathrm{mol}}/({M}_{\mathrm{mol}}+{M}_{\star })$, as a function of stellar mass for the protocluster galaxies, and other samples taken from the literature, including our reference sample. As can be seen, most of the ALMA-detected galaxies, represented by the blue and red filled circles in the figure, show gas fractions that resemble those estimated for coeval field galaxies and their scaling relations,¹³ which is illustrated by the gray shaded region (see also Appendix A). The individual upper limits derived from the ALMA nondetections for those galaxies with M_⋆ ≲ 5 × 10¹⁰ M_⊙ are also consistent with the field relation, but those of the most massive galaxies point toward reduced gas fractions when compared to the field (see also Appendix A).

Zoom In Zoom Out Reset image size

Different works from the literature have also reported gas fractions in protocluster galaxies in agreement with the field. For example, Lee et al. (2017) found that most of the observed galaxies in a z = 2.49 protocluster show gas fractions comparable to those of field galaxies. Their detections are plotted in Figure 4 (gold triangles), where it can be seen that their values are in agreement with our measurements.

Figure 4 also shows the results from Tadaki et al. (2019) and Gómez-Guijarro et al. (2019) who observed galaxies in several overdensities within z = 2.17–2.60. Besides the longer gas depletion timescales found by Tadaki et al. (2019) for some galaxies with $10.5\lt \mathrm{log}({M}_{\star }/{M}_{\odot })\lt 11.0$, sources with high gas mass fractions were also found (although they represent ≲50% of all the galaxies within this stellar mass range). Most of the remaining sources, and all the galaxies with M_⋆ ≳ 10¹⁰ M_⊙, have gas fractions (or upper limits) in agreement with the field relation. Similarly, Gómez-Guijarro et al. (2019) found some galaxies with large gas fractions (all of them with M_⋆ ≲ 6 × 10¹⁰ M_⊙) despite showing SFEs consistent with the field. Nevertheless, most of these gas-rich galaxies lie above the main sequence, and therefore, might be more representative of the starburst population. Galaxies with enhanced gas fractions have also been reported in more evolved (collapsed) clusters at lower redshifts, like those reported by Noble et al. (2017) and Hayashi et al. (2018) at z ∼ 1.5.

While some protoclusters show most of their members with gas fractions in agreement with coeval field galaxies, as those presented in this work (see also Lee et al. 2017), there is clear evidence that other structures have individual systems with large gas masses (e.g., Gómez-Guijarro et al. 2019; Tadaki et al. 2019), as discussed above. Discriminating between if these gas-rich systems exist preferentially in overdense environments or if they represent only outliers of the general population (note that this kind of outlier also exists in the field) requires deeper observations of complete samples, even deeper than the ones presented here. For example, the results of Noble et al. (2017), who present one of the best lines of evidence for the existence of sources with enhanced gas masses, are based on the detection of only seven individual sources from a parent sample of 49 cluster members. Similarly, the gas-rich galaxies found by Hayashi et al. (2018) come from 18 detections out of a parent sample of 65 sources. Although cluster-to-cluster variations, reflecting different evolutionary stages (e.g., Shimakawa et al. 2018; Gómez-Guijarro et al. 2019) are expected, it is still unclear if the bulk of the population in the aforementioned cluster also have enhanced gas fractions. Unfortunately, deeper ALMA spectroscopic observations would require several hours of on-source time per target, making spectroscopic observations of larger samples prohibitive. Continuum observations as a proxy for gas masses (Scoville et al. 2016) will therefore play a determinant role in our understanding of the star formation activity during the early phases of cluster assembly.

As shown in Figure 4, some of the most massive galaxies studied in this work show very low gas mass fractions, f_gas ≲ 6%–10% (see also Wang et al. 2018), resembling those estimated for quiescent galaxies (note, however, that most of the constraints on the gas fraction of passive galaxies are limited to z < 2; e.g., Sargent et al. 2015; Gobat et al. 2018; Spilker et al. 2018; Bezanson et al. 2019). These massive gas-poor galaxies might be in a transitional phase toward a quiescent mode, and hence, they are ideal sources to study the quenching mechanisms, especially the effects of environmental quenching.

In Figure 1 we highlight the most massive gas-poor systems (${M}_{* }\gt 4\times {10}^{10}\,{M}_{\odot }$) with black squares. As can be seen, most of them (seven out of nine) lie below the star-forming main sequence, supporting a possible quenching phase. The galaxy that lies above the main sequence is known to host an AGN (based on an X-ray detection; Laigle et al. 2016), and therefore it is likely that its SFR might be overestimated (note that only six of the whole sample are classified as AGN, as mentioned below).

To further explore their possible passive nature, Figure 5 shows the rest-frame U − V color of our sample of protocluster galaxies, which has been proven to be a good indicator to disentangle the quiescent and star-forming populations of galaxies via blue and red populations (e.g., Bell et al. 2004; Brown et al. 2007; Faber et al. 2007). This figure clearly shows that the most massive galaxies undetected by ALMA have red colors, as would be expected if they are in the quenching process.

Zoom In Zoom Out Reset image size

In Figure 6 we show instead the rest-frame (U − V) versus (V − J) color–color diagram. This UVJ plot is commonly used to distinguish between dusty, star-forming galaxies and quenched galaxies since both populations usually have red colors (e.g., Wuyts et al. 2007; Williams et al. 2009). While it is true that the massive gas-poor galaxies described above are not entirely within the quiescent parameter space, they lie very close to it. Additionally, it as been shown that some passive galaxies remain outside of the UVJ selection area well after their actual quenching (up to almost 0.5 Gyr), particularly, for those in which the SFR decreases abruptly, called post-starburst galaxies (Merlin et al. 2018; Belli et al. 2019). To illustrate this, we plot in Figure 6 the expected path on the UVJ diagram of a galaxy with a fast quenching process described by a tau model with a short timescale of τ = 100 Myr (see details in Belli et al. 2019). The colors predicted by this scenario (see also Merlin et al. 2018 for similar results) are in agreement with those measured for some of the massive gas-poor galaxies in our sample. The well-studied post-starburst galaxy at z ≈ 3 reported by Marsan et al. (2015), has, actually, very similar colors (see Figure 6).

Zoom In Zoom Out Reset image size

These findings, combined with the ALMA nondetections, suggest that the massive gas-poor sources shown in Figure 4 are evolved galaxies, most likely post-starburst or in transition to a quiescent mode.

Is the evolution of these high-mass gas-poor transition galaxies driven by the overdense protocluster environment? Quiescent massive galaxies have been found in the field up to z ≳ 3 (e.g., Straatman et al. 2014; Glazebrook et al. 2017), and therefore a comparison between the field and these protocluster galaxies is necessary to understand the impact of the environment on the quenching process. In Figure 7, we compare the ALMA continuum detection fraction of our protocluster galaxies to the one derived for our control sample (Scoville et al. 2016) and also to those found in blind blank-field observations (Bouwens et al. 2016; Dunlop et al. 2017), limiting these samples to galaxies within z ≈ 2–3. Note that all the observations have similar depths,¹⁴ making the comparison appropriate.

Zoom In Zoom Out Reset image size

Interestingly, the observations toward protocluster member galaxies show a lower detection fraction than those for the field samples, particularly for the high-mass galaxies (Figure 7). This is also true even if we only compare our results to those from the blind blank-field surveys, which have no biases due to selection effects. This difference is found in both the z = 2.10 and the z = 2.47 protocluster structures. These results imply that the most massive galaxies in protocluster environments experience an accelerated evolution compared to the field sources, given the higher fraction of quiescent gas-poor galaxies.

We also estimate an average environmental quenching efficiency, defined as ${\epsilon }_{q}\equiv ({f}_{\mathrm{passive},\mathrm{dense}}-{f}_{\mathrm{passive},\mathrm{field}})/{f}_{\mathrm{SFG},\mathrm{field}}$, which represents the fraction of field galaxies required to be quenched in order to match the observed quenching fraction in the protoclusters. For the two structures studied here, we estimate an average quenching efficiency of ${\epsilon }_{q}={0.45}_{-0.13}^{+0.16}$. This quenching efficiency is indeed in agreement with those found for more evolved clusters at z ≈ 1.0−1.6 (see the compilation by Nantais et al. 2016). All these results suggest that we are witnessing the first stages of the well-known environmental quenching found in lower redshift and/or more evolved clusters (e.g., Tran et al. 2009; Alberts et al. 2014; Newman et al. 2014; Nantais et al. 2016; Lee-Brown et al. 2017; Ji et al. 2018; Noirot et al. 2018; Paulino-Afonso et al. 2018; Shimakawa et al. 2018), which originates the excess of red sources in local galaxy clusters (Dressler 1984).

Understanding the mechanisms that drive this evolution requires further observations than are included here. However, it is clear that these mechanisms should involve gas consumption or removal, or at least a halting of gas accretion, given the small gas fractions derived for these red galaxies (see also Man et al. 2019). Note that counterexamples to this exist, i.e., passive galaxies with significant gas reservoirs (e.g., Rowlands et al. 2015; Suess et al. 2017; Gobat et al. 2018), although in a variety of galaxy environments from isolated systems to small groups. Some of the typically cited quenching processes in high-density environments are AGN feedback (e.g., Kauffmann et al. 2003; Fabian 2012; Shimakawa et al. 2018), major and minor mergers (e.g., Davis et al. 2019; Maltby et al. 2018), ram pressure stripping (meaning gas being removed from the galaxies by the IGM, e.g., Gunn & Gott 1972; Larson et al. 1980; Boselli & Gavazzi 2006; Foltz et al. 2018, although see Dannerbauer et al. 2017), and starvation (when gas accretion is stopped).

In Figure 5, we identify those galaxies that are likely to host an AGN based on their detection with Chandra X-ray Observatory (using the match reported by Laigle et al. 2016 to the Chandra COSMOS catalogs by Civano et al. 2016; Marchesi et al. 2016). Only a few galaxies are X-ray detected (all of them have M_⋆ ≳ 3 × 10¹⁰ M_⊙; see Tables 1 and 2), and there is not a clear trend between AGN activity and quenching since some of them are detected by ALMA while others are not. Nevertheless, we cannot rule out this mechanism as the main quenching processes given the incompleteness related to the AGN identification.

To shed light on other quenching mechanisms possibly involved, we estimate an upper limit on the quenching timescale adopting the average mass-weighted stellar age of the massive gas-poor galaxies derived from the SED fitting (see Section 3.2). Given that this quantity is related to the time since the onset of star formation,¹⁵ any quenching mechanism must have a shorter timescale. To be conservative, we adopt the longest age from the 97.5th percentile of the stellar age distribution, resulting in an upper limit on quenching timescale of ${\tau }_{q}\lt 1\,\mathrm{Gyr}$. Similar timescales have been reported for passive galaxies in lower redshift clusters (e.g., Muzzin et al. 2014; Davis et al. 2019; Socolovsky et al. 2018) and are usually associated with ram pressure stripping and/or mergers (e.g., Steinhauser et al. 2016). In line with these findings, Casey et al. (2015) and Hung et al. (2016) found slightly higher fractions of irregular and interacting galaxies in the z = 2.47 and 2.10 structures than in their control samples (although with a low statistical significance of ∼1.5σ), supporting the influence of galaxy mergers in the evolution of galaxies in overdense environments.

Given that the protocluster structures studied here are still in an assembly phase, these results indicate galaxy preprocessing, meaning the interactions between groups of galaxies prior to the settling in a cluster potential, might be also an important quenching mechanism (see also Zabludoff et al. 1996; Haines et al. 2015; Bianconi et al. 2018; Olave-Rojas et al. 2018).

Alternatively, the advanced evolutionary stage of some massive protocluster member galaxies can be explained by an earlier onset of the star formation activity (e.g., Steidel et al. 2005). Nevertheless, the mean stellar age of the red, massive gas-poor galaxies is consistent with the ages derived for the ALMA-detected galaxies (when using the same stellar mass threshold). Similarly, the stellar ages derived for the Scoville et al. (2016) sample, representative of the average field population, show values in agreement with the protocluster galaxies studied here. Although we cannot rule out specific sources with premature star formation, we conclude that environmental effects, as those discussed above, most likely nurture an accelerated evolution in these systems.