The Grism Lens-Amplified Survey from Space (GLASS). VIII. The Influence of the Cluster Properties on Hα Emitter Galaxies at 0.3 < z < 0.7

GLASS is a 140-orbit slitless spectroscopic survey with HST in cycle 21. It has observed the cores of 10 massive galaxy clusters targeted by the Hubble Frontier Fields (HFF; P.I. Lotz, Lotz et al. 2016) and by the Cluster Lensing and Supernova Survey with Hubble (CLASH; P.I. Postman, Postman et al. 2012) with the Wide-Field Camera 3 (WFC3) near-infrared (NIR) grisms G102 and G141 providing an uninterrupted wavelength coverage from 0.8 to 1.7 μm. Each cluster was observed at two position angles (PAs) approximately 90° apart to facilitate clean extraction of the spectra for objects in the crowded cluster fields. The sample of 10 clusters and their properties are presented in Table 1.

Table 1.  Cluster Properties

Cluster	Short	R.A.	Decl.	z	Phys Scale	${L}_{{\rm{X}}}$	${M}_{500}$	${r}_{500}$	PA1	PA2
	Name	(J2000)	(J2000)		(kpc/'')	(1044 erg s−1)	(1014 ${M}_{\odot }$)	(Mpc)
Abell2744	A2744	00:14:21.2	−30:23:50.1	0.308	4.535	15.28 ± 0.39	17.6 ± 2.3	1.65 ± 0.07	135	233
RXJ2248.7–4431	RXJ2248	22:48:44.4	−44:31:48.5	0.346	4.921	30.81 ± 1.57	22.5 ± 3.3	1.76 ± 0.08	053	133
Abell370	A370	02:39:52.9	−01:34:36.5	0.375	5.162	8.56 ± 0.37	11.7 ± 2.1	1.40 ± 0.08	155	253
MACS0416.1–2403	MACS0416	04:16:08.9	−24:04:28.7	0.420	5.532	8.11 ± 0.50	9.1 ± 2.0	1.27 ± 0.09	164	247
RXJ1347.5–1145	RXJ1347	13:47:30.6	−11:45:10.0	0.451	5.766	47.33 ± 1.2	21.7 ± 3.0	1.67 ± 0.07	203	283
MACS1423.8+2404	MACS1423	14:23:47.8	+24:04:40	0.543	6.382	13.96 ± 0.52	6.64 ± 0.88	1.09 ± 0.05	008	088
MACS1149.6+2223	MACS1149	11:49:36.3	+22:23:58.1	0.544	6.376	17.25 ± 0.68	18.7 ± 3.0	1.53 ± 0.08	032	125
MACS0717.5+3745	MACS0717	07:17:31.6	+37:45:18	0.546	6.400	24.99 ± 0.92	24.9 ± 2.7	1.69 ± 0.06	020	280
MACS2129.4–0741	MACS2129	21:29:26.0	−07:41:28.0	0.589	6.524	13.69 ± 0.57	10.6 ± 1.4	1.26 ± 0.05	050	328
MACS0744.9+3927	MACS0744	07:44:52.8	+39:27:24.0	0.686	7.087	18.94 ± 0.61	12.5 ± 1.6	1.27 ± 0.05	019	104

Note. J2000 coordinates, redshift, physical scale, X-ray luminosity, M₅₀₀ (from Mantz et al. 2010), r₅₀₀, and the two position angles.

Download table as:  ASCIITypeset image

Details on the observations and data reduction can be found in Schmidt et al. (2014) and Treu et al. (2015). Briefly, observations follow the dither pattern used for the 3D-HST observations and were processed with an updated version of the 3D-HST reduction pipeline¹⁵ described by Brammer et al. (2012) and Momcheva et al. (2016). All spectra were visually inspected with the publicly available GLASS inspection GUI, GiG¹⁶ (Treu et al. 2015), in order to identify and flag erroneous models from the reduction, assess the degree of contamination in the spectra, and flag and identify strong emission lines and the presence of a continuum.

As described in Treu et al. (2015), to determine redshifts, templates were compared to each of the four available grism spectra independently (G102 and G141 at two PAs each) to compute a posterior distribution function for the redshift. Then, with the help of the publicly available GLASS inspection GUI for redshifts (GiGz; Treu et al. 2015), we flagged which grism fits are reliable or alternatively entered a redshift by hand if the redshift was misidentified by the automatic procedure. Using GiGz we assigned a quality Q_z to the redshift (4 = secure; 3 = probable; 2 = possible; 1 = tentative, but likely an artifact; 0 = no z). These quality criteria take into account the signal-to-noise ratio of the detection, the possibility that the line is a contaminant, and the identification of the feature with a specific emission line. This procedure was carried out independently by at least two inspectors per cluster (see Treu et al. 2015 for details).

The full redshift catalogs from the inspection of the 10 GLASS clusters are available at https://archive.stsci.edu/prepds/glass/.

We make use of all 10 GLASS clusters. Virial radii ${r}_{500}$ have been computed from virial masses ${M}_{500}$ taken from Mantz et al. (2010):

$$\begin{eqnarray*}&&{r}_{500}=\sqrt[3]{\displaystyle \frac{3}{4\pi }\displaystyle \frac{{M}_{500}}{500{\rho }_{\mathrm{cr}}}},\end{eqnarray*}$$

where ${\rho }_{\mathrm{cr}}=\tfrac{3{H}^{2}}{8\pi G}=\tfrac{3{H}_{0}^{2}}{8\pi G}\times [{{\rm{\Omega }}}_{{\rm{\Lambda }}}+{{\rm{\Omega }}}_{0}\times (1+z{)}^{3})]$, with G being the gravitational constant = $4.29\times {10}^{-9}{(\mathrm{km}{\rm{s}}}^{-1})$ Mpc ${M}_{\odot }$.

Cluster-centric distances have been computed in units of r₅₀₀ from the peak of the X-ray distribution. For merging clusters, where more than one peak in the X-ray distribution can be detected, distances have been computed from the closest peak. Cluster mass maps were produced using the SWUnited reconstruction code described in detail in Bradač et al. (2005) and Bradač et al. (2009). The method uses both strong and weak lensing mass reconstruction on a nonuniform adapted grid. From the set of potential values, we determine all observables (and mass distributions) using derivatives. The potential is reconstructed by maximizing the log likelihood, which uses image positions of multiply imaged sources, weak lensing ellipticities, and regularization as constraints. Our team has at its disposal the cluster mass maps for all clusters, except for MACS0744, which is not ready yet. The X-ray images are based on Chandra data and are described in Mantz et al. (2010) and von der Linden et al. (2014). For the contours, the images have been adaptively smoothed after removing point sources identified in Ehlert et al. (2013). X-ray images are available for all clusters. To estimate the X-ray emission at the location of the galaxy, we masked the galaxy itself (which can emit in X-rays) and computed the average signal in an annulus around the galaxies with inner radius 2'' and outer radius 5''.

The local density of a galaxy is defined as the number of its neighbors per unit projected area: ${\rm{\Sigma }}=N/A$ in number of galaxies per Mpc⁻². The projected number densities have been estimated from the circular area containing the five closest objects: ${\rm{\Sigma }}=(5+1)/\pi {r}_{5}^{2}$, with r the radius of such an area (see Morishita et al. 2017). Local density estimates are available for only four clusters in our sample, A2744, MACS0416, MACS0717, and MACS1149, which are the first HFF clusters with complete data.

3.1.1. Hα Maps

3.1.2. Additional Galaxy Properties

We characterize the spatial distribution of the Hα emitters in terms of cluster-centric distance. We note that since the GLASS data set does not yield redshifts for passive galaxies, we cannot characterize the spatial distribution of all cluster members. The upper left panel of Figure 1 shows that galaxies are located within ∼0.5r₅₀₀, which roughly corresponds to the maximum coverage of all of the clusters, and do not seem to avoid the cluster cores, even though there might be possible projection effects. The distribution peaks around 0.2 r₅₀₀. We distinguish between relaxed (MACS1423, RXJ1347, MACS2129, RXJ2248, MACS0744, for a total of 38 galaxies) and merging or unrelaxed (MACS1149, MACS0717, A2744, MACS0416, A370, for a total of 38 galaxies) clusters. Since there is no unique and clear criterion to distinguish between the two categories, we assume that in unrelaxed clusters more than one X-ray peak is detected, as will be discussed in Section 4.2. Galaxies in unrelaxed clusters tend to be located closer to the cluster center than galaxies in relaxed clusters. Recall that for merging systems, where more than one peak in the X-ray distribution can be detected, distances have been computed from the closest peak. The median value for the former is 0.17 ± 0.01 and for the latter is 0.30 ± 0.02. We have also checked for mass segregation and computed the mean and median galaxy masses in bins of distance. We found that the typical stellar mass is similar at all distances from the cluster center, suggesting that the mass build-up is not very sensitive to the position of the galaxy in the cluster.

Zoom In Zoom Out Reset image size

The upper right panel of Figure 1 quantifies the relation between the radial offset (i.e., the offset between the peak of the Hα emission and the peak in the F475W filter projected along the cluster radial direction) and the distance of the galaxy from the cluster center. Most of the galaxies have offsets within ±0.5 kpc, but there are some showing a larger offset. The typical uncertainty on the offset estimates is ∼0.1 kpc. When considering the entire galaxy population as a whole, no dependencies on the cluster-centric distances are detected (Spearman correlation = −0.008 with 94% significance). Galaxies in relaxed and unrelaxed clusters have similar offsets. As also seen in the bottom left panel, 60% of cluster Hα emitters have a negative radial projected offset, and the distribution is clearly shifted toward negative values (the median of the distribution is −0.14 ± 0.07 kpc), indicating that for most of the galaxies the Hα peak points away from the cluster center. This finding might suggest that our galaxies are approaching the cluster center for the first time, and the weakly bound gas is left behind. However, the analysis of the skewness does not support the result: the ratio of the skewness to the standard error of skewness (SES)¹⁹ is 0.23/0.24 ∼ 0.83, suggesting that population data are neither positively nor negatively skewed. We will revisit and try to test quantitatively this hypothesis in the next section. In contrast, the distribution of the tangential offset peaks around zero (the median of the distribution is 0.01 ± 0.07 kpc, skewness/SES = 0.15/0.24 ∼ 0.62), indicating no preferential direction. A Kolmogorov-Smirnov test confirms that the two distributions are different (i.e., 4% probability of being drawn from the same parent distribution). If we consider only the 39/76 galaxies for which we have two orthogonal spectra and therefore the offset is better constrained, we find the same trends, indicating our results are robust against uncertainties. No strong differences are found for galaxies in relaxed and unrelaxed clusters.

The bottom right panel of Figure 1 correlates the tangential to the radial offset. As already noticed, there is no preferential direction for the tangential offset, while the radial offset is directed away from the cluster center. No differences emerge for relaxed and unrelaxed clusters

Galaxies with different Hα morphologies and experiencing different physical processes are highlighted in the upper right panel of Figure 1. Ram-pressure stripped galaxies with asymmetric morphology are preferentially found between 0.1 and 0.3 r₅₀₀ and tend to have negative radial offset, indicating that in these galaxies the Hα distribution is strongly influenced and shows a systematically different distribution than in the existing stellar population. In contrast, galaxies of the other types are not clustered.

4.1.1. Comparison of Galaxy Infall to Cosmological Simulations

In the previous section, we have found that the magnitude of the offset might give us an indication of the process operating on galaxies. In addition, it might also carry information about the orbit along which a galaxy is traveling through the ICM. In particular, if the offset is due to ram pressure, its direction is expected to trace the direction of the galaxy velocity. Therefore, the ratio ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ between the radial and the tangential offsets can be taken as a proxy for the ratio ${v}_{r}/| {v}_{\theta }|$ between the radial and the one-dimensional tangential components of the galaxy velocity at the time of the observation (v_r and v_θ are two of the three components of the velocity vector in spherical coordinates). As ${\mathrm{off}}_{r}$ is defined so that it is positive when the peak of the Hα emission is closer to the cluster center than the continuum emission, we expect that an infalling galaxy (${v}_{r}\lt 0$) has negative ${\mathrm{off}}_{r}$. In this section, we compare our observed radial and tangential offsets to the cosmological predictions of the orbits of satellites infalling onto galaxy clusters. For simplicity in what follows we just identify ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ with ${v}_{r}/| {v}_{\theta }|$, neglecting all possible sources of difference between the two quantities (for instance, the offset ratio is a projected quantity, while the velocity ratio is an intrinsic quantity). We note that the proxy is an underestimate of the real offset, since we would not measure any offset for objects that are infalling along the line of sight, yet they would have large ${v}_{r}/{v}_{\theta }$.

As a reference for the cosmological predictions, we take the results of Jiang et al. (2015), who studied the distribution of the orbital parameters of infalling satellite halos in a Λ cold dark matter (ΛCDM) cosmological N-body simulation. In particular, Jiang et al. (2015) provide the distributions of $V/{V}_{200}$ and ${V}_{r}/V$ as functions of host-halo mass and satellite-to-host halo mass ratio, where V is the satellite's speed at r₂₀₀, V_r is the radial component of the satellite's velocity at r₂₀₀, and V₂₀₀ is the host-halo circular velocity at r₂₀₀. Jiang et al. (2015) parameterized the distribution of $V/{V}_{200}$ with the three dimensionless parameters γ, σ, and μ and the distribution of ${V}_{r}/V$ with the dimensionless parameter B (see Section 3.4 in that paper). Here we fix $\gamma =0.05$, $\sigma =0.118$, $\mu =1.236$, and B = 3.396, which are the best-fitting values for host-halo mass ${10}^{14}{M}_{\odot }$ and satellite-to-host mass ratio 0.05–0.005 from Jiang et al. (2015) (note, however, that our results are not strongly dependent on this specific choice).

Assuming that the host halo is spherical and exploiting the fact that energy and angular momentum are conserved (neglecting tidal stripping and dynamical friction), for each orbit of given $V/{V}_{200}$ and ${V}_{r}/V$, it is straightforward to compute the ratio ${v}_{r}/| {v}_{\theta }|$ at each radius $r\lt {r}_{200}$. Specifically, we assumed that the host halo has a Navarro–Frenk–White (Navarro et al. 1996) density distribution with concentration ${c}_{200}=4$ (in this case ${r}_{500}/{r}_{200}\simeq 0.65$).

As a first comparison between the cosmological predictions and the observations, we look at the behavior of the offset and velocity ratios as functions of distance from the cluster center (for simplicity here we identify the projected observed cluster distance with the intrinsic orbital radius r). In Figure 2 the radial distribution of the observed offset ratios is compared with a few orbits characteristic of the cosmological orbit distribution. Since Jiang et al. (2015) find that the distribution of $V/{V}_{200}$ is relatively narrow, for simplicity, to compute the theoretical curves in Figure 2, we fix $V/{V}_{200}=1.236$ (the average value of the best fit of the distribution), and we sample the distribution in ${V}_{r}/V$ by selecting orbits corresponding to the 5th, 25th, 50th, 75th, and 95th percentiles. Figure 2 shows, as expected, that the observed sample (confined within $r/{r}_{500}\lt 0.5$) traces only the $\approx 25 \%$ most radial orbits of the cosmological distribution: the bulk of the cosmological orbits do not plunge deep enough into the cluster potential. One caveat is that in the analysis above we have neglected dynamical friction. However, this effect is negligible, at least for the first pericentric passage, as we verified by running N-body simulations for all of the orbits represented in Figure 2. In these simulations, run with the collisionless N-body code fvfps (Londrillo et al. 2003; Nipoti et al. 2003), we have followed, starting from r₅₀₀, the orbit of an infalling galaxy, represented as a particle with mass $0.005{M}_{200}$, in an isotropic NFW halo with concentration ${c}_{200}=4$, realized with $N\simeq {10}^{6}$ particles (M₂₀₀ is the total mass of the halo, which is truncated exponentially at r₂₀₀). The setup and technical characteristics of these simulations are identical to those described by Nipoti (2017).

Zoom In Zoom Out Reset image size

Figure 2 suggests that the observed Hα cluster galaxies might represent the radial-orbit selected tail of the distribution of cosmological satellites. If the cosmological prediction is correct, and if ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ is a proxy for ${v}_{r}/| {v}_{\theta }|$, at each radius, ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ should be distributed as the predicted ${v}_{r}/| {v}_{\theta }|$. To verify whether this is actually the case, we select from among the observed galaxies only the subsample (44 galaxies) with ${\mathrm{off}}_{r}\lt 0$, which, under our hypothesis, are infalling galaxies, for which the mapping between ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ and ${v}_{r}/| {v}_{\theta }|$ should be more justified. In principle, ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ might also be a proxy for ${v}_{r}/| {v}_{\theta }|$ for galaxies that are receding from the center of the cluster (${v}_{r}\gt 0$), but this model is too simple to describe a system that has already passed the pericenter. For comparison with this subsample, we generate a sample of 4400 mock galaxies with the same radial distribution. The orbital parameters of these mock galaxies are extracted from the distributions of $V/{V}_{200}$ and ${V}_{r}/V$ given by Jiang et al. (2015) with the values of the parameters reported above. The mock sample of galaxies can be used to numerically compute the distribution of ${v}_{r}/| {v}_{\theta }|$ at each observed radius to be compared with the observed values of ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$. In order to verify whether the observed and mock samples are consistent, we first compute the probability distribution of ${\mathrm{off}}_{r}/| {\mathrm{off}}_{\theta }|$ for the 44 observed galaxies and the probability distribution of ${v}_{r}/| {v}_{\theta }|$ for the 4400 mock galaxies. From Figure 3(a), where these distributions are plotted, it is apparent that there is qualitative agreement between the observed and theoretical histograms. When the cumulative distributions are considered (Figure 3(b)), there appears to be a discrepancy for large values of $| {\mathrm{off}}_{r}/{\mathrm{off}}_{\theta }|$ and $| {v}_{r}/{v}_{\theta }|$ ($x\lesssim -5$), but this discrepancy is not statistically significant because the number of observed galaxies in this tail of the distribution is small (see Figure 3(c)). This can be quantified with a K–S test, which gives a probability of 51% that the two samples are extracted from the same parent population. As a further statistical test, for each galaxy of the sample, we computed how it ranks within the distribution of mock galaxies at the same radius. If the distributions are consistent, the quantiles must be distributed uniformly. According to the K–S test, the probability that the quantiles are extracted from a uniform distribution is 25%.

Zoom In Zoom Out Reset image size

We repeated the above analysis for the subsample of 12 galaxies with ${\mathrm{off}}_{r}\lt 0$ that we visually classified as being affected by ram-pressure stripping (see Section 3.1.2, stars in Figure 2), for which our model is expected to work best (in this case we created a mock sample of 1200 galaxies). We find again that the cumulative distribution of the offset ratios (Figure 3(b)) is consistent with the theoretical expectation, as supported by the K–S test, which gives a probability of 71% that the two samples (mock and observed) are extracted from the same parent population. In this case, the probability that the distribution of the quantiles is uniform is 48%.

Bearing in mind the small sample size, we conclude that the observed distribution of offset ratios for the infalling galaxies is consistent with the cosmological predictions. This finding is even more significant when we consider only the infalling galaxies we labeled as being affected by ram-pressure stripping, so the cosmological predictions support our classification scheme.

Over the last years, there has been increasing evidence for a correlation between the efficiency of the stripping phenomenon and the presence of shocks and strong gradients in the X-ray intergalactic medium (e.g., Owers et al. 2012; Vijayaraghavan & Ricker 2013). In Paper V we found tentative trends between the X-ray counts and the radial offset, even though the correlations were not supported by statistical tests.

Figure 4 presents the color composite images of all of our clusters along with X-ray maps. Clearly, the clusters in our sample present very different X-ray emission morphologies: RXJ1347, RXJ2248, MACS1423, MACS2129, and MACS0744 show quite symmetric emissions and are relaxed, while A2744, A370, MACS0416, MACS1149, and MACS0717 have more than one main peak and extend along the north–south direction (A370), the northwest–southeast direction (A2744, MACS0717, MACS1149), or the northeast–southwest direction (MACS0416). Hα emitters with different Hα morphologies and experiencing different processes are highlighted.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Galaxies with all kinds of Hα morphologies and also experiencing all of the proposed physical processes are found in almost all clusters. Due to the low number statistics in each cluster, it is hard to detect solid trends with morphology and acting process. In MACS1423, Hα emitters are almost all at the same cluster-centric distance, where the hot gas density is nearly constant. However, this is not the case for the other relaxed clusters. In A2744 and MACS0717, characterized by multiple centers, Hα emitters tend to lie all in the same region of the cluster and avoid the second peak. In MACS0416 there are only two Hα emitters, so no solid conclusion can be drawn. The same is true for RXJ1347 and RXJ2248.

It is worth noting that some galaxies are found to correspond with a peak in the X-ray distribution. An example can be found in MACS0744. However, in this case, the Hα morphology seems not to be affected by the peculiar position of the galaxy: indeed, it has been visually classified as a galaxy with a regular Hα morphology where no strong process is occurring. We remind the reader that the classification has been performed blindly with respect to the cluster properties.

We note that we cannot know the exact three-dimensional locations of the galaxies with respect to the ICM structures, but in some cases the small projected distances from the X-ray peaks and shocks suggest that some galaxies may have recently been overrun by the shock-front subcluster gas. This indicates that a mechanism related to an interaction with these ICM features may be in some cases responsible for either the stripping of the gas or the triggering of the star formation, or both.

Similarly, Figure 5 shows the color composite images of nine of our clusters for which the surface mass density maps are available (see Section 2.2). These maps, based on lens modeling, provide an estimate for the total mass density of the cluster, composed mostly of invisible dark matter. Also from these maps the variety of structures in our sample emerges: A370, MACS0717, MACS1149, A2744, and MACS0416 present more than one peak in their distribution, the former extending along the northwest–southeast direction, the latter along the northeast–southwest direction. In contrast, MACS1423, MACS2129, RXJ2248, and RXJ1347 show nearly symmetric mass distributions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 6 correlates the projected radial offset with both the X-ray emission and the surface mass density. X-ray surface brightness has been corrected for cosmological dimming ($\propto \,{(1+z)}^{4}$). In both cases, Spearman rank-order correlation tests show that no correlation is present between these quantities (Spearman correlation = 0.004 with 80% significance). However, if we consider only galaxies in unrelaxed clusters, a weak correlation seems to emerge, in the sense that galaxies at higher X-ray counts and surface mass densities tend to have more negative offsets. The Spearman correlation test supports these findings. This result might suggest that in merging systems X-ray counts are a proxy for mergers between substructures (see, e.g., Poggianti et al. 2004), while in the relaxed ones they simply trace the density of the ICM, without inducing an alteration in galaxy properties.

Zoom In Zoom Out Reset image size

In addition, galaxies in unrelaxed clusters tend to be located systematically at higher X-ray counts and tend to avoid lower surface mass densities than galaxies in all clusters, indicating that in merging systems the gas temperature and the total dark matter are larger.

Trends with Hα morphology or acting process are hardly detected in both relaxed and unrelaxed clusters. Few Hα asymmetric, ram-pressure stripped candidates are indeed found at high values of the X-ray emission (in agreement with Owers et al. 2012; Vijayaraghavan & Ricker 2013), but we find others at intermediate values. Conversely, unstripped galaxies are found at high values of X-ray emission.

In order to better quantify the impact that the hot gas or the cluster total mass can have on galaxy properties, we have investigated the distribution of morphologies and Hα morphologies as a function of both X-ray emission and surface mass density distribution (plots not shown). While at low values of X-ray counts and surface mass density galaxies of all morphological types exist, only ellipticals are found at high values of surface mass density, and ellipticals and spirals are found at large X-ray count values. Focusing on Hα properties, galaxies with a regular Hα disk seem not to avoid very dense regions, where asymmetric and clumpy objects are also found.

It would be interesting to investigate the hot gas and surface mass density ranges over which the different physical processes take place, but given the small size of our sample, significant trends cannot be detected.

To conclude, even though some trends are only tentative, cluster properties like the hot gas density or the dark matter distribution seem to have an impact on the Hα morphology, and thus on the location of ongoing star formation, only in unrelaxed clusters. The lack of strong correlations prevents us from identifying a unique, strong environmental effect that originates from the cluster center.

In the next section, we will investigate whether some local effects, uncorrelated to the cluster-centric radius, play a larger role.

Figure 7 correlates the 2D distance between the peak of the Hα emission and that of the continuum, as traced by the F475W, to the projected local galaxy density. More precisely, it considers the absolute value of the offsets (not projected along the cluster-centric distance) in the two directions (obtained from the two different PAs) and, for the galaxies with both PAs, the real distance between the two peaks, obtained by combining the offsets.

Zoom In Zoom Out Reset image size

The magnitude of the offset does not seem to correlate with the local density (Spearman correlation tests are always inconclusive), even though there might be an excess at intermediate values of local density. Nonetheless, some segregation effects between local density and Hα properties (different colors in Figure 7) are visible.

Galaxies with concentrated Hα seem to be preferentially found at lower densities, while galaxies with asymmetric Hα seem to prefer denser environments. Galaxies with regular and clumpy Hα are found at intermediate values of local density. K–S tests confirm that each population is drawn from a different parent distribution with high significance levels (>90%), except for galaxies with regular and clumpy Hα.

Regular processes seem to operate at low to intermediate densities, as is also true for mergers. In contrast, ram-pressure stripping and unidentified processes tend to operate also at higher densities. A K–S test can reject the null hypothesis that a regular process and ram-pressure stripping are drawn from the same distribution at $\sim 90 \%$ confidence.

To conclude, despite the statistics limited to four out of 10 GLASS clusters, trends with local densities are stronger than trends with the other tracers.

Understanding the origin of the trends of star formation with cluster properties represents a significant step toward comprehending the link between galaxy evolution and environment. If galaxy properties depend on the mass of the system where they reside or have resided during their evolution, there should be a connection between the trends observed and the way cosmological structures have grown in mass with redshift.

In Vulcani et al. (2010) and Paper VII we have found that the SFR–mass relation depends on environment: while many galaxies in clusters can be as star-forming as galaxies in the field, in the more massive systems, a population of galaxies with a reduced SFR at fixed mass is detected. This result indicates that some cluster-specific processes that suppress star formation are taking place.

Here we just focus on clusters and search for differences in the typical star-forming properties for galaxies living in different conditions. To remove the influence of the stellar mass, we consider the sSFR, defined as the SFR per unit of galaxy stellar mass. As shown in Figure 8, the mean sSFR seems to depend on neither global nor local environment.

Zoom In Zoom Out Reset image size

Considering separately galaxies with different Hα morphology or experiencing different processes, we see no trends emerge.

Putting together these results and those presented in Vulcani et al. (2010) and Paper VII, we conclude that while differences emerge when comparing the most different environments in the universe (clusters versus field), the properties of the clusters do not seem to strongly influence the star formation in the cluster members. However, all of these trends will need to be confirmed or refuted with larger number statistics.

We stress that we do not have all measurements for all of our clusters, so our sample could be affected by incompleteness. Indeed, in principle, results from the clusters not included in our sample could differ from those presented here. Nonetheless, our findings are based on a random subset of the whole sample, and therefore, for each measurement, the incompleteness is not related to cluster properties, so we do not expect strong biases.

We also note that our lack of trends is most likely due to the fact that, as discussed in Paper VII, our sample has been assembled by selecting visually detected Hα emitters and therefore includes only highly active star-forming objects, which most likely are at the first infall. Their SFR might have therefore not been affected yet by the dense cluster environment. To properly characterize the effect of the environment, one should focus on galaxies that have been part of the system for a long time and reach lower SFR levels.

4.4.1. Comparison with Previous Work