HST/WFC3 Grism Observations of z ∼ 1 Clusters: Evidence for Rapid Outside-in Environmental Quenching from Spatially Resolved Hα Maps

The Gemini Cluster Astrophysics Spectroscopic Survey (GCLASS, see Muzzin et al. 2012; van der Burg et al. 2013) was the spectroscopic follow-up of 10 massive clusters in the redshift range 0.86 < z < 1.34, drawn from the 42 deg² Spitzer Adaptation of the Red-sequence Cluster Survey (SpARCS, see Muzzin et al. 2009; Wilson et al. 2009; Demarco et al. 2010). Extensive optical spectroscopy was obtained with the Gemini Multi-Object Spectrographs (GMOS) on both Gemini-South and -North. A spectroscopic redshift was obtained for 1282 galaxies in total, with 457 of these being confirmed as cluster members. There is also 12-band photometry available for these clusters. The details of the photometry in 11 bands (ugrizJK_s , 3.6 μm, 4.5 μm, 5.8 μm, and 8.0 μm) are summarized in Appendix A of van der Burg et al. (2013). Newly acquired photometry in the F140W band (wide JH gap) has been obtained as part of the GCLASS HST follow-up (see Section 2.3). The public release of all GCLASS data products is now available, and detailed in Balogh et al. (2020). Physical properties of the 10 clusters in the GCLASS survey are summarized in Table 1.

Table 1. The 10 GCLASS Clusters Used in This Study

Name	zspec	M200	R200	σv	Spec-z Members	Grism-z	Total Members
		(1014 M⊙)	(kpc)	(km s−1)	in HST FOV	Members	in HST FOV
SpARCS-0034	0.867	2.0 ± 0.8	888 ± 110	${609}_{-66}^{+75}$	37/33	17/16	54/49
SpARCS-0035	1.335	5 ± 2	977 ± 154	${941}_{-137}^{+159}$	22/22	26/18	48/40
SpARCS-0036	0.869	5 ± 2	1230 ± 129	${911}_{-90}^{+99}$	43/41	31/30	74/71
SpARCS-0215	1.004	3 ± 1	953 ± 103	${758}_{-77}^{+85}$	42/40	32/28	74/68
SpARCS-1047	0.956	3 ± 1	926 ± 138	${680}_{-86}^{+98}$	22/19	15/14	37/33
SpARCS-1051	1.035	1.2 ± 0.5	705 ± 102	${530}_{-65}^{+73}$	34/32	11/10	45/42
SpARCS-1613	0.871	13 ± 3	1663 ± 130	${1232}_{-93}^{+100}$	73/65	41/25	114/90
SpARCS-1616	1.156	3 ± 1	854 ± 107	${701}_{-73}^{+81}$	38/35	19/17	57/52
SpARCS-1634	1.177	4 ± 2	1008 ± 131	${835}_{-82}^{+91}$	38/34	15/15	53/49
SpARCS-1638	1.196	1.9 ± 0.9	769 ± 117	${585}_{-65}^{+73}$	26/23	13/9	39/32

Note. For full names of the clusters, we refer to Muzzin et al. (2012). R₂₀₀ is the radius at which the mean interior density is 200 times the critical density of the universe. M₂₀₀ is the mass enclosed within this radius. σ_v is the line-of-sight velocity dispersion (see Biviano et al. 2016). The numbers listed in the last three columns indicate total numbers before/after quality checks relevant to Matharu et al. (2019) were applied. Mass completeness limits (see Section 4.1.3) are not applied to these sample numbers.

Download table as:  ASCIITypeset image

A grism is the combination of a diffraction grating and prism. In the context of the G141 grism, a series of ridges on the glass surface of a prism act as the diffraction grating, dispersing light from an object at an angle. The prism directs the dispersed light toward the WFC3 detector, keeping the light at some chosen central wavelength undeviated. The result is an image of the galaxy at different wavelengths, in 46.5 Å increments spanning 10750 < λ/Å < 17000. An unresolved emission line in this two-dimensional spectrum manifests itself as an image of the galaxy in that line on top of the underlying continuum. The high spatial resolution of the WFC3 (FWHM ∼ 0141 at 14000 Å) and the low spectral resolution of the G141 grism (R ∼ 130) allow for spatially resolved ($0.130\leqslant {\rm{FWHM}}/{\rm{arcsec}}\leqslant 0.156$ for 11000 ≤ λ/Å ≤ 17000) emission line maps (Nelson et al. 2016).

The wavelength coverage of the G141 grism is such that Hα emission can be detected in galaxies within the redshift range 0.7 < z < 1.5. This redshift range coincides perfectly with the redshift range of the GCLASS clusters (0.86 < z < 1.34). One spectral resolution element for a galaxy at z ∼ 1 corresponds to a velocity range of approximately 1000 km s⁻¹. The majority of galaxies have line widths well below this. Therefore, structure in the emission line maps is due to spatial morphology, not kinematics (Nelson et al. 2016).

Designed deliberately to be analogous to 3D-HST observations, the HST follow-up to GCLASS (GO-13845; PI: Muzzin) provides WFC3 F140W imaging and G141 grism spectroscopy for all of the clusters to a two-orbit depth. 90% of this time is spent on the grism spectroscopy (see Section 2.2) and the remaining 10% on F140W imaging. This was done to ensure we could obtain spatially resolved Hα maps of the star-forming cluster galaxies with a sufficient signal-to-noise ratio, such that stacks containing ∼30–40 galaxies could have their Hα profiles measured to ∼10% accuracy (Nelson et al. 2013). The exposure time of the F140W imaging for 9 of the 10 clusters is ∼800 s, with a 5σ F140W limiting magnitude for galaxies of ∼26.6. The exception to this is SpARCS-0035, which has a varying depth, ranging from ∼800 to ∼7460 s. ¹⁷ The exposure times of G141 grism spectroscopy for the GCLASS clusters range from 4312 s (SpARCS-0215) to 5312 s (SpARCS-1616) per pointing.

The HST observations for 9 of the 10 clusters are taken with a WFC3 two-pointing mosaic with a random orientation. For the tenth cluster, SpARCS-1047 (see Table 1), a single pointing is obtained covering the center of the cluster. The observations cover approximately ∼8 arcmin² for each cluster. This corresponds to approximately a quarter of the observing area covered by GMOS in the GCLASS survey (Section 2.1). Four of the GCLASS clusters (SpARCS-0035, SpARCS-1616, SpARCS-1634, and SpARCS-1638; see Table 1) have imaging going out to ∼ R₂₀₀. Despite this smaller overlap with GCLASS, spectroscopic density is highest in the cores of clusters. 86% of galaxies confirmed as cluster members in GCLASS are in the HST fields of view. Furthermore, grism redshifts derived from the G141 grism spectroscopy (Section 2.2) allow for the identification of an additional 182 cluster members, increasing the number of members per cluster by an average of 53% of the GCLASS spectroscopic sample (Matharu et al. 2019). Many of these cluster members were identified using the Hα emission line and therefore result in a much more complete sample of Hα emitters than from the spectroscopic sample alone. A breakdown of the number of cluster members for each cluster can be seen in Table 1.

The Grism redshift & line analysis software for space-based slitless spectroscopy (Grizli, ¹⁸ Brammer 2019) is used to reduce the G141 data and create the Hα line maps. Grizli is software that allows for the full processing of space-based slitless spectroscopic data sets.

A detailed description of how Grizli works can be found in Simons et al. (2020). Here, we summarize the details relevant to the study presented in this paper. Grizli uses the HST proposal ID (GO-13845; PI: Muzzin) to retrieve the data from the MAST archive. Routines detailed in Gonzaga et al. (2012), Brammer et al. (2015), Brammer (2016), and Momcheva et al. (2016) are used to process the WFC3 F140W imaging and G141 grism data for a variable sky background, hot pixels, cosmic rays, flat-fielding, and sky subtraction.

A full-field contamination model for each two-dimensional (2D) grism exposure in each pointing is created to subtract contamination from overlapping spectra of adjacent objects. This contamination model is created by modeling the HST G141 mosaic for each cluster in an iterative process. The first pass creates a contamination model for all sources in the mosaic with F140W magnitude < 25. This is done by taking the segmentation map (Figure 1(A)) for each source, assuming that the continuum spectra are normalized at F140W and flat in units of flux density, f_λ . Then a second-pass contamination model for all sources with F140W magnitude < 24 is created by subtracting the contamination model from the first pass and then fitting a third-order polynomial to the spectrum of each source, from brightest to faintest. The resulting static contamination model of all neighboring sources is subtracted from the 2D spectrum of a given object of interest.

Zoom In Zoom Out Reset image size

Grizli is used to extract and fit all sources in the GCLASS HST field of view with signal-to-noise ratio S/N > 7 in F140W photometry and with F140W magnitude <26. The number of sources extracted per cluster ranges from 573 (SpARCS-1047) to 1484 (SpARCS-0035). For each source, the redshift is determined by simultaneously fitting the 12-band multiwavelength photometry (see Section 2.1) and the G141 grism spectrum. This is done by including the photometric redshift p(z) as a prior on the grism spectrum redshift fit that is multiplied by the spectroscopic redshift p(z). ¹⁹

2.4.1. Grism Redshift Determination

As part of the grism redshift determination process (see Section 2.4.1), a combination of continuum templates and line complex templates are used. The line complex templates include [O ii] + [Ne iii], [O iii] + Hβ, and Hα + [S ii]. As well as breaking redshift degeneracies, the redshift determination process leads to the generation of a full continuum + line model for each spectrum in 2D. This then allows the user to create drizzled ²⁰ continuum-subtracted narrowband maps at any desired output wavelength. Emission line maps are therefore created by deliberately choosing the wavelength of the detected emission line from the redshift determination process (e.g., λ = 1.245 μm for Hα in Figure 1). The full World Coordinate System (WCS) information for each individual grism exposure of the source is used.

Emission line maps are generated by subtracting the best-fit continuum model with the assumption that the direct image (Figure 1(A)) represents the morphology of the source in the continuum. For GCLASS, these emission line maps are generated with a pixel scale of 01 and dimensions of 80 × 80 pixels. Example emission line maps for a galaxy in the final cluster sample of our study can be seen in Figure 1(D).

2.5.1. Cleaning Hα Maps

The star formation rates (SFRs) used in our study are calculated from total Hα fluxes determined by Grizli using the same method that was used for the calculation of Hα-derived SFRs from the 3D-HST G141 grism spectroscopy in Whitaker et al. (2014).

As in Nelson et al. (2016), to account for the blending of Hα and [N ii], the Hα flux is scaled down by the normalization factor 1 + f([N ii]/Hα) before SFRs are calculated. The rescaled Hα flux, Hα_rescaled, is defined as

$$\begin{eqnarray}&&{\rm{H}}{\alpha }_{{\rm{rescaled}}}=\displaystyle \frac{{\rm{H}}{\alpha }_{{\rm{measured}}}}{1+f([{\rm{N}}\,\mathrm{II}]/{\rm{H}}\alpha )}\end{eqnarray} \tag{ 1 }$$

When calculating SFRs for individual galaxies (see Figure 2), a fixed value of 0.15 was used for f([N ii]/Hα). For the Hα stacks, the same relation that was used in Nelson et al. (2016), which is from Sobral et al. (2015) and shown below, is used to obtain the flux ratio between the [N ii] and Hα emission, f([N ii]/Hα), for each stack:

$$\begin{eqnarray}&&f(\left[{\rm{N}}\,\mathrm{II}]/{\rm{H}}\right.\alpha )=-0.296\times {{\rm{log}}}_{10}({{\rm{EW}}}_{{\rm{H}}\alpha +\left[{\rm{N}}\mathrm{II}\right]})+\mathrm{0.8.}\end{eqnarray} \tag{ 2 }$$

Zoom In Zoom Out Reset image size

For the equivalent width of the Hα + [N ii] emission, EW_{Hα+[N ii]}, we use the Hα equivalent widths calculated by Grizli for each galaxy.

For the measurements shown in Figure 2, after rescaling the Hα flux for the [N ii] contribution, we calculate SFRs using the Hα flux conversion from Kennicutt (1998), but adapt it from a Salpeter initial mass function (IMF) to a Chabrier (2003) IMF following the method in Muzzin et al. (2010):

$$\begin{eqnarray}&&\displaystyle \frac{{\rm{SFR}}({\rm{H}}\alpha )}{{M}_{\odot }\,{{\rm{yr}}}^{-1}}=\displaystyle \frac{1.7\times {10}^{-8}\,{L}_{{\rm{H}}\alpha }}{{L}_{\odot }}.\end{eqnarray} \tag{ 3 }$$

Stellar masses are estimated using FAST (Kriek et al. 2009) with the redshift of each cluster galaxy fixed to the spectroscopic redshift of its cluster's center. These stellar masses are then corrected for the difference between the total F140W flux from the photometric catalog and the total F140W flux as measured by GALFIT (see Section 4.4). An exponentially declining parameterization of star formation history is assumed. Further details on the calculation of stellar masses can be found in Matharu et al. (2019).

This section will explain how the sample selection was carried out for this study. However, a summary of the sample selection can be viewed in Table 2.

Table 2. Summary of Sample Selection

Sample Selection Criteria
0.86 < z < 1.34
Spec-z or grism-z secure cluster member
F140W magnitude < 25
Reff,F140W < 50 kpc
Log(M*/M⊙) ≥ 9.60
$\left\{\begin{array}{ll}{(U-V)}_{{\rm{rest}}}\lt 1.44 & {(V-J)}_{{\rm{rest}}}\lt 0.88\\ {(U-V)}_{{\rm{rest}}}\lt 0.81{(V-J)}_{{\rm{rest}}}+0.73 & 0.88\lt {(V-J)}_{{\rm{rest}}}\lt 1.57\\ {(V-J)}_{{\rm{rest}}}\gt 1.57 & {\rm{otherwise}}\end{array}\right.$
Hα S/N > 3
[b/a]F140W ≥ 0.3

Note. R_eff,F140W and [b/a]_F140W are the GALFIT half-light radius and axis ratio measured from the HST WFC3 F140W imaging, respectively.

Download table as:  ASCIITypeset image

4.1.1. Spec-z and Grism-z Cluster Member Selection

Our first step in sample selection is to select all the secure spec-z cluster members identified as part of the GCLASS survey that are in the HST fields of view (see Section 4.1 of Muzzin et al. (2012) for details on the cluster membership criteria). This amounts to 375 cluster galaxies in total (quiescent and star-forming) across all 10 GCLASS clusters (sixth column, Table 1, before quality check).

The G141 grism data for all extracted sources in the GCLASS HST field of view are quality-checked by eye. The criteria for this quality check are detailed in Section 2.1.4 of Matharu et al. (2019). Galaxies for which a good quality spectroscopic redshift as part of GCLASS and good quality G141 grism data were obtained were used to determine a "secure cluster" selection threshold on grism redshifts. This "secure cluster" selection threshold allowed for additional cluster members that do not have GMOS spectroscopy from GCLASS, but do have well-determined grism redshifts, to be identified. This selection threshold was

$$\begin{eqnarray}&&-0.02\lt \displaystyle \frac{{z}_{\mathrm{grism}}-{z}_{\mathrm{cluster}}}{1+{z}_{\mathrm{grism}}}\lt 0.02\end{eqnarray} \tag{ 4 }$$

where z_grism is the Grizli-determined grism redshift and z_cluster is the spectroscopic redshift of the cluster center. ²¹ We refer the reader to Section 3 of Matharu et al. (2019) for details on how this secure cluster selection threshold was determined. Cluster members that were flagged as "low confidence" in GCLASS were checked for good quality grism redshifts consistent with their respective cluster redshifts. This process adds 220 grism-z secure cluster members (seventh column, Table 1, before quality check) to the original 375 identified from GCLASS slit spectroscopy. This amounts to a total of 595 cluster galaxies (quiescent and star-forming).

4.1.2. Sample Selections Based on the Size Determination Process

4.1.3. Mass Completeness Limits

4.1.4. Rest-frame Colors

U − V and V − J rest-frame colors for each galaxy are obtained using EAZY (Brammer et al. 2008). Further details can be found in van der Burg et al. (2013). These rest-frame colors are used to determine the UVJ color separation for star-forming and quiescent cluster galaxies in the sample. The UVJ color separation technique has proved to be successful in separating star-forming and quiescent populations, even when the former is reddened by dust (Wuyts et al. 2007; Williams et al. 2009; Patel et al. 2012; Foltz et al. 2015). The star-forming GCLASS UVJ selection used is

$$\begin{eqnarray}\left\{\begin{array}{ll}{(U-V)}_{{\rm{rest}}}\lt 1.44 & {(V-J)}_{{\rm{rest}}}\lt 0.88\\ {(U-V)}_{{\rm{rest}}}\lt 0.81{(V-J)}_{{\rm{rest}}}+0.73 & 0.88\lt {(V-J)}_{{\rm{rest}}}\lt 1.57\\ {(V-J)}_{{\rm{rest}}}\gt 1.57 & {\rm{otherwise}}\end{array}\right.\end{eqnarray} \tag{ 5 }$$

After applying this selection to the sample of 474 cluster galaxies selected so far, we are left with 177 star-forming cluster galaxies.

4.1.5. Hα Map Quality Check

4.1.6. Cuts on Hα S/N and F140W Axis Ratio

To illustrate the range in redshift and Hα fluxes spanned by the final cluster sample, we show the star formation main sequence (SFMS) of the sample in Figure 2.

As detailed in Section 2.6, we follow the same method explained in Section 6.2 of Whitaker et al. (2014) to calculate Hα SFRs from our G141 grism spectroscopy. We also overplot the Hα SFMSs from Figure 8 of Whitaker et al. (2014) that happen to approximately span the redshift range of our sample. Our measurements and the SFMSs from Whitaker et al. (2014) are colored by redshift as shown by the color bar.

In general, it can be seen that the cluster star-forming galaxies in our sample are fairly typical, with an Hα SFMS following the field Hα SFMS (as also seen by many other works, e.g., Koyama et al. 2013; Nantais et al. 2020) from 3D-HST, which forms our comparative field sample.

Due to the difficulty in measuring reliable sizes from individual Hα emission line maps of two-orbit depth from WFC3 G141 grism spectroscopy, we stack our data as in Nelson et al. (2016) to boost S/N, allowing us to trace the Hα distribution out to large radii. It is worth noting that Grizli did not exist at the time the similar study to ours in the coeval field environment, 3D-HST (Nelson et al. 2016), was conducted. Due to the sophistication of Grizli, many of the cleaning and masking techniques employed in Nelson et al. (2016) are no longer required or are completed using improved methods in our study. We therefore follow the stacking method outlined in Nelson et al. (2016) as closely as possible, but deviate in places where Grizli provides improved solutions.

The stacking of Hα maps and F140W direct image thumbnails is done in three bins of stellar mass—$9.60\lt {\rm{log}}({M}_{\ast }/{M}_{\odot })\leqslant 10.05$, $10.05\lt {\rm{log}}({M}_{\ast }/{M}_{\odot })\leqslant 10.3$, and $10.3\lt {\rm{log}}({M}_{\ast }/{M}_{\odot })\leqslant 11.2$—ensuring an approximately equal number of galaxies in each stack. First, we convert the units of the F140W direct image thumbnails (which are in counts s⁻¹) to the same units the Hα maps are in (×10⁻¹⁷ erg s⁻¹ cm⁻²) using the pivot wavelength of the F140W filter. Bad pixels are masked and then the F140W direct image segmentation thumbnails (see example in Figure 1(A)) are applied to both the Hα maps and F140W direct image thumbnails to mask neighboring sources. It is worth noting here that the asymmetric double pacman mask devised in Nelson et al. (2016) is not required for our Hα emission line maps output by Grizli. This is because Grizli incorporates improved techniques for masking neighboring [S ii] emission and performing stellar continuum subtraction. Specifically, the flux of the [S ii] λ λ6717, 6731 doublet (assuming a 1:1 line ratio) is determined directly from the spectrum (as for Hα and the other lines), and the 2D [S ii] model, created assuming the same spatial morphology from the F140W thumbnail, is subtracted before drizzling the Hα map.

Each pixel in each Hα map and F140W direct image thumbnail is then weighted by its value in the corresponding weight (inverse variance) map. As in Nelson et al. (2016), each Hα map and F140W direct image thumbnail is weighted by its F140W flux density ²⁴ such that no stack is dominated by a single bright galaxy. For each stack, these normalized maps and direct image thumbnails are summed. The corresponding weight maps are also summed for each stack. Exposure-corrected stacks are created by dividing the summed science stacks by the summed weight stacks for each stellar mass bin. Variance maps for each stack are then simply ${\sigma }_{{ij}}^{2}=1/\sum {w}_{{ij}}$, where w_ij is the weight map for each object in the stack.

We do not rotate and align the thumbnails along the measured semimajor axis since Nelson et al. (2016) demonstrated that doing so did not change their results.

We account for the blending of Hα and [N ii] in each stack by rescaling the Hα flux using the method described in Section 2.6. The median equivalent width of Hα for each stack is used when doing this. The final F140W and Hα stacks can be seen in Figure 3.

Zoom In Zoom Out Reset image size

The size determination process for measuring the spatial extent of the F140W and Hα emission in each stack is conducted using GALFIT (Peng et al. 2002, 2010), following the two-GALFIT-run approach developed and used in Matharu et al. (2019).

GALFIT is software that fits two-dimensional analytic functions to light profiles in an image (Peng et al. 2002, 2010). We fit our stacks with a single-component Sérsic profile, defined as

$$\begin{eqnarray}&&I(r)=I({R}_{{\rm{eff}}})\exp \left[-\kappa \left({\left(\displaystyle \frac{r}{{R}_{{\rm{eff}}}}\right)}^{1/n}-1\right)\right]\end{eqnarray} \tag{ 6 }$$

where R_eff is the half-light radius. This is the radius within which half of the galaxy's total flux is emitted. ²⁵ n is the Sérsic index and κ is an n-dependent parameter. I(R_eff) is the intensity at the half-light radius.

4.4.1. PSF Construction

4.4.2. Size Measurements with GALFIT

We take the final measurements of half-light radius for each stack output by GALFIT in pixels and convert them to kiloparsecs. To do this, we first find the median redshift of the galaxies within each stack. We then calculate the half-light radius, R_eff, for each stack in kiloparsecs using the angular diameter distance to this redshift. Similarly, we also find the median stellar mass for the galaxies in each stack. The stellar mass–size relation for the F140W and Hα stacks can be seen in the right panel of Figure 4. In the left panel of Figure 4, we show the corresponding GALFIT results from the comparative field environment study using the 3D-HST survey (Nelson et al. 2016).

Zoom In Zoom Out Reset image size

Unlike in Nelson et al. (2016) where the errors are calculated from bootstrap resampling the stacks, we calculate our errors by jackknife resampling the stacks. This is due to the expensive nature of the two-GALFIT-run approach (see Section 4.4.2), which sometimes requires individual fits to be altered and rerun by hand. Nevertheless, both methods provide a reasonable estimate of the measurement error due to the stacking and size determination process.

Qualitatively, it can be seen that in both environments the general trend of Hα having a larger half-light radius than the stellar continuum at fixed stellar mass is prevalent (Figure 4). Quantitatively, we show the difference in the ratio of the Hα to stellar continuum half-light radius with environment in Figure 5 at the stellar masses of the cluster stacks. The corresponding Hα and stellar continuum measurements in the field are found by linearly interpolating the results of Nelson et al. (2016) in log space. Over the stellar mass range probed, we find that, on average, the ratio of the Hα to stellar continuum half-light radius is formally smaller in the clusters by (6 ± 9)% (horizontal purple shaded region), but consistent with environment within the 1σ uncertainties.

Zoom In Zoom Out Reset image size

In Section 5, we measured the half-light radii of z ∼ 1 cluster and field star-forming galaxies in Hα and the stellar continuum, finding that the half-light radius is larger in Hα than in the stellar continuum in both environments (Figure 4) and by the same amount (Figure 5) over the stellar mass range probed. This result suggests that inside-out growth via star formation (traced by Hα emission) is proceeding at the same rate in star-forming galaxies regardless of the environment they reside in.

Detailed low-redshift observations of star-forming cluster galaxies—particularly in the Virgo cluster—have provided us with a good understanding of how environmental quenching mechanisms affect the spatial distribution of Hα emission in these galaxies. These works have confirmed that ram pressure stripping (Gunn & Gott 1972), for example, quenches galaxies from the "outside-in," leaving the overall stellar distribution intact, but forces star formation—and therefore Hα emission—to become progressively confined toward the central regions of galaxies (Hodge & Kennicutt 1983; Athanassoula et al. 1993; Ryder & Dopita 1994; Kenney & Koopmann 1999; Koopmann & Kenney 2004b; Koopmann et al. 2006; Cortés et al. 2006; Crowl & Kenney 2006; Abramson et al. 2011, 2016; Vollmer et al. 2012; Gavazzi et al. 2013; Ebeling et al. 2014; Kenney et al. 2015; Lee et al. 2017; Bellhouse et al. 2017; Gullieuszik et al. 2017; Sheen et al. 2017; Gavazzi et al. 2018; Fossati et al.2018; Cramer et al. 2019; Boselli et al. 2020). This disk truncation is seen clearly in individual star-forming cluster galaxies at low redshift, particularly with integral field spectroscopy of "Jellyfish" galaxies as part of the GASP survey (Poggianti et al. 2017; Bellhouse et al. 2017; Gullieuszik et al. 2017; Fritz et al. 2017; Moretti et al. 2018; Jaffé et al. 2018; Vulcani et al. 2018, 2020; Poggianti et al. 2019a, 2019b) and narrowband imaging with long-slit spectroscopy in the VESTIGE (Fossati et al. 2018; Boselli et al. 2020, 2021) survey.

If outside-in quenching via ram pressure stripping is occurring in z ∼ 1 clusters, we should be able to measure the same Hα disk truncation as is measured in local cluster galaxies from our Hα stacks. We find that the ratio of the Hα to stellar continuum half-light radius in z ∼ 1 star-forming galaxies is smaller in the cluster environment by only (6 ± 9)%. Taken at face value, this result is statistically consistent with the coeval field environment, implying outside-in quenching is not occurring in clusters at z ∼ 1, is a subdominant environmental quenching mechanism at this epoch, or is occurring rapidly.

We know quenching is occurring in the GCLASS clusters and more effectively than in the coeval field, due to the high quenched fractions in the clusters (quenched fraction at log(M_*/M_⊙) ≥ 10 is 50% in the clusters versus 30% in the coeval field) and the overabundance of recently and rapidly quenched (or "poststarburst," PSB) galaxies in the clusters (Muzzin et al. 2012).

Over the past decade, a consensus has emerged that environmental quenching likely follows the "delayed-then-rapid" model proposed by Wetzel et al. (2013). In this model, galaxies continue forming stars as they are accreted onto the dark matter halo of another galaxy, but they do not immediately have lower SFRs than central galaxies of the same mass. This suggests that quenching does not begin immediately once a galaxy becomes a satellite, but that there is a "delay" in quenching. During the "delay" phase, there is a gradual decline in SFR. A "rapid" catastrophic event then quenches the satellite galaxy, making it difficult to measure any change in the average SFR. Wetzel et al. (2013) proposed this model to reconcile the high quenched fractions in z = 0 clusters with an environmentally independent SFMS. High-redshift studies have also seen a similar independence of the SFMS with environment (e.g., Figure 2 and Muzzin et al. 2012; Koyama et al. 2013; Nantais et al. 2020; Old et al. 2020), and within this framework of environmental quenching they have found that the delay time of environmental quenching is shorter than it is at z = 0 (Muzzin et al. 2014; Balogh et al. 2016; Foltz et al. 2018). Muzzin et al. (2014) found a delay time a factor ∼2 times shorter than that measured at z = 0 for the GCLASS clusters.

How do we then reconcile the high quenched fractions, overabundance of PSBs (Muzzin et al. 2012), and short delay times (Muzzin et al. 2014) in the GCLASS clusters with the indistinguishable Hα-to-continuum size ratio in the coeval field (Figure 5)? Since cluster galaxies are quenching from the outside-in at low redshift (Section 6.1), but may be doing so more rapidly at z ∼ 1, it makes sense to examine the recently quenched galaxies in the GCLASS clusters.

Under the hypothesis that delay times for environmental quenching are shorter at z ∼ 1, our stacks (Figure 3) would be dominated by star-forming cluster galaxies yet to undergo environmental quenching, washing out any potential outside-in quenching signal. This is because longer delay times allow for a measurable reduction in the Hα disk size, since a larger fraction of cluster star-forming galaxies can be "caught in the act" of environmentally quenching. Shorter delay times lead to cluster galaxies dropping out of the star-forming population more quickly, leaving only the star-forming cluster galaxies yet to undergo environmental quenching dominating the stacks. In support of the "outside-in" environmental quenching scenario, previous studies on the GCLASS clusters that have focused on the spectroscopically confirmed "recently quenched" or PSB cluster galaxy population have found evidence for rapid environmental quenching (Muzzin et al. 2012, 2014) leading to disk truncation in the stellar continuum (Matharu et al. 2020). These studies point toward ram pressure stripping being the likely quenching mechanism, acting within ${0.4}_{-0.4}^{0.3}$ Gyr after satellite cluster galaxies make their first passage of 0.5R₂₀₀, leaving the stellar component of the galaxy morphologically undisturbed (symmetrical, regular morphologies) but truncating its measured F140W half-light radius.

We perform the same analysis conducted on the star-forming cluster galaxies in this paper on the spectroscopically confirmed, recently and rapidly quenched cluster galaxies. 20 of the 23 PSBs studied in Matharu et al. (2020) surprisingly have Hα emission detected in their grism spectra by Grizli. Based on their Grizli-determined Hα fluxes, the mean SFR of the PSBs is 0.7 ± 0.1 M_⊙ yr⁻¹. This SFR is 8 ± 1 times lower than the mean SFR of star-forming cluster galaxies (6 ± 0.1 M_⊙ yr⁻¹) at a similar stellar mass (middle bin, Figure 2). These SFRs are calculated using the mean of the Grizli-determined Hα fluxes from individual detections. Due to the low S/N of the PSB Hα stack, we are unable to fit for both a half-light radius and Sérsic index simultaneously. We therefore follow the same size determination method as was used for the main analysis in this paper (Section 4.4.2), but fix the Sérsic index to the value determined for the Hα stack of the star-forming cluster galaxies at almost the same stellar mass as the median stellar mass of the PSBs. This is the Hα stack for the second stellar mass bin in the main analysis of the paper. The resulting fits are shown in the left panel of Figure 6.

Zoom In Zoom Out Reset image size

We do indeed find that the half-light radius of the Hα emission is significantly smaller, by (70 ± 4)%, than that of the stellar continuum. We caution the reader, however, that the error bar on the Hα measurement may be underestimated, due to four of the 20 jackknife resampled Hα stacks yielding unreliable size measurements. The ratio of the Hα to stellar continuum half-light radius for the PSBs is (81 ± 8)% smaller than for coeval star-forming field galaxies of the same stellar mass (right panel, Figure 6). This rather significant difference in the Hα to stellar continuum half-light radius ratio with environment for rapidly quenched cluster galaxies is consistent with the short delay times measured at high redshift in the framework of the delayed-then-rapid quenching model (Muzzin et al. 2014; Balogh et al. 2016; Foltz et al. 2018). It further suggests that most of the star-forming cluster galaxies in our main analysis are yet to undergo environmental quenching, and that significant changes in the disk occur just prior to or during the PSB phase. Interestingly, the PSB Hα disks are not only fainter in Hα but also smaller than their star-forming counterparts in the coeval field, implying environmental quenching does operate in an "outside-in" manner as seen in the local universe.

We therefore conclude that disk truncation due to ram pressure stripping is occurring in cluster galaxies at z ∼ 1, and is consistent with operating over shorter delay times than observed in z ∼ 0 clusters. This leads to the effects on the Hα distribution more prominently seen in the recently and rapidly quenched cluster galaxy population. Deeper, higher S/N Hα emission line maps of the PSBs from grism spectroscopy will be required for a more detailed comparison of the properties of star-forming disks in cluster star-forming and PSB galaxies. Moreover, there may be a small population of star-forming disks that are in a phase prior to the PSB phase and quenching outside-in. If so, deeper observations providing high S/N Hα emission line maps of individual galaxies would be key for identifying this population.

Our interpretation of the results (Section 6.2) fits well with the growing picture of environmental quenching in the literature, where the delay time of environmental quenching is found to decrease with both increasing redshift and host halo mass (see Foltz et al. 2018 and references therein). Furthermore, recent results from the GOGREEN survey (Balogh et al. 2020)—which includes five of the 10 GCLASS clusters used in our study—show a higher quiescent fraction in clusters with respect to the coeval field and an indistinguishable SMF for star-forming galaxies with environment (van der Burg et al. 2020). Similar results to those in GOGREEN were also found in more moderate overdensities at similar redshifts as part of the ZFOURGE and NEWFIRM Medium-Band surveys (Papovich et al. 2018). Higher quiescent fractions in clusters and an indistinguishable SMF for star-forming galaxies can also be explained by short delay times for environmental quenching and/or high environmental quenching rates building up a significant quiescent fraction in clusters at z ∼ 1. Short delay times would lead to very few star-forming cluster galaxies caught in the act of environmentally quenching (Section 6.2). High environmental quenching rates—regardless of how long or short the delay time is—would lead to the buildup of a large quiescent fraction in the clusters.

Possible evidence for outside-in environmental quenching operating over longer delay times emerges toward lower redshifts in the literature, using a similar technique to the one used in our study. Figure 7 shows a compilation of results from studies that have measured the spatial extent of current star formation versus the spatial extent of the integrated star formation history as a function of environment using various tracers. The most recent result of disk truncation in star-forming cluster galaxies outside the local universe was reported from the K-CLASH survey at 0.3 < z < 0.6, finding the Hα half-light radius to be smaller than the continuum half-light radius in clusters by (26 ± 12)% (Vaughan et al. 2020). Similar results with somewhat increasing significance have been reported at z ≤ 0.5, supporting the possibility that environmental quenching processes that quench cluster galaxies from the outside-in operate over longer delay times at lower redshifts (Koopmann et al. 2006; Bamford et al. 2007; Jaffé et al. 2011; Cortese et al. 2012; Bösch et al. 2013; Bretherton et al. 2013; Vulcani et al. 2016; Schaefer et al. 2017; Finn et al. 2018; Vaughan et al. 2020).

Zoom In Zoom Out Reset image size

The levels of disk truncation measured in cluster star-forming and PSB galaxies at z ∼ 1 straddle the range of disk truncation levels one can expect to measure from outside-in environmental quenching. Therefore, if outside-in quenching operates over longer delay times at lower redshift, we would expect to see levels of disk truncation measured that fall somewhere in between the values measured for star-forming and PSB cluster galaxies at z ∼ 1. This indeed seems to be the case when comparing the low-redshift literature values to our work (Figure 7). However, it is important to note that many of these studies use different tracers and methods to measure the spatial extent of current star formation versus the integrated star formation history. Furthermore, many define environment and select their sample of star-forming galaxies differently over a variety of stellar mass ranges.

Previous results at low redshift (z ≤ 0.5) show that the star-forming disks relative to continuum disks in star-forming galaxies are smaller in clusters by 18%–32% (Figure 7). Our measurement, whilst requiring the stacking of our HST data, has a 9% 1σ error bar. Therefore, we can rule out star-forming disks being 18%–32% smaller than their continuum disks in z ∼ 1 star-forming cluster galaxies at 2σ–3σ. Hence if there were no change in how outside-in environmental quenching operates out to z ∼ 1, our data have the sensitivity to confirm it. Instead, we see an evolution in disk truncation measurements by a factor of three between z ∼ 0 and z ∼ 1 (Figure 7, and see Noble et al. 2019 for similar levels of evolution measured in ratios of CO molecular gas to stellar continuum half-light radius at z ∼ 1.6), implying that the outside-in environmental quenching process evolves between z = 0 and z = 1. More specifically, the disk truncation measurements for star-forming cluster galaxies show that outside-in quenching either is a subdominant environmental quenching mechanism at z ∼ 1 or is operating over shorter delay times. The PSB result at z ∼ 1 confirms outside-in quenching is operating in z ∼ 1 clusters. Combining the PSB result with the known overabundance of PSBs (Muzzin et al. 2012) and shorter delay times (Muzzin et al. 2014) in the GCLASS clusters suggests shorter delay times for outside-in environmental quenching and/or higher outside-in environmental quenching rates at z ∼ 1 compared to z ∼ 0.

The F140W direct image thumbnails (see Figure 1(A)) have the same WCS and drizzle parameters as the emission line maps and can therefore be used to perform direct comparisons with the emission line maps (see second subpanel of Figure 1(D)). Note, however, that the direct image thumbnails are in units of electrons s⁻¹ and the emission line maps are in units of ×10⁻¹⁷ erg s⁻¹ cm⁻².

Within the same fits file as the direct image thumbnails, the user will find a segmentation map thumbnail (see Figure 1(A)), an inverse variance (or "weight") map thumbnail, and a WFC3 F140W PSF thumbnail for the source. Each thumbnail is 80 × 80 pixels, with a pixel scale of 01.

There are two data products for the 2D G141 grism spectra of a source. The first is a fits file (with file-ending beams.fits) containing each independent G141 grism + F140W imaging observation and their associated calibration images (see website for more details). There are often several observations for each individual PA of the G141 grism. Note that the GCLASS G141 observations were taken at one PA and do not have their contamination and continuum removed in this file.

The second data product provides the stacked 2D G141 grism spectrum for each individual PA and the PA-combined stack (see Figure 1(B)). In the case of GCLASS, all G141 observations were taken at a single PA. Therefore the PA-combined stack (bottom row of Figure 1(B)) is the same as the stacked individual PA spectrum (top row of Figure 1(B)). More details regarding the associated calibration images can be found on the data release website.

The collapsed 1D G141 grism spectra, along with their associated best-fit spectral energy distribution and calibration files, are also available in this data release. An example 1D G141 grism spectrum for a galaxy in our final cluster sample and its best-fit model can be seen in Figure 1(C).

The final data product contains the results of the redshift fit (first plot in Figure 1(C)) and the emission line maps for each source (Figure 1(D)). Note that the emission line maps already have the continuum and contamination removed, despite the calibration images being available in the same fits file. Each thumbnail is 80 × 80 pixels, with a pixel scale of 01. See Section A.1 for details regarding the direct image thumbnails in this file, a zoomed-in region of which is shown in the first subpanel of Figure 1(D).