Early Results from GLASS-JWST. I: Confirmation of Lensed z ≥ 7 Lyman-break Galaxies behind the Abell 2744 Cluster with NIRISS

We focus on the GLASS-JWST-ERS NIRISS observations of the central regions of the Abell 2744 cluster, obtained on 2022 June 28–29. The field was observed with ∼15 hr of wide field slitless spectroscopy (WFSS; Willott et al. 2022) at R ≈ 150 spectral resolution (at two orthogonal angles) and moderately deep (∼28.6–28.9 AB at 5σ; Treu et al. 2022) preimaging in three different filters (F115W, F150W, F200W). The wavelength range afforded by the three filters allows for the identification of a large variety of spectral features across a number of object types and redshifts, while the choice of two orthogonal grisms (GR150C and GR150R) facilitates the disentanglement of spectra in crowded environments (see Treu et al. 2022 for details on the observing strategy).

We reduce the entire data set using the latest set of available reference files ("jwst_0916.pmap", which includes in-flight calibrations) and the latest version the Grism Redshift & Line (Grizli ²² ; Brammer & Matharu 2021) analysis software, which incorporates the majority of Python routines from the STScI data reduction pipeline as well as custom routines for additional improvements (e.g., background subtraction, image alignment, and drizzling). Starting with the available count-rate files generated by the Detector1 STScI pipeline (which applies detector-level corrections such as the identification of bad pixels and cosmic rays, subtraction of dark current, and ramp fitting), we run Grizli's preprocessing pipeline, which performs World Coordinate System registration and astrometric alignment, flat-fielding, sky background subtraction, and pixel drizzling to provide fully reduced individual exposures and mosaics for both the preimaging and WFSS data sets. We align all our images to the LegacySurveys DR9 (Dey et al. 2019) astrometry, in order to match the astrometric reference frame used for the ALMA Lensing Cluster Survey's data reduction of HST data (Kokorev et al. 2022). For image drizzling, we produce two mosaics at 30 mas and at 60 mas, of which the former can be used for resolved studies and the latter for higher S/N studies of pointlike sources. We adopt the latter throughout this Letter. We calculate the 5σ limiting magnitudes for each of the stacked (and drizzled) preimages: we carefully place 10 circular, r = 01 apertures at representative locations across the images—ensuring these remain free of signal or rare image artifacts—and measure the median standard deviation from those, before scaling, and converting to the desired units. We find our stacked images reach depths of 28.2–28.5 AB (or 29.3 AB for a stacked IR image; see below), consistent with our estimates in Treu et al. (2022) adopting identically sized apertures.

One of the most significant challenges of slitless spectroscopy in clustered environments is to accurately remove overlapping spectra of nearby sources (e.g., Treu et al. 2015; Oesch et al. 2016; Schmidt et al. 2016), given the tendency for contamination to cause confusion in redshift estimations and physical interpretation of a source's spectrum. Considering the large density of sources in our field, such challenges are especially relevant. To identify the position of sources in the $\sim 2\buildrel{\,\prime}\over{.} 1\times 2\buildrel{\,\prime}\over{.} 1$ field of view (FOV), we begin by constructing a flux-weighted IR stack from the preimaging mosaics, which is used to generate a catalog of sources and associated segmentation map. For each detected source brighter than m_AB = 28, Grizli then fits and refines a seventh-order polynomial to the associated spectrum in each individual grism exposure, thereby creating a map with which to identify contaminating pixels around or nearby objects of interest. For F115W spectra (where spatial offsets between the target position in direct imaging and the spectral trace are negligible), we find the contamination model is able to adequately subtract the majority of nearby contaminants; however this becomes more challenging as a function of wavelength where offsets of the spectral trace relative to the source become larger. Improvements and imminent updates to existing reference files will further improve wavelength calibrations, spectral trace offset measurements, and flux calibrations for more precise measurements. Considering the above, all pixels used for extraction and modeling purposes are therefore weighted by w_pix according to a combination of their contamination level and flux uncertainty (defined by the error array in the flux-calibrated count-rate images). We define the weights as ${w}_{\mathrm{pix}}={\sigma }_{\mathrm{pix}}^{-2}\cdot \exp (-{f}_{{\rm{c}}}\cdot | {f}_{\lambda ,\mathrm{cont},\mathrm{pix}}| /\sqrt{{\sigma }_{\mathrm{pix}}^{2}})$, where f_λ,cont,pix is the contamination model flux, f_c is a constant down-weighting factor set to 0.2, and σ_pix is the flux noise for a pixel given by the error array.

For a specified target, all individual, "cleaned" (i.e., weighted according to the equation described above) 2D grism exposures provided to Grizli are modeled simultaneously, allowing for independent noise handling and the avoidance of additional complications resulting from stacking (e.g., morphological broadening and smearing of spectral traces from asymmetric pixel sizes, orthogonal dispersion directions, and varying trace offsets between grism setups), while retaining a collective S/N equivalent to that of a stacked mosaic. Using constraints from the collective sample of exposures, a 2D galaxy model is generated via a linear combination of galaxy templates and dispersed onto the plane of the sky according to the location and morphological structure of the source as measured from the preimaging exposures. To generate the model, we make use of a variety of galaxy templates (e.g., intermediate age spectral energy distributions (SEDs) with moderate 4000 Å breaks, older simple stellar population models to account for low-level absorption features, post-starburst galaxies from the UltraVISTA survey, and low-metallicity LBGs) incorporating continuum and emission line contributions, to be able to flexibly fit a large variety of spectra according to an allowed redshift range.

Given the common astrometric reference frame, in this study we also include for added constraints the full assortment of available HST photometry from the Frontier Fields, derived from the drizzled mosaics of Kokorev et al. (2022) that we resampled to 60 mas pixel size to match our NIRISS mosaics. Specifically, we include photometry from HST/WFC3-UVIS (F336W), HST/ACS (F435W, F606W, F814W), and HST/WFC3 (F105W, F125W, F140W, and F160W), which serve as valuable additional constraints to determine the location of the Lyman break. As an added check, for sources of interest, we also visually inspect each filter-grism stack and exclude combinations where we deem the spectra to be significantly impacted by lingering contamination (e.g., if they appear in only one grism orientation). Thus, we fit the resulting NIRISS grism exposures and HST photometry with models over a broad redshift range z = [0, 15], to allow for both low-z and high-z solutions. The best-fit model is considered the one with the highest P(z)/χ² value. Associated 1D spectra are then optimally extracted from the data by Grizli, using the morphological model of the source as a reference for the position and extension (in the spatial direction) of the spectral trace.

Here we focus on the WFSS follow-up of previously identified z ≥ 7 sources in the Abell 2744 cluster that (i) fall in the NIRISS FOV, and (ii) are sufficiently bright to be reliably detected via continuum measurements, with the aim of determining their spectroscopic redshifts. At z ≥ 7, the filters adopted by GLASS-ERS allow for spectral coverage of the Lyman break out to z ∼ 7.2–17.0. Additionally, the filters also cover emission lines such as Lyα and N v emission out to z ∼ 17.0, C iv out to z ∼ 13.5, He ii and O iii] out to z ∼ 12.5, and C iii] out to z ∼ 11.0.

We select our targets from a comprehensive compilation of photometrically selected sources from Zheng et al. (2014), Zitrin et al. (2014), Leung et al. (2018), Ishigaki et al. (2018), Bouwens et al. (2022b), and the public Hubble Frontier Fields catalogs of Shipley et al. (2018) and Castellano et al. (2016). Each of the compiled sources were originally selected as z ≥ 7 LBGs from NIR color cuts using HST ACS+WFC3 data and photo-z constraints.

In this initial study, aimed at continuum detections, we limit our search to galaxies with a reported F160W magnitude (or F125W if F160W is not reported) of m_AB < 26 and quality flags indicating robust flux and/or photo-z measurements. Such a choice is based on S/N predictions given by the Exposure Time Calculator, indicating >2σ detections in F115W (per spectral pixel, assuming a r = 01 aperture) for sources with m_cont ∼ 26 AB.

Accounting for overlap between studies, the compilation and cut comprises a total of six unique sources, three of which are sufficiently isolated from bright neighbors to be detected in our IR stack and segmentation map. The final sample of sources spans a range of observed F160W magnitudes ∼24.6–25.6 AB and photometric redshifts of z_phot ∼ 7.3–7.9. The galaxies—referred to as GLASSz8-1, GLASSz8-2, and DS7226—and their locations relative to the cluster center are shown in Figure 1, which displays an HST-NIRISS false-color image of the central cluster.

Zoom In Zoom Out Reset image size

Out of the three galaxy candidates compiled and identified here, we confirm two of them—GLASSz8-1 and GLASSz8-2 (formerly ZD2 and 2458 in Zheng et al. 2014 and Castellano et al. 2016, respectively) to lie at z ∼ 8 based on a clear drop in flux blueward of (rest frame) 1216 Å indicating substantial absorption by intervening neutral hydrogen along the line of sight. Both galaxies are sufficiently isolated as to be clear of contamination from nearby objects.

We show their 2D spectra and 1D extracted fluxes in Figure 2, along with the best-fit Grizli model, which places them at redshifts of z_grism = 8.04 ± 0.15 and z_grism = 7.90 ± 0.13, respectively. Redshift uncertainties are quoted as the 1σ standard deviation of the P(z) at their z ∼ 8 locus. Inspecting Figure 2, we find the break is clearly identified in the F115W filter, where significant flux is detected redward of the break and only noise is visible blueward of it. GLASSz8-2 displays some apparently rising flux at the bluest end of the F115W. However, this is very close to the edge of the filter's sensitivity curve and the enlarged error bars make this consistent with noise from imperfect background subtraction. As expected for z ≥ 7 galaxies, the continuum is also detected (at reduced levels) across the entire wavelength ranges of the F150W and F200W filters, further supporting the confirmation. The reduced continuum levels at increasing wavelengths suggest blue UV slopes and absolute 1500 Å magnitudes of $\beta =[-{1.69}_{-0.05}^{+0.04},-{1.33}_{-0.13}^{+0.14}]$ and ${M}_{\mathrm{UV}}=[-{20.37}_{-0.02}^{+0.02},-{19.68}_{-0.07}^{+0.05}]$ mag (for GLASSz8-1 and GLASSz8-2, respectively), corrected for magnification effects (μ ≈ 2; see below) using the model of Bergamini et al. (2022), indicative of young, star-forming systems.

Zoom In Zoom Out Reset image size

In both cases, we observe some excess flux at the expected position of Lyα; however, such an excess is marginal and would require deeper and higher-resolution spectroscopy (e.g., with NIRSpec) to confirm. Subtracting the best-fit continuum model from the flux, normalizing by the continuum model, and integrating over the apparent line profile points to 3σ equivalent width upper limits of <10–15 Å (rest frame <1.6 Å). We tabulate the galaxies' spectrophotometric properties in Table 1, where we show that both spectroscopic redshifts are consistent with previously estimated photo-z's. The remaining target displays a z ∼ 2.6 solution based on prominent emission line detections that classify it as a contaminant.

Table 1. A Summary of NIRISS/WFSS Spectroscopically Confirmed z > 7 Galaxies behind the Abell 2744 Cluster, and Their Spectrophotometric Properties

GLASS ID	R.A.	Decl.	H160	μ	MUV/μ	β	zgrism	zphot	Reference
	(deg)	(deg)	(AB)		(AB)
GLASSz8-1	3.60451	−30.38046	25.56	${2.00}_{-0.04}^{+0.05}$	$-{20.37}_{-0.02}^{+0.02}$	$-{1.69}_{-0.05}^{+0.04}$	8.04 ± 0.15	7.90	ZD2; Zheng et al. (2014)
GLASSz8-2	3.60135	−30.37921	24.79	${2.10}_{-0.04}^{+0.06}$	$-{19.68}_{-0.07}^{+0.05}$	$-{1.33}_{-0.13}^{+0.14}$	7.90 ± 0.13	7.55	2458; Castellano et al. (2016)

Note. The last column lists the reference from which the photometric redshift and H₁₆₀ magnitude are taken. For magnification factors, we use the model of Bergamini et al. (2022).

Download table as:  ASCIITypeset image

Both GLASSz8-1 and GLASSz8-2 are lensed by a factor of μ ≈ 2 (Bergamini et al. 2022) and are among the brightest of the target sample, with reported H₁₆₀ magnitudes of m_AB = 25.56 and m_AB = 24.80, respectively. We also find their positions on the plane of the sky place them within ∼7''–135 from the z ∼ 8 overdensity, possibly indicative of some association between themselves and the other galaxies.

The results shown here highlight the extraordinary potential of grism spectroscopy to determine spectroscopic redshifts independently of the Lyα emission seen in exceptional objects (e.g., Finkelstein et al. 2013; Oesch et al. 2015; Zitrin et al. 2015; Roberts-Borsani et al. 2016) and characterize unbiased samples of galaxies well into the Epoch of Reionization. To highlight this, in Figure 3 we show a representative compilation of spectroscopically verified LBGs at z ≳ 7 with clear continuum breaks, irrespective of emission line detections. The number of galaxies unsurprisingly decreases as a function of redshift and remains exclusive to the apparently brighter sources—the galaxies presented here increase the number of confirmed z > 7 continuum detections in the literature by a factor of 1.5×, thus serving as a powerful illustration of JWST. Such samples will prove crucial to determine a representative picture of the sources that governed the reionization process and in this regard additional, pure-parallel grism observations over blank regions of the sky (e.g., GO 1571 PASSAGE, PI Malkan) will prove especially useful to obtain conclusions over the general galaxy population.

Zoom In Zoom Out Reset image size

In addition to the clear Lyman break, for GLASSz8-1 we also report tentative (i.e., <5σ) detections of rest-frame UV lines at expected wavelengths. The clearest identified lines are He ii λ1640 Å, the (unresolved) O iii]λλ1661,1666 Å doublet, and the (unresolved) N iii]λλ1747,1749 doublet in the F150W filter. To quantify their statistical significance, we subtract the modeled continuum from the spectrum and measure the peak S/N of each line from the residual, using again the error array from the flux-calibrated count-rate image as the uncertainty. The He ii, O iii], and N iii] lines have peak S/Ns of 2.7σ, 2.8σ, and 3.8σ, respectively. After verifying the Gaussian nature of the noise distribution, we find only three other pixels in the full F150W 1D spectrum have S/N ≳ 2.75. Although the lines are not especially strong, we note none of them appear close to any nearby residual contamination and their observed wavelengths and separations relative to each other are consistent with the interpretation. If confirmed at higher significance, the detections of O iii] and N iii] at such high redshift would prove a first.

The presence of rest-frame UV lines in z > 7 galaxies is not unexpected, although have thus far been more prominently seen in strong Lyα emitters (e.g., Laporte et al. 2017a; Stark et al. 2017; Mainali et al. 2018). Placing the line detections into context, the presence of O iii] and N iii] can be explained by both star-forming sources with low metallicities (Z ∼ 0.001) or nonthermal sources as they require moderate photon energies (∼35 eV) to become ionized (Feltre et al. 2016). He ii, in contrast, requires significantly higher energies of ∼54 eV from strong ionizing continuum photons suggestive of more extreme ionizing conditions (e.g., active galactic nuclei or massive, low-metallicity stars). Low-metallicity systems are common among young, star-forming galaxies at z ∼ 8 (e.g., Strait et al. 2020; Roberts-Borsani et al. 2021); however only a handful of sources at z > 7 display high-ionization lines similar to those reported here (e.g., Laporte et al. 2017a; Mainali et al. 2018). More detailed and higher S/N observations of such lines will be required to confirm their prevalence among the general galaxy population.