The GLASS-JWST Early Release Science Program. I. Survey Design and Release Plans

Multiple lines of evidence indicate that the universe was reionized at z ≳ 6 (e.g., Planck Collaboration et al. 2020; Mason et al. 2018a, 2019a; Hoag et al. 2019a; Davies et al. 2018; Greig et al. 2019; Morales et al. 2021; Qin et al. 2021). However, the exact timeline of cosmic reionization has not yet been established and the primary sources that governed the process have not yet been identified. Whether reionization was driven by the more numerous, low mass sources with low ionizing photon escape fractions or rarer, high mass sources with enhanced ionizing escape fractions remains a source of debate (e.g., Robertson et al. 2010; Finkelstein et al. 2019; Mason et al. 2019a; Naidu et al. 2020; Matthee et al. 2022).

The galaxy UV luminosity function (LF) provides some clues. If ionizing photon escape fractions for Lyman-break galaxies (LBGs) are as low as inferred from z ∼ 3 observations (f_esc ∼10%, e.g., Steidel et al. 2018; Begley et al. 2022), providing sufficient photons to ionize the universe requires the UV LF to extend to fainter sources than HST is able to detect in blank fields (e.g., Finkelstein et al. 2012; Schmidt et al. 2014; Bouwens et al. 2015; Mason et al. 2015, though, see, e.g., Matthee et al. 2022). HFF data suggest that the LF might continue to rise steeply beyond the blank field limits, and thus testing this hypothesis requires deeper imaging. The intensity of Lyα detected in LBGs also yields hints to the progress of reionization: the transmission of Lyα emission seems to drop at z ≳ 6, suggesting an increasingly neutral intergalactic medium (Fontana et al. 2010; Treu et al. 2013; Schmidt et al. 2014; Pentericci et al. 2014; Mason et al. 2018a, 2019b; Hoag et al. 2019a).

However, interpreting these data requires an understanding of the interplay between the interstellar and intergalactic media (ISM and IGM): due to resonant scattering, Lyα transmission through the IGM depends on the frequency of ISM-escaping photons, which is imprinted by the velocity, density, and spatial distributions of galactic gas (e.g., Verhamme et al. 2006; Wofford et al. 2013; Henry et al. 2015; Rivera-Thorsen et al. 2015; Orlitová et al. 2018; Hoag et al. 2019b; Jaskot et al. 2019; Claeyssens et al. 2022). Recently, high rates of Lyα detection in z > 7.5 luminous galaxies have been reported (Stark et al. 2017). These rates could be explained if substantial H i ISM reservoirs in these systems result in Lyα emission that emerges from the ISM at substantially redshifted velocities (Mason et al. 2018b; Endsley et al. 2022). As such, measuring Lyα line widths and systemic velocity offsets motivates our requirement for high dispersion rest-frame UV/optical spectra with NIRSpec. Conversely, measuring the extent of Lyα requires spatially resolved 2D spectra, which we will obtain with NIRISS.

GLASS-JWST-ERS aims to pave the way to a new physical understanding of high-z galaxies. In the prime field, the combination of lensing magnification and NIRISS/NIRSpec's spatial and spectral resolution will enable the first studies of how Lyα propagates through the ISM/IGM at z ≳ 7. In the parallel fields, it will identify an HFF-comparable sample of 7 < z < 9 LBGs with NIRCam and more than double the census of currently known galaxies at z > 9. Using these data, the program will improve dramatically compared to previous estimates with the Spitzer Space Telescope on the rest-frame optical colors, sizes, and global properties of galaxies at z > 7.

The GLASS-JWST-ERS program was designed to shed light on these issues by enabling the following measurements:

• 1.  
  Lyα systemic velocity offsets at z > 6. Figure 2 highlights the need for a high-resolution (R ∼ 2700) grating with NIRSpec in order to obtain sufficient spectral resolution and wavelength coverage with which to directly measure Lyα velocity offsets and line profiles with respect to the galaxy systemic velocities traced by optical lines. Such data are crucial for determining ISM/IGM absorption and are currently completely unknown except for a handful of sources at these redshifts (e.g., Stark et al. 2017; Pentericci et al. 2016, 2018; Endsley et al. 2022).
• 2.  
  Lyα spatial extent. In combination with the velocity offsets, spatially resolved information is needed to differentiate between Lyα resonant ISM versus IGM scattering. This will be provided by the NIRISS observations.
• 3.  
  Lyα and UV continuum spatial offsets. These will also yield a slit-loss distribution for NIRSpec (Figure 3), which is unknown but crucial for interpreting z > 6 Lyα fluxes (Hoag et al. 2019b; Lemaux et al. 2021).
• 4.  
  Rest-frame optical line redshifts, and optical line measurements for LBGs with undetected Lyα. The vast majority of z > 7 dropouts lack spectroscopic confirmation. LFs are thus based entirely on the robustness of dropout selection techniques. Furthermore, Spitzer/IRAC data suggest that some z > 7 galaxies may have unusually strong optical emission lines indicative of especially young and highly star-forming systems (Laporte et al. 2014; Labbé et al. 2013; Smit et al. 2014; Roberts-Borsani et al. 2016; Castellano et al. 2017; De Barros et al. 2019), or even mature stellar populations governed by strong Balmer breaks (Hashimoto et al. 2018; Tamura et al. 2019; Roberts-Borsani et al. 2020; Strait et al. 2020; Laporte et al. 2021). Our NIRSpec data will provide the first wholesale empirical test of these hypotheses.
• 5.  
  UV emission line fluxes. We will detect or tightly constrain C iii] and C iv. These lines can be enhanced at high-z, possibly due to extreme metal-poor stellar populations (Senchyna et al. 2017, 2022; Stark et al. 2017; Hutchison et al. 2019; Berg et al. 2019; Vanzella et al. 2021; Feltre et al. 2020). In addition to metallicity and star formation conditions, the detection of these lines may aid in identifying high-z active galactic nuclei candidates, using diagnostic diagrams based on He ii, C iv, and C iii], and Lyα (Feltre et al. 2016; Nakajima et al. 2018; Laporte et al. 2017a). As an example, we show a simulated z = 8.38, ∼26.4 AB mag galaxy in the cluster field (Laporte et al. 2017b) as observed with our NIRISS setup in Figure 4, where each of these lines are visible.
• 6.  
  Based on current estimates of the LF (Mason et al. 2015; McLeod et al. 2015; Bouwens et al. 2015; Oesch et al. 2018), rest-frame UV/optical photometry and sizes for ∼100–200 LBGs at z ≳ 7 will be obtained with NIRCam in parallel mode. The images will provide new information on their abundances, stellar masses, and star formation histories (Roberts-Borsani et al. 2021), as well as on the size–luminosity relation (Grazian et al. 2012; Bowler et al. 2017; Bouwens et al. 2021, see also Yang et al. 2022, submitted). Based on the NIRcam point-spread function width at 2 μm of ∼80 mas, star-forming clumps can be probed down to 500–300 pc at z = 6–12, and down to <100 pc with the assistance of lensing magnification.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Why do some galaxies continue to form stars and others do not? What determines the relative growth of stellar disks and bulges? These questions relate to the baryon cycle; the competition between gravity and star formation/black hole-driven outflows (i.e., feedback), thought to regulate star formation in galaxies. Gas-phase metallicities are one of the key probes of this cycle as they are sensitive to the flow of material from galaxies into the circumgalactic medium (CGM; see Maiolino & Mannucci 2019; Kewley et al. 2019 for recent reviews). Outflows of enriched gas efficiently distribute metals throughout the ISM and CGM via galactic fountains (e.g., Roberts-Borsani & Saintonge 2019). Likewise, pristine gas accretion reduces ISM metallicities and enhances star formation.

The slope of the mass–metallicity relation and its star formation rate (SFR) dependence are sensitive probes of feedback and the baryon cycle (Davé et al. 2012; Henry et al. 2013a, 2013b, 2021; Sanders et al. 2021), as is the spatial distribution of metals (i.e., radial gradients; Jones et al. 2013; Anglés-Alcázar et al. 2014). Models make robust but conflicting predictions for such gradients and the mass–metallicity relation depending on the strength of feedback and its outflow properties (Gibson et al. 2013; Ma et al. 2017; Tissera et al. 2019; Hemler et al. 2021). Tension is highest at low masses ($\mathrm{log}{M}_{* }\lesssim 8$), especially at z > 2 (Pilkington et al. 2012; Pillepich et al. 2018; Wang et al. 2017, 2020). Inside-out growth models imply steep gradients that flatten at later times and at higher masses as disks grow. Other scenarios suggest metals should be well mixed by early feedback, then locked into steep gradients as winds lose the power to disrupt massive gas disks (McLure & Dunlop 2004; Few et al. 2012; Ma et al. 2017). Stronger feedback should also lead to a steepening mass–metallicity slope at low masses (Henry et al. 2013b, 2021; Wang et al. 2022). Thus, high-z observations of the low mass end of the mass–metallicity relation would be particularly valuable in discriminating between competing models.

Currently the z > 4 mass–metallicity relation is essentially unknown (Jones et al. 2020; Shapley et al. 2017), and most z > 1 spectroscopy is spatially unresolved. Recently, progress has been made in obtaining a large homogeneous sample of greater than or equal to kiloparsec resolution metallicity gradients at cosmic noon using HST WFC3/IR grism spectroscopy (Wang et al. 2017, 2020; Simons et al. 2021; Li et al. 2022) or ground-based integral-field spectroscopy supported by adaptive optics (Leethochawalit et al. 2016; Förster Schreiber et al. 2018). However, these efforts are exclusively focused on the z ≲ 2.5 universe, given the limitation of current instrumentation.

To distinguish what kind of feedback is active in which galaxies and when, we must map metallicity, dust, and SFR for systems spanning large mass and redshift ranges ($\mathrm{log}{M}_{* }\sim 6$–10 and z ≳ 2, when disks/bulges emerged and feedback was most active). Deep, spatially resolved spectra of large samples is the only route to this understanding. JWST uniquely provides sensitive, uninterrupted wavelength coverage and thus the necessary sets of multiple diagnostic lines required for this analysis.

GLASS-JWST-ERS was designed to shed light on these issues by enabling the following measurements.

• 1.  
  With NIRISS, ionized gas metallicity, dust extinction, and SFR maps will be obtained in ∼50 z ≲ 3.5, $\mathrm{log}{M}_{* }\gtrsim 6$ galaxies, sampling relatively low masses where current models diverge most. These maps will be crucial for interpreting NIRSpec data, which necessarily sample only subsets of galaxies (see Figure 5). The wavelength coverage will enable detection of the optical lines between [O ii] and Hα at 1.7 < z < 2.3, of the lines between [O ii] and [O iii] at 1.7 < z < 3.4.
• 2.  
  NIRSpec will spectrally resolve key diagnostic lines ([N ii] + Hα, the auroral [O iii] line at 4363 Å and Hγ, [Ne iii] + He i + Balmer lines, and doublets such as [S ii] and [O ii]), which are blended at the lower resolution of grism spectra. This addresses the major limitation of grism-only data by improving dust attenuation estimates and detecting weak AGN which might otherwise bias results.
• 3.  
  NIRSpec will measure metallicity, dust attenuation, and SFR in z ≳ 4 galaxies, probing the baryon cycle at these epochs for the first time (Henry et al. 2021; Wang et al. 2022).
• 4.  
  The combination of NIRSpec and NIRISS will cover multiple rest-frame optical metallicity diagnostics for each galaxy. While metallicity calibrations remain uncertain at high redshifts (Kewley et al. 2013), these data will enable comparisons between different diagnostics and support detailed photoionization modeling (Steidel et al. 2016; Chevallard et al. 2018). An additional goal is to detect or set stringent upper limits on the auroral [O iii] line at 4363 Å in the galaxies with the brighter lines, providing temperature-based metallicities in these cases.

The high spatial and spectral resolution public data our program will deliver will support a broad range of external investigations. Here, we detail a few examples, which will hopefully represent just a small subset of the community-driven investigations:

• 1.  
  NIRISS will provide Paβ, Paγ, and [S iii] maps for all of the cluster member galaxies selected from the Grism Lens Amplified Survey from Space (GLASS; Treu et al. 2015; Vulcani et al. 2016, 2017), and Hα—and thus SFR—Paβ and [S iii] maps for field galaxies at z > 0.5, enabling the extension of previous resolved SFR, dust, and metallicity studies to incorporate this critical piece of information pertaining to feedback processes.
• 2.  
  NIRSpec will provide Balmer decrements which—in combination with extant rest-UV photometry from HFF (and redder GTO NIRCam data)—will shed light on the dust content and distribution of high-z galaxies. This is a critical systematic in stellar mass and SFR estimation.
• 3.  
  NIRSpec will provide Paα and other IR lines (e.g., He i and Ca ii) for cluster galaxies, which will shed light on the properties of the central photoionization source, allowing studying in detail SFR and AGN.
• 4.  
  NIRISS + NIRSpec will support extensive refinements of photometric redshift estimation techniques by providing a robust spectroscopic redshift and template training database at higher redshifts than are currently available.
• 5.  
  NIRISS will enable unprecedented spectroscopic continuum studies, including measuring Balmer and metal absorption features (e.g., Mgb) at z > 2, and perhaps allowing their radial gradients to be inferred.
• 6.  
  The new redshifts and 2D spectra will improve lens models and help investigate the nature and distribution of dark matter.
• 7.  
  The comparison of our NIRISS data with previous HST images will provide another epoch for the detection of unusual transients, such as multiply imaged supernovae (Kelly et al. 2015; Treu et al. 2016) or highly magnified stars (Kelly et al. 2018).

As demonstrated by multiple programs, including the HFF campaign, observing a lensing cluster has multiple advantages over a blank field for understanding the high-redshift universe. Background sources are magnified and thus can be studied at higher depth and higher intrinsic resolution than in a blank field. Even at the median magnification of μ = 2.3, the gain in exposure time needed to reach the fainter galaxies, that are the most sensitive probes of reionization and of the baryonic cycle, is substantial (the required exposure time needed to match the depth without magnification would be longer by a factor μ²). In the regions with the highest magnification, lensing allows one to probe one or two orders of magnitude fainter sources and to achieve angular resolution as high as 10–20 pc (Vanzella et al. 2017; Bouwens et al. 2017; Cava et al. 2018). Furthermore, pointing at a cluster enables cluster science and dark matter science via lensing, within the same set of data used for background sources.

A2744 was selected as the GLASS-JWST-ERS primary target for the following reasons: (i) As an HFF cluster, it has exquisite ancillary ultra-deep HST imaging (Lotz et al. 2017), GLASS HST (Treu et al. 2015) and ground-based spectroscopy (Mahler et al. 2018); (ii) Its publicly available lens models at the time of selection (2017) were some of the most robust ever produced, being based on 83 spectroscopic multiple images (Mahler et al. 2018; Richard et al. 2021). A new detailed strong lensing model will soon become available (Bergamini et al., in preparation). The new model reproduces with a significantly lower rms offset (≈04) the observed positions of 90 spectroscopically confirmed multiple images, with secure optical counterparts from HST or VLT/MUSE, from 30 different sources with redshifts between 1.69 and 5.73. In this new study, the cluster total mass distribution is modeled by exploiting the measured stellar velocity dispersions of the brightest cluster members (85 among the 225 selected member galaxies, of which 202 are spectroscopic) and informative priors on the mass of three external substructures detected through an independent weak lensing analysis (Medezinski et al. 2016); (iii) It is visible during the ERS window; (iv) It is slated for GTO imaging, which will aid spectroscopic interpretation; (v) It has minimal galactic extinction (A_F125W = 0.0118), implying low infrared background.

Based on spectrophotometric catalogs from our team and the literature, the following groups of targets (ordered by decreasing priority) were identified at the time of submission. Target selection was updated and refined over the years to take advantage of new data, both spectroscopic (Richard et al. 2021) and imaging follow-up campaigns (Steinhardt et al. 2020). Sources collected from literature are, when necessary, cross matched to our HST catalog, which is aligned to the Gaia DR2 WCS frame, to refine the coordinates. Figure 6 shows the redshift, m_F160W, and stellar mass distributions of the primary samples, based on catalogs available at the time of this writing (early 2022). NIRSpec Microshutters Assembly (MSA; Ferruit et al. 2022) configurations will be designed around them, according to the constraints provided by the schedule. We anticipate of order 50 galaxy spectra will be included in the MSA configuration. In addition, spectra of relatively bright galaxies and stars will be included as secondary calibrators. Our primary science targets include the following populations. The design of the MSA configuration will aim to obtain spectra representing these classes, with emphasis on those relevant to the key science drivers.

• 11.  
  z > 5 spectroscopically confirmed galaxies (sample size 47).
• 12.  
  z > 5 extremely high Spitzer/IRAC-inferred EW(Hα) -and EW([O iii]) galaxies (5).
• 13.  
  z > 6 photometrically selected galaxies, including a z ∼ 8 protocluster (48).
• 14.  
  5 < z < 6 photometrically selected galaxies (46).
• 15.  
  1 < z < 2 spectroscopically selected emission line galaxies from GLASS (30).
• 16.  
  1 < z < 5 spectroscopically confirmed galaxies (327).
• 17.  
  1.1 < z < 3.4 photometrically selected galaxies (9).
• 18.  
  z > 3 photometrically selected galaxies outside the HFF WFC3 footprint (411).
• 19.  
  1.7 < z < 4 expected continuum sources from HFF imaging and GLASS spectroscopy (13).

Zoom In Zoom Out Reset image size

In addition, we include photometrically selected sources, as well as those located outside of the WFC3-IR coverage but within the ACS footprint, as fillers. We will also sample the intracluster light and if possible obtain spectra of some cluster members.

The two orthogonal R = 150 grisms in the F115W, F150W, and F200W filters will provide continuous wavelength coverage in the wavelength range 1–2.2 μm. The spectra will overlap with those obtained by NIRSpec at 1–2.2 μm, enabling direct, quantitative comparisons of spatially and spectrally resolved spectroscopy in that range.

NIRISS provides two orthogonal spectra, replicating the strategy adopted by GLASS (Schmidt et al. 2014; Treu et al. 2015) to mitigate contamination by nearby objects. This is especially important in the cluster core. With GLASS, 64% of the objects had at least one position angle that was free of contaminants (Treu et al. 2015). Since NIRISS spectra are shorter, we expect smaller levels of contamination: indeed, we find 84%–91% of all simulated spectra (see details of their construction below) have mild contamination, defined as such in the case of <10% of detector pixels containing contaminating (neighboring) galaxy flux at the 5σ level. The top end of this range represents values found for the F115W filter, while the lowest estimates represent levels found for the F200W, which covers by far the largest wavelength range (almost a factor of 2× larger than the F115W filter) and thus detector area. If we consider spectra with such mild contamination as clean, then harboring the power of the orthogonal setup afforded by the two grisms adopted by the ERS program, we find that ∼90%–97% of sources have at least one clean spectrum. The contamination level will of course increase for fainter objects, but these will largely be emission-line-only sources, for which contamination and deblending models are robust (as demonstrated by GLASS and NIRISS simulations we have carried out using the Grism Redshift & Line ²⁹ analysis software for space-based slitless spectroscopy).

The total exposure time per filter per grism is divided into four exposures of 1300 s each with small dithers between them, using both GR150R and GR150C grisms, for a total of 2.9 hours per filter. As part of the Wide Field Slitless Spectroscopy (WFSS) observing sequence we obtain four direct images per filter (before and after each orthogonal grism with extra dithers to cover the field of view (FoV) of both grisms) of 350 s each for a total observing time of 2830 s per filter, used to determine the flux, trace position, and wavelength in the grism data. The expected sensitivity of the images based on preflight ETC is given in Table 1.

To construct accurate simulations of our observational program, we utilize a combination of MIRAGE (see the Appendix) and Grizli. The construction of the simulated data sets is divided into three main stages, namely, the simulation of (i) accurate science and dark frames (both direct and dispersed), (ii) the post-processing through the JWST data reduction pipeline, and (iii) the injection of lensed, high-z background sources for estimations of emission line completeness.

4.2.1. Generation of NIRSS Simulations

4.2.2. Data Processing

As part of the second step to reduce the resulting uncalibrated exposures, we use the latest version (v1.4.3) of the JWST data reduction pipeline in an identical fashion to the steps described in the Appendix. Briefly, direct images are run through all of the available image pipelines to create a drizzled and astrometrically and photometrically calibrated image for each filter, with an output catalog of detected sources. That catalog is subsequently used to locate the position of sources in the spectroscopic exposures in order to correctly estimate and subtract contaminating background from the spectra. To obtain a final, co-added spectroscopic image, we align each of the WFSS exposures and take the median of the 2D images to obtain our final image with each filter/grism combination. Examples of the final images we produce (direct and dispersed using the F115W filter) are shown in Figure 7.

Zoom In Zoom Out Reset image size

4.2.3. Line Flux Completeness Simulation

Finally, for completeness estimates of emission lines we also simulate the recovery of mock luminous, high-z (z ≥ 8) sources. We use Lyα as a scientifically compelling example, but we stress that the sensitivity/completeness estimates presented here apply to any spectrally unresolved and spatially compact lines of the same flux and wavelength.

We begin by creating mock catalogs of background galaxies based on a grid of redshifts and Lyα emission. We create simple artificial SEDs of the high-z galaxies based on a simple line profile over a redshift grid of z = 8–15 in integer intervals (except for z = 13, where Lyα falls between the F150W and F200W bands) and an integrated flux grid of f(Lyα) = 1–5000 × 10⁻¹⁸ erg s⁻¹ cm⁻² (specifically f(Lyα/1 × 10⁻¹⁸) values of 5, 10, 50, 100, 500, 1000, and 5000 erg s⁻¹ cm⁻²). For simplicity, we assume Gaussian profiles for the line emission which is placed at the redshifted Lyα wavelength. To keep our results applicable to virtually any emission line of interest, we neglect IGM absorption effects which would typically affect the measured Lyα line flux. Source plane images of the galaxies are then created using postage stamps, where we assume circular profiles with an on-sky radius of ∼016 (corresponding to a median physical size of 0.8 kpc for luminous galaxies at z ∼ 8, as measured in the latest results from the Brightest of Reionization Galaxies Survey; Roberts-Borsani et al. 2022) and a pixel resolution of 005 pixel⁻¹, with an FoV of 150 × 150 arcsec² and 100 mock galaxies within the area.

Given their positions behind the lensing cluster, we aim to simulate lensing effects on the high-z galaxies by the low-redshift galaxy cluster through the use of the gravitational lensing code Lenstronomy and the (Navarro–Frenk–White) deflection maps of Zitrin et al. (2015). We begin by scaling the deflection maps by the lensing efficiency ratio at the desired redshift of the background galaxies (i.e., by the ratio of the angular diameter distance between the cluster and background galaxies, and the angular diameter distance to the background galaxies). Each galaxy is then lensed via a backward ray-tracing method according to the deflection maps, from the source plane to the image plane pixel grid given by the ASTRODEEP NIRISS image defined above. The lensed galaxy images and associated emission line spectra are then passed to Grizli for dispersion and orientation to each ERS pointing, in an identical fashion to the ASTRODEEP images, and subsequently added to the reduced image products after being converted to the appropriate units. The above procedure is repeated 100 times for each redshift/f(Lyα) combination.

To estimate the resulting completeness of our simulations, we define completeness as the median fraction (over the 100 iterations) of recovered sources in the direct images and the number of those recovered sources with detected Lyα. For the latter consideration, we cross match the direct imaging catalog (derived at the calwebb_image3 stage) with the dispersed spectra and integrate the background-subtracted spectrum over the expected position of Lyα while estimating the noise as the standard deviation of the sky adjacent to the source multiplied by the square root of the number of pixels considered. This method allows us to account, in part, for contamination of the high-z source by neighboring foreground sources if the former's flux were to be significantly contaminated by the latter.

The procedure is performed for both grism setups and we consider a source to be detected in Lyα if the extracted line profile has an integrated S/N >5σ in only one of the two configurations or >3σ in both, and display the results of our simulations in Figure 8. We find our imaging completeness remains largely constant at the ∼70%–80% level, primarily due to the apparent brightness of the sources which reach magnitudes of ∼21–27 AB with the strongest Lyα (>5 × 10⁻¹⁷ erg s⁻¹ cm⁻²). We find that even for very bright emission lines the completeness saturates at 80%: 20% of the lines are lost due to contamination by foreground cluster galaxies. By extension, we find our Lyα completeness estimates decrease toward fainter line fluxes as expected, dropping down to ∼40% for f(Lyα) = 1 × 10⁻¹⁷ erg s⁻¹ cm⁻² and down to ∼2%–5% for even fainter line fluxes. We note that, owing to the effect of lensing magnification, our completeness is nonzero down to fluxes of order 10⁻¹⁸ erg s⁻¹ cm⁻².

Zoom In Zoom Out Reset image size

While these simulations represent an idealized scenario in terms of the treatment and data reduction of direct and dispersed images from which sensitivity estimates may be derived, they also naturally fold in the effects of flux contamination and confusion by the massive cluster galaxies and ICL, thus representing realistic simulations of expected data sets. Additionally, while this exercise has been performed for high-redshift galaxies and Lyα emission to demonstrate a full suite of simulations accounting for ICL, lensing, and data reduction steps, the completeness results are sufficiently generic to be applicable to other emission lines at a variety of other redshifts.

GLASS-JWST-ERS will use the R = 2700 gratings G140H, G235H, and G395H, sufficient to resolve Lyα and measure systemic velocities. At this resolution, we expect most other nebular lines will be marginally resolved or unresolved (Masters et al. 2014). Hence, in these cases high resolution maximizes our sensitivity. Spectral coverage from 1–5 μm will enable the detection of features between [O ii] and Hα at 1.7 < z < 6.5, of the range between near-UV rest frame and Hα at 5 < z < 6.5, and of the range between Lyα and [O iii] at 7.3 < z < 9.

Based on our key science drivers we chose to expose for 5 hr in each of the three high-resolution gratings: G140H/F100LP, G235H/F170LP, and G395H/F290LP.

For dithering, we choose to nod in the 3-shutter slitlet. We expect that the standard pipeline subtraction of each nod will be inappropriate for the extended galaxies that we are observing. However, this approach is an efficient way to obtain in-slitlet dithering with lower overheads and higher multiplexing than would be possible with standard dithering via additional MSA configurations. At the time of flight-ready program submission, we will include extra background shutters in empty parts of the MSA, in order to ensure reliable background subtraction. These shutters will then be specified as the background in a customized reprocessing of the NIRSpec observations. We will also obtain an additional small sub-shutter dither ("2-POINT-WITH-NIRCAM-SIZE2") in each exposure, in order to improve point-spread function (PSF) sampling of our NIRCam parallel imaging.

We set up our NIRSpec exposures using NRSIRS2, which is recommended for deep observations. Each NIRSpec band is given 20 groups, one integration, and two exposures. This gives six exposure slots, for which we plan NIRCam imaging parallels. For NIRSpec, each dither/nod point is then 1473.5 s; with three nods, two sub-shutter dithers, and two separate exposures, the total time per band is 1473.5 s times 12, or 4.9 hr.

To simulate results expected from our NIRSpec exposures, we utilize the Python Pandeia exposure time calculation (ETC) engine, upon which the public JWST ETC is constructed. We provide the code with the precise setup designed for our exposures—i.e., an NRSIRS2 readout pattern with observations divided into 20 groups, one integration and two exposures per band, resulting in a total exposure time of ∼4.91 hr per configuration. To provide the most realistic forecasts, we display forecasts of known and spectroscopically confirmed sources in the cluster. To best illustrate the gain in information and variety of emission lines detected from each configuration, we highlight four real sources at redshifts of z_spec = 8.38 (Laporte et al. 2017b), z_spec = 5.054 (Mahler et al. 2018; Richard et al. 2021), z_spec = 2.327 (Mahler et al. 2018; Richard et al. 2021), and z_spec = 1.367 (Wang et al. 2020).

For each source and simulation, we use the high-resolution SEDs from the ASTRODEEP catalogs, while also adding a stacked and skewed Lyα profile from Pentericci et al. (2018) (normalized to the integrated flux reported by Laporte et al. 2017b) to the z = 8.38 object SED in order to illustrate the resolving power of our setups in determining line profiles. The resulting SEDs are subsequently passed to Pandeia for dispersion according to our JWST observations, assuming the sources are centered in their MSA shutter. We show the results for each of the four targets in Figure 9, where we plot the best-fit SED, as well as the (color coded) NIRSpec spectra. Additionally, for each target we also highlight zoomed-in portions of the resulting simulations to highlight resolved profiles of key emission lines or continuum features. We find for bright galaxies (m_F150W ≲ 26.5 AB) emission lines such as [O ii], [O iii], Hβ, and Hα are detected at >3σ, while the high-resolution spectroscopy delivered by our configurations ensures the lines are also resolved and line profiles of, e.g., Lyα can be characterized.

Zoom In Zoom Out Reset image size

GLASS-JWST-ERS takes advantage of NIRCam's capabilities, by simultaneously using the three wide filters in NIRCam's LW arm (F277W, F356W, and F444W) and four wide filters in the SW arm (F090W, F115W, F150W, and 200W). The same filter set and a similar balance between the seven filters will be adopted in the NIRSpec and NIRISS parallels, respectively.

The observational setup for the NIRCam imaging taken in parallel to NIRSpec (which is the deepest) is as follows. Images through filters F090W+F444W are taken in parallel to exposures 1 and 2, F115W+F444W in parallel to exposures 3 and 4, F150W+F277W in parallel to exposure 5, and F200W+F356W in parallel to exposure 6. For each filter, we use the DEEP8 readout pattern, with seven groups and one integration per exposure, with 6 dithers in each exposure slot. Thus each of the six exposure slots contains 8,245 s of imaging. In this way, we will obtain ∼4.6 hr each of F090W and F115W, ∼2.3 hr in F150W, F200W, and F356W, and ∼9.2 hr in F444W, reaching 5σ AB magnitudes limits for point-like sources in the range ∼29.2–29.7 in each filter, according to the prelaunch ETC v.1.7 (Table 2).

The observational setup for the NIRCam imaging taken in parallel to NIRISS is as follows. For each direct image exposure (four per grism angle per filter) in NIRISS we observe with one SW and one LW NIRCam filter in six groups of 311 s each with the SHALLOW4 readout mode, and during each NIRISS grism exposure (four exposures with small dithers) we observe with NIRCam for six groups totaling 4640 s in the DEEP8 readout mode. In total, we observe with F444W for ∼6.5 hr, F090W and F115W for a total of ≃3.2 hr each, F150W and F356W ≃1.7 hr each, and F200W and F277W for ≃1.5 hr each. Based on the prelaunch ETC v.1.7, with this strategy we expect to reach 5σ point-source AB magnitude limits of ∼29.0–29.5 in each filter (Table 2).

The simulations presented below are meant to provide the reader with an intuition of the data quality to be expected.

We performed extensive image reduction tests in preparation for the arrival of real data. We developed a simulation pipeline that, starting from the APT (Astronomer's Proposal Tool) file of the program, automatically creates the corresponding simulated NIRCam FoV. In practice, the simulation pipeline uses EGG (Schreiber et al. 2017) to create a mock observed galaxy catalog centered on the celestial coordinates of the FoV. EGG exploits CANDELS data (Grogin et al. 2011; Koekemoer et al. 2011) to empirically calibrate Monte Carlo realizations of galaxy samples mimicking observed scaling relations and statistical properties; the code can be used to obtain simulated catalogs listing positions, magnitudes, and morphological parameters (ellipticity, axis ratio, position angle) of galaxies observed with any set of filters in a given area of the sky and with any chosen depth. For a given APT configuration, we simulate a catalog that reaches about the 1σ depth of the final images, as estimated with the ETC, in order to include even objects below the limiting depth. In the case of the F150W filter shown here we extend the catalog to limiting magnitude AB = 31. We then add a stellar field, including the known bright stars in the FoV from the GAIA DR2 catalog, plus additional stars fainter than the GAIA limit using TRILEGAL (Girardi et al. 2005, 2012), a populations synthesis code for simulating the stellar photometry of any desired Milky Way stellar field (the positions of the stars were then distributed randomly within the area, after removing those that are so bright that would be detected by GAIA). The colors of the GAIA stars are transformed into those expected in the given JWST filter by assuming color transformations (m_JWST = f(m_BP − m_RP ) + m_RP ), which have been derived using TRILEGAL. The galactic and stellar catalogs are then formatted to be used as input to MIRAGE ³² (see the Appendix), to produce images with a pixel scale of 0031.

We then reduced the resulting images using the general STScI pipeline again as described in the Appendix. However, the current pipeline ³³ does not fully remove some horizontal and vertical striping patterns present in calibrated images. The striping is mainly due to the 1/f noise described by Schlawin et al. (2020), and it is partially correlated at a single amplifier scale. Four independent amplifiers are, indeed, used to simultaneously read out four vertical stripes of 512 pixels in each detector. Therefore, starting from a solution proposed by Schlawin et al. (2020), we added a supplemental step to remove this noise from calibrated images. The step subtracts the median value from each row slide read by each amplifier, applying a sigma clipping to reduce the contribution of extended sources. This method is able to remove almost completely the striping. A downside of the algorithm is that it introduces artifacts in case of very extended sources (see Figure 10). Alternative masking strategies are under study to avoid this unwanted side effect.

Zoom In Zoom Out Reset image size

The calwebb_image3 pipeline is then run on the images to produce the final image and the relevant variance image. We have performed this exercise in all seven filters, and we show in Figure 10 an example of the resulting scientific image F150W, computed in the case of the NIRSpec parallel pointings. We note that, owing to the small dithering patterns adopted by the primary NIRSpec observations, the gaps between detectors are very visible but cosmic rays and other defects are very effectively removed.

The simulations presented here are used to finalize the data reduction pipeline and prepare the analysis tools necessary to obtain the most accurate multiwavelength galaxy catalogs and estimate their completeness and reliability. A quick analysis of the resulting completeness obtained by injecting in the image fake PSF-shaped point sources (with the expected FWHM of 005) of known magnitude, and using SExtractor (Bertin & Arnouts 1996) to detect them and measure their photometry shows that the completeness in the F150W at m ∼ 29 is consistent with the ETC predictions, and recovered fluxes are similar to the input ones, although on average underestimated by approximately 0.1 mag. We emphasize that these are preliminary results focused only on presenting to the general user the expected data quality and illustrating the tools necessary for the analysis of real data. Final recipes and more detailed tests on real data will be reported by the team after the data have been acquired, calibrated, and reduced.

We have adopted a 7-band filter strategy with the primary goal of allowing a robust selection of galaxies at z > 7 via the Lyman-break technique and study their physical properties through a continuous sampling of the rest-frame optical emission. In particular, the F090W filter is essential to identify galaxies at 7 < z < 8.5, while the F115W plays a similar role at 9 < z < 12. However, the extended and continuous wavelength coverage and the high S/N of the resulting observations can deliver useful information for most of the observed galaxies, including those at lower z.

To quantify the performance of the photometric redshift with the seven JWST bands of GLASS-JWST-ERS we have conducted the following simulations. We have taken a conservative approach to minimize the risks of obtaining an artificially high accuracy by using the same galaxy templates for simulating both the sample and recovering the redshift. We have built the input catalog using again the EGG simulator to predict a realistic distribution of colors and magnitudes for galaxies up to z ∼ 6, augmented by a simulated catalog of LBGs computed using the BC03 (Bruzual & Charlot 2003) models, drawn from solar and subsolar metallicity models with random age, exponentially increasing SFR, Calzetti attenuation curve (Calzetti et al. 2000) with E(B − V) ≤ 0.2, Salpeter initial mass function (IMF) and IGM absorption from Inoue et al. (2014).

The resulting catalog of ∼30,000 sources has been perturbed with noise consistent with the observations and then fitted with the zphot code (Fontana et al. 2000; Merlin et al. 2021) adopting templates obtained from the Pegase 2 library (Fioc & Rocca-Volmerange 2019), with self-consistent treatment of the metallicity and dust evolution, a Rana-Basu IMF and the same IGM library. The resulting accuracy of photometric redshifts is shown in Figure 11.

Zoom In Zoom Out Reset image size

The JWST-only photometric redshifts appear reliable and precise over a broad redshift range even though the filter set was optimized for z > 7. This is not entirely surprising. At z > 1, the JWST filter set samples the entire near-UV-to-near-IR range, and is therefore able to track the position of the Balmer/4000 Å and other spectral breaks usually exploited by lower-z surveys. At lower z, it samples the position of the 1.6 μm peak of the stellar emission, which is also another strong redshift indicator (Sawicki 2002). Overall, the resulting accuracy of $\delta z=\tfrac{{z}_{\mathrm{phot}}-{z}_{\mathrm{spec}}}{(1+{z}_{\mathrm{spec}})}$ is σ(δ z) ∼ 0.035 at m_F200W < 29. The fraction of outliers is negligible at m_F200W < 28 and less than 5% at 28 < m_F200W < 29. These numbers should be taken with skepticism, due to some simplifying assumptions such as for example not including emission lines), and the simulations should be repeated based on real data and a specific science goal.

However, in conclusion, our simulations show the GLASS-JWST-ERS NIRCam parallel observations will enable the identification of galaxies at virtually any redshifts in the range z = 0−10 and provide unique information on their spectral energy distribution and rest-frame optical size and morphology.

Besides reduced data, we will release:

• 1.  
  An object-based exploration tool. Interactive visualization of NIRISS grism data + imaging + NIRSpec spectroscopy will be critical to understanding these complex data sets.
• 2.  
  An intuitive RA/DEC-based NIRISS forced extraction tool. The default Grizli + NIRISS pipeline will extract spectra for sources detected in a pre-image. Yet, sometimes—e.g., in the case of faint, high-z galaxies—forced extraction of undetected sources at specific locations is needed. We will develop a tool that allows the user to easily perform this within the Grizli infrastructure, and visualize the results using the GUI.
• 3.  
  Spectroscopic templates. Our program will be one of the first to obtain rest-frame optical spectra of z > 5 sources. Such spectra have never been seen before. Instead, studies have had to rely on redshifting spectra of much lower-z objects for forecasting/modeling. In order to enable the community to make more realistic forecasts and draw more accurate physical inferences, the core team will produce high S/N template spectra by co-adding NIRSpec data.
• 4.  
  Catalogs of basic spectral quantities such as line fluxes will be made available for the NIRSpec targets.
• 5.  
  NIRCam parallel catalogs. We will produce and release photometric catalogs of the parallel fields, focused on z > 7 galaxies.

The Stage II release will include updates to the Stage I products and a quantitative NIRISS/NIRSpec spectral comparison. As discussed above, these instruments provide each other with complementary spatial and spectral information. We will carry out a detailed comparison of the spectra obtained by two instruments for our primary samples, with the goal of characterizing the effects of slit losses and limited spectral resolution in the inference of emission line fluxes, star formation, dust extinction, and metallicity estimates.

The reduced MegaCam images of the field and associated catalogs will be publicly released together with the Stage II release.