JWST UNCOVER: discovery of z > 9 galaxy candidates behind the lensing cluster Abell 2744

ABSTRACT

We present the results of a search for high-redshift (z > 9) galaxy candidates in the JWST UNCOVER survey, using deep NIRCam and NIRISS imaging in seven bands over ∼45 arcmin² and ancillary Hubble Space Telescope (HST) observations. The NIRCam observations reach a 5σ limiting magnitude of ∼29.2 AB. The identification of high-z candidates relies on a combination of a dropout selection and photometric redshifts. We find 16 candidates at 9 < z < 12 and three candidates at 12 < z < 13, eight candidates are deemed very robust. Their lensing amplification ranges from μ = 1.2 to 11.5. Candidates have a wide range of (lensing corrected) luminosities and young ages, with low stellar masses [6.8 < log(M_⋆/M_⊙) < 9.5] and low star formation rates (SFR = 0.2–7 M_⊙ yr⁻¹), confirming previous findings in early JWST observations of z > 9. A few galaxies at z ∼ 9−10 appear to show a clear Balmer break between the F356W and F444W/F410M bands, which helps constrain their stellar mass. We estimate blue UV continuum slopes between β = −1.8 and −2.3, typical for early galaxies at z > 9 but not as extreme as the bluest recently discovered sources. We also find evidence for a rapid redshift-evolution of the mass-luminosity relation and a redshift evolution of the UV continuum slope for a given range of intrinsic magnitude, in line with theoretical predictions. These findings suggest that deeper JWST observations are needed to reach the fainter galaxy population at those early epochs, and follow-up spectroscopy will help better constrain the physical properties and star formation histories of a larger sample of galaxies.

1 INTRODUCTION

While the Hubble Space Telescope (HST) and ground-based observatories have uncovered more than two thousands galaxies at redshifts greater than z ∼ 6 (Atek et al. 2015; Finkelstein et al. 2015; Bouwens et al. 2021), only a handful of galaxies were known at z > 9 (Oesch et al. 2018; Bowler et al. 2020; Bagley et al. 2022a). This observational frontier is mainly due to the near-infrared (NIR) wavelength coverage of HST which is limited to λ < 2 |$\mu$|m, whereas the rest-frame ultraviolet (UV) light of early galaxies is increasingly shifted towards longer wavelengths. With its NIRCam (near-infrared camera) instrument covering the ∼1−5 |$\mu$|m domain, coupled with a significantly higher sensitivity compared to its NIR predecessors (Rigby et al. 2022; Rieke et al. 2023), JWST is poised to revolutionize our views of the early stages of galaxy formation.

In the early months of operation, several studies have reported the discovery of z > 9 galaxy candidates in the first JWST imaging observations: the Early Release Observations (ERO; Pontoppidan et al. 2022), Early Release Science (ERS) programmes CEERS (The Cosmic Evolution Early Release Science Survey; Bagley et al. 2022b) and GLASS (Through the Looking GLASS; Treu et al. 2022). Among these early results, Naidu et al. (2022) reported candidates at z ∼ 12–13, Finkelstein et al. (2022b) a candidate at z ∼ 12, while samples of z ∼ 9–16 candidates have been presented in Adams et al. (2023); Atek et al. (2023); Austin et al. (2023); Donnan et al. (2023); Harikane et al. (2023). Many of these galaxy candidates broke the previous distance record held by HST observations (Oesch et al. 2016). More recently, several programmes have started to spectroscopically confirm some of these high-redshift candidates (Morishita et al. 2022; Roberts-Borsani et al. 2022), with the highest-redshift galaxy located at z ∼ 13 (Curtis-Lake et al. 2022; Robertson et al. 2022). At the same time, the high-redshift solution of some of these candidates have been ruled out by NIRSpec follow-up observations. For example, the highest-redshift candidate at z ∼ 16.7 (Donnan et al. 2023) has been confirmed to be a dusty galaxy at z ∼ 4.9 with intense rest-frame optical emission lines (Arrabal Haro et al. 2023).

However, more than their distance, the most striking aspect perhaps is their combined number density and brightness. Indeed, the inferred number density is significantly larger than theoretical predictions based on galaxy formation models, or extrapolation of lower-redshift luminosity functions (Mason, Trenti & Treu 2022; Naidu et al. 2022; Bouwens et al. 2022a; Atek et al. 2023; Donnan et al. 2023). While some of these candidates have been confirmed at z ∼ 12 or 13 (e.g. Curtis-Lake et al. 2022; Arrabal Haro et al. 2023), others turned out to be low-redshift dusty interlopers, which always warrants caution in their interpretation. Also, a sample of red massive galaxies at z = 7–9, reported by Labbe et al. (2022), appear to have stellar masses approaching that of the present-day Milky Way, in potential tension with standard lambda cold dark matter (ΛCDM) models (Boylan-Kolchin 2022).

Several studies have attempted to understand these early observations and interpret these surprising results. In particular, the high number density of luminous galaxies at z > 12 has been attributed to the decreasing amount of dust attenuation at higher redshift (Ferrara, Pallottini & Dayal 2022), higher star formation efficiency in early galaxies and/or non-standard initial mass function (IMF; Mason et al. 2022; Ziparo et al. 2022), or even non-ΛCDM cosmologies (Boylan-Kolchin 2022; Menci et al. 2022). In the meantime, a larger area of deep JWST surveys and spectroscopic follow-up observations are needed to confirm this claim by increasing the sample size of confirmed z > 9 galaxies.

During the first cycle of JWST operations, our UNCOVER (Ultradeep NIRSpec and NIRCam ObserVations before the Epoch of Reionization) Treasury survey has obtained deep multiwavelength NIRCam imaging of the lensing cluster Abell 2744 (Bezanson et al. 2022). UNCOVER deep NIRCam imaging consists of a mosaic in seven filters for ∼4–6 h band⁻¹, reaching a magnitude limit of ∼29.2 AB. Following on the steps of the Hubble Frontier Fields (HFF), the programme relies on the gravitational lensing boost to push beyond blank fields limits. In fact, assuming an average amplification of 5, UNCOVER is intrinsically the deepest observing programme of Cycle 1. In addition, the programme will obtain spectra for the intrinsically faintest distant galaxies to date with 5–20 h of NIRSpec Prism follow-up observations. In addition to our programme, NIRISS imaging of A2744 was obtained as part of ERS GLASS programme, and NIRCam imaging as part of the DDT (Director Discretionary Time) programme ID 2756. We combined all these imaging data to increase the depth and the area of our survey.

In this paper, we present the detection of z > 9 galaxy candidates in the NIRCam and NIRISS imaging data, determine their physical properties, and compare them to theoretical predictions. Using imaging data in 15 broad-band filters, the identification of galaxy candidates is based on a combination of photometric dropout criteria and photometric redshifts derived from spectral energy distribution (SED) fitting with both beagle and eazy codes. We describe the imaging data set in Section 2 and the sample selection in Section 3. We present our estimates of the physical parameters and their redshift-evolution in Section 4. Our conclusions are given in Section 5. We use AB magnitudes (Oke & Gunn 1983) and a standard cosmology with H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_Λ = 0.7, and Ω_m = 0.3.

2 OBSERVATIONS

The UNCOVER observations are described in the survey article (Bezanson et al. 2022), which is accompanied by our first data release of the imaging mosaics, available at the UNCOVER webpage.¹ Here, we describe briefly the content of the data and their photometric characteristics.

The NIRCam data consist of short wavelength (SW) imaging in three broad-band filters (F115W, F150W, and F200W), long wavelength (LW) imaging in three broad-band filters (F277W, F356W, and F444W), and one medium band filter (F410M). The exposure times and the resulting magnitude limits in all filters are listed in Table 1. Simultaneously to the NIRCam observations, NIRISS imaging is obtained in parallel using five broad-band filters (F115W, F150W, F200W, F356W, and F444W). Our analysis also includes data from the GLASS survey (Treu et al. 2022) obtained with NIRISS, which adds the F090W band in a fraction of the UNCOVER area. In addition, we incorporate NIRCam imaging from the DDT programme ID 2756, which uses a similar set of filters to UNCOVER, except the F410M filter, and shorter exposure times. Using the grizli software (Brammer et al. 2022), the data were then reduced and drizzled into mosaics with a common pixel scale of 0.4 arcsec pix⁻¹ and a total field of view of ∼45 arcmin².

The cluster core of A2744 is covered by deep HST imaging from the HFF programme, and a slightly wider area with shallower observations from the BUFFALO (The Beyond Ultra-deep Frontier Fields and Legacy Observations ) programme (Steinhardt et al. 2020). The HST observations include Advanced Camera for Surveys (ACS) imaging in three filters (F435W, F606W, and F814W), and Wide Field Camera 3 (WCF3) in four filters (F105W, F125W, F140W, and F160W). Furthermore, the UNCOVER NIRISS parallels overlap with deep HST ACS F814W imaging in the A2744 parallel field. All these observations are drizzled to the same pixel scale and aligned to the UNCOVER images. Detailed characterization of the data are presented in Weaver et al. (2023).

3 HIGH-REDSHIFT SAMPLE SELECTION

3.1 Photometric catalogues

For our sample selection and analysis, we compared two photometric catalogues: (i) the general UNCOVER catalogue published in Weaver et al. (2023) which has been designed to fit most of the scientific investigations covered by this data set, (ii) a custom photometric catalogue specifically tailored to the detection of high-redshift galaxies. The main differences reside in the aperture size, the deblending parameters, and the aperture corrections.

3.1.1 General catalogue

The object detection and photometry are performed in images that were previously corrected for contamination from intracluster light (ICL) and bright cluster galaxies, following methods developed in Shipley et al. (2018). The approach is based on an iterative process of fitting and subtracting the bright cluster members and the ICL in the images. A detailed description is given in a companion paper by Weaver et al. (2023).

All images are matched to the point spread function (PSF) of the longest wavelength image in the F444W filter. The detection image consists of a co-addition of the three long-wavelength JWST filters F277W, F356W, and F444W. Photometry is measured in 0.32 arcsec apertures using the python version of source extractor sep (Bertin & Arnouts 1996; Barbary 2016). We adopted the following parameters for the sep extraction: a detection threshold of 1.5σ, a deblending threshold of 16, and deblending contrast of 3e-3. The total fluxes are estimated by applying a correction derived from elliptical Kron apertures (Kron 1980). Details of the PSF-matching procedure, and additional photometric corrections, are described in Weaver et al. (2023).

3.1.2 High-redshift catalogue

In addition, we produce a photometric source catalogue using the SExtractor tool (Bertin & Arnouts 1996) in dual mode on each available image, using the F444W as the detection image. We adopt a detection threshold of 0.9 (relative to the rms image), a minimum detection area of 6 pixels, and a deblending threshold of 3e-4. We measured individual fluxes in 0.24 arcsec circular apertures in each filter. The total fluxes were obtained by using a scaling factor derived from the ratio of the aperture flux to the auto_flux obtained from SExtractor in the F444W image. To account for the missing flux due to the PSF wings, we measured the aperture flux as a function of the aperture radius in the PSF of the F444W band. For each object, we computed the equivalent circularized radius as |$r = \sqrt{a \times b}~ \times$|kron_radius, and divided the encircled flux by the flux fraction in the PSF F444W for this radius. This correction typically increases all the fluxes by ∼10–20  per cent.

In the end, the comparison between the two catalogues shows that the latter is better suited for high-redshift sources, particularly in deblending the small objects, and in estimating the object and background fluxes.

3.2 Dropout selection

Following the colour–colour criteria defined in Atek et al. (2023) we select 9 < z < 11 galaxies that satisfy:

This selection window, illustrated in Fig. 1, has been designed to minimize potential contamination from low-redshift interlopers and cool stars. To determine the colour–colour space of these contaminants, we generated quiescent galaxy templates from the SWIRES (Spitzer Wide-area InfraRed Extragalactic Survey) library (Polletta et al. 2007), applied different dust attenuation values A_V = [0, 0.25, 1] assuming an SMC dust law, and computed synthetic photometry in the set of broad-band filters used in this paper. The resulting colour–colour tracks are shown in Fig. 1. We also compute the colour tracks of cold red stars and brown dwarfs using stellar templates from Chabrier et al. (2000) and Allard et al. (2001).

In addition to these selection criteria, we require that sources are detected in all LW filters with a minimum SNR = 5 and that they remain undetected in F090W, when available, and all HST optical bands at a 2σ level. For sources that are not detected in the dropout filter, we assign a 1σ lower limit corresponding to the filter limiting magnitude to ensure a minimum continuum break of one magnitude.

For z ∼ 12–15 candidates, we adopt the following criteria:

Similarly, we require high-significance detection in the LW filters and that none of these candidates are detected in the bands blueward of the Lyman break, i.e. in the HST, and JWST F090W and F115W filters.

3.3 Spectral energy distribution fitting

In parallel, we estimate photometric redshifts by running spectral energy distribution (SED) fitting. We apply 5 per cent error floor to all photometric measurements to reflect the calibration uncertainties of JWST NIRCam data. To minimize the propagation of lensing uncertainties, both of the following procedures are based on the observed flux densities without correction for magnification. The derived parameters are then unlensed a posteriori. We first use the eazy software (Brammer et al. 2022), over an allowed redshift range of 0.01 < z < 20, adopting a flat luminosity prior and the CORR_SFHZ set of galaxy templates, which include redshift-dependent SFHs informed by the most recent results of JWST observations of high-redshift galaxies (Larson et al. 2022a; Carnall et al. 2023). As shown in Weaver et al. (2023), these templates perform better than the default fsps_full library in recovering the true redshift. Improvements over the eazy fsps templates have also been presented in Larson et al. (), which include bluer UV slopes based on bpass (Stanway & Eldridge 2018) and nebular emission models.

Second, we run the beagle (BayEsian Analysis of GaLaxy sEds) SED fitting software (Chevallard & Charlot 2016) on the same photometric data. The procedure uses stellar population models from (Bruzual & Charlot 2003) and nebular emission models from Gutkin, Charlot & Bruzual (2016). In the first run of the SED fitting, we are mostly interested in the best redshift solution. We adopt a simple constant star formation history, uniform priors on the photometric redshift and dust attenuation optical depth of z_phot ∈ [0, 25] and |$\hat{\tau }\in [0,3]$|⁠, and log-uniform priors on the stellar mass and age of log(M_⋆/M_⊙) ∈ [5, 11] and log(t_age yr⁻¹) ∈ [7, t_Universe]. The metallicity is fixed to |$Z=0.1\, \mathrm{Z}_{\odot }$|⁠.

In order to identify spurious objects, or sources affected by artifacts, we flag objects that meet the following criteria: objects whose segmentation apertures overlap with the edge of the detector or are next to bright stars, are affected by bad pixels, whose size is 1 pixel or less, or are likely to be stars. For the latter, we combine information from both the sextractor stellarity parameter class_star and the χ² of the eazy SED fitting run using a set of dwarf star templates through the fit_phoenix_stars function. In addition, all the sources that pass these filters are visually inspected for potential contamination (diffraction spikes, detector artefacts, etc.)

All the candidates have best-fitting photometric redshifts consistent with the dropout selection. Conversely, when first selecting candidates using photometric redshift criteria, limiting the sample to best-fitting solutions with χ² < 30, 31 candidates satisfy the selection. The dropout selection rejects 13 candidates. For the majority of these rejected candidates, the high-redshift solution is due to fitting failures, where the best-fitting SED does not match the photometry. Some objects do not show a clear Lyman break, or are clearly detected in the blue bands. Also, in few cases, the signal-to-noise level in the detection bands is simply below our selection threshold.

The final sample of high-redshift candidates consists of a total of 16 galaxies in the redshift range 9 < z < 11 and 3 galaxies at 11 < z < 15. The complete list and properties of the high-redshift sample are reported in Table 2.

3.4 Gravitational lensing model

In order to compute the gravitational magnifications of the sample and the effective survey area, we use the new UNCOVER cluster mass model derived by Furtak et al. (2022). This parametric strong lensing model is based on existing and newly discovered multiple image systems in the deep NIRCam imaging of UNCOVER. Thanks to the wide NIRCam coverage, the total survey area with an amplification factor μ > 2 is about 3.5 arcmin², which is a significant improvement over the HFF-derived area of ∼0.9 arcmin². Fig. 2 shows the cumulative surface area as a function of magnification. The amplification factors shown in Table 2 range between μ = 1.2 and μ = 11.5, and were derived using the photometric redshifts.

4 RESULTS

4.1 High-redshift candidates

We find 16 candidates in the range 9 < z < 11, among which eight objects have best-fitting photometric redshifts above z = 10. Among the total sample, we identify seven robust candidates that have a narrow high-z solution with no secondary low-z solution. These candidates have about 70–90 per cent of their total probability enclosed within Δz = 1 around the best-fitting solution. The distribution of the full list of candidates in the UNCOVER field is shown in Fig. 3. Examples of these high-confidence candidates are shown in Fig. 4. These are ranked as the best candidates in the sample and are assigned a quality flag Q = 1 in Table 2. The lower quality Q = 2 sample includes galaxies that show a significant secondary solution, with a total probability within 50–70 per cent around their high-z peak. Four galaxies lie in this category. Finally, five sources belong to the Q = 3 category because of disagreements between their eazy and beagle best-fitting solutions or a total probability of less than 50 per cent around their high-z peak. Castellano et al. (2022b) recently identified seven candidates at z∼ 10 in the A2744 region. We recover six of their candidates in our sample. Five of these overlapping sources are in the Q = 1 category. One of their source, GHZ9, is ranked in the Q = 2 of our sample. Source GHZ4 in their catalogue is not included in our selection because it has SNR = 4.8, which is slightly below our colour–colour selection criteria. In Table 3, we compare the photometric redshifts of the sources that have been identified in both studies. Overall, there is a good agreement between these independent determinations, whose differences remain within the 1σ uncertainties, in most cases. Several studies have also used the GLASS imaging data to select high-redshift candidates. Object 26 928 (Q1) is also included in the high-redshift samples of Donnan et al. (2023), Naidu et al. (2022), and Bouwens et al. (2023).

We also identify three candidates in the redshift selection range 12 < z < 15 (Fig. 5). Among these, one candidate is classified as robust Q = 1, while the remaining two candidates have a Q = 3 score according to the criteria defined earlier. Object 70 846 (Q1) has been independently identified in other JWST studies (Naidu et al. 2022; Castellano et al. 2022a; Donnan et al. 2023; Harikane et al. 2023), with an estimated photometric redshift of z ∼ 12.2.

Overall, these candidates are among the highest-redshift candidates identified in recent JWST observations, as can be seen in Fig. 6. This sample spans a large dynamic range in intrinsic luminosity from M_UV = −17.6 to −21.7 mag.

We examined the recent JWST spectroscopic observations of A2744 as part of the GLASS programme, utilizing both NIRSpec, and NIRISS grism observations.

• Candidate ID 2065 in our sample has a spectroscopic confirmation at z = 9.3 as reported by Boyett et al. (2023), in good agreement with the estimated photometric redshift of |$z=9.50_{-0.08}^{+0.34}$|⁠. The NIRSpec observations of this spatially-resolved galaxy show prominent emission lines of O, Ne, and H, as well as a clear Lyman break. By combining the photometric and spectroscopic data, the best-fitting SED provides a stellar mass of log(M_⋆/M_⊙∼ 9.17).

• Candidate ID 10 619 is one of the three multiple images of a candidate galaxy previously identified in the HFF data by (Zitrin et al. 2014). It has the highest magnification (μ ∼ 11.5). It has been spectroscopically confirmed at z = 9.76 with NIRSpec prism spectroscopy by Roberts-Borsani et al. (2022). This value is in good agreement with our photometric redshift estimate of |$z=9.69_{-0.12}^{+0.33}$|⁠.

• Object ID 21 623 is a moderately magnified (μ ∼ 3.7) z ≳ 10 accreting black hole candidate. Recent deep (1.25 Ms) observations with the Advanced CCD Imaging Spectrometer on Chandra find a 4.2σ hard X-ray (2–7 keV observed) point source coincident with the position of 21623. The X-ray source is entirely undetected at softer energies, suggesting the presence of a heavily buried luminous (L_bol ∼ 5 × 10⁴⁵erg s⁻¹) quasar in this galaxy (see Bogdan et al. 2023).

• Candidates 2065, 52008, and 83 338 have tentative X-ray detections at 2.5–3σ levels.

4.2 Physical properties

In addition to computing photometric redshifts, we perform a second SED fitting run with beagle following the approach in Furtak et al. (2023) to refine our estimates of physical parameters, this time using a Gaussian prior for the photometric redshift based on the first-run photo-redshift solution. We adopt this time a more flexible SFH using a delayed exponential form SFR ∝ t exp(−t/τ), and a potential SF burst episode in the last 10 Myr. We use a constant metallicity of |$Z=0.1\, Z_{\odot }$|⁠, which has been shown to have little effect on the photometry of high-redshift galaxies (Furtak et al. 2021). We also assume an SMC extinction law, which is more appropriate for high-redshift galaxies (Capak et al. 2015; Reddy et al. 2018). We limit the fit to four physical parameters using the following priors:

• Stellar mass with a log-uniform distribution prior in the range log(M_⋆/M_⊙) ∈ [6, 10].

• SFR averaged over the last 10 Myr with a log-uniform distribution prior log(ψ/M_⊙ yr⁻¹) ∈ [−4, 4].

• Maximum stellar age for t = τ, with a log-uniform prior log(t_age yr⁻¹) ∈ [6, t_universe], where t_universe is the age of the universe at the redshift of the galaxy.

• Dust attenuation as traced by the optical depth measured in the V band with a uniform prior |$\hat{\tau }_{V} \in$| [0, 0.5]. The prior distribution is based on UV continuum slope values measured in Section 4.2.2, computed as in Furtak et al. (2023).

Most of the candidates have relatively low stellar masses in the range Log(M_⋆/M_⊙)=6.8−9.5. The data cover the redshifted Balmer break for most galaxies, which helps constrain the older stellar population. Indeed, the best constraints on the stellar mass, but also the age of the stellar population, are obtained for galaxies that show an excess in the F444W band indicative of Balmer break. For example, among the robust candidates, IDs 21623, 26 928, and 83338, show a significant Balmer break (Fig. 4) and small uncertainties on their derived stellar masses. It is also interesting to note that in comparison to eazy, and beagle attribute more broad-band flux to strong emission lines, which result in lower stellar masses. Strong [O ii]λλ3726, 3729, and [Ne iii]λ3869 emission lines can enhance both F410M and F444W fluxes and mimic a Balmer break. Such strong lines have been observed at z ∼ 10.6 for instance in the NIRSpec spectrum of GN-z11 (Bunker et al. 2023). This explains the difference observed between the eazy and beagle fits for a few objects, such as ID 21623. Candidates also show small SFR values, and young stellar ages between 10 and 100 Myr, confirming previous JWST results at similar redshifts (Furtak et al. 2021; Topping et al. 2022; Austin et al. 2023; Whitler et al. 2023). Taken at face value and considering a constant star formation, their stellar mass would imply older ages. It is clear that the current SFR derived from our SED fitting is not representative of the entire star formation history of these candidates. Intermittent episodes of intense star formation, or simply a higher SFR, likely occurred in the past in order to build up the estimated stellar mass so quickly.

4.2.1 Mass-luminosity relation

In addition to the stellar mass derived from the SED fitting, we computed the absolute rest-frame UV magnitude by combining the observed magnitude in F200W band and the photometric redshift (column 6 of Table 2). The M_⋆−M_UV relation provides insights into the stellar mass build up of galaxies and its evolution with redshift. Fig. 7 shows the M_⋆−M_UV best-fitting relation determined at z ∼ 6 (blue line; Furtak et al. 2021) and at z ∼ 9 (black line; Bhatawdekar et al. 2019) using HST and Spitzer observations. The redshift-evolution of this relation can already be observed. Properties of the present sample are shown with squares. The red squares indicate the robust subset of candidates (cf. Table 2). Our results are consistent with a redshift evolution, where galaxies are on average below the established relations at lower redshifts. They are also in agreement with other recent JWST constraints derived by Whitler et al. (2023) represented by purple circles. We note that for galaxy candidates at z ≳ 11, the absence of constraints on the Balmer break tend to underestimate the stellar mass we derive from SED fitting. Furthermore, the green shaded region and the red dashed line are theoretical predictions at z = 9 computed from hydrodynamical zoom simulations (Kannan et al. 2022) and semi-analytical models (Yung et al. 2019). Models also predict a rapid evolution with redshift, in line with our results.

4.2.2 UV continuum slopes

Next, we explore the UV continuum slope β, which is widely used to infer the dust attenuation in high-redshift galaxies. This parameter also encodes information about the age of the stellar population, where the contribution from young stars will result in a bluer UV slope. The β slope is measured by fitting a power law of the form f_λ ∝ λ^β to the rest-frame UV photometric measurements below 3000 Å in F200W, F277W, F356W bands. We fixed the redshift to the eazy best-fitting phot-z. To estimate the impact of redshift and photometric uncertainties, we performed a Monte Carlo sampling using a normal distribution around these fixed parameters in the fitting procedure. The results are provided in Table 2. Like most of the z > 9 galaxies recently uncovered in JWST observations, these candidates show blue UV slopes ranging from β = −1.8 to −2.3, which is expected if these candidates have younger stellar populations with low-dust attenuation and metallicities. Absence of dust has also been invoked to explain the high number density of high-redshift galaxies (Ferrara et al. 2022). We do not find, however evidence for extremely blue slopes like those that were reported in early JWST data (Topping et al. 2022; Adams et al. 2023; Furtak et al. 2023). When compared to literature results (see Fig. 8), these candidates show bluer UV slopes at a given luminosity than those observed in z ∼ 6 galaxies, for instance (Bouwens et al. 2014) or more recently from JWST observations (Nanayakkara et al. 2022). They follow the general trend of z > 9 galaxies (Curtis-Lake et al. 2022; Austin et al. 2023; Cullen et al. 2023; Whitler et al. 2023), for which we show only robust measurements with uncertainties below 0.3 dex. Similarly, the β-M_UV relation derived from the latest numerical simulations, such flares (Vijayan et al. 2021; Wilkins et al. 2023) and thesan (Kannan et al. 2022), show broad agreement with the observed β values.

In addition to the dependence with age, the UV slope is also affected by the nebular continuum, whose contribution is larger at longer wavelengths, which results in redder β values. Albeit with large uncertainties, these effects, combined with constraints from nebular recombination lines, can be used to indirectly infer the escape fraction of ionizing continuum f_esc in galaxies at the epoch of reionization (e.g. Zackrisson et al. 2017; Plat et al. 2019; Topping et al. 2022).

5 SUMMARY

In this paper, we presented the results of our search for z > 9 galaxy candidates in the JWST UNCOVER survey. We used deep NIRCam and NIRISS imaging from three observing programmes UNCOVER (Bezanson et al. 2022), GLASS (Treu et al. 2022), and a DD programme ID 2767, in addition to ancillary HST observations. We combined dropout selection and photometric redshift estimates from two independent codes, eazy and beagle, to identify high-redshift galaxy candidates. We report the detection of 16 candidates at 9 < z < 12 and three candidates at 12 < z < 13. According to our quality assessment, a total of seven candidates are deemed robust among this sample. Candidates span a wide dynamic range in luminosity, from M_UV ∼ −22 to −17.6 mag. Some of these sources are among the faintest galaxies discovered at z > 10, although most of their magnification factors are still relatively modest, i.e. below μ = 5. Two candidates have spectroscopic confirmation at redshift z = 9.76 and z = 9.3. Importantly, one source in this sample is an accreting black hole candidate at z ∼ 10, with a hard X-ray detection in deep Chandra observations.

In addition to photometric redshift, we ran refined SED fitting with beagle to constrain the physical properties of the sample, fixing the redshift, and focusing on four parameters: the stellar mass, the stellar age, the star formation rate averaged over the last 10 Myr, and the attenuation τ_V. We find that galaxies have young ages between 10 and 100 Myr and low star formation rates from ∼0.2 to ∼7 M_⊙ yr⁻¹. These results confirm previous findings in early JWST observations of z > 9 galaxies. Most of the candidates have low stellar masses in the range log(M_⋆/M_⊙) ∼ 6.8–9.5. We find evidence for a rapid redshift-evolution of the mass-luminosity relation, in line with recent observational results and theoretical predictions.

We also find that these galaxies have blue UV continuum slopes, between β = −1.8 and β = −2.4, although we do not find extremely blue β values as measured in recent z > 10 studies. The young ages we measure are consistent with these blue continuum slopes. We also see a redshift-evolution of the UV slope for a given intrinsic magnitude. The sample aligns with theoretical predictions and similar observational results for the β-M_UV relation at z > 9.

In the near future, JWST will continue to provide increasingly larger samples of rest-optical observations of galaxies at z > 9. Combined with follow-up spectroscopy, these data will help constrain more precisely the physical properties and star formation histories of these galaxies. In particular, ultra-deep observations of additional lensing clusters will help us push to fainter and more representative galaxies at those early epochs.

ACKNOWLEDGEMENTS

This work is based on observations obtained with the NASA/ESA/CSA JWST and the NASA/ESA Hubble Space Telescope (HST), retrieved from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute (STScI). STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5–26555. This work has made use of the CANDIDE Cluster at the Institut d’Astrophysique de Paris (IAP), made possible by grants from the PNCG and the region of Île de France through the programme DIM-ACAV+ . This work was supported by CNES, focused on the JWST mission. This work was supported by the Programme National Cosmology and Galaxies (PNCG) of CNRS/INSU with INP and IN2P3, co-funded by CEA and CNES. PD acknowledges support from the NWO grant 016.VIDI.189.162 (‘ODIN’) and the European Commission’s and University of Groningen’s CO-FUND Rosalind Franklin programme. LF and AZ acknowledge support by Grant No. 2020750 from the United States–Israel Binational Science Foundation (BSF) and grant no. 2109066 from the United States National Science Foundation (NSF). The BGU lensing group further acknowledges support by the Ministry of Science & Technology, Israel. This work has received funding from the Swiss State Secretariat for Education, Research and Innovation (SERI) under contract number MB22.00072, as well as from the Swiss National Science Foundation (SNSF) through project grant 200020_207349. The Cosmic Dawn Center (DAWN) is funded by the Danish National Research Foundation under grant no. 140.

DATA AVAILABILITY

The data underlying this article are publicly available on the Mikulski Archive for Space Telescopes² (MAST), under programme ID 2561. Reduced and calibrated mosaics are also available on the UNCOVER webpage: https://jwst-uncover.github.io/

Footnotes

References