JWST UNCOVER: Discovery of >9 Galaxy Candidates Behind the Lensing Cluster Abell 2744

The study aims to identify and characterize galaxy candidates at very high redshift (z > 9) to probe the early stages of galaxy formation, a regime that was sparsely sampled prior to JWST due to HST’s limited NIR coverage (<2 µm). Early JWST observations (ERO, ERS programs like CEERS and GLASS) revealed unexpectedly numerous and bright z > 9 candidates, with some spectroscopically confirmed up to z ~ 13, but also instances where high-z solutions were refuted by spectroscopy in favor of dusty low-z interlopers. The reported large number densities of luminous z > 12 galaxies and massive red galaxies at z = 7–9 potentially strain standard ΛCDM-based galaxy formation models, prompting theoretical investigations invoking reduced dust attenuation, enhanced early star formation efficiency, non-standard IMFs, or alternative cosmologies. To robustly assess the prevalence and properties of z > 9 galaxies, larger-area deep JWST imaging and spectroscopy are required. The UNCOVER Treasury survey targets the lensing cluster Abell 2744 with deep multi-band NIRCam and parallel NIRISS imaging, leveraging gravitational lensing to push to fainter intrinsic luminosities. This paper presents the detection and analysis of z > 9 candidates from combined JWST and HST imaging, using dropout color selection and photometric redshifts, and examines their physical properties and redshift evolution. AB magnitudes and a standard cosmology (H0 = 70 km s^-1 Mpc^-1, ΩΛ = 0.7, ΩM = 0.3) are adopted.

Prior to JWST, only a few galaxies at z > 9 were known (e.g., Oesch et al. 2018; Bowler et al. 2020; Bagley et al. 2022a), constrained by HST’s wavelength coverage. Early JWST programs (ERO, CEERS, GLASS) rapidly expanded z > 9 samples, with candidates at z ~ 12–16 (Naidu et al. 2022; Finkelstein et al. 2022b; Atek et al. 2023; Donnan et al. 2023; Harikane et al. 2023; Austin et al. 2023), some of which were spectroscopically confirmed (Roberts-Borsani et al. 2022; Morishita et al. 2022; Curtis-Lake et al. 2022; Robertson et al. 2022; Arrabal Haro et al. 2023) while others proved to be dusty z ~ 5 interlopers. The high observed number densities and brightness of early candidates sparked debate, with tensions relative to extrapolated luminosity functions and model predictions (Bouwens et al. 2022; Mason et al. 2022; Atek et al. 2023; Naidu et al. 2022; Donnan et al. 2023), and reports of massive red galaxies at z = 7–9 (Labbe et al. 2022) raising additional challenges (Boylan-Kolchin 2022). Theoretical explanations explored evolving dust content, high star formation efficiencies, non-standard IMFs (Ziparo et al. 2022; Mason et al. 2022), and even non-ΛCDM cosmologies (Menci et al. 2022; Boylan-Kolchin 2022). Larger-area deep JWST surveys and spectroscopy are needed to clarify these issues.

Observations: UNCOVER provides deep JWST NIRCam imaging over ~45 arcmin^2 with 7 filters (SW: F115W, F150W, F200W; LW: F277W, F356W, F444W; plus F410M) with ~4–6 hr per band, achieving ~29.2 AB (5σ). Parallel NIRISS imaging includes F115W, F150W, F200W, F356W, F444W, with additional F090W coverage from GLASS in part of the field. Additional NIRCam imaging from DDT program 2756 (similar filters, no F410M) is incorporated. HST imaging from HFF (ACS: F435W, F606W, F814W; WFC3: F105W, F125W, F140W, F160W) and BUFFALO complements JWST data. All data were reduced and drizzled with grizli to a common pixel scale of 0.04″/pix and aligned. Photometric catalogs: Two catalogs were compared. (i) A general UNCOVER catalog (Weaver et al. 2023) uses ICL and bright cluster galaxy subtraction (Shipley et al. 2018), PSF-matching to F444W, detection in a co-add of F277W+F356W+F444W, SEP with detection threshold 1.5σ, deblending threshold 16 and contrast 3e-3, 0.32″ apertures, and total fluxes via Kron corrections. (ii) A custom high-z catalog optimized for small, faint sources uses SExtractor in dual mode with F444W as the detection image, detection threshold 0.9×rms, minimum area 6 pixels, deblending contrast 3e-4, 0.24″ circular apertures, total fluxes scaled from F444W AUTO flux and corrected for PSF wings using encircled-energy curves; typical total-flux increases of ~10–20%. The custom catalog provided better deblending and flux/background estimates for high-z sources. Color-dropout selection: For 9 < z < 11, candidates must satisfy: (F115W − F150W) > 1.0; (F115W − F150W) > 1.5 + 1.4(F150W − F200W); (F150W − F200W) < 0.5. For z ~ 12–15: (F150W − F200W) > 1.5; (F150W − F200W) > 1.6 + 1.8(F200W − F277W); (F200W − F277W) < 0.5. Sources must have SNR ≥ 5 in all LW bands, and be undetected (≤2σ) in bands blueward of the Lyman break (HST optical and, as available, JWST F090W and F115W). For non-detections in the dropout band, a 1σ lower limit is assigned using the band’s limiting magnitude to enforce a minimum break of 1 mag. Contaminants (quiescent/dusty low-z galaxies and cool stars/brown dwarfs) were evaluated in color–color space using template libraries (Polletta et al. 2007; Chabrier et al. 2000; Allard et al. 2001) with SMC-like dust laws. Photometric redshifts and SED fitting: Photometry includes a 5% error floor. Photometric redshifts were derived with Eazy (0.01 < z < 20) using the CORR_SFHZ templates with redshift-dependent SFHs informed by recent JWST results, which outperform default FSPS templates for high-z. BEAGLE SED fitting (Bruzual & Charlot 2003; Gutkin et al. 2016) was also run with a constant SFR SFH, uniform log(age/yr) prior in [7, age_universe], and fixed metallicity Z = 0.1 Zsun to determine the best redshift solution. Star/artefact rejection and quality control: Candidates near image edges, bright stars, with bad pixels, or with unresolved sizes (≤1 pixel) were flagged. Stellar contaminants were flagged using SExtractor CLASS_STAR and Eazy chi^2 fits to dwarf-star templates (fit_phoenix_stars). All candidates were visually inspected for contamination (diffraction spikes, detector artifacts). Lensing and effective area: Magnifications and survey area were computed with the UNCOVER strong-lensing model (Furtak et al. 2022). The area with amplification μ > 2 is ~3.5 arcmin^2 (vs. ~0.9 arcmin^2 from HFF). Candidate selection workflow: Dropout-selected candidates were cross-checked with photometric-redshift solutions from both Eazy and BEAGLE; additional photo-z selected objects with good fits (chi^2 < 30) were tested against dropout criteria to exclude failures (lack of Lyman break, blue-band detections, insufficient S/N). Final sample sizes and quality flags were assigned based on the sharpness and dominance of the high-z PDF peak.

• Sample: 16 candidates at 9 < z < 11 (8 with best-fit z > 10) and 3 candidates at 12 < z < 15.
• Robustness: 7 robust candidates (Q=1) with narrow, dominant high-z solutions (70–90% of PDF within Δz = 1 around the best-fit). Four Q=2 with significant secondary solutions, and five Q=3 due to Eazy/BEAGLE disagreements or weak PDF dominance.
• Spectroscopic confirmations: Two candidates confirmed—ID 2065 at z = 9.3 (Boyett et al. 2023) consistent with photometric z = 9.50+0.34−0.08; and ID 10619 (a multiple image) at z = 9.76 (Roberts-Borsani et al. 2022), consistent with photometric z = 9.69+0.33−0.12.
• Lensing: Magnification factors span μ = 1.2–11.5; area with μ > 2 is ~3.5 arcmin^2.
• Photometric depth: NIRCam ~29.2 AB (5σ); total JWST+HST coverage ~45 arcmin^2.
• Physical properties: Lensing-corrected stellar masses log(M*/Msun) ≈ 6.8–9.5; SFRs ~0.2–7 Msun yr^-1; young stellar ages ~10–100 Myr. Several z ~ 9–10 galaxies show evidence of a Balmer break between F356W and F444W/F410M (e.g., IDs 21623, 26928, 83338), tightening mass constraints. BEAGLE tends to attribute more flux to strong emission lines (e.g., [O II], [Ne III]), yielding lower M* than Eazy in some cases.
• UV continuum slopes: Beta ≈ −1.8 to −2.3 (blue but not extremely blue), consistent with low dust and young ages, and broadly aligned with recent JWST measurements for z > 9.
• Scaling relations and evolution: Evidence for a rapid redshift evolution of the M*–MUV relation—candidates lie below relations established at lower redshift (z ~ 6, z ~ 9). UV slope–MUV relation also shows redshift evolution, with bluer slopes at fixed luminosity than typical z ~ 6 galaxies. Observations broadly agree with predictions from recent simulations and semi-analytic models.
• Luminosities: Intrinsic UV absolute magnitudes span about M_UV ≈ −17.6 to −21.7 mag (some among the faintest known at z > 10).

The combination of color-dropout criteria and photometric redshifts identifies a robust set of z > 9 candidates behind Abell 2744, with gravitational lensing enabling access to fainter intrinsic magnitudes. The derived low stellar masses, modest SFRs, and young ages, together with blue UV continuum slopes, are consistent with expectations for early galaxies with low dust content and low metallicities. Apparent Balmer breaks in several z ~ 9–10 objects provide leverage on older stellar components and M* estimates, though strong nebular lines can mimic breaks in broadband filters and must be carefully modeled. The observed positions of candidates relative to the M*–MUV relation suggest rapid evolution with redshift, in line with contemporary theoretical models. Similarly, the UV slope–luminosity trend indicates bluer populations at higher redshift for a given MUV. While the results support an emerging consensus from early JWST studies, spectroscopic confirmation remains limited and critical for resolving residual contamination and constraining nebular contributions. Deeper observations are needed to further probe the faint-end population at these epochs and refine constraints on star-formation histories and dust content.

Deep JWST NIRCam and NIRISS imaging from UNCOVER (augmented by GLASS and DDT data) combined with HST imaging reveals 19 z > 9 galaxy candidates behind Abell 2744 (16 at 9 < z < 11 and 3 at 12 < z < 15), with 7 robust (Q=1) cases and two spectroscopically confirmed at z ≈ 9.3 and 9.76. Candidates exhibit blue UV slopes (beta ~ −1.8 to −2.3), young ages (~10–100 Myr), low SFRs (~0.2–7 Msun yr^-1), and low stellar masses (log M* ~ 6.8–9.5). Several show evidence for Balmer breaks. The sample indicates a rapid redshift evolution in the M*–MUV relation and in the UV slope–luminosity relation, broadly agreeing with model predictions. These results reinforce trends seen in early JWST observations of z > 9 galaxies. Future work should focus on deeper JWST imaging, additional massive-lens fields to reach fainter intrinsic luminosities, and extensive spectroscopy to confirm redshifts, disentangle nebular-line contributions, and better constrain star-formation histories and dust properties.

• Reliance on photometric redshifts and color-dropout selection introduces potential contamination from dusty/quiescent low-z galaxies and cool stars, despite careful color–color vetting and star-template fitting.
• Limited spectroscopic follow-up: only two candidates are confirmed; most redshifts and physical parameters remain model-dependent.
• SED-fitting assumptions (SFH forms, metallicity, dust law) and template choices (e.g., treatment of nebular emission) affect inferred M*, SFR, ages, and can bias interpretations (e.g., Balmer break vs. emission-line boosts).
• Gravitational lensing magnification uncertainties propagate into intrinsic luminosities, M*, and SFR; although unlensed fits were used initially, posterior de-lensing still inherits lens-model systematics.
• The highest-z candidates lack strong constraints on the Balmer break, increasing M* uncertainties.
• Selection thresholds (SNR cuts, non-detection criteria) may bias against lower-surface-brightness or marginally detected sources.
• Survey area, while enhanced by lensing (~45 arcmin^2 total with ~3.5 arcmin^2 at μ > 2), remains modest, limiting statistical power and cosmic variance assessment.