Evidence for the Faint-End Suppression in the z = 6 $\sim$ 8 UV Luminosity Function: A Lensing Analysis of Abell 2744

TL;DR

This study tests the faint-end behavior of the rest-frame UV luminosity function at $z=6-8$ by leveraging JWST UNCOVER imaging behind Abell 2744 and a lensing-aware framework. By tying completeness to local depth, performing source-plane delensing, and computing lens-specific effective volumes with two independent mass models, the authors compare Schechter and turnover-extended LF forms using Poisson likelihood. They find decisive evidence for a faint-end turnover with $oldsymbol{α}oldsymbol{ o} ext{-}1$ and $oldsymbol{M_T}oldsymbol{ o} ext{≈}-18$, indicating suppressed ultra-faint contributions to the UV luminosity density and implications for the ionizing photon budget during reionization. The results remain robust to lens-model choices and pass internal magnification-bias cross-checks, suggesting a reduced role for ultra-faint galaxies in driving reionization unless escape fractions are higher or brighter galaxies dominate. The analysis highlights the importance of accounting for lens-model systematics via dual-model volumes and depth-tied completeness, paving the way for broader multi-field studies with JWST.

Abstract

We determine the $z=6-8$ ultraviolet (UV) LF in the JWST UNCOVER field behind Abell 2744, through a depth-tied completeness model and source-plane selection taking multiple images into account. We compute the intrinsic $M_{UV}$ LFs and lens-dependent effective volumes $V_{eff,\ell}(M_{UV})$ for two of them (CATS, GLAFIC), and construct binned LFs with statistical (Gehrels) errors presented separately from the lens-model spread. When fitting each lens model independently with both the original and turnover-extended form (while keeping $M^*=-20.5$), we obtain decisive model selection in favour of a faint-end turnover. The peak of the turnover posteriors is at $α\simeq -1$, $β\simeq 1.6$, and drive $M_T$ to the high side of our prior ($\sim -18$) and the difference in normalization between CATS-GLAFIC is absorbed into $φ^*$. A similar magnification-bias test in the image plane gives a $B(μ)$ trend matching to what is expected, which provides an internal consistency of our selection and volumes. Our results indicate an early suppression of the UV LF at $z=6-8$, and carry implications for the ionizing contribution of ultra-faint systems during reionization.

Figures (10)

• Figure 1: Magnification maps at $z\simeq7$ (CATS vs. GLAFIC). Place in §\ref{['subsec:lens']}.
• Figure 2: Inter-model magnification differences. Spatial pattern (left) and one-point statistics (right) of the magnification ratio between CATS and GLAFIC. Disagreements are modest over most of the field and increase near critical curves where both models are less stable; those extreme-$\mu$ regions are masked for the LF inference.
• Figure 3: UNCOVER NIRCam SW depth and coverage. Color shows the 5$\sigma$ point–source depth (AB mag) for a 0.32$"$-diameter aperture. This is computed from the public UNCOVER inverse-variance weight maps and taking, at each pixel, the maximum depth among the SW bands (F115W, F150W, F200W; “best of bands”). The contours outline the union footprint of the SW coverage. We will convert the spatially varying imaging depth into source–plane effective volumes later in §\ref{['subsec:veff']}. Data products are from the UNCOVER DR2 release UNCOVER_DR2.
• Figure 4: Effective comoving search volume $V_{\rm eff}(M_{\rm UV})$ (top) and the CATS/GLAFIC ratio (bottom) for sources behind Abell 2744 over $z\simeq6\text{--}8$. Solid and dashed curves denote the CATS and GLAFIC lens models, respectively; the ordinate of the top panel is logarithmic and the abscissa is $M_{\rm UV}$ (brighter to the left). We compute $V_{\rm eff}$ by integrating the source-plane selection function $S(\boldsymbol{\beta}\,|\,M,z)=1-\prod_k\!\left[1-C\!\left(m_{{\rm app},k}(M,z)\right)\right]$, where the per-pixel completeness $C(m)$ is derived from the UNCOVER SW "best‑of" depth map (F115W/F150W/F200W) using $m_{50}=m_5-\Delta$ with $\Delta=0.35$ mag and a roll‑off $\sigma_m=0.35$ mag. Apparent magnitudes include lensing magnification via $m_{\rm app}=M+\mathrm{DM}(z)-2.5\log_{10}|\mu|$. The bottom panel shows CATS/GLAFIC, with the grey dashed line marking unity. Cosmology: Planck18.
• Figure 5: Completeness and selection function $C(m,\boldsymbol{\theta})$ (best-of SW).Left: representative completeness curves $C(m)$ for depth quantiles (P10/P35/P65/P90) across the UNCOVER mosaic. We set the 50% point by $m_{50}(\boldsymbol{\theta})=m_{5}(\boldsymbol{\theta})-\Delta$ with $\Delta=0.35$ mag and adopt an error–function roll-off of width $\sigma_m=0.35$; the per-pixel $m_5$ is taken as the best-of NIRCam SW depth among F115W/F150W/F200W. Right: pixelized completeness map $C(m_{\rm ref},\boldsymbol{\theta})$ evaluated at a reference magnitude $m_{\rm ref}\equiv{\rm median}[m_{50}(\boldsymbol{\theta})]$ (value shown in the title). This spatially varying $C(m,\boldsymbol{\theta})$ is propagated to the source plane when computing $V_{\rm eff}(M_{\rm UV})$ in our LF analysis.