Exploring over 700 massive quiescent galaxies at z = 2-7: Demographics and stellar mass functions

TL;DR

This study assembles a large, photometrically selected sample of $N=743$ massive quiescent galaxies with $M_\star>10^{9.5}\,M_\odot$ across $z=2-7$ using $>$800 arcmin$^2$ of JWST data, enabling robust measurements of number densities, SMFs, and cosmic stellar mass density. By combining expanded UVJ-like colour selection with Bayesian SED fitting and careful contaminant rejection, the authors quantify strong redshift evolution and field-to-field variance, and they compare to state-of-the-art simulations. They find a persistent overabundance of high-$z$ quiescent galaxies relative to simulations, with no model reproducing the observed SMFs across $z=2-5$; selection biases and measurement scatter further shape the inferred SMFs. The results imply that current feedback prescriptions do not fully capture high-$z$ quenching and that galaxy quenching becomes dramatically more important between $z>5$ and $z\sim2$, motivating improved simulations and wide-area JWST surveys to resolve the discrepancies.

Abstract

High-redshift ($z>2$) massive quiescent galaxies are crucial tests of early galaxy formation and evolutionary mechanisms through their cosmic number densities and stellar mass functions (SMFs). We explore a sample of 743 massive ($\rm M_*> 10^{9.5}M_\odot$) quiescent galaxies from $z=2-7$ in over 800 arcmin$^2$ of NIRCam imaging from a compilation of public JWST fields (with a total area $>$ 5 $\times$ previous JWST studies). We compute and report their cosmic number densities, stellar mass functions, and cosmic stellar mass density. We confirm a significant overabundance of massive quiescent galaxies relative to a range of cosmological hydrodynamical simulations and semi-analytic models (SAMs). We find that no simulations or SAMs accurately reproduce the SMF for massive quiescent galaxies at any redshift within the interval $z=2-5$. This shows that none of these models' feedback prescriptions are fully capturing high-z galaxy quenching, challenging the standard formation scenarios. We find a greater abundance of lower-mass ($\rm M_*<10^{10}M_\odot$) quiescent galaxies than previously found, highlighting the importance of sSFR cuts rather than simple colour selection. We show the importance of this selection bias, alongside individual field-to-field variations caused by cosmic variance, in varying the observed quiescent galaxy SMF, especially at higher-z. We also find a steeper increase in the cosmic stellar mass density for massive quiescent galaxies than has been seen previously, with $ρ_*\propto (1+z)^{-7.2\pm0.3}$, indicating the dramatic increase in the importance of galaxy quenching within these epochs.

Figures (12)

• Figure 1: Upper panel: U-V and V-J colour selection colour-coded by redshift for the massive quiescent galaxies in our full sample. The black line corresponds to the Schreiber2015 quenching criterion with the grey line addition of the Belli2019 fast quenching criterion. The red dotted line corresponds to the selection criteria from Baker2025a which we use as our initial selection criteria in this work.
• Figure 2: Quiescent galaxy number densities of those with $\mathrm{M}_\star$>$\rm 10^{9.5}M_\odot$ from our sample (red points). We compare to Carnall2023 green points, Valentino2023 blue points, and the single galaxy from Weibel2024qgal as the grey point. We offset the points ever so slightly in redshift for the Valentino2023 and Carnall2023 number densities so as to make their errors visible within the figure.
• Figure 3: Quiescent galaxy number densities for $\mathrm{M}_\star$>$\rm 10^{10}M_\odot$ from our sample (red points). Upper panel: a comparison to other observational studies, Carnall2023, green points, a spectroscopically corrected sample, Baker2025a, orange points, a spectroscopic sample, Nanayakkara2025, light green point, and the Weibel2024qgal single galaxy point. Middle panel: a comparison to 5 cosmological hydrodynamical simulations, Eagle, Illustris-TNG100, Simba, Flamingo, and MagneticumLagos2025Baker2025aRemus2025 and 2 SAMs, Shark and Galform. Bottom panel: 4 cosmological hydrodynamical simulations and 2 SAMs, but with Gaussian scatter included in the stellar masses and star-formation rates. We see that simulations struggle to reproduce the observed number densities at most redshifts.
• Figure 4: Upper: observed stellar mass function ($\Phi$) versus stellar mass for massive quiescent galaxies in bins of redshift. Lower: Fitted stellar mass function ($\Phi$) versus stellar mass for massive quiescent galaxies in bins of redshift. The solid line is our best-fit stellar mass functions including the correction for Eddington bias, the dotted lines exclude the Eddington bias correction. The dotted lines show that we would significantly overestimate the number of most massive quiescent galaxies at $z=2-2.5$ without the correction.
• Figure 5: Comparison between our observed stellar mass function (solid lines) at $z=2-2.5$ (upper panel) $z=2.5-3$ (upper middle panel), $z=3-3.5$ (lower middle panel) and $z=4-5$ (lower panel), alongside 6 cosmological simulations denoted by dashed lines Lagos2025. The right hand panel shows the same, but with Gaussian scatter added to the stellar masses and SFRs for the 6 simulations. We see that the simulations cannot accurately reproduce the observed SMFs.