Witnessing downsizing in the making: quiescent and breathing galaxies at the dawn of the Universe

TL;DR

This study leverages JWST ASTRODEEP photometry to systematically identify $z>3$ quiescent galaxies with $lM_*>9.5$, using an SFH-diverse SED-fitting framework to separate quiescent solutions from star-forming interlopers. It reveals a clear downsizing signal: massive galaxies at $3<z<5$ are predominantly red-and-dead, while lower-mass systems are mostly post-starburst, suggesting a breathing quenching cycle in the early Universe. The authors quantify completeness and contamination with simulated mocks and validate against spectroscopic data, finding some tension between observed high-$z$ quiescent densities and state-of-the-art simulations, especially for $z>4$, though Magneticum offers partial agreement in certain regimes. The results imply rapid, early mass assembly and quenching in massive systems, with many low-mass galaxies temporarily quiescent before rejuvenation, highlighting the need for continued JWST-era observations and refined theoretical modeling of feedback and star-formation histories. Overall, this work advances our understanding of galaxy downsizing at the dawn of cosmic time and provides a benchmark for future models of early quenching.

Abstract

[Shortened for arXiv] We conduct a systematic search for $\log(M_\ast/M_\odot) \geq 9.5$ quiescent galaxies at $z > 3$ in six extragalactic deep fields observed with NIRCam, with the goal of extracting their physical and statistical features in a uniform and self-consistent manner. We exploit the ASTRODEEP-JWST photometric catalogs to single out robust candidates, including sources quenched only a few tens of Myr before the observation. We apply a SED-fitting procedure which explores three functional forms of star formation history and the $χ^2$ probabilities of the solutions, with additional checks to minimise the contamination from interlopers, tuning our selection criteria against available spectroscopic data from the DAWN archive and simulated catalogs. We select 633 candidates, which we rank by a reliability parameter based on the probabilities of the quiescent and alternative star-forming solutions, with 291 candidates tagged as "gold". According to the best-fit models, 79\% of the massive ($\log(M_\ast/M_\odot) \geq 10.5$) quiescent galaxies at $3 < z < 5$ stopped forming stars at least 150 Myr before the time of observation, while 89\% of low-mass sources have been quenched for less than 150 Myr. The abundance of low-mass old quiescent systems does not increase significantly with time from $z = 5$ to 3: low-mass objects seem to be experiencing a short episode of quenching followed by rejuvenation (``breathing''), consistent with a downsizing scenario of galaxy formation. We also find an abrupt drop in the density of massive quiescent candidates at $z > 5$. We derive estimates for the number density of early passive galaxies up to $z = 10$ and compare them against various models: tensions with data remain in the modeling of the observed bimodality of time passed since quenching as a function of mass.

Figures (20)

• Figure 1: Values of the D4000 and DBalmer spectral indexes for theoretical models and the quiescent candidates from this work with available spectroscopic data. Passive models are color-coded by their quiescence time $dt_q$. The candidates are depicted with different symbols depending on their best-fit properties: squares for RDs, diamonds for PSs; larger symbols for the "gold" selection. Known quiescent sources from literature are marked with a black circle. The dashed and dotted lines are used to select the outliers with respect to the locus of the models, for visual inspection (see Sect. \ref{['specomp']} for details).
• Figure 2: Best-fit stellar mass vs. redshift of the quiescent candidates. Each dot represents a galaxy, and it is color-coded by the logarithm of the time passed since quenching, $dt_q$; the size of the dot is a qualitative proxy for the reliability parameter $r$ (see Sect. \ref{['selec']}). "Gold" candidates are marked with a black border. Upward-pointing triangles indicate known LRDs from the lists cited in Sect. \ref{['limit']}, while downward-pointing triangles represent candidates with a secondary solution at lower redshift. Black crosses mark the sample of known spectroscopic quiescents discussed in Sect. \ref{['specomp']}.
• Figure 3: Distribution in time passed since quenching $dt_q$ of the high (lM$_*\geq10.5$, magenta) and low mass ($9.5\leq$ lM$_*<10.5$, green) quiescent candidates at $3<z<5$. The shaded histograms correspond to the "gold" selection (the grey area being the superposition of the red and blue histograms), the lines to the full sample.
• Figure 4: Half light radius vs stellar mass relation of the candidates. Symbols are as in Fig. \ref{['zm']}. Note that radii in M24 were estimated on the detection image, a F356W+F444W stack, so they sample different physical emissions at different redshifts. The two very massive and very compact source close to the bottom right corner PRIMER-UDS ID98234, most likely a peculiar AGN, and "Capotauro" Gandolfi2025, barely visible because of its low reliability; see Sect. \ref{['wow_objs_mass']}.
• Figure 5: Number of candidates as a function of redshift adding increasingly stringent selection criteria, from the initial full catalog to the final selection (see Sect. \ref{['nordz5']}).