A surprising abundance of massive quiescent galaxies at 3 < z < 5 in the first data from JWST CEERS

TL;DR

This study identifies a robust sample of 15 massive quiescent galaxies at $z>3$ in the first JWST CEERS data, including three robust systems at $z>4$, implying a higher-than-previously-thought abundance of early quenching. Using Bagpipes SED fitting with a double-power-law SFH and deep $z>3$ constraints, the authors derive stellar masses up to $\log_{10}(M_*/M_\odot)\approx11.1$ and formation redshifts $z_{\mathrm{form}}\sim9-12$ for the $z>4$ members, suggesting rapid, early mass assembly followed by quenching. They show that pre-JWST estimates underestimated the number densities by factors of $\sim3-5$ due to limited imaging depth at $\lambda>2\,\mu$m, and they compare results to previous EGS analyses, finding broader redshift solutions with earlier data. The work highlights the need for wider JWST surveys and NIRSpec follow-up to confirm redshifts and anatomize SFHs, and it points toward future discovery of $z>5$ quiescent galaxies. Overall, the findings place stronger constraints on early galaxy formation models and the efficiency and timing of quenching in the first billion years.

Abstract

We report a robust sample of 10 massive quiescent galaxies at redshift, $z > 3$, selected using the first data from the JWST CEERS programme. Three of these galaxies are at $4 < z < 5$, constituting the best evidence to date for quiescent galaxies significantly before $z=4$. These extreme galaxies have stellar masses in the range log$_{10}(M_*/$M$_\odot) = 10.1-11.1$, and formed the bulk of their mass around $z \simeq 10$, with two objects having star-formation histories that suggest they had already reached log$_{10}(M_*/$M$_\odot) > 10$ by $z\gtrsim8$. We report number densities for our sample, demonstrating that, based on the small area of JWST imaging so far available, previous work appears to have underestimated the number of quiescent galaxies at $3 < z < 4$ by a factor of $3-5$, due to a lack of ultra-deep imaging data at $λ>2\,μ$m. This result deepens the existing tension between observations and theoretical models, which already struggle to reproduce previous estimates of $z>3$ quiescent galaxy number densities. Upcoming wider-area JWST imaging surveys will provide larger samples of such galaxies and more-robust number densities, as well as providing opportunities to search for quiescent galaxies at $z>5$. The galaxies we report are excellent potential targets for JWST NIRSpec spectroscopy, which will be required to understand in detail their physical properties, providing deeper insights into the processes responsible for forming massive galaxies and quenching star formation during the first billion years.

Figures (8)

• Figure 1: Spectral energy distributions and cutout images for our three robust $z > 4$ quiescent galaxies. Our 10-band photometric data from HST ACS and JWST NIRCam are shown in blue and gold respectively. The posterior median Bagpipes models are overlaid in red. Posterior distributions for the redshifts and sSFRs of these galaxies are shown to the right of the main panels. The dashed vertical lines in the sSFR panels show the sSFR threshold for inclusion in our quiescent sample at the redshift of each object (see Section \ref{['method:sel']}). The inset RGB cutouts are composed of the F444W, F200W and F150W images respectively.
• Figure 2: Comparison of pre-JWST and new JWST CEERS fit results for object 101962, the only robust $z>4$ quiescent galaxy in our sample to be included in the CANDELS photometric catalogue of Stefanon2017. The best fit to the new JWST NIRCam data (gold) is shown in red, as in the middle panel of Fig. \ref{['fig:spectra']}. The best fit to the CANDELS data (blue) is shown in green. The corner plot to the left shows constraints on the stellar mass, redshift and sSFR of this galaxy from both datasets. It can be seen that previous data were unable to constrain these parameters, with an extremely large redshift uncertainty, $z=1.34^{+1.94}_{-0.39}$. This is largely due to the low SNR of previous data, in particular around the Balmer break, and a lack of data at $\lambda\simeq2.3-3.0\mu$m between $K_s$ and IRAC Channel 1.
• Figure 3: The rest-frame UVJ colour diagram, showing our magnitude-selected sample of massive galaxies (F200W < 26.5) in integer redshift bins spanning $3 < z < 6$. Points are coloured by sSFR, with star-forming galaxies denoted by circles without a border. Objects in our robust quiescent sub-sample are shown with black-bordered squares, whereas objects in our quiescent sample that do not meet our robust criteria (see Section \ref{['method:sel']}) are shown as black-bordered circles. SEDs for the three robust quiescent galaxies we identify at $z>4$ are shown in Fig. \ref{['fig:spectra']}, whereas the 7 robust quiescent galaxies at $3<z<4$ are shown in Fig. \ref{['fig:z3_seds']}.
• Figure 4: Number density estimates for high-redshift massive quiescent galaxies. Our estimate at $3 < z < 5$ derived from the JWST CEERS data are $3-5$ times higher than pre-JWST estimates, and, at $z\simeq3$, approach the result of McLeod2021 for the total galaxy population. Stellar masses derived by other authors have been converted to a Kroupa2001 IMF where necessary.
• Figure 5: Formation redshifts for $z>3$ massive quiescent galaxy sample, shown as a function of their observed redshifts. Our candidates at $z>4$ have formation redshifts from $9 < z_\mathrm{form} < 12$, whereas our $3 < z < 4$ galaxies are all younger, having formed at $z_\mathrm{form} < 6$.