2025MNRAS.539..557B

We use NIRSpec/MSA (Micro Shutter Assembly) spectroscopy and NIRCam (Near-Infrared Camera) imaging to study a sample of 18 massive ($\log M_\star /\mathrm{M}_\odot \gt 10$ dex), central quiescent galaxies at $2\le z \le 5$ in the GOODS (Great Observatories Origins Deep Survey) fields, to investigate their number density, star formation histories, quenching time-scales, and incidence of active galactic nuclei (AGN). The data depth reaches $\log M_\star /\mathrm{M}_\odot \approx 9$ dex, yet the least-massive central quiescent galaxy found has $\log M_\star /\mathrm{M}_\odot \gt 10$ dex, suggesting that quenching is regulated by a physical quantity that scales with $M_\star$. With spectroscopy, we assess the completeness and purity of photometric samples, finding number densities 10 times higher than predicted by galaxy formation models, confirming earlier photometric studies. We compare our number densities to predictions from FLAMINGO (Full-Hydro Large-scale Structure Simulations with All-sky Mapping for the Interpretation of Next Generation Observations), the largest box full-hydro-simulation suite to date. We rule-out cosmic variance at the 3$\sigma$ level, providing spectroscopic confirmation that galaxy formation models do not match observations at $z>3$. Using FLAMINGO, we find that the vast majority of quiescent galaxies' stars formed in situ, with these galaxies not having undergone multiple major dry mergers. This is in agreement with the compact observed size of these systems and suggests that major mergers are not a viable channel for quenching most massive galaxies. Several of our observed galaxies are old, with four displaying 4000 Å breaks with formation and quenching redshifts of $z\ge 8$ and $\ge 6$. Using tracers, we find that eight galaxies host AGN, including old systems, suggesting a high AGN duty cycle with a continuing trickle of gas to fuel accretion.