COSMOS2020: Discovery of a protocluster of massive quiescent galaxies at z = 2.77

The study investigates how galaxy properties, particularly quenching of star formation, relate to environment at high redshift. While protoclusters—progenitors of present-day massive clusters—have been identified out to z ~ 7–8, they have largely been discovered via overdensities of star-forming tracers (UV-bright, Lyα, Hα, IR-selected galaxies). Meanwhile, spectroscopic confirmations of massive quiescent galaxies now extend to z ~ 4, yet the environmental dependence of quiescence remains unclear. Identifying overdense structures through quiescent galaxies alone has been challenging due to their low number density at z ≥ 3. Leveraging the wide and deep multi-band COSMOS2020 dataset, the authors aim to identify overdense regions of quiescent galaxies at z ~ 3 and to spectroscopically confirm their nature, thereby probing whether mature, quiescently dominated protoclusters exist less than 2.5 Gyr after the Big Bang.

Prior protocluster discoveries (e.g., Steidel et al. 1998; Overzier et al. 2008; Toshikawa et al. 2018; Shimasaku et al. 2003; Jiang et al. 2018; Harikane et al. 2019; Hayashi et al. 2012; Darvish et al. 2020; Oteo et al. 2018; Miller et al. 2018) predominantly use star-forming galaxy overdensities as tracers. Massive quiescent galaxies at high redshift have been identified and spectroscopically confirmed to z > 3–4 (e.g., Kriek et al. 2009; Glazebrook et al. 2017; Forrest et al. 2020; Valentino et al. 2020; Tanaka et al. 2019; D'Eugenio et al. 2021; Marchesini et al. 2022), with evidence for rapid early star formation and quenching (Schreiber et al. 2018; Belli et al. 2019; Saracco et al. 2020). Environmental quenching at high-z remains debated; some works report elevated quiescent fractions in protoclusters or individual QGs within them (Spitler et al. 2012; McConachie et al. 2022; Kubo et al. 2021; Kalita et al. 2021). Photometric hints of red sequences in overdensities at z ~ 2–3 have been noted (Kodama et al. 2007; Uchimoto et al. 2012; Koyama et al. 2013), but spectroscopically confirmed red-sequence structures at z ~ 3 were lacking. The COSMOS surveys (Scoville et al. 2007; Laigle et al. 2016; Weaver et al. 2022) enable selection of QGs with improved photometric redshifts over large areas, facilitating searches for quiescent-dominated structures.

• Photometric selection and overdensity mapping: Quiescent galaxies at z ~ 3 were selected from the COSMOS2020 Classic catalog using SED fitting with MIZUKI across 40 bands (u to IRAC ch4). Selection criteria: (1) Ks (UltraVISTA, 3-arcsec aperture) < 24.8 mag (3σ in deep stripes) to homogenize depth; (2) quiescent defined by 1σ upper limit on sSFR from SED fitting: log(sSFR_1σ,upper/yr^-1) < -9.5; (3) stellar mass cut log(M/M☉) > 10.3 (90% completeness for QGs at z ~ 2.8 given the Ks limit). An overdensity map was constructed via Gaussian kernel density estimation within redshift slices of width δz = ±0.1 (comparable to photo-z scatter for z > 2 QGs), smoothing with a Gaussian kernel of σ = 3 arcmin (~5.5 cMpc at z ~ 2.8), masking regions around bright stars (COSMOS2020 FLAG_COMBINED), and correcting edge/mask effects following Chartab et al. (2020), ignoring areas needing corrections > 1.67. A significant QG overdensity was found in 2.74 < z_phot < 2.94 over ~14×8 arcmin^2 (~7×4 pMpc^2) with a 4.2σ peak and 14 QGs (all with log M > 10.5). Star-forming galaxies showed at most ~1σ overdensity and ~0σ at the QG peak. The QG fraction at log M > 10.3 is 0.34 ± 0.11 vs field 0.129 ± 0.009.
• Spectroscopy: Keck/MOSFIRE H-band observations targeted 9 QGs within the overdensity using two masks (centers at 10:00:08.26,+01:40:58.64 observed Apr 19, 2022; and 10:00:22.82,+01:41:36.54 on Apr 20, 2022). Individual exposures 120 s with ABBA nodding; total integration times 3.38 hr and 2.65 hr. Reduction with MOSFIRE DRP; flux calibration using A0V stars; slit-loss correction via UltraVISTA H-band magnitudes; noise estimated from blank-sky pixels. Redshifts were measured by fitting stellar continuum templates (as in EAZY) with SLINEFIT, convolved with Gaussian velocity dispersion from pPXF; where σ* was unavailable, an empirical M*–σ* relation (Schreiber et al. 2018; Belli et al. 2017) was used. Fit redshift range 2 < z < 4; skyline-contaminated pixels masked; spectra rebinned by 4× to 6.5 Å when needed.
• Additional analyses: Number density and overdensity computed assuming a cube volume from transverse (~4×1 pMpc^2) and line-of-sight dz ~ 0.03 (≈ 2400 km/s), yielding a structure volume ~1.1×10^3 cMpc^3. Color–magnitude diagram (J−Ks vs Ks; 2-arcsec aperture J−Ks and Ks Kron) assessed for red sequence; single-burst BC03 models with solar metallicity used to estimate formation redshift. Halo mass estimated via: (1) stellar-to-halo mass relation (Behroozi et al. 2013; Shuntov et al. 2022) using the most massive member; and (2) sum of stellar masses (excluding the widely separated QG280611) converted to M_halo via z ~ 1 relation (van der Burg et al. 2014). Comparison to structures in IllustrisTNG TNG-300 at z ≈ 2.73 used sSFR threshold log sSFR < -9.5 and M* > 10^11 M☉, searching for systems with ≥3 QGs within 1 pMpc plus at least one additional QG within 8 pMpc; halo mass evolution traced via merger trees (Group_M_TopHat200).

• Spectroscopic confirmation: Of 9 targets, four quiescent galaxies show multiple Balmer absorption lines with robust spectroscopic redshifts tightly clustered at z = 2.760–2.788. The integrated probability that the true redshift lies within ±0.01 of the best-fit is 96.4–100% for all four.
• Mass and star formation: All four have high stellar masses log(M/M☉) = 11.08–11.53 and low specific SFRs from SED fitting (−10.6 < log sSFR/yr^-1 < −9.7), >1 dex below the star-forming main sequence at this redshift. For QG281561, ALMA Band 7 non-detection implies log SFR(M☉/yr) < 1.74 (3σ), consistent with quiescence.
• Spatial concentration: The four QGs are concentrated within ~4×1 pMpc^2 transversely and dz ~ 0.03 (≈2400 km/s), with three lying within 1×1 pMpc^2 at virtually identical redshift (z ≈ 2.76).
• Overdensity: Assuming a cube-like volume (~1.1×10^3 cMpc^3), the number density within the structure is ~3.4×10^-3 cMpc^-3. In the COSMOS field, the average density of QGs with log M > 11.0 at 2.5 < z_phot < 3.0 is (4.7 ± 0.3)×10^-5 cMpc^-3 (360 galaxies). Thus, QO-1000 is 72.1 ± 3.8 times denser; conservatively, a lower limit of > 68× is adopted.
• Quiescent fraction and red sequence: The quiescent fraction at log M > 10.3 within the overdensity (0.34 ± 0.11) is ~3× higher than the COSMOS-wide average (0.129 ± 0.009). Members form a clear red sequence at J−Ks ≈ 1.85, tighter for spectroscopic members; a single-burst BC03 model suggests formation redshift z_form ≈ 3.7 (solar metallicity).
• Halo mass: From stellar-to-halo mass conversions, M_halo lower limits are log(M_halo/M☉) > 13.5 (Behroozi et al. 2013) or > 13.2 (Shuntov et al. 2022). Using the summed stellar mass of the central three spectroscopic members (excluding distant QG280611) and the z ~ 1 scaling (van der Burg et al. 2014) yields log(M_halo/M☉) ≈ 13.6. The study adopts a conservative lower limit log(M_halo/M☉) > 13.2.
• Simulation comparison: In TNG-300 at z ≈ 2.73, two analogous structures match QO-1000-like criteria; their host halos grow to log(M_halo/M☉) ≈ 14.8–15.0 by z = 0, exceeding the 84th percentile for halos hosting field QGs of similar stellar mass, consistent with QO-1000 evolving into a Coma-like massive cluster.

The concentration of four massive, spectroscopically confirmed quiescent galaxies at z ≈ 2.77, the elevated quiescent fraction, and the presence of a red sequence indicate that QO-1000 is a mature protocluster core in which quenching is already prevalent by z ~ 3. The overdensity (>68× field) in quiescent galaxies suggests strong environmental processes or accelerated galaxy evolution within a massive halo. Halo mass estimates (log M_halo > 13.2) and comparison to IllustrisTNG imply QO-1000 will evolve into a present-day very massive cluster (log M_halo ≳ 14.8). The likely evolutionary link is from z ~ 4 protoclusters dominated by dusty star-forming galaxies (with similarly high stellar masses) transitioning into quiescently dominated structures like QO-1000 as star formation rapidly declines. The compact triplet of QGs may trace the nascent core of the eventual cluster. This discovery provides direct evidence that environmentally influenced quenching and red-sequence assembly can be well underway at z ~ 3.

The study reports the discovery and spectroscopic confirmation of a quiescently dominated protocluster, QO-1000, at z = 2.77 in the COSMOS field. Identified as a 4.2σ photometric overdensity of 14 QGs, four massive members (log M > 11) have robust spectroscopic redshifts (z = 2.760–2.788) with low sSFRs, yielding a quiescent galaxy number density >68 times higher than the field. The structure exhibits a red sequence (J−Ks ≈ 1.85) and an elevated quiescent fraction (~3× field). Halo mass estimates place a conservative lower limit at log(M_halo/M☉) > 13.2, and analogs in IllustrisTNG are predicted to evolve to log(M_halo/M☉) ~ 14.8–15.0 by z = 0. QO-1000 thus represents a more mature protocluster at z ~ 3, likely in transition between star-forming protoclusters and quenched clusters. Future work will expand spectroscopy, constrain star-formation histories, analyze morphologies (e.g., with HST/3D-DASH), derive dynamical masses, and refine comparisons with simulations.

• Spectroscopic confirmation is limited to four quiescent members; the remaining photometric candidates lack spectroscopic redshifts, leaving membership uncertain.
• Spectral S/N constraints required rebinning for some targets, potentially limiting precision in kinematic measurements (to be presented elsewhere).
• The overdensity estimate assumes a simple cubical volume and is based only on confirmed QGs; true density may be higher but remains uncertain.
• Halo mass estimates rely on scaling relations (stellar-to-halo mass conversions; z ~ 1 cluster relations) and thus provide lower limits with systematic uncertainties.
• Star-forming galaxy overdensity is low; absence of a SF overdensity might reflect selection limits, depth variations, or environmental stage.
• Possible additional red galaxies near the structure with slightly offset photometric redshifts were not included as members due to photo-z uncertainties.