Downsizing does not extend to dwarf galaxies: identifying the stellar mass regimes shaped by supernova and AGN feedback

TL;DR

This study tests whether the downsizing paradigm extends into the dwarf galaxy regime by measuring the red/quenched fraction across $M_{\\star} \sim 10^{7}$–$10^{9.5}\,M_{\odot}$ using a mass-complete sample from COSMOS2020 at $z<0.15$. It combines deep multi-wavelength data, morphology from JWST/HSC/HST, and comparisons to NewHorizon and TNG50 simulations, with robust environment metrics based on photometric redshift PDFs. The key finding is a non-monotonic U-shaped relation, with a minimum red fraction near $M_{\\star} \sim 10^{8.5}\,M_{\odot}$; quenching below this mass is SN-dominated, while above it a mix of SN and AGN feedback (and possibly morphological quenching) suppresses star formation. The results challenge the notion of uninterrupted downsizing into dwarfs and show that current simulations diverge in reproducing the observed trend, highlighting the red fraction as a powerful calibration tool for galaxy formation models and motivating future deep-wide surveys to improve statistics in the dwarf regime.

Abstract

We explore how the fraction of red (quenched) galaxies varies in the dwarf galaxy regime (10^7 MSun < Mstar < 10^9.5 MSun), using a mass-complete sample of ~5900 dwarfs at z<0.15, constructed using deep multi-wavelength data in the COSMOS field. The red fraction decreases steadily until Mstar ~ 10^8.5 MSun and then increases again towards lower stellar masses. This 'U' shape demonstrates that the traditional notion of 'downsizing' (i.e. that progressively lower mass galaxies maintain star formation until later epochs) is incorrect -- downsizing does not continue uninterrupted into the dwarf regime. The U shape persists regardless of environment, indicating that it is driven by internal processes rather than external environment-driven mechanisms. Our results suggest that, at Mstar < 10^8 MSun, the quenching of star formation is dominated by supernova (SN) feedback and becomes more effective with decreasing stellar mass, as the potential well becomes shallower. At Mstar > 10^9 MSun, the quenching is driven by a mix of SN feedback and AGN feedback (which becomes more effective with increasing stellar mass, as central black holes become more massive). The processes that quench star formation are least effective in the range 10^8 MSun < Mstar < 10^9 MSun, likely because the potential well is deep enough to weaken the impact of SN feedback, while the effect of AGN feedback is still insignificant. The cosmological simulations tested here do not match the details of how the red fraction varies as a function of stellar mass -- we propose that the red fraction vs stellar mass relation (particularly in the dwarf regime) is a powerful calibrator for the processes that regulate star formation in galaxy formation models.

Figures (7)

• Figure 1: Example images of ETGs (row 1), LTGs (row 2), galaxies classified as compact (row 3) and those flagged as unclassifiable (row 4). The size of each image is 10 arcseconds on a side. The images from different instruments have different orientations - we intentionally keep it this way so that each galaxy is visually inspected at different orientations. In each image the first, second and third columns show the JWST, HSC and HST images respectively. Each individual image has six panels. Within each image the panels in the first row show the original images from the different instruments, while the second row shows their unsharp-masked counterparts.
• Figure 2: The galaxy red fraction calculated using the rest-frame $(g-i)$ colour, as described in Section \ref{['sec:red_fraction']}, in COSMOS2020 (red) and the SDSS NSA (blue).
• Figure 3: The galaxy red fraction for different terciles in each of the environment metrics described in Section \ref{['sec:environment_metrics']}. The lower, middle and upper terciles contain galaxies in the 0 - 33rd, 33rd - 66th and 66th - 100th percentile values of the metrics in question. The top, middle and bottom panels show the red fraction as a function of the simple and mass-weighted number density and tidal stress respectively. The metrics are calculated within 0.5 Mpc radius apertures. The grey shaded region shows the red fraction of the parent sample of galaxies. Uncertainties are calculated following Cameron2011.
• Figure 4: The galaxy red fraction as a function of the different morphological classes described in Section \ref{['sec:morphological_classification']} - all galaxies in the JWST footprint (black) ETGs only (red), ETGs and compact object (orange) and LTGs (blue). The grey shaded region shows the red fraction of the parent sample of galaxies. Uncertainties are calculated following Cameron2011.
• Figure 5: Comparison of the red fraction in COSMOS2020 (grey shaded region) to predictions from the NewHorizon (red) and TNG50 (blue) cosmological simulations. Uncertainties are calculated following Cameron2011.