Forged by Feedback: Stellar Properties of Brightest Group Galaxies in Cosmological Simulations

TL;DR

The paper compares how four cosmological simulations—Romulus, Simba, Simba-C, and Obsidian—produce brightest group galaxies (BGGs) and how their stellar masses, star formation rates, and ages compare to COSMOS X-ray selected group BGGs. By incorporating distinct AGN feedback prescriptions, the study demonstrates that feedback physics strongly shapes BGG populations, with Obsidian’s three-regime model providing the closest match to observations. Romulus exhibits inefficient quenching due to purely thermal feedback, while Simba and Simba-C show rapid quenching linked to jet activity, producing BGG populations that diverge from COSMOS in multiple aspects. The results thus emphasize the importance of physically motivated subgrid prescriptions for accurately modeling BGG growth and group environments, suggesting gradual quenching pathways are plausible in real galaxies.

Abstract

We investigate how different galaxy formation models impact the stellar properties of brightest group galaxies (BGGs) in four cosmological simulations: ROMULUS, SIMBA, SIMBA-C, and OBSIDIAN. The stellar masses, specific star formation rates, and mass-weighted stellar ages of the simulated BGGs are analysed alongside those of observed BGGs from X-ray-selected galaxy groups in the COSMOS field. We find that the global properties and underlying evolutionary pathways of simulated BGG populations are strongly impacted by the strength and mechanism of their respective active galactic nucleus (AGN) feedback models, which play a critical role in regulating the growth of massive galaxies. OBSIDIAN's sophisticated three-regime AGN feedback model achieves the highest overall agreement with COSMOS observations, matching stellar property distributions, quenched fractions, and the evolution of star formation in increasingly massive systems. We find evidence suggesting that BGG populations of OBSIDIAN and COSMOS undergo a gradual decline in star formation with stellar mass, in contrast to SIMBA and SIMBA-C, which display rapid quenching linked to the onset of powerful AGN jet feedback. By comparison, ROMULUS produces highly star-forming, under-quenched BGGs due to the inefficiency of its thermal AGN feedback in preventing cooling flows from fuelling BGG growth. The success of the OBSIDIAN simulation demonstrates the importance of physically motivated subgrid prescriptions for realistically capturing the processes that shape BGGs and their dynamic group environments.

Figures (9)

• Figure 1: Top: BGG redshift distributions. Simulated BGGs satisfying $\log(L_{\mathrm{X},\,0.1-2.4\,\mathrm{keV}}/\mathrm{erg}\,\mathrm{s}^{-1})\geq41.4$ are selected from nine snapshots to match the redshift distribution of the COSMOS sample (grey shaded histogram). The Romulus25 histogram is outlined in dotted red, Simba in dashed green, Simba-C in solid yellow, and Obsidian in dot-dashed blue. The vertical black lines mark the boundaries between redshift bins $z\in[0.08,\,0.18)$, $z\in[0.18,\,0.28)$, and $z\in[0.28,\,0.38]$. Bottom: BGG stellar mass as a function of redshift. COSMOS BGGs are shown as grey diamonds, Romulus25 BBGs as dark red hexagons, and BGGs from the Romulus zoom simulations as bright red hexagons. The Simba, Simba-C, and Obsidian BGGs are represented by $1\sigma$ and $3\sigma$ 2D histogram contour lines.
• Figure 2: $L_\mathrm{X}\!-\!M_*$ relations for observed and simulated BGGs. Top: The initial samples of simulated BGGs for all $L_\mathrm{X}$. The Simba, Simba-C, and Obsidian BGGs are represented by median lines, with outer lines and data points showing the 16th and 84th inter-percentile regions and outer scatter. The black line illustrates the COSMOS minimum $\log(L_{\mathrm{X},\,0.1-2.4\,\mathrm{keV}}/\mathrm{erg}\,\mathrm{s}^{-1})\simeq41.4$. All other formatting follows that of Figure \ref{['fig:mstar_vs_z']}. Bottom: The selected samples satisfying $\log(L_\mathrm{X}/\mathrm{erg}\,\mathrm{s}^{-1})\gtrsim41.4$. All Simba, Simba-C, and Obsidian BGGs are shown as data points in addition to their medians.
• Figure 3: BGG stellar mass distributions illustrated by normalised density histograms. The solid curves and vertical lines of corresponding colour are the $M_*$ distributions and sample medians respectively for the simulated BGG samples: Romulus25 (red, dotted median), Simba (green, dash-dotted median), Simba-C (yellow, solid median), and Obsidian (blue, dashed median). COSMOS is represented by the grey shaded distribution and solid grey median line. BGG stellar masses from the Romulus zoom simulations are shown in the bottom panel as bright red hexagons. The table contains the results of two-sample KS tests comparing the simulations' $\log(M_*/\mathrm{M}_\odot)$ distributions to that of COSMOS.
• Figure 4: BGG sSFR distributions. The top panel depicts the full BGG samples, the middle panel shows BGGs in the low-$M_*$ bin with $\log(M_*/\mathrm{M}_\odot)<11.23$, and in the bottom panel, BGGs in the high-$M_*$ bin with $\log(M_*/\mathrm{M}_\odot)\gtrsim11.23$ (see Table \ref{['tab:ngal']}). All BGGs with $\log(\mathrm{sSFR}/\mathrm{yr}^{-1})\leq-12$ are set to $\log(\mathrm{sSFR}/\mathrm{yr}^{-1})=-14$ to account for the minimum detection threshold of sSFR (see text). All other formatting follows that of Figure \ref{['fig:mstar_dist']}. The table contains the results of two-sided KS tests comparing the simulation samples' distributions of measurable $\log(\mathrm{sSFR}/\mathrm{yr}^{-1})>-12$ to that of the COSMOS sample. The top, middle, and bottom rows respectively compare $\log(\mathrm{sSFR}/\mathrm{yr}^{-1})>-12$ distributions in the full, low-$M_*$, and high-$M_*$ samples.
• Figure 5: BGG mass-weighted stellar age ($\mathrm{Age}_w$) distributions. Formatting follows that of Figure \ref{['fig:sfr_dist']}. The table contains the results of KS tests comparing the simulations' $\log(\mathrm{Age}_w/\mathrm{yr})$ distribution to that of COSMOS, where the top, middle, and bottom rows respectively compare the full, low-$M_*$, and high-$M_*$ samples.