Simulating galaxy formation with black hole driven thermal and kinetic feedback

TL;DR

The paper introduces a two-mode AGN feedback model in cosmological simulations, using thermal energy at high accretion and stochastic kinetic winds at low accretion to suppress star formation in massive halos without depleting gas excessively. Implemented in AREPO, the model shows self-regulated black hole growth, red and dead massive galaxies, and gas fractions and thermodynamic profiles in line with observations, resolving key tensions seen in earlier Illustris-type runs. The authors perform extensive parameter studies and resolution tests, finding overall robustness but highlighting critical dependence on seeding, threshold scaling, and wind burstiness. Together, the results demonstrate that kinetic feedback at low accretion is a crucial, physically motivated mechanism for galaxy quenching and provides a solid foundation for next-generation, large-volume simulations of galaxy formation. The work advances our ability to connect SMBH physics to observable galaxy properties across the mass spectrum and cosmic time.

Abstract

The inefficiency of star formation in massive elliptical galaxies is widely believed to be caused by the interactions of an active galactic nucleus (AGN) with the surrounding gas. Achieving a sufficiently rapid reddening of moderately massive galaxies without expelling too many baryons has however proven difficult for hydrodynamical simulations of galaxy formation, prompting us to explore a new model for the accretion and feedback effects of supermassive black holes. For high accretion rates relative to the Eddington limit, we assume that a fraction of the accreted rest mass energy heats the surrounding gas thermally, similar to the `quasar mode' in previous work. For low accretion rates, we invoke a new, pure kinetic feedback model which imparts momentum into the surrounding gas in a stochastic manner. These two modes of feedback are motivated both by theoretical conjectures for the existence of different types of accretion flows as well as recent observational evidence for the importance of kinetic AGN winds in quenching galaxies. We find that a large fraction of the injected kinetic energy in this mode thermalises via shocks in the surrounding gas, thereby providing a distributed heating channel. In cosmological simulations, the resulting model produces red, non star-forming massive elliptical galaxies, and achieves realistic gas fractions, black hole growth histories and thermodynamic profiles in large haloes.

Figures (15)

• Figure 1: Thin projection ($5$ kpc in depth, $25$ kpc on a side) of the $256^3$, $n=10^{-1}\,\text{cm}^{-3}$, $T=10^7\,\text{K}$ simulation after $5\,\text{Myr}$ of evolution. The panels show volume weighted density (top left), volume-weighted temperature (top right), absolute velocity (bottom left), and energy dissipation weighted Mach number (bottom right).
• Figure 2: Evolution of the different energy components after kinetic energy injection. The dotted lines show individual injection events, the solid line their average, both in the simulation initially with $32^3$ cells. The dashed line shows the average of the high resolution test with $256^3$ initial cells. On average, half of the feedback energy that was initially in kinetic form is thermalized after $0.5$ Myr. This behaviour is converged at the resolution of cosmological simulations.
• Figure 3: Energy dissipation as a function of shock Mach number $\mathcal{M}$ summed up over a simulation time of $5\,\text{Myr}$. The different colours denote different isobaric variations of the gas from relatively cool, dense ($1 \,\text{cm}^{-3}$, $T= 10^{6}\,\text{K}$) to hot, dilute ($\rho=10^{-3}\,\text{cm}^{-3}$, $T= 10^{9}\, \text{K}$). The solid lines show the simulation with an initial grid of $32^3$ cells, comparable to the resolution of cosmological simulations, while the dashed lines indicate simulations with $256^3$ cells to show the convergence of the analysis.
• Figure 4: Average ratio of the mass of haloes in the full physics simulation to the mass of the corresponding halo in the dark matter only run, as a function of halo mass. The shaded region indicates the $1\sigma$ scatter of the results of our default simulation. The dashed grey line represents the result for the Illustris simulation 2014MNRAS.444.1518V.
• Figure 5: Black hole mass as a function of bulge mass (upper plot) and stellar mass within twice the half mass radius (lower plot) for central galaxies in the high-resolution simulation. The size of the symbols is scaled with bulge mass for better visibility, and the assigned colour scale encodes the Eddington ratio. The dotted line is the fit to observational data. The symbols with error bars are observed ellipticals (black), and spirals or S0 galaxies with normal bulges (green), taken from 2013ARAA..51..511K. The bulge mass is estimated as twice the mass of the counterrotating fraction of stars within $0.1\,R_{200,c}$. We note that this might slightly underestimate the bulge mass in the case of rotating bulges.