Bridging scales: How much do supermassive black holes grow in the suppressed Bondi regime?

Abstract

The co-evolution of supermassive black holes (SMBHs) and their host galaxies remains one of the central open questions in cosmology, rooted in the coupling between accretion, feedback, and the multi-scale physics that links the event horizon to the circumgalactic medium. Here we bridge these scales by embedding a first-principles, GRMHD-informed prescription for black hole accretion and feedback--derived from multi-zone simulations that self-consistently connect inflows and outflows from the horizon to the Bondi radius--within cosmological magnetohydrodynamic zoom-in simulations of $\sim10^{14}\,M_\odot$ halos. These GRMHD results predict a "suppressed Bondi" regime in which magnetic stresses and relativistic winds strongly reduce effective accretion rates in a spin-dependent manner. We find that black holes cannot grow efficiently by accretion until they exceed $\sim10^{7}\,M_\odot$, regardless of the feedback strength. Beyond this threshold, systems bifurcate: low-spin ($η\!\sim\!0.02$) black holes continue to accrete without quenching star formation, while high-spin ($η\!\gtrsim\!0.3$) black holes quench effectively but become starved of further growth. Early, massive seeding partially alleviates this tension through merger-driven assembly, yet an additional cold or super-Eddington accretion mode appears essential to reproduce the observed SMBH population and the empirical black hole--galaxy scaling relations. Our results demonstrate that GRMHD-informed feedback models can account for the maintenance-mode behavior of low-luminosity AGN like M87*, but cannot by themselves explain the full buildup of SMBH mass across cosmic time. A unified, multi-regime framework is required to capture the evolving interplay between spin-dependent feedback, cold inflows, and mergers in driving co-evolution.

Figures (4)

• Figure 1: Stellar mass--halo mass relation (top) and black hole mass--stellar mass relation (bottom) for the light-seeding (left), fiducial-seeding (center), and early-seeding (right) models. The shaded region, which represents the empirical relation at $z = 0$, should be treated as a guide for the eye, as we overplot results from all redshifts. For seed black holes with $M_\mathrm{BH}<10^7\,M_\odot$, the black hole does not grow, and the galaxy is not quenched regardless of feedback efficiencies, resulting in excess stellar mass above $M_{\rm Halo}\gtrsim 10^{12}M_\odot$. In the fiducial seeding case, the black hole overgrows without feedback, grows moderately when the efficiency matches $a_*=0$ result, and fails to accrete when the efficiency corresponds to maximum spin ($\eta \gtrsim 0.3$); star formation is also quenched in this high-efficiency case. With early massive seeding, black holes still do not grow through accretion for $\eta \gtrsim 0.3$, but growth through mergers can reproduce the observed local black hole--stellar mass relation. Galaxy quenching is also more efficient, yielding better agreement with the local stellar mass--halo mass relation.
• Figure 2: The stellar mass (top), star formation rate (middle), and black hole accretion rate (bottom) of the most massive progenitor in the light-seeding (left), fiducial-seeding (center), and early-seeding (right) models. The shaded region shows the 16--84 percentile range of the accretion rate across all black holes in the run at each time. Black holes with $M_\mathrm{BH}\lesssim10^7\,M_\odot$ do not impact the star formation rate regardless of feedback efficiency. For $M_\mathrm{BH}\gtrsim3\times10^7\,M_\odot$, feedback efficiencies $\eta \gtrsim 0.3$ ($a_*\gtrsim0.9$) lead to efficient quenching of star formation, while efficiencies $\eta \lesssim 0.02$ remain ineffective. A feedback efficiency $\gtrsim0.3$ always suppresses black hole accretion, regardless of black hole mass.
• Figure 3: The evolution of accretion suppression (top, $\kappa \equiv \dot{M}_{\rm BH}/\dot{M}_{\rm Bondi}$), the corresponding gas temperature (middle), and the Bondi radius (bottom) for the light-seeding (left), fiducial-seeding (center), and early-seeding (right) models. The shaded regions show the full distribution and the 16--84 percentile range. In all cases, the suppression of black hole accretion relative to the Bondi rate falls within $\sim10^{-3}$--$10^{-2}$. Higher feedback efficiency produces hotter gas around the black hole and thus larger $\kappa$. More massive black holes, resulting from earlier or more massive seeding, lead to broader distributions of $\kappa$, temperature, and $r_B/r_g$, and systematically lower average $\kappa$ and temperatures.
• Figure 4: The gas temperature (upper half) and density (lower half) morphology at $z\sim1.2$ for seven runs with AGN feedback and one without. For $\eta = 0.02$, AGN feedback produces minimal changes to the gas morphology, while efficiencies $\eta \gtrsim 0.3$ generate visible outflows extending to large radii.