Supermassive Black Hole Growth in Massive Galaxies at Cosmic Dawn

TL;DR

This work probes how supermassive black holes grow in the most massive galaxies at Cosmic Dawn using RAMSES-based cosmological zoom-in simulations. By varying SMBH seed mass, maximal accretion rate, and AGN feedback strength, the study shows that SMBH growth is highly sensitive to the turbulent, multiphase ISM shaped by stellar feedback, producing stochastic feast and starvation accretion. Self-regulation via AGN feedback can occur when the SMBH becomes massive enough to heat and expel surrounding gas, but galaxy-scale quenching does not emerge in the explored parameter space. To reconcile with JWST-observed high-redshift AGN, the authors find that a seed mass ≳10^4 M⊙, mildly super-Eddington accretion (λ_edd ≳ 3), and very inefficient feedback coupling (ε_c ≲ 0.15%) are favorable, though resolution and missing physics (radiative transfer, non-equilibrium chemistry) remain critical factors for robust predictions.

Abstract

Among the emerging excess of massive, bright galaxies at Cosmic Dawn $z \gtrsim 9$ seen by the James Webb Space Telescope, several exhibit spectral features associated with active galactic nuclei (AGN). These AGN candidates suggest that supermassive black holes (SMBHs) grow rapidly in the early Universe. In a series of numerical experiments, we investigate how SMBHs grow within and influence the most massive galaxies at Cosmic Dawn using cosmological hydrodynamic zoom-in simulations run with the adaptive mesh refinement code RAMSES. Our suite of simulations explore how super-Eddington accretion, seed mass, and the strength of feedback influence SMBH-galaxy co-evolution in the early Universe. We find that SMBH growth is sensitive to stellar feedback which generates a turbulent-multiphase interstellar medium (ISM) that stochastically starves the SMBH. In the absence of AGN feedback, we find that the SMBH is starved $\sim 50\%$ of the time after the onset of star formation in the galaxy. SMBH growth can become self-regulated by AGN feedback if the SMBH becomes massive enough, either by accretion or seeding, for its feedback to dominate the surrounding nuclear region. We find no evidence of galaxy-scale, AGN-driven quenching in the star formation rate (SFR) across all simulations in our suite.

Figures (7)

• Figure 1: This figure depicts our zoom-in simulation of a massive Cosmic Dawn galaxy at $z = 9$ with a SMBH at the centre regulated by AGN feedback. The top row shows consecutive zooms into a slice of the gas density depicting gas filaments which feed onto the host galaxy. In the bottom row we show projected quantities: the gas surface density, temperature (weighted by gas density), and the stellar surface density superimposed on the dark matter surface density. The cyan star is the SMBH at the centre of the galaxy. The galaxy resembles a clumpy thick disk of stellar clusters.
• Figure 2: Mass evolution of the SMBH as a function of time for all simulations in our suite (see Table \ref{['tab:sims']}). Solid coloured curves denote simulations with AGN feedback and dashed coloured curves denote simulations without AGN feedback. Thin-black dashed lines correspond to exponential (super-)Eddington growth with a slope determined by $\lambda_{\rm edd.}$. The black dotted curve denotes the critical self-regulated mass $M^{\rm esc}_{\rm sink,crit}$ required for AGN feedback to unbind gas in the sink sphere from the halo. For context, we include observations of Cosmic Dawn AGN with JWSTmaiolino_small_2024natarajan_first_2024napolitano_dual_2025taylor_capers-lrd-z9_2025naidu_black_2025. The zoom-in panels show snapshots of the projected density field within a 500 pc region around the sink sphere at early and late times for different simulations. The cyan rings denote the sink sphere with radius $r_{\rm sink} = 40 \; \rm pc$. The arced arrows denote the suppression in the growth slope caused by the turbulent-multiphase ISM environment. We provide a heuristic diagram to show how the environment impacts SMBH growth in the bottom right. Simulations with no AGN feedback in the fiducial ISM model show the same amount of starvation ($\alpha \approx 0.5$) while simulations with the low stellar feedback model indicate no starvation ($\alpha = 1$).
• Figure 3: In the top panel we show the critical mass threshold for heating from AGN feedback to dominate over cooling $M^{\rm cool}_{\rm sink,crit}$ relative to the sink mass as a function of time for all simulations with AGN feedback enabled. In the bottom panel we show the gas density within the sink sphere for each simulation. For clarity, all curves shown are running averages with a window size of 5 Myrs.
• Figure 4: Sink sphere properties as a function of time. Each column represents a different set of simulations from the suite outlined in Table \ref{['tab:sims']}. In all panels, solid lines denote simulations with AGN feedback and dashed lines denote simulations without AGN feedback. The top and middle rows show the density (blue) and sound speed (red) within the sink sphere. The bottom row shows the accretion rate relative to the Eddington rate. For clarity all curves shown are running averages with a smoothing window of 5 Myrs.
• Figure 5: The SFR as a function of time over the duration of our simulations. This quantity is computed from stars within a 2 kpc box centred on the SMBH at the centre of the galaxy. All curves correspond to a running average with a window of 10 Myr. The black dash-dot curve is the measured SFR from the base fiducial simulation of andalman_origin_2025 with no SMBH. The solid coloured lines correspond to simulations with AGN feedback and the dashed coloured lines correspond to simulations without AGN feedback. In total there are SFRs for 9 simulations as a function of time.