Heavy seeds and the first black holes: Insights from the BRAHMA simulations

TL;DR

This study uses BRAHMA-CONSTRAINED simulations to explore how heavy seed formation, BH dynamics, accretion, and stellar/AGN feedback drive the assembly of the earliest black holes in extreme overdense environments. It finds a transition from merger-dominated growth at $z\gtrsim9$ to accretion-dominated growth at $z\lesssim9$, with seed properties and dynamical friction shaping high-redshift growth and accretion physics controlling later evolution. Reproducing the most massive $z\sim6$ quasars and $z\sim9$–11 JWST BHs remains challenging, often requiring very massive seeds, super-Eddington accretion, or significantly weaker feedback than in IllustrisTNG. The results highlight the critical interplay between seeding, dynamics, and feedback and provide a framework to interpret current and future high-redshift BH observations.

Abstract

From the luminous quasars at $z \sim 6$ to the recent $z \sim 9-11$ AGNs revealed by JWST, observations of the earliest black hole (BH) populations can provide unique constraints on BH formation and growth models. We use the BRAHMA simulations with constrained initial conditions to investigate BH assembly in extreme overdense regions. The simulations implement heavy seeds ($\sim 10^4-10^5 M_{\odot})$ forming in dense, metal-poor gas exposed to sufficient Lyman-Werner flux. With gas accretion modeled via Bondi-Hoyle formalism and BH dynamics and mergers using a subgrid dynamical friction scheme, we isolate the impact of seeding, dynamics, accretion, and feedback on early BH growth. With fiducial stellar and AGN feedback inherited from IllustrisTNG, accretion is strongly suppressed at $z \gtrsim 9$, leaving mergers as the dominant growth channel. Gas accretion dominates at $z \lesssim 9$, where permissive models (super-Eddington or low radiative efficiency) build $\sim 10^9\ M_{\odot}$ BHs powering quasars by $z \sim 6$, while stricter IllustrisTNG-based prescriptions yield much lower BH masses ($\sim 10^6-10^8\ M_{\odot}$). Our seed models strongly affect merger-driven growth at $z \gtrsim 9$: only the most lenient models (with $\sim 10^5\ M_{\odot}$ seeds) produce enough BH mergers to reach $\gtrsim 10^6\ M_{\odot}$ by $z \sim 10$, consistent with current estimates for GN-z11. Our dynamical friction model gives low merger efficiencies, hindering the buildup of $\gtrsim 10^7\ M_{\odot}$ BHs by $z \sim 9-10$, as currently inferred for GHZ9, UHZ1, and CAPERS-LRD-z9. If the BH-to-stellar mass ratios of these sources are indeed as extreme as currently inferred, they would require either very short BH merger timescales or reduced AGN thermal feedback. Weaker stellar feedback boosts both star formation and BH accretion and cannot raise these ratios.

Figures (9)

• Figure 1: Visualization of our two different types of constrained simulation boxes at $z=6$, showing the 2D gas density profile which gradually transitions into the 2D gas metallicity profile from left to right. At the center of each box, there is a $5\sigma$ overdensity peak (when smoothed over $1~\rm Mpc$ scales). In the left panel (5SIGMA_COMPACT), the overdensity peak has significantly higher compactness and low tidal field strength compared to a typical $5\sigma$ peak (5SIGMA_TYPICAL) that is shown in the right panel. Both simulations produce a $3\times10^{12}~M_{\odot}$ halo at $z=6$. Unless explicitly stated, most of the figures hereafter show results for 5SIGMA_COMPACT simulations.
• Figure 2: Evolution of the total mass (upper panel) and the stellar mass (lower panel) of the most massive halo across different redshift snapshots. The halo reaches a mass of $\sim3\times10^{12}~M_{\odot}$ at $z=6$, consistent with the observed $z\sim6$ quasars based on quasar-galaxy cross correlation measurements of 2024ApJ...974..275E shown as black circles. The stellar masses are consistent with the current estimates of the JWST AGN hosts, which makes this region a promising site for studying the assembly of $z\sim9-11$ BHs.
• Figure 3: Predicted number density of new BH seeds per unit redshift ($\mathrm{d}n/\mathrm{d}z$) for the various seed models explored in this work. All solid lines correspond to a fiducial seed mass of $M_{\mathrm{seed}} = 1.5 \times 10^5~M_{\odot}$, which produce peak seeding abundances ranging from $\sim10^{-3}-0.3~\rm Mpc^{-3}$ per unit redshift depending on the critical LW flux. The dashed lines correspond to the two higher-resolution supplementary runs that assume $M_{\mathrm{seed}} = 1.8 \times 10^4~M_{\odot}$, predicting peak seed abundances of $\sim0.8~\&~5~\mathrm{Mpc}^{-3}$.
• Figure 4: BH mass assembly for different seeding and accretion models: Evolution of the most massive BH in the most massive halo across redshift snapshots. The top row shows the BH mass ($M_{bh}$), where the solid line denotes the total BH mass growth (from both accretion and mergers), while dashed lines indicate the contribution from mergers alone. The middle and bottom rows display the stellar mass ($M_{*}$) and the $M_{bh}/M_*$ ratios, respectively. The blue, orange, and green curves in each panel represent the three seed models with our default seed mass of $1.5\times10^5~M_{\odot}$, corresponding to $J_{\rm crit}=10,\,100,\,\&\,300~J_{21}$. The four columns correspond to the four different accretion models. In the last column, we show an additional simulation with $J_{\rm crit}=10$ that uses BH repositioning instead of subgrid-DF (dark blue line). Faded grey circles mark observed $z\gtrsim6$ quasars, while dark pentagons highlight the subset with host galaxy mass measurements. The horizontal line indicates $10^9~M_{\odot}$, which we adopt as the minimum threshold for claiming that a simulation has successfully produced a $z\sim6$ quasar. The larger circle, square, triangles, and star correspond to the $z\sim9$–11 JWST AGN (see legend). BH growth is dominated by mergers at $z\gtrsim9$ and by gas accretion at $z\lesssim9$. As a result, a lenient accretion model significantly enhances the BH mass assembly at $z\lesssim9$, but has a negligible impact on the BH growth at $z\gtrsim9$. On the other hand, changing the seed model has the strongest impact at $z\gtrsim9$ as it determines the number of seeds that are available to fuel BH-BH mergers.
• Figure 5: Similar to the previous figure, but here we compare the main simulations with the default seed mass ($1.5\times10^{5}~M_{\odot}$), against the supplementary simulations with the lower seed mass ($1.8\times10^4~M_{\odot}$). All the lines assume the most lenient seeding ($J_{\rm crit}=10~J_{21}$) and accretion (TNG-BOOST-SE) model. Despite the lower seed masses forming at $4~\&~10$ times higher abundances (revisit Figure \ref{['seed_formation']}), they cannot merge efficiently due to weaker dynamical friction. As a result, they produce $\sim6$ times smaller BH masses at $z\sim10$.