Little Red Dots on FIRE: The Ability of Bursty Galaxies to Host an Abundant Population of High-Redshift AGN

The James Webb Space Telescope has unveiled an abundant population of potential active galactic nuclei (AGN) at high redshift ($z\gtrsim4$) known as little red dots (LRDs), which are likely hosted in relatively low-mass galaxies. However, previous theoretical models have highlighted the difficulty in continuously feeding massive black holes in the central regions of bursty, high-redshift galaxies because of repeated gas evacuation by stellar feedback. We analyze galaxies in high-redshift FIRE-2 simulations to understand whether they are capable of hosting the observed abundant population of high-redshift AGN. We use a gravitational torque-driven accretion (GTDA) model and a simple free-fall accretion model to derive black hole accretion rates and construct predicted AGN bolometric luminosity functions for $z=5-7$. The GTDA model and the free-fall model with black holes accreting $\lesssim 1$ percent of their central gas supply ($<100 \rm \ pc$) per free-fall time predict AGN abundances that are more than sufficient to explain the most recent LRD observations. The fiducial models, in fact, overpredict the number of low-luminosity AGN as compared with observations. We explore possible resolutions of this tension. A plausible, though likely not unique, scenario for alleviating the AGN overpredictions and which also provides a good match to the host-galaxy UV luminosity distribution suggests that LRDs are super Eddington-accreting, Eddington luminosity-limited, $M_{\rm BH}\gtrsim 2\times10^5 \ \rm M_\odot$ black holes residing in $M_\star \gtrsim 2\times10^7 \ \rm M_\odot$ galaxies.

💡 Research Summary

The paper addresses a striking tension that has emerged in the wake of JWST observations: the discovery of a surprisingly abundant population of compact, red‑optical sources at $z\gtrsim4$, dubbed “little red dots” (LRDs), which appear to be hosted by low‑mass, bursty galaxies yet exhibit luminosities that imply active galactic nuclei (AGN) powered by supermassive black holes (SMBHs). Previous theoretical work, especially using the FIRE suite of cosmological zoom‑in simulations, has argued that repeated stellar feedback in such galaxies evacuates the central gas reservoir, making sustained black‑hole growth difficult. The authors ask whether the central gas supplies in FIRE‑2 galaxies are nevertheless capable of feeding SMBHs at rates sufficient to explain the observed LRD abundance.

To answer this, they post‑process a set of high‑redshift FIRE‑2 zoom‑in runs (z = 5–7, halo masses $10^{10}$–$10^{12},M_\odot$) that resolve the multiphase interstellar medium but do not include black‑hole physics or AGN feedback. In each simulated galaxy they place a hypothetical black hole at the location of maximum stellar density within one virial radius and compute instantaneous accretion rates using two distinct sub‑grid prescriptions:

1. Gravitational‑torque‑driven accretion (GTDA) – based on the Hopkins & Quataert (2011) model, where the inflow rate scales with the disk mass fraction, total baryonic mass inside a chosen radius (fiducially $R=100,$pc), and a torque efficiency factor $\epsilon_T$. The authors adopt a relatively high mean $\bar\epsilon_T=5$ (consistent with reproducing $z=0$ $M_{\rm BH}$–$M_\star$ relations) and explore log‑normal scatter.

2. Simple free‑fall accretion – where the inflow rate is $\dot M_{\rm ff}= \epsilon_{\rm ff},M_{\rm gas}(<R)/t_{\rm ff}$, with $t_{\rm ff}$ the free‑fall time of the total mass inside $R$. The efficiency $\epsilon_{\rm ff}$ represents the fraction of gas that reaches the black hole per free‑fall time; the fiducial mean is $\bar\epsilon_{\rm ff}=0.01$, with scatter also examined.

Both models are deliberately insensitive to the assumed black‑hole mass because the central region is dominated by gas and stars rather than the black hole itself. The authors also introduce stochastic variability by adding log‑normal scatter (σ=0.5 dex) to the normalization parameters, mimicking unresolved sub‑parsec fluctuations.

The resulting accretion histories are highly intermittent, reflecting the bursty stellar feedback that periodically clears the central 100 pc. Nevertheless, when the GTDA model or the free‑fall model with $\bar\epsilon_{\rm ff}\lesssim0.01$ is applied, the predicted AGN bolometric luminosity function (BLF) at $z\sim6$ matches the high‑luminosity end ($L_{\rm bol}\sim10^{45},$erg s$^{-1}$) of the LRD‑derived BLF (Greene et al. 2025). Both models, however, overpredict the number of low‑luminosity AGN ($L_{\rm bol}\lesssim10^{44},$erg s$^{-1}$) relative to the observed LRD counts and to pre‑JWST X‑ray/UV AGN surveys (Shen et al. 2020).

To reconcile this excess, the authors explore several plausible modifications:

• Super‑Eddington, Eddington‑limited growth: If the black holes are relatively massive ($M_{\rm BH}\gtrsim2\times10^{5},M_\odot$) and accrete at rates that would nominally exceed the Eddington limit, radiative output would be capped at $L_{\rm Edd}$, effectively reducing the observable luminosity while still allowing rapid mass growth. This scenario yields a BLF that aligns with both the bright LRD bin and the observed UV luminosity distribution of host galaxies.

• Higher initial black‑hole seed masses: Starting from seeds larger than the canonical $10^{2-3},M_\odot$ reduces the required accretion efficiency, making it easier to reach the observed luminosities even with modest gas inflow.

• AGN feedback suppression of low‑luminosity accretion: Although not modeled directly, the authors argue that including AGN‑driven outflows could preferentially quench low‑level accretion episodes, thereby lowering the predicted counts in the faint BLF bins.

Sensitivity tests varying the aperture radius (50–200 pc), the $M_{\rm BH}$–$M_\star$ scaling (e.g., $M_{\rm BH}=0.005,M_\star$), and the scatter in $\epsilon_T$ or $\epsilon_{\rm ff}$ confirm that the high‑luminosity predictions are robust, while the faint‑end excess is most sensitive to the assumed central gas supply and the lack of AGN feedback.

The paper concludes that bursty, high‑redshift FIRE‑2 galaxies are indeed capable of hosting the abundant LRD population, provided that black holes are either relatively massive from the outset or can accrete in a super‑Eddington regime limited by radiation pressure. The overprediction of faint AGN likely signals missing physics—most notably AGN feedback and perhaps a more realistic treatment of the multiphase inflow on sub‑100 pc scales. Future work with hyper‑refined simulations that resolve the black‑hole sphere of influence and incorporate self‑consistent AGN feedback will be essential to fully close the gap between theory and JWST observations.