Overmassive No More: The Case for Little Red Dots Hosting Black Hole Seeds as Massive as Single Supermassive Stars

Wendy Q. Sun, Rohan P. Naidu, Hanpu Liu, Anna de Graaff, Jenny E. Greene, Jorryt Matthee, Chris Ashall, John Chisholm, Anna-Christina Eilers, Qinyue Fei, Kasper E. Heintz, Daichi Hiramatsu, Vasily Kokorev, Joel Leja, Zhaoran Liu, Priyamvada Natarajan, Pascal A. Oesch, Robert A. Simcoe, Alberto Torralba, Andrea Weibel

First listed 2026-09-10 | Last updated 2026-09-08

Abstract

Little Red Dots (LRDs) display singular properties unlike any known class of AGN or galaxies, motivating novel mass estimators for their central engines. Inspired by their similarities to stellar phenomena, here we interpret the LRD continuum as being produced by a pseudo-photosphere. We fit tailored stellar atmosphere models to host-subtracted LRD central engines ("black hole stars," BH*s) represented by stacks of $117$ objects. Typical BH* continuum spectra are well fit by models in a narrow range of temperatures ($T_{\rm eff}\approx4200-4800$ K), with bolometric luminosities $\approx10^{43-45}$ erg s$^{-1}$, implying pseudo-photospheric radii $\approx700-2000$ au. Based on these parameters, we explore four different approaches to deriving BH* masses: 1) using the surface gravity from atmosphere models; 2) appealing to the resemblance to super-Eddington phenomena; 3) approximating the escape velocity from the outflowing material; and 4) exploiting the lack of variability to bound the dynamical time. For the typical BH*, all of these methods yield remarkably consistent masses of $\approx10^{4-5}\,M_\odot$, implying a highly super-Eddington luminosity of $L_{\rm{bol}}/L_{\rm{Edd}}\sim5-50$. These mass estimates place BH*s within the scatter of the local scaling relation between black hole mass and host galaxy stellar mass, providing a self-consistent alternative to "overmassive" black holes that lie $2-3$ dex above it. Crucially, our derived masses are consistent with BH*s arising from single supermassive stars (SMSs), whose masses cannot exceed $\approx10^{5-6}\,M_\odot$ due to general relativistic instabilities. Furthermore, for our derived $L_{\rm bol}/L_{\rm Edd}$, the sharp cutoff of the LRD luminosity function matches the maximum theoretical mass of an SMS. With LRDs, we may therefore be directly observing the birth of heavy black hole seeds.

Short digest

Sun et al. recast the host-subtracted continua of 117 Little Red Dots as cool, optically thick pseudo-photospheres surrounding their central engines, fitting tailored stellar-atmosphere models to stacked spectra. The inferred BH* temperatures are tightly clustered at about 4200–4800 K, with luminosities of roughly 10^43–10^45 erg s^-1 and photospheric radii of about 700–2000 au. Four non-virial mass estimates, based on fitted surface gravity, super-Eddington analogies, outflow escape speeds, and the observed lack of variability, converge on central masses of about 10^4–10^5 solar masses and Eddington ratios of about 5–50. This removes the need for 2–3 dex overmassive black holes, places the sources near the local black-hole–host relation, and links both their masses and luminosity-function cutoff to the maximum mass of single supermassive stars, making LRDs plausible newly born heavy seeds.

Key figures to inspect

• Figure 2. This is the essential selection-and-decomposition figure: matching hosts by redshift and [O III] luminosity isolates the BH* component and reveals the red optical continuum and Balmer break that motivate the pseudo-photosphere interpretation.
• Figure 5. The HR-diagram comparison places BH*s in the unusually cold, luminous regime occupied by dense-wind pseudo-photospheres, visually connecting their narrow 4200–4800 K temperature range to hydrogen-recombination regulation.
• Figure 6. This figure makes the paper's central revision immediately clear by comparing the median non-virial BH* mass with local virial-calibration estimates at fixed host stellar mass, showing why the apparent overmassive-black-hole problem disappears.
• Figure 7. This three-substack synthesis is the conclusion-driving figure: all four mass-estimation approaches yield masses below the general-relativistic upper limit for single supermassive stars, directly supporting the heavy-seed birth scenario.