Black Hole Stars Across the Universe: Identifying Central Engine Dominated Little Red Dots at $z\sim1.5-9.5$

Andrea Weibel, Rohan P. Naidu, Pascal A. Oesch, Anna de Graaff, Raphael E. Hviding, Zhaoran Liu, Jorryt Matthee, Christina C. Williams, Gabriel Brammer, Alba Covelo Paz, Jenny E. Greene, Christian Kragh Jespersen, Zhiyuan Ji, Michael V. Maseda, David J. Setton, Wendy Q. Sun, Alberto Torralba, Callum Witten, Mengyuan Xiao

First listed 2026-06-17 | Last updated 2026-06-15

Abstract

Photometric selections of Little Red Dots (LRDs) largely rely on identifying their ``V-shaped'' spectral energy distribution (SED). Recent work suggests this V-shape stems from a combination of a central engine -- also referred to as a Black Hole Star (BH*) -- and a star-forming host galaxy. We present a new and highly complementary photometric selection that is based on incorporating BH* templates in the \texttt{eazy} redshift fitting code. Selecting compact sources where a BH* template contributes $>80$\% to the best fitting SED in the rest-optical, we compile a sample of 241 BH*-dominated candidates from $\sim1000\,{\rm arcmin}^2$ of legacy and pure parallel JWST imaging. Our selection does not require a blue UV-component, and it successfully identifies objects that resemble the paradigmatic sources ``MoM-BH*-1'' and ``The Cliff''. We find that BH*-dominated sources exist across a wide range of redshifts ($z\sim1.7-9.3$) and optical luminosities (log$(L_{5100}/{\rm erg}\,{\rm s}^{-1})\sim42-44.5$), and we measure a median Balmer break strength of $\sim3$, with some breaks reaching values $>10$. We estimate bolometric luminosities in the range log$(L_{\rm bol}/{\rm erg}\,{\rm s}^{-1})\sim42-45$, which, assuming accretion at the Eddington-limit, would translate to black hole masses of $M_{\rm BH}\sim10^4-10^7{\rm M_\odot}$, spanning the intermediate mass black hole to the quasar regime. The number density of BH*-dominated candidates peaks at $z\sim5-6$ ($\sim10^{-5}\,{\rm Mpc}^{-3}$) and it declines by an order of magnitude down to $z\sim2$. Tentatively, comparing to V-shaped LRD samples suggests that the fraction of BH*-dominated sources among the broader LRD population does not decrease towards lower redshift. Crucially, our work demonstrates that BH*-dominated sources are not merely an early-Universe phenomenon but rather persist at least until cosmic noon.

Short digest

This paper introduces a complementary photometric selection for little red dots by adding Black Hole Star (BH*) templates to eazy and selecting compact sources whose rest-optical SEDs are dominated at the >80% level by the central engine. Applied to roughly 1000 arcmin^2 of JWST imaging, the method yields 241 BH*-dominated candidates spanning z~1.7-9.3 and log(L_5100/erg s^-1)~42-44.5, including UV-faint objects analogous to The Cliff and MoM-BH*-1 that would often be missed by standard V-shape selections. The sample shows strong Balmer breaks with a median strength of about 3 and some values above 10, while inferred log(L_bol/erg s^-1)~42-45 implies black hole masses of about 10^4-10^7 M_sun if accreting near Eddington. Their number density peaks at z~5-6 and falls by about an order of magnitude toward z~2, arguing that BH*-dominated LRD-like sources are not confined to the reionization era but persist to cosmic noon.

Key figures to inspect

• Figure 3. Use this as the main observational validation figure. The PRISM spectra show that the template-based selection recovers sources that genuinely resemble the benchmark BH*-dominated objects The Cliff and MoM-BH*-1 across a wide redshift range, including newly presented cases, which directly supports the claim that this method finds real BH*-dominated LRD analogs rather than only photometric curiosities.
• Figure 5. This is the clearest figure for explaining why the selection matters. The Euler-diagram overlap and the F150W SNR versus redshift plane demonstrate that the BH*-template method is complementary to standard color-based V-shape selections, especially by recovering lower-redshift and rest-UV-faint candidates that blue-UV-dependent cuts tend to miss.
• Figure 10. This figure best summarizes the physical parameter space of the sample. It shows the broad spread in redshift and optical luminosity together with the distribution of Balmer break strengths, making it central for the paper's claim that BH*-dominated candidates span z~1.5-9.5, reach some of the strongest Balmer breaks known, and are not restricted to one narrow luminosity regime.
• Figure 11. Include this for the paper's central-engine interpretation. By converting the template-derived bolometric luminosities into implied black hole masses under an Eddington assumption, it ties the photometric selection to the intermediate-mass-black-hole through quasar-like regime and shows where the sample becomes sparse at the highest and lowest luminosities.
• Figure 13. This is the conclusion-driving population figure. The redshift-binned number densities establish the peak around z~5-6 and the decline toward z~2, which is the key evidence behind the headline result that BH*-dominated sources persist well past the early Universe and remain present through cosmic noon.