Summary

A new set of cosmological radiation-hydrodynamic simulations shows that heavy black hole seeds of about a million solar masses form naturally in dense protocluster regions, undergo a super-Eddington accretion phase that reproduces the spectral features of little red dots, and then grow into the overmassive black holes observed by the James Webb Space Telescope.

The origin of the supermassive black holes that power quasars less than a billion years after the Big Bang has long been a puzzle. The James Webb Space Telescope has added to the mystery by discovering an unexpectedly large number of overmassive black holes at redshifts greater than 4–6, as well as compact red sources dubbed little red dots, whose inferred black hole masses exceed local scaling relations. A new set of fully cosmological radiation-hydrodynamic simulations now offers a unified explanation: heavy black hole seeds of about a million solar masses form naturally in dense protocluster regions, undergo a brief phase of super-Eddington accretion that reproduces the spectral features of little red dots, and then grow into the overmassive black holes observed by Webb.

Contents

What the simulations show

The simulations follow the birth and early growth of supermassive black holes in a protocluster environment at a time when the Universe was only a few hundred million years old. Using the moving-mesh code AREPO with a self-consistent treatment of radiative feedback, the team tracked the collapse of gas in a halo that is exposed to intense far-ultraviolet radiation from a neighbouring star-forming galaxy. The calculation resolves, for the first time, the formation of black hole seeds with masses around 10^6 M☉—an order of magnitude more massive than the canonical ~10^5 M☉ seeds predicted by standard direct-collapse models. These heavy seeds then experience a short-lived but intense phase of super-Eddington accretion, during which the accretion rate reaches several to tens of times the Eddington limit. By redshift 8, the simulated black holes have grown to about 3 × 10^7 M☉, placing them well above the local black hole–stellar mass relation and matching the properties of observed overmassive quasars. The simulation also yields a comoving number density of at least 10^−4 Mpc^−3 for galaxies hosting such massive black holes, consistent with the observed abundance of little red dots.

In parallel, the researchers followed a light seed—a black hole of about 800 M☉ left over from a Population III star—that formed earlier in a different halo. That black hole barely grows, ending up orders of magnitude less massive than the heavy seeds, highlighting the necessity of the heavy-seed pathway to explain the overmassive population.

How heavy seeds form in protocluster environments

The key to forming a massive seed is the presence of a large, long-lived gas reservoir that can feed rapid star formation. In the simulation, the target halo sits about 10 kpc from a luminous neighbour that bathes it in a strong far-ultraviolet flux. This radiation suppresses molecular hydrogen cooling, delaying star formation until the halo reaches a virial temperature of roughly 4 × 10^4 K—much hotter than the typical 8,000 K threshold for direct-collapse black hole formation. The deep potential well of the massive halo keeps ionized gas bound, allowing a gas reservoir of several million solar masses to accumulate before collapse begins at redshift about 14.

Once collapse sets in, the central gas fragments hierarchically into a dense multiple system of protostars. Each protostar accretes mass at a rate of 0.1–1 M☉ yr^−1, and the most massive member reaches about 4 × 10^5 M☉ within 2 Myr. The high accretion rates keep the stars in an inflated supergiant phase, which weakens radiative feedback and prevents photoevaporation of the surrounding gas. The stars ultimately collapse into black holes roughly 2 Myr after their formation, with dense gas still tightly bound around them. The simulation identifies several such seed-forming sites within the same overdense region, suggesting that heavy-seed formation is a natural outcome of early structure formation in protoclusters.

From seeds to overmassive black holes

Immediately after the massive stars collapse, a dense, optically thick envelope assembles around the nascent black hole. Radiation becomes trapped: photons are advected inward faster than they can diffuse outward, driving accretion rates several to a few tens of times above the Eddington limit for less than a million years. This super-Eddington phase allows the black holes to grow rapidly to several million solar masses. As the gas supply begins to dwindle, the accretion rate falls to 0.1–1 times the Eddington rate, and by redshift 10 the black holes reach roughly 10^7 M☉.

At this point, the accretion rate drops sharply to less than 1% of the Eddington limit. The black holes pass through the central region of the massive neighbour halo, where their envelope gas is stripped by the surrounding hot medium. Dynamical friction then carries them into the galaxy centre, where they merge with the host system by redshift 8. At that epoch, the stellar mass of the galaxy is about 10^9 M☉, giving a black hole–to–stellar mass ratio of approximately 1%—an order of magnitude above the local relation. After settling, the black holes resume sub-Eddington accretion, maintaining slow but steady growth. For a brief period the accretion rate reaches the Eddington limit, producing luminosities and black hole masses consistent with those of observed high-redshift active galactic nuclei.

Reproducing the spectra of little red dots

The dense gas disks that surround the growing heavy seeds naturally produce the distinctive spectral signatures of little red dots. In the simulation, gas densities exceed 10^8 cm^−3 in the circum-black hole disk, which enhances the population of hydrogen atoms in the n=2 energy level and produces strong Balmer absorption lines. The large electron column densities also give Thomson optical depths sufficient to broaden the Hα emission line to widths greater than 1,000 km s^−1, matching the observed line widths of LRDs. The Hα luminosity in the early, heavily obscured phase is about 1.5 × 10^43 erg s^−1, comparable to values inferred for LRDs.

The system does not remain in this obscured state for long. After a few hundred thousand years, the Hα luminosity declines to about 5 × 10^39 erg s^−1, but the accretion rate stays near the Eddington limit, producing an inner disk luminosity of roughly 10^44 erg s^−1. About 0.1–1% of this emission is expected to emerge as Hα after Thomson scattering in the outer disk, yielding Hα luminosities of 10^41–10^42 erg s^−1, typical of active galactic nuclei. Thus the simulation naturally transitions from an LRD-like, heavily obscured phase to a more AGN-like, less obscured state on a timescale of 0.1–1 Myr.

The dense gas also explains why LRDs are often X-ray weak. In the early super-Eddington phase, the hydrogen column density exceeds 10^26 cm^−2 in all directions, strongly Compton-scattering any X-rays emitted by the central black hole. Even in the later, nearly Eddington phase, the column density remains around 10^25–10^26 cm^−2, making the system difficult to detect in X-rays, consistent with current observational constraints.

Why this pathway matters

This work provides the first self-consistent cosmological demonstration that little red dots and overmassive black holes are linked phases of a single evolutionary pathway. By following the birth, growth, and observable signatures of black hole seeds in a realistic protocluster environment, the simulation bridges the gap between the elusive seeds and the luminous quasars already observed at redshifts greater than 7. The predicted abundance of heavy seeds can account for the luminous quasar population seen by JWST, and the mergers of these black holes during subsequent galaxy assembly are expected to generate gravitational-wave signals detectable by the Laser Interferometer Space Antenna (LISA). The characteristic strains of 10^−17–10^−16 at millihertz frequencies imply LISA event rates of tens to hundreds over a three-year mission, offering an independent test of the heavy-seed scenario.

What the simulations do not capture

While the results are compelling, the simulation makes several simplifying choices that could affect the quantitative predictions. Most importantly, it does not include kinetic feedback from accreting black holes such as jets or disk winds. If such feedback is efficient, it could unbind the surrounding gas and reduce the accretion rate, potentially limiting the growth to below the Eddington limit. The efficiency of jet launching depends on black hole spin and magnetic field strength, both of which are highly uncertain and require resolving the stellar rotation at seed birth and the detailed magnetic dynamo processes during accretion.

The treatment of stellar evolution and the criteria for stellar collapse also rely on one-dimensional models with inherent uncertainties, particularly for supermassive stars with masses above 10^5 M☉. The simulation further assumes a simplified far-ultraviolet background spectrum and neglects the possible contribution of ionizing photons from neighbouring galaxies, which could alter the exact conditions for heavy-seed formation. Finally, the calculation covers only a small region around a single massive halo, so the derived number density of heavy-seed hosts is a lower limit; the true occurrence rate may be higher, but its exact value remains model-dependent.

Despite these limitations, the study establishes that heavy black hole seeds can form naturally in the early Universe and grow into the overmassive black holes observed by JWST, offering a promising direction for future simulations and observational tests.

Sources