🔭 astrophysics

How to raise a supermassive black hole: interpreting early JWST AGN with the AESOPICA simulations

The AESOPICA project demonstrates that the abundant, seemingly overmassive active galactic nuclei observed by JWST in the early Universe can be explained by efficient accretion across various black hole seed masses, with broad-line selection biases naturally reproducing the observed population while supporting a scenario where black holes assemble before their host stellar components are fully established.

Original authors: Sophie Koudmani, Jan Scholtz, Anthony J. Taylor, Ignas Juodžbalis, Debora Sijacki, Rachel S. Somerville, Roberto Maiolino, Emma Curtis-Lake, Steven L. Finkelstein, Martin A. Bourne, Francesco D'Eugeni

Published 2026-07-30

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Sophie Koudmani, Jan Scholtz, Anthony J. Taylor, Ignas Juodžbalis, Debora Sijacki, Rachel S. Somerville, Roberto Maiolino, Emma Curtis-Lake, Steven L. Finkelstein, Martin A. Bourne, Francesco D'Eugenio, Sophia Geris, Lucy R. Ivey, Hannah Übler

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, cosmic construction site. For decades, astronomers have believed that the massive black holes sitting at the centers of galaxies are like slow-growing giants. They thought these monsters started as tiny seeds—perhaps the remnants of the very first stars—and spent billions of years slowly eating gas, growing a little bit at a time, until they finally became the supermassive behemoths we see today. It was a story of patience, where the black hole and its host galaxy grew up together, hand-in-hand, following a strict set of rules about how big a black hole should be for a galaxy of a certain size.

But recently, a new telescope called the James Webb Space Telescope (JWST) started peering back in time to the very dawn of the universe. Instead of finding patient, slow-growing giants, it found something shocking: massive black holes that were already fully grown when the universe was just a toddler. These "early bloomers" are so big and so numerous that they break the old rules. They seem to have grown too fast, too early, and in galaxies that are too small to support them. This has left scientists scratching their heads: Did these black holes start as heavy seeds? Did they eat voraciously? Or are we just seeing the lucky few that got noticed?

This is where the AESOPICA project comes in. Think of this paper as a massive, digital time-travel experiment. The authors built a suite of twelve different computer simulations—essentially creating twelve parallel universes—to test every possible way these early black holes could have grown. They didn't just guess; they ran the numbers to see which "recipe" for black hole birth and feeding could actually produce the monsters JWST is seeing.

The Experiment: Twelve Parallel Universes

To solve the mystery, the team created a virtual universe in a box about 60 million light-years across. Inside this box, they started with the same initial conditions but changed three key ingredients to see what happened:

The Seed Size: They tried starting black holes as tiny as 100 times the mass of our Sun (light seeds, like a dead star) and as heavy as 100,000 times the Sun's mass (heavy seeds, like a collapsed gas cloud).
The Appetite: They tested how hungry the black holes were. Some models had them eating at a normal, steady pace (the "fiducial" model). Others let them gorge themselves, eating up to ten times faster than the theoretical limit (super-Eddington bursts).
The Neighborhood: They tweaked how much the surrounding stars pushed back against the black hole's growth (feedback), making it easier or harder for the black hole to grab its food.

They then ran these simulations forward in time and used a special tool called balmersopica to act like a virtual JWST. This tool asked: "If we looked at this simulated universe with our real telescope, which black holes would we actually see?" This is crucial because telescopes have blind spots; they usually only spot the brightest, most active objects, missing the quiet ones.

The Findings: How to Raise a Supermassive Black Hole

The results of these twelve universes revealed a few surprising truths about how to build a monster black hole in a hurry.

1. The "Appetite" Matters More Than the "Seed"
The biggest surprise is that it doesn't actually matter if you start with a tiny seed or a heavy one, as long as the black hole has a super-appetite. The simulations showed that if you let a tiny seed eat at a super-fast, super-Eddington rate, it can grow into a massive black hole just as quickly as a heavy seed that eats normally. However, if you start with a tiny seed and give it a normal, slow appetite, it simply cannot grow big enough in time to match what JWST sees. So, the paper suggests that if these early black holes started small, they must have been absolute gluttons, eating at breakneck speeds for short, intense bursts.

2. The "Tip of the Iceberg" Effect
When the team looked at the entire population of black holes in their simulations, they found a huge range of sizes. But when they applied the "JWST filter" (looking only at the bright, active ones), the picture changed. The telescope naturally picks out the most extreme, overgrown black holes. This explains why the ones we see look so "overmassive" compared to their host galaxies. We aren't seeing the average black hole; we are seeing the loudest, most active ones. The paper suggests that the "overmassive" population is likely just the observable tip of a much larger, more diverse iceberg.

3. The "Black Hole First" Scenario
The simulations also supported a specific story about how galaxies and black holes grow together. In the models that matched the observations best, the black hole grew to its full size before the galaxy finished building its stars. It's like a child growing up to be an adult before their parents finish building the house they live in. The black hole locked into the gravity of the galaxy early on, and then the stars caught up later. This explains why the black holes seem to follow a specific relationship with the galaxy's gravity (measured by how fast stars move) even when they seem too big for the amount of starlight we see.

4. The Metal Mystery
There is one catch. The paper found that while their models could explain the size and number of these black holes, they struggled to explain the "purity" of some of them. JWST has spotted some of these early black holes living in galaxies that are incredibly "pristine"—meaning they have almost no heavy elements (metals) in their gas. In the simulations, heavy seeds could create these conditions, but the most extreme, metal-free examples JWST found were still outside the reach of even the most optimistic models. This suggests that either our models of how gas behaves in these early galaxies are missing something (like complex, multi-phase gas clouds), or perhaps these black holes started in a way we haven't even simulated yet.

The Verdict

The AESOPICA simulations don't give us a single, definitive "winner" for how these black holes started. Instead, they show us that the universe is flexible. Whether black holes began as tiny seeds or heavy giants, the key to becoming a supermassive monster in the early universe was efficient, bursty eating.

The paper concludes that we likely need to look deeper. The current "tip of the iceberg" we see with JWST is biased toward the brightest objects. To truly solve the mystery, we need future, deeper surveys to find the fainter, smaller black holes that are hiding in the shadows. Only by seeing the whole population, not just the loud ones, can we finally figure out if these early giants were born heavy, or if they just learned to eat very, very fast.

Technical Summary: How to raise a supermassive black hole: interpreting early JWST AGN with the AESOPICA simulations

Problem Statement
The James Webb Space Telescope (JWST) has revealed a population of active galactic nuclei (AGN) at high redshifts ( $4 \lesssim z \lesssim 11$ ) that appear overmassive relative to local black hole–galaxy scaling relations. These findings challenge standard models of black hole formation and growth, which typically assume heavy seeds ( $\sim 10^5 M_\odot$ ) and Eddington-limited accretion. It remains unclear whether these objects trace an efficient early growth channel, represent the observable tail of a broader population biased by selection effects, or result from systematic uncertainties in mass estimation. Discriminating between these scenarios requires cosmological simulations that can systematically vary seed masses, accretion efficiencies, and feedback mechanisms while accounting for observational selection.

Methodology
The authors introduce the AESOPICA project, a suite of twelve mid-volume ( $L=60$ Mpc) cosmological simulations based on the FABLE galaxy formation model (itself derived from Illustris). The suite systematically explores the parameter space of black hole evolution by varying three key components:

Seed Mass: Four distinct initial masses are tested to represent theoretical formation channels: very light ( $10^2 M_\odot$ , Pop III remnants), light ( $10^3 M_\odot$ ), intermediate ( $10^4 M_\odot$ , runaway stellar collisions), and heavy ( $10^5 M_\odot$ , direct collapse).
Accretion Physics: Three accretion prescriptions are employed:
- Fiducial: Standard Bondi-Hoyle-Lyttleton accretion with a boost factor $\alpha=100$ and an Eddington limit ( $f_{\rm Edd, max}=1$ ).
- SE-Boost: Enhanced boost factor ( $\alpha=10^3$ ) allowing super-Eddington bursts up to $f_{\rm Edd, max}=10$ .
- SE-BoostMax: An "optimistic" scenario where the boost factor is inversely proportional to seed mass (up to $\alpha=10^5$ for $10^2 M_\odot$ seeds) to compensate for the $M_{\rm BH}^2$ dependency of Bondi accretion, and supernova feedback is reduced ( $\epsilon_{W,SN}=0.5$ ) for heavy seeds to facilitate early growth.
Observational Selection: The authors developed balmersopica, a mock JWST survey pipeline. This tool forward-models broad-line AGN detection by simulating NIRSpec spectra (R1000 and R2700 gratings) for various exposure times (1h to 100h), applying Bayesian fitting to determine detectability based on signal-to-noise ratios and line width constraints (FWHM $750-10,000$ km s $^{-1}$ ).

Key Contributions

Systematic Parameter Space Exploration: AESOPICA is one of the first hydrodynamical simulation suites to simultaneously vary seed mass, accretion efficiency (including super-Eddington bursts), and feedback strength across a cosmological volume.
Forward-Modeling Selection Effects: By applying balmersopica, the study moves beyond comparing raw simulation outputs to observations, instead generating synthetic catalogs that mimic specific JWST survey configurations (CEERS, RUBIES, JADES).
Decoupling Seeding from Growth: The simulations demonstrate that efficient accretion can rapidly erase the initial "imprint" of the seed mass, making the origin of high-redshift overmassive black holes degenerate with respect to seed mass unless specific conditions are met.

Results

Reproduction of JWST AGN: The bulk of the JWST-detected AGN population can be reproduced from any seed mass ( $10^2$ – $10^5 M_\odot$ ) provided the accretion efficiency is sufficiently high. However, very light seeds ( $10^2 M_\odot$ ) require the most favorable conditions (SE-BoostMax) to reach the observed mass scales.
Scaling Relations:
- $M_{\rm BH}$ – $M_{\rm stellar}$ : With fiducial accretion, only intermediate-to-heavy seeds reach the overmassive locus by $z \gtrsim 4$ . With SE-BoostMax, light seeds also populate this region, and seeding signatures are erased by $z \sim 6$ . At $z=0$ , all models converge to local relations for massive galaxies.
- $M_{\rm BH}$ – $\sigma_{\rm stellar}$ : Selected AGN lie on the local, weakly evolving $M_{\rm BH}$ – $\sigma_{\rm stellar}$ relation, supporting a "black hole first" scenario where black holes assemble before the stellar component is fully established.
Selection Bias: The apparent overmassiveness of JWST AGN is largely a consequence of observational selection. Broad-line selection naturally recovers the most massive and actively accreting black holes, which lie furthest above the intrinsic $M_{\rm BH}$ – $M_{\rm stellar}$ relation in inefficient accretion models.
Luminosity and Mass Functions: The inferred abundances of high-redshift AGN are consistent with intermediate and heavy seeds across all accretion models. Very light seeds are disfavored unless they undergo highly efficient, super-Eddington growth. The black hole mass function at the low-mass end ( $M_{\rm BH} < 10^7 M_\odot$ ) retains sensitivity to seed mass, whereas the high-mass end converges.
Metallicity Constraints: Gas-phase metallicity offers a potential discriminator. Only heavy seed models produce systems with $M_{\rm BH}/M_{\rm stellar} > 0.1$ in hosts with extremely low metallicity ( $Z < 0.01 Z_\odot$ ). However, even the most optimistic heavy seed models fail to reproduce the most extreme "pristine" AGN (e.g., QSO1), suggesting limitations in current ISM modeling or the need for alternative origins (e.g., primordial black holes).
Quenching and Dormant AGN: Strongly quenched massive galaxies at high redshift only appear in simulations allowing super-Eddington bursts. Dormant, overmassive black holes (e.g., GN-1001830) are naturally explained as systems observed during the declining phase of a rapid accretion burst, followed by strong feedback that shuts down both star formation and accretion.

Significance and Claims
The paper claims that the "overmassive" nature of early JWST AGN does not necessarily require a specific seeding mechanism (e.g., exclusively heavy seeds) if efficient accretion channels are permitted. Instead, the observed population is likely the "tip of the iceberg" of a broader distribution, heavily biased by selection effects that favor bright, broad-line sources.

The authors emphasize that while current observations can be matched by various models, future constraints will rely on:

Low-mass end statistics: Deep surveys probing the low-mass end of the black hole mass function, where seed mass signatures persist.
Metallicity measurements: High-resolution spectroscopy to test the $M_{\rm BH}$ –metallicity relation, which currently challenges even the most optimistic heavy seed models for the most extreme objects.
Improved Physics: The need for next-generation simulations that resolve the multi-phase ISM and accretion disc physics to replace phenomenological boost factors with physically grounded prescriptions.
Gravitational Waves: Future LISA observations of mergers in the $10^4$ – $10^7 M_\odot$ range will provide independent constraints on early assembly histories, free from electromagnetic selection biases.

The study concludes that disentangling the origin of the first supermassive black holes requires a combination of deeper observational censuses, more sophisticated subgrid physics, and multi-messenger constraints.

The Experiment: Twelve Parallel Universes

The Findings: How to Raise a Supermassive Black Hole

The Verdict

Technical Summary: How to raise a supermassive black hole: interpreting early JWST AGN with the AESOPICA simulations

More like this