Origins of Massive Black-Hole Seeds

JWST and X-ray observations now reveal a diverse population of accreting black holes at high redshift: luminous and apparently overmassive systems, disputed individual candidates near $z\simeq10$, and stacked lower-mass accretors at $3\lt z\lt7$. These discoveries sharpen the initial-seed problem but do not observe a seed at formation. At the evidence cutoff of 2026-08-20, light stellar remnants, direct-collapse or other heavy seeds, runaway clusters, and mixed channels remain viable once uncertain accretion, duty cycle, radiative efficiency, mergers, host scaling, and selection are propagated. The origin question is open and the channel evidence contested.

Helpful background. Massive Black-Hole Seeds develops the channel physics, characteristic formation masses and redshifts, growth equations, and proposed discriminants. This record asks what present observations identify after growth rather than treating an endpoint mass as an initial condition.

Separating a seed from its descendant

A seed mass $M_{\rm seed}$ at formation becomes an observed mass only after a growth history. Even a schematic law,

$$M(t)=M_{\rm seed}\exp\!\left[\frac{1-\epsilon}{\epsilon}\,f_{\rm duty}\,\lambda\,\frac{t}{t_{\rm Edd}}\right],$$

shows the degeneracy: radiative efficiency $\epsilon$, duty cycle $f_{\rm duty}$, Eddington ratio $\lambda$, formation time, mergers, and feedback can trade against the starting mass. Super-Eddington episodes and obscured phases widen the mapping further.

Five axes delimit an inference: seed mass at formation versus observed mass; physical channel and environment; accretion, feedback, and merger history; individual virial/X-ray/SED measurement versus population occupation or mass function; and host mass, metallicity, obscuration, lensing, and completeness. A large black-hole-to-stellar-mass ratio is evidence about a joint growth history, not a direct scale reading of the seed.

Assessment at 2026-08-20	Meaning
Question status	Open
Evidence status	Contested
Confidence	Moderate for the conclusion that several channels remain; lower for any detailed channel weight
Domain	High-redshift JWST and X-ray accreting-black-hole measurements plus low-mass occupation and mass-function implications under explicit growth and selection models
What could change the conclusion	A high-redshift population constraint or unambiguous low-mass occupation measurement

Entry project: quantify the seed–growth degeneracy

Choose this direction if you enjoy connecting simple evolution equations with uncertain observations. The first project needs cosmological time–redshift conversion, accretion energetics, and basic probability. It can be done on a laptop without hydrodynamical simulations. Spectral re-reduction, survey completeness, and population inference are separate projects that benefit from an observing or galaxy-formation group.

Read Maiolino et al. 2024, Fig. 3 and Methods, black-hole growth models, extracting the GN-z11 mass estimate, observed redshift, seed formation assumptions, and accretion prescriptions. Reconstruct one stated growth track from the mass-growth equation, retaining the difference between supplied rest mass, radiated energy, and retained black-hole mass. Do not assume a present accretion-rate estimate held throughout the object’s history. Then read Volonteri, Habouzit, and Colpi 2021, observational constraints and future directions and list observables that constrain something beyond the final mass.

Deliver a reproducible family of tracks reaching the same endpoint while varying seed mass, duty cycle, and radiative efficiency over explicitly chosen ranges. Show the input assumptions beside the plot and propagate the reported endpoint uncertainty. This training calculation should exhibit degeneracy; fitting several tracks to one mass measurement is not evidence that their seed populations are equally probable.

A candidate extension, with novelty unchecked, is to ask whether one additional observable, such as a host-mass constraint or a low-mass active fraction, breaks that degeneracy in a simple forward model. Define a selection probability before making the comparison. Stop if the inference is set mainly by the duty-cycle prior, if the spectral source classification is disputed, or if the proposed discriminator is not observable in the selected sample. Pivot to a sensitivity analysis instead of assigning a unique seed channel. Population synthesis and new data, rather than a more elaborate endpoint fit, are the next resource commitment.

What the current high-redshift objects establish

GN-z11 hosts a rapidly accreting black hole at $z=10.6$, demonstrating that substantial growth occurred very early Maiolino et al. 2024, Fig. 3 and Methods, black-hole growth models. Its virial mass, bolometric correction, and host estimate have uncertainties, but the source is a real constraint on formation time plus growth efficiency. Crucially, the paper’s own tracks show that a light seed with rapid or intermittent super-Eddington growth and a heavier starting seed can reach the endpoint. The observation rules out some slow-growth histories, not one entire seed family.

UHZ1 was reported as an X-ray-selected candidate near $z\simeq10$. Its inferred high black-hole-to-host ratio favors a massive-seed track in the original authors’ growth model Bogdán et al. 2024, Figs. 1–4 and heavy-seed discussion. That source classification is now disputed. A 2026 full-exposure Chandra reanalysis finds that the hard excess reaches only $2.3$–$2.9\sigma$ across plausible reductions, does not grow with the added exposure, and is accompanied by MIRI limits inconsistent with the originally inferred luminous buried AGN Zou et al. 2026, abstract, §§ II.1–II.2, Fig. 3, and conclusion. Deep MIRI/LRS spectroscopy independently finds a metal-poor star-forming spectrum with no rest-frame UV/optical AGN signature Álvarez-Márquez et al. 2026, abstract and 13.9-hour MIRI/LRS results. Both adverse records were preprints at the cutoff. UHZ1 is therefore a useful candidate-level stress test, not secure evidence for a heavy-seed class or a direct collapse-mass measurement.

The JADES stacked analysis changes the population picture by recovering broad H$\alpha$ evidence for lower-mass accretors at $3\lt z\lt7$, below the luminous “little red dot” samples Geris et al. 2026, §§ 2–9. It broadens the observed mass function and tests stellar or supernova alternatives. Stacking sacrifices object-level occupation information and retains virial and selection systematics.

Results and open questions at 2026-08-20

Claim or evidence	What it supports	What it does not establish
GN-z11 at $z=10.6$	Very early accretion and stringent seed-plus-growth timing	Unique light or heavy seed mass
UHZ1 claim and 2026 adverse reanalyses	A sharply testable high-redshift candidate and reduction-dependent X-ray tension	A robust luminous-AGN detection, natal mass, or heavy-seed class
JADES stacked low-mass accretors	A broader high-redshift active mass function	Object-by-object occupation or natal masses
Seed-formation reviews	Physically motivated light, heavy, and cluster channels with discriminants	Their realized cosmological branching fractions
Multiple growth tracks through one endpoint	Direct evidence of seed–accretion degeneracy	Equivalence of every channel under every prior
No unified public hierarchical package	Need for explicit selection and growth machinery	Adverse evidence against the observed objects

Population evidence receives more weight than any one spectacular endpoint, but only when selection and contamination are propagated. An individual object can exclude a narrow history; a channel family needs a demographic discriminant.

Strongest competing interpretations

The heavy-seed interpretation notes that forming a $10^4$–$10^6M_\odot$ object directly or through rapid cluster collapse reduces the number of required e-folds. Overmassive black holes in young hosts and a confirmed UHZ1-like ratio would arise naturally in some such tracks. This view becomes strongest when sustained super-Eddington accretion is disfavored and high-redshift abundance is large.

The light-seed interpretation emphasizes that stellar remnants may be common and can grow rapidly in dense, gas-rich environments. Intermittent or super-Eddington episodes erase much of the starting-mass difference, while selection preferentially finds actively accreting outliers. Inayoshi, Visbal, and Haiman show how formation channels and early growth produce overlapping observables Inayoshi, Visbal, and Haiman 2020, §§ 2–5 and Figs. 1–8.

A mixed-channel interpretation is increasingly natural: different environments may produce different seeds, and observed samples weight duty cycle as much as occupation. Volonteri, Habouzit, and Colpi argue that joint demographics, host relations, occupation, and future mergers are needed precisely because one luminous quasar cannot select among these histories Volonteri, Habouzit, and Colpi 2021, Figs. 1–4 and observational/future sections.

Counterexamples, null results, and failed routes

GN-z11 is the strongest counterexample to a unique inference: qualitatively different seed-plus-accretion tracks cross the same measured endpoint. This is not a weakness of the observation; it is a direct demonstration that the inverse problem is many-to-one. UHZ1 is a stronger candidate-versus-class warning: the original $4.2$–$4.4\sigma$ X-ray claim is not robustly reproduced in the full Chandra analysis, and current MIRI imaging and spectroscopy do not support the required luminous obscured nucleus Zou et al. 2026, abstract and conclusion. A preprint dispute does not by itself prove that no weak or variable nucleus exists, but it removes UHZ1 from any secure class-level heavy-seed count.

The JADES stack counters a census based only on luminous red sources by revealing lower-mass activity. Yet stacking cannot determine which individual galaxies host black holes, so it cannot by itself measure occupation. The failed route is to convert a virial mass or black-hole/host ratio into $M_{\rm seed}$ without a population growth model. Another failed route is to compare observed counts with a theoretical intrinsic abundance while omitting obscuration, line-detection completeness, and active duty cycle.

Which observations retain memory of formation

Seed information survives best where later growth has had less time to erase it: the low-mass end, very high redshift, occupation in shallow potential wells, and gravitational-wave merger demographics. Even there, feedback and selection are coupled. A low occupation fraction in dwarf galaxies could favor rare heavy seeds, but stellar dynamics and accretion searches have different completeness. A high occupation fraction could favor abundant light seeds, unless recoil or feedback removed them.

The ideal analysis therefore treats seed mass function, occupation, duty cycle, Eddington-ratio distribution, radiative efficiency, mergers, and host scaling jointly. It should compare several forward growth models rather than reconstruct one “best” seed mass from each object. The scientific output is a posterior over histories and channel fractions, with prior sensitivity visible.

What would change the assessment

A decisive result would measure a selection-corrected feature that an entire channel family cannot reproduce across independently calibrated growth priors. Examples include an unambiguous low-mass occupation pattern at high redshift, a robust abundance of massive black holes too early for even permissive light-seed growth, or a joint electromagnetic and merger distribution incompatible with heavy-seed rarity.

The strongest programme would preregister a hierarchical analysis combining contiguous JWST spectroscopy, X-ray and radio limits, and local dwarf occupation. It would release completeness injections and infer seed mass function together with duty cycle and accretion. A new luminous object alone would sharpen timing but would rarely resolve origin.

Research checklist

• Pin survey data release, calibration pipeline, target/exclusion table, redshift flags, lens model, and imaging products.
• Refit spectra or stacks with declared line-spread function, continuum, broad-line, stellar, supernova, and obscuration models.
• Propagate virial-factor, bolometric, lensing, host-mass, and completeness uncertainties rather than selecting one value.
• Distinguish intrinsic occupation, active fraction, line-detection probability, and sample selection.
• Forward-model several seed mass functions with duty cycle, Eddington ratios, efficiency, mergers, feedback, and recoil.
• Test recovery with injected populations and report prior sensitivity and posterior predictive checks.
• Combine individual-object and stacked evidence without treating their shared survey calibration as independent.
• No end-to-end population reproduction was performed for this assessment; spectral reruns and seed inference are separate prescribed stages.

Source and data map

Inayoshi et al. and Volonteri et al. provide channel and discriminant synthesis. Maiolino et al. provide the GN-z11 endpoint constraint. Bogdán et al. supply the original UHZ1 claim; Zou et al. and Álvarez-Márquez et al. supply current adverse Chandra/MIRI tests. Geris et al. provide the current low-mass stacked population evidence and its alternative-source tests.

A bounded reproduction can retrieve named JWST/JADES spectra and images, rerun line or SED fits, and reconstruct the declared stack selection. It must pin weights, exclusions, calibration context, lensing, and completeness. The cutoff search found no immutable package joining those steps to a seed-growth hierarchical likelihood. Reproducing spectra therefore does not reproduce channel identification; a public population/growth simulator and selection likelihood remain required.

References

• Álvarez-Márquez, J., et al. “PRISMS. UNCOVER-26185, a metal-poor SFG at $z=10.05$ with no evidence for a X-ray-luminous AGN.” arXiv:2602.02323v1 (2026). Open preprint.
• Bogdán, Á., et al. “Evidence for heavy-seed origin of early supermassive black holes from a $z\simeq10$ X-ray quasar.” Nature Astronomy 8, 126–133 (2024). DOI.
• Geris, S., et al. “JADES reveals a large population of low-mass black holes at high redshift.” Monthly Notices of the Royal Astronomical Society 545 (2026), staf1979. DOI.
• Inayoshi, K., Visbal, E., and Haiman, Z. “The assembly of the first massive black holes.” Annual Review of Astronomy and Astrophysics 58, 27–97 (2020). DOI.
• Maiolino, R., et al. “A small and vigorous black hole in the early Universe.” Nature 627, 59–63 (2024). DOI.
• Volonteri, M., Habouzit, M., and Colpi, M. “The origins of massive black holes.” Nature Reviews Physics 3, 732–743 (2021). DOI.
• Zou, F., et al. “Revisiting the Claim for a Direct-Collapse Black Hole in UHZ1 at $z=10.05$.” arXiv:2603.24893v1 (2026). Open preprint.