{"id":"39873a5f-f6f2-4b65-a892-9a214f7c6f1c","arxiv_id":"2602.22305","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"High-redshift JWST broad-line AGNs may be low-mass black holes accreting far above Eddington, whose steep, over-cooled coronal spectra explain their X-ray non-detections and make them appear overmassive.","lead":"Using 14 JWST-detected broad-line AGNs with no X-ray detections, the paper argues these are low-mass black holes accreting far above the Eddington limit, not the overmassive black holes implied by single-epoch mass estimates. The result matters because it would dissolve a major puzzle about how supermassive black holes grew so early in cosmic history.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central low-mass preference rests on the unvalidated L24 self-shadowed BLR model; X-ray upper limits alone do not select it.","rationale":"The reader identified the absence of X-ray absorption as the weakest assumption. That is a real and explicitly acknowledged caveat (Section 4), but it is not the least secure link in the chain. Even under the no-absorption assumption, the X-ray data do not select the low-mass super-Eddington branch: Fig. 2 demonstrates that many sources can be fit with standard spectra and M25 masses. The bimodality and the claimed preference for the low-mass branch are produced by the joint Hα fit, which depends entirely on the L24 self-shadowed BLR model. This model is self-cited, is not independently validated for z>6 sources, and its predictions enter the likelihood directly. A critical test is to rerun the joint MCMC with the standard R-L relation; if the preference vanishes, the conclusion is an artifact of the BLR geometry assumption. I also note the statistical comparison of the two modes by mean likelihood is not a valid model selection, which further weakens the 'strongly preferred' language. Therefore, while I do not object to the plausibility of super-Eddington accretion in these sources, I find the evidence insufficient to establish it as the explanation. My verdict remains CONDITIONAL (as the reader's), because the issues are addressable with additional modeling and future data.","tokens_in":14608,"tokens_out":13674,"duration_ms":147540,"concrete_test":"Re-run the joint MCMC (Section 3.2.2) with the same X-ray likelihood and priors, but replace the L24 self-shadowed BLR radius with the standard R_BLR ∝ L_Hα^0.5 relation (no geometric correction). Compute the Bayes factor between the low-mass (f_Edd > 1) and high-mass (f_Edd < 1) modes using nested sampling. If Δln Z < 5 in favor of the low-mass solution, the central claim is model-dependent and should be downgraded.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim — that the low-mass, super-Eddington solution is 'strongly preferred' — is established in Section 3.2.2 via a joint fit to Hα line luminosity/FWHM and X-ray non-detections, using the self-shadowed BLR prescription of Lupi et al. (2024b, L24). However, Section 3.1 (Fig. 2) shows that the X-ray upper limits alone are compatible with standard Γ=2 spectra and the M25 virial masses for most of the 14 sources. Thus the X-ray data do not by themselves require a super-Eddington explanation. What breaks the degeneracy is the L24 model's prediction that, for a given observed Hα luminosity, the inferred BH mass is much lower when the BLR is partially shadowed by the inner funnel. If this geometric correction is inaccurate or overestimated, the low-mass solution is an artifact. The L24 model is not independently validated here; the only external support cited is the GRAVITY+ measurement at z~0.4, which is encouraging but not a direct test of the extreme f_Edd regime at z>6. Moreover, the comparison of the two modes uses the ratio of mean likelihoods after an arbitrary split at f_Edd=1 (Section 3.2.2), not a proper Bayes factor, and can be biased by the hard X-ray cutoff in Eq. (2). A correct evidence calculation could substantially reduce the claimed preference.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reanalyzes 14 z>4 JWST broad-line AGNs by coupling the Madau & Haardt (2024) coronal over-cooling model with Kubota & Done (2019) slim-disc spectra. An MCMC explores black-hole mass, Eddington ratio, and spin, using either bolometric luminosities or Hα line luminosity/FWHM together with Chandra non-detections. The joint analysis yields strongly bimodal posteriors: a high-mass, very low-Eddington branch and a low-mass, highly super-Eddington branch. The authors argue the low-mass branch is strongly preferred, gives masses closer to local MBH–Mstar relations, and produces extreme bolometric corrections consistent with low-redshift super-Eddington AGNs.","tokens_in":14989,"tokens_out":3568,"duration_ms":39064,"significance":"If correct, the paper would materially change the interpretation of JWST high-redshift broad-line AGN masses and X-ray weakness: many objects could host ~1e6–1e7 solar-mass black holes accreting at f_Edd >> 1, with the apparent X-ray faintness caused by intrinsically steep, over-cooled coronal spectra rather than heavy obscuration. The paper's strengths are the explicit spectral library, the use of a full MCMC rather than point estimates, and the comparison with low-redshift super-Eddington samples. However, the central statistical claim of a 'strong preference' for the low-mass branch rests on an ad hoc likelihood comparison rather than a proper Bayesian model comparison, and the inference is heavily dependent on the L24 BLR geometry and MH24 coronal prescriptions, which are not independently validated in the regime considered. The paper is therefore valuable as a scenario-driving study, but its headline claim is not yet established at the level claimed.","major_comments":[{"comment":"The claim that the low-mass, super-Eddington solution is 'strongly preferred' is quantified by the ratio of mean likelihoods of MCMC samples after an arbitrary split at f_Edd,thin = 1. This is not a Bayesian model-comparison statistic. Mean likelihood over posterior samples is sensitive to the arbitrary split, to the prior weight, and to the hard likelihood cutoff of Eq. (2). The authors should compute marginal likelihoods (evidence) or posterior mass ratios for the two branches, and check robustness to the split threshold and to the assumed priors. Without this, the headline preference is not supported at the stated strength.","section":"Sec. 3.2.2, Figs. 5-6"},{"comment":"Figure 2 shows that, for most sources, the X-ray non-detections are compatible with standard Gamma=2 spectra and the M25 masses/spins. Thus the X-ray upper limits alone do not require the super-Eddington, low-mass branch. The preference for that branch is driven by the L24 broad-line model's self-shadowing correction. Since the L24 and MH24 models are from the same group and are not independently validated at f_Edd >> 1 or z>6, the paper should include a sensitivity test that replaces the L24 BLR prescription with a standard virial BLR calibration, or otherwise demonstrates that the low-mass preference is not an artifact of the adopted geometric model. The GRAVITY+ validation at z~0.4 is encouraging but does not test the extreme regime claimed here.","section":"Sec. 3.1, Fig. 2"},{"comment":"The likelihood applies a hard Heaviside cutoff, rejecting any model with predicted X-ray luminosity above the observed limit. Real X-ray non-detections are probabilistic, with background and calibration uncertainties; a step-function likelihood can bias posterior estimates, particularly when the model lies near the boundary. The authors should use a survival likelihood based on the actual upper limit distribution (e.g., Poisson likelihood), or at least verify that the bimodal structure and the preferred branch are unchanged when the cutoff is softened.","section":"Eq. (2), Sec. 2"},{"comment":"The paper explicitly states that the results assume the absence of extremely high gas column densities capable of absorbing the X-ray emission. This is a load-bearing assumption: if absorption is present, the non-detections could be explained without super-Eddington coronal suppression, and the low-mass inference would not follow. The authors should provide a quantitative justification — e.g., constraints on NH from X-ray stacking, spectral energy distribution fits, or the observed presence of broad lines requiring an unobscured line of sight — rather than only a caveat. Without this, the central inference remains conditional on an unverified assumption.","section":"Sec. 4"}],"minor_comments":[{"comment":"The caption reads 'Same as Fig. 6' but it should refer to Fig. 5. Also, the y-axis of the right panel should clarify that the Eddington ratio is log-scaled in the figure.","section":"Fig. 6 caption"},{"comment":"The piecewise mapping for theta_fun is stated to be 'adjusted to ensure continuity,' but the three branches do not appear exactly continuous at the boundaries (e.g., at f_Edd = 76.5). Please verify the constants or the intended continuity condition.","section":"Sec. 2, Eq. (1)"},{"comment":"The model names 'agnslim' and 'xspec' should be typeset consistently (e.g., in italics or with proper package formatting), and the fixed values f_c = 1 and p1 = 0.3 should be flagged as fixed assumptions rather than varied parameters in the main MCMC.","section":"Sec. 2"},{"comment":"The statement that the '10 highest-probability solutions' were used from the L24 run is vague. Please specify how these were selected and whether the results are robust to that choice.","section":"Sec. 3.2.1"}],"recommendation":"major_revision","confidential_remarks":"The paper's central conclusion depends on models developed by the same authors, and the comparison that establishes the 'strong preference' is not a proper Bayesian evidence calculation. These are fixable with additional analysis, but the current version overstates the strength of the result. I would not recommend rejection, as the underlying scenario is timely and the observational comparison is useful, provided the statistical and validation issues are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the paper is a serious attempt to connect two observations—JWST broad-line AGNs look overmassive and are X-ray weak—into one super-Eddington picture. Second, the central claim that the low-mass, super-Eddington solution is 'strongly preferred' is not actually established by the analysis as written. The preference comes from the authors' own BLR self-shadowing model (L24), and the model comparison uses an ad hoc split at f_Edd=1 followed by a ratio of mean likelihoods, not a proper Bayes factor. The X-ray upper limits alone, as the paper itself shows in Fig. 2, are compatible with standard Γ=2 spectra and the M25 virial masses for most sources. So the data do not independently demand the low-mass branch; the model supplies it.\n\nWhat's genuinely new: the paper builds a joint MCMC that couples the Madau-Haardt corona over-cooling with slim-disc spectra and the L24 BLR prescription, and produces posteriors for BH mass, Eddington ratio, and spin for all 14 objects. The cumulative spin distributions and the comparison with local super-Eddington AGNs (Laurenti 2022) are new and useful. The authors are transparent about the main caveats—no absorption, slim-disc assumption, the L24 model not tested at extreme f_Edd—so the paper reads like a hypothesis rather than an overclaim. That's a plus.\n\nThe soft spots are in proportion to how much the claim depends on them. The hard Heaviside cutoff on X-ray detections ignores measurement uncertainty and can bias the likelihood comparison. No code or data are shipped, which makes the L24-model sensitivity hard to check. And the no-absorption assumption is explicitly stated but remains a viable alternative. These are not fatal to the idea—the super-Eddington explanation for the X-ray weakness is physically motivated and worth taking seriously—but they are load-bearing for the 'strong preference' language. A referee should ask for a proper evidence calculation, tests of the L24 BLR model (e.g., by varying its free parameters), and ideally release of the spectral tables.\n\nWho's it for: anyone working on high-z BH growth, JWST AGN demographics, or X-ray surveys. It deserves a serious referee. I'd send it out, expecting revision.","headline":"Plausible but not established: the low-mass super-Eddington solution is built into the authors' own BLR model, and the 'strong preference' rests on an ad hoc likelihood ratio rather than a proper evidence calculation.","tokens_in":15445,"tokens_out":3448,"would_cite":true,"duration_ms":32693,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"JWST's X-ray-quiet broad-line AGNs may be low-mass black holes accreting far above Eddington.","keywords":["super-Eddington accretion","X-ray weakness","high-redshift broad-line AGN","black hole mass estimates","slim accretion discs","coronal over-cooling","bolometric corrections","Chandra non-detections"],"falsifier":"A decisive test would be a deep X-ray detection of one of the 14 sources: if the 2-10 keV spectrum is flat or absorbed and the bolometric correction is near the local value, the steep super-Eddington explanation is wrong; alternatively, measuring an absorbing column density above about 10^24 cm^-2 toward any of these sources would break the paper's no-obscuration premise and shift the mass inference.","tokens_in":14539,"feed_emoji":"🕳️","tokens_out":8564,"duration_ms":83710,"temperature":0.7,"pith_summary":"JWST has found many broad-line active galactic nuclei at redshift above 4 that are remarkably quiet in X-rays, and the simplest reading is that they host abnormally massive black holes. This paper argues that those same X-ray non-detections point the other way: in super-Eddington accretion the disc puffs up into a narrow funnel that cools the corona and steepens its X-ray spectrum, so the source becomes X-ray faint even while shining brightly in ultraviolet and broad-line emission. Jointly fitting broad H-alpha lines and the Chandra upper limits for 14 sources, the authors find a strongly bimodal answer and a clear preference for low-mass (~10^6 to 10^7 solar mass) black holes accreting far above Eddington. If correct, this lowers estimated black hole masses by 0.5 to 1.5 dex, eases the tension with local scaling relations, and removes the need for exotic heavy seeds growing at the Eddington limit for the whole age of the universe.","feed_headline":"X-ray-silent early AGNs may be small super-Eddington accretors","feed_subtitle":"JWST's X-ray-quiet galaxies may hide million-solar-mass black holes gorging above Eddington — easing mass puzzles.","key_machinery":"The carrier of the argument is the coupling between a super-Eddington slim accretion disc and its corona. In this picture the disc becomes geometrically thick and creates a narrow, self-shadowed funnel; soft disc photons entering the hot corona are repeatedly reflected off the funnel walls, enhancing Compton cooling and lowering the coronal electron temperature, which makes the emergent X-ray spectrum steeper and intrinsically weaker per unit bolometric luminosity. A piecewise relation maps the funnel half-opening angle to the Eddington ratio, and a grid of slim-disc spectra (with varying power-law slope and electron temperature) is used to tabulate bolometric and 2-10 keV luminosities. An M","core_discovery":"On its own terms, the paper's discovery is that the joint constraint from X-ray non-detection and broad-line emission is bimodal and almost always prefers a super-Eddington solution. The manuscript shows that adopting the standard single-epoch virial masses and bolometric luminosities cannot explain the faintest X-ray upper limits unless the spectral slope is already modified, and that once a full MCMC is run two families emerge: low-mass (10^6 to 10^7 solar mass) black holes accreting at f_Edd >> 1 with very steep intrinsic X-ray spectra, and high-mass (often > 10^9 solar mass) nearly dormant black holes. The high-mass branch is disfavoured by likelihood, by the implied black-hole-to-stella","pith_inferences":["A natural extension, not made in the paper, is to apply the same joint analysis to the broader 'Little Red Dot' population; if those objects are the same engine viewed differently, their X-ray faintness should follow the same steep-spectrum pattern with an additional absorbing component.","The model predicts a tight relation between H-alpha width and the depth of X-ray non-detection for a fixed bolometric luminosity; stacking deeper Chandra observations of these 14 sources should reveal a population with very steep photon indices (Gamma > 3) rather than an absorbed population.","If future 2-10 keV observations detect any of these sources at moderate bolometric correction, the low-mass branch would be falsified for that source, and the same framework could instead constrain how fast the corona over-cools as a function of funnel geometry.","The paper's stated caveat about non-advective radiation tori suggests a testable avenue: radiative-transfer simulations comparing funnel shapes could show whether the quantitative preferences for mass and accretion rate are robust, while leaving the qualitative super-Eddington preference intact."],"forward_implications":["If the low-mass solution is right, single-epoch virial black hole masses for many z > 4 broad-line AGNs are overestimated by roughly 0.5-1.5 dex.","The observed lack of X-rays becomes a signature of super-Eddington growth rather than a sign of unusually inactive or obscured black holes, so deep X-ray surveys should systematically miss this population.","The high-redshift sources would sit on or below local M_BH-M_star relations, removing the need for 'overmassive' black holes that challenge galaxy-formation models.","Because super-Eddington accretion allows rapid mass assembly from moderate seeds, the new masses reduce the required seed masses and duty cycles of Eddington-limited growth.","The inferred spin distribution differs sharply between the two branches: the disfavoured high-mass branch requires mostly low spins, while the preferred low-mass branch leaves spin largely unconstrained, a prediction testable with future observations."],"fun_headline_variants":["X-ray silence hints at super-Eddington black-hole growth","JWST's X-ray-quiet AGNs may be overfed, not overmassive","Small and fast: early black holes grow super-Eddington","Super-Eddington black holes explain JWST's X-ray faint AGNs"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The analysis assumes the absence of extremely high gas column densities capable of absorbing the AGN's X-ray emission; if such obscuration were present, the Chandra non-detections could be explained by absorption rather than by intrinsically steep super-Eddington spectra, and the low-mass inference would collapse.","fun_headline_variants_meta":{"raw":{"variants":["X-ray silence hints at super-Eddington black-hole growth","JWST's X-ray-quiet AGNs may be overfed, not overmassive","Small and fast: early black holes grow super-Eddington","Super-Eddington black holes explain JWST's X-ray faint AGNs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000737,"raw_usage":{"total_tokens":3227,"prompt_tokens":937,"completion_tokens":2290,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":681,"completion_tokens_details":{"reasoning_tokens":2210}},"tokens_in":681,"tokens_out":2290,"duration_ms":14327,"temperature":1.0,"reasoning_tokens":2210,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T20:44:14.848383+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test would be a deep X-ray detection of one of the 14 sources: if the 2-10 keV spectrum is flat or absorbed and the bolometric correction is near the local value, the steep super-Eddington explanation is wrong; alternatively, measuring an absorbing column density above about 10^24 cm^-2 toward any of these sources would break the paper's no-obscuration premise and shift the mass inference.","supporting_citations":[],"review_version":1}