{"id":"57d57b90-8c0c-46ee-9696-a7e2e479676a","arxiv_id":"2601.15207","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Super-Eddington black hole accretion is the COLIBRE ingredient that reproduces JWST's high-redshift massive quiescent galaxy abundance.","lead":"Simulations show that the COLIBRE galaxy-formation model matches JWST's surprising abundance of massive, already-quiet galaxies at early cosmic times only if its black holes are allowed to swallow gas faster than the Eddington limit. Models that cap black-hole feeding at the Eddington rate produce far too few of these quenched galaxies, pointing to super-Eddington growth as the driver of early shutdown.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Super-Eddington events may be resolution artifacts: the causal link to real super-Eddington accretion remains untested at higher resolution.","rationale":"The reader's weakest assumption identifies the same load-bearing point: the reality of the subgrid super-Eddington accretion rates. This is the correct focus because the paper's headline conclusion is causal, not merely predictive. The controlled comparison (f_Edd,max=100 vs 1 vs 0.1) is a strength: same ICs, same subgrid model, and a clear earlier-quenching sequence in Fig 1. But nothing in the paper directly validates that the f_Edd>1 bursts correspond to genuine super-Eddington accretion rather than to unresolved cold gas boosting the Bondi rate. The closing limitation statement effectively concedes this. The small-volume/cosmic-variance issue is real but secondary: even if the f=1 model could produce some MQGs in a larger volume, the internal same-galaxy comparison would still show that removing super-Eddington accretion delays quenching; the physical interpretation is what needs scrutiny. The concrete high-resolution rerun would settle whether the central mechanism survives resolution. Since this concern is already reflected in the CONDITIONAL verdict, no verdict change is needed.","tokens_in":13011,"tokens_out":6591,"duration_ms":75600,"concrete_test":"Re-run the same (100 cMpc)^3 initial conditions at m6 and m5 resolution for both f_Edd,max=100 and f_Edd,max=1, keeping all other subgrid parameters fixed. Check whether (i) the distribution of instantaneous f_Edd values, especially the f_Edd>1 events and the ~50% super-Eddington mass fraction at z~7 (Fig 4), is preserved; and (ii) the MQG number-density gap between f=100 and f=1 at z>6 persists. A cheaper complementary check: for the representative galaxy in Fig 1, recompute the Bondi-Hoyle rate using only gas within the resolved Bondi radius and compare with the full kernel rate; if most f>1 events disappear when unresolved gas is excluded, the super-Eddington attribution is a resolution artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that super-Eddington BH accretion (f_Edd>1, up to ~10) is what makes COLIBRE match JWST MQG counts at z>6. The controlled f_Edd,max variation in Sec 2.2 cleanly shows that the cap matters, but the interpretation of the f_Edd>1 events as physical super-Eddington accretion relies on the subgrid Bondi-Hoyle rate (Sec 2.1) being trustworthy at high density. At m7 resolution, the BH accretion rate is evaluated from SPH gas properties smoothed over scales of order the 1.4 pkpc softening; unresolved cold, dense gas can inflate the Bondi rate. If so, the 'super-Eddington events' in Figs 1 and 4 are numerical constructs, and Fig 3's only-fiducial-match is an artifact of the upper cap rather than evidence for real super-Eddington accretion. The paper's own caveat that 'BH physics in COLIBRE is necessarily captured rather crudely' (Conclusions) and the absence of a resolution comparison in this work leave this possibility open. External zoom-in simulations and the statement that super-Eddington accretion persists at m6/m5 are cited but not demonstrated in the controlled variation, so they do not resolve the concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses controlled COLIBRE simulations in a (100 cMpc)^3 volume at m7 resolution, varying only the maximum allowed Eddington fraction of black hole accretion (f_Edd,max = 10^2, 1, 0.1) while keeping all other physics and initial conditions fixed. It reports that the fiducial model that allows super-Eddington accretion produces massive quiescent galaxies (MQGs) at z>6 and matches recent JWST number-density estimates, whereas the capped models produce no or very few MQGs. It further quantifies that in the fiducial model about half of the high-redshift BH accreted mass is gained at f_Edd>1, even though such events are rare in time. The paper concludes that super-Eddington BH accretion is the key process enabling early quenching and resolving the tension with JWST observations.","tokens_in":13370,"tokens_out":5132,"duration_ms":56232,"significance":"The strength of the work is its clean experimental design: identical initial conditions, only the Eddington cap is varied, and the JWST data are external and were not used for calibration. If the result holds, it identifies the microphysical mechanism behind the success of COLIBRE in reproducing early massive quiescent galaxies. The paper also provides quantitative diagnostics (fraction of mass accreted in different f_Edd bins) that allow the mechanism to be inspected. However, the central claim rests on (i) number-density comparisons with only a handful of simulated and observed galaxies, and (ii) the assumption that the subgrid Bondi-Hoyle rates leading to f_Edd>1 are physically faithful. Both require scrutiny before the conclusion can be accepted as firmly established.","major_comments":[{"comment":"The paper claims that the f_Edd,max=1 and 0.1 models 'strongly undershoot' the JWST constraints and that only the fiducial model is consistent. However, the simulated volume yields only one MQG at z~7 and three at z~6.5 in the fiducial model, and zero in the f_Edd,max=1 model. The JWST data points themselves are based on one or few objects (e.g., Weibel et al. 2025). With Poisson statistics, zero events in a volume of 10^6 cMpc^3 is not formally inconsistent with an observed density of ~10^-6 cMpc^-3 at the 95% level. The paper does not provide a significance test or likelihood comparison. The conclusion of inconsistency therefore needs quantitative support, e.g., Poisson confidence intervals on the model predictions and a statement of the confidence with which the capped models are excluded by the current data.","section":"Sec. 3, Fig. 3"},{"comment":"The physical interpretation of the f_Edd>1 events as genuine super-Eddington accretion relies on the modified Bondi-Hoyle-Lyttleton subgrid prescription. At m7 resolution with a 1.4 pkpc gravitational softening, cold dense gas that is unresolved can inflate the Bondi rate, making the apparent super-Eddington episodes numerical artifacts rather than real accretion phenomena. The paper cites higher-resolution COLIBRE runs and external zoom-in simulations, but it does not present a resolution study of the f_Edd variation or of the specific super-Eddington events analysed here. Without such a demonstration, the claim that 'super-Eddington growth' is what resolves the JWST tension is not fully established; the controlled variation demonstrates only that the cap matters, not that the rates above the Eddington limit are physical. The authors should either add a resolution comparison (e.g., m6/m","section":"Sec. 2.1, Sec. 3, Figs. 1 and 4"},{"comment":"The comparison with observational number densities is affected by cosmic variance because the simulation volume is only (100 cMpc)^3, corresponding to an expected count of order unity at the claimed density of 10^-6 cMpc^-3. The paper acknowledges this and cites larger-volume runs from Chaikin et al. (2025b), but the model variations (f_Edd,max = 1 and 0.1) are only run in the small volume. The paper should at least discuss how cosmic variance could affect the inferred significance of the difference between the models, or ideally run the variations in a larger volume (if feasible) to increase the sample size.","section":"Sec. 3, Fig. 3"}],"minor_comments":[{"comment":"The caption says the shaded region shows 'the maximum of the 16th–84th percentile range and Poisson uncertainties'. It would be clearer to state explicitly whether the shaded band is the union of these two uncertainties or the larger envelope, and how the 300 bootstrap realizations are used in the Poisson term.","section":"Sec. 3, Fig. 3 caption"},{"comment":"The footnote explaining that f_Edd,max=10^2 is chosen for numerical reasons is useful, but the text around it could clarify that the cap is rarely reached in practice (as stated later in Sec. 3) to avoid the impression that the model is tuned to reproduce f_Edd~10 events.","section":"Sec. 2.1, footnote 2"},{"comment":"The black dotted lines for f_Edd=1 and 0.1 growth assume continuous accretion starting at z=10 and z=9. It would help to state in the caption whether this includes the seed mass and whether the starting redshift is chosen arbitrarily or motivated by the simulations.","section":"Fig. 1 top-right panel"},{"comment":"The cumulative fractions in Fig. 4 are computed over all BH particles. The text states that super-Eddington events contribute 50–70% of accreted mass at z≳7, but the exact redshift binning and the definition of 'accretion event' (time-step versus continuous) could be spelled out in the caption for reproducibility.","section":"Sec. 3, Fig. 4"},{"comment":"The paper repeatedly uses 'colibre' in lowercase in the running text and 'COLIBRE' in the title/abstract. Please standardise the capitalization, as journal style typically uses small caps for code names.","section":"Global"}],"recommendation":"major_revision","confidential_remarks":"The paper is well written and the controlled f_Edd,max variation is a nice experiment. The main concerns are statistical robustness (only a few galaxies) and the lack of a resolution test for the subgrid super-Eddington rates. If the authors can provide a significance analysis and a resolution check or a strong argument that the Bondi rates are physical, the paper would be acceptable. The current form oversells the 'strong' inconsistency between the models. I recommend major revision rather than rejection because the core methodology is sound and the issues are addressable within the scope of the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline is that this is a clean, controlled attribution study: same initial conditions, only f_Edd,max varies, and the difference in high-z MQG counts is stark. That is exactly the right way to isolate a physical ingredient, and it is new relative to the earlier COLIBRE papers, which reported agreement without identifying the cause. The paper shows the mechanism convincingly within its own model: the fiducial run (f_Edd,max=100) makes MQGs by z~7, the Eddington-limited run starts later, and the 0.1 run essentially never. The 50% super-Eddington mass fraction at high redshift is a useful quantitative summary. I think the reader's conditional verdict is about right. \n\nThe soft spots are real but proportionate. The stress-test concern about unresolved cold gas inflating the Bondi-Hoyle rate is legitimate. At m7 resolution with 1.4 pkpc softening, the subgrid accretion rate is not directly validated, and the paper's own caveat that BH physics is captured crudely is honest. That said, the companion claims that super-Eddington accretion persists at m6/m5 and in larger volumes are checkable and are cited, not hidden; the central logic of the controlled variation does not depend on proving that the events are physically real in nature, only that they are the cause within the model. If I have a stronger reservation, it is the statistical basis: the highest-redshift bins in Fig. 3 are built on a handful of galaxies, so the zero-count discrimination rests on small numbers. The authors acknowledge cosmic variance and point to larger-volume runs, which is fair. Data and code are not public yet, which limits reproducibility, but that is standard for this subfield and the release path is stated. \n\nOverall, this is a serious piece of work. The controlled design is the right tool for the question, the interpretation is appropriately hedged, and the paper will push other simulations to test super-Eddington accretion. It deserves peer review. My own verdict would be accept with minor-to-moderate revision: the referee should ask for a resolution test or at least a clearer statement of what higher-resolution runs show in this exact controlled set, but the main argument holds.","headline":"Clean controlled experiments make a solid case that super-Eddington accretion is the COLIBRE ingredient that explains the JWST MQG counts; the main caveat about subgrid artifacts is real but not fatal.","tokens_in":13842,"tokens_out":1546,"would_cite":true,"duration_ms":18481,"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":"Super-Eddington black hole growth is why JWST sees massive quiet galaxies so early","keywords":["super-Eddington accretion","massive quiescent galaxies","high redshift","AGN feedback","black hole growth","JWST","cosmological simulations"],"falsifier":"A direct check is to run the same COLIBRE setup at higher resolution (m6 or m5) in the same 100 cMpc box: if the super-Eddington bursts and the excess of quiescent galaxies disappear or shrink dramatically when the Bondi radius is better resolved, the central claim collapses. Observationally, a stacking analysis of z>6 massive galaxies that yields Eddington ratios clustered at or below unity for the actively accreting population would contradict the f_Edd∼10 events required by the model.","tokens_in":12903,"feed_emoji":"⚫️","tokens_out":2114,"duration_ms":25260,"temperature":0.7,"pith_summary":"This paper claims that the reason the COLIBRE galaxy-formation simulations match JWST observations of massive quiescent galaxies at redshift above 6 is that their black holes are allowed to accrete gas much faster than the Eddington limit. By rerunning the same simulations with the black-hole accretion rate capped at the Eddington rate and at 10 percent of it, the authors show that only the super-Eddington-allowing model reproduces the observed abundance of these early 'dead' galaxies. The enhanced black-hole growth triggers AGN feedback early enough to shut off star formation in massive galaxies. About half of the black-hole mass accreted at high redshift in the simulations is gained in short, rare super-Eddington bursts. A sympathetic reader would conclude that super-Eddington accretion is not a side detail but a load-bearing ingredient for explaining the early universe.","feed_headline":"Overfed black holes made the first quiet galaxies","feed_subtitle":"Simulations match JWST's early quiescent galaxies only when black holes accrete far above the Eddington limit.","key_machinery":"The central control parameter is the maximum allowed Eddington fraction f_Edd,max, the cap on the black-hole accretion rate in units of the Eddington rate. The accretion rate itself is a modified Bondi-Hoyle-Lyttleton formula with turbulence and vorticity corrections, capped at f_Edd,max = 10^2 in the fiducial model and at 1 and 0.1 in the variation runs. Varying this one parameter while holding all other subgrid physics fixed isolates the role of super-Eddington accretion in driving early black-hole growth, AGN feedback, and galaxy quenching.","core_discovery":"The fiducial COLIBRE model, which caps black-hole accretion at 100 times the Eddington rate, is consistent with JWST number-density constraints for massive quiescent galaxies at z≳6, whereas models capped at the Eddington limit or at 0.1 times it strongly undershoot the data. In the fiducial model, black holes in massive high-redshift galaxies undergo bursts with Eddington fractions f_Edd∼10, and by z≈6 roughly 50 percent of their accreted mass is gained in super-Eddington events, even though black holes spend only a few percent of their time in that regime. These bursts push black-hole masses up early, switch on AGN feedback sooner, and quench the host galaxies, producing the observed early","pith_inferences":["If super-Eddington accretion is genuinely at work, it eases the seed black-hole problem: heavy seeds are not required because a modest seed can balloon to 10^8–10^9 solar masses within a few hundred million years, consistent with JWST-discovered overmassive black holes at z>6.","A direct observational test is to measure Eddington ratios of z>6 massive galaxies: if X-ray, radio, or mid-infrared stacking shows that their black holes are accreting near or below Eddington, then the simulation's super-Eddington bursts would be ruled out as the driving mechanism.","The verdict on this mechanism is sensitive to the subgrid treatment of small-scale cold gas: higher-resolution runs (m6, m5) that resolve the Bondi radius should be checked to see whether the f_Edd>1 bursts persist or dissolve into ordinary Eddington-limited accretion, which would separate a physical effect from a numerical artifact."],"forward_implications":["Galaxy-formation models that forbid super-Eddington accretion will systematically underpredict the abundance of massive quiescent galaxies at z≥6, so such models need this physics to match JWST.","Super-Eddington accretion, though episodic, can dominate the mass budget of black hole growth at high redshift, implying that Eddington-limited growth assumptions are inadequate for early massive black holes.","The earlier onset of AGN feedback produced by rapid black-hole growth is the proximate cause of early quenching, connecting black-hole accretion physics directly to the star-formation histories of massive galaxies.","The models with and without super-Eddington accretion converge by z≈5, so the effect is specific to the first gigayear of galaxy formation rather than a general low-redshift feature."],"fun_headline_variants":["Super-Eddington black holes spawn early quiet galaxies","Overfed black holes birth the universe's first quiet galaxies","Black hole gluttony explains JWST's early massive quiescent galaxies","Greedy black holes quench massive galaxies in the early universe","Super-Eddington accretion drives early massive quiescent galaxies"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The modified Bondi-Hoyle formula with turbulence and vorticity corrections is assumed to produce physically meaningful super-Eddington accretion events, rather than being artificially inflated by unresolved cold, dense gas near the black-hole particle; if inflated, the early quenching and the JWST agreement would be numerical accidents.","fun_headline_variants_meta":{"raw":{"variants":["Super-Eddington black holes spawn early quiet galaxies","Overfed black holes birth the universe's first quiet galaxies","Black hole gluttony explains JWST's early massive quiescent galaxies","Greedy black holes quench massive galaxies in the early universe","Super-Eddington accretion drives early massive quiescent galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1425,"prompt_tokens":839,"completion_tokens":586,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":583,"completion_tokens_details":{"reasoning_tokens":499}},"tokens_in":583,"tokens_out":586,"duration_ms":6733,"temperature":1.0,"reasoning_tokens":499,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T08:54:37.541372+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct check is to run the same COLIBRE setup at higher resolution (m6 or m5) in the same 100 cMpc box: if the super-Eddington bursts and the excess of quiescent galaxies disappear or shrink dramatically when the Bondi radius is better resolved, the central claim collapses. Observationally, a stacking analysis of z>6 massive galaxies that yields Eddington ratios clustered at or below unity for the actively accreting population would contradict the f_Edd∼10 events required by the model.","supporting_citations":[],"review_version":1}