{"id":"73224c65-6c28-491c-aed9-08fff831ebc3","arxiv_id":"2506.03354","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"SHARK v2.0 predicts that SMBH masses at fixed galaxy stellar mass were about 1 dex higher at z=9 than at z=0, with morphology explaining most of the scatter.","lead":"This paper uses the SHARK v2.0 galaxy formation model to predict how the masses of supermassive black holes relate to galaxy stellar mass from 13 billion years ago to today. It finds strong evolution, with black holes about ten times heavier at fixed stellar mass at early times, and argues galaxy morphology drives most of the scatter.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The z=9-to-z=0 offset is anchored by the cold-gas accretion parameters fSMBH and eSB that were manually re-tuned (0.7x and 7.5x) with no sensitivity analysis, so the headline evolution could be an artifact of that retuning.","rationale":"I read the paper as making a testable SAM prediction, not as a measurement. The open-source SHARK code, the large PM volume, and the explicit parameter table are real strengths, and the authors are transparent that the z=0 M_bulge-M_BH normalisation is a secondary calibrator. The reader's weakest-assumption analysis identifies the correct soft spot: the z=9 anchor of the 1 dex offset is produced by the same cold-gas accretion prescription that was manually re-tuned by factors of ~0.7 (fSMBH) and ~7.5 (eSB) to fix PM-specific offsets. This is an internal robustness issue, not a disagreement with the observational consensus. A single rerun with the L24 values, or a small grid, would settle whether the evolution is robust. Until that is done, the conditional verdict is appropriate; I see no basis to reject the paper, and the non-evolution claims (morphology, scatter, BHMF) are supported by multiple comparisons. My read leaves the reader's verdict unchanged.","tokens_in":29926,"tokens_out":8769,"duration_ms":89839,"concrete_test":"Rerun SHARK v2.0 on Planck-Millennium with the L24 cold-gas parameters (fSMBH=0.01, eSB=15), re-fitting the six optim parameters to the z=0 and z=1 SMFs and to the z=0 M_bulge-M_BH normalisation, and measure the median log M_BH at Mstar = 10^10 and 10^11 M_sun at z=0, 3, 6, and 9. If the z=9-to-z=0 offset changes by more than ~0.4 dex relative to the fiducial ~1 dex, the evolution claim is not robust to the manual retuning; if it is stable across this run and a small grid (e.g., eSB=2,4,8,15 with fSMBH adjusted to preserve the z=0 normalisation), the high-z prediction is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline evolution claim (Section 5: SMBH masses ~1 dex lower at z=0 than at z=9 at fixed stellar mass) is anchored at high redshift by the cold-gas accretion channel. Equations (2)-(3) set this channel with fSMBH and eSB, and Table 2 shows these parameters were manually re-tuned from the L24 values (fSMBH: 0.01 -> 0.007; eSB: 15.0 -> 2.0) to remove PM-vs-SURFS offsets in the SMBH mass relations. Section 3.1 concedes that the z=0 normalisation is calibrated rather than predicted. Because Figure 9 shows cold-gas accretion dominates SMBH growth at z >~ 6, the z=9 end of the 1 dex offset is essentially set by this uncalibrated, manually adjusted prescription. In Eq. (3), tau_CG = eSB r_bulge/v_bulge, so the product fSMBH/eSB entering the cold-gas accretion rate changes by a factor ~5.25 between L24 and this work. No sensitivity analysis of the evolution claim to these parameters is provided, so the predicted offset could be an artifact of the retuning rather than a robust model prediction. The morphology-scatter claim is secondary and would survive softening; the evolution claim is the load-bearing result.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses the SHARK v2.0 semi-analytic model, re-calibrated to the Planck-Millennium N-body simulation, to predict the evolution of the M_bulge-M_BH and M_star-M_BH scaling relations from z=9 to z=0. The authors manually retune the cold-gas accretion parameters fSMBH and eSB (Table 2) to reproduce the local M_bulge-M_BH normalisation, and find that the model broadly matches observed BHMFs and scaling relations over the redshift range, predicts a ~1 dex decrease in SMBH mass at fixed stellar mass from z=9 to z=0, a factor of 2-5 increase in scatter toward z=0, and that bulge-to-total ratio is the dominant predictor of scatter in a random forest analysis. They interpret these results as evidence for strong cosmic evolution of the scaling relations and a central role for morphology, with environment playing only an indirect role.","tokens_in":30276,"tokens_out":8064,"duration_ms":85781,"significance":"If robust, the evolution prediction is a valuable, large-volume prediction from an open-source SAM that can be tested with 4MOST and JWST samples; the paper also provides useful qualitative predictions for the morphology-dependence of the relations and for the occupation fraction of low-mass black holes. Strengths include the transparency about primary vs secondary calibration (Section 2.2), the explicit statement that the z=0 normalisation is calibrated rather than predicted (Section 3.1), the test of seed-mass insensitivity (Section 2.1), and the public availability of the code. The high-redshift evolution claim, however, rests on the uncalibrated cold-gas accretion prescription, and the morphology-scatter claim is based on a predictive-correlation analysis rather than a causal decomposition; both caveats need to be addressed before the strong conclusions can be accepted.","major_comments":[{"comment":"The predicted ~1 dex offset in SMBH mass at fixed stellar mass between z=9 and z=0 is the central claim, and it is anchored at high redshift by the cold-gas accretion channel, which Figure 9 shows dominates SMBH growth at z>~6. This channel is controlled by fSMBH and eSB, which were manually re-tuned from (0.01, 15.0) to (0.007, 2.0) when moving the model to Planck-Millennium (Section 2.2). Because the cold-gas accretion rate in Eq. (3) is proportional to fSMBH/eSB, this retuning changes the high-redshift accretion rate by a factor of 5.25 relative to the L24 values. The paper does not report a sensitivity analysis of the evolution claim to these parameters, so it remains possible that a substantial part of the predicted 1 dex offset is an artifact of this manual adjustment rather than a robust prediction of the model. I request a sensitivity test, such as running the L24 parameter values on the PM trees or varying fSMBH and eSB by factors of order 2, and a discussion of how the z=9-to-z=0 offset responds.","section":"Section 2.2, Table 2, Eqs. (2)-(3), Section 5"},{"comment":"The abstract and conclusions state that galaxy morphology 'alone' explains most of the scatter around both scaling relations, but the evidence presented is random forest feature importance, which is a relative, model-dependent quantity and does not measure the variance in the scatter explained by morphology by itself. Because B/T is correlated with other features (M_bulge, Delta_SFMS, f_M_mergers), the Gini-importance ranking can overstate the unique role of morphology. To support the claim, the authors should quantify the predictive power of a regressor trained on B/T alone (e.g., R^2 or explained variance of the offset) and compare it with the full model, or use permutation importance; they should also phrase the conclusion in terms of predictive association rather than causation.","section":"Section 4.1 and Figure 11"}],"minor_comments":[{"comment":"'as been widely interpreted' should be 'has been widely interpreted'.","section":"Introduction, first paragraph"},{"comment":"The names SBHMR and SBHMRbulge appear to be swapped, since as written SBHMR uses the M_star,bulge-M_BH fit and SBHMRbulge uses the M_star-M_BH fit, the reverse of the convention in Figure 11.","section":"Appendix C, Eqs. (C4)-(C5)"},{"comment":"'Inferred SMBH occupation fraction at z~0 from observations show are shown' is grammatically incorrect; it should read 'are shown'.","section":"Figure 2 caption"},{"comment":"The symbols vSMBH and vvir are used but not defined in the text; please define them explicitly (vSMBH is later mentioned as a free parameter, and vvir is presumably the virial velocity).","section":"Equation (2)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the journal scope well and the main technical concern (missing sensitivity analysis) is addressable within the current framework. I would be comfortable with a major revision that adds the sensitivity test and tones down the causal language on morphology."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead this one. Bravo et al. run SHARK v2.0 on the Planck-Millennium simulation and predict the M_BH-M_star and M_BH-M_bulge relations from z=9 to z=0. The headline—SMBH masses ~1 dex lower at z=0 than at z=9 at fixed stellar mass—is a genuine model prediction, not a fit. It is also the first SHARK run on a volume this large (~1.7e7 galaxies at z=0), so the scatter and environment statistics are solid.\n\nThe paper is methodical. Calibration table is explicit, code is open (GitHub + Zenodo), and the comparison set is broad: BHMF, both scaling relations, morphology and central/satellite splits, against many observational datasets. The random forest feature importances are a useful way to rank scatter drivers, and the comparison to other hydro simulations and SAMs puts things in context.\n\nThe soft spot is exactly the one you flagged. The z=0 M_bulge-M_BH normalisation is a secondary calibration target, so the good local agreement is partly built in. The z=9 end of the 1 dex offset is anchored by the cold-gas accretion parameters fSMBH and eSB, manually re-tuned from 0.01->0.007 and 15->2 to fix PM-vs-SURFS offsets. Since cold gas accretion dominates at z>~6, the early normalisation is set by the product fSMBH/eSB, which changed by ~5x. No sensitivity analysis is shown. That doesn't make the claim wrong—some high-z data do sit where the model puts them—but the headline number is not robust to the manual re-tuning. A quick parameter variation would settle it.\n\nThe random forest conclusion that morphology 'explains' most of the scatter is correlational, and the paper does slide into causal language ('driven by') in the conclusions. Minor overstatement, not fatal.\n\nBottom line: it deserves peer review. It is a clear, reproducible, testable SAM prediction that will be used as a benchmark for 4MOST and JWST. Ask the authors for a sensitivity check on fSMBH/eSB and to soften the causality; neither is a desk-reject issue. I'd take it for reading group and cite it.","headline":"A solid, open, well-comparisoned SAM prediction paper whose headline 1 dex evolution claim rests on manually re-tuned cold-gas accretion parameters with no sensitivity analysis.","tokens_in":30790,"tokens_out":4300,"would_cite":true,"duration_ms":44895,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Using SHARK v2.0, a semi-analytic model of galaxy formation, this paper predicts that supermassive black holes at z=9 were about ten times more massive than black holes in galaxies of the same stellar mass today, and that galaxy…","keywords":["supermassive black holes","galaxy scaling relations","semi-analytic galaxy formation models","black hole mass function","galaxy morphology","AGN feedback","cosmic evolution","galaxy environment"],"falsifier":"Measure the M_star-M_BH and M_bulge-M_BH medians at z~4-6 with a large, selection-function-corrected sample (e.g., JWST and 4MOST); the paper's central claim would be contradicted if the median offset from local relations is much smaller than ~1 dex at fixed stellar mass once Lauer-type bias is corrected. A cheaper model-side falsifier is to rerun SHARK v2.0 with fSMBH varied over, say, 0.003-0.015 and eSB varied over 1-10; if the z=9-to-z=0 offset changes by more than ~0.5 dex, the evolution claim is an artifact of the manual calibration.","tokens_in":29739,"feed_emoji":"🕳️","tokens_out":8579,"duration_ms":86845,"temperature":0.7,"pith_summary":"SHARK v2.0, a semi-analytic model of galaxy formation run on a large-volume dark-matter simulation, is used here to predict how supermassive black hole masses relate to host stellar mass and bulge mass from redshift 9 to the present. The authors find the model reproduces observed scaling relations across a wide redshift range while predicting that the relations evolve strongly: at fixed stellar mass, SMBH masses are roughly 1 dex (a factor of ten) lower at z=0 than at z=9, and the scatter around the relations grows by a factor of 2-5 toward low redshift. They also find that the bulge-to-total stellar mass ratio, a proxy for morphology, is the dominant predictor of the scatter around both relations, with environment and star formation playing weaker, mostly indirect roles. The paper presents these as large-volume predictions that upcoming AGN and redshift surveys can test. The z=0 normalization of the bulge-mass relation was used as a secondary calibration target, so the slope and the cosmic-time evolution are the genuine predictions.","feed_headline":"Early black holes were 10x heavier for their galaxy size","feed_subtitle":"A galaxy-formation model predicts the black hole–galaxy mass relations evolve strongly from z=9 to today, not constant.","key_machinery":"SHARK v2.0's SMBH growth module: seeded black holes grow by Bondi-like hot gas accretion (Equation 1), by episodic cold gas accretion during starbursts triggered by mergers or disc instabilities (Equations 2-3, with free parameters fSMBH=0.007 and eSB=2.0 manually re-tuned for the Planck-Millennium run), and by SMBH-SMBH mergers. These channels set the normalization, slope, and scatter of the scaling relations; the cold-gas channel dominates high-redshift growth and sets the local M_bulge-M_BH normalization, while the merger channel increasingly contributes at low redshift. The random forest regression applied to offsets from the fitted relations is the tool that attributes the scatter primarily to morphology.","core_discovery":"The central discovery is a set of model predictions: the M_bulge-M_BH and M_star-M_BH relations in SHARK v2.0 are not universal but shift downward by about 1 dex in SMBH mass from z=9 to z=0 at fixed stellar mass, with the scatter increasing by a factor of roughly 2-5. The model produces a tight M_bulge-M_BH relation at all redshifts, with early- and late-type galaxies following distinct relations that separate by z~4; late-type galaxies have more massive SMBHs at fixed bulge mass but lower masses at fixed total stellar mass, and SMBHs are most massive in early-type galaxies. SMBH growth transitions from gas-accretion-dominated at high redshift to merger-dominated at low redshift, with a minimum in merger fraction around M_BH~$10^{7}$ Msun and M_bulge~$10^{10}$ Msun that does not evolve. Random forest regressions identify bulge-to-total ratio as the dominant feature explaining offsets from both relations at all redshifts, with central/satellite status having minor importance, implying environment acts indirectly through morphology.","pith_inferences":["Extension: If the 1 dex offset is physical, local SMBH mass predictors calibrated at z=0 will systematically underestimate high-redshift SMBH masses; a direct test is to apply the local relations to JWST-discovered z>4 galaxies and compare with independent estimates such as virial or reverberation mapping.","Extension: The dominance of morphology in the random forest suggests that merger history (wet versus dry) may be the underlying variable coupling SMBH growth to other galaxy properties; a controlled model experiment that shuffles merger histories at fixed morphology could isolate this.","Extension: The model's predicted satellite-central offset below ~10^10 Msun could be probed with upcoming 4MOST AGN surveys overlapping deep environmental metrics; detecting or ruling out the offset would constrain the role of ram-pressure and tidal stripping in SMBH feeding."],"forward_implications":["At fixed stellar mass, high-redshift galaxies should show SMBH masses up to ~1 dex above the local median, so deep surveys at z>4 can distinguish evolving from universal relations.","Estimates of SMBH masses from host stellar mass alone will carry large scatter; adding bulge mass or bulge-to-total ratio should markedly improve the inference.","The scatter around both relations should grow from z=9 to z=0 by a factor of 2-5, with a stellar-mass-dependent peak around M_star~10^10 Msun at low redshift.","Low-mass satellite galaxies should host more massive SMBHs than centrals of the same stellar mass by z=0 (up to ~0.5 dex), but this difference should vanish once morphology is controlled for.","The switch from gas-accretion-dominated to merger-dominated SMBH growth around z~4 should be reflected in the demographics of merging and active galaxies."],"supporting_citations":[{"why":"Supplies SHARK v2.0, its SMBH growth and AGN feedback prescriptions, and the baseline L210N1536 calibration that this work adapts.","marker":"L24"},{"why":"Introduced SHARK v1.1 and the original SMBH seeding, hot/cold gas accretion, and merger prescriptions that v2.0 updates.","marker":"L18"},{"why":"Provides the Planck-Millennium dark-matter simulation and merger trees on which SHARK v2.0 is run here.","marker":"Baugh et al. 2019"},{"why":"Supplies the merger-tree artefact fixes applied to the Planck-Millennium trees, affecting satellite populations and merger-driven SMBH growth.","marker":"Chandro-Gómez et al. 2025"},{"why":"Phenomenological cold-gas accretion model (Equations 2-3) that sets the normalization and evolution of the M_bulge-M_BH relation.","marker":"Kauffmann & Haehnelt 2000"},{"why":"Provides the hot-gas/radio-mode AGN accretion model used in Equation (1) for continuous SMBH growth.","marker":"Croton et al. 2006"},{"why":"Observational morphology-split scaling relations used to compare and interpret the model's early/late-type predictions.","marker":"Graham & Sahu 2023a"},{"why":"Hydrodynamic simulation predictions for M_star-M_BH evolution used as the main model comparison.","marker":"Habouzit et al. 2021, 2022"}],"fun_headline_variants":["Black holes 10x lighter per galaxy at z=0 vs z=9 in SHARK","Model: black hole–galaxy scaling relations evolve, not constant","Morphology, not environment, drives black hole–galaxy scatter","SHARK: SMBH masses drop 1 dex at fixed galaxy mass from z=9 to now","From z=9 to today, black holes shrink 10x relative to galaxies"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ~1 dex cosmic evolution claim rests on the assumption that the cold-gas accretion parameters fSMBH=0.007 and eSB=2.0, manually tuned to the local M_bulge-M_BH normalization, remain a valid extrapolation from z=0 to z=9, and the paper does not test how sensitive the evolution is to those choices.","fun_headline_variants_meta":{"raw":{"variants":["Black holes 10x lighter per galaxy at z=0 vs z=9 in SHARK","Model: black hole–galaxy scaling relations evolve, not constant","Morphology, not environment, drives black hole–galaxy scatter","SHARK: SMBH masses drop 1 dex at fixed galaxy mass from z=9 to now","From z=9 to today, black holes shrink 10x relative to galaxies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000253,"raw_usage":{"total_tokens":1636,"prompt_tokens":1086,"completion_tokens":550,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":702,"completion_tokens_details":{"reasoning_tokens":441}},"tokens_in":702,"tokens_out":550,"duration_ms":6294,"temperature":1.0,"reasoning_tokens":441,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:04:22.326193+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the M_star-M_BH and M_bulge-M_BH medians at z~4-6 with a large, selection-function-corrected sample (e.g., JWST and 4MOST); the paper's central claim would be contradicted if the median offset from local relations is much smaller than ~1 dex at fixed stellar mass once Lauer-type bias is corrected. A cheaper model-side falsifier is to rerun SHARK v2.0 with fSMBH varied over, say, 0.003-0.015 and eSB varied over 1-10; if the z=9-to-z=0 offset changes by more than ~0.5 dex, the evolution claim is an artifact of the manual calibration.","supporting_citations":[],"review_version":1}