{"id":"f4dfc0f5-9b73-4758-8157-263de7109165","arxiv_id":"2411.19332","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Black holes of 1e5 to 1e6 solar masses in local dwarf galaxies still carry information about how the first black hole seeds formed, according to the BRAHMA simulations.","lead":"Simulations that vary how the first black hole seeds form show that the differences survive in low-mass black holes in nearby dwarf galaxies. Future X-ray surveys that count such black holes could therefore test competing seed-formation models.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim's second pillar, merger-dominated growth up to ~1e6 Msun at z=0, depends on an optimistic BH dynamics treatment that the paper itself identifies as unresolved and should be tested before using these predictions to constrain seed models.","rationale":"The reader's weakest_assumption is exactly the load-bearing concern I find: the merger-dominated growth that preserves seed signatures to z=0 assumes optimistic BH dynamics. The paper is honest about this in Section 4.2, but honesty does not remove the dependence. The central claim is not internally inconsistent, and the simulations usefully isolate seeding by holding other physics fixed, so I do not see grounds for rejection. However, the claim's second pillar is not independently supported: the simulation data are not public, no variant with alternative BH dynamics is presented, and the only quoted dynamical-friction-equipped simulations (ROMULUS, Bellovary et al.) predict lower occupation fractions, which is at least qualitatively consistent with the concern. A single controlled rerun with modified BH dynamics would settle whether the predicted 1e5-1e6 Msun signatures are robust or are upper limits. This is a testable, specific condition, so the appropriate verdict remains conditional rather than accept. I therefore agree with the reader's conditional assessment and recommend no change to the verdict.","tokens_in":32353,"tokens_out":4402,"duration_ms":44959,"concrete_test":"Run one [18 Mpc]^3 box, e.g., SM5_LW10 or ESD:STOCHASTIC, with identical seed prescriptions but replace instantaneous repositioning/prompt merging with a subgrid dynamical friction prescription (e.g., Tremmel et al. 2017) or a merger delay based on the Chandrasekhar timescale. Then recompute the z=0 BHMF at 1e5-1e6 Msun and the occupation fractions in M*~1e9 Msun galaxies. If the seed-model separations at 1e5-1e6 Msun shrink by more than a factor of 2, or the 1e6 Msun occupation fractions fall below the Burke et al. lower limits, the central claim must be reframed as an upper-limit prediction. If the separations and observational consistency survive, the concern is refuted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's central claim has two pillars: (1) ungrown ~1e5 Msun relics of z~5-10 seeds, and (2) merger-dominated growth from ~1e5 to ~1e6 Msun down to z=0. The second pillar is directly controlled by the BH dynamics implementation described in Section 2: BHs are repositioned to the nearest potential minimum and are '.promptly merged when at least one of them is within the neighbor search radius.' This imposes instantaneous dynamical friction and zero wandering. Section 4.2 explicitly concedes that this is the '.most optimistic scenario for merger-driven BH growth' and that subgrid dynamical friction treatments in other simulations produce substantial wandering BHs, while gravitational recoil can eject BHs from galaxies. If real merger timescales are longer than the time between halo mergers at these masses, the z=0 merger fraction in Figure 6 is an upper limit, not a prediction. The consequences are direct: either growth to ~1e6 Msun would shift to gas accretion, which erases seed signatures, or the ~1e6 Msun abundance and occupation fractions (0.01-0.05 Mpc^-3 and 20-100%) would drop below current observational constraints. Both outcomes weaken the claim that local ~1e5-1e6 Msun BHs can discriminate seed models. The first pillar, ungrown relics, is less affected by merger efficiency, but the abstract's combined ~1e5-1e6 Msun signature and the observational predictions for the 1e6 population rest on the optimistic merger assumption.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses five [18 Mpc]^3 BRAHMA cosmological simulations run to z=0 with identical galaxy formation physics but different BH seeding models to predict local BH abundances, mass functions, AGN luminosity functions, scaling relations, and occupation fractions. The four 'DHS' models occupy heavy seeds with progressively stacked formation criteria, while the 'ESD' model initializes 1.5e5 M_sun BHs as extrapolated descendants of ~1e3 M_sun seeds using a stochastic model calibrated to high-resolution simulations. The central claim is that strong seeding signatures survive in ~1e5-1e6 M_sun local BHs hosted in M* < 1e9 M_sun dwarf galaxies, because these BHs are either ungrown relics of z~5-10 seeds or grow mainly by mergers to z=0. The predictions are compared with independent observational constraints from Shen et al. (2020), Merloni & Heinz (2008), Schutte et al. (2019), Greene et al. (2020), and others.","tokens_in":32716,"tokens_out":6670,"duration_ms":57024,"significance":"If the central claim survives scrutiny, the paper is a valuable contribution: it makes falsifiable, quantitative predictions for a local dwarf-galaxy BH population that upcoming facilities (AXIS, Athena) can test, and its five-box design isolates seeding from other galaxy-formation physics. The comparisons use published constraints rather than fitting to local BH data, and the paper is transparent about many caveats. The main limitation is that the size of the predicted signatures is controlled by an optimistic BH dynamics treatment, and the single realization per model leaves unknown cosmic variance; these issues currently limit how strongly the paper can claim to discriminate seed models.","major_comments":[{"comment":"The claim that growth from ~1e5 to ~1e6 M_sun is merger-dominated to z=0 (Figure 6) is directly controlled by the repositioning/prompt-merging prescription, which the paper itself calls the 'most optimistic scenario' and acknowledges suppresses wandering and recoil. Because wandering or delayed mergers would shift growth to accretion (erasing seed signatures) or reduce the 1e6 M_sun abundance and occupation fractions, the z=0 merger fraction and the ~1e6 M_sun predictions (0.01-0.05 Mpc^-3; 20-100% occupation in Figure 9) are upper limits, not robust predictions. Please run or cite a sensitivity test with a sub-grid dynamical friction treatment, or explicitly re-frame the abstract's two-pillar claim as contingent on optimistic BH dynamics.","section":"Section 2 (BH dynamics) and Section 4.2 (Caveats)"},{"comment":"All quantitative predictions come from a single 18 Mpc box per model, so the quoted abundance ranges and occupation fractions carry no error bars. The paper's argument that the limited volume is not important uses agreement with TNG100 at >1e7 M_sun, but the low-mass dwarf regime that is the focus of the paper could be more sensitive to cosmic variance. Please provide quantitative estimates of sample variance (e.g., bootstrap sub-boxes, multiple realizations, or analytic cosmic-variance calculations) and state uncertainties on the z=0 BHMFs and occupation fractions.","section":"Sections 3.2-3.3 and start of Section 3"},{"comment":"The ESD model is calibrated using the high-resolution BRAHMA simulations of Bhowmick et al. (2024b), but the paper does not state whether those calibration runs employ the same BH repositioning and prompt-merger treatment as the present boxes. If they do, the ESD predictions inherit the same optimistic-dynamics assumption, and the 'broad agreement' of the ESD model with local observations (Section 4.1) cannot by itself establish the viability of ~1e3 M_sun seeds independent of BH dynamics. Please clarify this and, if the same treatment is used, soften the viability claim accordingly.","section":"Section 2.1.2 (ESD stochastic seed model)"}],"minor_comments":[{"comment":"The model name 'SM_LW10_LOWSPIN_SPIN' appears to be a typo; the intended model is likely SM5_LW10_LOWSPIN or SM5_LW10_LOWSPIN_RICH depending on the argument.","section":"Section 4.1"},{"comment":"The sentence 'Greene et al. (2020) inferred > 50% occupation fractions for ≳ 109 galaxies' is missing the stellar-mass unit; it should read 'for M* ≳ 10^9 M_sun galaxies'.","section":"Section 3.6.1"},{"comment":"The phrase 'the subsequent the complex mass assembly history' contains a duplicated article and should be corrected to 'the subsequent complex mass assembly history'.","section":"Section 1 (Introduction)"},{"comment":"The artificial y-axis offsets used for the z=3 and z=5 curves are not quantified in the caption; please state the offsets or label the axes so that the reader can compare the curves across redshifts.","section":"Figure 6"},{"comment":"The phrase 'a lenient threshold of 5 Mseed' is ambiguous because Mseed is a mass; please clarify explicitly that this means 5 times the seed mass (7.5e5 M_sun) of dense metal-poor gas.","section":"Section 2.1.1"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about BH dynamics is legitimate and is the main reason for major revision. The paper is otherwise well-structured and the differential design is a strength; I do not see grounds for rejection, as the issues are fixable with caveats and sensitivity tests."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What's new here is the z=0 end of the BRAHMA suite: five 18 Mpc boxes that share everything except the BH seeding prescription. That clean isolation is the paper's main strength. Holding galaxy formation, accretion, and feedback fixed while varying only seed formation lets the authors attribute differences to seeding, and they use it to produce concrete local predictions: BH mass functions, number densities, AGN luminosity functions, M*–Mbh relations, and occupation fractions for 1e5–1e7 Msun BHs. The earlier BRAHMA papers focused on z>=5; the local analysis is new.\n\nThe headline claim—that strong seeding signatures survive in 1e5–1e6 Msun local BHs in dwarfs—rests on two mechanisms: ungrown relics of z~5–10 seeds, and merger-dominated growth up to ~1e6. The first is robust. The second depends on the BH dynamics treatment: repositioning to potential minima and prompt merging within the search radius. The paper says in Section 4.2 that this is the most optimistic scenario and that the occupation fractions are upper limits. That caveat is honest and well placed, but it means the quantitative predictions for the 1e6 population—0.01–0.05 Mpc^-3 and 20–100% occupation—are upper limits, not central predictions. If wandering, dynamical friction delays, or recoil matter, those numbers drop and the discriminating power weakens. A sensitivity test (a run with subgrid dynamical friction, or an analytic correction) would materially strengthen the claim. The 1e5 relics are less affected, so the qualitative conclusion—dwarfs retain seed information—survives.\n\nOther soft spots, in proportion: one realization per model in 18 Mpc boxes, so no error bars on the mass functions; the ESD low-mass seed model is calibrated to the authors' own high-res simulations, which is reasonable but not independent; and the faint-end AGN LFs sit a factor of ~10 above observations, a known issue they defer. The comparisons to independent observed constraints (Shen, Merloni & Heinz, Schutte, Greene, Burke) are fair and mostly even-handed.\n\nWho it's for: observers planning AXIS/Athena searches for IMBHs in dwarfs, and simulators working on seeding. It deserves a serious referee; the main fix is to frame the 1e6 predictions as optimistic upper limits, which the authors already mostly do. Send it to review.","headline":"Controlled BRAHMA seeding comparison makes a plausible but dynamics-dependent case that local dwarf BHs can discriminate seed models.","tokens_in":33266,"tokens_out":2072,"would_cite":true,"duration_ms":18662,"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":"The imprint of black hole seeding survives in local low-mass black holes inside dwarf galaxies, according to five simulations that differ only in how the seeds form.","keywords":["black hole seeding","intermediate-mass black holes","dwarf galaxies","cosmological simulations","black hole mergers","black hole occupation fraction","black hole mass function","direct collapse black holes"],"falsifier":"A complete local census of dwarf galaxies that measures the occupation fraction of $\\gtrsim10^6\\,M_\\odot$ black holes in $M_*\\sim10^9\\,M_\\odot$ galaxies would settle it: if the measured fraction falls below the lowest predicted value of about $20\\%$ and the local black hole mass function shows no steepening at $10^5$–$10^6\\,M_\\odot$, the predicted population of ungrown relics and merger-grown low-mass black holes would be absent.","tokens_in":32167,"feed_emoji":"🕳️","tokens_out":19149,"duration_ms":140730,"temperature":0.7,"pith_summary":"The paper tries to establish that the very first seeds of supermassive black holes leave measurable traces in the present-day Universe, specifically in black holes of roughly $10^5$ to $10^6\\,M_\\odot$ hosted by dwarf galaxies with stellar masses below $10^9\\,M_\\odot$. Five simulations of an $18\\,\\mathrm{Mpc}$ cube, identical except for the black hole seeding prescription, produce a local population of low-mass black holes that are either ungrown relics of seeds formed at $z\\sim5$–$10$ or systems whose growth to $\\sim10^6\\,M_\\odot$ is dominated by mergers all the way to $z=0$. Because gas accretion, which erases seed-model differences, only becomes the main growth channel above $\\sim10^7\\,M_\\odot$, the seeding signatures survive exactly in the mass range that local dwarf-galaxy surveys can target. If the prediction holds, counting these black holes and measuring how often dwarf galaxies host them can discriminate between competing seeding models.","feed_headline":"Black hole seeding imprints survive to today in dwarf galaxies","feed_subtitle":"Low-mass black holes in nearby dwarf galaxies could reveal how the first supermassive black holes formed","key_machinery":"The carrying mechanism is the split between two growth channels: black hole mergers versus gas accretion. The paper isolates this split by comparing five simulation boxes that share every aspect of the galaxy formation model except seeding: four 'direct heavy seed' models, which plant $1.5\\times10^5\\,M_\\odot$ black holes in gas that is dense and metal-poor and then progressively add conditions (Lyman-Werner radiation, low gas spin, rich environment), and one 'extrapolated seed descendant' model, which plants the same initial mass as a stand-in for the higher-mass descendants of unresolved $\\sim10^3\\,M_\\odot$ seeds. The merging fraction of the accumulated mass stays near unity for $10^5$–$10^6\\,M_\\odot$ black holes down to $z=0$, and a substantial population of seeds formed at $z\\sim5$–$10$ never grows at all; those two facts are what keep seeding differences visible in the local low-mass population.","core_discovery":"On the paper's own terms, strong signatures of seeding survive in local $\\sim10^5$–$10^6\\,M_\\odot$ black holes found in $M_*\\lesssim10^9\\,M_\\odot$ dwarf galaxies. Two effects preserve the signal: many of the smallest local black holes are seeds from $z\\sim5$–$10$ that have grown by less than about ten percent, and growth from $10^5$ to $10^6\\,M_\\odot$ is driven by black hole mergers rather than gas accretion all the way down to $z=0$. The seed models predict number densities of $\\gtrsim10^5\\,M_\\odot$ black holes spanning $0.02$–$0.4\\,\\mathrm{Mpc}^{-3}$, number densities of $\\gtrsim10^6\\,M_\\odot$ black holes of $0.01$–$0.05\\,\\mathrm{Mpc}^{-3}$, and occupation fractions of $\\sim20$–$100\\%$ in $M_*\\sim10^9\\,M_\\odot$ galaxies. The variations weaken for more massive black holes and disappear above $\\sim10^7\\,M_\\odot$, where accretion dominates; at $z\\gtrsim5$, by contrast, the signatures persist up to $\\sim10^8\\,M_\\odot$ because growth there is fully merger-dominated.","pith_inferences":["If the simulations' merger assumption overstates how often low-mass black holes coalesce, the predicted occupation fractions are upper limits, so a null detection in dwarf galaxies would tighten models of black hole dynamics as much as seeding.","The predicted steepening below $10^6\\,M_\\odot$ implies that extrapolating scaling-relation-based black hole mass functions into the dwarf regime could miss an entire population; targeted dwarf-galaxy surveys are the direct test.","Because the seed models differ most in the number of low-mass black holes available to merge, gravitational-wave event rates from $\\sim10^5$–$10^6\\,M_\\odot$ black hole mergers should differ across models, providing an accretion-independent probe of seeding.","Both the low-mass-seed model and the intermediate heavy-seed models fit current data, so distinguishing them will likely require combining local occupation measurements with high-redshift number densities rather than relying on either epoch alone."],"forward_implications":["Number densities of $\\gtrsim10^5\\,M_\\odot$ local black holes are predicted to span $0.02$–$0.4\\,\\mathrm{Mpc}^{-3}$ across seed models, so a census at the low-mass end can discriminate seeding scenarios.","Occupation fractions in $M_*\\sim10^9\\,M_\\odot$ galaxies range from $\\sim40$–$100\\%$ for $\\gtrsim10^5\\,M_\\odot$ black holes and $\\sim20$–$100\\%$ for $\\gtrsim10^6\\,M_\\odot$ black holes, giving a direct observable test.","The $z=0$ black hole mass function should be steeper at $10^5$–$10^6\\,M_\\odot$ than a Schechter extrapolation from higher masses; surveys that assume such an extrapolation will underestimate low-mass black hole abundances.","Faint local AGN with $L_{2-10\\,\\mathrm{keV}}\\lesssim10^{39}\\,\\mathrm{erg\\,s^{-1}}$, accessible to proposed X-ray instruments, are predicted to show strong seed-model variations, unlike high-redshift AGN luminosity functions.","The models that match observed high-redshift overmassive black hole scaling relations also match local black hole mass functions and dwarf-galaxy scaling relations, bracketing the viable range of seed formation efficiencies."],"supporting_citations":[{"why":"Builds the stochastic seed model that places 1.5e5 Msun BHs as extrapolated descendants of unresolved ~1e3 Msun seeds, defining the ESD box.","marker":"Bhowmick et al. 2024a"},{"why":"Provides the high-resolution simulations that explicitly resolve ~1e3 Msun seeds and their merger-dominated growth, calibrating the ESD model.","marker":"Bhowmick et al. 2024b"},{"why":"Constructs and runs the four heavy-seed boxes used here and reports their z>=5 BH populations.","marker":"Bhowmick et al. 2024c"},{"why":"Supplies the dense, metal-poor, low-spin direct-collapse criteria that the heavy-seed model stack is designed to emulate.","marker":"Lodato & Natarajan 2006"},{"why":"Gives the local observed BH mass function and AGN luminosity function used as the main z=0 comparison.","marker":"Shen et al. 2020"},{"why":"Provides observed BH masses in dwarf galaxies used to test the simulated M*-Mbh relations.","marker":"Schutte et al. 2019"},{"why":"Gives multi-wavelength lower limits on local BH occupation fractions against which the restrictive seed models are compared.","marker":"Burke et al. 2024"}],"fun_headline_variants":["Dwarf galaxies preserve imprints of early black hole seeds","Local black holes in dwarfs carry fingerprints of seeding","Black hole seed imprints survive in local dwarf galaxies","Mergers keep early black hole seed signatures until today"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The prediction assumes that low-mass black holes that grow by mergers actually meet and merge: the simulations park every black hole at the bottom of its galaxy's gravitational potential well and merge black holes as soon as they come close, so if wandering, slow dynamical friction, or gravitational recoil keeps many of them apart, the predicted low-mass populations and occupation fractions would be smaller and the seeding signatures weaker.","fun_headline_variants_meta":{"raw":{"variants":["Dwarf galaxies preserve imprints of early black hole seeds","Local black holes in dwarfs carry fingerprints of seeding","Black hole seed imprints survive in local dwarf galaxies","Mergers keep early black hole seed signatures until today"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00112,"raw_usage":{"total_tokens":4837,"prompt_tokens":1298,"completion_tokens":3539,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":914,"completion_tokens_details":{"reasoning_tokens":3474}},"tokens_in":914,"tokens_out":3539,"duration_ms":22911,"temperature":1.0,"reasoning_tokens":3474,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T10:16:42.931336+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A complete local census of dwarf galaxies that measures the occupation fraction of $\\gtrsim10^6\\,M_\\odot$ black holes in $M_*\\sim10^9\\,M_\\odot$ galaxies would settle it: if the measured fraction falls below the lowest predicted value of about $20\\%$ and the local black hole mass function shows no steepening at $10^5$–$10^6\\,M_\\odot$, the predicted population of ungrown relics and merger-grown low-mass black holes would be absent.","supporting_citations":[],"review_version":1}