{"id":"dfe4803a-449b-40be-bd0e-8263b29b9005","arxiv_id":"2506.09184","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Adding subgrid dynamical friction to BRAHMA simulations lowers predicted z>5 seed-black-hole merger rates by 4-10x relative to repositioning, to 100-1000 per year in the 9 Mpc boxes.","lead":"This paper simulates the motions of small black hole seeds in the early universe and shows that a friction force is needed to make them sink together and merge. It predicts how often such mergers might be seen by the future LISA gravitational wave mission.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fiducial merger-rate predictions rest on an uncalibrated dynamical seed mass (M_dyn_seed = 24 M_seed); this value is below the M23 model's recommended validity range, and Figure 14 shows rates drop by roughly a factor of 3 when it is lowered, so the 100-1000 yr^-1 headline is not yet robust.","rationale":"The paper's most defensible qualitative result is that BH repositioning overestimates prompt high-redshift seed mergers and that a subgrid DF treatment changes the dynamics: the controlled BRAHMA-4.5 comparisons in Figure 11 show DISCRETE_DF rates roughly an order of magnitude above ALL_NATURAL when gravitational boundedness is required, and independent work (Chen et al. 2022) found a similar suppression factor. That part of the central claim survives scrutiny. The load-bearing weak point is the conversion of this comparison into absolute z>5 merger rates and LISA-relevant numbers. The fiducial BRAHMA-9 rates use M_dyn_seed = 24 M_seed, an artificial enhancement whose physical meaning is only hand-waved as nuclear-star-cluster embedding. The M23 estimator is applied with an effective BH mass only about 2.2 times the DM particle mass, below the regime M23 itself recommends, and the paper supplies no independent calibration of this choice. Figure 14 demonstrates the sensitivity: lowering M_dyn_seed from 64 to 3 M_seed reduces z>5 merger counts by a factor of about 3. Thus the absolute 100-1000 yr^-1 range, and to a lesser degree the exact factor-4-10 suppression, are conditional on a parameter that is not yet physically anchored. The paper is unusually honest about this in its caveats, and it does frame the rates as upper limits, but the abstract and summary present the numbers without those qualifications. A resolved calibration test would settle whether 24 M_seed is physically motivated or merely a tuning choice. Until then the reader's CONDITIONAL verdict is appropriate, and the same weakest assumption is the correct focus.","tokens_in":34140,"tokens_out":8965,"duration_ms":102452,"concrete_test":"Calibrate M_dyn_seed with resolved idealized simulations: re-simulate roughly 20 representative seed-hosting halos from the BRAHMA-4.5 boxes (M_halo ~ 10^6-10^8 M_sun, z ~ 10-15) with gas and DM mass resolution m_p <= 200 M_sun << M_seed and softening <= 10 pc, so dynamical friction on the actual 2.2e3 M_sun seed is explicitly resolved, and measure sinking timescales for the same orbits as in the default runs. Then inject those same orbits into an M23-equivalent subgrid run at default BRAHMA resolution and find the M_dyn_seed value whose sinking time matches the resolved result. If the inferred value is below 24 M_seed, recompute the Section 3.4.4 rates; if the z>5 rates drop below the abstract's 100-1000 yr^-1 range, the central quantitative claim needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative headline (z>5 rates of 100-1000 yr^-1 and the factor-4-10 suppression) is controlled by the ad hoc dynamical seed mass introduced in Section 2.2.2. The M23 DF estimator in Eqs. (4)-(5) is evaluated with M_bh = M_dyn_seed; the fiducial BRAHMA-9 predictions in Section 3.4.4 use M_dyn_seed = 24 M_seed, about 2.2 times the DM particle mass in the default boxes, even though M23 recommends BH masses roughly 10 times the background particle mass. The actual seed mass is 2.2e3 M_sun, i.e., below the DM mass resolution, so the estimator is being used outside its calibrated regime. The paper's sensitivity test, Figure 14, shows the importance of this choice: at z>5 the DISCRETE_DF boxes yield 1223, 663, and 403 mergers for M_dyn_seed = 64, 8, and 3 times M_seed. The interpretation that M_dyn_seed represents an embedded nuclear star cluster is not modeled; an extended, tidally stripped cluster would not exert the same point-mass dynamical friction as a 24 M_seed BH. Since the absolute LISA-relevant rates and the precise suppression factor are extracted from this parameter, the central rate claim is not yet established independently of the tuning.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a new suite of BRAHMA cosmological hydrodynamic simulations that replace the previous BH-repositioning scheme with the Ma et al. (2023) subgrid dynamical-friction model. It studies the dynamics of ~2e3 M_sun BH seeds in (4.5 Mpc)^3 and (9 Mpc)^3 boxes at z>5, comparing three dynamics treatments (repositioning, subgrid DF, and purely natural dynamics), three merger criteria, four seed models, and dynamical seed masses ranging from 3 to 64 M_seed. The main findings are that subgrid DF is required for seeds to sink to halo centers and become gravitationally bound, that post-halo-merger BH merger timescales are roughly 50-1000 Myr, that z>5 merger rates are reduced by a factor of about 4-10 relative to the repositioning scheme, and that the different seed models predict merger rates of about 100-1000 yr^-1 at z~5, with implications for LISA event rates.","tokens_in":34416,"tokens_out":8582,"duration_ms":90779,"significance":"This is a valuable and timely study: it addresses a known limitation of the original BRAHMA predictions and provides a well-controlled comparison of BH dynamics treatments while holding seeding and galaxy physics fixed. The qualitative conclusion that BH repositioning overestimates prompt high-redshift seed mergers is well supported by the internal comparisons, and the relative suppression factor of roughly 4-10 is robust across the tested dynamical seed masses. The study also benefits from systematic tests of merger criteria and seed models, and it is transparent about many caveats. The absolute LISA-relevant rates, however, are conditioned on an ad hoc enhanced dynamical seed mass and should be read as upper limits with a substantial, only partially quantified systematic uncertainty.","major_comments":[{"comment":"The headline absolute rates in the abstract and in Figure 15 are not independent of the uncalibrated enhanced dynamical mass M_dyn_seed. In the default BRAHMA-9 boxes, M_dyn_seed = 24 M_seed corresponds to roughly 2.2 dark-matter particle masses, while the M23 model is stated to work best when the BH mass is about 10 times the background particle mass; the actual seed mass is 2.2e3 M_sun, below the DM particle mass resolution. Figure 14 shows that changing M_dyn_seed from 64 to 3 M_seed alters the z>5 merger counts by roughly a factor of three (1223 versus 403), and the fiducial BRAHMA-9 value lies in the middle of this range. Because the quoted 100-1000 yr^-1 rates are extracted from this parameter, the manuscript should either calibrate M_dyn_seed with an idealized sinking test or present the rates as a conditional band with the M_dyn_seed dependence made explicit. The relative suppression factor of 4-10 relative to repositioning is, by contrast, robust across the tested range and is not affected by this concern.","section":"§2.2.2, §3.4.3, §3.4.4 (Eqs. 4–5, Fig. 14, Fig. 15)"},{"comment":"The merger rates are plotted as events per year and are quoted without an explicit volume normalization or uncertainty budget. Because the production volumes are only (9 Mpc)^3 and Section 4.3 itself quotes a cosmic-variance factor of about two, the range 100-1000 yr^-1 across seed models cannot be taken as a precise prediction. I ask for error bands or at minimum an explicit statement in the figure caption and abstract that the rates carry at least this systematic uncertainty.","section":"§3.4.4, §4.3, Fig. 15"},{"comment":"The paper suggests that the enhanced dynamical seed mass may physically correspond to seeds embedded in nuclear star clusters, but no nuclear star cluster model is implemented. An extended, tidally stripped cluster does not necessarily exert the same point-mass dynamical friction as a more massive point particle, so this interpretation should be labeled as speculative rather than as a modeled process. This matters because the absolute rates are sensitive to the dynamical mass, as noted in the first major comment.","section":"§4.2, Summary bullet 5"}],"minor_comments":[{"comment":"The phrase 'merger rates between ~100-1000 events per year at z>5' is ambiguous: in the text the rates are evaluated at z~5. Please state explicitly whether the quoted number is a differential rate at z~5 or a cumulative rate integrated over z>5.","section":"Abstract, §3.4.4"},{"comment":"The left panels of Figure 14 show cumulative merger counts (1223, 663, and 403) while the text refers to them as 'merge rates'; please clarify the axes, the normalization, and the relation to the rates in Figure 15.","section":"Fig. 14, §3.4.3"},{"comment":"The BH mass in Eq. (4) is written as M_bh, but the calculations use M_dyn_seed; this substitution should be stated explicitly where the formula is introduced.","section":"§2.2.1, Eq. (4)"},{"comment":"The naming convention is inconsistent: the text and Figure 1 use both DF_DISCRETE and DISCRETE_DF for the same dynamics model; please standardize.","section":"Throughout"},{"comment":"There are several typographical errors, including 'consequentual' in Section 2, 'uncertainity' and 'scattening' in Section 4.3, and 'Thompson scattering' for 'Thomson scattering' near Eq. (3).","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for ApJ and the core qualitative result is solid. The main risk is the presentation of the absolute rates as robust predictions when they are strongly conditioned on M_dyn_seed. I would support publication after the rates are reframed as conditional upper limits with an explicit systematic band, ideally supplemented by an idealized sinking test. No concerns about novelty or attribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful new thing here is the combination: gas-based low-mass BRAHMA seeds plus the M23 subgrid dynamical friction model, with a systematic comparison across dynamics treatments and merger criteria. The controlled tests are done properly—multiple boxes, varied seed models, different merger criteria, and an honest caveats section. The central qualitative result is convincing: without subgrid DF, low-mass seeds do not sink and merge efficiently; with it, many do, but far fewer than under repositioning. The factor-of-4–10 suppression relative to repositioning at z>5 is a genuine methodological improvement over earlier BRAHMA papers, and the comparison to Chen et al. (2022) is fair and useful.\n\nThe soft spot is exactly where the stress-test note lands. The absolute rates (100–1000 yr^-1) and the precise suppression factor depend strongly on M_dyn_seed, an ad hoc enhanced dynamical mass introduced to avoid numerical heating. The paper is transparent about this—Section 2.2.2 and Figure 14 show the sensitivity—but the abstract presents the rates without that caveat, and the chosen fiducial value (24 M_seed) is below the M23 model's recommended range. Changing it from 64 to 3 M_seed changes z>5 rates by roughly a factor of three, so the headline number is not yet robust. This is a real limitation, though not a fatal one: the relative comparison between dynamics treatments is far less sensitive, and the qualitative conclusions about wandering and merger timescales hold across the tested values.\n\nOther weaknesses are minor in comparison: the quoted rates are per simulation volume with no error bars, cosmic variance is acknowledged but not quantified beyond a factor of ~2, and no code or data are released. The paper also overstates things slightly by calling the rates “LISA event rates” when they are really pair-formation rates at 0.2–0.4 kpc; the authors do note this in the caveats, so it is a presentation issue rather than a conceptual error.\n\nWho is this for? People building LISA event-rate predictions, especially those working with light-seed models, and simulators interested in subgrid DF prescriptions. It is a solid, useful contribution that deserves a serious referee. I would not desk-reject it. My recommendation: send to review, and ask the authors to either calibrate or marginalize over M_dyn_seed, quote volume-normalized rates with uncertainties, and soften the abstract to reflect that the rates are upper limits dependent on that parameter.","headline":"First BRAHMA light-seed runs with subgrid DF show repositioning overestimates z>5 merger rates; the qualitative conclusion holds, but the headline 100–1000 yr^-1 rates rest on a hand-tuned dynamical seed mass and should be treated as conditional.","tokens_in":35038,"tokens_out":1588,"would_cite":true,"duration_ms":19797,"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":"Subgrid dynamical friction cuts early black hole seed merger rates by a factor of 4-10, the paper contends.","keywords":["black hole seeds","dynamical friction","cosmological hydrodynamic simulations","high-redshift mergers","LISA gravitational wave rates","intermediate-mass black holes","BRAHMA simulations","merger-driven black hole growth"],"falsifier":"Run the same [9 Mpc]^3 boxes at a resolution where dark-matter particles are lighter than the seed mass and use the true seed mass in the gravity calculation; if the z>5 merger rate then falls below roughly 10 per year, the artificially enhanced dynamical mass is driving the predicted rates rather than real physics.","tokens_in":1658,"feed_emoji":"🕳️","tokens_out":1524,"duration_ms":56358,"temperature":0.7,"pith_summary":"This paper argues that unresolved dynamical friction governs whether low-mass black hole seeds sink to the centers of their small high-redshift halos and merge. Using a subgrid dynamical-friction prescription in the BRAHMA cosmological simulations, it finds that seeds do sink and merge on timescales of roughly 100-1000 megayears after their host halos merge, yielding z>5 merger rates of 100-1000 events per year across four seeding models. Those rates are a factor of 4-10 lower than the same simulations with black holes artificially pinned to halo centers. The paper concludes that merger-driven growth assembles these seeds to about $10^{4}$-$10^{5}$ solar masses by z~5, with gas accretion contributing less than 2 percent of the mass, and that LISA can use these events to distinguish seeding mechanisms.","feed_headline":"Pinning black holes to halo centers overstates merger rates 4-10x","feed_subtitle":"With unresolved dynamical friction added, seed mergers at 100-1000 per year at z>5 reshape LISA forecasts.","key_machinery":"The load-bearing mechanism is the M23 subgrid dynamical-friction estimator, a momentum-conserving sum over background mass elements (Eq. 5) that requires no free parameters and reduces to the Chandrasekhar formula under idealized conditions, applied together with artificially enhanced dynamical seed masses (3-64 times the true seed mass) to suppress numerical heating, and the MERGE_SOFTENING_BOUND merger criterion, which requires pairs to be within two gravitational softening lengths and to be gravitationally bound. This combination determines how far seeds sink, how long binary inspiral takes, and which pairs are counted as mergers.","core_discovery":"The central discovery is that the way a simulation treats black hole dynamics below the resolution scale changes the predicted early Universe merger history by an order of magnitude. Adding the Ma et al. 2023 subgrid dynamical-friction estimator lets a significant fraction of ~2.$2x10^{3}$ solar-mass seeds sink to subhalo centers by z~5 and become gravitationally bound pairs at separations of ~0.2-0.4 kpc, where they can merge. Removing both repositioning and subgrid friction leaves most seeds wandering and mergers rare, while the old repositioning scheme merges seeds promptly at kiloparsec separations and overproduces high-redshift events. Across four seed models, the authors find z>5 merger rates of about 100-1000 per year, with peak rates shifting to lower redshift than under repositioning, and these mergers dominate seed growth while gas accretion remains negligible below roughly $10^{5}$ solar masses.","pith_inferences":["If sub-softening dynamical friction contributes significantly, the true sinking efficiency may be higher than modeled here, pushing the rates toward the paper's upper end rather than the lower end.","The enhanced dynamical seed mass can be read physically as a nuclear star cluster; if such clusters are rare around z>5 seeds, the simulated merger rates would drop, which is a testable assumption for future small-scale simulations.","A LISA non-detection at the predicted rate would disfavor the most lenient seeding model or require that most 0.2-0.4 kpc pairs fail to reach gravitational-wave coalescence.","Applying the same dynamics treatment to heavy direct-collapse seeds would show whether the 4-10 x suppression factor persists across seed channels, which would sharpen the seeding discrimination power of LISA."],"forward_implications":["If the paper is right, the standard practice of pinning black holes to potential minima overestimates z>5 seed merger rates by factors of 4-10, so LISA event-rate forecasts must include unresolved dynamical friction.","Under subgrid dynamical friction, different seed models predict 100-1000 seed mergers per year at z>5, giving LISA a realistic chance of detecting seed-mass events during its lifetime.","Merger-driven growth builds seeds to 10^4-10^5 solar masses by z~5 with less than 2 percent of mass from gas accretion, implying steep mass functions in the intermediate-mass range that electromagnetic observatories cannot easily probe.","The paper treats its rates as upper limits to actual gravitational-wave events because inspiral below 0.2-0.4 kpc, loss-cone scattering, and gravitational-wave recoil are not modeled.","Because z>15 merger rates are suppressed by factors of 10-100 relative to repositioning, early-Universe seed mergers are much rarer than previously estimated but still numerous enough to be observable."],"supporting_citations":[{"why":"Supplies the subgrid dynamical-friction estimator (Eqs. 4-5) that the paper adopts as its fiducial dynamics model.","marker":"Ma et al. 2023 (M23)"},{"why":"Provides the earlier BRAHMA-9 repositioning merger-rate predictions used as the comparison baseline for the factor-of-4-10 suppression.","marker":"Bhowmick et al. 2024b"},{"why":"Contributes tests showing that dynamical friction is not resolved even at high BH-to-DM mass ratios, justifying the decision to apply the subgrid estimator without a double-counting correction.","marker":"Genina et al. 2024"},{"why":"Presents an independent test of dynamics treatments and merger criteria whose roughly factor-6 suppression rate is consistent with the paper's results.","marker":"Chen et al. 2022 (C22)"},{"why":"Introduced the MERGE_SOFTENING_BOUND criterion (separation below twice the softening length plus gravitational boundedness) used as the fiducial merger condition.","marker":"Bellovary et al. 2011"},{"why":"Provides semi-analytic predictions with dynamical delays that the paper compares against, showing its higher predicted rates stem from more permissive seed models.","marker":"Ricarte & Natarajan 2018 (R18)"}],"fun_headline_variants":["Subgrid friction cuts early BH merger rate 4-10x","Black hole seeds wander: merger rate drops 10-fold","Dynamical friction reshapes z>5 black hole merger forecasts","Realistic BH dynamics slash high-z merger rates 10-fold","Subgrid friction dims LISA's early seed merger rates"],"cache_read_input_tokens":36992,"weakest_assumption_plain":"The subgrid dynamical-friction force, computed with seeds artificially made 24 times heavier for the force calculation, accurately captures the real drag that ~$10^{3}$ solar-mass seeds would feel in halos of ~$10^{6}$-$10^{7}$ solar masses.","fun_headline_variants_meta":{"raw":{"variants":["Subgrid friction cuts early BH merger rate 4-10x","Black hole seeds wander: merger rate drops 10-fold","Dynamical friction reshapes z>5 black hole merger forecasts","Realistic BH dynamics slash high-z merger rates 10-fold","Subgrid friction dims LISA's early seed merger rates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000513,"raw_usage":{"total_tokens":2612,"prompt_tokens":1183,"completion_tokens":1429,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":799,"completion_tokens_details":{"reasoning_tokens":1342}},"tokens_in":799,"tokens_out":1429,"duration_ms":10068,"temperature":1.0,"reasoning_tokens":1342,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:54:21.943061+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same [9 Mpc]^3 boxes at a resolution where dark-matter particles are lighter than the seed mass and use the true seed mass in the gravity calculation; if the z>5 merger rate then falls below roughly 10 per year, the artificially enhanced dynamical mass is driving the predicted rates rather than real physics.","supporting_citations":[],"review_version":1}