{"id":"451513ac-7fe7-4c1c-9287-d9eccdec8575","arxiv_id":"2510.13976","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"The ASTRID simulation's present-day universe broadly matches observed black-hole–galaxy scaling relations, galaxy counts, cluster stellar masses, and clustering, with known excesses in low-redshift star formation and accretion.","lead":"This paper presents the present-day (z=0) output of ASTRID, one of the largest cosmological galaxy-formation simulations, and compares its simulated black holes and galaxies to telescope observations. It is a reference resource for upcoming gravitational-wave and X-ray surveys, although some low-redshift star formation and black-hole accretion rates are too high.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The artificial dynamical mass M_dyn=1e7 h^-1 M_sun assigned to all seed BHs may dominate early BH dynamics, making the claimed M_BH-M*/sigma agreement and scatter vulnerable to this free parameter.","rationale":"The reader identified the same weakest assumption: the artificial M_dyn and stochastic seed masses may produce MBH dynamics that are artifacts rather than physical. My analysis agrees and sharpens the concern: M_dyn is 30-300 times the seed mass, so for all BHs below 1e7 M_sun the sinking timescale is set by an artificial mass, potentially erasing seed-mass diversity and biasing growth. This is directly load-bearing for the abstract's claims of 'good agreement' and 'more realistic MBH diversity'. The paper provides no sensitivity test for M_dyn, only validation that the DF model itself matches semi-analytics. The proposed zoom-in experiment would settle whether the z=0 scaling relations depend on M_dyn within a factor of 3; if they do, the central claim is not a robust prediction. Since this is a plausible but unproven risk that qualifies the headline claim, the reader's CONDITIONAL verdict remains appropriate. No change to the verdict is needed; the concern reinforces the conditional status rather than overturning it. I credit the paper's public data release and large volume as independent support, but the specific subgrid parameter sensitivity remains untested.","tokens_in":38262,"tokens_out":8870,"duration_ms":73443,"concrete_test":"Re-simulate a representative sample of ~20 ASTRID halos (M_200c = 1e11-1e13 M_sun) as zoom-in runs at the same resolution, with all physics and seed masses unchanged but M_dyn set to 3e5, 1e6, 1e7 (default), and 3e7 h^-1 M_sun. At z=0, measure the median and 16-84th percentile scatter of M_BH at fixed M* (e.g., M*=1e10-1e11 M_sun) and at fixed sigma. If the median shifts by >0.3 dex or the scatter changes by >0.2 dex across this range, the claimed agreement and diversity are not robust to the dynamical mass parameter, and the central claim would need to be substantially qualified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that ASTRID 'successfully captures the co-evolution of MBHs and their host galaxies' rests on the realism of the subgrid BH dynamics. In Section 2.1.1, seeds are drawn stochastically from 3e4-3e5 h^-1 M_sun, but Section 2.1.2 immediately assigns each seed a dynamical mass M_dyn=1e7 h^-1 M_sun, used for the gravitational force and the Tremmel/Chen dynamical friction until M_BH exceeds M_dyn. This makes every seed behave dynamically as a 1e7 M_sun object, 30-300 times heavier than its true mass. Consequently, seed orbital decay and the ability to reach galactic centers and accrete are controlled by an artificial constant rather than by the seeded mass. The claimed diversity in M_BH-M* and M_BH-sigma, and the scatter 'more consistent with observations than previous simulations', could therefore be an artifact of this choice: the effective early growth might be homogenized by the fixed M_dyn, with the stochastic seed draw playing only a subdominant role. The paper's own Section 6.2 concedes results are 'partly determined by our BH seeding prescription.' The validations cited (Chen et al. 2022b; Zhou et al. 2025b) test the DF prescription in idealized or limited setups, but do not quantify how z=0 scaling relations respond to the value of M_dyn. Since the co-evolution claim is the paper's headline, this unvaried free parameter is the most load-bearing uncertainty.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents the z=0 output of the ASTRID cosmological SPH simulation (370 Mpc box, ~0.33 trillion particles), focusing on the massive black hole (MBH) population and its connection to galaxies and large-scale structure. It describes the subgrid MBH model (seeding, dynamical mass M_dyn, dynamical friction, Bondi accretion, two-mode AGN feedback), then compares z=0 MBH mass functions, AGN X-ray luminosity functions, M_BH-M* and M_BH-sigma relations, BH occupation statistics, and BH mass/accretion-rate densities with observations. It also reports the galaxy stellar mass function, sSFR, sizes, metallicities, dust-attenuated luminosity function and g-r colors, cluster stellar content, and MBH/galaxy clustering bias. The paper concludes that ASTRID captures MBH-galaxy co-evolution, reproduces the galaxy luminosity function after dust attenuation, hosts a large cluster population, and provides useful large-scale-structure tracers. Several tensions are acknowledged, including an overestimated z=0 BH accretion-rate density, a low quiescent fraction, and a 0.5 dex deficit near the stellar mass function knee.","tokens_in":38644,"tokens_out":5163,"duration_ms":51064,"significance":"If the results are robust, ASTRID is a major public resource for MBH and gravitational-wave science: it combines a large volume, low seed masses, full MBH trajectories, and a >3 million merger catalog, with direct applicability to LISA, PTA, and X-ray survey planning. The paper is unusually transparent about known disagreements, and it provides publicly available catalogs. The main caveats are that the MBH scaling-relation and scatter claims depend on unvaried subgrid choices, especially M_dyn and the seed mass distribution, and that the dust-attenuated luminosity-function and color results are calibrated to the same SDSS luminosity functions used for the comparison. These issues do not negate the value of the simulation, but they are load-bearing for the paper's headline claims.","major_comments":[{"comment":"The dust attenuation normalization κ_ISM is set to 10^3.0 by calibrating against observed SDSS u, g, r luminosity functions (Driver et al. 2012; Loveday et al. 2012). The right panel of Fig. 15 then presents the dust-attenuated luminosity function as being in 'good agreement' with those same observations. As written, this is a fit, not a prediction. The g−r color bimodality in Fig. 16 inherits the same calibration and therefore cannot be cited as independent support. Please either reframe these results as a calibrated post-processing model, or provide an out-of-sample test (e.g., a different photometric band, a different redshift, or number counts) that would demonstrate predictive power.","section":"§4.3, Eq. (6), Fig. 15 (right), Fig. 16"},{"comment":"Seeds are drawn stochastically from 3×10^4–3×10^5 h^−1 M_sun, but each seed is assigned a dynamical mass M_dyn = 10^7 h^−1 M_sun that is used for the gravitational force and the dynamical friction calculation until M_BH exceeds M_dyn. Thus early BH orbital decay, merging, and accretion are controlled by a constant that is 30–300 times the true seed mass. The paper cites Chen et al. (2022b) and Zhou et al. (2025b) for validation of the dynamical friction prescription, but no test is presented for the sensitivity of z=0 scaling relations, scatter, or occupation numbers to the value of M_dyn or to the seed mass draw. Given that §6.2 concedes that the clustering results are 'partly determined by our BH seeding prescription,' the headline M_BH–M* and M_BH–σ claims require either a convergence/sensitivity test or an explicit statement of which observables are robust to this choice.","section":"§2.1.1–§2.1.2, §3.3, §6.2"},{"comment":"The claim that ASTRID's scatter in the M_BH–M* and M_BH–σ relations is 'more consistent with observations than previous simulations' is supported only by visual comparison with gray observational contours. No quantitative measure of scatter is provided, and the observational contours include heterogeneous samples with different selection functions and measurement errors. Please provide a quantitative comparison, for example the rms scatter in log M_BH at fixed M* or σ, with observational errors accounted for, to justify this specific claim.","section":"§3.3, Figs. 6–7"},{"comment":"The paper acknowledges a >1 dex overestimate of the z=0 BH accretion-rate density and a 0.5 dex deficit near the stellar mass function knee, as well as a too-low quiescent fraction at M* ≈ 5×10^10 M_sun. These are not fatal by themselves, but they involve the same AGN feedback mechanisms used to explain the M_BH–M*/σ relations. The conclusion that ASTRID 'successfully captures the co-evolution' of MBHs and galaxies would be more convincing if the paper explicitly discussed how these tensions are related to the claimed co-evolution success, or tempered the headline claim accordingly.","section":"§3.5, Fig. 10; §4.2, Fig. 15 (left)"}],"minor_comments":[{"comment":"Typo: 'istropically' should be 'isotropically'.","section":"§2.1.4"},{"comment":"Typographical issues: 'with with' near the halo description in §2, and 'which which' in §5.1.","section":"§2, §5.1"},{"comment":"'M_cric' in the sSFR discussion should be 'M_crit'.","section":"§4.1"},{"comment":"The MILES stellar library is cited as 'Miller et al. 2015', but the reference list entry Miller et al. (2015) is the AGN occupation sample used in §3.4. Please check this citation; the MILES library should be credited to the appropriate original papers.","section":"§4.3"},{"comment":"The figure caption refers to a legend for the color-coded mass bins, but no legend is visible in the printed figure; please make the mass-bin color coding explicit in the caption or figure.","section":"§4.3, Fig. 16"}],"recommendation":"major_revision","confidential_remarks":"The reader's conditional verdict is consistent with my reading. The M_dyn issue is the main load-bearing uncertainty: it affects the very co-evolution and scatter claims that are highlighted in the abstract. I would ask for at least a sensitivity test or a clearly scoped robustness statement before publication. The dust-calibration circularity is a close second, but it is more local and can be fixed by reframing. No concerns about citation integrity or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi X,\n\nThis is a useful and mostly honest simulation paper. The real news is the public release of z=0 data from ASTRID: full histories of millions of MBHs, merger catalogs, and the new analyses—scaling-relation scatter, cluster stellar budget fits, and BH/galaxy bias. That resource is valuable.\n\nThe paper does well on confrontation with observations. They compare the SMF, quiescent fractions, sSFR, cluster stellar content, and bias to a wide range of data, and they are candid about failures: the BHAD is over an order of magnitude too high at z=0, the quiescent fraction is ~10% where observations say ~40% at 5e10 M_sun, and the SMF knee is underproduced by 0.5 dex. Those admissions are in the text and in the conclusion.\n\nThe listed soft spots are real but not fatal. First, the dust-attenuated luminosity function in Fig 15 is effectively a fitted result—κ_ISM is calibrated to the same SDSS LFs used for comparison. The authors state this plainly in Sec 4.3, but the abstract's 'agrees well' is too strong. The LF section should be framed as a calibration check, not a prediction.\n\nSecond, the M_dyn=1e7 h^-1 M_sun seed dynamical mass is a load-bearing free parameter that is never varied. The stress-test worry that it could homogenize early BH dynamics and inflate the claimed scatter is legitimate, but I don't think it sinks the paper. The authors already warn that the clustering results are partly seed-prescription dependent, and the same caveat applies to the scaling relations. A sensitivity test or a more explicit discussion of the expected M_dyn effect would improve the paper, but its absence does not make the resource misleading.\n\nThe abstract's 'successfully captures co-evolution' language is too smooth given the documented low-redshift feedback deficiencies. Still, the core deliverables—scale relations in the observed ballpark, realistic scatter for a simulation, and a public catalog—are genuinely useful.\n\nWho is this for? Anyone using ASTRID for GW predictions, AGN population studies, or clustering. I would cite it if I worked in those areas. It deserves a serious referee; I would send it out, with comments asking for a recalibrated LF framing and a moderated abstract.","headline":"A solid z=0 simulation characterization with a valuable public data release; the LF 'agreement' is partly calibrated rather than predictive, and the abstract overclaims, but the resource is worth refereeing.","tokens_in":39235,"tokens_out":3419,"would_cite":true,"duration_ms":30812,"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":"This paper presents the z=0 state of the ASTRID cosmological simulation and argues that its population of massive black holes genuinely co-evolves with host galaxies, reproducing observed black-hole–galaxy scaling relations and their scatte","keywords":["cosmological hydrodynamical simulation","massive black holes","black hole–galaxy scaling relations","dynamical friction","AGN luminosity function","galaxy stellar mass function","galaxy clusters","large-scale structure bias"],"falsifier":"Count the frequency of low-mass (1e6–1e8 Msun) central black holes in galaxies with M* ~ 1e10.5–1e11 Msun. ASTRID predicts this population is common; if deep X-ray or optical surveys show such galaxies almost always host black holes above 1e8 Msun, or almost never host black holes this small, the claimed scatter and diversity would be contradicted. Alternatively, run ASTRID with all seeds at a single mass instead of stochastic seed masses and compare the scatter of the M_BH–M* relation: if the scatter collapses, the stochastic seeding is doing the work.","tokens_in":38112,"feed_emoji":"🕳️","tokens_out":5003,"duration_ms":42879,"temperature":0.7,"pith_summary":"This paper presents the z=0 snapshot of ASTRID, one of the largest cosmological hydrodynamic simulations ever evolved to today's universe—0.33 trillion particles in a 370-Mpc cube. The central claim is that ASTRID's massive black holes, which span nearly seven orders of magnitude in mass, genuinely co-evolve with their host galaxies: the simulation reproduces observed M_BH–M* and M_BH–σ scaling relations, including a scatter that is closer to observations than earlier simulations produced. It also produces a local galaxy population matching the observed stellar mass function, dust-corrected luminosity function, and red-blue color bimodality, and it hosts thousands of massive groups and clusters whose stellar contents match observations. If these claims hold, ASTRID provides a credible cosmological laboratory for predicting black hole merger rates for gravitational wave observatories and for interpreting large surveys of galaxies and active galactic nuclei.","feed_headline":"ASTRID reproduces black hole–galaxy scaling relations","feed_subtitle":"A 370-Mpc, 0.33-trillion-particle simulation matches observed black-hole diversity, galaxy colors, and clustering.","key_machinery":"The load-bearing element is the subgrid dynamical-friction model plus a 'dynamical mass' M_dyn=1e7 h^-1 Msun assigned to freshly seeded black holes. The friction force dissipates the momentum of black holes moving through unresolved stars and dark matter, so their orbits decay toward the halo center; M_dyn prevents the lightest seeds (3e4–3e5 Msun) from being artificially kicked around by numerical heating. Together these replace the older black-hole repositioning algorithm, giving black holes physical trajectories and merger criteria based on separation and gravitational binding. The paper argues this mechanism is what allows a realistic diversity of black hole masses to survive in galaxies","core_discovery":"ASTRID is claimed to reproduce, at z=0, the co-evolution of massive black holes and their host galaxies: the scaling relations between central black hole mass and galaxy stellar mass, and between black hole mass and stellar velocity dispersion, agree with observations, and the scatter in these relations is substantially closer to observed scatter than in earlier large-volume simulations. The authors attribute this to treating black hole dynamics through a subgrid dynamical-friction force and omitting the usual repositioning algorithm, so black holes sink and merge more naturally. Beyond the scaling relations, the paper reports a black hole mass function matching observations above about 1e7","pith_inferences":["A direct test of the seeding prescription: if a twin run with uniform seed masses or a lower M_dyn drastically changes the scaling-relation scatter and occupation numbers, the claimed diversity is driven by the numerical choice rather than by realistic dynamics. The authors themselves acknowledge that some results are partly determined by their seeding prescription.","The z=0 overabundance of small black holes (below 1e7 Msun) and the deficit of bright AGN (LX > 1e45 erg/s) are places where next-generation X-ray surveys could falsify or confirm the subgrid accretion and feedback model.","The stable bias of 1e8-Msun black holes at z=0 suggests that future gravitational-wave source catalogs, once cross-correlated with galaxy surveys, could constrain cosmology—but only if the black-hole–galaxy connection remains this tight at higher redshifts.","The sSFR and quiescent-fraction discrepancies around M* ~ 1e10.5–1e11 Msun imply that lowering the kinetic-feedback critical mass would shift the quiescent transition to lower stellar masses; this is a testable consequence of the current feedback threshold."],"forward_implications":["ASTRID's public catalogs can be used to predict massive-black-hole merger rates and gravitational-wave source populations for LISA and pulsar timing arrays.","The z=0 black hole population, including wandering black hole occupation numbers, offers a test bed for interpreting AGN surveys and their luminosity functions.","The massive groups and clusters in ASTRID provide a census of central-galaxy, satellite, and intracluster-light stellar masses that can be compared with X-ray and optical cluster surveys.","Massive black holes above ~1e8 Msun and galaxies above ~1e10.5 Msun are reliable large-scale-structure tracers; their power spectra can be compared with observed clustering to constrain cosmology.","The predicted presence of ~20 wandering black holes in 1e11 Msun galaxies and over 1000 in 1e12 Msun galaxies means future X-ray observations may see off-center AGN in massive galaxies."],"fun_headline_variants":["ASTRID simulation reproduces black hole–galaxy relations","Giant simulation nails black hole scaling scatter","ASTRID tracks black holes to large-scale structure","Simulation captures black hole diversity at z=0","ASTRID: black hole growth and clustering reproduced"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the ad hoc numerical choices for seeding and subgrid black-hole dynamics—the small stochastic seed masses and the artificial dynamical mass M_dyn=1e7 h^-1 Msun used to prevent numerical heating—produce real black-hole behavior rather than artificial scatter; the authors themselves note in Section 6.2 that some results are partly determined by their seeding prescription.","fun_headline_variants_meta":{"raw":{"variants":["ASTRID simulation reproduces black hole–galaxy relations","Giant simulation nails black hole scaling scatter","ASTRID tracks black holes to large-scale structure","Simulation captures black hole diversity at z=0","ASTRID: black hole growth and clustering reproduced"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000166,"raw_usage":{"total_tokens":1181,"prompt_tokens":927,"completion_tokens":254,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":671,"completion_tokens_details":{"reasoning_tokens":178}},"tokens_in":671,"tokens_out":254,"duration_ms":3008,"temperature":1.0,"reasoning_tokens":178,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T09:38:28.325961+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Count the frequency of low-mass (1e6–1e8 Msun) central black holes in galaxies with M* ~ 1e10.5–1e11 Msun. ASTRID predicts this population is common; if deep X-ray or optical surveys show such galaxies almost always host black holes above 1e8 Msun, or almost never host black holes this small, the claimed scatter and diversity would be contradicted. Alternatively, run ASTRID with all seeds at a single mass instead of stochastic seed masses and compare the scatter of the M_BH–M* relation: if the scatter collapses, the stochastic seeding is doing the work.","supporting_citations":[],"review_version":1}