{"id":"72007ec4-e815-4ded-a0af-856faeacb99c","arxiv_id":"2412.14900","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Supermassive stars can form in metal-poor clouds with up to about 10^-3 solar metallicity, while more enriched clouds transition to forming dense star clusters.","lead":"This paper simulates how massive stars form in gas clouds with tiny amounts of metals under intense ultraviolet light. It finds that supermassive stars can still form if the metal content is below about one thousandth of the Sun's, while more enriched clouds instead make dense star clusters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Z~1e-3 threshold rests on a single cloud realization and a two-point metallicity bracket; if the [Z/H]=-3 filamentary inflow is not generic, the critical metallicity could shift substantially.","rationale":"The paper is a solid advance over Paper I: it adds self-consistent thermal and chemical evolution, stellar radiative feedback, and 2 Myr evolution, and the resolution checks in Appendix A support the total mass budget of the massive stars. The physical picture of super-competitive accretion is internally consistent, and I found no contradiction between the simulations and their stated mechanisms. My concern is not that the runs are wrong, but that the headline threshold is a generalization from one cloud and only two metallicity points that happen to straddle the transition. The reader identified exactly this as the weakest assumption, and I agree. Because the central claim is plausible but not yet robust to initial-condition variation, CONDITIONAL is the right verdict; nothing in my read pushes it to ACCEPT or REJECT. The concrete test of a second cloud or a perturbed turbulent realization would directly address the gap.","tokens_in":34546,"tokens_out":7572,"duration_ms":69084,"concrete_test":"Rerun the [Z/H] = -3 case with the same code and physics but a different realization: preferably the second SMS-forming cloud from Chon et al. (2016), or with the turbulent velocity seed of the Spherical Cloud resampled while preserving the power spectrum. If the most massive star at 2 Myr falls below 10^4 M_sun in either variant, the threshold is realization-dependent and the universal Z ~ 1e-3 claim is not established; if it remains above 10^4 M_sun, the single-cloud concern is substantially weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that SMSs form up to Z ~ 1e-3 Zsun is drawn from one initial condition, the Spherical Cloud of Chon et al. (2016), with all metallicities evolved from the same density, velocity, angular-momentum, and tidal field (Section 2.1). The decisive success case, [Z/H] = -3, works because a warm outer reservoir (T ~ 10^4 K) is delivered to the center through a pc-scale dense filament, while the cooler inner gas (T ~ few 100 K) accretes at a low rate early on (Figure 4 and Section 3.1). At [Z/H] = -2, fragmentation on 1-10 pc isolates the cores and starves the center. This sharp contrast is therefore produced by the particular cloud structure, and the paper provides no variation of the initial turbulent field, rotation, or tidal truncation that would show the outcome is a property of the physical mechanism rather than of this realization. Even accepting the cloud, the threshold is bracketed only by [Z/H] = -3 and -2, a decade apart; no run at [Z/H] = -2.5 is shown. Since the cosmological conclusion in Section 4 (seed BH number density comparable to local SMBHs) uses this threshold as input, a shift by even a factor of a few in Z could change the inferred abundance substantially. This is not a rejection of the mechanism: the long-term runs, resolution comparison in Appendix A, and agreement of the thermal tracks with one-zone models give real support. But the specific universal threshold at ~1e-3 Zsun is underdetermined by the current simulation set.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents three-dimensional radiation hydrodynamics simulations of star formation in strongly FUV-irradiated, metal-poor clouds, with metallicities from [Z/H] = -6 to -2. The authors follow collapse from a single cosmological initial condition (the Spherical Cloud of Chon et al. 2016), including non-equilibrium primordial chemistry, simplified C+ and O line cooling, dust cooling and heating, and radiative feedback from accreting protostars. Using paired high-resolution/short-term and low-resolution/long-term runs, they find that SMSs of 3-8e4 Msun form for [Z/H] <= -3, that the [Z/H] = -3 case catches up to the SMS regime only after about 1e5 yr via accretion from a pc-scale dense filament, and that at [Z/H] = -2 the central star reaches only about 2e3 Msun while a compact star cluster forms. The paper interprets this as a metallicity threshold Z_th ~ 1e-3 Zsun for SMS formation, proposes super-competitive accretion as the responsible mechanism, and uses this threshold to argue that massive seed black holes may be abundant enough to match local SMBH number densities.","tokens_in":34888,"tokens_out":4768,"duration_ms":47901,"significance":"If the result holds, it is an important step beyond Paper I: it relaxes the direct-collapse requirement from essentially pristine gas to Z up to roughly 1e-3 Zsun, roughly an order of magnitude above the earlier threshold, and it gives a concrete mechanism (super-competitive accretion) through which fragmentation does not prevent SMS growth. The simulations are substantially improved over Paper I by including self-consistent thermal/chemical evolution, stellar radiative feedback, and two-million-year runs, and the resolution study in Appendix A directly checks the main mass-budget claim. The comparison with one-zone thermal tracks in Figure 4 is a useful sanity check. The claimed transition from SMS formation to dense cluster formation at [Z/H] = -2 is interesting and potentially relevant to globular cluster and JWST compact-cluster observations. However, the central threshold claim is currently supported by only one cloud realization and by a two-point metallicity bracket, and the cosmological abundance estimate relies on an extrapolated collapse fraction; these issues make the quantitative threshold and the SMBH abundance conclusion less secure than the abstract suggests.","major_comments":[{"comment":"All five metallicity runs start from the same Spherical Cloud initial condition, with the same density field, velocity field, angular momentum, turbulence, and tidal environment. The decisive difference between [Z/H] = -3 and [Z/H] = -2 is attributed to the scale at which metal-line cooling triggers fragmentation (pc-scale filament versus 1-10 pc cores), but this outcome is sensitive to the specific structure of this single cloud. In particular, the late-time SMS growth at [Z/H] = -3 requires a warm outer reservoir delivered by a pc-scale filament; whether such a configuration is generic for FUV-irradiated, mildly enriched halos is not demonstrated. The paper should either add variations of the initial turbulent field, rotation, or tidal truncation, or explicitly restrict the claim to this cloud class and state that the universal threshold is not yet established.","section":"Section 2.1 and Section 3.1"},{"comment":"The claimed threshold near [Z/H] = -3 is bracketed only by the two metallicities [Z/H] = -3 and -2, a factor of ten in Z, with no run at an intermediate value such as [Z/H] = -2.5. This matters because the [Z/H] = -3 case is also the one where the simulation deviates most strongly from the one-zone model (the text attributes this to collapse faster than free-fall), so the success at [Z/H] = -3 is not backed by the same analytic expectation as the lower-metallicity cases. A run at [Z/H] = -2.5, or a quantitative analytic estimate of when 10-pc-scale fragmentation starts to starve the center, is needed before the abstract's 'approximately 1e-3 Zsun' threshold can be considered robust.","section":"Section 3.1 and Figure 4"},{"comment":"The cosmological conclusion that massive seed BHs reach number densities of 0.1-1 Mpc^-3 and can account for local SMBHs uses a fixed 5% success rate from Chon et al. (2016), applied to the candidate-halo sample of Chiaki et al. (2023). That 5% fraction was derived for the same cloud population from which the present initial condition is drawn, and it is not shown to apply to finite-metallicity, FUV-irradiated halos. Since the ordinate of Figure 15 is directly proportional to this correction factor, a metallicity-dependent or halo-dependent collapse fraction could shift the inferred abundance by orders of magnitude. The authors should either compute or bound f_coll for the finite-metallicity case, or present the abundance prediction as a conditional estimate with explicit sensitivity to this parameter.","section":"Section 4 and Figure 15"}],"minor_comments":[{"comment":"The text states that for [Z/H] = -2 the authors 'adopt a slope alpha of 1.5 instead of the value of 1 obtained from the fitting,' but Table 1 reports alpha = 1.09 for [Z/H] = -2; the discrepancy between 1 and 1.09 should be reconciled.","section":"Section 3.3.2 and Table 1"},{"comment":"The caption contains the typo 'reolution' for 'resolution' in the labels, which should be corrected.","section":"Appendix A, Figure A1 caption"},{"comment":"Metal-line cooling is limited to C+ and O, with C and O assumed fully ionized/neutral respectively; since the [Z/H] = -2 boundary is where the transition occurs, the absence of other coolants (e.g., Si+, Fe+) is a simplification that could shift the threshold. The text acknowledges the simplification, but a brief discussion of its expected direction and magnitude would help.","section":"Section 2.2"},{"comment":"The adiabatic cutoff at n_adia = 1e16 cm^-3 and sink formation at twice that density mean that the final SMS masses are determined partly by unresolved gas above this density. This is a standard opacity-limit treatment, and Appendix A shows total mass convergence between resolutions, but the paper should state more explicitly that the quoted final masses are the masses of the resolved sinks and could be affected by sub-grid accretion or disk physics.","section":"Section 2.5"}],"recommendation":"major_revision","confidential_remarks":"This is a solid and interesting simulation paper, and the central mechanism is credible. The main risk is not internal inconsistency but overgeneralization from a single initial condition and a two-point metallicity bracket. I would ask the authors to add at least one intermediate-metallicity run or an equivalent robustness argument, and to soften or carefully qualify the cosmological abundance claim, before publication. The paper is well within the scope of the journal and, with those changes, would be a strong contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The key result is that supermassive stars can form at Z up to roughly 1e-3 Zsun, an order of magnitude higher than the authors' own Paper I suggested, and that the same mechanism naturally yields dense clusters at higher metallicity. The new physics is real: self-consistent thermal and chemical evolution, radiative feedback, and runs to 2 Myr. The resolution comparison in Appendix A is solid, and the thermal tracks match one-zone models except for the interesting [Z/H]=-3 case. The 'super-competitive accretion' idea comes through clearly: despite sub-parsec fragmentation, the central stars win the gas.\n\nThe soft spots are proportionate to the claim. The threshold rests on one initial cloud from Chon et al. (2016), with all metallicities started from the same density, velocity, angular-momentum, and tidal structure. The [Z/H]=-3 case works because a warm outer reservoir is delivered through a pc-scale filament while the inner gas stays cooler and starves initially; at [Z/H]=-2, larger-scale fragmentation isolates the cores. That is a plausible mechanism, but it is a property of this realization until someone varies the turbulence, rotation, or tidal field. The metallicity bracket is also coarse: -3 vs -2, with no run at -2.5. The cosmological number-density estimate in Section 4 depends on a 5% collapse fraction from the authors' earlier simulation, so the 'explains all local SMBHs' line is a chain of assumptions, not a direct prediction. Approximations like the inverse-square radiation temperature and simplified C+/O cooling are standard for this kind of work, but they do add uncertainty.\n\nWho this is for: the direct-collapse and seed-BH community, and people working on globular cluster formation. It deserves a serious referee. I would ask for either a second initial condition or much stronger caveats about the threshold's robustness.","headline":"A plausible but underdetermined upward revision of the SMS metallicity threshold, built on one cloud realization and a two-point bracket.","tokens_in":35444,"tokens_out":2187,"would_cite":true,"duration_ms":18466,"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 paper argues that supermassive stars can form below a metallicity threshold of about $10^{-3}\\,Z_\\odot$, with the most massive reaching $3$–$8\\times10^4\\,M_\\odot$, and that above this threshold dense star clusters form instead.","keywords":["supermassive stars","direct collapse black holes","star cluster formation","metallicity threshold","FUV radiation","radiation hydrodynamics","super-competitive accretion","Population III"],"falsifier":"A simulation of a different FUV-irradiated cloud at $[Z/H] = -3$ that fragments into cores separated by more than a few parsecs and fails to feed a central star past $10^4\\,M_\\odot$, or a high-redshift cluster with $[Z/H] > -3$ that demonstrably formed without a very massive star, would contradict the claim that the threshold is general.","tokens_in":34301,"feed_emoji":"🌌","tokens_out":5825,"duration_ms":44017,"temperature":0.7,"pith_summary":"This paper tries to show that the direct-collapse pathway to supermassive black holes does not require chemically pristine gas. Using radiation hydrodynamic simulations run for about two million years across metallicities $Z=10^{-6}$ to $10^{-2}\\,Z_\\odot$, it finds supermassive stars above $10^4\\,M_\\odot$ forming whenever $Z$ is at or below roughly $10^{-3}\\,Z_\\odot$, with final masses of $3$–$8\\times10^4\\,M_\\odot$ for $[Z/H]\\lesssim -3$. The key process is 'super-competitive accretion': despite sub-parsec fragmentation, the central massive stars capture most of the inflowing gas, by direct accretion or mergers. At $[Z/H] = -2$ this fails: fragmentation on 1–10 pc scales, aided by ionizing feedback, limits the most massive star to about $2000\\,M_\\odot$ and produces a bound dense cluster instead. A sympathetic reader should care because this raises the expected density of massive seed black holes toward the observed local number of supermassive black holes and offers a unified origin for supermassive stars and dense star clusters.","feed_headline":"Supermassive stars can form in metal-enriched clouds","feed_subtitle":"Runs out to 2 Myr show 10^4-solar-mass stars still form below 0.1% solar metallicity; above it, dense clusters emerge.","key_machinery":"The central mechanism is 'super-competitive accretion': after fragmentation creates many protostars and low-mass stars, the gravitational potential of the central massive stars steers the remaining gas into them, both through direct inflow and through mergers of inward-migrating fragments, so the gas supply is not split evenly among competitors. The surrounding physics that makes this work is the FUV radiation field, which dissociates H$_2$ and keeps the low-density gas warm near $10^4$ K, sustaining high infall rates of order $1\\,M_\\odot\\,\\mathrm{yr}^{-1}$; at higher densities, metal-line and dust cooling cause fragmentation, but on scales small enough that the fragments still fall into the center. The paper also models the stellar radiative feedback (ionizing radiation, H$_2$ dissociation, H$^-$ photodetachment, and dust heating) and shows that its main effect at low metallicity is to suppress low-mass star formation rather than to halt supermassive star growth.","core_discovery":"The central discovery is that supermassive stars with masses exceeding $10^4\\,M_\\odot$ can form in strongly FUV-irradiated clouds even when the gas is metal-enriched, as long as the metallicity stays below roughly $10^{-3}\\,Z_\\odot$. In the simulations, the most massive stars reach $6$–$8\\times10^4\\,M_\\odot$ at $[Z/H]\\lesssim -4$ and about $3\\times10^4\\,M_\\odot$ at $[Z/H] = -3$, whereas at $[Z/H] = -2$ the most massive star barely exceeds $2000\\,M_\\odot$ and the system becomes a dense star cluster. The growth of the central stars is sustained by 'super-competitive accretion', in which the gravitational dominance of the central massive stars channels most of the gas reservoir toward them even while sub-parsec fragmentation produces many low-mass companions. Radiative feedback from the forming stars is ineffective at stopping this growth at low metallicity, because high accretion rates keep the surroundings dense enough to absorb ionizing photons, although dust heating does suppress low-mass star formation. These results revise the earlier expectation that metal enrichment at levels above about $10^{-5}\\,Z_\\odot$ would prevent direct-collapse seed formation.","pith_inferences":["Inference: because the simulations start from a single cloud realization, the $10^{-3}\\,Z_\\odot$ threshold may not be universal; clouds with different spin, turbulence, or tidal fields could shift the boundary by a factor of several.","Inference: if the seed black hole density really reaches $0.1$–$1\\,\\mathrm{Mpc}^{-3}$, the scenario may remove the need for super-Eddington accretion to explain the local supermassive black hole population.","Inference: nitrogen-rich compact clusters seen in the early universe could serve as observational tests; clusters with metallicities above the threshold that still show evidence of very massive stars would weaken the claim, while clusters below the threshold would support the supermassive-star nucleosynthesis route.","Inference: magnetic fields, which are not included here, tend to suppress disk fragmentation and could push the supermassive star formation channel to higher metallicities or to even larger final masses."],"forward_implications":["If the threshold holds, supermassive star formation becomes viable in clouds with a prior episode of star formation, not only in pristine gas, increasing the expected number density of massive seed black holes.","Using the semi-analytic halo counts and a 5% collapse-success correction, the paper estimates a seed black hole density of $0.1$–$1\\,\\mathrm{Mpc}^{-3}$ for $[Z/H]\\gtrsim -3$, comparable to the local supermassive black hole density.","At $[Z/H] = -2$, the outcome is a bound, compact cluster with a central very massive star of about $1000$–$2000\\,M_\\odot$ and a stellar surface density near $10^3\\,M_\\odot\\,\\mathrm{pc}^{-2}$, resembling young massive or globular clusters and matching the observed metallicity floor of Milky Way globular clusters around $[Z/H] \\sim -2.5$.","The simulated remnant population contains four to six massive seed black holes per system plus a population of intermediate-mass black holes, and the massive black holes often form binaries whose gravitational-wave mergers could be detectable at high redshift.","The low-metallicity mass spectra are composite: a Chabrier-like low-mass population peaking near $1\\,M_\\odot$, produced and shaped by stellar irradiation, plus a supermassive component of $10^4$–$10^5\\,M_\\odot$ that dominates the total stellar mass."],"supporting_citations":[{"why":"Introduced super-competitive accretion and first showed, with simplified thermal treatment and shorter runs, that supermassive stars may form in metal-enriched clouds; the present paper extends that result.","marker":"Chon & Omukai (2020)"},{"why":"Supplies the Spherical Cloud initial conditions, including its density, angular momentum, and tidal environment, and the 5% collapse-success estimate used for black hole number densities.","marker":"Chon et al. (2016)"},{"why":"Demonstrated that the same cloud forms a supermassive star above $10^4\\,M_\\odot$ in the metal-free case, providing the baseline for the metallicity scans here.","marker":"Chon et al. (2018)"},{"why":"Provides the supergiant protostar radius relation and the critical accretion rate that govern whether accreting protostars stay inflated or contract and become subject to ionizing feedback.","marker":"Hosokawa et al. (2013)"},{"why":"Provides the one-zone thermal evolution models showing metal-line and dust cooling thresholds, used to set particle-splitting densities and to validate the simulated temperature history.","marker":"Omukai et al. (2008)"},{"why":"Supplies the semi-analytic halo number densities as a function of threshold metallicity, from which the expected massive seed black hole density is derived.","marker":"Chiaki et al. (2023)"},{"why":"Supplies the H$_2$ self-shielding prescription used to compute how the strong FUV background dissociates molecular hydrogen in the collapsing clouds.","marker":"Wolcott-Green et al. (2011)"}],"fun_headline_variants":["Supermassive stars form in clouds with up to 0.1% solar metals","Below 0.1% solar metallicity, supermassive stars still emerge","Up to 0.1% solar metals: supermassive stars still form","Dense clusters replace supermassive stars above 0.1% solar metals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that one simulated cloud, with its particular density, spin, turbulence, and tidal environment, stands in for all FUV-irradiated metal-poor clouds; if other real clouds differ enough, the $10^{-3}\\,Z_\\odot$ threshold and the super-competitive accretion mode could shift or disappear.","fun_headline_variants_meta":{"raw":{"variants":["Supermassive stars form in clouds with up to 0.1% solar metals","Below 0.1% solar metallicity, supermassive stars still emerge","Up to 0.1% solar metals: supermassive stars still form","Dense clusters replace supermassive stars above 0.1% solar metals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001149,"raw_usage":{"total_tokens":4874,"prompt_tokens":1167,"completion_tokens":3707,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":783,"completion_tokens_details":{"reasoning_tokens":3618}},"tokens_in":783,"tokens_out":3707,"duration_ms":17469,"temperature":1.0,"reasoning_tokens":3618,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:49:02.306481+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A simulation of a different FUV-irradiated cloud at $[Z/H] = -3$ that fragments into cores separated by more than a few parsecs and fails to feed a central star past $10^4\\,M_\\odot$, or a high-redshift cluster with $[Z/H] > -3$ that demonstrably formed without a very massive star, would contradict the claim that the threshold is general.","supporting_citations":[],"review_version":1}