{"id":"9733724b-51f2-4dc4-8d7e-b2412d4a432d","arxiv_id":"2411.14209","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"Supernova feedback in minihalos ejects most metals beyond the virial radius until halos grow to about 10 million solar masses, delaying the start of enriched star formation.","lead":"Using a new star-by-star simulation of the early universe, this paper shows that the first supernovae blow most of their newly made metals out of the small dark-matter halos that host the first stars. As a result, those halos stay metal-poor and keep making Population III stars for longer than simple models assume.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Metal-retention threshold at ~10^7 Msun depends on excluded 1e53 erg pair-instability SNe; a single PISN energy exceeds the binding energy of a 10^7 Msun halo, so the headline threshold is model-dependent.","rationale":"The reader's weakest_assumption correctly identifies the PISN/IMF prescription as the least secure input to the headline threshold. I agree. The simulation itself is internally consistent: the loss-fraction bookkeeping, the ell(Mdm) evolution, and the qualitative agreement with earlier work (Kitayama & Yoshida 2005; Hicks et al. 2021; Muratov et al. 2013) all support the claim within the adopted model. The paper's Sec. 6.2 explicitly concedes that PISNe and vbc=0 can change the characteristic retention masses and the monotonicity of metallicity, yet these caveats are absent from the abstract. Because the concern is about model dependence rather than an internal error, it does not warrant rejection; it warrants keeping the conditional verdict. The proposed PISN rerun is the decisive test: it directly measures whether the 1e7 Msun threshold is an artifact of excluding 1e53 erg events. If the threshold is stable, the abstract can remain; if it shifts, the headline must be caveated. I would not change the reader's CONDITIONAL verdict.","tokens_in":22352,"tokens_out":10318,"duration_ms":103508,"concrete_test":"Run an AEOS variant in which Pop III stars with 140 < M < 260 Msun explode as 1e53 erg PISNe with the corresponding yields, keeping the IMF and all other physics fixed, and recompute the metal-loss curves ell(Mdm) and the characteristic metal-retention mass at z=14.3. If the retention mass moves above 10^7 Msun, or if ell for Mdm > 10^7 drops significantly relative to Figure 5, the abstract's quantitative threshold and 'regardless of number of supernovae' phrasing must be qualified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is conditional on the adopted Pop III IMF and SN prescription (Sec. 2.3): only 10–100 Msun stars explode as 1e51 erg CCSNe; 100–300 Msun stars collapse directly with no feedback; PISNe at 140–260 Msun are excluded. This is not a peripheral detail. With a Salpeter IMF alpha=2.3 and Mchar=20 Msun, the number fraction of 140–260 Msun progenitors is small (~2%), but their energy per event is 1e53 erg, roughly 100x a CCSN. The binding energy of gas in a 1e7 Msun halo at z~14 is of order 1e51 erg, so a single PISN carries enough energy to unbind gas from halos well above the claimed 1e7 Msun retention threshold. Section 6.2 item 2 explicitly concedes that including PISNe could increase the characteristic retention masses and the extent of IGM enrichment. Since the abstract states the threshold 'M* <~ 10^7 Msun' and the phrase 'regardless of the number of supernovae' without this caveat, the quantitative headline is not robust to this model choice. The vbc=0 choice (Sec. 6.2 item 3) similarly can raise the mass of first star-forming halos and suppress outflows, potentially making metallicity monotonic. The simulation is internally consistent, but the headline threshold is model-dependent in exactly the places the paper itself identifies.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses the Aeos star-by-star cosmological hydrodynamics simulation (1 comoving Mpc box, evolved to z = 14.3) to study how Population III stellar feedback transports metals out of minihalos. The central claim is that energy from core-collapse supernovae (CCSNe) expels a majority of the gas and injected metals beyond the virial radius of halos with dark matter mass Mdm <= 10^7 Msun, almost regardless of the number of supernovae, so that most minihalos do not retain their own nucleosynthetic yields until they grow to Mdm >= 10^7 Msun. The authors infer non-monotonic early metallicity evolution, metal deposition into the IGM followed by reaccretion or external enrichment of neighboring halos, and a delay of the Population III to Population II transition. A second component of the paper decomposes enrichment into 10 elements from CCSNe, Type Ia SNe, and AGB winds, finding CCSN dominance at z ~ 14.3 with a significant AGB contribution to the s-process elements Sr and Ba. The fiducial model adopts a Salpeter Pop III IMF with Mchar = 20 Msun, CCSNe at fixed 10^51 erg for 10-100 Msun stars, direct collapse without feedback for 100-300 Msun stars, no pair-instability supernovae, and zero streaming velocity; the authors describe these choices in Sections 2 and 6.2.","tokens_in":22655,"tokens_out":21025,"duration_ms":180314,"significance":"The metal-loss fraction (Eq. 2) is an emergent bookkeeping quantity rather than a fitted parameter, which is a genuine strength: the result that less than about 8% of injected metals reside within halos until z ~ 15.3, that 92% of halo metals are concentrated in the four most massive halos, and that metals are exchanged between neighboring halos (Halos 0 and 1) are concrete, falsifiable statements. The star-by-star treatment that resolves the Sedov-Taylor phase of most CCSNe at 1 pc resolution, the tracking of 10 elements through distinct nucleosynthetic channels, the candid robustness section (Sec. 6.2), and the quantitative comparison with the Kulkarni et al. (2021) critical mass all work in the paper's favor. If the central threshold holds, the implications for the Pop III to Pop II transition are substantial, because semi-analytic models assuming monotonic enrichment (e.g., Visbal et al. 2020; Hartwig et al. 2022) would need revision. The caveat is that the headline threshold and the non-monotonicity are conditional on the fiducial CCSN-only, vbc = 0 prescription, and the paper should present them as such.","major_comments":[{"comment":"The headline result, expulsion of gas and metals beyond rvir for halos with Mdm <= 10^7 Msun regardless of the number of supernovae, is stated categorically in the abstract and in Summary item 5, but it is established only for the fiducial prescription of Section 2.3: Pop III stars of 10-100 Msun explode as 10^51 erg CCSNe, stars of 100-300 Msun collapse directly with no feedback, and pair-instability SNe (140-260 Msun) are excluded. This model-dependence is load-bearing, not cosmetic. For a 10^7 Msun halo at z ~ 14 (Tvir ~ 3430 K, vc ~ 9.7 km/s, as quoted in Sec. 4), the gas binding energy is of order G Mdm Mg / rvir ~ 10^51 erg, comparable to a single CCSN; a single 10^53 erg PISN therefore exceeds the binding energy by two orders of magnitude and would unbind gas from halos well above the claimed threshold, assuming comparable coupling. This is precisely the point conceded in Section 6.2 item 2, which states that including PISNe could increase the characteristic retention masses and the extent of IGM enrichment. I recommend (i) making the abstract and Section 6.1 explicitly conditional, e.g., in the fiducial CCSN-only, vbc = 0 model, and (ii) adding a short analytic energy-budget estimate, injected energy per event versus gas binding energy as a function of Mdm for 10^51 and 10^53 erg events, so that the reader can see both the origin of the 10^7 Msun threshold and its sensitivity to the adopted SN prescription.","section":"Abstract; Secs. 2.3, 4, 6.2"},{"comment":"The abstract's claims of non-monotonic metallicity evolution and a delayed Pop III to Pop II transition are presented as general findings, but they rest on the choice vbc = 0. Section 6.2 item 3 concedes that for halos first forming stars at Mdm > 3 x 10^6 Msun, feedback may no longer expel gas and the halo metallicity could instead increase monotonically with time. Since realistic streaming velocities shift first star formation to larger and later-forming halos, the vbc = 0 assumption biases the simulation toward the paper's qualitative conclusions rather than away from them. The abstract and Summary item 3 should state the non-monotonicity result as conditional on vbc = 0 with a forward reference to the robustness discussion; as written, a reader of the abstract will take the non-monotonicity to be a robust feature of early enrichment rather than a fiducial-model outcome.","section":"Abstract; Sec. 6.2 item 3"},{"comment":"The N-independence of the metal loss fraction is the paper's principal argument for robustness to IMF variation (Section 6.2 item 1), but the evidence supports a narrower claim than the one stated. Figure 5 demonstrates approximately unit loss fraction independent of SN number only over the simulated range of roughly 0-10 CCSNe per halo, and the abstract's regardless-of-number phrasing applies to expulsion beyond rvir; Section 4.1.3 shows for Halo 2 that a clustered series of SNe drives metals beyond 2rvir and states that ejection at farther distances may depend on the number of SNe. The distinction matters because a PISN is not merely an additional event of the same kind: at 10^53 erg it is a roughly 100-fold increase in the energy budget per event, a regime that the N-independence test does not sample. As written, the robustness claim in Section 6.2 item 1 (no change in the characteristic retention mass so long as Pop III CCSNe are prevalent) conflates event count with energy budget. I recommend either scoping the claim to the fiducial 10^51 erg events and the rvir boundary, or adding a test that rescales the injected SN energy to 10^52-10^53 erg to map the threshold's sensitivity.","section":"Secs. 4.1.3, 5, 6.2 item 1; Fig. 5"}],"minor_comments":[{"comment":"The symbol M* used for the retention threshold in the abstract is ambiguous; the body (Section 4 and Figures 5-10) defines the threshold in terms of dark matter halo mass Mdm, and M* conventionally denotes stellar mass. Please align the abstract notation with the body.","section":"Abstract"},{"comment":"The statement that most of our Pop III stars will have masses between 20 and 300 Msun is a number-count statement whose truth depends on the normalization of the exponential cutoff below Mchar; please clarify whether most refers to the number fraction or the mass fraction of sampled stars.","section":"Sec. 2.2.1"},{"comment":"The caption contains the typo Alhough, which should read Although.","section":"Fig. 3 caption"},{"comment":"The Sr and Ba loss-fraction panels are known to be affected by the AGB-injection accounting artifact described in the text; please mark the affected panels in Figure 12 explicitly, or correct the injection-time accounting, so that their anomalous values of the loss fraction are not read as physical.","section":"Sec. 5.2; Fig. 12"},{"comment":"The inference that Halo 2 may be entering the metal-retention stage rests on the direction of a single final snapshot; a sentence acknowledging the single-timestep basis of this inference would strengthen the case study.","section":"Sec. 4.1.2; Fig. 7"},{"comment":"The simulation uses 1 pc maximum refinement and no pressure floor, with the Jeans length potentially unresolved; please state explicitly whether the metal-loss statistics are converged with resolution, with a reference to B24, or flag the resolution dependence as a caveat.","section":"Sec. 2.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the journal's scope, and the main result is worth publishing once the headline claims are brought into line with the paper's own robustness section. Two editorial notes. First, the simulation methods and the Mg yield calibration live in a companion paper (Brauer et al. 2025) with overlapping authorship; this is a normal companion-paper arrangement, but it means the referee of the present paper cannot fully audit the central physics from this manuscript alone, so please ensure that B24 receives equivalent scrutiny. Second, the regardless-of-the-number-of-supernovae phrasing carries a heavy load in the abstract; if the authors decline to add the energy-budget estimate suggested in my first major comment, the editor may wish to ask them at minimum to match the abstract's conditionality to Section 6.2."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a solid simulation paper with a genuinely new quantitative result—per-halo metal loss fractions from a star-by-star cosmological run—and an honest discussion of its own model dependencies. The headline threshold (~10^7 Msun) is real within the adopted model, but it is not as robust as the abstract implies.\n\nWhat's actually new: Aeos tracks individual Pop III and Pop II stars with yields for ten elements, and the paper computes element-resolved loss fractions as a function of halo mass. The finding that halos below roughly 10^7 Msun lose essentially all injected metals, independent of CCSN count, is clearly presented. The early AGB contribution to Sr and Ba is a nice result. The comparison with Kulkarni et al. (2021) for Pop III formation masses is useful. The bookkeeping quantity ℓ is emergent, not fitted, so the central analysis has real content.\n\nSoft spots, in order: (1) The central threshold is conditioned on excluding pair-instability SNe and on zero streaming velocity. A single 1e53 erg PISN can unbind gas from a 10^7 Msun halo, and the paper's own Section 6.2 concedes this would raise the retention mass. The abstract states the threshold and \"regardless of the number of supernovae\" without that caveat. That is an overreach, though not fatal—the qualitative result that minihalos lose most metals survives. (2) The vbc=0 choice could make metallicity monotonic in some halos; again acknowledged but not in the abstract. (3) The Sr/Ba loss fraction plots have the acknowledged accounting issue, so that part of the element analysis should be treated as preliminary. For the other elements the loss behavior is consistent.\n\nThe math and bookkeeping look fine. The yield tables come from standard sources, and the Mg boost is a calibrated input, but it does not drive the loss fraction. Self-citations to Brauer et al. (2025) are appropriate for the method.\n\nThis deserves a serious referee. The referee should push for sensitivity tests (or at least abstract caveats) on PISNe and vbc, and for a fix or explicit downgrade of the Sr/Ba loss analysis. With those changes, the paper is a solid contribution to early galaxy formation. I would bring it to reading group and would cite it.","headline":"Solid star-by-star simulation paper with a real quantitative threshold that is more model-dependent than the abstract admits; deserves refereeing with requested sensitivity caveats.","tokens_in":23279,"tokens_out":2182,"would_cite":true,"duration_ms":19669,"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 first supernovae blow most metals out of minihalos below about 10^7 solar masses, so early chemical enrichment is external rather than self-enrichment.","keywords":["Population III stars","chemical enrichment","minihalos","supernova feedback","metal transport","intergalactic medium","star-by-star simulation","s-process elements"],"falsifier":"Run the same Aeos initial conditions with pair-instability supernovae (140–260 $M_\\odot$, $\\sim 10^{53}$ erg) included and check whether halos below $10^7 \\, M_\\odot$ still lose essentially all their injected metals; a retained fraction in that case would show the central threshold depends on the assumed supernova energies.","tokens_in":22087,"feed_emoji":"💥","tokens_out":9628,"duration_ms":85086,"temperature":0.7,"pith_summary":"This paper uses the star-by-star cosmological hydrodynamics simulation Aeos to follow where the metals created by the first stars end up. It claims that the energy from core-collapse supernovae of Population III stars is enough to eject most of a halo's gas and newly produced metals beyond the virial radius whenever the dark matter mass is below about $10^7 \\, M_\\odot$, regardless of how many supernovae explode. As a result, most minihalos retain almost none of the yields they produce until they grow past that threshold, and halo metallicity does not rise steadily: it spikes with each supernova and then drops as the metals are blown out. If correct, the first supernovae delay the transition to enriched (Population II) star formation, because the metals that eventually form the next stellar generation fall back from the intergalactic medium rather than staying in their birth halo. The paper also resolves ten individual elements and finds that core-collapse supernovae dominate early enrichment, with asymptotic giant branch winds already contributing noticeably to the s-process elements strontium and barium.","feed_headline":"First supernovae blast most metals out of minihalos","feed_subtitle":"A star-by-star simulation shows halos under 10^7 solar masses lose nearly all their metal, delaying the first enriched stars.","key_machinery":"The central quantity is the metal loss fraction $\\ell = (M_{\\rm met,inj} - M_{\\rm met})/M_{\\rm met,inj}$, where $M_{\\rm met}$ is the gas-phase metal mass currently inside a halo's virial radius and $M_{\\rm met,inj}$ is the cumulative metal mass injected by supernovae and AGB winds within that halo or its ancestors. This single diagnostic lets the authors distinguish injection from retention and identify the dark matter mass at which halos switch from losing ~100% of their metals to rebuilding them. The simulation Aeos itself is the enabling object: it treats each star particle as an individual star sampled from an IMF rather than as a whole stellar population, samples Population III masses from a Salpeter-like IMF with characteristic mass 20 $M_\\odot$, and deposits $10^{51}$ erg per core-collapse supernova, with yields for ten elements from standard nucleosynthesis tables. The paper uses the loss fraction both for total metals and for individual elements to show that all elements share the same retention threshold until the halo reaches roughly $10^7 \\, M_\\odot$.","core_discovery":"Using the Aeos simulation, the authors find that supernova feedback from the first stars expels a majority of gas and injected metals beyond the virial radius of halos with $M_{\\rm dm} \\lesssim 10^7 \\, M_\\odot$, and that this loss is essentially independent of the number of supernovae. Most minihalos ($M_{\\rm dm} \\gtrsim 10^5 \\, M_\\odot$) therefore do not retain significant fractions of their own nucleosynthetic yields until they have grown to $M_{\\rm dm} \\gtrsim 10^7 \\, M_\\odot$. The resultant metallicity evolution is non-monotonic: the first supernova produces a rapid spike in metal mass that is immediately followed by a loss of two to three orders of magnitude, and later supernovae do not raise the halo's metallicity until metals begin to reaccrete from the intergalactic medium. Reaccretion is shown to be the dominant process rebuilding halo metals, including exchange between neighboring halos such as Halo 0 and Halo 1. On the timescale of the simulation, core-collapse supernovae dominate the production of all ten tracked elements, but asymptotic giant branch winds contribute significantly to the s-process elements Sr and Ba.","pith_inferences":["A direct test of the energy dependence would be to rerun Aeos with pair-instability supernovae (roughly $10^{53}$ erg) included; the authors themselves expect the retention threshold to move upward, which would make early enrichment even more external and reduce the fraction of halos that ever self-enrich.","If the threshold holds, the abundance patterns of the first enriched stars should often carry the signature of externally produced metals mixed in the IGM rather than a single local supernova; comparing Aeos predictions to the observed scatter in ultra-faint dwarf galaxy abundances could constrain how much mixing occurs between ejection and reaccretion.","The element-resolved loss fractions imply that at z > 14 s-process elements are not preferentially retained despite being injected by gentler AGB winds; at lower redshifts, zoom-in simulations could test whether the lower injection energy of AGB winds eventually makes s-process elements more centrally concentrated than iron-group elements.","An analytic escape condition could be derived from the simulation data: the binding energy of the baryons within the virial radius at the moment of first supernova, compared with the $10^{51}$ erg injection, should predict the $10^7 \\, M_\\odot$ threshold; if such a criterion works, semi-analytic models of the Pop III–Pop II transition can replace their monotonic-enrichment assumption with a simple"],"forward_implications":["Nearly all minihalos below $10^7 \\, M_\\odot$ are prevented from self-enriching; the metals of the first stars accumulate in the intergalactic medium and can later be accreted by other halos.","Early halo metallicity is non-monotonic, so any model that assumes retained ejecta and a steadily rising metallicity will trigger enriched star formation too early.","The loss of metals beyond the virial radius delays the onset of Population II star formation and extends the era in which Population III stars can keep forming.","Because gas and metals are lost at somewhat different characteristic masses ($3 \\times 10^6 \\, M_\\odot$ for gas, $10^7 \\, M_\\odot$ for metals), supernova ejecta do not efficiently mix with the cooler local gas into which they are injected.","All ten tracked elements, including s-process elements, show the same loss behavior and the same $10^7 \\, M_\\odot$ retention threshold at these early times."],"supporting_citations":[{"why":"Defines the Aeos simulation and its star-by-star methods, which this paper uses throughout.","marker":"B24"},{"why":"Provides the star-by-star feedback implementation on which Aeos is built.","marker":"E19"},{"why":"Supplies the Population III core-collapse supernova yields used for enrichment.","marker":"Heger & Woosley (2010)"},{"why":"Supplies the Population II core-collapse supernova yields and their rotation-averaged mixture.","marker":"Limongi & Chieffi (2018)"},{"why":"Supplies the AGB wind yields, the basis for the s-process element contribution.","marker":"Cristallo et al. (2015)"},{"why":"Provides the critical halo mass model used to validate the masses of the first star-forming halos.","marker":"Kulkarni et al. (2021)"},{"why":"Showed that low-mass halos can lose all their gas to supernova feedback, the mechanism extended here.","marker":"Kitayama & Yoshida (2005)"},{"why":"Measured metal retention fractions in minihalos and serves as the quantitative comparison for the retention threshold.","marker":"Muratov et al. (2013)"}],"fun_headline_variants":["First stars' supernovae blast metals out of minihalos","Supernovae from first stars strip minihalos of metals","Metal loss from minihalos delays first enriched stars","Population III supernovae eject metals past virial radius","Minihalos lose most metals to first-star supernovae"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is the adopted Population III stellar prescription: stars from 10 to 100 solar masses explode as ordinary $10^{51}$ erg core-collapse supernovae, stars heavier than 100 solar masses collapse silently, and pair-instability supernovae between 140 and 260 solar masses are left out; if the real first stellar population includes a substantial fraction of pair-instability supernovae at $10^{53}$ erg, the halo mass below which metals are lost would shift upward.","fun_headline_variants_meta":{"raw":{"variants":["First stars' supernovae blast metals out of minihalos","Supernovae from first stars strip minihalos of metals","Metal loss from minihalos delays first enriched stars","Population III supernovae eject metals past virial radius","Minihalos lose most metals to first-star supernovae"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000273,"raw_usage":{"total_tokens":1720,"prompt_tokens":1117,"completion_tokens":603,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":733,"completion_tokens_details":{"reasoning_tokens":515}},"tokens_in":733,"tokens_out":603,"duration_ms":4933,"temperature":1.0,"reasoning_tokens":515,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:25:44.531008+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same Aeos initial conditions with pair-instability supernovae (140–260 $M_\\odot$, $\\sim 10^{53}$ erg) included and check whether halos below $10^7 \\, M_\\odot$ still lose essentially all their injected metals; a retained fraction in that case would show the central threshold depends on the assumed supernova energies.","supporting_citations":[{"cited_title":"2018, ApJS, 237, 13","cited_arxiv_id":null,"evidence_quote":"Supplies the Population II core-collapse supernova yields and their rotation-averaged mixture."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the critical halo mass model used to validate the masses of the first star-forming halos."},{"cited_title":"2005, ApJ, 630, 675","cited_arxiv_id":null,"evidence_quote":"Showed that low-mass halos can lose all their gas to supernova feedback, the mechanism extended here."},{"cited_title":"L., Gnedin, O","cited_arxiv_id":null,"evidence_quote":"Measured metal retention fractions in minihalos and serves as the quantitative comparison for the retention threshold."}],"review_version":1}