{"id":"5530eca1-c51f-487b-a1be-ebcbaa5b9e19","arxiv_id":"2505.10206","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"Enhanced stellar winds near the Eddington limit in very massive stars change their temperatures, shrink final black holes, reduce binary mergers, and can produce 30-40 solar-mass black hole companions in LMC-like binaries.","lead":"Very massive stars lose mass through winds, and this paper adds stronger, Eddington-sensitive wind models to stellar simulations to see how they change the stars' evolution and the black holes they leave behind. The results are relevant because they shift which star-formation paths can explain heavy black holes detected by gravitational-wave observatories.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Calibration-validation overlap: Mdote transition parameters were fitted to Tarantula Of/WNh stars, the same stars used for the R136 age-match validation; the central observational support is not independent.","rationale":"The reader's weakest_assumption identifies exactly the concern I find most load-bearing. The central claim in the abstract and in Sec. 3.1 is that Gamma_e-enhanced single-star tracks match observed VMS properties better than standard winds. The evidence is the param fit in Table 3, where only Mdote yields R136 ages of 1.3-1.6 Myr. That inference is only as strong as the input prescription. The transition parameters entering Eqs. A.4-A.5 and Eq. 5 were calibrated by Sabhahit et al. (2022) from the average luminosity and terminal velocity of Of/WNh stars in the Tarantula Nebula, and the validation targets, R136a1-a3 from Brands et al. (2022) and Schneider et al. (2018), are drawn from that same cluster and population. Thus the 'better match' is not an independent confirmation; it is partly a self-consistency check. The paper does not perturb the calibrated constants, does not perform a leave-one-out test, and does not apply the recipe to an independent cluster such as the Arches or to a different metallicity, so the circularity remains unresolved. I considered but did not choose as the single concern the unvaried binary and remnant prescriptions behind the BBH mass claims in Secs. 3.5-3.6. Those are important caveats, but the paper's novel input is the wind prescription; if its calibration is not independently supported, the binary results inherit the same weakness. Conversely, if the calibration is validated independently, the binary differences are a plausible and interesting consequence. The paper also omits released tracks and parameter files, which hampers reproduction, but that is a reproducibility issue rather than the central logical vulnerability. The remainder of the modeling is internally coherent, and the rotating-track appendix (App. C) supports the claim that winds dominate VMS evolution for MZAMS >= 200 Msun. The appropriate verdict remains CONDITIONAL: the modeling is careful, but the headline observational match needs an independent-calibration check before it can be accepted as strong evidence. I therefore keep the reader's CONDITIONAL verdict, reflected here as UNCHANGED.","tokens_in":31994,"tokens_out":7083,"duration_ms":70152,"concrete_test":"Leave-one-out recalibration of the wind transition: (1) Obtain the Sabhahit et al. (2022) Of/WNh sample and recompute Gamma_e,trans and log Mdote_trans from the subset that excludes R136a1 (and, if possible, R136a2 and R136a3), rederiving f consistently. (2) Recompute the PARSEC Mdote tracks from 100 to 600 Msun at Z = 0.006 with the recalibrated transition parameters. (3) Rerun the param fits for R136a1, R136a2, and R136a3. If the inferred ages remain in the 1.3-1.6 Myr range and the Mdote-vs-Mrdw/ML/M ranking persists, the circularity concern is alleviated. If the ages shift significantly or the ranking inverts, the claimed observational validation is not independent evidence for the enhanced-wind recipe.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's key observational support for the new winds is the claim in Sec. 3.1 and Table 3 that only the Mdote tracks reproduce R136a1, R136a2, and R136a3 with ages of ~1.3-1.6 Myr, while standard winds and Mdote_L/M give ages below 1 Myr. However, the Mdote prescription is not an independent physical model: its transition point is empirically calibrated to the same population used for validation. Sec. 2.4.2 states that Sabhahit et al. (2022) obtained Gamma_e,trans = 0.42 and log Mdote_trans = -5.0 from the average luminosity, terminal velocity, effective temperature, and surface hydrogen fraction of Of/WNh stars in the Tarantula Nebula; R136a1-a3 are WNh stars in exactly that cluster and are the validation targets in Sec. 3.1. Eq. 5 sets the transition mass-loss rate from these observed L and v_inf values with the f = 0.6 correction, and Eqs. A.4-A.5 encode that calibration. The 'better match' with Mdote tracks is therefore partly a check that the code reproduces the calibration anchor, not an independent confirmation that VMS winds obey the enhanced recipe. The paper does not test alternative values of f, Gamma_e,trans, or v_inf, nor does it apply the recipe to an independent cluster or metallicity, nor perform a leave-one-out recalibration. Section 4 lists caveats about post-MS winds, rotation, PI-gap edges, and CE physics, but does not flag this calibration-validation overlap. Because the BH-population results in Secs. 3.5-3.6 are downstream of the same wind input, their credibility inherits the same unresolved issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper implements two Eddington-parameter-enhanced mass-loss recipes (Mdot_Gamma_e and Mdot_L/M, based on Sabhahit et al. 2022) into the PARSEC v2.0 stellar evolution code, computes single-star tracks for 100-600 Msun at LMC metallicity (Z=0.006), and uses them to interpret the Tarantula Nebula VMS population, in particular the R136a stars. The same tracks are then incorporated into the SEVN binary population synthesis code to study wind accretion, stellar mergers, BH remnant masses, and Hubble-time merging BBH populations. The main claims are (i) only the Mdot_Gamma_e tracks reproduce the R136a stars with ages ~1.3-1.6 Myr, consistent with independent cluster age estimates; (ii) enhanced winds suppress main-sequence mergers and reduce BH masses above the pair-instability gap; (iii) for merging BBHs, enhanced winds produce more primaries above 30 Msun and secondaries in the 30-40 Msun range, a feature absent with standard winds at LMC metallicity.","tokens_in":32327,"tokens_out":12174,"duration_ms":112567,"significance":"If the single-star validation were independent, the paper would provide strong evidence that Eddington-enhanced winds are required for VMS and would have direct implications for the interpretation of R136a and for LIGO/Virgo source modeling. The paper's strengths are the transparent description of the input physics, the use of public codes (PARSEC and SEVN), the inclusion of rotating tracks in an appendix, and the sharply formulated, falsifiable predictions for BH mass distributions and merger properties. However, the main observational support is weakened by the calibration-validation overlap: the transition parameters of the Sabhahit et al. recipe were fitted to the same Tarantula Of/WNh population later used for validation. The binary population synthesis results remain useful as a demonstration of sensitivity to wind physics, but they inherit the same calibration dependence.","major_comments":[{"comment":"The transition parameters Gamma_e,trans=0.42 and log Mdot_trans=-5.0 were obtained by Sabhahit et al. (2022) from average L, v_inf, Teff, and Xs of Of/WNh stars in the Tarantula Nebula, and the f=0.6 correction in Eq. (5) is also metallicity-calibrated. The validation in Sec. 3.1 uses the same stars (R136a1-a3 are WNh stars in that cluster) to conclude that only Mdot_Gamma_e reproduces ages ~1.3-1.6 Myr. This is therefore not an independent test of the enhanced-wind recipe: the tracks partly recover the calibration anchor. The paper should either demonstrate out-of-sample performance (e.g., Arches cluster, or a leave-one-out recalibration), test the sensitivity to f, Gamma_e,trans, and v_inf, or explicitly re-frame the single-star match as a consistency check rather than a validation. This point is load-bearing because the abstract and conclusion state that the enhanced tracks 'match observed VMS properties better.'","section":"Sec. 2.4.2 vs Sec. 3.1 (Eqs. 5, A.4, A.5; Table 3)"},{"comment":"The new recipes Mdot_Gamma_e and Mdot_L/M are defined by the max() switches in Eqs. (8)-(9), which do not include the Sander et al. (2019) Wolf-Rayet mass-loss prescription, whereas the standard recipe Mdot_rdw does. The paper does not state whether the Sander rate is still applied when a star with the new winds becomes a WR (Xs<0.3). If it is not, then the comparison between 'standard' and 'enhanced' winds conflates two effects: the Sabhahit enhancement on the O/WNh phase and the removal of the Sander WR rate in the later phases. This could explain, for example, why the Mdot_Gamma_e 200 Msun track retains a He-rich envelope (Sec. 3.2) while the Mdot_rdw track does not. The authors should clarify this point and, if Sander is not used in the new recipes, justify that choice or recompute the tracks with it included.","section":"Sec. 2.4.2 / Eqs. (8)-(9) vs Sec. 2.4.1"},{"comment":"The paper states that only the Mdot_Gamma_e tracks are consistent with the R136 cluster age, but it does not provide a formal model comparison (e.g., posterior probabilities or goodness-of-fit) among the three wind recipes. Given that the best-fit M_ZAMS for R136a1 with Mdot_Gamma_e is 389(+1,-64) Msun, near the upper end of the grid, a quantitative comparison would also clarify how strongly the data prefer Mdot_Gamma_e over the other recipes. The qualitative age argument is reasonable, but a statistical measure would strengthen the central observational claim.","section":"Sec. 3.1 / Table 3"}],"minor_comments":[{"comment":"The sentence contains a duplicated phrase: 'to estimate to estimate stellar parameters of interest'.","section":"Sec. 2.2"},{"comment":"In the description of the initial binary population, '10 7 binaries' should read '10^7 binaries'.","section":"Sec. 2.5"},{"comment":"The conclusion bullet in Sec. 5 refers to 'the new winds, Mdot_Gamma_e and Mdot_rdw'; this should be 'Mdot_Gamma_e and Mdot_L/M'.","section":"Sec. 3.1 and Sec. 5"},{"comment":"For R136a2 with Mdot_rdw, the fitted current mass (255 Msun) exceeds the ZAMS mass (240 Msun), which is inconsistent with single-star mass loss; please check the value and the fitting output.","section":"Table 3 and Table B.1"},{"comment":"The phrase 'Our results establish a strict upper limit on the possible initial mass for R136a1 since no models with an initial mass below 300 Msun fit the data' is misworded: the stated reasoning gives a lower limit on the initial mass, while the upper limit (about 390-400 Msun) follows from the best-fit mass and its uncertainty; please rephrase for clarity.","section":"Sec. 3.1"},{"comment":"The caption says 'three different rotation rates' but the table lists four columns (ω = 0.0, 0.4, 0.6, 0.8).","section":"Table C.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is careful in its modeling and the binary population synthesis results are interesting, but the single-star observational validation is not independent of the wind-recipe calibration. Please consider asking the authors to (i) explicitly discuss this calibration-validation overlap in Section 4, (ii) test at least a small grid of f and Gamma_e,trans variations, and (iii) clarify the treatment of the Sander WR mass-loss rate in the new recipes. The paper would be acceptable after these revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this if you care about VMS evolution or the progenitors of massive BH mergers. The genuinely new content is the PARSEC v2.0 implementation of the Sabhahit et al. enhanced-wind prescriptions and the first coupling of those tracks to SEVN binary population synthesis. The paper is technically careful: the track physics is described in enough detail to be reproduced, three wind recipes are compared consistently, and the binary results—fewer main-sequence mergers, limited BH formation above the pair-instability gap, 30-40 Msun secondaries in Hubble-time mergers—follow coherently from the stronger mass loss. I find the internal logic solid. If VMS winds are that strong, stars stay hot and compact, avoid Roche-lobe contact, and the remnant spectrum shifts the way they describe.\n\nThe soft spot is the one the stress-test note identifies, and I think it lands. Sec. 2.4.2 states that the transition parameters (Gamma_e,trans = 0.42, log Mdot_trans = -5.0, plus the f=0.6 correction) were fitted to the average properties of Of/WNh stars in the Tarantula Nebula. Sec. 3.1 then validates the Mdote tracks against R136a1-a3, which are WNh stars in that same cluster. That means the 'better match' is partly a consistency check, not independent confirmation. The paper does not test alternative calibrations, does not apply the recipe to another cluster or metallicity, and does not flag this overlap in the caveats section. The binary and GW predictions inherit the same uncertainty because they use the same calibrated input. I don't think this is disqualifying—Sabhahit et al. is a reasonable physically motivated prescription and the authors are clear about where it comes from—but the validation claim should be softened or supported with a sensitivity run.\n\nOther issues are minor and mostly already acknowledged: no released tracks or data products, unvaried fiducial CE and mass-transfer treatments, one metallicity. The paper's own discussion covers most of this, just not the calibration-validation overlap.\n\nWho is this for? Anyone modeling VMS or interpreting GW events at LMC-like metallicity. The paper deserves a serious referee. I'd send it to review, and I'd ask the authors to release the tracks and test how much the BH spectrum and merger predictions shift when the wind calibration parameters are varied.","headline":"A careful implementation of Sabhahit-style enhanced winds with genuinely new binary-population predictions, but the single-star validation is partly circular because the prescription was calibrated to the same Tarantula stars.","tokens_in":33017,"tokens_out":3110,"would_cite":true,"duration_ms":30071,"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":"Eddington-sensitive winds explain the Tarantula Nebula's most massive stars and reshape black hole mergers.","keywords":["very massive stars","stellar mass loss","Eddington parameter","black hole remnants","binary population synthesis","pair-instability mass gap","Tarantula Nebula","R136"],"falsifier":"Direct measurements of mass-loss rates for very massive stars in the Tarantula Nebula, using X-ray, radio, or H-$\\alpha$ emission, that do not show the steep $L^{4.77}$ dependence would falsify the prescription. A second test: the two variants diverge in their dependence on surface hydrogen abundance ($\\dot{M}_{\\Gamma_e}$ falls steeply as $X_s$ decreases while $\\dot{M}_{L/M}$ does not), so a sample of WNh stars spanning $X_s$ from about 0.3 to 0.75 would discriminate between them. A third test would be the same comparison at a different metallicity, such as an SMC cluster, where the $Z^{0.5}$ scaling of the recipe can be checked.","tokens_in":31704,"feed_emoji":"⭐","tokens_out":6469,"duration_ms":62165,"temperature":0.7,"pith_summary":"This paper argues that the standard wind recipes fail for stars above roughly 100 solar masses and that a new, Eddington-sensitive prescription is needed. It implements the Sabhahit et al. enhanced mass-loss rates, which depend steeply on the Eddington parameter and the luminosity-to-mass ratio, into the PARSEC stellar code and the SEVN binary population synthesis code. With these winds, single-star tracks stay hot and compact and spend about 1.6 Myr in the narrow temperature range of observed Tarantula Nebula very massive stars, and the fitted ages of R136a1, a2, and a3 agree with the cluster age, which standard winds cannot do. In binaries, enhanced winds suppress main-sequence stellar mergers and change the remnant spectrum: black hole binaries that merge within a Hubble time develop more primaries above 30 solar masses and secondaries in the 30-40 solar mass range at LMC metallicity. If correct, the standard wind recipe is insufficient for very massive stars, and high-mass black hole mergers can form at low metallicity without invoking extra dynamical channels.","feed_headline":"Eddington winds explain the biggest stars in the Tarantula","feed_subtitle":"The same mass-loss switch that keeps Tarantula's giants hot also shifts which black holes merge within a Hubble time.","key_machinery":"The load-bearing object is the Sabhahit et al. (2022) enhanced mass-loss prescription in its two variants, $\\dot{M}_{\\Gamma_e}$ and $\\dot{M}_{L/M}$, which scale as $\\dot{M} \\propto M^{0.78} \\Gamma_e^{4.77}$, equivalently $L^{4.77}/M^{3.99}$ with a surface-hydrogen or metallicity factor. The machinery uses a model-independent transition mass-loss rate $\\dot{M}_{\\rm trans} = f L_{\\rm trans}/(v_\\infty c)$ with $f = 0.6$, empirically pinned at $\\Gamma_{e,\\rm trans} = 0.42$ and $\\log \\dot{M}_{\\rm trans} = -5.0$ from the Tarantula Of/WNh stars. The code applies the maximum of the Vink et al. (2001) rate and the new rate, automatically switching at the transition point, and this steep dependence is what keeps tracks hot and compact on the main sequence and drives all downstream binary and remnant differences.","core_discovery":"The central claim is that very massive stars near the Eddington limit lose mass at rates far above the standard Vink et al. predictions, and that taking this seriously reproduces observed very massive star properties and changes predicted black hole populations. Using the Sabhahit et al. (2022) prescription, with a transition mass-loss rate calibrated on the Of/WNh stars of the Tarantula Nebula, the authors compute single-star tracks at LMC metallicity and embed them in the SEVN binary code. The Eddington-parameter-sensitive variant ($\\dot{M}_{\\Gamma_e}$) fits R136a stars with ages of about 1.3 to 1.6 Myr, consistent with independent age estimates, while standard winds fitting the same stars give implausibly young ages below 1 Myr. In binary evolution, the enhanced winds keep stars compact so fewer stars fill their Roche lobes; stellar mergers become rarer and black holes above the lower edge of the pair-instability gap (about 50 solar masses) are suppressed. The most striking population consequence is that among black hole binaries merging within a Hubble time, the enhanced-wind models yield many more primaries above 30 solar masses and secondaries at 30 to 40 solar masses, a range that standard winds at LMC metallicity do not produce, matching gravitational-wave detections of roughly 40-solar-mass secondaries.","pith_inferences":["The difference between the two wind variants offers a ready-made observational test: if WNh mass-loss rates track surface helium abundance, $\\dot{M}_{\\Gamma_e}$ is the better description, and if they do not, $\\dot{M}_{L/M}$ is preferred; the paper leaves this distinction to future data.","Extending the same recipe to solar metallicity, where the $Z^{0.5}$ scaling makes the enhanced winds even stronger, would likely suppress the massive-black-hole channel entirely at high metallicity, a consequence the LMC-only analysis does not test.","The paper's result that single-star and merger origins give similar current masses for R136a implies that current stellar mass alone cannot identify the formation route; post-merger surface composition, such as helium enrichment, could discriminate, but the authors note they cannot yet model it.","Because enhanced winds widen orbits and reduce Roche-lobe filling, the merger rate density of very massive binaries should drop; population-synthesis predictions for gravitational-wave event rates would shift accordingly."],"forward_implications":["R136a1's zero-age main-sequence mass is capped near 400 solar masses if it formed as a single star, and near 300 solar masses if a binary merger made it, lowering the implied upper end of the initial mass function.","Enhanced winds suppress main-sequence stellar mergers, so merger-driven formation of very massive stars is rarer at low metallicity; observed stars like R136a are more likely single-born or formed through stable Roche-lobe overflow pathways.","The pair-instability mass gap's lower edge near 50 solar masses acts as a harder ceiling for black hole production when winds are enhanced: stars that would form black holes above 100 solar masses under standard winds instead leave remnants of about 30-50 solar masses or end as pair-instability supernovae.","Black hole binaries merging within a Hubble time at LMC metallicity show more primaries above 30 solar masses and secondaries of 30-40 solar masses, so gravitational-wave events with roughly 40-solar-mass secondaries need not require dynamical formation or hierarchical mergers.","The $\\dot{M}_{\\Gamma_e}$ tracks give self-consistent stellar ages for R136 of about 1.3-1.6 Myr, supporting a cluster age near 1-2.5 Myr and resolving the too-young age problem that standard winds produce."],"supporting_citations":[{"why":"Supplies the two enhanced-wind prescriptions (Eqs. A.4 and A.5) and the empirical transition parameters that drive all the differences studied in the paper.","marker":"Sabhahit et al. (2022)"},{"why":"Provides the standard hot-star mass-loss recipe (Eq. A.1) that serves as the low-Eddington baseline and the comparison point for the new winds.","marker":"Vink et al. (2001)"},{"why":"Gives the high-Eddington scalings ($M^{0.78}\\Gamma_e^{4.77}$) on which the enhanced Sabhahit et al. recipe is built.","marker":"Vink et al. (2011)"},{"why":"Supplies the metallicity-dependent correction factor $f = 0.6$ used in the transition mass-loss rate.","marker":"Vink & Gräfener (2012)"},{"why":"Provides the observational luminosities, effective temperatures, and mass estimates of R136a1, a2, and a3 used for the Bayesian fitting.","marker":"Brands et al. (2022)"},{"why":"Supplies the VLT-FLAMES Tarantula Survey data of over 400 stars that define the observed effective-temperature range of very massive stars.","marker":"Schneider et al. (2018)"},{"why":"Provides the initial mass and age estimates of R136a stars against which the fitted masses and ages are compared.","marker":"Crowther et al. (2010)"},{"why":"Provides the pair-instability, pulsational pair-instability, and remnant-mass formalism used in SEVN to convert pre-supernova properties into black hole masses.","marker":"Spera & Mapelli (2017)"}],"fun_headline_variants":["Eddington winds reshape black hole masses from massive stars","Enhanced mass loss near Eddington limit alters binary evolution","Tarantula's brightest stars explained by Eddington-parameter winds","Near-Eddington winds suppress stellar mergers, boost BH masses"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Sabhahit et al. enhanced-wind prescription, including the empirically calibrated transition point (Gamma_e,trans = 0.42 and log Mdot_trans = -5.0) and the f = 0.6 correction factor, is a valid transferable description of very massive star mass loss, even though it was fitted to the same Tarantula Nebula Of/WNh stars later used for validation; if those calibration values are unrepresentative, the observational agreement is not independent evidence.","fun_headline_variants_meta":{"raw":{"variants":["Eddington winds reshape black hole masses from massive stars","Enhanced mass loss near Eddington limit alters binary evolution","Tarantula's brightest stars explained by Eddington-parameter winds","Near-Eddington winds suppress stellar mergers, boost BH masses"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000282,"raw_usage":{"total_tokens":1799,"prompt_tokens":1209,"completion_tokens":590,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":825,"completion_tokens_details":{"reasoning_tokens":519}},"tokens_in":825,"tokens_out":590,"duration_ms":6159,"temperature":1.0,"reasoning_tokens":519,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:15:22.162796+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Direct measurements of mass-loss rates for very massive stars in the Tarantula Nebula, using X-ray, radio, or H-$\\alpha$ emission, that do not show the steep $L^{4.77}$ dependence would falsify the prescription. A second test: the two variants diverge in their dependence on surface hydrogen abundance ($\\dot{M}_{\\Gamma_e}$ falls steeply as $X_s$ decreases while $\\dot{M}_{L/M}$ does not), so a sample of WNh stars spanning $X_s$ from about 0.3 to 0.75 would discriminate between them. A third test would be the same comparison at a different metallicity, such as an SMC cluster, where the $Z^{0.5}$ scaling of the recipe can be checked.","supporting_citations":[],"review_version":1}