{"id":"4443ad62-4d5f-4552-8af1-0c2f2ae79ede","arxiv_id":"1908.05169","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Using accretion-contaminated photospheres of 16 young stars, the authors infer that 89±8% of sulfur in inner protoplanetary disks is refractory, consistent with sulfide minerals such as FeS.","lead":"A new analysis of young star photospheres shows that about 89% of sulfur in planet-forming disks is locked in refractory minerals like iron sulfide, not in gases or ices. This closes a long-standing gap in the cosmic sulfur budget and gives observers a concrete target for where sulfur should hide.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reference-composition sensitivity: open-cluster S baseline is 0.55 dex above solar, and switching references shifts fS from 89±8% to 45±22%; this systematic is unquantified.","rationale":"The paper's central claim is fS = 89±8% for the refractory fraction of sulfur in protoplanetary disks, inferred from the correlation between S and Fe in the photospheres of accreting Herbig Ae/Be stars. The model (Eqs. 1–3) requires an external reference for the initial stellar composition, and the authors adopt the mean abundance of young open cluster stars (NGC 6250 and NGC 5460). Table 1 shows the open-cluster sulfur reference is -4.33±0.28 dex, 0.55 dex above the solar value. Since all six usable disk-host S abundances are below -4.66, the model interprets this as substantial S depletion, driving fS upward. The solar-reference fit gives fS = 45±22% (Section 3), a >2σ change. This is exactly the systematic the reader flagged. The paper's defence — that the open-cluster reference is more similar to the disk hosts — is plausible but not quantitatively tested; the Bayesian evidence argument is weakened by the different numbers of fitted elements in the two runs. The most direct test is to treat the reference S abundance as a free parameter and examine its posterior; if the posterior conflicts with the open-cluster value, the central claim is not robust. Therefore the CONDITIONAL verdict is appropriate: the result is interesting but should be accepted only if the reference systematic is shown to be small. No additional concern (e.g., non-uniform δd across elements) is needed to justify the conditionality, though the small number of S detections (6) and the unexplained volatile behaviour of oxygen (fO=2±2%) also temper confidence. In summary, the reader's weakest assumption correctly identifies the most load-bearing concern.","tokens_in":14476,"tokens_out":11698,"duration_ms":101729,"concrete_test":"Re-run the Multinest fit with the reference sulfur abundance left completely free (broad uniform prior) and inspect the posterior for S_ref. If the posterior is consistent with the open-cluster value -4.33±0.28, the reference choice is validated; if it is pulled toward the solar value -4.88, then the open-cluster reference is inconsistent with the disk-host data and fS is not robust. As a secondary check, recompute fS omitting the single star (HD 144432) that has Fe near the reference but S 0.45 dex below it; if fS shifts by more than the quoted uncertainty, the result is driven by the assumed reference.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2.4 defines the reference composition (X/H)_ref as the initial bulk composition of the disk-hosting stars, assumed equal to the mean of young open cluster stars. Table 1 shows the reference sulfur abundance is log(S/H) = -4.33±0.28, which is 0.55 dex higher than the solar value (-4.88). All six usable sulfur detections in the disk-host sample (Table 1) lie at or below -4.66, i.e., 0.3–0.8 dex below the open-cluster reference. The model (Eqs. 1–2) therefore interprets these stars as strongly S-depleted, driving fS to 0.89±0.08 (Section 3). Repeating the fit with the solar reference yields fS = 0.45±0.22 (Section 3) — a >2σ shift. The paper argues the open-cluster reference is more appropriate, but it does not quantify the systematic error from this choice; the Bayesian evidence comparison (log L ≈ -1401 vs -9912) is not decisive because the solar-reference fit includes additional elements and the evidence difference is dominated by the number of observations. Because the reference composition is itself a free parameter in the main Multinest fit, the posterior for S_ref could be pulled away from the open-cluster prior; if the data prefer a lower S_ref, the central value is not robust. This is the weakest link in the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies the contaminated A-star method (CAM) to photospheric abundance measurements of 16 young disk-hosting stars, using young open cluster stars as the reference initial composition, and performs Bayesian inference with the Multinest sampler to infer the refractory fractions fS, fNa, fO, and fZn in protoplanetary disks. The main result is that 89 ± 8 % of sulfur is refractory, with a minority in gas/ice; sodium is inferred at 35 ± 16 % refractory, oxygen at 2 ± 2 %, while a solar-reference fit gives fS = 45 ± 22 %. The paper interprets the refractory sulfur as sulfide minerals such as FeS and predicts a gas-plus-ice sulfur fraction of at most 11 ± 8 % in the inner disk. Synthetic-data tests in Appendix B support the inference machinery itself.","tokens_in":14805,"tokens_out":3164,"duration_ms":34496,"significance":"If the central value is robust, this is the first quantitative measurement of the refractory fraction of sulfur in protoplanetary disks, with direct implications for planetesimal composition, disk chemistry, and the sulfur budget of the terrestrial-planet-forming zone. The method is interesting and the Bayesian machinery is carefully tested on synthetic data, which is a genuine strength. However, the headline number depends strongly on the adopted stellar reference composition, and the paper does not yet quantify this systematic uncertainty; the gap between the open-cluster and solar-reference results is larger than the quoted statistical errors. The significance of the paper therefore stands or falls on whether the reference-composition choice can be made robust.","major_comments":[{"comment":"The reference-composition choice is the main load-bearing assumption. The open-cluster sulfur reference is log(S/H) = -4.33 ± 0.28, which is 0.55 dex above the solar value of -4.88, and all six usable sulfur detections in the disk-host sample lie at or below -4.66. The model therefore interprets these stars as strongly sulfur-depleted, driving fS to 89 ± 8 %; replacing the reference with the solar composition changes the result to 45 ± 22 %, a shift of more than 2 sigma. The paper argues that the open-cluster reference is more appropriate, but it does not quantify the systematic uncertainty associated with this choice. The authors should provide a systematic error budget, repeat the fit with several independent reference populations, or otherwise demonstrate that the open-cluster reference is unbiased for these particular disk-hosting stars.","section":"Section 2.4 and Section 3, Table 1"},{"comment":"The Bayesian evidence comparison used to prefer the open-cluster reference is not decisive as presented. The two fits are based on different numbers of elements and observations because the solar-reference fit additionally includes zinc, and the quantity reported as \"log L\" is not defined as either the log-evidence or the log-likelihood. The paper acknowledges that the evidence difference is affected by the differing observation count, but then dismisses this effect as negligible without a quantitative calculation. A fair comparison would require identical datasets with and without the extra element, or a predictive check such as cross-validation on held-out abundances.","section":"Section 3"},{"comment":"The text states that \"self-consistent posteriors on the reference composition\" make the Multinest fit superior, implying that the reference abundances are free parameters in the main fit. However, Section 2.4 only specifies priors for fS, fNa, δd, and fph; it does not state the prior on the reference composition or show the resulting posterior for S_ref. If the data pull S_ref away from the open-cluster mean, the central fS value may be biased in a way that the current uncertainty statement does not capture. The authors should report the posterior on the reference abundances and test how fS changes when the reference prior is widened or shifted.","section":"Section 3 and Section 2.4"}],"minor_comments":[{"comment":"There is a typo in the sentence defining δd: \"the level of change of the refactory dust mass\" should read \"refractory dust mass\".","section":"Section 2.4"},{"comment":"The caption says \"Each stage is normalised to its own standard\" and the figure is ordered from right to left, which is confusing because the text is read left to right; explicitly labeling each bar group with its normalization and adding left-to-right ordering would improve clarity.","section":"Figure 1 caption"},{"comment":"The description of the fph prior says the Gaussian is cut off at three standard deviations or at fph ≤ 1, but it is not stated whether the cut is applied before or after the log-normal transformation and how the normalization is adjusted; please clarify.","section":"Section 2.4"},{"comment":"The synthetic-data test validates the inference machinery using a solar-like reference, but it does not test the more dangerous scenario in which the assumed reference is offset from the true initial composition by a few tenths of dex; such a test would be directly relevant to the main result.","section":"Appendix B"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially an important result, but the reference-composition sensitivity is large enough that the central claim cannot be accepted as is. I would ask the authors for a rigorous treatment of the reference systematic and a fair model comparison. I do not see a correctness problem in the core statistical model itself, but the present evidence comparison and the unspecified reference prior need to be addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is a genuine first measurement, the method is sensible, and the sulfur result is probably real. But the headline fS=89±8% is not robust against the reference composition choice, and the paper doesn't quantify that.\n\nWhat's new: they use accretion contamination of A-type stars (CAM, from their prior work) to infer that 89% of S in inner disk material is refractory, with Na at 35±16%. No one had measured refractory S in disks before. The inference is Bayesian, tested on synthetic data, and they report the alternative fit with a solar reference honestly. The predicted gas-phase S ≤11% is testable. If true, FeS as the main carrier is a useful constraint for disk chemistry and planet formation.\n\nWhat's good: they don't hide the reference sensitivity—the solar-reference fit is right there in Fig. 3 and Section 3. The odd oxygen result (fO=2±2%) is flagged as needing further work rather than buried. For what it is, the paper is honest.\n\nSoft spots. The main one is the reference composition. The open-cluster S baseline is 0.55 dex above solar, all six usable disk-host S abundances lie 0.3–0.8 dex below it, and swapping references moves the result from 89±8% to 45±22%. The paper argues the open-cluster reference is more appropriate, and that may be right, but it does not give a systematic error for that choice, and the Bayesian evidence comparison is not decisive because the two fits use different elements and datasets. That is the load-bearing caveat. Second, six usable S detections is a small sample; the result leans on a handful of stars. Third, fph and delta_d are degenerate with fX to some degree—validation on synthetic data helps, but real systematics like imperfect dust filtering are not captured by the noise model. Fourth, no public code or data; \"available on request\" is weak for a paper whose central number is this conditional.\n\nWho it's for: astrochemists and planet formation modelers. It deserves a serious referee. I would send it out, but ask the authors to quantify the reference systematic, release the data and code, and soften the abstract claim so the reader sees that 89% is conditional on the reference choice. If the solar-reference test is truly a worse fit, they need to show that in a way that does not depend on poorly comparable evidence values.","headline":"First real measurement of refractory sulfur in planet-forming disks, with a plausible 89% central value; the caveat is that the result shifts to 45±22% with a solar reference, so the headline number needs a systematic-error qualifier.","tokens_in":15359,"tokens_out":2287,"would_cite":true,"duration_ms":22322,"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":"This paper measures, for the first time, that (89 ± 8)% of sulfur in planet-forming disks is locked in refractory solids such as FeS, not in gas or ice.","keywords":["sulfur","protoplanetary disks","refractory fraction","dust trapping","Herbig Ae/Be stars","stellar photosphere abundances","sulfide minerals","planet formation"],"falsifier":"A direct ALMA measurement of gas-phase sulfur carriers (H2S, CS, SO) in the inner disk of a dust-trapping system that finds sulfur in gas and ice well above the paper's predicted 11% of total would falsify the refractory-dominated picture; alternatively, expanding the open cluster reference sample to more coeval clusters and recomputing fS, if the cluster mean sulfur abundance shifts by more than ~0.1 dex, the 89% central value would not survive.","tokens_in":14279,"feed_emoji":"🪨","tokens_out":5620,"duration_ms":53786,"temperature":0.7,"pith_summary":"This paper reports the first measurement of the fraction of sulfur locked in solid, refractory form inside planet-forming disks. By comparing the surface compositions of 16 young disk-hosting stars with a reference sample of disk-free open cluster stars, and using the fact that early-type stars' photospheres record recently accreted disk material, the authors infer that (89 ± 8)% of elemental sulfur in the inner disk is refractory, with only about 11% in gas or ice. They also measure a lower refractory fraction for sodium, (35 ± 16)%, and find oxygen behaves almost entirely as a volatile, with fO = (2 ± 2)%. The result matters because sulfur's main reservoir in disks had previously escaped detection: gaseous molecules account for less than 1% of total sulfur, and the new measurement identifies the missing reservoir as refractory and consistent with sulfide minerals such as FeS.","feed_headline":"89% of disk sulfur is locked in rock, not gas","feed_subtitle":"First measurement of sulfur's reservoir in planet-forming disks points to sulfide minerals like FeS as the main carrier.","key_machinery":"The load-bearing mechanism is the Contaminated A-stars Method (CAM): in stars more massive than about 1.4 M_sun, the radiative envelope mixes slowly, so the photosphere (only ~1e-10 M_sun) can be almost entirely replaced on timescales of days by disk accretion, making the stellar surface a direct sample of inner-disk material. Dust trapping at a radial pressure bump (e.g., planet-induced) removes large grains before accretion, so the refractory component of each element is scaled by a per-star depletion factor δd, while the volatile component accretes freely. The abundance of element X in accreted material is $(X/\\mathrm{H})_{\\rm disk} = [(1-f_X) + f_X \\delta_d] \\times (X/\\mathrm{H})_{\\rm ref}$, with f_X the refractory fraction; combining this with the photospheric mixing fraction fph and fitting fS and fNa globally with Bayesian Multinest sampling extracts the refractory fractions from the iron–sulfur correlation.","core_discovery":"The central discovery is that sulfur in the terrestrial-planet-forming zone of protoplanetary disks is predominantly refractory: fS = (89 ± 8)%. The inference comes from the correlation between sulfur and iron abundances in accretion-contaminated photospheres of Herbig Ae/Be stars, using dust trapping at pressure bumps to fractionate the accreting material. Because the carrier must be far more refractory than water ice, the authors identify sulfide minerals, most plausibly FeS, as the main reservoir, with sulfur chains S_n playing at most a minor role. As a consistency check, imposing the condensation-model expectation fS < fNa leaves fS unchanged at 89% while shifting fNa to 97%, and an orthogonal distance regression fit using a solar reference gives fS = (75 ± 8)%. The result is stated as the first measurement of the refractory fraction of sulfur in protoplanetary disks.","pith_inferences":["If the 89% figure holds, models of prebiotic chemistry and habitable-zone volatile delivery should treat sulfur as a rock-derived element in the inner disk, with volcanic and impact processing, not ice sublimation, as its main pathway into atmospheres.","The same accretion-contamination technique could be extended to carbon, phosphorus, and chlorine, whose refractory fractions are poorly known; those elements' fX values would similarly be pinned down by correlations with iron in the same sample.","The large shift between open-cluster and solar reference compositions (89% vs 45%) implies the absolute value is sensitive to Galactic chemical evolution corrections; applying an age- and position-matched chemical evolution model could tighten or move the central value."],"forward_implications":["In the inner few to ten astronomical units around ~2–3 M_sun stars, almost all sulfur is available to be incorporated directly into rocky planetesimals, rather than being delivered later as ice.","Gas-phase and ice sulfur in the planet-forming zone is predicted to be at most (11 ± 8)% of the total, a concrete target for ALMA searches for H2S, CS, and SO.","The refractory carrier must survive temperatures well above 150 K, ruling out H2S, OCS, SO, and SO2 ices as main reservoirs and favouring FeS and related sulfide minerals.","In hot Jupiter atmospheres formed by core accretion of gas alone, little H2S is expected; significant H2S would point to late planetesimal accretion.","The solar-system pattern—volatile sulfur in cometary ices, sulfur in rocky meteorites as sulfides and chains—is a natural outcome of a general disk process rather than a local anomaly."],"supporting_citations":[{"why":"Established that dust-depleted inner disks correlate with refractory element depletion on the stellar surface, providing the observational basis for the method.","marker":"Kama et al. 2015"},{"why":"Supplies the calculation of photospheric contamination fraction fph and the mixing timescales the method depends on.","marker":"Jermyn & Kama 2018"},{"why":"Provides the stellar surface abundance measurements for the disk-hosting sample.","marker":"Folsom et al. 2012"},{"why":"One of the two open cluster reference abundance datasets (NGC 6250).","marker":"Martin et al. 2017"},{"why":"The other open cluster reference abundance dataset (NGC 5460).","marker":"Fossati et al. 2011"},{"why":"Solar composition used for the alternative reference and for zinc.","marker":"Asplund et al. 2009"},{"why":"Supplies condensation temperatures and the 50% condensation temperature of sulfur (664 K) used to classify element volatility.","marker":"Lodders 2003"},{"why":"Shows planet-induced pressure bumps trap dust, the mechanism assumed for fractionation.","marker":"Pinilla et al. 2012"}],"fun_headline_variants":["Sulfur in planet-forming disks is 89% rock, not gas","89% of disk sulfur is locked in minerals, not ice","First measure of sulfur in disks: 89% is in solids","Sulfur's main disk reservoir is sulfides, not gas","Disk sulfur: 89% is in refractory minerals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result rests on the assumption that the disk-hosting stars and the young open cluster stars used as a reference have the same initial bulk composition; if the reference is replaced by solar abundances, the inferred refractory sulfur fraction drops from 89% to 45%.","fun_headline_variants_meta":{"raw":{"variants":["Sulfur in planet-forming disks is 89% rock, not gas","89% of disk sulfur is locked in minerals, not ice","First measure of sulfur in disks: 89% is in solids","Sulfur's main disk reservoir is sulfides, not gas","Disk sulfur: 89% is in refractory minerals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001549,"raw_usage":{"total_tokens":6165,"prompt_tokens":888,"completion_tokens":5277,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":5190}},"tokens_in":504,"tokens_out":5277,"duration_ms":40887,"temperature":1.0,"reasoning_tokens":5190,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:21:39.435639+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct ALMA measurement of gas-phase sulfur carriers (H2S, CS, SO) in the inner disk of a dust-trapping system that finds sulfur in gas and ice well above the paper's predicted 11% of total would falsify the refractory-dominated picture; alternatively, expanding the open cluster reference sample to more coeval clusters and recomputing fS, if the cluster mean sulfur abundance shifts by more than ~0.1 dex, the 89% central value would not survive.","supporting_citations":[{"cited_title":"P., Bagnulo , S., Wade , G","cited_arxiv_id":null,"evidence_quote":"Provides the stellar surface abundance measurements for the disk-hosting sample."},{"cited_title":"J., Stift , M","cited_arxiv_id":null,"evidence_quote":"One of the two open cluster reference abundance datasets (NGC 6250)."},{"cited_title":"P., Bagnulo , S., et al","cited_arxiv_id":null,"evidence_quote":"The other open cluster reference abundance dataset (NGC 5460)."}],"review_version":1}