{"id":"a062cca0-827e-467d-b842-3a81d9aad48b","arxiv_id":"2507.11351","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Carbon abundances derived from CH 4300 A bands shift by up to 0.8 dex depending on which synthetic spectrum grid is used, even though internal precision and band-to-band agreement are excellent.","lead":"This paper measures carbon abundances in stars from CH molecular bands using two different synthetic spectrum grids, finding results for the same stars can differ by up to 0.8 dex depending on the grid. The X-shooter Spectral Library is a benchmark for stellar population models, so this caveat matters for any galaxy modeling that depends on carbon abundances.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The grid-to-grid [C/Fe] offset is real, but the paper's attribution to hidden carbon-model defects is not established: GAUGUIN's grid-tied pseudo-continuum normalisation alone can produce the same symptom, and the internal tests cannot separate these causes.","rationale":"The reader's weakest_assumption is directly on target: the grid offset is observed, but its cause is not identified. I agree with that framing. The paper's central empirical result—that [C/Fe] from CH at R≈10000 is strongly model-dependent—does not depend on the cause and is well supported by the two-grid comparison, the Mg control, and the solar/Arcturus tests. The conditional verdict is therefore appropriate: the paper should state the carbon-model attribution as a hypothesis, add a grid/normalisation systematic to the error budget, and avoid calling the catalogue 'unbiased' without qualification. My proposed test would settle the main ambiguity. No additional concerns warrant rejection: the data handling appears careful, the paper is transparent about its inability to decide which grid is right, and the literature comparison is honest about the heterogeneity of published [C/Fe] values.","tokens_in":24560,"tokens_out":8786,"duration_ms":110836,"concrete_test":"Pick ~30 XSL stars spanning the full range of BOSZ−Knowles offsets (≈0.05 to 0.8 dex). Normalise each observed spectrum once, without reference to either synthetic grid, by fitting a low-order spline through continuum windows in 4290–4315 Å that avoid strong CH absorption (or by using a high-resolution observed template). Feed this same common-normalisation spectrum into GAUGUIN with both the Knowles+21 and BOSZ grids. If the grid offset collapses to ≲0.2 dex, the discrepancy is a normalisation artifact rather than a carbon-model defect; if the offset persists at ≳0.5 dex, the carbon-model attribution is supported. As a cross-check, repeat on the HARPS/Vesta solar spectrum at R≈110000, where continuum placement is nearly unambiguous, and compare both grids' CH results with the CI-line value [C/Fe]≈+0.14.","verdict_should_be":"UNCHANGED","load_bearing_attack":"GAUGUIN's normalisation step is reference-grid dependent (Sect. 3.2, built on Santos-Peral et al. 2020): the observed spectrum is divided by a polynomial fit to the Synthetic/Observed residual before [C/Fe] is measured. If the Knowles+21 and BOSZ/MARCS synthetic spectra differ in broad line wings or in pseudo-continuum depression across the crowded 4300 Å region, those differences are partially absorbed into the normalisation, biasing each grid's zero-point. The paper's validation (band-to-band agreement, Monte Carlo parameter uncertainties, solar/Arcturus fits, [Mg/Fe] control) is performed within a single grid or on lines with a cleaner continuum; it therefore establishes precision, not the accuracy of the zero-point. The solar test is the clearest warning: the same solar spectrum yields [C/Fe]≈+0.12 with Knowles+21 and ≈−0.52 with BOSZ, a 0.64 dex shift that cannot be attributed to stellar parameters. Section 4.4's conclusion that 'hidden issues in carbon modeling' cause the grid offset is a plausible hypothesis but is not established, because the alternative—grid-tied normalisation—produces exactly the observed internally consistent, precise, but offset results. A second uncontrolled degree of freedom (the unspecified BOSZ microturbulence value ingested into GAUGUIN, given the different vmic prescriptions listed in Sect. 3.1.3) reinforces the need for a test that isolates the synthetic carbon models from the measurement pipeline.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper measures [C/Fe] abundances for ~200 stars of the X-shooter Spectral Library from two CH bands near 4300 Å using the GAUGUIN spectral-synthesis code, adopting two different 5D reference synthetic grids (Knowles et al. 2021 and the updated BOSZ/MARCS library of Mészáros et al. 2024). Within each grid the results appear precise and internally consistent: band-to-band agreement of ~0.01–0.03 dex, small Monte Carlo parameter uncertainties (Table 2), no trends with Teff or log g (Fig. 4), and a clean [Mg/Fe] control test (Appendix B). However, the two grids give systematically offset [C/Fe] values for the same stars, up to |Δ[C/Fe]| ~ 0.8 dex, with the solar spectrum yielding +0.12 (Knowles+21) versus −0.52 (BOSZ) and Arcturus +0.28 versus +0.46. The paper's central conclusion is that this offset reflects hidden problems in the carbon modeling of the crowded CH 4300 Å region, so that CH-based carbon measurements can be inaccurate without being detected by internal quality tests.","tokens_in":24779,"tokens_out":12579,"duration_ms":140092,"significance":"If the result holds, the paper delivers a genuinely useful warning to the stellar-population community: XSL is a benchmark empirical library, carbon is a key missing ingredient in population models, and a demonstrated 0.6–0.8 dex grid-dependence of CH-band [C/Fe] that is invisible to standard internal checks would explain a large part of the scatter among literature carbon abundances. The comparative design is a real strength: the [Mg/Fe] control (Appendix B) is clean (|Δ[Mg/Fe]| ~ 0.02 dex between grids), the internal precision tests (Table 2; Figs. 5 and 8) are convincing, and the authors honestly state in Sect. 4.3 that no objective criterion decides which grid is more accurate. The main gap is that precision is repeatedly presented as if it implied accuracy: the solar and Arcturus benchmarks favour different grids, and the only atomic-line anchor (Fig. 12) was run with a single grid. With the causal attribution either tested or weakened to an explicit model-dependence warning, the revised paper would be a valuable cautionary contribution.","major_comments":[{"comment":"The conclusion that the grid-to-grid offset is caused by \"hidden issues in the carbon modeling\" of the CH 4300 Å region is not established, because GAUGUIN's normalisation step is reference-grid dependent: as described in Sect. 3.2, the observed spectrum is divided by a polynomial fit to the Synthetic/Observed residual computed against the reference grid, so broad-band differences between the grids (line wings, pseudo-continuum depression in this crowded region) can be absorbed into each grid's own continuum zero-point. All internal validation tests (band-to-band agreement, Monte Carlo parameter uncertainties, absence of Teff/log g trends, and the Appendix B [Mg/Fe] control) are performed within a single grid or on a region with a cleaner continuum; they establish precision, not zero-point accuracy. Two facts in the paper point directly to this ambiguity: the Sun favours Knowles+21 (+0.12 vs. −0.52) while Arcturus favours BOSZ (+0.46 vs. +0.28, against a literature value of +0.43), which is inconsistent with a single grid-wide carbon-model error; and the high-resolution CI-line anchor (Fig. 12) was computed only with the Knowles+21 grid, so it cannot show whether the BOSZ offset is specific to CH or common to all carbon lines. To support the stated conclusion the authors should either weaken it to a model-dependence warning or add a test that isolates the synthesis models from the normalisation, e.g., fitting synthetic spectra drawn from one grid after normalising them with the other grid's residual polynomial.","section":"§4.4 and §5, with §3.2"},{"comment":"The manuscript describes the BOSZ grid as providing spectra at four fixed microturbulence values (0, 1, 2, and 4 km/s) but never states which value was adopted for the 5D subset ingested into GAUGUIN. Given that the paper itself notes (Sect. 3.1.3) that microturbulence can affect line strengths in this region by 1–2 Å, an unspecified vmic is a free parameter that could plausibly contribute to the measured offsets, particularly for the cool giants that dominate the sample. The adopted value must be stated explicitly, and a sensitivity test (e.g., re-fitting a subsample with the 1 vs. 2 km/s grids) is needed before the offset can be attributed specifically to carbon modeling rather than to an input-parameter mismatch.","section":"§3.1.3, item 3"},{"comment":"The statement that the two grids produce \"different and unpredictable [C/Fe] abundance results for the same star ... with no apparent reason\" is contradicted by the paper's own Fig. 9: the grid-to-grid difference Δ[C/Fe] (BOSZ − Knowles+21) shows a coherent dependence on the Knowles-grid value, with the most negative Knowles values receiving the largest positive BOSZ offsets. A structured, monotonic pattern of this kind is a testable signature (a zero-point shift, a scale factor, or a parameter-dependent mapping), and characterising it would help discriminate between candidate causes such as normalisation versus synthesis differences. The text describing this panel (\"a decreasing trend with [Fe/H]\") also mismatches its axes (x-axis: Knowles [C/Fe]; colour: [Fe/H]) and should be corrected.","section":"§4.4 and Fig. 9 (right panel)"},{"comment":"The catalogue is described as \"large and precise unbiased\" and the conclusions state that the method \"leads to inaccurate [C/Fe] abundance estimates ... without significantly affecting the measured high-quality precision.\" The evidence supports parameter-unbiasedness (no trends with Teff or log g, Fig. 4) and internal precision, but it does not support absolute accuracy: the solar and Arcturus offsets in Figs. 2, 7, A.1, and A.2 leave the zero-point unanchored, and Sect. 4.3 concedes that no objective criterion selects one grid. Since a 0.6–0.8 dex zero-point ambiguity is exactly what users of the catalogue need to know, the abstract and conclusions should carry an explicit grid-dependence caveat rather than the unqualified word \"unbiased.\"","section":"Abstract and §4.1"}],"minor_comments":[{"comment":"State the overlap sample size N for the grid-to-grid comparison, and confirm that the |Δ[C/Fe]| ~ 0.8 dex cases lie within the Teff coverage common to both grids (Knowles+21 covers 3500–6000 K, BOSZ 3500–6500 K); the two final catalogues in Figs. 3 and 9 have different Teff extents, so the intersection is not obvious.","section":"§4.2 and Fig. 9"},{"comment":"Explain why extending the [C/M] range of the Knowles+21 grid from [−0.25, +0.25] to [−0.75, +0.5] reduces the number of stars passing the quality criteria from 199 to 176.","section":"§4.1"},{"comment":"The repeated sentence \"The effect of the hydrogen atom seems to be negligible since [C/Fe] variations can be reproduced and measured at a given [Fe/H]\" is unclear, because hydrogen is not a varied dimension in either grid; please reword or remove it.","section":"§3.2, §4.2, and §5"},{"comment":"Provide a machine-readable table or access link for the final [C/Fe] catalogue; the conclusions present it as a deliverable, but the manuscript contains no table or data URL.","section":"§5"},{"comment":"Report the number of stars used in every panel of the band-to-band comparisons; N=151 and N=143 currently appear only in the first panels.","section":"Figs. 5 and 8"},{"comment":"The closing sentence \"Therefore, we did not find significant discrepancies among the models for the studied stellar sample\" directly contradicts the central result of the paper (offsets up to |Δ[C/Fe]| ~ 0.8 dex in Sect. 4.2); as written it appears to refer only to a visual comparison of synthetic spectra and should be reworded to avoid the appearance of internal inconsistency.","section":"§3.1.3"}],"recommendation":"major_revision","confidential_remarks":"Dear Editor: my assessment matches the reader's conditional verdict. The empirical grid-dependence result is robust and publishable; the problem is the paper's causal framing. The central claim is defensible but the load-bearing attribution (\"hidden issues in carbon modeling\") is not established because the grid-tied pseudo-continuum normalisation, the unspecified BOSZ microturbulence value, and the contradictory solar/Arcturus benchmark directions all provide alternative explanations that the paper's tests do not exclude. This is fixable within scope: either add a cross-normalisation or atomic-line test with both grids, or rewrite the abstract and conclusions as an explicit model-dependence and zero-point-anchoring caveat, which the data fully support. I therefore recommend major revision rather than rejection. The manuscript is candid about its limitations, and the Mg control is well designed; my concerns concern interpretation, not execution. No citation or novelty concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe genuinely new thing here is the controlled comparison: two modern 5D synthetic grids, same XSL sample, same GAUGUIN pipeline, offsets up to ~0.8 dex in [C/Fe] while [Mg/Fe] agrees to ~0.02 dex. The Mg control is the right experiment and it lands. The band-to-band agreement, the lack of parameter trends, and the solar/Arcturus checks all show the pipeline is internally precise. That is real evidence and the paper presents it clearly.\n\nThe soft spot is the conclusion. The paper attributes the grid offset to \"hidden issues in the carbon modeling\" without ruling out the normalisation itself. GAUGUIN normalises by dividing the observed spectrum by a polynomial fit to the synthetic/observed residual, so each grid carries its own pseudo-continuum. In a crowded region like CH 4300 Å, differences in broad wings or pseudo-continuum depression get absorbed into that fit. The result is exactly what you see: internally consistent, precise, but grid-offset abundances. The solar test is the red flag: same solar spectrum, [C/Fe]≈+0.12 with Knowles+21 and ≈−0.52 with BOSZ. The paper's own validation does not separate the synthetic carbon models from the measurement pipeline because every test runs within a single grid.\n\nAlso worth noting: the catalogue is described as \"unbiased\" while carrying an unresolved ~0.8 dex systematic. The quoted uncertainties are internal only and omit the dominant term. The paper never specifies which BOSZ microturbulence value GAUGUIN ingested, and the two grids use different vmic prescriptions. And the external comparison with APOGEE is too dispersed (σMAD≈0.24 dex) to arbitrate. So the central caveat is right—CH-based carbon abundances are strongly model-dependent—but the causal attribution is a hypothesis, not a demonstrated result.\n\nThis is a solid cautionary paper that needs a revision: add the grid offset to the error budget, temper \"unbiased\", and either test the normalisation hypothesis (e.g. fit with one grid's pseudo-continuum and measure with the other) or explicitly frame the carbon-model defect as a hypothesis. The core observation deserves referee time and likely publication.","headline":"The grid-dependence of CH-based [C/Fe] is real and well demonstrated, but the paper overreaches in blaming the synthetic carbon models without ruling out its own grid-tied normalisation.","tokens_in":25450,"tokens_out":2565,"would_cite":true,"duration_ms":29654,"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":"Carbon abundances measured from the CH 4300 Å band with the GAUGUIN spectrum-synthesis code can differ by up to ~0.8 dex for the same star depending only on which reference synthetic grid is used, even though all internal consistency…","keywords":["carbon abundances","CH 4300 Å band","spectral synthesis","X-shooter Spectral Library","model dependence","GAUGUIN","stellar abundances","G band"],"falsifier":"Re-derive [C/Fe] for the same XSL stars from the same two CH windows while forcing both grids to share one grid-independent continuum (e.g., pseudo-continuum anchored to spectral windows outside CH absorption), and check whether the up-to-0.8 dex offsets persist; alternatively, measure the CI lines at 5052 and 5380 Å at R ≥ 50000 for the same stars and see which grid's CH-based values match.","tokens_in":24285,"feed_emoji":"🔭","tokens_out":8353,"duration_ms":93717,"temperature":0.7,"pith_summary":"The paper measures carbon-to-iron ratios ([C/Fe]) for about 200 stars of the X-shooter Spectral Library by fitting two CH molecular bands near 4300 Å with the GAUGUIN spectrum-synthesis code. It runs the same stars and the same bands through two different grids of reference synthetic spectra — the theoretical library behind sMILES and the updated BOSZ library — each with the same [C/Fe] coverage. Every internal check looks healthy within each grid: the two CH bands agree, the values show no trend with stellar parameters, Monte Carlo uncertainties are small, and solar and Arcturus fits are good. Yet the same star can receive [C/Fe] values up to ~0.8 dex apart depending only on which grid is adopted, while [Mg/Fe] measured the same way agrees across grids. The paper's central claim is that CH-band carbon abundances from spectral synthesis are strongly model-dependent, so wrong answers can be delivered with high quoted precision and no warning.","feed_headline":"Carbon measurements shift up to 0.8 dex when the synthetic grid changes","feed_subtitle":"Same stars give [C/Fe] that shifts up to 0.8 dex between grids; internal checks cannot detect the error.","key_machinery":"The machinery is GAUGUIN, an automated spectrum-synthesis abundance code, applied to two narrow CH windows at 4301.5–4303.4 Å and 4307.1–4308.8 Å, with a 5-dimensional grid of synthetic spectra in effective temperature, surface gravity, metallicity, alpha-enhancement, and [C/Fe] as the reference. Two such grids are used: the theoretical library computed for sMILES (ATLAS9 atmospheres, ASSET/SYNSPEC, Allende Prieto et al. 2018 line lists) and the updated BOSZ library (MARCS atmospheres, newer SYNSPEC, Masseron et al. 2014 CH line list). Because the CH region is crowded, GAUGUIN defines a pseudo-continuum per grid: it divides the observed spectrum by the interpolated synthetic one, fits the residual with a third-degree polynomial, and fits [C/Fe] relative to that normalisation. The two grids therefore carry different pseudo-continua and differ in atmospheres, microturbulence treatment, molecular opacities, and SYNSPEC version — the small model differences that, the paper argues, expand into the large [C/Fe] excursions.","core_discovery":"The central discovery is that the reference synthetic grid, not the star, can set the measured carbon abundance. For identical spectral windows, identical stellar parameters, and the same [C/Fe] coverage in the two grids, GAUGUIN returns internally consistent but mutually incompatible catalogues: comparing the [C/Fe] derived with the Knowles et al. (2021) grid to that derived with the Mészáros et al. (2024) BOSZ grid gives a star-to-star scatter of about 0.3 dex and individual discrepancies up to |∆[C/Fe]| ≈ 0.8 dex, with no obvious cause in the fits themselves; the authors do note a decreasing trend with [Fe/H] in the offset. The same comparison for [Mg/Fe] is flat and agrees to about 0.02 dex. The authors conclude that small intrinsic differences between synthetic models in the crowded, blended CH 4300 Å region are amplified by the abundance-estimation procedure and produce large, unnoticed inaccuracies in stellar carbon measurements.","pith_inferences":["Editorial inference: the grid-to-grid offset may be produced by the pseudo-continuum normalisation itself, since each grid defines its own pseudo-continuum and all internal consistency tests stay inside one grid; if so, the paper's attribution of the offset to carbon modeling is not the only reading.","Editorial inference: if carbon opacity is the culprit, the offset should grow where CH features are stronger (cooler stars, lower gravity, higher [C/Fe]); the paper's Fig. 9 suggests a [Fe/H]-dependent offset that could be mapped against CH strength to test this.","Editorial inference: measuring the same XSL stars through other carbon molecules, such as C2 or CN, or through the 8727 Å [CI] line at high resolution, would separate a carbon-physics problem from a CH-band crowding problem.","Editorial inference: a simple decisive test is to fit both grids with a common, grid-independent continuum; the discrepancy should collapse if normalisation is responsible and persist if it is not."],"forward_implications":["Published [C/Fe] values derived from the CH 4300 Å band with any single synthetic grid may carry unrecognised grid-dependent offsets of order 0.3 dex, with outliers near 0.8 dex.","Stellar population models built on empirical libraries whose carbon abundances come from CH bands inherit that model dependence in their carbon-sensitive predictions.","The flat, dispersed [C/Fe] versus [Fe/H] trend seen in this work and in earlier CH-band studies is not a secure measurement of the Galactic carbon trend unless the grid dependence is understood.","Passing internal quality checks — band-to-band agreement, no parameter trends, small uncertainties — is not sufficient evidence that a CH-based [C/Fe] measurement is accurate.","Before CH-band carbon abundances are used as benchmarks, the same stars should be cross-checked with an independent grid or with high-resolution atomic CI lines."],"supporting_citations":[{"why":"Supplies one of the two reference synthetic grids (the theoretical library behind sMILES) whose [C/Fe] dimension is fit to the CH bands.","marker":"Knowles et al. (2021)"},{"why":"Supplies the second reference grid, the updated BOSZ/MARCS library; comparing it with the first produces the up-to-0.8 dex [C/Fe] offsets.","marker":"Mészáros et al. (2024)"},{"why":"Defines the GAUGUIN pseudo-continuum normalisation and abundance-estimation method used on the X-shooter spectra.","marker":"Santos-Peral et al. (2020)"},{"why":"Provides the original selection of clean CH 4300 Å windows that the paper adapts to X-shooter resolution.","marker":"Suárez-Andrés et al. (2017)"},{"why":"Provides the XSL sample, quality criteria, and the [Mg/Fe] catalogue used to validate consistency across grids.","marker":"Santos-Peral et al. (2023)"},{"why":"Provides the XSL stellar parameters (Teff, log g, [Fe/H]) fixed as GAUGUIN inputs for each star.","marker":"Arentsen et al. (2019)"},{"why":"The extended CH line list included in BOSZ, one of the concrete grid differences implicated in the carbon-modeling discrepancy.","marker":"Masseron et al. (2014)"},{"why":"The common atomic and molecular line list used by the sMILES grid, defining the carbon opacity to be varied.","marker":"Allende Prieto et al. (2018)"},{"why":"The Arcturus benchmark spectrum used to check CH-band [C/Fe] fits against a literature value at high resolution.","marker":"Hinkle et al. (2000)"}],"fun_headline_variants":["CH-band carbon abundances shift 0.8 dex with model grid choice","Grid choice swings stellar carbon by up to 0.8 dex","Carbon abundance measurements depend on synthetic grid","Same stars, different carbon: 0.8 dex grid effect","Stellar carbon values vary 0.8 dex across model grids"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise — stated in Sect. 4.4 — is that the grid-to-grid [C/Fe] offset is caused by hidden defects in how the synthetic spectra model carbon, and not by the GAUGUIN pseudo-continuum normalisation, which is defined separately from each grid (Sects. 3.2–3.3).","fun_headline_variants_meta":{"raw":{"variants":["CH-band carbon abundances shift 0.8 dex with model grid choice","Grid choice swings stellar carbon by up to 0.8 dex","Carbon abundance measurements depend on synthetic grid","Same stars, different carbon: 0.8 dex grid effect","Stellar carbon values vary 0.8 dex across model grids"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000181,"raw_usage":{"total_tokens":1392,"prompt_tokens":1112,"completion_tokens":280,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":728,"completion_tokens_details":{"reasoning_tokens":194}},"tokens_in":728,"tokens_out":280,"duration_ms":3633,"temperature":1.0,"reasoning_tokens":194,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:10:00.649896+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-derive [C/Fe] for the same XSL stars from the same two CH windows while forcing both grids to share one grid-independent continuum (e.g., pseudo-continuum anchored to spectral windows outside CH absorption), and check whether the up-to-0.8 dex offsets persist; alternatively, measure the CI lines at 5052 and 5380 Å at R ≥ 50000 for the same stars and see which grid's CH-based values match.","supporting_citations":[{"cited_title":"T., Sansom , A","cited_arxiv_id":null,"evidence_quote":"Supplies one of the two reference synthetic grids (the theoretical library behind sMILES) whose [C/Fe] dimension is fit to the CH bands."},{"cited_title":"2020, , 639, A140","cited_arxiv_id":null,"evidence_quote":"Defines the GAUGUIN pseudo-continuum normalisation and abundance-estimation method used on the X-shooter spectra."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the XSL sample, quality criteria, and the [Mg/Fe] catalogue used to validate consistency across grids."},{"cited_title":"2014, , 571, A47","cited_arxiv_id":null,"evidence_quote":"The extended CH line list included in BOSZ, one of the concrete grid differences implicated in the carbon-modeling discrepancy."},{"cited_title":"2000, Visible and Near Infrared Atlas of the Arcturus Spectrum 3727-9300 A","cited_arxiv_id":null,"evidence_quote":"The Arcturus benchmark spectrum used to check CH-band [C/Fe] fits against a literature value at high resolution."}],"review_version":1}