{"id":"c64c2c7f-4e0d-444c-90dc-4801e76e937f","arxiv_id":"2502.09707","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":14,"one_line_summary":"Interstellar calcium and sodium absorption in MaStar is modeled and removed, yielding corrected spectra and simple stellar population templates that reduce a systematic bias in galaxy stellar population analyses.","lead":"This paper removes Milky Way gas absorption lines from 80% of the spectra in the largest stellar library, MaStar, and releases cleaned spectra and model templates. It shows that without this correction, galaxy models would overestimate sodium and calcium absorption, biasing age and abundance measurements.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unresolved 3–4σ offset between BOSZ and empirical low-ISM coadds for T_eff>9000 K stars (Sec. 3.3) means the cleaned library may retain residual ISM or remove intrinsic absorption; the central cleaning and SSP claims are not settled in the regime where corrections are largest.","rationale":"The reader identified representativeness of the W_ISM model as the weakest assumption. The Section 3.3 BOSZ-versus-coadd discrepancy is a concrete, internal manifestation of that same risk: if the low-ISM sample is contaminated, the empirical replacement path is biased, and if BOSZ is wrong, the theoretical path is biased. The paper reports the offset and presents both possible interpretations honestly, but the ambiguity is unresolved, and the quantitative headline claims about hot-star cleaning and young/old SSP enhancements are conditional on which interpretation is correct. The low-ISM SSP consistency check in Section 3.6.4 does not fully resolve this because the cleaned SSPs and the low-ISM SSPs share the same low-ISM coadds as their building blocks. A high-resolution follow-up on overlapping stars would settle the direction of the bias. The reader's CONDITIONAL verdict remains appropriate: the data products are public, the methods are documented, and the limitation is explicitly acknowledged, so there is no basis for rejection, but the unresolved 3–4σ offset prevents full acceptance of the central quantitative claims. Verdict is therefore UNCHANGED relative to the reader's conditional assessment.","tokens_in":54631,"tokens_out":5295,"duration_ms":63119,"concrete_test":"Obtain high-resolution (R > 30,000) UVES/HIRES spectra for a sample of MaStar stars with T_eff > 9000 K, especially the stars that have both empirical and BOSZ replacements. Fit interstellar Ca II K and Na I D components and measure W_ISM; compare these measured values to the model predictions used for the low-ISM cut (W_ISM < 0.05–0.07 Å) and to the W_BOSZ - W_coadd offsets. If measured W_ISM is ~0.4–0.6 Å for stars labeled low-ISM, the empirical coadds retain residual ISM and the cleaned library is under-cleaned; if measured W_ISM is consistent with <0.1 Å, the BOSZ model is removing intrinsic absorption and the BOSZ-cleaned products should not be used for hot stars without correction. This directly adjudicates the 3–4σ offset in Section 3.3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is not the general representativeness of the literature W_ISM model, but the internal cross-check reported in Section 3.3. For the hot stars with both empirical and BOSZ replacements, the mean W_BOSZ - W_coadd is -0.57 Å for Ca II K and -0.43 Å for Na I D for the eleven stars with T_eff,med > 9000 K, with σ_W,coadd of 0.20 Å and 0.11 Å, respectively. The two cleaning prescriptions therefore disagree by roughly 3–4σ in exactly the parameter regime where the claimed ISM corrections are largest. The authors list two possible explanations: (1) the 'low-ISM' stars from which all empirical coadds are built still contain measurable interstellar absorption, or (2) the BOSZ models systematically underpredict intrinsic Ca II/Na I in hot stars. These interpretations have opposite consequences for the cleaned products. If (1), the 6342-star low-ISM sample is not truly low-ISM, the empirical replacement profiles inherit a residual ISM component, and the cleaned library does not fully remove Milky Way absorption; the reported mean reductions of 0.4–0.7 Å in Ca II K and 0.6–1.1 Å in Na I D for hot stars are lower limits, and the SSP fractional enhancements in Section 4.2 are underestimated. If (2), the BOSZ-cleaned spectra for 738 stars, including isolated hot stars, have had intrinsic stellar absorption removed, artificially weakening NaD and Ca II and biasing the hot-star SSP products in the opposite direction. Either way, the central claim that the released library and SSP templates correctly remove ISM while preserving stellar lines is not established in the regime where the corrections matter most. The low-ISM SSP comparison in Section 3.6.4 is not an independent validation: those SSPs are built from the same low-ISM coadds, so their agreement with the cleaned SSPs is partly by construction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper assesses the impact of Milky Way interstellar Ca II and Na I absorption on the SDSS-IV MaStar empirical stellar library. The authors build a model for W_ISM as a function of distance, Galactic latitude, and dust reddening from high-resolution literature samples (Sembach et al. 1993; Munari & Zwitter 1997; Welsh et al. 2010), apply it to 24,162 MaStar stars, identify 6,342 low-ISM sightlines, and replace the Ca II and Na I D profiles of 12,110 stars with coadded low-ISM spectra and of 738 hot stars with BOSZ theoretical templates. They release cleaned stellar spectra, hierarchically clustered templates, and SSP templates, and quantify the artificial enhancement of Ca II K and NaD indices in SSP models built from the original library.","tokens_in":55104,"tokens_out":7036,"duration_ms":74348,"significance":"If the cleaning procedure is reliable, this is a valuable data product: it would remove a previously unquantified Milky Way ISM contamination from the largest empirical stellar library, with direct consequences for galaxy spectral fitting and stellar population synthesis. The public releases of cleaned spectra, HC templates, and SSP templates are concrete strengths, and the demonstration that NaD enhancements persist at the >10% level at old ages is an important, falsifiable result. The paper also contains a careful treatment of telluric contamination (Appendix A). The significance is conditional on resolving a 3-4 sigma disagreement between the empirical and theoretical cleaning methods for hot stars, which is the central load-bearing issue.","major_comments":[{"comment":"This is a major comment.","section":"Section 3.3, Figure 14"},{"comment":"This is a major comment.","section":"Section 2.2-2.3"},{"comment":"This is a major comment.","section":"Section 3.6.4"}],"minor_comments":[{"comment":"This is a minor comment.","section":"Table 3 and Section 3.4"},{"comment":"This is a minor comment.","section":"Section 3.5"},{"comment":"This is a minor comment.","section":"Figure 13"},{"comment":"This is a minor comment.","section":"Equation (1) and Section 2.2"}],"recommendation":"major_revision","confidential_remarks":"This is a strong data-product paper with broad utility, but the unresolved 3-4 sigma discrepancy between the empirical coadd and BOSZ cleaning methods for hot stars is the central stumbling block. I recommend requiring the authors to either break the degeneracy with an independent test or to explicitly present the cleaned-library claims under both interpretations, rather than adopting one prescription in the released product without resolving the inconsistency."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a useful, practical paper that does what it says—quantifies and corrects Milky Way CaII/NaI absorption in MaStar—but it carries an unresolved 3–4 sigma disagreement between its two cleaning prescriptions for the hottest stars, exactly where the corrections are largest. The cleaned library and SSP products are still worth having, but the hot-star regime needs more work before I trust the quoted reductions as more than a range.\n\nWhat is new: this is the first ISM correction applied to a low-resolution empirical stellar library of this size. The authors release cleaned spectra, hierarchically clustered templates, and SSP templates, and show that original MaStar SSPs overestimate CaII K by >20% at young ages, NaD by >50% in starbursts, and NaD by >10% at old ages. The analysis is careful: they characterize intrinsic profile scatter versus stellar parameters, check telluric contamination in an appendix, and flag every star's treatment. That is real value, and the public data products will be used.\n\nThe main soft spot is Section 3.3. For the eleven hot stars with both empirical and BOSZ replacements, the mean difference W_BOSZ - W_coadd is -0.57 Å for CaII K and -0.43 Å for NaI D, with sigma_W,coadd around 0.20 and 0.11 Å. That is a 3–4 sigma offset. The authors list two interpretations—residual ISM in the low-ISM sample, or BOSZ models underpredicting intrinsic absorption in hot stars—which have opposite consequences for the cleaned products. They do not resolve which is right. Since the claimed corrections for hot stars are 0.4–1.1 Å, the ambiguity is comparable to the correction itself. The cross-check in Section 3.6.4, comparing cleaned SSPs to SSPs built only from low-ISM stars, is reassuring but partly circular because the low-ISM sample is defined by the same W_ISM model. Minor points: the supersolar low-ISM thresholds are hand-relaxed, and the W_ISM model extrapolates from literature sightlines to all MaStar coordinates, but that is clearly stated and the model is external, so there is no circularity.\n\nBottom line: the paper is honest about its limitation, and the flags in the released files let users test sensitivity. A serious referee should engage, but the authors should be asked to either validate the hot-star cleaning against independent high-resolution CaII/NaI measurements of a few MaStar stars, or to present the cleaned products and SSP biases under both interpretations so users can see how much the conclusions move.\n\nRecommendation: send to peer review. This is a solid, useful contribution with a load-bearing uncertainty that should be fixed or at least fully quantified.","headline":"Useful, practical ISM-cleaning of MaStar with one unresolved 3–4 sigma discrepancy between the two replacement methods for hot stars, exactly where the corrections are largest.","tokens_in":55776,"tokens_out":3504,"would_cite":true,"duration_ms":38158,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Milky Way interstellar gas plants spurious calcium and sodium lines in the largest star library, and the paper's cleaned version removes them, shifting galaxy ages, sodium abundances, and mass-function fits.","keywords":["interstellar absorption","Ca II H and K lines","Na I D doublet","MaStar stellar library","stellar population synthesis","empirical spectral templates","dust reddening","equivalent widths"],"falsifier":"Re-observe a sample of MaStar stars spanning the predicted contamination range at spectral resolution R ≳ 30,000, where interstellar Ca II and Na I components are narrow and kinematically separated from the broad stellar lines. If the measured interstellar equivalent widths disagree with the model's binned predictions for a substantial fraction of sightlines, or if the paper's 'cleaned' profiles do not match the interstellar-free photospheric lines seen at high resolution, the correction is wrong. A cheap version of this test is the paper's own internal check, in which SSPs built only from low-ISM stars agree with the fully cleaned SSPs at the 3–4% level: any regime where that agreement fails marks where the cleaning breaks down.","tokens_in":54422,"feed_emoji":"🌌","tokens_out":20243,"duration_ms":162803,"temperature":0.7,"pith_summary":"This paper establishes that cool gas in the Milky Way imprints its own calcium and sodium absorption onto the spectra of thousands of stars in MaStar, the largest empirical stellar spectral library, and that this contamination is strong enough to distort galaxy science built on the library. Using high-resolution measurements of interstellar absorption toward a few thousand stellar sightlines, the authors build a model that predicts the contaminating line strength from a star's distance, latitude, and dust reddening, and they use it to produce corrected spectroscopy for roughly 80% of the 24,162-star catalog. The central demonstration is that simple stellar population models constructed from the original library artificially strengthen the Ca II K line by more than 20% in populations younger than 400 million years, and strengthen the Na I D line by more than 50% in starbursting systems and by more than 10% even in 10-billion-year-old populations. If the paper is right, analyses using the uncorrected library infer galaxy ages that are too young, sodium abundances that are too low, and initial mass functions that are too bottom-heavy, and the released cleaned spectra, templates, and SSP models offer a direct way to correct those biases.","feed_headline":"Milky Way gas contaminates 80% of the largest star library","feed_subtitle":"Left uncorrected, the gas lines inflate galaxy ages, sodium abundances, and mass fits by 10% to over 100%.","key_machinery":"The central mechanism is an empirical scaling law for interstellar contamination: in each of six distance bins, the equivalent width of Ca II K and of each Na I D component is modeled as $W^{\\rm ISM} = \\beta + \\alpha \\log E_{\\rm DustMap}$, with slope and intercept fit by MCMC likelihoods that include upper limits and intrinsic scatter, anchored to the high-resolution samples of Sembach et al., Munari & Zwitter, and Welsh et al. This predictor converts three readily available stellar quantities — distance, Galactic latitude, and dust reddening — into a predicted contaminating width, which in turn sets the thresholds ($W^{\\rm ISM}$(Ca II K) < 0.07 Å, $W^{\\rm ISM}$(Na I 5891) < 0.05 Å) that define the 6342-star 'low-ISM' reference sample. Replacement matching runs on a three-dimensional stellar-parameter distance $\\Psi$ built from $\\log g$, $\\theta = 5040\\ \\mathrm{K}/T_{\\rm eff}$, and $[\\mathrm{Fe/H}]$; coadded low-ISM spectra are inserted into the windows 3912–3995 Å and 5873–5917 Å with Gaussian-tapered edges, and isolated hot stars instead receive ATLAS9-based BOSZ model profiles. The cleaned spectra are then passed through the established stellar-population construction to produce SSP templates whose line strengths, compared with SSPs built from the original library, quantify the bias.","core_discovery":"The paper argues that the Milky Way's interstellar medium leaves its own absorption signature inside the MaStar empirical stellar library, and that the signature is large enough to bias galaxy science. At the library's R ~ 1800 resolution, interstellar Ca II λλ3934, 3969 and Na I λλ5891, 5897 absorption is blended into the stellar photospheric profiles of those transitions. The authors construct a model of the interstellar equivalent width as a function of stellar distance, Galactic latitude, and dust reddening, fit as $W^{\\rm ISM} = \\beta + \\alpha \\log E_{\\rm DustMap}$ in six distance bins using high-resolution literature measurements. Applying this model to all 24,162 MaStar stars selects 6342 stars predicted to have negligible contamination ($W^{\\rm ISM}$(Ca II K) < 0.07 Å and $W^{\\rm ISM}$(Na I 5891) < 0.05 Å); for the remaining stars with close matches, the Na I D region (and, for stars with $T_{\\rm eff} > 9000$ K, the Ca II region too) is replaced with coadded spectra of similar low-contamination stars, and 738 isolated hot stars receive matched ATLAS9-based BOSZ model profiles instead. The result is a cleaned library whose mean $W$(Ca II K) falls by 0.4–0.7 Å and whose mean $W$(Na I D) falls by 0.6–1.1 Å for hot stars ($T_{\\rm eff} > 7610$ K) and by 0.1–0.2 Å for cooler stars. Constructed from the original library, simple stellar population models overestimate $W$(Ca II K) by ≥20% at ages below 400 Myr and overestimate the NaD index by ≥50% in starbursting systems and by ≥10% at ages above 10 Gyr. The cleaned stellar spectra, hierarchically clustered templates, and SSP libraries are released as public data products.","pith_inferences":["Editorial inference: the same contamination model could be applied to other moderate-resolution empirical libraries, re-deriving the NaD–[Na/Fe] calibration from a contamination-free sample to test how much the index–abundance slope shifts.","Editorial inference: the distance binning smooths over small-scale structure (the paper itself notes an absorbing 'wall' at 80 pc), so the model will be least reliable toward dense clouds and the Galactic plane; a version using the full 3D dust map rather than binned medians would directly test how often stars near such clouds are misclassified as low-ISM.","Editorial inference: the predicted interstellar widths for all 24,162 stars form a statistical absorption map of the local Milky Way that the paper does not exploit; cross-correlating these widths with known cloud and gas catalogs could calibrate Ca II and Na I columns against dust and 21 cm emission."],"forward_implications":["Stellar-population fits to young and starbursting galaxies made with the original library underestimate stellar ages, because interstellar Ca II K mimics the strong calcium absorption that diagnoses A-type stars.","[Na/Fe] estimates in early-type galaxies from the NaD Lick index are systematically low by roughly 0.1–0.2 dex, and the paper argues the true sodium enhancement of massive ellipticals may reach +0.7 to +1.0 dex.","Spectroscopic IMF-slope constraints anchored on Na I transitions shift toward less bottom-heavy (more Milky Way-like) slopes once templates are corrected, and lower-mass ellipticals may move toward bottom-light slopes.","Down-the-barrel studies of Na I D as a tracer of galactic winds will overestimate the stellar contribution and underestimate the host-galaxy interstellar component in starbursts.","The public cleaned spectra, hierarchical cluster templates, and SSP libraries let existing galaxy surveys be re-fit without new observations, putting a quantitative bound on this systematic for each survey."],"supporting_citations":[{"why":"Supplies the R~68,000 Ca II and Na I spectroscopy of 57 early-type stars beyond 1 kpc that anchors the intercepts of the $W^{\\rm ISM}$–reddening fits at large distance.","marker":"Sembach et al. 1993"},{"why":"Supplies the 32 O/B-star sightlines establishing the Na I D versus reddening relation across a wide reddening range.","marker":"Munari & Zwitter 1997"},{"why":"Supplies R>50,000 Na I and Ca II measurements toward 1857 nearby early-type stars, setting the distance dependence and the low-reddening behavior of $W^{\\rm ISM}$.","marker":"Welsh et al. 2010"},{"why":"Provides the 3D Milky Way dust map that yields $E_{\\rm DustMap}$ for every MaStar sightline, the predictor variable of the model.","marker":"Green et al. 2019"},{"why":"Provides the conversion $A_V = 3.31\\,E_{\\rm DustMap} - 0.076$ that calibrates literature reddenings to dust-map units.","marker":"Lazarz et al. 2022"},{"why":"Provides the Gaia EDR3 photogeometric distances that place each MaStar star in the model's distance bins.","marker":"Bailer-Jones et al. 2021"},{"why":"Supplies the ATLAS9/BOSZ theoretical spectrum grid used to replace the Ca II and Na I profiles of the 738 hot stars lacking close empirical matches.","marker":"Bohlin et al. 2017"},{"why":"Provides the isochrones and fuel-consumption methodology used to assemble the SSP templates from MaStar spectra, the products in which the ISM bias is quantified.","marker":"Maraston et al. 2020"}],"fun_headline_variants":["Gas lines contaminate 80% of largest stellar library","Interstellar smog inflates galaxy age and mass fits by ≥10%","MaStar library cleaned of Milky Way absorption in 80% of stars","Starburst NaD indices overstated by ≥50% without ISM fix"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire cleaning rests on one premise: that the high-resolution calcium and sodium absorption measurements made toward roughly two thousand stellar sightlines, binned by distance and reddening and converted through a dust map, correctly predict the interstellar absorption along every one of the 24,000 MaStar sightlines.","fun_headline_variants_meta":{"raw":{"variants":["Gas lines contaminate 80% of largest stellar library","Interstellar smog inflates galaxy age and mass fits by ≥10%","MaStar library cleaned of Milky Way absorption in 80% of stars","Starburst NaD indices overstated by ≥50% without ISM fix"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000337,"raw_usage":{"total_tokens":2111,"prompt_tokens":1439,"completion_tokens":672,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":1055,"completion_tokens_details":{"reasoning_tokens":593}},"tokens_in":1055,"tokens_out":672,"duration_ms":7149,"temperature":1.0,"reasoning_tokens":593,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T20:46:55.403377+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-observe a sample of MaStar stars spanning the predicted contamination range at spectral resolution R ≳ 30,000, where interstellar Ca II and Na I components are narrow and kinematically separated from the broad stellar lines. If the measured interstellar equivalent widths disagree with the model's binned predictions for a substantial fraction of sightlines, or if the paper's 'cleaned' profiles do not match the interstellar-free photospheric lines seen at high resolution, the correction is wrong. A cheap version of this test is the paper's own internal check, in which SSPs built only from low-ISM stars agree with the fully cleaned SSPs at the 3–4% level: any regime where that agreement fails marks where the cleaning breaks down.","supporting_citations":[{"cited_title":"1997, A&A, 318, 269 M¨ unch, G., & Zirin, H","cited_arxiv_id":null,"evidence_quote":"Supplies the 32 O/B-star sightlines establishing the Na I D versus reddening relation across a wide reddening range."}],"review_version":1}