{"id":"8fec97a7-0ccd-447c-96dc-5babcba6c04f","arxiv_id":"2509.10079","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"ALMA observations of 17 quiescent galaxies at z~0.4 show dust-to-molecular-gas ratios spanning roughly 1/700 to 1/40, implying independent dust and gas evolution after quenching.","lead":"Astronomers measured dust and molecular gas in 17 old, non-star-forming galaxies at z~0.4 with ALMA, finding dust-to-gas ratios that spread from 8 times below to 2.5 times above the standard value of 1/100. The result suggests dust and gas do not fade together after star formation stops, so dust brightness alone is a poor gauge of gas in quiescent galaxies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fixed CO(3-2) conversion factors (R31=0.5, alpha_CO=4.36) are the load-bearing assumption; if old QGs have different excitation or alpha_CO, the claimed above-canonical delta_DGR tail and 'dust/H2 independent evolution' conclusion are not secure.","rationale":"The reader's weakest assumption correctly identifies the fixed CO/H2 conversion factors as the most load-bearing point. My quantitative check from Table 1 confirms that the 'above canonical' delta_DGR values require only modest changes in R31 and alpha_CO to be erased, while the dust-mass cross-checks in Appendix B do not address this because they share a fixed T_dust=21 K and never vary the CO calibration. This is not a fatal flaw—the ALMA detections and the qualitative diversity in delta_DGR may survive—but it makes the headline '~2.5x above canonical' and the broader 'dust is an unreliable H2 tracer' conclusion conditional on unverified conversion assumptions. Since the reader already assigned CONDITIONAL, my assessment does not change the verdict; it sharpens the specific test that should be required. I agree with the reader's framing and find no additional load-bearing concern beyond this calibration issue.","tokens_in":17025,"tokens_out":7331,"duration_ms":85845,"concrete_test":"Recompute M_H2 and log delta_DGR for the 10 CO-detected QGs in Table 1 under observationally motivated alternative calibrations: (i) R31=0.3, alpha_CO=4.36; (ii) R31=0.5, alpha_CO=7.5; (iii) R31=0.3, alpha_CO=7.5; and (iv) R31=0.7, alpha_CO=3.0. If in any plausible QG scenario the number of systems with log delta_DGR > -2 drops from 3 to 0, or if the claimed spread relative to canonical 1/100 shrinks below ~0.6 dex, the 'above canonical' and 'independent evolution' conclusions are not secure. Ideally, obtain CO(1-0) or [CI](1-0) observations for hCOS-d3, hCOS-d10, and hCOS-d13 to directly measure R31 and independently constrain M_H2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—delta_DGR in QGs spans ~8x below to ~2.5x above canonical 1/100, implying dust and molecular gas evolve independently—rests on the H2 calibration in Section 2.3. M_H2 is derived from CO(3-2) assuming R31=0.5 and alpha_CO=4.36, both inherited from star-forming/Milky Way calibrations, with no quoted systematic uncertainty. The specific systems that drive the 'above canonical' result are hCOS-d3 (log delta_DGR=-1.64), hCOS-d10 (-1.49), and hCOS-d13 (lower limit >-1.53). Moving just these above -2.0 requires increasing M_H2 by factors of ~2.3-3.2. Such a shift is well within plausible QG systematics: e.g., R31=0.3 instead of 0.5 raises M_H2 by 1.67x, and alpha_CO=7.5 instead of 4.36 raises it by another 1.72x; combined, the three 'above canonical' systems fall to or below 1/100. If true alpha_CO is ~10 or R31 is ~0.2, the high tail disappears entirely. Appendix B only cross-checks dust mass methods, all adopting T_dust=21 K, so it does not constrain CO excitation or conversion. The ALMA detections themselves are credible, and the low-delta_DGR tail may be robust, but the novel 'above canonical' regime is directly contingent on the adopted H2 conversion. Thus the paper's strongest consequence—that dust continuum is an unreliable H2 tracer in QGs—would be substantially weakened if the high-delta_DGR sources are calibration artifacts.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents ALMA Band 6 continuum and CO(3–2) observations of 17 quiescent galaxies at z~0.4 in COSMOS, jointly constraining dust mass and molecular gas mass. The authors derive dust-to-gas ratios δ_DGR for 12 systems, reporting a spread from ~8× below to ~2.5× above the canonical SFG value of 1/100, and argue that dust and H2 evolve independently after quenching, that QGs do not follow a single ISM depletion path, and that dust continuum is not a reliable H2 tracer. The interpretation is framed with comparison to SIMBA fast and slow quenching tracks.","tokens_in":17491,"tokens_out":3141,"duration_ms":34652,"significance":"If the central empirical claims are robust, this is an important advance: it roughly triples the number of QGs with direct dust-to-molecular-gas estimates, opens a previously unexplored low-gas-fraction regime, and directly challenges the common practice of inferring H2 masses from dust continuum in quiescent systems. The paper’s strengths include a homogeneously selected sample, deep ALMA data with clearly reported fluxes and upper limits, mock SED recovery tests, and explicit comparison of dust-mass estimators. The conclusions are falsifiable and would motivate model refinements for ISM evolution in QGs.","major_comments":[{"comment":"The entire high-δ_DGR tail is calibrated through fixed CO conversion factors: R31=0.5 and alpha_CO=4.36 M_sun (K km/s pc^2)^-1, adopted from star-forming galaxy practice with no quoted systematic uncertainty. In Table 1, the systems with log δ_DGR above -2 are hCOS-d3 (-1.64), hCOS-d6 (-1.96), hCOS-d10 (-1.49), and hCOS-d13 (lower limit > -1.53). Moving these detections below -2 requires increasing M_H2 by factors of roughly 2.3–3.1; combinations such as R31=0.3 and alpha_CO=7.5 (factor ~2.9) or R31=0.2 and alpha_CO=10 (factor ~11) would erase the above-canonical tail entirely. Since old QGs may plausibly have lower CO excitation and higher alpha_CO than SFGs, the paper's headline '2.5× above canonical' claim is not yet secure. Please add an explicit sensitivity analysis over R31 and alpha_CO, or substantially soften claims about the high-δ_DGR regime.","section":"§2.3, Table 1"},{"comment":"The sentence 'This large spread ... regardless of how M_dust and M_H2 are estimated (see Appendix B)' is not supported by the cited appendix. Appendix B compares only dust-mass estimators (DL07 vs. modified blackbody vs. Scoville et al. 2016), and all three methods fix T_dust=21 K; it does not vary the H2 conversion factors R31 or alpha_CO. Thus the robustness of the δ_DGR spread against the main systematic axis—the CO-to-H2 calibration—is not demonstrated. Please correct this claim and either extend the appendix or state the conditional nature of the result.","section":"§3, Appendix B"},{"comment":"The identification of 'rapid' (~0.7 Gyr) versus 'mild' (~2 Gyr) dust decline is made by visually comparing observed f_dust–age points against two exponential tracks normalized to SFG values. No quantitative fit, goodness-of-fit statistic, or uncertainty is provided, and the sample includes upper limits. As a consequence, the statement that 'about half' of QGs exhibit rapid decline is a qualitative classification rather than a measured result. Please provide a quantitative classification (e.g., likelihood-based assignment to tracks or a fitted τ_dust with uncertainties), or clearly label this as an illustrative comparison.","section":"§3.1, Fig. 2"}],"minor_comments":[{"comment":"The text says 'two QGs emerge with log δ_DGR ≳ −2', but Table 1 lists three detected QGs above -2 (hCOS-d3: -1.64, hCOS-d6: -1.96, hCOS-d10: -1.49) in addition to the hCOS-d13 lower limit -1.53. Please reconcile the count.","section":"§3.2"},{"comment":"The axis label 'log(fdust = Mdust/M )' is missing the stellar-mass subscript; it should read M_★ for clarity.","section":"Fig. 2"},{"comment":"The phrase 'For the first time, we show that δ_DGR in QGs ranges from ~8× below to ~2.5× above' is a strong claim. If the sensitivity analysis recommended above weakens the upper end, the abstract should be reworded to emphasize the robust lower-DGR tail and the large scatter.","section":"Abstract/§3"},{"comment":"Typo: 'stellar paopulation age' should be 'stellar population age'.","section":"§1"}],"recommendation":"major_revision","confidential_remarks":"The paper reports a valuable ALMA dataset and the low-δ_DGR tail is likely robust, but the central 'above canonical' claim and the 'dust is an unreliable H2 tracer' conclusion rest on fixed SFG-like CO conversion factors. A sensitivity analysis is essential; without it, the manuscript should either be reframed as conditional on those assumptions or the high-δ_DGR results should be presented as tentative. The qualitative fast/slow classification in Fig. 2 also needs firmer support. These are fixable with additional analysis and careful rewording, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth reading for the dataset: 17 z~0.4 QGs observed with ALMA Band 6, 10 detected in both dust continuum and CO(3-2), plus honest handling of upper limits. That triples the number of QGs with direct dust-to-gas ratio estimates and opens a new parameter space of low gas fractions (median f_H2 ~4%). The wide spread in δ_DGR, from roughly 1/700 to 1/40, is a new empirical result and the variance appears real.\n\nWhere I part company with the abstract is the claim that δ_DGR in QGs reaches ~2.5× above the canonical 1/100, and the implication that dust evolves independently of H2. That above-canonical tail is carried by hCOS-d3, hCOS-d10, and hCOS-d13 (the last a dust-detected, CO-undetected limit). Their M_H2 comes from CO(3-2) with fixed R31 = 0.5 and α_CO = 4.36, both inherited from star-forming galaxy calibrations, with no quoted systematic uncertainty on either. If old QGs have R31 ~ 0.3 and α_CO ~ 7.5 — both within the range people find in low-excitation or CO-faint environments — those three systems drop to or below 1/100, and the 'first confirmation that δ_DGR can exceed SFG values' disappears. The stress test is right about this.\n\nAlso, the Appendix B cross-checks on dust mass all adopt T_dust = 21 K, so they do not test the temperature systematics that could shift M_dust. Calling the range 'robust regardless of how M_dust and M_H2 are estimated' is an overstatement.\n\nOn the positive side, the SED fitting is careful: the D_n4000 prior, the mock-injection checks, and the dust-mass comparisons agree within ~0.2 dex when the temperature is fixed. The selection is clean, and the authors are transparent about detections versus limits. The fast/slow quenching split in Fig. 2 is by-eye comparison to tracks rather than a formal fit, but that is a minor issue for a Letter.\n\nThe conclusion that dust is an unreliable H2 tracer does not die with the high tail; it rests mostly on the lack of correlation between f_dust and f_H2 across the sample, which is fairly robust to a uniform shift in the CO conversion. But the headline number is calibration-dependent, so the abstract and conclusions need to be toned down or accompanied by a sensitivity analysis.\n\nWho should read this: anyone working on cold ISM in quiescent galaxies. It is a dataset paper with real value, and the calibration issue is exactly the kind of thing a referee should push on. I would accept it for peer review, but I would ask for a systematic uncertainty budget on R31 and α_CO, and a version of the key figure showing how the above-canonical systems behave under alternative conversions. If the tail survives, the paper becomes much stronger. As is, the observations are solid and the interpretation is one step ahead of what the data can support.","headline":"New ALMA sample of 12 z~0.4 quiescent galaxies with direct dust and CO gives a real spread in dust-to-gas ratio, but the headline above-canonical tail depends on star-forming-galaxy conversion factors that could easily shift those systems back to normal.","tokens_in":18045,"tokens_out":5582,"would_cite":true,"duration_ms":56945,"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":"This paper shows that old, starless galaxies have widely varying dust-to-molecular-gas ratios, implying that dust and gas evolve independently after star formation stops and that dust alone cannot trace molecular gas.","keywords":["quiescent galaxies","dust-to-gas ratio","molecular gas","galaxy quenching","interstellar dust","CO(3-2)","ALMA","galaxy evolution"],"falsifier":"Observe CO(1-0) and an independent gas tracer such as [CI] in a subset of the galaxies with the highest and lowest reported dust-to-gas ratios. If the CO(3-2)/CO(1-0) ratio is significantly different from 0.5 or the CO-to-H2 conversion factor differs from the adopted value, the claimed spread in delta_DGR would shrink or disappear. Alternatively, measuring dust temperature in multiple submillimeter bands would test whether the dust-mass estimates are biased by the assumed temperature.","tokens_in":16934,"feed_emoji":"🌌","tokens_out":4164,"duration_ms":52964,"temperature":0.7,"pith_summary":"The paper uses deep observations of 17 quiescent galaxies (QGs) at redshift about 0.4 to measure both their dust and molecular gas content directly. It claims that the dust-to-molecular-gas mass ratio in these old, non-star-forming galaxies ranges from roughly 8 times below to 2.5 times above the canonical value of 1/100 used for star-forming galaxies. This spread is present even though their molecular gas fractions are uniformly low, with a median of about 4 percent. The authors interpret this as evidence that dust and molecular gas follow independent evolutionary paths after star formation is quenched, so one cannot be used as a reliable proxy for the other. The result matters because many previous estimates of cold gas in quiescent galaxies at high redshift relied on dust emission and a fixed ratio.","feed_headline":"ALMA finds dust and gas part ways in quiescent galaxies","feed_subtitle":"Old galaxies show dust-to-gas ratios from 8x below to 2.5x above the standard value, undercutting dust-only gas estimates.","key_machinery":"The central diagnostic is the dust-to-molecular gas mass ratio, delta_DGR = Mdust/MH2, which quantifies what fraction of the cold interstellar medium is locked in dust grains. It is obtained by pairing ALMA Band 6 dust continuum (which constrains dust mass via SED fitting to a physically motivated dust emission library) with CO(3-2) line emission (which traces molecular gas via fixed conversion factors). The ratio serves as a direct probe of whether dust and gas evolve together or separately after quenching, and it is the quantity that reveals the order-of-magnitude scatter in the sample.","core_discovery":"We present the first statistical sample of 17 massive quiescent galaxies at z~0.4 with simultaneous ALMA Band 6 dust continuum and CO(3-2) observations. By directly measuring both dust mass and molecular gas mass, we demonstrate that the dust-to-molecular-gas ratio in QGs spans about 1/700 to 1/40, deviating by up to ~8x below and ~2.5x above the canonical value of ~1/100. Despite uniformly low molecular gas fractions (median about 4.1 percent), the galaxies do not follow a single dust-depletion track: about half show a rapid exponential dust decline on a ~700 Myr timescale, while others retain elevated dust-to-gas ratios for more than 2 Gyr. This diversity supports the idea that dust and mo","pith_inferences":["If the decoupling holds generally, high-redshift quiescent samples whose gas masses rely on dust stacking will need to be re-examined with independent gas tracers such as CO(1-0) or atomic carbon lines.","A targeted follow-up measuring CO(3-2)/CO(1-0) ratios and [CI] emission in a subset of these galaxies would directly test whether the assumed conversion factors (R31=0.5, alpha_CO=4.36) are valid in old quiescent systems; deviations would shift the reported ratio spread.","The results suggest that galaxy formation models need to treat dust destruction and regrowth separately from molecular gas depletion rather than assuming a tightly coupled ISM.","One might expect the spread in delta_DGR to correlate with environment or merger history; checking whether these galaxies have companions or disturbed morphologies would test whether external processes drive the decoupling."],"forward_implications":["If dust and molecular gas evolve independently, dust-only measurements of molecular gas in quiescent galaxies can be wrong by up to roughly an order of magnitude.","Quiescent galaxies do not follow one universal interstellar-medium depletion path; rapid and slow dust-decline populations coexist at similar stellar ages.","The most massive systems (stellar mass above ~10^11 solar masses) consistently show dust-to-gas ratios at or below the canonical 1/100, suggesting efficient dust destruction in massive halos.","Elevated dust-to-gas ratios can persist more than a gigayear after quenching, so residual dust in old galaxies does not necessarily indicate recent star formation.","The observed scatter implies that quenching mode is not the sole driver of post-quenching ISM diversity."],"fun_headline_variants":["ALMA reveals dust-gas ratios vary 20x in quiescent galaxies","Rapid vs slow dust decline splits quiescent galaxies","ALMA finds no single dust-to-gas ratio for dead galaxies","Dust and gas part ways in old quiescent galaxies","Quiescent galaxies challenge dust as gas tracer"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The molecular gas masses assume that old quiescent galaxies have the same CO excitation and CO-to-H2 conversion factors as star-forming galaxies, so if either differs systematically, the reported dust-to-gas ratios would shift substantially.","fun_headline_variants_meta":{"raw":{"variants":["ALMA reveals dust-gas ratios vary 20x in quiescent galaxies","Rapid vs slow dust decline splits quiescent galaxies","ALMA finds no single dust-to-gas ratio for dead galaxies","Dust and gas part ways in old quiescent galaxies","Quiescent galaxies challenge dust as gas tracer"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000493,"raw_usage":{"total_tokens":2350,"prompt_tokens":930,"completion_tokens":1420,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":674,"completion_tokens_details":{"reasoning_tokens":1335}},"tokens_in":674,"tokens_out":1420,"duration_ms":11906,"temperature":1.0,"reasoning_tokens":1335,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T18:12:09.430274+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe CO(1-0) and an independent gas tracer such as [CI] in a subset of the galaxies with the highest and lowest reported dust-to-gas ratios. If the CO(3-2)/CO(1-0) ratio is significantly different from 0.5 or the CO-to-H2 conversion factor differs from the adopted value, the claimed spread in delta_DGR would shrink or disappear. Alternatively, measuring dust temperature in multiple submillimeter bands would test whether the dust-mass estimates are biased by the assumed temperature.","supporting_citations":[],"review_version":1}