{"id":"92e90c21-c5d0-4f66-b2b1-21b2aa3acac1","arxiv_id":"2506.10750","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Beam-corrected gas-to-dust size ratios are similar or slightly smaller in older Upper Sco disks than in younger Lupus disks, contrary to dust evolution predictions.","lead":"This paper measures gas disk sizes from ALMA carbon monoxide maps for twenty planet-forming disks, removes telescope blurring with model fitting, and combines them with dust sizes to compute gas-to-dust size ratios. The ratios do not grow with disk age, contradicting simple pebble drift models, and the result may change how we think pebbles build planets.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Beam-corrected gas sizes rest on an axisymmetric thin-disk model, yet 8/20 fits leave ≥5σ residuals (Sect. 4.2); an untested model-induced bias in R_CO,90% could create the apparent Lupus/Upper Sco reversal.","rationale":"The reader's weakest assumption is the right one: the headline reversal is only as good as the deconvolved sizes, and the paper itself shows signs that the axisymmetric thin-disk model is not a faithful description for a large fraction of the sample. I agree with the reader that the conclusion is overstated. The synthetic recovery test directly addresses whether the fitting pipeline returns unbiased sizes for the specific morphologies, S/N, and masking used here; it is the missing calibration that would turn the concern into a definite finding. I also note two secondary issues that the reader flagged: the region medians overlap within quoted uncertainties (3.02+0.33/-0.33 vs 2.46+0.53/-0.38, a ~0.9-sigma difference), and the internal numeric inconsistencies (2.78 vs 2.87; 2.46 vs 2.78) need correction. The paper deserves credit for a transparent method, an honest residual analysis, and a reproducible Zenodo implementation; the method is useful. But the central claim should be phrased as a null result absent the recovery test. If the recovery test shows small and region-symmetric biases, the CONDITIONAL verdict can be upgraded; if not, the 'contrary to models' claim should be softened.","tokens_in":28588,"tokens_out":9610,"duration_ms":109693,"concrete_test":"Build a synthetic-recovery benchmark: take the actual 12CO channel maps and replace the observations with model cubes generated from a grid of thermochemical disk models (e.g., Trapman et al. 2023) with known R_CO,90% spanning 0.2–5 arcsec, known inclinations (including flared emission surfaces with height 0.2–0.5 arcsec and 30–50% azimuthal asymmetries), observed with the AGE-PRO beams and noise, masked with the same Keplerian masks, and run the Section 2.1 astropy Nuker/Sersic fitting pipeline. Compare recovered versus input R_CO,90% per source, and compute the median bias separately for Lupus-like and Upper Sco-like configurations. If the median bias exceeds ~10% or differs by more than ~10% between regions, the size-ratio reversal is not robust; if the model inclination is recovered with >20° error for any well-resolved source, flag that source's size as unreliable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the R_CO,90% values in Table 1 be unbiased physical gas sizes. The pipeline (Sect. 2.1) fits a beam-convolved, azimuthally symmetric Nuker/Sersic profile to the moment zero map and then measures the unconvolved model's 90%-flux radius. Section 4.2 documents ≥5σ residuals for eight of twenty sources, with structured patterns: an elevated emitting layer (Lupus 10, Upper Sco 8), East-West asymmetries (Lupus 1, Upper Sco 7), and possible companion/cloud contamination (Lupus 6, Upper Sco 5, Lupus 2). For Upper Sco 8 the best-fit gas inclination is 16.2° versus 56.2° from continuum, so the model is clearly compensating for a flared surface with an incorrect geometry; this directly biases the deprojected radius. For the compact, marginally resolved disks (e.g., Lupus 4, 5, 8, 9; Upper Sco 2, 5), the outer Nuker slope is unconstrained by data yet sets R_90 after deconvolution; Appendix C only cross-checks between analytic profiles, not against the true morphology. If the bias differs between Lupus and Upper Sco (e.g., more flaring or more asymmetry in one region), the reversed median ratio could be a fitting artifact rather than an age trend. The reported numeric inconsistencies (median 2.78 in §3.2 vs 2.87 in §5; Upper Sco 2.78 in §3.2 vs 2.46 in §5) further undermine confidence in the exact medians.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents image-plane fitting of beam-convolved 2D Nuker and Sérsic profiles to 12CO J=2-1 moment-zero maps for ten Lupus and ten Upper Sco disks from the AGE-PRO ALMA Large Program. From the unconvolved best-fit models the authors measure gas disk sizes R_CO,90%, which they combine with FRANK dust sizes from Vioque et al. (2025) to compute gas-to-dust size ratios. They report a full-sample median ratio near 2.8, with Lupus (younger) at 3.02 and Upper Sco (older) at 2.46, and interpret the absence of an increasing ratio with age as evidence that pebble drift is halted in dust traps and/or that external photoevaporation truncates the Upper Sco gas disks. The paper also releases a public astropy-based implementation of the fitting method.","tokens_in":28939,"tokens_out":5494,"duration_ms":68224,"significance":"The AGE-PRO sample and the beam-correction method are valuable: the paper provides deconvolved gas sizes for 20 disks, makes the code available, and explicitly examines residuals and cloud-contamination effects. If the central result were robust, it would challenge the simple pebble-drift expectation that gas-to-dust size ratios increase with disk age and would strengthen the case for dust traps and external photoevaporation. However, the headline claim is not firmly established: the Lupus and Upper Sco median ratios overlap at the quoted 16th-84th percentiles, and the axisymmetric thin-disk fitting leaves significant residuals for 8 of 20 sources, so the apparent reversal could be a fitting artifact. The physical discussion is reasonable but is presented as an explanation of a trend whose statistical and systematic status is not yet secured.","major_comments":[{"comment":"The central claim is not supported by the quoted statistics. Section 3.2 states that \"within these uncertainties the size ratios of the two regions are the same,\" and the 16th-84th ranges (3.02+0.33/-0.33 for Lupus versus 2.46+0.53/-0.38 for Upper Sco) overlap substantially. Yet the abstract and Section 5 present the comparison as \"contrary to models of dust evolution\" and as a finding that Lupus has \"similar (or even larger)\" ratios. No significance test (e.g., a bootstrap p-value for the difference of the medians) is provided. The manuscript should either reframe the conclusion to state that no significant evolution is detected, or add a quantitative test that justifies the stronger claim.","section":"Abstract and §3.2/§5"},{"comment":"The beam-corrected gas sizes rest on an axisymmetric thin-disk model, but Section 4.2 documents multiple ≥5σ residual features for 8 of 20 disks, including elevated emitting layers (Lupus 10, Upper Sco 8), East-West asymmetries (Lupus 1, Upper Sco 7), and possible companions or cloud contamination (Lupus 6, Upper Sco 5). For Upper Sco 8, Table 1 gives a fitted gas inclination of 16.2° while the continuum inclination is 56.2° (Section 4.2), which indicates that the model is compensating for a flared surface with an incorrect geometry. This can directly bias the deprojected R_CO,90%. The paper does not quantify how these residuals propagate into R_CO,90% or demonstrate that the bias is region-independent. Because the central result is a comparison between regions, an injection-recovery test with flared or asymmetric synthetic disks, or a cross-check against a complementary modeling approach, is needed before the reversal can be considered physical rather than an artifact of the fitting procedure.","section":"§2.1, §4.2, Table 1"},{"comment":"The handling of upper limits in the median-ratio calculation is not specified. Section 3.2 notes that four Lupus and three Upper Sco sources have R_FRANK_dust,90% upper limits, and Table 1 lists Lupus 9 as having a gas-size upper limit of <0.09 arcsec. The Monte Carlo procedure described in Section 2.1 draws from \"its uncertainty,\" but no prescription is given for how a censored value is treated. If upper limits are replaced by point estimates or included as if they were Gaussian measurements, the region medians could shift systematically. This is load-bearing for the Lupus-versus-Upper Sco comparison and should be described explicitly, with a sensitivity test showing the effect of censoring versus replacement.","section":"§3.2, §2.1"}],"minor_comments":[{"comment":"There are internal numeric inconsistencies: Section 3.2 gives the full-sample median as 2.78+0.37/-0.32 and the Upper Sco median as 2.78+0.53/-0.38, while Section 5 gives 2.87+0.38/-0.36 and 2.46+0.53/-0.38, respectively; the abstract also quotes 2.46 for Upper Sco. These must be reconciled.","section":"§3.2 and §5"},{"comment":"In the first paragraph of Section 4.2, the sentence \"For Lupus 2 that is reason to include it from the analysis here\" appears to mean \"exclude it from the analysis\"; please clarify.","section":"§4.2"},{"comment":"The right panel of Figure 5 marks substructure status, but the text does not state how the categories \"yes,\" \"maybe,\" and \"no\" are assigned beyond a reference to Vioque et al. (2025). A one-sentence definition would help the reader interpret the panel.","section":"Figure 5 and Section 3.2"},{"comment":"The source labeling is inconsistent between \"Upper Sco\" in the text and \"UppSco\" in several figure panels and appendix figures; please standardize.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is part of a coordinated series; several key quantities (dust sizes, substructure classifications, external photoevaporation modeling) come from companion papers that are still in press. The referee report is written with the caveat that those papers are not yet available for independent verification. The fitting code and the careful residual analysis are strengths, but the central age-trend claim needs substantial reframing or additional validation before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the beam-corrected sizes for 20 AGE-PRO disks are a real resource, and the residual analysis is more careful than most. The headline physics claim, though, is softer than the conclusions imply: the Lupus and Upper Sco gas-to-dust medians overlap within uncertainties, and the per-region difference could come from the fitting model as much as from disk evolution.\n\nWhat is new: the paper implements a Sanchis-style beam-convolved profile fit in astropy, applies Nuker and Sersic models to 12CO moment zero maps, and delivers deconvolved R_CO,90% and gas-to-dust ratios for twenty disks. The code is on Zenodo, the masking of cloud-contaminated position angles is described, and Appendix C cross-checks Nuker against Sersic sizes. Those measurements will be useful to the disk community. The per-source residual discussion in Section 4.2 is genuinely careful: they identify elevated emitting layers, east-west asymmetries, and possible companions rather than sweeping them under the rug.\n\nSoft spots: the central age comparison is not established. The two medians (3.02 vs 2.46) overlap at the 16th-84th percentiles, and Section 3.2 explicitly says \"within these uncertainties the size ratios of the two regions are the same.\" The conclusions then drop that qualifier and call the ratio \"larger\" and the trend \"the opposite\" of models. The abstract's \"similar (or even larger)\" is fairer, but the abstract and conclusions should match. There are also numeric inconsistencies: Upper Sco's median appears as 2.78 in Section 3.2 and 2.46 in the abstract and conclusions; the overall median is 2.78 in Section 3.2 and 2.87 in Section 5. Those should be fixed before publication.\n\nThe deeper worry is model bias. Eight of twenty fits leave multiple >5-sigma residuals. Lupus 10 and Upper Sco 8 show the X-shaped signature of an elevated emitting layer that an axisymmetric, geometrically thin model cannot reproduce, and Upper Sco 8's fitted gas inclination (16 degrees) versus continuum (56 degrees) is a clear warning sign. For the compact, marginally resolved disks, the outer Nuker slope is unconstrained yet sets R_90. A region-dependent bias in these fits could create the apparent reversal. The paper acknowledges most of this, but still frames the result as contradicting dust evolution models.\n\nVerdict: worth publishing after minor-to-moderate revision. The measurements and method are solid enough to cite, and the sample is new. The authors should either soften the conclusion to \"no evidence for an increasing gas-to-dust size ratio with age\" or demonstrate robustness by removing the worst-fit sources and trying models with vertical structure. This deserves a serious referee.","headline":"A useful and honest measurements paper whose claimed reversal of the expected gas-to-dust size ratio trend is not actually supported by the data.","tokens_in":29617,"tokens_out":4749,"would_cite":true,"duration_ms":49987,"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":"Contrary to pebble-drift models, gas-to-dust size ratios do not grow with disk age; younger Lupus disks match or exceed older Upper Sco disks.","keywords":["protoplanetary disks","gas-to-dust size ratio","pebble drift","dust evolution","ALMA","12CO emission","Nuker profile","Lupus"],"falsifier":"Measure $R_{\\rm CO,90\\%}$ for the eight disks with significant residuals (including Lupus 1, Lupus 10, Upper Sco 7, and Upper Sco 8) using a model that includes an elevated emitting surface or an azimuthal asymmetry, or exclude these disks and recompute the two regional medians; if the Lupus-Upper Sco gap disappears or reverses, the central claim fails.","tokens_in":28384,"feed_emoji":"🔭","tokens_out":4428,"duration_ms":51364,"temperature":0.7,"pith_summary":"The paper tries to establish whether the gas-to-dust size ratio of protoplanetary disks grows as disks age, as pebble-drift models predict. To answer, it fits beam-convolved two-dimensional profiles to 12CO moment zero maps of twenty disks from the AGE-PRO ALMA survey, ten young Lupus disks (about 1-3 million years old) and ten older Upper Sco disks (about 2-6 million years old), then measures deconvolved gas disk radii that enclose 90 percent of the CO flux. Combined with continuum-derived dust radii, the ratios scatter between roughly 1 and 5.5. The central claim is that the younger Lupus disks have a median ratio (3.02) similar to or larger than the older Upper Sco disks (2.46), opposite to the predicted increase. The paper proposes halted pebble drift in dust traps plus external photoevaporation in Upper Sco as explanations, while also noting that survivorship bias could play a role.","feed_headline":"Young disks show wider gas-to-dust size gaps than older ones","feed_subtitle":"Beam-corrected ALMA sizes of 20 disks contradict the pebble-drift clock of disk evolution.","key_machinery":"A two-dimensional Nuker profile, an axisymmetric double power law meeting at a break radius, convolved with the clean beam and fitted pixel-wise to 12CO J=2-1 moment zero maps using astropy's Levenberg-Marquardt fitter, with a Sersic profile used for two cloud-contaminated sources. The work it does is to deconvolve the beam from the emission so that the radius enclosing 90 percent of the unconvolved model flux, $R_{\\rm CO,90\\%}$, is an unbiased gas disk size; the gas-to-dust ratio then compares that radius with the 90-percent-flux dust radius from continuum visibility fits.","core_discovery":"Beam-corrected CO sizes yield gas-to-dust size ratios that do not increase with disk age. From the unconvolved best-fit models, the median ratio for Lupus is $3.02^{+0.33}_{-0.33}$ while for Upper Sco it is $2.46^{+0.53}_{-0.38}$; within the propagated uncertainties the two regions are statistically consistent, and if anything the trend runs opposite to dust-evolution models. The result implies either that pebble drift stalls early in dust traps that fix the dust radius, that external photoevaporation in Upper Sco shrinks the gas disk, or that survivorship bias hides the predicted evolution.","pith_inferences":["If the flat ratio trend is real and general, inward pebble drift over the 1-6 million year window is far less efficient than standard models assume, which would lengthen the timescales for pebble-fed planet formation in the outer disk.","A natural test is to repeat the median ratio excluding the eight disks with significant model residuals; if the Lupus-Upper Sco gap closes, the apparent anti-trend is an artifact of profile mismatch rather than a physical age effect.","The photoevaporation explanation predicts that Upper Sco disks should show sharper outer CO cutoffs relative to their dust edges; comparing outer profile slopes between regions would test it.","Higher-resolution continuum imaging of Lupus could reveal the unresolved dust traps that the model requires, since the current roughly 22 au resolution may hide them."],"forward_implications":["Observed gas disk sizes in compact sources can be overestimated by about one beam major axis, roughly 0.3 arcseconds, when measured directly from images; model fitting corrects this.","For the AGE-PRO sample the median gas-to-dust size ratio is 2.78, consistent with earlier surveys of other regions.","Because the ratio does not rise from about 1-3 million years to about 2-6 million years, simple viscous-plus-drift evolution is not enough to describe these disks; dust traps or external photoevaporation must be invoked.","Larger gas disks (about 100 au and above) are the ones that show detected dust substructures, which may partly be a resolution effect.","The best-fit profiles of the larger disks show a shallow inner power law and a steep outer drop, matching the expected optically thick CO intensity shape."],"supporting_citations":[{"why":"Supplies the image-plane fitting approach and the Monte Carlo uncertainty method that this work extends to a wider family of profile shapes.","marker":"Sanchis et al. (2021)"},{"why":"Provides the FRANK continuum visibility fits and the dust disk radii $R_{\\rm dust,90\\%}$ used to compute the gas-to-dust ratios.","marker":"Vioque et al. (2025, in press.)"},{"why":"Establishes that a gas-to-dust size ratio above four is a clear sign of pebble drift while lower ratios can be explained by optical depth effects.","marker":"Trapman et al. (2019)"},{"why":"The previous survey that found no clear evolution of the size ratio with age, which this paper's older and younger region comparison supersedes.","marker":"Long et al. (2022)"},{"why":"One of the dust evolution models predicting that the gas-to-dust size ratio should increase with time as the dust disk shrinks.","marker":"Rosotti et al. (2019)"},{"why":"A model study predicting a steep increase of the ratio when viscous evolution is included, against which the observed flat trend is compared.","marker":"Toci et al. (2021)"},{"why":"Analyzes external FUV photoevaporation for the AGE-PRO Upper Sco disks and shows it can reduce $R_{\\rm CO,90\\%}$ and the size ratio.","marker":"Anania et al. (2025, in press.)"}],"fun_headline_variants":["Gas-to-dust size ratio fails to rise as disks age","ALMA sizes show pebble drift not tied to disk age","Disk gas-to-dust ratios contradict pebble drift models","Young disks have wider gas-to-dust size gaps than old","Beam-corrected sizes flip expected disk evolution trend"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"A single smooth, axisymmetric intensity profile, after deconvolution from the telescope beam, recovers each disk's true 12CO radial brightness well enough that the 90-percent-flux radius is unbiased.","fun_headline_variants_meta":{"raw":{"variants":["Gas-to-dust size ratio fails to rise as disks age","ALMA sizes show pebble drift not tied to disk age","Disk gas-to-dust ratios contradict pebble drift models","Young disks have wider gas-to-dust size gaps than old","Beam-corrected sizes flip expected disk evolution trend"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000323,"raw_usage":{"total_tokens":1867,"prompt_tokens":1053,"completion_tokens":814,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":669,"completion_tokens_details":{"reasoning_tokens":730}},"tokens_in":669,"tokens_out":814,"duration_ms":8891,"temperature":1.0,"reasoning_tokens":730,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:19:22.716345+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure $R_{\\rm CO,90\\%}$ for the eight disks with significant residuals (including Lupus 1, Lupus 10, Upper Sco 7, and Upper Sco 8) using a model that includes an elevated emitting surface or an azimuthal asymmetry, or exclude these disks and recompute the two regional medians; if the Lupus-Upper Sco gap disappears or reverses, the central claim fails.","supporting_citations":[{"cited_title":"T., Trapman , L., et al","cited_arxiv_id":null,"evidence_quote":"Provides the FRANK continuum visibility fits and the dust disk radii $R_{\\rm dust,90\\%}$ used to compute the gas-to-dust ratios."}],"review_version":1}