{"id":"9770ae28-a939-4b15-b502-60d7898402b0","arxiv_id":"2506.13899","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"At z=1-3, the CO-to-H2 conversion factor in simulated disks rises from about 1-5 in the center to over 100 at 15 kpc, making CO an unreliable tracer of extended molecular gas.","lead":"This paper makes simulated maps of carbon monoxide (CO) emission for hundreds of star-forming galaxies at the peak of cosmic star formation and finds that the amount of hydrogen gas per unit CO light rises sharply from each galaxy's center to its outskirts. The result warns that using one fixed CO-to-H2 conversion factor will undercount molecular gas in galaxy outskirts, and that different CO lines trace different parts of the disk.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The predicted αCO gradient and suprathermal CO excitation depend on an unsaturated, linear CR–ΣSFR scaling (Eq. 9) that Sec. 4.2 flags as likely overestimated in centers; the headline αCO range and size ratios lack a CR-sensitivity test.","rationale":"After reading the full manuscript, I find the most load-bearing step is the CR ionization prescription in Eq. (9), not the sub-resolution cloud model per se. The cloud model has independent support: Garcia et al. (2024) validated SLICK against z=0 α_CO–metallicity and M_H2–L'_CO relations, and Fig. 1 shows the localized ISRF improves agreement with local GMC data. At z=1–3, however, the CR field is unconstrained, and the paper itself identifies this sensitivity in Sec. 4.2: the line ratios are sensitive to CR modeling, the normalization is uncertain, and the linear scaling with Σ_SFR may not hold, with Krumholz et al. (2023) suggesting only moderate CR enhancement in starbursts. Sec. 4.1 makes the causal link explicit: CR heating raises Tg and CO(1–0) luminosity, lowering α_CO. Thus the central-to-outskirts α_CO contrast and the suprathermal r31 values both trace the CR assumption. This is a correctness-risk issue rather than an internal inconsistency; the paper's acknowledgment is honest but leaves the headline quantitative claims untested against a plausible alternative. The proposed rerun is inexpensive and would settle whether the α_CO gradient is a robust prediction or partly an artifact of the CR scaling. The qualitative conclusion that CO is a poor tracer of H2 in the outskirts may survive, but the specific ranges and size ratios are conditional. Hence I keep the reader's CONDITIONAL verdict unchanged, with confidence in the direction of the effect but not in the exact numbers.","tokens_in":24942,"tokens_out":13410,"duration_ms":127510,"concrete_test":"Re-run SLICK on the same Simba25 galaxy sample with two CR variants: (i) a fixed ζ_CR = 10^-17 s^-1 independent of Σ_SFR, and (ii) a saturated scaling, e.g., ζ_CR ∝ min(Σ_SFR, Σ_sat), with Σ_sat chosen per Krumholz et al. (2023). Recompute the median radial α_CO profiles (Fig. 4) and CO(3–2)/CO(1–0) profiles (Fig. 5). If central α_CO remains ≤5 and central r31 remains ≥0.9, the central claim survives; if central α_CO increases by more than a factor of 2–3 or the central-to-outskirts α_CO contrast drops below ~10, the headline range and the 27–29% size ratios are conditional on the CR prescription.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. (9) sets the cosmic-ray ionization rate ξ_CR,C = (Σ_SFR,C/Σ_SFR,MW) ζ_-16 ξ_CR,MW, with no saturation or attenuation. In the centers of the massive Simba disks, Σ_SFR is orders of magnitude above the solar neighborhood value, so the model produces ζ_CR ~ 10^-15–10^-14 s^-1. Sec. 4.1 states that the resulting CR heating raises the gas temperature and hence the CO(1–0) brightness, directly lowering α_CO in the centers. Sec. 4.2 then concedes that the Σ_SFR–ξ_CR scaling 'may not hold in all environments' and, citing Krumholz et al. (2023), that starbursts may have only moderately enhanced CR rates, implying central CR heating is likely overestimated. Because the headline α_CO range (1–5 inside 3 kpc vs >100 beyond ~10 kpc) and the 27–29% half-light-to-half-mass size ratios are derived from the same CO(1–0) luminosities, an overestimated central CR rate would flatten the predicted α_CO gradient and weaken the quantitative central claim. The suprathermal r31 > 1 values, used as supporting evidence for the excitation gradient, would be similarly affected. The paper defers alternative CR prescriptions to future work, so no test currently demonstrates that the central result is robust to this acknowledged uncertainty.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses the Simba cosmological simulation combined with the SLICK post-processing pipeline (Garcia et al. 2024) to generate synthetic, spatially resolved CO(1–0) to CO(5–4) maps for star-forming galaxies at z = 0.8–3.5, and compares the CO emission with the underlying H2 mass distribution. The central claims are that the CO-to-H2 conversion factor αCO rises from about 1–5 M⊙ (K km s−1 pc2)−1 in the central 1–3 kpc of massive galaxies to values exceeding 100 at ~15 kpc, that CO half-light radii are on average 27–29% smaller than H2 half-mass radii with roughly 30% of the H2 mass lying beyond two CO half-light radii, and that CO excitation ratios increase toward galaxy centers and with redshift, reaching suprathermal values in warm, dense gas. The paper argues that these results imply that a single αCO cannot be used to trace H2 across z = 1–3 disks and that combined CO plus H2 corrections will be needed for resolved observations.","tokens_in":25389,"tokens_out":12762,"duration_ms":121673,"significance":"If these predictions hold, they are directly relevant to the interpretation of current and upcoming resolved CO observations at z = 1–3: the paper provides a physically motivated, spatially resolved αCO prescription, predicts quantitative size-ratio corrections, and makes falsifiable predictions for CO line-ratio gradients. The work is carefully presented, with percentile spreads, a localized ISRF treatment that improves agreement with z = 0 GMC data, clear comparisons to observed sizes and line ratios, and an explicit list of caveats. The target z = 1–3 outputs are forward predictions rather than fits to the same data, so the exercise is not circular. The main weakness is that the headline quantitative claims depend strongly on the assumed cosmic-ray ionization rate scaling, which is both ambiguously written in Eq. (9) and acknowledged in Sec. 4.2 to be potentially overestimated in galaxy centers; no robustness test is provided. This needs to be addressed before the magnitude of the αCO gradient and the size ratios can be fully endorsed.","major_comments":[{"comment":"The central quantitative claims—the αCO gradient in Fig. 4 and the CO half-light to H2 half-mass radius ratios in Table 1—are directly affected by the assumed cosmic-ray ionization rate. Sec. 4.1 states that the higher gas temperature caused by increased ISRF and CR strength raises the CO(1–0) brightness and thus lowers αCO in the centers. The authors acknowledge in Sec. 4.2 that the linear Σ_SFR–ξ_CR scaling 'may not hold in all environments', and, citing Krumholz et al. (2023), that starburst systems may have only moderately enhanced CR rates, which would imply that central CR heating is overestimated in the present model. Yet no alternative CR prescription or sensitivity test is presented. I request a robustness test with a plausible alternative CR treatment (e.g., a capped CR enhancement or the Krumholz et al. prescription) recomputing at least the αCO(r) profiles, the r31(r) profiles, and the Table 1 size ratios. Without such a test, the magnitude of the predicted αCO range and the 27–29% size ratios are not demonstrated to be robust to an uncertainty the authors themselves flag.","section":"Sec. 4.2; Eq. (9); Fig. 4; Table 1"},{"comment":"The description of the CR normalization is internally inconsistent and invites a factor-of-10 misreading. The text says 'assuming ζ−16 = 0.1 is the CR ionization rate and ξ_CR,MW = 10−16 s−1 is the CR field in the solar neighborhood'. If ζ−16 is a dimensionless prefactor, calling it the ionization rate is wrong; if ζ−16 is already a rate, multiplying it by ξ_CR,MW double-counts the normalization. The product of the two values is 10−17 s−1, consistent with the '10^-17 s^-1' stated in Sec. 4.2, but a reader implementing Eq. (9) literally from Sec. 2.2.3 could easily adopt 10−16 s−1. Please rewrite the sentence and Eq. (9) so that ζ−16 is unambiguously a dimensionless scaling relative to ξ_CR,MW, or state the solar-neighborhood CR ionization rate directly and use only one symbol for it.","section":"Sec. 2.2.3; Eq. (9); Sec. 4.2"}],"minor_comments":[{"comment":"The number of galaxies entering each redshift–stellar mass bin after the merger and convergence cuts is not reported anywhere. Because some bins are strongly affected by the merger cut (the text notes only one high-mass z = 2.5–3.5 galaxy and omits that bin), the 16th–84th percentile spreads would be much more interpretable if N per bin were given in each panel or in Table 1.","section":"Sec. 2.3; Figs. 3–7; Table 1"},{"comment":"The definition of the molecular gas mass used for αCO includes the helium contribution (0.4×M_C) and is used interchangeably with M_H2. Since αCO is then effectively a molecular-gas-to-CO ratio rather than a pure H2-to-CO ratio, the comparison to literature αCO values (e.g., the observed values discussed in Sec. 4.1) should state explicitly whether the observational values include helium (i.e., the usual 1.36 correction).","section":"Sec. 2.2.2; Sec. 3.1"},{"comment":"The observational data points ('diamonds') in Fig. 8 are not individually identified or assigned uncertainties. Please add a legend or a caption note specifying which source (Kaasinen, Ikeda, Rizzo, Tadaki, and where applicable individual galaxies) corresponds to each symbol, so that the comparison is reproducible.","section":"Fig. 8"},{"comment":"The prose in Sec. 2.3 says the three bins are 'around z=1,2,3', while the figures use z=0.8–1.2, z=1.8–2.2, and z=2.5–3.5. This is clear in the figures, but the text should use the wider labels consistently to avoid confusion between the snapshot centers and the bin edges.","section":"Sec. 2.3"}],"recommendation":"major_revision","confidential_remarks":"I support publication after the requested robustness analysis. The CR-sensitivity test is essential: the manuscript itself identifies the linear Σ_SFR–ξ_CR scaling as a likely overestimate in the galaxy centers, and that assumption enters directly into the computation of CO(1–0) luminosities that define the headline αCO gradient and size ratios. On the other hand, the modeling is detailed, the caveats are mostly explicit, and the predictions are of genuine interest to the z = 1–3 molecular gas community. Please also ask the authors to report sample sizes per bin; the statistical interpretation of the percentile spreads currently depends on unknown N."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a read. This is the first systematic look at resolved CO emission through cosmic noon: hundreds of SIMBA galaxies run through the SLICK cloud/Despotic pipeline to yield CO(1-0) through CO(5-4) maps at 500 pc pixels. Prior work was one to three galaxies; this gives statistical radial profiles, size ratios, and excitation gradients. The headline results—alpha_CO rising from 1-5 in massive centers to >100 beyond 10 kpc, CO half-light radii ~27-29% smaller than H2 half-mass radii, high-J more compact, suprathermal r31 in dense central gas—are presented with percentile spreads, and the sample handling is honest: mergers removed, and the one high-mass z~3 galaxy is omitted rather than quoted. The paper is careful about instrumentation too: predicted surface brightnesses are matched to stacked observed values, and the discussion of which transition is most efficient for size measurements is practically useful. The writing is clear about what is model and what is emergent. Soft spots: the steep alpha_CO gradient leans partly on a linear, unsaturated scaling of the cosmic-ray ionization rate with local SFR surface density (Eq. 9). The authors themselves flag in Sec 4.2 that this may overestimate CR heating in starburst centers and cite Krumholz et al. 2023, but defer alternative prescriptions to future work. A CR-robustness test would materially harden the central claim. I'd call this a soft spot, not fatal: the outward decline in CO abundance and rise in alpha_CO are driven by density and chemical gradients that do not depend on CR alone. The reader's factor-10 CR normalization inconsistency doesn't land on reading the paper: Eq. 9 with zeta_-16=0.1 and xi_CR,MW=1e-16 gives 1e-17 s^-1, consistent with Sec 4.2. The notation is confusing, not wrong. Minor: sub-resolution model uncertainties are not propagated into alpha_CO or line ratios, though the quoted percentiles give some sense of spread. Who it is for: observers designing resolved CO surveys at z=1-3, and anyone interpreting CO half-light radii as H2 sizes. It deserves a serious referee. My recommendation: accept with minor revisions—ask for a CR sensitivity test, a clearer definition of zeta_-16, and ideally a data/code release, since the paper is otherwise reproducible in spirit.","headline":"Solid statistical advance in predicting resolved CO at z=1-3; the central gradient is likely robust, but the magnitude hinges on an unscaled CR prescription the authors themselves flag.","tokens_in":25864,"tokens_out":4660,"would_cite":true,"duration_ms":43208,"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 predicts a steep radial rise in the CO-to-H2 conversion factor in z=1–3 galaxies, from ~1–5 in the centers to >100 at ~15 kpc, so CO emission cannot be read as a uniform tracer of molecular gas across these disks.","keywords":["CO emission","molecular gas","H2 mass","CO-to-H2 conversion factor","high-redshift galaxies","cosmic noon","synthetic observations","galaxy sizes"],"falsifier":"Measure resolved CO(1–0) and CO(3–2) emission together with an independent, dust-based molecular-gas surface density map for a sample of z~2 main-sequence galaxies. If $\\alpha_{\\rm CO}$ stays near 3–5 $\\mathrm{M_\\odot}\\,(\\mathrm{K\\,km\\,s^{-1}\\,pc^2})^{-1}$ beyond 10 kpc, or if the CO(1–0) half-light radius matches the H2 half-mass radius instead of sitting about 28% smaller, the predicted steep gradient and size bias are ruled out.","tokens_in":24801,"feed_emoji":"🌌","tokens_out":8049,"duration_ms":74127,"temperature":0.7,"pith_summary":"The paper asks how reliably resolved CO emission traces molecular hydrogen in the peak star-forming epoch, z=1–3, by generating synthetic CO(1–0) through CO(5–4) maps for hundreds of main-sequence galaxies drawn from a cosmological simulation. It argues that the CO-to-H2 conversion factor $\\alpha_{\\rm CO}$ is not constant across these disks: it stays near 1–5 $\\mathrm{M_\\odot}\\,(\\mathrm{K\\,km\\,s^{-1}\\,pc^2})^{-1}$ in the central 1–3 kpc, then climbs steeply to values above 100 at roughly 15 kpc. The consequence is that using a single $\\alpha_{\\rm CO}$, or reading a CO half-light radius as the H2 size, systematically underweights the extended molecular gas: more than a quarter of H2 lies beyond twice the CO(1–0) half-light radius, and CO half-light radii are on average about 27–29% smaller than H2 half-mass radii. If correct, this matters because most resolved CO studies at these redshifts are effectively seeing only the dense centers of galaxies, not their full molecular reservoirs.","feed_headline":"One CO-to-H2 factor underpredicts galaxy outskirts by 100x","feed_subtitle":"Simulated resolved CO maps show the conversion factor rising from ~1-5 in centers to >100 at 15 kpc, making CO sizes too small.","key_machinery":"The carrying mechanism is the SLICK modeling pipeline: each gas particle in the cosmological simulation is treated as a virialized, pressure-confined cloud with a power-law density profile, split into concentric zones; a chemical network solves for H2 and CO abundances under an equilibrium between heating, from photoelectric, dust, cosmic-ray, and CMB processes, and cooling, and a spectral-line code computes CO line luminosities using escape-probability radiative transfer, with the interstellar radiation field and cosmic-ray field set locally from the 64 nearest neighboring clouds. This sub-resolution machinery is what converts the simulation's unresolved gas particles into resolved, roughly 500 pc CO maps, and the paper's radial $\\alpha_{\\rm CO}$ gradients, compact CO sizes, and excitation trends all follow from how these cloud properties vary with galactocentric radius.","core_discovery":"On the paper's own terms, the central discovery is that resolved CO emission in z=1–3 star-forming galaxies is not a uniform tracer of H2. Using sub-resolution cloud modeling plus an equilibrium line-radiation calculation on galaxies from a cosmological simulation, the authors predict that $\\alpha_{\\rm CO(1-0)}$ rises by one to two orders of magnitude from galaxy centers to about 15 kpc, that higher-$J$ CO transitions trace even more centrally concentrated gas, and that CO excitation, as measured by ratios like CO(3–2)/CO(1–0), declines from suprathermal values near 0.9–2 in the centers to roughly 0.1 in the outskirts, with line ratios increasing toward higher redshift and lower stellar mass. They attribute the $\\alpha_{\\rm CO}$ gradient primarily to a declining CO abundance per H2 molecule and to falling density, radiation field, and turbulence at large radius, rather than to metallicity alone. The predicted CO half-light radii are compact, around 1–5 kpc, comparable to current observations, but about 27–29% smaller than the H2 half-mass radii, with about 30% of H2 located outside twice the CO(1–0) half-light radius.","pith_inferences":["A direct observational test could come from combining resolved CO(1–0) with dust-based or other independent H2 maps in the same z~2 galaxies; if those maps confirm that CO half-light radii understate H2 sizes by about 30%, then integrated CO-based gas masses at high redshift may also be biased, not just resolved profiles.","Because the predicted $\\alpha_{\\rm CO}$ gradient is driven more by the local interstellar radiation field and density than by metallicity, galaxy-wide $\\alpha_{\\rm CO}$–metallicity scaling relations may hide large radial scatter, and resolved local-galaxy calibrations are the nearest analogues that could test the sub-grid assumptions.","A testable extension is to rerun the same pipeline with alternative cosmic-ray prescriptions, for instance with a flat rather than star-formation-proportional cosmic-ray field; the paper flags this as a key uncertainty, and since cosmic-ray heating boosts CO(1–0) brightness and high-$J$ excitation, the predicted $\\alpha_{\\rm CO}$ gradient and suprathermal ratios would weaken or strengthen accordin","The result implies that future high-resolution CO surveys of cosmic noon should either include low-$J$ CO(1–0) to capture the extended diffuse gas or explicitly model a radially varying $\\alpha_{\\rm CO}$; otherwise stacked CO sizes will be biased toward compact centers even when individual galaxies show a wide scatter."],"forward_implications":["Observed CO half-light radii at z=1–3 should be interpreted as sizes of the dense, star-forming molecular cores, not the full H2 disk; on average they are about 27–29% smaller than H2 half-mass radii.","Resolved studies that assume a constant $\\alpha_{\\rm CO}$ will systematically underestimate molecular gas surface densities in galaxy outskirts, by factors of tens to hundreds beyond roughly 10 kpc.","Tracing H2 beyond 3–5 kpc in cosmic-noon galaxies is predicted to be very difficult with current facilities because the CO surface brightness drops faster than the H2 density.","Higher-$J$ transitions such as CO(5–4) are progressively more compact and excitation-biased toward dense central gas, so different CO lines cannot be used interchangeably for individual galaxies.","CO excitation gradients, with line ratios peaking in centers and increasing with redshift, provide a way to infer gas physical state: suprathermal ratios require warm, roughly 30–100 K, and dense, above 100 cm$^{-3}$, gas."],"supporting_citations":[{"why":"Supplies the Simba cosmological simulation whose galaxy sample and gas particles are post-processed.","marker":"Davé et al. (2019)"},{"why":"Provides the SLICK pipeline combining sub-resolution cloud modeling with the line calculation code and anchors the method with a z=0 validation.","marker":"Garcia et al. (2024)"},{"why":"Provides the DESPOTIC equilibrium chemistry and line emission code that computes CO luminosities per cloud.","marker":"Krumholz (2014)"},{"why":"Defines the sub-resolution prescription for cloud radii and density structure that SLICK follows.","marker":"Popping et al. (2019)"},{"why":"Supplies the pressure-based cloud radius scaling and the galactic CO modeling approach used for the clouds.","marker":"Narayanan & Krumholz (2017)"},{"why":"Provides the 18-species chemical network used to compute H2 and CO abundances.","marker":"Gong et al. (2017)"},{"why":"Earlier simulation-based resolved CO predictions whose flatter $\\alpha_{\\rm CO}$ profiles the paper contrasts with its own steeper gradients.","marker":"Olsen et al. (2016)"},{"why":"Local resolved $\\alpha_{\\rm CO}$ measurements showing radial variations, used as observational motivation and comparison.","marker":"Sandstrom et al. (2013)"},{"why":"Reference compilation of $\\alpha_{\\rm CO}$ systematics, including dependence on metallicity and physical conditions, that the paper's findings extend.","marker":"Bolatto et al. (2013)"}],"fun_headline_variants":["CO-to-H2 conversion varies 100x across galaxy disks","Resolved CO maps reveal H2 hidden in galaxy outskirts","Single CO factor misses most galaxy H2 by 100x","CO excitation plummets from core to disk edge","Cosmic noon galaxies: CO underreports H2 by 100x"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on treating each unresolved gas particle in the simulation as a single virialized, pressure-confined molecular cloud with equilibrium chemistry and a cosmic-ray field that scales linearly with local star-formation surface density; if z=1–3 gas is not organized this way, the predicted radial $\\alpha_{\\rm CO}$ gradient, compact CO sizes, and excitation trends are model artifacts.","fun_headline_variants_meta":{"raw":{"variants":["CO-to-H2 conversion varies 100x across galaxy disks","Resolved CO maps reveal H2 hidden in galaxy outskirts","Single CO factor misses most galaxy H2 by 100x","CO excitation plummets from core to disk edge","Cosmic noon galaxies: CO underreports H2 by 100x"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00068,"raw_usage":{"total_tokens":3241,"prompt_tokens":1252,"completion_tokens":1989,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":868,"completion_tokens_details":{"reasoning_tokens":1903}},"tokens_in":868,"tokens_out":1989,"duration_ms":12575,"temperature":1.0,"reasoning_tokens":1903,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:56:47.018127+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure resolved CO(1–0) and CO(3–2) emission together with an independent, dust-based molecular-gas surface density map for a sample of z~2 main-sequence galaxies. If $\\alpha_{\\rm CO}$ stays near 3–5 $\\mathrm{M_\\odot}\\,(\\mathrm{K\\,km\\,s^{-1}\\,pc^2})^{-1}$ beyond 10 kpc, or if the CO(1–0) half-light radius matches the H2 half-mass radius instead of sitting about 28% smaller, the predicted steep gradient and size bias are ruled out.","supporting_citations":[{"cited_title":"2024, ApJ, 974, 197","cited_arxiv_id":null,"evidence_quote":"Provides the SLICK pipeline combining sub-resolution cloud modeling with the line calculation code and anchors the method with a z=0 validation."},{"cited_title":"S., Faisst, A","cited_arxiv_id":null,"evidence_quote":"Defines the sub-resolution prescription for cloud radii and density structure that SLICK follows."},{"cited_title":"C., & Wolfire, M","cited_arxiv_id":null,"evidence_quote":"Provides the 18-species chemical network used to compute H2 and CO abundances."},{"cited_title":"P., Greve, T","cited_arxiv_id":null,"evidence_quote":"Earlier simulation-based resolved CO predictions whose flatter $\\alpha_{\\rm CO}$ profiles the paper contrasts with its own steeper gradients."}],"review_version":2}