{"id":"140c3814-05f1-40de-890e-9464a0badb68","arxiv_id":"2509.04818","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Carbon monoxide line observations overestimate molecular cloud turbulent velocity dispersion by about 12-14% on average, and a correction factor R_CO of about 0.88 is derived from synthetic observations of one simulated cloud.","lead":"This paper uses a simulated collapsing turbulent cloud, post-processed with realistic carbon monoxide emission and radiative transfer, to test how opacity and chemistry distort turbulence measurements. It finds that CO-based measurements overestimate the turbulent velocity dispersion by about 12-14% and proposes a correction factor of about 0.88.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The quantitative R_CO correction is calibrated on one smoothing kernel and one simulation; applying it to previous CO studies is not yet supported without a kernel-sensitivity test.","rationale":"The concern is load-bearing because Eq. (7) tells observers to multiply their measured sigma_v,1D by R_CO, and Sec. 4.2 applies it retroactively to prior CO studies. If R_CO is sensitive to the smoothing scale or to the subtraction method, then the 0.88 value is conditional on a specific reduction recipe rather than a universal opacity/chemistry correction. The fix is inexpensive: all maps already exist, so a kernel sweep is a few hours of compute. I do not reject the paper: the qualitative sign (CO overestimates relative to Ideal in most configurations) has internal support, and the caveats section does limit R_CO to solar-neighbourhood-like conditions. But the abstract's sweeping statement about 'previous measurements' requires the kernel/method check. This partially overlaps with the reader's concern (one simulation, many conditions) but adds an internal, testable weakness. Hence the verdict remains conditional and unchanged from the reader's assessment.","tokens_in":19241,"tokens_out":7526,"duration_ms":72828,"concrete_test":"On the same PPV cubes, recompute R_CO for Gaussian smoothing FWHMs of L/4, L/3, L/2, 2L/3, and L (k=4, 3, 2, 1.5, 1) at both t_ff and 1.2 t_ff for all three LOS, keeping all other steps identical. Also compare the Gaussian-kernel result with a linear-gradient subtraction on the same maps, since Sec. 4.2 applies R_CO to prior work that used gradient subtraction. If the time/LOS-averaged R_CO changes by more than 0.05 (about half the quoted uncertainty), or if R_CO differs between the Gaussian and linear-gradient methods, then R_CO is kernel- and method-dependent and Eq. (7) cannot be applied to prior CO studies without a per-method calibration.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central numerical result, R_CO = 0.88 (+0.09/-0.08), is not an intrinsic property of CO lines; it is calibrated on a single MHD simulation and, within that simulation, on a single turbulence-isolation recipe: Gaussian smoothing of the first-moment map with FWHM = L/2 (k=2), described in Sec. 3.2. No test is shown for how R_CO changes with the smoothing scale, even though Appendix D shows the post-isolation power spectra of the Ideal and CO cases are not identical (slopes -3.1 vs -3.0). Because the measured residual dispersion is the square root of the integral of the filtered power spectrum, a different cutoff can change the ratio, and hence R_CO. Sec. 4.2 then recommends applying the same 0.88 factor to measurements by Gerrard et al. (2023, 2024), Brunt (2010), Menon et al. (2021), Sharda et al. (2022), and others, many of which used linear-gradient subtraction rather than the Gaussian filter used here. The bootstrap errors in Table 2 only sample noise within a fixed map; they do not cover the kernel choice or the method mismatch. Thus the headline 10-15% (and up to 40%) overestimate could be partly an artifact of an untested analysis choice.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates how opacity, radiative transfer, and chemistry affect measurements of the three-dimensional turbulent velocity dispersion from CO spectral-line observations. The authors post-process a chemo-dynamical MHD simulation of a collapsing, turbulent molecular cloud (L=2 pc, ~240 Msun, sonic Mach ~3, B=7.5 uG) with the non-LTE radiative-transfer code PyRaTE to produce synthetic PPV cubes for CO(1-0), CO(2-1), and an optically-thin \"Ideal\" case. They isolate turbulent velocity fluctuations in first-moment maps using Gaussian smoothing with FWHM=L/2, compute the 1D velocity dispersion sigma_v from the residual maps, and define a correction factor R_CO = sigma_v,Ideal / sigma_v,CO. The time- and line-of-sight-averaged values are R_CO,1-0 = 0.88 (+0.09/-0.08) and R_CO,2-1 = 0.88 (+0.10/-0.08). The paper argues that previous CO-based estimates of sigma_v were overestimated by about 10-15% on average, with potential 1-sigma overestimates up to 40%, and proposes combining R_CO with the Stewart & Federrath (2022a) factor C_SF to obtain corrected 3D dispersions.","tokens_in":19486,"tokens_out":3181,"duration_ms":27316,"significance":"If the proposed correction is robust, it would provide a simple rescaling that can be applied to a large body of CO-based turbulence measurements, affecting inferred Mach numbers and turbulence driving parameters in molecular-cloud studies. The paper has notable strengths: it uses a realistic chemo-dynamical simulation, a non-LTE RT treatment, direct measurements rather than fitted parameters, and bootstrap uncertainty estimates. The authors are also transparent about some limitations. However, the significance of the universal 0.88 correction hinges on external validity that is not yet established: the measurement comes from one simulation, one filtering scale, and two timesteps, and the paper recommends applying that correction to observations analyzed with a different turbulence-isolation method. The central numerical claim is internally consistent, but its generality is not yet demonstrated.","major_comments":[{"comment":"R_CO is calibrated on a single turbulence-isolation choice, the Gaussian low-pass filter with FWHM = L/2 (k = 2), and no sensitivity test is provided. Because sigma_v,1D is measured from the residual map after filtering, the ratio R_CO can depend on the filter scale; Appendix D in fact shows that the post-filter power spectra of the Ideal and CO cases are not identical (slopes -3.1 vs -3.0). The bootstrap errors in Table 2 only sample noise within a fixed map and do not cover this choice. I request a test with several filter scales or kernel widths (e.g., k = 1, 2, 3, or a range of FWHM values) showing how R_CO changes, or an explicit argument for why R_CO should be independent of the filter scale.","section":"Sec. 3.2 and Eq. (6)"},{"comment":"The application of R_CO to previous CO studies (e.g., Menon et al. 2021, Sharda et al. 2022, Gerrard et al. 2023, 2024) assumes that the correction is transferable across different turbulence-isolation methods. Many of the cited studies used linear-gradient subtraction rather than the Gaussian smoothing used here. Since the measured sigma_v,1D is defined with respect to the adopted isolation procedure, R_CO could be method-dependent. The manuscript should either add a comparison of R_CO obtained with linear-gradient subtraction on the same synthetic maps, or explicitly restrict the recommendation to measurements made with the same Gaussian-filter approach.","section":"Sec. 4.2 and Table 2"},{"comment":"The abstract's general statement that previous measurements were overestimated by 10-15% goes beyond what the single simulation supports. Table 2 shows R_CO ranging from 0.74 to 1.08 across line of sight and time, and Sec. 4.3.1 states that the correction factors 'strictly only apply to the conditions studied'. To support a general correction, the authors should either expand the simulation sample to cover different metallicities, radiation fields, Mach numbers, and column densities, or proportionally soften the abstract and conclusion claims to 'conditions similar to those simulated'. As written, the quantitative claim is presented as a universal correction factor.","section":"Sec. 4.3.1 and Abstract"}],"minor_comments":[{"comment":"The phrase 'small-scal variations' should be corrected to 'small-scale variations'.","section":"Sec. 3.1"},{"comment":"Stewart & Federrath (2022a) and Stewart & Federrath (2022b) are listed with identical journal, volume, and page numbers (MNRAS, 509, 5237); please clarify whether these are distinct papers or a duplicated reference entry.","section":"References"},{"comment":"The power-spectrum analysis would benefit from a sentence describing how the 2D power spectra are computed and azimuthally averaged, including any windowing or tapering applied before the transform.","section":"Appendix D"},{"comment":"The 'Time average of LOS averages' row reports arithmetic means, but it would be useful to state explicitly that this average is unweighted and to note that the quoted uncertainties correspond to the bootstrap percentiles rather than the scatter across LOS.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The core measurement is executed cleanly and the paper is within the journal's scope, but the gap between a single-simulation calibration and the broad applicability claim in the abstract is the main barrier. The requested kernel-sensitivity and method-comparison tests are feasible within the manuscript's scope and should resolve the concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: this is the first attempt to put a correction factor on CO-based turbulent velocity dispersion measurements that accounts for non-LTE radiative transfer and chemistry, and the internal measurement looks honest. The punchline, R_CO = 0.88 for both CO(1-0) and CO(2-1), implies prior CO first-moment estimates of sigma_v were overestimated by 10-15% on average, with tail cases up to 40%. I think that direction is right for the conditions studied, but the paper overreaches when it presents this as a universal correction for all previous CO measurements.\n\nWhat's genuinely new: nobody else has derived an R_CO calibration under non-LTE RT with a chemical network. The ratio is measured, not fitted — Eq. (6) is a direct ratio of dispersions from turbulence-isolated first-moment maps, and Eq. (7) is a chain-rule identity inserting R_CO into the Stewart & Federrath (2022a) pipeline. The bootstrap errors are reported, multiple LOS and two timesteps are included, and the authors are honest in Sec. 4.3.1 that the factors strictly apply only to the conditions studied — a single supercritical, solar-neighbourhood collapsing cloud. Appendix E, on why R_CO exceeds unity at late times along the collapse axis, is good physics.\n\nThe soft spots. First, the universal average rests on one simulation. Table 2 shows R_CO varying from 0.74 to 1.08 across LOS and time; real clouds with different densities, metallicities, or radiation fields could easily sit outside that range. The abstract doesn't carry the caveat that the main text does. Second — the more serious issue — the turbulence-isolation recipe is a single Gaussian smoothing with FWHM = L/2, and there's no sensitivity test for kernel size. The residual dispersion is the integral of the filtered power spectrum, and Appendix D shows the Ideal and CO spectra have slightly different slopes after isolation (-3.1 vs -3.0). A different kernel will change R_CO, and the bootstrap errors don't cover that choice. Third, the paper recommends applying the same factor to measurements by Gerrard, Brunt, Menon, Sharda, and others, most of which used linear-gradient subtraction, not the Gaussian filter used here. That method mismatch is untested. Fourth, code and data are not public, only available on request.\n\nNet: this deserves a serious referee. The central claim is probably correct for solar-neighbourhood conditions, but the blanket application to previous CO studies is premature. The authors should release the artifacts, add a kernel-sensitivity test, and/or soften the abstract to match the stated domain of validity. A moderate revision could make this a solid, widely-cited calibration paper.","headline":"A useful first RT-aware correction factor for CO turbulence inference, but the blanket application to all prior CO measurements is broader than the single-simulation evidence supports.","tokens_in":20041,"tokens_out":3599,"would_cite":true,"duration_ms":28687,"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":"CO spectral lines inflate turbulent velocity dispersions; a factor of 0.88 corrects them.","keywords":["turbulence","molecular clouds","CO spectral lines","velocity dispersion","radiative transfer","first-moment maps","MHD simulations","star formation"],"falsifier":"Compare the turbulence-isolated 1D velocity dispersion from $^{12}$CO(1-0) with that from a genuinely optically thin tracer (or dust-based velocities) in the same cloud, across a range of metallicities and densities; if the ratio departs systematically from $0.88^{+0.09}_{-0.08}$, the universal correction is falsified.","tokens_in":19017,"feed_emoji":"🌪️","tokens_out":17956,"duration_ms":139455,"temperature":0.7,"pith_summary":"Astronomers reading cloud turbulence from CO line widths must contend with the fact that CO is not an optically thin tracer: opacity, excitation, and chemistry can distort the measured velocity dispersion. This paper removes those distortions in synthetic observations by post-processing a magnetohydrodynamic (MHD) chemo-dynamical simulation of a collapsing cloud with a radiative transfer code and comparing CO(1-0) and CO(2-1) first-moment maps with an optically-thin 'Ideal' case. The result is a single conversion factor, $R_{\\rm CO}\\approx 0.88$, by which the CO-measured turbulent velocity dispersion must be multiplied to recover the optically-thin value. If the factor is correct, previous CO-based dispersions were overestimated by about 10-15% on average and up to about 40% at the 1-$\\sigma$ tail, and the corrected values can be inserted directly into the existing 3D turbulence reconstruction pipeline. The factor is calibrated for typical solar-neighbourhood conditions and may differ elsewhere.","feed_headline":"A 0.88 factor corrects CO turbulence measurements","feed_subtitle":"CO line opacity and chemistry bias turbulent velocity dispersions high by 10-15%; the paper's factor restores true values.","key_machinery":"The load-bearing object is the correction factor $R_{\\rm CO} = \\sigma_{v,1D}(\\mathrm{Ideal})/\\sigma_{v,1D}(\\mathrm{CO})$, the ratio of the turbulence-isolated 1D velocity dispersion in the optically-thin, density-weighted 'Ideal' case to the value measured from CO. To isolate turbulence, the method subtracts a Gaussian-smoothed version of each first-moment map, with a kernel FWHM of $L/2$ chosen so that the largest turbulent mode ($k=2$) is preserved while in-fall and rotation are removed; the standard deviation of the residual map supplies $\\sigma_{v,1D}$. The comparison is made possible by non-LTE (not assuming local thermodynamic equilibrium) radiative transfer post-processing of a chemo-dynamical MHD cloud simulation, which produces CO spectral-line data cubes and moment maps in which opacity, excitation, and chemical depletion are the only differences from the Ideal case encoded in $R_{\\rm CO}$.","core_discovery":"The paper's central claim is that 1D turbulent velocity dispersions measured from CO(1-0) and CO(2-1) first-moment maps are systematically larger than the optically-thin reference value, and that the CO value must be multiplied by $R_{\\rm CO}=0.88^{+0.09}_{-0.08}$ (both lines agree within uncertainties) to recover the 'Ideal' value. This gives the corrected 3D pipeline $\\sigma_{v,3D} = C_{\\rm SF} R_{\\rm CO} \\sigma_{v,1D}(\\mathrm{CO})$, where $C_{\\rm SF}$ is the reconstruction coefficient from the earlier optically-thin method. Consequently previous CO-based estimates of $\\sigma_v$ were overestimated by $\\sim10$-$15\\%$ on average, with up to $\\sim40\\%$ overestimation at the 1-$\\sigma$ tail. The average factor is derived for solar-neighbourhood interstellar medium conditions; along the collapse axis at late times CO depletion can push $R_{\\rm CO}$ above unity.","pith_inferences":["Not tested in this paper: applying the calibration to optically thinner isotopologues such as $^{13}$CO or C$^{18}$O would probably give correction factors closer to unity, a prediction testable with the same pipeline.","Not tested in this paper: because the turbulence-isolation step removes structure on scales below $k=2$, varying the smoothing kernel would show whether $R_{\\rm CO}$ depends on the assumed driving scale of turbulence.","If the bias is generic, published CO-based turbulent Mach numbers are systematically high, and models relating turbulence to star formation rates would shift toward less turbulent support once corrected."],"forward_implications":["Published CO-based 1D turbulent velocity dispersions should be multiplied by $R_{\\rm CO}\\approx 0.88$, lowering them by about 10-15% on average and up to about 40% at the 1-sigma tail.","Observers can now go from a CO first-moment map to a corrected 3D dispersion using $\\sigma_{v,3D} = C_{\\rm SF} R_{\\rm CO} \\sigma_{v,1D}(\\mathrm{CO})$, where $C_{\\rm SF}$ comes from the earlier optically-thin method.","Turbulent Mach numbers derived from CO decrease by the same factor, while turbulence driving parameters that scale as $b \\propto \\mathcal{M}^{-1}$ increase by roughly $1/R_{\\rm CO}\\simeq 1.14$.","With the current $0.08$-$0.10$ uncertainty in $R_{\\rm CO}$, the correction is modest but systematic and matters for high-precision comparisons.","The factor is not universal: $R_{\\rm CO}$ can exceed unity for lines of sight along the collapse axis at late times, when CO freezes onto dust grains, so applying 0.88 is only safe for typical solar-neighbourhood conditions."],"supporting_citations":[{"why":"Supplies the optically-thin 'Ideal' reconstruction step whose $C_{\\rm SF}$ factor appears in the corrected pipeline, and defines the baseline the paper recalibrates.","marker":"Stewart & Federrath (2022a)"},{"why":"Provides the chemo-dynamical MHD simulation of a collapsing, magnetised turbulent cloud that is post-processed into synthetic observations.","marker":"Tritsis et al. (2025a)"},{"why":"Describes the non-LTE radiative transfer code used to generate the CO(1-0) and CO(2-1) spectral-line cubes and moment maps.","marker":"Tritsis et al. (2018)"},{"why":"Supplies the molecular collisional and radiative data needed for the CO level-population and optical-depth calculations.","marker":"Schöier et al. (2005)"},{"why":"Introduces the Gaussian-smoothing subtraction used to remove large-scale motions and isolate turbulence in first-moment maps.","marker":"Gerrard et al. (2024)"},{"why":"Motivates the k=2 smoothing scale as the largest turbulent mode and the use of first-moment maps for turbulence statistics.","marker":"Federrath et al. (2016)"},{"why":"Provides the freeze-out rates that set CO depletion, which drives the late-time cases where the correction factor exceeds unity.","marker":"Hasegawa et al. (1992)"},{"why":"Provides the N(H2)/A_V conversion used in the chemical model, fixing where CO forms in the simulation.","marker":"Pineda et al. (2010)"}],"fun_headline_variants":["CO turbulence maps need a 0.88 correction factor","CO-based turbulence overestimated by up to 40%","Correcting CO turbulence: multiply by 0.88","CO opacity skews turbulence: apply 0.88 factor","Turbulence from CO? Correct with 0.88 factor"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The single average correction $R_{\\rm CO}=0.88$ rests on the assumption that one simulated collapsing, magnetised cloud at two ages and three viewing angles is representative of CO emission in real molecular clouds across the solar neighbourhood.","fun_headline_variants_meta":{"raw":{"variants":["CO turbulence maps need a 0.88 correction factor","CO-based turbulence overestimated by up to 40%","Correcting CO turbulence: multiply by 0.88","CO opacity skews turbulence: apply 0.88 factor","Turbulence from CO? Correct with 0.88 factor"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000711,"raw_usage":{"total_tokens":3261,"prompt_tokens":1065,"completion_tokens":2196,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":681,"completion_tokens_details":{"reasoning_tokens":2112}},"tokens_in":681,"tokens_out":2196,"duration_ms":12977,"temperature":1.0,"reasoning_tokens":2112,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:27:25.414418+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compare the turbulence-isolated 1D velocity dispersion from $^{12}$CO(1-0) with that from a genuinely optically thin tracer (or dust-based velocities) in the same cloud, across a range of metallicities and densities; if the ratio departs systematically from $0.88^{+0.09}_{-0.08}$, the universal correction is falsified.","supporting_citations":[],"review_version":2}