{"id":"216adb9d-b6b7-4def-a836-f4825fe90df6","arxiv_id":"2505.24129","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A compilation of 237 disc galaxies shows little evolution in gas velocity dispersion from z~1.5 to z~8, with ionized gas dispersions about 2 times higher than molecular gas at fixed gas mass.","lead":"This paper combines published measurements of gas motion in 237 faraway disc galaxies to trace how turbulent their gas is from when the universe was about one billion years old to about five billion years old. It finds that gas turbulence changes little over this time, and that hotter ionized gas is about twice as turbulent as colder molecular gas at the same gas mass.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 2x ionized/molecular offset and flat sigma(z) trend rest on dispersions from different tracers that are nearly perfectly confounded with analysis code and redshift, with no homogenization or matched-control check; the key table is not released.","rationale":"The paper is a valuable literature compilation and its updated Toomre model, built from independent sSFR and t_dep prescriptions rather than fit to the dispersion data, is a genuine prediction that lowers the circularity risk. The qualitative claim of mild evolution in sigma(z) is also broadly consistent with recent literature. However, the single most load-bearing problem is the unresolved confounding between gas phase, redshift, and kinematic extraction technique. The paper explicitly chooses not to homogenize methods in Section 2.3, and Section 4.2.1 acknowledges that the phase offset could partly be a method effect. Because the ionized and molecular samples are overwhelmingly measured with different codes on different lines at different redshifts, the approximately 2x offset in Fig. 6 and the apparent flatness of sigma(z) could both be produced by systematic, technique-dependent biases rather than by ISM physics. The small size and uncertain construction of the gas-mass-matched subsample, plus the absence of the full data table, make it impossible to audit how many points drive the offset. The reader's conditional verdict is therefore appropriate; my concern sharpens the same weakest assumption but does not change the verdict.","tokens_in":26235,"tokens_out":5645,"duration_ms":74308,"concrete_test":"Restrict the Fig. 6 phase-offset analysis to the redshift window z=1.0-1.6 where both H-alpha/KMOS and CO/3D-BAROLO samples are present, and recompute the ionized-to-molecular dispersion ratio at fixed M_gas. Separately, for a subset of galaxies with public data cubes, run a single forward-modeling code (e.g., 3D-BAROLO) on both tracers and compare recovered sigma; if the offset drops below ~1.5x or vanishes when code and redshift are matched, the phase-offset claim is not robust. Finally, require direct (not scaling-relation) molecular gas masses for the ionized sample and release the full Table A1/A2 so the sample can be audited.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative result, the ~2x ionized/molecular offset in Fig. 6, is a between-sample comparison in which gas tracer, redshift, and kinematic code are nearly perfectly confounded. Most ionized-gas dispersions come from KMOS H-alpha at z~0.6-2.7 fitted with DysmalPy; most molecular-gas dispersions come from ALMA CO/[CI]/[CII] at z>3 fitted with 3D-BAROLO. Section 2.3 states explicitly that methods are not homogenized, and Section 4.2.1 concedes that part of the offset 'may result from the different methods typically used to measure dispersion across samples,' citing Lee et al. (2024a) for code-dependent biases in low-S/N data. No matched analysis, correction, or control experiment is presented to show the offset survives tracer-independent measurement. The same confound threatens the flat sigma(z) trend: if high-z mm/FIR dispersions are systematically biased low by resolution and S/N effects, an intrinsically rising sigma(z) would appear flat. A second, compounding issue is that the ionized-gas gas masses in Fig. 6 are partly based on scaling relations (Tacconi et al. 2020) rather than direct observations (Sec. 4.2.1), and the full Table A1/A2 is not released, so the size and composition of the subsample driving the offset cannot be audited. The model update using independent sSFR and t_dep relations is a genuine prediction and the paper is transparent about its heterogeneity, but the headline quantitative claims are not yet isolated from these systematic confounds.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper compiles literature measurements of gas velocity dispersion and rotational velocity for 237 disc galaxies at z=0.5-8, with a subset of 63 galaxies having molecular gas mass estimates. The authors compare the compiled dispersions with a single-phase Toomre stability model, updating the earlier W15 model by adopting external prescriptions for sSFR(z) from Speagle et al. (2014) and depletion time from Tacconi et al. (2020). They report two central results: (1) once stellar mass and gas tracer are accounted for, disc velocity dispersions show little evolution between z~1.5 and z~8, consistent with the updated model; and (2) at fixed molecular gas mass, ionised gas dispersions are about twice molecular gas dispersions. They also discuss the mixed behaviour of [CII] kinematics as a function of star formation rate.","tokens_in":26557,"tokens_out":4050,"duration_ms":45741,"significance":"The compilation itself is a useful community resource, and the paper is unusually transparent about its heterogeneity: Section 2.3 explicitly states that kinematic methods are not homogenised, and Section 2.2 details the assumptions behind molecular gas mass conversions. A clear strength is that the updated Toomre model predictions are not fit to the dispersion data; they are built from independent empirical relations for sSFR and depletion time, making the comparison a genuine prediction. If the reported 2x ionised-to-molecular offset and the flat sigma(z) trend survive accounting for the confounds described below, the results would be an important constraint on multi-phase disc turbulence and on simulations of high-redshift disc structure. At present, however, the headline quantitative claims are not isolated from strong systematic effects in the heterogeneous sample, so the significance of the results is conditional on additional control analyses.","major_comments":[{"comment":"The central ~2x ionised-to-molecular offset at fixed Mgas is a between-sample comparison in which gas tracer, redshift, and kinematic fitting code are nearly perfectly confounded: most ionised-gas dispersions come from KMOS Halpha at z~0.6-2.7 fitted with DysmalPy, while most molecular-gas dispersions come from ALMA CO/[CI] at z>3 fitted with 3D-BAROLO. The paper explicitly declines to homogenise methods (Section 2.3) and concedes in Section 4.2.1 that part of the offset 'may result from the different methods typically used to measure dispersion across samples,' citing Lee et al. (2024a) for code-dependent biases at low S/N. As written, no matched analysis, common-redshift control, or quantitative propagation of these known biases is presented; I request such a test before the physical offset can be regarded as established.","section":"§2.3, Fig. 6"},{"comment":"The Mgas values used to define the offset include, for the ionised-gas sample, molecular gas masses derived from the Tacconi et al. (2020) scaling relations rather than from direct molecular gas observations. Since the offset is measured at fixed Mgas, any systematic error in these scaled masses moves the ionised points horizontally in Fig. 6 and can create or inflate the apparent vertical offset. The full Tables A1/A2 are not released in the arXiv manuscript, so the size and composition of the subsample driving the offset cannot be audited. I ask for release of the full tables, clear flags for which Mgas values are direct versus scaled, and a demonstration that the ~2x offset persists when restricted to galaxies with directly measured molecular gas masses.","section":"§4.2.1, Tables A1/A2"},{"comment":"The claim of 'little evolution' in sigma between z~1.5 and z~8 is supported only by visual inspection of scatter plots with considerable scatter and no quantitative model comparison. Because the high-redshift mm/FIR measurements are preferentially made with codes that can underestimate dispersions in low-S/N data (Section 2.3), an intrinsically rising sigma(z) could appear flat; a regression or binned comparison that includes redshift, tracer, stellar mass, and analysis code as covariates is needed to establish the evolution claim. Without this, the Conclusions statement that 'there is no evolution in sigma within the errors between z~1 and z~8' is stronger than the analysis presented.","section":"§3.4, §4.1, Figs. 2 and 5"}],"minor_comments":[{"comment":"There is a duplicated word in the cosmology sentence: 'We assume a Chabrier (2003) initial mass function and and assume a flat LambdaCDM cosmology.'","section":"§1"},{"comment":"There is a typo: 'withJWST will reduce this bias...' should read 'with JWST' in the sentence beginning 'Observations withJWST...'.","section":"§2.3"},{"comment":"The Conclusions list '[OI]' as a molecular gas tracer, but Section 4.2.1 and Fig. 6 classify only CO and [CI] as molecular tracers and treat [CII] and [OIII] separately; please reconcile this wording with the body of the paper.","section":"§5"},{"comment":"The statement 'The data compilation for this analysis will be provided online' should be updated to include the full machine-readable Tables A1/A2 in the arXiv version, particularly because Section 4.2.1 relies on a 63-galaxy subset that cannot be independently audited from the excerpted tables.","section":"Data Availability"},{"comment":"The caption states that errors are propagated assuming 0.3 dex on stellar mass and 30% on SFR, but the text does not describe how these uncertainties are combined with the kinematic and Mgas uncertainties; a sentence in Section 2 would clarify the error budget.","section":"Fig. 6 caption"},{"comment":"Table 1 lists Lelli et al. (2021) as a source, while Table A1 lists zC400569 and zC488879 with Lelli et al. (2023); please ensure the reference list and table entries are consistent.","section":"Table 1 and Table A1"}],"recommendation":"major_revision","confidential_remarks":"The paper's topic and compilation are well suited to the journal, and the authors are candid about limitations. The main barrier is not the absence of homogenisation per se—which the authors openly acknowledge—but that the two headline quantitative results (the 2x offset and the flat sigma(z) trend) are not yet demonstrated to be robust to the strong confounds between tracer, redshift, and analysis code. These are fixable with additional analysis and data release, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a genuinely useful paper: a literature compilation of 237 disc galaxies with homogenised gas masses, and a clear, honest model comparison. The SFR-dependent interpretation of [CII] kinematics is a fresh synthesis, and the updated Toomre model uses independent empirical relations for sSFR (Speagle et al. 2014) and depletion time (Tacconi et al. 2020), so the model prediction is not fit to the dispersion data. Low circularity. The authors also deserve credit for being upfront about heterogeneity — Section 2.3 says they did not homogenise methods, and Section 4.2.1 concedes part of the phase offset may come from different analysis codes. That transparency is real.\n\nBut the soft spot is central, and the stress-test note gets it right. The 2x ionized/molecular offset at fixed M_gas is a between-sample comparison in which tracer, redshift, and kinematic code are nearly perfectly confounded: most ionised dispersions are KMOS H-alpha at z~0.6-2.7 fitted with DysmalPy, most molecular dispersions are ALMA CO/[CI]/[CII] at z>3 fitted with 3D-BAROLO. The statement that no obvious code bias is seen in the figures is not testable from those figures, because the codes are nested within tracers. Known code-dependent biases (Lee et al. 2024a) could plausibly produce or inflate the offset. The same confound threatens the flat sigma(z) trend: if high-z mm dispersions are systematically biased low by resolution and S/N, an intrinsically rising sigma would look flat. The paper offers no matched-control check, no same-galaxy multi-tracer sample, and no systematic error budget. Also, the gas masses for the ionised gas sample at cosmic noon are partly from scaling relations rather than direct detections, which could shift the x-axis in Fig. 6. And the full data table is not released, only a few rows in the appendix, so the composition of the subsample driving the offset cannot be audited.\n\nA second, milder issue is that the 'little evolution' conclusion is drawn from scatter plots and a qualitative model-band comparison. There is no fit to the data, no treatment of the upper limits as censored points, and no quantitative test of whether the model actually matches the distribution. That is fine for a qualitative synthesis, but the abstract states it as a firm result.\n\nWho is this for? People working on high-z disc kinematics and multi-phase ISM. The compilation is a useful resource, and the paper is a fair summary of the current landscape. A serious referee should not reject it, but should push for either a matched-control analysis or a moderation of the quantitative claims, plus the full data release. I would send it to review and ask for major revision on the confound issue.","headline":"Useful compilation and honest model comparison, but the headline offset and flat trend rest on tracer/code/redshift confounds that the paper acknowledges but never quantifies.","tokens_in":27187,"tokens_out":2728,"would_cite":true,"duration_ms":34214,"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":"Disc gas velocity dispersion barely evolves between z≈1.5 and z≈8 once stellar mass and gas tracer are accounted for.","keywords":["galaxy evolution","disc kinematics","velocity dispersion","Toomre stability","ionised gas","molecular gas","[CII] kinematics","high-redshift galaxies"],"falsifier":"Re-fit every public data cube in the compilation with one forward-modelling code and a uniform beam, sensitivity, and spectral-resolution treatment. If the homogenised values erase the ~2× offset between ionised and molecular gas at fixed M_gas, or if σ_molecular rises steeply with redshift once beam-smearing and tracer differences are removed, the claims fail. A cheaper check: the same galaxies observed in both Hα and CO at z ≈ 1–2 should reproduce the offset within individual systems.","tokens_in":26031,"feed_emoji":"🌌","tokens_out":9985,"duration_ms":89499,"temperature":0.7,"pith_summary":"The paper assembles 237 rotation-supported disc galaxies from the literature, spanning redshifts 0.5 to 8, to ask whether the high gas turbulence seen at cosmic noon continues to rise toward the earliest epochs or flattens out. It argues that, once stellar mass and the gas phase being traced are taken into account, the velocity dispersion of disc gas shows little evolution between z~1.5 and z~8, in line with a simple Toomre marginal-stability model updated with modern gas-depletion and star-formation-rate relations. The same compilation shows that ionised gas (Hα, [OIII]) dispersions are about twice as high as molecular gas (CO, [CI]) dispersions at a fixed molecular gas mass, and that [CII] sits between these sequences, with its position set by star-formation rate. If these results hold, they dissolve an apparent tension between dynamically cold high-redshift molecular discs and turbulent ionised discs, and they imply that disc turbulence is governed mainly by gas supply and star formation rather than by cosmic epoch.","feed_headline":"237 galaxies show disc turbulence barely changes from z=1.5 to z=8","feed_subtitle":"Ionised gas is about twice as turbulent as molecular gas, and neither evolves much since z~1.5.","key_machinery":"The organising identity is the Toomre stability parameter, Q_gas = σ κ / (π G Σ_gas) ≈ 1, re-expressed as Q = (σ / V)(a / f_gas), where a describes the rotation curve shape and f_gas is the gas fraction. The paper feeds empirical relations for the specific star-formation rate and depletion time into this identity to predict σ(t) at fixed stellar mass. The updated prescriptions — Speagle et al. (2014) for sSFR and Tacconi et al. (2020) for depletion time — are the specific mechanism that flattens the predicted evolution relative to the earlier W15 model, which used a steeper sSFR relation that had previously driven the dispersion up with redshift.","core_discovery":"The paper's central claim is that, once sample properties (chiefly stellar mass and which gas phase is measured) are accounted for, the velocity dispersion of disc gas does not evolve significantly between z~1.5 and z~8. This directly follows from the authors' updated Toomre stability model: with the specific star-formation-rate relation of Speagle et al. (2014) and the depletion-time relation of Tacconi et al. (2020), the predicted dispersion at fixed stellar mass flattens beyond z~1, rather than rising toward ~100 km/s as the earlier W15 model suggested. The compilation also yields a phase-dependent result: at fixed molecular gas mass, ionised gas dispersions are roughly a factor of two higher than molecular gas dispersions (Hα/[OIII] versus CO/[CI]), and the 158-µm [CII] line scatters between the two sequences. The [CII] scatter is organised by star-formation rate, which the authors interpret as [CII] tracing molecular gas in high-SFR, low-metallicity systems and warmer, more turbulent gas in lower-SFR systems.","pith_inferences":["If the ~2× ionised-to-molecular dispersion offset is physical, newborn stars in these discs should inherit the molecular gas dispersion (~15–50 km/s), not the ionised value; this would raise the floor for subsequent disc heating and disfavour the strongest 'born-hot' scenarios for high-redshift disc stars.","A direct test is within reach: at z ≈ 1–2 both Hα and CO (or [CI]) are observable in the same galaxies, and a same-galaxy measurement of the phase offset would cleanly separate a physical tracer difference from a methodological one.","The SFR-dependence of [CII] kinematics predicts that dynamical masses derived from [CII] will be systematically biased in low-SFR galaxies (where the line traces a warmer, higher-dispersion phase), a bias that could be checked against CO-based masses of lensed systems at z ~ 4–6.","If the flat σ(z) is robust, much of the apparent 'disc settling' trend since z ~ 2 may be a selection effect of tracing progressively more molecular gas at later times, rather than a genuine decrease in ISM turbulence."],"forward_implications":["The seeming contradiction between dynamically cold molecular discs at z > 4 (σ ~ 15 km/s) and turbulent ionised gas at cosmic noon disappears: both populations are consistent with a single mildly evolving Toomre-unstable disc population once tracer and mass are controlled.","Ionised and molecular gas phases sit at different effective Toomre Q values (roughly Q = 0.67 for ionised gas versus Q = 0.3 for molecular gas), implying that multi-phase discs require phase-specific stability criteria.","[CII] kinematics should not be treated as a pure tracer of either cold or warm gas; the measured dispersion depends on the galaxy's star-formation rate because the line's dominant emission phase shifts with SFR.","The flat population-average evolution of σ(z) at fixed mass means individual discs can still be settling, heating, or merging; the average carries only weak information about any one galaxy's dynamical history.","The updated Toomre-model prediction of a flat, mildly mass-dependent σ(z) provides a new benchmark for interpreting the next generation of JWST and ALMA kinematic surveys at z > 4."],"supporting_citations":[{"why":"Supplies the previous Toomre-based model whose steep dispersion evolution the paper revises.","marker":"Wisnioski et al. (2015)"},{"why":"Derives the analytic form Q = (σ/V)(a/f_gas) used for all model predictions.","marker":"Genzel et al. (2011)"},{"why":"Provides the depletion-time scaling relation adopted for f_gas(t), whose flattening drives the updated model.","marker":"Tacconi et al. (2020)"},{"why":"Supplies the specific star-formation-rate relation adopted, consistent with the chosen depletion-time prescription.","marker":"Speagle et al. (2014)"},{"why":"Contributes PHIBSS data and CO luminosity conversions used in the kinematic and gas-mass compilation.","marker":"Tacconi et al. (2013)"},{"why":"Largest source of ionised gas dispersions at cosmic noon and the earlier phase-offset comparison.","marker":"Übler et al. (2019)"},{"why":"Reports a dynamically cold molecular disc at z ~ 4.2, the observation that motivates the tracer comparison.","marker":"Rizzo et al. (2020)"},{"why":"Provides local galaxies with both CO and Hα kinematics, giving the analogous ~2× phase offset.","marker":"Girard et al. (2021)"},{"why":"Calibrates the [CII] luminosity-to-molecular-gas-mass conversion used in the compilation.","marker":"Zanella et al. (2018)"},{"why":"Supplies the 3D-Barolo forward-modelling code used for many of the mm/FIR kinematic measurements in the sample.","marker":"Di Teodoro & Fraternali (2015)"}],"fun_headline_variants":["Disc gas turbulence steady from z=1.5 to z=8 in 237 galaxies","No evolution in disc velocity dispersion from z~1.5 to z~8","Ionised gas twice as turbulent as molecular in discs","Gas dispersion in discs constant from z~1.5 to z~8"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central results rest on the assumption that velocity dispersions measured with different instruments, emission lines, and fitting codes can be combined without homogenisation, and that the adopted depletion-time scaling for molecular gas holds out to z ~ 6.","fun_headline_variants_meta":{"raw":{"variants":["Disc gas turbulence steady from z=1.5 to z=8 in 237 galaxies","No evolution in disc velocity dispersion from z~1.5 to z~8","Ionised gas twice as turbulent as molecular in discs","Gas dispersion in discs constant from z~1.5 to z~8"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001023,"raw_usage":{"total_tokens":4375,"prompt_tokens":1069,"completion_tokens":3306,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":685,"completion_tokens_details":{"reasoning_tokens":3223}},"tokens_in":685,"tokens_out":3306,"duration_ms":22277,"temperature":1.0,"reasoning_tokens":3223,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:33:06.843954+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-fit every public data cube in the compilation with one forward-modelling code and a uniform beam, sensitivity, and spectral-resolution treatment. If the homogenised values erase the ~2× offset between ionised and molecular gas at fixed M_gas, or if σ_molecular rises steeply with redshift once beam-smearing and tracer differences are removed, the claims fail. A cheaper check: the same galaxies observed in both Hα and CO at z ≈ 1–2 should reproduce the offset within individual systems.","supporting_citations":[],"review_version":1}