{"id":"c0b8cf34-8e38-4e90-8512-90c3ce203a77","arxiv_id":"2607.14359","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Co-rotating cool CGM clouds in 21 galaxies are ~0.5 dex more metal-poor within 30 degrees of the major axis than at larger azimuthal angles.","lead":"By combining existing cloud-by-cloud absorption measurements with galaxy rotation speeds, this paper finds that gas falling into galaxies is roughly half a dex poorer in heavy elements along the disk plane than at larger angles. The result suggests cool inflow and enriched recycled gas are chemically distinct routes of the baryon cycle.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Detection-floor completeness is not quantitatively modeled; N(HI)-dependent censoring at high Φ may explain the 0.5 dex metallicity offset.","rationale":"The reader's weakest_assumption identifies the same load-bearing concern: the detection floor at low N(HI) for high-Φ clouds could bias the metallicity comparison, and the paper's dismissal is not quantitative. This is the central weak point because the entire physical interpretation—metal-poor inflow along the disk plane vs enriched recycled accretion at higher azimuth—rests on the 0.52 dex mean metallicity offset. The paper's own text flags the floor but does not model it, and the counter-argument based on rotation-inconsistent clouds is not a valid control since those clouds may have different N(HI) distributions. The fact that all derived quantities come from a joint Cloudy fit means the 'multiparameter coherence' (higher N(HI), higher density, lower non-thermal broadening) could also reflect correlated model outputs rather than independent physical evidence. A quantitative completeness model or a matched-N(HI) analysis would settle whether the offset survives. Given the small sample (43 clouds from 21 galaxies) and the already conditional verdict, this reinforces the need for the requested robustness checks without moving the verdict; the paper should not be accepted as demonstrating the effect until the floor is addressed. I agree with the reader's assessment that the claim is plausible but not established.","tokens_in":13097,"tokens_out":4101,"duration_ms":42515,"concrete_test":"Run an injection-recovery simulation using the Sameer et al. (2024) ionisation model and the actual survey detection thresholds. For each rotation-consistent PIE cloud in the low-Φ sample, rescale its N(HI) downward to match the high-Φ N(HI) distribution (preserving Z, nH, and bnt), then determine which clouds would still have detectable metal lines. Measure the recovered metallicity distribution for the rescaled sample; if a substantial fraction of low-Z clouds fall below detection, the observed high-Φ mean is biased upward by the floor. As a complementary check, recompute the §3.1 Φ<30°−Φ>30° mean metallicity difference after restricting both subsamples to clouds with N(HI) above the completeness limit of the high-Φ sample; if the difference becomes consistent with zero, the azimuthal metallicity claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that rotation-consistent PIE clouds at Φ<30° are ~0.5 dex more metal-poor than those at Φ>30°—depends on the assumption that the metallicity censoring floor is azimuthally uniform after kinematic selection. The paper itself states in §3.1 that 'the lower H I column densities at Φ>30° may raise the metallicity detection floor and reduce sensitivity to the most metal-poor clouds,' but dismisses this qualitatively: 'this effect alone is unlikely to explain the azimuthal metallicity difference because low-metallicity clouds are still detected at higher azimuthal angles when the kinematic selection is removed (see Figure 1).' That rebuttal is not decisive: the rotation-inconsistent high-Φ clouds are a different population, and their N(HI) distribution is not shown; the detection threshold depends on N(HI), density, and the full ionisation model, not on kinematic class. Moreover, Z, nH, N(HI), and bnt are jointly inferred from the same absorption lines by the Sameer et al. (2024) Cloudy modelling, so a low-N(HI) regime at high Φ could simultaneously (i) raise the Z upper-limit floor and (ii) bias the surviving Z estimates upward through modelling degeneracy (e.g., lower N(HI) can be partially compensated by higher Z when matching ionic ratios). The Kaplan-Meier survival analysis used to handle upper limits is only unbiased if censoring is non-informative—independent of the true Z—which is exactly what the N(HI)-dependent floor violates. With a 1.63 dex N(HI) offset between the two azimuthal bins, the plausible censoring bias is of the same order as the reported 0.52 dex difference, so the metallicity signal may be partly or wholly an artefact of selection.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This Letter cross-matches the MGHS cloud-by-cloud ionization catalog (Sameer et al. 2024) with galaxy rotation curves (Nateghi et al. 2024a,b) for 21 galaxies. Restricting to low-ionization PIE clouds whose line-of-sight velocities are consistent with galaxy rotation, the authors split the sample at azimuthal angle Phi=30 deg and report that major-axis clouds (Phi<30 deg) have lower metallicity by -0.52+/-0.20 dex, higher N(HI) by +1.63+/-0.45 dex, higher density by +0.71+/-0.22 dex, and lower non-thermal broadening by -5.92+/-2.32 km/s, with no significant temperature difference. They interpret this as evidence for coherent metal-poor inflow along the disk plane and enriched, turbulent recycled accretion at larger azimuthal angles. The higher-ionization TDP-low phase shows no significant azimuthal metallicity dependence, and the paper argues that angular-momentum selection is essential for revealing chemodynamical structure.","tokens_in":13339,"tokens_out":3815,"duration_ms":41300,"significance":"If the headline result is robust, the paper would be an important step beyond sightline-averaged and phase-mixed analyses: it identifies an angular-momentum-selected azimuthal metallicity difference in the cool CGM, with a multi-parameter signature that connects to simulations of cold accretion and recycled fountains. The authors are careful to use a cloud-by-cloud catalog, Kaplan-Meier estimates for censored data, and bootstrap uncertainties, and they explicitly acknowledge the limited power of the censored two-sample test. The azimuthal and kinematic cuts are not constructed from metallicity, so the central cross-tabulation is not circular. However, the current evidence is not yet at the strength claimed: the headline significance is 2.6sigma, the censored two-sample test is non-significant, the Phi=30 deg cut is data-driven without sensitivity analysis, the bootstrap treats clouds as independent despite clustering in galaxies, and the detection-floor/censoring issue is dismissed qualitatively. These concerns are fixable with additional analysis, but they are load-bearing for the central claim.","major_comments":[{"comment":"The headline difference Delta<log Z>=-0.52+/-0.20 (2.6sigma) is not supported by the censored two-sample test, which gives p=0.18+/-0.27; the text itself admits the test is underpowered. Yet the Abstract and Conclusions say the paper 'demonstrates' the gradient. Please report the number of detected metallicities versus upper limits in each azimuthal bin, specify which two-sample survival test is used (log-rank? Peto? etc.), and state whether the result persists if upper limits are simply excluded. A 2.6sigma bootstrap difference with a non-significant censored test cannot carry the weight of the current conclusions without additional sensitivity checks.","section":"Sec. 3.1, Table 1"},{"comment":"The split at Phi_cut=30 deg appears to be chosen after inspecting the data, and no sensitivity analysis is shown. The paper should report Delta<log Z> as a function of Phi_cut (e.g., 15, 20, 25, 30, 35, 40, 45 deg) for the rotation-consistent PIE population, and also show how the result depends on the velocity tolerance used to define 'rotation-consistent'. If the signal only appears at 30 deg, the interpretation as a physical bimodality is much weaker.","section":"Sec. 3.1, Fig. 1"},{"comment":"The detection-floor concern is not adequately addressed. The Phi>30 deg subsample has log N(HI) lower by 1.63+/-0.45 dex, so the metallicity detection floor is plausibly higher there; Kaplan-Meier estimation assumes non-informative censoring, which is violated if censoring depends on N(HI) (and thus potentially on Z). The rebuttal that low-metallicity clouds are seen at high Phi when the kinematic selection is removed (Fig. 1) is not decisive because those clouds are a different kinematic population with a different N(HI) distribution. Please provide a quantitative completeness model: e.g., simulate the Sameer et al. detection limits, compare N(HI) distributions of the two azimuthal bins, and test whether the Z offset survives after restricting to regions where the detection floors overlap. Without this, the -0.52 dex offset could be a censoring artefact.","section":"Sec. 3.1, Table 1"},{"comment":"The 43 rotation-consistent PIE clouds are drawn from only 21 galaxies, with multiple clouds per galaxy. The bootstrap resampling treats clouds as independent samples, which underestimates uncertainties if clouds from the same galaxy are correlated (e.g., through common environment, inclination, SED, or outflow geometry). Please repeat the main bootstrap using galaxy as the resampling unit (cluster bootstrap) and report the resulting uncertainties and significance. This applies to all rows of Table 1, not only metallicity.","section":"Secs. 2-3, bootstrap procedure"},{"comment":"The quantities Z, nH, N(HI), and bnt are jointly inferred from the same absorption lines via the Sameer et al. (2024) Cloudy modeling. The paper presents these as 'independent physical parameters' (Sec. 4.2), but systematic modeling degeneracies or shared priors could introduce correlated offsets. For example, at fixed ionic ratios a lower N(HI) can be partially compensated by a higher Z in the ionization solution. Please show posterior covariances for representative clouds, or test the sensitivity of the azimuthal differences to alternative ionizing SEDs/priors, to demonstrate that the Z and N(HI) offsets are not driven by a common modeling degeneracy.","section":"Secs. 2, 4.2"}],"minor_comments":[{"comment":"The language 'demonstrate' is stronger than the statistics support. Suggest 'suggest' or 'provide evidence for' given the p=0.18+/-0.27 censored test and the small sample.","section":"Abstract/Conclusions"},{"comment":"The TDP-low sample consists of 12 rotation-consistent and 2 rotation-inconsistent clouds. The claim of a 'similar metallicity distribution' to PIE clouds is not backed by a quantitative test. State the sample sizes and give a test or explicitly refrain from interpretation.","section":"Sec. 3.2"},{"comment":"Individual points are plotted without uncertainties. Add representative error bars for Z or state in the caption why they are omitted. This is important because many metallicities are upper limits and the figure currently suggests a clean bimodality.","section":"Fig. 1"},{"comment":"The sentence 'We also find that the differences between rotation-consistent and rotation-inconsistent PIE clouds for all of these properties are consistent with zero' is not accompanied by numbers. A small table or a sentence with the key values would improve transparency.","section":"Sec. 3.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest and the cross-tabulation is not circular, but the central claim currently rests on a 2.6sigma bootstrap difference and a non-significant censored test. The detection-floor and galaxy-clustering issues are the main risks; both are tractable with additional analysis. I would not reject, but I would ask for the sensitivity analyses described in the major comments before publication. The same-team lineage of the data products is not itself a concern; the methods seem appropriate and prior work is cited."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper reanalyzes the MGHS cloud catalogue with rotation kinematics and claims that rotation-consistent PIE clouds within 30° of the projected major axis are ~0.5 dex more metal-poor than those at larger azimuth, with higher N(HI), higher density, and lower non-thermal broadening. If true, it resolves the MEGAFLOW/MGHS tension by showing that kinematic selection is needed to see the azimuthal structure. The cross-tabulation is genuinely new—Sameer et al. (2024) saw no azimuthal dependence with the same clouds, and Nateghi et al. (2024b) only looked at co-rotation fractions per galaxy. The selection is velocity-based and geometric, not constructed from metallicity, so circularity is low.\n\nThe paper is transparent about its main weakness: the censored two-sample test gives p=0.18±0.27, and the 30° cut appears data-driven with no sensitivity analysis. The bigger worry is the completeness floor. High-Φ clouds have 1.63 dex lower N(HI), and the paper admits this may raise the metallicity detection floor, but dismisses it with a qualitative argument about low-metallicity clouds appearing at high Φ when the kinematic selection is removed. That rebuttal is not decisive: the rotation-inconsistent clouds are a different population, and the N(HI) distribution is not shown. The Kaplan-Meier estimates are only unbiased if censoring is non-informative, which is exactly what is violated if the floor depends on N(HI). A modest bias in the surviving Z estimates at high Φ could easily produce a 0.5 dex offset.\n\nThat said, the multi-parameter coherence (N(HI), density, bnt all shifting in the expected directions) is suggestive, and the absence of a temperature offset is a nice touch. The paper also correctly avoids overclaiming for the TDP-low phase. The main statistical issue is the untreated clustering: 43 clouds from 21 galaxies are bootstrapped as independent, which likely overstates significance. The 'demonstrate' language in the conclusions exceeds the evidence.\n\nWho benefits? People working on CGM surveys and baryon cycle observations. It's a solid, interesting re-analysis with a plausible but unproven central claim. It deserves a serious referee—the claim is important enough that we need someone to push on the completeness model and the binning choice. I'd accept it for review, but expect major revision before the gradient is established.","headline":"A plausible but not yet proven azimuthal metallicity gradient in co-rotating cool CGM gas — needs a quantitative completeness model before the claim holds.","tokens_in":14085,"tokens_out":3453,"would_cite":false,"duration_ms":31645,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper argues that the cool gas around galaxies shows a clear azimuthal metallicity split, but only among clouds whose velocities match the rotation of their host galaxy: co-rotating low-ionisation clouds near the projected disk plane a","keywords":["circumgalactic medium","quasar absorption lines","metallicity","galaxy rotation kinematics","cold inflow","recycled accretion","azimuthal angle","photoionisation modelling"],"falsifier":"Re-analyse the same 21 galaxies restricting both azimuthal bins to clouds above a common N(HI) detection threshold and recompute the mean metallicity offset; if Δ⟨log Z⟩ drops to ~0, the claim is falsified. Alternatively, inject synthetic metal-poor clouds at Φ>30° with the observed lower N(HI) into the modelling pipeline and show whether they would be detected; if they would be missed, the completeness floor alone can produce the offset.","tokens_in":12842,"feed_emoji":"🌌","tokens_out":4981,"duration_ms":48379,"temperature":0.7,"pith_summary":"The paper argues that the long-sought azimuthal metallicity structure of the cool circumgalactic gas becomes visible only when clouds are separated by whether their velocities match the host galaxy's rotation. Using 21 galaxy–quasar pairs with both cloud-by-cloud ionisation measurements and rotation curves, the authors find that low-ionisation clouds co-rotating with the disk have mean metallicity about 0.5 dex lower within 30 degrees of the projected major axis than at larger azimuths, together with higher neutral-hydrogen column density, higher gas density, and lower non-thermal line broadening. They interpret the major-axis population as dynamically cold, metal-poor inflowing gas, and the more enriched, turbulent co-rotating gas at larger azimuths as recycled accretion that preserves angular momentum. The result matters because previous sightline-averaged and cloud-by-cloud studies found no azimuthal metallicity gradient; this paper claims that gradient was hidden by mixing kinematically decoupled gas with the co-rotating component. If correct, metallicity and angular momentum jointly trace the baryon cycle, and kinematic selection becomes a necessary step in CGM abundance studies.","feed_headline":"Co-rotating halo gas reveals a 0.5-dex metal split","feed_subtitle":"Clouds that rotate with their galaxy are metal-poor near the disk axis and enriched away from it, exposing the baryon cycle.","key_machinery":"The load-bearing tool is kinematic selection applied to a cloud-by-cloud ionisation catalogue: each absorption cloud is first assigned an ionisation phase (cool photo-ionised 'PIE' clouds versus warmer time-dependent 'TDP-low' clouds) and a set of physical properties (metallicity, density, temperature, Doppler parameters) from Bayesian photoionisation modelling; then each cloud is classified as rotation-consistent or not by comparing its line-of-sight velocity with the host galaxy's rotation curve. The azimuthal angle Φ relative to the projected major axis, split at 30°, is applied only to the rotation-consistent PIE population. The paper's central mechanism is that angular-momentum selectio","core_discovery":"Among clouds in photo-ionisation equilibrium whose line-of-sight velocities are consistent with the direction of host-galaxy rotation, mean metallicity is log(Z/Zsun) = -0.77 ± 0.16 at Φ<30° versus -0.25 ± 0.13 at Φ>30°, a difference of -0.52 ± 0.20 dex at 2.6σ. The same co-rotating major-axis clouds have higher neutral-hydrogen column density (+1.63 ± 0.45 dex), higher hydrogen density (+0.71 ± 0.22 dex), and lower non-thermal Doppler broadening (-5.92 ± 2.32 km/s), with no significant temperature difference. No such azimuthal metallicity trend appears in rotation-inconsistent clouds or in the higher-ionisation warm phase. The authors interpret the multi-parameter separation as evidence tha","pith_inferences":["A straightforward extension is to re-analyse existing MgII and OVI absorber surveys with the same kinematic split; if the pattern repeats, the inflow/recycled-accretion dichotomy is generic rather than peculiar to this sample.","The 2.6σ significance and small sample (43 co-rotating PIE clouds from 21 galaxies) imply the true amplitude could be larger or smaller; a larger sample with dense rotation curves would sharpen the Φ-cut and test whether 30° is the physical boundary.","Because the high-Φ subsample has systematically lower N(HI), a quantitative completeness model could show whether metal-poor clouds at large azimuth are simply undetected; if so, the inferred metallicity offset would shrink.","The absence of a temperature difference while non-thermal broadening rises suggests turbulence/mixing, not heating; this could be tested in simulations by predicting bnt and metallicity jointly for recycled fountain gas."],"forward_implications":["If true, earlier null results for azimuthal metallicity gradients (both sightline-averaged and cloud-by-cloud) were diluted by kinematically decoupled gas; future surveys must combine metallicity with rotation kinematics.","The cool CGM near the disk plane is dominated by dense, metal-poor, dynamically cold gas, consistent with filamentary accretion retaining angular momentum.","Co-rotating enriched gas at larger azimuths is more likely recycled accretion than ballistic outflow, because it keeps rotational coherence.","The warm/high-ionisation phase is kinematically decoupled and chemically independent, so single-phase abundance measurements cannot represent the CGM baryon cycle.","Metallicity and angular momentum must be treated as joint observables when comparing observations to simulations of galaxy haloes."],"fun_headline_variants":["Co-rotating halo gas splits metals: axis poor, flanks rich","Halo rotation imprints 0.5 dex metallicity difference","Metal-poor inflow vs enriched recycled gas in galaxy halos","CGM co-rotation exposes baryon cycle in metal content","Angular momentum reveals cold inflow and recycled accretion"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing assumption is that the lower neutral-hydrogen column densities at Φ>30° do not bias the inferred metallicities upward: if metal-poor clouds at large azimuthal angles are preferentially missed by the detection floor (or if the joint photoionisation modelling degenerately raises metallicity at low N(HI)), the −0.52 dex offset could be partly or wholly an artifact, and the inflow/recycled-accretion interpretation would lose its foundation.","fun_headline_variants_meta":{"raw":{"variants":["Co-rotating halo gas splits metals: axis poor, flanks rich","Halo rotation imprints 0.5 dex metallicity difference","Metal-poor inflow vs enriched recycled gas in galaxy halos","CGM co-rotation exposes baryon cycle in metal content","Angular momentum reveals cold inflow and recycled accretion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000333,"raw_usage":{"total_tokens":1733,"prompt_tokens":836,"completion_tokens":897,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":810}},"tokens_in":580,"tokens_out":897,"duration_ms":8301,"temperature":1.0,"reasoning_tokens":810,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T02:19:12.414668+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-analyse the same 21 galaxies restricting both azimuthal bins to clouds above a common N(HI) detection threshold and recompute the mean metallicity offset; if Δ⟨log Z⟩ drops to ~0, the claim is falsified. Alternatively, inject synthetic metal-poor clouds at Φ>30° with the observed lower N(HI) into the modelling pipeline and show whether they would be detected; if they would be missed, the completeness floor alone can produce the offset.","supporting_citations":[],"review_version":1}