{"id":"5734654a-6b80-4c47-b202-4f5e6240c144","arxiv_id":"2507.14983","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A grid of non-equilibrium chemical models fit to JWST spectra of K2-18 b favors high metallicity (266x solar) and C/O > 2.1, with methane robustly detected.","lead":"Scientists modeled the atmosphere of the temperate exoplanet K2-18 b with a non-equilibrium chemistry code and JWST spectra, finding high metallicity, a high carbon-to-oxygen ratio, and a secure methane detection. The result matters because it shows how chemical models, not just free retrievals, are needed to interpret exoplanet spectra, while also showing current data cannot prove non-equilibrium chemistry.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The fixed P-T profile is the most load-bearing concern: the C/O lower limit and metallicity range depend on a thermal structure fitted to low metallicity, solar C/O, and may not hold at the retrieved high C/O and high metallicity.","rationale":"The reader's verdict identifies the fixed P-T profile as the weakest assumption, and I concur: it is the single assumption whose failure would most directly change the headline numbers. The paper is transparent about this limitation, but transparency does not remove the load-bearing role: the claimed 2σ C/O lower limit and the metallicity range are quoted as quantitative results in the abstract and conclusion, yet they are conditional on an unquantified and plausibly large systematic. I do not see an internal inconsistency in the grid method itself, and the CH4 detection is robust across equilibrium and non-equilibrium models, so the paper should not be rejected. However, the current CONDITIONAL verdict is appropriate: the authors should either quantify the P-T sensitivity (e.g., by recomputing profiles for a few grid points) or soften the abstract's quantitative claims. I note also that the paper's own evidence—equilibrium fit with ∆χ2 = -2, H2O-condensation bias, aerosol slope mismatch at <2 µm—reinforces that the fixed P-T assumption is not the only model uncertainty, but it is the most central one. A concrete, bounded test (self-consistent P-T recomputation for a subset of grid points) would settle whether the concern lands or whether the quoted constraints are robust.","tokens_in":16457,"tokens_out":1645,"duration_ms":15625,"concrete_test":"Re-run the grid (or at least the best-fit and 2σ boundary models) with self-consistent P-T profiles: for each (metallicity, C/O, Kzz) grid point, recompute the radiative-convective equilibrium profile (e.g., with the Blain et al. 2021 code or ATMO/PetitCODE) before running FRECKLL. If the resulting 2σ C/O lower limit and metallicity interval shift by more than ~0.3 dex or the C/O lower limit drops below ~1.5, the fixed-profile assumption is the dominant systematic and the headline constraints should be revised.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central claims—high metallicity (266 +291/-104 at 2σ) and C/O ≥ 2.1 at 2σ—rest on a grid of non-equilibrium models computed with a single fixed pressure-temperature profile taken from Blain et al. (2021), itself a best fit for metallicity=175, C/O=0.13, Kzz=1e6 (Section 2.3). The authors acknowledge 'some uncertainty' but do not quantify it. This is load-bearing because the thermal structure directly sets chemical equilibrium abundances, quench depths, and spectral band strengths. At the retrieved extremes (metallicity up to ~1000x solar, C/O up to 100), the background opacity, mean molecular weight, and radiative properties differ substantially from the fiducial 175x solar, C/O=0.13 case; a self-consistent P-T profile would shift the CH4 abundance and the CO/CO2/H2O column, which are exactly the species driving the C/O constraint. The paper's own equilibrium retrieval (Section 3.2) also finds a different best fit (metallicity ~71, C/O ~6.8) but uses the same fixed profile, so it does not test this assumption. Because the C/O lower limit is already fragile—removing H2O opacity drops the 2σ limit from 2.1 to 1.4 (§3.5)—a modest P-T shift could plausibly erase the C/O>2.1 result.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper compares a grid of 1D non-equilibrium chemical models computed with FRECKLL to the JWST transmission spectrum of K2-18 b, exploring metallicity (0.1–1000× solar), C/O (0.1–100), and Kzz (10^5–10^10 cm^2/s) under a fixed P–T profile taken from Blain et al. (2021). Using a chi-square grid search and polynomial fits to projected Δχ², it reports a best-fit metallicity of 266 (+291/−104) at 2σ, a C/O lower limit of 2.1 at 2σ, a robust CH4 detection (log10[CH4] ≈ −0.3 at 1 mbar), and no constraint on Kzz. A complementary equilibrium retrieval with TauREx yields a similar overall fit (metallicity ≈ 71, C/O ≈ 6.8) with a slightly lower χ². The paper concludes that the data favor a metal-rich, carbon-rich atmosphere but that non-equilibrium chemistry cannot be conclusively proven from these data.","tokens_in":16754,"tokens_out":6059,"duration_ms":60210,"significance":"If the metallicity and C/O constraints survive a self-consistent treatment, they would be important anchors for sub-Neptune formation and atmospheric evolution, and the demonstration that constant-abundance retrievals are insufficient is valuable. The study's strengths are the unusually wide C/O exploration (0.1–100), the use of a full kinetic network with photochemistry, and the candid acknowledgment of model limitations (fixed P–T, no condensation). The paper does not oversell the detection of minor species and explicitly states that only CH4 is robustly detected. However, the headline parameter constraints are conditional on modeling assumptions that are not fully tested, and the statistical reporting needs more care.","major_comments":[{"comment":"The grid spans metallicity 0.1–1000 and C/O 0.1–100, but every model uses the single fixed P–T profile from Blain et al. (2021), which is itself the best fit for metallicity = 175, C/O = 0.13, Kzz = 10^6. The paper states that composition-induced temperature changes are expected to be small, but provides no quantitative test; the retrieved solutions at metallicity ≈ 266 and C/O ≳ 2 are far from the fiducial composition, and the equilibrium retrieval of Section 3.2 uses the same fixed profile, so it does not validate the assumption. Since CH4, CO, and H2O abundances and the C/O constraint depend on quench temperatures, the headline 2σ C/O lower limit of 2.1 is not robust until the P–T sensitivity is quantified, for example with a self-consistent radiative-equilibrium grid or at least a perturbed-profile test.","section":"§2.3, Table 4"},{"comment":"The confidence intervals in Table 4 are derived by collapsing the 3D grid over Kzz via minimum χ², projecting the 2D Δχ² onto 1D, and fitting a polynomial to the projected values. This is neither a Bayesian marginalization nor a profile likelihood with stated coverage, and the reported intervals (e.g., metallicity 266 +291/−104) should not be presented as standard 1σ/2σ/3σ confidence intervals. The >4σ significance relative to a flat line also does not account for the number of grid points (30×58×6) or the implicit model selection. The paper should either re-derive the constraints with a statistically principled method (e.g., nested sampling over an interpolated forward model) or explicitly label the numbers as approximate grid-based ranges without coverage guarantees.","section":"§2.4"},{"comment":"The central claim 'C/O ≥ 2.1 at 2σ' is highly sensitive to the treatment of H2O: when H2O opacity is excluded (a crude proxy for condensation, which FRECKLL does not model), the 2σ limit drops to 1.4 and the 3σ limit to 0.57. Because the default model likely overestimates gas-phase H2O in the cool upper atmosphere, the high-C/O conclusion is conditional on an unmodeled process. The paper should either implement a condensation/rainout scheme or present the C/O lower limit as explicitly model-dependent rather than as a robust atmospheric constraint.","section":"§3.5"}],"minor_comments":[{"comment":"The CH4 abundance is reported as log10[CH4] = −0.3 +0.1/−1.7 without specifying the confidence level; Table 4 shows these are 3σ uncertainties, so the abstract should state this to avoid implying a 1σ range.","section":"Abstract"},{"comment":"Because the equilibrium retrieval has a slightly lower χ² than the best non-equilibrium model (Δχ² = −2.0), the statement 'Non-equilibrium models exceed 4σ confidence over a flat line' should clarify that this establishes the presence of spectral features, not disequilibrium chemistry; the title and abstract should be rephrased to avoid overclaiming.","section":"§3.2 / Abstract"},{"comment":"The Kzz entry 'All within 1σ' is ambiguous; please specify that all grid values are within 1σ of the best fit, i.e., Kzz is unconstrained by the data.","section":"Table 4"},{"comment":"The phrase 'consistent with the findings of Blain et al. (2021), where the impact was shown to be small' lacks a quantitative reference; please cite the specific figure or section of Blain et al. that supports this claim.","section":"§2.3"},{"comment":"Typographical artifacts such as 'su fficient', 'di fficult', 'W ASP-39 b', and inconsistent spacing in 'C /O' should be corrected in a final proofread.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript appears to have been accepted by A&A already (dates on the first page), so this report is relevant if the authors prepare a revision or a corrigendum. The main risk is that the abstract's headline numbers (metallicity 266, C/O ≥ 2.1) could be over-interpreted despite the paper's own caveats; the fixed P–T assumption and the statistical method are the key points to address. No concerns about novelty or citation practices."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nYou should know this paper before the next modeling meeting. It runs the FRECKLL non-equilibrium chemistry model over a much wider joint grid (metallicity 0.1–1000x solar, C/O 0.1–100, Kzz 1e5–1e10) than previous disequilibrium analyses of K2-18 b, and compares each model to the JWST spectrum. That grid is the real contribution. It lets them see a metallicity bimodality between low and high C/O regimes and puts a lower limit on C/O around 2.1 at 2σ, broadly consistent with Schmidt et al. (2025) but from a more complete parameter search.\n\nWhat they do well: the paper is transparent. It states clearly that the P-T profile is fixed from Blain et al. (2021), that condensation is not included, that the Kzz dimension is unconstrained, and that the equilibrium retrieval fits slightly better than the non-equilibrium grid (Δχ² = −2). That last point is the one to keep in mind: the title says \"non-equilibrium chemistry\" but the data do not favor it, and the authors admit it. The CH4 detection is robust, and the CO2 non-detection is consistent with Schmidt et al.\n\nThe soft spots are real but not fatal. The fixed P-T profile is load-bearing for the high-metallicity/high-C/O result. At 1000x solar and C/O 100 the thermal structure will not be the same as the 175x solar, C/O 0.13 profile they inherited. They note \"some uncertainty\" but don't quantify it; that is the main thing a revision should address, either with a self-consistent P-T iteration or a sensitivity test. The significance claims (4σ over flat line) ignore model complexity and the fact that the grid is not a proper retrieval; the Δχ² thresholds are applied after a polynomial fit to a projection, not a marginalization. That makes the 2σ numbers optimistic. The C/O lower limit also drops from 2.1 to 1.4 when H2O opacity is removed, so the constraint is softer than the abstract implies.\n\nStill, the paper is a useful exploration. It is worth citing for the grid range and for the behavior of minor species, and it deserves a serious referee. I would send it out with a request to clarify the CH4 abundance convention, add a sensitivity test on the P-T profile, and tone down the non-equilibrium framing to match the actual evidence.","headline":"Wide non-equilibrium grid for K2-18 b, honest about its limits, but the non-equilibrium versus equilibrium claim is not supported by its own fits.","tokens_in":17348,"tokens_out":1956,"would_cite":true,"duration_ms":20081,"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":"A non-equilibrium chemistry grid fit to JWST data makes K2-18 b metal-rich and carbon-rich, with methane as the only secure absorber.","keywords":["K2-18 b","sub-Neptune atmospheres","non-equilibrium chemistry","transmission spectroscopy","JWST","metallicity","C/O ratio","vertical mixing"],"falsifier":"A high-signal JWST NIRSpec G395H observation resolving the 4–5 $\\mu$m CO and CO2 bands, analyzed with a pressure-temperature profile recalculated for each grid point, would settle it: if secure CO2 near 1 percent and H2O above roughly $10^{-5}$ at 1 mbar are required to fit, the C/O ratio sits near solar and the >2.1 lower limit breaks.","tokens_in":16248,"feed_emoji":"🪐","tokens_out":13980,"duration_ms":123817,"temperature":0.7,"pith_summary":"K2-18 b, a temperate sub-Neptune in an M-dwarf habitable zone, is argued on the basis of a non-equilibrium chemistry grid to have a metal-rich, carbon-rich atmosphere. The best-fit models favor a metallicity of $266^{+291}_{-104}$ times solar and a C/O ratio above 2.1 at 2$\\sigma$, with methane securely detected at $\\log_{10}[\\mathrm{CH_4}] \\approx -0.3$ at 1 mbar. Carbon dioxide, water, ammonia, and other species remain unconfirmed, and vertical mixing ($K_{zz}$) is unconstrained. A constant-abundance retrieval cannot capture this chemistry, but the current data also cannot prove disequilibrium, since the equilibrium fit is statistically indistinguishable. The result matters because it changes what K2-18 b is made of and makes any simple habitable-ocean interpretation premature.","feed_headline":"K2-18 b favors a metal-rich, carbon-rich atmosphere","feed_subtitle":"Best non-equilibrium fit to JWST data secures methane and leaves CO2 uncertain.","key_machinery":"The machinery is FRECKLL, a one-dimensional steady-state non-equilibrium chemical model that solves production-loss-transport continuity equations for each species over a network of roughly 2000 reactions among more than 100 neutral species containing H, C, N, and O, with vertical transport set by an eddy diffusion coefficient $K_{zz}$ and photolysis driven by a stellar UV spectrum. TauREx 3's forward model converts the resulting abundance profiles into transmission spectra that are compared to the JWST data by $\\chi^2$. The load-bearing step is the three-dimensional grid in metallicity, C/O, and $K_{zz}$; the grid is collapsed along $K_{zz}$ by taking the lowest $\\chi^2$, and 1D polynomial fits to the projected $\\Delta\\chi^2$ produce the reported parameter values and limits.","core_discovery":"On the paper's own terms, the discovery is that exploring metallicity (0.1–1000 times solar), C/O (0.1–100), and $K_{zz}$ ($10^5$–$10^{10}$ cm$^2$/s) together collapses K2-18 b's chemical state to a high-metallicity, high-C/O corner of parameter space. The grid best fit sits at metallicity 280.7, C/O 90.0, and $K_{zz}=10^6$ cm$^2$/s; polynomial fits to the projected $\\Delta\\chi^2$ give a 2$\\sigma$ metallicity interval of $266^{+291}_{-104}$ and a 2$\\sigma$ lower limit on C/O of 2.1. CH4 is the only secure absorber, with $\\log_{10}[\\mathrm{CH_4}] = -0.3^{+0.1}_{-1.7}$ at 1 mbar, while CO2 can fall below $10^{-6}$, consistent with the recent reanalysis doubting its detection. The paper emphasizes that minor species such as CO, H2O, and NH3 may still be present at observable-to-ppm levels but are masked by CH4 and noise, and that a high C/O ratio favors aerosol formation. It also finds that the equilibrium retrieval, with metallicity $71^{+58}_{-20}$ and C/O $6.78^{+5.8}_{-2.0}$, fits the data marginally better ($\\Delta\\chi^2=-2$), so the case for disequilibrium rests on physical expectation rather than on statistical preference.","pith_inferences":["Editorial inference: The fixed pressure-temperature profile is the unquantified hinge; a self-consistent P-T grid could move the C/O limit by more than the quoted 1.4 shift from excluding H2O opacity, so 2.1 should be read as model-dependent.","Editorial inference: The bimodal metallicity behavior hints that K2-18 b's data currently permit two chemical families—metal-rich with near-solar C/O, and moderately metal-rich with very high C/O—so the next observation should target CO, not just CH4, to separate them.","Editorial inference: If the paper's picture holds, habitability arguments based on a liquid-water ocean at the surface must be reconsidered, since a carbon-rich, H2O-poor observable atmosphere is what the chemistry predicts.","Editorial inference: The result that $K_{zz}$ is unconstrained despite large non-equilibrium abundance changes suggests that 'disequilibrium' detections in similar planets will need abundance pairs with strong quenching sensitivity, not just the presence of CH4."],"forward_implications":["If the high-metallicity, high-C/O result is right, K2-18 b's missing water is expected: oxygen is tied up in CO and CO2, so the absence of a secure H2O detection is not evidence against water deeper down.","Methane becomes the only safe spectroscopic anchor; CO2, H2O, NH3, and CO claims from constant-abundance retrievals will need rechecking against chemistry-coupled models.","A C/O ratio above 2.1 favors carbon-rich aerosol and haze formation, which could explain the steep short-wavelength slope that the current models do not capture.","Because $K_{zz}$ is unconstrained, vertical mixing cannot be measured from CH4; useful constraints will come from species whose abundances change strongly with quenching depth.","Higher-signal 4–5 $\\mu$m spectroscopy is the concrete next step: CO and CO2 bands there are the leverage points for turning the C/O lower limit into a measurement."],"supporting_citations":[{"why":"Supplies the JWST NIRISS SOSS and NIRSpec G395H transmission spectrum of K2-18 b that all model fits are compared against.","marker":"Madhusudhan et al. (2023)"},{"why":"Presents FRECKLL, the non-equilibrium chemical model used to compute abundance profiles across the grid.","marker":"Al-Refaie et al. (2024)"},{"why":"Presents TauREx 3, the forward model and Bayesian retrieval framework used to generate spectra and run the equilibrium retrieval.","marker":"Al-Refaie et al. (2021)"},{"why":"Provides the fixed pressure-temperature profile assumed for every grid point and earlier disequilibrium chemistry predictions of CH4.","marker":"Blain et al. (2021)"},{"why":"Baseline disequilibrium model conclusion that CH4 is high and CO2 is low, which this paper's best fit reinforces.","marker":"Wogan et al. (2024)"},{"why":"The reanalysis casting doubt on the CO2 detection, invoked to support the paper's uncertain CO2 abundance.","marker":"Schmidt et al. (2025)"},{"why":"Supplies the C2-limited chemical reaction network used inside FRECKLL, carrying the kinetic scheme.","marker":"Venot et al. (2020)"},{"why":"Predicts roughly 100-400 times solar metallicity for planets of K2-18 b's radius, used to anchor the high-metallicity interpretation.","marker":"Fortney et al. (2013)"}],"fun_headline_variants":["K2-18 b: metal-rich, carbon-rich, methane secured","High metallicity and C/O on K2-18 b, methane confirmed","Non-equilibrium models tie K2-18 b to metal-rich mix","K2-18 b: methane definite, CO2 murky, atmosphere metal-rich","K2-18 b favors a metal-heavy, carbon-heavy composition"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes that one fixed temperature-pressure profile, originally fitted to a different composition, stays valid over the entire model grid; if the planet's thermal structure changes with metallicity or C/O, the retrieved abundances and the C/O lower limit could shift enough to break the central claim.","fun_headline_variants_meta":{"raw":{"variants":["K2-18 b: metal-rich, carbon-rich, methane secured","High metallicity and C/O on K2-18 b, methane confirmed","Non-equilibrium models tie K2-18 b to metal-rich mix","K2-18 b: methane definite, CO2 murky, atmosphere metal-rich","K2-18 b favors a metal-heavy, carbon-heavy composition"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001169,"raw_usage":{"total_tokens":5000,"prompt_tokens":1273,"completion_tokens":3727,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":889,"completion_tokens_details":{"reasoning_tokens":3628}},"tokens_in":889,"tokens_out":3727,"duration_ms":33303,"temperature":1.0,"reasoning_tokens":3628,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:44:07.389103+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A high-signal JWST NIRSpec G395H observation resolving the 4–5 $\\mu$m CO and CO2 bands, analyzed with a pressure-temperature profile recalculated for each grid point, would settle it: if secure CO2 near 1 percent and H2O above roughly $10^{-5}$ at 1 mbar are required to fit, the C/O ratio sits near solar and the >2.1 lower limit breaks.","supporting_citations":[{"cited_title":"F., Venot, O., Changeat, Q., & Edwards, B","cited_arxiv_id":null,"evidence_quote":"Presents FRECKLL, the non-equilibrium chemical model used to compute abundance profiles across the grid."},{"cited_title":"F., Changeat, Q., Waldmann, I","cited_arxiv_id":null,"evidence_quote":"Presents TauREx 3, the forward model and Bayesian retrieval framework used to generate spectra and run the equilibrium retrieval."},{"cited_title":"2021, A&A, 646, A15 Bézard, B., Charnay, B., & Blain, D","cited_arxiv_id":null,"evidence_quote":"Provides the fixed pressure-temperature profile assumed for every grid point and earlier disequilibrium chemistry predictions of CH4."},{"cited_title":"F., Batalha, N","cited_arxiv_id":null,"evidence_quote":"Baseline disequilibrium model conclusion that CH4 is high and CO2 is low, which this paper's best fit reinforces."},{"cited_title":"2020, A&A, 634, A78","cited_arxiv_id":null,"evidence_quote":"Supplies the C2-limited chemical reaction network used inside FRECKLL, carrying the kinetic scheme."},{"cited_title":"J., Mordasini, C., Nettelmann, N., et al","cited_arxiv_id":null,"evidence_quote":"Predicts roughly 100-400 times solar metallicity for planets of K2-18 b's radius, used to anchor the high-metallicity interpretation."}],"review_version":1}