{"id":"456c30fd-c76f-41dd-ab6d-f763174906c8","arxiv_id":"2506.12806","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Applying the TauREx-FRECKLL disequilibrium chemistry retrieval to 10 hot Jupiters changes retrieved metallicities and C/O ratios and demonstrates strong model dependence in HST results.","lead":"This paper re-analyzes Hubble Space Telescope observations of 10 hot-Jupiter atmospheres with a retrieval model that includes full disequilibrium chemistry, and finds that retrieved metallicities and carbon-to-oxygen ratios shift substantially compared to simpler chemistry assumptions. It shows that conclusions about exoplanet formation drawn from HST data depend heavily on the chemical model, retrieval priors, and data reduction choices.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed disequilibrium-driven parameter shifts are not statistically preferred: for WASP-74 b, the supersolar versus subsolar change has Δln E ≈ 0.5 and a nearly unconstrained log(Z) posterior, so the headline result may reflect prior volume rather than atmospheric composition.","rationale":"I read the paper as making a two-part central claim: (1) including disequilibrium chemistry materially changes retrieved metallicities and C/O ratios, impacting planet formation inferences; and (2) conclusions drawn from HST observations are strongly dependent on retrieval parameters, priors, and the assumed model. The second claim is well supported by the WASP-43 b experiments, where different priors and two independent data reductions yield divergent metallicities, thermal structures, and chemical profiles (Section 3, Figures 5-8). The first claim, however, is much less secure. The paper compares FRECKLL results against the free-chemistry retrievals of Changeat et al. (2022), not against equilibrium retrievals as the abstract states. The most dramatic example, WASP-74 b, shifts from subsolar to supersolar metallicity with a Bayesian evidence difference of only about 0.5, far below the paper's own Δln E > 5 preference threshold. Furthermore, the reported FRECKLL-only posterior for WASP-74 b is log(Z) = 1.74 (+0.98/-3.29), which is nearly flat across the allowed prior range; the lower uncertainty alone spans the entire subsolar regime. Similar patterns appear for TrES-3 b, WASP-19 b, and WASP-43 b, where the evidence differences between FRECKLL and free chemistry are small or even favor the free model. This means the 'significant alteration' is a statement about model sensitivity, not a statistically preferred description of these atmospheres. If the posterior is essentially unconstrained, then the central value is a prior-volume artifact and cannot support the claim that disequilibrium chemistry changes planet formation insights. This concern is more load-bearing than the reader's identified assumption about the Venot et al. (2020a) network accuracy, because even a perfect kinetic network would not rescue the conclusion if the data do not prefer the disequilibrium model and the retrieved parameters are not constrained. The paper does provide reproducible retrieval outputs and code, which is creditworthy, but the headline claim should be rephrased to emphasize model dependence and lack of statistical preference rather than a physical shift in atmospheric parameters. Since the reader's verdict is already CONDITIONAL and this concern is addressable through additional evidence comparisons and transparent posterior reporting, I recommend keeping the verdict unchanged.","tokens_in":32610,"tokens_out":4120,"duration_ms":48415,"concrete_test":"Re-run the WASP-74 b eclipse retrieval under an equilibrium-chemistry model (same temperature parametrization, opacity set, and priors as the FRECKLL run) and under FRECKLL, then compare the Bayesian evidences and the full log(Z) posterior distributions. If Δln E(FRECKLL - equilibrium) is less than 2 and the FRECKLL log(Z) posterior has substantial probability across the whole prior range (e.g., P(log Z < 0) > 0.3), the claimed supersolar shift is not a statistically supported finding, and the abstract should be softened to state that results are model-dependent and unconstrained by current HST data.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim that disequilibrium chemistry 'significantly altered' retrieved metallicities (Abstract; Section 3) rests on comparing FRECKLL results to the free-chemistry runs of Changeat et al. (2022), not to equilibrium models as the abstract states. The flagship example, WASP-74 b, moves from log(Z) = -2.1 (free) to +1.74 (FRECKLL) with Bayesian evidences of 202.8 and 203.29 respectively (Section 3; Table C.2). This Δln E ≈ 0.5 is far below the paper's own threshold of Δln E > 5 for preferring one model, so the data do not favor the disequilibrium interpretation. Moreover the FRECKLL-only posterior reported for WASP-74 b is log(Z) = 1.74 (+0.98/-3.29), with a lower error bar spanning the entire subsolar range; the posterior is essentially unconstrained across the prior. The same pattern holds for other planets whose metallicities are described as 'moved' (TrES-3 b, WASP-19 b, WASP-43 b), where the evidence differences are likewise small or negative. Thus the headline 'significant alteration' is a statement about model sensitivity, not a measured property of these atmospheres. What is robustly demonstrated is the weaker, cautionary claim that HST-based conclusions depend on priors and model choice, as shown by the WASP-43 b prior-restriction tests (Figures 5-8); that claim does not require FRECKLL to be physically accurate. The load-bearing premise that disequilibrium chemistry materially changes the inferred parameters, impacting planet formation insights, is therefore not established by the evidence presented.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper re-analyzes HST WFC3 eclipse and transit spectra of ten hot-Jupiter atmospheres using the TauREx 3.1 retrieval framework coupled with the FRECKLL disequilibrium chemistry model. The authors run a systematic set of retrievals (FRECKLL-only, FRECKLL+TiO, FRECKLL+VO, FRECKLL+TiO+VO), including tests with restricted metallicity priors for four bimodal cases, and compare the results with previous free-chemistry retrievals. They report that the disequilibrium chemistry assumption changes retrieved metallicities and C/O ratios substantially for several planets, that TiO/VO additions improve the fit for only two planets, that eclipse and transit temperature profiles are partly consistent at depth, and that retrieved parameters are highly sensitive to priors and data reduction. The paper concludes that HST-based conclusions depend on the retrieval model and priors, and that disequilibrium chemistry should be included in future JWST analyses.","tokens_in":33041,"tokens_out":5829,"duration_ms":62082,"significance":"If the robust parts of the analysis hold, the paper is a useful first demonstration of full kinetic disequilibrium retrievals on a sample of real HST spectra, with transparent reporting of broad posteriors and a public release of retrieval outputs. The strongest and most defensible result is the demonstration that retrieved parameters—especially metallicity—depend strongly on the chemistry model and priors, as shown by the WASP-43 b prior-restriction tests (Figures 5-8). However, the headline claim that disequilibrium chemistry 'significantly altered' retrieved metallicities and C/O ratios is not statistically supported by the evidence differences, and the abstract's comparison against 'equilibrium models' is not what the paper actually does. The value of the study is therefore primarily cautionary rather than a new measurement of hot-Jupiter compositions.","major_comments":[{"comment":"The abstract states that the disequilibrium approach 'significantly altered retrieved metallicity and C/O ratios compared to equilibrium models', but the comparisons in Section 3 are made against the free-chemistry retrievals of Changeat et al. (2022), not equilibrium models. More importantly, the evidence differences do not support 'significantly altered': for WASP-74 b, the flagship example, the Bayesian evidence changes from 202.8 (free) to 203.29 (FRECKLL-only), a difference of only 0.5, far below the paper's own threshold of Δln(E)>5 for model preference, and the FRECKLL-only posterior log(Z)=1.74 (+0.98/-3.29) is nearly unconstrained across the prior. The paper should be revised to frame the robust conclusion as model/prior dependence rather than a measured chemical shift.","section":"Abstract and Section 3"},{"comment":"For the four planets with bimodal metallicity distributions (HAT-P-2 b, HD 189733 b, WASP-19 b, WASP-74 b), the metallicity prior is split post hoc into two sub-ranges and separate retrievals are run with independent evidences. This procedure does not produce a valid posterior probability for the two modes because the prior over the two sub-ranges is not specified, and the comparison of evidences between the log(Z)>0 and log(Z)<0 runs does not account for the different prior volumes. The reported 'preferred' metallicities, such as WASP-74 b's log(Z)=1.74, are therefore not robust inferences; the authors should either use a mixture prior within a single nested-sampling run or clearly present the full-prior posterior without claiming a preferred mode.","section":"Section 2.2 and Appendix C"},{"comment":"The text states that both TrES-3 b and WASP-77 A b benefit from TiO addition with Δln(E)>5, but Table C.2 gives Δln(E)=117.12-111.21=5.91 for TrES-3 b and Δln(E)=199.5-195.31=4.19 for WASP-77 A b. The WASP-77 A b improvement is below the paper's stated threshold, yet the paper uses the FRECKLL-TiO result as the nominal fit and discusses it as a detected improvement. This internal inconsistency should be corrected, and the threshold policy for selecting best models should be applied consistently.","section":"Section 3 and Table C.2"},{"comment":"The claimed consistency between eclipse and transit temperature profiles in deeper atmospheric layers rests on a small number of planets and is contradicted by the strong model dependence of the transit retrievals themselves. For HD 209458 b and WASP-43 b, the isothermal and 5-point temperature profile runs give diametrically opposite C/O ratio outcomes, and for HD 209458 b the metallicity changes from log(Z)=+1.54 (5-point) to -1.66 (isothermal) (Table D.1). These are not reconciled by the FRECKLL model, so the conclusion that disequilibrium chemistry reconciles eclipse and transit views should be strongly qualified or removed.","section":"Section 3 and Appendix D"}],"minor_comments":[{"comment":"The phrase 'compared to equilibrium models' is inaccurate because the paper's comparisons are against free-chemistry retrievals from Changeat et al. (2022); this should be corrected to 'compared to free-chemistry models' or 'compared to other chemistry assumptions'.","section":"Abstract"},{"comment":"The sentence 'resulting of a ln(E) > 199' is ungrammatical; it should read 'resulting in ln(E) > 199', and the exact evidence difference should be reported rather than the rounded value.","section":"Section 3, paragraph on WASP-77 A b"},{"comment":"The notation for the restricted-metallicity runs is inconsistent: the text uses 'FRECKLL, log(Z)<0' while Figure 1's caption uses 'FRECKLL-log(Z)<0'; please unify the notation throughout.","section":"Figure captions C.1-C.9 and Figure 1 caption"},{"comment":"Table C.2 is extremely wide and difficult to read; splitting it into per-planet tables or using a landscape layout would improve readability.","section":"Table C.2"},{"comment":"The sample is explicitly preselected to be valid for the FRECKLL C0-C2 network (no thermal inversion, no TiO/VO signatures, Teq<2500 K), so the population-level statements in Section 4 should carry an explicit caveat that the conclusions do not extend to hot Jupiters outside this regime.","section":"Section 2.3 and Section 4"},{"comment":"The column headers in Table A.1 use inconsistent labels ('FRECKLL' vs 'FRECKLL TiO' without dashes); the table should match the notation used in the main text.","section":"Appendix A, Table A.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a transparent, well-documented application of a complex retrieval model to a sample of HST spectra, and the community will value the warning that retrieved compositions depend on the assumed chemistry and priors. However, the central claim of 'significantly altered' metallicities and C/O ratios is not supported by the evidence differences, which are typically far below the paper's own Δln(E)>5 criterion, and the abstract mischaracterizes the comparison baseline. The authors should substantially revise the abstract and conclusions to emphasize the model-dependence demonstration, or provide a statistically valid test for the disequilibrium shifts (e.g., a mixture-prior run for the bimodal cases). The paper would be acceptable after such a revision, but the current version overstates its findings."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the first application of the TauREx-FRECKLL kinetic chemistry plugin to real HST/WFC3 spectra across ten hot Jupiters. The paper also earns credit for shipping retrieval outputs, parameter files, and plotting code; that is reproducible evidence. A second useful contribution is the WASP-43 b prior-restriction experiment: showing that retrieved metallicity and thermal structure move drastically with the chosen prior, while the data reduction method plays a smaller role, makes the cautionary conclusion—that HST-based parameter claims are model- and prior-dependent—convincing. On that point, the paper holds up.\\n\\nThe soft spot is the headline claim. The abstract says the disequilibrium approach 'significantly altered retrieved metallicity and C/O ratios compared to equilibrium models,' but the actual comparison is against the free-chemistry retrievals of Changeat et al. (2022), not equilibrium. And in the flagship case, WASP-74 b, the reported shift from log(Z) = -2.1 to +1.74 has Bayesian evidences of 202.8 and 203.29—ΔlnE ≈ 0.5, far below the paper's own ΔlnE > 5 threshold. The FRECKLL-only posterior error bar spans most of the subsolar range, so the parameter shift is model sensitivity, not a statistical detection of disequilibrium chemistry. The same holds for TrES-3 b, WASP-19 b, and WASP-43 b, where the evidence differences are small or negative. The text sometimes says the effect is moderate, but the abstract does not carry that caveat, and the planet-formation framing is stronger than the evidence.\\n\\nTwo smaller issues, in proportion. The sample was preselected for the FRECKLL regime—no thermal inversions, no TiO/VO signatures, equilibrium temperature below 2500 K—so population-level statements inherit that restriction. And the bimodal metallicity cases are resolved by splitting the prior post hoc and comparing evidences, which is model selection done on the same data; it appears in the appendix but deserves an explicit sentence in the main text. These are addressable without redoing the analysis.\\n\\nWho this is for: retrieval practitioners and people working on hot Jupiter populations and formation indicators. They should read it for the feasibility demonstration and the prior-sensitivity warning, not for updated metallicities or C/O ratios. It deserves a serious referee; the required fixes are wording, framing, and explicit acknowledgment of the sample selection and prior splits.","headline":"First real-data demonstration that kinetic disequilibrium retrievals run on HST/WFC3 spectra, plus a robust prior-sensitivity warning; the stronger claim that disequilibrium chemistry significantly shifts retrieved metallicities and C/O ratios is not statistically supported.","tokens_in":33593,"tokens_out":2165,"would_cite":true,"duration_ms":26811,"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":"Using a disequilibrium kinetic chemistry model in retrievals materially changes the inferred metallicities and C/O ratios of hot Jupiters, moving planets between formation reservoirs at nearly unchanged fit quality.","keywords":["hot Jupiter atmospheres","atmospheric retrieval","disequilibrium chemistry","photochemistry","metallicity","carbon-to-oxygen ratio","HST WFC3 spectroscopy","Bayesian model comparison"],"falsifier":"Take WASP-74 b, whose retrieved metallicity jumps from log(Z) = -2.1 (free chemistry) to log(Z) = +1.74 (FRECKLL) with nearly equal Bayesian evidence, and measure its atmospheric abundances independently, for example through high-resolution ground-based cross-correlation of CO and H2O lines; a direct abundance near solar would indicate that the disequilibrium retrieval's large shift is an artifact of the kinetic network.","tokens_in":1786,"feed_emoji":"🪐","tokens_out":2175,"duration_ms":86014,"temperature":0.7,"pith_summary":"This paper tries to establish that the chemical model used in a retrieval changes what HST observations say about a hot Jupiter. Reanalyzing 10 eclipse and 4 transit spectra with the TauREx-FRECKLL coupled model, which replaces free or thermochemical-equilibrium abundances with a full kinetic disequilibrium network, the authors find that retrieved metallicities and C/O ratios shift substantially, sometimes moving a planet from subsolar to supersolar with nearly no change in Bayesian evidence. The consequence is that population-level conclusions about planet formation drawn from simple-chemistry retrievals are not robust until the chemistry assumption is tested. The paper also argues that adding TiO and VO helps only two planets, that transit and eclipse temperature profiles agree in deep layers but not upper layers, and that HST's narrow wavelength coverage leaves C/O difficult to constrain, motivating JWST-era disequilibrium retrievals.","feed_headline":"Kinetic chemistry flips hot-Jupiter metallicities","feed_subtitle":"A 108-species kinetic network moves planets between formation families even when the spectral fit stays equally good.","key_machinery":"The load-bearing object is FRECKLL, a chemical-kinetics model of 108 H/He/C/O/N species (up to two carbon atoms), 1906 reactions and 55 photodissociations, integrated by the TauREx 3.1 plugin until steady state starting from thermochemical equilibrium, with a constant-with-altitude eddy diffusion coefficient Kzz as a free parameter. It converts the retrieval parameters, metallicity Z, C/O ratio, Kzz and the 5-point temperature profile, into physically consistent abundance profiles during the Bayesian fit, so the forward model includes vertical mixing and photochemistry instead of assuming equilibrium or constant-with-altitude abundances.","core_discovery":"The central discovery is that a disequilibrium chemical kinetic model in spectral retrieval materially alters the inferred atmospheric parameters of hot Jupiters, and that conclusions drawn from HST observations depend first on retrieval parameters, priors and the chemical model, and only secondarily on the data reduction. The paper demonstrates this with WASP-74 b: free-chemistry retrieval gives a subsolar metallicity of log(Z) = -2.1 and C/O = 0.7, while the FRECKLL disequilibrium retrieval gives a supersolar metallicity of log(Z) = +1.74 and C/O = 0.47, with Bayesian evidence essentially unchanged (202.8 versus 203.29). Such shifts reassign planets between formation reservoirs and make the choice of chemistry model part of the measurement.","pith_inferences":["An implicit consequence is that the bimodal metallicity posteriors, which the paper handles by splitting priors, may indicate missing model flexibility (such as morning-evening limb differences in transit) rather than two real atmospheric states.","A testable extension is to run the same pipeline on simulated JWST spectra with known input metallicity, C/O and Kzz, quantifying how much of the HST-era scatter is systematic model bias rather than astrophysical signal.","If the pattern holds, HST-era population metallicity distributions built from free-chemistry retrievals should be re-derived with a kinetic model before being used to constrain core-accretion and migration scenarios.","Independent dynamical constraints on Kzz from 3D circulation models would shrink the eddy-diffusion prior and test whether the retrieved disequilibrium chemistry states are physically plausible."],"forward_implications":["If the central claim is right, metallicity and C/O inferred from HST are not stable observables: the same spectra can place a planet in opposite metallicity regimes depending on the chemistry assumption, so formation-reservoir assignments based on simple chemistry need revisiting.","The near-equal Bayesian evidence for the WASP-74 b solutions means fit quality alone cannot tell which chemical model is correct, so independent abundance constraints are required to break the degeneracy.","Temperature profiles from transit and eclipse retrievals become mutually consistent below roughly 10^4 Pa when a flexible 5-point thermal profile is used, but the upper atmosphere retains an approximately 1000 K bias, so 1D transit retrievals still misstate upper-level temperatures.","TiO and VO are not needed to fit most of the sample, but for WASP-77 A b the addition of TiO narrows the thermal constraints and changes the retrieved metallicity from supersolar to subsolar, so refractory opacities matter in specific cases.","JWST's broader spectral coverage is the natural next step because HST WFC3's narrow band cannot break the C/O degeneracies that the disequilibrium model exposes."],"supporting_citations":[{"why":"Supplies the TauREx 3.1 retrieval framework in which all forward models and Bayesian sampling are run.","marker":"Al-Refaie et al. (2021)"},{"why":"Provides the plugin that couples FRECKLL chemistry into TauREx, making kinetic retrieval computationally feasible.","marker":"Al-Refaie et al. (2024)"},{"why":"Defines the 108-species, 1906-reaction kinetic network that is the paper's disequilibrium chemistry assumption.","marker":"Venot et al. (2020a)"},{"why":"Provides the eclipse spectra and the free-chemistry retrieval baseline whose metallicities and C/O ratios are compared throughout.","marker":"Changeat et al. (2022)"},{"why":"Provides the transit spectra and the free-chemistry baseline for the four transit planets.","marker":"Edwards et al. (2023)"},{"why":"Supplies the Iraclis data reduction and the eclipse/transit light-curve fitting that produces the input spectra.","marker":"Tsiaras et al. (2016)"},{"why":"Provides the alternative WASP-43b reduction used to test how much results depend on the data reduction method.","marker":"Kreidberg et al. (2014b)"},{"why":"Underlies the MultiNest nested sampling used to compute Bayesian evidence, the paper's model-selection metric.","marker":"Feroz & Hobson (2008)"}],"fun_headline_variants":["Disequilibrium kinetics flip hot-Jupiter metallicity","Kinetic model flips hot-Jupiter metallicity sign","Hot-Jupiter chemistry model decides metallicity","Same spectra, different chemical models, different planets","For hot Jupiters, priors and kinetics beat data"],"cache_read_input_tokens":35584,"weakest_assumption_plain":"The argument rests on FRECKLL's 108-species H/He/C/O/N kinetic network with up to two carbon atoms and a constant-with-altitude eddy diffusion coefficient being a faithful description of these ten atmospheres; if that network or the constant-Kzz treatment misrepresents the real chemistry, the metallicity and C/O shifts are model artifacts, not atmospheric properties.","fun_headline_variants_meta":{"raw":{"variants":["Disequilibrium kinetics flip hot-Jupiter metallicity","Kinetic model flips hot-Jupiter metallicity sign","Hot-Jupiter chemistry model decides metallicity","Same spectra, different chemical models, different planets","For hot Jupiters, priors and kinetics beat data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000262,"raw_usage":{"total_tokens":1621,"prompt_tokens":995,"completion_tokens":626,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":611,"completion_tokens_details":{"reasoning_tokens":549}},"tokens_in":611,"tokens_out":626,"duration_ms":6911,"temperature":1.0,"reasoning_tokens":549,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:41:32.087961+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take WASP-74 b, whose retrieved metallicity jumps from log(Z) = -2.1 (free chemistry) to log(Z) = +1.74 (FRECKLL) with nearly equal Bayesian evidence, and measure its atmospheric abundances independently, for example through high-resolution ground-based cross-correlation of CO and H2O lines; a direct abundance near solar would indicate that the disequilibrium retrieval's large shift is an artifact of the kinetic network.","supporting_citations":[{"cited_title":"F., et al","cited_arxiv_id":null,"evidence_quote":"Provides the eclipse spectra and the free-chemistry retrieval baseline whose metallicities and C/O ratios are compared throughout."},{"cited_title":"2023, The Astrophysical Journal Supplement, 269, 31","cited_arxiv_id":null,"evidence_quote":"Provides the transit spectra and the free-chemistry baseline for the four transit planets."},{"cited_title":"& Hobson , M","cited_arxiv_id":null,"evidence_quote":"Underlies the MultiNest nested sampling used to compute Bayesian evidence, the paper's model-selection metric."}],"review_version":1}