{"id":"4a411061-c450-4eb2-9ceb-a9fe03f583ee","arxiv_id":"2506.16232","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"New analysis of HST plus JWST spectra finds near-solar water and carbon dioxide and a strongly sub-solar carbon-to-oxygen ratio in HD 209458b, but the C/O value depends on a ground-based CO prior.","lead":"This paper re-analyzes old Hubble Space Telescope observations of the hot Jupiter HD 209458b and combines them with JWST data, concluding that the planet's atmosphere has about solar metal abundance but a very low carbon-to-oxygen ratio. The result matters because it bears on how giant planets form and migrate, since the carbon-to-oxygen ratio records the planet's formation history.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline C/O and metallicity rest on an external CO prior in a retrieval the same data disfavour by 2.6σ; without that prior the C/O is ~1e-3, so the central claim is prior-driven rather than data-confirmed.","rationale":"The paper does careful work: a new HST reduction, clear detections of H2O and CO2, and a reasonable BMA treatment. I agree with the Reader's CONDITIONAL verdict. The most load-bearing weak point is exactly the CO prior: the headline C/O is not data-detected. The paper itself flags in Sect. 5 that CO is not detectable in the JWST spectrum and that simplified cloud models may hide it, so this is a self-identified limitation, not an external nitpick. The proposed synthetic closure test would settle whether the low C/O is real or an artifact of the prior and cloud assumptions. If the closure test separates the two inputs robustly, the claim would be strengthened; if not, the result should be reported as a conditional upper limit rather than a confirmed value. No change to the Reader's verdict is needed.","tokens_in":29237,"tokens_out":9623,"duration_ms":119202,"concrete_test":"Closure test on synthetic data: generate HST+JWST spectra from (a) the cloudy+CO prior best fit (C/O≈0.05) and (b) a solar-composition model with C/O=0.47 and the same retrieved H2O/CO2 abundances and cloud parameters, adding Gaussian noise at the published error bars. Run the full retrieval pipeline on both with and without the Brogi & Line CO prior. If the two inputs produce overlapping C/O posteriors (or if removing the prior collapses both to C/O<0.1), the low C/O is a prior artefact and the central claim fails. A complementary check is to replace the grey cloud deck with the complex-cloud model of Eq. (2) in the CO-prior retrieval and see whether the C/O moves above 0.2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.5 derives [M/H]=0.10 and C/O=0.054 from the single 'cloudy + CO prior' model. The same HST+JWST data without the Brogi & Line (2019) CO prior give [M/H]=-1.35 and C/O=1.3e-3, and the CO-prior model is disfavoured by 2.6σ in the paper's own evidence comparison (log Z = 365.4 vs 366.3). CO itself is not detected in the JWST spectrum; the paper's BMA only reports a 3σ upper limit of log χCO=-3.26, and Fig. 11 shows the CO band is obscured by H2O and the grey cloud deck. The CO prior therefore does the work: the retrieved log χCO=-4.15 is near the lower edge of the uniform mass-fraction prior U(-3.5,0) (converted to VMR, lower edge ≈ -4.6), not near the ground-based central value, and the data are still pushing CO down. The Discussion explicitly warns that the CO non-detection 'is particularly concerning' and that simplified cloud models may hide CO. Since the headline C/O is not robust to removing the prior or to the choice of cloud model, the claim 'confirming a low C/O' overstates what the data alone establish.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a new reduction of the original HST/WFC3 transmission spectrum of HD 209458b using the PACMAN pipeline, with a careful treatment of wavelength-dependent instrument systematics, and combines it with archival JWST/NIRCam data (Xue et al. 2024). Free-chemistry retrievals over 1.0–5.1 μm robustly detect H2O and CO2 (both >7σ), find a 3.6σ preference for a grey cloud deck over a clear atmosphere, and, after Bayesian model averaging, report H2O and CO2 abundances consistent with solar values within 1σ. In Section 4.5, the authors add a CO abundance prior from ground-based high-resolution spectroscopy (Brogi & Line 2019) to a cloudy retrieval and derive [M/H] = 0.10^{+0.41}_{-0.40} and C/O = 0.054^{+0.080}_{-0.034} with a 3σ upper limit of 0.454. They interpret these as evidence for solar metallicity and strong oxygen enrichment / carbon depletion during formation, and compare their results with the VULCAN 1D photochemistry model.","tokens_in":29543,"tokens_out":7264,"duration_ms":83947,"significance":"If the robust H2O and CO2 detections and the Bayesian-model-averaged abundances stand, the paper provides a valuable, high-quality benchmark for the atmospheric composition of HD 209458b and helps resolve earlier contradictory water-abundance claims. The careful re-reduction of the WFC3 data, the systematic exploration of cloud models, the use of Bayesian model averaging, and the explicit comparison with the VULCAN 1D model are methodological strengths, and the paper is commendably transparent about the model dependence of its results. However, the headline claims of solar metallicity and very low C/O are not as robust as the abstract and title suggest: they are derived from a single retrieval that is disfavoured by the paper's own model comparison, and the CO non-detection on which the low C/O largely rests is acknowledged as potentially an artefact of clouds. As it stands, the paper would be a solid joint retrieval study demonstrating prior and model sensitivity, but the 'confirming a low C/O' claim is not supported by the data alone.","major_comments":[{"comment":"The paper's central claim—solar metallicity [M/H]=0.10 and very low C/O=0.054—is taken exclusively from the 'cloudy + CO prior' retrieval, but the paper's own evidence comparison disfavours this model. The logarithmic Bayesian evidence for that retrieval is 365.4±0.2, compared to 366.3±0.0 for the cloudy retrieval with the broad CO prior (a difference of 2.6σ as stated in Sect. 4.5), and compared to 367.8±0.1 for the best model containing only H2O and CO2. Without the CO prior, the same data yield [M/H]=-1.35^{+1.25}_{-0.73} and C/O=1.33^{+4.79}_{-0.70}×10^{-3}. The abstract and conclusions present the prior-based values as the confirmed result, while the Discussion (Sect. 5) concedes that CO is not detected and that the apparent CO depletion may be an artefact of clouds. This is a load-bearing inconsistency: the headline claim is not robust to the model choices documented in the paper. The authors should either present the C/O and metallicity as explicitly conditional on the CO prior and cloud model, or provide a principled statistical argument for preferring the disfavoured model for inference.","section":"Abstract; Sect. 4.5 and Table 6"},{"comment":"The 'prior on the CO abundance from ground-based measurements' is implemented as a uniform prior on log mass fraction U(-3.5, 0.0), rather than as a measurement-informed posterior or likelihood. The retrieved log χCO = -4.15 lies close to the lower edge of this prior after conversion to volume mixing ratio (approximately -4.6), and well below the centre of the prior range; the data are therefore still pulling CO down against the prior. The phrase in the abstract, 'Combining these values with a prior on the CO abundance from ground-based measurements,' overstates the support the ground-based measurement provides for the specific C/O=0.054. The authors should report the Brogi & Line (2019) CO abundance value explicitly, and either use a properly informative Gaussian prior centred on that measurement or explain why a wide uniform range is being labelled 'informed'.","section":"Sect. 4.5 and Fig. A.5"},{"comment":"The low C/O is driven almost entirely by non-detections and upper limits: the 3σ upper limit on CO is log χCO = -3.26 (Sect. 4.4), CH4 is unconstrained, and the paper states that CO features are 'obscured by stronger water absorption' and 'hidden beneath the cloud deck,' concluding that the apparent CO depletion 'may be an artefact of the clouds.' Because the CO abundance is not measured, the derived C/O = 0.054 is an upper-limit-driven value, not a detected carbon depletion. The formation interpretation in Sect. 5—'strong enrichment in oxygen and depletion in carbon'—is therefore not supported by the data. The authors should propagate the CO upper limit through the C/O calculation in a way that transparently shows the non-detection, or remove the formation claim until CO is detected or constrained.","section":"Sect. 4.5, Sect. 5, and Fig. 11"}],"minor_comments":[{"comment":"The text reads 'our caluclated C/O'; 'caluclated' should be 'calculated'.","section":"Sect. 4.7"},{"comment":"The note 'log (Pcloud) is the the cloud base pressure' contains a duplicated article; it appears in both tables.","section":"Notes to Tables A.2 and A.3"},{"comment":"The caption states that 'The dashed spectra are best fits using the more complex cloud model described in Sect. 4.3,' but the figure legend does not clearly distinguish the 'complex clouds' and 'complex patchy clouds' variants; please make the line styles explicit.","section":"Fig. 8 caption"},{"comment":"The description of the recreated Deming et al. (2013) reduction is clear, but the statement that the Gaussian convolution 'was the crucial step to avoid the zig-zag pattern' could be better quantified: consider showing the pattern amplitude before and after convolution as a function of wavelength.","section":"Sect. 2.2"},{"comment":"In the text, the significances for the non-detected species are quoted as 'from 2.8σ to 3.0σ' and the Bayesian evidence disfavours their inclusion; please verify these sigma values are consistent with the Bayes factors in Table 6, since the table lists Zref/Z values that do not all map to those integers.","section":"Sect. 4.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is transparent and the data analysis appears careful, but the title and abstract substantially overstate the robustness of the C/O and metallicity results: the CO-prior model is disfavoured by the authors' own model comparison, yet it is the sole source of the headline numbers. This is fixable by reframing the paper as a model-dependent composition study and clearly separating the robust detections (H2O, CO2, clouds) from the prior-influenced C/O inference. I would also recommend that the authors run an additional retrieval with a Gaussian CO prior centred on the Brogi & Line (2019) central value and quantify the sensitivity of C/O to that choice; this would directly address the reviewer's main concern. No circularity or lack of originality is apparent; the key issue is interpretive framing rather than technical execution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid retrieval paper on the benchmark hot Jupiter, with a genuinely useful new HST/WFC3 reduction and robust (>7σ) H2O and CO2 detections from the combined HST+JWST spectrum. But the headline 'confirming low C/O and solar metallicity' is not actually confirmed by the data alone. The low C/O comes from the 'cloudy + CO prior' model, which the paper's own evidence comparison disfavours by 2.6σ relative to the broad-CO-prior model. Without the ground-based CO prior (Brogi & Line 2019), the same data give C/O ~1e-3 and [M/H] ~ -1.35, both essentially unphysical and driven by the CO non-detection. The paper is transparent about this — Section 4.5 and the Discussion flag that the CO non-detection is 'particularly concerning' — but the abstract and title still assert a confirmation.\n\nThe re-reduction is careful: they track down a zig-zag artifact and show that wavelength-dependent ramp fitting matters; the joint retrieval covers many species and uses BMA properly. The H2O and CO2 abundances at solar are probably the most reliable values yet for this planet. The VULCAN comparison is a benchmark, not an input, so no circularity. The treatment of previous work is fair; they cite Xue et al. (2024) and Verma et al. (2025) and situate differences.\n\nThe main problem is not the retrieval code, it's that the central claim is prior-driven. The CO prior is independent and physically motivated, so this is not circular reasoning, but the data are still pushing CO down even within the prior (retrieved log chi_CO = -4.15 is near the lower edge of the converted prior). The paper's own Fig. 11 shows CO band obscured by H2O and clouds. Either the cloud model is hiding CO or the ground-based and space-based measurements are inconsistent; the paper does not resolve that. That inconsistency is exactly why the title should not say 'confirming.' A second, related weakness: the BMA water abundance quoted in the abstract (0.95x solar) comes from the joint retrieval BMA, which is fine, but the C/O and metallicity are not BMA results — they come from a single, disfavoured model. That asymmetry is under-emphasized.\n\nWho is this for? Exoplanet atmosphere people working on hot Jupiters, retrieval methodology, and the HST-vs-JWST consistency question. It deserves a serious referee and probably a major revision, mostly to soften the claim and to add robustness tests: e.g., show how C/O changes under the more complex cloud model with the CO prior, and report the BMA C/O if one averages over priors. The re-reduction alone is worth publishing; the C/O is a hypothesis, not a confirmation.","headline":"Careful HST re-reduction and solid H2O/CO2 detections, but the headline C/O and metallicity rest on a disfavoured CO prior; accept with major revision.","tokens_in":30131,"tokens_out":1719,"would_cite":true,"duration_ms":17689,"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":"HD 209458b's atmosphere appears solar in metallicity ([M/H] = 0.10) yet extremely carbon-poor (C/O = 0.054), implying strong oxygen enrichment during formation.","keywords":["HD 209458b","hot Jupiter","transmission spectroscopy","atmospheric retrieval","carbon-to-oxygen ratio","water abundance","Bayesian model averaging","JWST NIRCam"],"falsifier":"A higher-resolution or longer-wavelength observation that resolves the CO band at 4.4–5.2 μm would settle the claim: detecting CO at near-solar abundance would push C/O up toward the stellar value of 0.47, and a retrieval without the CO prior that recovers CO directly from the space-based spectrum would make the low C/O claim unnecessary. A simpler check is to rerun the cloudy retrieval with a more flexible cloud model and the CO prior removed to see whether CO features emerge.","tokens_in":29034,"feed_emoji":"🪐","tokens_out":10112,"duration_ms":96374,"temperature":0.7,"pith_summary":"This paper re-reduces the original HST/WFC3 transmission spectrum of the hot Jupiter HD 209458b, explicitly modelling the wavelength-dependent instrument systematics, and combines it with archival JWST/NIRCam data to run free-chemistry atmospheric retrievals over 1.0–5.1 μm. The authors argue that the much-debated water abundance is close to solar, at 0.95+0.35−0.17 times solar, with CO2 also near-solar, and that the data prefer a cloudy atmosphere over a clear one by 3.6σ. Adding a ground-based high-resolution measurement of CO as a prior shifts the inferred composition to solar metallicity ([M/H] = 0.10+0.41−0.40) and an extremely low carbon-to-oxygen ratio (C/O = 0.054+0.080−0.034, with a 3σ upper limit of 0.454). If these values hold, HD 209458b formed with strong oxygen enrichment and carbon depletion, matching a formation story that involves pebble accretion and inward migration.","feed_headline":"Osiris atmosphere: solar metals, C/O near 0.05","feed_subtitle":"Re-reduced HST plus JWST data show solar water and CO2, but carbon is depleted, hinting at formation far out.","key_machinery":"The argument is carried by free-chemistry atmospheric retrievals that do not impose equilibrium chemistry, run on a jointly fit HST and JWST transmission spectrum spanning 1.0–5.1 μm, with a grey cloud deck at a retrieved pressure. Model uncertainty is handled with Bayesian model averaging over five cloud treatments, which widens the abundance uncertainties to include cloud assumptions. The load-bearing element is the 'cloudy + CO prior' retrieval, which puts a narrow prior on CO from a ground-based high-resolution detection, because CO is invisible in the low-resolution JWST band (masked by clouds and water). Without that prior the same data give a very metal-poor result ([M/H] = −1.35) and an even lower C/O (1.3 × 10−3), and the prior-based model is formally disfavoured by 2.6σ; the prior is what produces the solar-metallicity, carbon-depleted composition.","core_discovery":"The central claim is that HD 209458b's atmosphere contains near-solar water and carbon dioxide and overall has a solar-like metallicity with a very low carbon-to-oxygen ratio. H2O and CO2 are the only molecules firmly detected, at high significance, and clouds are preferred over a clear atmosphere. Bayesian model averaging over five cloud treatments gives water at 0.95+0.35−0.17 times solar and CO2 at 0.94+0.16−0.09 times solar. From the retrieved abundances, and only with a CO abundance prior taken from ground-based high-resolution spectroscopy, the composition is [M/H] = 0.10+0.41−0.40 and C/O = 0.054+0.080−0.034, with a 3σ upper limit of 0.454, below the stellar C/O of 0.47. The authors read this as a strong enrichment in oxygen and depletion in carbon during the planet's formation.","pith_inferences":["If the CO prior holds, other hot Jupiters whose low-resolution spectra yield extremely low C/O should be re-examined with ground-based CO measurements; some reported low C/O values may be an artefact of the same CO invisibility.","A direct extension of this work would be to predict that higher-resolution JWST observations of the 4.4–5.2 μm region will either detect CO (falsifying the grey-cloud suppression) or confirm that clouds hide it, testing the formation interpretation.","The strong sensitivity of C/O to the cloud prescription and the CO prior implies that formation-history inferences from single-planet C/O values deserve larger error bars than the quoted posteriors until CO is measured directly."],"forward_implications":["Earlier sub-solar water detections from HST data alone are not robust; the re-reduced spectrum combined with JWST puts H2O within 1σ of solar.","CO can be undetectable in a low-resolution JWST spectrum even when it is strongly present, because clouds and water occult its bands; space-based low-resolution spectra alone can miss a major carbon reservoir.","The inferred C/O of 0.054 is below the stellar value of 0.47, which the authors connect to oxygen-rich solid accretion and inward migration during formation.","Bayesian evidence alone is an unsafe criterion for choosing between cloud models of exoplanets; Bayesian model averaging better captures the true model uncertainty."],"supporting_citations":[{"why":"Original HST/WFC3 transmission spectrum that this paper re-reduces with wavelength-dependent systematics and uses for the joint fit.","marker":"Deming et al. (2013)"},{"why":"Source of the archival JWST/NIRCam spectrum and the equilibrium-chemistry retrieval results this work compares against.","marker":"Xue et al. (2024)"},{"why":"Ground-based high-resolution CO detection whose abundance is adopted as the prior in the 'cloudy + CO prior' retrieval that produces the headline C/O and metallicity.","marker":"Brogi & Line (2019)"},{"why":"Provides the free-chemistry radiative-transfer forward model used for all retrievals.","marker":"Mollière et al. (2019)"},{"why":"Data-reduction pipeline used to re-reduce the HST/WFC3 spectral data and model the light-curve systematics.","marker":"Zieba & Kreidberg (2022)"},{"why":"1D photochemistry model whose predicted abundances are compared with the retrieved H2O, CO2, and upper limits.","marker":"Tsai et al. (2021)"},{"why":"Supplies the adopted planet mass, radius, and orbital parameters used to build the retrieval model.","marker":"Bonomo et al. (2017)"}],"fun_headline_variants":["Osiris: solar metallicity, carbon-starved","HD 209458b: near-solar H2O & CO2, C/O = 0.05","Oxygen-rich Osiris: solar metals, tiny C/O","Solar-like metals, but carbon-poor in Osiris"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline C/O and metallicity rest on the assumption that the ground-based CO abundance used as a prior is correct; without it, the same data prefer a very metal-poor, even more carbon-depleted atmosphere, and the model that carries the paper's conclusion is formally disfavoured by the data's own model comparison.","fun_headline_variants_meta":{"raw":{"variants":["Osiris: solar metallicity, carbon-starved","HD 209458b: near-solar H2O & CO2, C/O = 0.05","Oxygen-rich Osiris: solar metals, tiny C/O","Solar-like metals, but carbon-poor in Osiris"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001674,"raw_usage":{"total_tokens":6764,"prompt_tokens":1197,"completion_tokens":5567,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":813,"completion_tokens_details":{"reasoning_tokens":5488}},"tokens_in":813,"tokens_out":5567,"duration_ms":49460,"temperature":1.0,"reasoning_tokens":5488,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T23:44:25.185204+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A higher-resolution or longer-wavelength observation that resolves the CO band at 4.4–5.2 μm would settle the claim: detecting CO at near-solar abundance would push C/O up toward the stellar value of 0.47, and a retrieval without the CO prior that recovers CO directly from the space-based spectrum would make the low C/O claim unnecessary. A simpler check is to rerun the cloudy retrieval with a more flexible cloud model and the CO prior removed to see whether CO features emerge.","supporting_citations":[{"cited_title":"2013, ApJ, 774, 95","cited_arxiv_id":null,"evidence_quote":"Original HST/WFC3 transmission spectrum that this paper re-reduces with wavelength-dependent systematics and uses for the joint fit."},{"cited_title":"model_ramp","cited_arxiv_id":null,"evidence_quote":"Data-reduction pipeline used to re-reduce the HST/WFC3 spectral data and model the light-curve systematics."},{"cited_title":"S., Desidera, S., Benatti, S., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the adopted planet mass, radius, and orbital parameters used to build the retrieval model."}],"review_version":1}