{"id":"35b6f3d2-35fe-417b-87e8-2e5d8031af00","arxiv_id":"2501.03998","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A uniform retrieval of 12 archival helium-triplet detections yields photoevaporative efficiencies of 0.34 +/- 0.13 and 0.75 +/- 0.21 for H/(H+He) = 0.90 and 0.99, supporting energy-limited mass loss.","lead":"Astronomers re-analyzed all public helium-triplet observations of 12 evaporating exoplanets with one uniform model, retrieving mass-loss rates and testing how they scale with stellar XUV irradiation divided by planet density. The fitted conversion efficiencies, 0.34 and 0.75 for two assumed helium abundances, support energy-limited atmospheric escape and calibrate a key parameter in planet evolution models.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Detection-only sample may imprint the Mdot–FXUV/rho trend; the population-level efficiencies and 'strong evidence' claim need a censored analysis including non-detections.","rationale":"The reader's weakest_assumption identifies the detected-only sample plus proxy-SED and density uncertainties as the key threat. I agree that the detection selection is the most load-bearing issue: it directly bears on whether the recovered Mdot vs FXUV/rho relation is a real population trend or a selection artifact. The paper's own Section 6 admission confirms the concern is real, not hypothetical. I partial-agree with the reader because the proxy-SED errors are also mentioned in the weakest_assumption, but I judge selection bias as the primary mechanism that could spuriously create or steepen the slope, while proxy errors mainly add scatter and could dilute or bias the slope less predictably. The proposed test -- incorporating non-detections through censored regression -- is a concrete, feasible check that would settle the issue. The paper otherwise deserves credit for the uniform p-winds reanalysis, Gaussian-process treatment of correlated noise, and open-source approach; the conditional verdict appropriately reflects that the population-level claim is promising but not yet robust to the acknowledged sample selection.","tokens_in":22833,"tokens_out":4434,"duration_ms":45934,"concrete_test":"Compile all published helium non-detections (Dos Santos 2021; Bennett et al. 2023; Vissapragada et al. 2024; Alam et al. 2024), compute their FXUV/rho and Mdot upper limits using the same p-winds + GP framework (or adopt published upper limits), and re-run the flat/linear/piecewise model comparison with these censored points. If the slope and epsilon remain consistent with the detection-only fit, the selection bias is not driving the trend; if the evidence for a slope drops below the strong threshold (Delta ln Z < 5) or epsilon shifts by more than 1 sigma, the 'strong evidence' conclusion must be softened to 'consistent with, but not yet proven by, the current sample.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central population claim in Section 5.2 and the conclusions rests on a sample that is exclusively helium detections: Section 2.1 explicitly removes all non-detections, and Section 6 acknowledges the bias toward large helium signatures and hence higher mass-loss rates. The helium-triplet detection probability is not uniform in Mdot or in FXUV/rho; weak outflows produce shallow signals that fall below the detection threshold, and at fixed Mdot, a larger FXUV/rho can enhance the line through stronger ionization and more extended outflows. If the detection threshold in Mdot is positively correlated with FXUV/rho, the fitted slope in Figure 3 and the derived efficiencies (0.34 +/- 0.13 and 0.75 +/- 0.21 for H/(H+He)=0.90 and 0.99) will be biased upward. The model comparison (flat vs linear vs piecewise) is performed on these 12 detections only, so the reported log-evidence preference for a slope (Delta ln Z ~ 5) could be an artifact of the truncated sample. The paper calls for future publication of non-detections but does not include already published ones, leaving the population-level 'strong evidence' conclusion unprotected against this selection effect.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a uniform reanalysis of archival helium-triplet transmission spectra of 12 exoplanets using the p-winds Parker-wind model with Gaussian-process modeling of correlated noise and nested-sampling parameter estimation. For each planet and for two assumed hydrogen fractions (H/(H+He)=0.90 and 0.99), the authors retrieve mass-loss rates, outflow temperatures, and line-of-sight velocities, and then fit these to the energy-limited scaling Mdot = 3 epsilon FXUV/(4 G rho_p). They report population-level efficiencies of epsilon = 0.34 +/- 0.13 and 0.75 +/- 0.21 for the two abundance cases, and conclude from Bayesian model comparison that the data provide \"strong evidence supporting energy-limited mass loss.\" The paper includes individual object discussions, comparisons with previous retrievals, and a brief discussion of evolutionary implications.","tokens_in":23113,"tokens_out":5452,"duration_ms":50652,"significance":"If the population-level efficiencies are robust, this would be a valuable uniform calibration of photoevaporative efficiency across a diverse sample of hot Jupiters, warm Neptunes, and mini-Neptunes, directly informing exoplanet evolution models and the interpretation of the radius gap. The paper's strengths are its homogeneous modeling framework, explicit treatment of correlated noise via Gaussian processes, use of open-source tools (p-winds, dynesty, celerite), and transparent reporting of archival data provenance and per-object posteriors (with appendix figures on Zenodo). The authors are also candid about limitations, including the bias toward large helium signals and the use of proxy SEDs. However, the central population claim currently rests on a detection-only sample and on x-axis quantities (FXUV and rho_p) whose uncertainties are not propagated, so the reported efficiencies and the \"strong evidence\" statement are not yet population-level robust.","major_comments":[{"comment":"The population-level conclusion in Section 5.2 and the abstract is based exclusively on the 12 detections; Section 2.1 removes all non-detections even though published upper limits exist, and Section 6 acknowledges the resulting bias toward large helium signatures and higher mass-loss rates. If the detection probability in Mdot is correlated with FXUV/rho, the fitted efficiencies (0.34 +/- 0.13 and 0.75 +/- 0.21) and the log-evidence preference for an energy-limited model (Delta ln Z ~ 5) can be imprinted by this selection. A censored analysis that includes non-detections as upper limits, or at minimum a quantitative demonstration that the selection does not bias the fitted slope, is required before claiming \"strong evidence supporting energy-limited mass loss\" for the population.","section":"2.1, 6"},{"comment":"For 8 of the 12 stars, the incident FXUV is derived from proxy SEDs (Table 1), and the text states that propagating the proxy uncertainty is difficult and \"does not significantly contribute to systematic errors\" (citing Zhang et al. 2024). No uncertainty on FXUV appears in Figure 3 or in the fitted efficiencies. Since Eq. (1) is linear in FXUV, a factor-of-two error in the proxy XUV flux translates directly into a factor-of-two error in epsilon, and correlated proxy errors across similar spectral types can systematically tilt the fitted slope. The paper should propagate or marginalize over SED uncertainties, or provide a sensitivity test that varies the assumed proxies.","section":"2.1, Eq. (1)"},{"comment":"The x-axis of Figure 3 uses rho_p, but the masses of TOI-1430b, TOI-1683b, and TOI-2076b are taken from the Wolfgang et al. (2016) mass-radius relation rather than measured, and Section 3.3 notes that these densities are uncertain. These three mini-Neptunes contribute to the spread in FXUV/rho, so their uncertain x-coordinates can affect the fitted slope and thus the derived efficiencies. At minimum, the fit should be repeated using the full mass-radius posterior or with densities varied within published uncertainties to show that the reported efficiencies are not driven by these assumptions.","section":"3.3, Eq. (1)"},{"comment":"The headline efficiencies quoted in the abstract and conclusions (0.34 +/- 0.13 and 0.75 +/- 0.21) are the pre-boundary slopes of the piecewise model, but the piecewise model is not significantly preferred over the simple linear model: the log-evidence differences are Delta ln Z = 0.19 for H/(H+He)=0.90 and Delta ln Z = -0.12 for H/(H+He)=0.99, with the latter actually disfavoring the piecewise model. The linear-fit efficiencies are 0.18 +/- 0.04 and 0.67 +/- 0.14. Given this weak or negative evidence for a break, the paper should either quote the linear-model efficiencies as the primary population values or justify why the piecewise pre-boundary value is the appropriate efficiency.","section":"5.2, Table 4"}],"minor_comments":[{"comment":"The phrase \"strong evidence supporting energy-limited mass loss\" is stronger than the model-selection results support; Benneke & Seager (2013) classify Delta ln Z ~ 5 as strong but the comparison is between flat and linear models, not against the piecewise model, and the detection-only sample weakens the inference.","section":"6"},{"comment":"Several references are duplicated in the bibliography, including Owen & Lai (2018) and Masson et al. (2024), which each appear twice; these should be consolidated.","section":"References"},{"comment":"The sentence \"WASP-52b contains large-amplitude correlated noise in its helium transmission spectrum, meaning a comparison with a pure Parker wind model should differ\" is unclear; it should specify that the comparison should be interpreted with caution or that the model difference is expected.","section":"4.5"},{"comment":"The phrase \"The GPs contributes 2 tunable parameters\" is grammatically incorrect and should read \"The GP contributes two tunable parameters.\"","section":"2.2"},{"comment":"The title in the posted draft reads \"T racing the Winds\"; this appears to be a typesetting artifact and should be corrected to \"Tracing the Winds.\"","section":"Title"}],"recommendation":"major_revision","confidential_remarks":"The uniform reanalysis and the explicit handling of correlated noise are genuine contributions, and the manuscript is transparent about its limitations. The central population claim, however, is not yet robust to the detection-only sample and to unpropagated uncertainties in FXUV and rho_p. I recommend major revision, with the expectation that the authors either add a censored analysis or substantially weaken the population-level claim in the abstract and conclusions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is the first uniform p-winds retrieval of all public helium-triplet detections, with a GP systematics treatment that is new for this dataset, and the resulting per-planet mass-loss rates are a useful resource. Second, the population-level efficiencies (0.34 ± 0.13 and 0.75 ± 0.21) should not be read as robust population constraints yet, because the analysis is restricted to detections only and the authors themselves acknowledge the resulting bias toward large helium signatures and high mass-loss rates.\n\nWhat the paper does well: the retrieval setup is careful — simultaneous GP and p-winds fitting, nested-sampling with dynamic live points, and model comparison with flat/linear/piecewise fits. The authors are transparent about the proxy SEDs, the mass-radius relation for mini-Neptunes, and the exclusion of non-detections. The engineering is reproducible in principle: p-winds is open-source, the input spectra are archival, and the appendix posteriors are on Zenodo. That is real credit.\n\nThe soft spots are exactly where the population claim lives. The detection threshold in a helium survey is unlikely to be uniform in Mdot or in FXUV/rho, so the fitted slope in Figure 3 and the log-evidence preference for a slope over a flat line could be imprinted by selection. The model comparison runs on the 12 detections only, so Δln Z ≈ 5 does not protect the conclusion. In addition, FXUV values from proxy SEDs come in without propagated uncertainty, and three of the mini-Neptune densities rely on a mass-radius relation; both can distort the scaling. The text notes these limitations, but the phrase “strong evidence supporting energy-limited mass loss” is stronger than the data justify.\n\nNone of this is fatal. The per-planet retrievals are worth having, and the efficiency values are clearly labeled as fitted slopes, so there is no circularity. A revision that includes or models published non-detections, propagates SED errors, and softens the population wording would materially improve the paper.\n\nWho this is for: exoplanet escape modelers and observers planning helium surveys. I would cite the per-planet mass-loss rates, and I would bring it to a reading group as a case study in selection effects, but I would not treat the efficiencies as settled. Send it to peer review: the analysis is serious, the limitations are identifiable, and the claims are testable.","headline":"A genuinely uniform reanalysis of helium detections gives useful per-planet mass-loss rates, but the 'strong evidence' for energy-limited scaling rests on a detection-only sample; treat the efficiencies as conditional.","tokens_in":23651,"tokens_out":1402,"would_cite":true,"duration_ms":16170,"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":"Twelve detected helium outflows in exoplanets follow the energy-limited mass-loss scaling, giving photoevaporative efficiencies of 0.34 and 0.75 depending on outflow hydrogen fraction.","keywords":["Exoplanet atmospheres","Extrasolar gaseous planets","Infrared astronomy","Exoplanet evolution","Atmospheric escape","Helium triplet","Photoevaporation efficiency","Energy-limited mass loss"],"falsifier":"Re-run the population fit including published helium non-detections as upper limits: if the slope of $\\dot{M}$ versus $F_{\\rm XUV}/\\rho_p$ flattens or vanishes once those bounds are added, the claimed trend is a selection artifact rather than a population law.","tokens_in":22619,"feed_emoji":"🪐","tokens_out":8675,"duration_ms":76286,"temperature":0.7,"pith_summary":"This paper asks whether planetary atmospheric escape follows the energy-limited law, which predicts that mass-loss rate is proportional to the XUV irradiation received from the host star divided by the planet's density. The authors re-analyze all twelve publicly available exoplanet helium-triplet detections with one uniform escape model, adding Gaussian-process noise modeling and nested-sampling parameter retrieval. They find that the retrieved mass-loss rates do track the ratio $F_{\\rm XUV}/\\rho_p$, with photoevaporative efficiencies of $0.34\\pm0.13$ for a solar-like hydrogen fraction and $0.75\\pm0.21$ for a hydrogen-rich outflow. If correct, this calibrates the efficiency of converting stellar XUV radiation into escaping gas and supports energy-limited escape as the governing process for these outflows. That efficiency is a key input for models of how sub-Jovian planets lose their atmospheres and evolve into super-Earths.","feed_headline":"Helium escape in 12 exoplanets follows energy-limited law","feed_subtitle":"A uniform re-analysis yields mass-loss efficiencies of 0.34 and 0.75, tying exoplanet evolution to XUV irradiation.","key_machinery":"The argument is carried by the energy-limited mass-loss identity $\\dot{M}=3\\varepsilon F_{\\rm XUV}/(4G\\rho_p)$, which ties the observable mass-loss rate to the ratio of incident XUV flux over planetary density through a single efficiency $\\varepsilon$. To measure $\\dot{M}$ uniformly, the paper feeds each archival transmission spectrum through one hydrodynamic Parker-wind escape model, treats correlated noise with a Gaussian-process covariance initialized outside the helium line, and searches posteriors with nested sampling. The efficiency is then obtained by fitting the linear relation, and a piecewise version with a break at $F_{\\rm XUV}/\\rho_p\\sim10^4$, to the twelve retrieved rates.","core_discovery":"The central claim is that the population of detected helium outflows is consistent with energy-limited mass loss. Fitting the relation $\\dot{M}=3\\varepsilon F_{\\rm XUV}/(4G\\rho_p)$ to the twelve retrieved mass-loss rates, the authors report a piecewise slope break near $F_{\\rm XUV}/\\rho_p\\sim10^4$; below that threshold the inferred efficiencies are $\\varepsilon=0.34\\pm0.13$ at H/(H+He)=0.90 and $0.75\\pm0.21$ at 0.99, dropping to $0.10\\pm0.06$ and $0.57\\pm0.19$ beyond it. Bayesian model comparison favors the linear and piecewise energy-limited models over a flat line, which the authors interpret as support for energy-limited escape. They also show that under these efficiencies the mini-Neptunes in the sample can lose a substantial fraction of their atmosphere in a gigayear, whereas hot Jupiters are mostly resilient.","pith_inferences":["Inference: If the calibration holds, the measured efficiency can be inserted into evolutionary models to predict which sub-Neptunes lose their envelopes and become super-Earths, tying the observed radius gap to a measured number rather than a free parameter.","Inference: The higher efficiency at H/(H+He)=0.99 suggests that measuring the helium abundance of an outflow, not just the line depth, is what matters most for population-level evolution, since a hydrogen-rich flow can lose mass several times faster at the same irradiation.","Inference: Because only detections entered the sample, the reported efficiencies are upper-regime values; adding published non-detections as upper limits could lower the population efficiency and sharpen the comparison with the Neptune-desert boundary.","Inference: The same Gaussian-process plus hydrodynamic-wind pipeline could be applied to Lyman-alpha and Balmer-line escape tracers to test whether the energy-limited efficiency is independent of the observing tracer."],"forward_implications":["If energy-limited escape holds, the mass-loss rate of an exoplanet is set mainly by $F_{\\rm XUV}/\\rho_p$, so planets with the same ratio should lose mass at the same rate.","The calibrated efficiencies give evolutionary models a concrete input: at solar helium abundance, roughly 34% of XUV energy goes into driving the outflow, while hydrogen-rich outflows convert about 75%.","The break near $F_{\\rm XUV}/\\rho_p\\sim10^4$ means the simple linear scaling overestimates mass loss for the most strongly irradiated planets.","Mini-Neptunes in the sample can lose a large fraction of their atmospheres within a gigayear, supporting their evolution into super-Earths, while hot Jupiters mostly retain their atmospheres.","Because the sample contains only detected outflows, the population trend chiefly applies to planets with substantial mass loss, and the paper calls for publishing non-detections to constrain the rest."],"supporting_citations":[{"why":"Supplies the Parker-wind metastable helium model from which the mass-loss rates are retrieved.","marker":"Oklopčić & Hirata (2018)"},{"why":"Provides the energy-limited formula $\\dot{M}=3\\varepsilon F_{\\rm XUV}/(4G\\rho_p)$ that the population fit assumes.","marker":"Caldiroli et al. (2022)"},{"why":"Earlier photometric population study whose efficiency constraint the paper's $\\varepsilon$ is compared against.","marker":"Vissapragada et al. (2022)"},{"why":"Source of the Keck/NIRSPEC spectra and mass-radius relation masses for the four mini-Neptunes.","marker":"Zhang et al. (2023)"},{"why":"Presents the hydrodynamic escape code and its HAT-P-11b retrieval, the baseline for the escape-model fits.","marker":"Dos Santos et al. (2022)"},{"why":"Supplies the WASP-52b and WASP-177b NIRSPEC data and prior mass-loss retrievals.","marker":"Kirk et al. (2022)"},{"why":"Sets the $F_{\\rm XUV}/\\rho_p\\sim10^4$ threshold where the piecewise fit changes slope.","marker":"Murray-Clay et al. (2009)"},{"why":"Comparison hydrodynamic retrievals for GJ-3470b and support for super-solar hydrogen abundances.","marker":"Lampón et al. (2021)"}],"fun_headline_variants":["Helium escape in 12 exoplanets follows energy-limited law","Energy-limited escape supported by helium triplet survey","XUV-to-density ratio drives exoplanet mass loss","Photoevaporation efficiencies 0.34 and 0.75 from He lines","Uniform helium analysis ties exoplanet winds to XUV"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the detected-only sample, whose XUV fluxes come partly from proxy stars and whose densities come partly from mass-radius relations, preserves the true population-level relation between mass-loss rate and $F_{\\rm XUV}/\\rho_p$.","fun_headline_variants_meta":{"raw":{"variants":["Helium escape in 12 exoplanets follows energy-limited law","Energy-limited escape supported by helium triplet survey","XUV-to-density ratio drives exoplanet mass loss","Photoevaporation efficiencies 0.34 and 0.75 from He lines","Uniform helium analysis ties exoplanet winds to XUV"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000209,"raw_usage":{"total_tokens":1422,"prompt_tokens":975,"completion_tokens":447,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":591,"completion_tokens_details":{"reasoning_tokens":361}},"tokens_in":591,"tokens_out":447,"duration_ms":4527,"temperature":1.0,"reasoning_tokens":361,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:42:34.512730+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the population fit including published helium non-detections as upper limits: if the slope of $\\dot{M}$ versus $F_{\\rm XUV}/\\rho_p$ flattens or vanishes once those bounds are added, the claimed trend is a selection artifact rather than a population law.","supporting_citations":[],"review_version":1}