{"id":"854b05ab-48b0-4f90-ad0e-2f8d955e74f9","arxiv_id":"2608.13473","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Four JWST NIRISS/SOSS transits detect no metastable helium absorption on LHS 1140 b, rejecting the reported ground-based detection at more than 3 sigma per visit and about 10 sigma combined.","lead":"Four JWST transits of the exoplanet LHS 1140 b saw no helium escaping from its atmosphere. The result casts doubt on a 2024 ground-based helium detection, unless the signal appears in fewer than half of all transits.","discovery_kind":"replication","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The >3σ rejection of the ground-based helium model depends on a ~600 ppm predicted SOSS signal derived from the pwinds best-fit model rather than from the measured equivalent width of the WINERED detection; if that EW is overestimated, the per-visit significances and the f<53% duty-cycle bound…","rationale":"The paper is a careful observational consistency check. The four-transit null result is supported by two independent reductions (FIREFLy and exoTEDRF), column-level light curves with adjacent-column checks, and a clear visual absence of the 1.0833 µm feature. The duty-cycle argument is honestly hedged. The most load-bearing step is the conversion of the reported ground-based detection into an expected SOSS signal (Sections 4.2-4.3). That conversion uses the pwinds best-fit model rather than a direct measurement of the equivalent width of the WINERED detection. If the model overestimates the EW, the predicted ~600 ppm pixel feature could be too large by a factor of two, which would reduce each visit's >3σ rejection to roughly 1-2σ and would widen the allowed variability fraction beyond f<53%. This does not undermine the primary nondetection, but it does weaken the quantitative claim that the ground-based model is rejected at high significance and the associated inference that the 2024 detection is likely spurious. A direct EW measurement from the archived WINERED spectrum would settle this. Because the paper's strongest quantitative claims are conditional on this calibration, I recommend CONDITIONAL acceptance rather than unconditional ACCEPT, with the condition being the addition of the EW-based check or a revised statement of the attainable significance.","tokens_in":17539,"tokens_out":10614,"duration_ms":106297,"concrete_test":"Take the archived WINERED spectrum from Cherubim et al. (2026) and measure the triplet equivalent width directly by integrating the in-transit/out-of-transit flux ratio over 1.082-1.085 µm. Convert this EW to the expected SOSS in-transit deficit using the measured SOSS pixel width (~0.7 nm) and the empirical SOSS PSF, then fit the central and ±1 columns jointly while allowing a free sub-pixel position. If the data still exclude this EW-based model at >3σ in each visit, the conclusion is robust; if not, the per-visit rejection significances and the f<53% bound should be recomputed and the paper's claim softened.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Section 4.3 predicts that the ground-based 1.24% absorption translates to a ~600 ppm SOSS pixel feature by convolving the pwinds best-fit model with a Gaussian kernel (R≈700) and binning to the pixel scale. The amplitude of this prediction is set by the equivalent width (EW) of the model line, not directly by the measured WINERED spectrum. If the model overestimates the EW (e.g., because the true line is narrower or the continuum and velocity structure differ), the expected SOSS signal would be smaller. Table 1's 2σ upper limits at SOSS resolution are 0.02-0.04% (200-400 ppm); a predicted signal of 300 ppm rather than 600 ppm would reduce each visit's rejection from >3σ to roughly 1-2σ, and the combined 9.9σ would drop to about 4-5σ. The duty-cycle bound f<53% also assumes that a 2024-like transit would be unambiguously detectable, so a smaller predicted signal would widen the allowed variability window. The paper's adjacent-column checks address sub-pixel displacement but not the absolute amplitude calibration of the conversion.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes four JWST NIRISS/SOSS transits of LHS 1140 b observed between 2023 and 2026, searching for metastable helium absorption at 1.0833 µm. Two independent reductions (FIREFLy and exoTEDRF) agree; the authors find no excess absorption in the helium light curves, in the column-level spectra, or in adjacent columns for any visit. They place 2σ upper limits on helium absorption depth at both SOSS and ground-based resolutions, translate the best-fit ground-based model of Cherubim et al. (2026) into an expected ~600 ppm SOSS feature, and report that each visit rejects this model at >3σ, with a combined rejection at 9.9σ. They also use pwinds to derive mass-loss rate upper limits, finding no trend over time and arguing that the 2024 reported detection is an outlier. The paper concludes that the ground-based detection may be spurious or, if time-variable, occurs in fewer than about 50% of transits.","tokens_in":17780,"tokens_out":6563,"duration_ms":62959,"significance":"If the conclusion holds, this is an important result that directly challenges a published detection of metastable helium escape from a high-profile temperate exoplanet, with implications for the planet's atmospheric composition and for the reliability of ground-based helium surveys. The paper is methodologically careful in several respects: it uses two independent reduction pipelines, checks adjacent columns for signal leakage, discusses wavelength-solution offsets explicitly, and makes its data products available. The strength of the nondetection is credible. However, the headline rejection significances depend on converting a model-dependent equivalent width into an expected SOSS signal and on a χ²/N-to-σ conversion that does not fully account for correlated systematics. These issues are fixable and do not undermine the basic finding that no helium is visible in these four transits, but they affect the paper's strongest quantitative claims.","major_comments":[{"comment":"The expected SOSS signal of ~600 ppm is computed by convolving the pwinds best-fit model from Cherubim et al. (2026) with a Gaussian SOSS PSF, rather than by directly integrating the equivalent width measured in the WINERED spectrum. If the true line profile has a smaller equivalent width than the model (e.g., a narrower core or different continuum placement), the predicted SOSS signal could be substantially lower, and the per-visit rejection significances would drop from >3σ to roughly 1–2σ. The adjacent-column checks in §4.1 mitigate sub-pixel centroid issues but do not calibrate the absolute amplitude of the prediction. Please either derive the expected SOSS signal from the measured equivalent width of the ground-based detection, or provide a sensitivity grid over plausible line widths, centroids, and triplet morphologies.","section":"§4.3, Figure 2"},{"comment":"The stated rejection significances are obtained by converting χ²/N to a Gaussian sigma via the chi-square distribution, under the assumption that the per-column errors are independent and Gaussian. The transmission spectra are subject to correlated systematics from 1/f noise and possible spot crossings, and the χ²/N values of 2.6–3.8 may reflect underestimated uncertainties rather than genuine model rejection. I recommend an injection-recovery test: add the convolved model at full amplitude (and at half amplitude) to the actual light curves and evaluate how often the recovered signal exceeds 3σ. This would place the per-visit and combined significances on a more robust footing.","section":"§4.3, Figure 2"},{"comment":"The mass-loss upper limits and the 'outlier' characterization of the 2024 detection are computed within pwinds varying only Ṁ while holding the other parameters fixed, as the authors acknowledge. Given the known degeneracies among Ṁ, T_wind, H:He, and XUV flux, and the non-monotonic behavior of the helium feature at high Ṁ due to self-shielding, the limits should be presented as conditional on the adopted model rather than as a direct measurement of the mass-loss rate. The text does include this caveat, but the abstract and conclusion statements ('no clear trend in mass-loss with time') could mislead readers who do not read the caveats.","section":"§4.4, Table 1, Figure 3"},{"comment":"The f<53% at 2σ bound assumes that a transit with the 2024 signal amplitude would always be detected in each of the four SOSS visits and that the four visits have equal sensitivity. Since the per-visit upper limits differ (0.02–0.04% at SOSS resolution) and the detection significance for a full-amplitude signal is only >3σ, the binomial calculation may underestimate the allowed variability window. Please incorporate the measured upper limits into the detection probability, or clearly label the bound as a first-order estimate.","section":"§5, duty-cycle bound"}],"minor_comments":[{"comment":"The claim that the FIREFLy reduction reproduces the Cadieux et al. (2024b) transmission spectrum for Visits 1 and 2 would benefit from a quantitative comparison (e.g., a plot or RMS of residuals) rather than a textual statement.","section":"§3.1"},{"comment":"The conversion of upper limits from SOSS to WINERED resolution assumes the same line morphology as the Cherubim et al. best-fit; please state this explicitly in the text.","section":"§4.2"},{"comment":"The χ²/N values are given in the panels, but the number of spectral points N is not stated; please report N in the caption or text.","section":"Figure 2"},{"comment":"The abstract says '≲50%' while the text gives f<53% at 2σ; please make the numbers consistent.","section":"Abstract and §5"},{"comment":"The word 'incontrovertible' is unnecessarily strong given the model-dependent caveats; please soften it.","section":"§5"},{"comment":"The in-text citation 'C. Cherubim et al. (2026)' should match the reference entry 'Cherubim, C., Vissapragada, S., Cunningham, T., et al. 2026' in author order if applicable.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The nondetection itself is solid and the paper is a valuable contribution. My main concern is that the headline claim (rejection of the reported detection at >3σ per visit and 9.9σ combined) rests on a model-dependent conversion and a χ²/N significance estimate that may be optimistic. If the authors can add an injection-recovery test and an amplitude-calibration check, the paper would be suitable for publication. The paper is likely to be heavily cited in the debate over this object, so the statistical rigor matters."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi,\n\nShort version: this is a careful null result that likely retires the reported ground-based helium detection on LHS 1140 b, or pushes it into the rare/stochastic camp. The paper adds two new JWST transits to two archival ones, reduces everything in two independent pipelines (FIREFLy and exoTEDRF), and finds no excess in the sub-pixel column, in adjacent columns, or in the binned spectra. That is real evidence, and the authors stay appropriately provisional about time variability. I would send it to review.\n\nWhat is new: four NIRISS/SOSS transits from 2023 to 2026, none contemporaneous with the 2024 WINERED detection but all sensitive at the reported level. Two of the four are new GO 7073 observations. The explicit comparison with the Cherubim et al. best-fit pwinds model, convolved to SOSS resolution, is the right frame for a consistency check. Data products are on Zenodo, which helps.\n\nWhat is soft: the headline significances (>3σ per visit, 9.9σ combined) come from converting chi-squared per degree of freedom to Gaussian sigma. That ignores correlated systematics and is not a full model comparison. The true rejection is probably weaker, though the light curves look clean enough that I doubt the main conclusion flips. The predicted SOSS amplitude (~600 ppm) is inherited from the pwinds best-fit model, not from the measured WINERED equivalent width; if that model's EW is off by a factor of two, per-visit rejections drop to 1-2σ and the combined significance to roughly 4-5σ. The adjacent-column checks address pixel offsets, not absolute amplitude calibration, so this caveat deserves attention. The duty-cycle bound (f<53%) is model-dependent for the same reason.\n\nThese are caveats, not fatal flaws. Six observations, five non-detections, and one outlier at 10^8 g/s make a strong pattern. The paper correctly notes it cannot distinguish the mini-Neptune versus water world scenarios and leaves that to future panchromatic work.\n\nAnyone working on the cosmic shoreline, helium escape, or JWST null results will get value from this. It deserves a serious referee. I would ask the authors for a more formal statistical treatment (injection-recovery or a likelihood approach) and a direct report of the transit-averaged line EW constraint, but I would accept this paper.\n\nBest.","headline":"Four independent SOSS transits make the LHS 1140 b helium claim hard to sustain—solid, cautious paper with soft statistical spots but a believable null result.","tokens_in":18375,"tokens_out":3340,"would_cite":true,"duration_ms":32506,"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":"Four JWST NIRISS/SOSS transits of LHS 1140 b show no metastable helium absorption, rejecting the reported ground-based detection of atmospheric escape.","keywords":["exoplanet atmospheres","metastable helium triplet","transmission spectroscopy","atmospheric escape","LHS 1140 b","JWST NIRISS/SOSS","non-detection","mass-loss rate"],"falsifier":"If an independent re-reduction of these four visits solves for the helium line position as a free parameter (allowing offsets beyond one pixel) or uses an empirically measured point-spread function and recovers the predicted ~600 ppm feature, the nondetection claim would collapse; a future transit showing helium at the reported depth would instead prove the signal is real but variable.","tokens_in":17352,"feed_emoji":"🔭","tokens_out":11791,"duration_ms":94445,"temperature":0.7,"pith_summary":"This paper reports four JWST NIRISS/SOSS transits of the nearby temperate planet LHS 1140 b, taken between 2023 and 2026, and finds no absorption from the metastable helium triplet at $1.0833\\,\\mu$m in any of them. The authors argue that if the 2024 ground-based detection of 1.24% helium absorption were real and persistent, JWST would have seen it, and they reject the ground-based model at $>3\\sigma$ in each visit and $9.9\\sigma$ combined. If correct, the ground-based signal is either spurious or a rare, time-variable event occurring in fewer than half of transits. This matters because LHS 1140 b is a key target for deciding whether small planets around M dwarfs retain atmospheres, and helium absorption is a direct tracer of atmospheric escape.","feed_headline":"JWST finds no helium on LHS 1140 b in four transits","feed_subtitle":"If right, the 2024 ground-based helium detection is spurious or recurs in under half of transits.","key_machinery":"The load-bearing observable is the metastable helium triplet at $1.0833\\,\\mu$m, a standard tracer of atmospheric escape, which falls within a single SOSS pixel. The analysis uses column-level light curves around the triplet, compared with the white-light curve and with neighboring columns, to maximize sensitivity to a narrow feature. To quantify the expected signal, the reported ground-based escape model is convolved with an idealized Gaussian kernel representing the SOSS point-spread function at $R\\approx700$, predicting a roughly 600 ppm feature at pixel resolution, and the predicted spectrum is tested against the data with a $\\chi^2$ statistic. The same escape framework, with all input parameters fixed to the ground-based study's values, is then used to translate nondetection depths into upper limits on mass-loss rate.","core_discovery":"Across four NIRISS/SOSS transits spanning three years, no excess absorption is seen at the metastable helium triplet in the column-level light curves or in the transmission spectra, and adjacent columns show no leaked signal. The best-fit escape model from the reported ground-based detection, convolved to SOSS resolution, is discrepant with each visit at $>3\\sigma$ and with the combined spectrum at $9.9\\sigma$. The resulting $2\\sigma$ upper limits on helium absorption correspond to mass-loss rates of roughly $5\\times10^6$ to $3\\times10^7$ g s$^{-1}$, several orders of magnitude below the reported $2\\times10^8$ g s$^{-1}$ detection. With six total observations over three years, the single 2024 detection is an outlier; if helium escape is genuinely variable, it must be active less than about half the time, with a duty cycle below 53% at $2\\sigma$.","pith_inferences":["Editorial inference: simultaneous ground-based and JWST observations of the same transit would directly test whether the 2024 detection was a real but rare escape event or a systematic artifact.","Editorial inference: the persistent nondetections indirectly favor a high mean-molecular-weight atmosphere over a hydrogen/helium envelope, since an escaping H/He envelope would normally produce a detectable helium signature.","Editorial inference: the template-convolution comparison used here could be applied to other ground-based helium detections to check their consistency with low-resolution space-based spectra.","Editorial inference: a re-analysis of the 2024 ground-based data with careful treatment of telluric and instrumental systematics would test whether the single detection could be spurious."],"forward_implications":["The 1.24% helium absorption reported from the 2024 ground-based transit is not persistent; four JWST transits reject the model at $>3\\sigma$ each and $9.9\\sigma$ combined.","If the signal is real but time-variable, it must occur in fewer than about 53% of transits at $2\\sigma$, making the escape episode rare or stochastic.","Under the adopted escape-model assumptions, mass-loss rates above roughly $10^7$ g s$^{-1}$ are excluded at $2\\sigma$, contradicting the $2\\times10^8$ g s$^{-1}$ inference from the detection.","LHS 1140 b's nature—mini-Neptune versus water world—remains open, since the helium nondetection does not discriminate between these interpretations."],"supporting_citations":[{"why":"Ground-based detection of 1.24% metastable helium absorption that this paper tests and rejects.","marker":"C. Cherubim et al. (2026)"},{"why":"Previous NIRISS/SOSS analysis of Visits 1 and 2 whose transmission spectrum and wavelength offset corrections this work reproduces and extends.","marker":"C. Cadieux et al. (2024b)"},{"why":"Provides the escape model framework used to predict the helium signal and convert absorption depths into mass-loss rates.","marker":"L. A. Dos Santos et al. (2022)"},{"why":"Supplies the primary reduction pipeline used to process all four transits at the column level.","marker":"Z. Rustamkulov et al. (2022)"},{"why":"Supplies the independent reduction pipeline used to cross-check the two newly acquired transits.","marker":"M. Radica et al. (2023)"},{"why":"Characterizes SOSS wavelength-solution visit-to-visit stability, the basis for sub-pixel uncertainty estimates.","marker":"T. Baines et al. (2023a)"},{"why":"Describes metastable helium formation and notes that very high mass-loss rates produce a decreasing signal, a caveat for interpreting nondetections.","marker":"A. Oklopčić (2019)"}],"fun_headline_variants":["No helium on LHS 1140 b in four JWST transits","JWST refutes helium detection on LHS 1140 b","Ground-based helium signal on LHS 1140 b not seen by JWST","LHS 1140 b shows no helium across four JWST observations","Four JWST transits find no helium absorption on LHS 1140 b"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion assumes the reported ground-based helium signal, if real, would land in the one pixel column the wavelength solution points to, which requires the signal's line shape, the telescope's focus, and the wavelength calibration all to be accurate to within about a pixel.","fun_headline_variants_meta":{"raw":{"variants":["No helium on LHS 1140 b in four JWST transits","JWST refutes helium detection on LHS 1140 b","Ground-based helium signal on LHS 1140 b not seen by JWST","LHS 1140 b shows no helium across four JWST observations","Four JWST transits find no helium absorption on LHS 1140 b"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000475,"raw_usage":{"total_tokens":2414,"prompt_tokens":1059,"completion_tokens":1355,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":675,"completion_tokens_details":{"reasoning_tokens":1255}},"tokens_in":675,"tokens_out":1355,"duration_ms":10250,"temperature":1.0,"reasoning_tokens":1255,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:13:58.964328+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If an independent re-reduction of these four visits solves for the helium line position as a free parameter (allowing offsets beyond one pixel) or uses an empirically measured point-spread function and recovers the predicted ~600 ppm feature, the nondetection claim would collapse; a future transit showing helium at the reported depth would instead prove the signal is real but variable.","supporting_citations":[],"review_version":1}