{"id":"441e16c6-c81f-4aea-8e44-d54ecec3636e","arxiv_id":"2412.16541","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Back-to-back transits of TRAPPIST-1 b and c allow an observed, model-free correction of stellar contamination in planet c's spectrum, reducing structured noise 2.5x below 2 micrometers.","lead":"Using two almost simultaneous transits of TRAPPIST-1 planets b and c seen by JWST, astronomers corrected the spectrum of planet c for the mottled surface of its star, cutting stellar noise by 2.5 times at short wavelengths. The technique could let scientists combine many observations to hunt for thin atmospheres on nearby rocky worlds.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Ratio correction assumes b and c see identical stellar contamination, but measured b–c differences (blue-slop residual; 110 ppm redward) are comparable to the 60–250 ppm target signals, making the 2–5 µm and stacking claims extrapolations.","rationale":"The reader's weakest assumption identifies the same load-bearing premise: the correction assumes identical stellar contamination for both planets and a featureless reference planet. My stress-test refines this into a sharper, empirically grounded concern. The paper's own results quantify the violation: a residual blue-slope difference and a 110 ppm (1.2σ) consistency floor redward of 2 µm. These residuals are not negligible compared with the 60–250 ppm secondary-atmosphere signals that motivate the method. The ratio correction removes any common-mode component, so the observed blue-wavelength noise reduction does not uniquely validate a stellar-contamination correction; it could equally reflect removal of shared instrumental systematics or of b's atmospheric signal. The 2.5× Allan-variance improvement is measured only where SNR is high, and the paper explicitly notes that the red-wavelength correction cannot be directly assessed from this single transit. The extrapolation to 2–5 µm, where the CO2 feature lies, is therefore the most load-bearing and least supported link in the argument. I agree with the reader that the verdict should remain conditional: the proof-of-concept at shorter wavelengths is credible, but the broader claim for enabling stacked searches of secondary atmospheres requires either additional transits or an explicit propagation of the b–c systematic difference into the final spectra. The injection-recovery test I propose directly quantifies how the observed b–c differences bias recovered atmospheric features, which would settle whether the method's central promise holds.","tokens_in":14842,"tokens_out":4609,"duration_ms":45221,"concrete_test":"Perform an injection-recovery test using the public MAST data (10.17909/4zcj-k904): inject a synthetic 100 ppm CO2 feature at 4.3 µm and a 100 ppm flat signal into the c transit light curve, then apply the full b/c ratio-correction pipeline exactly as in Section 3.2. Repeat over the five out-of-transit segments S1–S5, or bootstrap-resample epochs with the observed stellar surface-coverage variations, and measure the recovered feature amplitude and residual scatter in the 2–5 µm range. If the recovered amplitude is biased by more than ~20 ppm, or if the residual red noise exceeds the white-noise expectation, the method cannot support 60–250 ppm atmospheric searches by stacking.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central assumption of the correction—that stellar contamination is identical for the two transit chords and that planet b is featureless—is empirically violated at a level comparable to the signals the method aims to detect. Section 3.2 reports a residual blue-slope difference between b and c, attributed to evolving surface coverage or chord differences, and states that redward of 2 µm the two transit spectra agree only to within 110 ppm (1.2σ). Because the ratio correction divides c by b, any wavelength-dependent difference between the two chords—whether stellar, instrumental, or planetary—propagates directly into the corrected c spectrum. This systematic is epoch-correlated if it arises from spot evolution or chord geometry, so it will not average down as white noise when stacking epochs. The demonstrated 2.5× Allan-variance improvement at 0.8–2 µm is therefore not evidence that contamination is reduced in the 2–5 µm region; indeed, the bottom-right panel of Figure 3 shows the post-correction unbinned variance is higher there due to added photon noise from b. The paper's longer-wavelength benefit rests on an extrapolation from the blue, where SNR is high, to the red, where the target CO2 feature at 4.3 µm resides. With a measured ~110 ppm systematic floor between b and c, the claim that the method enables reliable searches for secondary atmospheres at 60–250 ppm is not established by this single epoch.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents JWST/NIRSpec PRISM observations of quasi-simultaneous transits of TRAPPIST-1 b and c on July 9, 2024, and proposes using the transit spectrum of TRAPPIST-1 b as an epoch-specific, empirical stellar-contamination reference to correct the transit spectrum of TRAPPIST-1 c. The authors report that the ratio correction removes the 1.4 and 1.8 µm water features, reduces the blue slope, and lowers the Allan-variance noise floor in the 0.8–2.0 µm range from about 200 ppm to about 80 ppm, a factor of 2.5. They also analyze out-of-transit spectra in five time segments and infer a two-component stellar photosphere with cool (~2000 K) and warm (~2600 K) regions whose covering fractions vary by about 0.1% per hour. Based on the short-wavelength success and the out-of-transit variability, the paper argues that contamination at 2–5 µm is likewise reduced, enabling co-addition of multiple epochs to search for secondary atmospheres at 60–250 ppm. The paper is framed as a proof-of-concept and repeatedly acknowledges the low SNR at longer wavelengths, but the abstract and conclusions go beyond the directly measured result.","tokens_in":15148,"tokens_out":6168,"duration_ms":84847,"significance":"If the method is validated, it would provide a valuable empirical route to stellar-contamination correction in multi-planet systems with an airless reference planet, directly addressing a known limitation of model-based corrections for late M dwarfs. The observation was strategically scheduled to test a specific prediction of the TRAPPIST-1 JWST Community Initiative, and the paper includes a custom independent reduction pipeline, time-resolved out-of-transit stellar analysis, and an honest discussion of residual slopes and model uncertainties. The short-wavelength demonstration—removal of the 1.4/1.8 µm features and a factor-of-2.5 reduction in the 0.8–2.0 µm Allan-variance floor—is a concrete, quantitative result. However, the broader claims about 2–5 µm correction and the ability to stack spectra to detect 60–250 ppm signals are not established by the single epoch analyzed here; they are extrapolations that need to be reframed or supported by additional propagation of systematics. The paper is a promising proof-of-concept, but the strength of the conclusions currently exceeds the strength of the evidence.","major_comments":[{"comment":"The long-wavelength benefit is asserted rather than measured. The paper states that in the 2–5 µm range the variance of the unbinned residuals is higher post-correction due to added photon noise from planet b, and that the lower SNR prevents clear detection of contamination or full assessment of mitigation. Despite this, the Abstract claims that contamination is reduced at longer wavelengths 'to a similar extent' and the Conclusions state that the correction 'reduces the challenge to primarily white noise' for atmospheric searches. The 2.5× reduction is demonstrated only in the 0.8–2.0 µm range; the reddest portion of the spectrum is consistent with white noise before correction, so no red contamination reduction is actually measured. The paper should present the 2–5 µm correction as an expectation supported by the short-wavelength analogy and the out-of-transit variability, not as a demonstrated result, and should quantify the sensitivity that can be claimed from a stacked epoch series given a redward systematic floor of order 110 ppm (1.2σ) rather than assuming it averages down.","section":"§3.2, Fig. 3 (bottom right); Abstract; §5"},{"comment":"The central correction assumes that b and c see identical stellar contamination and that b is featureless. These assumptions are only approximately met: the paper reports a residual blue-slope difference between the two planets, attributes it to evolving surface coverage or slightly different transit chords, and acknowledges in §4.1 that some atmospheric scenarios for b and c remain viable. Because the corrected c spectrum is formed by dividing the c spectrum by the b spectrum, any wavelength-dependent feature intrinsic to b—whether atmospheric or a chord-dependent stellar signal—enters the corrected c spectrum with sign reversed and is not removable by stacking. The 110 ppm (1.2σ) redward consistency is not an upper limit on such a systematic; at 1.2σ the true difference could be comparable to the 60–250 ppm target signals. The paper should propagate the uncertainty in the airless and identical-chord assumptions into the corrected spectrum, for example by injecting plausible b atmospheric models and spot configurations consistent with the Figure 4 constraint f1,chord = 0.495, and should explicitly state the residual systematic floor this places on the method.","section":"§3.2, §1, §4.1, Fig. 4"},{"comment":"The out-of-transit inference that TRAPPIST-1 has a two-component photosphere with covering fractions varying by ~0.1% per hour is used as part of the argument that contamination extends to red wavelengths. This inference comes from fitting SPHINX models to flux-calibrated spectra with no uncertainty inflation, and the best-fit cool component is pinned at the lower grid boundary of 2000 K while the model effective temperature (2324 K) is below the literature value (2566±26 K). The paper is appropriately cautious in §4.2 that these are not definitive, but the abstract presents the 0.1% per hour variability as a finding. Because the red-contamination argument depends on the time variability of the stellar-surface model, the model dependence and boundary effects should be quantified or the claim should be moved into the interpretation section rather than the abstract.","section":"§3.1.2, Fig. 2, §4.2"}],"minor_comments":[{"comment":"'Same transit cord' should read 'same transit chord'.","section":"§1"},{"comment":"The inclination posterior for planet c is reported as '89.83761834+0.09' with far more significant digits than the uncertainty; round to a consistent precision.","section":"Table 1"},{"comment":"The label 'Width of 4.3 µm CO2 feature' in the bottom-right panel is unclear; the caption should state the bin size or spectral width used for the Allan-variance comparison.","section":"Fig. 3 caption"},{"comment":"The statement that the two transit spectra are consistent 'to within 110 ppm (1.2σ)' should clarify whether 110 ppm is the 1σ uncertainty or the maximum difference, and should note that a 1.2σ consistency does not set a tight upper limit on a possible systematic.","section":"§3.2"},{"comment":"The symbols f1,disk, f2,disk, f1,chord, and f2,chord are defined in the text but not in the equation; adding an explicit definition after Eq. (1) would improve readability.","section":"§4.2, Eq. (1)"},{"comment":"The method is described as 'model-independent' in the abstract, but the long-wavelength discussion in §4 uses a best-fit stellar contamination model to estimate contamination amplitude. I suggest using 'empirical' or 'model-light' to describe the ratio step and reserving 'model-independent' for the short-wavelength correction only.","section":"Abstract and §5"}],"recommendation":"major_revision","confidential_remarks":"The short-wavelength proof-of-concept is solid and likely publishable after revision, but the abstract and conclusions overstate the long-wavelength and stacking results. The authors already identify most of the limitations in the text; the revision should simply align the claims with the measured data and add a quantitative propagation of the 110 ppm redward systematic and the airless/reference-planet assumptions. I would not require new observations for this revision, but I would require the paper to stop claiming demonstrated red-wavelength correction and 60–250 ppm detectability without a much more explicit accounting of the systematic floor."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this if you care about JWST transmission spectroscopy of M dwarfs. The paper is the first dedicated demonstration of the TJCI2024 idea: use a quasi-simultaneous transit of an airless planet (TRAPPIST-1 b) to divide out stellar contamination from a neighboring planet's spectrum (c). That is the genuinely new part, and the short-wavelength result is credible. The observed b and c spectra share the 1.4 and 1.8 µm water-like features and the blue slope; after the ratio correction those features disappear and the Allan variance floor drops from ~200 to ~80 ppm at 0.8–2 µm. That is a measured improvement, not a fitted one. The paper also does something useful out of transit: time-resolved S1–S5 spectra show the covering fraction of the cool component changing by ~0.1% per hour, which sets a floor on how identical two transit chords can be.\n\nThe soft spots are exactly where the paper hedges. Redward of 2 µm, the two transit spectra agree only to 110 ppm (1.2σ) in 0.1 µm bins, and the post-correction unbinned variance is higher there because b adds photon noise. The paper argues that the same physics must reduce contamination at long wavelengths, and says the correction should bring a ~200 ppm smooth feature down to ~80 ppm. That is an extrapolation from the blue where SNR is high, and 110 ppm is the same order as the 60–250 ppm signals the method is meant to enable. So the claim that this method makes secondary-atmosphere searches feasible at 4.3 µm is not demonstrated by this single epoch. Also, the entire correction hinges on b being featureless; the paper cites recent upper limits, and the residual blue-slope difference between b and c is a reminder that the chords are not identical.\n\nThe stellar heterogeneity inferences are honest but model-limited: the cool component sits at the 2000 K grid boundary, and the uncertainties are not inflated for model systematic, so take the exact covering fractions with salt. The custom pipeline Frida is described but not released, which is a real reproducibility gap.\n\nWho is this for? Anyone planning multi-transit JWST programs on TRAPPIST-1 or similar systems. The proof-of-concept is worth taking seriously; the quantitative claims about long-wavelength sensitivity need more transits. I would send it to referees—specialists can check the light-curve fitting and decide quickly whether the extrapolation is acceptable.","headline":"First real test of the back-to-back transit correction idea; the short-wavelength proof works, but the long-wavelength and stacking claims rest on an extrapolation that a single epoch cannot support.","tokens_in":15780,"tokens_out":2390,"would_cite":true,"duration_ms":20314,"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":"Back-to-back transits strip stellar contamination from TRAPPIST-1 c, cutting the structured noise by a factor of 2.5 below 2 µm.","keywords":["stellar contamination","transit spectroscopy","TRAPPIST-1","JWST NIRSpec PRISM","multi-planet systems","starspots","secondary atmospheres","transmission spectra"],"falsifier":"Take a longer-duration JWST observation of a back-to-back TRAPPIST-1 b/c pair with enough signal-to-noise to see contamination at 4–5 µm: if the corrected c spectrum still contains structured noise above about 80 ppm there, or if the same correction applied to b with c as the reference produces spectral features, then the identical-contamination and airless-anchor assumptions fail.","tokens_in":14623,"feed_emoji":"🪐","tokens_out":10115,"duration_ms":81077,"temperature":0.7,"pith_summary":"This paper claims a model-independent way to remove stellar contamination from exoplanet transit spectra: use the quasi-simultaneous transit of an airless companion planet as an epoch-specific reference. Applying the idea to JWST NIRSpec PRISM observations of TRAPPIST-1 b and c taken on July 9, 2024, the authors divide c's transit spectrum by b's, erasing the water-absorption features at 1.4 and 1.8 µm and flattening most of the blue slope while reducing the noise floor from about 200 ppm to about 80 ppm below 2 µm, a factor of 2.5. If correct, this shifts the obstacle to atmospheric characterization from stellar-model-limited red noise to white noise that can be beaten down by stacking transits, making secondary-atmosphere signals of 60–250 ppm accessible. The same data also constrain the star's surface, pointing to cold (about 2000 K) and warm (about 2600 K) regions that are well mixed across the disk and change their covering fractions by about 0.1% per hour.","feed_headline":"Back-to-back transits strip stellar contamination from TRAPPIST-1 c","feed_subtitle":"Using one airless planet to correct another shifts the challenge to white noise, opening 60–250 ppm atmosphere searches.","key_machinery":"The load-bearing object is the ratio of the two quasi-simultaneous transit spectra, $R(\\lambda)=T_c(\\lambda)/T_b(\\lambda)$, rescaled by the weighted mean transit depth of TRAPPIST-1 b over 2–5 µm. Because the two planets have nearly equal radii and impact parameters and transit within hours of each other, the stellar contamination $\\epsilon(\\lambda)$ is assumed to be common to both spectra; forming the ratio cancels it, leaving the planet-to-planet spectral ratio, and multiplying by b's mean depth restores the corrected transit depth of c. Supporting this are a joint white-light-curve fit with shared quadratic limb-darkening coefficients, a linear trend, and a Gaussian Process for correlated noise; wavelength-bin fits that fix the common parameters; and an out-of-transit modeling pipeline that reconstructs the stellar photosphere from one to four spectral components using SPHINX model spectra with 2MASS photometry as a constraint. Allan variance plots, which track residual scatter as a function of bin size, are then used to quantify whether the structured red noise has been converted into white noise.","core_discovery":"The paper's central discovery is that back-to-back transits of TRAPPIST-1 b and c carry essentially the same stellar contamination, so b's spectrum can be used as a template to correct c's without any stellar model. After the ratio correction, the 1.4 and 1.8 µm water features disappear and the blue-end slope is roughly halved; the residual-scatter analysis shows the structured red noise in the 0.8–2.0 µm range drops by a factor of 2.5, from roughly 200 ppm to 80 ppm at a bin size of ten. At 2–5 µm the single-transit signal-to-noise is too low to measure contamination directly, but the out-of-transit spectral variability across the whole wavelength range and the success of the blue-end correction lead the authors to argue that redder contamination is mitigated to a similar degree, down to about 80 ppm. They also use the shape of the contamination to infer that TRAPPIST-1's photosphere is well mixed between about 2000 K and about 2600 K components, with the cooler component slightly underrepresented in the transit chord, so the long-wavelength contamination amplitude is roughly 280 ppm rather than the 1740 ppm expected if all inhomogeneity lay outside the chord.","pith_inferences":["Inference: the same ratio technique could be transplanted to other compact multi-planet systems around active M dwarfs, provided at least one planet is confidently airless and the two transit chords are close; systems with larger impact-parameter differences would need a chord-aware correction factor.","Inference: the airless-anchor assumption is testable with this very dataset in reverse, because using c as the reference to correct b should produce a flat spectrum; if it instead shows spectral features, those would betray a thin atmosphere on c rather than stellar contamination.","Inference: the out-of-transit coverage evolution implies that contamination within a single transit may not be strictly constant, so modeling a time-dependent contamination term in the ratio could account for the residual blue slope that the authors attribute to stellar surface changes."],"forward_implications":["Stellar contamination no longer needs a definitive model of the host star: any multi-planet system with an airless reference planet observed in back-to-back transits can get an epoch-specific correction from the data itself.","Stacking corrected transit spectra across epochs becomes statistically meaningful because the residual noise is approximately white, which the paper says opens the way to searching for 60–250 ppm secondary-atmosphere features such as CO2 at 4.3 µm.","TRAPPIST-1 b and c become interchangeable reference planets, roughly doubling the scheduling windows for multi-transit observations of the system.","Long-wavelength contamination is expected to be broad and smooth, about 80 ppm after correction, so it should not mask sharp planetary spectral features like CO2.","The observed roughly 0.1%-per-hour evolution of the cold and warm surface coverage means a correction taken from a different epoch would be inaccurate, reinforcing the need for quasi-simultaneous reference transits."],"supporting_citations":[{"why":"It proposes the quasi-simultaneous airless-planet transit method and frames it as the alternative to stellar models that this paper tests.","marker":"TJCI2024"},{"why":"It supplies the TRAPPIST-1 system parameters, including the similar radii and impact parameters of b and c that justify the identical-contamination assumption, and the priors used in the light-curve fits.","marker":"Agol et al. 2021"},{"why":"It provides JWST secondary-eclipse evidence that TRAPPIST-1 b is likely airless, supporting its use as the reference planet.","marker":"Greene et al. 2023"},{"why":"It provides the corresponding evidence that TRAPPIST-1 c is likely airless or has a thin atmosphere, supporting the anchor-planet assumption.","marker":"Zieba et al. 2023"},{"why":"It introduces the multi-component stellar photosphere modeling method used for the out-of-transit spectral analysis.","marker":"Rackham & de Wit 2024"},{"why":"It establishes the standard stellar contamination formalism and the fully unocculted inhomogeneity baseline against which the well-mixed photosphere is compared.","marker":"Rackham et al. 2018"},{"why":"It provides the SPHINX model spectra grid from which the stellar component spectra are interpolated.","marker":"Iyer et al. 2023"},{"why":"It supplies the reference secondary-atmosphere model templates and the 60–250 ppm expected signal amplitudes used to frame the atmospheric search.","marker":"Lincowski et al. 2018"}],"fun_headline_variants":["Back-to-back TRAPPIST-1 transits cut stellar noise 2.5x","Airless b helps strip contamination from c's spectrum","JWST sees TRAPPIST-1 c clearer using b as template","Twin transits reveal star's patchy surface and clean spectra","TRAPPIST-1 c spectrum cleaned by sibling transit"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The correction holds only if TRAPPIST-1 b is truly airless and both planets occult the same distribution of stellar surface patches, so that any difference between the two transit spectra is stellar rather than planetary.","fun_headline_variants_meta":{"raw":{"variants":["Back-to-back TRAPPIST-1 transits cut stellar noise 2.5x","Airless b helps strip contamination from c's spectrum","JWST sees TRAPPIST-1 c clearer using b as template","Twin transits reveal star's patchy surface and clean spectra","TRAPPIST-1 c spectrum cleaned by sibling transit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000875,"raw_usage":{"total_tokens":3891,"prompt_tokens":1158,"completion_tokens":2733,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":774,"completion_tokens_details":{"reasoning_tokens":2651}},"tokens_in":774,"tokens_out":2733,"duration_ms":16645,"temperature":1.0,"reasoning_tokens":2651,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:29:06.887663+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a longer-duration JWST observation of a back-to-back TRAPPIST-1 b/c pair with enough signal-to-noise to see contamination at 4–5 µm: if the corrected c spectrum still contains structured noise above about 80 ppm there, or if the same correction applied to b with c as the reference produces spectral features, then the identical-contamination and airless-anchor assumptions fail.","supporting_citations":[],"review_version":1}