{"id":"7db3a543-42bd-4969-b8a3-c65b234938e4","arxiv_id":"2506.12140","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"No TRAPPIST-1 planet shows a Ly-alpha transit deeper than about 20%, and the noise limiting the search is traced to frequent ~10^29-erg microflares from the star.","lead":"Hubble spent 104 orbits over five years watching the seven TRAPPIST-1 planets pass in front of their star. It found no large hydrogen exospheres, but it did find that the star itself erupts in frequent small flares that can hide such signals.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The load-bearing step is the detection/attribution of microflares: STIS TIME-TAG candidates are selected with an unspecified significance threshold and no false-alarm analysis, while VLT corroboration is by-eye; if these excursions are artifacts, the claimed flare-driven noise floor and escape…","rationale":"The paper has real merit: 104 HST orbits, a multi-year baseline, a plausible rotation-period recovery, and an internally consistent non-detection analysis. However, the headline advance is the microflare interpretation, and the evidence for it rests on two under-specified pillars: (1) TIME-TAG count-rate excursions that are called 'flare candidates' without a defined detection statistic or false-alarm calibration; and (2) VLT g-band light curves whose flare candidates are identified by eye. The reader's weakest assumption (Section 3.4's waiting-time equivalence) is a real problem for the temperature and filling-factor numbers, but those numbers are secondary: even if Tfl and Xfl were discarded, the central claim that frequent microflares set the Ly-alpha detection limit could still stand if the count-rate excursions are real. The reverse is not true: if the excursions are artifacts or Poisson fluctuations, no amount of waiting-time modeling rescues the central claim. The paper itself uses hedged language ('flare-like events', 'candidates') in Section 3.2 while the abstract asserts 'we report flares', so the gap between candidate identification and claimed detection is the load-bearing soft spot. A reanalysis with a calibrated flare search and false-alarm reporting would settle it. Because the HST/STIS time-tag data are public and the FORS2 data appear reducible, this check is feasible. The reader's conditional verdict remains appropriate; I do not see grounds to reject the paper, only to require this check before the flare-existence claim is taken at face value.","tokens_in":15027,"tokens_out":7147,"duration_ms":82989,"concrete_test":"Re-run the STIS TIME-TAG analysis of the eight high-SNR visits with a pre-specified flare search (e.g., Bayesian blocks or a matched filter on 5-minute binned photon streams), calibrate the detection threshold by injecting artificial flare pulses into the airglow/background-only count-rate stream and requiring a false-alarm rate below one per visit, and publish the candidate list with per-event significance and recovered amplitudes. If fewer than one genuine >=300% Ly-alpha event per orbit survives this calibrated search, the microflare attribution in Section 5 is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.2 identifies 'flare-like events' and 'flare candidates' using TIME-TAG data, but the detection criterion is not specified: 'significantly exceeded the baseline trend' has no threshold, no signal-to-noise requirement, no false-alarm rate, and no candidate table. The only rejection of systematics is the absence of a concurrent airglow spike and an orbit-integrated correlation coefficient <0.5 with airglow/background; that does not test time-resolved PSF drift, wavelength-solution jumps, or detector artifacts on 5-minute timescales. Consequently, the quantitative claim in Section 5 that the baseline scatter is 'a manifestation of intrinsic stellar variability, consistent with frequent, short-duration microflares'—and the abstract's 'we report ~1e29-erg flares'—is not yet distinguished from a handful of unexplained count-rate excursions. Section 3.3's VLT/FORS2 'independent corroboration' is weaker than stated: the 1-2 flares/hour rate comes from four nights, with flare-like structures 'identified by eye' (Figure 4 caption), no automated detection or significance, and the g'-band amplitudes range 1-50% rather than clustering near the paired 4% value. Since the observations are not simultaneous, matching waiting-time distributions is the only link between the Ly-alpha and g' events, and this same link is what the Section 3.4 temperature/filling-factor inversion depends on. If the TIME-TAG excursions are not confirmed as stellar flares, the conclusion that the Ly-alpha exosphere detection limit is set by microflares (rather than by reduction systematics or unmodeled variability) loses its foundation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents 104 orbits of HST/STIS G140M Ly-α monitoring of TRAPPIST-1 from 2017 to 2022, with 3–6 transits per planet. Using PSF fitting, the authors derive Ly-α light curves, rule out transit depths ≳20% at 5σ, and convert the non-detections into escape-rate upper limits, reporting 1064 Earth-ocean-equivalent hydrogen per Gyr for planet b under an assumed XUV cross-section. The paper further attributes the excess baseline scatter to frequent microflares, reporting sub-hour Ly-α count-rate enhancements in STIS TIME-TAG data and 1–2 flares/hour in VLT/FORS2 g-band photometry. From paired Ly-α and optical amplitudes it derives a flare temperature of 11000 K and a filling factor of 0.011%, and it reports a 3.27 ± 0.04 day rotation period consistent with the known 3.295-day period.","tokens_in":15390,"tokens_out":3369,"duration_ms":152194,"significance":"If the microflare interpretation holds, this is an important result for UV exosphere searches: stellar microvariability, not photon noise, sets the practical detection limit for Ly-α transits around active M dwarfs. The underlying dataset is substantial and publicly available, and the transit non-detections and rotation-period recovery are valuable independent constraints: the rotation period is checked against the external K2 value and passes an injection–recovery test, and the 20% transit-depth limits are directly supported by the light curves. The flare temperature and filling factor, however, rest on a chain of assumptions that the manuscript does not yet quantitatively justify, so the significance of the microflare characterization is currently conditional.","major_comments":[{"comment":"The TIME-TAG flare detection criterion is never specified. The text states that candidates are epochs where the count rate 'significantly exceeded the baseline trend,' but no threshold, signal-to-noise requirement, false-alarm rate, or candidate table is provided. The only systematic check is the absence of a concurrent airglow spike and an orbit-integrated correlation coefficient <0.5 with background/airglow; that check does not address time-resolved PSF drift, wavelength-solution jumps, or detector artifacts on 5-minute timescales. As written, the Section 5 claim that the excess scatter is 'a manifestation of intrinsic stellar variability, consistent with frequent, short-duration microflares' is not distinguished from a small number of unexplained count-rate excursions. Please define the detection threshold, compute the false-alarm rate from scrambled or airglow-referenced time series, and provide a table of candidates with their significances.","section":"Section 3.2"},{"comment":"The VLT/FORS2 corroboration is weaker than stated. Flare-like structures are 'identified by eye' in the Figure 4 caption, with no automated detection, no significance threshold, and no completeness estimate. The 1–2 flares/hour rate comes from only four nights, and the reported g'-band amplitudes range from 1% to 50%, yet the temperature inversion in Section 3.4 pairs a ~4% optical event with a ~400% Ly-α event. If the optical excursions are not robustly detected, or if the 4% event is not representative, the independent-corroboration argument loses its quantitative force. Please add a reproducible detection algorithm for the optical data, report per-event amplitudes and durations, and quantify how the spread in optical amplitudes affects the pairing used in Section 3.4.","section":"Section 3.3"},{"comment":"The flare temperature and filling factor depend on the assumption that Ly-α and optical flares with similar waiting times share the same emission mechanism, citing Loyd et al. (2018). This assumption is load-bearing because it justifies pairing a ~400% Ly-α enhancement with a ~4% g'-band enhancement to solve a one-component blackbody. If the paired events belong to different flare populations, or if the 4% optical amplitude is not representative given the observed 1–50% range, then Tfl = 11000 K and Xfl = 0.011% are not supported. The manuscript should either provide a sensitivity analysis showing how Tfl and Xfl vary when the pairings and the continuum-to-Ly-α ratio (0.03+0.14–0.02) are changed, or clearly present these numbers as an illustrative estimate rather than a measured flare property. The current abstract presents them as a headline result without this caveat.","section":"Section 3.4"},{"comment":"The abstract quotes 1064 EO_H/Gyr for planet b as the headline upper limit, but the text shows that this value corresponds to (R_XUV/R_star)^2 = 0.20, while the fiducial adopted value of 0.10 yields 533 EO_H/Gyr. The paper itself notes that (R_XUV/R_star)^2 = 0.10 is an upper limit and that smaller cross-sections would reduce the loss rate significantly. The abstract and conclusions should therefore state the range explicitly and distinguish the assumption-dependent escape-rate estimate from the measured transit-depth upper limit. As written, quoting only the doubled value overstates the robustness of the escape-rate constraint.","section":"Section 2.5 and Abstract"}],"minor_comments":[{"comment":"The sentence 'We investigate its origin in Section 4' appears to be a cross-reference error: the variability origin is discussed in Section 3, not Section 4.","section":"Section 2.4"},{"comment":"Table 1 contains apparent anomalies: several visits appear twice (e.g., odhs33, odhs31, odhsd1) with different mean-flux values for the same visit, and some entries (e.g., '1.55-11', '1.50e-10') look like typos in the flux column. Please correct the table and ensure each visit appears exactly once with one consistent flux value.","section":"Table 1"},{"comment":"The filter is described as 'gHIGH (spectral range: 388–548 nm)' in Section 3.3, while the abstract, Section 3.4, and Figure 4 refer to g' or g-band observations. Please unify the filter nomenclature and specify whether the photometric band is the standard g' filter or the gHIGH filter.","section":"Section 3.3"},{"comment":"The text states that the filling factor 'aligned with values reported for TRAPPIST-2 flares' — this is likely a typo for TRAPPIST-1, since no TRAPPIST-2 system is otherwise referenced and the cited works concern TRAPPIST-1.","section":"Section 3.4"},{"comment":"The Figure 4 caption says 'A flare of ~50% amplitude is visible (trimmed here)' but the plotted light curve shows six orange-shaded features. Please clarify which feature is the 50% flare and whether the flare shown was trimmed to preserve the transit window.","section":"Figure 4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the scope of the journal and the core transit non-detection is solid, but the microflare detection and characterization sections need a substantial revision before the central claims can be accepted. The authors should be encouraged to add a quantitative flare-detection pipeline (threshold, false-alarm rate, candidate table) and to temper the temperature/filling-factor claims or clearly label them as conditional on the adopted pairing assumption."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a substantial dataset paper. 104 orbits of STIS Ly-alpha monitoring across all seven TRAPPIST-1 planets is genuinely new, and the core non-detection—no transit depths above ~20% at 5 sigma—looks well supported by the PSF-fitting analysis. The 3.27 ± 0.04 day periodicity matching the known 3.295-day rotation is also credible: the injection-recovery test is the right check and it passes. The excess scatter relative to photon noise is real, and the paper makes a plausible case that something stellar is driving it. That is the part I would trust: there is more going on in the Ly-alpha baseline than counting statistics, and two instruments show flare-like events at a similar rate.\n\nThe soft spots are mostly about how far the microflare story is pushed. Section 3.2's candidate selection is never quantified: 'significantly exceeded the baseline trend' has no significance threshold, no false-alarm rate, and no candidate table. The VLT corroboration is by-eye and non-simultaneous, so the waiting-time matching is the only link between the Ly-alpha and g'-band events. That matters because the 11000 K / 0.011% flare temperature and filling factor in Section 3.4 are derived from pairing a ~400% Ly-alpha enhancement with a ~4% g' event on the strength of that assumption. If the paired events are from different populations, the temperature is not supported. This is the weakest load-bearing step in the paper. Also, the abstract quotes 1064 EOH/Gyr for planet b, which is the (R_XUV/Rstar)^2 = 0.20 scenario; the text itself treats 0.10 as an upper limit and gives 533 for that case. Quoting the doubled number without flagging it overstates the escape-rate constraint. Finally, Table 1 contains duplicated visit rows with inconsistent mean flux values; that is a data-presentation error that needs fixing.\n\nNone of this sinks the central results. The non-detection and rotation period are likely correct and will be cited. The microflare detection is suggestive but not yet a quantitative flare frequency distribution; it needs a proper detection algorithm and a candidate list before the 10^29 erg rates are taken at face value. The paper deserves a serious referee: it is a large, hard-won dataset and the questions it raises about stellar microvariability as the noise floor for Ly-alpha exosphere searches are important for the exoplanet atmosphere community. I would send it out, with the expectation of moderate revision: tighten the flare detection statistics, re-frame the abstract's escape-rate number, and clean up the table.","headline":"Solid non-detection and rotation period, but the microflare statistics and the abstract's escape-rate limit need tightening before the quantitative claims stand.","tokens_in":16028,"tokens_out":2871,"would_cite":true,"duration_ms":33467,"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":"A five-year Hubble campaign finds no TRAPPIST-1 exospheres, and identifies the noise floor as frequent microflares.","keywords":["TRAPPIST-1","Lyman-alpha transits","exospheres","stellar microflares","M dwarf activity","HST/STIS","atmospheric escape","stellar rotation period"],"falsifier":"Simultaneous HST/STIS Lyman-$\\alpha$ and VLT g'-band monitoring of TRAPPIST-1 would settle it: if the 300-500 percent Lyman-$\\alpha$ bursts do not coincide with the few-percent g'-band brightenings predicted by an $11000$ K, $0.011\\%$ filling-factor blackbody, the inferred microflare properties are wrong.","tokens_in":14828,"feed_emoji":"✨","tokens_out":10464,"duration_ms":115398,"temperature":0.7,"pith_summary":"This paper tries to establish that TRAPPIST-1's own ultraviolet flickering, not photon noise, is what limits searches for exospheres around its seven planets. Across 104 Hubble orbits spanning five years, no planet shows a statistically significant Lyman-$\\alpha$ transit; the campaign rules out transit depths above about 20 percent and translates that into an upper limit near 1064 Earth-ocean-equivalent hydrogen per gigayear for planet b, consistent with an airless planet b. The excess scatter in the light curves is traced to frequent microflares of about $10^{29}$ erg: STIS time-tagged data show sub-hour Lyman-$\\alpha$ enhancements of 300-500 percent, and independent VLT g-band photometry shows one to two flares per hour. If correct, the finding changes how null exosphere searches around active M dwarfs should be read: the detection floor is the star, and thin atmospheres cannot yet be ruled out.","feed_headline":"Microflares, not noise, limit TRAPPIST-1 exosphere searches","feed_subtitle":"No Lyman-alpha transit appears around any of seven planets; frequent microflares set the detection floor.","key_machinery":"The load-bearing tool is the TIME-TAG mode of STIS, which records each detected photon's arrival time; it lets the authors look inside the $\\sim 45$-minute orbit and catch count-rate spikes that orbit-averaged photometry would smear out. They use the bright geocoronal airglow as a simultaneous reference to reject instrumental artifacts. The second mechanism is a waiting-time pairing rule taken from prior flare studies: flares of similar cadence are assumed to share an emission mechanism, so a Lyman-$\\alpha$ event and a $g'$-band event of the same waiting time can be combined and fit by a one-component blackbody, yielding the flare temperature and filling factor. The third mechanism is a Lomb-Scargle periodogram applied to the multi-year, orbit-averaged Lyman-$\\alpha$ fluxes, which recovers the stellar rotation period.","core_discovery":"On the paper's own terms, TRAPPIST-1's Lyman-$\\alpha$ line is so intrinsically variable that the star sets the measurement floor. In 104 HST/STIS orbits (24 visits) covering three to five transits per planet, none of the seven planets shows a statistically significant transit; the data exclude depths $\\gtrsim 20\\%$ at $5\\sigma$, place $2\\sigma$ upper limits near $30\\%$, and give an escape-rate upper limit of about $1064\\ \\mathrm{EO_H}/\\mathrm{Gyr}$ for planet b under the adopted XUV cross-section. The out-of-transit scatter is three to five times the photon-noise expectation, and STIS TIME-TAG data resolve it into count-rate increases of 300-500 percent on sub-hour timescales, about three per high-signal visit. VLT/FORS2 $g'$-band photometry independently shows one to two flares per hour with amplitudes up to 50 percent. Pairing a $\\sim 400\\%$ Lyman-$\\alpha$ enhancement with a $\\sim 4\\%$ $g'$-band enhancement of similar waiting time, the paper solves a one-component blackbody to obtain a flare temperature of $11000^{+4200}_{-3100}$ K and a filling factor of $0.011^{+0.03}_{-0.01}\\%$ of the stellar disk. A periodogram of the multi-year baseline gives $P = 3.27 \\pm 0.04$ days, matching the previously reported 3.295-day rotation period.","pith_inferences":["If microflares are as frequent as reported, the time-averaged XUV irradiation of the TRAPPIST-1 planets may exceed the quiescent value used in the escape model, making the derived water-loss upper limits conservative in the sense that true losses could be higher.","The same TIME-TAG analysis could be applied to archival STIS observations of other M dwarfs; the paper's method implies that the noise floor for exosphere searches is set by stellar microvariability rather than detector performance.","A direct test of the 11000 K inference would be simultaneous far-ultraviolet and g-band monitoring; if the strong Lyman-alpha bursts do not coincide with weak g-band brightenings at the expected amplitude ratio, the waiting-time pairing assumption would need revision.","If flare triggering is tied to the 3.27-day rotation, the microflare rate should vary with rotational phase, a prediction that could be checked by folding the TIME-TAG flare candidates on the recovered period."],"forward_implications":["For TRAPPIST-1, HST/STIS Lyman-alpha transit spectroscopy is now sensitivity-limited: exospheres producing transits shallower than roughly 20 percent cannot be detected, so null results do not prove the planets lack atmospheres.","The escape-rate upper limit near $1064\\ \\mathrm{EO_H}/\\mathrm{Gyr}$ for planet b supports the emerging picture that planet b is airless or has already lost its hydrogen reservoir.","TRAPPIST-1 flares extend down to about $10^{29}$ erg with sub-hour cadence, a decade below the $\\sim 10^{30}$-erg flares seen every $\\sim 6$ hours by JWST, so microflares are a persistent part of this star's ultraviolet environment.","The inferred flare temperature near $11000$ K and filling factor of about $0.011\\%$ place the microflares inside the usual range for solar and stellar flares.","The $3.27 \\pm 0.04$ day periodicity in Lyman-alpha flux independently confirms the 3.295-day rotation period, implying that ultraviolet activity is rotationally modulated."],"supporting_citations":[{"why":"Supplies the earlier Ly-alpha flux variability and water-loss history estimates that this campaign extends.","marker":"Bourrier et al. (2017a)"},{"why":"Reported marginal Ly-alpha transit detections for planets b and c, the prior claim this larger campaign tests and does not confirm.","marker":"Bourrier, V. et al. (2017)"},{"why":"Discovery paper that defines the TRAPPIST-1 planets and their transit geometry.","marker":"Gillon et al. (2017)"},{"why":"Provides the FUV continuum-to-Ly-alpha ratio and the waiting-time pairing assumption used to infer flare temperature and filling factor.","marker":"Loyd et al. (2018)"},{"why":"Gives the escape efficiency factors used to convert XUV luminosity into water-loss upper limits.","marker":"Bolmont et al. (2017)"},{"why":"Supplies the energy-limited escape equation used for the mass-loss estimates.","marker":"Bourrier et al. (2017b)"},{"why":"Provides the 3.295-day rotation period that the new 3.27-day detection validates.","marker":"Vida et al. (2017)"},{"why":"Defines the solar microflare energy range that motivates calling the ~10^29 erg events microflares.","marker":"Hannah et al. (2011)"},{"why":"Gives the comparable K2-25 b Ly-alpha depth limit used as a benchmark for the TRAPPIST-1 constraints.","marker":"Rockcliffe et al. (2021)"}],"fun_headline_variants":["No exospheres in TRAPPIST-1: microflares explain null result","Microflares set detection floor for TRAPPIST-1 exosphere searches","TRAPPIST-1's microflares dilute Lyman-alpha transit signals","104 HST orbits find no exospheres, only microflares","Stellar microflares, not exospheres, control TRAPPIST-1's Ly-alpha"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The flare temperature and filling factor rest on the assumption that a Lyman-alpha flare and a g-band flare with similar waiting times are the same kind of event, so their amplitudes can be combined in a single blackbody model; if the two bands see different flare populations, the 11000 K and 0.011 percent values do not follow.","fun_headline_variants_meta":{"raw":{"variants":["No exospheres in TRAPPIST-1: microflares explain null result","Microflares set detection floor for TRAPPIST-1 exosphere searches","TRAPPIST-1's microflares dilute Lyman-alpha transit signals","104 HST orbits find no exospheres, only microflares","Stellar microflares, not exospheres, control TRAPPIST-1's Ly-alpha"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000873,"raw_usage":{"total_tokens":3929,"prompt_tokens":1243,"completion_tokens":2686,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":859,"completion_tokens_details":{"reasoning_tokens":2589}},"tokens_in":859,"tokens_out":2686,"duration_ms":22548,"temperature":1.0,"reasoning_tokens":2589,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:58:17.605183+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simultaneous HST/STIS Lyman-$\\alpha$ and VLT g'-band monitoring of TRAPPIST-1 would settle it: if the 300-500 percent Lyman-$\\alpha$ bursts do not coincide with the few-percent g'-band brightenings predicted by an $11000$ K, $0.011\\%$ filling-factor blackbody, the inferred microflare properties are wrong.","supporting_citations":[{"cited_title":"E., et al","cited_arxiv_id":null,"evidence_quote":"Gives the escape efficiency factors used to convert XUV luminosity into water-loss upper limits."},{"cited_title":"G., Hudson , H","cited_arxiv_id":null,"evidence_quote":"Defines the solar microflare energy range that motivates calling the ~10^29 erg events microflares."},{"cited_title":"E., Newton, E","cited_arxiv_id":null,"evidence_quote":"Gives the comparable K2-25 b Ly-alpha depth limit used as a benchmark for the TRAPPIST-1 constraints."}],"review_version":1}