{"id":"9ebfce38-3fb8-457c-ae52-3835fb88c374","arxiv_id":"1908.08754","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Cross-correlation of CARMENES near-infrared transit spectra reveals water vapor in HD 209458 b with S/N 6.4, including a first individual detection from the 1.0 micrometer band.","lead":"This paper reports a likely detection of water vapor in the atmosphere of the hot Jupiter HD 209458 b using high-resolution near-infrared spectroscopy from the CARMENES instrument. The signal is strongest in a newly exploited band near 1.0 micrometers, offering a new probe for studying clouds and winds in giant exoplanet atmospheres.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"SYSREM iteration counts are selected using a 5x-injected model signal; the paper never verifies that the real, 1x signal survives at those same iterations, so the reported S/N may be biased.","rationale":"The central claim is a high-resolution cross-correlation detection of H2O in HD 209458 b, with novelty in the first individual 1.0 micron band detection. The most load-bearing condition is that the real planetary signal is not removed or distorted by SYSREM. The paper's only validation of the SYSREM iteration choice is an injection at 5x (or 10-12x) nominal strength, and the selected iteration counts are applied directly to the real data. This is a data-dependent tuning of a nuisance parameter, and a 5x signal surviving does not establish that a 1x signal survives at the same iterations. This is exactly the reader's weakest assumption, so I agree with the reader. Other concerns are real but secondary: the 6.4 S/N is high enough that a rough trial-factor correction would probably not erase it, and the band inconsistency is plausibly explained by the variable observing conditions, as the authors argue. However, if the SYSREM iteration selection is biased, both the significance and the band-to-band pattern become unreliable, so this concern should be explicitly tested before full acceptance. The paper has useful controls (other molecules show no signal, the drift is small, the peak appears near the expected KP), and the detection may well be real; the concern does not justify rejection. A conditional verdict remains appropriate, pending the proposed 1x injection-recovery check or an equivalent false-alarm test, so the reader's verdict is unchanged.","tokens_in":21413,"tokens_out":8815,"duration_ms":88089,"concrete_test":"Repeat the Section 3.1 injection-recovery test with the planetary model injected at 1x nominal strength, using exactly the same iteration counts selected from the 5x injections, and measure the recovered S/N at the injected velocity in each order and in the combined CCF. If the recovered 1x S/N is not approximately one-fifth of the 5x-injection S/N (accounting for noise), or if the 1x signal is not recovered at a significance consistent with expectations, then the iteration selection is signal-strength dependent and the reported 6.4 S/N on real data is not calibrated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1 selects the number of Sysrem iterations per spectral order by injecting a model at 5x the nominal planetary signal (10-12x in orders 60, 49, and 38) and keeping the iteration count that maximizes the recovered CCF peak (Fig. 4). The chosen counts are then applied to the real data. This is the load-bearing step for the central claim: the claim that the CCF peak is not a telluric artifact requires that the real, 1x signal is not significantly absorbed into the telluric/stellar PCA modes at the selected iteration counts. A 5x-stronger injection is not a surrogate for this. A signal five times brighter may survive iterations at which the real signal is already being fitted and removed by Sysrem, and the iteration count maximizing the injected signal can be shifted relative to the optimum for the real signal. The paper provides no injection-recovery test at 1x nominal strength and no bootstrap or permutation false-alarm estimate for the full iteration-selection plus cross-correlation pipeline. The band inconsistency (the strongest 1.4 micron band shows no signal while the weakest 1.0 micron band gives the largest S/N, Table 2) makes this concrete: if Sysrem at the selected iterations removes the planetary signal more aggressively in the stronger bands, the 1.0-micron-only detection could be a selection effect of the iteration tuning rather than a physical aerosol signature. The reported S/N of 6.4 is the maximum over a 2D KP-vwind grid, but the unresolved issue is the unvalidated Sysrem iteration choice.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes CARMENES near-infrared high-resolution transit spectroscopy of HD 209458 b, searching for water vapor by removing telluric and stellar lines with SYSREM and cross-correlating the residuals with atmospheric transmission models. It reports an H2O detection with maximum S/N of 6.4 and a Welch t-test sigma of 8.1, a net blueshift of -5.2 km/s interpreted as day-to-night winds, and a multi-band analysis claiming a S/N of 5.8 detection from the 1.0 micron band individually, hints from the 1.15 micron band, and no conclusive signal from the nominally stronger 1.4 micron band. The paper also reanalyzes CARMENES data of HD 189733 b with the same pipeline to support a qualitative comparison of haze levels.","tokens_in":21669,"tokens_out":7109,"duration_ms":75607,"significance":"If the detection holds, this would be the first individual detection of H2O from the 1.0 micron band in an exoplanet and would extend the high-resolution cross-correlation technique to a bluer spectral region. The consistent reanalysis of the HD 189733 b dataset is a useful comparative element, and the paper is otherwise careful about data reduction, masking, and the variable precipitable water vapor conditions. However, the central claim is currently not fully established because the SYSREM iteration count is optimized on injected signals without verifying survival of the real 1x signal, and because the reported significance is the maximum of a two-dimensional search without a trial correction. These issues are fixable and do not by themselves invalidate the interpretation, but they require additional analysis before the detection claim is secure.","major_comments":[{"comment":"The number of SYSREM iterations per spectral order is chosen by maximizing the recovered S/N of an injected model at 5 times the nominal planetary strength (10-12 times in orders 60, 49, and 38), but the paper does not show that the real, 1x signal survives at those same iteration counts. A 5x-stronger injection can persist at iteration counts where the real signal is already partially fitted and removed by SYSREM, and the iteration count maximizing an injected signal need not coincide with the optimum for the real signal; this is load-bearing because the claim that the CCF peak is not a telluric artifact depends on the adopted iterations. Please add injection-recovery tests at 1x nominal strength for each order, or demonstrate that the recovered CCF significance is stable over a range of iteration counts, and report the fraction of the injected signal recovered at the adopted iterations.","section":"Section 3.1, Fig. 4"},{"comment":"The quoted S/N of 6.4 and sigma of 8.1 are maxima of a two-dimensional map in KP and vwind, but no correction is made for the number of independent trials in that search. The paper should state the effective number of independent resolution elements in the KP-vwind grid and provide a false-alarm probability, for example from a bootstrap or permutation test that randomizes the in-trail/out-of-trail assignment or uses the negative-KP region as a null distribution. This need is reinforced by the authors' own statement that the Welch t-test may overestimate significance relative to the S/N calculation.","section":"Section 4.3, Fig. 6"},{"comment":"The band-by-band pattern - a S/N of 5.8 in the weakest 1.0 micron band, a hint at S/N of 2.8 in the 1.15 micron band, and no conclusive signal in the nominally strongest 1.4 micron band - is the opposite of the expectation stated in the paper and is currently explained only qualitatively by observing conditions. Because the SYSREM iteration counts are tuned on injected signals, this pattern could in part be a selection effect of the iteration tuning. Please demonstrate, via per-band injection-recovery at 1x strength at the adopted iteration counts or an equivalent sensitivity calculation, that the pipeline would have detected a real signal in the 1.15 and 1.4 micron bands at the level expected from the model, or quantify how much systematic removal suppresses those bands.","section":"Section 5.1, Table 2"}],"minor_comments":[{"comment":"The symbol for radial velocity is rendered as '3' in several places (e.g., '3sys', '3wind'); please use a standard v symbol and define it in the text.","section":"Equation (1), Table 1"},{"comment":"The abstract calls the blueshift a 'firm indication' of day-to-night winds, while Section 5.2 states that the large error bars make the wind measurements fully compatible with previous values; please align the wording with the actual uncertainties.","section":"Abstract and Section 5.2"},{"comment":"The CCF velocity step of 1.3 km/s is described as 'calculated by averaging the velocity step-size of the pixels' but it is not stated whether this is the mean over spectral orders or between pixel centers; please specify the exact computation, since it determines the effective number of independent trials discussed in the major comments.","section":"Section 4.2"},{"comment":"The template is computed with a single P-T profile and a single set of volume mixing ratios from Brogi et al. (2017); because the multi-band interpretation compares observed and expected band strengths, please state whether the main conclusions are robust to plausible variations in these model inputs.","section":"Section 4.1"}],"recommendation":"major_revision","confidential_remarks":"The central claim is plausible and within the scope of A&A, and the manuscript is a reasonable incremental contribution to high-resolution exoplanet spectroscopy. I recommend major revision rather than rejection because the main technical concern - the SYSREM iteration selection at 5x injected signal - can be addressed with additional injection-recovery and false-alarm tests within the scope of the paper. I do not see citation or novelty concerns that would affect the editorial decision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe one thing you should know: this paper reports the first individual water vapor detection from the 1.0 micron band in an exoplanet, on HD 209458 b, using CARMENES NIR transit spectroscopy, plus a day-to-night wind blueshift of -5.2 km/s. The result is plausible and the paper is honestly written, but the significance numbers are optimistic because they come from the maximum of a 2D search grid and a SYSREM iteration count tuned on injected signals five to twelve times stronger than the real one, with no recovery test at the real strength.\n\nWhat is genuinely new: the 1.0 micron detection and the reanalysis of HD 189733 b under the same pipeline, which gives a useful multi-band aerosol contrast between the two planets. The paper is methodical: externally retrieved PT profile and VMRs, KOPRA line-by-line models, careful a priori and a posteriori masking, drift correction, exclusion of low-velocity in-transit spectra, and null checks with CH4, NH3, HCN, and CO. The authors are transparent that the 1.4 micron band—the strongest in coverage—gives no signal while the weakest band gives the largest S/N, and they attribute that to variable PWV, high airmass, and falling S/N. That transparency earns credit. The citation pattern is appropriate: they acknowledge prior HST and CRIRES detections of H2O and only claim the 1.0 micron band as new.\n\nThe soft spots are real. The SYSREM iteration selection is the load-bearing step: choosing the iteration count that maximizes recovery of a 5x (or 10-12x) injected signal does not guarantee that the real 1x signal survives those iterations, since a stronger signal can persist after the algorithm starts sacrificing the weaker one. There is no 1x injection-recovery and no bootstrap or permutation false-alarm probability for the full pipeline, so the quoted S/N of 6.4 and sigma of 8.1 are upper bounds. The Welch t-test overestimation is acknowledged by the authors; adding the 2D grid search lowers the effective significance further. Also, relying on one band is fragile: the 1.0 micron band is the bluest and the least tested in model templates. That said, the peak appears at the expected KP (~150 km/s), is blueshifted, and is absent for other molecules, which argues against a pure telluric artifact.\n\nVerdict: a credible incremental result, not a breakthrough. It deserves peer review; a good referee should ask for the 1x injection-recovery and a permutation FAP. If those hold, the paper is worth publishing; if not, the claim should be softened. Either way, not a desk reject.","headline":"First individual 1.0 micron water vapor detection in HD 209458 b is plausible but the significance is optimistically reported, mainly due to SYSREM iteration tuning on 5x injected signals and no grid-wide false-alarm correction.","tokens_in":22441,"tokens_out":5349,"would_cite":true,"duration_ms":53951,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Water vapor is detected in the hot Jupiter HD 209458 b by extracting thousands of Doppler-shifted spectral lines, giving a signal-to-noise ratio of 6.4 and a significance of 8.1 sigma.","keywords":["hot Jupiter","transmission spectroscopy","water vapor","HD 209458 b","cross-correlation","SYSREM","near-infrared spectroscopy","exoplanet atmospheres"],"falsifier":"Re-observe a transit of HD 209458 b with the same instrument under stable, dry conditions; if the detection is real, the 1.15 and 1.4 $\\mu$m bands should reproduce the water signal at the same $K_\\mathrm{P}$ and blueshift, and injection-recovery tests at one times the nominal planetary strength should recover the signal at the chosen SYSREM iteration without being absorbed.","tokens_in":21204,"feed_emoji":"💧","tokens_out":9459,"duration_ms":84093,"temperature":0.7,"pith_summary":"The paper aims to establish that water vapor exists in the atmosphere of the hot Jupiter HD 209458 b by recovering the faint Doppler-shifted absorption of thousands of water lines from ground-based near-infrared transit spectra. It reports a detection with signal-to-noise ratio 6.4 and significance 8.1$\\sigma$, and claims this is the first time water is detected from the 1.0 $\\mu$m band alone in any exoplanet. If correct, the result places water vapor at the planet's terminator and yields a wind measurement from the signal's blueshift. Multi-band water measurements of this type allow a direct comparison of aerosol extinction and atmospheric dynamics between well-studied hot Jupiters.","feed_headline":"First water detection at 1 micron in HD 209458 b","feed_subtitle":"High-resolution spectra give S/N 6.4 and a blueshift hinting at day-to-night winds.","key_machinery":"The central machinery is cross-correlating high-resolution residual spectra with a synthetic transmission model after removing telluric and stellar lines with SYSREM, a principal component analysis algorithm. During transit the planet's orbital motion Doppler-shifts its absorption lines by tens of km s$^{-1}$, while telluric and stellar features stay nearly fixed, so the static components can be subtracted and the planetary signal, spread over thousands of weak lines, is accumulated into a single cross-correlation function. The number of SYSREM iterations per spectral order is chosen by injecting a model planetary signal at five times (in some orders ten to twelve times) nominal strength and keeping the iteration count that maximizes the recovered cross-correlation peak. The transmission templates are line-by-line models computed with a radiative transfer code using a previously retrieved pressure-temperature profile and assumed molecular abundances.","core_discovery":"The authors detect H$_2$O in the transmission spectrum of HD 209458 b with a maximum S/N of 6.4 and a $\\sigma$-value of 8.1 when all useful NIR spectral orders are combined. The cross-correlation peak appears at a planetary orbital velocity amplitude $K_\\mathrm{P} = 150^{+28}_{-25}$ km s$^{-1}$, consistent with the expected value, and is blueshifted by $-5.2^{+2.6}_{-1.3}$ km s$^{-1}$, which they interpret as day-to-night winds at the terminator. In the multi-band analysis, the 1.0 $\\mu$m band alone gives S/N 5.8, the 1.15 $\\mu$m band gives only a hint at S/N 2.8, and the 1.4 $\\mu$m band yields no conclusive signal. The authors argue that the weaker-than-expected signals from the two stronger bands reflect unfavorable observing conditions, including a sudden drop in precipitable water vapor, high airmass, and falling per-pixel S/N, rather than the absence of water. They also reanalyze CARMENES data of HD 189733 b with the same procedure and find water there mainly in the 1.15 and 1.4 $\\mu$m bands, which supports the picture of stronger aerosol extinction in HD 189733 b.","pith_inferences":["A direct test of the detection would be a second transit observation under stable and dry conditions; the 1.15 and 1.4 $\\mu$m bands should then show independent water signals at the same $K_\\mathrm{P}$ and blueshift if the interpretation is right.","The same iteration-optimization and cross-correlation pipeline could be applied to archived spectra of other hot Jupiters to see whether the 1.0 $\\mu$m band routinely detects water, turning the relative strength of the 1.0 and 1.4 $\\mu$m bands into a haze diagnostic.","Because the iteration count is tuned on an injected signal several times stronger than the real one, a cross-check with injections at one times the nominal strength, or with an independent telluric-removal method, would clarify whether the reported significance depends on that assumption."],"forward_implications":["If the detection is correct, water vapor is present in the terminator atmosphere of HD 209458 b, adding a high-resolution near-infrared confirmation to previous space-based measurements.","The 1.0 $\\mu$m band is demonstrated as a usable ground-based window for detecting water in hot Jupiters, the bluest such band detected individually to date.","The $-5.2$ km s$^{-1}$ blueshift of the combined signal indicates day-to-night winds at the terminator, consistent with global circulation model predictions for this highly irradiated planet.","The different band-by-band behavior of HD 209458 b and HD 189733 b supports the interpretation that HD 189733 b has stronger aerosol or haze extinction near 1 $\\mu$m.","The non-detections in the stronger 1.15 and 1.4 $\\mu$m bands are attributed to the unstable observing conditions of that night, implying that stable, higher-S/N re-observations should recover water in those bands as well."],"supporting_citations":[{"why":"First detection of sodium in HD 209458 b's atmosphere, establishing the transmission-spectroscopy method this paper applies to water.","marker":"Charbonneau et al. (2002)"},{"why":"Introduced high-resolution cross-correlation of transit spectra for HD 209458 b with CO and measured a wind-induced Doppler shift; the technique this paper extends to H2O.","marker":"Snellen et al. (2010)"},{"why":"Previous CARMENES multi-band water detection in HD 189733 b; provides the comparison dataset and methodology reanalyzed here.","marker":"Alonso-Floriano et al. (2019)"},{"why":"Source of the pressure-temperature profile and molecular abundances used to compute the transmission templates.","marker":"Brogi et al. (2017)"},{"why":"Established the practice of optimizing SYSREM iterations by injection-recovery and the CCF significance measures used in this paper.","marker":"Birkby et al. (2017)"},{"why":"Introduced the SYSREM principal component algorithm used to remove telluric and stellar lines.","marker":"Tamuz et al. (2005)"},{"why":"Global circulation model predicting strong day-to-night winds and blueshifted terminator signals that the measured wind is compared with.","marker":"Showman et al. (2013)"},{"why":"Space-based comparative spectra of HD 209458 b and HD 189733 b that motivate the aerosol-extinction interpretation of the band-by-band difference.","marker":"Sing et al. (2016)"}],"fun_headline_variants":["Water at 1 micron in HD 209458 b","First bluest-band water detection on HD 209458 b","HD 209458 b reveals water vapor and terminator winds","Water signal at 1 micron hints at winds on HD 209458 b","Blueshifted water on HD 209458 b points to winds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that the real, much weaker planetary water signal survives the SYSREM telluric and stellar subtraction at the same iteration counts that maximize recovery of an artificially injected signal five to twelve times stronger.","fun_headline_variants_meta":{"raw":{"variants":["Water at 1 micron in HD 209458 b","First bluest-band water detection on HD 209458 b","HD 209458 b reveals water vapor and terminator winds","Water signal at 1 micron hints at winds on HD 209458 b","Blueshifted water on HD 209458 b points to winds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00041,"raw_usage":{"total_tokens":2295,"prompt_tokens":1286,"completion_tokens":1009,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":902,"completion_tokens_details":{"reasoning_tokens":919}},"tokens_in":902,"tokens_out":1009,"duration_ms":10427,"temperature":1.0,"reasoning_tokens":919,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:30:19.386791+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-observe a transit of HD 209458 b with the same instrument under stable, dry conditions; if the detection is real, the 1.15 and 1.4 $\\mu$m bands should reproduce the water signal at the same $K_\\mathrm{P}$ and blueshift, and injection-recovery tests at one times the nominal planetary strength should recover the signal at the chosen SYSREM iteration without being absorbed.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced high-resolution cross-correlation of transit spectra for HD 209458 b with CO and measured a wind-induced Doppler shift; the technique this paper extends to H2O."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the SYSREM principal component algorithm used to remove telluric and stellar lines."},{"cited_title":"P., Fortney, J","cited_arxiv_id":null,"evidence_quote":"Global circulation model predicting strong day-to-night winds and blueshifted terminator signals that the measured wind is compared with."},{"cited_title":"K., Fortney, J","cited_arxiv_id":null,"evidence_quote":"Space-based comparative spectra of HD 209458 b and HD 189733 b that motivate the aerosol-extinction interpretation of the band-by-band difference."}],"review_version":1}