{"id":"d83c1f95-2513-473b-a00c-da1dce4e5ddd","arxiv_id":"2601.11296","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"An S2-only analysis of 7.83 tonne-years of XENONnT data finds no dark matter excess and sets 90% CL limits, e.g. 6.0e-45 cm^2 for spin-independent scattering at 5 GeV/c^2.","lead":"The XENONnT dark matter experiment searched for light dark matter using only the ionization signal left in liquid xenon, with 7.83 tonne-years of data. It saw no signal above background and set its strongest limits yet on several light dark matter models, bringing direct detection closer to the neutrino floor.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Post-unblinding relaxation of the SR2 temporal-isolation selection is the load-bearing point: the paper calls it conservative with negligible impact but provides no quantitative before/after comparison, leaving the signal efficiency—and thus all quoted limits—dependent on an unquantified correction.","rationale":"Reader identified the same load-bearing concern; I agree. The manuscript itself supplies the key evidence at Results: post-unblinding discrepancy, selection relaxed, 'conservative adjustment' with no quantification. This is exactly the kind of missing support that the review rules ask me to weigh. Independent support in the paper is genuine: blinded science ROI, calibration validation of the background model, and PCL reporting. Those support the null result. But the improved limit claim depends on the SR2 efficiency, and a post-unblinding change to the very selection that defines efficiency cannot be dismissed without a quantitative before/after comparison. The 11.2% systematic is not obviously applicable to a bias discovered after unblinding. The Discussion also flags that a fully simulation-driven cathode model is future work, which is a limitation; however, the cathode component is normalized to data sidebands and its uncertainty is propagated, so it is less decisive than the unquantified efficiency correction. Verdict remains CONDITIONAL as the reader set it.","tokens_in":15257,"tokens_out":4273,"duration_ms":47310,"concrete_test":"Recompute the SR2 signal efficiency and final 90% CL SI limit at 5 GeV/c^2 using a closed-loop simulation in which injected DM S2 signals are accompanied by realistic ambient S2s sampled from the same CNF/DE model used to generate backgrounds, under both the original and relaxed temporal-isolation selections. If the efficiency ratio (relaxed/original) deviates from unity by more than the 11.2% systematic, or if the 5 GeV/c^2 limit changes by more than 11.2%, the 'negligible impact' claim fails and the limits should be reported with the original blinded selection or the correction should be folded into the systematic.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—no excess and improved 90% CL limits—rests on the signal-efficiency estimate, since the limits scale directly with acceptance. The efficiency is derived by salting simulated DM S2 signals into real data streams (Signal section). The paper admits in Results that this method cannot reproduce ambient S2s generated by real DM events, so the temporal-isolation selection was less effective on salted waveforms than on real data, 'leading to an overestimation of signal efficiency.' A discrepancy was observed in SR2 for cS2<200 PE after unblinding, and the selection was relaxed to consider only ambient S2s preceding DM candidates. The paper states the adjustment is conservative and has 'negligible impact on the final limits,' but no numbers are given: no pre/post efficiency ratio, no change in the 5 GeV/c^2 limit, and no estimate of the residual bias. Because this is a post-unblinding change to a selection that directly multiplies the signal rate, the quoted limit 6.0e-45 cm^2 at 5 GeV/c^2 could shift by an unknown factor; the direction is likely toward weaker limits, so the null-result statement is probably safe, but the headline limit is not fully supported. The 11.2% systematic assigned to signal efficiency was presumably evaluated before this correction and therefore does not automatically cover it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a blinded S2-only (ionization-only) dark matter search in XENONnT using 7.83 tonne·year of exposure from three science runs. The analysis develops a full background model with cathode, delayed-electron (DE), accidental-electron (AE), and 8B CEνNS components, and uses a likelihood on the corrected S2 (cS2) spectrum. No significant excess is observed, and 90% CL upper limits are set for SI and SD DM-nucleon scattering, DM-electron scattering, axion-like particles, and dark photons. The headline result is a SI DM-nucleon cross-section limit of 6.0×10^-45 cm^2 at 5 GeV/c^2.","tokens_in":15608,"tokens_out":2853,"duration_ms":32755,"significance":"If the analysis is valid, this is a substantial experimental advance: it provides the first complete S2-only background model in XENONnT, extends ionization-only sensitivity to sub-keV nuclear and electronic recoils, and improves limits in several light-DM channels. The paper has real methodological strengths: the analysis is blinded, background models are validated against 220Rn/222Rn calibrations, the statistical inference uses power-constrained limits, and machine-checked or publicly available analysis tools (fuse, straxen, alea) and prior calibration data are cited. The central null result is credible, but the unquantified post-unblinding efficiency correction affects the headline limits and needs to be addressed before the results can be considered fully supported.","major_comments":[{"comment":"This paragraph is load-bearing but lacks quantitative support. The paper states that the salting method overestimated signal efficiency because it cannot simulate ambient S2s from real DM events, and that in SR2 for cS2<200 PE the temporal-isolation selection was therefore relaxed. It then asserts that the adjustment is conservative and has 'negligible impact on the final limits', but gives no before/after efficiency numbers, no change in the quoted 5 GeV/c^2 limit, and no estimate of residual bias. Since signal efficiency multiplies the signal rate, even a few-percent change at low cS2 can shift the limits, and the 11.2% systematic quoted in the Signal section may not cover a post-unblinding correction. Please provide a quantitative assessment: the corrected efficiency versus the pre-correction efficiency by cS2 bin, the impact on the reported limits, and an explicit statement of how th","section":"Results, paragraph beginning 'A discrepancy between data-driven and simulation-derived efficiency...'"},{"comment":"The paper acknowledges the fundamental limitation of the salting method: salted DM S2 signals do not produce the ambient S2s that a real DM event would produce, so the temporal-isolation selection is less effective on salted waveforms. This is an admitted shortcoming of the signal-efficiency evaluation. The paper states that the selection was relaxed to consider only ambient S2s preceding DM candidates, but it does not describe how the corrected efficiency is validated (e.g., with wall, cathode, or 37Ar calibration events) or how the 11.2% systematic uncertainty was modified after this correction. Without this, the central limit depends on an unquantified acceptance correction.","section":"Signal, first paragraph (efficiency evaluation) and Results"},{"comment":"The cathode background is the dominant background in the science ROI (Table I), yet its cS2 spectrum is not derived from first principles. The paper rescaled a simulated cathode cS2 spectrum to match a cathode-dominated sideband, using a simulation-driven sideband-to-ROI ratio in each cS2 bin. The systematic uncertainties from sideband statistics, the rescaling ratio, and background leakage are propagated to the inference. However, the validation of this rescaled spectrum in the science ROI and its impact on the final limits are not shown in the Letter. Given that the cathode rate is approximately 70% of the total background in SR0/SR1 and the cS2 shape is similar to signal, the shape systematic is a key ingredient for the limit; please provide a closure test or describe how the calibration data constrain the in-ROI cathode shape.","section":"Background modeling, cathode paragraph"},{"comment":"The background-only hypothesis in SR2 has a χ2 p-value of 0.013 in the cS2 dimension, attributed to a downward fluctuation. This is a mild tension, and the paper does not discuss whether this downward fluctuation artificially strengthens or weakens the reported limits, especially when combined with the post-unblinding efficiency correction in the same SR and cS2 region. Since the PCL procedure is used, please show the observed limit relative to the expected limit and the power-constraint threshold for SR2, and clarify whether the unblinded efficiency correction increases the uncertainty on the SR2 background rates beyond the quoted values.","section":"Results and Table I, SR2 p-value"}],"minor_comments":[{"comment":"The caption says 'black dashed (solid) lines show limits before (after) −1σ power-constrained limit (PCL)', which is confusing. Please rephrase to clarify what is dashed and what is solid, and whether the PCL is applied only when the observed limit is more than 1σ below the expected.","section":"Fig. 4 caption"},{"comment":"The 'Accidental electron' row lists a dash for SR2 but no explicit statement in the text that the AE component is zero because of the higher S2 area threshold; this is stated in the text, but adding '— (negligible)' in the table would improve readability.","section":"Table I"},{"comment":"The phrase 'negligible impact on the final limits' should be moved to a quantitative section or replaced with a specific reference to a table/figure. In the current form, it is an unsupported qualitative claim.","section":"Text, 'negligible impact' statement"},{"comment":"The paper uses several acronyms (DE, AE, CNF, BDT, PCL) that are defined internally, but a short glossary or a more explicit definition at first use would help readers outside the XENON analysis framework.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The core null result is likely correct, and the analysis is careful in many respects. The main issue is the post-unblinding efficiency correction in SR2: it is a selection change made after seeing data, it directly affects the signal acceptance, and the paper provides no quantitative before/after comparison. This is not a fatal flaw in the null-result claim, but it is a load-bearing point for the quoted limits. I recommend requiring the authors to add a quantitative treatment of this correction and its systematic uncertainty before publication. The cathode shape modeling should also be clarified, as it is the dominant background."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here’s your quick read on the XENONnT S2-only paper. It’s a solid, careful null result with genuinely new limits. The first complete S2-only background model in XENONnT is a real step up from the XENON1T analysis; the CNF-based delayed-electron modeling and cathode BDT are technically well done. The 7.83 tonne-year exposure, blind analysis, and calibration-driven validation give me confidence in the no-excess claim. The 6e-45 cm^2 at 5 GeV/c^2 SI limit and the new constraints on DM-electron, ALP, and dark photon channels are worth having.\n\nThe soft spot is the post-unblinding adjustment to the temporal isolation selection in SR2. The paper finds a discrepancy between data-driven and simulation-derived efficiency for cS2<200 PE, attributes it to the salting method’s inability to model ambient S2s from real DM events, and relaxes the selection to be conservative. It states the impact is “negligible” but gives no numbers. Since the exclusion limits scale directly with signal efficiency, a referee should ask for the before/after efficiency curves and the corresponding shift in the reported limits. The direction of the correction (relaxing a selection that had been cutting real signals) means the true limits are probably slightly weaker, not stronger, so the null result is not in jeopardy. But the headline limit should not be taken at face value until this is quantified.\n\nA second minor point: the dominant cathode background is not modeled from first principles; it’s a simulated spectrum rescaled to a sideband. The paper is transparent about this and includes systematic uncertainties, but the 11.2% efficiency systematic was presumably evaluated before the SR2 change and may not cover the resulting correction. That’s an argument for seeing the numbers, not for rejecting the paper.\n\nOverall, this is a credible experimental result that deserves full peer review. The main request to the authors should be a quantitative description of the SR2 efficiency correction and its impact on the limits. I’d cite this in any light-DM review.","headline":"Solid null result with improved light-DM limits, but the post-unblinding SR2 efficiency correction needs to be quantified before the headline limit is taken at face value.","tokens_in":17100,"tokens_out":5833,"would_cite":true,"duration_ms":56566,"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 7.8 tonne-year search using ionization-only signals in a liquid-xenon detector finds no excess and excludes spin-independent dark-matter cross sections above 6×10⁻⁴⁵ cm² at 5 GeV/c².","keywords":["dark matter","light dark matter","ionization-only signals","S2-only analysis","liquid xenon TPC","direct detection","upper limits","background modeling"],"falsifier":"A calibration measurement that generates real low-energy S2 signals with ambient S2 activity, then re-runs the full selection and compares the acceptance with the salted-simulation efficiency in the cS2 < 200 PE region, would settle whether the quoted limits are biased; if the true acceptance differs by more than the 11.2% systematic uncertainty, all cross-section limits in this search would need to be re-derived.","tokens_in":15136,"feed_emoji":"🌌","tokens_out":6349,"duration_ms":64665,"temperature":0.7,"pith_summary":"This paper tries to establish that light dark matter does not produce an observable ionization-only signal in 7.83 tonne-years of XENONnT data, and that this channel can push direct-detection sensitivity into a previously hard-to-reach low-recoil-energy regime. The search covers nuclear recoils from 0.5 to 5 keV and electronic recoils from 0.04 to 0.7 keV, with a complete background model that includes cathode radioactivity, delayed electrons, accidental electrons, and solar neutrinos. No significant excess is observed, and the resulting 90% upper limits improve on earlier constraints for spin-independent and spin-dependent scattering, dark-matter–electron scattering, axion-like particles, and dark photons. The result matters because it is the first S2-only XENONnT analysis with a full background model, and it brings sensitivity close to the point where solar-neutrino coherent scattering becomes an irreducible background.","feed_headline":"No light dark matter seen in 7.8 tonne-years of XENONnT data","feed_subtitle":"First complete background model for ionization-only events yields the detector's tightest low-mass limits yet.","key_machinery":"The load-bearing object is the S2-only channel in a dual-phase xenon TPC: when a scattering produces too few scintillation photons to be seen above threshold, the ionization electrons still drift, are extracted into the gas phase, and produce proportional scintillation (S2). The analysis uses the corrected S2 size (cS2) in the 80–500 PE range as the energy estimator and leaves the S1 requirement effectively open. Background is suppressed with waveform-shape classifiers, a machine-learning model of the delayed-electron 'ambience' that surrounds large S2s, and a spatial pattern likelihood that rejects accidental pile-up of single-electron signals. To measure signal efficiency, simulated S2 wav","core_discovery":"The central claim is a null result with improved limits: using only the ionization (S2) signal from the liquid-xenon time projection chamber, and after building the first complete S2-only background model for the detector, the experiment observes data consistent with background across three science runs. At 90% confidence, spin-independent dark-matter–nucleon cross sections above 6.0×10⁻⁴⁵ cm² at a dark-matter mass of 5 GeV/c² are excluded; dark-matter–electron scattering cross sections above 2.4×10⁻⁴¹ cm² at 0.4 GeV/c² are excluded; and for bosonic candidates, axioelectric coupling above 3.9×10⁻¹⁴ and kinetic mixing above 2.2×10⁻¹⁷ at 0.1 keV/c² are excluded. Signal efficiency is determined","pith_inferences":["If cathode radioactivity truly dominates, then a detector with a cleaner or shielded cathode could reach the neutrino floor at lower mass than the 5 GeV/c² point quoted here; this is a testable prediction for next-generation liquid-xenon detectors.","The SR2 efficiency discrepancy the paper reports suggests that any future low-threshold analysis using injected simulated signals should validate the temporal-isolation requirement with real low-energy calibration events that produce their own ambient S2s; otherwise quoted limits could become optimistic at the lowest cS2 values.","A natural extension, not fully explored here, would be to run the same background modeling down to single-electron S2s, which could extend sensitivity below the 3 GeV/c² mass floor of this search.","The paper's hint that isolated S2s in S1–S2 coincidence analyses come mainly from the cathode offers a concrete way to improve those searches: reject events correlated with cathode activity rather than treating such S2s only as random accidental background."],"forward_implications":["If the null result stands, the allowed parameter space for dark matter with masses around 3–8 GeV/c² is narrowed, especially for spin-independent and spin-dependent nuclear scattering.","The quoted limits push toward the coherent elastic neutrino-nucleus scattering floor, meaning further sensitivity gains at these masses will require distinguishing dark matter from solar neutrinos rather than simply accumulating more exposure.","The first complete S2-only background model identifies cathode radioactivity as the dominant residual background, directly motivating electrode design and material-purity improvements in future detectors.","The demonstration that machine-learning background models can handle the S2-only environment makes the channel usable for other low-threshold searches, including sub-GeV dark-matter–electron and bosonic dark-matter models.","The absence of an excess across 579 days of data is consistent with the standard halo model assumptions used to derive the limits; any dark matter in this mass range must have a smaller cross section than these bounds."],"fun_headline_variants":["XENONnT's 7.8 t-yr S2-only search tightens low-mass DM limits","S2-only XENONnT data: 7.8 t-yr, no light DM, tighter limits","Null result: XENONnT S2-only 7.8 t-yr improves light DM bounds","No light DM: XENONnT's 7.8 t-yr S2-only data set new bounds","Tighter light-DM limits from XENONnT's 7.8 t-yr S2-only run"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The analysis leans on the assumption that signal efficiency measured by injecting simulated S2 waveforms into real data matches the efficiency for real dark-matter events; the paper itself found this assumption failed in one run for small signals and had to relax the selection, so the corrected efficiency is the load-bearing premise.","fun_headline_variants_meta":{"raw":{"variants":["XENONnT's 7.8 t-yr S2-only search tightens low-mass DM limits","S2-only XENONnT data: 7.8 t-yr, no light DM, tighter limits","Null result: XENONnT S2-only 7.8 t-yr improves light DM bounds","No light DM: XENONnT's 7.8 t-yr S2-only data set new bounds","Tighter light-DM limits from XENONnT's 7.8 t-yr S2-only run"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000606,"raw_usage":{"total_tokens":2713,"prompt_tokens":848,"completion_tokens":1865,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":592,"completion_tokens_details":{"reasoning_tokens":1722}},"tokens_in":592,"tokens_out":1865,"duration_ms":13424,"temperature":1.0,"reasoning_tokens":1722,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T10:06:37.924664+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A calibration measurement that generates real low-energy S2 signals with ambient S2 activity, then re-runs the full selection and compares the acceptance with the salted-simulation efficiency in the cS2 < 200 PE region, would settle whether the quoted limits are biased; if the true acceptance differs by more than the 11.2% systematic uncertainty, all cross-section limits in this search would need to be re-derived.","supporting_citations":[],"review_version":1}