{"id":"192c849e-1a82-4e21-9bb2-82aeeb511acd","arxiv_id":"2501.13196","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A direct 3.4-year decay-curve measurement with the DEAP-3600 detector gives an argon-39 half-life of (302 ± 8stat ± 6sys) years, longer than the accepted value.","lead":"Using 3.4 years of data from the DEAP-3600 dark matter detector, the collaboration directly observed the decay of radioactive argon-39 and measured its half-life as 302 years, about 12% longer than the widely used value. The result matters because the argon-39 half-life is an input to argon-based dating and to background estimates in future argon dark matter experiments.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The light-yield correction factor in Sec. 3.1 appears inverted: as written it would amplify, not remove, PE drift, directly biasing the measured 39Ar half-life.","rationale":"The reader's weakest assumption correctly identifies the light-yield correction in Sec. 3.1 as the most load-bearing element of the analysis. The text and Sec. 6.1 together make a strong case that the factor Yj/Y0 is applied multiplicatively to event PE values. Given the standard definition of light yield as PE/keVee, this is the wrong direction: the correct correction is Y0/Yj. Because the ROI is a fixed PE window and the spectral fraction fi inside the ROI is steeply energy-dependent, an inverted correction would imprint a time-dependent efficiency on the observed trigger rate, directly mimicking or masking a change in the 39Ar decay rate. The differential light-yield systematic of 5.1 years is estimated by small perturbations of the correction factor, but perturbing an inverted factor does not sample the inversion itself, so the quoted systematic does not cover this error. The result would shift by an amount depending on the actual light-yield drift direction; Figure 6 shows clear run-to-run variations, so the effect is not negligible. No other concern appears as directly tied to the central claim: the statistical model, pile-up treatment, and background assumptions are described in detail with dedicated systematics, and the tension with NDS (p = 0.008) is physically plausible if the half-life is indeed longer. The paper is otherwise careful and reproducible in principle, but the sign ambiguity in the central correction must be resolved before the result can be used as a reference value. A conditional acceptance with a request to clarify or correct this point is appropriate, matching the reader's verdict.","tokens_in":14461,"tokens_out":4409,"duration_ms":49053,"concrete_test":"A decisive check is to apply the stated correction to the 40K calibration peak for each run: multiply each run's measured 40K peak PE by Yj/Y0 and confirm that the corrected peak position is constant across runs at Y0 × 1460 keV. If instead the corrected peak scales as Yj²/Y0 (i.e., it moves with the square of the light yield), the factor is inverted. In that case, re-run the full fit with PE values corrected by Y0/Yj; if the fitted half-life changes by more than the quoted 6-year systematic, the published result is not robust to the sign ambiguity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Sec. 3.1 the authors state: \"The PE value of each event in a run is then corrected by a factor of Yj/Y0,\" where Y is the light yield in PE/keVee. Since a higher light yield means more PE per keV, the correct mapping to the reference run is PE × (Y0/Yj), not PE × (Yj/Y0). Sec. 6.1 uses the same multiplicative form \"Yj/Y0 + ΔY\", confirming that the factor is applied as written. If applied literally, a run with Yj > Y0 would have its PE values stretched further, amplifying the very drift the correction is meant to remove. Because the ROI is defined in PE (700–1200 PE), this spurious scaling changes the fraction of the 39Ar spectrum inside the ROI as a function of time, mimicking a change in the decay rate. The dominant systematic (differential light-yield drift, 5.1 yr) is estimated by perturbing this factor around its nominal value, so an inversion of the factor is not covered by that uncertainty—it would shift the central value beyond the quoted ±6 yr systematic. The central claim therefore rests on the correction being implemented in the opposite direction from the one stated, or on a non-standard meaning of \"corrected by a factor\". This is a concrete, checkable ambiguity in the text, and it directly affects the measured half-life.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the first direct measurement of the 39Ar half-life by observing the decay curve with the DEAP-3600 liquid-argon detector over 3.4 years. The authors select electron-recoil events in a 700-1200 PE region of interest, model the ROI trigger rate as a combination of single, double, and triple 39Ar pile-up, 39Ar-Cherenkov pile-up, and a constant gamma background, and fit for the initial 39Ar rate and lifetime. The result is T1/2 = (302 ± 8stat ± 6sys) years, about 12% longer than the NDS value, with a reported p-value of 0.008 for disagreement. The systematic budget is dominated by light yield corrections.","tokens_in":14842,"tokens_out":7448,"duration_ms":81030,"significance":"If the result is correct, it is a valuable first direct decay-curve measurement of the 39Ar half-life, with implications for geochronology, radiochemistry, and argon-based dark matter detectors. The paper's strengths include a clean derivation of the fit model from Poisson statistics, a free-floating half-life (so no circularity in the central fit), very high statistics, explicit Monte Carlo and data cross-checks, and a reasonably complete systematic inventory. The main weakness is the light yield correction described in Section 3.1: as written, the direction of the correction appears inverted, and the dominant systematic is evaluated by perturbing that same factor. Because the expected decay signal over 3.4 years is only about 0.8% in rate, an error in this correction could change the central value well beyond the quoted uncertainties. The result is therefore conditional on resolving this point.","major_comments":[{"comment":"The stated correction factor appears inverted. Section 3.1 says 'The PE value of each event in a run is then corrected by a factor of Yj/Y0,' where Y is the light yield in PE/keVee. For a standard light yield definition, an event with a given energy has PE_j = Y_j * E, so the mapping to the reference run is PE_0 = (Y_0 / Y_j) * PE_j. The paper's formula multiplies by Y_j / Y_0, which would amplify, not remove, run-to-run light yield drift; Section 6.1 repeats this form as 'Yj/Y0 + ΔY'. Since the ROI is fixed in PE, an inverted correction changes the fraction of the 39Ar spectrum inside the ROI as a function of time, potentially mimicking a change in the decay rate. The quoted differential systematic (5.1 years) is obtained by perturbing this factor in magnitude, not by checking its sign, so it does not cover an inversion. The authors must clarify whether a non-standard definition of Y is being used, correct the formula if it is a typo, and show with a closure test (e.g., Monte Carlo injection of a known lifetime or a residual-rate versus light-yield check) that the implemented correction is in the direction that removes drift.","section":"Sec. 3.1 and Sec. 6.1"},{"comment":"The differential drift systematic is evaluated by adding a linear ramp of amplitude ΔY = 0.00076 to the correction factor and refitting. However, the calibration source light yields plotted in Fig. 6 show substantially larger, non-linear time variations (of order 0.1 PE/keVee over the dataset). While the nominal run-by-run correction may remove these variations if the calibration measurements perfectly track the detector response relevant to the ROI, the paper does not provide a direct closure check demonstrating that the corrected rates are insensitive to the observed non-linear light yield structure. Given that the physics signal is only about 0.8% in rate over the full dataset, a non-linear residual drift at the level visible in Fig. 6 could be comparable to the signal. The revision should include an explicit validation that the applied correction removes the observed time-dependent light yield changes, beyond the linear-ramp perturbation used for the systematic.","section":"Sec. 6.1 and Fig. 6"}],"minor_comments":[{"comment":"The manuscript header contains '3.4 Y ears of Data'; the word 'Years' should be restored.","section":"Title/header"},{"comment":"The text 'Individual runtimes Trun vary in the range of [ O(1minute ), ∼2 days]' has spacing and notation issues; it should read 'O(1 minute)' or a conventional notation for order-of-magnitude ranges.","section":"Sec. 3"},{"comment":"In the 'Constraints' column, entries such as 'Run-dependent N/A' are unclear: the table should specify that the light yield correction factors are run-dependent parameters whose uncertainties are evaluated as described in Section 6.1, and what 'N/A' refers to.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the direction of the light yield correction. If the Yj/Y0 factor is a typographical inversion and the implementation is actually correct, the paper is close to publishable after clarification and a closure test. If the implementation follows the written factor literally, the central value would be biased and the result could not stand. The disagreement with the NDS value is not by itself a reason to reject; a direct measurement may differ, but the systematic treatment must be credible. Please ask the authors for the exact implementation and a validation test."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is the first to measure the 39Ar half-life by directly watching its decay curve, using the DEAP-3600 detector. That is a real novelty: all prior determinations used isotopic or activity ratios. The statistical model is cleanly derived from Poisson statistics, with explicit pile-up terms and a fixed background rate, and the systematic accounting is unusually thorough. The result, (302 ± 8 ± 6) years, is in tension with the accepted 269-year value, and the paper acknowledges that with a p-value of 0.008.\n\nThe soft spot is the light-yield correction in Section 3.1. The text says each event's PE is \"corrected by a factor of Yj/Y0,\" where Y is the light yield in PE/keVee. If you multiply event PE by Yj/Y0, a run with higher light yield gets stretched further, which amplifies the drift instead of removing it. The correct normalization to the reference run would be a factor of Y0/Yj. Section 6.1 uses the same multiplicative form when estimating the systematic, which tells me this is likely what was actually implemented, not a typo. The differential drift systematic of 5.1 years is the dominant uncertainty, and it is evaluated by perturbing this same factor around its nominal value. An inversion is not covered by that perturbation; it would shift the central value beyond the quoted ±6 years. This is a concrete, checkable ambiguity that directly affects the reported half-life.\n\nThe rest of the analysis holds up well. The pile-up fractions, efficiencies, and background rates are estimated from data and MC without reference to the target half-life, so the circularity burden is low. The citation pattern looks normal, including reasonable citations of their own detector papers. The paper is worth serious referee time, but the referee should demand a clear statement of the correction direction and, ideally, a cross-check of the result with the factor inverted. If the correction is actually Y0/Yj, the measurement stands; as written, the central claim rests on an ambiguous sentence that might be a sign error.\n\nI would bring it to reading group because it is a nice case study of how a seemingly minor calibration detail can load-bear on a physics result. I would not cite it as a new reference value until the correction direction is resolved.","headline":"First direct 39Ar half-life measurement from a decay curve, but the light-yield correction as written is inverted and could shift the central value past the quoted systematic.","tokens_in":16039,"tokens_out":3230,"would_cite":false,"duration_ms":35338,"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":"DEAP-3600 directly observed the $^{39}$Ar decay curve over 3.4 years and measured the half-life as $(302 \\pm 8_{\\mathrm{stat}} \\pm 6_{\\mathrm{sys}})$ years, about 12% longer than the accepted 269-year value.","keywords":["argon-39","half-life","beta decay","DEAP-3600","liquid argon detector","pulse-shape discrimination","radioisotope dating","atmospheric argon"],"falsifier":"Re-fit the trigger-rate curve with the light-yield correction omitted and again with its sign reversed (dividing by $Y_j/Y_0$ instead of multiplying). If either variant shifts the half-life by more than the quoted $\\pm 6$-year systematic, the central result is controlled by the correction rather than by the decay; a second, independent check is to fit the two halves of the 3.4-year dataset separately and compare their $T_{1/2}$ values.","tokens_in":14297,"feed_emoji":"⏳","tokens_out":7626,"duration_ms":71636,"temperature":0.7,"pith_summary":"The paper reports the first direct measurement of the $^{39}$Ar half-life from continuous observation of its decay curve, using 3.4 years of data from the DEAP-3600 liquid-argon detector. The measured value is $T_{1/2}=(302 \\pm 8_{\\mathrm{stat}} \\pm 6_{\\mathrm{sys}})$ years, about 12% longer than the widely used $(269 \\pm 3_{\\mathrm{stat}} \\pm 8_{\\mathrm{sys}})$-year value from 1965 and the $(268 \\pm 8)$-year Nuclear Data Sheets evaluation. A direct decay-curve measurement matters because all previous estimates relied on isotopic-ratio or activity-ratio methods, none of which continuously tracked the decay. If correct, the longer half-life changes the inferred $^{39}$Ar abundance in atmospheric argon and recalibrates dating and radiochemistry applications that use this isotope.","feed_headline":"Argon-39 half-life measured 12% longer than accepted value","feed_subtitle":"Directly counting argon-39 decays over 3.4 years yields a half-life of 302 years, challenging the 269-year standard.","key_machinery":"The machine that carries the argument is the detector's ability to count $^{39}$Ar decays continuously and to model the trigger rate. DEAP-3600's liquid argon target is so rich in $^{39}$Ar that this isotope dominates triggers by two orders of magnitude; pulse-shape discrimination (the ratio of prompt to total charge, $F_{\\mathrm{prompt}}$) separates the electron-recoil band from nuclear recoils, and a data-cleaning cut plus livetime calculation converts raw triggers into rates. The rate model explicitly includes Poisson probabilities for uncorrelated pile-up (one, two, or three $^{39}$Ar decays landing in the same 10 $\\mu$s trigger window, and $^{39}$Ar plus Cherenkov light), so the exponential decay $R_{^{39}\\mathrm{Ar}}(t)=R_{^{39}\\mathrm{Ar}}e^{-t/\\tau_{^{39}\\mathrm{Ar}}}$ can be extracted with the mean lifetime $\\tau$ as a free parameter. A light-yield correction, derived from daily calibrations with $^{40}$K, $^{208}$Tl, and the $^{39}$Ar $\\beta$ spectrum, is applied to each event's PE value to remove time-dependent detector response changes.","core_discovery":"The central discovery is a half-life value obtained from a direct decay-rate fit. Selecting clean electron-recoil events with pulse-shape discrimination in the 700–1200 PE window, the collaboration measured the $^{39}$Ar trigger rate as a function of time and fit it with a model that includes single decays, double and triple pile-up, $^{39}$Ar–Cherenkov pile-up, and a constant ERB $\\gamma$ background. The fit yields $T_{1/2}=(302 \\pm 8_{\\mathrm{stat}} \\pm 6_{\\mathrm{sys}})$ years, and the paper reports a $p$-value of 0.008 for consistency with the NDS value, i.e. the two are in tension. When combined with the earlier DEAP specific-activity measurement, this corresponds to a $^{39}$Ar abundance in atmospheric argon of $(8.6 \\pm 0.4)\\times 10^{-16}$.","pith_inferences":["Not stated in the paper, the sign of the light-yield correction in Section 3.1 is worth an independent check: the text says each event's PE value is corrected by multiplying by $Y_j/Y_0$, and if the intended correction was to divide by that ratio, the drift used to extract the half-life would move in the opposite direction and could shift the central value by more than the quoted $\\pm 6$-year syst","Because the light-yield correction dominates the systematic budget, an external test would be to fit the 700–900 PE and 900–1200 PE sub-regions separately; a significant disagreement between the two inferred half-lives would indicate residual spectral-shape sensitivity.","The method of continuously counting a dominant endogenous decay and modeling pile-up statistics could be transplanted to other long-lived isotopes in large liquid detectors, such as $^{85}$Kr in liquid argon or $^{14}$C in organic scintillators, where no direct decay-curve half-life measurement exists."],"forward_implications":["If the 302-year value is right, the 269-year standard is about 12% too short, and the NDS evaluation (268 ± 8 years) disagrees at a p-value of 0.008.","Combining the new half-life with DEAP's measured specific activity gives an atmospheric $^{39}$Ar abundance of $(8.6 \\pm 0.4)\\times10^{-16}$, a direct estimate that does not depend on the older half-life.","Radiometric dating and tracer methods that use $^{39}$Ar would need to update their decay constants, shifting inferred ages by the corresponding factor.","Future dark-matter experiments using atmospheric argon would see a slightly higher $^{39}$Ar background rate per unit mass if the half-life is longer, which affects their background-model inputs."],"supporting_citations":[{"why":"Supplies the DEAP-3600 specific activity measurement (0.964 ± 0.001 ± 0.024 Bq/kg_atm) used to cross-check the fitted initial rate and to derive the $^{39}$Ar abundance.","marker":"[7]"},{"why":"Establishes the widely used 269 ± 3 ± 8 year half-life from activity ratios, the main historical baseline this measurement claims to supersede.","marker":"[14]"},{"why":"Provides the NDS 2018 re-evaluation (268 ± 8 years) against which the paper quotes a p-value of 0.008 for disagreement.","marker":"[17]"},{"why":"Gives the $^{39}$Ar beta-spectrum model used in the toy Monte Carlo to compute the fractions $f_i$ of each event type falling in the ROI.","marker":"[28]"},{"why":"Gives the 1996 mass-spectrometry half-life (276 ± 3 years), another prior measurement compared with this result.","marker":"[16]"},{"why":"Describes the DEAP-3600 detector, including the trigger and PMT calibration that underpin the decay-rate measurement.","marker":"[18]"},{"why":"Supplies the earliest half-life estimate (265 ± 30 years) from isotopic ratios, included in the comparison table.","marker":"[13]"},{"why":"Gives the 1960 activity-ratio result (325 ± 16 years) used as an earlier comparison.","marker":"[15]"}],"fun_headline_variants":["Direct argon-39 decay count puts half-life at 302 years","Argon-39 half-life 12% longer than accepted from DEAP-3600","Underground detector measures argon-39 half-life 302 years","Argon-39 half-life revised up to 302 years after long count"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the PE-scale correction of Section 3.1 completely removes time-dependent light-yield variations, so that the remaining change in trigger rate is purely the $^{39}$Ar decay; if the correction leaves, or introduces, a drift, the measured half-life changes by more than the quoted systematic uncertainty.","fun_headline_variants_meta":{"raw":{"variants":["Direct argon-39 decay count puts half-life at 302 years","Argon-39 half-life 12% longer than accepted from DEAP-3600","Underground detector measures argon-39 half-life 302 years","Argon-39 half-life revised up to 302 years after long count"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000285,"raw_usage":{"total_tokens":1700,"prompt_tokens":991,"completion_tokens":709,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":607,"completion_tokens_details":{"reasoning_tokens":627}},"tokens_in":607,"tokens_out":709,"duration_ms":6621,"temperature":1.0,"reasoning_tokens":627,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:22:42.846799+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-fit the trigger-rate curve with the light-yield correction omitted and again with its sign reversed (dividing by $Y_j/Y_0$ instead of multiplying). If either variant shifts the half-life by more than the quoted $\\pm 6$-year systematic, the central result is controlled by the correction rather than by the decay; a second, independent check is to fit the two halves of the 3.4-year dataset separately and compare their $T_{1/2}$ values.","supporting_citations":[{"cited_title":"Adhikari, et al","cited_arxiv_id":null,"evidence_quote":"Supplies the DEAP-3600 specific activity measurement (0.964 ± 0.001 ± 0.024 Bq/kg_atm) used to cross-check the fitted initial rate and to derive the $^{39}$Ar abundance."},{"cited_title":"Zeldes, et al., Phys","cited_arxiv_id":null,"evidence_quote":"Establishes the widely used 269 ± 3 ± 8 year half-life from activity ratios, the main historical baseline this measurement claims to supersede."},{"cited_title":"Stoenner, O.A","cited_arxiv_id":null,"evidence_quote":"Provides the NDS 2018 re-evaluation (268 ± 8 years) against which the paper quotes a p-value of 0.008 for disagreement."},{"cited_title":"Kostensalo, J","cited_arxiv_id":null,"evidence_quote":"Gives the $^{39}$Ar beta-spectrum model used in the toy Monte Carlo to compute the fractions $f_i$ of each event type falling in the ROI."},{"cited_title":"Baksi, D.A","cited_arxiv_id":null,"evidence_quote":"Gives the 1996 mass-spectrometry half-life (276 ± 3 years), another prior measurement compared with this result."},{"cited_title":"Seaborg, I","cited_arxiv_id":null,"evidence_quote":"Supplies the earliest half-life estimate (265 ± 30 years) from isotopic ratios, included in the comparison table."},{"cited_title":"Stoenner, O.A","cited_arxiv_id":null,"evidence_quote":"Gives the 1960 activity-ratio result (325 ± 16 years) used as an earlier comparison."}],"review_version":1}