{"id":"0f5de838-6b5a-454d-8857-5c87a29ae875","arxiv_id":"1908.10232","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A re-analysis of the Wu et al. axion search shows that the long-period bound is 9.5 times weaker than claimed because the unknown phase was averaged incorrectly.","lead":"This comment argues that an axion dark matter search with a nuclear spin comagnetometer overstated its sensitivity to ultra-low-mass axions by a factor of 9.5. The authors show that the original analysis mishandled the unknown oscillation phase, and they propose a simple least-squares framework that gives the correct scaling.","discovery_kind":"replication","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The factor 9.5 is asserted without derivation; the text only shows a τ/(πT)=6.0 blow-up, so the headline correction is unverified.","rationale":"Read in good faith: the comment is performing a valuable service if the original analysis indeed mishandles the unknown phase for τ > T. The general LLSA framework is standard and the blow-up argument is physically reasonable. I looked for the weakest point in the demonstration. It is not the statistical philosophy: even under Bayesian phase averaging, a conservative limit on amplitude must respect the least-favorable phase direction, so the original's treatment is at least questionable. The concrete weak point is the unexplained appearance of 9.5. A reader cannot check it from the text. The comment says the sine amplitude error blows up as τ/(πT) = 6.0 at the left edge, but never states how 6.0 becomes 9.5 in the final limit ratio. The likely path is to multiply by a factor from the faulty 1/sinφ averaging, but that step is not written down. Since the conclusion explicitly states a numerical factor and a comparison to previous limits, this is load-bearing. A reanalysis of the published figure would settle it. For now, the comment should be accepted only conditionally, with the derivation supplied.","tokens_in":1759,"tokens_out":13061,"duration_ms":140310,"concrete_test":"Digitize the leftmost point of the original Fig. 3 (Wu et al., PRL 122, 191302) and record the published g limit and the corresponding τ. Re-analyze the same time-series with a linear least-squares fit using the stated T=2.642e6 s, the actual sampling and noise, and basis functions sin(2πt/τ) and cos(2πt/τ). Compute the 95% upper limit on the marginalized amplitude, dividing by the LLSA sensitivity normalization at τ < T. Compare the ratio to 9.5. If the ratio differs by more than about 15%, the headline claim is not supported. A faster analytical check is to derive the ratio of the LLSA limit to the published phase-averaged limit using the formulas in the comment, showing explicitly how the 6.0 blow-up combines with the 1/sinφ correction to yield 9.5; if this derivation cannot be reproduced, flag the numerical claim as unsubstantiated.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's strongest claim is numerical: the exclusion limit at the left edge of Fig. 3 of arXiv:1901.10843 should be 9.5 times higher than published. The qualitative argument for long-period degradation is sound: in a linear least-squares fit, the sine and cosine basis functions become nearly degenerate for τ > T, and the marginalized amplitude uncertainty in the least-favorable phase direction grows as τ/(πT). The comment itself evaluates this factor as 6.0 at the left edge. However, the final factor 9.5 is never derived. It cannot be obtained directly from the 6.0 blow-up; one must additionally fold in the original paper's phase-averaging prescription (Eqs. S12-S17 of the Supplementary), but that combination is not shown. The ratio 9.5/6.0 is close to π/2, suggesting the origin in the 1/sinφ versus average-|sinφ| correction, but the comment does not state or justify this step. Because the comparison to the neutron-EDM limit (15% less constraining) rests entirely on the 9.5 value, an unverified factor leaves the central quantitative conclusion open. If the true factor were, for example, 6.0 or 12/π, the comparison to the prior limit would shift and the comment's punchline would change, even though the existence of a flaw would stand.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript is a comment on Wu et al. (arXiv:1901.10843), which reported a search for axionlike dark matter using a liquid-state nuclear spin comagnetometer. The comment argues that the long-period part of the published exclusion limit (left panel of Fig. 3) is invalid because the data span T=2.642×10^6 s is much shorter than the oscillation periods probed at the left edge. It proposes a linear least-squares analysis (LLSA) in which the sine and cosine quadrature basis functions become nearly degenerate for τ>T, causing the sine amplitude uncertainty to grow as τ/(πT); at the left edge of the published figure this factor is 6.0. It further criticizes the original supplementary-material phase treatment (Eqs. S12-S17) for averaging |sin φ| when the coupling-constraint expression contains 1/sin φ. Assuming the short-period analysis is correct, the comment concludes that the left-edge exclusion limit should be 9.5 times weaker, making it about 15% less constraining than the previous neutron-EDM limit.","tokens_in":2035,"tokens_out":5996,"duration_ms":55115,"significance":"If the arguments are correct, the comment identifies a genuine flaw in a published experimental upper limit and provides a simple, parameter-free explanation for the degradation of sensitivity at long periods. The LLSA framework is a useful cross-check, and the criticism of phase-averaging is statistically well founded. The main quantitative conclusion, however, depends on the factor 9.5, which is not derived in the manuscript; before the comparison to the neutron-EDM limit can be accepted, the derivation must be supplied. The paper otherwise introduces no free parameters and is consistent with other independent searches.","major_comments":[{"comment":"The factor 9.5 in the final paragraph is asserted without derivation. The text establishes a degradation factor τ/(πT)=6.0 for the sine amplitude at the left edge of Fig. 3 and separately argues that the original phase-averaging of |sin φ| is incorrect because the constraint scales as 1/sin φ, but it never shows how these two effects combine to produce 9.5. Since the concluding comparison to the neutron-EDM limit (∼15% less constraining) rests entirely on this number, the authors should state the explicit calculation, e.g., the product of 6.0 with the phase-correction factor and any additional normalization factors, and justify each step with reference to the original analysis equations.","section":"Final paragraph (9.5 factor)"},{"comment":"The claim that the extracted sine amplitude error 'blow[s] up as τ/(πT)' is stated after a short Taylor expansion, but the derivation of the factor is not shown. In particular, the normalization of the basis functions and the construction of the marginalized amplitude uncertainty from the correlation matrix are omitted. A concise derivation (or a reference to the analogous calculation in ref. [5]) would make the factor 6.0 verifiable and would support the subsequent 9.5 factor.","section":"LLSA example"}],"minor_comments":[{"comment":"The phrase 'the constraint actually becomes tighter as πT/τ < 2.5' is confusing; the inequality should be expressed in terms of ξ=τ/T for consistency with the rest of the text.","section":"Paragraph on 'strange long-period behavior'"},{"comment":"The comment would benefit from reproducing or referencing the specific panel of Fig. 3 that is criticized, since the 'left-hand edge' is not identifiable without the original paper at hand.","section":"General"},{"comment":"The sentence 'Note that approaches where the φ ia a free parameter will not have a pole for any non-zero frequency' contains a typo: 'ia' should be 'is'.","section":"Typo"}],"recommendation":"major_revision","confidential_remarks":"The comment is well within the scope of the journal and the criticism appears valid in its qualitative aspects. The missing derivation of the 9.5 factor is the only load-bearing gap; once it is supplied, the comment should be publishable. I would not reject the manuscript because the flaw is fixable in a short revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper: it finds a genuine statistical error in the published axion limit from Wu et al. The physical argument is sound, and the fix is probably in the right ballpark. But the headline factor of 9.5 is not actually derived in the text, so the quantitative punchline needs referee verification.\n\nWhat's new: the specific diagnosis of the phase-averaging mistake in the original supplementary material. The general point that long-period sensitivity degrades for τ > T is not new; refs [2-5] make it, including one by these authors. But connecting it to this specific experiment, and identifying the error in Eqs. S12-S17, is a service.\n\nWhat it does well: the LLSA explanation is clear. Fitting sine and cosine basis functions and showing that for long τ they become nearly degenerate or anticorrelated, so the marginalized amplitude uncertainty blows up as τ/(πT), is transparent and physically compelling. The 6.0 value at the left edge of Fig. 3 is a concrete, checkable number.\n\nSoft spots: the 9.5 factor. The text shows the 6.0 blow-up and separately notes the original authors averaged |sinφ| instead of using 1/sinφ, but never explicitly combines these to get 9.5. A reader has to guess: 6.0 × π/2 ≈ 9.4, which is close, suggesting that is the route, but the comment doesn't say it. That is a real gap in a Comment whose entire point is the corrected limit. Also, the comment assumes the τ < T analysis is correct; if that is wrong too, the corrected limit shifts. That is an honest caveat, but it means the 9.5 is conditional.\n\nThe citation pattern is fine. The references to prior long-period sensitivity work are relevant, and the inclusion of the authors' own paper [5] is proper given that it uses the LLSA method.\n\nWho is this for? Anyone working on axion dark matter searches, especially with comagnetometers or precision spin measurements. It is also a useful pedagogical example of long-period sensitivity in time-series fits. It deserves a serious referee: the issue matters, the argument is mostly clear, and the missing derivation can be fixed with a few lines.\n\nMy recommendation: accept for peer review with the explicit instruction to have the authors show the derivation of the 9.5 factor from the original analysis. If they can, it should be published.","headline":"Real flaw found in Wu et al.'s long-period axion limit, but the corrected factor is asserted, not derived.","tokens_in":2506,"tokens_out":3943,"would_cite":false,"duration_ms":39481,"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":"A 2019 axion-dark-matter limit is too strong by a factor of 9.5, a comment argues.","keywords":["axionlike dark matter","liquid-state nuclear spin comagnetometer","ultralight axion","long-period oscillations","linear least-squares analysis","phase marginalization","exclusion limit","neutron electric dipole moment"],"falsifier":"Reanalyze the original $2.642\\times10^6$ s time series using the linear least-squares fit described here, leaving the axion phase free, and read the upper limit at the left-hand edge of the period range; if the limit moves up by less than a factor of about 9.5 relative to the published curve, the comment's central number is wrong. A cheaper check is to simulate a pure-noise data set with the same duration and sampling and confirm that the marginalized amplitude error scales as $\\tau/(\\pi T)$ once $\\tau>T$.","tokens_in":1581,"feed_emoji":"🧲","tokens_out":9325,"duration_ms":93667,"temperature":0.7,"pith_summary":"This comment argues that a published search for axionlike dark matter using a liquid-state nuclear spin comagnetometer overstated its sensitivity to very low-mass axions. The problem is that the published constraint covers oscillation periods longer than the data span, where the analysis mishandles both the geometry of the measurement and the unknown phase of the axion signal. The comment shows that, for periods longer than the run duration, a correct least-squares treatment makes the uncertainty grow in a predictable way, and the low-mass edge of the exclusion curve should be 9.5 times weaker than reported. If the comment is right, the original experiment does not improve on the earlier neutron-EDM bound; that bound remains about 15% more constraining.","feed_headline":"Axion dark-matter limit is too strong by a factor of 9.5","feed_subtitle":"Corrected for run duration and axion phase, the low-mass limit is 9.5 times weaker; neutron-EDM data keep the lead.","key_machinery":"The load-bearing tool is the linear least-squares analysis (LLSA) of the time series, with two basis functions per assumed axion mass: one sine and one cosine at the axion frequency, transformed from celestial to laboratory coordinates. The LLSA reports the two quadrature amplitudes and their correlation; the crucial input is how that correlation behaves when the oscillation period exceeds the data span. The identity that carries the argument is coupling $\\propto \\Delta R/\\sin\\varphi$, which shows that the unknown phase must be treated as a free parameter rather than averaged out.","core_discovery":"On the paper's own terms, the original experiment's published constraint is not valid for axion oscillations whose period $\\tau$ is comparable to or longer than the data span $T = 2.642\\times10^6$ s. The correct treatment is a linear least-squares fit of the time series to two quadrature basis functions; when $\\tau/T>1$, these basis functions become nearly parallel or anti-correlated, so the marginalized amplitude uncertainty grows as $\\tau/(\\pi T)$. At the left-hand edge of the original Fig. 3 this growth is a factor of 6.0, and the supplementary analysis's phase averaging adds a further error because the coupling scales as $1/\\sin\\varphi$, not as the average of $|\\sin\\varphi|$. The combination means the low-mass exclusion limit should be 9.5 times higher than plotted, leaving the neutron-EDM bound about 15% more constraining.","pith_inferences":["If the comment is right, other axion-dark-matter searches that marginalized over the axion phase by averaging $|\\sin\\varphi|$ may need to revisit their long-period limits with a free-phase fit.","A testable extension: simulate the original run's sampling and noise, fit with LLSA for a grid of $\\tau/T$ values, and verify that the limit tracks $1/\\sin\\varphi$ and $\\tau/(\\pi T)$ in the long-period regime.","The comment's logic implies that for periods beyond the run length, extending the data span improves sensitivity faster than reducing per-measurement noise, a point the authors leave implicit."],"forward_implications":["The published low-mass edge of the original exclusion curve is not valid as stated.","A least-squares reanalysis would move that edge upward by a factor of 9.5.","The corrected limit is about 15% less constraining than the neutron-EDM limit, which therefore remains the strongest bound in this mass range.","The same sensitivity degradation applies to any oscillation search whose period exceeds the observation span, so finite run duration must appear explicitly in the error budget.","The short-period side of the original curve is assumed correct and unaffected."],"supporting_citations":[{"why":"Supplies the experimental data set, the run duration, and the Fig. 3 exclusion curve being criticized.","marker":"[1]"},{"why":"Earlier work cited as showing the expected loss of sensitivity for oscillations with periods comparable to the data span.","marker":"[2]"},{"why":"Another earlier search cited as corroborating the same long-period sensitivity reduction.","marker":"[3]"},{"why":"Provides the neutron-EDM bound against which the corrected limit is compared.","marker":"[4]"},{"why":"Provides the linear least-squares analysis method with quadrature basis functions that the comment uses to derive the correct scaling.","marker":"[5]"}],"fun_headline_variants":["Axion limit overstated by 9.5x due to data-length error","Corrected axion search: limit 9.5 times weaker","Dark matter axion bound revised: factor 9.5 too strong","Axion reanalysis: neutron-EDM bound still wins by 15%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes the unknown axion phase should be fitted as a free parameter, with uncertainty scaling as $1/\\sin\\varphi$; if the original paper's phase-averaging were a valid statistical construction, the factor 9.5 would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Axion limit overstated by 9.5x due to data-length error","Corrected axion search: limit 9.5 times weaker","Dark matter axion bound revised: factor 9.5 too strong","Axion reanalysis: neutron-EDM bound still wins by 15%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000231,"raw_usage":{"total_tokens":1402,"prompt_tokens":777,"completion_tokens":625,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":393,"completion_tokens_details":{"reasoning_tokens":544}},"tokens_in":393,"tokens_out":625,"duration_ms":5801,"temperature":1.0,"reasoning_tokens":544,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:22:40.697217+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Reanalyze the original $2.642\\times10^6$ s time series using the linear least-squares fit described here, leaving the axion phase free, and read the upper limit at the left-hand edge of the period range; if the limit moves up by less than a factor of about 9.5 relative to the published curve, the comment's central number is wrong. A cheaper check is to simulate a pure-noise data set with the same duration and sampling and confirm that the marginalized amplitude error scales as $\\tau/(\\pi T)$ once $\\tau>T$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the experimental data set, the run duration, and the Fig. 3 exclusion curve being criticized."},{"cited_title":"S12-S17) in the supplementary material","cited_arxiv_id":null,"evidence_quote":"Earlier work cited as showing the expected loss of sensitivity for oscillations with periods comparable to the data span."},{"cited_title":"Van Tilburg, N","cited_arxiv_id":null,"evidence_quote":"Another earlier search cited as corroborating the same long-period sensitivity reduction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the neutron-EDM bound against which the corrected limit is compared."},{"cited_title":"Abel et al","cited_arxiv_id":null,"evidence_quote":"Provides the linear least-squares analysis method with quadrature basis functions that the comment uses to derive the correct scaling."}],"review_version":1}