{"id":"5781f402-0df6-4f18-83d4-baf54bf78364","arxiv_id":"2501.08117","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Reanalysis of a comagnetometer bound on Lorentz violation yields the first laboratory limits on axion-nucleon coupling for axion masses below 10^-22 eV.","lead":"This paper reinterprets a 2014 comagnetometer measurement of Lorentz and CPT violation as a dark matter search, setting new limits on how strongly ultralight axions couple to nucleons. The derived bounds cover a mass range that laboratory experiments had not previously probed and are strong enough to beat the famous supernova SN1987A cooling limit.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Synthetic data, not a true reanalysis: Eq. (10) fits random Rice-distributed samples with a Gaussian |sin| likelihood, so the claimed 3-orders-of-magnitude limits are not established.","rationale":"The reader's weakest assumption was the phase marginalization, which the authors themselves flag as potentially overstringent in extreme cases. My concern is more fundamental: the paper does not actually reanalyze the original data. The likelihood in Eq. (10) is applied to random samples from a Rice distribution, not to the measured time series from Ref. [50]. This is stated explicitly in the text. Consequently, the abstract's phrase 'reanalyzing data' is misleading, and the derived limits are better described as projections from a synthetic calibration of the published B⊥ bound. The statistical model also has an internal inconsistency: Gaussian noise is assumed for quantities described as Rice-distributed, and the signal is modeled with |sin(...)| rather than the signed sine that enters the underlying Hamiltonian. Both issues disproportionately affect the low-mass end, where the paper claims the greatest improvement. If the concrete test shows that the synthetic likelihood faithfully reproduces a proper analysis of the real data, the paper could be salvaged as a reinterpretation, but as written the central claim is not supported. The phase issue, while acknowledged, would be a secondary caveat; the data-provenance issue is not acknowledged and invalidates the 'reanalysis' framing.","tokens_in":10378,"tokens_out":9335,"duration_ms":101449,"concrete_test":"Obtain the actual 3He-129Xe comagnetometer time series from Ref. [50] (or the authors' fitting code/data) and fit the signed axion model g sin(2πνa ti + φ) with the correct likelihood appropriate to the raw data. Then compute 68% upper limits on gaNN as in Fig. 3, both with and without phase marginalization. If the real-data limits differ from the synthetic limits by more than a factor of two, especially for ma ≲ 10^-22 eV, the 'reanalysis' central claim fails. As an internal consistency check, analytically propagate the published B⊥ bound to the slowly varying sidereal amplitude and compare the resulting phase-dependent gaNN bounds with Fig. 2.","verdict_should_be":"REJECT","load_bearing_attack":"The central claim requires that limits are derived by reanalyzing the actual 3He-129Xe comagnetometer data of Ref. [50]. In 'Derived constraints', however, the analysis does not use the original time series. The text states that 'bi are random samples from the generalized Rice distribution that the equatorial component follows [51]' and inserts these synthetic samples into Eq. (10), a Gaussian likelihood with mean g|sin(2πνa ti + φ)|. This raises two unaddressed problems. First, a Gaussian likelihood is misspecified for Rice-distributed data; the correct likelihood for the magnitude of a complex Gaussian amplitude is Rician. Second, the physical axion-induced frequency shift is proportional to sin(ωt+φ), not |sin(...)|; rectifying the signal can substantially bias amplitude estimates, especially in the low-frequency regime where the paper claims the largest improvement. More fundamentally, the published bound B⊥<8.4×10^-34 GeV at 68% CL from Ref. [50] is a fitted sidereal amplitude, not N≈3×10^5 independent Gaussian samples of the equatorial component. Without access to the original data or proof that the simulated likelihood reproduces the experiment's noise properties, the claim of 'reanalyzed data' is unsupported. The 3-orders-of-magnitude improvement over previous limits may be an artifact of the synthetic mapping rather than a property of the measurement. The phase-marginalization caveat acknowledged in the conclusion is real but secondary; the data-provenance and statistical-model issue is more fundamental and is not disclosed in the abstract.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to derive new limits on the axion-nucleon coupling over the mass range 10^-24 <= m_a <= 5e-21 eV by reanalyzing data from a 3He-129Xe comagnetometer measurement of Lorentz and CPT violation (Ref. [50]). The analysis in the 'Derived constraints' section does not use the original time series; instead, it generates synthetic samples from a generalized Rice distribution, fits them with a Gaussian likelihood using a |sin| template, and marginalizes over the unknown axion field phase. The authors report improvements of more than three orders of magnitude over previous laboratory limits, the first laboratory limits exceeding SN1987A cooling bounds in part of the mass range, and similar constraints for quadratic wind coupling and dark photon couplings.","tokens_in":10658,"tokens_out":7091,"duration_ms":67687,"significance":"If the reported limits were valid, they would constitute a significant advance: they would extend laboratory axion-nucleon constraints to masses below 10^-22 eV for the first time and surpass astrophysical bounds in the range 10^-22 to 5e-21 eV, with direct implications for the proposed ESS neutron-beam search. However, the central result currently rests on a statistical procedure that is misspecified and on simulated rather than actual data, so the claimed exclusion is not established. The paper does not provide reproducible code or the referenced supplemental material, and the primary figures present a single phase-marginalized curve with an acknowledged phase sensitivity. The significance is therefore conditional on a substantial revision that either uses the real data or reframes the results as a projection from the published bound.","major_comments":[{"comment":"The likelihood in Eq. (10) is evaluated on b_i defined as 'random samples from the generalized Rice distribution that the equatorial component follows,' not on the actual measured time series from Ref. [50]. The abstract and title say the limits come from 'reanalyzed data,' but the analysis uses simulated data. The published bound B_perp < 8.4e-34 GeV is a single fitted sidereal amplitude, not N ~ 3e5 independent measurements; the paper does not demonstrate that the Rice-distributed samples reproduce the experiment's noise, correlations, or systematics. As it stands, the limits in Figs. 3-5 are projections from a synthetic-data model, and the central claim of new limits from reanalyzed data is unsupported.","section":"Derived constraints (Eq. (10) and surrounding text)"},{"comment":"The signal template is written as |sin(2*pi*nu_a*t + phi)|, but the physical effective field in Eq. (8) is proportional to sin(2*pi*nu_a*t + phi), with no absolute value. Rectifying the sinusoid changes the mean of the signal and biases the amplitude estimator in Eq. (11), most severely at low frequencies where the paper claims the largest improvements. The authors do not justify the absolute value or quantify this bias.","section":"Eq. (10)"},{"comment":"The likelihood in Eq. (10) is Gaussian in b_i, yet b_i are explicitly drawn from a generalized Rice distribution. For magnitude or envelope data, the correct likelihood is Rician; using a Gaussian likelihood leads to incorrect confidence intervals even if the template were correct. No argument is given for why the Gaussian approximation is valid for N ~ 3e5 samples, and the misspecification directly affects the quoted 68% bounds.","section":"Eq. (10) and Eq. (11)"},{"comment":"The 68% upper limits in Fig. 3 are obtained by marginalizing over a uniform prior on the phase phi, and the authors state in the conclusion that for phi near 0 or pi the limits for m_a < 10^-21 eV 'may become overly stringent.' Since the phase is unknowable a priori, a single marginalized curve does not constitute a valid exclusion for all phase values; the phase-dependent results of Fig. 2 should be the primary presentation, or a conservative worst-case envelope should be quoted. The abstract and Fig. 3 present the marginalized curve without this caveat.","section":"Conclusion and discussion; Fig. 3"},{"comment":"The limit B_perp from Ref. [50] is derived by demodulating at the sidereal frequency Omega. The paper extends the bound to axion frequencies up to 2*pi*nu_a ~ 0.1*Omega without providing a quantitative transfer function or leakage analysis for the comagnetometer's frequency response. The statement that the bound applies 'when 2*pi*nu_a is significantly less than Omega, for instance, by an order of magnitude' is an assumption, not a derivation; the validity of the 5e-21 eV endpoint is therefore not established.","section":"Basic idea / Derived constraints, m_a <= 5e-21 eV"}],"minor_comments":[{"comment":"Ref. [51] is cited as 'Supplemental Material' but the supplement is not included in the arXiv posting; please include it so the Rice distribution parameters and the averaging method can be checked.","section":"References"},{"comment":"The conclusion states that the new limits 'exceed the projected reach' of the ESS proposal, while the abstract says they are 'nearly equivalent' to that projection; please reconcile these statements.","section":"Conclusion and discussion"},{"comment":"The figure captions do not specify the 68% confidence level, the assumed local dark matter density, or the exact definition of the shaded exclusion regions; please add these details for reproducibility.","section":"Figures 2-5"}],"recommendation":"major_revision","confidential_remarks":"The synthetic-data issue is the main concern. If the authors cannot obtain the actual time series, they should reframe the paper as a reinterpretation of the published bound, which would change the significance claims. I recommend major revision rather than rejection because the underlying mapping is plausible and correctable, but the current manuscript's title and abstract overstate what is demonstrated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's headline claim does not survive contact with its own methods. The authors say they 'reanalyzed' the 3He-129Xe comagnetometer data of Ref. [50], but the likelihood in Eq. (10) is applied to 'bi are random samples from the generalized Rice distribution' — i.e., synthetic draws, not the measured time series. That alone makes the 3-orders-of-magnitude improvement over previous limits unsupported. On top of that, the signal model uses |sin(2πνt+φ)|, but the physical axion-induced frequency shift is sin(2πνt+φ) without the absolute value. Rectifying the signal biases amplitude estimates; at the low-frequency end where the claimed gain is largest, this matters a lot. The Gaussian likelihood is also misspecified for Rice-distributed data. So the paper is not a reanalysis; it is a simulation with a published bound used only for calibration.\n\nWhat is genuinely new, and worth credit, is the idea to take the published sidereal bound B⊥<8.4×10^-34 GeV and map it onto axion-nucleon couplings for ma below 10^-22 eV. That mapping is plausible in principle, and the paper properly flags the phase-marginalization caveat in the conclusion — they admit the limits can become 'overly stringent' for unlucky phases. The comparison with the proposed ESS experiment is also useful.\n\nThe soft spots are not minor. (1) Data provenance: no actual data are used, and no demonstration that the synthetic Rice likelihood reproduces the experiment's noise properties or the published bound. (2) Signal model: |sin| is physically wrong. (3) Phase marginalization: the 68% CL quoted in Fig. 3 mixes data- and phase-uncertainty in a way that is not a true confidence interval; a conservative result would be a function of φ. (4) The abstract overstates robustness, omitting the synthetic-data issue entirely.\n\nThe citation pattern is fine, and there is no self-citation loop. The paper is honest in its conclusion about the phase issue, which suggests the authors are not trying to hide the main caveat. But the synthetic-data problem is hidden in the text's wording.\n\nI would send this back for major revision, not desk reject, because the underlying idea is salvageable. To make the claim, the authors need to either get access to the original time series and fit a proper sine model, or clearly present the result as a projection/sensitivity estimate based on the published bound, with no 'reanalysis' language. The quadratic-wind and dark-photon constraints suffer the same issue.\n\nFor a reader: this is a cautionary example of how 'reinterpretation' can go wrong when the underlying data are simulated. I would not cite the limits. Bring it to reading group only to discuss what counts as a reanalysis.\n\nRecommendation: serious referee, but expect heavy revision; if data access is impossible, the paper should be reframed as a sensitivity study.","headline":"A promising idea undermined by a synthetic-data 'reanalysis': the limits are not established as written.","tokens_in":11211,"tokens_out":3448,"would_cite":false,"duration_ms":34027,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Reanalyzing a 129Xe+3He co-magnetometer run that originally bounded Lorentz/CPT violation sets the strongest laboratory limits on the axion-nucleon coupling in the 10^-24 to 5e-21 eV mass range.","keywords":["axion dark matter","axion-nucleon coupling","ultralight axions","Lorentz violation","CPT violation","co-magnetometer","nuclear spin precession","SN1987A cooling"],"falsifier":"Pin down the local axion phase at these frequencies, or rerun the analysis quoting the worst-case phase rather than the phase average: if the phase is near 0 or π, the Fig. 3 exclusion region for ma < $10^{-21}$ eV should shrink or disappear, so the claimed limits would be contradicted by a measurement with independent sensitivity in the same mass range.","tokens_in":10120,"feed_emoji":"🌌","tokens_out":3968,"duration_ms":37223,"temperature":0.7,"pith_summary":"This paper takes a published dataset from a dual-species 129Xe+3He co-magnetometer that was originally used to bound Lorentz and CPT violation and reinterprets its bound on an equatorial effective magnetic field as a limit on ultralight axion dark matter coupled to nucleons. It claims the first laboratory constraints on the axion-nucleon coupling for axion masses below $10^{-22}$ eV, and for $10^{-22}$ to 5×$10^{-21}$ eV limits more than three orders of magnitude stronger than previous lab searches, exceeding supernova SN1987A cooling bounds for the first time in the laboratory. The analysis pays explicit attention to the local phase of the axion field, which matters at these ultralow frequencies, and marginalizes over it under a uniform prior. The same reanalysis also yields much stronger constraints on a quadratic wind coupling and on dark-photon-nucleon interactions.","feed_headline":"Reanalyzed magnetometer data tighten axion limits 1000-fold","feed_subtitle":"First laboratory constraints below 10^-22 eV, surpassing supernova SN1987A bounds for 10^-22 to 5e-21 eV.","key_machinery":"The load-bearing object is the effective magnetic field Ba = (2gaNN/γ)√(2ħcρa) sin(2πνa t + ϕ) va induced on nuclear spins by the coherently oscillating axion dark matter field, whose equatorial component is modulated at the sidereal frequency as the Earth rotates. The previous comagnetometer bound on that equatorial component supplies the data, and a Gaussian likelihood over roughly 3×$10^{5}$ samples gives the estimator ĝ for each mass and phase; the reported limit is the 68% upper quantile after phase marginalization.","core_discovery":"The paper's central claim is that data from a $10^{6}$ second run of a 129Xe+3He free-spin-precession co-magnetometer, with the upper limit B⊥ < 8.4×$10^{-34}$ GeV on the sidereal-modulated equatorial component of an effective magnetic field at 68% CL, translate into new upper limits on the axion-nucleon coupling gaNN over $10^{-24}$ ≤ ma ≤ 5×$10^{-21}$ eV. For axion masses below $10^{-22}$ eV these are the first laboratory constraints; in the overlapping band $10^{-22}$ to 5×$10^{-21}$ eV they improve on the PSI neutron EDM and NMR comagnetometer limits by more than three orders of magnitude and for the first time beat the SN1987A cooling bounds. The limits are derived by a likelihood analysis of time-series data that includes the unknown axion phase ϕ, with the 68% upper bound obtained after marginalizing ϕ uniformly over [0,2π].","pith_inferences":["If the phase-averaging assumption is conservative, the same data could be folded to give phase-dependent exclusion plots; publishing worst-case phase bounds would let other experiments compare in a prior-free way.","The method should carry over to longer co-magnetometer datasets or multiple runs, where partial oscillations of lower-frequency axions would reduce phase sensitivity and push limits to even lower masses.","The same reinterpretation logic may be applied to other Lorentz/CPT-violation searches with different species, potentially covering mass ranges between 5×10^-21 eV and the sidereal frequency cutoff.","Because the constraints on gdMDM mirror gaNN, improvements here automatically tighten dark-photon magnetic coupling limits without additional analysis."],"forward_implications":["The axion-nucleon coupling in 10^-22 ≤ ma ≤ 5×10^-21 eV is now more tightly constrained by a tabletop lab measurement than by SN1987A cooling, closing a long-standing gap between lab and astrophysics.","The first lab limits below 10^-22 eV make the regime inaccessible to other searches testable already with existing data.","The derived sensitivity is comparable to the projected one-year HIBEAM neutron-beam experiment at ESS, so the reanalysis demonstrates that current datasets can reach design sensitivity without new apparatus.","The quadratic wind coupling gquad is bounded about two orders of magnitude more tightly than the best previous lab result and more than four orders better than SN1987A."],"supporting_citations":[{"why":"Supplies the reanalyzed dataset and the 68% CL bound B⊥ < 8.4×10^-34 GeV on the equatorial effective field.","marker":"[50]"},{"why":"PSI neutron EDM experiment whose previous laboratory limits on gaNN are surpassed by more than three orders of magnitude.","marker":"[36]"},{"why":"NMR-based comagnetometer experiment that set the prior best laboratory bounds and introduced the phase approximation approach.","marker":"[19]"},{"why":"ESS HIBEAM neutron-beam proposal whose projected sensitivity the derived limits are comparable to.","marker":"[39]"},{"why":"SN1987A cooling astrophysical bounds that this work exceeds for the first time in the laboratory.","marker":"[52, 53]"},{"why":"Supplemental material describing the likelihood analysis and the averaging method used to extract limits.","marker":"[51]"}],"fun_headline_variants":["First lab axion limits below 10^-22 eV from reanalysis","Reanalysis sets 1000x tighter axion-nucleon bounds","Axion limits beat SN1987A for first time in lab data","Lab data reanalysis yields best axion-nucleon limits yet","First lab bounds on axion-nucleon coupling below 1e-22 eV"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported limits assume the unknown local phase of the ultralight axion field is uniformly distributed over [0,2π]; if the real phase is near 0 or π for masses below $10^{-21}$ eV, the bounds become overly stringent and would not be valid exclusions.","fun_headline_variants_meta":{"raw":{"variants":["First lab axion limits below 10^-22 eV from reanalysis","Reanalysis sets 1000x tighter axion-nucleon bounds","Axion limits beat SN1987A for first time in lab data","Lab data reanalysis yields best axion-nucleon limits yet","First lab bounds on axion-nucleon coupling below 1e-22 eV"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000999,"raw_usage":{"total_tokens":4237,"prompt_tokens":964,"completion_tokens":3273,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":3175}},"tokens_in":580,"tokens_out":3273,"duration_ms":23932,"temperature":1.0,"reasoning_tokens":3175,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:30:21.277818+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Pin down the local axion phase at these frequencies, or rerun the analysis quoting the worst-case phase rather than the phase average: if the phase is near 0 or π, the Fig. 3 exclusion region for ma < $10^{-21}$ eV should shrink or disappear, so the claimed limits would be contradicted by a measurement with independent sensitivity in the same mass range.","supporting_citations":[{"cited_title":"Allmendinger, W","cited_arxiv_id":null,"evidence_quote":"Supplies the reanalyzed dataset and the 68% CL bound B⊥ < 8.4×10^-34 GeV on the equatorial effective field."},{"cited_title":"Abel et al., Search for Axionlike Dark Matter through Nuclear Spin Precession in Electric and Magnetic Fields, Phys","cited_arxiv_id":null,"evidence_quote":"PSI neutron EDM experiment whose previous laboratory limits on gaNN are surpassed by more than three orders of magnitude."},{"cited_title":"Wu et al., Search for Axionlike Dark Matter with a Liquid-State Nuclear Spin Comagnetometer, Phys","cited_arxiv_id":null,"evidence_quote":"NMR-based comagnetometer experiment that set the prior best laboratory bounds and introduced the phase approximation approach."},{"cited_title":"Fierlinger, M","cited_arxiv_id":null,"evidence_quote":"ESS HIBEAM neutron-beam proposal whose projected sensitivity the derived limits are comparable to."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplemental material describing the likelihood analysis and the averaging method used to extract limits."}],"review_version":1}