{"id":"30f3255b-90c3-4bc0-a789-8bbac634b813","arxiv_id":"2507.04344","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A haloscope search with an 8-cell cavity and a quantum-noise-limited amplifier excludes axion-photon couplings above 1.2 × 10^-14 GeV^-1 for axion masses 24.11-24.57 micro-eV at 90% confidence.","lead":"Scientists searched for axion dark matter around a mass of 24.4 micro-electronvolts using an eight-cell microwave cavity in a strong magnetic field, and found no signal. They set the strongest limit yet on the axion's coupling to photons in this mass window, approaching but not reaching the KSVZ model prediction.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unexplained negative excesses in 5.87–5.88 GHz and the ad hoc substitution of supplementary data call into question the 90% CL coverage of the quoted limit in that band.","rationale":"The paper's central claim is a numerical exclusion limit with a stated confidence level. For that claim to hold, the statistical noise model must be correct in every frequency bin contributing to the limit. The manuscript itself identifies strong negative excesses in 5.87–5.88 GHz whose origin is unknown, and the analysis excises and replaces those bins. This is precisely the condition the reader singled out as the weakest assumption. My reading agrees: the ad hoc handling of an unexplained non-Gaussian feature is load-bearing because the limit's stated 90% CL coverage in that band depends on the replacement data having the same noise properties as the original scan, which is not demonstrated. The observation of 85 candidate bins above a 3.47σ threshold after normalization, when roughly one would be expected, reinforces that the noise is not fully understood. A coverage test with injected signals would directly settle whether the quoted limit is valid. Since this is an addressable but serious issue rather than a demonstrated error, the reader's CONDITIONAL verdict remains appropriate, with no change.","tokens_in":8952,"tokens_out":10167,"duration_ms":117694,"concrete_test":"Perform a signal-injection coverage test restricted to the 5.87–5.88 GHz band. Inject synthetic KSVZ axion signals with known gaγγ values into the raw spectra before preprocessing, then run the published analysis pipeline (including the negative-excess exclusion and substitution with the supplementary scan) and determine the fraction of trials in which the injected coupling is excluded at 90% CL. Repeat without the exclusion/substitution step, using the original preprocessed data. If either the coverage differs significantly from 90% (e.g., below 85%) or the two limits differ by more than 20%, the quoted limit in this band is not robust and the anomalous features must be understood before a reliable limit can be stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a 90% CL exclusion limit, which presupposes a well-understood Gaussian noise model over the full scanned band. The data-analysis section states: \"We observed several strong negative excesses deviating from the expected Gaussian distribution, particularly in the 5.87–5.88 GHz range... Although the origin of these features remains unknown, a follow-up supplementary scan with twice the statistics showed no such excesses. Accordingly, we excluded the anomalous bins from the original preprocessed spectra and substituted them with data from the supplementary scan.\" This is an explicit, unresolved limitation. The quoted limit in that band is therefore not derived from a single dataset with known noise, but from a composite spectrum assembled by excising bins whose behavior contradicts the Gaussian assumption and replacing them with data from a different run. If the negative features are a real instrumental response that also contaminates (even weakly) the supplementary scan, or if the two runs differ in gain, system noise, or baseline shape, the effective noise distribution in 5.87–5.88 GHz is not the one used to compute the 90% CL, and the coverage of the limit in this band is unvalidated. The concern is not merely cosmetic: after normalization to N(0,1), the grandspectrum still produced 85 bins above a 3.47σ threshold, far more than the O(1) expected for pure Gaussian noise, confirming that the noise model has unexplained tails. The statement that no candidate was persistent does not repair the underlying noise-model uncertainty, because the limit is derived from the same noisy spectrum.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a haloscope search for axion dark matter near 5.9 GHz using an 8-cell copper cavity in an 8 T magnet, read out with a flux-driven Josephson parametric amplifier and a sideband-summing analysis. The scan covers 5.83–5.94 GHz, finds no persistent candidate, and sets a 90% CL upper limit on the axion-photon coupling of 1.2 × 10^−14 GeV^−1 over the axion mass range 24.11–24.57 μeV, claimed as the most stringent limit in this range and approaching the KSVZ benchmark. The analysis includes in-situ noise-temperature calibrations, Savitzky–Golay baseline removal, sideband correlation, grandspectrum construction, and follow-up scans of candidate bins.","tokens_in":9218,"tokens_out":4041,"duration_ms":43589,"significance":"If the limit is robust, this is a valuable null result in a mass range that is difficult to reach with conventional single-cell cavities, and the 8-cell resonator is a useful technical advance. The paper provides detailed experimental characterization: in-situ JPA gain monitoring, 24-hour stability checks, injection-based measurement of the Savitzky–Golay SNR efficiency (84%), and explicit follow-up of all threshold-crossing candidates. However, the statistical interpretation is weakened by untreated non-Gaussian tails and by the ad hoc substitution of data in the 5.87–5.88 GHz band; these issues must be resolved before the 90% CL claim can be accepted as stated.","major_comments":[{"comment":"The exclusion of anomalous bins in the 5.87–5.88 GHz range and their replacement with data from a supplementary scan is not accompanied by a demonstrated instrumental mechanism or by a quantitative check that the two datasets have identical noise properties. Because the quoted result is a 90% CL exclusion, the effective noise distribution in this band must be validated: please either explain the anomalies as understood instrumental effects, or present a conservative limit that does not rely on the substitution (for example, a limit computed with the original bins included, or with an enlarged noise model in that band). As written, the coverage of the claimed limit in 5.87–5.88 GHz is unvalidated.","section":"Data analysis (paragraph beginning 'We observed several strong negative excesses...')"},{"comment":"The paper states that after normalization to N(0,1) the grandspectrum still produced 85 bins above the 3.47σ threshold, far more than the O(1) expected for a pure Gaussian, and that follow-up scans showed none was persistent. This indicates that the noise model used to define the 90% CL threshold has unexplained tails. The expected number of threshold crossings under the null hypothesis, including the effect of Savitzky–Golay bin-to-bin correlations, should be computed and compared with the observed 85 before the threshold can be regarded as calibrated. If the excess is due to correlations or residual systematics, that must be quantified; otherwise the 90% coverage of the limit is not established.","section":"Data analysis (hypothesis testing and grandspectrum)"},{"comment":"No systematic uncertainty budget is provided for the final upper limit. The limit should propagate the uncertainties in the system noise temperature, the JPA gain and its stability, the form-factor degradation (quoted as below 1%), the Savitzky–Golay SNR efficiency (84%), the sideband-summing weights and correlation coefficient, and the grandspectrum normalization. Even if the total systematic is negligible relative to the statistical power, a quantitative statement is required to support the single quoted value of 1.2 × 10^−14 GeV^−1.","section":"Data analysis and result (Fig. 4)"}],"minor_comments":[{"comment":"In the conclusion, 'we exclude axion-photon couplings gaγγ ≳ 1.2 × 10^−14 GeV^−1' uses the wrong inequality direction; the abstract and Fig. 4 correctly state the limit as an upper bound ('down to' or '<').","section":"Conclusion"},{"comment":"The decomposition 'Tsys = Tcav + Trcv' is immediately followed by definitions using Ton and Toff without connecting Trcv to those quantities; please clarify the notation so that Eq. (2) is self-contained.","section":"System noise temperature paragraph"},{"comment":"The statement that 'The distribution of δ was consistent with the standard normal distribution' is difficult to reconcile with the later reported grandspectrum width of 0.85 and the 85 threshold-crossing bins; please specify the quantitative consistency test used and elaborate on how these observations are consistent.","section":"Preprocessing paragraph"},{"comment":"The author list of Ref. [26] contains 'i. m. c. b. u. Kutlu', which appears to be a metadata corruption of the author name 'C. Kutlu'; please correct it.","section":"Reference [26]"},{"comment":"The text contains 'with an built-in down-converter' (should be 'a built-in'), and the Fig. 3 caption uses 'the idle signals' where 'idler signals' is the standard terminology for the JPA-generated intermodulation tone.","section":"Experimental setup and Fig. 3 caption"}],"recommendation":"major_revision","confidential_remarks":"The experimental work appears technically sound and the central issue is addressable with additional analysis; I recommend major revision rather than rejection because the missing noise-model validation in the 5.87–5.88 GHz band and the unexplained threshold-crossing rate are load-bearing for the quoted 90% CL but can plausibly be fixed within the scope of the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"New result: a haloscope exclusion in the 24.11–24.57 µeV window down to gaγγ ≈ 1.2×10^-14 GeV^-1 at 90% CL, the best limit there and approaching the KSVZ line. The experiment side is described well: 8-cell cavity with measured form factor, JPA near the quantum limit, sideband-summing with the correlation coefficient estimated from data, in-situ noise temperature, injected-signal efficiency of 84%, and follow-up scans for every candidate. That is a serious, careful run.\n\nThe soft spots are real but mostly fixable. The most important is the 5.87–5.88 GHz band. The paper reports strong negative excesses of unknown origin, excises those bins from the original spectra, and substitutes data from a supplementary scan. That turns the final spectrum into a composite whose noise properties in that band are assumed, not measured. The stress-test note is right that the 90% CL coverage there is unvalidated. And the fact that the normalized grandspectrum still produced 85 bins above the 3.47σ threshold — far more than expected for Gaussian noise — tells you the noise model has unexplained tails. Calling them statistical fluctuations and moving on does not repair the coverage statement.\n\nThe hypothesis-testing paragraph is also confusing: 'null hypothesis that assumes the presence of a KSVZ axion signal with SNR = 5' is not what a null hypothesis is. I suspect they mean a threshold chosen for 90% CL, but the wording needs fixing. There is also no explicit systematic uncertainty budget for the final limit; the limit might be dominated by statistics, but they should say so.\n\nNone of this is load-bearing. The central measurement is a new constraint in a region that had no strong limits, and the techniques are sensible. I would send it to a serious referee. The right referee will ask for (1) a demonstration that the substituted band has the same noise model, or a systematic uncertainty attached to that exclusion, (2) a correct hypothesis-testing statement, and (3) a systematic budget. Those are addressable.\n\nWho is this for? Axion dark matter experimentalists and theory groups checking KSVZ/DFSZ parameter space. It deserves a proper review rather than a desk rejection.","headline":"Solid new haloscope limit near 5.9 GHz, but the unexplained 5.87–5.88 GHz excesses and the odd hypothesis-testing language need work before the 90% CL is trustworthy in that band.","tokens_in":9872,"tokens_out":2533,"would_cite":true,"duration_ms":25950,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["14.80.Va"],"model":"deepseek-v4-flash","headline":"No axions found: 8-cell cavity sets strictest limit near 5.9 GHz","keywords":["axion dark matter","haloscope","8-cell cavity","Josephson parametric amplifier","quantum-limited readout","sideband summing","KSVZ axion","high-frequency cavity search"],"falsifier":"Run a dedicated high-statistics scan with independent calibration over the 5.87–5.88 GHz band; if the negative excesses reappear at a level inconsistent with the noise model, the exclusion limit in that band is not valid.","tokens_in":8753,"feed_emoji":"📡","tokens_out":8054,"duration_ms":72049,"temperature":0.7,"pith_summary":"The paper reports a haloscope search for axion dark matter in the 5.83–5.94 GHz band, corresponding to axion masses of 24.11–24.57 µeV. It uses an eight-cell microwave cavity to reach frequencies about three times higher than a conventional single-cell cavity in the same magnet bore, and a flux-driven Josephson parametric amplifier operating near the quantum noise limit to read out the converted photons. With no statistically significant excess observed, it excludes axion-photon couplings $g_{a\\gamma\\gamma}$ down to $1.2 \\times 10^{-14}\\,\\mathrm{GeV}^{-1}$ at 90% confidence, the most stringent limit in this mass range to date. This approaches the KSVZ benchmark prediction, so a KSVZ axion in this window at the upper end of the allowed coupling would have been detectable.","feed_headline":"No axions found: 8-cell cavity sets strictest limit near 5.9 GHz","feed_subtitle":"An 8-cell cavity plus quantum-limited amplifier rules out axion couplings to 1.2e-14 GeV-1 at 90% confidence.","key_machinery":"The central object is the eight-cell microwave cavity, a cylindrical resonator split into eight coupled cells that supports a TM$_{010}$-like mode near 6.0 GHz with a total detection volume of 3.1 L and an average magnetic field of 6.97 T, reaching a frequency about three times that of a single-cell cavity in the same bore. The second key component is the flux-driven Josephson parametric amplifier, a SQUID-terminated coplanar waveguide that provides phase-preserving gain near the quantum noise limit and generates an idler tone at $f_i = f_p - f_s$; the analysis then uses sideband summing, Eq. (3), to coherently combine the signal and idler power excesses $\\delta_k$ and $\\delta_{-k}$ with optimized weights and a measured correlation coefficient $\\rho$. The form factor, tuned by rotating alumina rods in each cell, ranges from 0.5 to 0.7, with fabrication asymmetries mitigated to below 1% degradation. The machinery operates together to convert axion-photon power, expressed by Eq. (1), into a measurable excess power spectrum.","core_discovery":"The central claim is that no axion dark matter signal exists in the axion mass range 24.11–24.57 µeV for couplings above $g_{a\\gamma\\gamma} = 1.2 \\times 10^{-14}\\,\\mathrm{GeV}^{-1}$ at 90% confidence. The experiment demonstrates that an eight-cell cavity architecture, which supports a TM$_{010}$-like mode near 5.9 GHz while preserving a 3.1 L detection volume in an 8 T magnetic field, can sustain a haloscope search at frequencies roughly three times the reach of a single-cell cavity. Combined with a flux-driven Josephson parametric amplifier and a sideband-summing analysis that exploits the JPA's signal and idler modes, the search achieves system noise temperatures of 380–500 mK and an overall SNR efficiency of 84%, yielding the most stringent limit in this frequency range and sensitivity approaching the KSVZ benchmark. The null result is robust to follow-up scans of 85 candidate excesses, none of which persisted.","pith_inferences":["Extending the same 8-cell architecture with higher quality factors or larger volume could push the limit below the KSVZ line, allowing a definitive test of the model across a broader mass range.","The unexplained negative excesses in the 5.87–5.88 GHz band, present in two independent experiments, may point to a systematic tied to the cavity or amplifier configuration; identifying it could improve the robustness of future analyses.","The sideband-summing gain, though modest here, could become larger when the JPA idler noise is reduced or when the signal and idler have very different noise temperatures, making the technique more valuable in other searches.","The successful operation at 5.9 GHz suggests that similar multi-cell cavities could be pushed to even higher frequencies, for example with smaller cells or higher-order modes, to probe axion masses beyond 25 µeV."],"forward_implications":["The 8-cell cavity extends the reachable axion mass for a given magnet bore roughly threefold, opening higher-mass windows to haloscope searches.","The sideband-summing technique yields a consistent 4% SNR improvement over conventional analysis, which will be relevant wherever JPAs are used.","The 90% CL exclusion of $g_{a\\gamma\\gamma} < 1.2 \\times 10^{-14}\\,\\mathrm{GeV}^{-1}$ over 24.11–24.57 µeV is the most stringent limit in this mass range.","The achieved sensitivity approaches the KSVZ benchmark prediction, so the window at higher couplings is closed and only sub-benchmark couplings remain.","The null result constrains axion dark matter models with hadronic (KSVZ) couplings in this mass window."],"supporting_citations":[{"why":"Introduces the multi-cell cavity concept that extends the resonant frequency of a haloscope while preserving detection volume.","marker":"[16]"},{"why":"Demonstrates the multi-cell cavity in an axion search, providing the design basis for the 8-cell resonator used here.","marker":"[17]"},{"why":"Documents the tuning-rod optimization that keeps form-factor degradation below 1%, protecting the quoted sensitivity.","marker":"[19]"},{"why":"Presents the Josephson parametric amplifier design (SQUID-terminated resonator) that provides near-quantum-limited readout.","marker":"[20]"},{"why":"Characterizes the flux-driven JPA used in this experiment, establishing its gain and noise behavior.","marker":"[21]"},{"why":"This paper's companion search describes the sideband-summing technique (Appendix C) used to combine signal and idler bins and reports similar negative excesses.","marker":"[26]"},{"why":"Provides the dark matter signal lineshape used to construct the grand spectrum weights.","marker":"[27]"},{"why":"Reports the same kind of unexplained negative excesses in another haloscope, supporting the decision to treat them as artifacts.","marker":"[28]"},{"why":"Gives the analytical prediction for Savitzky-Golay filter SNR efficiency used to normalize the grand spectrum.","marker":"[29]"}],"fun_headline_variants":["8-cell haloscope rules out axions near 5.9 GHz","New best axion limit at 5.9 GHz from 8-cell cavity","No axion signal: tightest bound yet at 5.9 GHz","Quantum-limited search excludes KSVZ axions at 5.9 GHz","Eight-cell cavity sets stringent axion limit near 5.9 GHz"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quoted limit depends on the assumption that the unexplained negative excesses seen near 5.87–5.88 GHz are instrumental artifacts, not signals, and that the supplementary scan used to replace those bins has the same systematic behavior.","fun_headline_variants_meta":{"raw":{"variants":["8-cell haloscope rules out axions near 5.9 GHz","New best axion limit at 5.9 GHz from 8-cell cavity","No axion signal: tightest bound yet at 5.9 GHz","Quantum-limited search excludes KSVZ axions at 5.9 GHz","Eight-cell cavity sets stringent axion limit near 5.9 GHz"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000203,"raw_usage":{"total_tokens":1389,"prompt_tokens":949,"completion_tokens":440,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":565,"completion_tokens_details":{"reasoning_tokens":340}},"tokens_in":565,"tokens_out":440,"duration_ms":5322,"temperature":1.0,"reasoning_tokens":340,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:50:40.079058+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a dedicated high-statistics scan with independent calibration over the 5.87–5.88 GHz band; if the negative excesses reappear at a level inconsistent with the noise model, the exclusion limit in that band is not valid.","supporting_citations":[{"cited_title":"Jeong, S","cited_arxiv_id":null,"evidence_quote":"Documents the tuning-rod optimization that keeps form-factor degradation below 1%, protecting the quoted sensitivity."},{"cited_title":"Flux-driven Josephson parametric amplifier","cited_arxiv_id":"0808.1386","evidence_quote":"Presents the Josephson parametric amplifier design (SQUID-terminated resonator) that provides near-quantum-limited readout."},{"cited_title":"Characterization of a flux-driven Josephson parametric amplifier with near quantum-limited added noise for axion search experiments","cited_arxiv_id":"2101.08496","evidence_quote":"Characterizes the flux-driven JPA used in this experiment, establishing its gain and noise behavior."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This paper's companion search describes the sideband-summing technique (Appendix C) used to combine signal and idler bins and reports similar negative excesses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the dark matter signal lineshape used to construct the grand spectrum weights."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the analytical prediction for Savitzky-Golay filter SNR efficiency used to normalize the grand spectrum."}],"review_version":1}