{"id":"dd91c2eb-baec-4299-9804-68ba19145d42","arxiv_id":"2509.03423","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"ECHo-1k's 163Ho calorimetric spectrum yields m_nu_e < 15 eV/c^2 (90% credible interval), a factor of 1.8 improvement over the HOLMES limit.","lead":"The ECHo-1k experiment analyzed 200 million electron-capture decay events from embedded holmium-163 and set a new upper limit on the effective electron neutrino mass of 15 eV/c^2, nearly halving the previous best limit. The independently derived endpoint energy agrees with a Penning-trap measurement, supporting the calorimetric method.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 15 eV limit hinges on the ad hoc A2(E) tail model; model-mismatch systematics are not quantified, so the result is conditional.","rationale":"I read the paper's central claim as conditional on the A2 spectral model. The reader's weakest assumption identifies the same issue: the ad hoc A2(E) tail is not backed by a quantified systematic uncertainty, and the same data determine both the tail parameters and the neutrino-mass limit. My concern does not change the reader's verdict. The analysis is careful, the derived Q value agrees with the independent Penning-trap measurement, and the A1/A2 difference in the mass limit is small (16 vs 15 eV). However, the lack of a model-mismatch systematic and the absence of an independent tail cross-check mean the 15 eV limit should be regarded as conditional on the A2 parametrization. This is a standard model-uncertainty concern, not a fatal flaw, and does not call into question the authors' integrity or the basic experimental achievement.","tokens_in":12697,"tokens_out":6159,"duration_ms":69294,"concrete_test":"Simulate ECHo-1k-like spectra from the ab initio model of Braß & Haverkort [37] (or from a plausible variant with, e.g., a power-law or additional Lorentzian tail) with m_nu = 0 and with m_nu = 15 eV, using the quoted exposure, background, and sigma = 2.8 eV response. Refit with the exact A2/Stan pipeline. If the frequentist coverage of the 90%/95% credible intervals is not within, say, ±5%, or if the median reconstructed Q shifts by >4 eV, the reported limit is not robust to the A(E) parametrization. A second, cheaper check: fit the last 100 eV below Q alone with A2 and examine the residual chi2; large residuals would directly indicate tail misspecification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result m_nu_e < 15 eV (90% C.I.) comes from the A2 fit in Eq. (2), whose high-energy tail is a Lorentzian plus a phenomenological two-exponential SA(E) term. The SA parameters (kA, EA, E1, E2) are extracted from the same 2250-2750 eV spectrum that is then used, in the joint Bayesian fit, to constrain m_nu_e and Q. The neutrino-mass sensitivity lives in the last tens of eV below Q, which is precisely an extrapolation of the SA tail. The A1 vs A2 comparison is not an adequate systematic check: A1 is explicitly judged inappropriate at this statistics and requires narrow priors, and both parametrizations share the same data, so a common-mode tail error would affect both. The ab initio model [37] cannot currently be used to fit the tails, so there is no independent physics-based A(E). A misspecified tail can shift the endpoint shape and bias the m_nu_e posterior; the paper gives no uncertainty attached to this model choice.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The ECHo collaboration analyzes the endpoint region of a calorimetric 163Ho electron-capture spectrum containing about 200 million events acquired with the ECHo-1k detector array. Using a Bayesian fit with two analytical approximations for the atomic physics factor, A1(E) and A2(E), they extract the endpoint energy Q and the effective electron neutrino mass. With the A2 model they obtain Q = 2862(4) eV, in good agreement with the independent Penning-trap value Q = 2863.2(6) eV, and an upper limit m_nu_e < 15 eV/c^2 at 90% credible interval (18 eV at 95%). The A1 model gives compatible but less reliable results, requiring narrow priors. The paper reports a very low background rate of B = 9.1(1.3) x 10^-6 /eV/pixel/day and claims an improvement by almost a factor of two over the previous best limit from the HOLMES collaboration.","tokens_in":13018,"tokens_out":5306,"duration_ms":58633,"significance":"If the result is correct, this is the lowest direct limit on the effective electron neutrino mass from a calorimetric 163Ho experiment, and it strengthens the case for 163Ho-based neutrino mass searches. The strengths of the paper include a high-statistics spectrum, a carefully characterized detector and data-reduction chain, a low and well-measured background, and a fitted Q value that is consistent with an independent Penning-trap determination. The A1/A2 cross-check and the Monte Carlo study of per-pixel energy-resolution differences are useful consistency checks. The main limitation is that the central limit depends on the phenomenological A2(E) tail model, and the paper does not quantify the associated systematic uncertainty.","major_comments":[{"comment":"The claimed limit m_nu_e < 15 eV (90% C.I.) is obtained from a Bayesian fit using A2(E), whose high-energy tail is a Lorentzian plus a two-exponential term with parameters kA, EA, E1, E2 fitted from the same 2250-2750 eV spectrum that is later used to constrain m_nu_e and Q. The neutrino-mass sensitivity is concentrated in the last tens of eV below Q, i.e., in the extrapolated tail of this phenomenological term. The paper does not provide a systematic uncertainty associated with the choice of A(E). The A1 vs A2 comparison is not an adequate model-mismatch test: A1 is explicitly judged inappropriate at this statistics, and any common-mode tail error would affect both parametrizations. Please quantify the model systematic, for example by repeating the fit with alternative tail functions (power law, modified Lorentzian, or template based on the ab initio model [37]) and reporting the spread","section":"Main text, Eq. (2) and following paragraph"},{"comment":"The statement that the A1 and A2 results are compatible is weakened by the very different treatment of priors: A1 requires narrow priors constrained to grid uncertainties, and the authors state that A1 is not appropriate at the present statistics. Moreover, the A1 Q value is 2866.3(2.0) eV, which is less consistent with the Penning-trap value than the A2 result. Thus the A1 comparison does not bound the systematic error from the tail parametrization. In addition, the A2 grid fit used to initialize the A2 parameters assumes m_nu_e = 0 and fixes Q to the Penning-trap value; if the tail model is misspecified, this procedure can absorb part of the mass signal into the shape parameters. Please address this explicitly and show that the final posterior for m_nu_e is robust to the model choice, not merely that two models give overlapping intervals.","section":"Main text, paragraph on A1 and A2 results and Fig. 4"}],"minor_comments":[{"comment":"The neutrino mass unit is sometimes written as 'eV' without '/c^2' (e.g., 'm_nu_e < 15 eV' in the abstract and main text); please use consistent notation. Also, Fig. 4 axis labels use 'm_e' instead of 'm_nu_e'.","section":"Abstract and throughout"},{"comment":"The 'relative residual' panels would benefit from including error bars or a shaded uncertainty band so that the reader can judge the fit quality in the endpoint region, which is the critical region for the mass limit.","section":"Fig. 3"},{"comment":"The relationship between the quoted background rate B = 9.1(1.3) x 10^-6 /eV/pixel/day and the fitted constant background b_const = 0.073(8) /2eV is not explicitly derived; please state the effective exposure and conversion clearly.","section":"Main text, background discussion"},{"comment":"The sentence 'The theoretical calculations, which best describe the observed spectral features are based on an ab-initio approach' should have a comma after 'features'. There are several other minor grammar and punctuation issues throughout.","section":"Appendix A"}],"recommendation":"major_revision","confidential_remarks":"The experimental achievement is impressive and the result is plausible, but the central limit rests on an unquantified phenomenological model for the atomic physics tail. This is a load-bearing issue that, in my view, requires additional analysis rather than a simple textual revision. If the authors can provide a convincing quantitative bound on the A2 model-mismatch systematic, the paper would likely be acceptable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is the new best direct limit on the electron neutrino mass from 163Ho electron capture, m_nu_e < 15 eV at 90% CI, from about 200 million events. That is a factor of ~1.8 better than HOLMES and the ECHo program's best result to date. The paper is honest about its own limitations, and the Q value it extracts, 2862(4) eV, agrees nicely with the Penning-trap value of 2863.2(6) eV. That agreement is the strongest piece of evidence that the endpoint region is not badly distorted.\n\nWhat is actually new is the dataset and the improved limit. The analysis recipe follows HOLMES (the A2 function, Bayesian fit), so methodological novelty is modest, but this is a legitimate step forward for the field. The background control is genuinely good: 9.1 x 10^-6 /eV/pixel/day, and the unresolved pile-up is treated via auto-convolution. I also give them credit for being explicit that the A1 exponential model is not appropriate at this statistics and for not overclaiming the cross-check.\n\nNow the soft spots, and here I partly agree with the stress-test note. The neutrino-mass sensitivity lives in the last tens of eV below the endpoint, which is exactly where the A2 function's exponential tail SA(E) is doing the work. Those tail parameters are fit from the same 2250–2750 eV region used to constrain m_nu_e and Q. A1 vs A2 is not a real systematic check: A1 requires narrow priors, shares the same data, and the paper itself says it is inappropriate. So there is a genuine model-mismatch uncertainty that is not quantified. That said, the Q agreement with Penning trap gives me some confidence that the slope near the endpoint is about right; if the tail were badly wrong, Q would likely shift more than 4 eV. I would call this a moderate concern, not a fatal one.\n\nOther minor issues: no raw data or analysis code released, and the data reduction discards a fair fraction of events (25% for Ag pixels) with some judgment involved. The authors do perform ratio checks, but those are not fully propagated into the final systematic budget. These are addressable in revision.\n\nBottom line: this is a solid, careful experimental result that deserves a serious referee. The main question a referee should push on is whether the A2 tail choice can be validated or at least bracketed with an alternative—maybe using the ab initio model in a limited way or a more flexible parametrization with a prior on the model difference. I would accept it with that request.\n\nI'd cite this if I work in neutrino mass / 163Ho. I'd bring it to a reading group for the analysis choices, not for the AI-generated novelty.","headline":"New 163Ho calorimetric result improves neutrino mass limit to <15 eV, but the ad hoc spectral model leaves a real, unquantified systematic.","tokens_in":13777,"tokens_out":1594,"would_cite":true,"duration_ms":19574,"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":"The ECHo-1k analysis of about 200 million holmium-163 electron-capture events yields an upper limit of 15 eV/c² (90% credible interval) on the effective electron neutrino mass, with endpoint energy 2862(4) eV consistent with the Penning-tra","keywords":["effective electron neutrino mass","163Ho electron capture","calorimetric neutrino mass","metallic magnetic calorimeter","endpoint spectrum","Bayesian posterior","ECHo-1k"],"falsifier":"Re-fit the same 200-million-event endpoint spectrum using the full ab initio atomic-physics calculation (or a third independent parametrization) in place of A2(E), and check whether the 90% interval for m_nu stays below about 20 eV and Q remains within 4 eV of 2863 eV; if the result moves materially, the limit is set by the tail model rather than by the neutrino mass.","tokens_in":12622,"feed_emoji":"⚛️","tokens_out":8367,"duration_ms":81465,"temperature":0.7,"pith_summary":"This paper reports the ECHo-1k experiment's measurement of the electron-capture decay spectrum of holmium-163, recorded with metallic magnetic calorimeters that absorb the full decay energy. From about 200 million events at a background rate below 10^-5 per eV per pixel per day, the collaboration analyzes the high-energy endpoint of the spectrum and sets an upper limit of 15 eV/c² (90% credible interval) on the effective electron neutrino mass—the combination of neutrino mass eigenstates that governs electron-neutrino kinematics. The endpoint energy extracted in the same fit, Q = 2862(4) eV, agrees with the value 2863.2(6) eV obtained independently by Penning-trap mass spectrometry, an important cross-check of the spectral model. This matters because it is the tightest direct bound from a 163Ho experiment to date and because the method depends only on energy and momentum conservation, offering a path to compare neutrino and antineutrino masses without cosmological assumptions.","feed_headline":"Electron neutrino mass limit drops to 15 eV with holmium calorimetry","feed_subtitle":"Two hundred million electron-capture events set the tightest direct bound from a 163Ho experiment.","key_machinery":"The load-bearing object is the endpoint spectral model of Eq. (1): the decay rate is the product of an atomic-physics factor A(E) and the phase-space factor F_PS(E,Q), convolved with a Gaussian detector response and added to two backgrounds. Neutrino mass enters only through F_PS: near the endpoint, the spectrum is proportional to (Q - E) sqrt((Q - E)² - m_nu²), so a finite mass rounds and shifts the endpoint downward. Because the full ab initio atomic calculation does not yet describe the measured tails, the analysis uses the analytic approximation A2(E)—a Lorentzian for the dominant M1 capture line plus a two-exponential high-energy tail—and samples the posterior via Hamiltonian Monte Carl","core_discovery":"The central result is a nearly factor-of-two improvement over the previous most stringent 163Ho limit: m_nu_e < 15 eV/c² at 90% credible interval (18 eV at 95%), based on about 200 million electron-capture events. The analysis uses the mass-sensitive phase-space factor F_PS(E,Q) = (Q - E) sqrt((Q - E)² - m_nu²); the upper limit comes from the posterior of m_nu_e under a flat prior over 0–100 eV, with the atomic-physics tail represented by an analytic Lorentzian-plus-two-exponentials function A2(E). In the same fit the decay energy is Q = 2862(4) eV, compatible with the Penning-trap value, and the background rate is 9.1(1.3) × 10^-6 /eV/pixel/day. The authors state that a simpler single-expon","pith_inferences":["Inference: because the posterior peaks at zero mass, the data do not demand a nonzero m_nu; the 15 eV figure is an experimental cap, and scaling the exposure by an order of magnitude should push the reach toward a few eV if systematics stay subdominant.","Inference: the next systematic frontier is the atomic-physics tail—once a full ab initio calculation reproduces the measured A(E) tails, the need for the ad-hoc A2 function disappears and the fit becomes less sensitive to model choice.","Inference: the same fitted endpoint model could be used to constrain non-standard endpoint distortions, such as additional neutrino mass eigenstates or spectral anomalies, because the analysis already marginalizes over background and atomic parameters."],"forward_implications":["The 15 eV/c² bound is the strongest direct limit on the electron neutrino mass from a calorimetric 163Ho measurement, roughly half the previous HOLMES limit and almost an order of magnitude better than the earlier ECHo result.","The Q-value agreement (2862(4) eV vs 2863.2(6) eV) confirms that the endpoint energy can be controlled at the few-eV level by combining calorimetry with Penning-trap mass spectrometry.","At current statistics the limiting factors are statistics and background, not detector energy resolution, since the simple-exponential A1 fit fails while the richer A2 model succeeds.","Together with the HOLMES result, this establishes 163Ho calorimetry as a viable route toward sub-eV sensitivity, and combined with tritium limits it opens the possibility of comparing neutrino and antineutrino masses as a CPT test."],"supporting_citations":[{"why":"Provides the independently measured Q-value 2863.2(6) eV against which the fitted endpoint is compared and which anchors the grid fits.","marker":"[1]"},{"why":"Previous best 163Ho limit (27 eV/c²) that this paper improves, and the source of the A2 tail parametrization.","marker":"[2]"},{"why":"Best tritium-based limit (0.45 eV/c²), the comparison point for the overall mass-scale program and for the proposed CPT test.","marker":"[17]"},{"why":"Ab initio calculation of the 163Ho electron-capture spectrum that motivates the exponential high-energy tail approximated by A2.","marker":"[37]"},{"why":"Describes the ECHo-1k metallic magnetic calorimeter array and gives the detector response and performance used to acquire the 200-million-event spectrum.","marker":"[26]"},{"why":"Earlier ECHo spectrum and 150 eV limit that the present analysis improves by nearly an order of magnitude.","marker":"[27]"},{"why":"Statistical platform implementing the Hamiltonian Monte Carlo sampler used for the posterior inference on Q and m_nu.","marker":"[40]"},{"why":"Data reduction concept for identifying single 163Ho events and removing background and pile-up from the ECHo-1k spectrum.","marker":"[36]"},{"why":"Ab initio calculation of the capture spectrum used for identifying the resonance lines and their positions.","marker":"[20]"}],"fun_headline_variants":["Neutrino mass bound halved to 15 eV in holmium decay","200M events tighten neutrino mass limit to 15 eV","Holmium data nearly halve neutrino mass ceiling","Electron neutrino mass <15 eV from 163Ho spectrum","Tightest 163Ho bound: neutrino mass <15 eV"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The analytical approximation A2(E) in Eq. (2) adequately represents the atomic-physics factor A(E) in the energy window used to constrain the neutrino mass, and the data themselves are used to set its parameters, so a mismatch in the tail shape near the endpoint could shift the mass limit.","fun_headline_variants_meta":{"raw":{"variants":["Neutrino mass bound halved to 15 eV in holmium decay","200M events tighten neutrino mass limit to 15 eV","Holmium data nearly halve neutrino mass ceiling","Electron neutrino mass <15 eV from 163Ho spectrum","Tightest 163Ho bound: neutrino mass <15 eV"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000876,"raw_usage":{"total_tokens":3646,"prompt_tokens":783,"completion_tokens":2863,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":527,"completion_tokens_details":{"reasoning_tokens":2777}},"tokens_in":527,"tokens_out":2863,"duration_ms":18774,"temperature":1.0,"reasoning_tokens":2777,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:54:10.667478+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-fit the same 200-million-event endpoint spectrum using the full ab initio atomic-physics calculation (or a third independent parametrization) in place of A2(E), and check whether the 90% interval for m_nu stays below about 20 eV and Q remains within 4 eV of 2863 eV; if the result moves materially, the limit is set by the tail model rather than by the neutrino mass.","supporting_citations":[{"cited_title":"Schweiger, M","cited_arxiv_id":null,"evidence_quote":"Provides the independently measured Q-value 2863.2(6) eV against which the fitted endpoint is compared and which anchors the grid fits."},{"cited_title":"Aker et al","cited_arxiv_id":null,"evidence_quote":"Best tritium-based limit (0.45 eV/c²), the comparison point for the overall mass-scale program and for the proposed CPT test."},{"cited_title":"Hammann, A","cited_arxiv_id":null,"evidence_quote":"Ab initio calculation of the 163Ho electron-capture spectrum that motivates the exponential high-energy tail approximated by A2."},{"cited_title":"Mantegazzini, A","cited_arxiv_id":null,"evidence_quote":"Describes the ECHo-1k metallic magnetic calorimeter array and gives the detector response and performance used to acquire the 200-million-event spectrum."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier ECHo spectrum and 150 eV limit that the present analysis improves by nearly an order of magnitude."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Statistical platform implementing the Hamiltonian Monte Carlo sampler used for the posterior inference on Q and m_nu."},{"cited_title":"Drung, C","cited_arxiv_id":null,"evidence_quote":"Data reduction concept for identifying single 163Ho events and removing background and pile-up from the ECHo-1k spectrum."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Ab initio calculation of the capture spectrum used for identifying the resonance lines and their positions."}],"review_version":1}