{"id":"14a77670-4f28-4678-94cb-2df8377dc467","arxiv_id":"2506.13829","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The leading-order neutrino-mass bias from an inhomogeneous KATRIN source potential comes from potential-dependent shifts between spectra of different electron scattering multiplicities, observable with co-circulated krypton-83m.","lead":"A theory paper derives how an uneven electric potential inside the KATRIN tritium source shifts the measured electron spectrum and biases the neutrino mass. It shows that this bias is dominated by shifts between scattered and unscattered electrons, and that krypton-83m calibration lines can constrain the effect.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Quantitative susceptibility coefficients in Eq. (52) rely on a simplified response model and unweighted ΔqU averaging; the reported a1 maximum mismatch with the full simulation means the '1 eV·Δ10' scaling is not yet established.","rationale":"The reader's weakest assumption identified exactly the same load-bearing point: Eq. (52) is derived under a surplus-energy-only response, a simplified response model, and unweighted ΔqU averaging, and one of the two plotted coefficients visibly disagrees with the full simulation. I see no reason to move away from the reader's conditional verdict. The qualitative central claim—that energy-loss shifts Δᵢ₀, rather than Gaussian broadening σ₀, drive the leading-order neutrino-mass bias in the presence of scattering—is well supported by the covariance formalism in Section 2.5 and by the linear-versus-quadratic scaling argument. That claim is robust to factor-of-2 uncertainties in εᵢ, because σ₀ effects are quadratic in a 30 meV potential and the Δᵢ₀ effects are linear with coefficients of order 1 eV, a separation of roughly an order of magnitude. However, the specific numerical prediction 'Δm² ∝ 1 eV·Δ₁₀' is exactly as strong as the computed ε₁ value. The paper is honest about the limitations of the analytic calculation, but the abstract's phrasing 'establishes the leading-order observables' risks overstating what has been validated. The conditional verdict—requiring either a full-model check of the coefficients or a more explicitly qualitative statement—is appropriate. The proposed concrete test, recomputing aᵢ and εᵢ with the full response model and real measurement-time weights and comparing to the Asimov results, would settle whether the quantitative scaling is trustworthy or should be presented as an estimate only.","tokens_in":16118,"tokens_out":4882,"duration_ms":60823,"concrete_test":"Recompute a₁ and ε₁ from Eq. (52) using the full KATRIN response model of [5] (including energy- and angle-dependent scattering probabilities and transmission) and the actual measurement-time weighting over retarding energies used in a KNM campaign, instead of the simplified model and unweighted ΔqU average. Compare the resulting ε₁(ΔqU) curves with the Asimov-fit values reported in [6]. If ε₁ changes by more than a factor of 2 relative to the simplified calculation, or if the a₁ maximum disagreement persists, the conclusion's '1 eV·Δ₁₀' scaling must be downgraded to a qualitative estimate and the abstract should not claim the leading-order bias magnitude is established without the full-model validation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claim is that the leading-order neutrino-mass bias is Δm² ≈ −ε₁Δ₁₀ with ε₁ on the 1 eV scale, and that this shift effect dominates the Gaussian σ₀ broadening. That claim depends on the susceptibility coefficients aᵢ and εᵢ computed in Section 3.3. Equation (52) is derived under three explicit approximations: the response function depends only on surplus energy ε = E − qU; a simplified response model is used; and the ΔqU-average is unweighted. The paper itself reports that the maximum of a₁ in Fig. 9 does not match the full Asimov simulation in [6]. Because the central quantitative statement 'Δm² ∝ 1 eV·Δ₁₀' rides on ε₁, an unvalidated coefficient makes that particular number a motivated estimate rather than an established result. The qualitative claim that Δᵢ₀, not σ₀, dominate at leading order is supported by the covariance bound |Δᵢ₀| ≤ κᵢσ₀ and by the linear-versus-quadratic scaling, and would survive moderate errors in εᵢ. The load-bearing gap is therefore not the existence of the shift effect, but the numerical magnitude claimed for it. The paper's own wording, 'results are only qualitative', is consistent with this assessment, but the abstract's 'establishes the leading-order observables' can be read as going further than the validated coefficients allow.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes how a non-constant electric potential in the KATRIN tritium source biases the measured neutrino mass and beta-spectrum endpoint. It introduces starting-potential distributions for each scattering multiplicity, defines their moments (mean, standard deviation, and mutual shifts Δ_i0), and derives a covariance bound |Δ_i0| ≤ κ_i σ_0 together with an ellipsoidal parameter space for the shifts (Section 2). In Section 3 it argues that, because of electron scattering, the leading-order neutrino-mass bias is caused by the first-moment shifts Δ_i0 rather than by the Gaussian broadening σ_0, and it derives susceptibility coefficients a_i and ε_i (Eq. 52) that relate ΔE_0 and Δm² to Δ_i0. The paper concludes that Δm² scales roughly as 1 eV × Δ_10, and that the krypton-83m measurement of Δ_10 can constrain this systematic.","tokens_in":16432,"tokens_out":3524,"duration_ms":41038,"significance":"If the central qualitative claim holds, the paper corrects a potentially important systematic treatment for KATRIN: the dominant source-potential bias would come from shifts between spectra of different scattering multiplicities, not from a simple Gaussian broadening. The covariance and ellipsoid mathematics in Section 2.5 are clean and appear correct: the bound |Δ_i0| ≤ κ_i σ_0 follows directly from Cauchy-Schwarz, and the ellipsoid parameterization is a sound geometric result. The distinction between shift-type and broadening-type observables is conceptually valuable and likely to influence how source-potential systematics are parametrized and constrained by krypton-83m calibration data. However, the quantitative susceptibility coefficients in Eq. (52) are not yet validated: they rest on a simplified response model and unweighted ΔqU averaging, and Fig. 9 reports a visible mismatch between a_1 and the full Asimov simulation of reference [6]. The paper itself repeatedly describes the calculation as qualitative, so the strength of the contribution lies in the qualitative hierarchy and the parameter-space geometry, not in the specific 1 eV coefficient.","major_comments":[{"comment":"The central quantitative claim, Δm² ≈ −ε_1 Δ_10 with ε_1 on the 1 eV scale, rests entirely on the coefficients a_i and ε_i derived in Eq. (52). The derivation is only sketched (\"only the basic ideas are described\"), uses a simplified response model, and assumes an unweighted ΔqU average. Figure 9 shows that the maximum of a_1 deviates visibly from the full Asimov simulation in [6]; the authors attribute this to the approximations. As presented, ε_i is an unvalidated estimate, not an established result. Please either derive Eq. (52) from stated assumptions in an appendix and validate both a_1 and ε_1 against the full simulation, or explicitly downgrade the 1 eV scaling to a motivated heuristic and remove it from the abstract and conclusions.","section":"§3.3, Eq. (52)"},{"comment":"Equations (45) define the bias as an average over all retarding potentials U with a weight that must contain the measurement-time distribution, and the footnote adds that a χ²-minimization would modify the result. The evaluation in Fig. 9 instead uses an unweighted ΔqU average. This is a different weighting from the actual KATRIN analysis, so the numerical values of a_i and ε_i are not tied to a real measurement procedure. Please state the relationship between the unweighted average and the experimental analysis weighting, and estimate the resulting uncertainty on the coefficients, or restrict the conclusions to the unweighted model.","section":"§3.2, Eq. (45) and footnote 2"},{"comment":"The normalization step is calibrated with the constant-potential case ΔE_0 = q⟨V⟩_0 and Δm² = 0, adding a correction term proportional to the endpoint. This is an ad-hoc convention that enters the derivation of Eq. (52). It is not shown that the final coefficients are independent of this calibration or that the 1 eV scale is robust under a different normalization choice. Please make the calibration dependence explicit and demonstrate that the quoted scale is not an artifact of this convention.","section":"§3.3, Normalization"},{"comment":"The perturbative truncation at second order relies on the statement that moments of order n ≥ 3 are suppressed by (σ_i/w)^n, justified only by a reference to Hölder's inequality without showing the argument. Since this truncation underlies the entire moment-based expansion, please provide the derivation or a precise citation; otherwise the stated suppression factor is unsubstantiated.","section":"§2.3, suppression of higher moments"}],"minor_comments":[{"comment":"There are several typographical errors, including \"througout\" (abstract or Section 2) and \"previous believe\" in the conclusions; these should be corrected.","section":"Throughout"},{"comment":"The caption refers to an \"N123 spectrum\" and lines \"N1\", \"N2\", \"N3\" without defining these labels in the text; please add a brief definition.","section":"Fig. 5 caption"},{"comment":"The figure caption says \"good qualitative agreement\" for ε_1 but \"the position of its maximum deviates visibly\" for a_1; please quantify the deviations (e.g., shift in eV and relative amplitude) so the reader can judge the level of agreement.","section":"§3.3, Fig. 9"},{"comment":"The definition of the shape operators ρ̂_i uses σ_0[V] both in the numerator denominator; consider clarifying that σ_0 is the standard deviation of the unscattered electron distribution, as stated around Eq. (25).","section":"§2.3, Eq. (31)"},{"comment":"The propagation of the krypton Δ_10 constraint to tritium via Eq. (67) would benefit from a brief derivation or a reference to the covariance calculation, as the expression with the ± sign is not immediately obvious.","section":"§4.3, Eq. (67)"}],"recommendation":"major_revision","confidential_remarks":"The paper's main novelty relative to the author's PhD thesis [6] should be clarified; as written, it is not entirely clear how much of Section 2.5 already appeared in [6]. The title and abstract claim to \"establish\" leading-order observables, but the quantitative part is explicitly qualified as qualitative; the editorial decision should weigh whether a corrected version that either validates the coefficients or clearly limits the claims would be suitable for EPJ C. The covariance/ellipsoid part is solid and could stand as the core contribution if the susceptibility discussion is reframed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper makes a genuinely useful conceptual point—once you have electron scattering, the leading-order neutrino-mass bias from a non-constant source potential comes from the shifts Δi0 between scattering multiplicities, not from the Gaussian broadening σ0. That is something KATRIN's predecessor analyses missed, and the moment-based formalism makes it precise. The covariance bound |Δi0| ≤ κiσ0 and the ellipsoid parameterization are clean and correct as far as I can tell, and the connection to the krypton-83m measurement is practical. The paper is honest about the limits of the quantitative part: Section 3.3 says the calculation is 'only qualitative' and recommends Asimov studies for numbers. The abstract's 'establishes the leading-order observables' is fine if read as establishing the structure, not the exact coefficients.\n\nWhere it gets soft: the coefficients ai and εi in Eq. 52 are derived with a simplified response model and unweighted ΔqU averaging, and the paper itself reports that the a1 maximum does not match the full simulation in the author's thesis. So the claim 'Δm² ≈ −ε1Δ10 with ε1 ~ 1 eV' is a motivated estimate, not a demonstrated result. If the approximate model is off on the position of a1, it could be off on ε1 too. The qualitative conclusion—that first moments dominate—does not depend on those numbers, because it follows from the covariance structure and the linear-versus-quadratic scaling. So the soft spot is real but load-bearing only for the number, not for the idea.\n\nI would send this to a serious referee. The formalism is likely to be used in KATRIN systematics analysis, and the paper deserves scrutiny on its own terms. The author is clearly thinking carefully, and the self-citation to the thesis is legitimate—the thesis has the simulations, and the paper contributes the analytic generalization. A referee should push for a more careful statement of the quantitative reach, maybe suggest moving the '1 eV scale' claim into a caveated part, but the core is worth refereeing.","headline":"The paper's real contribution is reframing the KATRIN source-potential bias as a first-moment problem; the quantitative size of the effect is still open, but the qualitative argument is solid.","tokens_in":16939,"tokens_out":2549,"would_cite":true,"duration_ms":25641,"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":"KATRIN's neutrino-mass bias from a non-constant source potential is set by shifts between scattered and unscattered spectra, not by Gaussian broadening.","keywords":["KATRIN","neutrino mass","source potential","krypton-83m","energy-loss shift","starting-potential distribution","beta spectrum endpoint","MAC-E filter"],"falsifier":"Run the full KATRIN simulation with a known antisymmetric potential proportional to the boundary shape $Q_1(z)$ and fit the generated spectrum with the zero-potential response over the standard analysis range; if the recovered $\\Delta m^2$ deviates from $-\\varepsilon_1 \\Delta_{10}$ by more than the statistical precision, the leading-order coefficient is wrong. A more direct check is to compare the predicted location of the maximum of $a_1$ with the full simulation, since the paper reports a visible mismatch there.","tokens_in":1916,"feed_emoji":"⚛️","tokens_out":3351,"duration_ms":104903,"temperature":0.7,"pith_summary":"The paper tries to establish that the leading-order neutrino-mass bias from an inhomogeneous electric potential in the KATRIN tritium source comes from first moments of the starting-potential distribution, specifically the mutual shifts between spectra of different electron-scattering multiplicities, rather than from the Gaussian broadening of the potential. It derives a perturbative expansion in the statistical moments of the starting-potential distribution and shows that the allowed parameter space of these shifts is an ellipsoid constrained by a single measured standard deviation. If correct, this means the krypton-83m calibration spectrum, where the once-scattered line sits about 13 eV below the unscattered line, is the right instrument for the dominant source-potential systematic: the distance between the two lines directly measures the shift. The paper estimates that the squared neutrino-mass bias scales roughly as a 1 eV coefficient times the shift, turning a tens-of-millivolt potential asymmetry into a substantial systematic, and projects that measuring the shift will reduce this uncertainty by a factor of two to three. This matters because the experiment's sensitivity goal depends on controlling every spectrum-shaping systematic near the endpoint.","feed_headline":"Krypton line shift, not broadening, biases KATRIN neutrino mass","feed_subtitle":"A new analysis says the dominant source systematic is a measurable shift between scattered and unscattered spectra.","key_machinery":"The central object is the starting-potential distribution $\\mathrm{SPD}_i(\\nu) = \\langle \\delta(\\nu - V(z)) P_i(z) \\rangle$, which, because convolution is commutative and associative, enters the KATRIN response exactly like an additional energy-loss distribution $f'_i = \\mathrm{SPD}_i \\otimes f_i$. The argument is carried by the first two moments of these distributions, especially the energy-loss shifts $\\Delta_{i0}[V] = \\mathrm{Cov}_0[(P_i - P_0)/P_0, V]$, the bound $|\\Delta_{i0}[V]| \\le \\kappa_i \\sigma_0[V]$, and the ellipsoidal shape-operator space $\\vec{\\rho}^\\mathsf{T} P^{-1} \\vec{\\rho} = 1$. The paper then linearizes the response-function moments to obtain the susceptibility coefficients $a_i$ and $\\varepsilon_i$ that convert the shifts into endpoint and squared-mass biases.","core_discovery":"The paper's central claim is that, once electron scattering in the KATRIN source is taken into account, the leading-order neutrino-mass bias from an inhomogeneous source potential is not the familiar Gaussian broadening $\\sigma_0[V]$ but the energy-loss shifts $\\Delta_{i0}[V] = \\langle V \\rangle_i - \\langle V \\rangle_0$, which are the first moments of the starting-potential distributions for each scattering multiplicity. Because scattered electrons come preferentially from the rear of the source and unscattered electrons from the front, even a small antisymmetric potential produces such shifts, and the squared neutrino mass is biased approximately by $\\Delta m^2 \\propto 1\\,\\mathrm{eV} \\cdot \\Delta_{10}$. The paper proves these shifts are constrained by the measured standard deviation through $|\\Delta_{i0}[V]| \\le \\kappa_i \\sigma_0[V]$, with the full vector of shifts lying on an ellipsoid, and it identifies the $^{83\\mathrm{m}}$Kr conversion-electron lines as the observable that directly measures $\\Delta_{10}$.","pith_inferences":["If the first-moment logic is correct, other gaseous-source tritium endpoint experiments using MAC-E filters should re-examine their source-potential systematics in terms of scattering-multiplicity shifts, not just variance broadening.","The same formalism suggests that calibration lines with higher scattering multiplicities, beyond the once-scattered line, offer additional handles on the potential shape if their widths can be resolved.","A direct experimental test would be to compare the fitted neutrino-mass bias in simulated KATRIN data with a known asymmetric potential against the prediction $-\\varepsilon_1 \\Delta_{10}$; the paper already notes a mismatch in the location of the $a_1$ maximum, making such a test decisive.","The distinction between $\\sigma_0$ and $\\Delta_{10}$ implies that the source potential should be monitored by line-position shifts between scattered and unscattered calibration lines, not only by line broadening."],"forward_implications":["The dominant source-potential systematic can no longer be treated as a Gaussian broadening; future KATRIN analyses must include the first-moment shifts as separate parameters.","Measuring $\\Delta_{10}$ from the $^{83\\mathrm{m}}$Kr spectrum, with the projected few-millivolt uncertainty, can shrink the neutrino-mass systematic by roughly a factor of two to three.","The ellipsoidal parameter space provides a practical way to constrain all $\\Delta_{i0}$ simultaneously from a single measured $\\sigma_0$.","The effective single-parameter treatment of the shifts is discouraged because it corrects the mass bias but leaves residual structure in other observables; the full ellipsoid parametrization is recommended instead.","A measured $^{83\\mathrm{m}}$Kr $\\Delta_{10}$ near zero would only weakly constrain the potential, because antisymmetric potentials dominate $\\Delta_{10}$ while symmetric ones contribute mostly to $\\sigma_0$."],"supporting_citations":[{"why":"Supplies the response-function model, including scattering probabilities, transmission functions, and convolved energy-loss spectra, on which the perturbative expansion is built.","marker":"[5]"},{"why":"Provides the full Asimov Monte Carlo simulation against which the analytic susceptibility coefficients are compared and from which the generalization to arbitrary scattering numbers is taken.","marker":"[6]"},{"why":"Establishes the Taylor-expansion result that an arbitrary energy fluctuation shifts the endpoint by its mean and the squared neutrino mass by $-2$ times its variance, the baseline the paper extends to scattering multiplicities.","marker":"[10]"},{"why":"The parallel $^{83\\mathrm{m}}$Kr measurement that determines $\\sigma_0^2 = 1.0(3) \\times 10^{-3}\\,\\mathrm{eV}^2$ and supplies the projected uncertainty on $\\Delta_{10}$ used to constrain the ellipsoid.","marker":"[11]"},{"why":"The original derivation that a Gaussian energy spread biases the squared neutrino mass by $\\Delta m^2 = -2\\sigma^2$, the previous understanding the paper overturns for the source-potential case.","marker":"[12]"}],"fun_headline_variants":["Krypton-83m shows source shift, not spread, biases KATRIN","Neutrino mass bias at KATRIN is a shift, not a broadening, from source","Krypton-83m lines measure the potential shift biasing KATRIN","Source potential shift, not spread, is the leading KATRIN bias"],"cache_read_input_tokens":19072,"weakest_assumption_plain":"The result depends on the response function depending only on surplus energy $\\epsilon = E - qU$ and on evaluating the susceptibility coefficients with a simplified response model and unweighted averaging over the retarding-energy scan; if the real energy-dependent scattering or the actual measurement-time weighting breaks that form, the numerical coefficients, including the claimed $\\Delta m^2 \\propto 1\\,\\mathrm{eV} \\cdot \\Delta_{10}$ scaling, would change, and the paper itself notes that the maximum of $a_1$ does not match the full simulation.","fun_headline_variants_meta":{"raw":{"variants":["Krypton-83m shows source shift, not spread, biases KATRIN","Neutrino mass bias at KATRIN is a shift, not a broadening, from source","Krypton-83m lines measure the potential shift biasing KATRIN","Source potential shift, not spread, is the leading KATRIN bias"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000718,"raw_usage":{"total_tokens":3216,"prompt_tokens":928,"completion_tokens":2288,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":544,"completion_tokens_details":{"reasoning_tokens":2207}},"tokens_in":544,"tokens_out":2288,"duration_ms":18008,"temperature":1.0,"reasoning_tokens":2207,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:37:15.207704+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the full KATRIN simulation with a known antisymmetric potential proportional to the boundary shape $Q_1(z)$ and fit the generated spectrum with the zero-potential response over the standard analysis range; if the recovered $\\Delta m^2$ deviates from $-\\varepsilon_1 \\Delta_{10}$ by more than the statistical precision, the leading-order coefficient is wrong. A more direct check is to compare the predicted location of the maximum of $a_1$ with the full simulation, since the paper reports a visible mismatch there.","supporting_citations":[{"cited_title":"Kleesiek et al","cited_arxiv_id":null,"evidence_quote":"Supplies the response-function model, including scattering probabilities, transmission functions, and convolved energy-loss spectra, on which the perturbative expansion is built."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the Taylor-expansion result that an arbitrary energy fluctuation shifts the endpoint by its mean and the squared neutrino mass by $-2$ times its variance, the baseline the paper extends to scattering multiplicities."},{"cited_title":"Acharya et al","cited_arxiv_id":null,"evidence_quote":"The parallel $^{83\\mathrm{m}}$Kr measurement that determines $\\sigma_0^2 = 1.0(3) \\times 10^{-3}\\,\\mathrm{eV}^2$ and supplies the projected uncertainty on $\\Delta_{10}$ used to constrain the ellipsoid."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The original derivation that a Gaussian energy spread biases the squared neutrino mass by $\\Delta m^2 = -2\\sigma^2$, the previous understanding the paper overturns for the source-potential case."}],"review_version":1}