{"id":"310f2d50-05dc-4093-a823-55135e1737cd","arxiv_id":"1908.06992","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Allowing electron-tau non-standard neutrino interactions in the T2K and NOvA data removes the preference for normal mass ordering and even creates a mild preference for inverted ordering.","lead":"The 2.4 sigma hint that neutrinos have a normal mass ordering, seen in the T2K and NOvA experiments, disappears once the data are reinterpreted with non-standard neutrino interactions in the electron-tau sector. The paper argues that any claim of a mass ordering discovery must account for such exotic interactions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"IO washout is achieved at |εeτ|≈0.39, above the paper's own 90% bound (0.36); the relaxation argument is self-referential, so an independent global constraint could restore the NO preference.","rationale":"The reader identified the external-bound dependence as the weakest assumption, and my reading agrees: the numerical washout in the inverted ordering happens at |εeτ|≈0.39, which sits above the 90% bound quoted in the paper itself. The paper's proposed relaxation of that bound is inferred from the very T2K/NOvA data that are being fit, so it does not settle whether independent atmospheric, COHERENT, or global constraints would allow the large coupling. This is the single most load-bearing concern because the central claim is not just that NSI can degrade ordering sensitivity (which is expected for sufficiently large NSI), but that the current 2.4σ NO preference is completely erased for e-τ NSI allowed by data. If independent constraints cap |εeτ| near or below the old 0.36 value, the IO fit loses its main advantage and the preference for NO could re-emerge, making the washout an artifact of neglecting those constraints. The paper is otherwise well structured: the analytical probability decomposition is instructive, the GLoBES-based numerical analysis is plausible, and the DUNE projection gives a concrete falsifiable statement. No internal inconsistency is apparent; the issue is that the conclusion is stronger than the included dataset can support. Therefore the appropriate verdict remains conditional, unchanged from the reader's assessment.","tokens_in":10813,"tokens_out":7269,"duration_ms":79439,"concrete_test":"Reproduce the T2K+NOvA GLoBES analysis described in the paper and add an external penalty on |εeτ| derived from a global fit that excludes the newest T2K/NOvA appearance samples, e.g., enforcing the quoted 90% bound |εeτ|≤0.36 (equivalently, a one-sided Gaussian prior with σ≈0.22). Recompute the IO best fit and Δχ²_SM+NSI,NO − Δχ²_SM+NSI,IO. If the IO best-fit |εeτ| drops below ~0.35 or Δχ² becomes ≳2 in favor of NO, the 'complete washout' conclusion is not robust. An alternative check is to add Super-K atmospheric and COHERENT likelihoods to the same setup and see whether the global best-fit |εeτ| in IO remains near 0.39.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central washout claim requires that the e-τ NSI coupling needed to fit the inverted ordering, |εeτ|≈0.39, is compatible with independent constraints. The paper itself quotes a 90% upper bound |εeτ|≲0.36 from a global analysis (Theoretical framework, Sec. 2) but argues that the newer T2K/NOvA data would relax it. This argument is self-referential: the data used to infer the relaxed bound are the same appearance data whose ordering preference is being erased. The paper does not actually perform a full global fit including atmospheric (Super-K, IceCube) or COHERENT data, so the size of the external bound remains an unvalidated input. If those datasets constrain |εeτ| below roughly 0.3, the IO best fit would be penalized and the χ² difference between NO and IO would shift back toward NO, potentially restoring the 2.4σ preference. This is not an internal inconsistency, but it is a robustness gap in the central claim: the washout is only established within a partial fit that excludes the most constraining independent datasets. The paper's own caveat about the bound relaxation makes this dependence explicit, strengthening the need for a quantitative check.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper re-analyzes the combined T2K and NOvA data using the GLoBES package with the public NSI tool, asking whether the ~2.4sigma preference for the normal mass ordering (NO) survives when neutral-current non-standard interactions of the flavor-changing e-tau type are allowed. In the standard three-flavor fit the authors reproduce the NO preference (chi^2_SM,NO - chi^2_SM,IO ~ -5.6), while in the SM+epsilon_e-tau fit the difference becomes chi^2_SM+NSI,NO - chi^2_SM+NSI,IO ~ +0.5, i.e., a mild preference for inverted ordering (IO), thereby washing out the NO indication. The paper also presents an analytical expansion of the nu_mu->nu_e probability to explain the effect, and DUNE projections in the Supplemental Material. The authors explicitly note a caveat: the IO best fit requires |epsilon_e-tau| ~ 0.39, above the 90% bound quoted from the earlier global analysis, and they argue that the newer T2K/NOvA data would relax that bound.","tokens_in":11125,"tokens_out":6808,"duration_ms":66383,"significance":"If the central claim holds, it would demonstrate that the current normal-ordering hint from T2K and NOvA is not robust against a well-motivated class of new physics, which is an important message for the community. The analysis is reproducible in principle: it uses publicly released experimental data, the public GLoBES framework, and the public NSI simulation tool. The analytical decomposition of the appearance probability is a pedagogically useful illustration of why epsilon_e-tau can alter the ordering discrimination. However, the quantitative significance of the washout is conditional on the assumption that independent constraints on |epsilon_e-tau| are weak enough to accommodate the large IO best-fit value; this assumption is not validated inside the paper.","major_comments":[{"comment":"The IO best fit is located at |epsilon_e-tau| ~ 0.39 (Fig. 2, right panel), which is above the 90% C.L. upper bound |epsilon_e-tau| <~ 0.36 quoted in the Theoretical framework section from Refs. [41] and [39]. The authors anticipate that including the new T2K/NOvA data would sensibly relax the bound, but the data used to infer this relaxation are the same appearance data whose ordering preference is being erased. The paper does not include the most constraining independent probes, in particular atmospheric neutrino data (Super-K, IceCube) and COHERENT, even though the Conclusions state that complementing the study with atmospheric data would be interesting. This matters because if independent constraints keep |epsilon_e-tau| below roughly 0.3, the IO fit would be penalized and the NO preference would likely survive. I ask the authors to quantify this dependence, e.g., by profiling over a range of external upper bounds on |epsilon_e-tau| or by adding a representative atmospheric constraint, and to adjust the abstract and conclusions if the washout disappears under such a bound.","section":"Theoretical framework and Fig. 2"},{"comment":"The DUNE sensitivity projection assumes a true value |epsilon_e-tau| = 0.2, described as intermediate between the best-fit values found for NO and IO. In the real-data analysis the NO best fit is |epsilon_e-tau| ~ 0.09 and the IO best fit is ~ 0.39, so the chosen true value is not derived from the fit. If the true value were closer to the IO best fit, the DUNE sensitivity could be even worse than shown; if closer to the NO best fit, it could be better. The stated conclusion that DUNE can resolve the ambiguity only for delta near 1.5pi is therefore sensitive to an unquantified assumption. The authors should show the DUNE sensitivity as a function of the assumed true |epsilon_e-tau| (including values near the IO best fit) and, ideally, with a prior based on an updated external bound.","section":"Supplemental Material, Fig. S1"}],"minor_comments":[{"comment":"The text says 'three mixing angles theta12, theta13, theta13'; the third angle should be theta23.","section":"Introduction"},{"comment":"The analytical probability expansion is first-order in |epsilon| and is justified by the claim that |epsilon| is small, of order 0.2, yet the same paragraph anticipates best-fit values up to |epsilon| ~ 0.4. The authors should state explicitly that the analytic expression is illustrative and that the numerical GLoBES results use the full Hamiltonian.","section":"Eq. (11) and surrounding discussion"},{"comment":"The sentence 'We have also marginalized over Delta m^2_31 in the test, where we have taken the true value equal to 2.49 x 10^-3 eV^2 obtained from the real data analysis' is confusing; presumably the true value is fixed and the test value is profiled.","section":"Supplemental Material"},{"comment":"The contours in Fig. 2 are drawn for 1 d.o.f., but the text quotes the significance of the preference for nonzero |epsilon_e-tau| as 2.5sigma; the connection between the contour level and the quoted significance (e.g., the Delta chi^2 between the best fit and |epsilon_e-tau| = 0) should be made explicit.","section":"Fig. 2 and Numerical Results"},{"comment":"A brief comment on why epsilon_ee and epsilon_mu-tau are not included in the scan would be helpful, especially since epsilon_mu-tau is tightly bounded by atmospheric data while epsilon_ee is not obviously negligible.","section":"Theoretical framework"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The punchline: this paper shows that the 2.4σ normal-ordering preference from T2K/NOvA collapses once you allow flavor-changing e-tau NSI. The Δχ² between normal and inverted ordering flips from −5.6 to +0.5, leaving a 0.7σ lean toward inverted. That is the new quantitative result, and it is worth knowing.\n\nWhat is actually new is the application to the 2019 public datasets. The general idea that NSI can create mass-ordering degeneracies is in the earlier literature, but the specific demonstration that the current data lose the NO preference under a specific NSI sector is new. The paper is also careful: the probability decomposition is explicit, the GLoBES fit uses public data, and the authors state their analysis setup clearly. The DUNE projection gives a concrete, testable expectation: if δ is not near 1.5π, DUNE will not resolve the ordering in the presence of NSI.\n\nWhere the paper is softer: the inverted-ordering washout sits at best-fit |ε_eτ| ≈ 0.39, which is above the 90% bound of 0.36 quoted from the global fit. The authors argue that the bound would be relaxed by the newer T2K/NOvA data, but that argument is self-referential—the relaxation is inferred from the same data whose ordering preference is being erased. If independent constraints (atmospheric, COHERENT, or a full global fit) keep |ε_eτ| below roughly 0.3, the NO preference may survive. The paper flags this and lists atmospheric data as future work, so it is not a hidden flaw, but it does mean the central claim is conditional on an external bound that is not quantitatively checked. A second, more minor caveat: only ε_eτ and ε_eμ are varied one at a time; simultaneous variation with ε_ee or ε_μτ is not explored.\n\nThe math checks out as far as I can tell, and the citation pattern is fair, with the relevant prior work cited. The paper is a legitimate robustness study rather than a discovery claim, and it is honest about its scope. I would bring it to the reading group, and I would cite it. For peer review, I would definitely accept it—the phenomenology is well executed and the fragility point is important for the field, even if the quantitative robustness relies on an external constraint that deserves closer scrutiny.","headline":"A well-executed robustness study: allowing e-tau NSI flips the T2K/NOvA mass-ordering preference from 2.4σ normal to 0.7σ inverted, though the washout size hinges on a self-referential relaxation of the 90% NSI bound.","tokens_in":11644,"tokens_out":3277,"would_cite":true,"duration_ms":34093,"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":"Allowing electron-tau flavor-changing neutral-current interactions completely erases the current 2.4σ normal-ordering preference from T2K and NOvA; with those interactions, inverted ordering is even mildly favored.","keywords":["neutrino mass ordering","non-standard interactions","e-tau NSI","T2K","NOvA","long-baseline neutrino oscillations","matter effects","CP violation"],"falsifier":"Re-run the same T2K+NOvA fit with an external 90% C.L. constraint $|\\varepsilon_{e\\tau}|\\le0.36$ imposed as a hard prior. Since the inverted-ordering best fit sits at $|\\varepsilon_{e\\tau}|\\approx0.39$, that minimum would move to the boundary and pay a penalty of roughly $((0.39-0.36)/\\sigma_{\\rm ext})^2$ in $\\chi^2$; if independent data (e.g., atmospheric neutrinos or coherent elastic neutrino-nucleus scattering) tighten the bound to $|\\varepsilon_{e\\tau}|\\lesssim0.3$, the penalty would exceed 4 and the normal-ordering preference would re-emerge. A published global fit with updated T2K/NOvA data that keeps $|\\varepsilon_{e\\tau}|<0.3$ at 90% C.L. would falsify the washout claim.","tokens_in":10586,"feed_emoji":"⚛️","tokens_out":6028,"duration_ms":59606,"temperature":0.7,"pith_summary":"The paper is trying to establish that the current ~2.4σ preference for normal neutrino mass ordering, extracted from T2K and NOvA long-baseline data under the standard three-flavor assumption, disappears as soon as neutral-current non-standard interactions (NSI) of the flavor-changing e-tau type are allowed in the fit. Concretely, the difference in fit quality between normal and inverted ordering changes from Δχ² ≈ −5.6 (favoring normal ordering) in the standard model to Δχ² ≈ +0.5 (mildly favoring inverted ordering) once εeτ is included. The reason this matters is that the neutrino mass ordering is one of the main open questions in particle physics, and experiments are being designed around a discovery that this result shows would be contingent on the absence of a specific class of new physics.","feed_headline":"Mass-ordering hint vanishes if e-tau neutrino interactions allowed","feed_subtitle":"T2K and NOvA's 2.4σ normal-ordering edge flips to a 0.7σ tilt toward inverted ordering.","key_machinery":"The argument runs on an analytic decomposition of the νμ→νe appearance probability, $P_{\\mu e}\\simeq P_0+P_1+P_2$, where $P_0$ is the standard leading term set by $\\theta_{13}$, $P_1$ is the standard solar-atmospheric interference, and $P_2$ is a matter-induced NSI interference term proportional to $V_{CC}\\,|\\varepsilon|$ and to the CP-phase combination $\\delta+\\varphi_{\\varepsilon}$. For $\\varepsilon_{e\\tau}$ the coefficients in $P_2$ enter with opposite signs for the two mass orderings because the sign of $\\Delta m^2_{31}$ flips, while the larger matter potential in NOvA ($v\\simeq0.14$) relative to T2K ($v\\simeq0.05$) makes the term big enough to reshape the predicted event ellipses. This extra interference term supplies the freedom that lets inverted ordering reproduce the data.","core_discovery":"On the paper's own terms, the central discovery is an ambiguity: the same T2K and NOvA appearance data that select normal ordering in the standard three-flavor framework can be described just as well by inverted ordering once a complex flavor-changing coupling εeτ between electron and tau neutrinos is introduced. The best fit in inverted ordering sits at |εeτ| ≈ 0.39 with phase φeτ ≈ 1.30π, and the preference for nonzero NSI in that ordering is 2.5σ; in normal ordering the corresponding preference is only 0.7σ. Because both orderings then fit the data equally well, the inferred ordering is no longer determined by these two experiments.","pith_inferences":["If the large $|\\varepsilon_{e\\tau}|\\simeq0.39$ preferred in inverted ordering is real physics, then the current mass-ordering tension is arguably the first hint of non-standard neutrino interactions, and the normal-ordering preference is an artifact of assuming their absence.","The same $P_2$ mechanism should also contaminate global fits that combine reactor, accelerator, and atmospheric data; a testable extension is to redo the global NMO fit with $\\varepsilon_{e\\tau}$, $\\varepsilon_{ee}$, and $\\varepsilon_{\\mu\\tau}$ varied simultaneously, which could shift or strengthen the ambiguity.","Because the paper fixes solar parameters and treats $\\theta_{13}$ through a reactor prior, the stability of the washout could be re-examined with those parameters free; that is a direct check of whether the conclusion changes under a different statistical treatment.","A practical consequence is that future long-baseline analyses should present the ordering preference as a function of NSI parameters rather than quoting a single significance, since the significance is prior-dependent."],"forward_implications":["Any claim that T2K and NOvA favor normal ordering is conditional on the standard model (or at least on small $\\varepsilon_{e\\tau}$); the 2.4σ preference is not a property of the data alone.","The washout is specific to the e-tau sector: for $\\varepsilon_{e\\mu}$, normal ordering keeps a 2.5σ preference, so the confusion is not generic to all NSI.","DUNE can in principle break the degeneracy, but only if the CP phase $\\delta$ stays close to its current best-fit value around $1.5\\pi$; otherwise its ordering sensitivity drops below 2σ over broad ranges of $\\delta$.","Simulations of T2HK and T2HKK show little ability to resolve the confusion, making DUNE the most promising future long-baseline experiment.","Complementary probes such as atmospheric neutrinos and JUNO, which are less affected by matter-enhanced NSI, become necessary to pin the ordering down."],"supporting_citations":[{"why":"Supplies the latest T2K event samples used in the fit.","marker":"[46]"},{"why":"Supplies the latest NOvA event samples used in the fit.","marker":"[47]"},{"why":"Quoted source of the 90% bound $|\\varepsilon_{e\\tau}|\\lesssim0.36$ that the large inverted-ordering best-fit coupling must evade.","marker":"[7]"},{"why":"More recent global analysis of NSI bounds; its older T2K/NOvA datasets motivate the caveat that the bound may relax once new data are included.","marker":"[39]"},{"why":"Source of the perturbative decomposition of $P_{\\mu e}$ into $P_0+P_1+P_2$ used to isolate the NSI interference term.","marker":"[42]"},{"why":"Definitions of the $f$ and $g$ oscillation functions used in the probability expressions and in the matter-parameter notation.","marker":"[44]"}],"fun_headline_variants":["e-tau NSI erases 2.4σ neutrino mass-ordering hint","Neutrino ordering hint evaporates with non-standard interactions","NOvA/T2K ordering preference vanishes if e-tau NSI allowed","Mass ordering: NSI makes normal versus inverted a coin flip"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The washout relies on allowing $|\\varepsilon_{e\\tau}|$ to rise to about 0.39, well above the 90% bound (≈0.36) quoted from current global analyses; that bound can only be relaxed if the same T2K/NOvA data driving the NSI preference are imported into the constraint, so if independent data hold the coupling below roughly 0.3, the inverted-ordering fit is penalized and the normal-ordering preference survives.","fun_headline_variants_meta":{"raw":{"variants":["e-tau NSI erases 2.4σ neutrino mass-ordering hint","Neutrino ordering hint evaporates with non-standard interactions","NOvA/T2K ordering preference vanishes if e-tau NSI allowed","Mass ordering: NSI makes normal versus inverted a coin flip"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000669,"raw_usage":{"total_tokens":2978,"prompt_tokens":799,"completion_tokens":2179,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":415,"completion_tokens_details":{"reasoning_tokens":2100}},"tokens_in":415,"tokens_out":2179,"duration_ms":18141,"temperature":1.0,"reasoning_tokens":2100,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:29:46.541596+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same T2K+NOvA fit with an external 90% C.L. constraint $|\\varepsilon_{e\\tau}|\\le0.36$ imposed as a hard prior. Since the inverted-ordering best fit sits at $|\\varepsilon_{e\\tau}|\\approx0.39$, that minimum would move to the boundary and pay a penalty of roughly $((0.39-0.36)/\\sigma_{\\rm ext})^2$ in $\\chi^2$; if independent data (e.g., atmospheric neutrinos or coherent elastic neutrino-nucleus scattering) tighten the bound to $|\\varepsilon_{e\\tau}|\\lesssim0.3$, the penalty would exceed 4 and the normal-ordering preference would re-emerge. A published global fit with updated T2K/NOvA data that keeps $|\\varepsilon_{e\\tau}|<0.3$ at 90% C.L. would falsify the washout claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the latest T2K event samples used in the fit."},{"cited_title":"˙Zarnecki, Talk at Les Rencontres de Physique de la Valle d’Aoste, 10-16 March 2019, La Thuile, Italy (2019)","cited_arxiv_id":null,"evidence_quote":"Supplies the latest NOvA event samples used in the fit."},{"cited_title":"Correlations and degeneracies among the NSI parameters with tunable beams at DUNE","cited_arxiv_id":"1812.10290","evidence_quote":"More recent global analysis of NSI bounds; its older T2K/NOvA datasets motivate the caveat that the bound may relax once new data are included."},{"cited_title":"In that case, however, the nature of the new interference term P2 is kinematical, and it is operative also in vacuum","cited_arxiv_id":null,"evidence_quote":"Definitions of the $f$ and $g$ oscillation functions used in the probability expressions and in the matter-parameter notation."}],"review_version":1}