Joint neutrino oscillation analysis from the T2K and NOvA experiments

The paper is internally coherent as a statistical combination of two oscillation analyses within the standard three-flavor framework. Definitions of oscillation parameters are consistent throughout: Δm^2_ij ≡ m_i^2 − m_j^2, normal/inverted ordering are used consistently, and the conditioning of results on ordering is clearly maintained. The distinction between results marginalized over mass ordering and results conditional on a fixed ordering is mostly respected; for example, δ_CP intervals are separately reported for normal and inverted ordering, and Bayes factors are explicitly quoted for ordering and octant comparisons. The use of a shared oscillation parameter set with mostly independent nuisance parameters is logically consistent with the stated assumptions after the correlation studies. The paper also consistently states that conclusions about CP violation in the lepton sector are conditional on assuming inverted ordering, which avoids overclaiming.

The main logical weakness is that several important choices are justified empirically rather than derived from a unified probabilistic model. Most notably, the decision to neglect almost all inter-experiment systematic correlations is based on stress tests ('individual parameter correlations,' 'nightmare parameters,' and out-of-model variations) rather than on an explicit joint covariance construction. This is not a contradiction, but it does mean the logical chain is conditional: the inference is valid only insofar as those tests are representative. Similarly, the statement that values around +π/2 are outside the 3σ credible interval uses Bayesian highest-posterior-density intervals while also speaking the language of '3σ'; that is acceptable if interpreted as 99.73% credible regions, but it mixes Bayesian and Gaussian-style terminology in a way that could confuse interpretation. Overall, however, I do not see internal contradictions in the mathematical-statistical structure presented.

Core formulae and definitions that do appear are mathematically standard and dimensionally consistent: the oscillation-phase structure sin^2(Δm^2_{32} L /(4E)) is dimensionless in natural units (ℏ=c=1), and the sign convention for Δm^2_{ij} is explicitly defined. The Jarlskog invariant expression shown in Fig. 4 caption, J_CP = sinδ_CP · cosθ23 sinθ23 · cosθ12 sinθ12 · sinθ13 cos^2θ13, matches the usual PDG form (equivalently (1/8)sin2θ12 sin2θ23 sin2θ13 cosθ13 sinδ_CP), so no algebraic error is evident there.

However, the submission (as provided) is predominantly a descriptive experimental-combination paper and omits many mathematical details necessary for a full rigor check: the explicit likelihood function, the precise parameter vector, the nuisance-parameter priors/covariances, the treatment of discrete ordering as a model index in Bayes factors, and the definition of the “posterior predictive p-value” test statistic are not written out. Consequently, many key claims (e.g., the exact Bayes factor values 1.3, 2.5, 3.5; the 3σ credible intervals; the statement that “values around +π/2 are outside our 3σ credible intervals”; the goodness-of-fit p-values and look-elsewhere/Bonferroni procedure) cannot be independently verified from equations in the text excerpt.

There are also a few technical-mathematical ambiguities: (i) The prior specification says “flat in Δm^2_{32}” but the ordering is a discrete choice with Δm^2_{32} sign; a flat prior over a signed interval together with a separate ordering hypothesis can double-count or require a precise statement of how the prior mass is split between NO and IO. They do report Bayes factors conditional on using different reactor constraints, but the mathematical definition of the prior over the model index (ordering) is not given. (ii) The “flat in sinδ_CP” alternative prior is mentioned; that is not a smooth one-to-one reparameterization on [−π,π] because sin is not injective, so “flat in sinδ_CP” must be defined carefully (piecewise density in δ_CP) to be mathematically well-posed; the paper asserts robustness but does not specify the induced density. (iii) Several quantitative thresholds for out-of-model tests (≤10% interval-width change; center shift ≤50% of systematic uncertainty) are heuristic criteria; that is acceptable, but they are not derived and do not constitute a mathematical guarantee of coverage.

Thus, what is written is largely correct, but the mathematical substantiation of the statistical procedure is incomplete in the provided text, limiting validation to spot-checks of definitions and dimensional consistency.

This work makes clear, testable predictions about neutrino oscillation parameters that can be experimentally verified. The paper provides specific quantitative measurements (e.g., ∆m²₃₂ = 2.43⁺⁰·⁰⁴₋₀.₀₃ × 10⁻³ eV² for normal ordering) and defines clear criteria that would falsify specific hypotheses. The analysis explicitly states what would provide evidence against current assumptions (e.g., CP symmetry violation in the inverted ordering case). The goodness-of-fit tests with specific p-value thresholds demonstrate rigorous falsifiability criteria. While this is experimental data analysis rather than new theoretical predictions, the methodology allows for clear discrimination between competing models of neutrino behavior.

For a technically dense large-collaboration paper, the manuscript is generally clear and well organized. The scientific motivation is stated early, the complementary roles of T2K and NOvA are communicated effectively, and the bi-event discussion gives readers an intuitive picture of why a joint analysis helps disentangle δ_CP and mass ordering effects. The paper also does a commendable job distinguishing core results from methodological detail, moving many technical robustness studies into the Methods while still summarizing their conclusions in the main text. Importantly, it is careful with conditional statements—for example, that the CP-violation evidence arises only if inverted ordering is assumed within the three-flavor framework. That said, clarity is somewhat limited by the sheer scale and specialization of the submission. Some conclusions about systematic-correlation handling rely on dense prose and would benefit from a more compact schematic summary of what is correlated, what is uncorrelated, and why. The communication around Bayesian evidence, credible intervals, and the meaning of 'evidence for CP violation' could also be more explicit for readers outside the immediate subfield, especially to avoid confusion between conditional inference and an unconditional discovery claim.

The main novelty is not a new physical theory but the first fully joint oscillation analysis of T2K and NOvA data using integrated likelihood machinery from both collaborations. That is a meaningful and nontrivial contribution: it goes beyond a qualitative comparison or posterior-level overlay and instead combines the experiments at the likelihood level while preserving each collaboration's detector response, near-detector constraints, nuisance treatment, and MCMC infrastructure. The manuscript also contributes methodological novelty through its containerized cross-framework inference setup and its explicit program of testing whether inter-experiment systematic correlations can be neglected at current exposure. The 'nightmare parameter' studies and cross-framework redundancy are particularly original as communication-worthy methodological elements. This is not a foundationally new oscillation paradigm, so the novelty should not be overstated; the physics model remains standard three-flavor oscillation phenomenology with external reactor/solar constraints. But within experimental neutrino analysis, the synthesis and execution appear genuinely new and scientifically valuable.

This paper is highly complete for its stated goal: performing and interpreting the first joint oscillation analysis of T2K and NOvA data within the standard three-flavor framework. The central physics quantities are defined before use (mixing angles, mass-squared splittings, δ_CP, mass ordering, Bayes factors, credible intervals), the statistical framework is explicitly described, and the paper clearly states what constraints are applied externally (reactor θ13, plus θ12 and Δm^2_21). The manuscript does not merely report headline numbers; it explains why a joint analysis is useful, how the two experiments are combined technically, how likelihoods and priors are handled, and how redundancy between the ARIA and MaCh3 frameworks is used as a cross-check.

A major strength in completeness is the treatment of systematic uncertainties and inter-experiment correlations, which are the most obvious place where a joint analysis could otherwise be underdeveloped. The paper explicitly separates flux, detector, and cross-section uncertainties; explains why some are treated as uncorrelated and one ν_e/ν̄_e normalization term is correlated; and then performs multiple stress tests: individual-parameter correlation studies, constructed 'nightmare' parameter studies, out-of-model variations, goodness-of-fit tests, and alternate-prior checks. Limitations are also stated rather than hidden: conclusions about neglected correlations are explicitly said to apply to the current exposures/models and may need reevaluation for future data sets. Edge cases such as normal vs inverted ordering, with/without reactor constraint, flat prior alternatives, and rate-only versus full spectral fits are addressed. The only minor incompleteness is practical rather than logical: some implementation details are necessarily deferred to prior NOvA/T2K papers and collaboration software, and raw data/code are not publicly distributed, which limits independent reproduction. But as a scientific paper reporting a collaborative experimental analysis, the argument itself is fully developed and well-supported internally.