The T2K experiment in Japan has released new results that bring together improved measurements of how neutrinos change type and detailed studies of how they interact with nuclei. The work uses a larger dataset and the first measurements from Super‑Kamiokande after loading its water with gadolinium. The team emphasizes that precise interaction data from near detectors is essential to reduce the main systematic uncertainties in oscillation results.

T2K sends a beam of muon neutrinos 295 km from the J‑PARC accelerator to the Super‑Kamiokande (SK) water detector. The beam is aimed 2.5° off axis so its energy peaks near 0.6 GeV, where oscillation effects are largest. SK has been loaded with 0.03% gadolinium sulfate by mass. Gadolinium improves tagging of neutrons produced in interactions. Neutron tagging gives a new handle to tell neutrino and antineutrino events apart and to reduce some background from atmospheric neutrinos.

A complex of near detectors sits 280 m from the production point to measure the beam before oscillation and to study neutrino–nucleus interactions. The suite includes the magnetized ND280 tracker and the WAGASCI–BabyMIND system, and an upgraded ND280 is already taking data. Early studies with the upgraded detector show large gains: the photon background in the electron‑neutrino charged‑current selection drops from about 30%, and some muon samples reach about 90% purity, including for backward‑going tracks. These improvements are expected to tighten constraints on interaction models used in the oscillation analysis.

The oscillation analysis uses 21.4×10^21 protons on target (POT), a 9% increase in neutrino‑mode data. Including the first gadolinium‑loaded SK results, the nominal T2K-only analysis excludes exact charge‑parity (CP) conservation at the 90% confidence level. The data show a mild preference for the normal mass ordering (Bayes factor NO/IO = 3.3) and a slight preference for the upper octant of the mixing angle θ23 (Bayes factor 2.6). The atmospheric mass splitting |Δm^2_32| is determined to about 2% precision at 1σ, with a central value near 2.5×10^−3 eV^2. Joint fits that combine T2K with NOvA and with SK atmospheric data give additional, though not decisive, pulls: one joint fit prefers inverted ordering and another prefers normal ordering, and the exclusion of CP conservation ranges from about 1.9σ to 3σ depending on the dataset and assumptions. The θ23 octant ambiguity remains unresolved.