arXiv2024
While the origin of neutrino masses remains unknown, several key neutrino mass generation models result in a non-unitary three-neutrino mixing matrix. To put such models to test, the deviations of the mixing matrix from unitarity can be measured directly through neutrino oscillation experiments. In this study, we perform a Bayesian analysis of the non-unitary mixing model using the recent public data from atmospheric and reactor neutrino experiments - namely IceCube-DeepCore, Daya Bay, and KamLAND. The novelty of our approach compared to the preceding global fits for non-unitarity is in the detailed treatment of the atmospheric neutrino data, which for the first time includes the relevant flux and detector systematic uncertainties. From the Bayesian posteriors on the individual mixing matrix elements, we derive the non-unitarity constraints in the form of normalisations and closures of the mixing matrix rows and columns, assuming either a fully unconstrained matrix or a physically motivated submatrix scenario. We find comparable constraints for electron and tau row normalisations as other similar studies in literature, and additionally reveal strong correlations between muon and tau row constraints induced by the atmospheric systematic uncertainties. We find that the current data is well described by both unitary and non-unitary mixing models, with a strong preference for the unitary mixing indicated by the Bayes factor. With the upcoming IceCube-Upgrade and JUNO detectors, both featuring superior energy resolution compared to the current atmospheric and reactor neutrino experiments, our constraints on the row normalisations in the submatrix case are expected to improve by 25%, 40%, and 20% in the electron, muon, and tau sectors respectively.