Until recently, neutrinos were believed to be massless — a prediction of the standard model. The SNO experiment measured electron neutrinos created in the sun turning into muon and tau neutrinos. This, combined with data from the Super-K experiment definitively proved that neutrinos change flavour, an effect called neutrino oscillation. Oscillation can only happen if neutrinos have mass, a discovery for which the 2015 Nobel prize was awarded.
SNO and Super-K used natural sources of neutrinos: solar and atmospheric. Contemporary to this were searches for oscillation using man-made neutrinos. LSND was one such experiment, using the LAMF particle accelerator at Los Alamos National Laboratory. The experiment saw evidence of muon anti-neutrinos turning into electron anti-neutrinos.
Neutrino oscillation depends on the difference in the masses of the neutrinos. This parameter is called the mass splitting, and SNO and Super-K constrain two splittings to be small. The LSND data, on the other hand, requires a very large mass splitting, as the detector is located a short distance from the beam target. This leads to a contradiction with only three types of neutrinos, and for this reason the LSND result is called a “short base-line anomaly”. Adding a fourth, much heavier kind of neutrino would allow this large mass splitting.
This fourth kind of neutrino is called a sterile neutrino, as it does not interact with the strong, weak or electromagnetic forces. It’s existence must be inferred through their effect on the standard neutrino oscillations. While LSND is the most well known signal for sterile neutrinos, there are more experiments that have tested this hypothesis. Some have seen evidence (such as Bugey and Gallex/SAGE), while others have not (such as IceCube).
To make sense of all this data, Janet Conrad’s group does a simultaneous global fit of data from many experiments. Building on the previous work of previous students (Christina Ignarra, Georgia Karagiorgi and Michel Sorel), I rewrote the group’s fitting framework in C++. I used a parallel tempering affine invariant Markov chain Monte Carlo based on the Emcee package to sample the parameter space. This has allowed us to show Bayesian credible regions alongside frequentist confidence regions for the first time.
The results of fits to 3+1 and 3+2 models shows that the sterile neutrino hypothesis provides a better fit to the data when compared to the null hypothesis (no sterile neutrinos). However, there is significant tensions in the data which must be resolved before we can claim the existence of a sterile neutrino.
IceCube’s latest sterile neutrino search has given the best limit on muon neutrino disappearance to date. This result is important for constraining the theoretical models behind sterile neutrinos.
Our latest fitting paper shows how IceCube’s data changes the global fit landscape. While the best fit point remains constant, the 90% confidence regions have been moved to higher mass splittings, favouring heavier sterile neutrinos. The IceCube analysis is also sensitive to the mixing between sterile and tau neutrinos. This allows us to show the complete 3+1 mixing matrix for the first time.
