Muon neutrino to tau neutrino oscillations with the OPERA experiment

The OPERA experiment played the key role of demonstrating unambiguously the muon neutrino to tau neutrino oscillation. It took data from 2008 to 2012, integrating a beam intensity of 1.8*10^20 protons on target. Five tau neutrino candidates were found satisfying stringent kinematical cuts, with 0.25 expected events of background: the absence of signal was excluded with a significance of 5.1?, thus claiming the discovery of muon neutrino to tau neutrino oscillation in the CNGS beam. I contributed to the validation of the fifth tau neutrino candidate, in the Naples emulsion Laboratory, with a specific analysis for large angle track search (tan(?)< 3 rad) that I had implemented, meant to strongly suppress the hadron re-interaction background hypothesis. Given the full Monte Carlo validation with data in all corners of the parameter space, a likelihood approach is well justified and therefore the Collaboration has decided to define a new strategy with looser cuts for the selection of tau neutrino candidates. Looser cuts have the advantage of reducing the statistical uncertainty on the measurement of ?m^2_{23}. I defined a new strategy for the tau neutrino selection, replacing the stringent kinematical cuts with minimum bias ones, optimised to minimise the statistical uncertainty on the measurements. Applying these criteria, 6.5 tau neutrino candidates and 2.0 background events are expected. At the end of data taking, 5603 neutrino events had been fully analysed and have been used for this analysis. Within this sample, 10 tau neutrino candidate events satisfied the minimum bias criteria. The number of observed tau neutrino candidates after background subtraction is a function of the product of tau neutrino cross-section and ?m^2_{23}. Assuming the maximal mixing sin^2 2(?_{23})=1, and fixing tau neutrino cross-section to the SM value, it is possible to measure ?m^2_23, while fixing ?m^2_23 to the present global best fit it is possible to measure tau neutrino CC cross-section. The measurement of ?m^2_{23}, done for the first time in appearance mode, is (2.8±0.6)*10^-3 eV^2 at 68% C.L. The result is consistent within 1? with the measurements done in disappearance mode by other experiments and with the PDG 2016 best fit. tau neutrino cross-section is proportional to the energy, with a constant term?0 and a kinematic suppression factor K(E), which takes into account the suppression due to the heavier tau lepton mass. I did the first ever evaluation of?0, so far measured only by the DONuT experiment, with no separation between tau neutrinos and tau anti-neutrinos. It results to be (8+4-3)*10^-39 cm^2 GeV^-1, compatible with the Standard Model prediction within 1?. To perform signal to background separation, I used the Boosted Decision Tree (BDT) method, a multivariate machine learning method used to classify observations. With this analysis, two events in the tau?1h decay channel were classified as background-like, in agreement with expectations, while the other eight were signal-like. Combining the Poisson counting method and BDT results in the tau?mu decay channel, including also the background due to the muon anti-neutrino beam contamination, I have obtained the first observation of the leptonic number with tau neutrinos with a significance of more than 3?. Finally, I evaluated the significance of the observation of neutrino oscillations with this new analysis using three different approaches: the first one based on Poisson statistics and p-values, the second one on the likelihood ratio and the third one combining in the extended likelihood function the Poisson statistics and the BDT probability density functions. All results lead to a significance above 5?.