First sub-electronvolt direct neutrino mass measurement with the KATRIN experiment

The mass of the neutrino is an important parameter in astroparticle physics and cosmology. While neutrino oscillation experiments have proven that neutrinos have mass, they are not sensitive to its absolute scale. The Karlsruhe TritiumNeutrino (KATRIN) experiment is designed to measure the effective electron anti-neutrino massmν with a sensitivity of 200 meV at 90% confidence level (CL) using the kinematics of tritium β-decay. First data used for neutrino mass analysis was taken in spring 2019, with a second measurement phase following in autumn of the same year. This thesis explains a novel approach to infer the neutrino mass confidence interval from the data based upon full Monte Carlo propagation of uncertainty. We show that our approach gives consistent results with the well-known nuisance parameter method in all cases currently used by the KATRIN collaboration. Applying our Monte Carlo propagation approach to the data of the first two measurement campaigns, we find a combined best fit value of $m_{\nu }^2$ = 0.12${}_{-0.33}^{+0.32}$ eV${}^2$ including all relevant statistical and systematic uncertainties. This leads to the first sub-electronvolt laboratory measurement of the neutrino mass, limiting it to $m_{\nu }$< 0.8 eV (90% CL). In addition, we describe a future proof analysis method by approximating the KATRIN physics model with a neural network (NN). Our NN model shows no significant bias at the statistical sensitivity expected for KATRIN and is able to reproduce the results of analysing the first two measurement phases while reducing the computation time by several orders of magnitude. This makes it a promising approach to analyse future KATRIN neutrino mass data.