Semi-inclusive neutral current neutral pion production selection at the NOvA (numi off-axis electron neutrino appearance) near detector using prong level convolutional neural networks

Abstract

The NOνA neutrino experiment based in Fermilab is designed to measure νµ→νe neutrino oscillations. This experiment will give us insight into the properties of massive neutrinos. Neutral current (NC) νµ,e neutral pion production events can mimic the νµ→νe oscillation signal and therefore are an important background for NOνA to understand. Neutral pions decay into two photons which can fake a single electron shower (νe appearance signal) in two ways: either the 2 photons can merge together or one of them may escape detection. In order to constrain this background, NOνA utilizes the Near Detector to measure neutral current neutrino interactions. In this analysis, neutrino-Nucleus (νµ→N) NC π⁰ interactions with total pion energy greater then 0.3 GeV are studied by selecting two prong events with two final state photons as determined by prong based Convolutional Visual Networks (CVN). The analysis is performed on 3.54x10²¹ Protons On Target (POT) of NOvA Near Detector simulated data and compared to 8.09x10²⁰ POT of data. Optimization of the selection based on fractional cross-section uncertainty and an initial energy resolution study of the final sample are presented. The final 2 prong selection using prong based CVN gave a purity of 74%, selection efficiency of ∼ 1.8%, and an expected (NC) νµ,e neutral pion cross section of ∼ 15.2%.