Measurement of Atmospheric Neutrino Oscillation on IceCube using Convolutional Neural Network Reconstructions

The IceCube Neutrino Observatory, instrumenting a cubic kilometer of Antarctic ice, is sensitive to both astrophysical and atmospheric neutrinos. IceCube’s DeepCore subarray extends the detected energy down to GeV scales, providing additional resolution for neutrino oscillation to be observed. Neutrino oscillation is dependent on the distance the neutrino travels (L) and the energy of the neutrino (E). IceCube allows us to probe oscillations in unique ranges of L and E using 8 years of atmospheric neutrinos that travel distances through the earth at the 10 GeV-scale. Typical oscillation analyses on IceCube data use likelihood-based reconstruction methods to determine the incident neutrino’s energy, direction of travel, and interaction vertex. We have recently developed a convolutional neural network (CNN) to replace all likelihood-based reconstructed variables. This talk explores the sensitivity to the muon neutrino disappearance measurement in IceCube using a convolutional neural network reconstruction of the neutrino interactions.