Speaker: Joshua Henry Peterson, Physics PhD Graduate Student
Abstract: The electrons in Earth's matter affect the oscillation of atmospheric neutrinos and antineutrinos differently depending on the neutrino mass ordering. As more neutrinos than antineutrinos are expected to be detected in the IceCube Neutrino Observatory, this matter effect can be used to probe the NMO. The fraction of energy transferred to the nucleon during a neutrino interaction, known as the inelasticity, has a different distribution for neutrinos and antineutrinos. This can in theory be used to statistically separate neutrinos from antineutrinos, but hasn't been exploited in IceCube DeepCore or the upcoming IceCube Upgrade. Processing photomultiplier tube voltage waveforms from detections of photons is a computationally intensive process that needs to be done before any science can proceed. We explore methods to port this process into the digital optical modules of future IceCube expansions, and find multiple ways to increase the speed of the waveform processing. The processed photon information can then be used for neural network reconstructions. New two dimensional convolutional neural networks are developed for neutrino flavor identification and inelasticity reconstruction in IceCube DeepCore, which outperform other convolutional neural network algorithms. Model stacking methods with boosted decision trees are developed to combine neural network algorithms for further improve performance. We also developed a graph neural network for inelasticity reconstruction for the IceCube Upgrade. The inelasticity reconstructions are then used as a fourth observable, along with the particle energy, direction and neutrino flavor, to calculate new neutrino mass ordering sensitivities and determine the impact of adding the inelasticity in the measurement of the neutrino mass ordering with IceCube DeepCore and upcoming IceCube Upgrade detectors. It is found that incorporating inelasticity improves the sensitivity to the neutrino mass ordering for both DeepCore and the Upgrade, for both the normal and inverted orderings, especially for the normal ordering in the second octant of $\theta_{23}$. We also find that small improvements to the current IceCube Upgrade inelasticity reconstruction can lead to significant increases in the sensitivity to the neutrino mass ordering.