Improving the angular reconstruction uncertainty of IceCube’s real-time alerts.

Extremely high energy astrophysical neutrinos are essential to modern day multimessenger astronomy, revealing unique information about their origins: violent cosmic phenomena such as supernovae and black holes. The IceCube Neutrino Observatory was designed to detect these astrophysical neutrinos in order to study their properties and origins. Located beneath the South Pole, IceCube observes Cherenkov light emitted by products of neutrino-ice interactions with a cubic kilometre array of optical sensors. The IceCube real-time alert system notifies telescopes and observatories around the world when a high energy neutrino, likely to have originated from an astrophysical source, is detected in the ice. IceCube’s full sky sensitivity allows highly directional telescopes to observe spatially localised transient phenomena they would otherwise miss. The angular uncertainty of IceCube alerts is crucial for the correlation of associated astrophysical objects. Ascertaining an accurate estimate of a neutrino’s trajectory from Cherenkov light flashes is non-trivial. Reconstruction algorithms generate a likelihood space of possible neutrino trajectories, representing the probability a neutrino with a certain trajectory and energy would have generated the light output observed by the detector. The likelihood space is rarely simple, often containing more than one local maximum. Whilst a global minimum can be found, no procedure exists to map likelihoods onto p values. Instead, large sets of simulated events are used for calibration. Approximate p values can be generated for events according to the distribution of simulated likelihood landscapes. This is expected to depend on an event’s properties. Currently, all real-time alerts are calibrated from the same set of simulated events. This set mimics the energy and trajectory of an historical IceCube event, IC160427A. In this thesis we investigate the relationship between an event’s properties and how accurately its trajectory is reconstructed. We establish which properties are most relevant and begin the process of simulating sets of events to represent these properties in future reconstructions. These sets will represent the spread in event properties observed in the IceCube real-time alerts, such as neutrino energy, trajectory, depth within the detector, and energy loss stochasticity. The final product will allow real-time alerts to have a more accurate mapping from their likelihood spaces to p values, enabling more formal and accurate statistics for IceCube and other observatories globally