• neutrino
• multimessenger astrophysics
• IceCube
• statistical analysis

IceCube is a neutrino observatory with 5160 optical sensors embedded in 1 km^3 of the Antarctic ice sheet close to the Amundsen-Scott South Pole Station. IceCube detects neutrino interactions with the surrounding ice or nearby bedrock through detection of Cherenkov radiation of charged secondary particles produced in the interaction. The IceCube Neutrino Observatory announced a significant detection of a diffuse astrophysical neutrino flux above 30 TeV in 2013. However, having not yet identified any sources at the 5-sigma level, their origin is still unknown. Multimessenger observations can play a significant role in this search. IceCube identifies events with a higher probability of being astrophysical and updates the astronomical community with realtime alerts acting as triggers for follow-up observations. However, an accurate localization with appropriate angular errors is necessary.
Neutrinos have three flavors, electron, muon, and tau, and can interact with charged or neutral currents (CC or NC). In the case of a CC interaction, the respectively charged lepton is produced. At high energies, the charged lepton and neutrino directions are strongly correlated. A high-energy muon (with energies larger than TeV) can travel in ice for several kilometers and emit light along its track via the Cherenkov effect and indirectly from the particles produced in stochastic energy losses. Thus, the recorded light from an ultra-relativistic muon has a track signature. The best estimations of neutrino directions result from the reconstruction of muon tracks.
The system currently used by IceCube to estimate angular uncertainties is known as Millipede. Maximizing a likelihood, Millipede fits the stochastic energy losses among the muon's track. This system implemented in IceCube has two main disadvantages: it is hard to study and characterize using large Monte Carlo samples because of its high computational cost, and it strongly depends on systematic uncertainties. It is possible to reconstruct the direction using a completely different system. Instead of considering the stochastic energy losses, it is possible to fit the Cherenkov light emission of the neutrino-generated muon. By assigning primary relevance only to the first-detected photons, this system provides results, which are more robust and less dependent on systematic uncertainties, besides being much faster. This alternative reconstruction method is known as SplineMPE, and this work explores its application to the problem of angular-uncertainty estimation in IceCube realtime alerts.
The main results of the thesis are the implementation of SplineMPE developed in this work for studying angular errors in the last section of chapter 3 and the comparison, in chapter 4, of the estimation of angular uncertainties and their dependence on systematics using the two reconstruction methods: Millipede, explored in precedent studies, and SplineMPE, studied exclusively in this work. The structure of the thesis is the following: In chapter 1, I introduce the IceCube Neutrino Observatory and describe the hardware that permits, from photons emitted by ultra-relativistic particles in ice, to obtain the data of an IceCube event. In chapter 2, I introduce the various muon-track-reconstruction algorithms used in IceCube and describe in detail the various analysis anticipating the final estimation performed in real-time. In Chapter 3, I cover the different approaches used in IceCube for estimating the angular uncertainty relative to the reconstructed direction. I explore first the simplest and fastest ones. Then, I introduce a likelihood scan, which is a more sophisticated and computationally expensive system currently used for uncertainty estimation, together with the Millipede reconstruction algorithm. In the last section of chapter 3, I introduce the implementation of the likelihood scan with SplineMPE (the one defined and used in this work). Chapter 4 compares Millipede and SplineMPE on their dependence on systematics and on their estimation of angular uncertainties.
The outcomes of the applications of this work to IceCube data show that, besides being faster and more robust, SplineMPE produces even much tighter containment regions. On the one hand, smaller angular uncertainties give much more relevance to the astronomical phenomena within the uncertainty areas. On the other, this could exclude many sources from the follow-up studies. However, even if less than Millipede, this system still depends on systematic uncertainties. As this could strongly affect follow-up observations, it will be necessary in future studies to characterize SplineMPE's dependence on peculiar event features which can build on the results of this work.