Investigating the theoretical limit of muon track reconstruction in IceCube

Details

Supervisors

de Wasseige, Gwenhaël ; Jansson, Karl

Faculty

Faculté des sciences

Degree label

Master [120] en sciences physiques, à finalité approfondie

Abstract

enAccurate muon-track reconstruction is essential for determining the arrival direction of high-energy neutrinos detected by IceCube. The direction of a muon is inferred from the Cherenkov light recorded by the detector’s Digital Optical Modules. However, the sparse and irregular detector geometry, the depth-dependent optical properties of the ice and the limited amount of detected light make this reconstruction strongly dependent on the event configuration. This thesis investigates the theoretical performance attainable by TriplePandel, a likelihood-based reconstruction framework using a neural-network model and the hypothesis of an infinite muon emitting light continuously along its trajectory. Controlled matched-model closure tests are performed in which event generation and reconstruction use the same geometry, ice description and detector response. The track depth, azimuth, zenith, distance from the detector centre and local position relative to the optical modules are systematically varied. Performance is evaluated through angular-error percentiles, aligned three-dimensional directional bias, detected charge, hit-module multiplicity and reconstruction success. The results show that well-illuminated and well-contained tracks can reach sub-degree median angular errors. This performance is nevertheless configuration dependent. Dust-rich ice, peripheral detector regions and poorly distributed optical information produce broader angular-error distributions, larger directional shifts and, in the most extreme cases, a substantial loss of reconstructed events. Local studies also show that outer detector regions dominate the largest angular-error tails. These results provide a controlled best-case benchmark for the continuous infinite-muon model, rather than an experimental IceCube angular resolution. They identify where the reconstruction performs well, where it becomes limited and which effects should be examined next through realistic stochastic energy losses, model mismatch and direct comparisons with existing reconstruction methods.