Master Thesis defense by Simon Christian Debes

Title: Calibration of IceCube simulations through reconstruction of photons

Abstract: This project highlights the data and simulation (MC) agreement in IceCube. A large sample of stopped muons is selected using the graph neural network model DynEdge, and a pulsemerging algorithm is developed to merge pulses that are believed to be incorrectly split in the data recording process. The algorithm is shown to improve the agreement between the charge distributions of data and MC in hard local coincidence hits, and does not affect soft local coincidence hits.
For each DOM in each event a photon vector is reconstructed, making it possible to determine the efficiency of the detector as a function of distance, zenith angle, azimuth angle, and depth.

Each reconstructed photon vector is also used to calculate the stopping time of each of the muons, tA. The variation of tA indicates the detector's resolution and timing and reveals how much light scatters in the ice. The resolution is measured at different distances to the DOMs and is fitted with a linear function, revealing the contribution of the ice to the resolution. The process is done for predicted and true values of MC, which shows how much the reconstruction affects the resolution.