Simon Hiesl ; Prof. Dr. Christian Kiesling

2024
Max-Planck-Institute for Physics Munich

Abstract: The target luminosity of the Belle II experiment, located at the SuperKEKB electron-positron collider in Tsukuba, Japan, is 6x10^{35} cm^{-2}s^{-1}. For this world-record luminosity, high levels of beam background are expected. Thus, an efficient and robust trigger system at the hardware level, capable of selecting annihilation events from the interaction point ("z=0"), is indispensable. For this purpose, a novel track trigger using the wire information of the central drift chamber predicts the z-vertex utilizing a neural network. While this trigger is very successful in suppressing background, recent high luminosity physics runs produced many fake tracks, increasing the trigger rate close to the maximum limit. To keep the L1 trigger rate below the design value of $30 \kHz$, an upgrade of the Neuro Trigger is proposed in this work. In this thesis, the preselection algorithm for the neural network, called the 3DFinder, which creates three-dimensional track candidates using a three-dimensional Hough transformation, is extensively analyzed and upgraded. For the analysis, both simulated Monte Carlo data and real data from the latest Belle II runs are utilized. The upgrade includes a new clustering algorithm and new parameters to reduce the number of fake tracks. The proposed new algorithms for the 3DFinder allow for an implementation on the trigger hardware while achieving high efficiency for single IP tracks and a low fake rate.

Note: Presented on 12 04 2024
Note: MSc