000003620 001__ 3620
000003620 005__ 20230526083515.0
000003620 037__ $$aBELLE2-MTHESIS-2023-021
000003620 041__ $$aeng
000003620 100__ $$aElia Schmidt
000003620 245__ $$aDeveloping a Displaced Vertex Trigger for Dark Matter Searches at the Belle II Experiment
000003620 260__ $$aMunich$$bMax-Planck-Institute for Physics$$c2023
000003620 300__ $$a104
000003620 500__ $$aPresented on 13 03 2023
000003620 502__ $$aMSc$$bMunich, Ludwig-Maximilians-University$$c2023
000003620 520__ $$aFor many well-motivated dark matter models, discovery potential is predicted at highintensity particle colliders such as the Belle II experiment. A particularly interesting scenario is inelastic dark matter which features a long-lived scattering partner in addition to the dark matter particle. At Belle II it manifests itself through the signature of large missing energy with two associated tracks that originate at vertices far displaced from the primary collision point with almost no Standard Model background. However, in large parts of parameter space the Belle II trigger system is inefficient, and by exhausting the capabilities of the detector, new limits could be set on many dark matter models. In this thesis, a trigger system is presented that significantly increases the sensitivity of the Belle II experiment towards displaced vertex signatures. Like the standard track triggers of the experiment, the displaced vertex trigger relies on the Hough transform of 2D hits into a generalised parameter space, where tracks can be found using clustering algorithms. The Hough transform is performed in parallel on multiple vertex hypotheses selecting a subgroup of probable vertex candidates. These candidates are then analyzed by a Neural Network in order to reliably discriminate background tracks. The algorithm is written in python and prepared to be implemented on programmable trigger boards installed at the Belle II facility, guaranteeing a low latency and a fast trigger response. The trigger is tested in Monte Carlo (MC) simulations on example models, achieving an event efficiency between 40% and 80% at a fake rate of 1kHz. The efficiency does not significantly change in the entirety of the available parameter space and is found to be robust despite challenging background conditions.
000003620 700__ $$aProf. Dr. Christian Kiesling$$edir.
000003620 8560_ $$fcmk@mpp.mpg.de
000003620 8564_ $$uhttps://docs.belle2.org/record/3620/files/BELLE2-MTHESIS-2023-021.pdf
000003620 980__ $$aTHESIS