Sumitted to PubDB: 2024-04-21

Category: Bachelor Thesis, Visibility: Public

Tags: -

Authors	Stefan Wallner, Stephan Paul, Arina Katscho
Date	Jan. 1, 2024
Belle II Number	BELLE2-UTHESIS-2024-001
Abstract	We develop the event selection of τ ∓ → K∓π∓π±ντ for data from the Belle II electron-positron collider experiment at SuperKEKB, Japan. The aim is that the event selection will be used for further spectroscopy analysis. We tested two event selection approaches to achieve a highly pure τ ∓ → K∓π∓π±ντ selection sample. In the first approach we trained Boosted Decision Trees (BDT) based on the previous preselection with a 3x1 prong topology, i.e. one τ decays to three charged final states particles (3-prong) and the other τ to one charged final state particle (1-prong). To this end, we tuned the hyperparameters of the BDT for an optimal event selection performance. Considering future acceptance corrections, we also tested a second event selection approach where we applied manual cuts on the particle identification (PID) variables and combined them with a BDT that was trained with the remaining variables. Performance comparison showed the first approach achieves significantly higher performance of 80% purity for 6.7% signal efficiency. We also studied acceptance effects and remaining background kinematics in the selected sample. The background is composed of various decays, mainly τ -decays to other hadrons. Noteworthy, the acceptance is not strongly modulated despite the restrictive selection.
Conference	Garching

• BELLE2-UTHESIS-2024-001.pdf (versions: 1)