Speaker
Description
Decay products of long-lived particles are an important signature in dark sector searches in collider experiments. The current Belle II tracking algorithm is optimized for tracks originating from the interaction point at the cost of a lower track finding efficiency for displaced tracks. This is especially the case for low-momentum displaced tracks that are crucial for dark sector searches in low mass regions.
For the expected high beam-background and high occupancy in the upcoming data taking of the Belle II experiment, new track finding methods must be developed. Graph Neural Networks (GNN) show promising tracking results in a high occupancy environment, which is especially challenging for low-momentum particle tracks.
We develop a GNN-based tracking algorithm for the drift chamber tracking detector of the Belle II experiment. A graph representation of detector hits, including the 2D position, timing information and pulse height, is used to find tracks and assign hits to them. In order to identify the varying number of tracks per event, we use GNN-based object condensation for track finding.
The goal of this project is to improve the track finding for the offline analysis of displaced tracks in the Belle II Analysis Software Framework. Furthermore, we also implement track fitting simultaneously to the track finding, to investigate if this GNN approach can also be used in real-time application in the level 1 trigger system.
| Consider for long presentation | Yes |
|---|
