Track Reconstruction using Geometric Deep Learning in the Straw Tube Tracker (STT) at the PANDA Experiment

TL;DR

This paper introduces a geometric deep learning pipeline for track reconstruction in the PANDA experiment's Straw Tube Tracker, achieving over 95% accuracy and minimal fake tracks, improving particle detection in complex detector geometries.

Contribution

It is the first application of geometric deep learning to track reconstruction in the PANDA experiment, demonstrating high efficiency and low fake rate.

Findings

Reconstructs over 95% of particle tracks

Creates less than 0.3% fake tracks

Shows promise as a candidate algorithm for PANDA

Abstract

The PANDA (anti-Proton ANnihilation at DArmstadt) experiment at the Facility for Anti-proton and Ion Research is going to study strong interactions at the scale at which quarks are confined to form hadrons. A continuous beam of antiproton, provided by the High Energy Storage Ring (HESR), will impinge on a fixed hydrogen target. The antiproton beam momentum spans from 1.5 GeV {Natural units, c=1} to 15 GeV \cite{physics2009report}, will create optimal conditions for studying many different aspects of hadron physics, including hyperon physics. Precision physics studies require a highly efficient particle track reconstruction. The Straw Tube Tracker in PANDA is the main component for that purpose. It has a hexagonal geometry, consisting of 4224 gas-filled tubes arranged in 26 layers and six sectors. However, the challenge is reconstructing low momentum charged particles given the complex detector geometry and the strongly curved particle trajectory. This paper presents the first application of a geometric deep learning pipeline to track reconstruction in the PANDA experiment. The pipeline reconstructs more than 95\% of particle tracks and creates less than 0.3\% fake tracks. The promising results make the pipeline a strong candidate algorithm for the experiment.