Erlangen 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 28: Data, AI, Computing, Electronics III

T 28.8: Talk

Tuesday, March 17, 2026, 18:00–18:15, KH 00.024

Secondary Particle Tracking with Graph Neural Networks for the ATLAS Experiment — •Hannah Schlenker, Sebastian Dittmeier, and André Schöning — Physikalisches Institut, Universität Heidelberg, Germany

The increased number of simultaneous proton-proton collisions in the upcoming High Luminosity LHC will increase the computational demands for charged particle track reconstruction in the new Inner Tracker (ITk) of the ATLAS Experiment. To reduce computing resources, the usage of parallel architectures like GPUs are investigated, and new track reconstruction algorithms based on machine learning are in development. Track finding on the basis of Graph Neural Networks (GNNs) has been shown to be promising [1]. This method constructs a graph based on all hits of an event using a module map, assigns a score to each edge and finds track candidates based on these scores.

Previous work focused on particle tracks originating near the interaction region [1]. Secondary particle tracks, which are produced all over the detector, have not been targeted. A good track finding efficiency for these particles is important for enhanced energy measurements, characterisations of material interactions and potential searches of long-lived particle decays.

This talk discusses the secondary tracking performance of current models and of new models developed specifically for this task.

[1] ATLAS Collaboration, Optimizations of the ATLAS ITk GNN reconstruction pipeline, tech. rep., CERN, 2025, URL: https://cds.cern.ch/record/2948192

Keywords: Track Finding; Machine Learning; Graph Neural Networks