Physics Performance of the ATLAS GNN4ITk Track Reconstruction Chain
Résumé
Graph-based techniques and graph neural networks (GNNs) in particular are a promising solution for particle track reconstruction at the HL-LHC. Simulations of the HL-LHC environment produce noisy, heterogeneous and ambiguous data. We present an upgrade to the ATLAS GNN4ITk pipeline that allows detector regions to be handled heterogeneously. We perform direct comparisons of our results with those of existing tracking algorithms on a range of physics metrics, including reconstruction efficiency, track reconstruction performance in dense environments, and track parameter resolutions. By integrating this solution within the offline ATLAS Athena framework, we also explore different reconstruction chain configurations, for example using the GNN4ITk pipeline together with traditional techniques for track cleaning and fitting.
| Licence |
|---|