Göttingen 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 18: Methods in Particle Physics I (Calo, Jets, Tagging)
T 18.6: Vortrag
Montag, 31. März 2025, 18:00–18:15, VG 4.101
Material interactions in ATLAS jet flavour tagging — Diptaparna Biswas1, Beatrice Cervato1, Markus Cristinziani1, Carmen Diez Pardos1, Ivor Fleck1, Arpan Ghosal1, Gabriel Gomes1, Jan Joachim Hahn1, Vadim Kostyukhin1, •Nils Krengel1, Buddhadeb Mondal1, Stefanie Müller1, Sebastian Rentschler1, Elisabeth Schopf1, Katharina Voss1, Wolfgang Walkowiak1, Adam Warnerbring1, and Tongbin Zhao1,2 — 1Experimentelle Teilchenphysik, Center for Particle Physics Siegen, Universität Siegen — 2Shandong University, China
Jet flavour tagging plays a crucial role in understanding particle physics processes. In the continuous effort to enhance flavour tagging performance the ATLAS Collaboration is currently deploying deep learning transformer models.
Jets originating from a b-quark are easy to tag because of the characteristic bottom hadron decays. Due to long lifetimes, B-hadrons decay far from the primary event vertex, producing a significant number of tracks with big impact parameters. However, this feature can be mimicked by interactions of particles with the detector material, also producing displaced tracks.
This presentation will demonstrate how material interactions may lead to the misidentification of jets originating from quarks of lighter flavour as b-jets, and it will discuss first results of an attempt to mitigate the influence of material interactions. This attempt consists of adding an auxiliary task, which identifies these interactions, to the flavour tagging machine learning model.
Keywords: Jet flavour tagging; Detector material interaction; Machine learning; Auxiliary task; Transformer model