Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 69: DAQ and Trigger 3

T 69.9: Vortrag

Mittwoch, 23. März 2022, 18:15–18:30, T-H28

Jet Tagging in the Level-1 Trigger of CMS for the HL-LHC — •Philipp Rincke, Karim El Morabit, Gregor Kasieczka, and Artur Lobanov — Institut für Experimentalphysik, Universität Hamburg

With the upcoming upgrade of the High-Luminosity LHC, triggering in the CMS experiment will become more challenging as more particles will be present in each event. A possible solution to the increased complexity could be to employ trigger algorithms that use inputs from all sub-detectors that will become available in future upgrades and the increased computing power of the FPGAs on which the algorithms of the Level-1 Trigger (L1T), the first trigger level in CMS, are implemented.

At the L1T track parameters and particle identification of some of the jet constituents will be available, making it possible to evaluate the jet substructure. Many high-energy physics analyses require jet flavour identification, for which the substructure information can be exploited. By attempting this in the L1T a higher fraction of interesting events could be recorded or thresholds could be lowered. One big challenge is that neural networks, often used in jet tagging, are not straightforward to deploy on FPGAs.

Besides the tagging performance, strict timing and resource limitations need to be considered, which results in a compromise between network architecture and size. In this talk we present studies on how jet tagging can be used in the L1T. We consider a simple multilayer perceptron architecture as well as graph-inspired network architectures.