The algorithm of the CMS Level-1 Overlap Muon Track Finder trigger

Abstract

The CMS Level-1 trigger, based on custom electronics built around FPGA devices, was upgraded in 2016 to achieve the required performance with almost two times higher LHC luminosity than originally designed. The upgraded Level-1 muon trigger merges data from the three muon detectors in the CMS (DT, CSC, and RPC) in the track reconstruction stage — contrary to the legacy trigger which comprised three separate systems each based on one muon detector. This approach allows better use of the detector redundancy and improved measurement of the muon transverse momentum . However, it is particularly challenging in the barrel-endcap transition region, where up to 18 muon chamber layers are present, the detector geometry is complex, and the magnetic field bending the muon tracks is heterogeneous. For this region the Overlap Muon Track Finder (OMTF) uses a novel algorithm based on a naive Bayes classifier. The algorithm identifies the muon tracks and measures their momentum by calculating the probabilities of matching the detector hits to defined transverse momentum hypotheses. The algorithm was tailored to the needs of muon measurements based on the detector data (e.g. to cope with missing hits or multi-muon events) and implementation in FPGA technology. The algorithm details, performance, and optimization methods are discussed in this report. • A novel algorithm was developed for the upgrade of the CMS L1 muon trigger. • It is based on a naive Bayes classifier - a classic machine learning algorithm. • The algorithm performance is significantly better than the legacy muon trigger.