Neural network-based algorithm for the identification of bottom quarks in the CMS Phase-2 Level-1 trigger

The Phase-2 upgrade of the CMS detector for the High Luminosity upgrade of the LHC (HL-LHC) includes the introduction of many new capabilities into the Level-1 trigger, including tracking and the new high- granularity calorimeter. The inclusion of tracking in particular offers the possibility of developing an algorithm to identify jets originating from bottom quarks (b-tagging) for use in the Level-1 trigger for the first time at CMS. However, only the outer tracker contributes information to the Level-1 trigger, while the sensitive pixel detector which is used for offline b-tagging algorithm does not contribute. In order to make effective use of the new information CMS will implement specialized versions of the particle flow (PF) and Pileup Per Particle Identification (PUPPI) algorithms. The particles reconstructed using the PUPPI algorithm will then be used to construct higher level objects, such as jets. The results presented here show the performance of an algorithm using a neural network (NN) that is feasible to implement on current trigger hardware in the Correlator Layer-2 system where jets are built. This implementation is also capable of operating within the budgeted latency requirements of the Level-1 trigger environment. The algorithm runs on the PUPPI particles in each jet, and is capable of discriminating between jets originating from bottom quarks and jets originating from light quarks or gluons. The expected performance and physics implication of this algorithm is shown using Monte Carlo samples with 200 pileup interactions, simulating the conditions of the HL-LHC.