Jet Reconstruction with the Seeded Cone algorithm 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) [CMS-TDR-021] includes the introduction of many new capabilities into the Level-1 trigger, including tracking and the new high-granularity calorimeter. The combination of the precise information coming from the muon, calorimeter and tracking system will allow particle candidates to be reconstructed with the Particle Flow algorithm. This reconstruction will be performed in the Correlator subsystem, where the Pile-Up Per Particle Identification (PUPPI) algorithm will also be applied to suppress particles that originate from pile up interactions. From these particles, a number of final state objects, such as jets, will be reconstructed. The Seeded Cone jet algorithm is an approach to reconstructing jets from these particles using the full available granularity, and has been implemented in the Xilinx Ultrascale+ FPGA used in the Correlator system. The algorithm latency is within the one microsecond constraint of the subsystem. The algorithm begins by finding the highest pT particle in the event, then finding all particles within a cone of this seed in (eta, phi) to form the first jet. The constituents of the jet are removed from the processing before the process is repeated in order to find up to 16 jets in the event. The jet axis is computed as the sum over the constituents of the pT and the pT-weighted sum of their eta and phi. Preceding the jet algorithm itself is a step, known as the 'deregionizer', that aggregates the particles as they arrive into the jet processing FPGA board. Several upstream FPGAs transmit particles over optical fibres in a geometric order over 6 bunch crossings (150 ns). The deregionizer aggregates up to 128 input particles, storing them ready for processing by the jet finding algorithm. In this note we show the multiplicity of objects arriving to the deregionizer, indicating the size of any potential truncation effect. We also describe the physics performance of the seeded cone jet finding algorithm, and compare this against the benchmark offline jet algorithm anti-kt [JHEP04(2008)063], applied to the same L1T particles. The Seeded Cone algorithm has been implemented to perform jet reconstruction with two jet radius parameters, producing R=0.4 and R=0.8 jets (SC4, SC8 respectively). The SC4 jets are used for the baseline performance, while the SC8 jets are made available for new trigger development.