The CMS HGCAL pion energy reconstruction using GNNs and handling large data rates

As a part of the Phase-2 upgrade, the CMS collaboration is designing new endcap calorimeters, referred to as the High-Granularity Calorimeter (HGCAL), in an effort to deal with the challenges of unprecedented in-time event pileup and cumulative radiation exposure during the high-luminosity phase of the LHC. The HGCAL features fine longitudinal & transverse segmentation in both electromagnetic and hadronic sections and corresponds to six million independent readout channels. It uses silicon sensors as active materials in regions of high radiation exposure and plastic scintillator tiles equipped with on-tile silicon photomultipliers elsewhere. These sensors will sample the electromagnetic and hadronic particle showers using 47 longitudinal layers with fine lateral granularity, and provide five-dimensional measurements of energy, position, and timing. The process of going from conceptualization to realisation of such a state-of-the-art system involves prototyping, testing, and updating design choices based on feedback from electronics, data acquisition, and physics performance studies.

In this talk, I will give a brief overview of the various activities going on to finalise the frontend and backend data handling. I will also summarise the performance of the HGCAL prototype using single pion beams conducted at beam test experiments at CERN using Graph Neural Network (GNN)-based energy reconstruction, which has shown a significant improvement in the energy resolution of single hadrons compared to traditional rule-based methods.