$\pi^{-}$ energy reconstruction in HGCAL Beam Test prototype detector using Graph Neural Networks

The current CMS endcap calorimeter will be replaced by a high granularity sampling calorimeter (HGCAL) for the high luminosity operation of the LHC. The HGCAL is a sampling calorimeter based on silicon sensors and scintillator tiles directly readout by SiPMs for the active media with very fine transverse and longitudinal granularity. The absorbers are a combination of lead and Cu/CuW in electromagnetic section, and steel in hadronic section. As part of the development of the detector and its readout electronic components, a section of a silicon-based HGCAL prototype detector along with a section of the CALICE AHCAL prototype was exposed to muons, electrons and charged pions in beam test experiments at the H2 beamline at the CERN SPS in October 2018. The energy deposition of hadron showers results from particles of different nature with a multiplicity, energy and spatial distributions which is subject to large fluctuations owing to the admixture of complex interaction processes. The reconstruction of the energy of such showers can be furthermore complicated by potential leakage if the detector does not provide sufficient coverage. A novel algorithm for the energy reconstruction of charged pions collected in the beam test experiment is presented. The algorithm is based on graph neural networks (GNNs), makes use of a dynamic reduction network (DRN) architecture, and improves significantly the energy resolution with respect to classic approaches to the problem.