Use of Artificial Neural Networks for Improvement of CMS Hadron Calorimeter Resolution

The Compact Muon Solenoid (CMS) experiment features an electromagnetic calorimeter (ECAL) composed of lead tungstate crystals and a sampling hadronic calorimeter (HCAL) made of brass and scintillator, along with other detectors. For hadrons, the response of the electromagnetic and hadronic calorimeters is inherently different. Because sampling calorimeters measure a fraction of the energy spread over several measuring towers, the energy resolution as well as the linearity are not easily preserved, especially at low energies. Several sophisticated algorithms have been developed to optimize the resolution of the CMS calorimeter system for single particles. One such algorithm, based on the artificial neural network application to the combined electromagnetic and hadronic calorimeter system, was developed and applied to test beam data using particles in the momentum range of 2-300 GeV/c. The method improves the energy measurement and linearity, especially at low energies below 10 GeV/c.