Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 46: Calorimeter / Detector Systems II

T 46.3: Talk

Tuesday, March 21, 2023, 17:30–17:45, WIL/C133

Shower Separation in Five Dimensions using Machine Learning — •Jack Rolph and Erika Garutti — University of Hamburg, 22761, Luruper Chaussee 149, Hamburg, Germany

To fulfil the requirements for BSM physics searches and Higgs precision measurements at future linear colliders, a final state jet-energy resolution of 3-4 % for jet energies in the range 150-350 GeV is mandatory. Particle Flow Calorimetry (PFC) is a method expected to provide this resolution, which relies upon highly granular sampling calorimeters and sophisticated clustering techniques. In addition, the PFC technique requires excellent separation of single particles. This study presents the performance of three published neural network models to separate the energy deposited by a single charged and single pseudo-neutral hadron estimated from a charged shower, observed with the highly granular CALICE Analogue Hadronic Calorimeter (AHCAL). The neural networks use spatial and temporal event information from the AHCAL and energy information, which is expected to improve sensitivity to shower development and differences in the time development of the hadron shower. Neutral hadron showers with energy 5-120 GeV were separated from charged showers at a variable distance of 0.2-658 mm by the neural networks. It is found that the best-performing network reconstructed events with a Mean90 energy in agreement within 5% of the known shower energy and with an average RMS90 of 1.6 and 1.4 GeV without and with 100 ps timing information from AHCAL, respectively. The improvement due to timing information is attributed to the superior clustering of the hadron shower core.