Erlangen 2026 – scientific programme

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

T: Fachverband Teilchenphysik

T 28: Data, AI, Computing, Electronics III

T 28.5: Talk

Tuesday, March 17, 2026, 17:15–17:30, KH 00.024

Machine-Learning based Energy Regression of Muon Detector Showers in CMS — •Mascha Hackmann, Ayse Asu Guvenli, Karim El Morabit, and Gregor Kasieczka — University of Hamburg, Hamburg, Germany

Exotic Long-Lived Particles (LLPs), predicted by many extensions of the Standard Model, can travel macroscopic distances before decaying. Decays inside the muon system of the CMS detector produce hadronic showers instead of isolated muon tracks, leading to unconventional signatures beyond the scope of standard muon reconstruction.

This talk presents a machine learning approach to analyze Muon Detector Showers (MDS) originating from LLP decays in the CMS experiment. For the analysis of the MDS, the Cathode Strip Chambers are interpreted as a sampling calorimeter. Low-level hit information is used as input to train a ParticleNet based Dynamic Graph Convolutional Neural Network (DGCNN) to regress the energy of the LLP. These studies provide insight to improve future searches for LLPs that decay in the muon system of the CMS.

Keywords: Long-lived particles; Muon detector showers; Machine learning; Energy regression