Enhancing the Search for Long-Lived Particles in the Muon System at CMS with Graph Neural Networks

Since the inception of the CERN Large Hadron Collider (LHC), the standard model (SM) has been substantially validated, yet it still leaves several key questions, such as the origins of neutrino mass, the matter-antimatter imbalance, and the nature of dark matter, unanswered. Among the various extensions to the SM, theories like supersymmetry, hidden valley, and little Higgs models suggest the existence of new long-lived particles (LLPs), potentially producible at the LHC. These LLPs are hypothesized to generate detectable decay cascades within the tracker, calorimeters, and muon subdetectors of the compact muon solenoid. A significant component of our research involves the study of hit clusters recorded in the muon system, believed to be a result of these decay showers. By analyzing these clusters, we aim to enhance the detection capabilities of the muon system, particularly for LLPs. The central objective of this research is to develop a practical and efficient graph neural network (GNN)--based classifier for LLP detection. This classifier is designed to process the spatial relationships within the recorded hit clusters. Our approach involves representing the recorded hits as nodes in a graph convolutional neural network, with their spatial relationships forming the edges. This network architecture is tailored to manage the intricacies of muon system data, enabling improved classification of LLP events. The network itself is trained on centrally produced CMS Monte Carlo samples. Preliminary results presented in this poster demonstrate the initial steps toward achieving a reliable GNN-based LLP classifier.