Particle Identification at STCF DTOF Detector Based on Classical/Quantum Convolutional Neural Network

Excellent particle identification (PID) capability is one of the key requirements for detector performance to support the physics analysis of highenergy physics experiments. The rapid advancement of computational power has enabled deep learning to provide robust data-driven solutions for PID tasks. At the same time, quantum computing is introducing a new computational paradigm with potential for quantum-enhanced learning models. This study focuses on π/K identification at the DTOF detector of the Super Tau-Charm Facility, proposing a convolutional neural network-based PID algorithm to process Cherenkov image representations. The trained model achieves over 99% signal efficiency with a 2% misidentification rate for p < 2 GeV/c on the π/K sample, demonstrating its effectiveness for PID. A quantum convolutional layer is designed and compared with its classical equivalent at the same parameter level to evaluate the feasibility of Hybrid Quantum-Classical Convolutional Neural Networks. The results show that both architectures exhibit comparable performance.