Improving the time resolution of the MRPC detector using deep-learning algorithms

The multi-gap resistive plate chambers (MRPCs) will be used as the Time-of-Flight (ToF) system in the Solenoidal Large Intensity Device (SoLID). The time resolution required by the experiment for the MRPC system is 20 ps in order to make a 3 $\sigma$ separation of the $\pi/K$ created in the collisions. To achieve this goal, the whole system including the MRPC detector, the front-end electronics and the readout system will be upgraded. Based on the new system, a time reconstruction algorithm using a combined LSTM (ComLSTM) neural network is proposed. The best time resolution achieved with this algorithm in a cosmic ray test is 16.8 ps, which largely improves the timing ability of the MRPC detector and well satisfies the requirement of the SoLID.