Machine learning for precise hit position reconstruction in Resistive Silicon Detectors

F. Siviero,R. Arcidiacono,N. Cartiglia,M. Costa,M. Ferrero,L. Lanteri,C. Madrid,L. Menzio,R. Mulargia,V. Sola

DOI: https://doi.org/10.1088/1748-0221/19/01/c01028

2024-01-20

Journal of Instrumentation

Abstract:RSDs are LGAD silicon sensors with 100% fill factor, based on the principle of AC-coupled resistive read-out. Signal sharing and internal charge multiplication are the RSD key features to achieve picosecond-level time resolution and micron-level spatial resolution, thus making these sensors promising candidates as 4D-trackers for future experiments. This paper describes the use of a neural network to reconstruct the hit position of ionizing particles, an approach that can boost the performance of the RSD with respect to analytical models. The neural network has been trained in the laboratory and then validated on test beam data. The device-under-test in this work is a 450 μm-pitch matrix from the FBK RSD2 production, which achieved a resolution of about 65 μm at the DESY Test Beam Facility, a 50% improvement compared to a simple analytical reconstruction method, and a factor two better than the resolution of a standard pixel sensor of equal pitch size with binary read-out. The test beam result is compatible with the laboratory ones obtained during the neural network training, confirming the ability of the machine learning model to provide accurate predictions even in environments very different from the training one. Prospects for future improvements are also discussed.

instruments & instrumentation

What problem does this paper attempt to address?

This paper aims to solve the problem of using machine - learning techniques to improve the accuracy of reconstructing the hit positions of ionizing particles in Resistive Silicon Detectors (RSDs). Specifically, RSDs are silicon sensors based on the Low - Gain Avalanche Diode (LGAD) technology, with a filling factor of 100% and operating through the principle of AC - coupled resistive readout. These characteristics make RSDs very promising as 4D trackers in high - energy physics experiments, being able to provide picosecond - level time resolution and micron - level spatial resolution.

However, the accuracy of traditional analytical models gradually decreases as the number of readout pads increases when dealing with multi - electrode readout signals. To this end, researchers introduced machine - learning algorithms, especially neural networks, to improve the accuracy of position reconstruction. By training neural networks with laboratory data (such as data set up using the Transient Current Technique (TCT)) and validating the effectiveness of the model on test beam data, the results show that the method using neural networks improves the position reconstruction accuracy by about 50% compared to simple analytical models and is superior to standard pixel sensors of the same size.

In addition, this study also demonstrated the generalization ability of the machine - learning model in different environments, that is, the model can maintain good prediction performance in test environments that are very different from the training data set. This result is of great significance for the future application of RSDs in high - energy physics experiments.