First application of machine learning algorithms to the position reconstruction in Resistive Silicon Detectors

TL;DR

This paper presents the first use of machine learning algorithms for position reconstruction in Resistive Silicon Detectors, achieving sub-micrometer spatial resolution and outperforming traditional analytical methods.

Contribution

It introduces a novel machine learning approach to RSD position reconstruction, demonstrating significant improvements in spatial resolution over standard techniques.

Findings

Machine learning achieves <2 μm resolution in RSDs.

ML method outperforms traditional analytical reconstruction.

Potential for further improvements discussed.

Abstract

RSDs (Resistive AC-Coupled Silicon Detectors) are n-in-p silicon sensors based on the LGAD (Low-Gain Avalanche Diode) technology, featuring a continuous gain layer over the whole sensor area. The truly innovative feature of these sensors is that the signal induced by an ionising particle is seen on several pixels, allowing the use of reconstruction techniques that combine the information from many read-out channels. In this contribution, the first application of a machine learning technique to RSD devices is presented. The spatial resolution of this technique is compared to that obtained with the standard RSD reconstruction methods that use analytical descriptions of the signal sharing mechanism. A Multi-Output regressor algorithm, trained with a combination of simulated and real data, leads to a spatial resolution of less than 2 $\mu m$ for a sensor with a 100 $\mu m$ pixel. The prospects of future improvements are also discussed.