Improved algorithms for determination of particle directions with Timepix3

TL;DR

This paper compares regression algorithms for 3D particle trajectory determination using Timepix3 detectors, demonstrating high accuracy and efficiency in both simulated and real-world radiation fields, including LHC data.

Contribution

It introduces and evaluates new regression methods for particle track analysis with Timepix3, optimizing accuracy and computational speed for radiation characterization.

Findings

Best methods achieved ~2° angular error

Fastest method takes 0.02 ps per track

Validated on LHC experimental data

Abstract

Timepix3 pixel detectors have demonstrated great potential for tracking applications. With $256\times 256$ pixels, 55 $\mathrm{\mu}$m pitch and improved resolution in time (1.56 ns) and energy (2 keV at 60 keV), they have become powerful instruments for characterization of unknown radiation fields. A crucial pre-processing step for such analysis is the determination of particle trajectories in 3D space from individual tracks. This study presents a comprehensive comparison of regression methods that tackle this task under the assumption of track linearity. The proposed methods were first evaluated on a simulation and assessed by their accuracy and computational time. Selected methods were then validated with a real-world dataset, which was measured in a well-known radiation field. Finally, the presented methods were applied to experimental data from the Large Hadron Collider. The best-performing methods achieved a mean absolute error of 1.99{\deg} and 3.90{\deg} in incidence angle $\theta$ and azimuth $\varphi$, respectively. The fastest presented method required a mean computational time of 0.02 ps per track. For all experimental applications, we present angular maps and stopping power spectra.