Simulation of angular resolution of a new electromagnetic sampling calorimeter

TL;DR

This paper presents a simulation study of an electromagnetic sampling calorimeter's angular resolution for photons between 100 MeV and 2 GeV, utilizing Geant4 and machine learning for reconstruction.

Contribution

The study introduces a novel simulation approach combining Geant4 and XGBoost to evaluate and optimize the angular resolution of a new EM calorimeter design.

Findings

Angular resolution follows a specific energy-dependent formula.

Performance depends on detector configuration and machine learning accuracy.

Results are consistent across various incident angles.

Abstract

We report on the simulation results for the angular resolution of an electromagnetic (EM) sampling calorimeter with photons in the range of 100~MeV to 2~GeV. The simulation model of the EM calorimeter consists of alternating layers of a 1-mm-thick lead plate and a 5-mm-thick plastic scintillator plate. The scintillator plates are alternately segmented into horizontal and vertical strips. In this study, we obtain energy deposits in individual strips using Geant4 simulations and reconstruct the incident photon angles using XGBoost with gradient-boosted decision trees. The performance of the angle reconstruction depends on the detector configuration and the accuracy of machine learning. The angular resolution is well described by the expression $0.24^{\circ} \oplus 1.25^{\circ}/\sqrt{E_{\gamma}}$, where $E_{\gamma}$ is the incident photon energy in GeV, for strips of 15 mm and 32 layers. This energy dependence is consistent for different incident angles in the range of 10$^{\circ}$ to 40$^{\circ}$.