Sensor Size Optimization for Particle Physics

Particle physics experiments often involve the precise measurement of energy and position of particles, such as photons, interacting with detectors like electromagnetic calorimeters. Optimizing the sensor size of these detectors is crucial for achieving accurate reconstructions of particle properties while considering cost-effectiveness. In this research, we explore the optimization of sensor size for particle physics experiments, focusing on the reconstruction of photon energy and position within calorimeter cells. We propose using deep learning methods to address this task and evaluate various sensor granularities ranging from 10x10 to 40x40 keeping the same physical size of the calorimeter. Our approach involves training deep learning models, including ResNet18 and Vision Transformer architectures, and comparing their performance against baseline models such as analytical formula-based model and Linear Regression. We employ metrics like Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and weighted versions of these metrics to assess the reconstruction accuracy across different sensor granularities. Additionally, we investigate the impact of loss functions, training strategies, and model parameters on the performance of the deep learning models. Our findings highlight the importance of selecting an optimal sensor size that balances reconstruction accuracy and cost-effectiveness for particle physics experiments.