Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 63: Detectors 6 (calorimeters)

T 63.5: Vortrag

Mittwoch, 6. März 2024, 17:00–17:15, Geb. 30.23: 2/1

Fast Hadron Shower Simulation using Generative Adversarial Networks with the CALICE AHCAL Prototype — •André Wilhahn, Julian Utehs, and Stan Lai for the CALICE-D collaboration — II. Physikalisches Institut, Georg-August-Universität Göttingen, Deutschland

Extensive simulations of particle showers are crucial for high energy physics experiments, since they allow for a sensible interpretation of recorded calorimeter data. As many calorimeters are designed with increasing granularity, while having to cope with higher energy deposits and higher luminosity conditions, the accurate simulation of particle showers in a computationally efficient manner is of utmost importance. This talk describes preliminary investigations into a data-driven fast calorimeter simulation, based on machine learning techniques, that is meant to describe particle showers accurately.

We start by investigating pion showers in the CALICE AHCAL (Analog Hadron Calorimeter) prototype, which is a highly granular hadronic calorimeter comprising a total of 38 active layers embedded in a stainless-steel absorber structure. Each active layer contains a grid of 24×24 scintillator tiles that are read out individually via silicon photomultipliers. Based on energy distributions, Generative Adversarial Networks have been trained on testbeam data, aiming at creating a neural network that is able to generate and recreate energy distributions from random input noise, while also preserving correlation factors between individual detector layers.

Keywords: Calorimeter; AHCAL; Simulation; Neural Networks