Título: Redes Neurais Artificiais como algoritmo de reconstrução da informação depositada no Calorímetro Eletromagnético do Experimento ATLAS reduzindo os efeitos de Crosstalk
Autores: Marton S. dos Santos, Paulo C. M. A. Farias, Eduardo F. de Simas Filho, Bertrand Laforge, Jose M. de Seixas
Resumo: ATLAS is the largest experiment at the LHC accelerator complex in CERN. It is situated at one of the collision points in the accelerator tunnel. The experiment consists of a collection of specialized sub-detectors designed to characterize the particles generated by collisions in the LHC. One of these specialized detectors is the liquid argon (LAr) calorimeter, which contains approximately 187k sensor cells used to characterize electromagnetic showers. The LAr calorimeter has a high cell density, which, combined with the high collision rates and the mechanical and electronic structure of the detector readout, leads to crosstalk (XT) effects between adjacent sensor cells. Crosstalk degrades the accuracy of energy and time reconstruction for incoming particles. To address this challenge, an electromagnetic shower simulator based on the ATLAS LAr calorimeter was developed, along with a machine learning approach to mitigate the XT effects. The results demonstrate that the energy and time-of-flight of particles can be reconstructed very closely to the desired values. The proposed approach reduces the error fluctuations in energy estimation by at least 37 times compared to the standard algorithm. Additionally, it decreases the error fluctuations in time estimation by three orders of magnitude
Palavras-chave: ATLAS, Electromagnetic Calorimeter, Liquid Argon (LAr), Crosstalk, Neural Networks, Simulator, Regression.
Páginas: 8
Código DOI: 10.21528/CBIC2023-062
Artigo em pdf: CBIC_2023_paper062.pdf
Arquivo BibTeX: CBIC_2023_062.bib