Implementação embarcada em FPGA de métodos visando a reconstrução online de energia no calorímetro hadrônico do experimento ATLAS

Abstract

Research laboratories in Experimental High Energy Physics have been collaborating on significant advances in science and technology. Within these laboratories, modern apparatuses are constructed for the detection and study of particles, where interactions occur between subatomic particles with high energies. The primary objective is to investigate various characteristics of elementary particles. The Large Hadron Collider (LHC), currently the world’s largest particle accelerator, is undergoing a gradual upgrade to increase the luminosity of the collisions, aiming to increase the probability of rarer events occurring. This directly impacts the instrumentation systems of detectors, particularly the Hadronic Calorimeter of the ATLAS Experiment, where the increased occurrence of adjacent collisions leads to the phenomenon of pile-up. Due to the current algorithm used in signal reconstruction in this calorimeter not considering the pile-up in its formulation, various alternatives for energy estimation are being proposed. These include iterative methods of signal deconvolution based on Sparse Data Representation and non-linear methods based on Artificial Neural Networks. These methodologies have a high computational cost, and the primary challenge is currently the development of algorithms capable of operating online. In this context, this work describes the implementation of algorithms on dedicated processors in FPGA with a multicore architecture, aiming to perform online energy reconstruction in High Energy Calorimeters, focusing on the Hadronic Calorimeter of the ATLAS Experiment in pile-up scenarios. The results demonstrate the possibility of operating such processors at the ATLAS first-level trigger, respecting the LHC collision rate of 40 MHz, and with logical resource consumption within the limits for implementation in modern FPGAs.