Deep learning and signal processing applied to 4 dimension track reconstruction for the HL-LHC

• Grant number: 23/18484-6
• Support Opportunities: Scholarships in Brazil - Doctorate (Direct)
• Start date: April 01, 2024
• End date: March 31, 2029
• Field of knowledge: Physical Sciences and Mathematics - Physics - Elementary Particle Physics and Fields
• Principal Investigator: Marco Aurelio Lisboa Leite
• Grantee: Rodrigo Estevam de Paula
• Host Institution: Instituto de Física (IF). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated research grant: 20/04867-2 - High energy physics and instrumentation with the LHC-CERN, AP.ESP

Abstract

This work proposes the study of new methods for track reconstruction and track-vertex association in the ATLAS experiment at the LHC-CERN for the LHC high luminosity operation phase (HL-LHC). To operate during this phase, where a larger number of simultaneous collisions will occur, the ATLAS experiment is building a new track reconstruction system to completely replace the present Inner Detector (ID) with higher granularity and spatial coverage, the Inner Tracker (ITk) and a totally new detector with excellent temporal resolution to cover the forward region, the High Granularity Timing Detector (HGTD). The present track reconstruction methods use adaptive filters in which the processing time increases quadratically with the number of points. It is expected that, with the significant increase of the number of proton-proton collisions and with the increased segmentation of the future detectors, such an approach will not be able to satisfy the criteria for track-vertex association required by the HL-LHC research program. To overcome such limitations it is important to devise new reconstruction methods that utilizes the ITk segmentation and the timing information from HGTD together, and capable of processing the data within an acceptable time in this challenging high luminosity scenario. New methods of deep learning and signal processing suitable to be implemented in heterogeneous architectures (CPU+GPU) will be evaluated, seeking a reduction of the processing time . Moreover, these methods will be benchmarked for use in the ATLAS experiment through physics simulations including the new detectors (ITk and HGTD). (AU)