Analysis of a deep neural network for missing transverse momentum reconstruction in ATLAS

• Main Author: Leigh, Matthew
• Other Authors: Yacoob, Sahal
• Format: Dissertation
• Language: English
• Published: Faculty of Science 2020
• Subjects: Physics; Particle Physics
• Online Access: http://hdl.handle.net/11427/32401

The ATLAS detector is a multipurpose particle detector built to record almost all possible decay products of the high energy proton-proton collisions provided by the Large Hadron Collider. The presence and combined kinematics of unobserved particles can be inferred by the observed momentum imbalance in the transverse plane. In this work, a deep neural network was trained using supervised learning to measure this imbalance. The performance of this network was evaluated in MC simulation and in 43 fb⁻¹ of data recorded at ATLAS. The network offered superior resolution and significantly better pileup resistance than all other pre-existing algorithms in every tested topology. The network also provided the best discriminator between events that did and did not contain neutrinos. The potential gain insensitivity to new physics was demonstrated by using this network in a search for the electroweak production of supersymmetric particles. The expected sensitivity to observe the production of said particles was increased by up to 26%.