Learning physics of collider events with machine

Abstract

Collider experiments have confirmed that the Standard Model (SM) of particle physics has been highly successful in predicting the existence and properties of subatomic particles (hereinafter “particles”). The Higgs boson discovered in 2012 was the final piece of the SM whose existence had remained to be confirmed. Despite such success, however, some questions pertaining to the SM remain unanswered. This may imply the existence of physics beyond the SM. To improve the capability of colliders to explore the underlying physics, dedicated strategies are needed for data analysis, especially for the ones involving hadronic processes where the produced particles are rich. In this thesis, we first introduce the concept of persistent homology to characterize the topological structure of jets. These topological invariants measure the multiplicity and connectivity of jet branches at a given scale threshold, and their persistence indicates the evolution of each topological feature as this threshold varies. With the knowledge provided by these measurements, we can reconstruct the topological phylogenetic tree for each jet. Such a technique opens a new angle to look into jet physics. We then introduce a cosmic microwave background(CMB) like observable scheme for use in future e+e− colliders to address the deformation and loss of information in jet clustering and thus improve the precision of measuring hadroncic events. In this scheme, the event-level kinematics is encoded as Fox–Wolfram moments at the leading order and as multi-spectra of spherical harmonics at higher orders. We show that the difficulty measuring hadronic events can be well addressed by synergizing the event-level information with the data analysis using a deep neutral network. At last, the directions for future exploration are discussed.

Date of Award	2023
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Tao LIU (Supervisor)