Pushing into the Unknown: A Search for light long-lived particles with the ATLAS detector

New long-lived particles (LLPs) are predicted by various extensions of the Standard Model that address unresolved questions such as dark matter, neutrino masses, and the matter-antimatter asymmetry in the universe. These LLPs can be investigated at the Large Hadron Collider (LHC). LLPs exhibit distinctive detector signatures, and require novel search strategies and analysis techniques to aid in their discovery. This thesis investigates a search for light, pseudoscalar LLPs using 140, $fb^{−1}$ of proton-proton collision data collected by the ATLAS detector from 2015 to 2018 at $\sqrt{s} = 13 TeV$. The search focuses on hadronically decaying LLPs with masses between 5 and 55 GeV, with bench- mark models including both Higgs boson decays to LLP pairs ($H \to ss$) and axion-like particle (ALP) models. No significant excess above the expected background is observed. Upper limits are
placed on the branching ratio of Higgs boson to pairs of LLPs, the cross-section for ALPs produced in association with a vector boson, and, for the first time, on the branching ratio of the top quark to an ALP and a u/c quark. The obtained limits for the $H \to ss$ model are the most stringent so far for LLP masses ms < 40, GeV and lifetimes of $10^{−1} –10^{-3}$ m. This thesis also presents the performance of the large-radius tracking algorithm, which was optimized for Run 3 data-taking period. This algorithm is an additional pass of the track reconstruction process with relaxed collision vertex pointing requirements to improve sensitivity to LLPs in the ATLAS inner detector. Additionally, a transformer-based graph neural network algorithm is also studied in this thesis for hard-scatter vertex selection and is found to improve selection efficiency by approximately 5–10% compared to the existing approach for various physics processes.