Exploring New Frontiers in High Energy Physics: Boosted Resonances Decaying To Quarks, Foundation Models, and Heterogeneous Computing at the CMS Experiment

Abstract

In this thesis, we introduce machine learning (ML) tools to optimize data taking and analysis at data-intensive scientific experiments, focusing on the CMS experiment at the Large Hadron Collider (LHC). A path to a foundation model for LHC physics is described, where self-supervised learning is enabled through the re-simulation of decaying partons. The first experiments with remote operation of GPUs in LHC experiments are presented. These tools will help equip experiments at the High-Luminosity LHC (HL-LHC) to perform precision measurements and searches for new physics, for example, low mass resonances decaying to quarks. In this context, a search for narrow resonances decaying into quarkantiquark pairs produced with high transverse momentum is presented. The analysis is based on data collected in Run 2 with the CMS detector at the LHC in proton-proton collisions at √ 𝑠 = 13 TeV. Resonance candidates are reconstructed as large-radius jets and identified using a state-of-the-art jet tagging algorithm. This analysis presents the most sensitive limits for new spin-1 bosons coupling universally to quarks and spin-0 bosons coupling preferentially to heavier quarks. The invariant jet mass spectrum is probed for a potential narrow peaking signal over a smoothly falling background. Upper limits at 95% confidence level are set on the coupling of narrow resonances to quarks as a function of the resonance mass. For masses between 50 and 300 GeV, these are the most sensitive limits to date on all possible mediators. Using conventions on s-channel dark matter mediators, limits are set on dark photons and dark matter in the context of the relic density.