Phenomenology of scalar particles assisted by machine learning

TL;DR

This thesis employs machine learning techniques to analyze scalar particles beyond the Standard Model, enhancing detection sensitivity at collider experiments and identifying promising parameter regions for future discoveries.

Contribution

It introduces ML-enhanced collider analysis methods for scalar particles, focusing on specific models and improving signal-background discrimination for future experiments.

Findings

Identification of parameter regions consistent with recent anomalies

Potential for 5σ discovery at future collider luminosities

ML techniques significantly improve signal detection and analysis robustness

Abstract

In this thesis, we explore the phenomenology of scalar particles within Beyond Standard Model frameworks, using Machine Learning (ML) techniques to enhance sensitivity and discovery potential at current and future collider experiments, the Large Hadron Collider (LHC) and the High-Luminosity LHC (HL-LHC). Specifically, we study scalar extensions of the Standard Model such as the Two Higgs Doublet Model Type-III (2HDM-III) and the Froggatt-Nielsen Flavon model. We perform a detailed collider analysis focusing on charged Higgs boson pair production within the 2HDM-III, examining final states involving muons, neutrinos and quark jets. Our studies identify parameter regions consistent with recent experimental anomalies reported by ATLAS collaboration, particularly in charged Higgs decays involving charm-bottom quark transitions, and suggest concrete scenarios for achieving statistically significant signals of 5$\sigma$ at future luminosities. In the context of the Flavon model, we analyse potential signatures of a new scalar called Flavon decaying into a Higgs boson and a pair of bottom quarks, followed by the channels where the Higgs decays into a pair of bottom quarks or a pair of photons. Additionally, we analyse Lepton-Flavour-Violating processes, both of them achieving discovery level significances of up to $5\sigma$ at the HL-LHC. Using multivariate analysis techniques, specifically Boosted Decision Trees, we demonstrate a significant improvement in signal discrimination. Throughout this thesis, ML methodologies have been integral, notably enhancing the signal from background separation and significantly improving the robustness of phenomenological predictions. The methods and analyses presented here contribute to clarifying the flavour structure mysteries of the SM and offer actionable targets for future experimental searches.