Searches for scalar resonances using diphotons in association with leptons with the ATLAS detector at the LHC

Abstract

In the first project, we examine recent studies that have identified growing excesses in scalar resonances with diphoton (γγ) final states at the Large Hadron Collider (LHC). These results suggest the presence of scalar particles beyond the Standard Model (BSM), motivated by multi-lepton anomalies observed in Run 2 data from the ATLAS and CMS experiments. Phenomenological studies of the Higgs spectrum side- bands have indicated the presence of a new scalar S with a mass of 152 ± 1 GeV, originating from a heavier scalar H, with a global significance of 5.3σ. ATLAS has launched a new analysis using the easyjet analysis framework with Run 3 data. This analysis searches for scalar resonances produced via γγ channels in association with leptons, inspired by the Two-Higgs Doublet with a Singlet (2HDM+S) and triplet models. Given the complexity of the signal, we carefully account for the dominant background processes, including γγ + multijets, V γγ and t¯tγγ production. The study uses Monte Carlo (MC) samples corresponding to the 2022 ATLAS Run 3 data, with an integrated luminosity of 35.7 fb−1 at √s = 13.6 TeV. The objective is to provide a comprehensive analysis of the significance of this channel, to develop and implement advanced multivariate analysis techniques including deep neural networks, and to con- tribute to the understanding of potential new scalar particles and their implications for physics beyond the SM. In the second project, we build upon the ATLAS and CMS discovery of the Higgs boson (h) decaying into a Z boson and a photon (γ), a rare process within SM that occurs at next-to-leading order (NLO) with a 3.4σ significance. Even with NLO quantum chro- modynamics (QCD) corrections and signal-background interference, SM predictions exceed the observed signal rate by 2.2σ ± 0.7, indicating possible new physics effects or systematic uncertainties that validate further investigation. We investigate this rare process using machine learning (ML) techniques. We use the XGBoost classifier to analyse the production modes of the h → Zγ process, particularly the gluon-gluon fusion (ggF). We use the kernel density estimation (KDE) technique to visualise and iii compare the probability density functions of the classification scores for the generated signal and background events. This provides insights into their separation and overlap within the framework of the Standard Model Effective Field Theory (SMEFT). We aim to constrain the six-dimensional Wilson coefficients and shed light on potential deviations from SM prediction. These methodological innovations demonstrate the effective integration of advanced ML techniques in precision Higgs physics and BSM searches.