Search for new physics with one Higgs boson in the $b\bar{b}\gamma \gamma$ channel with the ATLAS detector

The Standard Model (SM) is the current theory that describes the elementary particles and their interactions. The Higgs boson discovery in 2012 at the Large Hadron Collider (LHC) at CERN marked a remarkable success of its predictive power even if the SM has shortcomings. Since then, the Higgs boson and the Higgs mechanism have been thoroughly studied in the hope to find a hint for new physics. These researches are made possible by high quality experimental infrastructures like the LHC and the ATLAS detector. Its current Inner Detector will be replaced by the brand new Inner Tracker to maintain a high level of tracking and object reconstruction performance in the harsher environment of High-Luminosity LHC (HL-LHC). This thesis presents the adaptation of a neural network based flavour tagging algorithm in the HL-LHC configuration. The track selection used in the training and the resampling method have notably been optimised. A research for new scalar particles in the Higgs sector $X$ and $S$ in the $X\to SH\to b\bar{b}\gamma\gamma$ channel is also presented. The analysis is performed using 140~fb$^{-1}$ of ATLAS Run~2 data at $\sqrt{s}=13$~TeV. Parameterised Neural Networks (PNNs) are used to probe a large range of masses $m_X$ and $m_S$. Results show a local (global) excess of 3.55$\sigma$ (2.0) with respect to the background only hypothesis. 95\% CL upper limits between 0.09 and 39~fb are set on the signal production cross section $X\to SH$ in the $b\bar{b}\gamma\gamma$ final state.