Heavy-Flavour Tagging and Measurement of the $t\bar{t}H$ Production Cross-Section in the $H \rightarrow b \bar{b}$ Decay Channel at the ATLAS Experiment

With the final missing piece of the Standard Model of particle physics (SM), the Higgs boson, discovered in 2012 by the ATLAS and CMS experiments at the Large Hadron Collider (LHC) at CERN, both experiments have shifted their focus towards a precise determination of the properties of the new boson. The large datasets collected by both experiments during the second run of the LHC, Run~2, enable such precision measurements of the Higgs boson properties. One of the properties under study is the Yukawa coupling of the top quark, the heaviest of all particles in the SM, to the Higgs boson. The top-Higgs Yukawa coupling, the largest in the SM, is varied in many postulated theoretical extensions of the SM and is sensitive to the effects and the possible presence of physics Beyond the SM (BSM). The coupling itself can be measured in an indirect way, e.g using $H \rightarrow \gamma \gamma$ measurements in which the coupling is involved via virtual loops, and in a direct way, which can be measured in the rare process where a Higgs boson is produced in association with a pair of top quarks, $t\bar{t}H$. Exploiting the dominant decay mode of the Higgs boson into a pair of $b$-quarks, $t\bar{t}H \left(H \rightarrow b\bar{b}\right)$, this thesis investigates the direct measurement of the top-Higgs Yukawa coupling in a challenging yet promising final state with at least four $b$-quark initiated jets. The $t\bar{t}H \left(H \rightarrow b\bar{b}\right)$ legacy analysis presented in this thesis is part of the re-analysis of the complete Run~2 dataset of $140\,\text{fb}^{-1}$ at $\sqrt{s} = 13\,\text{TeV}$ recorded by the ATLAS experiment between 2015 and 2018. The focus lies on enhancing and integrating the latest physics object reconstruction algorithms, as well as employing an improved modelling of the main background process, $t\bar{t} + b\bar{b}$. Alongside state-of-the-art classification and reconstruction neural networks, a new estimate for the fake-lepton contribution in the single-lepton channel is derived. Further, the derivation of data-driven correction factors for the mis-modelled $H_\text{T}$, observed in various $t\bar{t} + \text{jets}$ processes, are presented. Finally, studies of the performance of profile likelihood fits as employed in the full analysis are presented with blinded data, yielding an expected significance of the $t\bar{t}H$ process at $5.5\,\sigma$. Effective separation of the signal from the expected background requires dedicated and high-performing $b$-jet identification algorithms, the so-called $b$-taggers. In this thesis, the development and integration of a novel track-based $b$-tagger, DIPS, is presented. In addition, the DIPS tagger was added as a component to the ATLAS high-level $b$-tagger DL1r, replacing the previous track-based $b$-tagger, RNNIP. The resulting version, DL1d, is the new recommended $b$-tagger in ATLAS, thanks to its significantly improved performance compared to its predecessor. Furthermore, an extension of the DIPS $b$-tagger is introduced in this thesis, which is referred to as DIPS Tau which incorporates a new jet-class: $\tau$-jets.