$b$-Tagging and Evidence for the Standard Model $H \to b\bar{b}$ Decay with the ATLAS Experiment

The observation of a new particle consistent with the Standard Model Higgs boson by the ATLAS and CMS experiments in 2012 heralded a new era in the understanding of the Standard Model. To date, the direct couplings of the Higgs boson to the bosonic and leptonic sectors of the Standard Model have been observed with a significance of at least 5 standard deviations. With a mass of 125~GeV, the largest decay mode of the Higgs boson is predicted to be to a pair of bottom-quarks, $H\to b\bar{b}$, accounting for approximately 58\% of all decays. A search for the Standard Model $H\to b\bar{b}$ decay has been conducted in the $VH(\to b\bar{b})$ production channel, with $V$ corresponding to either a $W$ or $Z$ boson, using $36.1~\text{fb}^{-1}$ of Run-2 $pp$ collisions, recorded by the ATLAS detector at the Large Hadron Collider. Three channels are considered to target each of the leptonic decays of the vector boson, $Z\to\nu\nu$, $W\to\ell\nu$ and $Z\to\ell\ell$, known as the 0-lepton, 1-lepton and 2-lepton channels, respectively. The identification of the $H\to b\bar{b}$ candidate is a vital aspect of the analysis, with techniques known as $b$-tagging algorithms used to identify the $b$-quark content of jets. Understanding the performance of $b$-tagging algorithms in collision data is necessary to correctly model the performance in simulated processes. A measurement of the $b$-jet tagging efficiency in data has been conducted using a likelihood fitting procedure in \ttbar\ events. Multivariate analysis techniques have been implemented to reduce the dominant systematic uncertainties in the measurement of the $b$-jet tagging efficiency, which improved the sensitivity of the $VH(\to b\bar{b})$ analysis. To enhance the $VH(\to b\bar{b})$ analysis, studies have been carried out to improve the understanding of the modelling of background processes in the 1-lepton channel, with the development and implementation of a dedicated control region for events where a $W$ boson is produced in association with a $b$- or $c$-quark. A boosted decision tree has been trained to improve the signal sensitivity of the $VH(\to b\bar{b})$ analysis in the 1-lepton channel, and numerous cross-checks have been carried out to improve the performance. From the analysis of the Run-2 dataset, an observed signal significance of $3.5$ standard deviations for the $VH(\to b\bar{b})$ process over the background-only model, compared to an expectation of 3.0 standard deviations, is measured. This corresponds to a ratio of the signal yield to the Standard Model expectation of $1.20 ^{+0.24}_{-0.23} \mathrm{(stat.)} ^{+0.34}_{-0.28} \mathrm{(syst.)}$. The combination of this result with the Run-1 analysis yields an observed signal significance of $3.6$ standard deviations, compared to an expectation of 4.0 standard deviations, with the ratio of the measured signal yield to the Standard Model expectation is equal to $0.90 \pm 0.18 \mathrm{(stat.)} ^{+0.21}_{-0.19} \mathrm{(syst.)}$. This result is the first single-experiment evidence for the $H\to b\bar{b}$ decay and the $VH$ production mode.