Highlighting the bottom: light jet mistag efficiency calibration with ML-based b-tagging algorithms and b-quark forward-backward asymmetry measurement with the ATLAS Experiment

The Standard Model (SM) of particle physics has proven to be remarkably successful, with predictions tested to per-mille precision. Nonetheless, several open questions remain, such as the dominance of matter over antimatter, the existence of dark matter, non-zero neutrino masses, the hierarchy problem, and the strong CP problem. Electroweak precision measurements are powerful tools to probe the SM and possible beyond-the-SM (BSM) effects. A persistent ~3σ tension exists between the two most precise determinations of sin^2 θW,eff, from the LEP measurement of the forward–backward asymmetry in b-quark production, Afb(b), and from the SLD measurement of A_l. Clarifying this discrepancy requires new experimental input. This thesis presents preliminary studies, based on simulated events and on the LHC Run-2 and partial Run-3 dataset collected by ATLAS, aiming at a new experimental determination of Afb(b). The analysis explores methods for reconstructing the b-quark charge and building the asymmetry observable, with unfolding techniques used to evaluate statistical and systematic uncertainties. An essential part of the work concerns jet flavour tagging, which is essential both for this measurement and for many other ATLAS analyses, including Higgs and top-quark physics. Recent improvements in b-jet tagging are highlighted, in particular the GN2 algorithm based on transformer architectures. Compared to the previous DL1d algorithm, GN2 achieves more than a factor of two improvement in rejecting light- and c-jets. The calibration of the b-tagging algorithms for Run-2 and Run-3 analyses, and in particular the strategies adopted to measure light-jet mistagging rates, are discussed. A further look into the next steps of the algorithm development is also shown. Looking ahead, these developments set the stage for the HL-LHC era, where the increased luminosity and the upgraded ATLAS Inner Tracker (ITk) will provide unprecedented opportunities for precision measurements. The advances in flavour tagging and the methodology explored in this work will be key to fully exploiting the HL-LHC dataset, further refining electroweak precision tests and enhancing ATLAS sensitivity to new physics. Together, these studies demonstrate the potential of ATLAS to refine SM precision physics and to strengthen the tools needed for searches for new physics.