Thesis defense of Sonja Zeißner

In this thesis, the development and calibration of an algorithm used to identify jets from strange quarks as well as a measurement constraining the CKM matrix elements $|V_{ts}|$ and $|V_{td}|$ in top quark decays are presented. The thesis considers data from proton-proton collisions at the Large Hadron Collider at a center-of-mass energy of 13 TeV recorded by the ATLAS Experiment during Run 2. First, the maximally achievable separation between jets from strange quarks and jets from down quarks at hadron colliders given different idealized detector designs is studied using recurrent neural networks containing Long Short-Term Memory layers. Afterwards, an algorithm to select jets from strange quarks for the application at the ATLAS Experiment is developed using deep neural networks. Its efficiency for these jets from strange quarks and mis-tag rates for jets of other flavors is determined in semileptonic decays of top-antitop pairs selected from data. The algorithm to identify jets from strange quarks is then applied in events containing decays of top-antitop pairs with an electron and a muon of opposite-sign electric charge in the final state in order to study its potential to constrain the CKM matrix elements $|V_{ts}|$ and $|V_{td}|$ in the two-dimensional plane spanned by them. In this study, limits of $|V_{ts}|^2+|V_{td}|^2<0.06$, $|V_{ts}|<0.21$ and $|V_{td}|<0.24$ are derived at 95% confidence level assuming unitarity of the CKM matrix.