Thesis defense of Björn Wendland

The rich datasets collected by the ATLAS and CMS experiments in proton-proton collisions of the Large Hadron Collider at a center-of-mass energy of 13 TeV allow high-precision measurements of top-quark properties and enable the experimental exploration of several rare top-quark processes for the first time. The t-channel single-top-quark production in association with a photon is such a process. Measurements of this signal process directly probe the top-quark-photon interaction, which is a cornerstone of electroweak physics. In this work, this process is examined using events selected from the data collected by the ATLAS experiment that contain either an electron or a muon, at least one photon, high missing transverse momentum, and a particle jet initiated by a bottom quark. 
Studying final states with photons, such as that of the signal process, using proton-proton collision data is challenging, as the majority of reconstructed photon candidates are background photons arising from hadronic activity. This requires excellent rejection of such candidates, which is achieved by applying photon-isolation and photon-identification requirements. Precise measurements of their efficiencies are vital for ensuring consistently high performance and for accurate modeling of the contribution of signal photons in simulation. This work introduces improvements to the method used for measuring the photon-identification efficiency at high photon energies and presents the respective measurement results. 
The largest background contributions to the selected data arise from the production of top-quark pairs in association with a photon, from events where an electron is misidentified as a photon, and from the production of a W boson in association with a photon and particle jets. Background contributions arising from events with background leptons and background photons are estimated using data-driven techniques. 
Only a small fraction of the selected events originate from the signal process. Deep neural networks are employed to efficiently discriminate between signal and background contributions. This thesis presents the first observation of the signal process, achieved with a statistical significance of 9. 3 standard deviations using these neural networks. This observation represents a milestone in electroweak physics. Two fiducial cross sections of the signal process are measured with a precision of 110/o. Their values are found to be compatible with Standard Model predictions at next-to-leading order accuracy in quantum chromodynamics at the level of 2.0 and 2.1 standard deviations.