Fellow, affiliation:
1.8.23-31.7.25: Experimental Particle Physics
Measurements of the CKM matrix elements |Vts| and |Vtd| using an improved s-tagging algorithm with the ATLAS detector at the Large Hadron Collider.
The direct measurement of the parameters Vts and Vtd, two of the elements of the Cabibbo–Kobayashi–Maskawa matrix, is one of the most promising ways to search for signatures of new physics beyond the Standard Model.
The top quark is the heaviest fundamental particle and predominantly decays to a bottom quark and a W boson. However, a small fraction of them decay to strange or down quarks, as well. These highly energetic quarks form narrow cones of hadrons and other particles produced through hadronization, known as jets. The direct measurement of Vts (or Vtd) is obtained from the branching fraction of the top quark decaying to a strange (or down) quark, which requires distinguishing between the jets originating from strange and down quarks. The identification of jets originating from strange quarks is known as s-tagging. Previous efforts of building an s-tagging algorithm have been carried out using different neural-network architectures. While the recurrent neural network based models provide all the flexibility required for accommodating the jet-level and constituent particle-level parameters for a variable number of particles in jets, such models are not inherently permutation invariant, introducing an artificial dependence on the ordering of the constituent particles. Recently, two new approaches, i.e. graph neural networks and deep sets with an attention mechanism, have been introduced to tackle this problem and have been found promising in the field of b-tagging.
My goal is to investigate all these different approaches and build a robust general-purpose s-tagger for the ATLAS experiment and demonstrate its capability by directly measuring Vts and Vtd.