Arianna García Caffaro successfully defends thesis, "Probing the Higgs CP Structure in the H to tau tau Decay and Quantifying QCD Systematics in Machine Learning-Driven Jet Substructure Techniques"

On June 2, Arianna García Caffaro successfully defended the thesis, “Probing the Higgs CP Structure in the H to tau tau Decay and Quantifying QCD Systematics in Machine Learning-Driven Jet Substructure Techniques” (advisor: Paul Tipton).

García Caffaro explained, “During my PhD, I worked on a precision measurement of the Higgs boson, a fundamental particle discovered in 2012. In particular, my research studied how the Higgs boson interacts with tau particles. Within this interaction, we looked for a subtle effect called CP violation. The Standard Model predicts the Higgs interaction with taus to be CP symmetric; any deviation from this prediction could be a sign of new physics.”

García Caffaro continued, “My main contribution was developing and implementing a new Neural Network-based observable that is preliminarily 40% more sensitive than previous measuring techniques. This improved sensitivity will help us better distinguish whether the Higgs interaction is purely CP symmetric or contains a mixture of CP states. If CP violation were observed in this interaction, it might help us explain why the Universe contains so much more matter than antimatter, an open puzzle known as the baryon asymmetry problem.”

García Caffaro will be a postdoctoral researcher at the Enrico Fermi Institute of the University of Chicago, continuing to work at the ATLAS experiment, as well as joining the Muon Collider Collaboration. 

Thesis Abstract: Since the Higgs boson’s discovery in 2012, the collider physics community has centered its efforts on thoroughly studying this particle’s properties. As a step towards that goal, the ATLAS Collaboration has undertaken a study of the CP nature of the Higgs Yukawa coupling to the tau lepton.  An analysis with LHC Run 2 data has measured the CP mixing angle to be 9 ± 16°, excluding the pure CP-odd Higgs hypothesis at 3.4 sigma.  The next iteration of this analysis with partial Run 3 data is underway.  Building on earlier analysis techniques, a new machine learning-based method for measuring the CP mixing angle, which achieves a preliminary ~40% sensitivity improvement, will be discussed.  Furthermore, in view of the rapidly increasing interest in jet substructure measurements, a study on the robustness of Neural Networks (NNs) commonly used for jet tagging will be presented.  As a step towards better understanding the role QCD systematics play in Machine Learning applications, this study proposes new metrics to quantify the NNs’ robustness against non-perturbative uncertainties originating from the hadronization process.

Thesis Committee: Paul Tipton (advisor), Sarah Demers, Ian Moult, Karsten Heeger, and Jahred Adelman (Northern Illinois University)