Precision phenomenology of the PDF-BSM interplay

Abstract

The Standard Model (SM) is one of the most successful theories ever conceived, representing a triumph of human intellect and collaboration. The SM provides a very good description of visible matter, with the proton being a central part of this understanding. The structure of the proton can be parametrised in terms of parton distribution functions (PDFs), essential inputs for theoretical predictions at the Large Hadron Collider (LHC) which cannot be computed from first principles and must be obtained from fits to experimental data. Despite the remarkable success of the SM, many fundamental questions remain unanswered, motivating the search for physics beyond the Standard Model (BSM). These searches often rely on precise comparisons between theoretical predictions and experimental measurements. However, a significant challenge arises because PDFs are typically determined under SM assumptions. This approach can introduce potential biases and inconsistencies in BSM searches. Additionally, signals of BSM physics may be absorbed by the PDFs during the fitting procedure, further complicating the interpretation of the results. This PhD thesis explores the interplay between PDFs and BSM physics. We introduce the open-source release of SIMUnet, a deep learning framework designed to perform simultaneous fits of PDFs and BSM parameters, as well as to assess the potential absorption of BSM signals by the PDFs. We study the PDF-BSM interplay in the top sector and in a global dataset, and assess the absorption of BSM physics in widely used models. Additionally, we explore how BSM effects can alter the evolution of PDFs, and demonstrate the potential of interpretable machine learning techniques in symbolic regression for collider physics in general.