ADVANCEMENTS IN PDFS DETERMINATION: INCORPORATION OF QED EFFECTS AND NEW THEORETICAL IMPROVEMENTS IN A MODERN DEEP LEARNING FRAMEWORK

The focus of this Ph.D. thesis is the improvement in the determination of parton distribution functions (PDFs) through the inclusion of new theoretical effects in the theoretical predictions that are compared to experimental data. Such effects are the inclusion of Quantum Electrodynamics (QED) corrections, the inclusion of the so-called missing higher orders uncertainties (MHOUs) and the inclusion of approximate next-to-next-to-next-to-leading order contributions (aN$^3$LO). This work aims to overcome the NNPDF4.0 PDFs determination, that did not include such effects and whose methodology is at the basis of this analysis. We first present the methodology used in the NNPDF4.0 PDFs fit. Then, we explain how the QED effects have been added in the NNPDF4.0 fitting framework, describing the details of the implementation, the resulting PDFs determination and the impact on physical processes of such PDFs. Moreover, we explain how the MHOUs and aN$^3$LO effects have been included in NNPDF4.0, presenting the results of the two effects both separately and combined. We also show how the QED effects have been combined to MHOUs and aN$^3$LO and we will present the results of the PDFs determination that includes all the three effects. In conclusion, we will show an application to Monte Carlo event generator of the implementation of the QED effects and of the development of a more flexible and efficient theory pipeline for the computation of the theoretical predictions that has been interfaced to the NNPDF4.0 code and is at the basis of all the work described in this thesis.