Nucleons, protons and neutrons, are the building blocks of all nuclei, and hence of most of the visible matter in the Universe. Understanding their fundamental structure and dynamics in terms of their partonic constituents – quarks and gluons – is currently one of the main challenges in hadronic physics. Such an understanding is encapsulated in the Standard Model (SM), in which Quantum Chromodynamics (QCD) is the field theory that describes the strong interaction of quarks and gluons. Because energy grows with separation between colour charges (a fundamental characteristic of quarks and gluons), they cannot exist in isolation, but only in neutral colour combinations called hadrons, among which are nucleons.
Nucleons are probed in scattering experiments whereby they are bombarded with high energy beams made of other particles (leptons or protons). The analysis of the remnants of these collisions allow scientists to derive information on the distribution of quarks and gluons in protons. This information is encapsulated in Parton Distribution Functions (PDFs) and in Fragmentation Functions (FFs). They encode the probability that a given quark or gluon carrying a given fraction of the parent nucleon’s momentum is struck from the incoming or outgoing nucleon, respectively.
This Action aimed at achieving ultimate precision of PDFs and FFs, possibly to match the requirements for discovery at current colliders. One of such colliders, the Large Hadron Collider, is the most powerful particle accelerator ever built by mankind. The determination of PDFs and FFs cannot be reliably achieved from first principles. Instead, they are modelled by means of some parametrisation, which is then optimised by comparing the PDF-dependent prediction of a variety of hadronic cross sections to their actual measurements, a procedure that is called (global) QCD analysis (or fit). In this sense, PDF/FF determinations can be labelled generally as a nonlinear regression problem, whereby one has to learn a set of functions from data. This goal was achieved by developing and utilising innovative artificial intelligence and machine learning techniques, an essential part of the research plan. These techniques allow one to achieve a statistically sound and faithful representation of PDF/FF uncertainties, a feature which is crucial for a characterisation of the SM and for discoveries beyond it.
