Precision studies for the partonic kinematics calculation through Machine Learning

Resumo

High Energy collider experiments are moving to the highest precision frontier quickly. The predictions of observables are based on thefactorization formula, which helps to connect small to large distances. These predictions can be contrasted with experimental measurementsand the success of this phenomenological approach is based on the correct description of nature. The application of the method to proton-proton colliders brings new challenges due to the proton structure and the detectors efficiency on reconstructing hadrons. Furthermore, sincethe non-perturbative distribution functions takes an important role to describe the experimental distributions, the presence of them makes theinformation of the partons diluted. At Leading Order (LO) in perturbative calculations, the momentum fractions involved in hard scatteringprocesses are known exactly in terms of kinematical variables of initial and final states hadrons. However, at Next-to-Leading Order (NLO)and beyond, a closed analytical formula is not available. Furthermore, from the pure theoretical calculation, the exact definition of themomentum fraction is very challenging. In this work, we report a methodology based on Machine Learning techniques for the extractionof momentum fractions forp+p→π++γusing a Monte Carlo simulation including quantum corrections up to Next-to-Leading Orderin Quantum Chromodynamics and Leading Order in Quantum Electrodynamics. Our findings point towards a methodology to find thefundamental properties of the internal structure of hadrons because the reconstructed momentum fractions deeply relate our perturbativemodels with experimental measurements