Study of the production of a W boson in association with two b-jets using machine learning with the ATLAS experiment at the LHC

This master thesis is focused on the study of the production of a W boson in association with jets generated from b-quarks (b-jets) in proton-proton collisions data using the ATLAS detector at the Large Hadron Collider. The final state is constituted by a W boson that decays leptonically into a muon and a neutrino in association with two b-jets. Therefore muons, jets, b-jets, and missing transverse energy are the primary physics objects used in this analysis. The measurement of differential cross-section in this final state has never been performed at the Large Hadron Collider so far. The main challenge is constituted by the huge background containing events originating from top-quark pair production.

This thesis presents a strategy optimized for the rejection of the top background using a powerful machine learning approach based on the development of a Neural Network. This study marks the first step for the first differential cross-section measurement of this process in the near future. The study is based on simulated Monte Carlo samples of the signal and the backgrounds officially produced by the ATLAS Collaboration. The simulated samples reproduce the experimental conditions of the full Run-2 dataset, collected by the experiment from 2015 until 2018 at the centre-of-mass energy of $\sqrt{s} = 13 TeV$. This dataset represents the largest and best understood dataset available so far at the ATLAS experiment and corresponds to an integrated luminosity of $140fb^{−1}$. The simulated samples are processed using the full detector simulation and analysed with the same reconstruction tools used for real collision data, therefore they serve as a faithful emulation of real collision data.