Study of b-quark production asymmetry with quantum machine learning techniques at the LHCb experiment

The precise determination of the b-quark pair-production asymmetry is important not only as a test of the Standard Model of Particle Physics but also to investigate possible sources of New Physics. At the LHCb experiment, that takes data at the LHC proton-proton collider at CERN, where b-quarks are copiously produced, the Forward-Central asymmetry in the b-quark pair-production has been measured at the center of mass energy of 7 TeV. The result is consistent with the Standard Model expectation within the experimental error. However, additional measurements of such an observable need to be performed at higher energies and with an increased number of events, in order to reduce the uncertainties and to put constraints on New Physics models. The final uncertainty on asymmetry is strongly related to the ability of identifying the flavor (u,d,s vs c vs b) and the charge (b vs anti-b) of the quark producing the jet in the detector (jet tagging). At the LHCb experiment, jet tagging exploits the excellent capability of the detector to precise reconstruct vertices displaced with respect to the interaction point, which allow to distinguish the long-lived heavy quarks from light quarks jets. A standard technique, called Muon Tagging, performs charge tagging by exploiting the charge correlation between the b-quark and the muon produced in the semi-leptonic decay. However, the small branching ratio of this decay strongly limits the efficiency of the algorithm. For this reason other inclusive algorithms, based on Machine Learning techniques, are being developed, which aim to exploit the full jet particles substructure. In this thesis, I present new approaches to the charge tagging of b-jets, based on Quantum Machine Learning techniques: as a general paradigm, data are embedded in a quantum circuit through a quantum feature map; then the initial state gets processed by a variational quantum circuit with trainable parametrized gates; finally, measurements of observables on the final state are mapped to a binary classification label (b-jet or anti-b-jet). The models are trained on official LHCb simulated data at the center of mass energy of 13 TeV and the tagging performance is compared with the Muon Tagging algorithm and a classical Deep Neural Network model. Finally, the precision on the Forward-Central asymmetry is evaluated applying the different tagging algorithms on a sample of simulated data corresponding to the integrated luminosity of the Run2.