This thesis presents a dual contribution to the study of heavy-ion collisions at the LHC, combining physics analyses of charm meson production in fixed-target proton-neon collisions with the development of machine learning methods for charge particle in the second phase of the LHCb detector upgrade program.
The first part reports measurements of D+ and Ds+ meson production and charge asymmetry in p-Ne collisions at sqrt(s_{NN}) = 68.5 GeV.
Cross-sections and asymmetries are obtained as a function of transverse momentum and rapidity, and results are compared with previous measurements to explore hadronization universality.
Although the integrated cross-section production ratios support this hypothesis, the differential measurement of the D+-D– asymmetry hints toward a hadronization process dependent on the distance to the beam rapidity, questioning the charm fragmentation functions universality.
The second part focuses on the use of graph neural networks (GNNs) for segment classification in the future LHCb Upstream Pixel detector.
An optimization of the graph construction from simulated data is performed, and a model specially implemented is trained and evaluated on the produced graph dataset.
The obtained segment classification exhibits a clear increase of the purity while keeping a high efficiency.
LQGP Quark-Gluon Plasma Laboratory
DPhN LQGP