Quantification of the Quark-Gluon Plasma with statistical learning
Date:
Invited talk to accept the 2024 American Physical Society Dissertation Award in Nuclear Physics
Abstract:
The results of the first global Bayesian analysis of relativistic Pb + Pb collisions at 2.76 TeV using a multistage model with an IP-Glasma initial state are presented. The hybrid model, also including a viscous fluid dynamical evolution and a hadronic transport final state, represents state-of-the-art physics modeling of the evolution of quark gluon plasma. The observables from the soft sector hadronic final state are systematically compared to data with the first use of novel transfer learning, design space sampling, and non-uniform prior distributions in heavy ion collisions. Theoretical consistency of the model is demonstrated with Bayes Factor closure tests, also for the first time in heavy ion collisions, as well as thorough model validation for Bayesian inference. The resulting analysis reveals new sensitivity of experimental observables to the properties of quark gluon plasma. Precise predictions and accurate post dictions using the most realistic available dynamics of pre-equilibrium strongly-interacting matter are shown and demonstrate the resounding success of Bayesian methods in heavy ion physics.