Speaker
Description
We use Bayesian inference with deep-learning surrogate models to quantitatively extract the in-medium heavy quark (HQ) potential from bottomonium nuclear modification factors ($R_{AA}$) measured in Pb--Pb collisions at $\sqrt{s_{NN}}=5.02$ and $2.76$ TeV and Au--Au collisions at $\sqrt{s_{NN}}=200$ GeV. The real and imaginary parts of the potential are parametrized and embedded in a time-dependent Schr\"odinger equation for the evolution of $b\bar b$ dipoles in hydrodynamic QCD media. To accelerate the inverse problem, we train collision-system-specific CNN surrogate models that map the potential parameters to bottomonium $R_{AA}$ observables. The training data are generated from Schr\"odinger-model calculations and augmented using Principal Component Analysis (PCA) and Gaussian Process Regression (GPR). We then perform a joint Stochastic Gradient Langevin Dynamics (SGLD) posterior sampling of the potential parameters using all available RHIC and LHC bottomonium data. Within the present parametrization and hydrodynamic background, the real part of the in-medium potential remains close to the vacuum Cornell form, corresponding to a relatively weak screened Debye mass across RHIC to LHC energies. By contrast, the imaginary part is more strongly constrained by the data and provides the dominant contribution to bottomonium suppression from RHIC to LHC energies.
