Mapping properties of the quark gluon plasma in Pb-Pb and Xe-Xe collisions at energies available at the CERN Large Hadron Collider

TL;DR

This paper uses a Bayesian, machine learning-enhanced fluid dynamics model to analyze experimental data from Pb-Pb and Xe-Xe collisions at the LHC, extracting properties of the quark-gluon plasma.

Contribution

It introduces a novel Bayesian inference framework with neural networks to determine QGP properties from collision data, incorporating chemical and kinetic freeze-out separation.

Findings

Estimated shear and bulk viscosity to entropy ratios.

Determined initial entropy density and freeze-out temperatures.

Quantified uncertainties in QGP property extraction.

Abstract

A phenomenological analysis of the experimental measurements of transverse momentum spectra of identified charged hadrons and strange hyperons in Pb-Pb and Xe-Xe collisions at the LHC is presented. The analysis is based on the relativistic fluid dynamics description implemented in the numerically efficient \textsc{Fluid{\it u}M} approach. Building on our previous work, we separate in our treatment the chemical and kinetic freeze-out, and incorporate the partial chemical equilibrium to describe the late stages of the collision evolution. This analysis makes use of Bayesian inference to determine key parameters of the QGP evolution and its properties including the shear and bulk viscosity to entropy ratios, the initialisation time, the initial entropy density, and the freeze-out temperatures. The physics parameters and their posterior probabilities are extracted using a global search in multidimensional space with modern machine learning tools, such as ensembles of neural networks. We employ our newly developed fast framework to assess systematic uncertainties in the extracted model parameters by systematically varying key components of our analysis.