Speaker
Description
Understanding the non-perturbative properties of the quark-gluon plasma (QGP) formed in ultrarelativistic heavy-ion collisions remains a central challenge in QCD phenomenology. Conventional quasi-particle approaches rely on analytic parametrizations that often fail to capture the complex thermal structure of the medium, particularly near the deconfinement crossover region. To address this, we develop a deep-learning-assisted quasi-particle model (DLQPM) [1] in which residual neural networks are trained directly on lattice QCD thermodynamic data, enabling a robust, non-perturbative extraction of the temperature-dependent medium properties of the QGP without the assumptions inherent to perturbative frameworks.
The machine-learning-derived medium properties are employed to construct a complex-valued medium-modified Cornell potential that encodes both color screening and Landau damping effects, thereby driving the dissociation of heavy quarkonia in the deconfined medium. Heavy quarkonia, including charmonium (J/psi, psi(2S)) and bottomonium (Upsilon(1S), Upsilon(2S)), serve as important probes of QGP formation owing to their sequential suppression pattern experimentally observed at RHIC and LHC [2, 3], directly reflecting the temperature and density conditions of the created medium. The present framework offers a flexible, data-driven, and physically well-motivated path toward bridging first-principles lattice QCD constraints with quarkonium suppression measurements, with direct relevance to the ongoing heavy-ion programs at ALICE, CMS, and ATLAS.
References:
[1] M. Y. Jamal, F. P. Li, L. G. Pang and G. Y. Qin, Phys. Rev. C 113, 034915 (2026).
[2] T. Matsui and H. Satz, Phys. Lett. B 178, 416 (1986).
[3] S. Chatrchyan et al. (CMS Collaboration), Eur. Phys. J. C 72, 1945 (2012).