Speaker: Longgang Pang (CCNU)
DATE: April 03, 2025, Thursday, 10:30 am(UTC+8)
Online: Zoom meeting ID:421173735 Passcode: 644179
Abstract:
Relativistic heavy-ion collisions recreate an exotic state of nuclear matter-0Gp-mirroring theprimordial matter that permeated the universe within its first microseconds after the Big Bang. TheQGP is the hottest substance ever created on Earth, hosts record-breaking vorticity and magneticfields, and approaches the theoretical limit of minimal viscosity (n/s ~ 1/4m). Despite its fleetinglifetime (~10-23 seconds), the QGP encodes profound insights into findamental physics, accessiblethrough the momentum distribution of final-state hadrons. These observables unveil criticalfeatures of QCD, including the phase structure, equation of state, transport coefficients, and theinitial-state nuclear configurations. In this talk, I will present our recent work at the intersection ofhigh-energy nuclear physics and artificial intelligence (Al). We leverage Al to revolutionize
scientific inquiry across multiple frontiers:Solving inverse problems to reconstruct the OGP properties and the initial state of HICsDesigning physics-informed neural networks for effective theory in QCDSimulating medium response dynamics with unprecedented precision.Developing large language model (LLM) agents to automate and enhance scientific workflowsThis interdisciplinary study bridges computational physics and machine learning, opening newparadigms for understanding strongly coupled quantum systems
About the speaker:
Longgang Pang is a professor at CCNU, specializing in highenergy nuclear physics and AI-driven scientific computingHe earned his B.Sc.(2006)and Ph.D.(2012)from USTCwith a joint program at LBNL. After postdocs at Institute oParticle Physics(China),FIAS(Germany)and UC Berkelev(US), he joined CCNU in 2019.He developed CLVisc(GPUaccelerated OGPhydrodynamics)andcontributedtoSMASH,pioneered Al for QGP tomography and LLMdriven research automation. His current focus are:CLVis(3.0,Al for QGP inverse/generative problems and LLMpowered scientific tools.