Large-scale molecular dynamics simulations of electron nonlocal transport

Hot, dense plasmas produced during the implosion of inertial confinement fusion (ICF) capsule are characterized by steep temperature and density gradients. Because of these gradients, the thermal mean-free path of the electron becomes large with respect to the characteristic scale length of the temperature. Thus, electrons with mean-free path collisional larger than the scale length can escape gradients before being scattered and depositing their energy into the plasma, leading to a distortion of the distribution function away from Maxwellian. This leads to nonlocal transport of the electrons energy as the electrons are delocalized. In this work, reduced nonlocal electron transport model proposed by Schurtz, Nicolai and Busquet (SNB) [Phys. Plasmas 7, 4238 (2000)] is compared to classical molecular dynamics (MD) simulations. The nonequilibrium classical MD simulations are performed with particles interactions modeled by quantum statistical potentials. Using machine learning algorithms, we improve the SNB model by finding accurate and efficient kernel for the electron heat conduction.