Machine Learning for Classifying Hydrodynamic Breakdown in HED Plasmas

Correct prediction of shock width is important for evaluating atomic mixing in high energy density plasma experiments, e.g. shell-fuel mixing in inertial confinement fusion capsules. In this study we use machine learning to attempt to identify when 1d fluid descriptions of a multi species plasma reminiscent of a capsule interface fail to adequately describe a standing shock, and determine when a kinetic description is appropriate. We compare the effectiveness of various combinations of several different metrics for non-equilibrium behavior, such as shock width, species mixing, temperature anisotropy, Knudsen numbers, and deviations from Maxwellian distributions between a fluid and kinetic model to efficiently train a perceptron identify the breakdown regime.