Quantum-accurate molecular dynamics shock simulations at experimental time and length scales

Recent experiments using powerful lasers, pulsed power and bright X-ray sources enable transformative insights into behavior of matter at extreme conditions. To obtain significant science return from these sophisticated and expensive experiments, predictive atomic-scale simulations of dynamic compression at experimental time and length scales are urgently sought. In this talk I describe recent advances in developing machine learning interatomic potentials that provide description of bond-breaking and remaking at extreme conditions with unprecedented quantum accuracy. Efficient implementation of these potentials on GPU based supercomputers opens up an unprecedented opportunity to perform quantum accurate molecular dynamics simulations at time and length scales that allows direct comparison with laser-driven shock experiments. Shock MD simulations of micrometer thick diamond samples uncovered atomic-scale mechanisms of anomalously high strength and orientational dependence of inelastic deformations. The latter persist up to the point of complete melting along diamond Hugoniot.