Feature extraction from nonequilibrium molecular dynamics data for high-speed flows

Nonequilibrium molecular dynamics (MD) simulations are useful for capturing the detailed dynamics of high-speed flows including shock waves and chemical reactions. Extracting continuum-scale information from MD simulations is challenging due to the stochastic nature of molecular motions stored in unstructured Lagrangian data. In this study, we investigate the use of data decomposition techniques to extract continuum-scale dynamics from nonequilibrium MD simulation data. This method is based on principal component analysis of the time series of molecular velocities. We extract the coherent dynamical information in a way similar to modal decomposition of Eulerian flow field data. The method is tested and validated by extracting two counter-traveling acoustic waves excited by moving pistons in a rectangular domain. We then apply this method to analyze flow structures in our recent MD data of piston-driven reactive shock waves in a stoichiometric $H_2/O_2$ mixture, at effective Mach numbers in the range of $2 \le M_s \le 5$. We extract the collective motion of each species in the post-shock region and discuss its implications to shock-induced chemical reactions.