Data-Driven Reduced-Order Modeling of Plasma Charge Dynamics in Electromagnetic Particle-in-Cell Simulations

Electromagnetic particle-in-cell (EMPIC) kinetic plasma simulations are widely used in simulation of high-energy, laser ignited devices, high-power microwave sources, particle accelerators, study of ionosphere, magnetosphere phenomena, and many more applications. EMPIC simulations have been historically popular due to their ability of capturing the nonlinear wave-particle interaction accurately. However, particle-in-cell simulations requires a large number of particles (typically in several thousands or millions) to produce accurate and physically meaningful results. The requirement for large number of charged particles often becomes the computational bottleneck for EMPIC simulations. The dynamics of these charged particles are manifested through the charge density and current density at the mesh nodes and mesh edges respectively. In this work, we present a data-driven approach based on dynamic mode decomposition (DMD) [1]–[3] for reduced-order modeling of the time evolution of charge density at the mesh nodes. Such DMD based reduced-order model helps in rapid prediction of charge density in future time at any location of the solution domain, thus reducing the overall computation time. The extracted features from the data such as DMD modes and frequencies, also help in analysis and diagnostic purposes.