Ab initio PIC simulations and data-driven techniques for multi-fidelity or reduced plasma models
Luis Oliveira e Silva
Instituto Superior Tecnico, University of Lisbon
Physics
Particle-in-cell (PIC) simulations provide an ab initio kinetic description of plasmas with minimal physics approximations and well-characterized numerical properties. Recent advances in high-performance computing now enable direct, quantitative comparisons between high-fidelity PIC simulations and experiments,
I will highlight the physical fidelity of the PIC approach through recent large-scale simulation–experiment comparisons, and discuss how these simulations can serve as reference models in a multi-fidelity modeling framework. I will then describe emerging data-driven and machine-learning techniques that exploit PIC data for discovery science, reduced modeling, and the acceleration of kinetic plasma simulations, and conclude with an outlook on future opportunities at the interface of PIC, multi-fidelity methods, and learning from data.