HYPERS: A unidimensional asynchronous framework for multiscale hybrid simulations

Treatment of disparate timescales remains a major challenge in computational science. Previously we introduced a new asynchronous approach to the explicit time integration of multiscale numerical systems based on partial differential equations and particle techniques – self-adaptive discrete-event simulation (DES). In DES time increments for updates of numerical variables (events) are predicted by imposing small but finite bounds to their changes, and event synchronization requirements are defined with physical rules. The feasibility and superior metrics of DES were demonstrated for several different physical problems in one dimension. Here we extend DES to multiple dimensions by introducing a unidimensional infrastructure for asynchronous simulations on logically uniform meshes. As the first example of this infrastructure we present a new event-driven electromagnetic hybrid code, HYPERS (HYbrid Particle Event-Resolved Simulator). This code is validated in two dimensions against a state-of-the-art time-stepping hybrid code on a numerically challenging problem which describes the interaction between the magnetized plasma flow and a magnetic dipole obstacle. We find that HYPERS achieves significant speedups and remains physically accurate in a broad mesh resolution range, including coarser resolutions where the time-driven code produces numerical artifacts.