Breaking the Radiation Frequency Limit in PIC Codes

Radiation emission plasmas is often a result of collective effects associated with the dynamics of relativistic charged particles. A common numerical approach to model their motion involves the Particle-In-Cell scheme which solves the full set of Maxwell's equations and the relativistic Lorentz force for the charged particles. The recently developed Radiation Diagnostic for OSIRIS (RaDiO) can retrieve the emitted spatiotemporal electromagnetic field structure of the emitted radiation in OSIRIS simulations, even at wavelengths smaller than the PIC resolution, by relying on the Liénard-Wiechert Potentials. OSIRIS can run with a high level of efficiency in most of the largest CPU-based supercomputers. Nevertheless, in recent years, GPU accelerator boards have been employed in supercomputers to the point where some of the most powerful machines nowadays are GPU-based systems. In these architectures, the CPU nodes are equipped with additional GPU boards that can be used to perform highly parallelizable computations such as obtaining the emitted radiation in a PIC code. This work describes the implementation of the radiation algorithm in the GPU architecture and its integration into OSIRIS. Several performance benchmarks confirm the efficiency and applicability of this diagnostic.