GPU-accelerated particle-in-cell (PIC) simulations of laser-ablated HED plasmas with the PSC code

The PSC code is a modern C++14 based PIC code that has been optimized for running large-scale 2-D and 3-D simulations of HED plasmas on GPU-based machines like ORNL's Summit. These end-to-end simulations, which self-consistently generate the magnetic fields through the Biermann-battery effect and follow their evolution until they reconnect as the plasma plumes collide are challenging to run efficiently due to their large and evolving load imbalance. We describe enhancements to the PSC code, including the integration with the gtensor library for performance portability, using pooled GPU memory allocation using the RMM library, and a defragmentation algorithm.  As an application we present a large 3-d simulation of plumes that collide and their intrinsic anti-parallel magnetic fields reconnect. We have also added the option to add a separate seed population of high-energy electrons to investigate Fermi acceleration.