FBPIC: A spectral, quasi-3D, multi-GPU Particle-In-Cell code for plasma accelerators

Plasma accelerators are compact high-energy particle sources. Excited by an ultra-short laser or particle beam driver, the strong electric fields inside a plasma allow to accelerate electron bunches to GeV energy levels on mm length scales. Modelling the complex acceleration process requires computationally demanding Particle-In-Cell codes that self-consistently solve the electromagnetic particle interaction inside the plasma. Here, we present FBPIC - a highly parallel Particle-In-Cell code featuring a spectral electromagnetic solver that eliminates numerical instabilities common to traditional field solvers, a quasi-3D geometry that greatly reduces the computational costs and the Lorentz-boosted frame technique that scales down the required simulation time by orders of magnitude. The code is written in pure Python using Just-In-Time compilation to generate machine code supporting CPU multi-threading and multi-GPU execution.