Gpu-accelerated 3d Time-domain Simulation of Rf Fields and Particle Interactions∗

The numerical simulation of electromagnetic fields and particle interactions in accelerator components can consume considerable computational resources. By performing the same computation on fast, highly parallel GPU hardware instead of conventional CPUs it is possible to achieve a 20x reduction in simulation time for the traditional 3D FDTD algorithm. For structures that are small compared to the RF wavelength, or that require fine grids to resolve, the FDTD technique is constrained by the Courant condition to use very small time steps compared to the RF period. To avoid this constraint we have implemented an implicit, complex-envelope 3D ADI-FDTD algorithm for the GPU and demonstrate a further 5x reduction in simulation time, now two orders of magnitude faster than conventional FDTD codes. Recently, a GPU-based particle interaction model has been introduced, for which results are reported. These algorithms form the basis of a new code, NEPTUNE, being developed to perform self-consistent 3D nonlinear simulations of vacuum electron devices.