Progress in 3d Space-charge Calculations in the Gpt Code

The mesh-based 3D space-charge routine in the GPT (General Particle Tracer, Pulsar Physics) code scales linearly with the number of particles in terms of CPU time and allows a million particles to be tracked on a normal PC. The crucial ingredient of the routine is a non-equidistant multigrid Poisson solver to calculate the electrostatic potential in the rest frame of the bunch. The solver has been optimized for very high and very low aspect ratio bunches present in state-of-the-art high-brightness electron accelerators. In this paper, we introduce a new meshing strategy based on a wavelet decomposition of the space-charge density. The numerical results show that the number of particles with large numerical error, typically located at the edges of the bunch, can be reduced with this new approach enormously.