Speaker
Description
Space-charge effects can limit beam quality and accelerator performance in high-intensity and high-brightness machines. This contribution focuses on OPALX, the GPU-enabled successor framework to OPAL, as a scalable and performance-portable platform for modeling space-charge effects in conventional accelerators.
For beams with large energy spread, electrostatic space-charge solvers is no longer accurate. The existing OPAL energy-binning model approximates this interaction by partitioning particles into N uniform energy bins. Electrostatic fields are then solved independently for each bin and combined to approximate relativistic space-charge effects. This model is used during particle emission and before bunch compression.
In this paper, we extend the existing OPAL energy-binning model in OPALX with adaptive energy binning. The method no longer requires a fixed number of uniform bins. Instead, it chooses the number of energy bins and their boundaries from the particle distribution. The chosen binning minimizes a prescribed cost criterion. This allows direct control of the trade-off between modeling error, accuracy, and solver cost.
The current implementation starts from a fine energy histogram. It then merges bins using dynamic programming and an information-theoretic cost function. We show that this approach preserves the accuracy of the OPAL binning model. At the same time, it substantially reduces the number of space-charge solver calls in SwissFEL and AWA-gun electron-gun simulations.