Fast, Accurate Simulations for Free-Electron Laser Oscillators

The current generation of high-gain free-electron lasers require long undulators to achieve saturation power, and can extract only 1% of the electron beam power. High-efficiency recirculating free-electron laser concepts require substantially shorter undulators and may be able to achieve more than 10% extraction efficiency, but will require improved modeling and controls to realize their efficiency. We will use reduced models to dramatically reduce the simulation time for a single pass through the free-electron laser undulator and the optical beamline, for faster simulations of the multi-pass free-electron lasers. We will use these tools to develop machine learning based methods for operating and optimizing the performance of a proof-of-concept TESSO high-efficiency recirculating free-electron laser experiment. We derived reduced models to simulate the free-electron laser amplifier and optical beamlines, and are actively benchmarking implementations of these models against well-established codes. We have established that the optical beamline model is over 500× faster than a paraxial optics code, and the free-electron laser amplifier code promises to be at least 100× faster than existing codes. During Phase II RadiaSoft will develop these reduced models into documented, usable codes. We will use these codes to build machine learning models, which we will use for virtual diagnostics and feedback in controls to optimize an experimental demonstration of the TESSO recirculating free-electron laser concept. Commercial Applications and Other Benefits: The codes will enable more efficient design for future free-electron laser systems and optical beamlines. The machine learning diagnostics and operations tools built will reduce down time and improve performance for industrial and medical accelerators.