Speaker
Description
In the design of high-power attosecond X-ray free-electron laser (XFEL) pulses, strongly coupled collective effects and many tunable parameters turn layout and parameter choice into a challenging multi-objective optimization issue. Conventional evolutionary approaches such as NSGA-II and NSGA-III require a very large number of high-fidelity start-to-end simulations, which makes systematic optimization prohibitively expensive for state-of-the-art XFEL facilities. We propose a data-driven surrogate framework for high-dimensional multi-objective optimization in this setting. A machine-learning surrogate model is trained on a limited set of high-fidelity simulations and then replaces most simulation calls in the optimization loop. As a first application, we optimize the AttoSHINE scheme for the Shanghai High Repetition Rate XFEL and Extreme Light Facility (SHINE). The resulting Pareto-optimal solutions reveal non-trivial trade-offs in the AttoSHINE design and identify parameter regions that support terawatt-level attosecond pulses at greatly reduced computational cost compared with direct NSGA-II or NSGA-III optimization. The proposed framework offers an efficient and flexible route to multi-objective design of attosecond XFEL beamlines and, more broadly, complex accelerator systems.
| In which format do you inted to submit your paper? | LaTeX |
|---|