DESC0023887

STATEMENT OF THE PROBLEM Structure-based wakefield accelerators promise orders of magnitude improvements in accelerator gradient over conventional radiofrequency-based technologies, and have been identified as a candidate technology for future applications ranging from compact free electron lasers to high energy colliders. However, achieving the desired peak energies and beam quality requires meter-scale structures with tighter tolerances, introducing new constraints on structure and beam characteristics. Accurate simulations of these structures require tools with both greater sophistication and performance in order to realize these advances. GENERAL STATEMENT OF HOW THE PROBLEM IS BEING ADDRESSED We propose to build on recent advances in numerical analysis and machine learning to develop novel surrogate model implementations for capturing the wakefield response of arbitrary structures in conjunction with self-consistent beam dynamics. The resulting tools will be integrated into community beam dynamics codes to support global optimization of beamlines leveraging wakefield accelerator technology. Our vision is to provide fast, flexible, and user-friendly software tools for the design and optimization of beamstructure interactions. WHAT IS TO BE DONE IN PHASE I? During Phase I, we will develop key algorithms and prototype implementation of novel surrogate models for computationally efficient simulation of beam-structure interaction in common structure geometries, thereby establishing the feasibility of the overall approach. We will then integrate these new surrogate models into an open source community beam dynamics code, to enable start-to-end simulations of accelerator structures. The resulting software will be benchmarked against first-principles simulations and experimental results at a federally-funded research facility. COMMERCIAL APPLICATIONS AND OTHER BENEFITS The proposed research will drive innovation in burgeoning accelerator technologies, benefiting National Laboratories and university facilities by improving particle accelerator performance while reducing experimental design and planning costs. Complementary efforts to develop efficient data representation schemes via machine learning will deliver value to broader scientific computing and research applications. Successful deployment of these technologies will enable broad commercial applications including industrial accelerators for food safety and waste treatment, medical accelerators for imaging and treatment, and defense accelerators for inspection of and protection against threats.