Artificial Intelligence Application in Nonlinear Beam Dynamics Study for Future HEP Accelerators

Yue Hao, Michigan State University, Principal Investigator

To support the scientific goals of high energy physics (HEP), the future HEP accelerators aim at higher beam intensities and involve more complicated dynamics, which requires a deeper understanding of nonlinear dynamics to ensure long-term beam stability.  As the concepts of the nonlinear integrable optics are being tested at the IOTA test facility in Fermi National Laboratory, it is essential to develop theoretical tools to analyze the nonlinear dynamics in both the traditional nonlinear dynamics problem based on a linear lattice and the new nonlinear integrable optics.  In this proposal, we propose to study an AI-based method that can be applied in both cases.  The proposed method adopts the concept of the Koopman operator represented by an autoencoder.  Meanwhile, it also exploits physics knowledge, such as the KAM theory and the normal form, to reduce the demand for training data and improve explainability.  The method, if successfully demonstrated, paves the road to systematically applying AI-based methods in designing and optimizing the nonlinear dynamics problem in future HEP accelerators.