Surrogate Modeling of Ion Acceleration with Invertible Neural Networks

The interaction of overdense plasma with ultra-intense laser pulses presents a promising approach to enable the development of very compact ion sources. Prospective applications of high-energetic protons and ions include, but are not limited to, medical applications (in particular ion beam radiotherapy), laboratory astrophysics and nuclear fusion. However, current records for maximum proton energies (94 MeV, 2018) are still below the required values for the aforementioned applications (typically in the range of 150-250 MeV), and especially challenges such as stability and spectral control remain unsolved to this day. In particular, significant effort per experiment and a high-dimensional design space renders naive sampling approaches ineffective. Furthermore, due to the strong nonlinearities of the underlying laser-plasma physics, synthetic observations by means of particle-in-cell (PIC) simulations are computationally very costly, and the maximum distance between two sampling points is strongly limited as well. Consequently, in order to build useful surrogate models for future data generation and experimental understanding and control, a combination of highly optimized simulation codes (where we employ PIConGPU), powerful data-based methods, such as artificial neural networks (ANNs), and modern sampling approaches are essential. Specifically, we employ invertible neural networks for bidirectional learning of input (parameter) and output (observables) and convolutional autoencoder to reduce intermediate field data to a lower-dimensional latent representation.