Multi-parameter Bayesian optimisation of laser-driven ion acceleration in particle-in-cell simulations

Abstract

High power laser-driven ion acceleration produces bright beams of energetic ions that have the potential to be applied in a wide range of sectors. The routine generation of optimised and stable ion beam properties is a key challenge for the exploitation of these novel sources. We demonstrate the optimisation of laser-driven proton acceleration in a programme of particle-in-cell simulations controlled by a Bayesian algorithm. Optimal laser and plasma conditions are identified four times faster for two input parameters, and approximately one thousand times faster for four input parameters, when compared to systematic, linear parametric variation. In addition, a non-trivial optimal condition for the front surface density scale length is discovered, which would have been difficult to identify by single variable scans. This approach enables rapid identification of optimal laser and target parameters in simulations, for use in guiding experiments, and has the potential to significantly accelerate the development and application of laser–plasma-based ion sources.

Keywords

• laser-driven ion acceleration
• Bayesian optimization
• machine learning