Machine Learning Optimization of Laser-Driven Ion Beam Characteristics

We develop a machine learning framework, leveraging Gaussian Process Bayesian Optimization, to maximize user-defined beam characteristics for a high-intensity laser-driven ion beam. Our method utilizes a particle-in-cell simulation to generate beam data according to specific criteria, which then serves as the input for machine learning-driven optimization. First, we perform an initial parameter sweep of the input space. We then explore this parameter landscape using a computationally efficient Gaussian Process surrogate model. Following this, we employ Bayesian Optimization to efficiently identify the parameters that maximize or minimize the desired beam characteristic.