Fast optimization for betatron radiation from laser wakefield acceleration based on Bayesian optimization

Abstract

Laser wakefield acceleration, a highly nonlinear process, can provide high-quality electron beams and betatron radiation. However, it has always been a challenge to optimize the photon number of betatron radiation because X-ray production is the secondary process followed by electron generation and many influential parameters are highly coupled with each other. To balance these parameters and optimize target value, we use a machine-learning algorithm called Bayesian optimization that ignores the complicated middle process and so is suitable to solve this kind of problems. Without designed initialization, we optimize betatron radiation in experiments by changing plasma density and laser focal position. The X-ray photon number doubles compared with initial condition within 10 iterations, which shows the rapidity of this algorithm. PIC simulation also illustrates the complexity of the optimization and reveals the reason why X-ray photon number increases.