Numerical analysis of the plasma density profile of a laser wakefield accelerator using multi-objective Bayesian optimization

Laser wakefield acceleration (LWFA) provides a compact pathway to generate high-energy electron beams, but its performance depends on a large set of coupled, nonlinear parameters. Machine-learning approaches, particularly Bayesian optimization (BO), have become valuable for efficiently exploring this space. In this study, we examine how the longitudinal plasma density profile affects the properties of LWFA electron beams using high-fidelity quasi-3D particle-in-cell simulations guided by BO. We focus specifically on the influence of the injection gradient and the plasma density down-ramp length. This work establishes a systematic methodology for multi-parameter and multi-objectives optimization of LWFAs and offers design guidance for the upcoming laser-plasma acceleration center installation. The framework is general and adaptable to other accelerator configurations and optimization goals.