Bayesian optimisation for ultrafast x-ray generation from dual-stage laser-plasma accelerators

Laser wakefield acceleration (LWFA) has proven highly effective in producing high-energy electron beams, which then generate bright femtosecond X-ray pulses through betatron oscillations. A drawback of this method is its low conversion efficiency and relatively low photon flux. One promising approach to increase the brightness of X-ray radiation is the so called passive plasma lensing. In this method, electrons are injected and accelerated in the first gas jet – the accelerator – and the transverse oscillations of the electron beam are then amplified in the second jet with higher density – the radiator. It has been shown both theoretically and experimentally that in the radiator, the electron beam is focused, and the magnitude of the betatron radiation is increased. We demonstrate how Bayesian optimisation can be used to optimise the parameters of this multi-jet scheme using multidimensional particle-in-cell simulations.