Intelligent multi-objective optimization for high-performance betatron radiation sources

Betatron radiation, generated by electrons undergoing transverse oscillations in laser wakefield acceleration, represents a promising compact X-ray source. It features micrometer-scale source size, femtosecond pulse duration, and high peak brightness comparable to that of third-generation synchrotron facilities. These properties make it highly suitable for phase-contrast imaging and ultrafast dynamic probing. However, betatron radiation sources still need improvements in photon number, critical energy, and source size, and optimization is challenging owing to the complex coupling and trade-offs among these key performance parameters. Traditional single-objective optimization strategies often fail to achieve a global optimum. This paper proposes a multi-objective Bayesian optimization framework integrated with particle-in-cell simulations to address this challenge. By systematically tuning critical parameters including laser focal position, plasma density, and plasma length, our method efficiently navigates the parameter space and identifies the Pareto front within 90 iterations. Results demonstrate that photon number and critical energy can be simultaneously optimized, while the source size is the primary limiting factor. The established predictive model further aids in physical analysis and application-specific source design, paving the way for intelligent, closed-loop control of high-performance betatron radiation sources.