Efficient optimization of plasma surface high harmonic generation by an improved Bayesian strategy

Abstract

Plasma surface high-order harmonics generation (SHHG) driven by intense laser pulses on plasma targets enables a high-quality extreme ultraviolet source with high pulse energy and outstanding spatiotemporal coherence. Optimizing the performance of SHHG is important for its applications in single-shot imaging and absorption spectroscopy. In this work, we demonstrate the optimization of laser-driven SHHG by an improved Bayesian strategy in conjunction with particle-in-cell simulations. A traditional Bayesian algorithm is first employed to optimize the SHHG intensity in a two-dimensional space of parameter. Then an improved Bayesian strategy, using the Latin hypercube sampling technique and a dynamic acquisition strategy, is developed to overcome the curse of dimensionality and the risk of local optima in a high-dimensional space optimization. The improved Bayesian optimization approach is efficient and robust in three-dimensionally optimizing the harmonic ellipticity, paving the way for the upcoming SHHG experiments with a considerable repetition rate.

Figures (6)

• Figure 1: The workflow of Bayesian optimization based on Gaussian process regression.
• Figure 2: (a) Distribution of the electric field E$_y$ of the attosecond pulse in the 2D parameter space (incident angle $\theta$ and preplasma length $L_{\rm{pre}}$). The data is obtained using a grid search method in conjunction with PIC simulations. The '$\times$' symbols mark the predicting points during the BO routine and the optimal condition for maximizing intensity is marked with 'O'. (b) Dependence of electric field E$_y$ of the attosecond pulse on the iteration number using BO. The blue points indicate the initial dataset and the green ones are the iteration points. The red smoothed line is plotted using the average of every 5 neighbor iteration points. The optimal point is marked with a star.
• Figure 3: The electric field E$_y$ of the attosecond pulses for different parameters (a) $\theta=38.75$ degree, L$_{\rm{pre}} = 0.423~\lambda_L$; (b) $\theta=50.05$ degree, L$_{\rm{pre}} = 0.651~\lambda_L$. The latter case is the optimized one predicted by BO.
• Figure 4: Dependence of the ellipticity of the attosecond pulse on the iteration number using (a) an improved DAS-BO and (b) a classical BO. The blue points indicate the initial dataset and the green ones are the iteration points. The purple smoothed line is plotted using the average of every 5 neighbor iteration points.
• Figure 5: (a) The ellipticity and (c) the structure of the attosecond pulse for a case of low ellipticity $\epsilon=0.17$ using the parameters: incident angle $\theta=7.01$ degree, inclination angle $\alpha=46.60$ degree, L$_{\rm{pre}} = 0.72~\lambda_L$. (b) and (d) are the same plots but for a case of high ellipticity $\epsilon=0.98$ using the parameters: incident angle $\theta=46.27$ degree, inclination angle $\alpha=50.68$ degree, L$_{\rm{pre}} = 0.76~\lambda_L$. In (a) and (b), the red line is for the normalized intensity of the attosecond pulse while the green line is for the ellipticity. Since the ellipticity changes with time, we use its value when the intensity peaks.