Comparative study on algorithmic strategies for achromatic metalens design

The design of achromatic metalens must simultaneously balance computational efficiency with physical performance. Addressing the limitations of machine learning methods, such as high data dependence and elevated memory costs, this study systematically compares the performance boundaries of particle swarm optimization (PSO), differential evolution (DE), Bayesian optimization (BO/BY) and their hybrid strategies. Through small-aperture achromatic metalens simulations, a Bayesian-guided hybrid algorithm employing boundary contraction strategy achieves over 50% speed improvement compared to single-algorithm approaches while maintaining controlled computational errors and significantly enhanced stability. Results demonstrate that traditional algorithms retain practical advantages in parameter optimization through lightweight architectures and physical interpretability, with hybrid strategies effectively balancing efficiency-performance trade-offs, thereby providing a cost-effective technical pathway for scalable manufacturing.