Efficient Genetic Algorithm Framework for Inverse Design of Multilayer Spaceplate

초록

The miniaturization of optical systems has accelerated the development of planar optical components. In particular, spaceplate technology, which compresses the free space between optical elements, offers a promising route to novel flat optics. However, existing inverse-design approaches for spaceplates often require long computation times, rely on continuous optimization, or fail to guarantee sufficient transmission efficiency, limiting their practical usability. In this study, we propose a simple and efficient framework that integrates a genetic algorithm with the transfer-matrix method and angular spectrum method. Optimization is performed under a consistent angular design condition (approximately ±10°) to ensure stable convergence. To improve convergence, we introduce an optimization filter and a compact yet effective figure of merit. The algorithm successfully optimizes multilayer spaceplates with high performance within an average of 15 minutes on standard computing hardware. To assess its statistical robustness, we analyzed convergence probability, quality, and time as functions of the Gaussian beam’s Rayleigh range, and extended the same evaluation to spherical waves formed by a lens. The results demonstrate that the proposed framework enables rapid and reliable inverse design of multilayer spaceplates, offers a practical tool for efficient design, and contributes to the realization of ultra-thin optical systems.

키워드