|
In our previous work, a meta-surface was designed using adjoint-based topology optimization for high-efficiency. However, since the design pattern was complicated and the size of element was small, it was hard to maintain in the actual process and was not appropriate for commercialization. To solve this problem, we use the adjoint inverse design method by applying the local curvature filter algorithm and curvature flow technique in this study. The meta-surface design was conducted using Fourier modal method, based on rigorous coupled-wave analysis. It is assumed that the meta-surface is composed of pixels having a relative dielectric constant of the dielectric material. The dielectric constant distribution in the meta-lattice region is a design variable for optimization, and the dielectric constant of the air layer is converted over the duration. Traditionally, Gaussian filter was used to change to a processable pattern. However, since it does not maintain high efficiency characteristics, we used the local curvature filter (LCF) method. As a result, the results of efficiently finding and filtering small and complex patterns while maintaining characteristics were acquired. The LCF detects it as a local area according to the degree of curvature. The detected area is filtered using a Bernstein filter, and then combined with the global pattern again. In addition, we re-progress filtering for smooth patterns using the curve flow to adjust the curve threshold value to design the meta-surface. Consequently, we propose the method and theory of novel algorithm as a way of maintaining the high efficiency properties of the meta-surface in practice.
|
