Infrared broadband quarter-wave and half-wave plates synthesized from anisotropic Bézier metasurfaces

In this work a new technique for synthesizing metamaterials using Bézier surfaces is introduced. First, the computational efficiency for the optimization of a reconfigurable Bézier quarter-wave plate metasurface is compared to the popular technique of optimizing pixelized surfaces via a binary Genetic Algorithm (GA). For the presented design methodology, a real valued optimization technique is employed which is based on the Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES). When compared to the GA, the optimizations of Bézier surfaces using CMA-ES are shown to consistently arrive at better solutions with an order of magnitude reduction in the required number of function evaluations. Additionally, more examples of Bézier metasurfaces are presented in the form of broadband quarter-wave and half-wave plate designs operating at optical wavelengths, subsequently exhibiting bandwidths which outperform metasurface designs found in the current literature.