Efficient Full Characterization of Centimeter-Scale Metasurfaces by Accurate Segmentation Using Augmented Partial Factorization

Metasurfaces are arrays of nanostructures that enable precise control of light, providing a compact, planar form factor with the potential to surpass conventional optical devices. However, fully characterizing large-area, multichannel metasurfaces presents significant challenges due to high memory and time consumption. Conventional methods for modeling metasurfaces either scale drastically with the size of the structure, making the full characterization of large-area or 3D metasurfaces impractical, or rely on overly simplistic approximations, leading to inaccuracy especially at large angles. Recently, the augmented partial factorization technique has enabled significant speed improvements. However, this method still encounters memory limitations when characterizing very large structures. In this work, we propose a novel approach to divide metasurfaces into smaller segments, characterize them independently, and then carefully stitch them together to obtain a full characterization. This method remarkably reduces computational costs while maintaining a high accuracy less than 1% for all incident angles. With this approach, we successfully enabled the characterization of extremely large 2D metasurfaces up to 1 cm in size in less than 21 min, while using maximum 97 GiB of memory with less than 1% error.