Abstract:
Our study presents an innovative approach to metasurface design, marrying Fourier techniques from image processing with artificial intelligence (AI). Metasurfaces, vital in compact optical system creation, have been a focus. Conventional topological optimization methods show promise but face challenges in computational efficiency, especially with large-scale devices. AI-driven techniques, though effective, are often limited to devices with few design parameters. Our proposed methodology addresses these issues, offering a robust design framework for expansive metasurfaces. We interpret unit cell dimensions as Fourier series coefficients, simplifying design complexities and addressing periodicity concerns. By utilizing AI on the captured Fourier series coefficients, we drastically reduce design parameters, facilitating specialized AI metasurface applications. This fusion of Fourier and AI methodologies promises breakthroughs in metasurface design, enriching optical engineering possibilities.
|