Global optimization of optical metasurfaces using the RCWA method

Abstract

A metaatom is a nanoparticle that interacts with light such that it creates a phaseshift. Through using these metaatoms as building blocks, one may create a metasurface. By simulating a metasurface with the S4 program, which uses the RCWA (Rigorous Coupled Wave Analysis) method, one may see how well it performs. When optimizing the structure, simulations were done such that the positions, radii and rotation of the elliptical cylinder metaatoms used for the metasurface were varied. It is possible to optimize a grating that bends normal incident light with an angle given by the periodicity of the metasurface in one direction. When slowly varying the periodicity of a metasurface grating (metagrating), it is possible to create multiple metagratings that later may be combined into a lens by combining the metagratings in large rings. These rings are then placed around a center piece, which uses a phase-mapping approach instead. By comparing results with the metasurface lens that J. Byrnes et al. [1] had done, the results are similar. Since they also used the RCWA method, a grating optimized with the FDTD (Finite difference time domain) method by Paniagua-Domínguez et al. [2] was simulated with the S4 method and compared to their FDTD simulations. The comparison gave a similar overall shape of the polarisation efficiencies, but the RCWA graphs had more noise in them. Through balancing the computational time against the convergence of the efficiencies, 400 vectors for simulations were used in order to optimize three new de flection gratings. The gratings were optimized for deflecting light with a wavelength of 1064 nm in angles of 30°, 45° and 60° in air respectively. These gratings were then fabricated and tested experimentally. The 30° grating did not work as well as in simulations, the 45° grating worked moderately similar, and the 60° grating worked very close to what was expected from simulations. Some more studies would be necessary in order to understand why the simulated efficiencies did not agree with the experimental efficiencies for the 30° grating.