|
Multifunctional metasurfaces perform different functions depending on the wavelength, polarization, or wavefront of the incident light. Designing such metasurfaces require more degrees of freedom (DOF) than what is available in a single layer metasurface, and stacking metasurface layers is one of the approaches for achieving the required DOF for realizing multifunctional metasurfaces. In the conventional metasurface design technique used for designing single layer metasurfaces, the couplings among the meta-atoms are ignored; however, the meta-atoms in multi-layer metasurfaces exhibit significant mutual couplings and multiple scattering phenomena are not negligible. As a result, multi-layer metasurfaces designed using the conventional techniques have low efficiencies. In this talk, we will present an inverse design technique that is suitable for designing efficient large-scale multi-layer metasurfaces. The method is based on a combination of the gradient descent optimization and the adjoint sensitivity techniques and is used to design efficient parametrized multifunctional metasurfaces. The design of multifunctional metasurfaces is cast as a multi-objective optimization problem and the optimal values of meta-atom geometrical parameters are found through an iterative approach. The sensitivities of the objective function and the metasurface response are computed using full-wave simulations; therefore, the mutual interactions and the multiple scattering effects are accurately considered. To demonstrate the effectiveness of the method, we present a bi-layer double-wavelength metasurface composed of more than 2,000 amorphous silicon nano-posts that are embedded in silicon dioxide and arranged in two stacked layers. The bi-layer metasurface projects two different patterns with more than 65% efficiency when illuminated with two different wavelengths.
|
