Neural-adjoint method for the inverse design of all-dielectric metasurfaces

All-dielectric metasurfaces exhibit exotic electromagnetic responses, similar to those obtained with metal-based metamaterials. Research in all-dielectric currently uses relatively simple unit-cell designs, but increased geometrical complexity may yield even greater scattering states. Although machine learning has recently been applied the design of impressive results, much more challenging task finding a geometry that yields desired spectra remains largely unsolved. We propose and demonstrate method capable accurate solutions ill-posed inverse problems, where conditions existence uniqueness are violated. A specific example metasurface which radiant exitance matching external quantum efficiency gallium antimonide is demonstrated. also show how neural-adjoint can intelligently grow search space include designs increasingly accurately approximate response. The not restricted case demonstrated be plasmonics, photonic crystal, other artificial materials.