Doubling the Convergence Speed of Planar Topology Optimization Using the Multigradient

Metasurfaces and metalenses promise the potential to be transformative in the way we design optical and nanophotonic systems. Because high performance is essential in these areas, metasurface designers have sought out device parameterizations which can reach very high efficiencies. Topology Optimization (TO) has been one such method owing to the freeform structures it is capable of producing. Indeed, TO is notable that despite facilitating very high degrees-of-freedom, as a local, gradient-based optimization technique, it nevertheless converges quite quickly compared with global optimization methods. In this paper, we share our recent work on improving the convergence of TO for the specific class of planar, lithographically fabricated metasurfaces using the multigradient. The new method is shown to converge twice as fast as the conventional method for a representative supercell design problem. We believe this method has the potential to transform how nanophotonic TO is done, with potential applications extending well beyond planar devices.