Freeform metasurface design based on topology optimization

Metasurfaces are thin-film electromagnetic devices with subwavelength-scale geometric structuring. They can be tailored to produce a broad range of optical functions due to the strong relationship between electromagnetic response and geometric shape. An open challenge has been understanding how to produce an ideal metasurface design when presented with a desired electromagnetic response. This article discusses the use of topology optimization as a design platform for high-performance, freeform metasurfaces. Two types of topology optimizers are covered-local gradient-based optimizers that leverage the adjoint variables method, and global population-based optimizers that reframe the optimization process as the training of a generative neural network. It is anticipated that these inverse design concepts will push metasurface performance to the physical limits of structured media and enable new functionalities in electromagnetic systems.