Deep Neural Networks for the Prediction of the Optical Properties and the Design of Metamaterials

Paper in proceeding, 2021

We will present our work on using deep neural networks for the prediction of the optical properties of free-form nanophotonic structures and for their inverse design. We designed neural networks and created training data, which were labelled with the respective optical properties and the degree of manufacturability. Furthermore, a cGAN network with 5 neural networks was developed to overcome problems with non-uniqueness and mode collapse, and to increase the experimental feasibility of the generated structures.