Data from: Deep learning for accelerated all-dielectric metasurface design

Deep learning has risen to the forefront of many fields in recent years, overcoming challenges previously considered intractable with conventional means. Materials discovery and optimization is one such field. However, the metamaterials community currently lacks access to large datasets for training models. This dataset aims to fill this gap by providing approximately 21,000 simulated spectra for different geometric configurations of all-dielectric metasurfaces. For the geometry of the metamaterial, it is a supercell ADM comprised of four cylindrical unit cells. Each unit-cell cylinder is parameterized by a radius and a height (two parameters), yielding an 8-dimensional parameterization, x, for the geometry of the supercell. We select silicon (Si) for the metasurface material and, with no loss of generality, embed it within a vacuum for faster training set acquisition.