Metamaterials enable tailoring of light–matter interactions, driving discoveries which fuel novel applications. Deep neural networks (DNNs) have shown marked achievements in metamaterials research, however they are black boxes, and it is unknown how they work. We present a causal DNN where the learned physics is available to the user. Here, the condition of causality is enforced through a deep Lorentz layer which takes in the geometry of an all-dielectric metamaterial, and outputs the causal frequency-dependent permittivity and permeability. The ability of the LNN to learn metamaterial physics is verified with examples, and results are compared to theory and simulations.

© (2023) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.

Omar Khatib, Simiao Ren, Jordan Malof, and Willie J. Padilla, "Metamaterial design with physics informed neural networks," Proc. SPIE PC12431, Photonic and Phononic Properties of Engineered Nanostructures XIII, PC124310O (Presented at SPIE OPTO: February 01, 2023; Published: 17 March 2023); https://doi.org/10.1117/12.2661089.6321431321112.