Opportunity: Machine-learning Driven Inverse Design of Optical Metasurfaces

The next generation of optical materials will achieve novel performance by utilizing wavelength-scale geometric features to manipulate light in ways beyond the capability of traditional bulk optical materials. However, advanced optimization strategies coupled with computational electromagnetic simulation must be employed to accelerate the design process due to the near-infinite design space for such optical metamaterials. Machine learning approaches have demonstrated the potential to overcome the limitations of ansatz and/or iterative approaches by inverting the design process, delivering optical metamaterial designs with desired properties. This opportunity will allow for the investigation and creation of machine learning models capable of designing optical metasurfaces to achieve the desired performance.