A Data‐Mining‐Assisted Design of Structural Colors on Diamond Metasurfaces

Artificial colors generated by light–matter interactions are widely achieved usingmetal or dielectric nanostructures. However, those systems are susceptible toreported issues like high optical loss and insufficient robustness, limiting theirpractical deployment. To address these limitations, the concept of a low-loss andhigh-saturation color pixel design based on diamond metasurfaces is proposedand demonstrated. The numerical simulation and analytic evaluation areperformed to demonstrate the feasibility of realizing high-quality structural colorswith diamond metasurfaces. Moreover, a tensor completion-based algorithm isintroduced to assist the design of vivid colors, shortening the required runtime ofsimulation by four orders of magnitude. Such a tensor-based algorithm canprecisely predict a complete set of generated colors through partially simulatedconditions. Particularly, when the missing rate is lower than 0.8, the obtainedprediction accuracy is no less than 85.1%, 90.6%, and 90.9% for cube-, cylinder-,and cone-shaped diamond metasurfaces, respectively. The current studyprovides a solid theoretical foundation for achieving all-diamond optical meta-surfaces, and the present data-mining approach could be expanded to otherphotonic structure designs and parametric optimization.RESEARCH ARTICLEwww.adpr-journal.comAdv. Photonics Res.2022,3, 21002922100292 (1 of 9)© 2022 The Authors. Advanced Photonics Research published by Wiley-VCH GmbH