Related Experiment Video
Updated: Dec 10, 2025
09:33
Demonstration of Equal-Intensity Beam Generation by Dielectric Metasurfaces
Published on: June 7, 2019
6.5K
Non-Empirical Large-Scale Search for Optical Metasurfaces
1Research Center for Functional Materials, National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba 305-0044, Japan.
Nanomaterials (Basel, Switzerland)
|September 5, 2020
Summary
Researchers developed a novel AI-driven method to discover high-performance optical metasurfaces without human design bias. This approach yields unpredictable, efficient all-dielectric metasurfaces and a unique light-focusing device.
Area of Science:
- Photonics and Materials Science
- Artificial Intelligence in Design
Background:
- Metasurfaces are engineered thin films for optical functions, typically relying on human expertise for complex designs.
- Existing methods often involve optimization or modification of known structures.
Purpose of the Study:
- To introduce a non-empirical, large-scale structural search method for discovering novel optical metasurfaces.
- To bypass human knowledge and experience in the design process.
- To identify high-performance metasurfaces with unprecedented functionalities.
Main Methods:
- Implementing a large-scale, non-empirical structural search algorithm.
- Utilizing machine search to explore vast design spaces for metasurfaces.
- Demonstrating the method's capability to find optically functional structures.
Main Results:
- Discovery of unpredictable, optically high-performance, all-dielectric metasurfaces.
- Identification of a novel light-focusing metadevice with diffraction-limited performance.
- Observation of a unique, invariant focal length property in the designed metadevice.
Conclusions:
- The non-empirical search method effectively generates advanced optical metasurfaces.
- AI-driven discovery offers a powerful alternative to knowledge-based design in photonics.
- The found metasurfaces and devices showcase significant potential for future optical applications.
