Specific wavelength peak emulation with amorphous metastructures.

Area of Science:

• Nanophotonics
• Metamaterials
• Computational Physics

Background:

• Metasurface design is traditionally slow and resource-intensive.
• Developing new optical materials requires efficient design methodologies.

Purpose of the Study:

• To employ a deep convolutional generative adversarial network (DCGAN) for generating nanohole metastructure designs.
• To achieve desired transmittance spectra in the visible range efficiently.

Main Methods:

• Utilized a DCGAN model trained on optical properties and geometric parameters.
• Generated diverse and manufacturable nanohole metastructure designs.
• Fabricated and experimentally characterized the generated designs.

Main Results:

• The DCGAN model successfully generated designs matching target transmittance spectra.
• Generated designs were diverse, manufacturable, and optically accurate.
• Experimental validation confirmed the model's effectiveness.

Conclusions:

• The DCGAN approach significantly accelerates metasurface design.
• This AI-driven method offers high efficiency and generalizability for metamaterial development.
• The approach bypasses the need for prior knowledge of structure-property relationships.