Predicting Laser-Induced Colors of Random Plasmonic Metasurfaces and Optimizing Image Multiplexing Using Deep...

Area of Science:

• Plasmonics and Nanophotonics
• Machine Learning Applications
• Materials Science

Background:

• Plasmonic metasurfaces offer bright, durable structural colors but face fabrication challenges for customized, large-area designs.
• Laser processing shows promise for precise metasurface fabrication, yet lacks predictive models for optical properties based on nanostructure statistics.

Purpose of the Study:

• To develop a predictive model for optical properties of laser-induced plasmonic metasurfaces.
• To enable accurate, large-area fabrication of metasurfaces for customized structural colors and multiplexed imaging.
• To improve laser processing technology for high-end applications.

Main Methods:

• Trained deep neural networks using thousands of experimental data points from laser-processed plasmonic metasurfaces.
• Utilized statistical properties of random metallic nanostructures for model input.
• Validated the deep learning approach through experimental demonstration of two-image multiplexing.

Main Results:

• Achieved high training accuracy, surpassing the perceptual just noticeable change.
• Enabled prediction of spectra and colors for metasurfaces in various observation modes.
• Successfully demonstrated laser-induced two-image multiplexing using predicted color charts.

Conclusions:

• Deep learning models can accurately predict optical properties of laser-processed plasmonic metasurfaces.
• This approach significantly enhances laser processing for color printing and multiplexing parameter optimization.
• The study provides a versatile algorithm for multiplexing implementation across various printing technologies and applications.