Nanophotonics Breakthrough Achieves 99.9% Accuracy in Structural Color Design via Novel Sampling Network

New Mixture Probability Sampling Network enables ultraprecise structural color design with 99.9% accuracy, advancing augmented reality and encryption tech.

By: AXL Media

Published: Apr 8, 2026, 11:19 AM EDT

Source: Information for this report was sourced from EurekAlert!

A Breakthrough in Computational Nanophotonics

An international team of scientists has introduced a Mixture Probability Sampling Network, known as MPSN, to fundamentally change how light interacts with nanostructures to create color. Unlike traditional pigments that rely on chemical properties, structural colors are derived from the physical geometry of surfaces at the nanoscale. According to the research team led by Professors Xinbin Cheng and Yuzhi Shi, this development addresses the persistent difficulty of inverse design, where finding the exact physical structure required to produce a specific target color has historically been hindered by the one to many relationship between geometry and light.

Overcoming the Hurdles of Solution Degeneracy

Existing neural network models frequently struggle with accuracy when multiple structural solutions could theoretically produce the same color outcome. The MPSN architecture overcomes this by combining a mixture density network with a pretrained forward network to sample multiple structural solutions simultaneously. This framework ensures that training stability remains unaffected by solution degeneracy, allowing the system to identify the most effective match for any given design objective. The researchers note that this end to end framework allows for the generation of multiple structural parameter sets without losing precision or consistency.

Exceptional Precision in Metasurface Testing

The system underwent rigorous testing on square ring pillar metasurfaces, yielding results that redefine the benchmarks for high performance in the field of nanophotonics. Data indicates a prediction accuracy of 99.9%, while maintaining an exceptionally low mean absolute error. Furthermore, the technology demonstrated the ability to cover more than 100% of the sRGB color space, signaling a significant expansion in the available color gamut for digital and physical applications. Experimental validation included the successful fabrication of a 16 color palette and specific institutional logos, which confirmed the high fidelity between computational design and physical measurement.