Inverse design of dynamically tunable phase-change material based metamaterial absorber induced structural color

Structural colors using phase change material based metasurfaces/metamaterials have enabled high saturation, high resolution, and wide-gamut color printing. The wide color gamut with high resolution is achieved by changing the geometry of the metamaterials. However, finding the design parameter for desired color is computationally costly. Designing such parameters to precisely show the required color is therefore essential for producing metamaterials for use in real-world applications. In this work, we report a tunable reflective metamaterial using Ge2Sb2Te5 (GST) for the generation of structural colors. Exploiting the large contrast in refractive index of GST with the change in phase we can dynamically tune between two different colors. Further, we report an inverse design of tunable reflective metamaterial structure using deep neural network. The inverse design based on bidirectional artificial neural network can accurately predict the design parameter for a desired color. Our results show that tunable color filters using GST nanostructures can be designed accurately and efficiently using deep neural networks and could find potential applications in color printing and display technologies. •Dynamically tunable reflective metamaterial using Ge2Sb2Te5 for the generation of structural colors is investigated.•The colors produced can be tuned by varying the phase of GST and the geometrical parameters of the metamaterial.•A bidirectional neural network was developed to accurately predict the geometrical parameters for a desired color.•The forward model showed very high accuracy with ΔE2000 = 0.16 and the inverse design displayed ΔE2000 = 1.27.