Ultra-broadband and polarization-insensitive metasurface absorber with behavior prediction using machine learning

The solar spectrum energy absorption is very important for designing any solar absorber. The need for absorbing visible, infrared, and ultraviolet regions is increasing as most of the absorbers absorb visible regions. We propose a metasurface solar absorber based on Ge2Sb2Te5 (GST) substrate which increases the absorption in visible, infrared and ultraviolet regions. GST is a phase-changing material having two different phases amorphous (aGST) and crystalline (cGST). The absorber is also analyzed using machine learning algorithm to predict the absorption values for different wavelengths. The solar absorber is showing an ultra-broadband response covering a 0.2–1.5 µm wavelength. The absorption analysis for ultra-violet, visible, and near-infrared regions for aGST and cGST is presented. The absorption of aGST design is better compared to cGST design. Furthermore, the design is showing polarization insensitiveness. Experiments are performed to check the K-Nearest Neighbors (KNN)-Regressor model’s prediction efficiency for predicting missing/intermediate wavelengths values of absorption. Different values of K and test scenarios; C-30, C-50 are used to evaluate regressor models using adjusted R2 Score as an evaluation metric. It is detected from the experimental results that, high prediction proficiency (more than 0.9 adjusted R2 score) can be accomplished using a lower value of K in KNN-Regressor model. The design results are optimized for geometrical parameters like substrate thickness, metasurface thickness, and ground plane thickness. The proposed metasurface solar absorber is absorbing ultraviolet, visible, and near-infrared regions which will be used in solar thermal energy applications.