Optimization of Broadband Solar Metamaterial Absorber Based on Deep Neural Network

We propose a broadband solar metamaterial absorber based on a Ti-SiO₂-Ti-SiO₂-Fe₂O₃ five-layer configuration. The optimized parameters of the configuration are obtained from the deep neural network with Keras in Python. The average absorptance of the absorber can achieve 98.70% in the wavelength range from 170 nm to 900 nm under the normal incidence. The absorber is not sensitive to the incident angle of the polarized light. When the incident angle of light is not greater than 40°, the average absorptance of the absorber can reach over 90% for both x- and y- polarized lights. In addition, different polarization angles in the range of 0–90° have almost no effect on the absorptance of the absorber. The proposed absorber has potential applications in many fields, such as solar energy absorption, ultraviolet protection and near-infrared detection.