Topological insulator-based wideband infrared absorber: Optimization and behaviour prediction using machine learning models

Abstract

A numerical investigation of the multi-layered perfect absorber structure for the wide band infrared spectrum is analyzed using the wide wavelength spectrum. The proposed structure is formed with the topological insulator, metal, and dioxide layers and the different resonator arrays. This structure is investigated over the wavelength spectrum of 0.2 μm–1.6 μm of the spectrum. Different physical parameters have been identified to achieve the optimized results of the wideband absorption spectrum. The proposed absorber structure is investigated over the different incident conditions (TE/TM) modes with wide-angle values ranging from 0⁰ to 80⁰. The proposed structure is polarisation non-sensitive up to 70⁰ of the incident angle. The structure is also investigated for the different resonator arrays consisting of circular, squared and rectangular resonators. The effect of the different unit cells was also examined to identify the optimized size of the wideband absorber structure. The behaviour of the overall structure is predicated by the different machine learning models such as kNN, Random Forest, Neural Network, Gradient Boosting and AdaBoost. The machine learning parameters such as MSE, RMSE, MAE, MAPE, and R2 are compared to identify suitable models for accurately predicting the absorption spectrum. The R² value is the minimum on the Neural network of up to 0.83 and the maximum of up to 0.96 in all four other models. The scatter plots and these values can help identify the suitable model that generates the prediction of the absorption spectrum more accurately. The proposed results in this manuscript can be helpful for various sensing and quantum dot-based photonics applications.