Inverse design of anisotropic and multi-resonant absorbers based on black phosphorus via residual neural network

Black phosphorus (BP), a new type of two-dimensional material, has attracted extensive attention because of its excellent properties. The anisotropy of BP makes its physical properties vary greatly in different directions, which increases the complexity of the design of BP metamaterials. We present a residual neural network on the basis of the improved adaptive batch normalization algorithm to achieve the inverse design of a multilayer thin film structure based on BP, and we adopt the characteristic matrix method to obtain perfect optical absorption samples. The prediction accuracy of the neural network model is more than 95% for absorbing structures with both single and multiple resonances. This method has the advantages of a fast rate of convergence and high precision of prediction and achieves the design target on the basis of the structure of a BP metamaterial.