Topological band optimization of nested photonic crystals based on deep learning

Photonic crystals are favored in the field of topological photonics due to their rich structural characteristics and topological properties. However, the accumulation of geometric errors during their manufacturing process often poses challenges to device performance. In this paper, a deep learning (DL) convolutional neural network is introduced to efficiently predict and correct manufacturing errors. The final corrected results by the model are highly consistent with the band structure results from optical simulations of the target design, verifying the effectiveness of DL algorithms in manufacturing optimization and optical performance prediction. This work is expected to provide an intelligent solution for high-precision manufacturing and high-performance design of topological photonic crystals, further advancing the interdisciplinary research between topological photonics and DL.