Inverse design of dual-band photonic topological insulator beam splitters for efficient light transmission

Abstract The study of optical topological insulators (PTIs) has revealed intriguing optical properties that diversify the ways in which light can be manipulated, offering significant potential for a wide range of applications. This paper presents a machine learning (ML)-based approach for the reverse design of optical PTIs. Using finite element methods, the paper addresses the challenge of computing the band structure of a dual-band model, enabling the construction of a dataset suitable for ML training. With the goal of maximizing dual-band bandgaps, the study employs the random forest algorithm to predict target parameters and further designs topological edge states. Leveraging these boundary state patterns, two different optical PTI beam splitters are devised, and their transmission coefficients and losses are computed. The results demonstrate that optical devices designed using topological boundary states exhibit enhanced stability and robustness. This approach offers a reliable solution for applications in fields such as optical communication and optical sensing.