Design of high group velocity dispersion optical fiber assisted by cascaded neural networks

The growing demand for high-speed optical communication requires advanced dispersion management to suppress pulse broadening and nonlinear effects. This work proposes a dual-core optical fiber with ultrahigh group velocity dispersion, optimized using cascaded deep neural networks. By leveraging the anti-crossing between the LP 01 mode in the central core and the LP 02 mode in a side core, controlled mode coupling enables dispersion enhancement. Finite element method simulations show that a dual-core fiber design derived from the standard G.652 fiber achieves an ultrahigh chromatic dispersion of approximately 11,000ps/(nm⋅km) at the 1550 nm wavelength window, with tunable bandwidth via core spacing adjustment. This methodology establishes a paradigm for AI-driven photonic device design, offering significant potential for advancing optical communication and ultrafast laser technologies.