Intelligent self-configuration optical cross-connect of heterogeneous orbital angular momentum fibers

Optical communication technology has become indispensable in modern communication systems due to its high bandwidth, low latency, and long-distance capabilities. With the exponential increase in data demand, advancements in various types of optical fibers have played a crucial role in enhancing both transmission efficiency and network flexibility. Yet, managing heterogeneous fiber connections, particularly between different fiber structures, remains a challenge. This paper presents the design and implementation of an intelligent optical cross-connect system based on a silicon photonic neural network chip. By integrating first-order and second-order orbital angular momentum (OAM) modes generated by (de)multiplexers and dynamically managing them through neural networks, the system facilitates real-time mode conversion and cross-connect between heterogeneous multi-mode and ring-core fibers. The experimental results validate its performance, demonstrating effective mode management and low crosstalk, thus indicating potential applications in data centers and metropolitan networks. This system has broad prospects, enabling scalable interconnection between heterogeneous fiber sub-networks, thereby improving network flexibility and efficiency.