In a significant advancement for the field of optical fiber communication, a groundbreaking study has been published in Engineering. Titled “A Neuro Metasurface Mode-Router for Fiber Mode Demultiplexing and Communications”, the research presents a neuro-meta-router (NMR) that has the potential to revolutionize the way we handle data transmission in fiber-optic systems.

As emerging technologies such as AI, cloud computing, and the Internet of Things continue to drive an unprecedented demand for information-transmission capacity, optical communication has faced the challenge of approaching the Shannon limit. Space-division multiplexing techniques, especially those using few-mode fibers (FMFs), have been explored to boost communication capacity. However, developing ideal mode multiplexers and demultiplexers with outstanding performance, scalability, and compatibility remains a hurdle.

The researchers, led by Guoxing Zheng and a team of experts from various institutions, propose an innovative solution by integrating metasurfaces and deep-learning techniques into a fiber communication system. The NMR they designed contains a single metasurface and no compensation technologies or correlation devices, yet it enables dual-mode division and dual-channel mode-division multiplexing (MDM) communication.

The working mechanism of the NMR is quite ingenious. Fiber modes act as Internet Protocol (IP) addresses, with each mode carrying a set of information. The neural network’s algorithmic framework aligns with the routing processor, and the metasurface’s modulation capability corresponds to the routing protocol. This setup allows for the organized and efficient transmission of multiple sets of information in independent channels.

To optimize the phase distribution of the metasurface, the team employed a neural network. They considered various spatial transformations of the mode fields, such as position translation, angular rotation, and phase divergence, to enhance the NMR’s anti-jamming and alignment tolerance capabilities. The training process used a stochastic gradient descent algorithm and carefully designed loss functions to reduce background noise and inter-mode crosstalk.

The experimental validation of the NMR was comprehensive. The fabricated metasurface successfully achieved spatial division among the two modes of the FMF with accurate division, minimal crosstalk, and a certain degree of robustness. An NMR-based optical communication system was also built, which demonstrated a remarkable transmission capacity of up to 100 gigabits per second (Gbps) for dual channels and a symbol error rate down to the order of 10⁻⁴. The system was even able to simultaneously and independently transmit two sets of diverse images, further proving its practicality and flexibility.

This research not only offers a promising solution for high-speed, high-capacity fiber communication but also paves the way for future developments. The scalability of the NMR for a larger number of modes is theoretically achievable, and there is potential to incorporate polarization multiplexing and wavelength division multiplexing schemes. Overall, the neuro-meta-router represents a significant step forward in the quest for more efficient and advanced optical communication technologies.

The paper “A Neuro Metasurface Mode-Router for Fiber Mode Demultiplexing and Communications,” authored by Yu Zhao, Huijiao Wang, Zile Li, Tian Huang, Chao Yang, Ying Qiu, Yuhan Gong, Zhou Zhou, Congling Liang, Lei Yu, Jin Tao, Shaohua Yu, Guoxing Zheng. Full text of the open access paper: https://doi.org/10.1016/j.eng.2024.11.012. For more information about the Engineering, follow us on X (https://twitter.com/EngineeringJrnl) & like us on Facebook (https://www.facebook.com/EngineeringJrnl).