High-speed all-optical neural networks empowered spatiotemporal mode multiplexing

Area of Science:

• Optics and Photonics
• Optical Communications
• Machine Learning Applications

Background:

• Orbital angular momentum (OAM) beams offer high-capacity optical communication via mode-division multiplexing (MDM).
• Traditional spatial-only OAM systems face complexity, crosstalk, and scalability issues.
• Hybrid OAM multiplexing with wavelength/polarization shows promise but is limited by continuous-wave lasers.

Purpose of the Study:

• To develop a novel OAM-based spatiotemporal multiplexing (OAM-STM) technique.
• To leverage pulsed laser sources for enhanced optical communication robustness and speed.
• To overcome the repetition-rate bottleneck in OAM systems.

Main Methods:

• Synergistic implementation of pulsed OAM beams with a diffractive deep neural network (D2NN).
• Utilizing optical fiber delay lines to project spatial mode information into the temporal domain.
• Experimental demonstration of an OAM-based spatiotemporal demultiplexer.

Main Results:

• Successful projection of spatial OAM information into the temporal domain.
• Achieved demultiplexing speed limited by photodiode bandwidth, enabling high-speed operation.
• Demonstrated system scalability to GHz rates, compatible with high-repetition-rate OAM sources.

Conclusions:

• Established a foundational framework for high-speed, all-optical, high-capacity OAM-STM systems.
• Pulsed OAM-STM overcomes limitations of traditional OAM systems by activating the time dimension.
• Promising implications for free-space, underwater, and other complex optical communication environments.