In high-speed optical communications, traditional orbital angular momentum (OAM) multiplexing systems face fundamental limitations, including exponentially increasing spatial-domain complexity, aggravated modal crosstalk, and strong dependence on continuous-wave lasers. These challenges hinder scalability and robustness in complex environments.
To address this, a research team led by Professor Fu Feng and Professor Xiaocong Yuan from Zhejiang Lab has developed a novel OAM-based spatiotemporal multiplexing (OAM-STM) architecture. This approach couples pulsed OAM beams with a diffractive deep neural network (D2NN) and optical fiber delay-line arrays, establishing a "space encoding–time decoding" transmission link. In this design, pulsed OAM states are spatially separated into distinct "activation regions" by the D2NN and then mapped into the time domain via fiber delay lines before being detected by a single-pixel photodetector.
The experimental demonstration achieved 3-bit data transmission using ultrafast pulsed lasers with 10 ps pulse width. A digital micromirror device (DMD) generated OAM beams carrying binary data patterns ("001" to "111") with topological charge l ∈ [1, 3]. After D2NN modulation, the beams were focused into three activation regions and transmitted through fiber delays (2 m, 4 m, 6 m), producing distinct temporal pulse sequences with a 9.48 ns delay difference. By applying an intensity threshold of 0.6, the system accurately decoded the 3-bit data.
Although the experimental system operated at kHz rates due to DMD switching speed (10,752 Hz), the OAM-STM architecture is inherently compatible with high-repetition-rate OAM sources. In principle, its demultiplexing speed is limited only by photodiode bandwidth, enabling scalability to the GHz regime. Increasing D2NN layer count or neuron density can further enhance bit capacity.
This breakthrough overcomes the low time-domain utilization and high demultiplexing complexity of conventional OAM communications. By integrating a temporal multiplexing dimension into an all-optical framework, each laser pulse can carry multiple times more data, alleviating the throughput bottleneck caused by pulse repetition rate limits. Furthermore, all-optical decoding avoids the latency and losses of electronic signal processing, enabling more compact and efficient high-speed systems.
Looking forward, the researchers envision upgrades in three directions: (1) adopting high-repetition-rate pulsed OAM generators and lasers to directly achieve GHz-level transmission rates; (2) developing multilayer D2NN and on-chip integrated delay lines (e.g., high-index spiral waveguides) for miniaturized devices suitable for 5G all-optical networks; and (3) extending the technique to more challenging environments such as long-distance free-space links, underwater optical wireless channels, and quantum communications.
This work provides a new paradigm for spatiotemporal multiplexing in optical communications, fostering deeper integration between all-optical neural networks and OAM technologies, and paving the way toward next-generation high-capacity, high-adaptability optical information systems.
