Nonreciprocal surface plasmonic neural network for decoupled bidirectional analogue computing

Machine learning algorithms enable computers to learn autonomously by emulating human intelligence, but they require considerable electronic computational resources, increasing costs for big data processing. To address the burgeoning demand for computing capacity in artificial intelligence, researchers have explored optical neural networks that show advantages of ultrafast speed, low power consumption, ultra-high bandwidth, and high parallelism. However, such neural networks capable of mimicking the unidirectional behavior of biological neural networks remain largely unexplored. A significant challenge lies in achieving independent data processing in bidirectional paths. Here, we present a nonreciprocal deep neural network leveraging the magneto-optical effect in ferrites to decouple forward and backward paths, thus enabling independent control over weight matrices for multiplexed bidirectional microwave processing. Moreover, the computing function of the network can be flexibly modulated by the magnetization orientation in ferrites and variations in operating frequency. We demonstrate broadband bidirectional decoupled image processing across various operators, where the operator configuration can be precisely designed by encoding the input signals. Furthermore, matrix-solving operations can be facilitated by incorporating feedback waveguides for desired recursion paths. Our findings open pathways to nonreciprocal architectures for independent bidirectional algorithms in analogue computing.