On-Chip Integrated Ultra-Compact Microscale Optical Logic Operations Based on Diffractive Neural Networks

Optical logic operations are considered as important components of optical computing, overcoming the inherent limitations of traditional electronic systems in transmission bandwidth and power, thus enabling applications in high-speed signal processing, parallel computing, and all-optical communication systems. The traditional optical logic gates implemented by methods such as semiconductor optical amplifiers, highly nonlinear optical fibers and micronano waveguides suffer from instability, difficulty in miniaturization, and the precise control of the input optical signals. Recently, diffractive neural networks have emerged as a new framework for implementing optical logic operations because of their high parallelism, low energy consumption and antinoise ability. In this study, we demonstrate three-dimensional (3D) microscale optical logic operation structures by the two-photon polymerization printed diffractive neural network. Specifically, the diffractive neural network featuring a volume size of 100 × 100 × 50 μm3 and neural size of 2 μm can execute the seven optical logic operations at the visible wavelengths with an accuracy of 100%. Furthermore, the logic operation neural network can be readily printed on commercially available CMOS chips, enabling ultracompact and miniaturized integrated devices. This study provides a feasible path for scaling optical logic components into practical optical computing systems by leveraging the existing CMOS-compatible platform.