Diffractive deep neural network (DDNN) is an optical machine learning framework that combines deep learning with optical diffraction and light-matter interaction to create diffractive surfaces that perform optical computation at the speed of light. A diffractive neural network is first designed in a computer using deep learning techniques, then physically fabricated using 3-D printing or lithography techniques. Because the connection between a diffractive neural network’s input and output planes is established through light diffraction through passive layers, the inference process and associated optical computation consume no power other than the light used to illuminate the object of interest.
Diffractive optical networks provide a low power, low latency, and highly scalable machine learning platform with numerous applications in robotics, autonomous vehicles, and the defense industry. Diffractive neural networks have been used to design deterministic optical systems, such as a thin imaging system, and provide statistical inference and generalization to data classes.
Researchers created diffractive networks that can process data across various wavelengths, extending this all-optical computation framework into broadband optical signals. The researchers demonstrated the success of this new framework by developing a series of optical components that filter broadband input light into desired sub-bands. Deep learning-based diffractive systems also control the precise location of each filtered radiation band at the output plane, demonstrating spatially controlled wavelength de-multiplexing in the electromagnetic spectrum’s terahertz (THz) region.