Free-space and on-chip diffractive deep neural network for next-generation optical computing

Key Points

• Summarize the core principles, design methodologies, and recent advances in free-space and on-chip diffractive deep neural networks for next-generation optical processing.
• Reviewed foundational physics and design frameworks for multi-layered trainable diffractive neurons operating via light diffraction and interference.
• Surveyed deployment strategies across computational tasks including image classification, 3D perception, and edge computing.
• Diffractive neural architectures perform deep network operations passively at the speed of light by exploiting intrinsic optical parallelism.
• Optical setups achieve ultrafast inference speeds and virtually zero computational energy consumption during light propagation.