Abstract:On-chip photonic processors for neural networks have potential benefits in both speed and energy efficiency but have not yet reached the scale at which they can outperform electronic processors. The dominant paradigm for designing on-chip photonics is to make networks of relatively bulky discrete components connected by one-dimensional waveguides. A far more compact alternative is to avoid explicitly defining any components and instead sculpt the continuous substrate of the photonic processor to directly perform the computation using waves freely propagating in two dimensions. We propose and demonstrate a device whose refractive index as a function of space, $n(x,z)$, can be rapidly reprogrammed, allowing arbitrary control over the wave propagation in the device. Our device, a 2D-programmable waveguide, combines photoconductive gain with the electro-optic effect to achieve massively parallel modulation of the refractive index of a slab waveguide, with an index modulation depth of $10^{-3}$ and approximately $10^4$ programmable degrees of freedom. We used a prototype device with a functional area of $12\,\text{mm}^2$ to perform neural-network inference with up to 49-dimensional input vectors in a single pass, achieving 96% accuracy on vowel classification and 86% accuracy on $7 \times 7$-pixel MNIST handwritten-digit classification. This is a scale beyond that of previous photonic chips relying on discrete components, illustrating the benefit of the continuous-waves paradigm. In principle, with large enough chip area, the reprogrammability of the device's refractive index distribution enables the reconfigurable realization of any passive, linear photonic circuit or device. This promises the development of more compact and versatile photonic systems for a wide range of applications, including optical processing, smart sensing, spectroscopy, and optical communications.

Easily Scalable Photonic Tensor Core Based on Tunable Units with Single Internal Phase Shifters

Clements-Enhanced Complex-Valued Coherent Mesh with Balanced Detection Units for Photonic Neural Networks

A Reconfigurable Photonic Computing Chip with Easily Scalable Topology

Digit Recognition Based on Programmable Nanophotonic Processor

Optical mode division multiplexing inspired photonic neural network accelerator

Photonic Neural Networks and Its Applications

Integrated WDM-compatible Optical Mode Division Multiplexing Neural Network Accelerator

Ultra-high Density and Nonvolatile Nanophotonic Convolution Processing Unit

Photonic Neural Network Fabricated on Thin Film Lithium Niobate for High-Fidelity and Power-Efficient Matrix Computation

Scaling on-chip photonic neural processors using arbitrarily programmable wave propagation

Electrically Programmable Phase-Change Photonic Memory for Optical Neural Networks with Nanoseconds in Situ Training Capability

Computing Dimension for a Reconfigurable Photonic Tensor Processing Core Based on Silicon Photonics

Programmable low-threshold optical nonlinear activation functions for photonic neural networks

Multicore Photonic Complex-Valued Neural Network with Transformation Layer

120 GOPS Photonic Tensor Core in Thin-film Lithium Niobate for Inference and in-situ Training

Ultra‐Compact and NonVolatile Nanophotonic Neural Networks

Parallel Photonic Acceleration Processor for Matrix-Matrix Multiplication

Control-free and efficient integrated photonic neural networks via hardware-aware training and pruning

Photonic tensor processing unit with single dataflow and programmable high-precision weighting control

Integrated multi-operand optical neurons for scalable and hardware-efficient deep learning

On-chip Spiking Neural Networks Based on Add-Drop Ring Microresonators and Electrically Reconfigurable Phase-Change Material Photonic Switches