Nonlinear germanium-silicon photodiode for activation and monitoring in photonic neuromorphic networks

Abstract

Silicon photonics is promising for artificial neural networks computing owing to its superior interconnect bandwidth, low energy consumption and scalable fabrication. However, the lack of silicon-integrated and monitorable optical neurons limits its revolution in large-scale artificial neural networks. Here, we highlight nonlinear germanium-silicon photodiodes to construct on-chip optical neurons and a self-monitored all-optical neural network. With specifically engineered optical-to-optical and optical-to-electrical responses, the proposed neuron merges the all-optical activation and non-intrusive monitoring functions in a compact footprint of 4.3 × 8 μm2. Experimentally, a scalable three-layer photonic neural network enables in situ training and learning in object classification and semantic segmentation tasks. The performance of this neuron implemented in a deep-scale neural network is further confirmed via handwriting recognition, achieving a high accuracy of 97.3%. We believe this work will enable future large-scale photonic intelligent processors with more functionalities but simplified architecture.

Advanced all-optical classification using orbital-angular-momentum-encoded diffractive networks

“…[2][3][4][5][6] Compared with traditional electronic hardware computing, optical computing offers several advantages, including ultrafast computing speed, 7,8 ultralow energy consumption, 9 and significant potential for parallel computing. 10,11 In recent years, with the rapid development of deep learning, 12 optical computing based on deep learning with different implementation schemes has been increasingly applied to various tasks, 13 such as vowel recognition, 9 image classification, 11,[14][15][16][17] mathematical operations, 7 and matrix operations. [18][19][20][21][22][23][24][25] A diffractive deep neural network (D 2 NN) is a series of successive diffractive layers designed in a computer using error backpropagation and stochastic gradient descent methods.…”

Ultra‐Compact and NonVolatile Nanophotonic Neural Networks

“…For example, Raoux et al achieved 10 11 cycles by switching PCM states [42] Zheng et al demonstrated that Ge 2 Sb 2 Te 5 can reversibly trigger over 1000 large-area phase transitions with almost zero additional loss [41] and for the new family PCM Sb 2 Se 3 , Delaney et al demonstrated a stable switching endurance of over 4000 cycles [43] Compared with ENNs, PNNs have broader bandwidth and faster speed, but still face challenges such as low-density integration. Very recently, Shi et al proposed new on-chip neurons using nonlinear germanium silicon photodiodes, whose nonlinear parts are concentrated in a compact size ≈4.3 × 8 μm 2 [46] Bai et al proposed an microcomb-driven chip-based photonic processing unit, and show a preeminent photonic-core compute density of over 1 trillion of operations per second per square millimeter (TOPS mm −2 ) [47] Furthermore, high-density photonic integrated circuits (PICs) are essential when considering tens of millions of possible photonic components in a monolithic chip. Digital nanophotonics is an effective method for realizing high-density PICs [48] and inverse-designed digital nanophotonics could have an ultra-small footprint, breaking the limitations of traditional periodic metamaterials and allowing arbitrary topological structures [49][50][51] Here, we propose a nanophotonic neural network (N-PNN) architecture that combines nonvolatile O-PCM (Sb 2 Se 3 ) and the inverse design of digital nanophotonics.…”

“…Very recently, Shi et al. proposed new on‐chip neurons using nonlinear germanium silicon photodiodes, whose nonlinear parts are concentrated in a compact size ≈4.3 × 8 µm 2 [ 46 ] Bai et al. proposed an microcomb‐driven chip‐based photonic processing unit, and show a preeminent photonic‐core compute density of over 1 trillion of operations per second per square millimeter (TOPS mm −2 ) [ 47 ] Furthermore, high‐density photonic integrated circuits (PICs) are essential when considering tens of millions of possible photonic components in a monolithic chip.…”

Adaptive Convolutional Neural Networks for Enhanced Memory Retention and Restoration in Optoelectronic Vision Devices

“…[15,[23][24][25] The optoelectronic device responds differently to the different wavelengths of light depending on the properties of the light-responsive semiconductor used. [25][26][27] As such, the devices rely on capturing the light at particular wavelengths which results in the formation of electron-hole pairs. In certain materials, a phenomenon of persistent photoconductivity (PPC) can also be exploited for mimicking synaptic plasticity and implementing a neural network for processing of information for tasks such as pattern/image storage, recognition, and classification.…”