An ITO–graphene heterojunction integrated absorption modulator on Si-photonics for neuromorphic nonlinear activation

Abstract

The high demand for machine intelligence of doubling every three months is driving novel hardware solutions beyond charging of electrical wires, given a resurrection to application specific integrated circuit (ASIC)-based accelerators. These innovations include photonic-based ASICs (P-ASICs) due to prospects of performing optical linear (and also nonlinear) operations, such as multiply–accumulate for vector matrix multiplications or convolutions, without iterative architectures. Such photonic linear algebra enables picosecond delay when photonic integrated circuits are utilized via “on-the-fly” mathematics. However, the neuron’s full function includes providing a nonlinear activation function, known as thresholding, to enable decision making on inferred data. Many P-ASIC solutions perform this nonlinearity in the electronic domain, which brings challenges in terms of data throughput and delay, thus breaking the optical link and introducing increased system complexity via domain crossings. This work follows the notion of utilizing enhanced light–matter interactions to provide efficient, compact, and engineerable electro-optic neuron nonlinearity. Here, we introduce and demonstrate a novel electro-optic device to engineer the shape of this optical nonlinearity to resemble a leaky rectifying linear unit—the most commonly used nonlinear activation function in neural networks. We combine the counter-directional transfer functions from heterostructures made out of two electro-optic materials to design a diode-like nonlinear response of the device. Integrating this nonlinearity into a photonic neural network, we show how the electrostatics of this thresholder’s gating junction improves machine learning inference accuracy and the energy efficiency of the neural network.

Free-space optical multiplexed orbital angular momentum beam identification system using Fourier optical convolutional layer based on 4f system

“…Both Fourier transform steps are performed entirely passively, overwhelming the electrical time and power consuming convolution. Whats more, in this project, we managed to load two kernels , where achieve parallel convolution via nature diffraction by DMD, [117][118][119][120][121][122][123][124] this property increase the computing throughput to a much higher level. After completion of the Fourier optical convolution and the domain crossing, the signal is electronically post-processed by two fully connected layers with the rectified linear unit (ReLU) activation function to conclude the classification.…”

Graphene on Silicon Photonics: Light Modulation and Detection for Cutting-Edge Communication Technologies

“…Most recently, a high-performance modulator utilizing a 2D-3D dielectric integration in a high-quality encapsulated graphene device was used to exhibit a ~39 GHz bandwidth, resulting in a three-fold increase in modulation efficiency when compared to previously reported high-speed modulators [20]. Meanwhile, the electro absorption graphene modulator could also be an ideal candidate as a photonic neuron thresholder, as reported by Volker's group in 2021 [51]. Besides the intensity modulator, the phase modulator based on graphene with a modulation efficiency of 0.28 V•cm and an operation bandwidth of 5 GHz has also been demonstrated [52].…”

Photonic tensor core for machine learning: a review

“…To improve this aspect, and by so reducing, even more, energy consumption, one possibility is to switch to ITO-based modulators, that can achieve VπL as low as 95Vµm, [41][42][43][44] or ITO-graphene devices, that can work with a bandwidth of over 130GHz. 45 The strategy of mixing multiple materials to achieve on-chip activation function was discussed by Miscuglio et al, 46 where a road-map identify the major challanges were proposed.…”