Programmable chalcogenide-based all-optical deep neural networks

Abstract

We demonstrate a passive all-chalcogenide all-optical perceptron scheme. The network’s nonlinear activation function (NLAF) relies on the nonlinear response of Ge2Sb2Te5 to femtosecond laser pulses. We measured the sub-picosecond time-resolved optical constants of Ge2Sb2Te5 at a wavelength of 1500 nm and used them to design a high-speed Ge2Sb2Te5-tuned microring resonator all-optical NLAF. The NLAF had a sigmoidal response when subjected to different laser fluence excitation and had a dynamic range of −9.7 dB. The perceptron’s waveguide material was AlN because it allowed efficient heat dissipation during laser switching. A two-temperature analysis revealed that the operating speed of the NLAF is ≤ 1 $\le 1$ ns. The percepton’s nonvolatile weights were set using low-loss Sb2S3-tuned Mach Zehnder interferometers (MZIs). A three-layer deep neural network model was used to test the feasibility of the network scheme and a maximum training accuracy of 94.5% was obtained. We conclude that combining Sb2S3-programmed MZI weights with the nonlinear response of Ge2Sb2Te5 to femtosecond pulses is sufficient to perform energy-efficient all-optical neural classifications at rates greater than 1 GHz.

Optoelectronic integrated circuits for analog optical computing: Development and challenge

“…Due to the flexibility and simple implementing, the OEO method was widely applied [37,[140][141][142][143][144]. However, the requirement for the computing speed and power consumption has promoted alloptical nonlinear activation functions in recent years [66, 139,[145][146][147][148][149][150]. Here, we summarized the implementations of the OEO and all-optical nonlinear activation functions in recent years, and the characteristics of each implementation are depicted in Table 1, including the physical mechanism, the type of nonlinearity function, power consumption/activation threshold, reconfigurability, and integratability.…”

“…Frontiers in Physics frontiersin.org in a silicon ring resonator, the optical resonance condition of the ring was controlled by the phase state of PCMs and a ReLU-like nonlinear activation function was obtained with a contrast ratio of 9 dB. Afterwards, Teo et al demonstrated a passive allchalcogenide ONN scheme consisting of Sb 2 S 3 -programmed MZI weights and the nonlinear response of Ge 2 Sb 2 Te 5[139].…”

Deep Learning Image Analysis of Nanoplasmonic Sensors: Toward Medical Breath Monitoring

“…To thoroughly study the correlations between the sensor responses (RGB values) with various concentrations of acetone and other interfering analytes, a deep learning neural network model named SensingNet was developed in this work to intelligently analyze the change of RGB values in each sensor image and accurately predict the NPP sensing results. Deep learning neural networks have been developed and applied to address challenges in many areas. − …”

Characterizing optical phase-change materials with spectroscopic ellipsometry and polarimetry

“…Similarly, time resolved optical methods have been applied to study both phenomena, the ones happening above (single shot approach) and the ones below the fluence threshold in the reversible state of the system. Single-shot transmission, reflection, and imaging spectroscopy has been used to study irreversible changes occurring during the phase transitions of Ge 2 Sb 2 Te 5 [44,[50][51][52][53] and Sb 2 Se 3 [54]. Meanwhile, in the reversible state, time-resolved optical methods such as time-resolved absorption and reflection has been used to probe the loss of resonance in the bonded states of Ge 2 Sb 2 Te 5 prior to the phase change transition [44] and to study [55] the creation and annihilation of light induced defects in GeSe 2 thin films.…”

“…5 shows a scheme of this technique and recent results of its application to the observation of transient effects in crystalline PCM thin film. A wide range of events happens during light-induced phase transitions of PCM between time scales from 0.1 -100 ps [42][43][44][45][46][47][48][49][50][51][52][53][54] and time-resolved ellipsometry in complement with ultrafast structural methods are ideal tools for probing these dynamics.…”