The Photoelectric Synaptic Device with Sensing‐Memory‐Computing Function Regulated by All‐Optical Pulse
Yongle Zhang, Qianwen Guo, Yingfeng Duan, Feng Yang, Feng Xue, Mingli Zheng, Junmeng Guo, Gang Cheng, Zuliang Du
Abstract
Abstract The photoelectric synaptic devices with the bidirectional optical response and multilevel nonvolatile all‐optical pulse control are the basis for constructing the sensing‐memory‐computing hardware network. However, the current all‐optical‐pulse‐regulated photoelectric synaptic devices cannot realize the bidirectional and multistage nonvolatile regulation of photocurrent responsivity under the detection light excitation. Here, p‐Si/n‐ZnO heterojunction synaptic devices are constructed, and the bidirectional, nonvolatile control of the synaptic device under visible light excitation is realized with up to 60 states of photocurrent responsivity by ultraviolet (UV) and near‐infrared (NIR) irradiation. Based on the performance of the synaptic device, the artificial feedback network can be built to realize the classification and recognition of colored handwritten digits. The recognition accuracy for optical handwritten digits at 415, 530, and 970 nm is 78.1 ± 1.2%, 79.4 ± 1.9% and 79.0 ± 1.6%, respectively. Additionally, the 3 × 3 convolution kernel constructed by the synaptic device can realize image‐processing functions like mean filtering, Gaussian filtering, and edge enhancement. The development of sensing‐storage‐computing all‐optical control artificial vision hardware networks can be greatly aided by the new approach to bidirectional and multistage nonvolatile all‐optical pulse regulation of photoelectric synaptic devices that is proposed in this paper.