Fully UV Modulated Artificial Synapses with Integrated Sensing, Storage and Computation
Shuyi Sun, Tao Zhang, Shuoqi Jin, Xinhua Pan, Jianguo Lü, Zhizhen Ye, Bin Lu
Abstract
Abstract Optical synaptic devices are considered a preferred substitute to electrical synapses for low‐energy, high‐efficiency neuromorphic computing (NC). However, most optical synapses irreversible optical response severely restricts their application scenarios. In this study, a fully photon‐modulated synaptic device constructed from Si‐doped beta‐gallium oxide ( β ‐Ga 2 O 3 )/ZnO heterojunction is demonstrated. The device exhibits reversible conductance changes and mimics a range of synaptic behaviors, which are modulated by two different wavelengths (255 and 370 nm) of UV light. Owing to the device's plasticity, four types of logic functions including “AND”, “OR”, “NOR” and “NAND” are implemented in the device by optical stimulation. Furthermore, an artificial neural network (ANN) is simulated based on the device's excitatory and inhibitory characteristics, and an optical programming scheme is employed to enhance the network's performance, resulting in a high recognition rate of 95.5% for handwritten digits. The fully photon‐modulated synapse development strategy and programming scheme, as presented in this work, advance the utilization of optical artificial synapses in ANN.