Artificial Synaptic Performance with Learning Behavior for Memristor Fabricated with Stacked Solution-Processed Switching Layers
Zongjie Shen, Chun Zhao, Chun Zhao, Tianshi Zhao, Wangying Xu, Yina Liu, Yanfei Qi, Ivona Z. Mitrović, Li Yang, Ce Zhou Zhao, Ce Zhou Zhao
Abstract
As one of the promising next-generation electronics, brain-inspired synaptic resistive random access memory (RRAM) devices with stacked solution-processed (SP) spin-coated resistive switching (RS) layers were fabricated in this work. Compared with the RRAM device with a single SP-RS layer (Ag/SP-AlOx/ITO), the device with stacked SP-RS layers (Ag/SP-GaOx/SP-AlOx/ITO) is induced by the metal conductive filament performed with lower power consumption (∼±0.6 V operation voltage), larger read and write capability (∼2 × 104 ON/OFF ratio), and enhanced stability (>2 × 104 s retention time and >1000 endurance cycles). Multiple conductance states with long-term potentiation and depression (200 pulses) were obtained on Ag/SP-GaOx/SP-AlOx/ITO RRAM devices, which resulted in the human brain-like behavior (learning–forgetting–relearning) of a matrix comprising of RRAM devices with SP-GaOx/SP-AlOx layers. Based on the synaptic performance of Ag/SP-GaOx/SP-AlOx/ITO RRAM devices, an image recognition process based on a neuron network was conducted and the average recognition accuracy was close to 90%.