记忆电阻器
材料科学
神经形态工程学
钙钛矿(结构)
双层
电极
MNIST数据库
光电子学
纳米技术
电子工程
计算机科学
人工神经网络
化学工程
人工智能
物理化学
工程类
生物
遗传学
化学
膜
作者
Jiuchao Feng,Yanwei Fan,Yue Wang,Qing Song,Yang Liu,Yonghua Chen,Deli Li,Wei Huang
标识
DOI:10.1002/adfm.202420547
摘要
Abstract Perovskite memristors hold great promise for neuromorphic computing due to their ease of fabrication and sensitivity to light and electrical stimuli. However, the common use of expensive metal electrodes, such as gold (Au) and unstable silver (Ag), limits their stability and broader application. In this study, a cost‐effective perovskite memristor utilizing a novel Ag/Bismuth (Ag/Bi) bilayer electrode, which serves as a viable alternative to Au while maintaining excellent performance, is presented. This design prevents electrochemical reactions and the formation of unstable metallic filaments, enabling controlled mixed electronic/ionic conductivity. Moreover, the low work function of the Ag/Bi bilayer reduces the built‐in voltage, facilitating the formation and retention of conductive filaments, which improves device performance and stability. The memristor exhibits a high on/off ratio (10 2 ), excellent endurance (≈800 cycles), long retention (>10 4 s), and storage stability comparable to Au‐based devices. Furthermore, it demonstrates neuromorphic synaptic behaviors, including long‐ and short‐term plasticity (STP), potentiation/depression, and spike‐timing‐dependent plasticity (STDP). When integrated into a spiking neural network (SNN) for digital image recognition using the MNIST dataset, the device achieves an accuracy of 86.68%. This work demonstrates the potential of the cost‐effective Ag/Bi bilayer electrode in enhancing the stability and performance of perovskite memristors.
科研通智能强力驱动
Strongly Powered by AbleSci AI