神经形态工程学
材料科学
电导
堆栈(抽象数据类型)
电极
光电子学
可扩展性
铁电性
MNIST数据库
纳米技术
纳米尺度
非易失性存储器
纳米
限制
隧道枢纽
制作
电压
记忆电阻器
纳米电子学
突触重量
Spike(软件开发)
计算机科学
电容
快速切换
领域(数学)
电子工程
作者
Seungjoon Jeong,Huiseong Shin,Myeongjae Choi,Changwoo Han,Hyeonjung Park,Junseok Kim,Y.J. Choi,Changhwan Shin
标识
DOI:10.1021/acsami.6c00989
摘要
Ferroelectric tunnel junctions (FTJs) are promising synaptic devices for neuromorphic computing owing to their compact two-terminal structure and ability to support multilevel conductance modulation. However, many FTJ synapses rely on complex multilayer stacks or additional insertion layers to stabilize interfacial transport, increasing device complexity and limiting scalability. Here, we demonstrate a structurally simple metal-ferroelectric-metal (MFM) FTJ based on an ITO/Hf0.5Zr0.5O2/WOx stack in which the interface and electrode properties are deliberately engineered to achieve stable switching without additional layers. By tuning the oxygen stoichiometry of the WOx bottom electrode, a controlled trap-rich interfacial region is formed that enables polarization-modulated trap-assisted tunneling. In addition, the use of an ITO top electrode redistributes the electric field across the junction, improving programming reliability and breakdown tolerance. As a result, the optimized FTJ exhibits a resistance ratio of ∼100, switching endurance exceeding 108 cycles, and 64 well-resolved conductance states. The device further demonstrates stable spike-dependent plasticity and reliable analog weight modulation suitable for neuromorphic operation. Neural-network simulations based on experimentally extracted conductance characteristics achieve 91.5% accuracy on the MNIST data set, highlighting the potential of simple MFM FTJ synapses for scalable neuromorphic and in-memory computing hardware.
科研通智能强力驱动
Strongly Powered by AbleSci AI