神经形态工程学
记忆电阻器
计算机科学
物理不可克隆功能
加密
人工神经网络
材料科学
调制(音乐)
云计算
突触重量
计算机体系结构
电子工程
电阻随机存取存储器
异质结
嵌入式系统
晶体管
突触后电流
电压
保险丝(电气)
钥匙(锁)
压阻效应
尖峰神经网络
计算机硬件
深度学习
人工智能
物联网
锌黄锡矿
电阻式触摸屏
非易失性存储器
突触
作者
Fengxia Yang,Hao Sun,Xiang Zhang,Feng Sun,Xiaofei Dong,Jianbiao Chen,Xuqiang Zhang,Jiangtao Chen,Yun Zhao,Yan Li
摘要
High‐performance optoelectronic co‐modulated synaptic devices hold pivotal significance in advancing the practicalization of neuromorphic computing, yet challenges remain regarding their capacity to perform high‐level tasks in hardware neural networks. Herein, an optoelectronic synaptic memristor based on a kesterite heterostructure is innovatively proposed. The device enables precise regulation of carrier transport and resistive switching behaviors under dual light and electrical stimuli, achieving a high switching ratio of up to 10 4 , and allowing dynamic modulation of the switching window (2 × 10 2 ∼3 × 10 4 ) and switching voltage (0.16 ∼0.91 V) in a non‐contact manner. Furthermore, it faithfully emulates diverse bio‐synaptic functions, including excitatory postsynaptic current (EPSC), spike‐rate‐dependent plasticity (SRDP), multiparametric modulation of long‐term potentiation/depression (LTP/LTD), and brain‐like neuromorphic learning behaviors, manifesting superior bionic performance. Notably, this device supports two high‐level hardware applications: (1) a multi‐level authorized encryption computing framework relying on light/electrically induced physical unclonable functions (PUF) in memristors, suitable for secure cloud communications; (2) serving as neural network activation units for high‐performance complex pattern classification, outperforming its traditional software‐implemented counterparts. The results clarify the core potential of heterostructures in developing advanced multimodal synaptic memristors and provide a viable strategy toward next‐generation optoelectronic‐integrated neuromorphic systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI