神经形态工程学
材料科学
晶体管
光探测
计算机科学
光电子学
无定形固体
卷积神经网络
宽带
逻辑门
纳米技术
纳米线
节点(物理)
冯·诺依曼建筑
人工神经网络
卷积(计算机科学)
计算机体系结构
柔性电子器件
活动层
非晶硅
纳米孔
光电二极管
电子工程
电介质
图层(电子)
和大门
集成电路
紫外线
电子线路
栅极电介质
光子学
记忆电阻器
作者
Jinkui Si,Qiong Zhuang,Rengjian Yu,Junling Zhou,Lan Yang,Changjie Zhou,Huili Zhu,Minshuai Wang,Xiaomei Cai,Qiubao Lin,Wangying Xu,Shuqiong Lan,Huipeng Chen
摘要
Multifunctional neuromorphic hardware has emerged as a promising paradigm to transcend the physical separation of sensing, memory, and computing inherent in traditional von Neumann architectures. The central challenge lies in developing integrated devices with high functional density for in‐situ fusion of sensing, memory, and computing. Here, we propose a multifunctional vertical bulk‐heterojunction transistor (MVBT) based on amorphous GaO x dielectric layer, integrating four functions within a single device: deep‐ultraviolet (DUV) photodetection, threshold neuronal firing, broadband optoelectronic synaptic function, and reconfigurable logic gates. In the two‐terminal (gate‐source) mode, the ultraviolet‐selective photoresponse and electrochemical reactions of amorphous gallium oxide (a‐GaO x ) enable deep‐ultraviolet photodetection and neuronal firing, respectively. In the three‐terminal mode, the organic bulk‐heterojunction active layer supports broadband optoelectronic synaptic function from ultraviolet to near‐infrared wavelengths. Asymmetric interfacial carrier‐injection barriers further enable bias‐reconfigurable AND/OR logic functions. In addition, a pixelated MVBT array demonstrates dual‐mode image sensing and visual preprocessing. Building upon these multifunctional capabilities, an MVBT‐based convolutional spiking neural network (CSNN) hybrid simulation architecture achieves a fingerprint recognition accuracy of 97.4%. This work provides a new strategy for constructing high‐functional‐density neuromorphic hardware integrating sensing, memory, computing, and decision‐making.
科研通智能强力驱动
Strongly Powered by AbleSci AI