神经形态工程学
材料科学
晶体管
光电子学
纳米复合材料
纳米技术
能源消耗
人工神经网络
二氧化钛
计算机科学
电介质
调制(音乐)
突触
电子工程
电压
制作
突触重量
噪音(视频)
光开关
记忆电阻器
干扰(通信)
像素
长时程增强
逻辑门
量子隧道
尖峰神经网络
作者
Qun‐Gao Chen,Rou‐Yi Li,Hsiu‐I Liu,Chu‐Chen Chueh,Wen‐Ya Lee
出处
期刊:Small
[Wiley]
日期:2026-11-05
摘要
ABSTRACT Artificial optical synapses featuring low‐energy consumption, high speed, and noise tolerance are critical for visual neuromorphic electronics. Most existing devices rely on complex multilayer structures or require high supply voltages to induce carrier tunneling effects. Therefore, we propose a photoactive nanocomposite dielectric (PND) layer, thereby enabling an embedded‐layer‐free, low‐voltage, low‐power optical organic synaptic transistor (OST) capable of all‐optical, optical‐electrical, and all‐electrical modulation. We achieve this by functionalizing titanium dioxide nanoparticles (TiO 2 NPs) with nitrile butadiene rubber to fabricate a photo‐curable, patternable PND layer. Under electrical and optical stimulation, the PND‐based OST exhibits outstanding synaptic characteristics, including spike‐width‐dependent plasticity, spike‐number‐dependent plasticity, spike‐rate‐dependent plasticity, as well as long‐term potentiation and long‐term depression. Notably, this device achieves a low‐energy consumption of 10.09 fJ at V D = ‐1 mV. We further propose a new energy assessment framework to quantify the energy consumption of synaptic weight modulation in neural network computations, providing a more practical evaluation metric. In a neuromorphic image recognition, integrating optically tunable synaptic weights into a deep neural network achieves a recognition accuracy of 92.7% under high‐noise conditions through all‐optical modulation. Overall, PND structure not only simplifies device fabrication but also offers a promising solution for developing efficient, low‐power neuromorphic electronic devices.
科研通智能强力驱动
Strongly Powered by AbleSci AI