神经形态工程学
材料科学
联轴节(管道)
航程(航空)
量子
量子计算机
订单(交换)
光电子学
纳米技术
化学物理
计算机科学
量子力学
人工神经网络
物理
人工智能
财务
经济
复合材料
冶金
作者
Zhiqing Wang,Jie Shen,Keqiang Chen,Jing Yang,Qiao Wang,Zhiwen Yin,Zhi‐Yi Hu,Jianrong Zeng,Pengchao Zhang,Wen Chen,Jing Zhou
标识
DOI:10.1002/adma.202509083
摘要
Abstract Bio‐inspired neuromorphic computing based on memristors holds significant potential for performing massively parallel computational tasks with high accuracy. However, its practical application is significantly limited by poor reliability, primarily due to instability in carrier transport. Here, long‐range ordered quantum dot (QD) superlattices with strong quantum coupling is presented to enable carrier transport stability and improve device reliability. Leveraging a data‐assisted QD synthesis optimization loop, Cu 12 Sb 4 S 13 QDs are synthesized with precisely controlled growth kinetics, crystal orientation, and surface chemistry. These QDs self‐assemble into long‐range ordered superlattices on flexible substrates, achieving a 56% reduction in inter‐dot spacing (to 0.92 nm), aligned lattice orientations, and a 4.4‐fold increase in carrier mobility. This architecture enables strong quantum coupling, effectively overcoming the limitations imposed by localized quantum‐confined states. As a result, the QD‐based memristors exhibit remarkable reliability, with variations below 0.1% over 8.4 × 10 7 s of continuous operation and 10 6 rapid read cycles. They further demonstrate linear potentiation and depression characteristics ( v p = 2.03 and v d = 2.33), a wide conductance range (G max /G min = 264), and high recognition accuracy (93.31%) as validated by chip‐level convolutional neural network simulations. This work establishes a robust and flexible platform for memristor‐based neuromorphic computing, offering a promising route to overcoming critical challenges in device reliability and computational performance.
科研通智能强力驱动
Strongly Powered by AbleSci AI