Owing to the synaptic behaviors and functions, memristors are intensively studied as a critical component for the neuromorphic computing system which is considered as an effective scenario to tackle the performance bottleneck existing in modern computers based on the von Neumann architecture. A novel synaptic device base on the CuI memristor prepared with a solution-based process is proposed in this work, and a set of synaptic features are emulated. The electrochemical metallization contributes to the resistive switching behavior and small operating voltages ( = 0.67 V and = −0.33 V). Moreover, the CuI memristor exhibits the small switching energy (23 pJ) and ultraquick switching speed (100 ns). A large CuI memristor array is realized experimentally, by which the hardware-based multiply-and-accumulate (MAC) operation is implemented. Furthermore, the hardware-based convolution computation in image processing is realized based on the MAC operation, demonstrating the potential applications in image processing requiring the high degree of parallelism.

中文翻译：

---

基于溶液的工艺制备纳秒级 CuI 突触忆阻器

由于忆阻器的突触行为和功能，忆阻器作为神经形态计算系统的关键组件而受到深入研究，这被认为是解决基于冯·诺依曼架构的现代计算机中存在的性能瓶颈的有效方案。这项工作提出了一种基于 CuI 忆阻器的新型突触器件，采用基于溶液的工艺制备，并模拟了一组突触特征。电化学金属化有助于电阻开关行为和小工作电压（= 0.67 V 和= -0.33 V）。此外，CuI忆阻器具有较小的开关能量（23 pJ）和超快的开关速度（100 ns）。实验实现了大型CuI忆阻器阵列，实现了基于硬件的乘法累加（MAC）运算。此外，基于MAC运算实现了图像处理中基于硬件的卷积计算，展示了在需要高度并行性的图像处理中的潜在应用。