A high-performance memristor induced by the synergistic effect between oxygen vacancies and Ag filaments for pattern recognition and error correction circuit design

With the rapid development of big data and artificial intelligence, conventional computing chips can no longer meet the demands of modern applications. Memristor-based neuromorphic computing, as a novel computing paradigm inspired by brain computing, has the potential to fundamentally overcome the computational bottleneck of traditional von Neumann architectures and improve computational efficiency. Herein, we demonstrate a biomimetic synaptic device based on an Ag/ZnO_x/FTO memristor, which excels in conventional storage, non-volatile conductance modulation, and logic operations. The switching mechanism is governed by the synergistic effect of oxygen vacancies and Ag filament formation. This device achieves over 1.2 × 10⁴ reversible write/erase cycles and maintains stable performance across five distinct conductance states, demonstrating promising memory characteristics. Moreover, its linearly tunable conductance enables synaptic weight updates in convolutional neural networks (CNNs), achieving an 89 % recognition accuracy on the MNIST dataset. Finally, an error correction circuit based on the ZnO_x memristor is designed for error correction of traffic signals and local blurring of images. This multifunctional ZnO_x-based memristor thus provides a viable hardware support for bio-inspired neuromorphic computing.