High-performance BST:MgO memristor with optoelectronic synaptic plasticity for bio-inspired visual memory

Abstract

In the era of big data, traditional computing paradigms are confronted with numerous challenges. However, artificial neural networks offer an effective approach to break through these bottlenecks by emulating the information processing mechanisms of the human brain. This work presents a typical memristor based on the Pt/BST:MgO/LSCO/STO structure, featuring excellent cycling performance, concentrated distribution of high/low resistance states, and superior retention capability. The device enables stable modulation of conductance under electrical regulation, while its current variation under light pulse stimulation correlates with neural excitability and achieves the transition from short-term memory to long-term memory. Furthermore, the device demonstrates sensitivity to both light intensity and duration. A 5 × 5 array was further constructed to mimic synaptic behaviors analogous to human visual perception by utilizing this characteristic. Finally, by integrating synaptic plasticity with convolutional neural network, the system achieves up to 96% accuracy in image recognition and classification tasks. This study paves the way for high-performance memristors in advanced neuromorphic computing and artificial intelligence applications.