Stable retention in SrTiO3/SrRuO3 heterostructure-based memristive devices

Abstract

The switching dynamics in memristive devices resembles biological synapses, which makes these devices promising candidates for the development of artificial neural networks. The retention of the synaptic weight is a key parameter in performing artificial intelligent tasks, particularly the inference process. However, many memristive devices show retention loss over time, especially the oxide device with switching behavior caused by oxygen migration. In this work, we report a memristive device based on the structure of Pt/SrTiO3/SrRuO3, which greatly improves the retention time of the device. Based on the investigation of the electrical transport mechanism and interface microstructure, the retention improvement in present devices is due to the restriction of oxygen vacancy migration through the SRO-rich interface layer formed on the surface of the SrRuO3 bottom electrode. Given the patternable bottom electrode, linear conductance modulation, and excellent performance in neural network simulation, the present device has shown great potential for hardware neuromorphic computing applications.