Study of SnSe and SrTiO3 thin films for memristor devices

Abstract

Due to the bottleneck limitation of von Neumann architecture-based hardware, the current Artificial Intelligence (AI) needs a large and complicated structure and a huge number of energy and memory accessing time to support the learning and thinking operation. Inspired by the operation of human brain, neuromorphic computing system built with circuit elements to mimic the neurobiological activities is a good concept to meet the challenge. Toward this application, resistive random-access memory (RRAM) has attracted great interest for non-volatile, low-power consumption, non-destructive readout, and high-density memories. In this work, RRAM devices based on SnSe and SiTrO3 (STO) thin films and their heterostructures with Ag and Cu electrodes are fabricated by pulsed-laser deposition, and their different structures and memory characteristics are investigated. Growth temperature-dependent microstructural study reveals that, the SnSe thin films grown at 300℃ on a Pt-coated silicon substrate possesses crystallized structure with better multi-level switching and endurance memory characteristics compared to that of the lower temperature grown film. With the STO buffer layer, the SnSe/STO heterostructure device shows much better performance on reliability test. Multi-level switching studies reveal that the Cu/SnSe/STO/Pt heterostructure device can present 6 resistance states and the Ag/SnSe/STO/Pt device presents 8 resistance states; all states remain stable after 10000 seconds, suggesting very good retention property. These results suggest that the Ag/SnSe/STO/Pt heterostructure device may function well in traditional multilevel memory applications. The memristive performance of SnSe/STO heterostructure suggests that these heterostructure devices are promising for applications in neuromorphic computing as a synaptic device.