High-Stability Memristive Devices Based on Pd Conductive Filaments and Its Applications in Neuromorphic Computing.

Abstract

Memristive devices with high-density and high-speed performance have considerable potential for neuromorphic computing applications in data storage and artificial synapses. However, current memristive devices that are based on conductive filaments, such as silver, are unstable owing to the high mobility and low thermodynamic stability of the filaments. A high-quality SnSe film was deposited using the pulsed laser deposition technology, and high-performance Pd/SnSe/NSTO devices were fabricated. High-stability memristive devices can not only implement simple arithmetic function but also exhibit the centralized distribution of SET/RESET voltage and cell-cell uniformity. The SET/RESET power can achieve approximately 4.1 and 61 μW power. The possibility of Pd filament formation and Pd2+ diffusion in SnSe thin films is first confirmed by combining high-resolution transmission electron microscopy, energy-dispersive spectrometer mapping, and first principle calculation. The formation and destruction process of Pd filaments can simulate the influx and extrusion kinetics of K+, Ca2+, or Na+ in biological synapses and implements considerable synaptic functions. This study thus provides a new idea for improving device performance using different filament materials, which can greatly facilitate the development of neuromorphic computing.