Demonstration of synaptic and resistive switching characteristics in W/TiO2/HfO2/TaN memristor crossbar array for bioinspired neuromorphic computing

In this study, resistive random-access memory (RRAM)-based crossbar arrays with a memristor W/TiO2/HfO2/TaN structure were fabricated through atomic layer deposition (ALD) to investigate synaptic plasticity and resistive switching (RS) characteristics for bioinspired neuromorphic computing. X-ray photoelectron spectroscopy (XPS) was employed to explore oxygen vacancy concentrations in bilayer TiO2/HfO2 films. Gaussian fitting for O1s peaks confirmed that the HfO2 layer contained a larger number of oxygen vacancies than the TiO2 layer. In addition, HfO2 had lower Gibbs free energy (Delta G degrees = -1010.8 kJ/mol) than the TiO2 layer (Delta G degrees=-924.0 kJ/mol), resulting in more oxygen vacancies in the HfO2 layer. XPS results and AG degrees magnitudes confirmed that formation/disruption of oxygen-based conductive filaments took place in the TiO2 layer. The W/TiO2/HfO2/TaN memristive device exhibited excellent and repeatable RS characteristics, including superb 10(3) dc switching cycles, outstanding 10 7 pulse endurance, and high-thermal stability (10(4) s at 125 degrees C) important for digital computing systems. Furthermore, some essential biological synaptic characteristics such as potentiation-depression plasticity, paired-pulse facilitation (PPF), and spike-timing-dependent plasticity (STDP, asymmetric Hebbian and asymmetric antiHebbian) were successfully mimicked herein using the crossbar-array memristive device. Based on experimental results, a migration and diffusion of oxygen vacancy based physical model is proposed to describe the synaptic plasticity and RS mechanism. This study demonstrates that the proposed W/TiO2/HfO2/TaN memristor crossbar-array has a significant potential for applications in non-volatile memory (NVM) and bioinspired neuromorphic systems. (C) 2022 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.

Automated summary

In this study, RRAM-based crossbar arrays with a memristor structure were fabricated through ALD to investigate synaptic plasticity and RS characteristics for bioinspired neuromorphic computing. The experimental results showed that the device exhibited excellent RS characteristics and successfully mimicked biological synaptic features. Additionally, a physical model involving migration and diffusion of oxygen vacancies was proposed to describe synaptic plasticity and RS mechanism.