Enhanced Nonvolatile Electrochemical Random-Access Memory and Artificial Synapse Characteristics through Oxygen Ion-Exchange Engineering in an Atomic-Layer-Deposited HfO2-x Gate Insulator and a Zinc Oxide Channel Layer

Abstract

Enhanced nonvolatile memory and artificial synapse characteristics are achieved in oxygen ion-based ECRAM consisting of a low-temperature atomic layer-deposited (ALD) oxygen-deficient hafnium oxide (HfO2-x) ion-exchange layer and zinc oxide (ZnO) channel layer. The drain current modulation of the device reaches a few orders of magnitude upon application of positive programming and negative erasing gate bias. Also, the device exhibits nonvolatile retention of modulated current up to >10(4) higher than the initial value for 24 h. Nonvolatile modulation of channel conductance results from oxygen ion exchange between the HfO2-x ion-exchange layer and ZnO channel layer in the nanometer scale, facilitated by using oxygen-deficient HfO2-x deposited at a low temperature (LT-HfO2-x) and ZnO layers as well as the use of UV/ozone treatment on LT-HfO2-x. Additionally, it presents various synaptic characteristics including analog, linear, and symmetric potentiation and depression behaviors upon repeating >10(4) pulses, paired-pulse facilitation depending on the pulse number, amplitude, and width, and short-term and long-term plasticity. These synapse characteristics are benchmarked to have MNIST pattern recognition accuracy over 93% using a CrossSim simulator. These enhanced nonvolatile memory and artificial synaptic characteristics verify the potential application of the proposed ECRAM for high-density stand-alone nonvolatile memory and artificial synapses for brain-inspired neuromorphic computing systems