Reliable Multivalued Conductance States in TaO x Memristors through Oxygen Plasma-Assisted Electrode Deposition with in Situ-Biased Conductance State Transmission Electron Microscopy Analysis

ACS Appl Mater Interfaces. 2018 Sep 5;10(35):29757-29765. doi: 10.1021/acsami.8b09046. Epub 2018 Aug 2.

Abstract

Transition metal oxide-based memristors have widely been proposed for applications toward artificial synapses. In general, memristors have two or more electrically switchable stable resistance states that device researchers see as an analogue to the ion channels found in biological synapses. The mechanism behind resistive switching in metal oxides has been divided into electrochemical metallization models and valence change models. The stability of the resistance states in the memristor vary widely depending on: oxide material, electrode material, deposition conditions, film thickness, and programming conditions. So far, it has been extremely challenging to obtain reliable memristors with more than two stable multivalued states along with endurances greater than ∼1000 cycles for each of those states. Using an oxygen plasma-assisted sputter deposition method of noble metal electrodes, we found that the metal-oxide interface could be deposited with substantially lower interface roughness observable at the nanometer scale. This markedly improved device reliability and function, allowing for a demonstration of memristors with four completely distinct levels from ∼6 × 10-6 to ∼4 × 10-8 S that were tested up to 104 cycles per level. Furthermore through a unique in situ transmission electron microscopy study, we were able to verify a redox reaction-type model to be dominant in our samples, leading to the higher degree of electrical state controllability. For solid-state synapse applications, the improvements to electrical properties will lead to simple device structures, with an overall power and area reduction of at least 1000 times when compared to SRAM.

Keywords: TaOx memristors; artificial solid-state synapses; in situ STEM analysis; multivalued atomic switching.