Point-contact enabled reliable and low-voltage memristive switching and artificial synapse from highly transparent all-oxide-integration

摘要

Mimicking brain-like functionality for enabling higher-level artificial intelligence is one of the ultimate goals in neuromorphic computing, which could be achieved by two-terminal memristors. However, conventional memristors are suffering from severe shortcomings such as temporal (cycle-to-cycle) and spatial (device-to-device) reproducibility along with high operative voltage, albeit all these are crucial for accurate and quick information processing. Here, we demonstrate point-contact enabled reproducible and reliable bipolar resistive switching from all-oxide-based highly transparent memristors with low operating voltage (<0.5 V) and long retention times. Analogous to bio-synapse, memristor mimics the functions like short-term potentiation and short-term depression under the action of applied pulses. Conductive atomic force microscopy unambiguously revealed the formation of the localized conducting channels as well as nanoscale dynamics. Further, finite element simulation confirms that the tip-enhanced electric field could generate localized conduction channels, in contrast to the uniform electrode in which edges are the preferred sites for conduction. These results represent an important milestone toward the use of point-contact to design all-oxide-based highly transparent and reproducible resistive switching devices as well as artificial synapse. (C) 2020 Elsevier B.V. All rights reserved.