ECS Meeting Abstracts, 27(MA2019-02), p. 1226-1226, 2019
DOI: 10.1149/ma2019-02/27/1226
Full text: Download
Memristor devices are promising potential hardware components for computer memory and neural network computing. Advances over the last years on understanding and implementing memristor technology had positioned them as a major candidate to overcome current bottlenecks in current electronic-based transistors in terms of downscaling capabilities and energy consumption. In particular, current challenges preventing a widespread implementation of oxygen-based memristors in today’s integrated circuits include the need to address cycle-to-cycle and device-to-device variabilities. These occur due to the random nature of nanoionic conductive filaments, whose rupture and formation govern the device operation. Changes in the filament location, shape and chemical composition during resistance switching are responsible for the cycle-to-cycle variability. Here we tackle the variability challenge by employing a double stabilization mechanism, where we spatially confine the conductive filaments with Ni nanoparticles that are capable of oxygen exchange between the set and reset cycles. Ni nanoparticles are fabricated on the bottom La0.2Sr0.7Ti0.9Ni0.1O3-δ thin film electrode by an exsolution process, during which Ni dopant metal cations exsolve from the perovskite backbone at high temperatures and reducing conditions. This method of nanoparticle fabrication offers great control over particle size and density tuned by the exsolution conditions. Particles with average diameters ranging from 10 to 60 nm were fabricated at 900 and 1000°C under a 5% H2 atmosphere. Exsolution also insures a good particle anchorage into the bottom electrode preventing particle agglomeration and movement. SrTiO3 is deposited on top of the Ni-decorated bottom electrode as the switching material followed by and the device is toped with a platinum top electrode. Cyclic voltammetry shows, that the ratio between the ON and OFF resistance states increases from single units to 180 as the particle diameter increases. Additionally, the variability of the low resistance state decreases below 5% as the average particle diameter increases to 60 nm. These findings offer a straightforward strategy on improving cycle-to-cycle variability in memristive devices, which is a major challenge for commercial application of the technology.