Investigation of Defect-Driven Memristive and Artificial Synaptic Behaviour at Nanoscale for Potential Application in Neuromorphic Computing

摘要

Abstract The need for nano-sized memristive devices as synaptic emulators is increasing day-by-day to build neuromorphic devices on a chip with a large integration density. Therefore, nanoscale fabrication and subsequent device assessment are exigent to make further advancements in the field. In this report, by employing atomic force microscopy, we demonstrate highly dense electronic nano-synapses in filamentary metal oxide-based memristors, wherein the tip works as one of the electrodes. Different metal oxide-based memristors with varying defect concentrations have been studied to demonstrate the efficacy of atomic force microscopy-based local probe techniques in extracting the defect-dependent performance metrics of the same at nanoscale. Some of the basic biological synaptic characteristics such as potentiation/depression and spike’s property-dependent plasticities are accessed through different pulsed measurements in conductive atomic force microscopy mode. The current maps acquired using conductive atomic force microscopy measurements on the films confirm the filamentary resistive switching behaviour to prevail and point towards a possible decrease in the operating voltage with an increase in defects density. Kelvin probe force microscopic analyses, on the other hand, suggest that to improve the switching stability, resistivity of the active medium needs to be optimized. Overall, our results substantiate the efficacy of local probe microscopy-based methods to optimize the performance of memristive synapses at nanoscale for potential application in neuromorphic computing.