Adaptive threshold in TiO2-based synapses

Abstract

Abstract: We measured and analyzed the dynamic and remnant current-voltages curves of Al/TiO2/Au and Ni/TiO2/Ni/Au memory devices in order to understand the conduction mechanisms and their synapselike memory properties. Current levels and switching threshold voltages are strongly affected by the metal used for the electrode. We propose a non-trivial circuit model which captures in detail the currentvoltage response of both kinds of devices. We found that, for the former device, the voltage threshold can be maintained constant, independently of the applied voltage history, while for the latter, a limiting resistor controls the threshold voltages behavior, being the origin of their dependence on the resistance value previous to the switching. The identification of the conduction mechanisms across the device allows optimizing the memristor performance and determining the best electrode choice to improve the device synapse-emulation abilities.