Impact of sub- and supra- threshold switching in the synaptic behavior of TiO2 memristors

Abstract

Al/TiO2/Au memory junctions grown by reactive sputtering were fabricated and characterized. Usual hysteretic features were obtained, spanning an order of magnitude ratio between high and low resistance states. We explored sub-threshold and supra-threshold stimuli, to describe “binary” and “analogic” synapses (i.e. synaptic emulators) characteristics in detail. We show that each type of behavior can be chosen regarding the dependence of the switching on the amplitude of voltage stimuli. The sub-threshold switching originating the analog mode is explained by the drift-diffusion of oxygen vacancies while the binary mode is attributed to the Joule-heating-assisted electric-field-induced movement of oxygen vacancies. In the supra-threshold limit we found that “excitatory” and “inhibitory” synaptic behavior can be tailored by controlling the pulsing current amplitude. We discuss this result within the artificial neural network topology framework, as changing the behavior type of the synapses results in its reconfiguration.