Memristive artificial synapses based on Au–TiO2 composite thin film for neuromorphic computing

Abstract

Neuromorphic computing architecture is considered to be a highly desirable next-generation computing architecture as it simulates the way the brain processes information. The basic device supporting such an architecture is called an artificial synapse, which possesses synapse-like functionalities. Here in this work, an Au–TiO2 composite thin film (Au nanoparticles embedding into TiO2 matrix) based memristive artificial synapse has been fabricated with excellent interface-type resistive switching (RS) characteristics. The conductivity of the device can be continuously tuned by applying different sequences of pulses, which could be analogous to the weight change of synapses. Various synaptic behaviors have been emulated, such as long-term potentiation/depression, short-term/long-term memory, learning-forgetting process, and paired-pulse facilitation. Finally, an artificial neural network for hand-written digits recognition has been constructed with an accuracy level as high as ∼90%. The excellent performance of the Au–TiO2 based device demonstrates the availability of incorporating the second phase to tune RS properties and shows its potential in a memristor for artificial synapses and neuromorphic computing with enhanced performance.