Pulse Tunable Quantum Conductance States in Oxygen Vacancy Engineered TiO_{2−Δ<i>x</i>} Memristor for Artificial Neural Network Applications

Ballistic transport of conduction electrons in atomic constrictions is of great interest for nanoscale device engineering. The conducting filament (CF) in memristors when confined to the atomic scale dimensions instigates ballistic transport of electrons that leads to quantized conductance (QC) states. Here, tunable QC states are demonstrated in an oxygen vacancy engineered TiO2-Δx memristor for artificial neural network computation. Synchrotron based energy dependent X-ray Photoelectron spectroscopy measurements are carried out to authenticate the oxygen vacancy gradient across the depth of the film. The TiO2−Δx memristor exhibits stable and reliable QC states which are controlled with DC and pulsed I–V measurements. More than 100G0 states are achieved by carefully optimizing the pulse operation. Incorporating quantum properties of CF in a physics based equivalent electrical circuit model, the correlation between the QC states and the applied pulse is established. Moreover, the device can recognize handwritten digits with an accuracy of 94.3% being predicted utilizing the QC states of the device as the synaptic weights in artificial neural network simulations. Demonstration of pulsed controlled QC states and physics based analytical modeling provide a unique opportunity to create multilevel analogue memory and neuromorphic computing circuits.