A Neural Synapse Based on Ta2O5 Memristor

Abstract

The main purpose of this paper is to propose an improved memristor-based synaptic scheme, containing a re-sistor-memristor current divider and a differential amplifier with Metal Oxide Semiconductor (MOS) transistors. The memristor is made of tantalum oxide, doped by oxygen vacancies. The synaptic circuit contains only one memristor and produces positive, zero and negative weights. The applied tantalum oxide memristor model is based on the classical Hewlett-Packard model with several modifications and simplifications. Owing to the applied optimizations, the considered memristor model is faster than the corresponding original model. The synaptic weights of the considered memristor scheme, applied in a neural network are adjusted by voltage pulses and its operation is analyzed in L TSPICE environment.