Reconfigurable Memristive Devices based on Phase Change Materials for Artificial Synapses and Neuromorphic Computing

Phase change materials such as vanadium oxide (VO₂) possess great potentials for artificial neuron application because of its inherent nonlinear characteristics such as metal-to-insulator (MIT) phase transition with an applied electric field and quick relaxation. Recent studies have demonstrated that the VO₂ based artificial neurons achieve superior biological neuronal activities at lower energy as compared to those based on the filamentary type devices. Our recent study on the VO₂based memristive devices have suggested that a reconfigurable resistive switching (RS) i.e., both volatile (as neurons) and non-volatile (as synapse) can be achieved, enabling the creation of more energy and area efficient artificial neural networks for neuromorphic computing. In this talk, I will present the case of the VO₂/La0.7Sr0.3MnO3 (LSMO) heterostructure devices and their dual RS properties. The movement of oxygen vacancies and the MIT phase change are deemed to be the main driving forces for such dual RS modes. A dual mode RS in a single device holds great promise for the application in memristive neural networks and neuromorphic computing.