Artificial optoelectronic synaptic characteristics of Bi2FeMnO6 ferroelectric memristor for neuromorphic computing

Abstract

Photoelectric synapse is a new type of artificial synapse that combines light pulse excitation and electrical pulse excitation, and is an important component for the development of neuromorphic hardware. In this work, the Bi2FeMnO6 ferroelectric artificial synapse with optoelectronic response is reported. It can generate an excitatory post-synaptic current through ultraviolet light irradiation and then modulate the excitatory post-synaptic current through ferroelectric polarization, thereby realizing a photo-ferroelectric dual-control synapse. A training method based on ultraviolet excitation and ferroelectric enhancement is proposed, which greatly reduced the number of training pulses and increased the plasticity of the synapse. The optimized spike scheme based on ferroelectricity is proposed, which made the synapse have very linear weight adjustment function. This synapse is used to build a convolutional neural network, which can achieve 98.80% recognition accuracy on the MNIST dataset. In addition, the ferroelectricity and multi-level resistive switching effect of synapse have also been reported.

Keywords

• Bi2FeMnO6
• Ferroelectric
• Photoelectric
• Resistive switch
• Artificial synapse
• Neuromorphic computing