Multilayer Cross-Point Synapses Using Ga-Sn-O Thin Films for Neural Network

Abstract

We have developed multilayer cross-point synapses using Ga-Sn-O (GTO) thin films for neural networks. Twenty intermediate layer metal lines and ten lower and upper layer metal lines make 400 cross-point synapses integrated on a glass substrate. By continuously applying a constant voltage to the GTO thin films, the current value is changed. As a result, we obtained degradation that can be applied to the modified Hebb's learning rule.