Non-volatile memory behavior of interfacial InOx layer in InAs nano-wire field-effect transistor for neuromorphic application

Abstract

Nano-wire (NW) field-effect transistor (FET) is expected to be a promising device in the semiconductor industry owing to its scalability and enhanced gate-controllability. Particularly, III-V compound semiconductor-based NW FETs enable low power consumption because of their high mobility. In this study, a novel charge injection memory (CIM) device is presented using intrinsic InAs NW with high electron mobility. A simple combination of InAs native oxide and SiO2 stack accommodates charge trapping sites to store a bit information, resulting in a memory window of over 5 V. This charge-trapping behavior of the InAs NW FET is confirmed for more than 1000 s at room temperature. The disclosure of the charge-trapping effect in the InAs CIM provides a glimpse of the simplified non-volatile memory devices based on III-V NWs. Additionally, the synaptic behavior of InAs CIM is investigated for neuromorphic application. Utilizing the synaptic characteristics of the InAs CIM, an artificial neural network is implemented for simple handwritten digit recognition. This indicates that the InAs NW FETs can be used as the hardware for neuromorphic computational architectures.