Multistate resistive switching behaviors for neuromorphic computing in memristor

Conventional Von Neumann computing systems encounter increasing challenges in the big-data era due to the constraints by the separated data storage and processing. Resistive random-access memory provides dual functionalities of data storage and computing at the same position without data transmission. This is one of the most promising candidates for energy efficient neuromorphic computing. The key points to realize neuromorphic computing are the selection of functional materials, the design of multistate devices, and a complete logic function implementing in-memory computing. Here, we demonstrate a memristor device, formed by Al/TiO2-few-layer Graphene-DNA/Pt layers, with stable intermediate multistate resistive switching behaviors. Asynchronous conduction by either oxygen vacancies migration or injected electron transfer is responsible for the multistate resistive switching behaviors. For neuromorphic computing, a pixel data stored and 2-bit parallel logic computations are simulated based on the multistate resistive switching behaviors. Compared with traditional memristor devices, this device can achieve theoretically double the data storage. This work provides a new horizon on the memristive memory and the complete logic hardware. (C) 2020 The Author(s). Published by Elsevier Ltd.

Automated summary

Resistive random-access memory provides dual functionalities of data storage and computing at the same position, making it a promising candidate for energy efficient neuromorphic computing. The key points to realize neuromorphic computing include selection of functional materials, design of multistate devices, and complete logic function implementation in-memory computing. A memristor device with stable intermediate multistate resistive switching behaviors has been demonstrated, which enables simulated pixel data storage and 2-bit parallel logic computations for neuromorphic computing.