Two modes of bipolar resistive switching characteristics in asymmetric TaO x -based ReRAM cells
- PDF / 1,175,685 Bytes
- 7 Pages / 432 x 648 pts Page_size
- 72 Downloads / 213 Views
MRS Advances © 2019 Materials Research Society DOI: 10.1557/adv.2019.316
Two modes of bipolar resistive switching characteristics in asymmetric TaOx-based ReRAM cells Toshiki Miyatani, Yusuke Nishi, and Tsunenobu Kimoto Department of Electronic Science and Engineering, Kyoto University, Kyoto 615-8510, Japan
ABSTRACT
Impacts of a forming process on bipolar resistive switching (RS) characteristics in Pt/TaOx/Ta2O5/Pt cells were investigated. We found that the forming resulted in a transition from an initial state to a particular high resistance state (HRS) in most of the Pt/TaOx/Ta2O5/Pt cells. Evaluation of electrical characteristics after the transition to the particular HRS revealed that two modes of bipolar RS with the conventional polarity based on valence change mechanism and with the opposite polarity could be selectively obtained by adjusting the magnitude of the applied voltage. Moreover, the cell resistance decreased gradually during set processes in the bipolar RS with the opposite polarity.
INTRODUCTION The era of the Internet of Things (IoT) generates more and more digital data all over the world every day. Processing of these big data requires further development of AI (artificial intelligence) technology. Deep neural network (DNN) has been studied as one of the AI technology [1]. Multiply-accumulate (MAC) operation in the DNN is conventionally performed by utilizing a high-performance parallel computing hardware [2-3]. However, a conventional von Neumann-type computer architecture has physical limit of dramatic improvement in power and speed of computation, so-called the von Neumann bottleneck [4, 5]. One of novel computing architectures to solve the problem is neuromorphic computing, which enables low-power and high-speed computation by mimicking the human brain [6, 7]. The neuromorphic computing requires synaptic elements to store analog weights of the MAC operation [1, 8]. A resistive random access memory (ReRAM) cell is a strong candidate not only for emerging nonvolatile memories but also for the synaptic elements, because ReRAM can control the cell resistance as synaptic weights by the applied voltage [9, 10]. Nevertheless, the challenge of ReRAM
Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 07 Aug 2019 at 10:46:25, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1557/adv.2019.316
as synaptic elements is abrupt decrease in the cell resistance especially at a set process [7, 11]. Although multi-level resistance at the set process is controllable by a current limiting resistor of an external transistor [12, 13], the external circuit is not preferable in terms of high-density integration. Therefore, analog resistive switching (RS) is required at the set processes as well as reset processes. Our previous study revealed that forming characteristics had a significant influence on nonpolar RS characteristics based on thermo-chemical mechanism (TCM) [14, 15]. However, the nonpolar RS normally exh
Data Loading...
