Data-driven services which create value for customers by processing big data using AI (Artificial Intelligence) are receiving a lot of attention. Contactless IC cards are one of the most promising IoT (Internet of Things) devices for collecting big data in view of ease of use.

In this paper, we have developed the world’s first nonvolatile embedded SrBi2Ta2O9(SBT)-based FeRAM (Ferroelectric Random Access Memory) with ultra-low power, which is most useful for application to contactless IC cards contributing to the establishment of traffic infrastructure. To expand its application, Mbit-class high density embedded memory has been required. However, poly-crystalline SBT thin films have the essential issue that only <100> domains of thin films show strong polarization, limiting the reduction of the area of SBT-based memory capacitors. In order to overcome the issue, we have developed a high-density nonvolatile embedded TaOx-based ReRAM（Resistive Random Access memory）which consists of nano-crystalline tantalum oxide thin films as resistive layers.

A highly reliable nonvolatile ReRAM operation based on the resistance of the thin films strongly requires controlling the density of oxide vacancies in the conductive filament which is formed in thin film layers. In order to meet this requirement, we have developed resistive memory cells with two-layer thin films and embedded integration techniques incorporated on standard 40nm CMOS LSIs. The fabricated 4Mbit-ReRAM memory shows the characteristics of more than 10,000 endurance cycles and over 10-years retention at 85°C. In addition, we have proposed a neuromorphic device utilizing the analog variable resistance of ReRAM cell as weights, which potentially shows application to cutting-edge AI devices in view of ultra-low power operation. This paper deserves the IEICE Best Paper Award because it provides useful information to understanding the current status regarding FeRAM and ReRAM, and also shows the future prospects for embedded ReRAM applications.