Stable After More Than 10 Million Write and Erase Cycles
Performance Demonstrated in an 11×11 Memory Cell Crossbar Array

The Gwangju Institute of Science and Technology (GIST) announced on the 30th that the research team led by Professor Lee Sanghan from the Department of Materials Science and Engineering has developed next-generation device technology that can significantly improve the integration density and process efficiency of neuromorphic semiconductors. This achievement was made through joint research with teams from Seoul National University and the University of Southern California (USC).

Group photo of the international joint research team from GIST, Seoul National University, and USC. Provided by GIST

Group photo of the international joint research team from GIST, Seoul National University, and USC. Provided by GIST

View original image

Neuromorphic computing, a core technology of artificial intelligence (AI) semiconductors, is a technique that mimics the neural networks of the human brain to perform ultra-low-power and high-speed parallel processing. It mainly employs a “crossbar array” structure, where memory devices are arranged in a lattice pattern. However, nonlinear characteristics—where information is stored irregularly according to electrical stimuli—and the phenomenon of “leakage current,” where electricity escapes to adjacent devices, have limited computational accuracy.


To address these issues, conventional methods required each device to have an additional “selector” component to block leakage current. This increased the structural complexity and reduced integration density.


The joint research team overcame these challenges by precisely controlling the movement of oxygen vacancies through doping barium (Ba) into the ferroelectric material bismuth ferrite (BiFeO3). As a result, they succeeded in integrating memory functions for information storage and the “self-rectifying” function—which allows current to flow in only one direction—into a single device.


This newly developed device raised the rectification ratio by more than a million times, effectively blocking leakage current even without a separate selector. It also maintained stability after more than 10 million repeated write/erase cycles and demonstrated linear information storage characteristics, where information is accumulated within a constant range in response to electrical stimuli.


In an actual experiment, the research team implemented a crossbar array consisting of 121 memory cells (11×11) and succeeded in stably recording and controlling both text and image pattern information without the need for selector devices.


Professor Lee Sanghan of GIST stated, “This research is significant both academically and industrially, as it simultaneously solves two major challenges in neuromorphic semiconductors—nonlinearity in information storage and the problem of leakage current—within a single device. It will serve as a foundational technology for realizing highly efficient neuromorphic computing and accelerating the commercialization of ultra-low-power AI semiconductors.”



This research, supported by the Ministry of Science and ICT and the National Research Foundation of Korea, has been published online in Nature Communications, a highly respected international academic journal.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing