KAIST strikes again!
The "heterovalent ion doping" method??
View attachment 68757
Korean Researchers Discover Method to Enhance Next-Gen. Neuromorphic Computer Performance
- Editor Jasmine Choi
- 2024.06.21 16:52
Print URL Copy Fonts Size Down Fonts Size Up
facebook(으)로 기사보내기 twitter(으)로 기사보내기 URL Copy(으)로 기사보내기 링크드인(으)로 기사보내기 Send to Email Share Scrap
Neuromorphic computing technology, which mimics the human brain to implement artificial intelligence (AI) operations. (Photo by Getty Images Bank)
A team of South Korean researchers has developed a technology that enhances the reliability and commercialization of next-generation neuromorphic computing devices by addressing their irregular characteristics.
Professor Choi Sin-hyeon from the Korea Advanced Institute of Science and Technology (KAIST) and his team, in collaboration with researchers from Hanyang University, announced on June 21 that they have developed a heterovalent ion doping method that improves the reliability and performance of next-generation memory devices.
Neuromorphic computing is a technology that implements AI operations by emulating the human brain. It uses memristors as basic units, which are advantageous for their low power consumption, high integration, and efficiency. Memristors, a portmanteau of memory and resistor, are memory devices that retain all previous states. However, due to their unstable characteristics, memristors often have low reliability.
The research team developed a “Heterovalent ion doping method” to improve the uniformity and performance of these devices. Heterovalent ions are ions that have a different valency from the atoms that originally existed, with valency being a measure of bonding.
The team proved the performance of heterovalent ion doping through atomic-level simulation analysis. The doped heterovalent ions attracted vacancies in nearby oxygen, creating stable device operation. Additionally, the space near these ions was expanded, allowing for faster device operation. According to the team's analysis, the performance of memristors doped with heterovalent ions improved in both crystalline and amorphous environments.
Professor Choi Sin-hyeon stated, "The heterovalent ion doping method can enhance the reliability and performance of neuromorphic devices,” and added, “It can contribute to the commercialization of next-generation neuromorphic computing based on memristors."
The results of this study were published in the international academic journal “Science Advances” on June 7.
A team of South Korean researchers has developed a technology that enhances the reliability and commercialization of next-generation neuromorphic computing dev
www.businesskorea.co.kr