On December 20, Hanyang University announced a groundbreaking development by Professor Park Hui-joon from the Department of Organic and Nano Engineering and his research team, in collaboration with Professor Kim Kyeong-hak from the Department of Chemical Engineering and Professor Baac Hyoung-won’s research team from Sungkyunkwan University. Together, they developed a high-performance and highly reliable semiconductor synapse array based on a 3D ion moving channel.

With the advent of Big Data, a next-generation computing technology that can quickly and effectively process various forms of unstructured data is being spotlighted. Neuromorphic computing systems, in particular, have garnered attention for their ability to efficiently process large amounts of data by executing data storage and computation within artificial intelligence networks.

While multilevel resistive random-access memory has been used to construct artificial intelligence networks, ensuring the performance and reliability of the synapse element was difficult because of the switching medium’s inability to control the formation and rupture of metal filaments deliberately. 

To address this issue, the collaborative research team introduced the PDMS  head-end process to develop a metal oxide thin film with a few nanometer-sized, uniform 3D ion moving channels. Utilizing this film as a switching medium in multilevel resistive random-access memory could form a high-performance and highly reliable synapse array with the optimized conducting filament mechanism. Subsequently, the research team demonstrated precise control over synapse weight, which is crucial for enhancing the learning and recognition accuracy of neuromorphic computing systems within the proposed synapse array. 

Going further, the research team made a stretchable double-layered photoacoustic strain sensor based on carbon nanotubes. This sensor, attachable to hands, facilitated the creation of a neuromorphic sensory system capable of distinguishing random hand movements with a 97.4% accuracy through learning ultrasonic signal changes via a synapse-based artificial neural network.

Professor Park Hui-joon, who led the research, said, “We succeeded in applying a sensor that can detect various signals to a neuromorphic system. This breakthrough can be utilized as a core technology for metaverse and will contribute to developing the next-generation neuromorphic sensory system.”

This research was conducted under the support of the National Research Foundation’s PIM Artificial Intelligence Semiconductor Core Technology Development (Device) Project and Basic Research in Science & Engineering Support Project and published in the ACS Nano, a journal for materials science.

■ Title: Highly Reliable 3D Channel Memory and Its Application in a Neuromorphic Sensory System for Hand Gesture Recognition

■ Authors: Kim Do-hyung (first author, Hanyang University, master’s and doctorate integrated course), Lee Cheong-beom (first author, Hanyang University, master’s and Ph.D. integrated course), Park Kyu-kwan (first author, Ph.D. program, Sungkyunkwan University), Phuoc Loc Truong (Ph.D. program, Gachon University), Lee Jong-min (Hanyang University, master’s and Ph.D. integrated course), Jeong Bum-ho (Hanyang University, master’s and Ph.D. integrated course), Professor Won Sang-min (Sungkyunkwan University), Professor Kim Do-hwan (Hanyang University), Professor Lee Dae-ho (Gachon University), Professor Ko Jong-hwan (Sungkyunkwan University), Professor Baac Hyoung-won (corresponding author, Sungkyunkwan University), Professor Kim Kyeong-hak (corresponding author, Hanyang University), Professor Park Hui-joon (corresponding author, Hanyang University)