POSTECH Breakthrough: Hidden Mechanisms in Next-Generation AI Memory Device Unveiled

POSTECH researchers have achieved a breakthrough in understanding Electrochemical Random-Access Memory (ECRAM), uncovering hidden operating mechanisms that could make AI technologies significantly faster and more efficient.

Professor Seyoung Kim and Dr. Hyunjeong Kwak from POSTECH's Materials Science & Engineering and Semiconductor Engineering departments, collaborating with IBM's Dr. Oki Gunawan, published their groundbreaking findings in Nature Communications on April 25, 2025.

Current computing systems separate data storage from processing, causing significant energy consumption through data transfers. "In-Memory Computing" addresses this by enabling calculations directly within memory, and ECRAM is critical for implementing this concept.

Using a multi-terminal ECRAM device with tungsten oxide and the "Parallel Dipole Line Hall System," researchers observed internal electron dynamics from ultra-low temperatures (-223°C) to room temperature. They discovered that oxygen vacancies create shallow donor states (~0.1 eV), forming "shortcuts" for free electron movement.

Rather than simply increasing electron quantity, ECRAM inherently creates environments facilitating easier electron transport. This mechanism remained stable even at extremely low temperatures, demonstrating device robustness and durability.

Prof. Seyoung Kim emphasized: "This research experimentally clarified ECRAM's switching mechanism across various temperatures. Commercializing this technology could lead to faster AI performance and extended battery life in smartphones, tablets, and laptops."

ECRAM features three-terminal structure (source, drain, gate) where gate voltage controls ion movement, with channel conductivity read through source and drain. The Parallel Dipole Line Hall System uses rotating cylindrical dipole magnets to generate strong, superimposed magnetic fields for enhanced sensitivity in observing internal electron behaviors.

This work, supported by K-CHIPS (Korea Collaborative & High-tech Initiative for Prospective Semiconductor Research) and funded by Korea's Ministry of Trade, Industry & Energy, represents a major step toward commercializing energy-efficient AI hardware.