Yiyang Li1
University of Michigan1
Data-intensive processes like artificial intelligence consume substantial amounts of energy. In-memory computing using analog resistive memory can be substantially more efficient that conventional digital approaches. In this work, we discuss our work on developing the electrochemical random-access memory (ECRAM) as the "synapse" for in-memory computing. ECRAM is a type of resistive memory that electrochemically shuttles oxygen vacancy point defects between two transition metal oxides through a solid electrolyte. Its electrochemical process yields ~1V switching and very low currents. Importantly, ECRAM can also attain high-temperature operation and the potential for several years of retention time at 85C.