Artificial Synaptic Properties in Oxygen-Based Electrochemical Random-Access Memory with CeO Nanoparticle Assembly as Gate Insulator for Neuromorphic Computing.

Beyond the von Neumann architecture, neuromorphic computing attracts considerable attention as an energy-efficient computing system for data-centric applications. Among various synapse device candidates, a memtransistor with a three-terminal structure has been considered to be a promising one for artificial synapse with controllable weight update characteristics and strong immunity to disturbance due to decoupled write and read electrode. In this study, oxygen ion exchange-based electrochemical random-access memory consisting of the ZnO channel and CeO nanoparticle (NP) assembly as a gate insulator, also as an ion exchange layer, is proposed and investigated as an artificial synapse device for neuromorphic computing. The memtransistor shows a tunable and reversible conductance change via oxygen ion exchange between ZnO and CeO NPs upon gate voltage application. The use of CeO enables efficient oxygen ion exchange with the ZnO channel due to its inherent property of easily absorbing and releasing oxygen ions by altering the valence state of the Ce cation. Additionally, the porous structure of the CeO NP assembly supports the oxygen reservoir function while retaining its insulating properties as a gate insulator, ensuring reliable device operation. Also, its porous nature enhancing oxygen ion exchange permits high-speed operation within tens of microsecond range. Based on the facilitated oxygen ion exchange, a highly linear and symmetric conductance modulation is achieved with good endurance over 10 pulses and excellent nonvolatile retention. Furthermore, the memtransistor mimics representative functions of the biological synapse such as paired-pulse facilitation, short-term (STP) and long-term plasticity (LTP), and the transition from STP to LTP as repeating learning cycles.