Solid-state Proton Devices for Neuromorphic Information Processing

Biological neural systems achieve real-time information processing by exploiting dynamics spanning multiple time scales, ranging from seconds to minutes, which enables effective interaction with the surrounding environment. In contrast, conventional electronic circuits have been predominantly designed for high-speed operation, and as a result, there exist few scalable electronic devices capable of operating on the same time scale as environmental dynamics. In this study, we focus on proton transport in amorphous WO3 and develop a vertical two-terminal protonic device that exhibits volatile resistance changes with relaxation times on the order of seconds to minutes. The proposed device operates at low voltages below 1V, features high impedance, and shows area-scalable behavior. We further demonstrate that the device can function as a short-term memory element in an Echo State Network (ESN). Using real-time prediction of blood glucose time-series data as a representative application, we demonstrate the potential of ultra-low-power neuromorphic computing in which slow material relaxation phenomena are directly exploited for information processing.