Real-time information processing via volatile resistance change in scalable protonic devices

Biological neural systems operate on multiple time scales, enabling real-time interaction with their environment. However, replicating this in electronic systems is challenging, as scalable devices that operate on similar time scales, particularly from seconds to minutes, are lacking. This study addresses this gap by exploiting proton dynamics to achieve volatile resistance changes over long time scales, and developed a neuromorphic system that can predict biomedical data, such as blood glucose levels, in real time. By applying a low voltage (below 1 V) to a two-terminal device, the Pd electrode hydrogenates or dehydrogenates the mixed conductor WO3, resulting in significant changes in electronic resistance. The device is scalable due to the uniform proton distribution, leading to high impedance and minimal power consumption. Utilizing these volatile protons for short-term memory in an Echo State Network (ESN), this approach demonstrates low-power, efficient real-time information processing, paving the way for future neuromorphic computing applications. Real-time processing and exchange of information with the surrounding environment is characteristic of biological neural systems but is difficult to achieve in on-chip analog devices due to time-scale mismatch. Here, two-terminal protonic devices based on Pt/WO3/Pd structures address this issue and demonstrate neuromorphic computing short-term memories for real-time prediction of biomedical data such as blood glucose level.