Electrochemical ohmic memristors for continual learning

Developing versatile and reliable memristive devices is crucial for advancing future memory and computing architectures. The years of intensive research have still not reached and demonstrated their full horizon of capabilities, and new concepts are essential for successfully using the complete spectra of memristive functionalities for industrial applications. Here, we introduce two-terminal ohmic memristor, characterized by a different type of switching defined as filament conductivity change mechanism (FCM). The operation is based entirely on localized electrochemical redox reactions, resulting in essential advantages such as ultra-stable binary and analog switching, broad voltage stability window, high temperature stability, high switching ratio and good endurance. The multifunctional properties enabled by the FCM can be effectively used to overcome the catastrophic forgetting problem in conventional deep neural networks. Our findings represent an important milestone in resistive switching fundamentals and provide an effective approach for designing memristive system, expanding the horizon of functionalities and neuroscience applications.