Physically Transient Optic-Neural Synapse for Secure In-Sensor Computing

Herein, we demonstrate an artificial optic-neural synapse based on physically transient memristor. The physically transient optic-neural synapse with structure of W/MgO/ZnO/Mo exhibits remarkable long-term plasticity under light with various wavelengths and intensity including the red, green, and blue lights. In addition, a complete electro-thermal resistive switching model accurately simulates the migration and redistribution of oxygen vacancies in the memristor. Importantly, the long-term plasticity can disappear when the devices were immersed in deionized (DI) water, demonstrating the physically transient nature of the device. Finally, an optical convolutional neural network based on the physically transient optic-neural synapse was successfully simulated. These results show that the transient optical synapse has a great prospect for in-sensor computing and secure electronics.