A fast-switching threshold memristor based on AlN/Ag/AlN stacked device for mental disorder electroencephalogram recognition

Abstract

As artificial intelligence technology advances, volatile threshold switching (TS) devices gain increasing attention for neuromorphic applications; however, most conventional TS devices suffer from slow switching speeds, limiting their suitability for high-speed neuron circuits. This work introduces an additional thin Ag layer into the traditional Ag/AlN/n-Si structure, forming an Ag/AlN/Ag/AlN/n-Si configuration that accelerates filament formation and significantly improves threshold switching characteristics. The device achieves a steep turn-on slope of 0.1 V/decade and fast switching with 40 ns turn-on and 25 ns turn-off times, showing competitive performance with reported TS devices. Interestingly, this work demonstrates neuron activation and deactivation and enables a leaky integrate-and-fire model to construct an artificial neural network with ∼91% accuracy in mental disorder electroencephalogram recognition. These results highlight the device's potential for high-speed artificial neuron circuits and neuromorphic systems.