Ultra-Low Power and Robust BiSeO Films Optoelectronic Memristors for Bio-Visual Perception Computing Systems.

Optoelectronic memristors have broad application prospects in the fields of artificial intelligence (AI) and the internet of things (IOT) because they can dynamically process a large number of spatiotemporal optoelectronic signals in complex environments. However, it is still a challenge to develop optoelectronic memristors with low power consumption and fast response while maintaining the efficiency and robustness of the system. Here, a photoelectric memristor based on BiSeO thin film is proposed. This device has an ultra-fast resistance switching speed (≈9.5 ns) and ultra-low synaptic event power consumption (≈1.36 fJ). The stable instantaneous light on/off response behavior within 4000 cycles and the very fast photoresponse speed (≈28 ns) also confirm the excellent applicability of the device as a photodetector. More importantly, a bionic visual perception and computing system is designed. The system not only uses devices as photodetectors to achieve wireless communication with a speed of more than 160 kb s, but also uses the synaptic effect of devices under bias voltage to realize the optical reservoir computing (RC) network at the hardware level. The accuracy rate of digital recognition is 90.5%, which provides an ultra-fast and low-power method for developing widely used and performance enhanced bionic visual systems.