Coexistence of unipolar and bipolar resistive switching in optical synaptic memristors and neuromorphic computing

Abstract

The human brain possesses a highly developed capability for sensing-memory-computing, and the integration of hardware with brain-like functions represents a novel approach to overcoming the von Neumann bottleneck. In this study, Ga₂O₃ photoelectric memristors were successfully fabricated, enabling efficient visual information processing and complex recognition through the integration of optoelectronic synapses with digital storage. The memristors with a Pt/Ga₂O₃/Pt sandwich structure exhibit the coexistence of unipolar resistive switching (URS) and bipolar resistive switching (BRS), coupled with an impressive switching ratio and stable retention characteristics. The device demonstrates robust photo-responsive properties to ultraviolet (UV) light, which enables the realization of an array of 16 photoconductor types through the manipulation of four-timeframe pulse sequences. Exposure of the device to UV light elicits stable synaptic behaviors, including paired-pulse facilitation (PPF), short-term memory (STM), long-term memory (LTM), as well as learning-forgetting-relearning behavior. Moreover, the device exhibits outstanding image sensing, image memory, and neuromorphic visual pre-processing capabilities as a neuromorphic vision sensor (NVS). The integration of light pulse potentiation with electrical pulse depression yields a remarkable 100 conductances with superior linearity. This advanced functionality is further validated by the ability of the device to facilitate the recognition of 85.3% of handwritten digits by artificial neural networks (ANNs), which underscores the significant potential of artificial synapses in mimicking biological neural.