Polarity-tunable persistent photoconductive behaviors in BiFeCrO3/TiO2 heterojunctions optoelectronic memristor for neuromorphic computing

Abstract

Neuromorphic computing is regarded as a key approach to overcoming the limitations of traditional computing architectures, with its core objective focusing on the development of artificial synaptic devices capable of replicating the plasticity of biological synapses. Herein, the high-quality heterostructured BiFeCrO3 (BFCO)/TiO2 thin films were prepared using a combination of the sol-gel method and magnetron sputtering technique. The BFCO/TiO2 heterojunctions optoelectronic memristor exhibited stable digital resistive switching behavior (stability up to 103 s) and demonstrated synaptic-like weight potentiation and depression under consecutive voltage sweeps. Furthermore, the BFCO/TiO2 optoelectronic memristor device successfully achieved the key synaptic functions including long-term potentiation, long-term depression, paired-pulse facilitation, and spike-timing-dependent plasticity, with relaxation time constants closely matching those observed in biological synapses. Interestingly, leveraging the photosensitive properties, polarity-tunable persistent photoconductive behaviors were demonstrated in the BFCO/TiO2 optoelectronic memristor under ultraviolet light pulse stimulation at different bias polarities, accompanied by a transition from short-term memory to long-term memory during learning processes. Based on nonlinear normalized conductance data, a memristor based neural network was constructed, achieving recognition accuracies of 97.30% for handwritten digit classification and 82.25% for clothing classification. These results strongly highlight the broad application potential of BFCO-based heterojunctions optoelectronic synapses in low-power intelligent sensing and brain-inspired computing systems.