Identical optical stimulus-enabled multi-level conductance under zero field in ferroelectric optoelectronic field-effect transistor for neuromorphic computing

Abstract

Neuromorphic computing, particularly bio-inspired visual neural simulation, demands devices with simplicity, multi-level tunability, and high stability. Here, we demonstrate a ferroelectric optoelectronic field-effect transistor with an InGaZnO/Al2O3/PbZr0.52Ti0.48O3 heterostructure, which realizes multi-level conductance modulation under identical optical stimulation and zero field via the nonvolatile nature of ferroelectric polarization, mimicking multi-level synapses. The device shows an on/off ratio up to 104, stable performance over 300 cycles, 7-bit multi-level conductance retention for 100 s, and highly linear symmetric modulation (optical: αp = 0.051, αd = 0.133; electrical: αp = 0.052, αd = 0.058). Accordingly, the neural network built on it achieves nearly 98% motion recognition accuracy and 88.0% (optical) and 94.3% (electrical) handwriting digit recognition accuracy. Moreover, the device emulates the anchoring effect of historical stimuli on emotion, where previous stimuli induce the release of dopamine to modulate neuroplasticity, allowing individuals to exhibit multi-level emotional states when encountering the same scene after experiencing historical stimuli. This study proposes a novel strategy for implementing highly integrated and stable artificial photoelectric synapses, promoting the development of next-generation neuromorphic vision systems.