Optoelectrical Conditioning in Graphene Quantum Dot‍-‍Vanadium Oxide Artificial Synapse

Neuromorphic computing requires artificial synaptic devices capable of adaptive learning and memory retention. Here, we present a monolithically integrated optoelectronic synaptic device based on a 0D/3D Ag/graphene quantum dot-vanadium oxide (GQD-VO₂)/Ni heterojunction exhibiting stimulation-history-dependent conductance modulation. The device supports synaptic modulation under electrical, optical, and combined optoelectrical stimulation, with enhanced optoelectronic response compared with the bare VO₂ device. Repeated optoelectrical stimulation modifies the retained device state and subsequent potentiation-depression characteristics, a process described here as optoelectrical conditioning. This evolution is evaluated using retained current (I), fitted potentiation-depression time constants (τ), and nonlinearity coefficients. Wavelength-dependent optical associative learning with positive and negative reinforcement is further demonstrated under different optical stimuli. Performance evaluation using Fashion-MNIST and Digit-MNIST datasets show improved classification accuracy with the progression of optoelectrical conditioning learning cycles, increasing from 81.53% to 84.13% and from 91.71% to 95.58%, respectively. Monte Carlo simulations further indicate improved classification accuracy across multiple runs following repeated learning cycles. Furthermore, this behavior is illustrated through reconstructed Fashion-MNIST image outputs with the repeated learning cycles. These findings highlight the potential of Ag/GQD-VO₂/Ni device for neuromorphic applications, advancing AI-integrated bio-inspired computing.