Dual-modal synaptic operation in a van der Waals heterostructure for emotion-inspired neuromorphic processing

Substantial progress has been made in artificial emotion-recognition systems; however, reproducing neurotransmitter-mediated emotional processing remains a major challenge for neuromorphic computing. In this study, we report bimodally modulated artificial synapses based on a van der Waals-stacked graphene quantum dot (GQD)/MoS₂ heterostructure that integrates dopamine and optical stimuli for emotion-inspired neuromorphic processing. Incorporation of GQDs increases the ON-state current and dynamic range and improves relearning behavior and synaptic weight modulation. The GQD/MoS₂ heterostructure exhibits dopamine-dependent electrical responses at concentrations of 1-5 nM and stable synaptic plasticity with retention maintained over the 1000 s measurement period. Optical stimulation in the presence of dopamine produces postsynaptic current (PSC) increases that exceed the additive expectation calculated from the individual optical-only and dopamine-only responses, supporting synergistic enhancement under the tested conditions. Device-informed facial emotion recognition simulations achieved an accuracy of 73.4% on the FER-2013 dataset using a convolutional neural network (CNN). Separately, classification of PSC-derived temporal signals using electroencephalography (EEG)-inspired processing achieved an accuracy of 76.6% with a support vector machine (SVM). These findings highlight the potential of this approach for affective computing and human-machine interaction.