Optosynapse Habituation in WO_x-C₆₀ Heterojunction for Efficient Human Faces Recognition

Neuromorphic optoelectronic devices show great potential in processing complex tasks, but state-of-the-art optoelectronic devices possess an accumulative photocurrent response that causes high consumption and requires additional circuit to stress spike overshoot. Here we proposed a WO_x-C₆₀ heterojunction optosynaptic device that can execute efficient human faces recognition through a falling photocurrent response. Such WO_x-C₆₀ heterojunction integrates the positive photoresponse of WO_x film and negative photoresponse of C₆₀ film to build reliable synapse habituation behaviors, enabling low energy consumption ∼25.0 pJ and 256 distinguished photoresponse states. Reservoir computing (RC) is based on the synapse habituation effect in the WO_x-C₆₀ heterojunction, yielding over 95% accuracy for human faces recognition. This work discloses a significant optosynapse mechanism as well as an optoelectronic hardware for in-sensor computing.