Multifunctional artificial synapse enabled by CuInP2S6/graphene heterojunction nanosheets with fast speed and high recognition accuracy

A wide range of conductance in artificial synapses is of great significance for regulating electronic properties and realizing neuromorphic computing. In this work, we fabricated a CuInP2S6/graphene (CIPS/G) two-dimensional (2D) heterojunction device for simulating biological neural networks. By combining CIPS with graphene, the tuning range of the synaptic weight is expanded, and the conductance can reach a saturated state more quickly in a pulse operation time of 200 ns. Moreover, the simulation of biological synaptic functions of long-term potentiation/depression is achieved through the update of conductance weights, achieving recognition accuracies of 96.1% and 91.7% in digital recognition and sound feature recognition, respectively. Edge recognition of images is realized using the Bienenstock–Cooper–Munro learning rule. Our 2D heterojunction device shows potential applications in information recognition and image processing.