Optoelectronic Synapses Based on Hot-Electron-Induced Chemical Processes.

Highly efficient information processing in the brain is based on processing and memory components called synapses, whose output is dependent on the history of the signals passed through them. Here, we have developed an artificial synapse with both electrical and optical memory effects using chemical transformations in plasmonic tunnel junctions. In an electronic implementation, the electrons tunneled into plasmonic nanorods under a low bias voltage are harvested to write information into the tunnel junctions via hot-electron-mediated chemical reactions with the environment. In an optical realization, the information can be written by an external light illumination to excite hot electrons in the plasmonic nanorods. The stored information is nonvolatile and can be read either electrically or optically by measuring the resistance or inelastic-tunneling-induced light emission, respectively. The described architecture provides a high density (∼10 cm) of memristive optoelectronic devices which can be used as multilevel nonvolatile memory, logic units, or artificial synapses in future electronic, optoelectronic, and artificial neural networks.