Optoelectronic Synaptic Transistors Based on Colloidal CdSe Nanowires for Energy-Efficient Neuromorphic Computing

Neuromorphic computing─which mimics biological synaptic functions─has garnered significant attention as a promising candidate for overcoming the limitations of conventional von Neumann computing. Various synaptic devices exhibiting short- and long-term plasticity characteristics have been developed for neuromorphic device fabrication, and field-effect transistor (FET) structures have been actively researched for their ability to implement sophisticated computing. This study reports the first neuromorphic thin-film transistor (TFT) based on colloidal semiconductor nanowires (NWs). Cadmium selenide (CdSe) NWs exhibit synaptic characteristics in response to both electrical and optical stimuli when fabricated as synaptic thin-film transistors (STFTs) owing to their persistent photoconductivity and large transfer-curve hysteresis characteristics. The device exhibits both short-term plasticity features, including excitatory and inhibitory postsynaptic currents alongside paired-pulse facilitation, as well as long-term plasticity behavior, such as long-term potentiation and depression. Notably, a single NW of these STFTs was calculated to consume approximately 8.848 fJ per synaptic event, approaching the energy efficiency of biological synapses. A spiking neural network implemented through the spike-timing-dependent plasticity characteristics of the CdSe NW STFT successfully learned handwritten digits from the Modified National Institute of Standards and Technology database with over 80% accuracy. The combination of biologically similar learning mechanisms and ultralow energy consumption─comparable to that of biological synaptic events─makes these STFTs highly promising for the development of energy-efficient neuromorphic computing systems.