Abstract
Neurons of the ventral tegmental area of the brain contain single axon terminals that release excitatory and inhibitory neurotransmitters, creating reconfigurable synaptic behaviour. Artificial synaptic transistors that exhibit similar excitatory and inhibitory behaviour-and hence synaptic function reconfiguration-could provide diverse functionality and efficient computing in various applications. However, some of these applications, such as soft robotics and wearable electronics, require synaptic devices that are mechanically soft and deformable. Here we report an elastic and reconfigurable synaptic transistor that exhibits inhibitory and excitatory characteristics even under mechanical strain. The synaptic device uses a top-gated configuration and is made using a stretchable bilayer semiconductor and an encapsulating elastomer as the gate dielectric. The device exhibits memory characteristics when operating with a presynaptic pulse of only 80 mV, resulting in a low specific energy consumption. When applied to a model artificial neural network for dual-directional image recognition of the Modified National Institute of Standards and Technology dataset, a recognition accuracy of over 90% is achieved even when the transistors are stretched by 50%.
An artificial synaptic transistor that uses a stretchable bilayer semiconductor as the channel and an encapsulating elastomer as the dielectric can exhibit both excitatory and inhibitory synaptic behaviour, even when under 50% strain.