Multimode transistors and neural networks based on ion-dynamic capacitance
Xiaoci Liang, Yiyang Luo, Yanli Pei, Mengye Wang, Chuan Liu
Abstract
Abstract Electrolyte-gated transistors can function as switching elements, artificial synapses and memristive systems, and could be used to create compact and powerful neuromorphic computing networks. However, insight into the underlying physics of such devices, including complex ion dynamics and the resulting capacitances, remains limited. Here we report a concise model for the transient ion-dynamic capacitance in electrolyte-gated transistors. The theory predicts that plasticity, high apparent mobility, sharp subthreshold swing and memristive conductance can be achieved—on demand—in a single transistor by appropriately programming the interfacial ion concentrations or matching the scan speed with ion motions. We then fabricate such multimode transistors using common solid-state electrolyte films and experimentally confirm the different capabilities. We also show in software that the multimode devices could be used to create neural networks that can be switched between conventional artificial neural networks, recurrent neural networks and spiking neural networks.