Improvement of Synaptic Properties in Oxygen-Based Synaptic Transistors Due to the Accelerated Ion Migration in Sub-Stoichiometric Channels

The oxygen ion-based synaptic transistor (OIST) is a promising candidate for next-generation synaptic devices for neuromorphic computing. However, the key process parameters that control synaptic characteristics have not yet been studied. In this study, the authors report near-ideal synaptic characteristics by adopting oxygen-deficient transition metal-oxide (TMO) channels and yttria-stabilized zirconia electrolytes. With decreasing oxygen stoichiometry in the TMO channel by controlling the O-2 flow rate during reactive sputtering, the oxygen ion supply and migration are accelerated by lowering the migration barrier. The reduced barrier energy and improved ion diffusivity characteristic in the oxygen-deficient TMO channels are confirmed experimentally through cyclic voltammetry analysis. As a result, improved weight-update linearity and wide conductance range can be achieved in OIST devices with an oxygen-deficient TMO (WO2.7, TiO1.7) channel. The authors confirm the excellent pattern recognition accuracy, which can be explained by the improved synaptic characteristics of the OIST synapse array.

Automated summary

In this study, near-ideal synaptic characteristics are achieved in OIST devices by adopting oxygen-deficient transition metal-oxide channels and yttria-stabilized zirconia electrolytes. By decreasing the oxygen stoichiometry in the TMO channel, oxygen ion supply and migration are accelerated, leading to improved weight-update linearity and wide conductance range. The excellent pattern recognition accuracy in the OIST synapse array can be explained by the improved synaptic characteristics.