Dual-gated ambipolar oxide synaptic transistor for multistate excitatory and inhibitory responses

Developing tunable and multi-input artificial synaptic devices is a significant step to realize diverse functionalities inspired by a bio-neural network and is essential to advance the development of multifunctional human-like neuromorphic devices. This study developed an artificial synaptic device exhibiting tunable and multi-state excitatory and inhibitory responses by using a dual-gated (DG) ambipolar boron-doped SnO thin-film transistor. We demonstrated dynamic modulation of multi-state potentiation/depression responses in both reconfigurable excitatory and inhibitory modes by the DG operation in a single ambipolar transistor. In comparison with conventional single-gate devices, the DG configuration improved the linearity and the symmetricity of synaptic weight updates in addition to the capability of conduction level tuning. Therefore, the presented DG ambipolar oxide synaptic transistor exhibited distinct advantages in learning-accuracy and energy-efficiency, and high pattern recognition accuracy over 90% and low energy operation of similar to 200 pJ per pulse in excitatory and inhibitory responses were achieved. It demonstrates the high potential of the DG ambipolar oxide synaptic transistor for next-generation energy-efficient multi-input neuromorphic devices to emulate diverse functionalities in bio-neural network systems.

Automated summary

This study developed an artificial synaptic device that exhibits tunable and multi-state excitatory and inhibitory responses. By using a dual-gated ambipolar boron-doped SnO thin-film transistor, the device demonstrated dynamic modulation of multi-state potentiation/depression responses in both reconfigurable excitatory and inhibitory modes. Compared to conventional devices, the dual-gated configuration improved the linearity and symmetry of synaptic weight updates and allowed for conduction level tuning. This oxide synaptic transistor shows high potential for energy-efficient multi-input neuromorphic devices.