Controllable and Cost-Efficient Three-Terminal GaN Nano-Synapse for Brain-Inspired Computing

Although the development of highly controllable and low-cost three-terminal synaptic nano-devices is essential for advancing neuromorphic electronics, achieving precise alignment of single nanowire and stable electrical gating still remains severely challenging. Here, we propose and demonstrate a three-terminal artificial synaptic nano-device based on the GaN nanowire successfully, enabled by a dielectrophoretic-assisted assembly strategy that ensures controllable nanowire placement. Benefiting from this cost-efficient method and with an engineered gate-coupled interface, the nano-device exhibits robust and gate-tunable synaptic plasticity, including short-/long-term memory transition, paired-pulse facilitation, and spike-timing-dependent plasticity. By modulating optical spike parameters and gate voltages, the key cognitive behaviors such as learning-forgetting-relearning are effectively emulated, with negative gating significantly accelerating memory reinforcement. They are mainly attributed to the gate-regulated optoelectronic mechanisms, particularly carrier modulation and oxygen-vacancy-induced persistent photoconductivity. Thanks to the excellent regulatory capability of electrical gating, the postsynaptic current of nano-device can be enhanced over 1,000%. Furthermore, the recognition accuracy can surpass 95% accuracy by gating modulation when integrated into a spiking neural network. This work highlights the promise of three-terminal nano-synapses as effective and cost-efficient building blocks for next-generation neuromorphic systems.