Electric-stimulated controllable synaptic GaN nanodevice for neuromorphic computing

Due to a range of distinctive advantages, synaptic nanodevices hold substantial promise in the fields of neuromorphic computing systems that parallel the functionality of the human brain. In this work, we have demonstrated a cost-efficient method for preparing the artificial synaptic nanodevice based on gallium nitride (GaN) nanowire successfully. By applying appropriate alternating current, the proposed dielectrophoretic alignment method is simple to be performed. The GaN nanowire can be accurately aligned at the predetermined position on the metal electrodes within 10 min. Such a facile method can not only reduce the cost but also save the time. The systematic experimental results prove that the synaptic nanodevice can mimic the functions of biological synapses, including the excitatory postsynaptic current, paired-pulse facilitation, spike-timing-dependent plasticity, short-term and long-term memory, etc. Furthermore, a threshold voltage of about 7 V was found to exist in the electric-stimulated synaptic GaN nanodevice. The memristive response can be modulated by adjusting the operational voltage, pulse width, inter-pulse interval and pulse quantity. Based on experimental conductance, the simulated three-layer neural network could achieve a stable recognition accuracy of 95% after just 13 training cycles. Notably, such a nanodevice is also capable of reducing image noise to enhance image quality. The findings of this study pave a new path for the development of neuromorphic computing hardware and artificial intelligence systems based on synaptic nanodevices.