In recent years, artificial intelligence, big data and the Internet of Things have the higher requirements for the speed and efficiency of data processing. However, due to the separation of storage and computing structures, the conventional computing systems have the bottleneck problems in power consumption and efficiency when processing intensive-data and artificial intelligence tasks. The human brain is one of the most complex computing systems, which is capable of storing, integrating and processing a large amount of data and information simultaneously by the densely coordinated synaptic and neural networks. It has the advantages of both high speed and low-power consumption. Inspired by the human brain, artificial synaptic devices have received widespread attention due to their ability to process and remember data simultaneously, which are expected to be the core components in the next generation of neuromorphic computing systems.

GaN-based nanowires have several advantages such as large surface volume ratio, high stability and continuously adjustable energy bands. However, whether they can be used as an ideal material to prepare low-power artificial synaptic devices for simulating biological synaptic characteristics is a problem worth studying. A group led by Prof. Shulong Lu from the Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, has developed a micro artificial synaptic device based on a single GaN nanowire successfully. By the light-response characteristics and small-size advantages of a single GaN nanowire, this device can effectively simulate neural synaptic characteristics under light stimulation. With another advantage of low-power consumption, the device can help promote the development of neuromorphic computing systems. The relevant research results were published in Photonics Research Volume 11, No. 10, 2023 (Min Zhou, Yukun Zhao, Xiushuo Gu, Qianyi Zhang, Jianya Zhang, Min Jiang, Shulong Lu. Light-stimulated low-power artificial synapse based on a single GaN nanowire for neuromorphic computing[J]. Photonics Research, 2023, 11(10):1667).

As shown in Fig. 1(a), the incident light, bilateral electrodes and photogenerated carriers simulate the action potential, presynaptic/postsynaptic membranes and neurotransmitters in biological synapses, respectively. The scanning electron microscopy (SEM) image of the device in Fig. 1(b) agrees with the design in Fig. 1(a). This device has neural synaptic characteristics. Its single pulse energy consumption can be as low as 2.72 × 10^-12 J, which helps develop low-power neural network computing systems. Meanwhile, the team constructed a neural network to simulate the recognition of digital images. The recognition results of 20 randomly selected images in the sample library are shown in Fig. 1(c). As illustrated in Fig. 1(d), the recognition accuracy for the entire sample library (10000 images) can reach as high as 93% after 30 training cycles.

Fig.1 (a) Schematic diagram of the structural design of a single-nanowire synaptic device. (b) SEM physical image of a single-nanowire synaptic device. (c) The recognition results of randomly selected numbers from the MNIST database. (d) The curve between recognition accuracy and training frequency in simulation models.

Yukun ZHAO, Associate Research Fellow of Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, has said "The GaN-based nanowires with ultra-small size can not only reduce the power consumption, but also have good stability. In this work, the single nanowire has been demonstrated to simulate various functions of biological synapses, including peak dependence, light-intensity dependence and learning ability. Therefore, the artificial synaptic device based on a single nanowire has great application potential in the fields of neuromorphic computing systems and artificial intelligence".

Currently, compared to that of conventional devices, the device has a certain advantage in the metrics of power consumption. In the further study, this team will further reduce the power consumption of devices and promote the application process of the new optoelectronic devices.