Optoelectronic synapses possess the potential to seamlessly integrate perception, storage, and neuromorphic computation, thereby offering advantages in constructing artificial vision systems. Gallium nitride (GaN) materials, renowned for their exceptional optoelectronic characteristics and the compatibility of Silicon-based GaN (GaN-on-Si) with CMOS technology, contribute significantly to realizing large-scale optoelectronic neuromorphic circuits. In this study, we present an optoelectronic dual-synapse based on a structurally-improved AlGaN/GaN MIS-HEMT with ring-shaped SiO barriers embedded surrounding the source and drain regions. This device facilitates simultaneous signal transmission to the source and drain through current-limiting electrical breakdown beneath the gate. This integration of two discrete optoelectronic synapses within a single device enhances neural signal transmission nodes and branching connectivity, facilitating the construction of large-scale neuromorphic circuits. Under the co-influence of electrical and optical pulses, the device not only demonstrates excellent optoelectronic synaptic characteristics but also exhibits the ability to modulate the critical wavelength of synaptic responses by varying the source/drain voltage, effectively simulating the light signal recognition observed in biological retinas. The synergistic operation of the two sub-synapses enables a successful transition from short-term potentiation (STP) to long-term potentiation (LTP). Furthermore, leveraging the device's STP, we emulate the function of artificial retina through a GaN optoelectronic neuromorphic array, successfully achieving image recognition in conjunction with artificial neural networks (ANNs). The devices developed in this study, based on GaN-on-Si allow wafer-level manufacturing, provding a promising avenue for the large-scale production of optoelectronic synapse devices and the realization of artificial vision systems.

中文翻译：

---

基于包含嵌入式 SiO2 势垒层的晶圆级 GaN-on-Si 器件的光电双突触

光电突触具有无缝集成感知、存储和神经形态计算的潜力，从而为构建人工视觉系统提供了优势。氮化镓（GaN）材料以其卓越的光电特性以及硅基氮化镓（GaN-on-Si）与CMOS技术的兼容性而闻名，为实现大规模光电神经形态电路做出了重大贡献。在这项研究中，我们提出了一种基于结构改进的 AlGaN/GaN MIS-HEMT 的光电双突触，在源极和漏极区域周围嵌入了环形 SiO2 势垒。该器件通过栅极下方的限流电击穿促进信号同时传输到源极和漏极。这种在单个设备中集成两个离散光电突触增强了神经信号传输节点和分支连接性，促进了大规模神经形态电路的构建。在电脉冲和光脉冲的共同影响下，该器件不仅表现出优异的光电突触特性，而且还表现出通过改变源/漏电压来调制突触响应的临界波长的能力，有效模拟了生物中观察到的光信号识别。视网膜。两个子突触的协同运作能够成功地从短期增强（STP）过渡到长期增强（LTP）。此外，利用该设备的STP，我们通过GaN光电神经形态阵列模拟人工视网膜的功能，成功地实现了与人工神经网络（ANN）结合的图像识别。本研究开发的基于硅基氮化镓的器件可实现晶圆级制造，为光电突触器件的大规模生产和人工视觉系统的实现提供了一条有前景的途径。