Optical synapses offer a promising solution to the high energy consumption of von Neumann architectures. Despite significant research, existing photoconductivity modulation methods are typically unidirectional, and inhibitory behavior still depends on electrical stimulation. To address this, a two-terminal planar fully optically modulated synapse device based on a ZnAlSnO/SnS heterostructure, demonstrating bidirectional optical response is presented. This device exhibits an excitatory postsynaptic current (EPSC) when exposed to 370 nm UV light and generates an inhibitory postsynaptic current (IPSC) under 630 nm red light. Continuous potentiation and depression stimuli reveals the stability of fully optically modulated artificial synapses. Leveraging its fully optically modulated conductance, a three-layer artificial neural network is implemented for handwritten digit and clothing recognition, achieving accuracies of 91.12% and 78.22%, respectively. Additionally, based on its unique electrical response to UV light pulse, the development process of the Polaroid camera is well simulated. This work not only enriches the content of optical synapses, but also contributes to advancements in artificial intelligence, brain-like computing, and image-processing technologies.

中文翻译：

---

用于图像处理的基于 SnS 的 ZnAlSnO 基全光调制人工突触装置


光学突触为冯诺依曼架构的高能耗提供了一种有前景的解决方案。尽管进行了大量研究，现有的光电导调制方法通常是单向的，抑制行为仍然依赖于电刺激。为了解决这个问题，提出了一种基于 ZnAlSnO/SnS 异质结构的两端平面全光学调制突触器件，展示了双向光学响应。该器件在暴露于 370 nm 紫外光时表现出兴奋性突触后电流 (EPSC)，并在 630 nm 红光下产生抑制性突触后电流 (IPSC)。连续的增强和抑制刺激揭示了完全光学调制的人工突触的稳定性。利用其完全光学调制电导，实现了三层人工神经网络用于手写数字和服装识别，分别实现了 91.12% 和 78.22% 的准确率。此外，基于其对紫外光脉冲的独特电响应，可以很好地模拟宝丽来相机的开发过程。这项工作不仅丰富了光学突触的内容，还为人工智能、类脑计算和图像处理技术的进步做出了贡献。