Xie Pengshan, Li Dengji, Wang Weijun, Ho Johnny C
Department of Materials Science and Engineering, City University of Hong Kong, Hong Kong SAR, 999077, China.
State Key Laboratory of Terahertz and Millimeter Waves, City University of Hong Kong, Hong Kong SAR, 999077, China.
Small. 2025 Jul 23:e2503717. doi: 10.1002/smll.202503717.
The von Neumann architecture faces significant challenges in meeting the growing demand for energy-efficient, real-time visual processing in edge applications, primarily due to data-transfer bottlenecks between processors and memory. Two-dimensional (2D) materials, characterized by their atomic-scale thickness, adjustable optoelectronic properties, and diverse integration capabilities, present a promising avenue for advancing in-sensor computing. These material systems, which include ferroelectric 2D materials, topological insulators, and twistronic systems, enhance the device's ability to handle perception, computation, and storage efficiently. This review provides a comprehensive overview of the latest advancements in 2D material systems, exploring their operational mechanisms and key visual perceptual functions, such as polarization sensing and spectral selection. The potential applications of visual neural synaptic devices within current material systems are also examined, highlighting ongoing efforts to integrate various deep learning algorithmic architectures with innovative device integration strategies. This includes everything from demand-side design to the selection of appropriate material systems. By merging device and materials innovation with neuromorphic engineering, 2D materials hold the promise of overcoming the limitations of the von Neumann architecture, paving the way for the development of intelligent vision systems that harness the power of in-sensor computing.
冯·诺依曼架构在满足边缘应用中对节能实时视觉处理日益增长的需求方面面临重大挑战，这主要是由于处理器与内存之间的数据传输瓶颈所致。二维（2D）材料以其原子级厚度、可调节的光电特性及多样的集成能力为特征，为推进传感器内计算提供了一条很有前景的途径。这些材料体系，包括铁电二维材料、拓扑绝缘体和扭曲电子体系，增强了器件高效处理感知、计算和存储的能力。本综述全面概述了二维材料体系的最新进展，探讨了它们的运行机制以及关键的视觉感知功能，如偏振传感和光谱选择。还研究了当前材料体系中视觉神经突触器件的潜在应用，突出了将各种深度学习算法架构与创新器件集成策略相结合的持续努力。这涵盖了从需求端设计到合适材料体系选择的方方面面。通过将器件与材料创新与神经形态工程相结合，二维材料有望克服冯·诺依曼架构的局限性，为开发利用传感器内计算能力的智能视觉系统铺平道路。
