综述：基于二维材料的类脑突触器件

以MOSFETs作为集成电路的核心器件，计算机为现代信息社会奠定了基础并见证了许多领域的技术创新。然而，由于被称为冯·诺依曼瓶颈的内存与中央处理器(CPU)的分离，传统计算机在海量数据处理方面面临着挑战。在大数据时代，这一瓶颈在实现物联网（IoT）方面将变得尤为突出。受人脑具有高度并行计算和自适应学习能力的启发，人工神经网络被提出来解决传统计算架构的冯·诺依曼瓶颈问题，并取得了一系列突破性成果。然而，人工神经网络（ANNs）算法及相关软件仍在传统计算机上运行，导致计算能力有限、能源效率低等问题。例如，AlphaGo由1200个中央处理器(CPU)和180个图像处理器(GPU)实现，功耗高达数十万瓦。相比之下，人脑包含≈1000亿个神经元和1000万亿个突触时，功耗仅为20 W，每个刺激的操作消耗仅为1-100 fJ。模拟人脑的信号传输和信息处理，为开发高效能的神经形态计算提供了重要解决方案，显示了人工智能应用的巨大潜力，如海量数据处理、机器学习和模式识别。

自人工智能(AI)、物联网(IoT)和机器学习(ML)出现以来，对计算能力的需求呈指数级增长，这些领域都需要新的计算基元。类脑神经形态计算系统，能够实现设备级的模拟计算和数据存储的结合，近年来引起了极大的关注。此外，模拟生物突触的基本电子器件也取得了重大进展。由于二维材料的原子厚度和较弱的屏蔽效应，其物理性质可以很容易地被调节，这对突触的应用非常有利。本文以高性能和多功能神经形态计算应用为目标，对基于二维材料的突触器件进行了全面介绍，包括二维材料和异质结的优点、多种多功能二维突触器件以及相关的神经形态应用。讨论了二维突触器件未来发展面临的挑战和策略。这篇综述将对深入了解二维突触器件的设计和制备及其在神经形态计算中的应用提供帮助。

The demand for computing power has been increasing exponentially since the emergence of artificial intelligence (AI), internet of things (IoT), and machine learning (ML), where novel computing primitives are required. Brain inspired neuromorphic computing systems, capable of combining analog computing and data storage at the device level, have drawn great attention recently. In addition, the basic electronic devices mimicking the biological synapse have achieved significant progress. Owing to their atomic thickness and reduced screening effect, the physical properties of 2D materials could be easily modulated by various stimuli, which is quite beneficial for synaptic applications. In this article, aiming at high-performance and functional neuromorphic computing applications, a comprehensive review of synaptic devices based on 2D materials is provided, including the advantages of 2D materials and heterostructures, various robust multifunctional 2D synaptic devices, and associated neuromorphic applications. Challenges and strategies for the future development of 2D synaptic devices are also discussed. This review will provide an insight into the design and preparation of 2D synaptic devices and their applications in neuromorphic computing.