Li Zebin, Lee Jihea, Yao Fei, Sun Hongyue
Department of Industrial and Systems Engineering, University at Buffalo, The State University of New York, Buffalo, NY, USA.
Department of Material Design and Innovation, University at Buffalo, The State University of New York, Buffalo, NY, USA.
Nanoscale. 2021 Sep 23;13(36):15324-15333. doi: 10.1039/d1nr03802h.
Machine learning (ML) techniques have been recently employed to facilitate the development of novel two-dimensional (2D) materials. Among various synthesis approaches, chemical vapor deposition (CVD) has demonstrated tremendous potential in producing high-quality 2D flakes with good controllability, enabling large-scale production at a relatively low cost. Traditionally, the quality of CVD-grown samples can be manually evaluated based on optical images which is labor-intensive and time-consuming. In this paper, we explored a data-driven unsupervised quality assessment strategy based on image clustering integrating self-organizing map (SOM) and -means methods for optical image analysis of CVD-grown 2D materials The high matching rate between the clustering results and material experts' labels indicated a good accuracy of the proposed clustering algorithm. The proposed unsupervised ML methodology will provide materials scientists with an effective tool kit for efficient evaluation of CVD-grown materials' quality and has a broad applicability for various material systems.
机器学习（ML）技术最近已被用于促进新型二维（2D）材料的开发。在各种合成方法中，化学气相沉积（CVD）在生产具有良好可控性的高质量二维薄片方面显示出巨大潜力，能够以相对较低的成本进行大规模生产。传统上，基于光学图像对化学气相沉积生长的样品质量进行人工评估既费力又耗时。在本文中，我们探索了一种基于图像聚类的数据驱动无监督质量评估策略，该策略集成了自组织映射（SOM）和K均值方法，用于对化学气相沉积生长的二维材料进行光学图像分析。聚类结果与材料专家标签之间的高匹配率表明所提出的聚类算法具有良好的准确性。所提出的无监督机器学习方法将为材料科学家提供一个有效的工具包，用于高效评估化学气相沉积生长材料的质量，并且对各种材料系统具有广泛的适用性。
