当前位置: X-MOL 学术IEEE Trans. Semicond. Manuf. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Deep Learning for Classification of the Chemical Composition of Particle Defects on Semiconductor Wafers
IEEE Transactions on Semiconductor Manufacturing ( IF 2.3 ) Pub Date : 2020-02-01 , DOI: 10.1109/tsm.2019.2963656
Jared O'Leary , Kapil Sawlani , Ali Mesbah

Manual classification of particle defects on semiconductor wafers is labor-intensive, which leads to slow solutions and longer learning curves on product failures while being prone to human error. This work explores the promise of deep learning for the classification of the chemical composition of these defects to reduce analysis time and inconsistencies due to human error, which in turn can result in systematic root cause analysis for sources of semiconductor defects. We investigate a deep convolutional neural network (CNN) for defect classification based on a combination of scanning electron microscopy (SEM) images and energy-dispersive x-ray (EDX) spectroscopy data. SEM images of sections of semiconductor wafers that contain particle defects are fed into a CNN in which the defects’ EDX spectroscopy data is merged directly with the CNN’s fully connected layer. The proposed CNN classifies the chemical composition of semiconductor wafer particle defects with an industrially pragmatic accuracy. We also demonstrate that merging spectral data with the CNN’s fully connected layer significantly improves classification performance over CNNs that only take either SEM image data or EDX spectral data as an input. The impact of training data collection and augmentation on CNN performance is explored and the promise of transfer learning for improving training speed and testing accuracy is investigated.

中文翻译:

用于半导体晶片上粒子缺陷化学成分分类的深度学习

人工对半导体晶圆上的颗粒缺陷进行分类是劳动密集型的,这导致解决方案缓慢,产品故障的学习曲线更长,同时容易出现人为错误。这项工作探索了深度学习对这些缺陷的化学成分进行分类的前景,以减少分析时间和由于人为错误造成的不一致性,从而可以对半导体缺陷来源进行系统的根本原因分析。我们基于扫描电子显微镜 (SEM) 图像和能量色散 X 射线 (EDX) 光谱数据的组合研究了用于缺陷分类的深度卷积神经网络 (CNN)。包含粒子缺陷的半导体晶片部分的 SEM 图像被输入 CNN,其中缺陷的 EDX 光谱数据直接与 CNN 的全连接层合并。所提出的 CNN 以工业实用的准确性对半导体晶片颗粒缺陷的化学成分进行分类。我们还证明,与仅将 SEM 图像数据或 EDX 光谱数据作为输入的 CNN 相比,将光谱数据与 CNN 的全连接层合并可以显着提高分类性能。探讨了训练数据收集和增强对 CNN 性能的影响,并研究了迁移学习在提高训练速度和测试准确性方面的前景。所提出的 CNN 以工业实用的准确性对半导体晶片颗粒缺陷的化学成分进行分类。我们还证明,与仅将 SEM 图像数据或 EDX 光谱数据作为输入的 CNN 相比,将光谱数据与 CNN 的全连接层合并可以显着提高分类性能。探讨了训练数据收集和增强对 CNN 性能的影响,并研究了迁移学习在提高训练速度和测试准确性方面的前景。所提出的 CNN 以工业实用的准确性对半导体晶片颗粒缺陷的化学成分进行分类。我们还证明,与仅将 SEM 图像数据或 EDX 光谱数据作为输入的 CNN 相比,将光谱数据与 CNN 的全连接层合并可以显着提高分类性能。探讨了训练数据收集和增强对 CNN 性能的影响,并研究了迁移学习在提高训练速度和测试准确性方面的前景。
更新日期:2020-02-01
down
wechat
bug