摘要
大量实验表明了在集成电路 (IC)制造过程中形成的缺陷团之间具有很强的相关性 ,然而现用于划分缺陷团的方法均忽略了缺陷之间的相关性 .因此 ,得到的缺陷分布规律不能有效地反映缺陷在圆片上的分布 .为了提高IC成品率和可靠性仿真和设计的精度 ,提高缺陷分布模型的准确性 ,该文提出了划分缺陷团的聚类算法 .该方法依据缺陷形成的动力学基础 ,充分考虑了缺陷团之间的相关性 .实验证明该算法用于划分
In all available methods of partitioning IC's defects into clusters, the correlation among defects is ignored. Therefore, the available defect spatial distributions cannot reflect effectively the distribution of the defects on a wafer. In fact, on a wafer not only relevant are defects to defects, but also defects to clusters, and clusters to clusters. Firstly, based on the idea of the correlation among defects, the definition of the subjection of the i th defect X i to the j th defect cluster with its center at V j , and the definition of fuzzy correlative coefficient and the correlation degree between the j th defect cluster and the k th defect cluster are presented in this paper. Secondly, the step change fuzzy clustering algorithm is presented based on fuzzy clustering method of fuzzy mathematics according to the correlation coefficient among defect clusters. Finally, the step change fuzzy clustering algorithm of partitioning into clusters is used to deal with the 100 maps of defect distribution sampled in a production line and every sample map must be divided into some square grids. To guarantee the correlation among clusters is poorly, and it is required that at most one defect is in every square grid when the 100 maps of defect distribution are divided into m×m square grids. The statistic histogram of defects among wafers, the statistic histogram of the number of defect in a defect cluster, the distribution of the cluster number on a wafer, and so on, are given in this paper. The experimental results show that this algorithm is effective to partition IC's defects into clusters.
出处
《计算机学报》
EI
CSCD
北大核心
2002年第6期661-665,共5页
Chinese Journal of Computers
