Developing a New Defect Cluster Index

新晶圓缺陷群聚指標之建構

蔡文傑(Wen-Jie Tsai) ； 唐麗英(Lee-Ing Tong) ； 王春和(Chung-Ho Wang)

《工業工程學刊》 25卷1期 (2008/01) Pp. 18-30

https://doi.org/10.29977/JCIIE.200801.0003

摘要

晶圓的良率(yield)是衡量積體電路製造業者生產能力的一個重要指標，影響良率高低的因素有很多，其中晶圓上缺陷點(defect)數的多寡以及缺陷點的群聚(clustering)程度皆是決定晶圓良率高低的兩個重要因素。近年來隨著晶圓面積不斷增大，晶圓上缺陷點的群聚現象越來越明顯，此群聚現象會導致良率預測模式(yield model)錯估良率，或缺陷點管制圖(c-chart)誤判製程失控，此乃因兩片缺陷點數相同之晶圓，群聚程度嚴重的晶圓其缺陷點會集中在少部份的晶片上，良率反而會較群聚程度不嚴重之晶圓為高，因此探討晶圓上缺陷點的群聚程度在良率的管理上是非常重要的。許多中、外文獻雖然提出了各種不同的群聚指標來衡量缺陷點群聚的嚴重程度，然而這些群聚衡量指標各有不完善之處。因此本研究另發展一個新的群聚指標，此指標不僅較目前中、外文獻所提之指標能更準確地偵測出缺陷點群聚的嚴重程度，且能辨認出群聚之圖案。本研究最後以數個模擬之12吋晶圓實驗及一個實際6吋晶圓案例來說明所提出之新群聚指標確實有效可行。

關鍵字

晶圓 ； 積體電路 ； 缺陷點 ； 群聚現象 ； 群聚指標

並列摘要

The wafer yield is a critical index for evaluation of productivity of integrated circuit manufacturers. There are many factors affecting wafer yield. However, defect number and defect clustering are two determinants of wafer yield. Defect clustering tends to grow with increasing wafer size, making yield prediction erroneous when using yield model to predict wafer yield. Additionally, the false alarm will be produced in monitoring manufacturing process using C control chart. Even the number of defects is identical on respective wafer; the wafer yield will not be the same, because the defect pattern and defect clustering degree are different on each wafer. Normally, a serious defect clustering have higher yield than low defect clustering in the case of the same defects number, because defects are concentrated on a small number of chips. Therefore, evaluating defects clustering on wafer is an important issue in determination of wafer yield. Although many studies regarding development of defect clustering indexes had been released, some drawbacks still exists. Therefore, this study primarily developed a new clustering index to retrieve these drawbacks. The developed new clustering index not only can identify the clustering pattern but also can accurately evaluate the defect clustering given different type of typical clustering patterns. Several simulated experiments based on 12 inches wafer demonstrate the effectiveness of the developed clustering index. Finally, a wafer yield model based on a real case of 6 inches from semiconductor manufacturer in Taiwan was employed using the developed index to verify the improvement in predicted accuracy.

並列關鍵字

Wafer ； integrated circuit ； defect clustering ； clustering index

參考文獻

Cunningham, J. A.(1990).The use and evaluation of yield models in integrated circuit manufacturing.IEEE Transactions on Semiconductor Manufacturing.3,60-71.

Friedman, D. J.,M. H. Hansen,V. N. Nair,D. A. James(1997).Model-free estimation of defect clustering in integrated circuit fabrication.IEEE Transactions on Semiconductor Manufacturing.10,344-359.

Hong, C.,L. Milor,M. Choi,T. Lin(2005).Study of area scaling effect on integrated circuit reliability based on yield models.Microelectronics Reliability.45,1305-1310.

Jun, C. H.,Y. Hong,S. Y. Kim,K. S. Park,H. Park(1999).A simulation-based semiconductor chip yield model incorporating a new defect cluster index.Microelectronics Reliability.39,451-456.

Liou, J. J.,Q. Zhang,J. McMacken,J. R. Thomson,K. Stiles,P. Layman(2002).Statistical modeling of MOS devices for parametric yield prediction.Microelectronics Reliability.42,787-795.