標題: 半導體晶圓瑕疵圖形之分析與偵測
Analysis and Detection of Spatial Defect Patterns on Semiconductor Wafer
作者: 陳星戎
Shing-Rong chen
周志成
Chi-Cheng Jou
電控工程研究所
關鍵字: 空間統計;空間瑕疵分佈對;自相關分析;spatial statistic;spatial defect distribution;autocorrelation
公開日期: 2007
摘要: 半導體製造過程包含繁複製作步驟,製作完成後,對晶圓切割下的每顆晶片做測試來確認其是否能正常運作。每批產品的正確率被稱為良率,而提高良率是半導體廠營運重要的課題之一。要使得產品的良率提高,必須找出製造過程中的錯誤來加以改進,然而製程變數非常多,在此情況下,較為便利的方式便是藉由晶圓圖上的瑕疵圖形去推測製造過程中發生的問題。目前分析晶圓圖的工作仍然是靠著目視去判斷,這是一項依靠經驗的工作,不同的工程師對於同一張晶圓圖可能會判斷出不同的結果,導致分類缺乏一致性,而且缺乏效率。為了使上述的不可抗力因素降低,我們希望憑藉資料分析的方法,從目前所擁有的分類實例中學習判斷法則,建立快速、有效、一致的自動化系統來分辨晶圓上的錯誤圖形。本論文提出自動化分析晶圓圖的方式,將瑕疵圖形分類為,幾何型,重複型,隨機型三種瑕疵圖形。改善現有方法無法將重複型以及隨機型圖形分類的缺點,並利用空間統計分析的方式將幾何型瑕疵自動分類,再對其加以改善,提出了新的作法,使得分類的效果更好。
Product yield is very important for integrated circuit (IC) fabrication. Because of the complex fabrication processes, it is critical to identify possible faults (ex. machine problems or parametric errors) through discriminating the failure patterns on wafer. But there will be a lot of problems if we recognize defect pattern by human operation. So designing an automatic system to classify defect pattern by standard rules efficiently is very imperious. In this paper, we develop a method to classify defect patterns to three different types (repeat, geometric, and random) automatically. That new method can identify repeat type and random type, and there is not any current method can do this. It also d make the spatial statistic method better than now, therefore we can get better result in classifying geometric and random patterns.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009412567
http://hdl.handle.net/11536/80700
Appears in Collections:Thesis