建構晶圓缺陷點與缺陷群聚現象之適應性管制流程

Abstract

由於晶圓表面缺陷點（defects）多寡是決定晶圓良率（yield）的一個重要因素，積體電路業界多使用缺陷點數管制圖（c-chart）來管制晶圓表面之缺陷點，但若依照傳統抽樣方式，每批量僅抽取一片晶圓，c-chart僅能監控製程中不同批量間的變異，而無法同時監控晶圓間的變異，以及晶圓表面缺陷點之群聚現象，因此c-chart不適合用於複雜且具多種變異來源的晶圓製造系統。隨著科技的進步，產品良率是製造商亟欲提昇的目標，在提昇良率的同時亦希望能降低製造成本。在品質管制技術與成本都需改進的情況下。本論文利用缺陷群聚指標（cluster index）、Hotelling T2管制圖及適應性管制圖（adaptive control chart）發展出一套完整之晶圓缺陷點適應性管制流程，此流程不僅能有效地監控晶圓表面缺陷點之分佈及晶圓間缺陷點之變異，還可監控不同批量間之變異，並能依缺陷點及群聚嚴重程度以動態取樣之方式抽取適當之晶圓數目，來節省成本。本論文最後以模擬之晶圓資料來說明本流程之有效性，不但能迅速偵測出製程中各種偏移狀況，亦可以明確的指出造成異常產品的品質特性。

Number of defects on a wafer is a key factor of wafer yield. The Integrated-circuits(IC) manufacturers often adopt c-chart to monitor the wafer defects. The traditional sampling scheme is to sample just one wafer per lot, therefore c-chart can only monitor the lot-to-lot variation. The c-chart is unable to monitor simultaneously the wafer-to-wafer variation and clustering phenomenon of defects on a wafer. Consequently, c-chart is inappropriate to be employed in the complicated wafer manufacturing process with various sources of variation. Moreover, because the rapid development of the manufacturing techniques, manufacturers want to promote the wafer yield and curtail the budget. An adaptive control chart can be employed to reduce the monitoring cost. The objective of this study is to utilize a cluster index to describe the defect clustering phenomenon. Hotelling T2 control chart is employed in this study to detect number of total defects and clustering phenomenon simultaneously. This study further develops a complete monitoring procedure in which the adaptive control chart is utilized. The proposed procedure can not only monitor the defect clustering, but also monitor the wafer-to-wafer variation and the lot-to-lot variation. Finally, simulated cases of wafer defects are utilized to demonstrate the effectiveness of the proposed procedure.