晶圓缺陷點變異管制圖

Abstract

摘要 晶圓品質是積體電路產業成為擁有市場競爭優勢的一個重要因素。在晶圓製造過程中，穩定的製程是影響晶圓品質的一個重要因素。統計製程管制中之缺陷點數管制圖(c-chart)，常被用來監控晶圓品質之穩定性。目前由於晶圓製造技術的精進，晶圓表面積愈作愈大，使得晶圓表面的缺陷點群聚(clustering)現象越來越明顯，而c-chart未能考慮到此群聚現象，晶圓業者若仍使用c-chart來管制缺陷點數，會產生過多的假警報，因此在監控晶圓是否有變異時必須先考慮到晶圓表面缺陷點之群聚現象。此外，在晶圓生產過程中品質變異的來源有許多種，除了有不同批量間的變異(lot-to-lot variation)外，還有同一批量內晶圓間的變異(wafer-to-wafer variation)，如果不能釐清變異的來源，將無法有效的控管晶圓品質。因此本研究以晶圓表面的缺陷點數當作品質特性，提出一套晶圓缺陷點數變異來源之辨識流程。一旦晶圓缺陷點數發生變異時，工程人員能即時辨認出變異來源。本研究首先進行檢定晶圓表面上缺陷點數是否呈現隨機分佈，若缺陷點之分佈不隨機時，則合併具群聚現象之缺陷點以消除群聚現象，然後再進行監控同一批晶圓間的變異以及不同批量間的變異。應用本研究提出之流程可以快速辨識變異的來源。本研究最後以模擬之晶圓數據及新竹科學園區之實際晶圓數據來驗證本研究方法的實用性。

Abstract Wafer quality is an important factor for companies’ competitive advantage in the Integrated circuits(IC) industry. The stability process is vital to the quality of wafer in IC. The c-chart is employed to monitor the stability for the wafer quality in statistic process control (SPC). The cluster phenomenon of defect appears on wafer surface becomes more pronounced as the wafer size increases. However, the conventional c-chart does not consider the cluster phenomenon, the wafer manufacturers utilize the c-chart to monitor defects would result in too many false alarms. Therefore, the cluster phenomenon should be monitored on the wafer surface before monitoring between wafer variations. There are various sources of quality variations appeared wafer in the wafer production process. Besides lot-to-lot variation, the wafer variation may also be caused by wafer-to-wafer variation. If the source of variation cannot be identified correctly, the quality of wafer will not be maintained at a certain level. Therefore, the purpose of this study is to develop a control procedure, which uses defects on the wafer surface as the quality characteristic to effectively identify the variation source. The study uses defect on wafer surface as a quality characteristic and proposes a procedure of identifying variation source of wafer defect. At first, a random test is performed to check whether the defects scatter randomly on the wafer surface. If the defects are not randomly distributed on a wafer, the clustered defects will be combined to monitor the wafer-to-wafer variation and lot-to-lot variation. The proposed procedure can effectively identify the source of variation of wafer defects. Finally, this study uses simulated and real-world cases to verify the effectiveness of the proposed procedure.