應用模糊理論構建晶圓缺陷點與群聚現象之管制圖

Abstract

積體電路（Integrated Circuit,IC）製造商在應用傳統之缺陷點管制圖（c chart）來管制晶圓之缺陷點數（defect）時，須假設晶圓上的缺陷點數會呈現隨機分佈且服從卜瓦松（Poisson）分配。然而隨著晶圓製造技術的精進，晶圓面積逐漸增大，晶圓上缺陷點的分佈會出現群聚（clustering）現象，導致缺陷點呈現卜瓦松（Poisson）分配的假設不再成立，若仍使用傳統缺陷數管制圖來監控晶圓，則會產生過多的假警報。為了解決缺陷群聚造成傳統缺陷數管制圖失效的問題，許多中外文獻提出修正之缺陷數管制圖，然而這些修正之管制圖各有一些缺失，實務上仍無法有效的監控晶圓表面之缺陷點。本研究的主要目的即是在考慮缺陷出現群聚的狀況下，應用模糊理論（fuzzy theory）建構出一個可以包含專家智慧與工程人員經驗之模糊管制圖，以有效且簡便的管制晶圓表面的缺陷數。本研究所提之管制圖經新竹科學園區某積體電路公司之實際資料驗證後，顯示此管制圖不但與良率吻合，且可以較現有之修正缺陷數管制圖更快速的判斷出異常的原因，確實能有效地減少假警報的發生。

Integrated circuits (IC) manufacturers usually employ c-charts to monitor wafer defects. As the surface area of the wafer increases, the clustering phenomenon of wafer defects becomes apparent. The defect clustering causes the Poisson based c-chart to have many false alarms. Although many modified c-charts have been developed to reduce the false alarms, these control charts still have some drawbacks for practical use. This study proposes a process control chart using Fuzzy theory and engineering experience to monitor wafer defects with clustering phenomenon. The proposed method is simpler and more effecient than the modified c-charts. A case study from an IC company is presented to illustrate the proposed method.