A Three-staged Controlling Method to Improve the Single-Wafer Process
|關鍵字:||晶圓內變異;晶圓間變異;晶圓批量間變異;線上的品質;調整損失函數;適應性管制圖;Within-wafer variation;wafer-to-wafer variation;run-to-run variation;On-line quality;adjusted loss function;adaptive control chart|
|摘要:||在複雜的晶圓製造過程中，品質變異的來源有三種：同片晶圓上不同量測位置的變異(within-wafer variation)、同生產批量內晶圓與晶圓間存在的變異(wafer-to-wafer variation)以及不同批量間存在的變異(run-to-run variation)。目前多數半導體商僅使用靜態(static)的 管制圖，即採用固定(fixed)抽樣樣本數與抽樣間隔的管制方法來檢測晶圓製造的過程是否在管制範圍內，但隨著科技的進步，在品質管制技術與成本都需改進的情況下，晶圓製造商無不盡力提升其生產線上的品質(on-line quality)，以提供品質穩定且交期準確的產品給顧客。本論文利用一個能同時監控平均數與變異數的調整損失函數(adjusted loss function, AL)， 並結合根據變動(variable)抽樣樣本數與抽樣間隔之概念，發展出一套適應性管制圖(adaptive control chart)，此管制流程不僅能有效地監控晶圓的三種變異來源，且能同時管制樣本平均數與變異數，此外動態的抽樣方式較靜態之抽樣方式更能為企業節省大量的抽樣成本。本論文最後以模擬之晶圓資料來說明本流程，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偏移的變異來源。|
In order to implement control charts effectively in Integrated Circuits(IC) industry, the identification of various sources of variation is necessary. There are three sources of variation for the single wafer manufacturing process：within-wafer variation, wafer-to-wafer variation and run-to-run variation. Currently most semiconductor industries only use static control chart, for example, chart, which uses a fixed sampling size and a fixed sampling interval to assess whether or not the wafer production is under control. However, as current technology is in rapid progress, wafer manufacturers improve their on-line quality by enhancing their quality control techniques and lowering the costs to provide consumers reliable products. Therefore, the objective of this study is to utilize the adjusted loss function, AL, to monitor sample mean and sample variance simultaneously. Further, an adaptive control chart is a variable control chart which can adjust sampling sizes and sampling intervals according to the information obtained from the previous samples. Consequently, the study proposed a three-staged control process which combines the AL and the adaptive control chart. The proposed procedure can monitor the three types of wafer variations effectively. Also, it can save a lot of sampling cost and time by simultaneously monitoring the sample mean and variance through using adaptive sampling methods. Finally, some simulated cases of wafer quality characteristics are utilized to demonstrate the effectiveness of the proposed procedure.