監控晶圓表面量測值區塊變異之流程

Abstract

隨著晶圓面積逐漸增大，晶圓上的量測值會因複雜的製造過程，在晶圓表面上出現不同的區塊現象，即同一區塊內之量測值差異小，不同區塊間之量測值差異大。工程師在得到晶圓上量測值的區塊分佈圖後，便可依照此晶圓表面量測值的區塊分佈結果，針對各區塊某些具代表性的量測點數目與量測位置進行品質管制。實務上，目前積體電路公司在管制晶圓品質時最常使用修華特管制圖(Shewhart control charts)，但此管制圖之抽樣方式是假設每片晶圓量測值區塊之分佈非常相似，然迄今在中外文獻中尚未找到有任何偵測晶圓量測值區塊變異的方法，因此本論文即是針對已經取得之晶圓區塊分佈的歷史資料，提出兩套不同方法來監控晶圓表面量測值的區塊分佈變異。其中之一為利用區別分析法(Discriminant Analysis)來對晶圓表面量測點進行分類，建立一個區別函數(Discriminant Function)以決定某特定量測位置的量測值所屬之區塊，然後再針對新量測之晶圓資訊，依晶圓區塊相似度 (Wafer Block Similarity)來提供晶圓內區塊分佈是否發生變異的訊息。另外一個方法為利用多變量指數加權移動平均(Multivariate Exponentially Weighted Moving Average, MEWMA)管制圖來監控晶圓區塊分佈之變異。本論文最後以模擬之八吋晶圓的資料來驗證本研究所提之兩個方法確實有效可行，並分析兩方法之優缺點及建議其使用之時機。

As the size of the wafers increases, the process complexity results in the measurement blocks on the wafer surface. Measurements in each block are similar to each other, but different from the measurements of other blocks. According to the map of measurement blocks on wafer, engineers can make decide to determine the number and positions of some representative measurements on wafers. The Integrated-circuits(IC) manufacturers often adopt Shewhart control charts to monitor wafer manufacturing process under the assumption that the locations of measurement blocks are similar for each wafer. Techniques for detecting the variation of measurement blocks on wafers are not developed yet. This study proposed two different procedures for monitoring the variation of measurement blocks on wafers based on the historical data. One method is to use the Discriminant Analysis to construct a Discriminant Function to classify the measurements which belong to the same blocks. The variation of measurement blocks on wafer surface is significant if the pattern of the wafer block is changed. Another method is to use Multivariate Exponentially Weighted Moving Average (MEWMA) control chart to monitor the variation of measurement blocks on wafers. Finally simulated cases of wafer measurements are utilized to demonstrate the effectiveness of the proposed procedures.