考量耗材磨耗下化學機械研磨製程之控制研究

Abstract

隨著電子元件尺寸的日益縮小，化學機械研磨（Chemical Mechanical Polishing ,CMP）是現在唯一能夠提供VLSI製程全面平坦化的技術。化學機械研磨中除了所牽扯到的參數非常多，耗材的磨耗更會使得製程產生變異，所以CMP製程看似簡單但實際技術門檻卻非常高。 本論文利用實驗設計中，區集劃分法（Blocking）以及反應曲面法（Response Surface Method, RSM）找出輸入變數與輸出變數之間的關係，藉由這些關係討論當元件逐漸磨耗時，為了維持製程最佳平坦度輸入配方所需的變化；並使用實驗結果建立批次控制的預測模型。 最後提出以本實驗室所開發之雙互質分解干擾觀測器（DCFDOB）架構改善元件磨耗對於平坦度的影響，並與MIMO double-EWMA批次控制器進行模擬比較。經過模擬驗證後得知使用DCFDOB能有效抑制製程干擾並有良好的強健性能對抗耗材磨損對於製程造成影響。

With the continuing shrink of electronic components, the Chemical Mechanical Polishing (CMP) process is the only way to provide global planarization of VLSI process. In the CMP process, not only a lot of parameters should be considered but also the abrasion of consumables will cause the process variance. Therefore, CMP process seems to be simple but the included technology is actually difficult. In this thesis, we used the Blocking method and the RSM (Response Surface Method) from DOE (Design of Experiment) to find out the association between the input and output variables with different consumable lives. With these results, we discussed the changes of optimal recipes and, furthermore, established the predictive process model. In order to decrease the variance which was caused by consumable abrasions, the DCFDOB structure was applied in this thesis. The performance of DCFDOB and MIMO double-EWMA were compared through simulations. The result showed the performance of DCFDOB was better than MIMO double-EWMA obviously.