迴歸分析法在晶圓拋光效益之應用研究

Abstract

深次微米半導體製程中，嚴格要求晶圓曝光表面的全區域性平坦化，而非區域性平坦化。至目前為止，唯有化學機械研磨技術能夠達到全區域性平坦化的層次。 以迴歸分析法應用於晶圓拋光效益，主要在於對於晶圓的磨耗率､平坦度所得之實驗數據，作有效模型的建立，驗證理論與實驗所產生誤差的。不但為驗證過去曾被提出之理論的可信度，同時也促成許多新的機構及模式的建立。 這份研究，經由田口法(Taguchi Method)的實驗設計，應用於晶圓拋光製程效用上的品質改善，藉以發掘問題之癥結並尋求確認影響製程品質的重要因子，同時建立此類要因造成晶圓邊際效應（Edge Effect)觀測缺陷的迴歸方程式。 本研究結果，由統計分析，就容許變因中，歸納得到實用的結論：(1)在不考慮邊際效應下，研磨壓力在40psi，內徑大小為7公分的壓環的研磨設計下，晶圓研磨的全區域平坦度達到最佳的水準。(2)在研磨頭上對晶圓考慮施予20 psi的主壓及3.5 psi的背壓，晶圓的邊際效應範圍能縮 短0.1~0.2 mm之間，邊際部位局部壓力的最大值則可以從30.1 psi降至 25.1psi。此歸納成果，有助於化學機械研磨製程的改善，而研究方法與 程序則可供相關製程參數最佳值訂定時的參考。

In the quarter sub-micronmeter process of the semi-conductor production. We aim to achieve the planarization of the full field severely on the exposal surface, not only the planarization of local fields. Up to this time now, only the Chemical Mechanical Polishing(CMP) technology could reach the level of the planarization of the full field. We applied the method of the regression analysis to the effect of CMP, subject to establish the better model from the experimental data of the polishing rate and the planarization on a wafer. We investigated the difference between the results of the experiment and analytical simulation of CMP from some typical types of the wafer polishing examples. The metrical data not only prove the reliability of the theories which were propound in the past, but also actuate the new gadgets and models to be established. This study is applied to prove the quality of the wafer polishing effect through the experiment design of Taguchi Method. Then find out the issues for some problems, and confirm the important factors affecting the quality in the CMP’s process. At the same time, we try to establish the regression equation of the edge effect resulting from some critical factors to observe these defects. The results of the study are compressed to obtain several practical conclusion with the statistical analysis in the permissible variables. (1) When the edge effect is not considered, the experiment could be achieved the best level of the the planarization of the full field on a wafer polishing if the polishing pressure is fixed at 40psi, and the diameter of the O-ing at 17.78 inch. (2) We considered to apply the primary pressure at 20psi and air pressure at 3.5psi on a wafer, then we can shorten the area of the edge effect reaching 0.1~0.2mm on a wafer. At the same time, the relatively maximal pressure would be degraded from 30.1psi to 25.1psi. On the other hand, edge of the wafer. These results could contribute to improve the process of CMP. The several methods and procedures of the study are expected to be references assisting to design the best variables on a relevant process.