郭瑞祥2006-07-252018-06-292006-07-252018-06-292002http://ntur.lib.ntu.edu.tw//handle/246246/3065本研究針對製程非均勻度做系統性的研 究。運用合理分群的方式找出影響非均勻度的 自然變異以及特殊變異,並利用系統性的分 析,提出改善非均勻度的方法。合理分群的主 要目的,乃是根據製程的特性,藉由統計的方 法,將資料作系統性的抽樣與分群,期能將製 程發生特殊變異與自然變異分離開來,這樣的 變異分離提供了下列的好處(1 )增加特殊變異 被偵測出的敏感性,(2 )針對不同變異來源, 便於施予不同的消除對策。依據上述的目的, 在本研究中,我們以快速熱氧化製程中晶圓表 面沉澱薄膜之厚度為研究載具,將非均勻度的 統計模型依其變異來源加以分解後提出三種不 同的非均勻度模型1 NU ,2 NU ,3 NU 。其中, 1 NU 代表晶圓表面整體的非均勻度,2 NU 代表 將晶圓表面分區之後,各區內的粗糙程度, 3 NU 代表各區厚度之間的非平坦度,且觀察出 其相互之間的關係為 2 3 2 2 2 1 NU NU NU + » 。此 外,更利用了數理及模擬分析的方法,驗證了 我們所建立之非均勻度之統計特性,並且針對 其不同特性,提出改善的對策。In this research, we focus on the process non-uniformity and its variation causes. We apply the concept of rational sub-grouping to the systematic search for the special causes and propose methods for non-uniformity reduction. The advantages of rational sub-grouping are twofold: (1) higher sensitivity for special causes (2) effective classification of variation sources. We used the oxidation layer’s non-uniformity of a rapid thermal process (RTP) as an example. By using the rational sub-grouping method, we develop three non-uniformity statistics: NU1, NU2, and NU3, to represent different variation sources. Among the three statistics, NU1 represents the overall non-uniformity and is a statistic usually adopted in practice. We also propose two novel statistics. NU2 represents the surface roughness while NU3 is an indication of the surface unevenness. We find the relationship among the three statistics to be: 2 3 2 2 2 1 NU NU NU + » In addition, the statistical properties of the three statistics are analytically investigated and their properties are further verified using simulations. Results provide in-depth information for process improvement strategies.application/pdf332906 bytesapplication/pdfzh-TW國立臺灣大學工商管理學系非均勻度合理分群變異分解non-uniformityrational subgroupingvariation decomposition製程非均勻度的多重變異分解與合理分群之研究Multiple Variation Decomposition and Rational Sub-grouping on Modeling Process Non-uniformityreporthttp://ntur.lib.ntu.edu.tw/bitstream/246246/3065/1/902416H002030.pdf