Statistical Experimental Methods used to Obtain Edge Control of 300mm Oxide Wafers During Chemical Mechanical Planarization (CMP)

摘要

Statistical experimental methods are a structured approach to experimentation that provides investigators with a powerful tool for maximizing the benefits from experimental projects. These methods have been used in a variety of applications and are most useful in cases where no definite theory exists about the relationship between factors and responses; thus, an empirical approach must be used. The statistically evaluated experiment discussed in this paper was conducted using 300 mm blanket oxide wafers with the objective of obtaining a model for edge control (outer 10 mm) post oxide CMP. An augmented experiment was conducted with 5 factors: ratio of membrane to retaining ring pressure, slurry flow, conditioning down force, head rotation rpm, and platen rotation rpm. Then, the factors were reduced to the four factors that influence the edge range. Next a central composite design was done to investigate these four significant factors, from which models for edge range, removal rate, and % non-uniformity were generated. A verification experiment was run at levels obtained from the analysis of the central composite design.