In the semiconductor field, the Within Wafer Non-Uniformity (WIW NU ) of the thin film process has always been the most critical factor in quantifying thin film performance. The wafers produced by the Fab should uniformly meet the required thin film performance. However, due to the lack of a thorough WIW-NU indicator to effectively solve the problem, we are faced with a huge challenge of improvement and optimization. The existing WIW-NU processing methods can only provide basic wafer morphology distribution diagrams and sample averages and standard deviations, but cannot accurately determine the cause of the problem. Therefore, it is a strong demand of semiconductor customers and manufacturers to establish a smarter WIW-NU indicator to shorten the process cycle of solving the problem. As a powerful data analysis platform, JMP can be used to decompose the overall WIW-NU problem into four dimensions of failure modes: ( 1 ) Radial ; ( 2 ) Angular ; ( 3 ) Off Centering ; ( 4 ) Outlier . At present, the subject has been analyzed methodologically from the following aspects: transformation of raw data to create the required statistical data; definition of WIW-NU standards; hypothesized mean test, normal distribution test and normal distribution violation mode; GRR and Cpk/Ppk analysis; establishment of I-MR and Xbar-R charts to capture special changes; use of Data Mining platform to analyze the impact of different dimensions on WIW-NU ; deriving predictive models to guide process adjustments; using Query Builder , Group Scripts , Journal , and Dashboard Builder to establish a comprehensive WIW-NU database, as well as automatic JMP run programs.
