Hello, everyone. My name is Shin Changjun. I'm a Process Support Engineer from Applied Materials China, West Kong. Charles Chen is my co-presenter. Today, our topic is yearly implement of a MOSFET product based on JMP WIW non-uniform analysis. This paper focus on MOSFET product, which was treated with metal deposition due to metal film or stress value variation may induce a Vth shift, which caused device yield loss. DOE test were carried out to identify post metal deposition treatment process on the Vth shift include DOE number one, baseline condition, number two, number three, number four. Vth results showed DOE number one, the baseline condition, Vth was NG, and the other splits are okay. Why Vth was NG upon DOE number one, the baseline condition? Next, a JMP analysis software was adopted to identify the correlation of a bow and stress value with Vth. First, JMP data transformation. Three types of packaging non-uniformity metrics were used: radial non-uniformity, planner or angular non-uniformity, and the off-center non-uniformity.

Then, for DOE for stress and bow value were transformed into a descriptive statistic. The result show that bow value non-uniform variable DOE number one, the baseline condition was the worst one, which may be related to Vth shift.

So next, we mainly focus on bow value analysis. Firstly, we did a bow ANOVA analysis for radial. 6 radio, zero to C5 of VAPR DOE number one, the baseline is different from each other, according to ANOVA analysis. In addition, with this equal variance, normal distribution and the independence, the three indexes are on match ANOVA criteria. Secondly, we did bow variability chart analysis. We for both, variability chart was conducted based on a radius, angular, and pair. The result show bow radial C4 and C5, the region of a V4, DOE number one, based on condition was the worst one, which may be a key factor caused by the Vth shift. Thirdly, we did data mining multivariate correlation analysis. At the beginning, we verify if bow value is proportional to wave order. The result show bow value This is confirmed a new correlation with wave-range order. Then we did data mining. Data mining multivariate correlation analysis confirmed. We were doing number one, the condition, bow radial non-uniformity contributes the most to overall saturation.

According to the above bow value analysis, we summarize the baseline condition Vth shift field model. We propose from intrinsic or meta-deposition and then ball to post-metta-deposition treatment.

Do it number one, baseline condition yield loss attributed to bow value. Overall, standard vision was worst one, especially at ratio C4 and C5. Even post-treatment about the DOE number one condition leads to standard deviation of we propose a ratio C4 and C5. Even post-treatment about the DOE number one condition leads to standard deviation of we propose a ratio C4 and C5. Because of the Vth degradation . Finally, we use hardware and recipe CIP, so the issue of low yield. In conclusion, in this topic, JMP analysis of variables adopted to identify correlation of a bull and stress value with Vth. Three types of pattern non-uniform metrics were used in JMP for bull and stress analysis, include radial, planar or angular and the over-center are non-uniform. Bull and stress raw data was transformed into a disruptive statistic. Defined in-way for non-uniform criteria, conducts a lower analysis, including eco-variants, normal distribution, and independent violation models. Carry out variability chart analysis, based on radial, angle, and pair.

Utilize this mining, multivariate correlation analysis to explore the root cause of the issue. Finally, great thanks to my mentor Charles Chen and Yang Jiayi for their mentoring. Thank you to my manager, Hank Qian, Tony Lu, Kang Jian for support me to this program.

Thank you to Applied Materials China West MDP PSE team great support. That's all for my presentation. Thanks for your listening.