Surface roughness is a key indicator of defectivity in the semiconductor and display industries because it can degrade the electrical performance of devices and affect their lifetime. The causes of surface roughness process failure may vary during processing, but the lack of an effective model that can comprehensively evaluate and analyze surface roughness indicators has made it a huge challenge to improve the process. This project aims to establishJMP analysis packages link surface roughness measurements to non-normal distribution models for efficient root cause analysis and process adjustments. This project used the JMP Random Simulation platform to generate roughness Z-profile data and classify them into six non-normal distribution models: (1) normal; (2) uniform; (3) heavy tail; (4) right skewed; (5) bimodal; (6) outliers (3%). Five surface roughness measurements (Ra, Rz, Rp, Rv, Rk) and eight statistical parameters describing the distribution were calculated. Different clustering methods, including Hierarchical Cluster, K-means Cluster, Normal Mixtures, Cluster Variables, Multivariate Correlation, Principle Components Analysis , and Model Driven Control Charts were compared to see if they could effectively classify the six non-normal distribution models into light-tailed clusters and discrete points, and classify the 13 variables into peak clusters, asymmetric clusters, and light-tailed clusters. The relationship between surface roughness measurements and non-normal distribution models was effectively distinguished and predicted. The JMP Surface Roughness Data Mining project can effectively detect the causes of process failures, thereby reducing process debugging time. In addition, we have built a strong cross-functional JMP surface roughness analysis team within the company to build databases (roughness, raw Z- profile, roughness measurement, process adjustments) and develop predictive models for the entire Applied Materials company .
