Statistical and Experimental Analysis of Correlated Time-Varying Process Variables for Conditions Diagnosis in Chemical–Mechanical Planarization

Abstract

During chemical-mechanical planarization (CMP) of semiconductor wafers, chemical and mechanical process variables are strongly correlated and jointly affect polishing performance. The correlation among these process variables could potentially be utilized to characterize process conditions for the purpose of diagnosis. However, process variables measured during CMP, such as the temperature distribution and coefficient of friction between wafer and pad, vary with time and present in a functional form. This significantly increases the complexity of analyzing correlation patterns and relating them with process conditions. The focus of this paper is therefore twofold: 1) experimental investigation of the correlation between sensing process variables and the implication of correlation pattern changes on process conditions and 2) statistical analysis of correlation patterns between process variables in functional form. In the designed CMP experiment, we investigated two failure modes during CMP process: pad failure and slurry failure. Slurry failure was generated by reducing the percentage of oxidizer to investigate its effects on polishing performance and heat generation on the pad. Pad failure was due to variation of diamond abrasive sizes in the conditioner. The post-CMP study of nonuniformity and defects such as scratches on the wafer was conducted to characterize process conditions. The experimental and statistical results support the investigation of correlation among process variables for condition diagnosis.