Abstract
As the number of metal levels and the wafer size increase, the global planarity and effective removal of metal overlay across the wafer becomes more crucial. Chemical-mechanical polishing (CMP) has been recognized essential to achieve this goal. Accurate in situ endpoint detection and monitoring method significantly improves the yield and throughput. Previous methods have been proposed, which either require the rearrangement of the machine set-up, or can only be implanted on certain types of machines. In this study, a model for pad temperature capable of predicting the endpoint of CMP in situ is established based on the total consumed kinematic energy between wafer and pad. Limited assumptions of thermal and kinematic conditions are made. The model of temperature rise uses the integral of the relative polishing speed and is verified by on-line measurement. Since the coefficient of friction between the pad and dielectric layer is distinguishably lower than that between the pad and the metal layer, the pad temperature increases milder than polishing the metal layer. In use of the proposed regression method applied to the measured temperature rise, the endpoint of the process can be detected.