Modeling and control of material removal and defectivity in chemical mechanical planarization

Chemical Mechanical Planarization (CMP) is a necessary step in semi conductor manufacturing. Since its introduction it has been able to provide better local and global planarization. CMP has found applications in emerging technologies such as shallow trench isolation, damascene technologies. As device size shrinks CMP has become increasingly prominent.;CMP process has been analyzed at different length scales such as particle, feature and wafer scales. Models have been developed for each scale. The models initially have been deterministic, accounting for material removal by a single isolated particle. These models predicted the quality control parameters such as material removal rate and planarization at global scales. Recently, probabilistic models have been developed to describe material removal rate from particle scale. However there do not exist models which capture the interaction between parameters at different length scales.;The focus of this thesis is to develop a multi-scale model that considers the interaction between parameters at different length scales. The interaction between macroscale property pH of the slurry on the micro-scale phenomenon such as particle agglomeration has been studied. Interaction between pad asperity distribution and particle diameter distribution has also been studied.;In the existing probabilistic models the effect of asperities on the material removal has been studied assuming that the pad asperities are supported on a rigid base. In this thesis the pad cellular structure under the asperities is considered to predict scratch performance and material removal rate.;The effect of pad structure and slurry pH on scratch propensity and wafer scale material removal rate is studied. The model prediction of material removal rate has been validated against experimental data. The predicted linear dependence of material removal rate on pressure has been verified. The underlying cellular structure of the pad has been found to have little impact on the dependence of material removal rate on pressure.;Defectivity, as defined by scratch propensity has been studied in this thesis. Scratch performance has been found to be effected mainly by the proximity of slurry pH to the isoelectric point of the slurry particles. The scratch performance has worsened as the pH of the slurry becomes closer to the isoelectric point of particles.;A parametric study has been undertaken to study the effect of pad and slurry evolution on scratch propensity and material removal rate. Based on the study, suggestions have been made to improve scratch performance and material removal rate. Aggressive pad conditioning has been found to improve scratch propensity while it also has been found to maintain material removal rate, which decreases if the pad is allowed to evolve. On the other hand, pad evolution has been found to alleviate scratch problem.;Based on the insights gained in the parametric study, few suggestions about operating conditions have been made. It has been suggested to keep the slurry pH away from the isoelectric point of particles to alleviate scratch propensity. It is suggested that the pad be conditioned when the wafer surface is away from the target profile and the pad be allowed to evolve when the wafer surface is closer to target profile. This strategy maintains the material removal rate during the beginning of the process while avoiding the killer defects towards the end of the process.