UC San Diego
The effects of chemistry on the colloidal behavior of alumina slurries and copper nanohardness for copper chemical mechanical planarization
- Author(s): Ihnfeldt, Robin Veronica
- et al.
Chemical mechanical planarization (CMP) is used in integrated circuit manufacturing to remove excess material and provide a globally planarized wafer surface. The CMP process requires slurry containing nanometer-sized abrasive particles and chemical additives which produce a mechanical and chemical synergistic effect that is responsible for the material removal rate (MRR). Because copper has become the interconnect material of choice, the focus of our research is on copper CMP. The chemical additives in the slurries control the state of the copper (CuO, Cu²⁺, etc.) on the surface of the wafer and in the slurry and also affect the dispersion characteristics of the abrasives. This research investigated the influence of common additives (glycine, H₂O₂, etc.), solution pH, and presence of copper on the colloidal behavior of alumina suspensions. The colloidal behavior was characterized through measurement of zeta potential and agglomerate size distributions. The effects of common slurry additives and solution pH on the nanohardness and etch rate of the copper surface were also studied. It was found that with the addition of copper into the slurry, an increase or decrease in agglomeration of the alumina was observed depending on the state of the copper in the solution. With the addition of chemical additives and changes in the pH of the solution, the nanohardness of the copper film was observed to range from 0.05 - 20 GPa, due to the formation of different films (CuO, Cu₂O, etc.) and/or changes in the compactness of the surface film from complexing reactions or dissolution. Additionally, experimental results were incorporated into a model of CMP to predict MRR and predictions were compared to experimental copper CMP data. The CMP model accounts for the chemical activity of the slurries through the abrasive size and distribution, hardness and chemical etch rate parameters. The model MRR predictions only agreed with experiment for slurries with pH>8 and small etch rates. However, for acidic slurries and slurries with large etch rates, model predictions did not agree with experiment, most likely due to using nanohardness measured under quiescent conditions