Abstract: The influence of the components of an alkali polishing slurry and the mutual influences on the Cu polishing rate were investigated by a CMP polishing rate prediction model established with a modified artificial neural network based on the artificial bee colony algorithm. The quantitative method of sensitivity analysis was employed to fulfill the purpose of quantizing the influence on the polishing rate. The result of the analysis indicates that under certain CMP conditions, the Cu polishing rate was controlled by the silica abrasives, the FA/O chelating agent, the surfactant and the oxidant agent in the polishing slurry. Such factors showed the different sensitivity coefficients with 0.78, 0.53, 0.29 and 0.19 respectively on all the sample points. The mutual influence between the FA/O chelating agent and the oxidant agent on the polishing rate seemed obviously strongest when the proportion of them was 2 to 7, with the global sensitivity coefficients between 5 to 9; the mutual influence of silica abrasives and oxidant on the polishing rate was greater as the proportion of the above additives was beyond 5, with the global sensitivity coefficients between 2.5 and 6; the mutual influence of the surfactant and oxidant on the polishing rate was not obvious, with global sensitivity coefficients less than 3. Thus, it provides a kind of effective method for quantitating the influence with the components of the CMP alkali slurry on the polishing rate.

Key words: polishing slurry, polishing rate, sensitivity analysis, artificial neural network, artificial bee colony algorithm