CMP Friction as a Function of Slurry Silica Nanoparticle Concentration and Diameter

Next generation integrated circuits (IC’s) will require the use of porous dielectric materials with low shear strengths and at the same time still require processing with chemical mechanical polishing (CMP). CMP polishes the substrate material (e.g. SiO₂) by rubbing the surface with nanometer scale SiO₂ particles. The particles are suspended in an aqueous slurry and are rubbed on the substrate by a porous polyurethane pad. The high friction of CMP can damage porous dielectric materials. This research is defining the source of this friction to enable development of CMP which will not damage porous (low dielectric constant) materials. Experiments were done to determine the contributions of the SiO₂ particles and bare pad asperities to the total friction with an SiO₂ substrate. Further experiments were done to determine the change in friction for various SiO₂ particle sizes. This system level friction data was obtained using a bench-scale tribometer at contact stresses near that of commercial CMP. Very low speeds (∼ 10⁻³ m/s) were used to eliminate confounding hydrodynamic effects. AFM (Atomic Force Microscope) and instrumented indentation experiments will be used to determine the friction contribution of a single SiO₂ particle/substrate contact. The bench-scale experiments showed that the coefficient of friction increased by as much as 10% with the increasing weight percent of SiO₂ particles in the slurry. This data suggests that a constant real area of contact is distributed between bare pad asperity contacts and higher friction SiO₂ particle contacts. As more SiO₂ particles are added, their contact area increases as the pad asperity contacts decrease thereby increasing the coefficient of friction of the system. Further experiments revealed a trend of increasing coefficient of friction with smaller particle diameter. This friction increase is consistent with theories which suggest that particle real contact area is a function of the particle cross-sectional area as opposed to a purely Hertzian model. The results demonstrate a possible route to controlling CMP friction by varying the relative combinations of the pad friction, the percentage of abrasives in the slurry, and the abrasive particle size.