No abstract

The time evolution of a semiconductor wafer's feature-scale topography during chemical-mechanical polishing applied to copper films deposited over patterned dielectrics depends strongly on the bulk copper removal rate and the time at which copper clears from the up areas in the pattern. To first order, the overall evolution scales with the copper removal rate. But after the bulk copper is cleared locally, the evolution of the underlying dielectric and recessed copper depends strongly on the local clearing time.Complexities in analyzing the evolution data can be reduced by referencing that time. A new method of characterizing featurescale planarization is described that removes the first-order clearing time and removal rate dependence from the topographical evolution graph. By doing this, the topographical evolution is understood in terms of initial topography and pad bending. The characterization is normally expected to be the same across the wafer, from die to die, and differences are attributed to nonuniform local pressures, pad condition, or other padwafer interface conditions, such as the local chemistry. Ideal cases and easily understood metrics named clearing offset, overpolish sensitivity, saturation level, and ''clearing nonuniformity'' are introduced to provide a conceptual underpinning and utility of the method. The method is demonstrated using data measured on selected Cu structures in three different die locations during polishing on an Auriga polisher with Cabot 5001 slurry.

No abstract

Non-Prestonian behavior is defined as a non-linear correlation between the removal rate and the polishing pressure. It is believed to be related to dishing, erosion and self-stopping behavior that is observed on pattern wafers. Nojo et al. has reported that “self-stopping polishing” can be achieved when the non-Prestonian behavior was observed.  Significant research effort has been made in this field, following this initial report. For example, the effect of chemical additive concentrations and molecule weight of additives, along with abrasive concentration and size on non-Prestonian behavior have been studied.  Primarily, blanket wafer performance data has been presented in the majority of published work. Although some researchers had conducted modelling to calculate the effect of non-Prestonian behavior on pattern performance, a comparison between pattern and blanket wafer performance in not available.  This is needed to enable a more clear understanding of non-Prestonian behavior. In this presentation, Prestonian behavior was modulated by adjusting chemical concentrations and / or formulation pH. Pattern performance, including oxide removal rate and step height reduction rate, were collected and evaluated.  It was determined that non-Prestonian behavior does not necessarily lead to good pattern performance. From the learnings of this systematic study, a high-performance, oxide-polish slurry for stop-in-film applications was successfully developed.