Figures 6 & 7. Results are shown from a simulation of copper polishing in a hydrogen peroxide-glycine solution containing no particles. The drawings show concentration contours of glycine in a small cavity on the pad of a height H = 30 micrometers and of the same width, surrounded on either side by 15 micrometer long regions where the top of an asperity is separated from the wafer by a distance of 0.6 micrometers. This distance would normally be determined by the size of the particles if they are present. Chemical removal occurs over the entire wafer surface. The drawings show concentration contours of glycine with time. The concentrations are normalized based on the concentration of glycine in the fresh slurry. The symbol Re represents the Reynolds number, and the symbols D, a, and l correspond to geometric parameters.

Physical Modeling of Patterned Wafer Cleaning

Professor Ahmed Busnaina and his group are investigating the cleaning of surfaces and submicron deep trenches, which pose a tremendous challenge in semiconductor manufacturing. Following many BEOL (Back End of Line) processes, contaminants or chemicals can remain in the trenches. According to the international technology roadmap for semiconductors, industry will face the challenge of cleaning 100 nanometer trenches with high aspect ratios (30-60) in the next five years. However, it is difficult for SC-1 to remove contaminant from deep micron scale trenches by following typical cleaning processes.

Professor Busnaina and his group have been working on the physical modeling of the rinsing of submicron trenches on patterned wafers using pulsating flow. Simulation of oscillating flow past a series of cavities has been verified and shows excellent agreement with the numerical and experimental results of Perkins. Compared to steady flow rinse, oscillating flow rinse shows a significant advantage in cleaning and rinsing of patterned wafers. See Figure 5. The vortex oscillation and breakup mechanism significantly enhances the mixing in the submicron trenches. The best cleaning efficiency depends on the frequency, trench size and velocity . For the different size cavities considered, optimum Strouhal numbers were the same at St=0.133. Rinsing of smaller cavities requires a higher frequency oscillating rinse flow.


Modeling of CMP

Professor Ahmadi and his students are developing an abrasive particle scale model for describing the chemical-mechanical polishing and planarization processes. The forces, acting on each abrasive particle in contact with the wafer, are evaluated and the effects of both abrasive ware as well as adhesive ware mechanisms are included. The developed model also accounts for the adhesion force between the particle and wafer. The results show that the removal rate follows the Preston equation if the pad has a random roughness, while sublinear removal equations follow for pads with a wavy roughness pattern.

Professor R. Shankar Subramanian is working on various aspects of modeling of Chemical Mechanical Polishing. He is interested in predicting overall removal rates from blanket wafers, and phenomena such as dishing and erosion in patterned wafers. Along with graduate student, Lu Zhang, he recently developed a detailed transport model in which the convective and diffusive transport of chemical species in the pores of the polishing pad is accommodated using a two-dimensional repeating cell description. The appropriate conservation equations for momentum and species, along with the boundary conditions, are solved numerically in this approach. The model can predict the removal rate as a function of relative speed between the wafer and the pad and geometric parameters. Abrasive removal of material can be accommodated in the transport model as well. He and graduate student Lu Zhang have recently used the model to predict copper removal rates in a Strasbaugh 6CA polishing tool by a slurry containing hydrogen peroxide and glycine. See Figures 6 and 7. They compared these predictions with experimentally measured rates. The model correctly predicts the trend of the removal rate plotted against the relative velocity between the wafer and the polishing pad over a range of glycine concentrations, displaying a non-linear dependence on the velocity. The removal rate is small at low relative velocities and approaches an asymptotic rate at larger values of the relative velocity that corresponds approximately to purely chemical removal in a well-stirred reactor. The work has been submitted for publication.

Professor Subramanian also is interested in modeling the process by which mechanical removal of material occurs at the microscopic level. Here, the issues are the role of the mechanical properties such as the relative hardness of the wafer, abrasive particle, and the pad, the role of asperities on the pad, and the coupling of the chemistry to the mechanical removal process. In collaboration with graduate student, Lirong Guo, he is exploring mechanical removal rates from copper disks in a Struers polishing tool as a function of operating conditions.

Professor Subramanian is collaborating with Professor Don Rasmussen and graduate student Saurabh Jain, on a project in which the mechanical and wetting properties of polishing pads are being explored. In this work, contact angles are measured for a variety of slurries on the polishing pad, and the hardness of polishing pads is measured using a nanoindenter. By characterizing the pads with respect to these properties which are important in determining their polishing behavior, it is anticipated that polishing pad designs can be improved.


Go to Page