Slurry Assay and Flow Characterization in Chemical Mechanical Planarization

Abstract

This thesis includes three studies which, even if independent of one another, have the same underlying goal: optimization of slurry use in chemical mechanical planarization (CMP). Slurry is one of the major concerns in CMP, both for economical and environmental reasons. On the one hand, it represents more than 50% of the cost of ownership of the process, which translates into more than $1 billion per year. On the other hand, it makes waste management very complicated and costly, since it contains nanoparticles and chemicals that are highly hazardous and traditional wastewater treatment is not effective. Therefore, the studies presented in Chapters 4 to 6 aim to optimize slurry distribution on the pad and slurry in-situ monitoring to optimize its use and reduce its consumption. In the first study, a method to visualize and quantify the slurry bow wave that forms at the leading edge of the retaining ring during polishing is developed. This technology relies on high-speed videography and image analysis, and it overcomes many limitations of other slurry visualization methods. Previous work was performed using UV light, so it needed a dark environment and black pads and it suffered from photobleaching. Our technique avoids all those problems: it can be used with common industrial equipment in normally lit environment and it provides real-time data, with a data acquisition speed of 240 frames per second. We apply our method to polishing experiments with different flow rates and we are able to quantify the slurry bow wave width. Two different regions on the pad are analyzed, one closer to the center and another one closer to the edge. The bow wave width in the region closer to the edge is not significantly affected by changes in flow rate. We believe that in this region most of the slurry has already been transported through the slots or to the waste stream, so the bow wave is governed by a hydrodynamic boundary layer on the retaining ring wall, which remains approximately constant. However, in the region closer to the center, the bow wave is formed due to the excess slurry availability in this region, and its width changes under different conditions. After obtaining the spectral fingerprint, we can successfully attribute the bow wave fluctuations to kinematics of the system. The second study is a continuation of the previous one, where we employ the new method we have developed to analyze the differences in bow wave (i.e., in slurry distribution) under different polishing conditions and with different consumables. As such, we perform experiments with different retaining ring slot designs, conditioning disc designs and conditioning modes (i.e., in-situ vs. ex-situ). A retaining ring with rounded slots results in wider bow wave than a ring with sharp slots. When in-situ conditioning is employed, four peaks corresponding to the conditioner sweeping frequency are observed, which are not present using ex-situ conditioning. A conditioner disc with vanes on its surface results in a thinner bow wave than a full-faced conditioner disc. We are able to explain all those variations based on the slurry availability and flow dynamics in each case. As in the previous study, we also obtain the spectral fingerprint to elucidate the root causes of the observed bow wave fluctuations. The third study focuses on slurry monitoring and it investigates the possibility of introducing refractive index (RI) measurements to obtain a better control on slurry density and composition. Slurry storage, handling and blending often introduce changes in composition and can lead to density non-uniformity, which can impact CMP performance. Commonly employed densitometers are often not precise enough to detect those changes and can potentially introduce contamination in the slurry due to metal corrosion. We study three industrially employed silica-based slurries and find a very strong linear correlation between density and refractive index, with regression coefficients in the excess of 0.95. When we compare the approximate limit of detection of both instruments, the refractometer can detect changes in slurry concentration that are up to 50% smaller than the densitometer. The introduction of RI measurements to monitor slurry properties could lead to an earlier detection of slurry changes in composition and improved performance. Overall, the work included in this thesis contributes to the effort my research group has been doing to develop methods and consumables that reduce COO and EHS impact in CMP, focusing this time on the concerns derived from slurry use.