CMP Pad Conditioning and Applications to Type 1 Impregnated Felt Polishing Pads

Abstract

Pad conditioning is the process of “dressing” the surface of the polishing pad, generally with a highly engineered diamond abrasive disc. Conditioning not only refreshes the pad surface, the cutting characteristics of the conditioner determine the structure of the asperities that are in contact with the wafer. By tuning the cutting characteristics of the conditioner, polish performance can be manipulated over a wide range. With the most advanced pads, specific conditioning protocols are typically required to obtain the designed performance. Optimization of these protocols is enabled by an understanding of the pad surface statistics.Figure 1 illustrates the two types of polishing pads that will be compared in this talk. Suba 800 is an example of a Type 1 pad. Type 1 pads consist of a “non-woven” needled polyester felt impregnated with polyurethane. The pores in a Type 1 pad are formed from the spaces between the impregnated fibers and have a length scale of 100s of microns. The pores in a Type 1 pad tend to be interconnected. VP5000 is a Type 3 pad. Type 3 pads consist of discrete pores in a matrix of polyurethane. The matrix properties can be varied over a wide range and pore sizes range from single to 10s of microns. Type 3 pads should be isotropic in terms of their physical property and porosity distribution. Type 3 pads typically exhibit Shore D hardness in the range of 20-70. Type 1 pads are measured on softer hardness scales or in terms of compressibility.Pad conditioning of Type 3 pads in semiconductor wafer CMP is at a technically mature stage, and the knowledge base can be directly applied to substrate polishing with Type 1 pads. The decoupled Gaussian-Exponential model which has been successfully used in Type 3 pads can be directly applied to Type 1 pads. As Figure 2 illustrates, Type 1 pads exhibit an exponential porosity signature, but with a decay length about an order of magnitude larger than a typical Type 3 pad, consistent with the difference in length scales of the pores of the two pad types. The core of the distribution (accounting for most of the total surface) can be modeled as a Gaussian.Figure 3 provides examples of both Suba 400 and Suba 800 and differences in both the intrinsic porosity (exponential decay constant, Eτ) and Gaussian core width are evident. At least part of the difference in the porosity signature is due to the Suba 800 receiving a double impregnation of polyurethane, compared to a single impregnation for Suba 400, resulting in a comparatively shallower pore structure. The difference in the Gaussian core width is most likely the result of a different buffing process or media used to impart a surface finish at the factory.Figure 4 illustrates the range of texture that can be developed on the DuPont Suba 800 impregnated felt pad, one of the Type 1 pads most widely used in substrate polishing applications. As the figure illustrates, the hwhm of the Gaussian core can be driven from about 10 to 50 µm, while maintaining the deep background porosity that characterizes this pad. The ability to drive the texture into rougher and smoother regimes, compared to a new pad, implies a potential to drive the substrate process into previously unattainable performance spaces. Initial feedback from the field indicates that smoother pad surfaces can contribute to improved substrate surface finish. Rougher surfaces may contribute to improved throughput through polish rate increase.Figure 5 illustrates how key pad roughness metrics of Suba 800 are correlated with a basic conditioner diamond design parameter. Most roughness measurements are highly correlated, although one of the most widely used roughness parameters, the average roughness Ra, does not correlate well at all in this case. This is an excellent example where pad height distribution modeling provides critical insight into the drivers of the textural changes that effect polish performance, and how that performance might be manipulated through conditioning design.Signatures of surface textural degradation from the stress of polishing and accumulation of debris can be observed on Type 1 pad surfaces. It is well known that restoration of the pad surface structure through conditioning is a primary mechanism of stability in Type 3 pads. Data from the field indicates improvements in process performance and consumable life with Type 1 pads when conditioning has been implemented versus processes with no pad conditioning. The examples given here will be expanded upon and additional applications will be discussed. Figure 1