Inho Yoon

A Mixed-Lubrication Approach to Predicting CMP Fluid Pressure Modeling and Experiments

Chemical mechanical polishing (CMP) is a manufacturing process used to remove or planarize metallic, dielectric, or barrier layers on silicon wafers. During polishing, a wafer is mounted face up on a fixture and pressed against a rotating polymeric pad that is flooded with slurry. The wafer also rotates relative to the pad. The combination of load on the wafer fixture, relative speed of rotation, slurry chemistry, and pad properties influences polishing rates. Prior work has shown that an asymmetrical subambient pressure, which exceeds that expected from the applied load, can develop at the interface between the fixture and a plane pad. The spatial distribution of this pressure can be measured and then simulated using a specially designed fixture with water as the slurry. A mixed-lubrication approach to modeling the fluid pressure was developed by including the contact stress, frictional behavior, and fluid film thickness. For a given fixture/pad separation, the contact stress can be determined using a Winkler model approximation. The film thickness can be approximated as the distance from the fixture surface to the mean asperity plane. Once the fluid film thickness is known, the fluid pressure can be determined from the two-dimensional polar Reynolds equation using finite-differencing. The theoretical pressure solution was found to match the experimental pressures when the system of forces and moments were balanced. The iterative secant numerical method was employed to compute the appropriate fluid film thickness that accommodates a balanced system of forces and moments produced by the fluid/solid interactions. After the fluid pressure is determined from an initially assumed separation, all shear and normal forces are computed from the solid contact stress and hydrodynamic fluid pressure. The results agree with the experiments.

Tilt and Interfacial Fluid Pressure Measurements of a Disk Sliding on a Polymeric Pad

Previous experimental work has shown that negative fluid pressure does develop at the disk/pad interface during chemical mechanical polishing. However, these studies dealt with one-dimensional measurement and modeling. To better understand the problem, two-dimensional pressure mapping is carried out. In addition, the orientation of the disk is measured with a capacitive sensing technique. Results reveal a large negative pressure region at the disk/pad interface that is skewed toward the leading edge of the disk. The disk is also found to be leaning down toward the leading edge and toward the center of the pad. A mixed-lubrication model based on the Reynolds equation and taking into account the disk orientation angles has been developed. Modeling and experimental results show similar trends, indicating the tilting of the disk as a dominant factor in causing the negative pressure phenomenon.

Wafer-Bending Measurements in CMP

Wafer curvature, contact pressure, and film stresses have been a subject of interest to many researchers who are working on the modeling of within-wafer nonuniformity in chemical mechanical polishing (CMP). Wafer shape and film stresses prior to and after CMP have been measured before in order to correlate film stresses with removal rate distribution across the wafer. In this paper we describe measurements of wafer bending under static loading and in dynamic conditions. A finite element analysis is also carried out to model wafer bending under static loading.

Pad Soaking Effect on Interfacial Fluid Pressure Measurements During CMP

Pad Soaking Effect on Interfacial Fluid Pressure Measurements During CMP Prior work has shown that there exist a sub-ambient fluid pressure at the interface between a rigid flat and the polishing pad during chemical mechanical polishing (CMP). This sub-ambient fluid pressure can have a significant impact on the polishing process since its magnitude may be similar to the applied load, depending on conditions. Further results have shown that there is a relationship between pad soaking time and the magnitude of this sub-ambient fluid pressure. This paper addresses measurements of the pad soaking time versus the magnitude of the sub-ambient interfacial fluid pressure. Experiments utilized a Rodel IC1000 polishing pad made of foamed polyurethane with average void size of 30 to 50 microns. Pad soaking tests indicated that the weight of the pad increased with soaking time due to water absorption. There is a high rate of water absorption initially before the pad becomes saturated and the mass of the pad stabilizes. It is also observed that the pad material is impermeable to water and most of the water penetrated only the topmost layer of voids in the material. These experiments suggest that the water progressively softens the top layers of the pad during the soaking and causes the sub-ambient fluid pressure to increase in magnitude. A model of the sub-ambient fluid pressure increasing as the elastic modulus of the pad decreases is also suggested.

An Analysis of Mixed Lubrication in Chemical Mechanical Polishing

Pressure and shear flow factors (Patir and Cheng, 1978) were used to take into account the roughness of the pad surface in the modeling of the interfacial fluid pressure during chemical mechanical polishing. An attempt was made to explain the physical meaning of the flow factors in this particular application. Additionally, a parametric study was carried out to see the effect on the model after the incorporation of the flow factors. The pressure and shear flow factors were found to have a competing effect on the magnitude of the sub-ambient fluid pressure.

Dynamics of the Formation of the Subambient Pressure Distribution During Chemical-Mechanical Polishing

Chemical-Mechanical Polishing is used to polish silicon wafers in the manufacturing of integrated circuits. Wafers are pressed, electronics side down, onto a rotating pad that is flooded with a slurry containing abrasive particles. The slurry is entrained in the interface and the abrasive particles slide against the silicon and polish it. Our previous work has shown that subambient pressures develop at the silicon/pad interface and we have measured this pressure and its distribution over the wafer surface (1). However, our experiments have been limited to those conditions where the pad rotates and the wafer slides on the pad but the wafer itself does not rotate. Our experiments showed a skewed pressure distribution. This paper describes experiments and pressure distribution measurements where the wafer, as well as the pad/platen is rotated (2). Specifically-designed wireless electronic transmitters and receivers were built and used to measure the interfacial pressures at the silicon/pad interface. Subambient stress maps and temperatures have been measured and Figure 1 shows an example of a skewed pressure distribution when the silicon is not rotated and Figure 2 shows the pressure distribution for the same wafer while it is rotating. The subambient pressures develop over a 2 second time period from when the rotation started. The pressure distributions are symmetric in spite of the lean and tilt of the wafers. The rotational speed and other variables have a big influence on the polishing rate and this will be discussed in the talk.Copyright

Dishing and Erosion in Cu-CMP

Chemical Mechanical Polishing (CMP) of copper in trenches and vias of patterned silicon wafers is routinely used in CMOS processes as well as MEMS applications. Although the main goals of CMP are to achieve a planar surface at the nano-scale without scratches, it is generally the case that copper is preferentially polished, a condition called dishing and erosion, relative to the pattern geometry. We have measured dishing and erosion of electroplated copper on patterned silicon wafers with specially-designed patterns containing combinations of line-width and density. One hundred millimeter diameter wafers were patterned using a standard etching process and electroplated with copper. The polishing was done on a modified laboratory-scale bench top polisher which allows ranges of normal loads and velocities. A commercial pad (Rodel IC1000 plain) and slurry were used along with a slurry-delivery rate fixed by a peristaltic pump. The pad conditioning and other process parameters were chosen to represent those used in standard industrial practice. Dishing and erosion were measured as a function of the pattern geometry and polishing conditions. The measured dishing and erosion were then compared to other models.© 2005 ASME

Nanoparticulate and Interfacial Mechanics in Confined Geometries Typical of Chemical-Mechanical Planarization

Chemical-mechanical planarization (CMP), a surface preparation process used widely in integrated circuits manufacture, is currently the leading nanoscale manufacturing process worldwide, with an annual economic impact well in excess of

MechanicalModeling of the 2D Interfacial Slurry Pressure in CMP

1 billion. Originally developed for glass polishing, CMP is used by the microelectronics industry to create silicon, silicon oxide, tungsten and copper surfaces with average roughnesses of O(10 mm). The process typically involves shearing a dilute abrasive silica or ceria nanoparticle-laden “slurry” between a compliant rough surface (the “pad”) and the surface to be polished (the “wafer”). The composition of the slurry can greatly affect material removal rates. Despite its importance, however, a lot still remains to be discovered about the fundamental mechanisms involved in this process. A multidisciplinary effort at Georgia Tech has focused upon the interfacial mechanics of this process and how nanoparticles chemomechanically wear SiO2 , Si and Cu surfaces. It has been found, for example, that the wear rate of dielectric varies approximately as the particle diameter. The entrapment of particles at the asperity/dielectric interface is thought to produce the polishing, but the exact nature of this interaction is still unknown. An evanescent-wave visualization technique has therefore been developed to visualize the dynamics of fluorescent 300–500 nm diameter colloidal silica and polystyrene particles within a particle diameter of the “wafer” surface in a simplified model pad-wafer geometry. The technique has been used for the first time to the authors’ knowledge to directly measure the velocity and concentration of the interfacial particles—which presumably interact with and wear the wafer. Although the pad speeds in these studies are much lower than those encountered in the actual CMP process, the initial results suggest that there is negligible “slip” between the particle and fluid phase velocities at the wafer surface. The number of particles at the wafer surface appears, however, to be strongly affected by particle properties, including particle density and size.Copyright