Interfacial transfer between copper and polyurethane in chemical-mechanical polishing
The interactions between a copper and urethane polishing pads were characterized to investigate the effects of friction on removal mechanisms of a polishing system of copper interconnect wafers in water. In-situ characterization of polished copper and urethane were conducted using Auger and x-ray photoelectron spectroscopy (XPS) analysis techniques. These techniques pinpointed the chemical interactions immediately during polishing. Results indicated that, because of the stimulation of friction, the molecules from the pad transferred to the copper surface, and the oxidized copper surface was transferred to the urethane surface. Without friction, however, such a transformation did not occur and passivation of the copper surface took place. This evidence proves a possible new chemical-mechanical polishing (CMP) mechanism. In addition to the formation and removal of a passivation layer, a transformation layer is formed during CMP because of friction stimulation. This layer is found on both copper and pad surfaces with different chemical bonds. Understanding the transformation layer helps to understand the formation of defects, pad conditioning, and pad life.
Key words: Cu CMP, electrochemistry of Cu, tribology, nanoabrasive particles, Cu passivation, Cu oxidation, friction-stimulated chemical wear Chemical-mechanical polishing (CMP) is a synergetic-planarization process that undergoes kinetic combination of three different components: wafer, slurry, and pad. It is a complicated system that involves more than one mechanism. It has been well accepted that the tungsten-metal CMP is based on the formation and following abrasion of a passivation layer.1 This mechanism is extended to other metal CMP, such as copper and aluminum. There are reports found indicating that this might not be the only case.2 A previous study extrapolating and estimating existing data concluded that abrasive particles abraded a soft surface layer from wafer surfaces.3 Copper oxides exist during CMP,4 and the friction depends on the type of oxides.5 In the past, we have proven that copper CMP does not reach the hydraudynamic regime during polishing.6-9 This means that the polishing pad always contacts the copper through asperities.
The motivation for this study is to pinpoint the chemical interactions caused by mechanical stimulation during polishing. We focus on the interactions between copper and urethane to investigate the effects of friction. This can only be done in a system that includes a friction test and surface analysis instantaneously. At surfaces against one another, different chemical interactions take place.10-12 A tremendous amount of work has been done on lubricating additives in studying the tribochemical wear.13 To distinguish the chemical reactions from the friction-stimulated reactions, a polishing experiment was performed in a vacuum with a controlled amount of water-vapor pressure. Under a vacuum, the sensitivity of the surface analysis is as high as a few nanometers in depth. We evaluate the friction value and surface-bonding nature. Such a study shall bring insights of removal mechanisms of copper and pad behavior in conditioning and polishing.
EXPERIMENTAL
Polishing experiments were conducted on a pinon-disk tribometer located in a vacuum chamber, as shown in Fig. 1, with well-controlled environments and surface analysis tools attached. This system was designed to conduct in-situ surface analysis that cannot be otherwise obtained. As shown in Fig. 1, the system includes two chambers in which a pressure as low as 10^sup -8^ mbar can be achieved. The tribometer is located at the center of the chamber and the surface analytical tools are at the top. Partial pressures of pure gases, water vapor in this study, can be introduced into the main chamber using a leak valve. The pressure can be controlled in the range from 10^sup -7^-10^sup 4^ mbar. The gas concentration is monitored using a residual-gas analyzer.
Surface-analysis equipment was attached to the vacuum chamber for Auger electron spectroscopy (AES) and the x-ray photoelectron spectroscopy (XPS). The AES has an electron probe sizing down to 0.5 (mu)m in diameter. The XPS is performed with a nonmonochromatized source and has a dual anode for MgK(alpha) or AlK(alpha) irradiation. The area of the x-ray is around 1 cm^sup 2^. The electron spectrometer (VS 220i) includes a set of inlet and outlet lenses and an energy analyzer. The electrons emitted by the surface of the sample during AES or XPS analysis are collected by the inlet lenses, energy-filtered by the analyzer, and detected by six-channel electron multipliers.
Ion etching was conducted for cleaning before each test. The Ar ion beam was scanned with 100 tim in diameter. A scanning electron microscope (SEM) was generated using an electron gun and the secondary electron detector, with a lateral resolution near 0.5 (mu)m. The analysis is conducted inside and outside the tested area to study the effects of friction.
The initial vacuum of the chamber was 10-8 mbar (10^sup -6^ Pa). At the beginning of the test, the copper was ion etched to remove the existing oxide layer. This was confirmed with the Auger analysis. Water vapor was then introduced into the chamber so that the vacuum reached 10^sup -6^ mbar (10^sup -4^ Pa). Under a vacuum environment, water molecules were introduced into the chamber as a vapor phase. A couple layers of water molecules were estimated to have formed on the copper and urethane surfaces uniformly.
The copper material has a purity of 99.999 wt.%. It is cut into a 5-mm-diameter pin with a sphere head. The pin is fixed on a holder and moves back and forth on a polyurethane polishing pad. The size of this disk is 8 mm x 12 mm x 3 mm. The pad was not conditioned before it was put into the chamber in order to observe the surface-change reflecting friction. The disk was cleaned and soaked with deionized water and dried in air. It continuously dries under vacuum. The polishing pad was fixed on a vertical shaft attached to an XYZ manipulator. The pad moves in a linear-reciprocating motion. The pad speed was fixed at a slow speed of 0.5 mm/sec to avoid frictional heating. The applied force was 1 N, which is relevant to the pressure of 542 Pa . The speed was 500 jim/sec; the travel length for reciprocal motion was 2 mm. The total number of cycles for the tests was 600. During testing, the frictional force was recorded. The AES and XPS were performed on and off tested areas for chemical analysis.
The friction coefficient as a function of time estimated in dry and in water-vaporized (wet and in vacuum) environments is shown in Fig. 2. This figure shows that the friction-coefficient value changes during sliding in two different conditions: dry vacuum at a pressure of 10^sup -5^ MPa and wet vacuum at pressures of 10^sup -2^ MPa. The open circles are data points recorded when tested in the dry vacuum and filled symbols are that in the wet environment. The frictional behavior depends on the materials and surface properties, and it is generally shown as the change at the beginning of tests and the stabilization after. According to Fig. 2, the wet surfaces produce increased friction at a low rate. The friction increases in the first 40 cycles before it reaches a maximum. The friction in dry conditions increases immediately after the test starts and after eight cycles reaches a maximum. In three tests, the friction coefficient lies in the range of 0.6-0.75. As shown in Fig. 2, the friction is relatively stable for the dry environment, and a few water molecules were introduced. The large variation of friction is in the middle of three potential causes: the softening of the urethane pad surface, the change of the pad surface roughness, and the change of the copper surface. It is known that urethane is sensitive to water.14 When water changes the bonding structure of the urethane surface, the urethane’s shear stress reduces. This change will affect the change in friction.