Oxide CMP mechanisms
07/01/1997
Oxide CMP mechanisms
Minoru Tomozawa, Rensselaer Polytechnic Institute, Troy, New York
Chemical mechanical polishing (CMP) is an increasingly important planarization process for microelectronic devices with multilevel integration [1]. CMP of oxides is reliable and efficient when executed properly, but like optical-glass polishing, much of the CMP process is an art rather than a technology.
For a process to qualify as a technology, its basic mechanism needs to be understood so that it can be logically optimized. Since oxide CMP is similar to optical-glass polishing, a great deal can be learned about oxide CMP by studying the literature developed over the long history of optical-glass polishing.
Optical-glass polishing
The optical-glass polishing rate, R, defined as the thickness of the glass removed per unit time, is proportional to the applied pressure, P, and the linear velocity, v, at which the pad moves relative to the optical glass specimen, i.e.
R = k Pv (1)
where k is a constant under a given experimental condition. This relation is known as Preston`s law [2] and the same relationship is commonly observed in oxide CMP.
Numerous mechanisms have been proposed for oxide glass polishing. T. Izumitani [3, 4] investigated the glass polishing mechanism by comparing the polishing rates and various physical and chemical properties of different optical glasses. The polishing rate correlated reasonably well with the chemical durability; glasses with reduced chemical durability exhibited greater polishing rate. The polishing rate did not correlate with either the indentation hardness or the softening point of a glass.
The polishing rate was dependent upon the indentation hardness of the hydrated glass surface formed by chemical reaction with water or acid. From these observations, Izumitani concluded that the optical-glass polishing, "proceeds by the formation of a hydrated layer by means of a chemical reaction between the glass surface and water and then the removal of this hydrated layer by the abrasive particles" [3, 4].
 |
This proposed mechanism is consistent with the fact that the polishing rate of optical glasses is extremely low without water present to cause the hydration [3, 4]. The hydrated layer of an optical-glass surface is usually produced by ion-exchange between hydrogen (or hydronium) ions and alkali (or alkaline earth) ions (Table 1), and generally has much lower hardness and mechanical strength. Therefore, plastic deformation of the hydrated layer results in mechanical removal when it is rubbed against harder abrasives.
Tribology
A relation similar to Preston`s law was theoretically derived in the field of tribology. The rate of mechanical wear was shown to follow an exactly analogous relation to Preston`s law when plastic deformation takes place at a point of contact [5]. The wear volume/unit sliding distance, Q, arising from all the asperity contacts, becomes proportional to a normal load, W, and reciprocally proportional to the indentation hardness, H, by
where K is a dimensionless constant called the "wear coefficient." The same relation can be obtained for abrasive wear through plastic deformation [6], where the surface is plowed by a cone (of semi-angle, a) being dragged across the surface (Fig. 1).
When both sides of Eqn. 2 are multiplied by the sliding distance and divided by the sliding time and the nominal contact area, then Eqn. 1 is derived. From a comparison of the two equations, the proportionality constant, k, in Eqn. 1 is reciprocally proportional to the hardness, H, by
Q = KW/H (2)
The finding [3, 4] that the polishing rate of optical glasses decreased with increasing hardness of the hydrated glass surface is consistent with Eqn. 3. The optical-glass polishing mechanism seems to involve the plastic deformation wear of the hydrated surface.
 |
Figure 1. Schematic diagram of a conical abrasive producing wear by plastic deformation of a planer surface; a) side view, b) top view [6].
CMP mechanisms
A similar polishing mechanism is expected for oxide CMP. The polishing rates of both SiO2 and optical glasses are much greater in the presence of water [3, 4, 11], indicating that hydration plays an important role in both materials. However, there is one important difference between the two materials. Optical glasses usually contain alkali or alkaline earth elements, which can exchange with the hydrogen (or hydronium) ions in water to produce the hydrated surface layer.
 |
Figure 2. Schematic diagram of the expected plastic deformation between an abrasive and a planar surface under static compression. Ha is the hardness of the abrasive and Hs is the hardness of the planar surface; a) planar surface deformation, b) abrasive deformation. For oxide CMP without heating, Ha = Hs and situation b) is applicable [6].
SiO2, on the other hand, does not contain any components which can exchange with hydrogen, and therefore its hydration or water entry is difficult (see "Difficulties in determining CMP mechanisms"). Since oxide CMP abrasives are also SiO2, the same material being polished, it is not likely that the oxide layer is worn by plastic deformation. Under a static compressive stress, SiO2 abrasive particles are more likely to exhibit plastic deformation than the oxide layer on a wafer (Fig. 2) [6].
Heating
Two inter-related processes, however, can lead to wear by plastic deformation of the oxide layer during CMP. One process is the heating of the wafer by friction. When SiO2 abrasive is rubbed against an SiO2 layer, even in the presence of water, the poor thermal conductivity of amorphous SiO2 will create localized heating.
Since the water in the slurry is the coolant, and SiO2 abrasive particles have a much greater surface-to-volume ratio than the SiO2 layer on the wafer, the abrasive will be cooled more effectively than the oxide layer. Therefore, during CMP, the silica layer on the wafer will be at a higher temperature than the silica abrasives. Since the hardness of silica decreases with increasing temperature [7], frictional heating leads to the plastic deformation of the oxide layer.
The temperature rise of a wafer during CMP was confirmed [8], and it was found that the polishing rate rose with temperature. The researchers also found that wafer heating during polishing was reduced when the slurry was replaced with water. This temperature reduction implies that the primary heating source is the friction between abrasives and the oxide layer. They attributed the increased polishing rate to an unspecified chemical reaction between the slurry and the oxide layer. It is more likely, however, that the softer oxide at higher temperatures [7] produces greater plastic deformation.
Surface hydration
The second contributing process is the hydration (or water entry) of the oxide that occurs during plastic deformation. Even though water diffusion is slow in SiO2 at room temperature, water readily enters the oxide during the plastic deformation caused by diamond indentation [9-11]. This complex phenomenon has been confirmed by observing a water-related band in the FTIR spectrum from an indented portion of the oxide.
The amount of water in silica increases with increasing volume of plastic deformation (Fig. 3) [10]. The commonly observed lowering of oxide hardness with indentation time, called indentation creep [12], is due to water entry into the deformed oxide during indentation. From the wavenumber of the IR absorbance signal, it appears that H2O primarily enters in the form of molecular water at room temperature [13].
 |
Figure 3. IR spectra at various locations of an indentation on silica glass showing water entry into silica during Knoop indentation under 100 g load for 30 sec in water. The absorbance scale is the same for all four locations but is vertically displaced for clarity. The diagram at the bottom indicates the approximate location on the indentation where the IR beam probed [10].
 |
Figure 4. IR spectra of silica glass samples hydrated at various temperatures under 0.467 water vapor pressure for 400 hours. The spectra are normalized by the maximum absorbance value [13].
H2O + ?Si-O-Si?????Si-OH + HO-Si? (4)
It is believed that molecular water enters into SiO2 and reacts with the silica network in the following manner
Keq = [?Si-OH]2 / [H2O][?Si-O-Si?] (5)
where H2O represents molecular water in silica glass. At high temperatures, this reaction rapidly proceeds and reaches an equilibrium defined by the equilibrium constant,
such that almost all of the water-related species exist in the form of hydroxyl (Si?OH). At low temperatures (such as room temperature) however, there is typically a large fraction of molecular water observed in silica glass, due either to the slow kinetics of Eqn. 4 or to a shift in the equilibrium of Eqn. 5. The relative fraction of molecular water increases gradually as the heat-treatment temperature decreases (Fig. 4) [13].
Water incorporation strongly influences material properties. Water generally decreases oxide hardness [11] due to increased plasticity, elastic modulus, mechanical strength and chemical durability. The relative importance of hydroxyl and molecular water in property alteration of oxides is not currently understood, but molecular water seems to have a greater influence on mechanical strength.
The likely oxide CMP mechanism is oxide hydration during the plastic deformation caused by abrasives. Plastic deformation is assisted by frictional heating, and the removal of the resulting softer (mechanically weaker) hydrated surface layer takes place by the plowing action of the abrasive particles.
Some sort of chemical mechanism (such as SiO2 dissolution) is generally considered to be a cofactor in explaining CMP. Various chemical processes have been proposed [15], and the higher polishing rate in higher pH slurry is often used as evidence for silica dissolution [16]. The pH effect does not necessarily imply a chemical reaction, however, since the change of pH can alter other properties. Hayashi et al. [17] showed that the polishing rate of SiO2 became higher when pH was reduced from 9 to 6 for a slurry containing ammonium salt. They attributed the pH effect to the agglomeration of SiO2 abrasives, caused by the reduction of the surface charge at lower pH.
Pad viscoelasticity also plays an important role in CMP. Izumitani found [4] that there is an optimum viscoelastic pad modulus for optical-glass polishing. It is probable that the polishing pad has to be soft enough to grip the abrasives, but hard enough to transmit the load to the oxide.
Conclusion
The mechanisms required to explain oxide CMP are complex. Polishing pressure and friction heat the oxide surface, and soften it. The surface is then further weakened by hydration. The oxide particles in the slurry easily scrape off the hydrated surface.
Acknowledgments
This research was supported by CAIST under contract IC-448. Careful readings of the manuscript by S.P. Murarka, S.N. Crichton, C.G. Kallingal, and Y.-L. Peng of Rensselaer are greatly appreciated.
References
1. W.L. Patrick, W.L. Guthric, C.L. Standley , P.M. Schiable, "Application of Chemical Mechanical Polishing to the Fabrication of VLSI Circuit Interconnections," J. Electrochem. Soc., 138, 1778, 1991.
2. F.W. Preston, "The Theory and Design of Plate Glass Polishing Machines," J. Soc. Glass Technol., 11, 214, 1927.
3. T. Izumitani. S. Harada, "Polishing Mechanism of Optical Glasses," Glass Technol., 12, 131, 1971.
4. T. Izumitani, "Polishing, Lapping, and Diamond Grinding of Optical Glasses," Treatise on Materials Science and Technology, Vol. 17, Glass II, ed. by M. Tomozawa and R.H. Doremus, Academic Press, New York, p. 115, 1979.
5. J.F. Archard, "Contact and Rubbing of Flat Surfaces," J. Appl. Phys. 24, 981, 1953.
6. I.M. Hutchings, in Tribology, Friction and Wear of Engineering Materials, CRC Press, Boca Raton, pp.136 and 142, 1992.
7. J.H. Westbrook, "Hardness Temperature Characteristics of Some Simple Glasses," Phys. Chem. Glasses, 2, 32, 1960.
8. F. Sugimoto, Y. Arimoto, T. Ito, "Simultaneous Temperature Measurement of Wafers in Chemical Mechanical Polishing of Silicon Dioxide Layer," Jpn. J. Appl. Phys. 34, 6314, 1995.
9. K. Hirao, M. Tomozawa, "Microhardness of SiO2 Glass in Various Environments," J. Am. Ceram. Soc. 70, 497, 1987.
10. M. Tomozawa, K. Hirao, "Diffusion of Water into Oxides during Microhardness Indentation," J. Mat. Sci. Letters, 6, 867, 1987.
11. M. Tomozawa, K. Yang, H. Li, S. Murarka, "Basic Science in Silica Glass Polishing," in Advanced Metallization for Devices and Circuits-Science, Technology and Manufacturability, MRS Symposium Vol. 337, ed. by S.P. Murarka, A. Katz, K.N. Tu, K. Maex, Materials Research Society, Pittsburgh, PA, p. 89, 1994.
12. J.H. Westbrook, P.J. Jorgenson, "Indentation Creep of Solids," Trans. AIME, 223, 425, 1965.
13. K.M. Davis, M. Tomozawa, "An Infrared Spectroscopic Study of Water-Related Species in Silica Glass," J. Non-Cryst. Solids, 201, 177, 1995.
14. C.-W. Liu, B.-T. Dai, C.-F. Yeh, "Characterization of Chemical Mechanical Polishing Process Based on Nanoindentation Measurement of Dielectric Films," J. Electrochem. Soc. 142, 3098, 1995.
15. L. Cook, "Chemical Processes in Glass Polishing,J. Non-Cryst. Solids, 120, 160, 1990.
16. R.K. Iler, Chemistry of Silica, Wiley, New York, p. 65, 1979.
17. Y. Hayashi et al., "Ammonium Salt Added Silica Slurry for the Chemical Mechanical Polishing of the Interlayer Dielectric Film Planarization in ULSIs," Jpn. J. App. Phys. 34, 1037, 1995.
MINORU TOMOZAWA received his PhD degree in materials science and metallurgy from the University of Pennsylvania in 1968. He is a professor of Materials Science and Engineering at Rensselaer Polytechnic Institute. Tomozawa joined the faculty at Rensselaer in 1969. He has published extensively in the area of glass science and edited several books on the subject. He is chair of the Glass and Optical Materials Division, and a fellow, of the American Ceramic Society. Rensselaer, School of Engineering, Materials Engineering Department, Troy, NY 12180-3590; ph 518/276-6659, fax 518/276-8554, e-mail [email protected].