SURFACE-RELATED PHENOMENA IN THE DIRECT BONDING OF SILICON AND FUSED-SILICA WAFER PAIRS

Transcription

1 Philips J. Res. 49 (1995) SURFACE-RELATED PHENOMENA IN THE DIRECT BONDING OF SILICON AND FUSED-SILICA WAFER PAIRS by G.A.C.M. SPIERINGS, J. HAlS MA and T.M. MICHIELSEN Philips Research Laboratories, Prof Holstlaan 4, 5656 AA Eindhoven, The Netherlands Abstract Direct bonding is the result of a complex interaction between chemical, physical and mechanical properties of the surfaces to be bonded and is therefore strongly correlated with the surface state of the materials. Phenomena characteristic ofthe actual bonding process are (a) the formation of an initial bond area, (b) bond energy, and (c) bond-front velocity. The effects of variations in surface state on these process characteristics have been investigated for silicon, oxidized silicon and fused-silica wafer pairs. The surface bond energy of hydrophilic wafers is in the range of J/m 2 and is largely determined by the hydrogen bonds formed. The bond energy of hydrophobic wafers is a factor of 10 smaller and is determined by Van der Waals attractive forces. The bond-front velocity is determined by the surface state and the stiffness of the wafer. Both bond energy and bond-front velocity show ageing effects. Keywords: direct bonding, silicon, fused silica, bond energy, bond-front velocity, surface modifications. 1. Introduetion The direct bonding of silicon and oxidized silicon wafer pairs is without doubt the most widely studied topic in the field of direct bonding. Surveys of the technological aspects, mechanisms and applications involved can be found in a number of recent reviews [1, 2] and in this issue of Philips Journal of Research. The principles underlying the direct-bonding phenomenon have been known in surface science [3] for decades already, but it is only recently that developments in surface-preparation techniques have made it possible to prepare macroscopie surfaces that are sufficiently clean, smooth and flat to allow attractive surface forces to draw two surfaces together over areas exceeding 100 cnr', Philips Journalof Research Vol. 49 No. 1/

2 G.A.C.M. Spierings et al. Fundamental aspects of direct bonding have been discussed in a number of papers [1, 2, 4]. This paper will report the results of our work concentrating on the basic concepts of the phenomenon, which, involving the interaction of physical, chemical and mechanical properties, are quite complex.. Direct bondingtakes place when two smooth, clean surfaces free of particles approach one another to within distances of nm and the attractive forces between the two surfaces then become sufficiently large to draw the two surfaces together, as described in an accompanying paper [5]. The initial bond-formation process and the bond-front velocity can be used to characterize the bonding process. Both are by nature closely related to the chemical (surface groups), physical (defects) and mechanical (roughness, flatness) surface states ofthe two surfaces, which also determine the bond energy of the direct bond. The bond energy and the bond-front velocity of standard hydrophilic silicon, oxidized silicon and fused-silica wafer pairs will be discussed in this paper. The effect of intentional surface modifications on the two parameters will be described in Section Bond energy and attractive forces in direct bonding The surface forces involved in direct bonding and their effects on the bond energy have been discussed in an earlier paper [4];they will be briefly reviewed below. The attractive Van der Waals forces are responsible for the snapping together of the two surfaces. They are the result of collective dispersion forces of electrical influences between the atoms or molecules of one body and those of another body. For two flat surfaces which approach each other to within distances smaller than about 20 nm the attractive force F at a distance d is given by: where Al2l is the Hamaker constant [6], which is about 25 x and 5.4 x J for Si and Si0 2 (fused silica), respectively.. The Van der Waals attractive energy can be estimated by integrating eq. (1) from d = 0.4 nm to d = 00. The minimum value of 0.4 nm for distance d is the distance at which the attractive forces are assumed to be cancelled by the Born repulsion forces as proposed by Krupp [7]. This results in a 'Van-der-Waalsforce-related' surface bond energy ofo.0075 J/m 2 for a fused-silica wafer pair. The Van der Waals forces (and other physical interactions) are long-range forces (> 1 nm) on an atomic scale. If the distance between the two surfaces becomes sufficiently small, chemical bond formation can contribute (1) 48 Phillps Journal of Research Vol.49 No.IJ2 1995

3 Surface-related phenomena in direct bonding substantially to the strength of the adhesive bond. The surfaces of standard silicon and fused silica (e.g. the thermaloxide on oxidized silicon wafers) have been polished and cleaned in aqueous environments. The resulting surfaces are hydrophilic because they are covered with OH groups with a density of 4-6 OH groups/nmê [8]. Under ambient conditions the surface OH groups adsorb a monolayer of water molecules. After direct bonding, the water molecules and OH groups are trapped at the interface and hydrogen bridges are formed, which link the two surfaces across the interface. The bridging effect increases the bond energy and the strength of the bond [9]. The contribution of the hydrogen bonds to the bond energy can be calculated theoretically. The bond energy of bonded hydrophilic silicon, oxidized silicon and fused-silica wafer pairs may be expected to equal twice the surface energy of a hydrated silica surface. Using a theoretical model of the structure ofthe monolayer of water, Stengl et al. [9] calculated a surface bond energy of J/m 2 This value is at least one order of magnitude larger than the bond energy caused by the Van der Waals forces ( J/m 2 ). 3. Bond energy of standard-cleaned hydrophilic wafers A wedge, e.g. a razor blade, inserted between two bonded wafers will cause the interface to separate until an equilibrium between the bond force and the bending force of the wafers is reached. The bond energy T can be obtained from the resulting crack length L according to: 3Et 3 l T = 32L4 (2) where E is the modulus of elasticity and tand y are the thicknesses of the wafer and blade, respectively [1]. As long as hydroxyl groups or a monolayer of water are present at the surface, the bond energy is largely determined by the hydrogen bonds formed. In the literature, bond energies in the range of J1m 2 have been reported many times for such standard, cleaned hydrophilic wafers; we obtained similar values in our studies. The value obtained for the bond energy depends on the measurement method used and the surface state of the materials involved. As both may vary considerably, the bond energy values cover a very wide range. This is due to the following circumstances: direct bonding is usually effected under ambient conditions, which means that the relative humidity, and to a lesser extent the temperature, may vary. The thickness of the adsorbed water film is known to vary with the relative humidity; PhilJps Journal of Research Vol. 49 No. 1/

4 G.A.C.M. Spierings et al. wafer holder wedge holder wafer holder wafer Fig. 1. Setup for measuring the bond energy in a reproducible manner with the aid of the wedgeinsertion method. the surface roughness of each wafer may vary on a nanometre scale [10]; the blade is usually inserted manually; this operation is therefore not accurately reproducible; the time elapsed between the final cleaning step and the actual bonding and the time elapsed between the bonding and the measurement of the bond energy are often not taken into account; deviations in flatness may vary significantly in the case of standard wafers. Because of that, the degree to which the wafers have to accommodate one another geometrically during bonding is different for each wafel' pair,,> A dedicated measurement test was carried out in order to eliminate the nonreproducible effects of the blade insertion (see Fig. I). The crack length L at t = 0 was measured within the first hour after the direct bonding process. The TO values obtained cover a relatively large range, as can be seen in Table I; this was the same for all three wafer pairs. When the bond energy was measured after the bonded wafer pair had been stored for some time, it was 50 PhlUpsJournnl of Research Vol.49 No. 1/2 1995

5 Surface-related phenomena in direct bonding TABLEI Bond energies obtained shortly after the direct-bonding process (TO) and the equilibrium value (Too), obtained after the bonded wafers had been stored for hours. As TO was measured in the first hour after the directbonding process, substantial variations were found and therefore only a range is given in this table. Wafer combination Si-Si Si (ox.)-si (ox.) fused silica-fused silica fused silica-si (ox.) ± ± ± ± found to increase, reaching an equilibrium value, Too, h after the formation of the bond. Too proved to be far more reproducible than TO' The observed variations in bond energy can be minimized to an even greater extent by using flat surfaces, which implies less accommodation of the surfaces. Unlike thin wafers, thick blocks of material can be ground and polished to a high degree of flatness. By fixing a wafer (via direct bonding) to a flat, 28-mm fused-silica block, as shown in Fig. 2, it is possible to grind and polish the wafer to a flatness similar to that of the thick block (20 nm over 100 mm). After a standard cleaning procedure, we caused the fixed fused-silica, silicon and oxidized silicon wafer to bond directly to other (thin) wafers. That way, the second wafer had to accommodate itself to a significantly smaller extent. The bond energy of a silicon wafer bonded to such a fixed silicon wafer, as a function ofthe time elapsed after bonding, is shown in direct bonded interface \ ~ \_ direct bonded interface \._ direct bonded interface Fig. 2. Basic setup for the bonding experiment with rigid wafers. Philips Journal of Research Vol. 49 No. 1/

6 G.A.C.M. Spierings et al / ~. > I El ~ 0.12 _f al I. "0 c:: r. ~ 0.10 r-._.,~ _ Time (hr) Fig. 3. Bond energy as a function of the time elapsed between direct bonding and the measurement of the bond energy of a silicon wafer bonded to another silicon wafer which has been fixed on a fused-silica block in order to minimize variations in flatness. Fig. 3. It can be seen that the reproducibility is much better than that of the results obtained for two thin wafers. The final bond energy, 7 00, is, however, the same in both cases! The ageing effects of the bond energy shown in Fig. 3 can be understood by taking a closer look at the bonded interface. The two bonded surfaces are not perfectly flat on an atomic scale [10] and therefore there are parts of the surfaces where, at the moment of bonding, hydrogen bond links cannot be formed between the water molecules adsorbed at the surfaces. After the initial bond between the surfaces has been formed, water molecules will diffuse to those regions where the gap between the surfaces is too large for hydrogen bonding. This process is probably driven by internal attractive pressure at the interface. The effect is schematically represented in Fig. 4. The diffusion enlarges the area of hydrogen bonds and thus increases the bond energy, which results in the observed relation between bond energy and time. The equilibrium bond energies 7 00 of the three types of bonded wafer pair are given in Table I. The bond energy of the Si/Si wafer pair is J/m 2, i.e. twice the value obtained for a fused-silica wafer pair (0.083 J/m 2 ), the value of the oxidized silicon wafers lying between the two. In view of the fact that the surfaces of the three standard, cleaned wafers are all hydrophilic 52 PblJlps Journal of Research Vol. 49 No. 1/2 1995

7 Surface-related phenomena in direct bonding t = Fig. 4. Representation of the direct-bonded interface at room temperature. The surfaces can be seen to be in atomic contact, linked by hydrogen bridges, in some parts, whereas the Van der Waals attractive forces dominate in the parts that are not in atomic contact. and all contain OH groups and adsorbed water molecules on a layer of Si0 2 (the hydrophilic Si surface contains a 2-nm native oxide layer), this difference is not easily understood. The explanation is to be found partly in differences in the surface states, such as roughness, and also in cleanliness (of the monolayer), and partly in differences in physical properties such as elasticity (E si = 16.6 X 10 8 J/m 2, E fused silica = 7.25 X 10 8 J/m 2 ). Furthermore, the Van der Waals attractive force of Si is about 5 times that of fused silica [6] due to its higher Hamaker constant, A12!, which means that the contribution of the Van der Waals forces is greater in the case of Si-Si bonding. The bond energy can be increased by several orders of magnitude, up to values corresponding to those of purely chemical bonds in the starting materials, by subjecting the bonded wafers to annealing treatments at temperatures from 900 to 1200 C. The resulting fusion of the two surfaces is convincingly represented in Fig. 5, which shows the surfaces of two bonded and annealed oxidized silicon wafers. The interference pattern (represented in shades of grey here) indicates a brittle fracture after the insertion of a wedge to separate the wafers. This fracture moves randomly through the Si0 2 layer and does not follow the initial interface. 4. Bond-front velocity of standard, cleaned hydrophobic wafers The bond-front velocity (the rate at which the bonded area expands) oftwo transparent fused-silica wafers can be observed thanks to the difference in reflectivity between the bonded and the unbonded areas. The bond-front velocity of silicon wafers (with low dopant concentrations) can be observed using an infra-red sensitive camera. By recording the moving bond front with the aid of a VCR its velocity can be measured. The bond-front velocity was found to PhIIlps Journal of Research Vol. 49 No. 1/

8 G.A.C.M. Spierings et al. SO I bond ruptures split wafer pair complementary interferometric thickness images almost complete disrupture in Si0 2 1ayer Fig. 5. Complementary images of the interference patterns of two oxidized silicon surfaces after direct bonding and annealing. The silicon-on-insulator wafer pair was separated by inserting the wedge several times. remain constant throughout the entire bonding process [4]. Bond-front velocities of mmjs have been reported for standard, cleaned hydrophobic silicon, oxidized silicon and fused-silica wafer pairs (wafer thicknesses 625 Mm) [4, 9]. These wafers bond spontaneously, i.e. the initial bonded area is formed without additional pressure and expands freely. No theory has yet been developed for relating the bond-front velocity to chemical, physical and mechanical properties of the materials involved. Such a theory should consider the equilibrium between bond energy, stiffness 54 Philips Journalof Research Vol. 49 No. 1/2 1995

9 Surface-related phenomena in direct bonding ~ silicon bonded to silicon oxidized silicon bonded to oxidized silicon ~ "13 o a:i >... c: o.:;: "0 c: o CD 1.0 fused silica bondedto fused silica '-- '-- '--...J o 2 Zero, one or two rigid holders --_- Fig. 6. Bond strengths ofwafer pairs consisting oftwo thin wafers (0), one thin and one rigid wafer (1) and two rigid wafers (2). and inertia of the wafers and the fluid dynamics of the film between the wafers. The influence of the elasticity of one or both wafers can be eliminated by making one or both significantly thicker and hence more rigid (20 mm instead of 625 /Lm). In our experiments we made rigid wafers by bonding fused-silica, silicon and oxidized wafers to fused-silica holders. The bond-front velocity was determined as quickly as possible after drying (within 1 h); the results are shown in Fig. 6. The highest bond-front velocity is that observed for the Si/Si pair, the lowest that observed for the fused silica/ fused silica pair. This corresponds to the results obtained in the determination of the bond energy (Table I). The bond-front velocity was found to be lower in the experiments in which rigid, thick blocks were used (Table 11).The experimental results suggest a linear relation between thickness and bond-front velocity. This is in conflict with what is predicted by wave-propagation theory for plates only. In the case of pure bending waves [11] the propagation rate increases with the thickness by a factor of 3/2. Thick blocks are too rigid to deform so as to be able to accommodate each other.. That means that 28-mm thick blocks can only bond if elastic processes occur at the bond front where the two surfaces are Philips Journal of Research Vol. 49 No. 1/

10 G.A.C.M. Spierings et al. TABLE 11 Bond-front velocities of wafer-to-wafer, wafer-to-block and block-to-block combinations of silicon, oxidized silicon and fused silica. The wafer thickness was mm, the block thickness 28 mm. Materials Wafer-wafer Wafer-block Block-block silicon 29.7 mm/s 17.0 mm/s 4.5 mm/s fused silica 16.9 mm/s 10.3 mm/s 3.3 mm/s oxidized silicon 21.5 mm/s 10.7 mm/s 2.9 mm/s close together and are drawn together in a kind of adhesive avalanche. The data indicate a significant influence of the elasticity of the material; silicon has the highest elasticity, fused silica the lowest. We then decided to compare the asymptotic bond energy, Too, with the bond-front velocity. This seemed to be an ambiguous comparison because t N 0.16 ~ c:... ~ 0.14 e ti 0.13 " oxidized silicon to oxidized silicon ID fused silica to fused silica L--...l l- ---I Bond-front velocity (cm/s) Fig. 7. Bond strength Too plotted as a function of the bond-front velocity for a fused silica, an oxidized silicon and a silicon wafer pair. 56 PblIlps Journal ocresearch Vol.49 No.1/2 1995

11 Surface-related phenomena in direct bonding we were comparing an asymptotic value in time (r) with a phenomenon occurring at the initial stage of bonding. Fig. 7 shows the results. To a first approximation a linear relation was found for three pairs of wafers of different materials, i.e. fused-silica, oxidized silicon and (natively oxidized) silicon wafers bonded to wafers of the same material. That also indicates a common factor, which is probably the elasticity. 5. Intentional surface modifications Many chemical and physical treatments are available for modifying the chemical surface state of materials such as Si and Si0 2 The effects of some of these modifications on the direct-bonding characteristics at room temperature are reported in this section. First of all, the effect of the removal of the hydroxyl groups and water molecules, which have a strong influence on the bond strength at room temperature, will be discussed in Sections 5.1 and 5.2. We will then look at the effect of chemical surface modifications (5.3) and modifications in the drying method (5.4) Removal of native oxide layer The native oxide layer containing the surface =Si-OH groups can be removed in a 1% HF solution, which results in a hydrophobic surface containing =Si-H, =Si-CH 3 and a few =SiF groups [12]. The wafers will then bond either under additional pressure [4] or spontaneously [13], depending on the treatment undergone. At room temperature a bond energy of ± J/m 2 is obtained, which is substantially lower than the J/m 2 that was obtained for our hydrophilic silicon wafers cleaned according to a standard method. The low bond energy is the result of an interaction dominated by Van der Waals forces. The bond-front velocity of hydrophobic wafers that have been cleaned through immersion in an HF solution, followed by rinsing with water, is 3 mm/s, which is a factor of ten smaller than that of hydrophilic wafers, after the initial contact has been brought about with the aid of pressure. No ageing effects were observed between the immersion in the HF solution and the bonding, which implies a more or less stable surface condition [4]. The bondfront velocity of surfaces that had been immersed in the HF solution, but had not been rinsed with water, was found to depend on the HF concentration [13] Removal of hydroxyl groups and water molecules The hydroxyl groups and water molecules can be removed from the surface Philips Journal of Research Vol.49 No. 1/

12 G.A.C.M. Spierings et al. ~ EE ~ 10-1 o Q) ;: 10.2 c ol....,_ -g 10.3 o al o Time (hr) Fig. 8. Bond-front velocities of an annealed silicon (0) and an oxidized silicon (e) wafer bonded to a standard, cleaned fused-silica wafer. The delay time is the time elapsed between the final wafer treatment and the actual bonding. ofthe wafers by annealing the wafers at 900 C (for 30 minutes in N 2 ) [12, 14]; in the case of silicon, oxidized silicon and fused-silica wafers =Si-O-Si= (siloxane) surface groups are then formed [12]. Two annealed silicon wafers formed an initial bond only when additional pressure was applied. The wedge test showed that the strength of the bond formed at room temperature was very low, i.e ± J/m 2, which resembles the value obtained for wafers that had been immersed in an HF solution. This bond energy is hence also dominated by Van der Waals forces. Shortly after the annealing, an initial bond was formed only when additional pressure was applied, and the bond-front velocities of both silicon and oxidized silicon were very small. Figure 8 shows the effect of storing the wafers for some time under ambient conditions: the bond-front velocity can be seen to increase as a function of the time elapsed between the annealing and the formation of the bond. This phenomenon, specific to annealed wafers, is caused by the rehydration of surface siloxane (=Si-O-Si=) groups, which form =SiOH groups under ambient conditions. In that way the hydrophillic surface is gradually restored [14] Chemical surface modifications When silicon wafers are treated with a vapour-phase silane compound 58 PhiUpsJournal of Research Vol.49 No. 1/2 1995

13 Surface-related phenomena in direct bonding + Fig. 9. NH-CH 2 -CHOH bond formed in the condensation reaction of amino and epoxy end groups of silane compounds. (hexamethyldisilazane, or HMDS), the surface =SiOH groups are replaced by =Si-O-Si-(CH 3 h groups [15] and the surface hence becomes hydrophobic. In our experiments we succeeded in bonding two wafers that had been treated with HMDS, but the bonded area expanded only when pressure was applied, though less pressure was required than in the previous case (Section 5.2). In a second series of experiments we bonded two more silicon wafers. One of the wafers had been treated with a silane compound with an amino end group (3-amino propyldimethylethoxysilane), the other with a silane with an epoxide end group (3-glycidoxypropyldimethylethoxysilane). After the initial bond had been formed, some of the two end groups were close enough together to be able to react at low temperatures because of the stress at the interface. In this reaction an NH-CH 2 -CHOH bond was formed, which bridged the initial interface as illustrated in Fig. 9. An energy of about 80 kcal/mol is TABLE III Results of the bond-energy measurements of two bonded oxidized silicon wafers. 'UV-cleaned' means that the surfaces had been treated in a UVactivated ozone cleaner. 'Modified' means that the surfaces had been treated with the silane compounds shown in Fig. 9. Modified UV-cleaned TO (J/m 2 ) no no yes no no yes yes yes Pbillps Journal of Research Vol. 49 No. 1/

14 G.A.C.M. Spierings et al. TABLEIV Results ofbond-energy measurements of an oxidized silicon wafer bonded to a fused-silica wafer. 'UV-cleaned' means that the surfaces had been treated in a UV -activated ozone cleaner. 'Modified' means that the surfaces had been treated with the silane compounds shown in Fig Modified UV-cleaned TO (J/m 2 ) no no yes no no yes yes yes required to break such a bond. Tables III and IV show the bond energies obtained in experiments with and without the modification described above. As can be seen, the bond energies obtained after the modification, in combination with UV cleaning, are much higher Modifications of drying method Hydrophilic wafers cleaned according to the standard method are often subjected to a Marangoni drying process [16] before bonding. The wafers are removed from a water bath over which a flow of, for example, isopropanol vapour is passed. The wafers are then dried at the moment they emerge from the liquid, due to surface tensions. Usually the water is neutral (ph = 6-7). Figure 10 shows the effects of variations in the ph of the water bath on the bond-front velocity of a silicon wafer bonded to a fused-silica wafer; the ph was varied from 2 to 12 by adding Hel or NaOH. As the ph increases to 7, the bond-front velocity remains almost constant, at a value of about 20 mm/so As the ph increases above 10, the bond-front velocity increases to about 50 mm/s at ph = 12. At ph values> 2 the native oxide surface layer is negatively charged in the water bath due to the dissociation of surface =Si-OH into =Si-O- units. The surface concentration of =Si-O- groups increases with.increasing ph [17]. At phvalues < 7,the charged surface will be.neutralized by H30 + ions when it is withdrawn from the water bath and =SiOH and adsorbed H20 will be formed. At ph values> 7, Na + ions will be present and the surface will hence be (partly) covered with =Si-O- -Na + units and a monolayer of water. The higher bond-front velocity observed at higher ph values is presumably the result of the formation of a surface with =Si-O-- Na +, i.e. the result of chemical effects such as the high mobility ofthe charged Na + ions and the possibility of the formation of surface Na2Si Phllips Journal of Research Vol.49 No. 1/2 1995

15 Surface-related phenomena in direct bonding t 5 4 ~ 3. E ~ _y: ï3 2 0 Cii >... c: I "0 c: al ~ Fig. 10. Bond-front velocity as a function of the ph of a Marangoni drying liquid. ph 6. Conclusions Characteristics ofthe physical description ofthe direct-bonding process are: (a) the ease with which an initial bonded area is formed, (b) the rate at which this initial area expands (the bond-front velocity or contact-wave velocity), and (c) the bond energy. All three characteristics are strongly dependent on the surface state of the materials to be bonded. The rigidity of the substrate bodies also has an effect on the direct bonding behaviour. In the case of hydrophilic silicon, oxidized silicon and fused-silica wafer pairs with adsorbed hydroxyl groups and monolayer(s) of water, the initial bond is formed spontaneously, the bonded area expands at a rate of mm/s and a bond energy in the range of J/m 2 is obtained, the exact value depending on the surface-preparation method used and the properties ofthe material. When the hydroxyl groups and the adsorbed monolayer of water are removed from the silicon surface, the wafers' tendency to bond directly decreases significantly. The bond-front velocity and the bond energy both decrease by a factor of about one, and only the Van der Waals attractive forces still have an influence. In simplified terms, this means that hydrophilic wafers are bonded by hydrogen forces, hydrophobic wafers by Van der Waals forces. There is evidence that, after surface modification, chemical bonds formed in chemical reactions between groups present at one of the surfaces with other groups present at the other surface contribute to the bond strength. Philips Journal of Research Vol. 49 No. 1/

16 G.A.C.M. Spierings et al. In experiments with pairs of thick blocks with rigidities about 10 5 times higher than that of standard loo-mm wafers, bond-front velocities were measured which were about 10-20% of that of a standard wafer pair. That indicates that the two surfaces are drawn together during elastic processes, such as adhesive avalanche, occurring at the bond front prior to the bonding. Acknowledgments The experiments were carefully conducted by Douwe Bergsma, Ivo Camps and Miss Judith Migchels. Our colleagues Joost Horsten and Jaap Wijdenes read the manuscript critically. Their contribution is gratefully acknowledged. REFERENCES [I] W.P. Maszara, Silicon-on-insulator by wafer bonding; a review, J. Electrochem. Soc., 138, (1991). [2] S. Bengtsson, Semiconductor wafer bonding; a review of interfacial properties and applications, J. Electronic Mater., 21, (1992). [3] Lieng-Huang Lee (ed.), Fundamentals of Adhesion, Plenum Press, New York, [4] G.A.C.M. Spierings and J. Haisma, Diversity and interfacial phenomena in direct bonding, Proc. 1st Int. Symp. on Semiconductor Wafer Bonding; Science, Technology, and Applications, Phoenix, AR, USA, 1992, pub!. The Electrochemical Society, pp [5] J. Haisma, G.A.C.M. Spierings, T.M. Miehielsen and C.L. Adema, Surface preparation and phenomenological aspects of direct bonding, Philips J. Res., 49, (1995). [6] S. Ross and I.D. Morrison, Colloidal Systems and Interfaces, John Wiley & Sons, New York, [7] H. Krupp, Particle adhesion; theory and experiment, Adv. Colloïd Interface Sci., 1, (1967). [8] D.R. Bassett, E.A. Boucher and A.C. Zettlemayer, Adsorption studies on hydrated and dehydrated silicas, J. Colloid Interface ScL, 27, (1968). [9] R. StengI, T. Tan and U. Gösele, A model for the silicon wafer bonding process, Jpn. J. Appl. Phys.,28, (1989). [10] W.P. Maszara, B.L. Jiang, A. Yamada, B.A. Rozgonyi, H. Baumgart and A.J.R. de Koek, Role of surface morphology in wafer bonding, J. Appl. Phys., 69, (1991). [11] S. Timoshenko, Vibration Problems in Engineering, D. van Nostrand Co. Inc., New York, [12] M. Grundner and H. Jacob, Investigations on hydrophilic and hydrophobic silicon (100) wafer surfaces by X-ray photoelectron and high-resolution electron energy loss-spectroscopy, Appl. Phys., A39, (1986). [13] K. Ljungberg, A. Soderbarg and Y. Bäcklund, Spontaneous bonding ofhydrophobic silicon surfaces, Appl. Phys. Lett., 62, (1993). [14] Y.Ya. Davydov, A.V. Kise1ev and L.T. Zhuravlev, Surface and bulk hydroxyl groups of silica by infrared spectra and D 2 0-exchange, Trans. Faraday Soc., 60, (1964). [15] J.J. Ponjeé, V.B. Marriott, M.C.B.A. Michielsen, F.J. Touwslager, P.N.T. van Velzen and H. van der Wel, The relation between lift-off photoresist and the surface coverage of trimethylsiloxy groups on silicon wafers: A quantitative time-of-flight secondary ion 62 Pbllip. Journal of Research Vol. 49 No. 1/2 1995

17 Surface-related phenomena in direct bonding mass spectroscopy and contact angle study, J. Vac. Sci. Techno!., B8, (1990). [16] A.F.M. Leenaars, I.A.M. Huethorst and J.J. van Oekel, Marangoni drying: A new extremely clean drying process, Langmuir, 6, (1990). [17] G.R. Weise, R.D. lames and T.W. Healy, Discreteness of charge and solvation effect in cation adsorption at the oxide/water interface, Discuss. Faraday Soc., 52, (1971). PhIlips Journal of Research Vol. 49 No. 1/