July 09, 2003 - -Journal of Electronic Materials--
This paper reports on the effects of dimple and metallic coating of the Cu-alloy lead frame on interfacial adhesion with an epoxy- molding compound. Round dimples of varying number are introduced on one side of the lead frame by chemical etching. The plating materials studied include bare-Cu alloy and microetched Cu, Ag, Ni, Pd/Ni, and Au/Ni coatings. The surface characteristics, such as wettability, surface roughness, and element compositions, were evaluated based on several characterization tools, which, in turn, are correlated with adhesion performance. The dimples enhanced the interfacial-bond strengths through improved mechanical interlocking of the molding compound, depending on the type of coating. The improvement was much more significant for the coatings with inherently weak interfacial adhesion (e.g., microetched Cu and Ni coating) than those with inherently strong adhesion characteristics (e.g., Au and Pd coatings). The wettability of the metal surface represented by the surface energy or interfacial energy played a dominant role in the resulting interfacial adhesion. Elemental analysis of the fracture surface indicates that the silicon content had roughly a linear relationship with the interfacial-bond strengths for different coatings. The surface roughness was insensitive to the interfacial-adhesion performance. The silicon content measured from the lead-frame fracture surface was shown to directly correlate to the interfacial-bond strength. Higher silicon content was a reflection of larger surface-area coverage by the molding compound associated with cohesive failure.

Key words: Interfacial adhesion, lead frame, plastic packages, metal coatings, surface energy

INTRODUCTION

Delamination at various interfaces is one of the most critical reliability issues in plastic packages. Delaminations, especially those occurring between the lead frame and the molding compound in plastic packages, often lead to popcorning during/after the solder- reflow process because of the presence of absorbed moisture within the plastic encapsulant, resulting in the cracking of the entire package.1 Copper alloys have been widely used as lead-frame material because of their high electrical and thermal conductivity and relatively low cost. In addition, the coefficient of thermal expansion (CTE = 17-18 ppm/[degrees]C) of Cu alloys matches much better with encapsulant material (16-20 ppm/[degrees]C) than Alloy 42 (about 4.5 ppm/[degrees]C). Thus, the danger of delamination caused by thermal mismatches between copper-alloy lead frames and molding compound is less than for Alloy 42 lead frames. The copper surface, however, is very susceptible to oxidation during the package-manufacturing process, significantly impairing the adhesion performance. Copper has a high affinity to oxygen and is readily oxidized when exposed to elevated temperatures. The epoxy-molding compound chemically adheres to Alloy 42 better than to copper alloys, probably, because Alloy 42 does not form a passivating- oxide film as easily as copper alloys do. Significant research efforts have been directed toward improved performance of interfacial adhesion for copper-alloy lead frames. Several techniques have been introduced in the industry to improve the delamination and popcorning performance. They include (1) cross- shaped, window-frame, or slot die-pad design to reduce the contact area between the die pad and molding compound; (2) the introduction of holes or dimples on the die-pad surface to increase the mechanical-anchoring effect with the molding compound; (3) prebaking the package to remove moisture before further processing; and (4) incorporation of a vent hole at the bottom of the plastic package to allow evaporation of absorbed moisture.2 Another promising approach to overcome the delamination problem is by altering the molding- compound formulations for low-moisture solubility and high strength and low stress at high temperatures.3-6 This is an ideal solution because the molding-compound performance is the ultimate test for popcorning and delamination.

The techniques based on the principle of mechanical interlocking have been widely used to improve adhesion. This is often realized by introducing a number of tiny holes or dimples etched on the lead frame,7-10 which can allow the molding-compound resin to flow in during the molding process. The molding compound entrapped within the dimples provides an interlocking mechanism, which, in turn, helps prevent large-scale delamination. The shape of the dimple was critical for the interlocking mechanism: square and round dimples were effective, while the pyramid-shaped were rather inefficient in providing an interlocking effect. The molding compound within the square and round dimples sheared off, while the molding compound within the pyramid dimples tended to pull out easily. The size and number of dimples are restricted by the die-pad space availability after the die-attach process. In a similar approach, holes were also introduced. The most common position for the lead-frame holes is in the short lead stubs that are just inside the dam bars. The pin- shape molding compound entrapped within these holes prevents lateral movement of the molding compound against the lead. Small tabs were also introduced on the lead walls to promote mechanical interlocking of the lead into the molding compound. These tabs are more common on thin leads, which are not wide enough to accommodate holes.

Apart from the foregoing remedies for improving the delamination/ popcorning performance, a more fundamental approach based on plating techniques has also been widely employed to improve the interfacial adhesion.11,12 One of the main functions of plating is to reduce the growth of oxide films during the assembly processes that are detrimental to adhesion. The degree of oxidation is a function of temperature, environmental conditions, heating duration, presence of surface impurities or contaminants, and the surface finish of the metal.13,14 The lead-frame plating material should adhere well to the base material as well as minimize corrosion and enhance solder- ability. Several plating materials used with varying degrees of success include Ni, Ag, Au, Pd, and black copper oxide.

The present study forms part of a larger project regarding the optimization of interfacial characteristics for plastic encapsulated integrated-circuit packages. This paper reports the results from interfacial-adhesion studies between molding compound and surface- finished lead frames. Special emphasis was placed on evaluation of the effect of dimples etched on the lead-frame surface. Several surface-characterization techniques, including contact-angle measurements, x-ray photoelectron spectroscopy (XPS), and atomic force microscopy (AFM) were used, and the results are correlated with the interface-bond strengths measured from the lead-pull test.

EXPERIMENTS

Materials and Lead-Pull Test

The lead-frame material used was a copper alloy, C-194, which typically contains 2.35% Fe, 0.03% P, and 0.12% Zn. Dimples of 0.2 mm in diameter and approximately 76 [mu]m in depth were introduced by chemical etching on one side of the lead frame. The process consisted of chemical etching, stripping, cleaning, and drying of the panels. Etching was done in a ferric-chloride solution at 70[degrees]C for 2 min at a speed of 1.5 m/min. The etched panels were then rinsed with water, followed by stripping of photoresist using potassium hydroxide. The panels were cleaned using deionized water and dried at 55[degrees]C for 1 min. The number of dimples varied from 12 to 20 to 32, which corresponded to 1.5%, 2.5%, and 4.4% of the total surface area. Figure 1 shows a typical leadframe surface with dimples and the corresponding cross-sectional micrograph across the dimple, presenting the relative sizes of dimple and silica-particle fillers added in the molding compound.

Lead frames with six different surface finishes were studied: as- received, bare-Cu sheet (Group A); microetched Cu by means of a chemical process to provide a rough surface (Group B); 3.8-[mu]m- thick spot Ag finish on one (nondimple) side of the lead (Group C); 1-[mu]m-thick Ni coating (Group D); 1-3-[mu]m-thick Ni coating, followed by 0.08-[mu]m-thick Pd plating (Group E); and 1-[mu]m- thick Ni coating, followed by 0.5-[mu]m-thick Au plating (Group F). The silver "spot" plating has been widely used to provide the Cu lead frame with an appropriate surface for die attachment and wire bonding. Following encapsulation with the molding compound, the external lead is typically dipped into a Sn-Pb molten solder to preserve the solderability. Palladium plating over a Ni undercoat has recently been developed15 to circumvent the use of toxic chemicals, waste water, and multiple processing steps associated with the Ag and solder coatings. The Pd/Ni plates have been proven to be suitable for forming reliable solder joints as well as wire bonds. Apart from the Pd plate, the lead-frame finishes also use Ni electroplate over the substrate to provide a barrier layer to base- metal migration. While these impurities can cause severe deficiencies in processing, the Ni finish alone is not a solution. The Ni tends to react readily with the atmosphere, forming oxides that are detrime\ntal to adhesion so that other noble metals, such as Au, are applied over the Ni plate. Although used in economically small amounts, the Pd and Au layers preserve the lead-frame finish and allow a more efficient packaging-assembly operation. Silver was plated selectively because of high cost, concerns about silver migration, and impact on processing yields. The Ag-plating process consisted of an electrolytic clean, activation, thin Cu strike, Ag plating (at 65[degrees]C and 5 A for 1 min) using Ag-jet chemicals, and drying at 100[degrees]C for 1 min. The Au-plating process over Ni was carried out at 70[degrees]C and 5 A for 4 min using an Au- plating solution. The Pd-plating process involved cleaning, activation, Ni plating, activation, and finally Pd plating with rinsing after each step. The Ni was plated at 60[degrees]C and 20 A for 5 min, whereas the Pd plating was performed at 65[degrees]C and 8 A for 1 min.

The microphotographs for a typical lead-frame surface with dimples and the corresponding cross-sectional micrograph across the dimple.

All lead frames were placed in an oven at elevated temperatures to simulate two front-end assembly processes before the encapsulation of the molding compound, namely, die attach at 175[degrees]C for 90 min and wire bonding at 220[degrees]C for 1 min. These processes are known to significantly influence the adhesion performance, as the lead-frame surfaces are contaminated through oxidation. The lead frames were encapsulated with a molding compound (6300 HJ supplied by Sumitomo, Japan) using a transfer- molding machine at 175[degrees]C. The 6300 HJ is an O-Cresol novalac- based epoxy with a 71-74 wt.% silica-filler content. The specimens were post-moldcured at 1757deg;C for 6 h in an air-circulated oven. Figure 2 illustrates the lead-pull specimen. The interface-bond strength between the molding compound and lead frame was measured in terms of the maximum force required to debond the lead frame from the encapsulant. The lead-pull test was performed using a screw- driven universal testing machine. An external force was applied to the upper end while clamping the lower end at a fixed position. All tests were carried out at room temperature at a crosshead speed of 2.54 mm/min. The load-displacement curves were recorded, from which the maximum values were taken as the debond load. At least five specimens were tested to obtained the average debond load.

The schematic of the lead-pull specimen.

Contact-Angle Measurements and Surface Energy

The contact-angle measurement was used to evaluate the wettability of solid surfaces with different probing fluids. Contact angles of coated surfaces were measured using two liquids, namely, deionized water (a polar liquid) and methylene iodide (a nonpolar liquid), using a goniometer (Kruss G10 contact-angle measuring system). The polar and dispersive surface energies are 46.8 mN/m and 26 mN/m, respectively, for deionized water, and 6.7 mN/m and 44.1 mN/ m, respectively, for methylene iodide (CH^sub 2^I^sub 2^). Microsyringes were used to dispense droplets of 2-4 [mu]L on the lead-frame surface. With the aid of an illuminator and a camera, the images of droplets were captured, which were then analyzed using an image analyzer to determine the contact angles. The contact angles were read within 20 sec of droplet formation to avoid evaporation of liquid.

The spreading coefficient is defined as the decrease in surface- free energy per unit area covered by the liquid phase 1 on the solid phase 2. A positive S^sub 12^ value is required for spontaneous spreading of the liquid phase onto the solid surface. In other words, the work of adhesion, W^sub a^, should be greater than the cohesive energy of the liquid, 2^sub [gamma]1^, for complete wetting of the solid by the liquid phase. The polarity, the ratio of the polar component to the total surface energy, [gamma]^sup P^^sub 2^/ ([gamma]^sub P^^sub 2^ + [gamma]^sup d^^sub 2^, was also determined.

The interfacial-bond strength as a function of number of dimples for different coatings: (a) bare Cu, (b) microetched Cu, (c) Ag coating, (d) Ni coating, (e) Pd coating, and (f) Au coating.

Surface Roughness and Elemental Analyses

The surface profiles and roughness of lead frames were characterized using a scanning probe microscope, which consisted of a double parallel-leaf spring-tip assembly, a force sensor, and a piezoelectric tube scanner. The surface roughness was measured using a scan area of 100 [mu]m x 100 [mu]m square with a 1-[mu]m/cycle wavelength after the assembly process. The power spectral-density function approach was used to calculate the arithmetic average- roughness values, R^sub a^. The XPS was used to analyze the chemical compositions of the lead-frame surfaces before and after the simulated assembly processes. The surface morphology of the lead frame before molding and after the lead-pull test was characterized using a scanning electron microscope (SEM).

RESULTS AND DISCUSSION

Effect of Dimple

The interfacial-bond strengths of lead frames with different coatings and dimples are plotted in Fig. 3. The interfacial-bond strength increased with increasing the number of dimples, depending on the type of metal coating. The microetched Cu and Ni coated specimens with inherently weak interfacial adhesion displayed significant improvements with the interfacial-bond strength varying in a linear manner with the dimple number. The bare Cu and Ag- and Au-coated surfaces had only marginal improvements in interfacial- bond strength. In sharp contrast, the Pd-coated specimen with an inherently strong adhesion displayed a negligible effect of the dimple, resulting in an almost constant interfacial-bond strength regardless of the dimples. These observations suggest that the lower the inherent adhesion performance of the metal coating, the greater the benefits to be gained from dimples.

The typical lead-pull load-displacement curves: (a) lead frame without dimple and (b) lead frame with 32 dimples.

The improved interfacial-adhesion performance caused by dimples is further studied based on the typical lead-pull load-displacement curves shown in Fig. 4. Apart from the difference in maximum debond load, another important difference between the specimens with and without dimples is that there was a large, sudden load drop after the peak for those without dimple, and the load drop was minimal and gradual for those with dimples. The sudden large load drop occurs when there is an unstable relative displacement between the two materials caused by the complete chemical/physical separation (i.e., debonding or delamination) along the boundary in the specimens without dimples. The ultimate failure at the peak load may also be unstable for the specimens with dimples because the fracture of encapsulant entrapped in the dimples often occurs almost simultaneously. However, the rough fracture surface of the broken encapsulant was able to induce high frictional resistance against lead pull after debonding, which is mainly responsible for the gradual diminution of the load after the maximum.

The typical SEM photographs in Fig. 5 show matching fracture surfaces on the molding compound and lead-frame sides. The mechanical interlocking/friction and crack-tip bridging mechanisms provided by the dimples worked well, which can be explained as follows with reference to the schematic illustrations presented in Fig. 6. At the initial stage of lead-pull loading, plastic-shear deformation takes place against the molding compound entrapped within the dimples. Upon continuous loading, cracks initiate around the dimple edges. The fracture surface would not be perfectly flat, but rather rough on the microscopic scale, because of the nature of failure (i.e., shear) and the presence of silica fillers in the molding compound. Upon complete interfacial debonding and fracture of the entrapped molding compound, the lead frame is pulled out against friction, leaving significant scratches on the lead-frame surface. If the failure mode were predominantly in tension, as in the double cantilever-beam test,17 the unbroken encapsulants entrapped in the individual dimples would have acted as bridges of the fracture surfaces even after the crack tip traversed past, requiring much energy to be absorbed before ultimate fracture occurred. In summary, the debonding mechanism during the lead-pull test of dimpled specimens is controlled by the combination of cohesive failure of the molding compound in the dimpled region and adhesive failure along the lead-frame/molding-compound interface.

The SEM fractographs of typical lead-pull specimens: (a) lead-frame side and (b) encapsulant side.

Effect of Metallic Coatings

Figure 7 presents the maximum debond loads for different coatings measured using lead-frame specimens without dimples. The interfacial- bond strengths were higher in the descending order of Pd, Ag, bare copper, Au, and microetched Cu and Ni coatings. The high bond strengths for the noble metal coatings, including Pd, Ag, and Au, arise partly from their ability to resist oxidation. While the microetching process increased slightly the surface roughness of the Cu surface, the corresponding interfacial-bond strength was lower than the bare Cu, partly because of the lower wettability represented by the lower surface energy (especially the polar component of the surface energy) associated with severe oxidation of the pure copper surface. The Ni-plated Cu lead frame showed the lowest pull adhesion among all coatings studied, which is consistent with the previous report.11

The schematic illustrations of failure mechanisms taking place around a dimple.

The interfacial-bond strengths in terms of maximum debond load for different coatings.

An attempt is made in the following to correlate the adhesion performance of these coating materials with various surface characteristics\. The contact angles were measured, and the surface energies and the corresponding interfacial energies were calculated. The interfacial-bond strengths for the individual coatings are plotted as a function of total surface energy and interfacial energy in Fig. 8, which exhibit approximately a linear relationship within the data scattering. The spreading coefficients calculated based on Eq. 3 were in the range of 4-12, satisfying the requirement of wetting of the all metal surfaces by the molding component. Judging from the fact that the dispersive components of surface energy were almost identical (i.e., 32-39 mN/m) for all metal surfaces studied, the polarity is compared between these coatings in Fig. 9. The relatively small polar component compared to the dispersive component was also noted for the bare-Cu surface previously.12 The interfacial-bond strength increased as the polarity increased with a close correlation between the two parameters, indicating the surface wettability and hydrogen bonding represented by the polar component of the surface energy played an important role in interfacial-bond strength. This occurs despite the fact that the surface energies of all coatings were dominated by the dispersive component, and that the dispersive component is mainly responsible for improved delamination performance.18 It was also suggested that the interface with a high polar component exhibited high affinity to moisture, and the interface was susceptible to debonding caused by hydrolysis reaction. The amine-cured molding compound was relatively nonpolar with a near-zero polar component of the surface energy (its polar and dispersive components, respectively, are [upsilon]^sup P^^sub 2^ = 0.54 mN/m and [upsilon]^sup d^^sub 2^ = 29.0 mN/m). This may be due to the state of the molding compound; thus, in its solid state, the hydrogen bond may not be very strong. The preceding results suggest that the Pd and Au coatings should have better adhesion with the molding compound than the Ni coating or bare-Cu surface. The Pd plating gives rise to a higher interfacial strength partly because of its higher free-surface energy than the bare-copper surface.15

The relationship between debond load and wettability of the metal surface: (a) surface energy and (b) interfacial energy.

The relationship between debond load and polarity.

The SEM photographs of different coated surfaces, which correlate well with the AFM images (not shown here). The surface of the bare Cu exhibited directionality and microcracks along that direction, which may be beneficial to interfacial bonding, as these corrugations are partly filled by the molding compound, providing some mechanical-interlocking effect. The Ni and Au coatings were rough with nodular grains of uniform size all over the surface, whereas the Pd-coating surface was smoother without nodular features. Figure 11 presents the interfacial-bond strength as a function of the arithmetic-average roughness measured from the AFM, showing no particular correlation between the two parameters. Other surface roughness parameters, such as the surface area and the surface-area difference, had similarly no correlation with the interfacial-bond strength. Although the Ni and Au coatings showed similar surface profile and roughness, a significant difference existed in interfacial-bond strength between these coatings. The Pd surface had the highest interfacial-bond strength even though its surface roughness was among the lowest. This strongly suggests that the coating-surface morphology may not be the dominant factor that determined the interfacial adhesion amongst many parameters studied. In fact, the range of surface roughness for all coating materials was narrow (i.e., in the range of 60-170 nm) and below 0.2 [mu]m. Assuming the average size of the smallest silica fillers contained in the molding compound was approximately 2-3 [mu]m, the potentials for these rigid fillers to be entrapped in the coating surface would be very slim. This may suggest that the contribution of mechanical interlocking, caused by the rough surface, to interfacial adhesion was small compared to the other bonding mechanisms.

The SEM photographs of different coated surfaces.

The interfacial-bond strength as a function of the arithmetic-average roughness measured from AFM analysis.

The XPS spectra for the Au/Ni surface obtained before molding and after pull test.

The XPS analysis was conducted to obtain information about the chemical state of various coating surfaces. There exists a significant difference between the typical XPS spectra obtained before molding and after pull test, as shown in Fig. 12. A significant peak was observed at the binding energy of 104 eV, indicative of Si (Si^sub 2p^), on the fracture surface after the lead-pull test, while the same was absent on the surface analyzed before molding. It is thought that Si originates from the organosiloxane compound, a silane coupling agent, that was added in the molding compound to improve the interfacial adhesion with the lead frame. The XPS analysis on the cured-mold compound indicated approximately 11 at.%Si. Figure 13 presents the correlation between the interfacial-bond strength and Si content obtained from the XPS analysis: the higher the silicon content, the greater the interfacial-bond strength. A higher Si content may be interpreted as a larger surface area being covered with the molding compound because of the high interfacial adhesion.

The interfacial-bond strength as a function of silicon content obtained from XPS analysis.

The lead-pull strength was higher for the bare Cu than the microetched counterpart so that the susceptibility of the two Cu surfaces to oxidation taking place during exposure to the simulated assembly environments was different. It appears that the oxide thickness was much greater for the microetched Cu than for the bare- Cu surface. This is partly evidenced by the large oxygen content for the former Cu surface found in the XPS elemental analysis, as shown in Table I. While the oxide thickness present on the bare-Cu surface was within the range of the optimal thickness 20-30 nm required for the best interfacial-adhesion performance,19-21 it appears that the oxide thickness found on the microetched-Cu surface far exceeded the optimal range. The weaker interfacial adhesion for the microetched- Cu surface can be related to the internal voids grown along the oxide/copper (Cu^sub 2^O/Cu) interface.19 It was shown that, with increasing oxide thickness, the interfacial adhesion between Cu^sub 2^U and Cu becomes weaker because of the higher potential for large voids. Meanwhile, the microetching process also resulted in an increase in surface roughness, which may be beneficial to interfacial adhesion. However, the enhanced surface roughness was too small, i.e., 10% higher than the bare-Cu surface, to make any significant contribution to interfacial adhesion,19 suggesting that the original aim of increasing the surface roughness for improved mechanical interlocking was not specifically achieved. This is particularly true when the roughness of the microetched-Cu surface (Fig. 10b) is much smaller than the average size of the silica particles (Fig. 1). Therefore, the chemical-etching process may have some detrimental effects on interfacial adhesion to the molding compound because of rapid oxidation.

Elemental Compositions of Selected Lead-Frame Surfaces (1) before and (2) after the Assembly Process

There were reports15,22 with conflicting results regarding the adhesion performance of Pd-coated Cu lead frames. The Pd coating gave rise to an almost twofold increase in lead-pull strength because of the high surface energy of the Pd coating,15 consistent with the present study. In contrast, the Pd coating also resulted in lower interfacial adhesion than the bare-Cu lead frame,22 which was attributed to a higher percentage of carbon present on the Pd- plating surface after the simulated assembly process, such as die attach and wire bonding. In the present study, the carbon content on the Pd coating surface was even lower than the Cu surfaces (Table I). It is suspected that there were other unknown factors that adversely affected the interfacial adhesion in their study.22

CONCLUDING REMARKS

The effects of dimple and coating material on interfacial-bond strength between the lead frame and molding compound were studied. Several techniques were employed to evaluate the surface characteristics of the coated surface, which, in turn, were correlated to the interfacial adhesion. The following can be highlighted from the experimental study.

* Dimples introduced on the lead frame increased significantly the lead-pull strength against the molding compound through mechanical interlocking/friction mechanisms. The interfacial adhesion improved with increasing the number of dimples, depending on the type of metal coating. The improvement in interfacial adhesion was higher for the coatings with inherently weak interfacial adhesion (e.g., microetched Cu and Ni coating) than those with inherently strong adhesion characteristics (e.g., Au and Pd coatings).

* The interfacial-bond strength exhibited approximately a linear correlation with the surface energy, suggesting the wettability of the metal surface plays a significant role in constituting the interfacial adhesion. While the dispersive component of surface energy was dominant, the polar component is found to control the wettability of the metal surfaces studied.

* No specific relationship was noted between the surface roughness of the coating as measured from the AFM analysis and the interfacial-adhesion performance.

* The silicon content measured from the lead-frame fracture surface was shown to directly correlate to the interfacial-bond strength. Silicon is contained in the molding compound as the adhesio\n promoter, and a higher silicon content corresponds to a larger surface area being covered with the molding compound because of predominant cohesive failure.

ACKNOWLEDGEMENTS

This project was supported by the Research Grant Council of Hong Kong Special Administration Region (Project No. HKUST6014/98E) and the HKUST Postdoctoral Fellowship Matching Fund. QPL Ltd. provided partial financial support for the first author (ML) and supplied lead-frame samples. Most experiments were carried out with the technical supports from the Material Characterization and Preparation Facilities and EPack Lab. of HKUST.

Journal of ELECTRONIC MATERIALS, Vol. 32, No. 6, 2003

(Received May 7, 2002; accepted October 11, 2002)

REFERENCES

1. L.T. Nguyen, ASME J. Electron. Packaging 115, 346 (1993).

2. I. Fukuzawa, S. Ishiguro, and S. Nanbu, Int. Reliability Physics Symp. (Piscataway, NJ: IEEE, 1985), pp. 192-197.

3. S. Kuroki and K. Oota, 39th Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1989), pp. 885-890.

4. N. Mogi and H. Yasuda, 42nd Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1992), pp. 1023-1029.

5. S. Asai, U. Satura, and M. Tobita, J. Appl. Polym. Sci. 51, 1945 (1994).

6. T. Nishioka, M. Nagasawa, K. Igarshi, M. Kohmoto, and S. Ito, 40th Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1990), pp. 625-631.

7. T.O. Steiner and D. Suhl, IEEE Trans. Comp. Hybrids Manuf. Technol. 10, 209 (1987).

8. R. Lin, E. Blackshear, G. May, G. Hamilton, and D. Kirby, 7th Ann. Conf. of International Electronic Packaging Society (1988), pp. 995-1010.

9. L.T. Nguyen and M. Michael, 42nd Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1992), pp. 930-938.

10. S. Altimari, S. Golwalkar, P. Boysan, and R. Foehringer, 42nd Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1992), pp. 945-950.

11. O. Yoshioka, N. Okabe, S. Nagayana, G. Murakami, and R. Yamagishi, 39th Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1989), pp. 464-471.

12. S. Kim, IEEE Trans. Comp. Hybrids Manuf. Technol. 14, 809 (1991).

13. S. Yi, C.Y. Yue, H. Hsieh, L. Fong, and S.K. Lahiri, J. Adhes. Sci. Technol. 13, 789 (1999).

14. S. Yi, J.K. Kim, C.Y. Yue, and J.H. Hsieh, J. Adhes. 73, 1 (2000).

15. D.C. Abbott, R.M. Brook, N. McLelland, and J.S. Wiley, IEEE Trans. Comp. Hybrids Manuf. Technol 14, 567 (1991).

16. S. Wu, J. Adhes. 5, 39 (1973).

17. J. Guzek, H. Azimi, and S. Suresh, IEEE Trans. Comp. Packaging Manuf. Technol. Part A 20, 496 (1997).

18. G.S. Ganesan, G.L. Lewis, and H.M. Berg, Adv. Electron. Packaging ASME 10, 450 (1995).

19. S.J. Cho, K.W. Paik, and Y.G. Kim, IEEE Trans. Comp. Packaging Manuf. Technol. Part B 20, 167 (1997).

20. C.T. Chong, A. Leslie, L.T. Beng, and C. Lee, 45th Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1995), pp. 463- 469.

21. H. Ohsuga, H. Suzuki, T. Aihara, and T. Hamano, 44th Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 1994), pp. 141-146.

22. K.C. Chan, T.C. Chai, YZ. Yang, and R. Gopalakrishnan, 50th Electronic Components Technology Conf. (Piscataway, NJ: IEEE, 2000), pp. 175-182.

M. LEBBAI,1 JANG-KYO KIM,1,2 and MATTHEW M.F. YUEN1

1.-Department of Mechanical Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong. 2.-E- mail: mejkkim@ust.hk