An investigation of the flip-chip bonding process for application in MEMS devices was ... flip-chip assembly were found to correlate well with simulation results.
W. Salalha E. Zussman Manufacturing Systems Laboratory, Faculty of Mechanical Engineering, Technion-Israel Institute of Technology, Haifa, Israel 32000
P. Z. Bar-Yoseph Computational Mechanics Laboratory, Faculty of Mechanical Engineering, Technion-Israel Institute of Technology, Haifa, Israel 32000
1
Investigation of Flip-Chip Bonding for MEMS Applications An investigation of the flip-chip bonding process for application in MEMS devices was carried out. Finite element analyses of axisymmetric and non-axisymmetric solder joint geometries were performed. It was found that in typical cases of MEMS devices in which the solder volume is small 共 BoⰆ1, where Bo is the Bond number), the finite element solution of the axisymmetric solder joint is well approximated by a surface of revolution whose generating meridian is a circular arc. Experimental results of solder joints in flip-chip assembly were found to correlate well with simulation results. 关DOI: 10.1115/1.1646427兴
Introduction
One of the critical problems in MEMS is to integrate sensing and actuating processes with existing electronics and packaging technology. This integration usually demands highly accurate assembly operations in order to fulfill the functional requirements of the system. A common integration approach uses flip-chip bonding technology 关e.g., 关1,2兴兴. The aim of this work is to investigate the flip-chip bonding process as it may potentially be applied in a solder joint system. A system consists of a liquid column of solder that is located on a substrate and which supports a chip 共see Fig. 1兲. The problem of calculating the shape of capillary surfaces has been discussed by many authors 共e.g., Finn 关3兴兲. Goldman 关4兴 studied a two-dimensional solder joint in which the joint geometry model was assumed to be a truncated sphere. Heinrich et al. 关5兴 and Chiang and Chen 关6兴 presented a model that assumed a solder joint with a surface of revolution whose generating arc was a circular arc. Katyl and Pimbley 关7兴 calculated the shape of an axisymmetric solder joint connected to circular-shaped pads. A three-dimensional model for a solder joint was presented by Patra et al. 关8,9兴. In their model, solder joints with circular and noncircular shaped pads were considered. Finite-element analyses of solder joint geometries have been suggested by Barake 关10兴, Nigro et al. 关11兴, and Subbarayan 关12兴. In this study the proposed model includes optimization of the final height of the joint, and usage of a Laplace-Young equation to calculate the equilibrium state. Finite elements to calculate axisymmetric and nonaxisymmetric solder joint shapes under quasistatic conditions are briefly presented in the paper 共for details see 关13兴兲. The model was tested and verified through flip-chip bonding experiments. The paper is organized as follows: The finite element formulation used to calculate solder joint shape parameters is described briefly in Section 2. Experimental results of a flip-chip assembly are presented in Section 3 and a discussion provided in Section 4.
⌸⫽ ␥ •As⫺  •Ap⫹
Theoretical Joint Model
The proposed model considers the total energy of a solder joint and determines its minimum energy. The variational problem associated with this system is handled using a constant solder volume constraint. For our model we have assumed that: the solder joint wets the entire pad, the surrounding material is perfectly non-wettable, and the solder pads are perfectly aligned when the solder reaches the minimum energy state. The total energy associated with the solder joint is, Contributed by the Electronic and Photonic Packaging Division for publication in the JOURNAL OF ELECTRONIC PACKAGING. Manuscript received July 2003. Associate Editor. Y. C. Lee.
48 Õ Vol. 126, MARCH 2004
⍀
•g•d⍀⫹W•H
(1)
The first three terms in Eq. 共1兲 represent surface energy, adhesion energy, and potential energy, respectively. The last term represents the potential energy of the chip. The shape of the solder joint with volume V and height H under a quasi-static equilibrium condition is obtained from the solution of the variational problem that involves minimizing Eq. 共1兲 subject to a constant volume constraint. By using the Lagrange multiplier approach we can introduce the volume constraint by forming another functional, ¯ ⫽⌸⫹• 共 ˜V ⫺V 兲 ⌸
(2)
The finite element formulation considers the total energy of the solder joint system without ignoring the solder joint’s potential energy (Bo⬇O(1), 关14兴, e.g., the solder column’s diameter is about 0.8 关mm兴 using a tin-lead alloy兲. An axisymmetric finite element model using 1-D linear elements is employed for solving the solder joint radius r(z) as well as the stand-of-height, H, thereby defining the solder joint’s shape in an equilibrium state. In this state, the height, H, is divided into N linear elements to determine the free surface of the solder. For a 3-D solder joint with a non-axisymmetric cross section, a two-dimensional model using linear triangular elements is employed. In this case, the height, H, of the solder joint is divided into N solid slices, where each slice is divided into M angles of degree 关13兴. Typical results for an axisymmetric solder joint formed on circular pads were obtained through simulation in the Matlab® environment 共see Fig. 2兲. A comparison of the proposed axisymmetric model results with previous studies is presented in Table 1. As expected 共for BoⰆ1), there was close agreement with the results obtained by Heinrich et al. 关5兴, who assumed that the generating arc of the solder joint’s surface of revolution was circular.
3 2
冕
Experimental Results
The experiments were conducted in an experimental flip-chip bonding setup in a residue-free active atmosphere 共gaseous formic acid, 0.55%, as used by Lau 关15兴 among others兲. Testing was performed using three different chips, labeled as Glass-Chip, Silicon-Chip and Quasi-MEMS Chip, various pad geometries and dimensions, and varying solder joint dimensions 共see Table 2兲. The solder material was a ‘‘solder sphere’’ with a diameter variation of 1% 共Micro-Swiss Ltd兲. Electron micrographs of the solder joints and chips were obtained using a SEM and tested with Energy Dispersion Spectroscopy 共EDS兲, 共Philips XL30 SEM including an Oxford thin window EDS system兲. Typical bonded Glass-Chips are shown in Fig. 3. The accuracy of the horizontal alignment of the chips in their final stage, after
Copyright © 2004 by ASME
Transactions of the ASME
Fig. 3 Typical picture of bonded Glass-Chips with pad diameter, r p Ä300 † m‡: „a… top view of the bottom chip, „b… close-up of the assembled chips before reflow process, „c… close-up of the bonded chips after self-alignment termination Fig. 1 Cross section of an axisymmetric solder joint with height, H , and pad radius, r p loaded by a chip with weight, W
Fig. 2 Concave shape of solder joint. Simulation parameters: W ÄÀ2.5Ã10À5 †N‡, V Ä1.4137Ã10À3 †mm3 ‡, r p Ä0.175 †mm‡, ␥ Ä480 †dyneÕcm‡, number of elementsÄ22 Table 1 Comparison of joint stand-of-height, H , predicted by the present axisymmetric finite element solution and other independent solutions. Simulation parameters: V Ä0.2413 †mm3 ‡, r p Ä0.3175 †mm‡, Ä0 †gÕcm3 ‡, ␥ Ä325 †dyneÕcm‡, number of elementsÄ100, „ … * Ä9 †gÕcm3 ‡ H 关m兴 W 关 ⫻105 N兴
Nigro et al. 关11兴
Brakke 关10兴
Heinrich et al. 关5兴
Present study
⫺23
469
453
0
526
32
750
472 (470) * 528 共524兲 749 共746兲
454 共453兲 522 共521兲 752 共743兲
524 753
the self-alignment process, could be determined using the alignment marks 共4 rectangular marks on the bottom chip, and a cross on the top chip兲. The average final misalignment was ⬃2 关 m兴 with a standard deviation of ⫽0.4 关 m兴 . Typical results of the stand-of-height, H, of the solder joint for different pads and the solder volumes are presented in Table 3. Representative magnified pictures and scanning electron fractographs of a Silicon-Chip are shown in Fig. 4. The average standof-height, H, of a group of 15 devices, performed after the reflow ¯ ⫽133.3 关 m兴 , with a standard deviation of process was H ⫽1.7 关 m兴 . A similar examination was made of a Quasi-MEMS device with the results shown in Fig. 5. The average stand-of-height, H, of 5 ¯ ⫽84.6 关 m兴 measured devices after the reflow process was H with a standard deviation of ⫽1.8 关 m兴 . Intermetallics compound consisting of Cu and Sn were found in the solder joints of Quasi-MEMS device by EDS analyzing 共e.g., see 关16兴兲.
4
Discussion
A table comparing the solder joint stands-of-height, H, obtained through the experiments 共Glass-Chip, see Table 3兲 and the proposed model is presented in Fig. 6. There is a clear difference 共of about 8%兲 between the experimental and numerical results when small pad radii, r p , are considered. Nevertheless, for large pad radii, r p , the experimental results and the presented finite element models show similar values. For the Quasi-MEMS chip with a solder joint with square pads, the predicted stand-of-height was H⫽86.6 关 m兴 . The experimen¯ tal average height 共for 5 devices兲 measured with a SEM was H ⫽84.6 关 m兴 .
Table 2 Experimental chip data
Chip type
Chip Size 关mm兴
Weight ⫻105 关N兴
UBM material, thickness 关m兴
Glass Chip
4⫻3⫻0.2
1
Cr/Cu,Cu,Ni,Au 0.5,1,1,0.1
Pb37Sn67
Silicon Chip
2.2⫻2.2⫻0.67
6.244
Pb37Sn67
Quasi-MEMS Chip
0.78⫻3.6⫻5.78
37
Cr/Cu,Cu,Ni,Au 0.5,1,1,0.1 Cr/Cu,Cu,Au 1,2,0.1
Journal of Electronic Packaging
Solder material
Pb37Sn67
Pad shape; No. of pads; Size 关m兴
Solder diameter, 关m兴
Circle; 4; 200,300,350, 400,600 Square; 15; 150⫻150 Square; 9; 200⫻200
200,300, 350,400 200 200
MARCH 2004, Vol. 126 Õ 49
Table 3 The experimental results for various glass-chips Pad radius, r p 关m兴
Solder radius, r s 关m兴
Solder Stand-of-height, ¯ 关m兴 H
175 200 300
150 150 200
122 105 120
The solder spheres used in the experiments varied by 1% in their diameter; an example of the result of this variability is a decrease of H by 2– 4 m for a 250 m diameter solder sphere. Other sources of process inaccuracy are the solder solidification rate, and the intermetallic created during bonding.
5
Conclusion
A study of the flip-chip bonding process using numerical and experimental methodologies was carried out. Axisymmetric and a non-axisymmetric finite element formulations have been developed to study the effect of variation in de-
Fig. 6 Glass-Chip: Comparison of typical solder joint standof-height, H , results obtained by: „a… experiment, „b… axisymmetric FE model „number of elementsÄ20….
sign parameters 共vertical loading, solder joint volume, and pad radius兲 on the geometrical shape of a joint under quasi-static conditions. It was found that in typical cases of MEMS devices in which the solder volume is small (BoⰆ1), the finite element solution of the axisymmetric solder-joint geometry can be well approximated by a surface of revolution whose generating meridian is a circular arc. The experimental results of the investigation correlated well with those produced by the simulation model. The results indicate the possible application of flip-chip bonding for MEMS applications, which require high accuracy in fabrication, where typical solder joint heights are in the range of tenths of microns.
Acknowledgment This research was partially supported by the Fund for the Promotion of Research at the Technion, and Rafael Ltd.
Nomenclature
Fig. 4 Typical pictures and SEM fractographs of SiliconChips: „a… substrate chip, „b… assembled chips, „c… chip with solder bumps after reflow, „d… SEM picture of solder bump cross section
Ap ⫽ wetted solder area 共Pad surface兲 关 mm2 兴 As ⫽ free solder surface 关 mm2 兴 Bo ⫽ bond number, gravity to surface tension forces (⬇ gr s2 / ␥ ) 关14兴 g ⫽ gravity acceleration 关 cm/sec2 兴 H ⫽ stand-of-height of solder joint 关mm兴 ¯ ⫽ average of stand-of-height of solder joints 关mm兴 H r s ⫽ solder radius 关mm兴 r p ⫽ pad radius 关mm兴 k ⫽ 2-D finite element–a single element angle 关rad兴 V ⫽ solder volume 关 mm3 兴 ˜V ⫽ computed volume 关 mm3 兴 W ⫽ load on top pad 共upward is positive兲 关N兴  ⫽ adhesion coefficient 关dyne/cm兴 ⫽ solder density 关 g/cm3 兴 ␥ ⫽ surface tension coefficient 关dyne/cm兴 ⫽ Lagrangian multiplier 关 dyne/cm2 兴 ⌸ ⫽ total energy functional ¯ ⫽ total energy functional including a volume constraint ⌸ ⍀ ⫽ domain
References Fig. 5 Typical pictures and SEM fractographs of Quasi-MEMS chips: „a… top view of a quasi MEMS chip, „b… assembled device, „c… cross section of assembled device with two solder joints, „d… cross section of a single solder bump with stand-ofheight, H Ä86.6 † m‡.
50 Õ Vol. 126, MARCH 2004
关1兴 Harsh, K. F., Su, B. Z., Zhang, W. G., Bright, V. M., and Lee, Y. C., 2000, ‘‘The Realization and Design Considerations of a Flip-Chip Integrated MEMS Tunable Capacitor,’’ Sens. Actuators A, 80共2兲, pp. 108 –118. 关2兴 Salalha, W., Zussman, E., Meltser, M., and Kaldor, S., 2000, ‘‘Prediction of Yield for Flip-chip Packaging,’’ Proc. of the 10th CIRP Design Seminar, Israel, pp. 259–263. 关3兴 Finn, H., 1990, Equilibrium Capillary Surfaces, Springer-Verlag, NY.
Transactions of the ASME
关4兴 Goldmann, L. S., 1969, ‘‘Geometric Optimization of Controlled Collapse Interconnections,’’ IBM J. Res. Dev., 13, pp. 251–265. 关5兴 Heinrich, S. M., Schaefer, M., Schoroeder, S. A., and Lee, P. S., 1996, ‘‘Prediction of Solder Joint Geometries in Array-Type Interconnects,’’ ASME J. Electron. Packag., 118共3兲, pp. 114 –121. 关6兴 Chiang, K. N., and Chen, W. L., 1998, ‘‘Electronic Packaging Reflow Shape Prediction for the Solder Mask-Defined Ball Grid Array,’’ ASME J. Electron. Packag., 120共2兲, pp. 175–178. 关7兴 Katyl, R. H., and Primbley, W. T., 1992, ‘‘Shape and Force Relationships for Molten Axisymmetric Solder Connections,’’ ASME J. Electron. Packag., 114共3兲, pp. 336 –341. 关8兴 Patra, S. K., and Lee, Y. C., 1991, ‘‘Quasi-Static Modeling of the SelfAlignment-Part 1: Single Solder Joint,’’ ASME J. Electron. Packag., 113共3兲, pp. 337–342. 关9兴 Patra, S. K., Sritharan, S. S., and Lee, Y. C., 1995, ‘‘Quantitative Characterization of Flip-Chip Solder Joints,’’ ASME J. Appl. Mech., 62共2兲, pp. 390– 397.
Journal of Electronic Packaging
关10兴 Brakke, K., 1992, ‘‘The Surface Evolver,’’ Exp. Mech., 1, pp. 141–165. 关11兴 Nigro, N. J., Zhou, F. J., Heinrich, S. M., Elkouh, A. F., Fournelle, R. A., and Lee, P. S., 1998, ‘‘Parametric Finite Element Method for Predicting Shapes of Three-Dimensional Solder Joints,’’ ASME J. Electron. Packag., 118共3兲, pp. 142–147. 关12兴 Subbarayan, G., 1996, ‘‘A Procedure for Automated Shape and Life Prediction in Flip-Chip and BGA Solder Joints,’’ ASME J. Electron. Packag., 118共3兲, pp. 127–133. 关13兴 Salalha, W., Zussman, E., and Bar-Yoseph, P. Z., 2002, ‘‘Modeling and Simulation of Flip-Chip Bonding,’’ Technical Report, TME # 476, Faculty of Mechanical Engineering, Technion—Israel Institute of Technology, Haifa, Israel. 关14兴 Myshkis, A. D., Babskii, V. G., Kopachevskii, N. D., Solobozhanin, L. A., and Tyuptso, A. D., 1986, Low-Gravity Fluid Mechanics, Springer-Verlag, NY. 关15兴 Lau, J. H., 1995, Flip Chip Technologies, McGraw-Hill, NY. 关16兴 Jang, S.-Y., and Paik, K.-W., 1998, ‘‘Eutectic Sn/Pb Solder Bump and Under Bump Metallurgy: Interfacial Reactions and Adhesion,’’ Soldering & Surface Mount Technology, 10共3兲, pp. 29–37.
MARCH 2004, Vol. 126 Õ 51
