Selecting a Temperature Time History for Predicting Fatigue Life

Journal of Thermal Stresses, 24:1063^1083 , 2001 Copyright # 2001 Taylor & Francis 0149-5739 /01 $12.00 + .00

SELECTING A TEMPERATURE TIME HISTORY FOR PREDICTING FATIGUE LIFE OF MICROELECTRONICS SOLDER JOINTS Cemal Basaran UB Electronic Packaging Laboratory University at Buffalo, SUNY Buffalo, New York, USA Terry Dishongh Intel Corporation Hillsboro, Oregon, USA Ying Zhao Analog Devices Norwood, Massachusetts , USA

Temperature cycling tests are standard industry practice for determining the thermomechanical fatigue life of solder joints. Industry-standar d temperature pro¢les usually start from room temperature, then go to a high temperature, then to a cold temperature, and then back to room temperature. In addition, most of the time, the temperature pro¢le contains dwell times at the highest and lowest temperatures. The dwell time at a high temperature corresponds to the on-state storage of the device, and the cold-temperature dwell time simulates the off-state storage of the device. In this study, the actual temperature history of a Ball Grid Array (BGA) package in a laptop computer was measured in situ. Experimental reliability studies were conducted using the in situ measured temperature history as well as industry-standard temperature history. This article presents the in£uence of temperature history on solder joint fatigue life. In order to measure deformations in the solder joint under cycling loading, a new Moire¨ interferometry grating replication technique was developed to be able to measure strain ¢eld during fatigue testing.

Received 7 November 2000; accepted 25 November 2000. This research project was partially sponsored by Intel Corporation and partially by a grant from the National Science Foundation GOALI program CMS-9908016 under the supervision of Dr. Jorn Larsen-Basse. Address correspondence to Professor Cemal Basaran , UB Electronic Packaging Laboratory, 212 Ketter Hall, North Campus, Buffalo, NY 14260. E-mail: [email protected]

1063

1064

C. BASARAN ET AL.

Driven by trends toward miniaturization and a higher level of integration, the complexity of microelectronic packaging grows continually. The reliability of assembly components is critical to manufacturing a high-quality ¢nal product. Due to an increasing number of circuits and a shortage of real estate in a processor, BGA packaging is becoming one of the most popular packaging options in the microelectronics industry. The Lead/Tin (Pb/Sn) eutectic alloy is widely used as a joining material in the electronic industry. It is well known that the dominant failure mode for solder joints is low-cycle thermal fatigue, which is caused by a thermal expansion mismatch between the bonded components and heat dissipated by devices during operation. The level of heat generated by the device during service may £uctuate due to the variation of the work intensity, which results in temperature £uctuation in the package. Such mini-temperature £uctuations can also be caused by changes in the power level and the ambient temperature. Although thermal cycling has been studied by many researchers, the impact and effect of temperature £uctuations during dwell times on BGA solder joint fatigue life has never been studied or reported in the literature. The main questions in reliability studies have always been, ``Is temperature cycling appropriate both in time history pro¢le and magnitude if we are unsure of the damage mechanism and strain ¢eld in the solder joint?’’ and ``What is the baseline of inelastic strains developed in thermal cycling and usage?’’ Until now these questions have remained unanswered due to the lack of a technique to measure strain ¢eld in a solder joint in fatigue cycling. This article is aimed at providing reasonable answers to these questions that were raised and shed some light on BGA packaging design and failure analysis. Temperature cycling is used as a standard industry practice for determining fatigue life and reliability of solder joints [1^11]. A thermal cycling test, also known as a highly accelerated stress test (HAST), is a vehicle used to accelerate the fatigue process in packaging. The test yields the number of cycles to failure value for a given package. There is a continuously increasing interest in the experimental analysis of solder joint responses to thermomechanical loading. Moire¨ interferometry (MI) provides whole ¢eld maps of in-plane deformation contours with submicron resolution and both normal and shear strain information. Such a capability is extremely useful in studying the thermomechanical behavior of solder joints in electronic packaging. A new MI technology developed in the UB Electronic Packaging Laboratory allows for recording inelastic strain accumulations for cycling thermomechanical loading. Since inelastic deformation is directly related to fatigue life of solder joints, the inelastic strain accumulation can be used to evaluate solder joint fatigue life performance [12^14].

EXPERIMENTAL PROCEDURE A standard practice in the industry is to use thermal cycling for fatigue life determination of BGA solder joints. Conventional temperature pro¢les have a jigsaw shape with dwell times at maximum and minimum temperatures (Figure 1a). The conventional temperature time histories are based on many assumptions with

PREDICTING FATIGUE LIFE OF SOLDER JOINTS

Figure 1. Temperature time histories used in this study.

1065

1066

C. BASARAN ET AL.

no technical or empirical basis. During this study several laptop computers were wired with thermocouples in several different locations for a three-month period. Temperature time histories were recorded by ¢eld studies. The observation results indicate that the actual temperature pro¢le experienced by the package is different than traditional temperature time histories used for temperature cycling tests in the industry. Figures 1b,c were obtained from in situ ¢eld measurements. The purpose of this study is to compare the in£uence of the temperature time history pro¢le on fatigue life of BGA solder joints. In the study, the industry-standard temperature pro¢le as well as the in situ measured actual temperature time histories are used for thermal cycling. Three different temperature pro¢les are used. Test 1 has a conventional thermal cycling pro¢le (Figure 1a). For Tests 2 and 3 temperature pro¢les include temperature £uctuations at the highest temperature (Figures 1b,c). For each test (temperature pro¢le), 5 specimens were used to validate the results. Test conditions for three tests are: Test 1. Temperature pro¢le is between 0¯ and 75 ¯ C with 30 minutes of dwell on both sides of the curve. Test 2. Temperature pro¢le is between 0¯ and 75¯ C with 30 minutes of 5 minicycles at 10¯ C on the hot side and 30 minutes of dwell on the cold side. Test 3. Temperature pro¢le is between 0¯ and 75¯ C with 30 minutes of 15 minicycles at 5 ¯ C on the hot side and 30 minutes of dwell on the cold side. The temperature pro¢les are shown in Figures 1a^c. The temperature cycling period is 70 min/cycle. Humidity in the environmental chamber was 60% for all tests. Each test was performed independently on ¢ve BGA packages, and each sample was cycled 100 thermal cycles. For each sample, Moire¨ interferometry strain ¢eld measurement was performed at every 20 cycles. SEM micrographs were taken at the beginning and the end of cycling. Moire¨ fringes give us the inelastic strain accumulation, and SEM micrographs give us the microstructure coarsening behavior.

Moire¨ Interferometry The Moire¨ interferometry and the imaging systems used for submicron displacement measurement have been described in detail in [14, 15]. The most important feature of the MI system developed in this study is that it allows usage for fatigue studies. Until now Moire¨ has only been used for monotonic loading tests [16, 17] or for a cycling test up to four cycles [18]. A special optical grating replication technique developed during this study allows for measuring strain ¢elds under cyclic fatigue loading. The major advantages of MI are its high sensitivity, high resolution, and the whole ¢eld view of deformation distribution of the specimen surface. Brie£y, the optical diffraction grating is replicated on the specimen surface. The light from a single laser source is split into four coherent beams. The four coherent beams incident on the specimen grating at a particular angle so that the strongest diffraction order emerges at the direction perpendicular to the specimen grating surface. The diffracted beams recombine in the normal direction and create interference patterns.


1067

When the specimen surface deforms, the optical diffraction grating deforms with the specimen and the diffracted beams in the normal direction generate interference patterns that represent the in-plane displacement contour maps. When the two vertical beams are blocked, the two horizontal beams interfere and generate a horizontal deformation ¢eld, called the U-¢eld. When the two horizontal beams are blocked, the two vertical beams interfere and generate a vertical deformation ¢eld, called the V-¢eld. The one-dimensional setup of MI is shown in Figure 2a, and the optical table setup is shown in Figure 2b. This scheme applies to both the horizontal and vertical directions, so deformation in the two perpendicular directions can be obtained. The fringe pattern can be related to in-plane deformation quantitatively as given by [20] Uˆ

Nx f

…1†

Vˆ

Ny f

…2†

where U is the displacement in the x-direction; V is the displacement in the y-direction; f ˆ 2fs, where fs is the frequency of specimen diffraction grating. Nx is the horizontal fringe order and Ny is the vertical fringe order. In this project, fs ˆ 1200 lines/mm; thus, each fringe represents 1=f ˆ 0:417 mm in this study. Once the displacement data are available, the inelastic strains are computed by differentiation of the displacement distributions with respect to the two basic directions: horizontal …x†and vertical …y†. The strains are given by

ex ˆ

"

#

…3†

"

#

…4†

@ U ˆ 1 @ Nx f @x @x

@V ˆ 1 @Ny ey ˆ f @y @y

gxy ˆ

"

@U ‡ @V ˆ 1 @Nx ‡ @Ny f @y @y @x @x

#

…5†

Specimen Preparation The sample used in this study has a typical multilayered structure. The top layer is polymer with an inserted copper foil connector array. The bottom layer is ¢ber-reinforced printed wiring board (FR-4 PWB) substrate. The two layers are interconnected with BGA 63Sn^37Pb solder joints. In order to test ¢ve samples consistently, the boundary conditions should not vary from one sample to another during cycling testing. Furthermore, testing should simulate the actual service con-

1068

C. BASARAN ET AL.

(a)

(b) Figure 2. (a) Two-beam optical setup for Moire¨ interferometry; and (b) Moire¨ interferometry optical table.


1069

ditions. Based on these considerations, a special ¢xture is designed to hold the sample in boundary conditions identical to the in-service conditions. The sample and the ¢xture pair are shown in Figure 3. The specimen preparation for MI follows the same steps as described in [14, 15]. The sample is cut through the center of a row of BGA solder joints and then the cross section is polished, cleaned, and dried. An innovative optical diffraction grating replication procedure is developed to transfer a cross-line grating (1200 lines/mm) pattern onto the specimen cross-section.

Testing Procedure The testing procedure follows gradual steps for each specimen: . . . . . . .

Optical diffraction grating preparation; Initial Scanning Electron Microscope (SEM) picture of the sample; Aligning the initial null ¢eld of Moire¨ system; Registering the specimen position; Thermal cycling; Optical measurement of deformation ¢elds: every 20 cycles; Solder joint integrity is examination by SEM.

MI is used to record the deformation ¢eld. The typical Moire¨ fringe patterns are shown in Figures 4a^d for the U-¢eld and V-¢eld, respectively. The accuracy of MI measurement relies on the initial reference ¢eld. After the optical system is aligned for the null ¢eld, the optical system is protected from any position change by means of a register. A new specimen position register is designed to ensure that the specimen always occupies exactly the same optical space. A commercially available kinematic platform was modi¢ed to add functionality as a position register as well as a specimen positioning platform. Two translation stages are used to facilitate the adjustment of the initial position of the specimen on the optical table. To the best of our knowledge this is the ¢rst time ever in published literature that MI was used for fatigue testing. This was possible due to a new optical grating replication technique developed in this study. In order to prove the accuracy of the new replication technique and to support the test results, a supportive test on steel bar was done. This supportive test and its results are presented later in this article.

Figure 3. Specimen on ¢xture.

1070

C. BASARAN ET AL.

Figure 4. Moire¨ interferometry images: (a) U-¢eld before cycling, (b) U-¢eld after 100 cycles, (c) V-¢eld before cycling, and (d) V-¢eld after 100 cycles.


1071

RESULTS AND DISCUSSION During testing it was observed that the shear strain is very large compared to peeling strain. Therefore, shear strain is the major concern of the extended analysis and comparison in this study. To extract meaningful information from the raw test data and to ¢nd the global trend, the effect of some random factors needs to be reduced to a minimum. Such random factors include manufacturing defects, sample preparation, solder material itself, and voids in solder balls, which results in varying solder geometry, volume, material properties, and the strength of intermetallic connections. Furthermore, the even-numbered solder joints have a slightly different con¢guration from the odd-numbered solder joints due to the fact that the copper connectors joined to the neighboring solder joints are oriented in a different direction. Based on these considerations, the test data obtained from the ¢ve samples for each test were averaged based on the same type of cycling, the same joint number, and the same number of cycles. Comparisons of the inelastic shear strain accumulation of the three tests from joints 1^6 are shown in Figures 5a^5f. Figure 5a shows the inelastic shear strain accumulation during thermal cycling for solder joint 1. Results for test 1 (standard temperature time history), test 2 (temperature time history with 10¯ C £uctuations during high-temperature dwell) and test 3 (5¯ C temperature £uctuations at high-temperature dwell) are plotted in the same ¢gure for comparison purposes. The results indicate that a traditional industry-standard temperature pro¢le induces higher inelastic strain than test temperature pro¢les with £uctuation during the dwell time. The difference is more pronounced between tests 1 and 3. Also, the difference becomes more distinct as the number of cycles increases. But most importantly at the end of 100 cycles, the inelastic strain level reaches an asymptotic value for tests 2 and 3 while the test 1 curve has a much larger positive slope. Many researchers, such as in [12^14, 19, 20] have shown that inelastic strain is directly proportional to damage accumulation in a structure. Figures 5b^f show the inelastic strain accumulation during thermal cycling for solder joints 2^6. The strain is the largest in joint 1 (outer edge joint); hence, usually failure initiates at joint 1, for the con¢guration given in Figure 3. Figures 6a^c show the averaged shear strain distribution for each test at 20 cycles and 100 cycles. It is obvious that the shear strain distribution curves at 20 cycles and 100 cycles have very similar shape, which indicates a similar distribution trend. The factors affecting the distribution of the stress and strain in the solder joints show their in£uence at the very beginning and usually persist through the joint fatigue life. Therefore, improper packaging design or manufacturing process usually can be found at the early stage. Testing, if properly designed and conducted, should not signi¢cantly alter such a trend. Figures 7a,b show the averaged inelastic shear strain distribution in three tests at the end of 20 cycles and at the end of 100 cycles. By observation, at the end of the 100 cycles, test 1 developed higher inelastic shear strain in solder joints than tests 2 and 3. The difference between tests 1, 2, and 3 for solder joints numbered

1072

C. BASARAN ET AL.

Figure 5 Comparison of inelastic shear strain accumulation versus number of thermal cycles for different temperature pro¢les at solder joints 1^6.


1073

Figure 5 (continued). Comparison of inelastic shear strain accumulation versus number of thermal cycles for different temperature pro¢les at solder joints 1^6.

1074

C. BASARAN ET AL.

Figure 5. (continued). Comparison of inelastic shear strain accumulation versus number of thermal cycles for different temperature pro¢les at solder joints 1^6.


1075

Figure 6. Inelastic shear strain accumulation versus solder joint number at thermal cycles 20 and 100 for temperature pro¢les 1^3.

1076

C. BASARAN ET AL.

(a)

(b)

Figure 7. Inelastic shear strain accumulation versus solder joint number for tests 1^3 at the end of (a) 20 thermal cycles and (b) 100 thermal cycles.


1077

6 and higher is small because strain is zero in the middle of the package. Joint number 1 usually determines the reliability of the package. Therefore, results for other joints have no practical signi¢cance. Figure 8 gives the comparison of the maximum shear strain accumulation in the solder joints in three tests, which shows that test 1 has the maximum shear strain accumulation, test 2 gives the second largest, and test 3 has the smallest value. There are, we believe, two major reasons for this difference. The ¢rst one is due to load and the second one is due to metallurgical reasons. It is very well known that creep damage dominates the failure mechanism and creep is the largest component of inelastic strain in Pb/Sn solder joints [22, 23]. Creep is a plastic £ow process that best occurs under constant load with no reversal in direction. Constant shear load, which exists in the test 1 temperature pro¢le, is reversed due to temperature £uctuations in tests 2 and 3. The frequency of temperature £uctuations makes a difference in the level at which the creep process slows down. High-frequency £uctuations slow down the creep process more than low-frequency £uctuations. Therefore, test 3 has the smallest inelastic strain accumulation. The second reason for smaller inelastic strain accumulation in the presence of temperature £uctuations during dwell time is the fact that temperature £uctuations speed up the microstructural evolution and coarsening process in the solder joint. This latter statement was experimentally validated during this project (see Table 1). It is very well known that due to coarsening the solder material increases its resistance to the creep £ow process [24^26]. When the coarsening occurs faster, creep strain accumulation slows down. Hence, in Figure 8 we see that test 1 has a higher inelastic strain accumulation in 100 cycles as compared to tests 2 and 3.

Figure 8. Average inelastic shear strain accumulation versus number of thermal cycles.

1078

C. BASARAN ET AL.

Table 1 Pb/Sn Grain size coarsening (evolution)

Before test ( mm) After test ( mm) Increment (%)

Test 1

Test 2

Test 3

3.141 3.859 22.86

3.346 4.852 45.01

3.369 5.183 53.84

Microstructural Coarsening and Metallurgical Evidence Scanning electron microscope pictures were taken before testing and after 100 cycles in order to study the microstructural coarsening behavior of eutectic Pb/Sn solder joints. Test results explain the creep behavior of solder joints under three different cycling temperature pro¢les from the metallurgical point of view. In this study grain size was determined via the intersection method, as detailed in [27]. Test 1 has the smallest grain-size-coarsening ratio. Alloys with ¢ner grain size are easier to creep. On the other hand, tests 2 and 3 result in higher coarsening rates, which correspond to earlier stabilization in creep and microstructural evolution. Morris and Reynolds [29] have shown that manufactured eutectic slowly cooled Pb/Sn solder is a thermodynamically unstable material in parallel plate structure. As a result of the straining process it evolves into an equiaxed granular structure. These results are in agreement with earlier work by Kashyap and Murty [25].

A Supportive Test on a Steel Bar A supportive test on a steel bar has been done to verify the testing and measurement process. A description of this test and its results follow. There are several points that need to be veri¢ed regarding the MI technique developed in this project in order to provide a proof of the ¢delity of the test results. Such considerations include: . . . . .

Is the optical grating replication procedure developed actually working? Is there a relative deformation between grating and specimen? Is the temperature cycling appropriate for grating material? Is the measurement procedure well established? How stable is the optical system throughout the cycling fatigue test?

To answer these questions, a supportive test has been conducted on an A36 steel bar. A steel bar was polished ¢ne and £at. The same diffraction grating replication technique was used to replicate a layer of optical grating on its surface; see Figure 9. Before thermally cycling the steel bar, the MI optical system was aligned to obtain the initial null ¢eld using the specimen grating on the steel bar and the images of initial U-¢eld and V-¢eld were recorded. The observation area was marked (an area 12 mm £5 mm).


1079

Figure 9. Cross section of steel bar used for veri¢cation.

The steel bar was then taken away from the optical table for three consecutive thermal cycles with a more severe temperature pro¢le ranging between 125 ¯ C and ¡55¯ C, as shown in Figure 10. No constraints were applied to the steel bar through cycling. The steel bar was then taken back to the optical table to record the after-cycling fringe ¢eld images. The steel bar should behave elastically under such a thermal cycling because it is allowed to expand and contract freely. Furthermore, its melting point is above 1000 ¯ C, therefore, within the temperature range of ¡55¯ C to 125 ¯ C it should behave completely elastic. Based on such reasoning, the Moire¨ images before and after cycling should have negligible differences. Figures 11^14 show the initial V-¢eld, after-cycling V-¢eld, initial U-¢eld, and after-cycling U-¢eld, respectively. It is obvious that the initial fringe ¢elds and the after-cyclin g fringe ¢elds are almost the same. There are no fringes as a result of cycling. (The variation in darkness in the pictures is due to printing not fringes.)

Figure 10. Temperature time history.

1080

Figure 11. Initial V-fringe ¢eld.

Figure 12. V-¢eld after cycling.

Figure 13. Initial U-¢eld.

C. BASARAN ET AL.


1081

Figure 14. Fringe ¢eld after thermal cycling.

CONCLUSIONS Results indicate that industry-standard temperature pro¢les are conservative and underestimate the fatigue life of the BGA solder joints. From the data averaged on ¢ve samples for each test, the following conclusions can be inferred. . Temperature £uctuations during dwell time reduces inelastic strain accumulation on the ¢rst six solder joints by 17.1% in test 2 and 17.9% in test 3 (averaged for six solder joints). . Using minicycles during dwell time reduces maximum inelastic strain accumulation in solder joint 1 by 14.5% in test 2 and by 23.5% in test 3. Inelastic strain accumulation is directly proportional to fatigue life; therefore, using minicycles increases the BGA solder joint fatigue life in testing. Creep is the major characteristic of eutectic solder joint damage behavior; the temperature cycling pro¢le (frequency and holding time) signi¢cantly affects creep mechanism. Most of the inelastic strain is accumulated during the ¢rst 20 cycles; the increments become smaller and smaller through more cycles, gradually approaching a stable pointöthe upper limit. The global inelastic strain distribution trend is similar in all three tests. Minitemperature cycles reduce inelastic strain accumulation. Factors affecting local variation include void, geometry, actual design layout, solder microstructure, and specimen preparation.

REFERENCES 1. P. Hall, Forces, Moments and Displacements during Thermal Chamber Cycling of Leadless Ceramic Chip Carriers Soldered to Printed Wiring Boards, IEEE Trans. Comp. Hybrid Manuf. Tech., vol. 7, no 4, pp. 314^327 , 1984. 2. P. Hall and W. Sherry, Materials, Structures and Mechanics of Solder-Joints for Surface-Mount Microelectronics Technology, Proc. of the Lectures of the 3rd Int. Conf. on Tech. de connexion en electronique, Welding Society, Fellbach, Dusselfdorf, Germany, 1986. 3. J. H. Lau, D. W. Rice, and D. A. Avery, Elasto Plastic Analysis of Surface Mount Solder Joints, IEEE Trans. Comp. Hybrid Manuf. Tech., vol. CHMT-10, no. 3, 1987.

1082

C. BASARAN ET AL.

4. A. Solomon, Low Cycle Fatigue of 60/40 Solder Plastic Strain Limited vs. Displacement Limited Testing, Electron. Packaging: Mat. Processes, pp. 29^47, 1989. 5. D. Jeannotte and R. H. Goldman, Package Reliability, in Tummalo and Rymaszewski (eds.), Microelectronics Packaging Handboo k, Van Nostrand Reinhold, 1989. 6. A. Zubelewicz and L. M. Keer, Micromechanically Based Constitutive Modeling of Crystalline Materials, Int. J. Solids Structures, vol. 25, pp. 797^801 , 1989. 7. E. D. Riemer, Prediction of Temperature Cycling Life for SMT Solder Joints on TCE-Mismatched Substrates, IEEE Proc. Electron. Comp., pp 418^423 , Piscataway, NJ, 1990. 8. J. LeMaitre, A Course on Damage Mechanics, Springer-Verlag, New York, 1996. 9. T. Pan, Thermal Cycling Induced Plastic Deformation in Solder Joints: Part III: Strain-Energy Based Fatigue Life Model and Effects of Ramp Rate and Hold Time, Proc. ASME Winter Annual Meeting, New York, 1991. 10. Y. H. Pao, K. L. Chen, and A. Y. Kuo, A Nonlinear and Time Dependent Finite Element Analysis of Solder Joints in Surface Mounted Components under Thermal Cycling, Proc. Mat. Res. Soc. Symposium, vol. 225, 1991. 11. J. K. Tien, B. C. Hendrix, and A. I. Attarwala, Understanding the Cyclic Mechanical Behavior of Lead/Tin Solder, J. Electron. Packaging, vol. 113, pp. 115^120 , 1991. 12. D. R. Frear, H. Morgan, S. Burchett, and J. Lau, The Mechanics of Solder Alloy Interconnections, Van Nostrand Reinhold, New York, 1994. 13. B. Ozmat, A Nonlinear Thermal Stress Analysis of Surface Mount Solder Joints, Proc. 40th ECTC Conference, May 20^23, Las Vegas, NV, 1990. 14. C. Basaran and C. Yan, A Thermodynamic Framework for Damage Mechanics of Solder Joints, J. Electron. Packaging, vol. 120, no. 4, pp. 379^384 , 1998. 15. Y. Zhao, C. Basaran , A. Cartwright , and T. Dishongh, Thermomechanical Behavior of Micron Scale Solder Joints: An Experimental Observation, J. Mech. Behavior Mat., vol. 10, no. 3, pp. 135^146 , 1999. 16. Y. Zhao, C. Basaran , A. Cartwright , and T. Dishongh, Thermomechanical Behavior of Micron Scale Solder Joints under Dynamic Loads, Mech. Mat., vol. 32, no. 3, pp. 161^173 , 2000. 17. H. Xie, K. Satoshi, S. Liu, and D. Zou, Thermal Strain Measurement of Solder Joints in BGA Packages by Using Electron Moire Method, Proc. 99-IMECE/EEP-30, ASME Nashville, TN, Nov. 14^19, 1999. 18. B. Han, Deformation Mechanism of Two-phase Solder Column Interconnections under Highly Accelerated Thermal Cycling Condition: An Experimental Study, J. Electron. Packaging, vol 119, pp. 189^196 , 1997. 19. B. Han and Y. Guo, Determination of an Effective Coef¢cient of Thermal Expansion of Electronic Packaging Components: A Whole-Field Approach, IEEE Trans. Comp. Pack. Manuf. Tech., Part A, vol. 19, no. 2, pp. 240^247 , 1996. 20. K. C. Valanis, Some Thoughts on Thermodynamics of Internal Variables, Arch. Mech, vol. 49, no. 2, pp. 443^445 , 1971. 21. C. L. Chow and Y. Wei, Constitutive Modeling of Material Damage for Fatigue Failure Prediction, Int. J. Damage Mech., vol. 8, pp. 355^375 , 1999. 22. D. Post, B. Han, and P. Ifju, High Sensitivity Moire¨, Springer-Verlag, New York, 1994. 23. C. Basaran, A Comparison of Viscoplastic and Plastic Constitutive Models for Pb/Sn Solder Alloys, Structural Anal. Microelectron. Fiber Optics, EEP-vol. 16, pp. 149^154 , 1996. 24. Y. Wei, C. L. Chow, H. E. Fang, and M. K. Nielsen, Characteristics of Creep Damage for 60Sn^40Pb Solder Material, Proc. ASME 99-IMECE/EEP-15, 1999. 25. P. Kashyap and G. S. Murty, Experimental Constitutive Relations for the High Temperature Deformation of a Pb^Sn Eutectic Alloy, J. Mat. Sci. Engrg. vol. 50, pp. 205^213 , 1981. 26. P. Haswell and A. Dasgupta, Viscoplastic Characterization of Constitutive Behavior of Two Solder Alloys, Proc. ASME 99-IMECE/EEP-13, Nashville, TN, Nov. 14^19, 1999. 27. C. Basaran and R. Chandaroy, Mechanics of Pb40/Sn60 near Eutectic Solder Alloys Subjected to Vibrations, Appl. Math. Modelling, vol. 22, pp. 601^627 , 1998. 28. T. Sayama, T. Takayanagi , and T. Mori, Analysis of Phase Growth Process in Sn/Pb Eutectic Solder Joint, Adv. Electron. Packaging , vol. 126-1, ASME-EEP, pp. 13^21, 1999.


1083

29. J. W. Morris, Jr. and H. L. Reynolds, The In£uence of Microstructure on the Mechanics of Eutectic Solders, Adv. Electron. Packaging. EEP-vol. 19-2, pp. 74^81, 1992. 30. P. L. Hacke, A. F. Sprecher, and H. Conrad, Microstructure Coarsening during Thermomechanical Fatigue of Pb/Sn Solder Joints, J. Electron. Mat., vol. 26, pp. 774^782 , 1997.