Thermomechanical behavior of micron scale solder jointsunder dynamic loads

Y. Zhao a, C. Basaran a,*, A. Cartwright a, T. Dishongh b

a SUNY at Bu�alo, Electronic Packaging Laboratory, School of Engineering and Applied Sciences, 212 Ketter Hall, North Campus,

Amherst, NY 14260-4300, USAb Intel Corporation, Hillsboro, OR, USA

Received 21 July 1999

Abstract

Recent trends in reliability and fatigue life analysis of electronic devices have involved developing structural integrity

models for predicting the operating lifetime under vibratory and thermal environmental exposure. Solder joint reli-

ability is the most critical issue for the structural integrity of surface mounted electronics. Extensive research has been

done on thermal behavior of solder joints, however, dynamic loading e�ects to solder joint fatigue life have not been

thoroughly investigated. The physics of solder joint failure under vibration is still not very clear. This paper presents a

test program which was performed to study inelastic behavior of solder joints of BGA packages. A concurrent loading

unit is used which consists of a thermal environmental chamber and an electrodynamic shaker. Laser Moir�e Inter-

ferometry was used to measure the whole deformation ®eld of the prepared specimen surface. The corresponding in-

elastic strain ®eld is then calculated. It is found that at elevated temperature, vibration and shock can cause the

accumulation of inelastic strains and damage in solder joints. In this paper, contrary to the popular belief that all

vibration-induced strains are elastic, it is shown that vibration can cause signi®cant inelastic strains. Ó 2000 Elsevier

Science Ltd. All rights reserved.

Keywords: Solder joint; Electronic packaging; Vibration; Moir�e interferometry

1. Introduction

Thermal behavior of electronic packaging hasreceived extensive attention and has been studiedby many researchers. Due to its application pur-poses, electronic devices are manufactured asmultilayered structures with very di�erent me-chanical as well as electrical properties pertainingto each layer mostly designed to meet the re-

quirements of electronic service. Fast developingtechnologies bring in computer chips with veryhigh power density for better performance, whichmakes the thermal expansion mismatch problemeven severe. It has been generally recognized thatthermal loading is one of the major causes to thefailure of electronic devices during its service life.Especially the solder joints that act as mechanicalsupport as well as electrical interconnection, arevery important parts for the integrity of electronicdevices.

However, a number of categories of electronicdevices have to endure even severe working

Mechanics of Materials 32 (2000) 161±173www.elsevier.com/locate/mechmat

* Corresponding author. Tel.: +1-716-645-2114/2429; fax: +1-

716-645-3733.

E-mail address: cjb@eng.bu�alo.edu (C. Basaran).

0167-6636/00/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved.

PII: S 0 1 6 7 - 6 6 3 6 ( 9 9 ) 0 0 0 5 3 - 8

environments that involve not only thermal butalso dynamic loading conditions. The US AirForce estimates that vibration and shock cause20% of the mechanical failures in airborne elec-tronics. Empirical fatigue life prediction of com-ponents, component leads and joints has beendetermined based on the dynamic displacement ofprinted circuit board (PCB) (Marstein, 1987).However, the nonlinear stress±strain behavior ofsolder joint under vibration is still not clear. Dy-namic loading e�ect to solder joint fatigue life hasnot been systematically studied. And the role ofvibration in the life of solder joint has not beenrecognized su�ciently.

According to the literature survey, vibrationfatigue analyses which have been done by now aremainly model testing, mode analysis, and cyclicfailure investigation. Only very few existing fatiguemodels are based on stress or strain analysis.Steinberg (1988) developed an empirical relationto estimate the fatigue life of components mountedon a PCB based on testing and experience. Lau etal. (1990) studied solder joint reliability undershock and vibration. They conducted in-planerandom vibration testing, in-plane shock testing,out-of-plane vibration testing, and out-of-planeshock testing. Their testing was limited to deter-mine the failure status after certain cycles of dy-namic loading. Wong et al. (1991) reportedexperimental modal analysis and dynamic re-sponse prediction of PCBs under out-of-plane vi-bration. Ham and Lee (1996) also conductedreliability testing of electronic packaging undervibration. An experimental method was developedto measure the changes in electrical resistance inthe lead which is used to indicate a fatigue life.They discussed a relationship between the loadingforces and the high-cycle fatigue life for the leadpart of spider gullwing type surface-mountedcomponent. Liguore and Followell (1995) alsoreported a test program to obtain structural fa-tigue data for SMT solder joints exposed to anout-of-plane random vibration environment. Dy-namic response characteristics of the PCBs wasstudied, and empirical fatigue model for compo-nents, component leads and solder joints was dis-cussed in their paper. Sidharth and Barker (1995)reported vibration fatigue life estimation of corner

leads of peripheral leaded components using linearelastic ®nite element procedure.

This paper presents the results of a series oflaboratory tests dedicated to the investigation ofdynamic behavior of 63Sn/37Pb eutectic solderjoints in a new generation high performance CPUchip with ball grid array (BGA) con®guration. Thepurpose of this project is to study the nonlinearinelastic response of the solder joints under vi-bration at certain temperature, and the role ofdynamic loading to the fatigue life of the solderjoints.

The tested BGA package has the structure andmaterial composition shown in Fig. 1. In the shocktests, continued shock was applied at room tem-perature and at temperature range ramping from20°C to 130°C. In harmonic vibration tests, sine-wave vibrations were conducted at 20°C and100°C. The plastic deformation in each solder jointafter each shock and vibration test was measuredby laser Moir�e Interferometry.

2. Experimental procedures

There are two relatively independent parts inthe entire testing procedure. One is the loadingpart, and the other is the optical measurementpart. For the ®rst part, an environmental thermalchamber and an electrodynamic shaker are usedsimultaneously. For the second part a laser Moir�e

Fig. 1. Cross-section of the BGA specimen.

162 Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173

Interferometry device was designed and manufac-tured for deformation measurement at the reso-lution of half of the light wavelength (submicron).

3. Concurrent loading unit

A high capacity Super AGREE environmentalchamber by Thermotron (SA-36-CHV-30-30) isused to control the environmental temperature,which is shown in Fig. 2. It is capable to maintainthe temperature within the range of �1°C, and itsfastest ramping rate is 30°C/min.

The electrodynamic shaker system is UN-HOLTZ-DICKIE model S092HS, shown in Fig. 3.It consists of a power ampli®er, electrodynamicshaker, DC ®eld supply, and remote-coolingblower. The shaker armature incorporate a multi-ribbed table for maximum sti�ness and minimumweight, and is ideal for application requiring highg (up to 360 g shock and 100 g sine) level testinginvolved in this project. A personal computer isspeci®cally equipped for the shaker system for

function generating, test control, test monitoring,and data acquisition, and form a closed loop to-gether with the power ampli®er, shaker, acceler-ometer and charge ampli®er. It is capable ofperforming random vibration, sinusoidal vibra-tion, and shock.

To combine the shaker and the chamber, theshaker is inserted underneath the chamber and theslip table of the chamber is replaced by a special¯exible thermal barrier. An aluminum round platewith the same con®guration as the shaker arma-ture table is made to hold the specimen ®xtureinside the chamber. The plate is connected to thearmature table through the thermal barrier. Fig. 2shows the concurrent loading unit, and Fig. 4shows the special specimen ®xture for the con-current testing.

4. Moir�e Interferometry

The Moir�e Interferometry (MI) and the imagingsystems used for submicron displacement mea-surement have been described in detail in ZhaoFig. 2. Concurrent thermal-dynamic loading unit.

Fig. 3. Electrodynamic shaker.

Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173 163

et al. (1999). The major advantage of MI is its highsensitivity, high resolution, and the whole ®eldview of deformation distribution of the specimensurface. Brie¯y, the optical di�raction grating isreplicated on the specimen surface. The specimengrating di�racts the incident two coherent beamswith certain incident angle, and in the directionnormal to the specimen surface, two strong di�r-acted beams are obtained. When the specimensurface deforms, the optical di�raction gratingdeforms with the specimen, and the di�racted twobeams in the normal direction generate featureinterferometry pattern that represents the in-planedisplacement distribution. The one-dimensionalsetup of MI is shown in Fig. 5. This scheme appliesto both horizontal and vertical direction, so thatdeformation in the two perpendicular directionscan be obtained. The feature fringe pattern gen-erated by the vertical two beams represents thevertical deformation ®eld, and the fringe patterngenerated by the horizontal two beams representsthe horizontal deformation ®eld. Fig. 6 shows theoptical table for real MI designed for this project.

The fringe pattern can be related to in-planedeformation quantitatively as given by Post et al.(1994):

U � Nx

f; �1�

V � Ny

f; �2�

where, U is the displacement in x direction; V isthe displacement in y direction; f � 2fs, fs is thefrequency of specimen di�raction grating, in thisstudy, fs � 1200 lines/mm; Nx is the horizontalfringe order, and Ny is the vertical fringe order.Thus, each fringe represents 1=f � 0:417 lm inthis study.

Once the displacement data are available theplastic strains are computed by di�erentiation ofthe displacement distributions with respect to thetwo basic directions: horizontal (x) and vertical(y). The strains are given by

ex � oUox� 1

foNx

ox

� �; �3�

ey � oVoy� 1

foNy

oy

� �; �4�

cxy �oUoy� oV

ox� 1

foNx

oy

�� oNy

ox

�: �5�

5. Specimen and ®xture

The specimen preparation follows the similarsteps as described in Zhao et al. (1999). Thespecimen was cut by a high precision diamond

Fig. 5. Optical setup for Moir�e Interferometry.

Fig. 4. Specimen ®xture for the concurrent test.

164 Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173

wheel saw right through the center of the solderjoints of interest to expose the two rows of solderball joints. There are totally 30 joints in the bottomrow, and 24 joints in the upper row. The exposedcross-section was then polished ¯at, cleaned anddried thoroughly before a thin layer of epoxy wasapplied on the cut surface to transfer the di�rac-tion grating pattern to the specimen surface.

A special ®xture is designed to hold the speci-men tightly at two ends of the middle FR-4 layer.This ®xture is then mounted on the shaker arma-ture table. The boundary condition and the load-ing direction are shown in Fig. 7. The accelerationfor all the tests is applied in-plane. The estimatednatural frequency of the chip specimen under thiscon®guration is well above 1000 Hz (Lau et al.,1990; Chandaroy, 1998; Basaran and Chandaroy,1999a,b).

6. Test results and discussion

Both shock and harmonic sine vibration wereconducted in this project.

6.1. Results of shock testing

Two sets of testing were done to examine thedynamic response of BGA solder joints. In the ®rstset of test, shock pulses were continuously appliedat room temperature 20°C for 5 min. In the otherset of tests, shock pulses were applied with con-current temperature ramping from 20°C to 130°Cfor 5 min, and then the specimen was air cooled toroom temperature. The same type of shock pulsesare applied continuously with peak acceleration10g, shock duration 11 ms, total duration 97 s, anddelay 103 ms. The loading period is 5 min.

The results of room temperature tests areshown in Fig. 8(a)±(d). Since the specimen istightly ®xed at two ends for the purpose of testing,the initial deformation ®eld is not nulled at thearea close to the end. The initial ®eld is subtractedfrom the ®nal deformation ®eld, therefor the re-sults should not be a�ected by the initial ®eldFig. 7. Boundary condition and dynamic loading direction.

Fig. 6. Optical table for Moir�e Interferometry.

Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173 165

which has trivial strains. The measured deforma-tion ®elds after 5 min room temperature continu-ous shock testing are given in Fig. 8(c) and (d).These ®gures show that little change has occurredduring the testing which means little inelastic de-

formation happened. Obviously, the dynamic re-sponse of the solder joints at room temperature inthe shock testing is elastic.

The results in the second set of tests are given inFig. 9(a)±(d). The bottom row of solder joints

Fig. 8. (a) Initial U-®eld before shock testing, (b) U-®eld after 5 min shock at room temperature, (c) initial V-®eld before shock testing

and (d) V-®eld after 5 min shock at room temperature.

166 Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173

Fig. 9. (a) Initial U-®eld before concurrent thermal and shock testing, (b) U-®eld after concurrent thermal and shock testing, (c) initial

V-®eld before concurrent thermal and shock testing and (d) V-®eld after concurrent thermal and shock testing.

Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173 167

developed signi®cant amount of plastic deforma-tions. Since the bottom row is between the FR-4layer and the polymer layer, whose coe�cients ofthermal expansion (CTE) are very di�erent, thethermal e�ect is very large, and one cannot tell thedynamic e�ect by now. However, the ®rst row ofsolders seem to show some dynamic e�ect. Thisrow of solders are between two identical materiallayers-FR4, their CTE's are practically the same,whereas their fringe pattern shows signi®cantamount of plastic deformation.

6.2. Results of harmonic vibration testing

Also two sets of tests were done to examine thevibration response and the e�ect of nonlinear re-sponse to solder joint fatigue life under harmonicvibration. One set of test was conducted at tem-perature 20°C (which is 0.64 of the melting tem-perature), and the other tests are conducted at100°C (which is 0.82 of the melting temperature).Each test lasted 42 min. Di�erent frequency valueswere applied during each set of tests. The appliedvibration is of sine wave with peak acceleration of30g in all cases.

7. Room temperature (20°C) testing

In the ®rst set of tests, harmonic sine wave vi-brations were applied with frequencies of 2000,500, and 50 Hz in sequence. The deformation re-sponses indicate that the solder joints in this BGApackage behave elastically under vibrations atroom temperature.

The MI fringe patters that represent the defor-mation ®elds of the ®rst set of tests are shown inFig. 10(a)±(f). It can be seen that the fringe pat-terns do not change from the initial ®elds. Theslight di�erences of the fringe pattern at solderballs are in the error range of MI.

8. Elevated temperature (100°C) testing

Since most semiconductor devices experienceheat dissipation during service, studying the e�ectof vibrations at elevated temperatures is more re-

alistic. For the second set of tests, the specimenwas heated to 100°C and then was subjected tosine wave vibration at the same temperature. Sincethe specimen is a multilayered structure, and thetwo ends of the middle board were ®xed, heatingprocess introduces thermal stresses and thermalstrains. Thermal strains must be subtracted in or-der to study the vibration-induced deformation.Therefore, the specimen was ®rst heated from20°C to 100°C in 5 min, and then was hold at100°C for another 6 min to allow full relaxation ofthermal stresses. It was then air cooled to roomtemperature for measurement of the initial defor-mation ®elds. Therefore, the initial ®elds actuallyrecorded the plastic deformation caused by heat-ing. Thermal strains were subtracted from elevatedtemperature vibration test strains to get the vi-bration induced plastic deformation.

The specimen was then put back to 100°C forfurther testing. 50, 500, 1000 and 700 Hz fre-quencies were applied in sequence. And the de-formation fringe patterns for each test are shownin Fig. 11(a)±(j). It is obvious, that the fringepatterns change, and deformation became largeras the fringe patterns become denser. Fig. 12shows the calculated net shear strain induced ineach vibration test at 100°C at di�erent frequen-cies. It seems that beyond 1000 Hz, the solderjoints subject to vibration behave elastically.However, at lower frequencies, signi®cant plasticstrain is developed under vibration. Speci®cally,smaller frequencies cause larger plastic deforma-tion as shown in Fig. 12. However, the lower fre-quency limit of 700 Hz needs to be furtherexplored. It should be pointed out that 700 Hz isvery close to 750 Hz reported by Basaran andChasaroy (1998) for the boundary between elasticand plastic behavior. In Fig. 12 solder joint num-ber is in the increasing order from the most sideone towards the center ones. In such laminatedstructures, outer solder joints experience largershear stress and strain than the center parts (Ba-saran and Zhao, 1999). Lau and Rice (1985) haveshown that the modulus of elasticity increases withfrequency and decreases with increased tempera-ture. At higher temperature, solder material alsosoftens, and yield point is much lower than that atroom temperature. Therefore with the same

168 Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173

Fig. 10. (a) U-®eld after 42 min 2000 Hz dynamic cycling at 20°C, (b) V-®eld after 42 min 2000 Hz dynamic cycling at 20°C, (c) U-®eld

after 42 min 500 Hz dynamic cycling at 20°C, (d) V-®eld after 42 min 500 Hz dynamic cycling at 20°C, (e) U-®eld after 42 min 50 Hz

dynamic cycling at 20°C and (f) V-®eld after 42 min 50 Hz dynamic cycling at 20°C.

Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173 169

Fig. 11. (a) Initial U-®eld of isothermal temperature at 100°C, (b) initial V-®eld of isothermal temperature at 100°C, (c). U-®eld after

42 min dynamic cycling of 50 Hz at 100°C, (d) V-®eld after 42 min dynamic cycling of 50 Hz at 100°C, (e) U-®eld after 42 min dynamic

cycling of 500 Hz at 100°C, (f) V-®eld after 42 min dynamic cycling of 500 Hz at 100°C, (g) U-®eld after 42 min dynamic cycling of 1000

Hz at 100°C, (h) V-®eld after 42 min dynamic cycling of 1000 Hz at 100°C, (i) U-®eld after 42 min dynamic cycling of 700 Hz at 100°C

and (j) V-®eld after 42 min dynamic cycling of 700 Hz at 100°C.

170 Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173

Fig. 11. Continued.

Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173 171

frequency, solder material responds plastically athigh temperature while it responds elastically atlower temperature. Therefore, vibration at highhomologous temperature is very harmful to solderjoint fatigue life.

Essentially the solder behaves elastically forhigher frequencies and inelastically for lower fre-quencies. This observation is consistent with theexperience reported by Steinberg (1988) ``Solderjoints that are exposed to low-rate alternatingstresses will fail more quickly than solder jointsthat are exposed to high-rate alternating stress''.Basaran and Chasaroy (1998) have shown that thisis due to the fact that at lower frequencies theperiod of the loading is higher. The time depen-dent inelastic deformations are directly related tothe period of the load. When the load has a longerperiod the material has more time to creep. Con-sequently for vibrations with small frequenciescreep dominates the response.

9. Conclusion

Both shock and harmonic vibration tests wereconducted at room temperature and at elevatedtemperature with di�erent acceleration and fre-quency levels. Testing is performed on real lifeelectronic packages that provide true perspectiveof reliability and failure analysis of BGA solderjoints. MI is a very useful and highly sensitivemethod in measuring submicron deformation andproviding a whole ®eld distribution ofdeformation.

It can be summarized from the test data thatsolder joints respond elastically at room tempera-ture under either shock or sine vibration loading.

Inelastic deformation may be still accumulatingalbeit very slowly. At elevated temperature, bothshock and vibration induce signi®cant inelasticshear deformation in solder joints, thus shortenthe fatigue life of solder joints. The solder jointssu�er more severe damage at lower vibration fre-quencies than at higher ones. At 100°C, the testdata show that dynamic loading with frequencieshigher than 1000 Hz have little e�ect to solderinelastic response.

Thus it can be concluded that contrary topopular belief, the solder alloy does not remain inthe elastic region. With a low melting point(183°C) and high viscoplastic characteristics, eu-tectic solder alloy shows creep behavior duringdynamic loading process which should not be un-derestimated in solder joint fatigue life prediction.

Acknowledgements

This research project is sponsored by the De-partment of Defense O�ce of Naval ResearchYoung Investigator Award to Dr. C. Basaran.Helpful discussions with Dr. Roshdy Barsoum,Director of Solid Mechanics Program at ONR aregratefully acknowledged.

References

Basaran, C., Chadaroy, R., 1998. Mechanics of Pb40/Sn60 near

eutectic solder alloys subjected to vibrations. Applied

Mathematical Modeling 22, 601±627.

Basaran, C., Chadaroy, R., 1999a. Nonlinear dynamic analysis

of surface mount interconnects: Part I. Theory. ASME

Journal of Electronic Packaging 121 (1), 8±12.

Basaran, C., Chadaroy, R., 1999b. Nonlinear dynamic analysis

of surface mount interconnects: Part II. Applications.

ASME Journal of Electronic Packaging 121 (1), 12±18.

Basaran, C., Zhao, Y., 1999. Closed form vs. ®nite element

analysis of laminated stacks. International Journal of

Finite Elements in Analysis and Design, accepted for

publication.

Chandaroy, R., 1998. Damage mechanics of microelectronic

packaging under combined dynamic and thermal loading.

Ph.D. dissertation, SUNY at Bu�alo.

Ham, S.-J., Lee, S.-B., 1996. Experimental study for reliability

of electronic packaging under vibration. Experimental

Mechanics 36 (4), 339±344.

Fig. 12. Plastic strain vs. frequency.

172 Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173

Lau, J.H., Rice, D.W., 1985. Solder joint fatigue in surface

mount technology: state of the art. Solid State Technology,

91±104.

Lau, J., Powers-Maloney, L.M., Baker, J.R., Rice, D., Shaw,

B., 1990. Solder joint reliability of ®ne pitch surface mount

technology assemblies. IEEE Transactions on CHMT 13

(3), 534±544.

Liguore, S., Followell, D., 1995. Vibration fatigue of surface

mount technology (SMT) solder joints. In: 1995 Proceed-

ings Annual Reliability and Maintainability Symposium,

pp. 18±26.

Markstein, H.W., 1987. Designing electronics for high vibration

and shock. Electronic Packaging and Production, 40±43.

Post, D., Han, B., Ifju, P., 1994. High Sensitivity Moir�e.

Springer, New York.

Sidharth, Barker, D.B., 1995. Vibration induced fatigue life

estimation of corner leads of peripheral leaded compo-

nents. In: 1995 ASME International Mechanical Engineer-

ing Congress and Exposure.

Steinberg, D.S., 1988. Vibration Analysis for Electronic

Equipment, Wiley, New York.

Wong, T.-L., Stevens, K.K., Wang, G., 1991. Experimental

model analysis and dynamic response prediction of PC

boards with surface mount electronic components. Journal

of Electronic Packaging 113, 244±249.

Zhao, Y., Basaran, C., Cartwright, A., Dishongh, T., 1999. An

experimental observation of thermomechanical behavior of

BGA solder joints by Moir�e Interferometry. Journal of

Mechanical Behavior of Materials, 10 (3), 135±146.

Y. Zhao et al. / Mechanics of Materials 32 (2000) 161±173 173