Effect of Size of Lid-Substrate Adhesive on Reliability of Solder Balls in Thermally Enhanced Flip Chip PBGA Packages

718 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 29, NO. 4, DECEMBER 2006

Effect of Size of Lid-Substrate Adhesive onReliability of Solder Balls in Thermally

Enhanced Flip Chip PBGA PackagesYi-Ming Jen, Chih-Kai Fang, and Yun-Hsin Yeh

Abstract—Six design cases of lid-substrate adhesive with variouscombinations of widths and heights were analyzed to investigatehow the size of the adhesive affects the reliability of the solderballs of thermally enhanced flip chip plastic ball grid array (FC-PBGA) packages in thermal cycling tests. Analysis results werecompared with data on the reliability of conventional FC-PBGApackages. Thermal-mechanical behavior was simulated by the fi-nite element (FE) method and the eutectic solder was assumed toexhibit elastic-viscoplastic behavior. The temperature-dependentnonlinear stress/strain relationship of the adhesive was experimen-tally determined and used in the FE analysis. Darveaux’s modelwas employed to obtain the predicted fatigue life of the solder ball.Simulation results reveal that the fatigue life of the solder balls inthermally enhanced FC-PBGA packages is much shorter than thatin conventional FC-PBGA packages, and the life of solder balls in-creases with both the width and the height of the adhesive. How-ever, the effect of the width of the adhesive on the reliability of thesolder ball is stronger than that of the height. Moreover, increasingeither the width or the height reduces the plastic strain in the adhe-sive at critical locations, indicating that the reliability of the adhe-sive can be improved by its size. The predicted results of the life ofsolder balls for some selected studied packages are also comparedwith experimental data from thermal cycling tests in the paper.

Index Terms—Finite element (FE) method, flip-chip plasticball grid array (FC-PBGA), lid-substrate adhesive, reliability,thermal cycling test, thermal enhancement, thermal fatigue life,viscoplastic.

I. INTRODUCTION

AMONG the advantages that flip chip packages have overconventional packaging include higher input/output (I/O)

density, shorter interconnection, smaller size, lower dielectricconstant, and higher reliability. Conventional flip chip deviceshave been developed to be mounted on expensive ceramic sub-strates because the mismatch between the coefficients of thermalexpansion (CTE) of the silicon chip and of the ceramic sub-strates appears to be low. However, the flip-chip plastic ball gridarray (FC-PBGA) package has recently attracted more interest

Manuscript received March 19, 2004; revised Revised June 23, 2006. Thiswork was supported by the National Science Council of Taiwan, R.O.C., underContract NSC 93-2212-E-216-008. This paper was recommended for publica-tion by Associate Editor R. Lee upon evaluaton of the reviewers’ comments.

Y.-M. Jen is with the Department of Mechanical Engineering, Chung HuaUniversity, Hsinchu 30012 Taiwan, R.O.C. (e-mail: [email protected]).

C. K. Fang is with the Department of Breaker Engineering, Shihlin Electricand Engineering Corporation, Hsinchu, Taiwan, R.O.C.

Y.-H. Yeh is with the Assembly Process Development Department, Pow-ertech Technology, Inc., Hsinchu, Taiwan, R.O.C.

Digital Object Identifier 10.1109/TCAPT.2006.885927

than the flip-chip ceramic ball grid array (FC-CBGA) packageowing to its higher second-level reliability and lower cost. Ba-sically, the reliability of flip chips on organic substrates is themain factor in determining the design because the CTE mis-match between the silicon die and the plastic substrate is large.The space between the die and the substrate is filled with un-derfill encapsulants to reduce the thermal stress/strain in the C4solder bumps and thus overcome this shortcoming. However,FC-PBGA packages have other failure modes and some experi-mental and numerical simulation methods have been applied toelucidate them. Previous studies have investigated how parame-ters such as the size of the chip, the edge distance, the thicknessof the substrate, the height of the bump, the underfill characteris-tics and the CTE of the substrates, affect the thermal fatigue lifeof the solder balls that connect the substrate to the printed circuitboard (PCB) [1]–[6]. The failure modes of die edge crackingand underfill cracking have been demonstrated to be associatedwith the underfill properties, the size of the underfill fillet, theangle of the fillet, the thickness of the substrate, the initial edgedefects in the chip, the C4 height, global and local CTE mis-matches, the hygroscopic swelling, and the absorption of mois-ture [7]–[10]. Additional efforts have attempted to enhance in-vestigative methods, including integrating flow-thermomechan-ical and reliability analysis of FC-PBGA and FC-CBGA pack-ages in power cycle tests [11], [12].

Packages are small and the interconnections withinFC-PBGA packages are dense, explaining why effectivelymanaging the heat transfer is important in the area of de-sign. Some thermal enhancements have been developed forFC-PBGA packages to improve thermal performance; theseinclude heat spreaders, metallic lids, molding compounds, andheat sinks. The efficiency of heat transfer and flow for thesetypes of thermally enhanced packages has been addressedelsewhere [13], [14], using computational fluid dynamicstools. However, thermal enhancement alters the stiffness ofthe packaging structure and redistributes the stress/strain,which is governed by the mutual constraints and the mis-match between CTE of the enhanced parts and the other partsof the packages in thermal environments. Aside from thoseof the reliability of conventional thermally enhanced pack-ages [15]–[18], thermal-mechanical reliability of thermallyenhanced FC-PBGA packages has seldom been studied. Apioneering work established that the shape of the fillet, CTE,and the modulus are important in determining the reliabilityof FC-BGA packages with heat spreaders [19]. However, inthat study, the property of adhesive was assumed to be linearlyelastic in a temperature-dependent fashion, and the effect of

1521-3331/$20.00 © 2006 IEEE

JEN et al.: EFFECT OF SIZE OF LID-SUBSTRATE ADHESIVE 719

these factors on the thermal fatigue life of solder balls wasunclear. Exactly how the local design parameters, including thematerial characteristics of thermal enhancements, the propertiesof the adhesive, the size of the adhesive, and the shape of thefillet affect the reliability of the solder joint remains unknown.The failure of adhesive at a high temperature becomes anotherproblem that is associated with package assembly. The delam-ination of the lid-adhesive or the substrate-adhesive, and thefatigue failure of the adhesive have been observed to occur oftenduring thermal cycling tests (TCTs) and thermal shock tests(TSTs). A practical means of solving the problem of the fatigueof adhesive is to increase its height and width, but doing soincreases the cost of the lid-substrate adhesive. Therefore, thesize of the adhesive required in designing thermally enhancedFC-PBGA packages is an important area of study.

This work studies the effect of the size of the lid-substrate ad-hesive on the reliability of the solder balls of FC-PBGA pack-ages with a copper lid in TCT. The stress/strain behavior ofthe packages in TCT is simulated using the finite element (FE)method. The temperature-dependent mechanical characteristicsof the adhesive are experimentally determined and used in theanalysis because the nonlinear characteristics of a resin-basedadhesive are apparent, especially at the operating temperature.The induced thermal stress/strain is important to the thermal fa-tigue failure of both the solder balls and the adhesive in the TCT.Six combinations of widths and heights of the adhesive are se-lected to investigate the effect of size on the fatigue life of thesolder ball. Finally, the reliability of the adhesive in variouslysized designs is studied by investigating the effective plasticstrain, thus providing a parametric understanding of the relia-bility of the design.

II. PROBLEM FORMULATION

Fig. 1(a) and (b) present the FC-PBGA packages studiedherein. The schematics are not to scale to save space.Fig. 1(a) reveals that the conventional FC-PBGA packagewas constructed by flip chip bonding the silicon die to a Bis-maleimide–Trianze (BT) epoxy resin substrate with 55 55 C4bump interconnects with a pitch of 0.22 mm. The epoxy-typeunderfill was used to fill the space between the chip and sub-strate, to reduce the thermal stress/strain associated with thethermal mismatch between the die and substrate. The substratewas attached to the FR-4 PCB using 27 27 full-matrix63Sn/37Pb eutectic solder balls with a pitch of 1.27 mm. Thesolder balls were on the solder mask-defined (SMD) copperpads. Table I lists the relevant dimensions.

The thermally enhanced FC-PBGA package was developedby mounting a copper lid on the back of the chip using thermalpaste, which provides paths for the removal of heat from the die,as depicted in Fig. 1(b). The lid was simultaneously mountedperipherally on the organic substrate using bismaleimide resinwith boron–nitride-filled, non-electrically conducting lid-sub-strate adhesive. Given a fixed length of each edge of the lid,six design cases with various peripheral adhesive heights andwidths were investigated to study the effect of the size of theadhesive on the fatigue life of the solder balls. Fig. 1(b) showsthe dimensions of the adhesive utilized in the six design cases.Notably, the peripheral width of the lid changed with the width

Fig. 1. Schematics of (a) the conventional FC-PBGA package and (b) the ther-mally enhanced FC-PBGA package.

TABLE IDIMENSIONS AND MATERIAL PROPERTIES OF PACKAGING MATERIALS

of the adhesive, and the space between the lid and the chip wasfilled with thermal paste.

III. ANALYSIS

The FE method was initially used to simulate the elastic-vis-coplastic behavior of the packages in CTC. The FE code adoptedto perform the simulation was ANSYS [20]. A three-dimen-sional (3-D) 1/8th model was used for FE analysis because ofsymmetry. Fig. 2(a) and (b) show the example FE models of theconventional FC-PBGA package and the thermally enhancedFC-PBGA package of case D, respectively. The model did notincorporate elements of the thermal paste because the paste hadlow stiffness. Eight-node solid elements were used in the anal-ysis. Table II lists the numbers of nodes and elements used inthe FE models. Notably, the mesh density of the lid-substrateadhesive was the same in all cases, to enable a fair compar-ison to be made. The assumed boundary condition was that theout-of-plane displacements were constrained on the planes of


Fig. 2. FE model of (a) the conventional FC-PGBA package and (b) the ther-mally enhanced FC-PGBA package in case D.

TABLE IINUMBERS OF NODES AND ELEMENTS USED IN THE FE MODELS

symmetry of the 1/8th model. Furthermore, the displacementsin all directions of the bottom node on the intersectional linesegment of the two planes of symmetry were also constrained.

The material properties of die, the copper pad, and thecopper lid were assumed to be linearly elastic, independent oftemperature. The coefficients of thermal expansion (CTE) ofthe above materials were constants. The properties, includingYoung’s Modulus, Poisson’s ratio and the coefficient of thermalexpansion (CTE), of the BT substrate and those of FR-4 PCBin in-plane directions differed from those in out-of-plane di-rections. Table I presents the thermal and linear mechanicalproperties of all packaging materials.

The property of the polymer-based compound underfill usedin the study was assumed to be linear-viscoelastic, as presentedby Feustel et al. [21]. The time and temperature dependenttensile behavior can be described by a series of generalized

TABLE IIIPARAMETERS OF VISCOELASTIC UNDERFILL MODEL [21]

Fig. 3. Temperature-dependent stress-strain curves for lid-substrate adhesive.

Maxwell models:

(1)

where and are the tensile modulus and the retardation timeof the th Maxwell model at time , and is the tensile mod-ulus as time approaches infinity. In (1), three Maxwell models

3 were used to express the time and temperature depen-dent tensile property, and and are given by the followingrelation:

(2)

(3)

where is the activity energy; is the gas constant, andis the absolute temperature. The corresponding parameters

expressed in (1)–(3) are found elsewhere [21] and listed inTable III.

The lid-substrate adhesive was assumed to exhibit temper-ature-dependent elastic-plastic behavior, in the FE analysis.Fig. 3 plots the nonlinear stress-strain curves at various temper-atures, experimentally obtained using strip-shaped specimenstested on a microtester machine at a strain rate of 0.001 1/s. Allcurves were simulated multi-linear types in the elastic-plasticFE analysis. The kinematic hardening rule and the von-Misesyield criterion were applied in the nonlinear analysis.

The nonlinear FE analysis of 63Sn/37Pb eutectic solderbumps and solder balls involved Anand’s viscoplastic consti-tutive behavior. The strain hardening and softening behavior isdescribed by one flow rule and three corresponding evolutionrules, which can be expressed as follows.

The flow rule is

(4)


TABLE IVPARAMETERS USED IN ANAND’S CONSTITUTIVE

MODEL OF 62Sn/37Pb/2Ag SOLDER [22]

Fig. 4. Temperature profile in thermal cycling test.

and the three evolution equations are

(5)

(6)

(7)

where is the plastic strain and is the stress. For the em-ployed eutectic solder, the constants in (4)–(7) are given else-where [22] and presented in Table IV.

Fig. 4 displays temperature profile in the thermal cycling test.The temperature ranges from 55 C to 140 C, and both theramp time and the dwell time are 5 min. Thermal loading beganat 25 C and the FE models were assumed to be stress-free at thistemperature. The isothermal simulation was used for the tem-perature of the package was assumed to be uniform. The non-linear behavior of the package was simulated over three thermalcycles, since in each case, the stress-strain behavior tended tobe stable after three thermal cycles. The simulated stress/strainresults in the third cycle were used in subsequent fatigue lifeanalysis.

According to the FE results based on Anand’s viscoplastictheory, the predicted characteristic life to 63.2% failure of the

TABLE VPARAMETERS USED IN DARVEAUX’S MODEL [22]

critical solder ball can be obtained by adding the crack initia-tion life with the crack propagation life. In the expression ofDarveaux’s model [22], the prediction parameter is the plasticstrain energy density accumulated in the stable cycle. The av-erage plastic strain energy density accumulated during the stablecycle for the interfacial layer elements of the critical ball, calcu-lated by the element volume averaging technique, was used asDarveaux’s prediction parameter to solve the problem of mesh-sensitivity, and yield the predicted life of the solder ball. Theaverage plastic strain energy density accumulated in the stablecycle for interface elements is given by

(8)

where is the plastic strain energy density of the thinterfacial layer element accumulated in the stable cycle, andis the volume of the th interfacial layer element. The predictedcrack initiation life and the crack propagation rateare related to the average plastic strain energy density

(9)

(10)

where is the crack length and is the number of cycles. Theconstants , , , and in (9) and (10) are the parametersthat are dependent on the analysis variables used in the FE codeANSYS. According to Darveaux’s study [22], these dependedanalysis variables include the FE model configuration, the sim-ulation method, the constitutive model of the eutectic solder,the code version, the temperature change command, the timestep scheme and the thickness of the interfacial element. In thepresent study, a 1/8 symmetric model was established. Anand’sconstitutive model in ANSYS 7.0 was used to describe the non-linear behavior of solder ball. The BF command was applied tospecify the ramped change in temperature, and the fine time stepscheme was utilized in the thermal loading condition. The thick-ness of the first interfacial layer element was 0.0125 mm (about0.5 10 in). Table V gives the corresponding parameters inDarveaux’s model under these analytic conditions. Combiningthe crack initiation life with the crack propagation life yields acharacteristic life to 63/2% failure for the critical solder ballthat is given by

(11)

where is the diameter of the interface between the solder balland the copper pad ( 0.466 mm).

IV. RESULTS AND DISCUSSION

Fig. 5(a) and (b) present the deformation shapes of theconventional FC-PBGA package and the thermally enhanced


Fig. 5. Deformation of (a) conventional FC-PBGA package and (b) thermallyenhanced FC-PBGA package in case D at the end of the high-temperature dwellin the third cycle.

FC-PBGA package in case D, respectively, at the end of thehigh-temperature dwell in the third cycle. The shadowed andunshadowed contours represent the deformed and undeformedshapes, respectively. The deformed shape of the conventionalFC-PBGA package is concave, while that of the lid of thethermally enhanced FC-PBGA packages is convex, becausethe CTE of the copper lid exceeds that of the substrate. Thesimulated deformation status, the von Mises stress and theequivalent inelastic strain imply that the most critical locationof the conventional FC-PBGA package is the upper-right cornerof the fifth solder ball on the diagonal plane. This critical ballis under the edge of the chip. The most critical locations ofthe six thermally enhanced FC-PBGA packages are all theupper-right corners of the balls, furthest from the neutral pointon the diagonal plane.

Fig. 6(a) and (b) plot the history of von Mises stress at themost critical locations of all studied FC-PBGA packages. Thevon Mises stress of thermally enhanced FC-PBGA packagessignificantly exceeds that of conventional FC-PBGA packages.The height of the adhesive in the thermally enhanced FC-PBGApackage is found to affect very slightly the stress behavior, givena particular width of the adhesive, because the correspondingcurves in the figures are almost convergent. However, a com-parison of Fig. 6(a) and (b), which have the same scale, showsthat increasing the width of the adhesive from 2 to 3 mm reducesthe von Mises stress.

Fig. 7(a) and (b) plot the von-Mises stress/plastic strain hys-teresis loops at the most critical locations of the solder balls incases with fixed adhesive widths of 2 and 3 mm, respectively.The shape and the area of the loop in the third cycle tend tobe stable. The area of the stress/strain loop in the conventionalpackage is much smaller than the areas of the loops for the ther-mally enhanced packages. The areas of the loops of the packages

Fig. 6. Von Mises stress history at the most critical locations of the solder ballsin all studied packages with (a) 2-mm-wide adhesive and (b) 3-mm-wide adhe-sive.

with adhesive widths of 3 mm are smaller than those with adhe-sive widths of 2 mm, indicating that increasing the width of theadhesive reduces the fatigue damage of the solders during onecycle, from the perspective of strain energy density.

Fig. 8(a) and (b) plot the behaviors of accumulative effectiveplastic strain at the critical locations in cases with fixedadhesive widths of 2 and 3 mm, respectively. The accumula-tive effective plastic strain at a given time can be obtained bysumming the increments of the effective plastic strain in eachloading step

(12)

where is the increment of the von-Mises type effectiveplastic strain in the ith step. Fig. 8(a) and (b) show that the accu-mulative effective plastic strains at the most critical locations ofthe solder balls in thermally enhanced FC-PBGA packages ex-ceed those of the conventional FC-PBGA packages, apparentlybecause the copper lid constrains the thermal expansion of the


Fig. 7. Von Mises stress-plastic strain hysteresis loops at the most critical lo-cations of the solder balls in all studied packages with (a) 2-mm-wide adhesiveand (b) 3-mm-wide adhesive.

substrate, increasing the deformation of the solder balls locatedbetween the substrate and PCB.

Comparing the simulated plastic strains of six thermally en-hanced FC-PBGA packages with variously sized adhesives, aspresented in Fig. 8(a) and (b), reveals that for a fixed adhe-sive width, the accumulative effective plastic strain at the crit-ical location slightly increases as the height of the lid-substrateadhesive decreases. Comparing Fig. 8(a) and (b), which havethe same scale, demonstrates that increasing the adhesive widthmarkedly reduces the effective plastic strain. The results revealthat enlarging the adhesive typically reduces the thermal stress/strain, and the effect of the adhesive width on the stress/strainbehavior of the solder ball exceeds apparently that of the adhe-sive height.

Fig. 9(a) and (b) shows the behaviors of the accumulativeplastic strain energy density at the critical locations incases with fixed adhesive widths of 2 and 3 mm, respectively.With a definition that is similar to that of accumulative effec-tive plastic strain, the accumulative plastic strain energy density

is obtained by adding the increments of plastic strain en-

Fig. 8. Accumulative effective plastic strain behavior at the most critical loca-tions of the solder balls in all studied packages with (a) 2-mm-wide adhesiveand (b) 3-mm-wide adhesive.

ergy density in all loading steps. It is given by

(13)

where is the increment of viscoplastic strain energy den-sity in the th loading step. It is given by

(14)

where is the stress tensor in the th step; indices andrefer to the directions, and is the increment of the plasticstrain component in the two sequential steps.

Fig. 9(a) and (b) indicate that, for a fixed adhesive width, theaccumulative plastic strain energy density at the critical locationslightly increases as the height of the adhesive decreases. Fur-thermore, comparing Fig. 9(a) and (b) shows that increasing theadhesive width reduces the accumulative plastic strain energysignificantly. In summary, according to the simulated results,the effect of the size of the adhesive on the plastic strain energydensity follows a similar trend to the effective plastic strain.


Fig. 9. Accumulative plastic strain energy density behavior at the most criticallocations of the solder balls in all studied packages with (a) 2-mm-wide adhesiveand (b) 3-mm-wide adhesive.

Table VI lists the prediction parameters and the predictedfatigue lives of solder balls in all cases, obtained using theDarveaux’s model. Fig. 10 displays how the size of the lid-sub-strate adhesive affects the thermal fatigue life of the solder ball.The estimated fatigue strength of the solder balls of thermallyenhanced FC-PGBA packages is substantially lower than thatof conventional FC-PBGA packages. The height of the adhe-sive has a weaker effect than the width on the fatigue life ofthe solder balls. The fatigue life increases slightly from 2060to 2211 cycles and from 3168 to 3405 cycles, as the height ofthe adhesive increases from 0.05 to 0.15 mm at fixed adhesivewidths of 2 and 3 mm, respectively. However, the width ofthe adhesive greatly affects the fatigue life of solder balls: thefatigue life increases markedly as the width is increased from 2to 3 mm for a fixed height of adhesive.

TCTs for some of the studied packages were performed usingthe same temperature profile as displayed in Fig. 4 to confirm thesimulated thermal fatigue life of the solder balls. Daisy-chainedruns for the conventional FC-PBGA components and thermallyenhanced FC-PBGA components of cases A, C, and F were con-ducted separately. For each selected type of package, twelve

TABLE VIPREDICTION PARAMETERS AND FATIGUE LIFE PREDICTED

USING DARVEAUX’S MODEL IN ALL CASES

Fig. 10. Effect of size of the lid-substrate adhesive on the thermal fatigue lifeof the solder balls.

components were assembled on a FR-4 PCB. Two circuits wereused to connect the solder balls. One circuit connected the ballsthat were located under the chip and the other circuit connectedthe other peripheral balls. The electrical resistance was moni-tored during the tests, and the detection of a 20% increase fromthe initially measured value was regarded as failure. The two-pa-rameter Weibull cumulative distribution function was adopted inthe reliability study

(15)

where is the fraction of failure; is the number of cyclesto failure; is the characteristic life to 63.2% failure, andis Weibull’s shape parameter. Fig. 11 presents the Weibulldistribution of the reliability of the various studied pack-ages. Table VII lists the corresponding Weibull’s parameters.Although the experimental data for the thermally enhancedpackages of cases B, D, and E are not available, the test resultsreveal some trends. Comparing the experimental data of char-acteristic life indicates that the fatigue life of the solder ball ofthe conventional FC-PBGA package is much longer than thatof the solder ball of the thermally enhanced package. This ob-servation is consistent with the predicted results. Additionally,the characteristic life of the solder ball for thermally enhancedpackages increases from 2411 to 4577 cycles as the adhesivewidth increases from 2 to 3 mm at a fixed adhesive height of0.15 mm. This trend is also elicited from the predicted results.


Fig. 11. Weibull distribution of reliability of the studied FC-PBGA packagesin TCTs.

TABLE VIITEST RESULTS FOR TCTS AND CORRESPONDING WEIBULL DATA

However, the characteristic life of the solder ball for thermallyenhanced packages remains nearly unchanged as the adhesiveheight increases from 0.05 to 0.15 mm at a fixed adhesive widthof 2 mm. The effect of adhesive height is not apparent from theanalysis of experimental data. The predicted slight increase ofthe fatigue life of solder joint with the adhesive height is notconfirmed experimentally.

Fig. 10 also plots the experimental characteristic life to 63.2%failure of the solder balls, to examine the validity of Darveaux’sprediction model. The maximum deviation between the pre-dicted life and the experimental data of characteristic life is25.6%, indicating that Darveaux’s model based on the resultsof viscoplastic FE analysis yields reasonable predictions.

The accumulated effective plastic strain at the critical locationin the adhesive was studied and plotted in Fig. 12(a) and (b) forvarious adhesive heights and widths, to examine the reliabilityof the adhesive in thermally enhanced FC-PBGA packages.Table VIII presents the accumulated effective plastic strain atthe critical location during the third cycle. The potential failurelocations are the same in all cases, at the lower left corners ofthe adhesive on the diagonal plane. The effective plastic strainincreases as the adhesive height decreases, for a fixed adhesivewidth. Moreover, increasing the width of the adhesive from 2 to3 mm considerably reduces the plastic strain, as demonstratedby Fig. 12(a) and (b). These figures reveal that the reliability ofthe lid-substrate adhesive increases with its size.

V. CONCLUSION

This study investigated how the size of lid-adhesive affectsthe thermal fatigue life of solder balls in thermally enhanced

Fig. 12. Accumulative effective plastic strain at the most critical location in theadhesive in all studied cases, with adhesive widths of (a) 2 mm and (b) 3 mm.

TABLE VIIIEFFECTIVE PLASTIC STRAIN AT THE MOST CRITICA

LOCATION IN THE ADHESIVE IN THE THIRD CYCLE

FC-PBGA packages. The 3-D FE method was used to simulatethe stress/strain behavior in the packages, and the results wereused to predict the life by correlation with the Darveaux’s refer-ence data for eutectic solders. The solder ball fatigue life of theFC-BPGA package with a copper lid is less than that of the con-ventional ones because the lid limits the thermal expansion ofthe substrate. Enlarging the adhesive increases the fatigue lifeof the solder balls. The width influences the fatigue life morestrongly than does the height. A comparison of the FEM resultsacross all design cases demonstrates that increasing the size ofthe adhesive reduces the plastic strain in the adhesive, implyingthat the size of the adhesive is a design factor that should bedetermined by considering both cost and reliability. The exper-imental results of TCTs also show that the Darveaux’s model


based on the elastic-viscoplastic FE analysis can predict the fa-tigue life of solder balls of the thermally enhanced FC-PBGApackages within a reasonable deviation range.

REFERENCES

[1] A. Schubert, R. Dudek, B. Michel, H. Reichl, and H. Jiang, “Materialsmechanics and mechanical reliability of flip-chip assemblies on organicsubstrates,” in Proc. Int. Symp. Third Adv. Packag. Mater., Mar. 1997,pp. 106–109.

[2] L. L. Mercado and V. Sarihan, “Predictive design of flip-chip PBGA forhigh reliability and low cost,” in Proc. 49th Electron. Comp. Technol.Conf., Jun. 1999, pp. 1111–1115.

[3] E. S. Drexler, “Plastic strain in thermally cycled flip-chip PBGA solderballs,” in Proc. IEEE/CPMT 24th Electron. Manufact. Technol. Symp.,Oct. 1999, pp. 239–244.

[4] L. L. Mercado, V. Sarihan, Y. Guo, and A. Mawer, “Impact of solderpad size on solder joint reliability in flip-chip PBGA packages,” IEEETrans. Adv. Packag., vol. 23, no. 3, pp. 415–420, Aug. 2000.

[5] K. Verma, S.-B. Park, B. Han, and W. Ackerman, “On the design of flip-chip PBGA package assembly for optimum solder reliability,” IEEETrans. Comp. Packag. Technol., vol. 24, no. , pp. 300–307, Jun. 2001.

[6] H. Chan, S. Alvarez, and G. Carson, “Material and process consider-ations for reliable overmolded flip-chip PBGAs,” in Proc. IEMT 27thElectron. Manufact. Technol. Symp., Jul. 2002, pp. 23–26.

[7] J. B. Shim, E. C. Ahn, T. J. Cho, H. J. Moon, T. G. Chung, J. H. Lyu, H.K. Kwon, S. Y. Kang, and S. Y. Oh, “Mechanisms of die and underfillcracking in flip-chip PBGA package,” in Proc. Adv. Packag. Mater.,Mar. 2000, pp. 201–205.

[8] L. L. Mercado and V. Sarihan, “Evaluation of die edge cracking in flip-chip PBGA packages,” in Proc. ITHERM Electron. Packag. Technol.Conf., May 2000, pp. 73–78.

[9] E. H. Wong, S. W. Koh, R. Rajoo, and T. B. Lim, “Underfill swellingand temperature-humidity performance of flip-chip PBGA package,”in Proc. 3rd Electron. Comp. Technol. Conf., Dec. 2000, pp. 258–262.

[10] B. Stone, J. M. Czarnowski, and J. R. Guajardo, “High performanceflip-chip PBGA development,” in Proc. 51st Electron. Comp. Technol.Conf., May 2001, pp. 997–1002.

[11] B. Z. Hong and T. D. Yuan, “Integrated flow-thermomechanical andreliability analysis of a low air cooled flip-chip PBGA package,” inIEEE 48th Electron. Comp. Technol. Conf., May 1998, pp. 1354–1360.

[12] ——, “Integrated flow-thermomechanical analysis of solder joints fa-tigue in a low air flow C4/CBGA package,” J. Microcirc. Electron.Packag., vol. 21, pp. 137–144, Second Quarter 1998.

[13] K. Ramakrishna and T.-Y. T. Lee, “Prediction of thermal performanceof flip chip-plastic ball grid array (FC-PBGA) packages: effect ofsubstrate physical design,” in Proc. ITHERM Thermal Thermomech.Phenom. Electron. Syst., May 2002, pp. 528–537.

[14] T. Y. Lee, “An investigation of thermal enhancement on flip chipplastic BGA packages using CFD tool,” IEEE Trans. Comp. Packag.Manufact. Technol., vol. 23, no. 3, pp. 481–489, Sep. 2000.

[15] L. M. Eyman and G. B. Kromann, “Investigation of heat sink attachmethodologies and the effects on package structural integrity and in-terconnect reliability of the 119-lead plastic ball grid array,” in Proc.47th Electron. Comp. Technol. Conf., May 1997, pp. 1068–1075.

[16] Z. Z. Celik, D. Copeland, and A. Mertol, “Thermal enhancement andreliability of 40 mm EPBGA packages with interface materials,” inProc. IEEE/CPMT 21th Electron. Manufact. Technol. Symp., Oct.1997, pp. 376–385.

[17] T. Kobayashi and S. Hayashida, “A study on reliability modeling forthrough hole cracking failure in thermal enhanced PBGA laminate,” inProc. 50th Electron. Comp. Technol. Conf., May 2000, pp. 1658–1660.

[18] A. Kaisare, Z. Han, D. Agonafer, and K. Ramakrishna, “Prediction ofeffect of heat sink adhesive on mechanical reliability of a wire bondedplastic ball grid array package,” in Proc. ITHERM 8th Thermal Ther-momech. Phenom. Electron. Syst., Jun. 2002, pp. 926–931.

[19] H. F. Hsiao, “Role of adhesive modulus on reliability of FCBGA withheat spreader,” in Proc. 4th Electron. Mater. Packag., Dec. 2002, pp.161–166.

[20] “ANSYS User’s Manual, Revision 7.0,” Swanson Analysis System,Inc., 2004.

[21] F. Feustel, S. Wiese, and E. Meusel, “Time-dependent material mod-eling for finite element analysis of fillip chips,” in Proc. IEEE 50thElectron. Comp. Technol. Conf., May 2000, vol. 21–24, pp. 1548–1553.

[22] R. Darveaux, “Effect of simulation methodology on solder joint crackgrowth correlations,” in Proc. IEEE 50th Electron. Comp. Technol.Conf., May 2000, vol. 21–24, pp. 1048–1058.

Yi-Ming Jen received the B.S. and Ph.D. degrees inpower mechanical engineering from National TsingHua University, Hsinchu, Taiwan, R.O.C., in 1990and 1995, respectively.

He joined the Hull Department, United Ship De-sign and Development Center, Taipei, Taiwan, as aStructure Design Engineer in 1997. In 1998, He was aConsultant with the Department of Computer AidedEngineering, Cadmen Company, Taipei, to promotethe application of FE codes in related industries. He iscurrently an Assistant Professor with the Department

of Mechanical Engineering, Chung Hua University, Hsinchu. His specialties aremultiaxial fatigue analysis, FE simulation and experimental stress analysis. Hiscurrent technical interests include the thermal stress simulation, failure analysis,and reliability study of the advanced electronics packaging.

Chih-Kai Fang received the B.S. and M.S. degrees inmechanical engineering from Chung Hua University,Hsinchu, Taiwan, R.O.C., in 2002 and 2004, respec-tively.

He is currently an Assistant Engineer in theDepartment of Breaker Engineering, Shihlin Electricand Engineering Corporation, Hsinchu, Taiwan. Hisresearch interests are in thermal analysis of electricequipments and design and development of breakersand switchgears.

Yun-Hsin Yeh received the B.S. degree in mechan-ical engineering from Chung Yuan Christian Univer-sity, Chung Li, Taiwan, R.O.C., in 2000 and the M.S.degree in mechanical engineering from Chung HuaUniversity, Hsinchu, Taiwan, R.O.C., in 2002.

He joined the Assembly Process Development De-partment, Powertech Technology, Inc., Hsinchu, as aSenior Engineer in 2005. His current research inter-ests are at thermal and structural analysis of multichipmodules and flip-chip packages.

Documents

Effect of Size of Lid-Substrate Adhesive on Reliability of Solder Balls in Thermally Enhanced Flip Chip PBGA Packages