Effects of Current Stressing on Shear Properties of Sn-3.8Ag-0.7Cu Solder Joints

J. Mater. Sci. Technol., 2010, 26(8), 737-742. Effects of Current Stressing on Shear Properties of Sn-3.8Ag-0.7Cu Solder Joints X.J. Wang 1), Q.L. Zeng 1), Q.S. Zhu 1), Z.G. Wang 1) and J.K. Shang 1,2) 1) Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, Shenyang 110016, China 2) Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA [Manuscript received April 8, 2009, in revised form May 20, 2009] Effects of electromigration on microstructure, shear strength, and fracture behavior of solder joints were investigated by single-ball shear samples of eutectic Sn-3.8Ag-0.7Cu (SAC) joined by Cu plates at two sides. The electromigration tests were conducted at a current density of about 1.1 10 3 A/cm 2 and a working temperature of about 83 C. The results showed that the shear strength and flow stress decreased greatly after current stressing. Such a decrease was associated with no significant loss of the fracture strain at short electromigration but a great reduction in the fracture strain after long-term current stressing. The variation of the fracture strain with the electromigration time was shown to result from the shift of the fracture surface from the center of the solder towards the intermetallic compound (IMC) interface at the cathode. KEY WORDS: Strength; Fracture strain; Voids; Solder 1. Introduction As the rapid growth of microelectronic technology has driven the solder joints in flip chip devices to a very small scale, solder joints must endure not only stresses induced by the mismatch of thermal expansion coefficients, but also current induced electromigration [1 3]. During electromigration of the Sn-3.8Ag-0.7Cu (SAC) solder joints, copper atoms are driven by the electron wind force to move from the cathode to the anode, while a reverse flux of vacancies moves from the anode to the cathode [4 6]. Consequently, voids collect at the cathode between the interface of the solder and the intermetallic compound (IMC) due to the vacancy accumulation [7,8], and the hillocks emerge on the surface near the anode [9 11]. At the same time, a thicker IMC layer forms at the anode side and a thinner at the cathode. The higher the current density, the stronger the electromigra- Corresponding author. Prof., Ph.D.; Tel.: +217 333 9268; Fax: +217 333 2736; E-mail address: jkshang@illinois.edu (J.K. Shang). tion effects [4,6,12]. As future devices take on additional miniaturization and multiple functions, local current crowding caused by solder joint s structure and size may greatly promote the electromigration induced failures [13,14]. While either thermal stresses or current could result in the failure of a solder joint by itself, they may also combine to produce an accelerated failure. For microelectronic packages, since both stresses and currents may co-exist, it is important to the device reliability that the combined effects of the mechanical stresses and the electric current are understood. Recently, Tu s group [12] reported that both the solder joint strength and the plastic strain decreased following the brittle fracture at the cathode after electromigration under coupled current and force. The brittle fracture was related to the voids near the cathode. On the other hand, Shang s group [9] conducted tensile tests on the electromigration-damaged solder joints and found the softening of the solder with no loss of the fracture strain. The fracture was confined in the solder and occurred in a ductile manner. They

738 X.J. Wang et al.: J. Mater. Sci. Technol., 2010, 26(8), 737 742 ascribed this softening to the excess vacancies in the solder joints. By comparing the results from those two studies, it should be noted that besides the three hours difference under current, 48 and 51 h, respectively, there was one order of magnitude difference in the current density. In addition, they also took different sample types and loading rates. These differences made it difficult for an effective comparison of the results and prompted the need of a detailed study to understand the leading factors, which have caused current-induced reductions in the mechanical properties of the solder joints. In this study, we conducted a systematic study on the effect of the electromigration on mechanical properties of SAC solder joints. Since the solder joints are predominantly subject to shear forces in their service [15], a single-lap solder joint was used and the experiments were conducted under a wide range of electromigration conditions with current stressing time up to 3100 h. The results show strong effects of electromigration time on microstructures, shear strength and fracture behavior of solder joints. A transition from ductile solder softening to brittle interfacial fracture was found as the electromigration time increased. 2. Experimental A single lap joint of Sn-3.8Ag-0.7Cu (SAC) solder was designed and fabricated as shown in Fig. 1 to combine the effects of electromigration and mechanical force. First, copper substrate was cut into plates of 15 mm 3 mm 0.5 mm. A 0.02 mm thick solder mask was made to keep a bare copper opening of 0.9 mm in diameter, which was used as the pad for solder bonding. An SAC solder ball of 1 mm dipped in a rosin mildly activated (RMA) flux was then mounted on the bonding pad before the solder balls and the substrates were reflowed at 260 C for about 5 min. After the reflow cycle, the specimens were cleaned using an ultrasonic cleaner with a flux remover. Electromigration tests were carried out by applying a constant direct current at an average current density of 1.1 10 3 A/cm 2. During the test, electromigration samples were immersed in heat-conducting oil to minimize oxidation of the sample surface and to dissipate the Joule heating (Fig. 1). The working temperature of the current stressing was 83 C and the time under current stressing was selected as: 0, 325, 1000, 1300 and 3100 h, respectively. After the electromigration, the shear tests were conducted at a shear rate of 0.3 mm/min. The load-displacement curves of the samples were recorded during the shear test (Fig. 2). For microstructural analysis, the specimens before and after fracture were mounted in cold epoxy, ground with SiC paper through a series of solder balls and polished with 0.5 µm diamond paste. The specimens were observed by scanning electron microscopy Fig. 1 Schematic diagram of the electromigration testing Fig. 2 Schematic diagram of the shear test after electromigration (SEM, JEOL, Tokyo, Japan) in back-scattered electron imaging (BSE) mode. 3. Results and Discussion 3.1 Microstructure The polarity effect of the electromigration on the thickness and morphology of the IMCs at the cathode and anode in the solder joint is shown in Fig. 3. The same IMC of Cu 6 Sn 5 forms at the solder/cu interface with or without applying electric current. The IMC formed after the initial reflow has a scallop-type morphology (see Fig. 3(a)) but becomes much flatter with electromigration (Fig. 3(b) and (c)), as previously reported [4 6], called layer-like IMC for convenience. The electromigration enhances the growth of IMC at the anode and inhibits the growth at the cathode as compared with the no-current case. For the sample under the current density for 325 h (Fig. 3(b) and (c)), Cu 6 Sn 5 layer grows at the anode side with the application of the current. The total thickness approaches 7 10 µm after 325 h. This thickness is comparable with that grown at 150 C for 1700 h without current [6]. At the cathode side, shown in Fig. 3(c), the IMC grows much more slowly than that at the anode side, to only 2 3 µm. This polarity reaction under current stressing is the result of the electromigration, where the copper atoms are driven to the anode side by the electronic wind force, filling the interspaces of the scallop-type IMCs and turning the IMC into the layer-like morphology. The IMC at the cathode side, however, becomes flatter due to the dissolution of the

X.J. Wang et al.: J. Mater. Sci. Technol., 2010, 26(8), 737 742 739 Fig. 4 Coarsening of Ag 3Sn in solder joints before (a) and after (b) electromigration for t=325 h Fig. 3 IMCs before electromigration (a) and after electromigration for 325 h ((b) and (c)) scallop tips of IMC. Besides the polarity effect at the electrode interfaces, the microstructural coarsening is also severe, as shown in Fig. 4. It can be seen that the fine Ag 3 Sn particles dispersed in the Sn-rich matrix coarsen appreciably after 325 h of the electromigration. As a result, the average size of the approximately spherical Ag 3 Sn particles grows from 0.4 to 1 µm with the increase of the related inter-particle spacing, but the coarser Cu 6 Sn 5 particles grow at a much slower rate. 3.2 Current effect on mechanical properties In lap-shear test, the average shear stress in the solder was taken as the applied axial load divided by the solder joint area, which contacted the substrate. The shear strain was obtained by using the applied displacement divided by the solder thickness as follows: γ= L D, where L is the extension of a sample along axial load direction and D is the height of the solder joint. The shear strain thus obtained is the nominal value, γ, given the complex deformation fields involved in such type of loading. The effect of the electromigration on the shear strength of the solder joint is shown in Fig. 5. Before and after electromigration, the solder joint displays Engineering stress / MPa 35 30 25 20 15 10 5 0 Original 325 h 1000 h 1300 h 3100 h -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 Engineering strain Fig. 5 Effect of the current on the shear properties of the solder joints (shear speed of 0.3 mm/s) quite different stress-strain behavior. For all cases, the shear strength, yield strength and flow stress are decreased to some extent because of the electromigration, which indicates that the application of the electric current reduces the deformation resistance of the solder joint. Such a strength drop is far more severe than the softening caused by simple isothermal aging of the solder alloy at the same temperature [6]. It is worth noting that the curves before and after 1300 h in Fig. 5 are different. The curves before 1300 h show almost the similar characteristics as the response of the bulk solder alloy, namely shaped like a dome consisting of a linear elastic region, a plastic rising dome including the steady state flow, a gradual decay followed by the fracture. The extensive plastic dome disappears when the current application time is up to 1000, 1300 and 3100 h shown in Fig. 5 correspondingly. When the shear properties are plotted against the current application time, Fig. 6 reveals

740 X.J. Wang et al.: J. Mater. Sci. Technol., 2010, 26(8), 737 742 Nominal shear strength / MPa 28 26 24 22 20 18 16 14 12 10 8 0 0 500 1000 1500 2000 2500 3000 3500 Current stressing time / h 160 140 120 100 80 60 40 20 Nominal fracture strain / % Fig. 6 Variation of the joint properties with electromigration time two distinct responses. Below 1300 h, the decrease in the shear strength is gradual, and there is more softening at long current application time. The fracture strain also remains high, indicating that the fracture is preceded by extensive plastic deformation, typical of a ductile fracture process. By contrast, above 1300 h, there is a sudden drop of the fracture strain to a very low level and the shear strength tends to level off to a very low value. Therefore, short term electromigration only caused softening of the solder joint with little loss of the fracture strain, while long term electromigration resulted in severe weakening and embrittlement of the solder joint. As electromigration damage accumulated in the solder joint, a ductile to brittle transition took place in the solder mechanical behavior. In the early stage of the electromigration, the ductile behavior is not surprising in the absence of any visible voids in the electromigration samples. Only decrease in the flow stress is observed. At 1000 h, the flow stress decreases from the original 27 MPa without electromigration to about 16.2 MPa, around 39% drop while the strains at fracture are comparable. Such current-induced softening of the SAC solder joints in shear observed in this study is quite similar to the phenomenon reported in the previous studies by Zhang et al. [9]. As the time increased to 1300 and 3100 h, the flow stresses decreased to 12 and 11.2 MPa, respectively. At the same time, the strains at fracture are reduced by a factor of more than 7. Such currentinduced embrittlement of SAC solder joint is consistent with the results reported by Ren et al [12]. 3.3 Fracture morphology Figure 7 shows the top views of the fracture surface of the joint samples after current stressing up to Fig. 7 Shear fracture morphologies after current stressing: (a) t=0 h, (b) t=325 h, (c) and (e) t=1300 h, (d) t=3100 h

X.J. Wang et al.: J. Mater. Sci. Technol., 2010, 26(8), 737 742 741 Fig. 8 Macro-fracture path after current stressing Fig. 9 Microstructure and fracture morphology after current stressing for 325 h: (a) optical photograph, (b) SEM backscattering images of the fracture surfaces 3100 h. The macroscopic views of the broken joints are shown in Fig. 8. It is clearly seen that the solder joints fail in solders at the current stressing time in the range of 1000 h, as seen in Fig. 7(a) and (b) and Fig. 8. The ductile dimples in Fig. 7(a) are elongated along the shear direction, indicating a macroscopically ductile fracture behavior of the solder joint, which is consistent with the shear stress-strain curves showed in Fig. 5. Prior to fracture, extensive plastic deformation takes place (Fig. 8). Below 1000 h, the joints failed in the solder and the fracture was solder-controlled. The microstructural coarsening should contribute to the strength reduction but not the leading factor, because in the mechanical studies by Zhang et al. [9], the decrease was nearly completely recovered by thermally annealing. In this case, the decrease in the mechanical properties of the solder interconnect may be related to the excess vacancies. Though the vacancies cannot be seen directly, their role is confirmed by thermal annealing at the same temperature after exposure to the electric current. It is suggested that the loss of the joint strength owing to the electromigration and the subsequent recovery by annealing are caused by the strong and weak interactions of vacancy with dislocations in the solder joint. The fracture morphologies after 325 h of currentstressing are different from the dimples formed without current stressing. Figure 9(a) is the microstructure of solder joint after 325 h of current application. The average grain size is around 5 µm, which is comparable with the grain size after current stressing. Figure 9(b) shows the fracture surfaces after the current stressing, where the grain boundaries are clearly visible. Therefore, the current also induced the weakening of grain boundaries. However, since no significant grain growth was observed in electromigration, the effect of the grain boundary was not directly responsible for the ductile to brittle transition. Voids are observed on the fracture surfaces after 1300 h of current-stressing, as shown in Fig. 7(c) (e). Thereinto Fig. 7(e) is the related part in Fig. 7(c) at higher magnification. Unlike excess vacancies, which may be annealed out, voids represent permanent dam-

742 X.J. Wang et al.: J. Mater. Sci. Technol., 2010, 26(8), 737 742 age to the solder joint. As expected, void density should be relatively low when they first appeared. Since most ductile fracture of metals undergoes void nucleation, growth and coalescence, the appearance of the voids should not lead to immediate fracture. Therefore, the fracture surface and the corresponding stress-strain curve in Fig. 5 still retain the ductile features. After 3100 h of current-stressing, as shown in Fig. 7(d), fracture occurs at the interface, through Cu 6 Sn 5 IMC. Voids are also found in the solder residuals exposed on the fracture surface. In the studies by Ren et al. [12], the decrease in the mechanical properties of the solder interconnect was related to the void formation at the cathode induced by electromigration as the fracture followed the interfacial voids [16,17]. Since voids could also develop in the IMC layer by the Kirkendall effect, failure would likely run through the IMC as seen in Fig. 7(d). From the observations above, the following scenario has emerged. When the current-exposure time is short, the mechanical loss is mainly caused by the vacancies induced by the electromigration. With increasing current-exposure time, mass transport process is enhanced. Thus, the vacancies are driven to the cathode and aggregate as voids near the cathode. Consequently, the voids turn into a major factor in reducing the joint strength. As the strength reduction moves from vacancy-dominated to void-controlled, a ductile to brittle transition occurs in the solder joint as observed in Figs. 5 7. 4. Conclusions Mechanical behavior of Sn-3.8Ag-0.7Cu solder joints exposed to the current density of 1.1 10 3 A/cm 2 for different time has been investigated by lap shear samples. The conclusions can be drawn as follows. (1) The electromigration had a strong influence on the mechanical behavior of the solder joints. With increasing electromigration time, the shear strength of the solder joint decreased by up to more than a factor of two and the fracture behavior went through a ductile to brittle transition. (2) The effect of the electromigration manifested in two distinct ways. In the early stage of the electromigration, only solder softening was observed with no significant loss in the fracture strain. In the late stage of the electromigration, weakening of the solder joint was accompanied by significant drop in the failure strain. (3) The early softening by electromigration was related to the generation of excess vacancies while the later embrittlement of the solder joint resulted from void accumulation at the cathode side of the solder joint. Acknowledgement This study was supported by the National Basic Research Program of China (Grant Nos. 2004CB619306 and 2010CB631006). REFERENCES [1 ] A. LaLonde, D. Emelande, J. Jeannette, C. Larson, W. Rietz, D. Swenson and D.W. Henderson: J. Electron. Mater., 2004, 33, 1545. [2 ] L.P. Lehman, S.N. Athavale, T.Z. Fullem, A.C. Giamis, R.K. Kinyanjui, M. Lowenstein, K. Mather, R. Patel, D. Rae, J. Wang, Y. Xing, L. Zavalij, P. Borgesen and E.J. Cotts: J. Electron. Mater., 2004, 33, 1429. [3 ] T.R. Bieler A.U. Telang, J.P. Lucas, K.N. Subramanian, L.P. Lehman, Y. Xing and E.J. Cotts: J. Electron. Mater., 2004, 33, 1412. [4 ] H. Gan and K.N. Tu: Effect of Electromigration on Intermetallic Compound Formation in Pb-free Solder- Cu Interfaces, Impact Reliability of Solder Joints, in IEEE Electronic components and Technology Conference, 2002, 1206. [5 ] H. Gan, W.J. Choi, G. Xu and K.N. Tu: JOM, 2002, 54, 34. [6 ] J.W. Yoon and S.B. Jung: J. Alloy. Compd., 2008, 458, 200. [7 ] I.A. Blech: J. Appl. Phys., 1976, 47, 1203. [8 ] C.C. Lu, S.J. Wang and C.Y. Liu: J. Electron. Mater., 2003, 32, 1515. [9 ] L. Zhang, Z.G. Wang and J.K. Shang: Scripta Mater., 2007, 56, 381. [10] J.A. Nucci, A. Straub, E. Bischoff, E. Arzt and C.A. Volkert: J. Mater. Res., 2002, 17, 2727. [11] J. Bohm, C.A. Volkert, R. Mönig, T.J. Balk and E. Arzt: J. Electron. Mater., 2002, 31, 45. [12] F. Ren, J.W. Nah, K.N. Tu, B.S. Xiong and L.H.L. Pang: Appl. Phys. Lett., 2006, 89, 141914. [13] Everett C.C. Yeh, W.J. Choi and K.N. Tu: Appl. Phys. Lett., 2002, 28, 580. [14] K.N. Chiang, C.C. Lee and K.M. Chen: Appl. Phys. Lett., 2006, 88, 072102. [15] K.S. Siow and M. Manoharan: Mater. Sci. Eng. A, 2005, 404, 244. [16] M. Date, T. Shoji, M. Fujiyoshi, K. Sato and K.N. Tu: Impact Reliability of Solder Joints, in Proceedings of the 54th ECTC, Las Vegas, NV, 2004, 668. [17] F. Ren, L. Xu, X. Zhang, J. Nah, J.H.L. Pang and K.N. Tu: In-situ Study of the Effect of Electromigration on Strain Evolution and Mechanical Property Orange in Lead-free Solder Joints, in Proceedings of the 2006 ECTC Meeting, San Diego, CA, 2006.