Materials Transactions, Vol. 45, No. 3 (2004) pp. 714 to 720 Special Issue on Lead-Free Soldering in Electronics #2004 The Japan Institute of Metals Effect of Cu ddition to -g Lead-Free Solder on Interfacial Stability with Chi-Won Hwang* and Katsuaki Suganuma ISIR, Osaka University, Ibaraki 567-0047, Japan The interfacial reactions and the interface microstructures between -3.5g-(0.7Cu) lead-free solders and substrate were investigated at the reaction temperature of 250 C. Joint strength was also evaluated. The -3.5g joint shows the double reaction layers of Fe 2. Ni from dissolves into molten -3.5g solder during soldering and forms Ni 3 4 compounds with the eutectic network of g 3 /- in solder layer during solidification. For the -3.5g-0.7Cu jont, the Cu addition in -3.5g changes the microstructure of solder and interfacial reaction layers. Ni dissolved into solder melt reacts with and Cu to form a -phase of Cu-Ni- compound during the soldering without the formations of Ni 3 4 compound. t the reaction interface, fine particles of Fe 2 can be found near to the reaction layer instead of the faceted compound of Fe 2 in the second reaction layer for the -g joint. -3.5g-0.7Cu joint increases the joint strength by about 30 40 MPa higher than that of -3.5g joints. (Received September 19, 2003; ccepted January 21, 2004) Keywords: lead-free solder, tin-silver-copper solder, iron-42 nichkel, reaction interface, tin, silver, nichkel, copper 1. Introduction s development of lead-free solders, several third or fourth elements have been tried as additive components to -based lead-free solder to improve the solderability on various substrates. mong those additives, Cu has been regarded as one of the most attractive elements to -g or -Zn solder alloys to decrease melting temperature but not to sacrifice their mechanical properties. mong lead-free solder alloys, near eutectic -g-cu has been already applied to electronics packaging industry widely. 1,2) Various metallic or metallized substrates such as Cu, Ni, Fe, Ni-P, u, g, and Fe-Ni have been used as useful electrodes or lead frames because of their excellent electrical conductivity. Occasionally, thermal mismatches of these metal substrates with Si devices or ceramic chips induce joint failure after soldering. To escape such thermal mismatch influence, alloy has been used as the useful material because its coefficient of thermal expansion is much closer to those of Si devices than the others such as Cu or Ni. 3,4) During soldering, principally reacts with substrates to form different interfacial reaction products or layers in a few micrometer scales or less. 5,6) Sometimes, an alloying element such as Zn in -Zn solder reacts with substrate elements of Cu or u, in preference to, resulting in the formation of Zn-Cu or Zn-u compound. 6,7) Cu even of a small amount in the -g-cu alloy also influences the formation of interfacial reaction layer with Ni substrate. 8) On the other hand, in practical assembly processes, multiple reflows are often required. During the processes, the elements dissolved from a substrate into a solder can affect the interfacial reaction and interface microstructures. 9) For these reasons, when one adopts these lead-free solders, it is important to understand the interfacial reaction of a lead-free solder with a substrate and to consider the effect of a substrate on the soldered microstructure to select a better substrate as well as to attain suitable process conditions. 10 12) number of *Corresponding author, E-mail: hwang12@sanken.osaka-u.ac.jp studies have been carried out on the interfacial reactions between -g-(cu) solders and various substrates such as Cu, Ni, and Ni-P. 10,11,13 15) However, little is known about the interfacial reaction and interface microstructure between - g-(cu) and. 16 18) Thus, it is worthwhile to evaluate the jointability of -g-(cu) with, but also to investigate the effect of substrates on the interface microstructure during the long term reaction. Therefore, the purpose of the present work is to investigate the effects of the additive elements of g and Cu into pure and of substrates on the interfacial reaction and interface microstructure. The strength of -g-(cu) joints for each reaction condition is also evaluated to access suitable soldering conditions. 2. Experimental Procedures Two types of -3.5mass%g and -3.5mass%g- 0.7mass%Cu solders were prepared and the chemical compositions are listed in Table 1. (hearafter, all unit mass% measures are omitted) invar alloy plates (15 mm 1 mm), of which chemical compositions are also shown in Table 1, were used as the substrates. The faces of the platelets were polished with 0.3 mm l 2 O 3 powders. The solder ingots were melted and held in an l 2 O 3 bath at reaction temperature of 250 C in air. Two plates were preheated up to around 170 C with rosin middle activated type flux (RM Delta Flux 533z, Senju Metals Co., Ltd.) in order to prevent oxidation. The RM flux also applied to the molten solders to eliminate the oxide layer of Table 1 Chemical compositions of solder and substrate alloys. (Unit: mass%) Composition g Cu Fe Ni Co al -3.5g 96.53 3.44 0.03-3.5-0.7Cu 95.6 3.55 0.77 0.08 56.7 42.1 0.1 1.1
Effect of Cu ddition to -g Lead-Free Solder on Interfacial Stability with 715 the surfaces. Then the two preheated plates were immersed into each solder melt for about 3 s. fter they were withdrawn from the bath, the plates were immediately joined together on an l 2 O 3 plate being kept at the reaction temperature. These /solder/ joints were held inside the furnace for 2 min to. For maintaining a solder layer uniformly, Mo wires of 200 mm in thickness were placed as a spacer. fter removing the joint from the furnace, it was cooled in air. Microstructural observation and analysis on the solder and interface layers were carried out with scanning electron microscopy (SEM), electron probe microanalysis (EPM), and transmission electron microscopy (TEM, Hitachi H- 8900T, operated at 200 kv) equipped with EDS. SEM and EPM samples were prepared by polishing with 0.05 mm l 2 O 3 particles, followed by etching with diluted HCl. TEM samples were prepared by ultramicrotomy with a Leica Ultracut T Ultramicrotome (i) by cutting a cross-sectioned cubic sample (about 0.5 mm in side length), (ii) by molding with commercial epoxy, (iii) by setting the molded sample in an ultramicrotomy arm and trimming the cross-section surface with a diamond knife, and (iv) by slicing the trimmed sample by 15 nm in sectioning thickness using a boat type diamond knife (Ultra, DiTOME). For tensile test, cylinders (15 mm 10 mm) were used instead of the plates. The joints were prepared in the same way mentioned above. The tensile test was carried out for the cross-sectioned samples (1 mm 3 mm 20 mm 3 ) at a crosshead speed of 0.5 mm/min. eutectic network β- 2 min reaction layer 3. Results and Discussions 3.1 -g/ reaction system Figure 1 shows the microstructures for the -3.5g joint with the substrate reacted for 2 min and utes, respectively. The microstructure after reaction for 2 min in Fig. 1 shows the typical -g eutectic morphology composed of primary - grains/eutectic network of fine g 3 particles (or fibrous one) precipitated in - matrix. 15) fter reaction for, both g 3 particles and the eutectic network are slightly coarsened as shown in Fig. 1. Some primary crystal compounds and platelets indicated as and, respectively, are also observed in the solder layer. etween the -3.5g solder and the, an interfacial reaction layer in Fig. 1 can be found and grown up to a few mm in thickness. For the identification of element distributions on the corresponding area of Fig. 1, EPM analysis results on Ni, g,, and Fe are shown in Fig. 2. From the typical EPM maps, it is found that Ni- compounds indicated as and appear with g 3 eutectic networks in a solder layer, while Fe- compounds are formed at the reaction interface region. This result is similar to that for the pure joint with the. 16) In the previous work, the present authors evaluated the interface microstructure of the pure / reaction system to understand the basic formation and growth mechanisms of the reaction layers. In brief, Fe of the primarily reacts with to form double reaction layers with same phase of Fe 2 containing 2 6 at% Ni substituted as an impurity without phase change. In contrast, Ni dissolves into liquid to form the platelets of Ni 3 4 on solidification. Fig. 1 SEM microstructures of -3.5g solder with reacted for 2 min and, respectively. Fig. 2 EPM element maps for Ni, g,, and Fe on the corresponding region of Fig. 1.
716 C.-W. Hwang and K. Suganuma Secondary Ni 3 4 plate C 10 µm (quenching) 3 µm Fig. 4 SEM microstructure of the reaction interface for -3.5g joint after a reaction for. 10 µm Fig. 3 SEM microstructures of the solder layers reacted for after cooling in air and after the quench, respectively. To investigate the formation step of Ni 3 4 large rods and plates in the solder layer, a joint sample was quenched into an ethyl alcohol kept at 15 C after the reaction at 250 C for 60 minutes. Two SEM photographs in Fig. 3 show the typical microstructures of the solder layer after the cooling in air and after the quench, respectively. Figure 3 shows the secondary Ni 3 4 plates grown up about 10 mm in size or more, while there is no any Ni 3 4 compound in the quenched solder layer, only exhibiting fine or immature eutectic networks. The fact implies that both the primary and the secondary Ni 3 4 compounds appear in the solder layer during air-cooling or solidification, not during the soldering. The sequence of solidification of the -g solder melt containing Ni from the after a long period of reaction can be also inferred easily because the microstructure is similar to that of the -g-cu system, presenting large primary Cu 6 5 crystals in solder layer. 19) In the present reaction system, primary Ni 3 4 crystal rods may appear at first in liquid solder. Then, secondary Ni 3 4 platelets precipitate with g 3 fine particles in the eutectic region. In Fig. 1, the coarsening of g 3 particles and the eutectic networks with increasing reaction time implies that dissolved Ni into the -3.5g expands the pasty range of the solder melt. In other words, the time for the growth of g 3 compound is increased during the cooling or the solidification. s can be seen in Fig. 1, the reaction layer reacted for 2 min is too thin to identify the microstructure. fter a reaction for, the reaction layer grows up about 5 mm in thickness. s explained, principally reacts with Fe from the and forms double reaction layers of Fe 2. However, the EPM maps in Fig. 2 show the reaction interface detected with the elements of not only and Fe but also g and Ni. n enlarged SEM image in Fig. 4 shows the microstructure of double reaction layers formed after reaction for. The first layer indicated as adjacent to the substrate exhibits a uniform layer in thickness, while the second layer indicated as is composed of faceted grains of a few mm in size and small particles. Such a characteristic difference between the first and the second reaction layers was regard as the difference in formation mechanism governed by different diffusion mechanisms. 16,20,21) Fast diffusion occurs for the second reaction layer along the surfaces or grain boundary of the Fe 2 compound. Slow diffusion takes place inside the first layer through Fe 2 grain or lattice. Fast diffusion principally leads to the growth of faceted Fe 2 products into solder liquid. In contrast, slow diffusion contributes to the formation of the first reaction layer and small reaction particles attaching on to the first reaction layers. lthough such double reaction layers are also observed at the interface for the -3.5g joint, some of the reaction products indicated as C are detached from the reaction layer into a solder matrix. This detachment of coarsened reaction products from the interface may be caused by the formation of many small particles below them to push the coarsened one into solder. The detection of g and Ni elements near to the reaction layer in EPM maps of Fig. 2 can be attributed to the reaction products separation from the interface during soldering.
Effect of Cu ddition to -g Lead-Free Solder on Interfacial Stability with 717 2 min -g eutectic network β- -g-cu eutectic network β- 10 µm Fig. 6 SEM microstructure with two different network morphologies. Ni Fig. 5 SEM microstructures of -3.5g-0.7Cu with reacted for 2 min and, respectively. Cu g 3.2-3.5g-0.7Cu/ reaction system Figure 5 shows the representative microstructures of the -3.5g-0.7Cu with the substrate after reactions for 2 min and, respectively. s can be seen in Fig. 5, the soldered layer reacted for 2 min shows the typical eutectic network of the near eutectic -g-cu alloys. Kim et al. analyzed the microstructure of the -g-cu alloys and showed that fine Cu 6 5 þ g 3 particles network is formed in their eutectic region with large primary g 3 platelets about several tens mm in the solder. 22) For the present reaction system, however, no primary g 3 platelet was observed in the solder layer. Instead, two different morphologies in the network structure can be observed as shown in Fig. 5. n enlarged SEM micrograph corresponding to the box area in Fig. 5 is shown in Fig. 6. The coarsened particles network is formed in the left side, while the fine particles network appears in the right. s can be seen in Fig. 7, the element maps on the region of Fig. 6 reveals that the coarsened particles network is composed of mainly g 3 compounds without Cu 6 5 particles, which is the typical eutectic microstructure of eutectic -g alloy. In contrast, the fine particles network consists of fine g 3 and Cu 6 5 compounds, which is a typical eutectic morphology of the -g-cu ternary alloy. lthough the formation mechanism of the two different network morphologies is yet to be fully clarified, no formation of primary large g 3 Fig. 7 EPM element maps for, Ni, Cu, and on the corresponding region of Fig. 6. plate in a solder layer may improve the jointability because the existence of large g 3 platelets seriously decreases the mechanical property especially such as fatigue resistance. 22,23) fter a reaction for in Fig. 5, large compound grains appear in the solder layer without Ni 3 4 platelets. With EPM analysis, it is found that the grains are composed of mainly Cu, Ni, and. To identify the phase, TEM observation and analysis on the solder layer was carried out, and the results are shown in Fig. 8. The Cu-Ni- compound is identified as -phase of Ni 3 2 or Cu 6 5 not but Ni 3 4. This finding is similar to the change in the interfacial phase between -g-cu and Ni-P plating. 8) It is difficult to make a strict distinction between Ni 3 2 and Cu 6 5 because both - phases have the same structure type with similar lattice parameters as listed in Table 2. oth of Ni 3 2 and Cu 6 5 have solubility ranges of about 37 40 at% and about 43
718 C.-W. Hwang and K. Suganuma η-phase of Cu-Ni- compound (111) (121) Ni, Cu Cu 5 µm (221) Ni Cu Fig. 8 TEM photograph and diffraction patterns taken from the primary Ni-Cu- compound with an EDS spectrum. Table 2 Crystallographic data of -Cu 6 5 and -Ni 3 2 intermetallic compounds. Phase Pearson symbol, Lattice Structure Space group parameters type (no.) (nm) -Cu 6 5 a ¼ 4:1922 Nis hp4, P6 3 =mmc c ¼ 5:0372 (Hexagonal) (194) ¼ 120 Ni 2 In or a ¼ 4:146 hp6, P6 3 =mmc -Ni 3 2 Nis c ¼ 5:253 (194) (Hexagonal) ¼ 120 Table 3 Quantitative analysis results on the -phase of Cu-Ni- compound. (Unit: at%) Contents Ni Cu Fe Fraction 34.6 19.1 46.1 0.2 44 at%, respectively. This stoichimetry difference can be caused by the typical -phase of Nis structure type, which has extra sites for interstitial atoms at the bipyramidal sites without changing the basic crystalloid property. 5) To clarify each element quantity on the present -phase, EDS analysis was performed and the result shows a content occupying about 34 at% as listed in Table 3. lthough the content of in the EDS analysis result is much closer to that of (Cu,Ni) 3 2 than (Cu,Ni) 6 5, the present authors would like to leave the identification of the Cu-Ni- compound as a subsequent work because some unexpected factors such as electron shower beam can influence the EDS quantitative analysis. For the further work, a thermodynamic consideration needs to be performed to identify the compound. To examine the formation step of the -phase, a quenched sample was prepared at the same condition of that for the - 3.5g joint. Figure 9 shows the microstructure after quenching, and several -phase compounds still exist inside solder layer with fine eutectic network. The fact means that the -phase products are formed during the soldering, not on (quenching) Cu-Ni-compounds Fig. 9 SEM microstructure of the solder layer for -3.5g-0.7Cu joint reacted for after the quench. the solidification. The formation of -phase can affect the change of Cu content in the molten solder during the long period reaction by the consuming of Cu. t the interface of -3.5g-0.7Cu joint, an apparent change in the microstructure of reaction layer is found, comparing with that of -3.5g. Figure 10 shows a high magnification SEM image on the reaction layer. Fine particles can be found near to the reaction interface instead of faceted grains of Fe 2 in the second reaction layer for - g joint. For the details, TEM observation and analysis were carried out, and the results are shown in Fig. 11 with two diffraction patterns take from the first reaction layer and a fine particle indicated as and, respectively. The crosssectioned TEM image shows a first reaction layer and fine particles dispersed in matrix near to the reaction layer. oth reaction layer and fine particle have same Fe 2 phase without phase change. This fining of Fe 2 products is observed for all samples regardless of reaction time condition. Unfortunately, this fining mechanism by Cu addition to -g is yet to be cleared, and further work should be more studied on how Cu affect the morphology change.
Effect of Cu ddition to -g Lead-Free Solder on Interfacial Stability with 719 120 Reaction interface Fine particles 3 µm Joint Strength, σ/mpa 100 80 60 40 20 0 pure -3.5g -3.5g-0.7Cu 0 20 40 60 Reaction time (min) Fig. 10 SEM microstructure of the reaction interface for -3.5g-0.7Cu joint after a reaction for. Fig. 12 Joint strengths for the -3.5g and -3.5g-0.7Cu joints as a function of reaction time, comparing with the result of the pure reaction system. Fe 2 Fine particles (102)Fe 2 Fe 2 reaction layer 500 nm (102)Fe 2 Fig. 11 TEM micrograph and two diffraction patterns taken from () the Fe 2 reaction layer and () a Fe 2 fine particle, respectively. 3.3 Interface strength Lead-free soldering accompanies formations and a growths of various intermetallic reaction layer, 5 8) essential to obtain a proper solderability such as wetting or joint strength. Simultaneously, the microstructure change of reaction layer such as coarsening or fining of a reaction product can also affect the joint reliability due to their brittle nature. In the present work, there is an apparent difference of interface microstructure between -g and -g-cu joints. In this section, the effects of the microstructure change by Cu addition to -3.5g on joint strength are discussed. Figure 12 shows the strength changes of -3.5g and -3.5g-0.7Cu joints as a function of reaction time, comparing with the result 16) of the pure joint. Through all reaction time, -3.5g joint has similar results with that of the pure joint. In contrast, -3.5g-0.7Cu joint increases the joint strength by about 30 40 MPa for all reaction time conditions comparing with that for -3.5g joint. careful consideration of fracture mode can give a useful explanation for the joint strength change. Figure 13 presents the representative cross-sectioned images of fractured samples for -3.5g and -3.5g-0.7Cu joints, respectively. -3.5g joint shows a crack propagating along the interface between the first reaction layer and the second reaction layer
720 C.-W. Hwang and K. Suganuma -3.5g -3.5g-0.7Cu Second layer Crack First layer 5 µm 0.7Cu joint, Cu addition to -g affects the microstructure changes of solder and reaction interface. Ni dissolved from into molten solder reacts with and Cu, and forms a -phase of Cu-Ni- compound during the soldering without formations of Ni 3 4 compounds. -3.5g-0.7Cu joint shows fine particles near to the reaction layer instead of faceted Fe 2 grains in the second reaction layer for the - g joint. -3.5g-0.7Cu joint increases the joint strength by about 30-40 MPa higher than that of -3.5g joints. cknowledgements The present work was carried out with the support of a Grant-in-id for Scientific Research () in 2001 2002, and in 2002 of the 21COE program of The Japan Ministry of Education, Culture, Sports, Science and Technology. These grants and aid are greatly appreciated. The authors would also like to thank Prof. K. Niihara and Prof. H. Mori for their helpful advice. REFERENCES for all samples regardless of reaction time as shown in Fig. 13. In contrast, the crack for the -3.5g-0.7Cu joint progresses through the solder matrix as shown in Fig. 13 near to the fine particle region, not inside the reaction layer. This fining of the reaction products in the second reaction layer by Cu addition to -3.5g stops the interface crack. This may also increase the joint strengths for -3.5g- 0.7Cu joints. 4. Conclusions Crack Fine particles Reaction interface 5 µm Fig. 13 Two typical cross-sectioned SEM images on the fractured samples for -3.5g joint and -3.5g-0.7Cu joint, respectively. The interface microstructures of -3.5g and -3.5g- 0.7Cu with were investigated at the reaction temperature of 250 C. Joint strength was also evaluated for various reaction time conditions. For the -3.5g joint, Fe of reacts with to form double reaction layers of Fe 2. Ni from substrate dissolves into molten - 3.5g and the melt forms Ni 3 4 compounds and g 3 eutectic network during the solidification. For -3.5g- 1) K. Suganuma: Current Opinion in Solid State and Mater. Sci. 5 (2001) 55 64. 2) M. McCormack and S. Jin: J. Electron. Mater. 23 (1994) 635 640. 3) M.. Kinna: Int. SMPLE Electron. Conf. 6 (1992) 547 555. 4) D. R. Stelmak and J. I. Vesce: Natl. Electron. Packag. Prod. Conf. 1 (1995) 286 291. 5) C. W. Hwang, K. Suganuma, J. G. Lee and H. Mori: J. Electron. Mater. 32 (2003) 52 62. 6) C. W. Hwang, K. S. Kim and K. Suganuma: J. Electron. Mater. 32 (2003) 1249 1256. 7) K. Suganuma, K. Niihara, T. Shoutoku and Y. Nakamura: J. Mater. Res. 13 (1998) 2859 2865. 8) C. W. Hwang, K. Suganuma, M. Kiso and S. Hashimoto: J. Mater. Res. 18 (2003) 2540 2543. 9) S. Chada, R.. Fournelle, W. Laub and D. Shangguan: J. Electron. Mater. 29 (2000) 1214 1221. 10). J. Sunwoo, J. W. Morris and G. K. Lucey: Metal. Trans. 23 (1992) 1323 1332. 11) Y. C. Chan, lex C. K. So and J. K. L. Lai: Mater. Sci. Eng. 55 (1998) 5 13. 12) S. hat, L. Du, M. Sheng, L. Luo, W. Kempe and J. Freytag: J. Electron. Mater. 29 (2000) 1105 1109. 13) J. D. ernal: Nature 122 (1928) 54. 14) S. K. Kang and V. Ramachandran: Scr. Metal. 14 (1980) 421 424. 15) K. Suganuma and Y. Nakamura: J. Japan Inst. Metals 59 (1995) 1299 1305. 16) C. W. Hwang, K. Suganuma, J. G. Lee and H. Mori: J. Mater. Res. 18 (2003) 1202 1210. 17) C. W. Hwang, K. Suganuma, E. Saiz and. P. Tomsia: Trans. JWRI 30 (2001) 167 172. 18) E. Saiz, C. W. Hwang, K. Suganuma and. P. Tomsia: cta Mater. 51 (2003) 3185 3197. 19) I. E. nderson,.. Cook, J. H. Harringa and R. L. Terpstra: J. Electron. Mater. 31 (2002) 1166 1174. 20) D. Gur and M. amberger: cta mater. 46 (1998) 4917 4923. 21) J. V. Castell-Evans and S. Wach: J. Iron and Steel Inst. 211 (1973) 880 889. 22) K. S. Kim, S. H. Huh and K. Suganuma: Mater. Sci. Eng. 333 (2002) 106 114. 23) D. W. Henderson, T. Gosselin,. Sarkhel, S. K. Kang, W. K. Choi and D. Y. Shih: J. Mater. Res. 17 (2002) 2775 2778.