Cross-Interaction Between Au and Cu in Au/Sn/Cu Ternary Diffusion Couples

Cross-Interaction Between Au and Cu in Au/Sn/Cu Ternary Diffusion Couples C. W. Chang 1, Q. P. Lee 1, C. E. Ho 1 1, 2, *, and C. R. Kao 1 Department of Chemical & Materials Engineering 2 Institute of Materials Science & Engineering National Central University Jhongli City, Taiwan (* E-mail: kaocr@hotmail.com Phone/Fax: +886-3-4227382) Both Au and Cu are so-called fast diffusers in Sn, and can diffuse very long distances in Sn in a relatively short time. In this study, the cross-interaction between Au and Cu across a layer of Sn was investigated through the use of the Au/Sn/Cu ternary diffusion couples. A seven µm Au layer and a 100 µm Sn layer were electroplated over Cu foils to produce the Au/Sn/Cu diffusion couples. Aging at 200 o C revealed that cross-interaction could occur in as short as 10 minutes. Evidence of this cross-interaction included the formation of (Cu 1-x Au x ) 6 Sn 5 on the Au side of the diffusion couples as well as on the Cu side. The reaction products on the Au side included the Au-Sn binary compounds. Between the Au-Sn compounds and the Sn was (Cu 1-x Au x ) 6 Sn 5. The reaction products on the Cu side initially was only (Cu 1-x Au x ) 6 Sn 5, but a layer of Au-free Cu 3 Sn eventually formed between (Cu 1-x Au x ) 6 Sn 5 and Cu. A detailed atomic flux analysis showed that the Cu flux through the Sn layer was about 2-3 times higher than the Au flux at any moment. The results of this study show that the cross-interaction of Au and Cu in solders is extremely rapid, and cannot be ignored in those solder joints that have both elements present. Keywords: Au-Sn-Cu, cross-interaction, lead-free solder. 1

1. Introduction In electronic packages, the surface of the soldering pads is often coated with an Au layer to provide oxidation protection. This Au layer will rapidly dissolve into solder during reflow and form AuSn 4 particles dispersed inside the solder joint after reflow [1-3]. These AuSn 4 particles, when present in a large amount, are known to cause reliability problems, such as the brittle fracture of a solder joint [4-12]. The element Cu is also an element which is often in direct contact with the solders because it is the most common metal used as the conducting lines on the Si devices, substrates, and printed circuit boards. In fact, Cu is one of the constituents in the Sn-Ag-Cu lead-free solder family. Both Au and Cu are very mobile in solders, and can diffuse across very long distances in a relatively short time. Consequently, Cu and Au might interact even if they are separated by a thick layer of solder. The objective of this study is to investigate the cross-interaction between Au and Cu across a layer of Sn. In other words, we will study the Au/Sn/Cu ternary diffusion couple. A literature search revealed that there was no such study reported. Nevertheless, there are preliminary evidences showing that Au and Cu do interact in solders [13, 14]. The Au/Sn/Cu diffusion couple is a rather ideal system for studying such type of cross-interaction. This is because both the Au/Sn and Sn/Cu binary reactions have been well studied. In addition, the AuCuSn ternary isotherm is also well characterized. For a summary of the Au/Sn and Sn/Cu binary reactions, see a book chapter written by Zakel and Reichl [15]. The AuCuSn ternary isotherm at 190 o C, determined by Zakel [16], is reproduced in Fig. 1. The Au/Sn/Cu ternary diffusion couples used in this study were prepared by sequentially electroplating a layer of Sn and a layer of Au over the Cu substrate. There are two benefits by using such a method to prepare the diffusion couples. The first is that the distance between Au and Cu (or the Sn thickness) can be controlled at will. The second is 2

that the diffusion couples will not experience any high temperature processing step before the diffusion couples are subjected to the planned interdiffusion. If the diffusion couple had been prepared by melting the solder to bond Au and Cu, then a certain degree of cross-interaction might have occurred during bonding. The cross-interaction during the bonding stage could be, in fact, substantial because it had been reported that Cu and Ni could cross-interact during the reflow [17]. Consequently, in those diffusion couples bonded by melting, it will be difficult to distinguish to the extent of the cross-interactions that have occurred during the bonding stage from that during the reaction stage. 2. Experimental The Au/Sn/Cu diffusion couples were prepared by electroplating a seven µm Au layer and a 100 µm Sn layer over the Cu foil substrate. The Cu foil was 400 µm thickness and 99.6 wt% pure. Before electroplating, each Cu foil surface was polished, with one µm diamond abrasive used in the last polishing step. The Cu foil was then lightly etched in a 5 vol% HNO 3 -CH 3 OH solution for one minute, and cleaned with deionized water. After electroplating the Au and Sn layers, the foil was cut into 4 mm 8 mm diffusion couples. The diffusion couples were then subjected to aging at 200 o C. After aging, these samples were mounted in epoxy, sectioned by using a low-speed diamond saw, and metallographically polished to reveal the microstructure. The reaction zone for each diffusion couple was examined by using the optical microscopy (OM) and the scanning electron microscopy (SEM). The compositions of the intermetallic compounds were determined by the electron probe micro-analysis (EPMA). In EPMA analysis, the concentration of each element was measured independently, and the total weight percentage 3

of all elements was within 100 ± 1% in each case. The average value from at least three measurements was then reported. 3. Results The as-deposited Au/Sn/Cu diffusion couple is shown in Fig. 1. As can be seen here, there was no evidence of interfacial reaction during the electroplating of Au and Sn. Nevertheless, the interfacial reaction occurred very quickly at 200 o C, as shown in Fig. 2 for a diffusion couple that had been aged for 10 minutes. Figure 2 (a) shows that three Au-Sn binary compounds, AuSn, AuSn 2, and AuSn 4, and an Au-Cu-Sn ternary compound existed at the original Au/Sn interface. Basing on the results that are to be presented later, we propose this Au-Cu-Sn compound has the Cu 6 Sn 5 crystal structure with a small amount of Au dissolved in the Cu sublattice, i.e. (Cu 1-x Au x ) 6 Sn 5. The fast Au-Sn reaction is not surprising, as such rapid reaction had been reported before [15, 18]. However, the formation of (Cu 1-x Au x ) 6 Sn 5 after only 10 minutes of aging is unexpected. As the only source of Cu was from the opposite side of the diffusion couple, the formation of (Cu 1-x Au x ) 6 Sn 5 on the Au side suggested that the cross-interaction between the Au/Sn and the Sn/Cu interface had occurred across a 100 µm layer of Sn in 10 minutes. The original Sn/Cu interface also shows evidence of this cross-interaction. We had detected strong Au signal that could not attributed to background noise from the Cu 6 Sn 5 phase between Sn and Cu in Fig. 2(c). In other words, Au atoms had diffused through the Sn layer to be incorporated into Cu 6 Sn 5. It should be stressed again that during the sample preparation step, Sn did not go through the molten state. The results above were purely from a solid-state interdiffusion at 200 o C. The cross-interaction proceeded even further when the aging time reached 400 minutes, 4

as shown in Fig. 3. The entire Au layer had been consumed, and even the AuSn phase had disappeared completely. Bulk of the remaining Au-Sn binary compound was AuSn 4, and a few isolated AuSn 2 embedded in AuSn 4 still existed. The amount of (Cu 1-x Au x ) 6 Sn 5 on the Au side of the original diffusion couple increased and became a continuous layer. On the opposite side (Cu side) shown in Fig. 3 (c), the thickness of (Cu 1-x Au x ) 6 Sn 5 also increased. Moreover, a layer of Cu 3 Sn formed between (Cu 1-x Au x ) 6 Sn 5 and Cu. When the aging time reached 24 hours (Fig. 4), AuSn 2 had disappeared completely as shown in Fig. 4 (a). Interestingly, there was a discontinuous Sn phase between AuSn 4 and (Cu 1-x Au x ) 6 Sn 5. An EPMA line scan was performed along the line ab marked in Fig. 4 (a), and the concentration profile is shown in Fig. 5. From the EPMA measurement, the composition of (Cu 1-x Au x ) 6 Sn 5 was determined to be 14.7 at.%, Au, 35.8 at.% Cu, and 49.5 at.% Sn. According to the 190 o C Au-Cu-Sn isotherm determined by Zakel [16] and reproduced in Fig. 6, this composition corresponded to the composition of (Cu 1-x Au x ) 6 Sn 5 at the Sn-AuSn 4 -(Cu 1-x Au x ) 6 Sn 5 three-phase equilibrium. As for the reaction on the Cu side, the thickness of (Cu 1-x Au x ) 6 Sn 5 as well as Cu 3 Sn increased as the reaction time increased, as shown in Fig. 4 (c). The (Cu 1-x Au x ) 6 Sn 5 phase now had substantial amount of Au dissolved (x= 0.23~0.25), but Au was still not detected in Cu 3 Sn. In Fig. 4 (b), one also noticed that there were AuSn 4 particles inside the Sn layer. The existence of these particles will be explained in the discussion section. As shown in Table I, the composition of (Cu 1-x Au x ) 6 Sn 5 did not change with the reaction time. The thickness of (Cu 1-x Au x ) 6 Sn 5 on the Au side as well as that on the Cu side is plotted against the square root of time in Fig. 7. Also shown in Fig. 7 is the thickness of Cu 6 Sn 5 from the binary Sn/Cu diffusion couple. The configuration, preparation, and aging treatment of these binary diffusion couples were the same as those of the Au/Sn/Cu ternary diffusion 5

couples, except that the Au layer was not deposited. As can be seen in Fig. 7, (Cu 1-x Au x ) 6 Sn 5 on the Au side was the thickest, while (Cu 1-x Au x ) 6 Sn 5 on the Cu side is the thinnest. The data in Fig. 7 is summarized in Table II. 4. Discussion The Au-Sn-Cu phase diagram had been studied by Roeder at 170 o C [19] and by Zakel at 190 o C [16]. The one determined by Zakel (redrawn in Fig. 6) is to be used for the discussion as the temperature (190 o C) is closer to our experimental condition. In Figure 6, a diffusion path for the diffusion couple that had been aged for 24 hours is also shown. One question naturally arise is that why AuSn 4 on the Au side was covered with a layer of (Cu 1-x Au x ) 6 Sn 5, while those AuSn 4 particles inside the Sn layer was not, even though these particles located closer to the Cu source? The diffusion path shown in Fig. 6 clearly rationalized that such diffusion path was allowed. In other words, such feature was not forbidden for ternary diffusion couples. The thicknesses of (Cu 1-x Au x ) 6 Sn 5 under various conditions were shown in Fig. 7. The three sets of data all seemed to follow the parabolic kinetics. The (Cu 1-x Au x ) 6 Sn 5 on the Au side was the thickest, while that on the Cu side was the thinnest. The total consumption of the Cu foil corresponded to the formation of (Cu 1-x Au x ) 6 Sn 5 on the Au side and the Cu side. Comparing the amounts of this compound formed for the case with and without the Au layer, the existence of the Au layer clearly accelerated the Cu consumption. Apparently, the existence of the Au layer provided extra driving force, which accelerated the Cu consumption. The data in Fig. 7 and also enables us to calculate the Au flux and Cu flux across the Sn 6

layer. This is because the solubilities of Au and Cu in Sn are very low, about 0.2 at.% for Au in solid Sn at 200 o C [20] and 0.01 at.% for Cu in solid Sn at 227 o C [21], and consequently the amounts of Cu and Au present in the Sn layer can be ignored. We also neglected the Au present in the form of AuSn 4 inside the Sn layer. From the EPMA measurement, the compositions of the (Cu 1-x Au x ) 6 Sn 5 phase on both sides were known. With these assumptions, the integration of the Au flux over time correspond to the amount of Au that was incorporated into (Cu 1-x Au x ) 6 Sn 5 on the Cu side per unit area, and the integration of the Cu flux over time corresponded to the amount of Cu presented in (Cu 1-x Au x ) 6 Sn 5 on the Au side per unit area. Therefore, the derivative of the thickness versus time data will produce the flux at any instant. The resulting Au flux and Cu flux calculated this way versus time are shown in Fig. 8 and Table III. In calculating Fig. 8, the density for (Cu 1-x Au x ) 6 Sn 5 was needed. The binary compound Cu 6 Sn 5 was reported to have a density of 8.28 g/cm 3 [22]. The density of (Cu 1-x Au x ) 6 Sn 5 will increase slightly because Au is a heavier atom. We estimated the density of (Cu 0.7 Au 0.3 ) 6 Sn 5 to be about 8.8 g/cm 3, by using the lattice constant values for (Cu 0.7 Au 0.3 ) 6 Sn 5 measured in this study (a=0.4229 nm and c=0.5170 nm). It can be seen that the Cu flux through the Sn layer was about 2-3 times higher than the Au flux at any moment. According to the literature data, at 200 o C the diffusivity of Au along the c axis of Sn is 4.8 10-12 (m 2 /sec) [23] and that of Cu along the c axis of Sn is 5.4 10-11 (m 2 /sec) [24]. The diffusivity of Cu in Sn is one order of magnitude larger than the diffusivity of Au. The relative values for Au and Cu along other directions exhibits similar ration. Apparently, the difference in the values of the diffusivity is larger than that of the atomic flux. The difference between the relative magnitude of the flux and the diffusivity is probably due to the different driving forces for the diffusion of Au and Cu across the Sn layer. In Fig. 4 (a), there was Sn phase between AuSn 4 and (Cu 1-x Au x ) 6 Sn 5 interface. These Sn 7

formed only after AuSn 2 had disappeared and AuSn 4 became the only Au-Sn binary compound. We believe these Sn came from the decomposition of the AuSn 4 phase. The AuSn 4 phase had to decompose to release Au, so that Au can reacted with the incoming Cu to form (Cu 1-x Au x ) 6 Sn 5. In other words, (Cu 1-x Au x ) 6 Sn 5 grew in the expense of AuSn 4. In fact, if we let the reaction continue, all AuSn 4 will be consumed, and the only compound left at that interface will be (Cu 1-x Au x ) 6 Sn 5. The released Sn from the decomposition accumulated and became the Sn phase. 5. Conclusion At 200 o C, the Cu atoms could diffuse through the 100 µm Sn layer and resulted in the formation of the (Cu 1-x Au x ) 6 Sn 5 on the Au side in 10 minutes. Under the same conditions, the Au atoms could diffuse through the Sn layer and became incorporated in the Cu 6 Sn 5 compound. These are evidences showing that Au and Cu can cross-interact across a distance of 100 µm even when Sn is in the solid state. During the joining process by soldering, the solder becomes molten and the cross-interaction becomes even more substantial. Consequently, the results of this study shows that the cross-interaction of Au and Cu in solders is extremely rapid, and cannot be ignored in those solder joints that have both elements present. Acknowledgment. This work was supported by the National Science Council of R.O.C. through grants NSC-93-2214-E-008-008 and NSC-93-2216-E-008-010. The authors thank Chung-Yuan, Kao (NTU) for assistance in EPMA measurements. 8

References 1. C. E. Ho, Y. M. Chen, and C. R. Kao, J. Electron. Mater., 28, p.1231, 1999. 2. C. E. Ho, R. Zheng, G. L. Luo, A. H. Lin, and C. R. Kao, J. Electron. Mater., 29, p.1175, 2000. 3. S. W. Chen, and Y. W. Yen, J. of Electron. Mater., 30, p.1133, 2001. 4. C. E. Ho, S. Y. Tsai, and C. R. Kao, IEEE Trans. Adv. Packaging, 24, p.493, 2001. 5. C. E. Ho, W. T. Chen, and C. R. Kao, J. Electron. Mater., 30, p.379, 2001. 6. C. M. Liu, C. E. Ho, W. T. Chen, and C. R. Kao, J. Electron. Mater., 30, p. 1152, 2001. 7. T. Y. Lee, W. J. Choi, K. N. Tu, J. W. Jang, S. M. Kuo, J. K. Lin, D. R. Frear, K. Zeng, and J. K. Kivilahti, J. Mater. Res., 17, p.291, 2002. 8. L. C. Shiau, C. E. Ho, and C. R. Kao, Solder. Surf. Mt. Tech., 14, p.25, 2002. 9. C. E. Ho, L. C. Shiau, and C. R. Kao, J. Electron. Mater., 31, p.1264, 2002. 10. C. B. Lee, J. W. Yoon, S. J. Suh, S. B. Jung, C. W. Yang, C. C. Shur, and Y. E. Shin, J. Mater. Sci., 14, p.487, 2003. 11. C. E. Ho, Y. L. Lin, J. Y. Tsai, and C. R. Kao, J. Chin. Inst. Chem. Eng., 34, p.387, 2003. 12. T. Laurila, V. Vuorinen, T. Mattila, and J. K. Kivilahti, J. Electron. Mater., 34, p.103, 2005. 13. J. Y. Tsai, C. W. Chang, Y. C. Shieh, Y. C. Hu, and C. R. Kao, J. Electron. Mater., 34, in press (February issue), 2005. 14. Y. L. Lin, W. C. Luo, Y. H. Lin, C. E. Ho, and C. R. Kao, J. Electron. Mater., 33, p.1092, 2004. 9

15. E. Zakel and H. Reichl, in Flip Chip Technologies, ed. By J. H. Lau, Chapter 15, McGraw-Hill, New York, 1995. 16. E. Zakel, thesis, Technical University-Berlin, 1994. 17. C. M. Tsai, W. C. Luo, C. W. Chang, Y. C. Shieh, and C. R. Kao, J. Electron. Mater., 33, p.1424, 2004. 18. P. G. Kim and K. N. Tu, J. Appl. Phys., 80, p.3822, 1996. 19. J. Roeder, Doctoral Dissertation, Lehigh University, 1988. 20. H. Okamoto and T. B. Massalski, eds., Phase Diagram of Binary Gold Alloys, ASM International, Metals Park, OH, p.278, 1978. 21. C. E. Homer and H. Plummer, J. Inst. Met., 64, p.169, 1939. 22. H. P. R. Frederikse, R. J. Fields, and A. Feldman, J. App. Phys., p2879, 1992 23. B. F. Dyson, J. Appl. Phys., p. 2375, 1966. 24. B. F. Dyson, T. R. Anthony, and D. Turnbull, J. Appl. Phys., p. 3408, 1967. Figure Captions 10

Fig. 1 The microstructure of the as-deposited Au/Sn/Cu ternary diffusion couple. The Au layer was seven µm thick, and the Sn layer was 100 µm. Fig. 2 The backscattered electron micrographs of a diffusion couple that had been aged at 200 o C for 10 minutes. (a) The Au side of the couple, (b) the whole couple, and (c) the Cu side. Fig.3 The backscattered electron micrographs of a diffusion couple that had been aged at 200 o C for 400 minutes. (a) The Au side of the couple, (b) the whole couple, and (c) the Cu side. Fig. 4 The backscattered electron micrographs of a diffusion couple that had been aged at 200 o C for 24 hours. (a) The Au side of the couple, (b) the whole couple, and (c) the Cu side. Fig. 5 The EPMA line scan across the line ab shown in Fig. 4 (a). Fig. 6 The Au-Cu-Sn ternary isotherm at 190 o C. This isotherm was redrawn from the one determined by Zakel [16]. Fig. 7 The thickness of (Cu 1-x Au x ) 6 Sn 5 on the Au side, the thickness of (Cu 1-x Au x ) 6 Sn 5 on the Cu side, and the thickness of Cu 6 Sn 5 from the binary Sn/Cu diffusion couple versus the aging time. Fig. 8 The Au flux and the Cu flux across the Sn layer versus the aging time. Table I - The values of x in (Cu 1-x Au x ) 6 Sn 5 after 1, 24, and 189 hours. 11

x value in (Cu 1-x Au x ) 6 Sn 5 1 hrs 24 hrs 189 hrs Au side 0.30 0.29 0.31 Cu side 0.25 0.23 0.23 12

Table II Thickness of (Cu 1-x Au x ) 6 Sn 5 in Au/Sn/Cu diffusion couple and Sn/Cu diffusion couple at 200 o C. Time (Cu 1-x Au x ) 6 Sn 5 at Au/Sn/Cu (µm) Cu 6 Sn 5 at Sn/Cu (µm) (min) Cu side Au side 10 0.9 0.7 0.8 30 1.2 1.4 1.5 70 1.3 2.6 1.6 100 2.3 3.0 2.1 150 2.2 3.5 2.0 200 3.4 4.2 3.2 300 2.4 5.9 4.1 400 4.2 6.1 5.1 500 4.2 6.4 5.5 720 6.7 9.1 7.1 1080 7.8 9.6 9.0 1440 8.5 10.2 8.9 13

Table III- The Au flux and the Cu flux across the Sn layer in the Au/Sn/Cu diffusion couple at 200 o C. Time (min) Au flux Cu flux (atoms/cm 2 sec) (atoms/cm 2 sec) 10 5.2 10 14 1.3 10 15 30 3.0 10 14 7.6 10 14 70 2.0 10 14 5.0 10 14 100 1.6 10 14 4.2 10 14 150 1.3 10 14 3.4 10 14 200 1.2 10 14 3.0 10 14 300 9.5 10 13 2.4 10 14 400 8.2 10 13 2.1 10 14 500 7.3 10 13 1.9 10 14 720 6.1 10 13 1.5 10 14 1080 5.0 10 13 1.3 10 14 1440 4.3 10 13 1.1 10 14 14