Void formation at the interface in Sn/Cu solder joints - Semantic Scholar
Microelectronics Reliability 51 (2011) 2314–2318
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
Void formation at the interface in Sn/Cu solder joints Yang Yang a,⇑, Hao Lu a,b, Chun Yu a, Yongzhi Li a a b
School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China Key Lab of Shanghai Laser Manufacturing and Materials Modification, Shanghai Jiao Tong University, Shanghai 200240, PR China
a r t i c l e
i n f o
Article history: Received 14 March 2011 Received in revised form 1 June 2011 Accepted 21 June 2011 Available online 23 July 2011
a b s t r a c t The effect of electroplated Cu (EPC), electroplated Sn (EPS) and Cu addition (0.7 wt.%) on the void formation at the reaction interface was investigated through the reaction of solders with Cu substrates. The voids were observed at the Cu3Sn/EPC interface in the Sn/EPC joints after aging at 150 °C, while not at the Cu3Sn/high purity Cu (HPC) interface in the Sn/HPC joints even after aging at 180 °C for 720 h. In the EPS/HPC joints, the voids appeared at both Cu6Sn5/Cu3Sn and Cu3Sn/HPC interfaces after long time aging at 150 °C. The formation of these voids may be induced by the impurities, which were introduced during the electroplating process. The addition of Cu could reduce the interdiffusion of Cu and Sn at the interface and retard the growth of Cu3Sn layer. Consequently, the formation of voids was suppressed. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction Recent years, the interfacial reactions between solder and copper under bump metallurgy (UBM) have been paid great attention by researchers due to the reliability concerns—intermetallic compounds (IMCs) and micro-voids. Especially, the formation mechanism of micro-voids is rather complicated, on which there are many different views. Early studies [1,2] attributed the formation of voids to the unbalanced diffusion rates of different diffusing species without considering the effect of impurities (both organic and inorganic). The recent review [3] also did not take this factor into account. The experimental work carried out by Yang et al. [4] took the research on micro-voids to a new high. They reported that the microvoids can be observed at the interface between Sn–Ag solder and EPC, but not between Sn–Ag solder and rolled Cu foil. Finally, they suggested that the voids were possibly caused by the excessive hydrogen introduced during the electroplating process. Actually, impurities and defects have been detected in the electroplated Cu film, which could induce the formation of voids within the film [5– 7]. Laurila et al. [8] in their review discussed the correlation of void formation with impurities in the deposited Cu film, and confirmed that the voids did not form at all at the interface of Sn/HPC and Sn–Ag–Cu/HPC joints. Later, Kim and Yu [9–12] clearly pointed out the detrimental effect of residual impurities in electroplated Cu UBM on the void formation, and especially studied the effect of S
⇑ Corresponding author. Tel./fax: +86 21 34202548. E-mail address: [email protected] (Y. Yang). 0026-2714/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.microrel.2011.06.026
element. Liu et al. [13,14], proceeding from the source of impurities, systematically studied the influence of electroplating parameters on void formation at the interface. According to the different voiding mechanism, various methods for preventing the formation of voids at the interface were proposed: (1) The suppression of the heterogeneous nucleation sites, such as impurities, grain boundaries, dislocations [15]. (2) The minor addition of alloying elements. (i) The elimination of harmful elements such as S [11]. (ii) Retarding the growth of Cu3Sn layer since the voids always generated with the formation of Cu3Sn layer, but seldom appeared within the Cu6Sn5 layer [16–22]. Yin and Borgesen [23] carried out the high temperature annealing experiment of Cu foil before the reflow process, eliminating the heterogeneous nucleation sites in the foil. The reaction interface using the heat-treated foil was almost void-free after subsequent soldering and thermal aging. Laurila et al. [22] reported that some alloying (such as Ni, Co and Zn) could dissolve in either of the Cu–Sn IMCs and retard the growth of Cu3Sn layer. Accordingly, the formation of voids was depressed. In addition, Wang et al. [24] suggested that the addition of Cu could increase the Cu concentration in solder, reduce the driving force for Cu diffusion and inhibit the formation of voids. In this work, the Sn/EPC and EPS/HPC joints were prepared to investigate the mechanism of void formation and the correlation between Cu3Sn phase and void. The effect of Cu on the void formation was also studied.
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
2. Experimental procedures The solders used in this study were commercial Sn (99.99%) and Sn0.7Cu (wt.%). Two types of Cu substrates were introduced, high purity foil (99.99%, 0.1 mm in thickness) and electroplated Cu film (approximately 10 lm in thickness) deposited directly on the surface of foil. The electroplating solution contained H2SO4 + CuSO4 + Cl + PEG (ethylene glycol) [13]. Both kinds of foils then were cut into 10 mm 10 mm square pieces. The solder joints were prepared by melting solders on the square pieces at 260 °C. Prior to the reflow process, these pieces were deoxidized and degreased in 5 wt.% NaOH and 5 vol.% HCl solutions sequentially, rinsed in de-ionized water after each step, and treated with flux. To investigate the interfacial microstructure in the solder joints, the isothermal aging for the as-reflowed samples was performed at 150 °C and 180 °C respectively. Then the samples were mounted in epoxy and metallurgically polished. The interesting zones at the interface were observed by using a scanning electron microscope (SEM).
2315
In addition, there were numbers of cracks existed within the Cu6Sn5 layer, which may be produced during the preparation of samples. 3.2. Effect of electroplated Sn on the formation of void In the case of Sn/EPC joints, the voids were mainly located in the Cu3Sn layer near the EPC, as seen in Fig. 3. If a layer of Sn is electroplated on the HPC, what would happen at the reaction interface after aging? Would the voids appear within the Cu6Sn5 layer? In order to verify this assumption, the corresponding experiments were carried out. A layer of Sn (approximately 10–15 lm
3. Results and discussion 3.1. Effect of UBM on the formation of void The cross-sectional microstructure of as-reflowed Sn/HPC and Sn/EPC joints are shown in Fig. 1. In both kinds of joints, a layer of scallop-shaped Cu6Sn5 formed at the interface without void. A very thin layer of another phase (possibly Cu3Sn) was also observed at the interface of Cu6Sn5/Cu [2]. After aging at 150 °C for 720 h, two thick IMC layers, Cu6Sn5 and Cu3Sn, were developed in the Sn/HPC joints, and the reaction interface remained void-free (Fig. 2a). With the same aging period, there were still no voids after raising the aging temperature up to 180 °C (Fig. 2b). In the Sn/EPC joints, a few voids were scattered within the Cu3Sn layer after aging at 150 °C for 240 h (Fig. 3a), and these voids covered around 1.6% of the area of the Cu3Sn phase. Extending the aging period to 480 h, numerous voids appeared (Fig. 3b), covering approximately 6% of the area of the Cu3Sn layer.
Fig. 1. Back scattered electron (BSE) images of as-reflowed solder joints. (a) Sn/HPC joint; (b) Sn/EPC joint.
Fig. 2. BSE images of Sn/HPC joints. (a) 150 °C, 720 h; (b) 180 °C, 720 h.
Fig. 3. BSE images of Sn/EPC joints after aging at 150 °C. (a) 240 h; (b) 480 h.
2316
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
Fig. 4. BSE images of EPS/HPC joints after aging at 150 °C. (a) 0 h; (b) 240 h; (c) 720 h; (d) 720 h.
in Fig. 6a. Prolonging the aging period to 480 h, the number and size of the voids had no obvious change though the Cu3Sn layer became much thicker (Fig. 6b). Thus, the area ratio of voids to Cu3Sn became lower, around 2%. 3.4. Discussion
Fig. 5. BSE images of as-reflowed Sn0.7Cu/EPC joint.
in thickness) was deposited on the surface of HPC using a commercially available plating solution. Then the samples were cut into 10 mm 10 mm square pieces and then aged at 150 °C. The preparation and characterization for the aged samples were the same as those in the second section (Experimental procedures). It was interesting that the Cu6Sn5 and Cu3Sn layers had been formed at the reaction interface of the as-electroplated sample, as shown in Fig. 4a. The IMC layers may grow during the electroplating process and subsequent storage at room temperature. Besides, the lost parts of the film (in the circle, Fig. 4a) were also produced during the preparation of samples. After aging at 150 °C for 240 h, both Cu6Sn5 and Cu3Sn layers became thicker, while Sn film became thinner (Fig. 4b). A few voids were observed not only at the Cu3Sn/Cu interface, but also at the Cu6Sn5/Cu3Sn interface. Prolonging the aging period to 720 h, more voids were formed at the Cu6Sn5/Cu3Sn interface and only a small amount of Sn was left (Fig. 4c and d). 3.3. Effect of Cu on the formation of void Similar to the as-reflowed Sn/Cu joints, a thin layer of Cu6Sn5 appeared at the interface without voids, as seen in Fig. 5. After aging at 150 °C for 240 h, a few voids can be observed, which covered approximately 4.6% of the area of Cu3Sn layer, as shown
During the isothermal aging, the formation of intermetallic compounds (IMCs) will cause a significant volume contraction, which can be estimated by the structural data [25–28]. A 9.97% reduction of volume will happen when Cu reacts with Sn forming Cu3Sn crystal, and a reduction of 5.44% in volume when forming g0 –Cu6Sn5 crystal (low temperature phase). The change of volume may result in the generation of inner stress in the solder joints [29]. The volume contraction, Kirkendall effect (KE) and residual stress generated during soldering were common in the solder joints. They were all potential factors of void formation at the reaction interface. However, the fact that voids appeared in the joints using EPC, while not in the joints using HPC [4,8,9,19,30,31], implied that the factors above might be not enough to induce the formation of voids, and it should be the Cu substrate that was responsible for this. In the process of electroplating, a number of impurities and defects may be introduced into the deposited Cu film. Lee and Park [32] reported that the organic impurity, poly ethylene glycol (PEG), was absorbed at the grain boundary of EPC film. Besides, some impurity elements S, Cl, C and O, were detected in the electroplated film by using SIMS (secondary ion mass spectrometry) analysis [9,13,23]. Contradictory viewpoints on the mechanism of void formation were put forward according to the different impurities. Kim and Yu [9] suggested that the segregation of S to the Cu/Cu3Sn interface would lower the free energy barrier for void nucleation and accelerate the process. However, Yin and Borgesen [23] considered that the void formation was associated with a complex of Cu+–PEG–Cl, but not S. The vacancy sinks at the interface may be pinned down by the organic complexes, and excess vacancies were left [33]. The locally high stress originated from the complexes would also accelerate the diffusion of Cu atoms around them (Cu vacancies were
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
2317
left) [32]. Moreover, these complexes could act as the heterogeneous sites and lower the nucleation barrier for the voids. Based on the discussion above, two necessary conditions for the nucleation of voids should be satisfied: (1) The unbalanced diffusion. In the Sn/Cu couple annealed at 150 °C, Sn was the slightly fast diffuser in the Cu6Sn5 layer, while Cu was the main diffusing species in the Cu3Sn layer [34]. (2) Vacancy supersaturation, that is
E > Ec
ð1Þ
where Ec is the critical energy for void nucleation and E is the energy possessed by the active area for voiding. The Ec at the Cu3Sn/ EPC interface was decreased by the introduced impurities. During the thermal aging, the Cu3Sn/EPC interface shifted towards the substrate gradually through the consumption of Cu film. The impurities contained in the film would get into the reaction interface. The Cu3Sn/EPC interface was the most active area for the voiding, which also served as the heterogeneous sites. With the further migration of Cu3Sn/EPC interface toward EPC, the voids originally formed were left behind, trapping inside the Cu3Sn layer. That may be the reason why most of the voids located in the Cu3Sn layer close to the EPC. Similar to the Cu plating solution, the solder plating solution also contained some usual additives (like PEG) which might be incorporated into the electroplated film [35,36]. However, after thermal aging for the EPS/HPC joints, the voids were formed at the Cu6Sn5/Cu3Sn and Cu3Sn/Cu interfaces, while not within the Cu6Sn5 layer though the Cu6Sn5 layer had the similar conditions with Cu3Sn layer in the Sn/EPC joint (close to the electroplated film). This indicated that the voids were inclined to form and grow in the Cu3Sn layer, while not in the Cu6Sn5 layer, which may be related to the microstructure of the two Cu–Sn IMCs. The impurities may be introduced into the Cu6Sn5 layer through the consumption of EPS film, and then into the Cu6Sn5/Cu3Sn interface during the growth of Cu3Sn toward the Cu6Sn5 layer. The joints with EPS film were different from the Sn/HPC joints which had a large amount of Sn. The possible reactions at the Cu6Sn5/Cu3Sn interface are
Fig. 6. BSE images of Sn0.7Cu /EPC joints after aging at 150 °C. (a) 240 h; (b) 480 h.
for the Sn/EPC joints, 0.792 (240 h) and 1.229 (480 h) respectively. The Cu3Sn layer in the Sn0.7Cu/EPC joints was thinner than that in the Sn/EPC joints, which was in agreement with the results obtained by Lee et al. [37]. Accordingly, the formation of voids was depressed owing to the correlation between Cu3Sn phase and void. 4. Conclusions
Cu6 Sn5 þ 9½Cu ! 5Cu3 Sn
ð2Þ
In this work, the void formation in the Sn/Cu joints was studied, taking EPC, EPS and Cu addition into consideration. After thermal aging, the voids were observed at the reaction interface in the Sn/EPC and EPS/HPC joints while not in the Sn/HPC joints. The voids were highly correlated with the Cu3Sn phase. These voids may be mainly caused by the impurities incorporated during electroplating. The addition of Cu element (0.7 wt.%) retarded the growth of Cu3Sn layer and depressed the formation of voids.
Cu6 Sn5 ! 3½Sn þ 2Cu3 Sn
ð3Þ
Acknowledgements
where [Cu] and [Sn] are diffusing atoms. With the extension of aging period, four different phases at the interface (Sn, Cu6Sn5, Cu3Sn and Cu) would evolve into three (Cu6Sn5, Cu3Sn and Cu) owing to the limited amount of Sn, as seen in Fig. 4c. It can be noted in Eqs. (2) and (3) that there was still enough driving force for [Cu] atoms to move from Cu3Sn layer to Cu6Sn5 layer for the formation of Cu3Sn needed a large number of [Cu] atoms. However, the supply of [Sn] atoms was inadequate. The unbalanced diffusion and low Ec caused by the impurities may induce the formation of voids at the Cu6Sn5/Cu3Sn interface. In addition, the large voids at the EPS/Cu6Sn5 interface may be caused by the volume contraction, as shown in Fig. 4c (in the circle). The addition of Cu would reduce the diffusion of Cu and Sn, especially for Cu [24]. The degree of the unbalanced diffusion between Cu and Sn was also decreased. Therefore, the formation of voids was suppressed. On the other hand, the Cu addition had a significant effect on the interfacial morphology. The total thickness of IMC layer in the Sn0.7Cu/EPC joints was close to that in the Sn/ EPC joints, as observed in Figs. 3 and 6. However, the thickness ratios of Cu3Sn to Cu6Sn5 for the Sn0.7Cu/EPC joints, 0.342 (aging for 240 h) and 0.716 (480 h) respectively, were much lower than those
This work was supported by the National Natural Science Foundation of China (Grant Nos. 50975176 and 50475021). The authors would like to thank Prof. C. R. Kao at National Taiwan University for his kind help (corrosive agents, BSE). Reference [1] Liu CY, Tu KN, Sheng TT, Tung CH, Frear DR, Elenius P. J Appl Phys 2000;87:750–4. [2] Zeng K, Stierman R, Chiu T-C, Edwards D, Ano K, Tu KN. J Appl Phys 2005;97:024508. [3] Kim D, Chang J, Park J, Pak JJ. J Mater Sci Mater Electron 2000. doi: 10.1007/ s10854-011-0357-2. [4] Yang W, Messler RW, Felton LE. J Electron Mater 1994;23:765–72. [5] Moriyama M, Konishi S, Tsukimoto S, Murakami M. Mater Trans, JIM 2004;45:3172–6. [6] Stangl M, Dittel V, Acker J, Hoffmann V, Gruner W, Strehle S, et al. Appl Surf Sci 2005;252:158–61. [7] Uedono A, Mori K, Ito K, Imamizu K, Hachiya T, Kamijo H, et al. Jpn J Appl Phys 2007;46:L483–5. [8] Laurila T, Vuorinen V, Kivilahti JK. Mater Sci Eng 2005;R49:1–60. [9] Yu J, Kim JY. Acta Mater 2008;56:5514–23. [10] Kim JY, Yu J. Appl Phys Lett 2008;92:092109. [11] Kim JY, Yu J, Kim SH. Acta Mater 2009;57:5001–12.
2318
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
[12] Kim S, Yu J. J Appl Phys 2010;108:083532. [13] Liu Y, Wang J, Yin L, Kondos P, Parks C, Borgesen P, et al. J Appl Electrochem 2008;38:1695–705. [14] Wafula F, Liu Y, Yin L, Bliznakov S, Borgesen P, Cotts EJ, et al. J Electrochem Soc 2010;157:D111–8. [15] Cogan S, Holmes DS, Rose RM. Appl Phys Lett 1979;35:557–9. [16] Tsai JY, Hu YC, Tsai CM, Kao CR. J Electron Mater 2003;32:1203–8. [17] Anderson IE, Harringa JL. J Electron Mater 2006;35:94–106. [18] Ho CE, Yang SC, Kao CR. J Mater Sci Mater Electron 2007;18:155–74. [19] Cho MG, Kang SK, Shih DY, Lee HM. J Electron Mater 2007;36:1501–9. [20] Gao F, Nishikawa H, Takemoto T. J Electron Mater 2008;37:45–50. [21] Wang YW, Lin YW, Kao CR. Microelectron Reliab 2009;49:248–52. [22] Laurila T, Vuorinen V, Paulasto-Kröckel M. Mater Sci Eng 2010;R68:1–38. [23] Yin L, Borgesen P. J Mater Res 2011;26:455–66. [24] Wang YW, Lin YW, Kao CR. J. Alloys Compd. 2010;493:233–9. [25] Suh I-K, Ohta H, Waseda Y. J Mater Sci 1988;23:757–60.
[26] Deshpande VT, Sirdeshmukh DB. Acta Crystallogr 1962;15:294–5. [27] Burkhardt W, Schubert K. Z Metallkd 1959;50:442–52. [28] Larsson AK, Stenberg L, Lidin S. Acta Crystallogr, Sect B: Struct Sci 1994;50:636–43. [29] Belyakov S, Atkinson HV, Gill SPA. J Electron Mater 2010;39:1295–7. [30] Vianco PT, Erickson KL, Hopkins PL. J Electron Mater 1994;23:721–7. [31] Deng X, Piotrowski G, Williams JJ, Chawla N. J Electron Mater 2003;32:1403–13. [32] Lee C-H, Park C-O. Jpn J Appl Phys 2003;42:4484–8. [33] Liu PL, Shang JK. Scripta Mater 2005;53:631–4. [34] Paul A, Ghosh C, Boettinger WJ. Metall Mater Trans A 2011;42:952–63. [35] Homma T, Sato H, Kobayashi H, Arakawa T, Kudo H, Osaka T, et al. J Electroanal Chem 2003;559:143–8. [36] Tsai YD, Hu CC. J Electrochem Soc 2009;156:D490–6. [37] Lee TY, Choi WJ, Tu KN, Jang JW, Kuo SM, Lin JK, et al. J Mater Res 2002;17:291–301.
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
Void formation at the interface in Sn/Cu solder joints Yang Yang a,⇑, Hao Lu a,b, Chun Yu a, Yongzhi Li a a b
School of Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China Key Lab of Shanghai Laser Manufacturing and Materials Modification, Shanghai Jiao Tong University, Shanghai 200240, PR China
a r t i c l e
i n f o
Article history: Received 14 March 2011 Received in revised form 1 June 2011 Accepted 21 June 2011 Available online 23 July 2011
a b s t r a c t The effect of electroplated Cu (EPC), electroplated Sn (EPS) and Cu addition (0.7 wt.%) on the void formation at the reaction interface was investigated through the reaction of solders with Cu substrates. The voids were observed at the Cu3Sn/EPC interface in the Sn/EPC joints after aging at 150 °C, while not at the Cu3Sn/high purity Cu (HPC) interface in the Sn/HPC joints even after aging at 180 °C for 720 h. In the EPS/HPC joints, the voids appeared at both Cu6Sn5/Cu3Sn and Cu3Sn/HPC interfaces after long time aging at 150 °C. The formation of these voids may be induced by the impurities, which were introduced during the electroplating process. The addition of Cu could reduce the interdiffusion of Cu and Sn at the interface and retard the growth of Cu3Sn layer. Consequently, the formation of voids was suppressed. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction Recent years, the interfacial reactions between solder and copper under bump metallurgy (UBM) have been paid great attention by researchers due to the reliability concerns—intermetallic compounds (IMCs) and micro-voids. Especially, the formation mechanism of micro-voids is rather complicated, on which there are many different views. Early studies [1,2] attributed the formation of voids to the unbalanced diffusion rates of different diffusing species without considering the effect of impurities (both organic and inorganic). The recent review [3] also did not take this factor into account. The experimental work carried out by Yang et al. [4] took the research on micro-voids to a new high. They reported that the microvoids can be observed at the interface between Sn–Ag solder and EPC, but not between Sn–Ag solder and rolled Cu foil. Finally, they suggested that the voids were possibly caused by the excessive hydrogen introduced during the electroplating process. Actually, impurities and defects have been detected in the electroplated Cu film, which could induce the formation of voids within the film [5– 7]. Laurila et al. [8] in their review discussed the correlation of void formation with impurities in the deposited Cu film, and confirmed that the voids did not form at all at the interface of Sn/HPC and Sn–Ag–Cu/HPC joints. Later, Kim and Yu [9–12] clearly pointed out the detrimental effect of residual impurities in electroplated Cu UBM on the void formation, and especially studied the effect of S
⇑ Corresponding author. Tel./fax: +86 21 34202548. E-mail address: [email protected] (Y. Yang). 0026-2714/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.microrel.2011.06.026
element. Liu et al. [13,14], proceeding from the source of impurities, systematically studied the influence of electroplating parameters on void formation at the interface. According to the different voiding mechanism, various methods for preventing the formation of voids at the interface were proposed: (1) The suppression of the heterogeneous nucleation sites, such as impurities, grain boundaries, dislocations [15]. (2) The minor addition of alloying elements. (i) The elimination of harmful elements such as S [11]. (ii) Retarding the growth of Cu3Sn layer since the voids always generated with the formation of Cu3Sn layer, but seldom appeared within the Cu6Sn5 layer [16–22]. Yin and Borgesen [23] carried out the high temperature annealing experiment of Cu foil before the reflow process, eliminating the heterogeneous nucleation sites in the foil. The reaction interface using the heat-treated foil was almost void-free after subsequent soldering and thermal aging. Laurila et al. [22] reported that some alloying (such as Ni, Co and Zn) could dissolve in either of the Cu–Sn IMCs and retard the growth of Cu3Sn layer. Accordingly, the formation of voids was depressed. In addition, Wang et al. [24] suggested that the addition of Cu could increase the Cu concentration in solder, reduce the driving force for Cu diffusion and inhibit the formation of voids. In this work, the Sn/EPC and EPS/HPC joints were prepared to investigate the mechanism of void formation and the correlation between Cu3Sn phase and void. The effect of Cu on the void formation was also studied.
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
2. Experimental procedures The solders used in this study were commercial Sn (99.99%) and Sn0.7Cu (wt.%). Two types of Cu substrates were introduced, high purity foil (99.99%, 0.1 mm in thickness) and electroplated Cu film (approximately 10 lm in thickness) deposited directly on the surface of foil. The electroplating solution contained H2SO4 + CuSO4 + Cl + PEG (ethylene glycol) [13]. Both kinds of foils then were cut into 10 mm 10 mm square pieces. The solder joints were prepared by melting solders on the square pieces at 260 °C. Prior to the reflow process, these pieces were deoxidized and degreased in 5 wt.% NaOH and 5 vol.% HCl solutions sequentially, rinsed in de-ionized water after each step, and treated with flux. To investigate the interfacial microstructure in the solder joints, the isothermal aging for the as-reflowed samples was performed at 150 °C and 180 °C respectively. Then the samples were mounted in epoxy and metallurgically polished. The interesting zones at the interface were observed by using a scanning electron microscope (SEM).
2315
In addition, there were numbers of cracks existed within the Cu6Sn5 layer, which may be produced during the preparation of samples. 3.2. Effect of electroplated Sn on the formation of void In the case of Sn/EPC joints, the voids were mainly located in the Cu3Sn layer near the EPC, as seen in Fig. 3. If a layer of Sn is electroplated on the HPC, what would happen at the reaction interface after aging? Would the voids appear within the Cu6Sn5 layer? In order to verify this assumption, the corresponding experiments were carried out. A layer of Sn (approximately 10–15 lm
3. Results and discussion 3.1. Effect of UBM on the formation of void The cross-sectional microstructure of as-reflowed Sn/HPC and Sn/EPC joints are shown in Fig. 1. In both kinds of joints, a layer of scallop-shaped Cu6Sn5 formed at the interface without void. A very thin layer of another phase (possibly Cu3Sn) was also observed at the interface of Cu6Sn5/Cu [2]. After aging at 150 °C for 720 h, two thick IMC layers, Cu6Sn5 and Cu3Sn, were developed in the Sn/HPC joints, and the reaction interface remained void-free (Fig. 2a). With the same aging period, there were still no voids after raising the aging temperature up to 180 °C (Fig. 2b). In the Sn/EPC joints, a few voids were scattered within the Cu3Sn layer after aging at 150 °C for 240 h (Fig. 3a), and these voids covered around 1.6% of the area of the Cu3Sn phase. Extending the aging period to 480 h, numerous voids appeared (Fig. 3b), covering approximately 6% of the area of the Cu3Sn layer.
Fig. 1. Back scattered electron (BSE) images of as-reflowed solder joints. (a) Sn/HPC joint; (b) Sn/EPC joint.
Fig. 2. BSE images of Sn/HPC joints. (a) 150 °C, 720 h; (b) 180 °C, 720 h.
Fig. 3. BSE images of Sn/EPC joints after aging at 150 °C. (a) 240 h; (b) 480 h.
2316
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
Fig. 4. BSE images of EPS/HPC joints after aging at 150 °C. (a) 0 h; (b) 240 h; (c) 720 h; (d) 720 h.
in Fig. 6a. Prolonging the aging period to 480 h, the number and size of the voids had no obvious change though the Cu3Sn layer became much thicker (Fig. 6b). Thus, the area ratio of voids to Cu3Sn became lower, around 2%. 3.4. Discussion
Fig. 5. BSE images of as-reflowed Sn0.7Cu/EPC joint.
in thickness) was deposited on the surface of HPC using a commercially available plating solution. Then the samples were cut into 10 mm 10 mm square pieces and then aged at 150 °C. The preparation and characterization for the aged samples were the same as those in the second section (Experimental procedures). It was interesting that the Cu6Sn5 and Cu3Sn layers had been formed at the reaction interface of the as-electroplated sample, as shown in Fig. 4a. The IMC layers may grow during the electroplating process and subsequent storage at room temperature. Besides, the lost parts of the film (in the circle, Fig. 4a) were also produced during the preparation of samples. After aging at 150 °C for 240 h, both Cu6Sn5 and Cu3Sn layers became thicker, while Sn film became thinner (Fig. 4b). A few voids were observed not only at the Cu3Sn/Cu interface, but also at the Cu6Sn5/Cu3Sn interface. Prolonging the aging period to 720 h, more voids were formed at the Cu6Sn5/Cu3Sn interface and only a small amount of Sn was left (Fig. 4c and d). 3.3. Effect of Cu on the formation of void Similar to the as-reflowed Sn/Cu joints, a thin layer of Cu6Sn5 appeared at the interface without voids, as seen in Fig. 5. After aging at 150 °C for 240 h, a few voids can be observed, which covered approximately 4.6% of the area of Cu3Sn layer, as shown
During the isothermal aging, the formation of intermetallic compounds (IMCs) will cause a significant volume contraction, which can be estimated by the structural data [25–28]. A 9.97% reduction of volume will happen when Cu reacts with Sn forming Cu3Sn crystal, and a reduction of 5.44% in volume when forming g0 –Cu6Sn5 crystal (low temperature phase). The change of volume may result in the generation of inner stress in the solder joints [29]. The volume contraction, Kirkendall effect (KE) and residual stress generated during soldering were common in the solder joints. They were all potential factors of void formation at the reaction interface. However, the fact that voids appeared in the joints using EPC, while not in the joints using HPC [4,8,9,19,30,31], implied that the factors above might be not enough to induce the formation of voids, and it should be the Cu substrate that was responsible for this. In the process of electroplating, a number of impurities and defects may be introduced into the deposited Cu film. Lee and Park [32] reported that the organic impurity, poly ethylene glycol (PEG), was absorbed at the grain boundary of EPC film. Besides, some impurity elements S, Cl, C and O, were detected in the electroplated film by using SIMS (secondary ion mass spectrometry) analysis [9,13,23]. Contradictory viewpoints on the mechanism of void formation were put forward according to the different impurities. Kim and Yu [9] suggested that the segregation of S to the Cu/Cu3Sn interface would lower the free energy barrier for void nucleation and accelerate the process. However, Yin and Borgesen [23] considered that the void formation was associated with a complex of Cu+–PEG–Cl, but not S. The vacancy sinks at the interface may be pinned down by the organic complexes, and excess vacancies were left [33]. The locally high stress originated from the complexes would also accelerate the diffusion of Cu atoms around them (Cu vacancies were
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
2317
left) [32]. Moreover, these complexes could act as the heterogeneous sites and lower the nucleation barrier for the voids. Based on the discussion above, two necessary conditions for the nucleation of voids should be satisfied: (1) The unbalanced diffusion. In the Sn/Cu couple annealed at 150 °C, Sn was the slightly fast diffuser in the Cu6Sn5 layer, while Cu was the main diffusing species in the Cu3Sn layer [34]. (2) Vacancy supersaturation, that is
E > Ec
ð1Þ
where Ec is the critical energy for void nucleation and E is the energy possessed by the active area for voiding. The Ec at the Cu3Sn/ EPC interface was decreased by the introduced impurities. During the thermal aging, the Cu3Sn/EPC interface shifted towards the substrate gradually through the consumption of Cu film. The impurities contained in the film would get into the reaction interface. The Cu3Sn/EPC interface was the most active area for the voiding, which also served as the heterogeneous sites. With the further migration of Cu3Sn/EPC interface toward EPC, the voids originally formed were left behind, trapping inside the Cu3Sn layer. That may be the reason why most of the voids located in the Cu3Sn layer close to the EPC. Similar to the Cu plating solution, the solder plating solution also contained some usual additives (like PEG) which might be incorporated into the electroplated film [35,36]. However, after thermal aging for the EPS/HPC joints, the voids were formed at the Cu6Sn5/Cu3Sn and Cu3Sn/Cu interfaces, while not within the Cu6Sn5 layer though the Cu6Sn5 layer had the similar conditions with Cu3Sn layer in the Sn/EPC joint (close to the electroplated film). This indicated that the voids were inclined to form and grow in the Cu3Sn layer, while not in the Cu6Sn5 layer, which may be related to the microstructure of the two Cu–Sn IMCs. The impurities may be introduced into the Cu6Sn5 layer through the consumption of EPS film, and then into the Cu6Sn5/Cu3Sn interface during the growth of Cu3Sn toward the Cu6Sn5 layer. The joints with EPS film were different from the Sn/HPC joints which had a large amount of Sn. The possible reactions at the Cu6Sn5/Cu3Sn interface are
Fig. 6. BSE images of Sn0.7Cu /EPC joints after aging at 150 °C. (a) 240 h; (b) 480 h.
for the Sn/EPC joints, 0.792 (240 h) and 1.229 (480 h) respectively. The Cu3Sn layer in the Sn0.7Cu/EPC joints was thinner than that in the Sn/EPC joints, which was in agreement with the results obtained by Lee et al. [37]. Accordingly, the formation of voids was depressed owing to the correlation between Cu3Sn phase and void. 4. Conclusions
Cu6 Sn5 þ 9½Cu ! 5Cu3 Sn
ð2Þ
In this work, the void formation in the Sn/Cu joints was studied, taking EPC, EPS and Cu addition into consideration. After thermal aging, the voids were observed at the reaction interface in the Sn/EPC and EPS/HPC joints while not in the Sn/HPC joints. The voids were highly correlated with the Cu3Sn phase. These voids may be mainly caused by the impurities incorporated during electroplating. The addition of Cu element (0.7 wt.%) retarded the growth of Cu3Sn layer and depressed the formation of voids.
Cu6 Sn5 ! 3½Sn þ 2Cu3 Sn
ð3Þ
Acknowledgements
where [Cu] and [Sn] are diffusing atoms. With the extension of aging period, four different phases at the interface (Sn, Cu6Sn5, Cu3Sn and Cu) would evolve into three (Cu6Sn5, Cu3Sn and Cu) owing to the limited amount of Sn, as seen in Fig. 4c. It can be noted in Eqs. (2) and (3) that there was still enough driving force for [Cu] atoms to move from Cu3Sn layer to Cu6Sn5 layer for the formation of Cu3Sn needed a large number of [Cu] atoms. However, the supply of [Sn] atoms was inadequate. The unbalanced diffusion and low Ec caused by the impurities may induce the formation of voids at the Cu6Sn5/Cu3Sn interface. In addition, the large voids at the EPS/Cu6Sn5 interface may be caused by the volume contraction, as shown in Fig. 4c (in the circle). The addition of Cu would reduce the diffusion of Cu and Sn, especially for Cu [24]. The degree of the unbalanced diffusion between Cu and Sn was also decreased. Therefore, the formation of voids was suppressed. On the other hand, the Cu addition had a significant effect on the interfacial morphology. The total thickness of IMC layer in the Sn0.7Cu/EPC joints was close to that in the Sn/ EPC joints, as observed in Figs. 3 and 6. However, the thickness ratios of Cu3Sn to Cu6Sn5 for the Sn0.7Cu/EPC joints, 0.342 (aging for 240 h) and 0.716 (480 h) respectively, were much lower than those
This work was supported by the National Natural Science Foundation of China (Grant Nos. 50975176 and 50475021). The authors would like to thank Prof. C. R. Kao at National Taiwan University for his kind help (corrosive agents, BSE). Reference [1] Liu CY, Tu KN, Sheng TT, Tung CH, Frear DR, Elenius P. J Appl Phys 2000;87:750–4. [2] Zeng K, Stierman R, Chiu T-C, Edwards D, Ano K, Tu KN. J Appl Phys 2005;97:024508. [3] Kim D, Chang J, Park J, Pak JJ. J Mater Sci Mater Electron 2000. doi: 10.1007/ s10854-011-0357-2. [4] Yang W, Messler RW, Felton LE. J Electron Mater 1994;23:765–72. [5] Moriyama M, Konishi S, Tsukimoto S, Murakami M. Mater Trans, JIM 2004;45:3172–6. [6] Stangl M, Dittel V, Acker J, Hoffmann V, Gruner W, Strehle S, et al. Appl Surf Sci 2005;252:158–61. [7] Uedono A, Mori K, Ito K, Imamizu K, Hachiya T, Kamijo H, et al. Jpn J Appl Phys 2007;46:L483–5. [8] Laurila T, Vuorinen V, Kivilahti JK. Mater Sci Eng 2005;R49:1–60. [9] Yu J, Kim JY. Acta Mater 2008;56:5514–23. [10] Kim JY, Yu J. Appl Phys Lett 2008;92:092109. [11] Kim JY, Yu J, Kim SH. Acta Mater 2009;57:5001–12.
2318
Y. Yang et al. / Microelectronics Reliability 51 (2011) 2314–2318
[12] Kim S, Yu J. J Appl Phys 2010;108:083532. [13] Liu Y, Wang J, Yin L, Kondos P, Parks C, Borgesen P, et al. J Appl Electrochem 2008;38:1695–705. [14] Wafula F, Liu Y, Yin L, Bliznakov S, Borgesen P, Cotts EJ, et al. J Electrochem Soc 2010;157:D111–8. [15] Cogan S, Holmes DS, Rose RM. Appl Phys Lett 1979;35:557–9. [16] Tsai JY, Hu YC, Tsai CM, Kao CR. J Electron Mater 2003;32:1203–8. [17] Anderson IE, Harringa JL. J Electron Mater 2006;35:94–106. [18] Ho CE, Yang SC, Kao CR. J Mater Sci Mater Electron 2007;18:155–74. [19] Cho MG, Kang SK, Shih DY, Lee HM. J Electron Mater 2007;36:1501–9. [20] Gao F, Nishikawa H, Takemoto T. J Electron Mater 2008;37:45–50. [21] Wang YW, Lin YW, Kao CR. Microelectron Reliab 2009;49:248–52. [22] Laurila T, Vuorinen V, Paulasto-Kröckel M. Mater Sci Eng 2010;R68:1–38. [23] Yin L, Borgesen P. J Mater Res 2011;26:455–66. [24] Wang YW, Lin YW, Kao CR. J. Alloys Compd. 2010;493:233–9. [25] Suh I-K, Ohta H, Waseda Y. J Mater Sci 1988;23:757–60.
[26] Deshpande VT, Sirdeshmukh DB. Acta Crystallogr 1962;15:294–5. [27] Burkhardt W, Schubert K. Z Metallkd 1959;50:442–52. [28] Larsson AK, Stenberg L, Lidin S. Acta Crystallogr, Sect B: Struct Sci 1994;50:636–43. [29] Belyakov S, Atkinson HV, Gill SPA. J Electron Mater 2010;39:1295–7. [30] Vianco PT, Erickson KL, Hopkins PL. J Electron Mater 1994;23:721–7. [31] Deng X, Piotrowski G, Williams JJ, Chawla N. J Electron Mater 2003;32:1403–13. [32] Lee C-H, Park C-O. Jpn J Appl Phys 2003;42:4484–8. [33] Liu PL, Shang JK. Scripta Mater 2005;53:631–4. [34] Paul A, Ghosh C, Boettinger WJ. Metall Mater Trans A 2011;42:952–63. [35] Homma T, Sato H, Kobayashi H, Arakawa T, Kudo H, Osaka T, et al. J Electroanal Chem 2003;559:143–8. [36] Tsai YD, Hu CC. J Electrochem Soc 2009;156:D490–6. [37] Lee TY, Choi WJ, Tu KN, Jang JW, Kuo SM, Lin JK, et al. J Mater Res 2002;17:291–301.
