Microstructure and Thermal Cycle Reliability of Sn–Ag–Cu–In–Sb Solder Joint

Abstract

In this study, the microstructure and the thermal fatigue life of Sn–3.0 mass%Ag–0.5 mass%Cu–6.0 mass%In–1.0 mass%Sb (SAC305–6In–1Sb) were investigated and they were compared with those of Sn–3.0 mass%Ag–0.5 mass%Cu (SAC305). As the test pieces, chip resistors were joined on a printed wiring board by reflow soldering with each solder. Although precipitates of Ag and Cu were present in both solder joints, the precipitation state and size of Ag precipitates were different for each solder joint. In the solder joint with SAC305–6In–1Sb, the precipitate of In and Sb was also confirmed. In addition, In and Sb were solid-soluted in Sn. Solder joints with SAC305–6In–1Sb were polycrystal with approximately 20 crystal grains per cross section of the joint. On the other hand, the joint of SAC305 was consisted of a single crystal. Average of thermal fracture life of SAC305–6In–1Sb was approximately 4.3 times longer than that of SAC305. In addition, the variation in the life of SAC305–6In–1Sb was smaller.

1. Introduction

Development of autonomous driving and advanced driver assistance systems for automobiles has been active. As these systems are installed in a vehicle, number of electronic control unit (ECU) installed in the vehicle increases.¹⁾ ECU is required to be downsized due to the securing of the passenger compartment space and the limitation of the installation position. ECU is composed of printed wiring boards and various electronic components, and solder is used to join them. In addition to adopting small electronic components, reduction of solder fillets is also being considered so as to realize miniaturization of ECU.^2,3) One of the problems associated with reducing solder fillets is the decrease in the bonding life.^4,5) In Japan, lead-free solder Sn–3.0 mass%Ag–0.5 mass%Cu (SAC305) recommended by the Japan Electronics and Information Technology Industries Association has been widely used as a bonding material for electronic components.^6,7) Lately, studies of lead-free solders which have better reliability than SAC305 by precipitation strengthening and solid solution have been carried out.^8,9) It has been reported that Sn-based solders have variations in mechanical properties due to the anisotropy of Sn.^10–12) Since current solder joints are very fine, the solder joints of a ball grid array joint (BGA) with SAC305 have only one crystal grain.¹⁰⁾ Therefore, the anisotropy of Sn may affect the variation of the solder joint life. If the life of the solder joint varies widely, it becomes difficult to guarantee the reliability of electronically controlled products and to design them. It has been reported that Sn–Ag–Cu–In–Sb solder forms polycrystals depending on the composition, and among them, Sn–3.0 mass%Ag–0.5 mass%Cu–6.0 mass%In–1.0 mass%Sb (SAC305–6In–1Sb) is stably polycrystallized.^9,13) The tensile strength of SAC305–6In–1Sb is higher than that of SAC305, and the variation in tensile strength is reduced.¹³⁾ Therefore, SAC305–6In–1Sb may not only improve the joining life of the solder joint, but also reduce the variation in the joining life. Hence, in this study, the thermal fatigue life and the variation in it of SAC305–6In–1Sb was investigated by the thermal cycle test.

2. Experimental Procedure

2.1 Test piece

Figure 1 shows the appearance of the test piece used for the evaluation. The test piece had 10 chip resistors soldered on a printed wiring board. The size of the printed wiring board was 80 mm width, 80 mm length and 1.6 mm thickness. The base material was FR-4 (linear expansion coefficient: 14.1 ppm/°C). Size of the chip resistor was 1.6 mm width and 3.2 mm length. The chip resistor used in this study was model number RK73B2BTTD-202J manufactured by KOA CORPORATION. Two types of solder shown in Table 1 were used for soldering. SAC305 was used for comparison. On the test pieces, 10 chip resistors and a set of pads for the measurement of resistance are connected in series. Therefore, the resistance of the whole circuit can be acquired by measuring resistance using the pads.

Fig. 1

Appearance of test piece used for evaluation.

Table 1 Chemical compositions of solder investigated in this study.

2.2 Preparation of test pieces

Figure 2 shows the appearance of the printed wiring board and Cu pad before the chip resistors are soldered. Solder paste consisting of solder balls with an average particle size of φ20 µm and flux was applied to the Cu pad on the printed wiring board by screen printing. Then, the resistors were placed on the top of the solder paste by a mounter. Finally, they were soldered by heating and melting the solder in a reflow oven replaced with nitrogen. Regardless of the type of solder, they were soldered with the heating profile shown in Fig. 3.

Fig. 2

Appearance of printed wiring board and Cu pad.

Fig. 3

Heating conditions for soldering.

2.3 Microstructure and crystal orientation observation

Microstructural observations and element mapping of the solder joints were carried out by the following methods. Firstly, a portion to be observed was cut out from the test piece and it was encased in room temperature curing type epoxy resin. Secondly, it was polished by # 500, # 800, # 1200, # 2400 and # 4000 SiC abrasive papers until the cross section of the test piece shown in Fig. 4 was exposed. It was buffed by alumina abrasive with 0.5 µm particle size, and OPS (SiO₂ of 0.04 µm in particle size). Finally, the specimens were processed by a cross section polisher. After these processes, the microstructures were observed by scanning electron microscopy (SEM) and element mapping was carried out with an electron probe micro analyzer (EPMA). The grain map analysis and crystal orientation observation were performed by an electron backscatter diffraction (EBSD) system. In grain map analysis in this study, crystal grains whose orientation differs by more than 15° from adjacent crystal grains, namely, with high-angle grain boundaries, were defined as different crystals. In order to investigate the variation in the number of crystal grains and the crystal orientation, EBSD observation was performed on two specimens for each solder. The specimens used for EBSD observation was encased using conductive room temperature curing type epoxy resin.

Fig. 4

Observed cross section. (a) Cutting position, (b) Cross section image.

2.4 Thermal cycle test

Figure 5 shows the temperature conditions in the thermal cycle test conducted in this study.

Fig. 5

Temperature conditions of the thermal cycle test.

The test started at 25°C and the test piece was held at −40°C and 125°C for 30 minutes respectively, then was reached 25°C again in one cycle. This heat load was repeated until the end of life of the test piece. The life was defined as the number of cycles when the resistance value was doubled with respect to the initial value. The resistance value was continuously measured by the 4-terminal measurement method. The number of test pieces was 10 for each solder. Ten chip resistors were soldered to one board, but if even one solder joint of the chip resistor breaks, the resistance value doubled. Therefore, the number of cycles in which one solder joint was fractured was equal to the life of the test piece. The test piece whose resistance value was doubled was immediately taken out from the test chamber, and the fractured part was identified by observing the appearance and measuring the resistance value of each chip resistance by a multimeter.

3. Results and Discussions

3.1 Microstructures of joints

Figure 6 shows results of microstructure observation by SEM and element mapping by the EPMA. Intermetallic compound layer of Cu–Sn with thickness of approximately 1 to 2 µm was confirmed at the interface between the solder fillet and Cu pad in both solder joints. Precipitates of Ag and Cu were present in both SAC305 and SAC305–6In–1Sb. These are considered to be Ag₃Sn and Cu₆Sn₅ intermetallic compounds (IMC).^14,15) Although Ag precipitates were dispersed in SAC305–6In–1Sb, they were aggregated in SAC305. In SAC305, the majority of Ag precipitates are 1 µm or less length. In contrast, in SAC305–6In–1Sb, there are many deposits of approximately 3 µm. In addition, precipitates of In and Sb were present in SAC305–6In–1Sb. These are considered to be In–Sn and In–Ag IMCs.^9,16,17) In SAC305–6In–1Sb, In and a small amount of Sb were solid-soluted in Sn.

Fig. 6

Microstructures and element mapping analysis results for each solder joints.

3.2 Grain map and crystal orientation

Figure 7 shows the grain maps and the inverse pole figure (IPF) of each solder joint. From the grain maps and IPF maps, it was found that SAC305 was a single crystal and has different crystal orientation depending on the test piece. On the other hand, SAC305–6In–1Sb was polycrystal with approximately 20 crystal grains per cross section. It seems that those test pieces also polycrystallized by passing through the solid-liquid coexistence region where the β phase and the γ phase exist and solidifying as shown in the previous study.¹³⁾ The crystal orientation of SAC305–6In–1Sb differed depending on the test piece as well as SAC305.

Fig. 7

Grain maps and IPF maps of solder joints.

3.3 Thermal cycle test

Figure 8 shows the result of the thermal cycle test. The thermal fatigue life of SAC305 was 1183 to 2455 cycles, and that of SAC305–6In–1Sb was 7279 to 8372 cycles. The average thermal fatigue life of SAC305 was 1857 cycles, while that of SAC305–6In–1Sb was 8122 cycles, showing a 4.3-fold improvement. The reason seems to be that the material properties of SAC305–6In–1Sb were improved by precipitation strengthening and solid solution. Next, when comparing the range of fracture life, SAC305–6In–1Sb was approximately 200 cycles smaller than SAC305. Ratio of the range to the average was 68% for SAC305 and 13% for SAC305–6In–1Sb. In other words, the variation in the fracture life of SAC305–6In–1Sb was smaller than that of SAC305. This result is considered to be affected by the number of crystal grains. Since Sn, which is the base metal, shows anisotropy, it seems that SAC305, which is a single crystal, was affected by the anisotropy of Sn. On the other hand, since SAC305–6In–1Sb was a polycrystal, the anisotropy was canceled out and the lifetime variation seemed to be smaller than that of SAC305. However, since the life range of SAC305–6In–1Sb was approximately 1000 cycles, the cause was investigated using the test pieces No. 1 (life: 7297 cycles) and No. 9 (life: 8319 cycles) shown in Fig. 8. The result is described in the next section. Furthermore, the same investigation was conducted for the SAC305 test pieces No. 1 (life: 1183 cycles) and No. 10 (life: 2455 cycles) shown in Fig. 8. The result is described in section 3.5.

Fig. 8

Result of thermal cycle test.

3.4 Causes of variation in fracture life of SAC305–6In–1Sb

Firstly, the appearance of the test pieces were observed using an optical microscope. Figure 9 shows the observation result. The solder joints surrounded by the dotted line in Fig. 9 were fractured. There is no difference between the position of the fractured chip resistor on the test piece No. 1 and No. 9 since both chips were mounted on the second place from the outside. Next, the fracture surfaces of the solder joints were analyzed by SEM. Figure 10 shows secondary electron (SE) images of the fracture surfaces of the solder joints on the printed wiring board side after the chip resistor is peeled off. The part surrounded by the dotted line in Fig. 10 was soldered. It was found that cracks grew along the chip resistance in both test pieces, leading to breakage. In addition, from Fig. 10, traces of voids were present in both test pieces No. 1 and No. 9 was confirmed. It is probable that while the test piece No. 9 had three voids of 100 µm or more, the test piece No. 1 had one void of 100 µm or more. Larger void areas reduce the soldered area and can lead early breakage, but that does not coincide with the test results. Furthermore, there was a non-soldered area in the test piece No. 1. When the soldered area in the area surrounded by the dotted line in the test piece No. 9 was 100%, that in the test piece No. 1 was approximately 95%. Since this area ratio was different from the life ratio of those test pieces, it cannot be considered that the life is different only by the difference in the soldered area. Figure 11 shows the fracture morphology investigated by SEM. Since striations were observed in both test pieces No. 1 and No. 9, the fracture form of both test pieces was confirmed to be thermal fatigue fracture. However, the striation interval of test piece No. 9 was approximately 0.1 µm while that of test piece No. 1 was approximately 0.8 µm. This result suggests that the stress generated in each test piece was different, which may have caused a difference in life. From these results, it is considered that the variation in the life of SAC305–6In–1Sb is due to the difference in stress depending on the area of the solder joint and the effect of voids.

Fig. 9

Appearance of investigated test pieces of SAC305–6In–1Sb. (a) No. 1, (b) No. 9.

Fig. 10

Fracture surfaces of SAC305–6In–1Sb solder joints on the printed wiring board side. (a) No. 1, (b) No. 9.

Fig. 11

Investigation results of fracture morphology of SAC305–6In–1Sb. (a) No. 1, (b) No. 9.

3.5 Results of investigating fracture morphology of SAC305

Figure 12 shows the results of observing the appearance of the test piece with an optical microscope. The solder joints surrounded by the dotted line in Fig. 12 were fractured. In the test piece No. 1, there was a part where the solder was not wet. In other words, the short life test pieces of SAC305 and SAC305–6In–1Sb were low quality. There is no difference between the position of the fractured chip resistor on the test piece No. 1 and No. 10 since both chips were mounted on the second place from the outside. Figure 13 shows SE images of the fracture surfaces of the solder joints on the printed wiring board side. It was found that cracks grew along the chip resistance in both test pieces, leading to breakage. The path of these cracks is similar to the result of SAC305–6In–1Sb. In addition, from Fig. 13, traces of voids were present in both test pieces No. 1 and No. 10 was confirmed. It is probable that while the test piece No. 1 had two voids of 100 µm or less, the test piece No. 10 had five voids of 100 µm or less. In other words, it seems that the test piece No. 10 had more voids. Both SAC305 and SAC305–6In–1Sb tended to have more voids in the long-life test pieces. However, it seems that SAC305–6In–1Sb had a larger void. Figure 14 shows the fracture morphology investigated by SEM. Since striations were observed in both test pieces No. 1 and No. 10, the fracture form of both test pieces was confirmed to be thermal fatigue fracture. However, the striation interval of test piece No. 10 was approximately 1.3 µm while that of test piece No. 1 was approximately 1.9 µm. This result suggests that the stress generated in each test piece was different, which may have caused a difference in life. The cause of this difference seems to be the effects of anisotropy, solder non-wetting and voids. There was variation in the quality of the test pieces regardless of the solder type. From this, it seems that the main difference between SAC305 and SAC305–6In–1Sb is the crystal orientation. In other words, the main reason why the life variation of SAC305–6In–1Sb is smaller than that of SAC305 is probably because SAC305–6In–1Sb is polycrystalline.

Fig. 12

Appearance of investigated test pieces of SAC305. (a) No. 1, (b) No. 10.

Fig. 13

Fracture surfaces of SAC305 solder joints on the printed wiring board side. (a) No. 1, (b) No. 10.

Fig. 14

Investigation results of fracture morphology of SAC305. (a) No. 1, (b) No. 10.

4. Conclusions

In this study, using a test piece soldered with a chip resistor and a printed wiring board by SAC305–6In–1Sb, the microstructure and thermal fatigue life were investigated. In addition, the results were compared with those of SAC305.

The following conclusions were obtained.

1. (1)    Precipitates which are considered to be Ag₃Sn and Cu₆Sn₅, were confirmed at both solder joints. Although these precipitates containing was dispersed in SAC305–6In–1Sb, they were aggregated in SAC305.
2. (2)    Precipitates which are considered to be In–Sn and In–Ag IMCs were also present in SAC305–6In–1Sb. In addition, In and a small amount of Sb were solid-soluted in Sn.
3. (3)    The SAC305–6In–1Sb joint was polycrystal with approximately 20 crystal grains per cross section. On the other hand, the SAC305 joint was a single crystal, and it has different crystal orientation depending on the test piece.
4. (4)    Average fracture life of SAC305–6In–1Sb was approximately 4.3 times longer than that of SAC305. Moreover, the range of fracture life of SAC305–6In–1Sb was approximately 200 cycles smaller than that of SAC305.
5. (5)    There was variation in the quality of the test pieces regardless of the solder type. From this, it seems that the main difference between SAC305 and SAC305–6In–1Sb is the crystal orientation. In other words, the main reason why the life variation of SAC305–6In–1Sb is smaller than that of SAC305 is probably because SAC305–6In–1Sb is polycrystalline.

REFERENCES

• 1)  Y. Onuma, Y. Terashima, S. Nakamura and R. Kiyohara: 2018 International Conference on Information Networking (2018) pp. 298–303.
• 2)   P.  Zhang,  S.  Xue and  J.  Wang: Mater. Des. 192 (2020) 108726. doi:10.1016/j.matdes.2020.108726
• 3)   S.  Annuar,  R.  Mahmoodian,  M.  Hamdi and  K.-N.  Tu: Sci. Technol. Adv. Mater. 18 (2017) 693–703. doi:10.1080/14686996.2017.1364975
• 4)  C. Han and B. Song: 8th Electronics Packaging Technology Conference (2006) pp. 781–786.
• 5)   C.  Han and  B.  Han: J. Mech. Sci. Technol. 28 (2014) 879–886. doi:10.1007/s12206-013-1154-z
• 6)   Y.  Nomura,  S.  Minamoto and  S.  Nomoto: ISIJ Int. 50 (2010) 1920–1924. doi:10.2355/isijinternational.50.1920
• 7)   A.  Sharif and  Y.C.  Chan: Thin Solid Films 504 (2006) 431–435. doi:10.1016/j.tsf.2005.09.061
• 8)   S.A.  Belyakov,  J.  Xian,  G.  Zeng,  K.  Sweatman,  T.  Nishimura,  T.  Akaiwa and  C.M.  Gourlay: J. Mater. Sci. Mater. Electron. 30 (2019) 378–390. doi:10.1007/s10854-018-0302-8
• 9)   K.  Hine,  S.  Norimine,  K.  Mihara,  S.  Akiyama,  A.  Furusawa and  K.  Uenishi: Trans. Japan Inst. Electr. Packaging 10 (2017) E16-006-1–E16-006-10. doi:10.5104/jiepeng.10.E16-006-1
• 10)   Y.  Hirai,  K.  Oomori,  H.  Morofushi and  I.  Shohji: Mater. Trans. 60 (2019) 909–914. doi:10.2320/matertrans.MH201807
• 11)   J.G.  Lee,  A.  Telang,  K.N.  Subramanian and  T.R.  Bieler: J. Electron. Mater. 31 (2002) 1152–1159. doi:10.1007/s11664-002-0004-z
• 12)   Y.  Kariya,  S.  Tajima and  S.  Yamada: Mater. Trans. 53 (2012) 2067–2071. doi:10.2320/matertrans.MB201204
• 13)   Y.  Hirai,  K.  Oomori,  H.  Morofushi and  I.  Shohji: Mater. Sci. Forum 1016 (2021) 553–560. doi:10.4028/www.scientific.net/MSF.1016.553
• 14)   M.  Reid,  J.  Punch,  M.  Collins and  C.  Ryan: Solder. Surf. Mt. Technol. 20 (2008) 3–8. doi:10.1108/09540910810902651
• 15)   P.M.  Kumar,  G.  Gergely,  D.K.  Horváth and  Z.  Gácsi: Powder Metall. Progress 18 (2018) 49–57. doi:10.1515/pmp-2018-0006
• 16)   A.A.  El-Daly and  A.E.  Hammad: J. Alloy. Compd. 509 (2011) 8554–8560. doi:10.1016/j.jallcom.2011.05.119
• 17)   S.  Mosaad,  A.R.  Mohamed and  M.  Kamal: J. Adv. Phys. 12 (2016) 4231–4243. doi:10.24297/jap.v12i2.56