Effects of Impurity on Void Formation at Interface between Sn-3.0Ag-0.5Cu and Cu Electroplated Film


 Void formation is a critical reliability concern for solder joints in electronic packaging. The control of microstructures and impurity quantities in Cu electroplated films significantly affects the void formation at the joint interface, but the studies for their comparison are seldom. In this study, three Cu films (termed as A, B, and C) are fabricated using an electroplating process. The Cu A film has a facted grain texture embedded with twins while Cu B and C have a similar columnar texture. After thermal aging at 200°C for 1000 h, the SAC 305 (Sn-3.0Ag-0.5Cu) solder joints with Cu A and B exhibit a robust interfacial structure without voids. However, microstructural collapse is observed in the solder joint of SAC 305/Cu C, where many crevives parallel to the interface are formed. Based on the microanalysis, the concentration of impurity is higher in the Cu C film than those in Cu A and B. Moreover, discrete voids rather than continuous crevices are presented in the SAC305/Cu C system when the impurity concentration in Cu C is reduced. The findings demonstrate that the impurity control in the Cu electroplated film is critical for the control of void/crevice formation in the electronic solder joints.


Introduction
Electroplating copper (Cu) is an important technology to fabricate the metallizations and conducting traces for signal transmission in electronic devices. One important application is the Sn-rich solder/Cu interconnection in advanced electronic packaging [1,2]. Functional organic compounds as additives in the plating solutions play a vital role to control the Cu microstructure and property for speci c purposes.
However, due to strong adsorbability to the Cu electrodeposits the organic additives inevitably co-deposit with the reduced Cu atoms to some extent, leading to the impurity incorporation in the Cu electrodeposits.
Excessive incorporation of impure species affects the atomic arrangements and leads to the formation of defects in the Cu electroplated lm. Vacancy is one of the most critical defects that can signi cantly change the material property. For example, vacancy accumulation to form voids at the interface between the Sn-rich solder and Cu lm during thermal aging decreases the soldering strength [3,4].
Some studies have been dedicated to decreasing the level of the impurity incorporation in electroplated Cu to suppress the void formation through the optimization of additive formulation [1,4,5]. On the other hand, plenty of studies presented the void formation at the interfaces of solder joints could be suppressed by constructing the Cu microstructures with coherent twin boundary or larger grain size [6][7][8]. These speci c microstructures enabled to block the paths for vacancy accumulation, making voids hardly to nucleate. Although the control of impurity quantity and Cu microstructure can both suppress the void formation at the Sn-rich solder/Cu joints, the studies that take into consideration both factors are rarely seen. Therefore, which factor dominating the void formation is still unclear. In this study, three different Cu lms (termed as Cu A, B, and C) were prepared using electroplating, and their impurity quantity and grain microstructure were adjusted by ne-tuning the formulation of organic additives. The Cu A lm possessed a similar impurity concentration with Cu B but a different microstructure from Cu B.
The Cu C lm possessed an impurity concentration much higher than Cu A and B, but its microstructure is similar to that of Cu B. The results of this study revealed that the impurity concentration played a more signi cant effect on the void formation at the interface between Sn-3.0Ag-0.5Cu (SAC305) solder and electroplated Cu.

Experimental Procedures
Si of 1 cm × 1 cm, with a sputtered Ta layer of 100-nm thick and a sputtered Cu seed layer of 200-nm thick, was used as the substrate for the deposition of Cu electroplated lm. An electroplating bath of Cu was constructed using the Ta/Cu-coated Si substrate (cathode), a Cu plate with 0.04 wt.% P (anode), and an electrolyte with high-purity CuSO 4 , H 2 SO 4(aq) of 5 vol.%, and proper H 2 O 2(aq) , as shown in Fig. 1.
Additionally, different concentrations of an organic additive were added into the plating solution as the key to control the impurity concentration and grain microstructure in the Cu electroplated lms. The current density was set as 40 mA/cm 2 for all additive formulation using a power supply (Autolab PGSTAT302N, Netherlands) for 30 min. In this study, three different Cu lms were fabricated using the above-mentioned electroplating parameters and the main difference was the additive concentration.
Mechanical stirring of 1000 r.p.m was required for uniform deposition of Cu.
Microstructural analysis of each Cu electroplated lm was performed by electron backscattered diffraction (EBSD, Oxford, UK), equipping a scanning electron microscope (SEM, JEOL JSM-7800F, Japan), and focus ion beam (FIB, JEOL JIB-4601F, Japan) for grain orientations and cross-sectional ionbeam images with clear contrasts of the grain and twin boundary, respectively. The identi cation of each impurity (such as carbon, oxygen, and sulfur) in the Cu lms and their respective intensities were carried out using a secondary ion mass spectrometer (SIMS, ims 4f, CAMECA, France). Then, the Cu lms went through a soldering process at 250°C with an Sn-3.0 wt.% Ag-0.5 wt.% Cu solder (SAC305, Senju Metal Industry Co., Japan). The SAC305 solder was 22 mg and re owed upon a soldering area of 1.8-mm diameter de ned by a heat resistant tape on each Cu lm, conducting on a hotplate for 30 s (Fig. 1).
Solid-state aging in a furnace at 200°C for 1000 h was performed for the SAC305/Cu solder joints to induce the void formation. Then, the cross-sectional microstructures of the SAC305/Cu interfaces were observed using SEM after grinding and polishing the samples. Energy-dispersive X-ray spectrometer (EDX) equipped to SEM was used to determine the compositions of the reaction products.  Table 1. Hence, the grain orientations of all Cu lms in ND can all be regarded as highly < 110>-oriented, as shown in their OIMs and IPFs-DI.  [11]. On the other hand, "Kirkendall effect" is a theory to explain the vacancy diffusion through an interface between two dissimilar solid materials, and "Kirkendall voids" are formed at one side material with a higher diffusivity. This is the reason that void formation frequently appeared at the interface between Sn-based solder and Cu because the diffusion of Cu into Sn is much faster than that of Sn into Cu [3,6,12].

Results And Discussion
Consequently, the control of the Cu surface orientation is very important to investigate the formation of Kirkendall voids at the SAC305/Cu interface. Because the three Cu electroplated lms were all {110}preferred on their surface, the effect of surface diffusivity on the void formation can be ignored in this study. In other words, the effects of microstructure and impurity concentration on the void formation can be sure compared. Figure 3 shows the cross-sectional FIB images of the Cu A, B, and C lms. The grains in Cu A were embedded with a few twin boundaries (Fig. 3a). In Cu B and C, their grains were columnar-shaped as shown in Figs. 3b and c. Therefore, the microstructure of Cu B was similar to that of Cu C, but they were dissimilar to that of Cu A without the columnar grains. Some articles emphasized that the microstructures of the Cu lms signi cantly induced or suppressed the Kirkendall effect [7,8,12]. Among them, the smaller grain size in a Cu lm provided numerous grain boundaries for Cu fast diffusion into the Sn-rich solder [7,13], and the grain boundaries between columnar grains were the paths for vacancy aggregation on the Cu surface [8]. They enhanced the Kirkendall effect at the Sn-rich solder/Cu interface unless Cu nanotwin existed in the Cu lms [8,12]. According to the above-mentioned inference, although the Cu A lm possessed twin boundary, but the number of twins was much lower than that previously reported capable of suppressing the void formation. The Cu B and C lms were with silimar columnar microstructure that enhanced the formation of Kirkendall voids. Therefore, from the microstructural pointof-view the three Cu lms all faced high risks of the void formation. Cu. The Ag 3 Sn particles embedded at the SAC305/Cu 6 Sn 5 interface were precipitates from the SAC305 solder matrix. At the SAC305/Cu A interface (Fig. 4a), a void-free interlayer was observed after the thermal aging, which was similar to that at the SAC305/Cu B interface (Fig. 4b), although the microstructural difference between Cu A and B was signi cant (Fig. 3a versus 3b). The Cu B lm with columnar grains provided several channels for the diffusion of vacancies from the Cu interior toward the lm surface. However, voids were not observed at the SAC305/Cu B interface. Conversely, at the SAC305/Cu C interface where the Cu C lm possessed a similar microstructure with Cu B, plenty of voids were formed after thermal aging for 1000 h. Kirkendall effect was clearly observed in the SAC305/Cu C system. Moreover, these voids even propagated across the entire SAC305/Cu interface and connected in series to form many crevices (or cracks) parallel to the interface, forming a IMC/crevice alternating structure. Based on the above examinations, it was concluded that the Cu microstructure is not the only essential factor to induce the formation of Kirkendall voids at the interface between SAC305 and Cu.
On the other hand, some impure species originating from the organic substances in the plating solution usually co-deposit in a Cu electroplated lm and segregate in the grain boundaries. An increase in the atomic disarrangement due to the impurity incorporation gives rise to plenty of vacancies in the Cu lm, and the void formation is induced by the segregation of vacancies during thermal aging. Therefore, the impurities are considered as another possible source of producing the voids in the Cu lms [13]. Herein, the incorporation of impurities such as oxygen (O), sulfur (S), and carbon (C) in the Cu A, B, and C lms was examined using SIMS as shown in Fig. 5. Obviously, the incorporation levels of O, S, and C were in the same order of Cu C > Cu B ≈ Cu A. In Fig. 4, the SAC305/Cu A and SAC305/Cu B solder joints were void-free. Although the microstructure of Cu B resembled that of Cu C, serious void propagation occurred in the interlayer between SAC305 and Cu C during thermal aging. The results demonstrated that, rather than the microstructures of the Cu lms, the impurity level in the Cu lms were highly related to the void formation in the SAC305/Cu solder joints thermally aged at 200 °C. The impurities incorporated in the Cu lms played the roles to produce plenty of vacancies nucleating at the SAC305/Cu C interface. Segregation of the impure species or their derivatives in the grain boundaries might also annihilate the vacancy sinking sites which accelerated the accumulation of vacancies to form voids [14][15][16]. Severe void propagation gave rise to the formation of continuous crevices or cracks which was a kind of volumetric defects in crystal caused by void nucleation during annealing [17,18].
To further investigate the impurity effect, a new Cu electroplated lm (termed as Cu C*) was prepared by reducing the additive concentration. The Cu C* lm possessed a microstructure similar to that of Cu C but a lower impurity concentration as shown in the SIMS patterns in Fig. 5. Figure 6 shows the crosssectional SEM image of the SAC305/Cu C* interface after thermal aging at 200 °C for 1000 h. Discrete voids were found but these voids had not propagated and aggregated together to form continuous crevices as those in the SAC305/Cu C system shown in Fig. 4(c) and 4(d). Such different void evolution can be attributed to the level of impurity incorporated in the Cu lms. As shown in Fig. 5, the impurity level in the Cu C* lm was lower than that of Cu C, so the formation rate of voids caused by the oversaturation of vacancies was lower in the Cu C* lm than in Cu C. Therefore, voids were formed discretely instead of continuous crevices in the SAC305/Cu C* system. In other words, the impurity concentration in the Cu electroplated lms plays a crucial role to dominate the void evolution in the form of discrete voids or continuous crevices caused by severe void propagation.

Conclusions
Void formation at the interface between solder and Cu is critical for the reliability of solder joints. In this study, the Cu A, Cu B, and Cu C electroplated lms all possessed {110}-preferred orientations, identi ed by EBSD OIM and IPFs-DI, thus the surface diffusion of these Cu lms was similar to each other. The microstructural observation of FIB image shows that the Cu B lm consisted of numerous columnar grains, which was similar to that of Cu C but was different from that of Cu A with faceted grains embedded with few twin boundaries. After thermal aging at 200 °C for 1000 h, there was no void formation in both the SAC305/Cu A and SAC305/Cu B solder joints although their microstructures were different from each other. Conversely, void formation was observed in the interlayer of the SAC305/Cu C system with a Cu microstructure similar to the Cu B lm after thermal aging. Moreover, these voids even propagated inside the IMC layers and connected in series to form crevices (or cracks) periodically inside the IMC layers. Based on the microanalysis of SIMS, the void formation at the solder/Cu interface was rationally explained by the level of impurities incorporated in the Cu electroplated lms (Cu C > Cu B ≈ Cu A), rather than their grain microstructures. These results demonstrated the signi cant effect of impurity on the Kirkendall effect occurring at the interface between SAC305 and Cu lm. The ndings also suggest that the impurity control of the Cu electroplated lm is important for the void suppression in the solder joints of electronic packaging.

Con ict of Interest
The authors declare that they have no con ict of interest.