Solders serve as electrical and mechanical interconnects in electronic packaging. The need to develop environmentally benign electronic packages has generated great interest in Pb-free solder alloys.[1–6] The demand for smaller portable devices also drives a reduction in solder bump size and pitch. Pb-free solders are typically Sn rich, with small alloying additions of Ag and/or Cu. When solder interconnects are fabricated, the Sn-based alloy is melted between two substrates with metallization layers, such as Cu or Ni. From the reaction between Sn and Cu, a Cu6Sn5 intermetallic compound (IMC) layer is formed at the solder/Cu interfaces. The overall thickness and morphology of the IMC layer is known to have a significant influence on the mechanical behavior of solder joints under quasi-static[7] and mechanical shock conditions.[8–10] Indeed, Deng et al.[7] showed that not only does the IMC layer serve as a stress concentration site, but also this phenomenon can be exacerbated by the morphology of the IMC layer. Therefore, understanding the growth behavior and morphology of the IMC layer is critical to developing reliable electronic packages.

The IMC layer thickness and morphology is affected by multiple factors, including the cooling rate of the solder joint upon solidification, isothermal aging, and extended reflow.[7,11] During extended reflow, the solder joint is held above the melting temperature for extended periods of time, allowing the liquid solder to react with the substrate and form thick IMC layers. In this article, we report on a novel IMC layer growth behavior. Sn-3.9Ag-0.7Cu solder joints were held in extended reflow for up to 168 hours. The resulting IMC layers exhibited an asymmetry with respect to the thicknesses of the top and bottom interfaces. Characterization of this novel growth behavior and discussions of possible mechanisms follow.

Sn-3.9Ag-0.7Cu solder (Indium Corp., Utica, NY) was reflowed between two oxygen-free, high-conductivity copper bars (25 mm long and 6.35 mm diameter) to form butt joints. Copper bars were mechanically polished to a 0.05 μm colloidal silica finish. A mildly activated rosin flux was then applied to the end of the copper bars to promote wetting between the solder and Cu during reflow. A jig was used to create a reproducible joint thickness of ~500 μm. Temperatures during reflow and aging were measured using a thermocouple placed near the joint. The jig and solder joints were heated in a box furnace to a nominal reflow temperature of 503 K (230 °C). As-reflowed solder joints were held for 40 seconds above the melting temperature, removed from the oven, and air cooled on an aluminum block to obtain a cooling rate of about ~1 °C/s. The as-reflowed solder joints were an experimental control against which the changes in microstructure for extended reflow solder joints were compared. Extended reflow solder joints were held at the reflow temperature for 3, 12, 24, 96, and 168 hours, then removed from the oven, and cooled at ~1 °C/s. Solder joints were mounted in epoxy and cross sectioned. Solder joint cross sections were polished to a 0.05 μm colloidal silica finish. Scanning electron microscopy (SEM; FEI-XL30; FEI Corporation, Hillsboro, OR) was used to characterize the solder joint microstructure. Average intermetallic layer thickness was computed with at least 500 measurements from SEM micrographs. Energy-dispersive X-ray spectroscopy (EDS) was used to identify compositions of the solder constituents.

The microstructural evolution of the IMC layers was characterized using backscattered SEM (BSEM) with solder joints that were reflowed for various times. BSEM micrographs of solder joints reflowed for 0 hours (as-reflowed), 3 hours, 12 hours, 24 hours, 96 hours, and 168 hours are shown in Figures 1(a) through (f), respectively. The as-reflowed Cu6Sn5 IMC layer thicknesses at the top and bottom interfaces were both 2.5 ± 0.6 μm, as shown in Figure 1(a). The IMC layer morphology is nodular, due to a thermal grooving process that occurs in the liquid state.[12] At 3 hours extended reflow, the top IMC layer became slightly irregular, and a continuous Cu3Sn IMC layer was detected, as shown in Figure 1(b). At 12 hours, large gaps began to form between the nodules in the top IMC layer, and the Cu3Sn layer was no longer continuous, as Figure 1(c) shows. The interface between the top IMC layer and the Cu substrate also took on a wavy morphology. At 24 hours, the top IMC layer does not show a uniform thickness anymore, as Figure 1(d) shows. At 96 and 168 hours, the top IMC layer did not seem to growth any further, and Cu3Sn was found only in small patches beneath larger nodules. In contrast, the bottom IMC layer of all the samples continued to grow with a nodular morphology and remained continuous with respect to both Cu6Sn5 and Cu3Sn. It is clear that the IMC layers at the top and bottom interfaces did not grow at equal rates, which resulted in a significant asymmetry in growth.

Fig. 1
figure 1

Backscatter SEM micrographs showing growth behavior of IMC layer for: (a) as-reflowed solder joint, and joints reflowed for (b) 3 h, (c) 12 h, (d) 24 h, (e) 96 h, and (f) 168 h

Full size image

The asymmetric growth behavior was characterized by tracking the growth of the top and bottom IMC layers as a function of several reflows. The average thicknesses of the top and bottom IMC layers (Cu6Sn5 + Cu3Sn) were quantified and are shown in Figure 2(a). It was observed that while the bottom IMC layer grows in a typical t 1/2 manner, the top layer only behaves this way for reflow times under 3 hours. If growth is controlled by bulk diffusion in the IMC layer, then it goes as ~t 1/2, whereas if it is controlled by grain boundary diffusion, then it goes as ~t 1/3.[11–13] After 3 hours, the top IMC layer growth begins to slow down and eventually reaches a plateau value. The value at which the IMC layer thickness begins to plateau to may be considered a critical IMC thickness. Asymmetric IMC layer thickness could degrade solder joint mechanical reliability since it would cause nonuniform deformation in the solder joints.

Fig. 2
figure 2

Quantification of growth behavior at top and bottom interfaces, for (a) the total IMC layer, (b) Cu6Sn5 IMC, and (c) Cu3Sn IMC

Full size image

Before a solution for mitigating the asymmetric growth behavior could be found, the driving forces of the phenomenon had to be determined. Initially, it was thought that the mechanism causing asymmetric growth was gravity driven since there were no other forces acting on the solder joints. Thermocouples were placed near the top and bottom of the joint during extended reflow, and these confirmed that no significant temperature gradient was present either. One proposed mechanism was sedimentation of Cu solute in the liquid Sn during extended reflow. To test this hypothesis, EDS line scans were made across the thickness of as reflowed solder joints, and joints reflowed for 3, 12, 24, 96, and 168 hours. Multiple reflow durations were chosen so that the time-dependent dynamics of the sedimentation process might be captured. Figure 3 shows the EDS line scan results for Sn (Figure 3(a)), Cu (Figure 3(b)), and Ag (Figure 3(c)). The line scans showed that, while there were slight variations in wt pct, there were no concentration gradients present for any of the solder alloy constituents. The slight variations are most likely due to the solder microstructure, which is composed of Sn-rich dendrites and a eutectic mixture of Ag3Sn and Cu6Sn5. Therefore, it was concluded that a solute sedimentation process was not the cause of the asymmetric IMC layer growth.

Fig. 3
figure 3

EDS line scans across the thickness of solder joints reflowed for various times, showing wt pct of (a) Sn, (b) Cu, and (c) Ag. No concentration gradients were evident

Full size image

Spallation of the top IMC layer was another possible mechanism considered. The evidence for this was found in some micrographs that showed large pieces of Cu6Sn5 IMC in the interior of the solder joint, as shown in Figure 4. These large Cu6Sn5 had morphology and size similar to that of Cu6Sn5 nodules in the IMC layer. The problem of IMC layer spallation in electronic interconnects has been studied extensively. In those cases, the metallization upon which the solder is reflowed is a thin film. After multiple reflows or during isothermal aging, the IMC layer that forms begins to consume the thin film metallization.[14–18] Once the entire metallization has been consumed, the entire IMC layer spalls off all at once since it does not adhere well to the substrate beneath. This form of spallation is called layer type massive spalling. This mechanism, however, does not explain the results presented in this study because in our case, the entire IMC layer does not spall all at once. And the substrate essentially constituted a nearly infinite source of Cu since the volume of the Cu bars (~3165 mm3) was many times greater than the volume of the solder (~16 mm3).

Fig. 4
figure 4

Backscatter SEM micrographs showing evidence for spallation of IMC layer. The top micrograph shows a spall Cu6Sn5 nodule sinking, and bottom micrograph shows a spalled nodule that has settled on the bottom layer

Full size image

An alternative mechanism was a gravity-driven spallation process. This mechanism is not well studied in the literature. Therefore, a simple experiment was conceived to test this hypothesis. Because spallation might be gravity driven, if there were a way to rotate the joints during reflow, then the asymmetric IMC layer growth would be mitigated. The experimental difficulty encountered in this experiment was how to rotate a molten Sn joint without destroying the joint in the process. The solution was to use a shaft collar, with the same inner diameter as the solder joint. The shaft collar was coated with graphite to prevent reaction with the solder. Then, it was placed around the solder joint perform (fluxed Cu bar-solder-fluxed Cu bar assembly prior to reflow) and secured to the copper bars using set screws. This procedure essentially locked the joint into place, and because the shaft collar inner diameter was equal to the solder joint diameter, it prevented the molten solder from flowing out. Using this shaft collar assembly, a solder joint was reflowed for 168 hours and rotated every 24 hours. Rotation effectively mitigated the spallation of the IMC layer and produced a more symmetric IMC layer thickness on both sides of the joint as shown in the micrograph in Figure 5.

Fig. 5
figure 5

Comparison of solder joints that were reflowed for 168 h. The top micrograph shows a solder joint that was not rotated. The bottom micrograph shows a solder joint that was rotated at equal time intervals to achieve more uniform IMC layer thicknesses at the top and bottom interfaces. The method of solder joint rotation using shaft collar is shown at right

Full size image

The phenomenon presented in this study has a potential impact on the reliability of solder joints in electronic packages. The asymmetric IMC layer growth became significant after 3 hours of reflow time. The move toward Pb-free solder alloys may also exacerbate this phenomenon since it is known that Sn-rich Pb-free alloys are more reactive, with respect to Cu substrates, than Pb-containing alloys.[19] And the drive to reduce solder pitch and squeeze more interconnects in smaller areas means that the solder joint volume will decrease as well. The microstructural evolution and tendency for spallation of IMC layers has been shown to have some dependence on solder volume as well.[20–23]