A study shows two significant factors are solder paste flux chemistry and BGA alloy ball wetting.

Head-in-pillow (HIP), also known as head-on-pillow or ball-in-socket, is a solder joint defect in which the solder paste deposit wets the pad, but does not fully wet the ball. This results in a solder joint with enough connection to have electrical integrity, but lacking sufficient mechanical strength. Because of the lack of solder joint strength, these components may fail under minimal mechanical or thermal stress. This potentially costly defect is not usually detected in functional testing, but rather shows up as a field failure after the assembly has been exposed to physical or thermal stress.

HIP defects have become more prevalent since BGA components have been converted to Pb-free alloys. The defect possibly can be attributed to a chain reaction of events that begins as the assembly reaches reflow temperatures. Components generally make contact with solder paste during initial placement, and start to flex or warp during heating, which may cause some individual solder spheres to lift. This unprotected solder sphere forms a new oxide layer. As further heating takes place, the package may flatten, again making contact with the initial solder paste deposit. When solder reaches the liquidus phase, there isn’t sufficient fluxing activity left to break down this new oxide layer, resulting in possible HIP defects. Since component warpage is unpredictable and inconsistent, the focus must turn to the interaction of process variables and those that can be altered to reduce the incidence of HIP defects. These variables include BGA ball alloy, reflow process type, reflow profile and solder paste chemistry. Each of these variables is studied and discussed below.

With the need for better drop resistance, many Pb-free BGAs are being made in alloys other than SAC 305. Because SAC 305 has significantly lower drop resistance than SnPb37 (Figure 1), component manufacturers have been moving away from this type of alloy and toward alternatives such as SAC 105 (composed of tin plus 1% silver and 0.5% copper). Many varieties of SAC 105 include a fourth element, often referred to as a dopant, such as antimony, magnesium, nickel, cobalt or indium. These additives create finer grain boundaries and reduce the intermetallic formations of the tin with silver or copper, resulting in a more reproducible grain, as well as a more uniform grain formation in the Pb-free alloy. These also yield a different oxide and surface condition, depending on the element used and cooling rate during assembly. This different oxide and surface condition can cause issues with the flux activity and impact solder wetting and complete joint formation of the BGA.


Solder sphere (ball) issues. Figures 2 and 3 are analyses of BGAs known to have had HIP problems. Using SEM, it was determined there are very distinct grain structure variations within the balls (Figure 2). Inspecting these components reveals three distinct classifications of balls on the component; these were labeled shiny, matte and spotted (Figure 3). (As a point of clarification, the large dimples on the ball surfaces are from test probes that easily penetrated any of the surface irregularities or containments during component testing by the manufacturer.)



The spectrums of the BGA balls also are different (Figure 4). On this single BGA there exist three different grain structures and surface elements. One theory explains this is caused by variations in cooling rates when the solder ball was initially formed.


We developed a test procedure to understand the interaction of these elements with specific paste chemistries. This permitted a classification of reactivity levels of some of these dopants. It was discovered very low levels of magnesium (in the 30 ppm level) directly affect standard solder paste flux chemistries, while indium affects them in the 500 ppm range, nickel and cobalt in the 400 ppm range, and antimony in the 1000 ppm range. Although the grain structures all appeared similar, the flux interaction was different. This difference was determined by a viscosity test conducted while the paste medium was in contact with the solder alloy doped with the aforementioned elements.

Other factors that appear to influence HIP include types of reflow, reflow profiles and solder paste chemistry.^1,2 Some data obtained suggest vapor phase reflow may result in more HIP defects than convection reflow. It is not clear whether this is truly related, however, as it has only been seen as a trend.

An experiment was performed to measure the impact of reflow profile on HIP solder joint formation. The experiment utilized two different reflow profiles. The first profile was a standard ramp-soak-spike (Figure 5a). The second profile (Figure 5b) included a hotter soak zone and longer dwell time at liquidus. There was no perceived difference in the defect rate, depending on the profile; each resulted in random cases of HIP, depending on the component tested.


Solder paste chemistry was the next factor tested to determine impact on HIP. During this experiment, it was found solder paste chemistry appears to have the single greatest effect on the HIP defect. When changing from an older Pb-free solder paste to a new higher-temperature activation paste, the defect, in many cases, was eliminated. In other cases, it was more difficult to remove. However, in the experiment run, the solder paste chemistry appears to have the largest impact on HIP.

An experiment was conducted using various solder paste chemistries to measure their effect on HIP incidents. It was determined that, regardless of the reflow profile used, the novel solder paste eliminated HIP. Although this solder paste is halide-free, a solder paste containing >0.5% halide also was used in this experiment, and the defect was once again eliminated.

This indicates solder pastes with an activation system able to provide sustainable high-temperature fluxing activity are capable of creating a homogenous connection beyond the ball and the paste alloy interface. Figure 6 shows a HIP formed using a lower-temperature activation system. Figure 7 is a joint formed with a high-temperature activation system; no evidence of HIP is seen.


Based on these experiments, Table 1 was generated to show the relative impact of variable(s) that contribute to HIP, rated on a scale of one to 10, with 10 as the most likely to eliminate the issue.


Conclusions

Based on this preliminary study, it appears the two significant factors are solder paste flux chemistry and BGA alloy ball wetting. Frosty, non-uniform structures appear to perform the worst for BGA HIP. This is logical, as these are intermetallic regions on the surface of the solder ball. The intermetallic connection of AgSn and CuSn possesses much higher melting temperatures than the alloy themselves. They are also crystalline in structure and can repel wetting. Although additional studies are necessary to corroborate these results, there is a strong indication this surface structure is one of the leading causes of the HIP defect.

References

1. Chrys Shea, “HOP-ping Mad,” Circuits Assembly, July 2008.
2. American Competitiveness Institute, “Stop the HOP,” Circuits Assembly, August 2008.

Karl Seelig is vice president of technology at AIM (aimsolder.com); kseelig@aimsolder.com.