Factors affecting voiding in Pb-free solder joints: voiding is sensitive to a particular flux constituent, and copper pad finishes more susceptible than AuNi or immersion Ag.

Concerns have been raised that Pb-free solder joints show higher overall levels of voiding than corresponding SnPb joints. With the widespread availability of x-ray inspection equipment, voids are easily observed. Moreover, there is widespread perception that voids are deleterious to joint reliability. Although both published experimental work and theoretical considerations show no effect of normal levels of voiding on reliability (1), there is demand to minimize voiding, which in turn requires better understanding of factors affecting voiding.

The tendency to higher levels of voiding has been rationalized on the basis of differences in physical properties between SnPb and Pb-free solders. Pb-free alloys' lower density means a reduced buoyancy force to expel bubbles from the molten solder, and their higher surface tension means an increased kinetic barrier for bubbles to escape during reflow. The higher melting temperature of Pb-free alloys--217[degrees]C for SAC alloys vs. 183[degrees]C for SnPb--may cause more volatile material to be evolved, leading to more trapped voids.

The intuitive assumption that the presence of voids in a joint degrades joint reliability has been refuted by a recent IPC report (1), which concluded "there is no evidence that solder joint voiding has any significant impact on solder joint reliability." The same conclusion has been reached in other studies; in one case (2), it was shown that for a constant solder volume in the joint, eliminating voids actually lessened reliability by a factor of two, probably because the reduced joint volume decreased the standoff height. Despite these findings, reduced voiding requirements remain.

In our previous study (3, 4), the significance of reflow profile and solder paste flux medium parameters were investigated using DoE methodology, and it was found that while the flux parameters had at best only a marginal effect on voiding, changing the reflow profile from a short linear profile to a soak profile significantly reduced the voiding level.

In the current study, we have largely ignored the process parameters and have focused on investigating the effects of some material parameters on voiding, including Pb-free alloy type, PCB pad finish and content of a specific flux medium constituent. Moreover, we have compared a SnAg solder alloy, and looked at different PCB pads; bare copper, OSP, immersion silver and AuNi finishes, and four different levels of the medium component.

Principal findings suggested that amount of voiding was sensitive to a particular constituent of the flux medium, and that copper pad finishes gave more voiding than AuNi or immersion Ag finishes. A SnAg solder showed much reduced voiding as compared to a SAC solder under otherwise identical conditions. Possible reasons for these findings are discussed below.

Experiment

Voiding in BGAs. Voiding levels were assessed in solder joints/bumps from three different Pb-free BGA components populated on a specific test PCB. The BGA56 and TBGA132 were 56 and 132 bump devices with a peripheral ball matrix at 0.5 mm pitch; the UCSP98 had a full array ball matrix (also at 0.5 mm pitch). The surface finishes came in bare copper, OSP coated copper (Shikoku Glicoat FX), immersion silver and ENIG or AuNi (as referred to throughout this report). The surface pads for all three BGA components were 0.3 mm diameter round pads. Figure 1 shows an example of a populated test PCB.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

Each test PCB was prepared by first stencil printing the solder paste onto the PCB using a MPM AP-Hie stencil printer incorporating a 150 [micro]m thick laser-cut stainless steel foil. The printing parameters were optimized for each solder paste type to yield good paste roll, stencil clearing and drop-off from the squeegee blades.

A standard Pb-free flux medium incorporating SAC387 (SnAg3.8Cu0.7) alloy was used to assess the effect of PCB pad finish on voiding, and a SnAg3.6 alloy incorporated into the same standard flux medium was used to assess the effect of alloy type. The solder pastes used to assess the different flux formulations on voiding also incorporated the SAC387 alloy. The fluxes had similar chemistries, with only small adjustments made to specific components.

After stencil printing, the BGA components were placed on the PCB using a Mydata MY9 placement machine, and the populated PCBs were reflowed in air using a short, linear-like reflow profile. This particular reflow profile was chosen, as it represents a typical, relatively fast ramp, linear-like profile used in many SMT applications. Moreover, our previous work (3,4) has shown that this type of profile is more susceptible to yield a larger amount of voiding. The reflow profile and corresponding parameters are shown in Figure 2 and Table 1, respectively.

Each set of populated PCBs with different combinations of pad finish/paste type was repeated five times to ensure consistency of results.

Solder bumps and joints were inspected for voiding using an X-Tek 2-D vertical x-ray system. The x-ray settings were initially optimized to yield an accurate assessment of the level of voiding and then kept constant throughout the experiment.

It is worth noting that from this study and other experience, voiding was always significantly higher in solder bumps than in joints. Solder bumps are reflowed solder paste deposits on the BGA pads that do not contain a placed component, while BGA solder joints are produced by reflowing after components are placed on the printed solder paste deposits.

The level of voiding was assessed by:

* Measuring the projected area of voids in each BGA bump/joint.

* Counting the number of voids in each BGA bump/joint.

* Displaying the average void area (total) in each bump/joint as a boxplot.

* The average void size (single) in each BGA bump/joint was calculated by dividing total void area by number of voids.

Figure 3 displays a typical x-ray image used to assess voiding. In some bumps, there is more that one void; however, the total projected void area in that bump would be the sum of the multiple voids.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

Results

Voiding vs. PCB pad finish. Figure 4 shows typical x-ray images from USCP98 solder bumps on four different pad finishes. It is clear from these images that the level of voiding is greatest for the bare copper pads; i.e., more voids per solder bump and larger voids. The levels of voiding decrease as the pad finish changes from OSP to silver, and finally AuNi, where the level of voiding is lowest. In the latter case, there are far fewer voids, with the majority of the bumps void-free. This observation is quantified and clearly shown in Figure 5. Additionally, the behavior is repeated in the solder bumps of the TBGA132 (Figure 6).

These results are the first indication that voiding is sensitive to the copper content of the bump/joint. In the case of the bare copper PCB pad, the copper would be available to dissolve into the molten solder during reflow and form CuSn intermetallics. The amount of copper dissolved into the solder would be minimal for the AgNi surface finish where the nickel would act as a barrier layer.

Moreover, this could also explain why voiding is much lower in the solder joints compared to solder bumps; i.e., Figures 7 and 8. The BGA's balls would have a dilution effect to the copper content during reflow, and hence, less copper content in the joints may explain the lower level of voiding.

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

[FIGURE 10 OMITTED]

Voiding vs. alloy in solder paste. Figure 9 shows typical x-ray images from USCP98 solder bumps on copper and silver pads using standard SAC387 and SnAg paste. It is clear from these images that the alloy used has a surprisingly large effect, and that the alloy without copper shows much lower levels of voiding than the SAC alloy. This is quantified in Figure 10. The significant reduction in voiding is observed in the boxplots and histogram in Figure 11, where the SnAg alloy displays mostly balls with zero voiding.

Voiding vs. flux formulation. A matrix of mediums was designed (Figure 12) whereby a specific flux medium component level was reduced from a high level to zero, i.e., M1-M4. In addition, other mediums were produced that had the same chemistry but contained different levels of the rest of the flux medium constituents--again like the M1-M4 set, the level of a specifically chosen component was decreased from a high to low level (i.e., M6-M9 and M11-M14). In essence this equated to three flux chemistries in which the level of one particular component was altered.

Figure 13 displays the voiding in the UCSP98 bumps for the M series flux formulations. It is clear that the level of voiding decreases with the targeted flux component. The different surface finishes in which pastes M1 and M4 are compared (Figure 14) showed the same result. Finally, the SnAg alloy was introduced into formulation M1 and the voiding level again decreased dramatically. In this case, it is clear that the flux manufacturer can control the level of voiding.

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

[FIGURE 13 OMITTED]

[FIGURE 14 OMITTED]

Conclusions

The results demonstrate that the solder alloy has a surprisingly large effect and that the alloy without copper shows much lower levels of voiding than does the SAC alloy. Among the PCB pad finishes, AuNi where no copper comes into contact with the solder exhibits the least voiding, followed by immersion silver, where coating of silver is over the base copper. Next is OSP. Bare copper showed the most voiding. These results indicate a connection between copper in the solder and voiding. One explanation is that the addition of copper to SnAg solder leads to more large Ag3Sn plates in the joint (5), and that these plates hinder the escape of voids from the joint during reflow. Further work is planned to identify several candidate mechanisms for the formation of such voids.

The effect of specific flux constituent was that voiding increased proportionally with concentration. The same trend was observed with paste manufactured using three different flux mediums.

The magnitude of the effect on voiding was greatest when the alloy was changed; the magnitudes of the effects of flux formulation and surface finish were similar.

References

1. IPC Solder Products Value Council, "Round Robin Testing and Analysis of Lead Free Solder Pastes with Alloys of Tin, Silver and Copper," IPC Solder Products Value Council Final Report, July 2005.

2. Donald R. Banks, et al, "The Effects of Solder Joint Voiding on Plastic Ball Grid Array Reliability," Surface Mount International Proceedings, August 1996.

3. G. J. Jackson, P. Hedges and B. J. Toleno, "Step-7 Soldering," SMT, August 2006.

4. G. J. Jackson, P. Hedges and B.J. Toleno, "Assessing Solder Paste Design and Process Factors on Voiding Levels in 0.5mm pitch Pb-free BGA Solder Joints," Proceedings of the 2005 Soldertec Conference, June 2005.

5. H.Y. Lu, H. Balkan and K.Y. Simon Ng, "Solid-Liquid Reactions: The Effect of Cu Content on Sn-Ag-Cu Interconnects," Journal of Metals, vol. 57 (6), pp. 30-35, June 2005.

Ed.: This article was originally presented at IMAPS 2006 and is reprinted here with permission from the author.

Dr. Gavin J. Jackson is product development manager and Dr. Hector A. H. Steen is research associate at Henkel Corp. (henkel.com); [e-mail].

Table 1. Reflow Profile Parameters for Profiles Used in Voiding
Assessment

Time to 150[degrees]C/sec.                     94
Time between 150[degrees]-170[degrees]C/sec.   30
Time between 170[degrees]-217[degrees]C/sec.   58
Time at 217[degrees]C                          45
Peak temperature/[degrees]C                   241