Reliability of Lead Free Solder

Did you ever go to a conference and compare lead free reliability results presented by different groups? Often looks like ‘lots of noise’, doesn’t it?

Click on a picture for larger image and explanation

Current lead free solders are fundamentally different from SnPb. It isn’t just that materials properties are different. They also depend on things that didn’t use to matter, or at least not as much, for SnPb¹. This means that general statements such as ‘SAC solder joints are more reliable than SnPb under thermal excursions’ are not always true (more), but there are lots of other consequences too.

What’s so challenging about reliability assessment?

Modeling -- We are, for example, used to assuming that constitutive relations etc. are independent of solder volume, time, and so on. However, that is not true for current lead free solders. Properties vary with solder joint size, pad finishes, process, history, … Modeling is particularly complicated by solder properties varying with joint location and time in cycling (details). Or rather it would be, if anyone took this into account. To be fair, however, they can’t do that until quantitative data become available. That is going to take a while, among other because of the proliferation of lead free solder alloys (details).

Test Protocols – Current thermal cycling test protocols may in fact be strongly misleading, but we believe to be on track of a solution to the most obvious challenge (details). Otherwise, reliability test protocols should also be defined to account for effects of time (temperature) and cyclic loading on solder properties (details). This is far from always trivial. Aging may, for example, soften the solder but also affect its fatigue resistance. Intermetallic strength and robustness under a cyclic load have been seen to both increase and decrease in aging.

Data Interpretation – Are you sure you know how to interpret your experimental results? Your difficulties depend on what you want to know. However, you almost always want to extrapolate your results to much earlier failures. Maybe you do that implicitly without realizing it, but hopefully you do it. The problem is here that lead free solder results are almost certain to not follow for example a Weibull distribution (details).

What can you do? Click here for options.

References

¹ P. Borgesen, T. Bieler, L. P. Lehman, and E. J. Cotts, “Pb-Free Solder: New Materials Considerations for Microelectronics Processing”, MRS Bulletin 32 (4), April 2007, pp. 360-365