There should now be a couple of pages from a SMT magazine article on ABSE.
Posted in response to Phil's post in the voids in solder joint thread:
" This reference:
suggests that a fractional strain of 1% gives a fatigue life of 1000 cycles to failure with PbSn eutectic, and that the lifetime goes as
1/(epsilon**2). Lead-free solder has about a quarter of the fatigue life of PbSn. "A single test showing separation of the solder from the pad is hardly justification for the conclusion that "Lead-free solder has about a quarter of the fatigue life of PbSn."
Solder joint reliability is a complex phenomenon involving surface finish, processing parameters, solder creep and fatigue properties, etc. The only test that matters is a test of your particular processed assemblies, usually thermal cycling. The most challenging connection to make reliably is usually large BGAs, where the corner balls are most highly strained by thermal cycling. While I can't find them now, I have seen a few published test results comparing different lead free solders with tin-lead solders in large BGA thermal cycle to failure tests, and in these tests some lead-free solders underperformed tin-lead by about 20%, and some outperformed it by over 50% (not the cheap ones of course).
High reliability soldering is a complex issue, and if you want to understand it you should follow a few of the manufacturing trade rags like SMT.
While searching for the BGA test results I also happened to notice that the EMS division spun off by IBM has folded. Too much effort "proving" the unsuitability of lead-free and not enough time developing processes that work well, perhaps. All of the surviving large EMS companies have reported improvements in yield and reliability accompanying the conversion to lead-free, and virtually all exempt high-rel manufacturers are in the process of qualifying lead-free materials and processes, not because of any need to comply with ROHS but because these materials offer significantly improved reliability when selected and used in an optimal manner.
Tin-lead is still easiest to use, and I would not suggest that small exempt manufacturers convert due to the high cost of qualifying new processes. But those who have done the work are achieving excellent results.