Being metallurgist and having worked with the intermetallics (IMCs) formed by
soldering or solid state diffusion for many years. I may add some
experimental information on the subject.

As mentioned in previous e-mail, a solder joint can be formed by rubbing
(ultrasonic) a solder coated lead against freshly cleaned copper.  In  my
tests these joints were roughly 1/3 the peel strength of a similar fusion
formed solder joint.  After thermal treatment between 80 and 100C and 130 to
150C these joints formed the normal CuSn intemetallics and improved in
strength, but were not up to the same value as the fusion joint.  Analysis
indicated that more copper surface area reacted when the joint was formed by
fusion.

As mentioned the solder joint depends on the formation of the IMCs to produce
wetting, but that the it mayb be that the thicker the intermetallic the worse
the joint.  When using very "high purity" copper and tin/lead solder for the
joint the peel strength remained roughly the same after long or short
immersion times in the molten solder such as HASL and while the copper
thickness decreased, IMC thickness remain nearly constant, and the ratio of
the two intermetallic layers remained the same.  (i.e. If left in the HASL
over lunch the intermetallic remained the same thickness but all of the
copper was dissolved).  In solid state growth (150C for 4 hours to 192 hours)
the intermetallic grew following the square root of time formula for about 48
hours until the residual lead layer appeared to inhibit diffusion of the
copper into the solder. I tried this with molten pure tin and the square root
of time held.

The thickness of the intermetallic formed by thermal aging had little effect
on the peel strength until the intermetallic thickness reached above 25
microns (0.001 inch).  Then the lead rich layer appeared to be a major the
contributor, it formed large lead crystals the fractured easily along grain
boundaries.

Failures by brittle fracture at the intermetallic layer were produced by
shorter term thermal aging and thinner (2-5 microns [0.000080-0.0002 inch])
by dosing a copper plating bath with specific amounts of organic plating
additives, brighteners etc.  The higher the additive content of the bath the
thinner the IMC had to be before the fracture occurred.  This additive
content and early fracture phenomena occurred with pyro copper, fluoroborate
copper and sulfuric acid copper with pyro and fluoroborate copper being the
worse.  GE and Raytheon in articles published in the 1970's found similar
results, wherein surface mounted components were popping off after final bake
or burn-in of the finished assembly.

In tests run with copper foil in which the foil was solder dipped, then
thermally aged and rolled on a 1/4 inch mandrel, the intermetallic layer had
slightly more vertical cracking as the layer got thicker, but separation of
the copper to solder at the IMC layer was not seen on pure copper foils.  For
plated copper foil made with different additive amounts in the plating bath
the separation was concentration related.  A further note, when a copper
plating solution had been used extensively with little care or or
purification, the increase in separation was greatly increased.  Analysis of
the separated/fractured surface of the sample that was plated using high
additive plating solution showed an increased amount of Sulfur and Carbon.
Theory proposed was that when the organic concentration in the copper plating
reached an amount, the copper left behind a debris layer as it diffused into
the solder to form the intermetallic and the .sparation/fracture occurred
along the debris layer.  Kirkendall voids were probably not a factor as the
pure copper samples did not exhibit the separation.

Phil Hinton.