I just received a customer inquiry regarding a phenomenon that is little studied and even less quantified; “How many times can I reflow a solder alloy before damaging the solder joint?”
As you may already know, each time you bring a solder alloy above its liquidus temperature, it continues to dissolve the metallizations on the substrate to which you are soldering, as well as the metallizations on the leads of the component being attached. With modern processes, a 3-time “excursion” is common, especially with double-sided reflow and rework. Typically, the solder applied with paste is not reflowed again during the wave, because, through the use of pallets or selective soldering, it doesn't get quite hot enough to melt. That said, in such a case, the solder joint may become hot enough to receive some damage. To me, the most interesting thing with crystalline intermetallic layers, is that they don’t need to reach liquidus to form larger crystals. So a temperature excursion close to the liquidus may also increase the crystal structure size.
Another factor is surface metallizations, especially easy-to-solder surfaces such as gold or HASL. With gold, molten Sn/Pb solder at 200°C will dissolve at 35u-inch/s. So, a fine flash layer, such as 3-5u-inch, is gone within the first second, and the actual intermetallic is formed to the underlayment; most commonly nickel (Ni). This is similar with HASL, as the HASL layer is consumed into the solder joint at liquidus, and the intermetallic layer is formed with what is beneath the HASL.
The intermetallic layer will increase with time above liquidus (TAL) and also with temperature, with hotter dissolving more of the surface. This is why there are operating temperature limitations on the final solder joint, such as no more than 90% of the solidus of the alloy, in degrees Kelvin.
Another factor that affects grain structure, besides TAL and peak temperature, is cool-down rate. A faster cool-down rate will form a smaller crystalline grain structure, but keep in mind that a too-fast cool down rate may result in stresses being trapped in the grain structure from the CTE mismatch between the component and the substrate.
It has been our experience that 3 temperature excursions is the accepted limit (by most companies that I work). But, the only recommendation that we can offer is that you try “worst-case” scenarios, and have ALT testing and SEM cross-sections performed on real-world products in which 3, 4, and 5 excursions have taken place. Your particular case may be unique - it is well-worth determining your particular situation.