No abstract

Solder joints of leadless surface mounted components, which were thermally cycled, were studied to determine their failure mode and to compare the results with work done on bulk solder specimens. Previous studies done on the characterization of the mechanical properties of cyclically tested bulk Pb/Sn solder of different compositions show that their deformation behavior is a result of an interaction between creep and fatigue damage processes. Fractographic forensics of the failed solder joints also showed damage due to both creep and fatigue processes. Fatigue striations were observed side by side with creep voids and extensive intergranular cracking. The crack growth rate (da/dn) was calculated to be around 0.25 μm/cycle. It is also established that the failure is crack propagation controlled. It was found that the crack initiation site on at least one of the joints was a tin rich area. The size of the voids varied from 0.1 μm to 0.5 μm.

For lead/tin solders, room temperature is hot enough that deformation is controlled by creep. Creep deformation is thermally activated and is a function of time. Therefore, the appropriate parameters which can describe cyclic behavior are the ones which are functions of temperature and time, such as strain rates, ramp rates, and hold times. The traditional cyclic loading parameters of strain amplitude and stress amplitude are inadequate and can even be misleading as they do not take into consideration the rate at which strain or stress was applied. In addition to permanent creep strains, anelastic strains are also stored as a function of time during cyclic loading. Anelastic strains are recoverable and thus can be nondamaging. During cyclic loading, anelastic strains are stored along with creep strains during the time spent on load. If the hold times in accelerated cycling are shortened to the extent that a significant portion of the total strain stored during on-load is anelastic (around 3 seconds) then the damage storage rate decreases rather than accelerating. These factors need to be considered both when accelerated testing is contemplated and in the development of life prediction equations.

No abstract

No abstract