No abstract

The extrapolation and generalization of accelerated test results for lead free solder joints require the identification of a damage function that can be counted on to apply beyond the region of the test. Individual ball grid array (BGA) scale Sn3Ag0.5Cu (SAC305) solder joints were subjected to isothermal shear fatigue testing at room temperature and 65 °C. The resulting mechanical response degradation and crack behavior, including strain hardening, crack initiation, and propagation, were correlated with the inelastic work and effective stiffness derived from load–displacement hysteresis loops. Crack initiation was found to scale with the accumulated work, independently of cycling amplitude and strain rate. The subsequent damage rate varied slightly with amplitude.

The fatigue behavior of solder joints in realistic service applications is still not well understood. Service life prediction based on conducting accelerated tests and extrapolating test results therefore involves a high potential for error. Understanding both the evolution of solder joint properties and the damage accumulation has proved to be critical to reliability modeling. Damage accumulation in isothermal cycling is shown to scale with the accumulated inelastic work even in complex cycling scenarios, so that the life of a solder joint ends upon accumulation of a given amount of work. Individual ball grid array solder joints were cycled in shear fatigue experiments with different load amplitudes and strain rates. The effects of load amplitudes and strain rates on the work accumulation and fatigue life were systematically addressed. The correlation between different loading scenarios and the accumulated work to failure was also discussed. The results showed that the accumulated work until the development of a major crack is constant regardless of the load amplitude. After that the accumulated work to failure is lower for larger load amplitudes. For some reason, a larger fraction of the work appears to be dissipating as heat at lower load amplitude, but only during crack growth. On the other hand, the strain rate affects the fraction of the work going to heat even before the development of a major crack.

In the absence of manufacturing defects the ultimate life of a microelectronics assembly in long-term vibration is likely to be limited by solder joint fatigue. So far relatively limited efforts have been dedicated to the quantitative prediction of this life. However, even day to day 'engineering tests' may not be very useful if they do not offer at least a ranking of alternative materials, designs or processes in terms of performance in service.

No abstract