Thermal cycle fatigue life prediction for flip chip solder joints

Finite element analysis was used to predict the thermal cycle fatigue life of flip chip solder joints. Three different data sets of measured thermal cycle fatigue life were simulated to assess model correlation. Variables in the measured data included solder resist opening diameter, UBM diameter, joint height, solder volume, underfill material, package structure, and temperature range. Viscoplastic strain energy density (SED) per cycle was used as the damage indicator. Three different post processing schemes were compared for extracting the damage indicator: maximum elemental SED value; maximum SED² / average SED; and layer averaged SED. It was found that all three methods could achieve some degree of correlation between measured and predicted life for a given data set. The best and worst legs of an experiment could be predicted well, but the ranking of all legs in a given experiment could not be perfectly predicted. For the experiment on joint design, max SED ² / avg SED gave the best prediction, followed by max SED, and worst was layer averaged SED. For the experiments on power cycle conditions and package structure, all methods gave similar results. Finally, the three data sets were plotted together with additional predictions from previous work on WLCSPs. There was a large discrepancy in the absolute prediction from one data set to the next. However, the ability to predict the relative fatigue life within a data set was consistent, and the results did match the previous work on WLCSP solder joints.