Reliability of Multiple Pb-free Solder Ball Alloys

Joe Smetana, Richard Coyle, Peter Read, Alcatel-Lucent
Thomas Koshmeider, Freescale Semiconductor
Dave Love, Sun Microsystems
Mark Kolenik, Philips Medical
Jennifer Nguyen, Flextronics

Thermal cycling tests were conducted using two different ceramic ball grid array (CBGA) test vehicles having balls comprised of nine different Pb free solder alloys. The experiment was designed as a screening experiment to obtain data comparing thermal fatigue reliability of the various solder ball alloys. The test matrix was dominated by commercial SnAgCu (SAC) alloys but also included other high, low, and no-Ag alloys. The surface mount assembly was done with SAC305 (Sn3Ag0.5Cu) solder paste. The thermal cycling data, Weibull analyses, and metallographic failure analysis indicate that the best thermal fatigue performance was obtained with higher Ag alloys

A thermal cycling study was conducted to obtain data comparing thermal fatigue reliability of the various Pb free solder ball alloys. The following conclusions can be drawn from the experimental results of this screening evaluation:

•Despite a number of anomalies in the experimental data, the thermal cycling data, Weibull analyses, and metallographic failure analysis indicate that the best thermal fatigue reliability performance was obtained with two alloys containing the highest Ag content. The SAC405 and binary eutectic Sn3.5Ag had a lower incidence of anomalies and consistently were the best performers with both the full and depopulated arrays.

•The SAC310 (the other high Ag alloy) data contained anomalies that produced a moderate reduction in the apparent characteristic lifetimes. However, this behavior could be explained largely by solder assembly quality defects detected in the failure analysis. Even with those data anomalies, SAC310 had performance in full and depopulated arrays, generally exceeding that of the lower Ag alloys.

•The SAC205N (Ni addition) performed nearly as well with the full array as the higher Ag alloys but the depopulated array performance was low.

•Based on previous results from the literature and direct comparisons with low Ag alloys in this work, the SN100C full array performed unexpectedly well. This is an inexplicable result and perhaps requires further study.

•The anomalies, particularly within the depopulated array data, preclude consistent, accurate rank ordering of the reliability performance of the alloys. With the exception of the three alloys with the highest Ag content, all other alloys had very low characteristic lifetimes when tested with the depopulated array.

•Atypical non-fatigue mixed mode failures located at the PCB side of the solder joints could account for much of the anomalous experimental observations including: 1) rank order alloy variations between full and depopulated arrays, 2) atypically low Weibull slopes (<3.0), 3) poor correlation between Ag content and reliability performance, and 4) early failures or indications of multiple failure modes in the Weibull plots.

•There is no correlation between alloy composition or specific alloying elements and the incidence of non-fatigue or mixed mode failures.