Pb-free Solder Joint Reliability

Joe Smetana, Alcatel-Lucent, Plano, TX
Richard Coyle, Alcatel-Lucent, Murray Hill, NJ
Thilo Sack, Celestica, Toronto, Canada
Ahmer Syed, Amkor, Chandler, AZ
David Love, Oracle, Santa Clara, CA
Danny Tu, Huawei, China
Steve Kummerl, Texas Instruments, Dallas TX

Two different temperature cycling profiles were used to compare the thermal fatigue reliability of Pb free and SnPb solder joints in 16 different, high strain surface mount (SMT) packages. In some applications, high strain Pb free (SnAgCu) components are expected to fail earlier than the equivalent SnPb version. In this program, test results were compared for a 0-100°C thermal cycle used often for evaluating high reliability applications to those of a comparatively mild accelerated thermal cycle condition of 20-80°C.

A total of 6957 cycles was completed in the 0-100°C testing and a total of 9792 cycles was completed in the 20-80°C thermal cycling. The program was completed after over more than two years of elapsed test time. Weibull analysis, acceleration factors between the tests and failure analysis are included. The results indicate that the SnAgCu (SAC) components that create a high strain in thermal cycling tend to perform worse in 0-100°C testing than identical SnPb soldered components.

However, when the strain is reduced by testing with the reduced ΔT in the 20-80°C cycle, the SAC thermal fatigue performance is equal to or better than that of identical SnPb soldered components. The encouraging SAC performance in the 20-80°C cycle tends to mitigate Pb free reliability concerns because the lower strain test conditions are closer to actual service conditions.

The following specific conclusions can be drawn from the results of the thermal cycling study:

1. The SnPb solder did not unilaterally outperform the SAC solder in 0-100 °C testing as might be expected for high strain components.. This suggests that the reliability of SAC is not substantially worse than SnPb for many components, even when testing is done with the 0-100 °C.


2. The 20-80°C test results confirm the working hypothesis for this project that thermal fatigue performance of high strain SAC assemblies approaches that of comparable SnPb assemblies when the thermal cycling is performed using a milder thermal cycle. Only two of the high strain components, the QFN64 and the 5x7 ceramic oscillator, exhibited poorer performance with SAC solder and in the case of the 5x7 oscillator, the reliability of the SAC is not measurably lower than the SnPb (~ 3% difference).


3. The SAC characteristic life is always better than SnPb for all the BGA components evaluated in this study. This is an important result because of the long standing concern regarding the reliability of SAC ball grid arrays. Adequate BGA reliability with SAC solder is vital to the successful transition to Pb free design and manufacturing.


4. In view of the aforementioned conclusions, SnPb acceleration factors should enable a conservative estimate of performance when used on data generated at 0-100°C as long as the thermal cycle end use environment is no more severe than 20-80°C. This conclusion provides a path to confirming the reliability of Pb free product designs. As originally published in the IPC APEX EXPO Proceedings.A wider distribution of data, and associated lower Weibull beta for SAC was found in the 0-100°C testing. The same effect was not seen in the 20-80°C data, where the strain levels were lower. This suggests that conditions of higher strain or higher peak temperature lead to more variability in SAC thermal cycling performance than in SnPb. However, with conditions of lower strain or lower peak temperature, the variability seen in thermal cycling performance of the two solders is similar.


5. The statistical analysis suggests that a second failure mode may be present in at least one component (and possibly a few others). The indication is that the second failure mode occurs with 0-100°C thermal cycling but does not occur with the 20-80°C thermal cycling. Failure analysis was not successful in identifying the actual failure mode and further work is suggested to study this behavior.