Detecting Brittle Fracture Failures
11/01/2007
Correlating High-speed Bond Testing and Drop Testing
BY STEPHEN CLARK, Ph.D., Dage Precision Industries
The increased use of lead-free solder in BGA packages, widely used in portable devices, makes them susceptible to brittle fracture failures at the solder ball to pad interfaces when subjected to mechanical shock. Brittle fractures at the interfaces between solder balls and package substrate bond pads are considered unacceptable.
In principle, this kind of solder joint reliability is characterized by board-level drop testing, but such testing has several drawbacks. Each drop test consumes several packages and hundreds of solder joints, incurring considerable expense. In addition, cracks in the solder joint may close after impact, resulting in an undetectable failure unless there is a high-speed real-time data acquisition system available for monitoring. Finally, analysis of the data is time consuming, adding significant expense. Therefore, finding alternative methods for evaluating solder joint integrity under mechanical shock loading conditions is imperative.1 A comparison of high-speed bond testing with board-level drop testing of BGA packages using lead-free solder balls and various package substrate surface finishes suggests that the former might be a viable alternative.
A study was conducted using multiple BGA package constructions using various combinations of solder alloys, surface finishes, substrate material, solder ball size, and package dimensions. A typical device tested was a 316 PBGA (27 mm×27 mm) construction, using Sn 4.0%/Ag 0.5%/Cu (SAC 405) solder balls, fabricated with different substrate surface finishes including electroless nickel immersion gold (ENIG) and organic solderability preservative (OSP).
The 316 PBGA samples used standard 0.76-mm diameter spheres. The package substrates were composed of BT laminate, with a thickness of 0.36 mm. The solder bond pads were solder-mask-defined with an opening of 0.635 mm in diameter. The solder balls were attached to the substrates in a hot-air convective reflow oven using a lead-free soldering profile of 150° C ± 2° C pre-heat, with a peak reflow temperature of 260° C.
Samples were divided into groups which were subjected to thermal aging at 125°C (0 to 500 hours) to accelerate the formation of intermetallic compound (IMC) at the package substrate/solder-joint interface. The high-speed ball-shear tests ranged from 10 mm/s to 3,000 mm/s, and the high-speed ball-pull tests ranged from 5 mm/s to 500 mm/s. An advanced, high-speed bond testing machine was used, equipped with control and analysis software and next-generation force transducers, which are able to evaluate the fracture energy of solder balls in both ball-shear and ball-pull tests.
The second part of the study involved board-level drop testing, when records of electrical resistance, circuit board strain, and fixture acceleration were recorded. Detailed analyses were performed to identify the failed solder joints and corresponding failure modes. The failure modes and loading speeds of solder ball-shear and -pull tests were cross-referenced with the mechanical drop tests for comparison. Also, the energy absorption value recorded during solder ball-shear and -pull tests was considered an effective index to interpret the solder joint failure mode.
Thermal aging to accelerate IMC growth was conducted at 125°C in an oven for several time durations of 100, 300, and 500 hours. After thermal aging, some PBGA specimens with solder balls were molded, cross-sectioned and etched, and subsequently inspected and analyzed by scanning electron microscope (SEM). Similar BGA samples were assembled on test boards and dropped using a dual-rail guided device. Some board level test samples were also subjected to thermal aging, as above. All samples were equipped with daisy chains and subjected to real time data acquisition monitoring.
Detailed SEM analysis performed on both complementary surfaces of brittle fracture failures from both shear and pull test samples showed similarity to those generated during board-level drop testing, and high-speed shear and pull testing. Furthermore, there was strong correlation between various bond testing parameters at which brittle fractures occurred, and the number of drops-to-failure observed with board-level drop testing. 2
From the comparison of cross-sections and fracture surfaces, it is clear that brittle fracture interfaces from drop testing show a striking similarity with those from high-speed ball-shear and pull tests. As a result, it is suggested that brittle fractures obtained in high-speed bond testing are a strong indicator of board-level drop testing behavior. A distinctive feature of the current work is the effort which has led to direct comparison of the physical characteristics of brittle failures from high speed bond testing with those from board level drop testing.3
Drop Testing
Test boards subjected to board-level drop testing were fabricated with both non-solder-mask-defined (NSMD) and solder-mask-defined (SMD) pad geometries. In both cases, the solder-wetted pad diameter was 0.684 mm. Although NSMD is more typical of actual production circuit boards, SMD has the advantage for this correlation study in that the board-level drop testing fracture locations are more likely to occur on the package side. This is significant because solder ball-shear/pull testing can only evaluate package-side fractures as the component is not attached to a PCB. The appearance of the brittle fractures at the surface of a solder joint subjected to drop testing are shown in Figure 1.
An abbreviated summary of the drop testing results (8 assemblies per data point) of the drops-to-failure for test board assemblies illustrated a more rapid degradation with thermal aging of the OSP package substrate surface finish than those with an ENIG finish.
High-speed Bond Testing
Brittle fractures failures with ENIG surface finish specimens were typically induced between the IMC and Ni layers. For the OSP specimens without aging subjected to two times reflow, the brittle fracture failures were found between the Cu6Sn5 IMC and Cu layers. The brittle fracture failures of the OSP specimens after thermal aging occurred between the Cu6Sn5 and Cu3Sn IMC phases. The appearance of brittle fractures at the surface of a solder joint subjected to high-speed shear and pull testing is shown in Figures 2 and 3.
Previous evaluations of high-speed solder ball-shear and pull testing have observed brittle fractures that appeared similar to the brittle fracture mode observed in board-level drop-testing assemblies, but little definitive cross-sectional evidence has been provided. This is partially due to the difficulty of such studies, both in terms of retrieving individual sheared or pulled balls and matching them to their corresponding pad, and the subsequent cross-sectional work.
Direct Correlation
In addition to the microstructural correlations observed, a strong correlation has been observed between the test parameters for high speed bondtesting and board level drop testing. The mathematical correlations relating solder ball-shear/pull and drop-test results are highly complex. Nonetheless, an innovative approach is graphically summarized in Figure 4. These graphs relate the brittle fracture percentages from shear and pull solder ball testing to the drops-to-failure for specific packages and drop-test conditions used in this study. Briefly, this plot is achieved by plotting the drops-to-failure number for each time point against the equivalent data from shear or pull testing, followed by power law curve fitting. Each curve corresponds to one solder ball-shear or pull-test speed. These curves can be used to estimate the drops-to-failure number from the brittle fracture percentage obtained in either the ball shear or pull test at a specific test speed, and may provide some predictive capability to estimate drop-test results from high-speed ball-shear or pull-test data.
Conclusion
Detailed comparison of brittle fracture interfaces produced by drop testing and those produced by high-speed shear and pull tests show a striking similarity. Therefore, based on micro-structural observations, there is a direct correlation between high-speed bond testing and conventional drop testing. This supports the correlation between test parameters for the two methods showing that high-speed solder ball shear and pull testing can be used as a predictor of drop-test performance.
Acknowledgments
The work described in this article was carried out by Fubin Song, Ph.D. in the Center for Advanced Microsystems Packaging under the supervision of Professor Ricky Lee at Hong Kong University of Science and Technology. Thanks are due also to Keith Newman of Sun Microsystems.
REFERENCES
- K. Newman, “BGA Brittle Fracture-Alternative Solder Joint Integrity Test Methods,” Proc. 55th Electronic Components & Technology Conference, Orlando, FL, June (2005), pp. 1194-1200
- F. Song, S. Lee, K. Newman, B. Sykes and S. Clark, “High-Speed Solder Ball Shear and Pull Tests vs. Board Level Mechanical Drop Tests: Correlation of Failure Mode and Loading Speed,” Proc. 57th Electronic Components & Technology Conference, Reno, NV, June 2007, pp. 1504-1513
- F. B. Song, S. W. R. Lee, K. Newman, B. Sykes and S. Clark, “Brittle Failure Mechanism of SnAgCu and SnPb Solder Balls during High Speed Ball Shear and Cold Ball Pull Test,” Proc. 57th Electronic Components & Technology Conference, Reno, NV, June 2007 pp.364-372]
STEPHEN CLARK, Ph.D. bondtester product manager, may be contacted at Dage Precision Industries, Inc., 48065 Fremont Boulevard, Fremont, CA 94538; 510/683-3930; E-mail: [email protected].