Electromagnetic and thermomechanic characterization and reliability evaluation

Long-term reliability testing under field conditions and evaluation of acceleration factors#

© Fraunhofer ENAS
Characteristic damaged SnAgCu solder joints after 4 ½ years of thermal cycling (6500 field cycles).
© Fraunhofer ENAS
Characteristic initial damage at Innolot solder joint after 3 ½ years of thermal cycling (4800 field cycles).
© Fraunhofer ENAS

Cost-effective long-term fatigue testing in a former mine

Microsystems application in harsh environments, i.e. outdoor or offshore use of electronic or sensing devices, demand for long-term reliable and efficient functionality. However, there is a lack of testing data for long-term thermal cyclic under benign cyclic conditions as well as a on a reliable predictive model for the relation between accelerated tests failure and failure under use conditions, i.e. fatigue acceleration. This issue can be addressed by long-term testing. Relevant loadings are thermal cycling and moisture exposure. However, long-term testing requires expensive equipment and in case of thermal cycling a huge amount of electrical energy, which sums up to a big amount of running cost.

A unique solution was found to lower cost: why not using natural conditions as given in former mines and install a long-term reliability lab there? An almost constant environmental temperature of 8 – 12 °C, a humidity of 70 – 98 % and a high constant air flow, which allows for cooling without artificial freezing devices, is provided for free in that particular environment. Temperature chambers were installed in a former mine »Sankt Anna Fundgrube« dating back to the 18th century, which allow exposure of microelectronic systems to long-term thermal cycling environments at high relative humidity. Temperature cycling is computer-controlled and can be accessed via internet.

Testing of functional electronic assemblies subjected to automotive »under the hood« conditions

Long-term fatigue of lead-free solder joints was investigated at conditions of test coming close to automotive applications. Industrially produced electronic boards soldered with either tin-silver-copper alloy (SnAgCu 305) or Innolot were subjected to field cycles 23 °C / 93 °C, 6 hours cycle time. After 3 ½ and 4 ½ years or 4800 and 6500 cycles, respectively, the test boards were analyzed by electrical testing, shear testing and cross sectioning of selected samples. Although the new creep-resistant solder material Innolot was developed to enhance the limits of temperature, however, its performance at characteristic field temperatures is also of interest.

After 6500 field cycles almost complete fatigue failure was observed from cross sectioning for large ceramic resistors (Fig. 1) and beginning fatigue failure for smaller resistors soldered with SnAgCu. The field cycling induced damage patterns were compared to those from test cycling -40 / 150 °C. The microstructural degradation observed for the field reliability tests suggested that the failure mechanism of SnAgCu is similar to that seen with test cycles. Recrystallization causes a fine grain structure and intergranular fracture occurs, sometimes with crack branching. Comparably little fatigue effects were seen with Innolot for the same components, however, different and new types of brittle cracking have been identified for test- und field cycling. For field cycling cracks occurred at the solder meniscus (Fig. 2), which are a new type of failure. In the case shown, they do not affect the electrical functionality, but similar cracks can result in a failure risk for other solder joint geometries.

Comparison of theoretical predictions for fatigue life to testing results

For SnAgCu acceleration of thermal shock cycling with respect to field cycling was seen to be approximately AF = 11. This acceleration factor was compared with calculations based on different published acceleration laws (Norris/Landzberg equations) and on numerical simulations. Evaluation by acceleration laws resulted in a broad range of acceleration factors which differed by several hundred percent, what can lead to a significant overestimate of field fatigue life. Numerical simulations combined with Coffin/Manson or Morrow type criteria underestimated acceleration by up to two hundred percent, i.e. these predictions are on the safe side.