Sn whiskers: material, design, processing, and post-plate reflow effects and development of an overall phenomenological theory

Abstract

Replacement of Pb/Sn terminations on electronic devices with pure Sn has proven to be much more difficult than expected. The main problem is Sn whisker formation. Sn whiskers are typically single-crystal, mechanically strong, metallic filaments that can nucleate and grow over time in such a way as to lead to device failure in the field. In this study, we present the results of Sn whisker formation and growth studies for plated matte-Sn on Cu Pb frames. We include data on the effects of Sn-thickness, post-plate annealing, and simulated board attach reflow. Also included are data on two different bath base chemistries, methane sulfonic acid (MSA) and mixed acid, three different plating bath suppliers, and three different lead-frame plating suppliers. The effects of trim and form as well as lead-frame alloy type were also investigated. Whisker testing was completed by storing devices for at least 1 year at room ambient, 60/spl deg/C/93%RH, and 60/spl deg/C/<10%RH. Additional tests were done by subjecting devices to temperature cycling between -55/spl deg/C and +85/spl deg/C. Finally, condensing environmental testing and bias/temperature/humidity tests were completed. The results indicate that the propensity for whisker formation and growth decreases with increasing Sn thickness, and with a post-plate heat treatment. The data imply that the improvement is one of increasing the incubation time, not one of complete elimination of the potential for whiskers. Analysis of the data also indicates the interpretation of Sn whisker formation and growth data taken from a small set of experimental conditions and or small number of devices or leads may lead to incorrect conclusions when compared to a larger data set. When taken in totality, no exact systematic dependency on the propensity for whisker formation was found between dependent variables of trim and formed and nontrim and formed lead frames or for stamped versus etched lead frames. With the exception of 60/spl deg/C/93%RH exposure, no systematic interaction with process conditions and/or process/stress conditions were found. For stressing at 60/spl deg/C/93%RH, there was a clear indication that the trim and form devices were significantly more susceptible to whisker formation than nontrim and formed devices. The propensity for whisker growth was found to be similar at room ambient and 60/spl deg/C/<10%RH. The propensity for whisker formation is lowest for the room temperature and 60/spl deg/C/<10%RH storage, followed by temperature cycling, followed by noncondensing 60/spl deg/C/93%RH storage. A condensing 60/spl deg/C/93%RH environment was found to result in an accelerated corrosion assisted whisker growth mechanism producing locally dense clusters of whiskers or "flowers". No effects of electrical bias on the propensity for whisker growth were observed, and it is concluded that electrical bias itself does not appear to be a promary driving mechanism for Sn whisker growth. Finally, the application of a post-plate solder reflow is found to significantly impact the propensity for whisker growth. In particular, the incubation time for whisker growth in 60/spl deg/C/93%RH was found to be of order 10-15 weeks for all devices subjected to a simulated board-attach reflow process at peak temperatures greater than 232/spl deg/C. Whereas for devices not subjected to a simulated reflow, the incubation time was 26 weeks for 7-/spl mu/m-thick films and 61 weeks for 15-/spl mu/m-thick films. We have observed whisker growth on all samples independent of lead-frame alloy type, plating chemistry, and plating supplier. Finally, we have found that a Ni-underplate, if implemented correctly, eliminated whisker growth for at least 52 weeks at 60/spl deg/C/93%RH independent of simulated reflow process. We present a model that reveals the improvements afforded by the post-plate anneal is one of extending the incubation time not one of complete elimination of whiskers. Finally, we propose a model tha