Tin Whisker Risk Assessment for Electronic Products

By Tony Fang, Sony Mathew, Michael Osterman and Michael Pecht, Electronic Products and Systems Center, CALCE (Center for Advanced Life Cycle Engineering)

Introduction

Driven by government legislation and market forces, the global electronics industry has been migrating to lead-free electronics [1][2]. Companies who fail to move to lead-free electronics may be excluded from the global markets. In response, a large number of electronic part manufacturers have adopted pure tin and high tin lead-free alloy finishes, as a replacement to lead-alloyed finishes. This selection is based on the low cost, corrosion resistance and compatibility with both lead and lead-free solders.

A major drawback of using lead-free tin finishes is tin whisker formation. A tin whisker is a conductive tin crystal, which grows spontaneously from tin finished surfaces, often in a needle-like form. Whisker related field failures, resulting in millions of dollars lost, have been reported by the electronics industry since 1990[3]. The major failure risks are current leakage and shorting due to bridging adjacent.

CALCE Electronic Products and System Center at the University of Maryland, has developed a methodology and software to assess tin whisker failure risk with time. The methodology uses experimental data and Monte Carlo simulation to probabilistically quantify the risk. It provides a dynamic risk trend with time because the algorithm incorporates distributional data of whisker growth and the distributional data as a function of time [4]. The software was developed with the goal to provide the electronics industry a practical way to assess and predict tin whisker risk for the pure tin and high tin lead-free alloy finished products.



Experiment Samples

In order to experimentally generate tin whisker growth data, samples of matte tin over copper were monitored for whisker length and density parameters. These samples were annealed at 150ºC for one hour one week after plating and were thereafter aged in a temperature/humidity chamber at 60ºC and 95% RH for two weeks. The samples were then stored in a room ambient environment, and whisker length and density data was collected over 18 months. The JEDEC whisker test requirements [5] were used to select the experimental sites to monitor whisker growth and collect data in order to construct a standardized database. The data of whisker length was fitted to a lognormal distribution[6] while the density data is a normal distribution. Growth rates of mean of length and density were then determined from the data.

The growth angle data was fitted as step-wise uniform distribution and uniformly distributed in four ranges of 0 to 20, 20 to 40, 40 to 60, and 60 to 90 degrees with the probability of 0.071, 0.146, 0.244, 0.539 respectively. Based on the observations over a period of 18 months it is also found that the whisker growth angle is independent of time and the rates of mean of length and average density decrease with time. Based on the length and density data and trend whisker growth values for longer time periods was extrapolated.



Methodology to Assess Risk

Tin whisker risk is quantified by the probability of a conductive whisker growing across adjacent electrically isolated conductors, resulting in unintended electrical leakage. The risk assessment algorithm is based on whisker growth characteristics, the geometry of the product at risk, failure criterion and time.

Tin whisker growth parameters include whisker density, whisker length, and whisker growth angle, which refers to the angle between a whisker and its orthotropic projection against the finished surface from which the whisker develops. All of the growth parameters are considered functions of time. The growth parameters are modeled in terms of probabilistic time dependent distributions[6] and quantified based on experimental data. Table 1 shows an example of experimental data for 24 months of tin whisker growth, where measurements were made at regular intervals.

Table 1: Tin Whisker Length Data for Matte Over Copper Finish

Time (Months)	Mean (mm)	Standard Deviation	Maximum Length Observed (mm)
14	7.2	3.3	14
18	13.24	1.41	25
24	15.52	1.68	35.1

The geometry parameters include spacing between adjacent conductors and available conductor area. For the phenomenon of tin whiskering to occur, at least one conductor should have a pure tin or high tin finished surface.

A bridging short is assumed to occur if a whisker has sufficient length and the proper angle to span the space between a defined pair of conductors.

l_w * Sin(q) > l_s           (1)
where q is the whisker growth angle, lw is the length of the whisker, and ls is the pitch spacing between the two adjacent conductors. This definition can also be applied to any shape of surfaces, and is not limited to leaded conductors.

The risk of failure due to tin whiskers, PRi, is defined as the ratio of the number of failures, Nf, per number of potential failure opportunities, Nop, at a particular time. In the algorithm, Nf represents the number of failures and Nop is the sample size for Monte Carlo simulation. The final risk at a particular time is:

P_Ri = N_f / N_op           (2)
It is assumed that the product will fail immediately once a tin whisker bridge occurs, so if a failure occurs during a run of the simulation, the simulation will continue in order to avoid double counting a failure. If there is more than one type of part in a product, assuming no redundancy, the total risk from tin whiskers for the product is:

          (3)
where j is the part type, m is the number of the part type and P_Product is the total failure risk posed by tin whiskers to the product.



Implementation of the Methodology

The methodology was used to assess the tin whisker risk of a printed circuit board assembly (PCB) in operation. The board is a six layered board with components surface mounted on the top and bottom layers. There are a total of 3 QFPs, 3 SOICs, 18 SOTs, 9 diodes, 117 resistors and 43 capacitors. The life requirement of the product is 20 years in an industrial environment.

The conductor pairs analyzed included adjacent terminals (leads) of individual components as well as terminals of components within close proximity of each other. The shortest distance between adjacent conducted was considered for simulation. The area of the conductor pairs used for assessment included the surface area of the side of the lead and the top surface area.

The assessment was performed for matte tin over copper. This assumption was made since the actual lead finishes were unknown and because we have long term test data. Matte tin finish, without an undercoat or without annealing was assessed to provide a conservative risk assessment. The risk of failure due to tin whiskers was also calculated for a period of 5 years, 10 years and 20 years.



Results and Recommendations

For the case of all components having matte tin over copper lead finish the whisker failure risk is 0.05% over a period of 15 years, and is 4% at the 20 year mark. It was recommended that to reduce the risk of tin whisker failure the leads of the QFP, SOIC and SOT components on the printed circuit board could be conformally coated. References

• [1] S. Ganesan and M. Pecht, Lead-free Electronics, 2004 Edition, CALCE EPSC Press, University of Maryland, College Park, MD.


• [2] European Union, "DIRECTIVE 2002/96/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 27 January 2003 on waste electrical and electronic equipment (WEEE)", Official Journal of the European Union, pp. L37/24-38.


• [3] J. Brusse, "Tin Whisker Observations on Pure Tin-plated Ceramic Chip Capacitors", AESF SUR/FIN Proceedings, June 24-27, 2002.


• [4] T. Fang, M. Osterman, S. Mathew, and M. Pecht, "Tin Whisker Risk Assessment ", Circuit World, May. 2006.


• [5] JEDEC Standard JESD22A121, "Measuring Whisker Growth on Tin and Tin Alloy Surface Finishes", JEDEC Solid State Technology Association, May 2005, Arlington, VA.


• [6] T. Fang, M. Osterman, and M. Pecht, "Statistical Analysis of Tin Whisker Growth", Microelectronics Reliability, Vol. 46, No. 5-6, May/June, 2006, p 846-849.