Tin Whisker Workshop
Overview (from workshop handout)
With the move toward Pb-free electronics, a popular finish for component terminations is 100% tin. However, pure Sn coatings have a tendency to grow small filaments known as "whiskers" that can bridge adjacent terminals causing system failures. The National Electronics Manufacturing Initiative (NEMI) formed three projects to work on this phenomena: Accelerated Test Development (attempt to develop tests to predict tin whiskers); Modeling (attempt to understand basic cause of whiskers); User Group (how can high reliability, long life systems or critical applications be protected with today's knowledge). This workshop reviewed results of three years of work on these projects. The results and the slides used by presenters were assembled into a CD given to attendees and available through NEMI and the CPMT members-only Internet pages.
Approximately 100 attendees gathered for the all day June 1 workshop moderated by Dr. Carol A Handwerker, Chief of Metallurgy for NIST. She pointed out that Tin Whiskers have come again. Tin Whiskers were a plague at first in the 60s but they were quickly minimized by adding some Pb to the "tinning" and in adopting processing parameters that minimized whisker growth opportunities. The legislation of Pb-free is bringing whiskers back out of the tinning. In fact Cd and Zn whiskers were noticed back when vacuum tubes ruled electronics. Some organizations such as NASA have performed detailed analysis on field failures and discovered that $100M systems were destroyed by Tin Whiskers. On the other hand these organized efforts of NEMI to study whiskers were plagued by the fact that you usually can not grow whiskers when you want to.
George Galyon of IBM started the workshop with a history and introduction to a bibliography that his group has put together during this study. One result is a 146 reference bibliography on the NEMI website. The first reference is from 1946 concerning Cd whiskers. In the 50s it was determined that the whisker growth was correlated to compressive stresses in the thin film. However, there was often an incubation time before the whiskers started growing; a time that varied from seconds to years. This made it very hard to do predictive formulas. The studies show the whiskers grow from their base not top. The growth rates vary but 0.1 A/sec is typical. Many studies now point at the micro-stress state as being the driver but there is still good arguments suggesting macro-stress is important. The development of FIB (Focused Ion Beam) cross sectioning has resulted in much discovery of the growth mechanisms. The old polishing sample preparation often changed the morphology of Sn due to simple local heating. Galyon indicated many of the old studies included thousands of samples and more than 10 years of aging. No company or institution has the money to repeat studies of this scale. In fact, probably no company has invested at a rate of more than 1 engineer on this project given today's lean economics. However by using new tools (computer modeling and FIB, for example) we can probably gain enough insight to make safe processes. He also pointed out that the nature of Matte and Bright Tin plating have changed so much over the decades that one can not easily relate to the past studies.
(picture of workshop speakers Joe Smetana and George Galyon)
Joe Smetana of Alcatel discussed the state of Tin coating from the process standpoint. It is pure tin or high tin content that are the risk of whisker growth. By adding enough Pb or Bi such growth is highly discouraged. However, there is a push by manufacturers to use pure Sn if at all possible. The exact finish put on Tin is again a judgement not a science. The best a user can expect in the near future is some mitigation practices including some accelerated testing and process controls on certain steps. Today, there is no standard test that directly relates to field use.
Some options include:
**Using NiPdAu or NiPd plating which are harder to mold to and can be moisture sensitive.
**In the past we added Pb to the Tin to cut way back on whisker growth.
**One can add a non-porous Ni layer between the Cu base and Sn film. Two micron thickness is recommended.
**Fusing Sn plating in hot oil right after the plating process (stress relief?)
**Immersion Sn (not plating) results in a very thin film (can't grow big whiskers) and works for PCBs not components (which flex in application).
**Hot dry Tin (SnAgCu) used in relays and connectors have avoided whisker failure for years (bigger gaps)
**Annealing at 150C for 1 hour of one micron matte on Cu lead frames (another example of a specific solution by controlling a few of the zillion variables).
**Matte tin with larger grain sizes (1-5 microns)and carbon (.005-.05%) which results in lower stresses.
**SnBi alloys (2-10%) suppresses whisker growth. But if any Pb around there is a nasty low-temp phase.
**Thicker Sn finishes to get lower stress. 10 microns nominal if no underplating
**Anything that controls the macro0stress level so that the tin film is in tensile state
**Bias Voltage Testing is not understood but looked at for bright Tin more often
**Alloy 42 lead-frames has not acceptable plating yet -- perhaps low porosity NiPdAu
In all cases more data is needed and since there are lots of variables in whisker growth, a lot more process control is needed by industry. Joe also mentioned that for high frequency circuits whiskers have been more of a field problem.
Next Nhat (Nick) Vo of Motorola-Free Scale spoke on test results.
Their tests have indicated even low levels of impurities in metal
or plating can result in low stress levels and low whisker growth.
In fact, whisker growth appears so multi-factorial that it is
not yet clear what to control and what changes accelerate growth.
However, being a large team, a compromise set of 3 tests were
** -55 to +85 C air to air cycles (20 minute/cycle) for 1500 cycles -- good test for plating on different CTE substrate.
** 60C with 93% RH for 1000 hours -- good test for initial built in stress on common CTE materials
** 20-25C at 30-80% RH for 5000 hours -- good test for ambient storage in field.
Different samples are used for each test (don't put on sample through all tests)
One of the highly interactive audience members (Nancy?) mentioned rules they had developed over the decades. 1.) control of additives are the key to eliminate 75% of the problems, 2.) control of the exact alloy, surface preparation, and cleaning is a good part of the rest, 3.) must anneal right after the plating, not a day later. Some mention was made that use of stronger de-scum steps where appropriate also could greatly decrease whisker growth. Also she mentioned that whiskers became a bigger problem not just because Pb was eliminated as a coating additive but because much finer pitch leads are used today compared with decades ago so even a small whisker can do in a system.
Carrying on in much the same tone, Ichizo Sakamoto of Omron/JEITA gave his paper by phone. He said their work concentrated on gaining repeatability of the Sn chemistry and substrate plating process. For example, there is often Zn in the original Cu leads and it gets to the surface of plating so you want to control it so it is always the same with the same consequences. Similarly in a humid storage environment the Ni underlayer that diffuses to the top will oxidize and this often sparks whisker growth. Again control the process to avoid this stress build up in the Tin layer. Again Ichizo emphasized that temperature cycling will bring out the worst of any mismatch in thermal expansion so pick materials based on temperature excursions expected in manufacturing and use.
Next Mark Dittes of Infineon reported on results that paralleled those from the European "Protin Project". Their studies indicated that the whiskers grow from irregularities of substrate grains on Cu surfaces but CTE mismatch on FeNi42 alloy. For copper a post bake of one hour at 150C forms a transition layer between the irregular grainy copper surface and the tin plating giving an effectively smooth grain boundary and suppressing whisker growth. With alloys such as FeNi42 the CTE mismatch can grow whiskers under temperature cycle even with PbSn plating.
Chen Xu of Cookson Electronics brought out that Sn has the highest CTE of common metals. As a result on Cu one develops -.23 MPa/C or a sum of 14 MPa when heating from 25 to 85 C. Sn on solid Ni results in 0.29 MPa/C or -17.4 MPa build up when reaching 85 C. However, if one uses a NiFe alloy you exceed 50 MPa in thermally generated stress over the same excursion. Since Ni as a substrate or thick plating places the Sn film in tensile stress this is often the best option. If you use a Ni plating make sure it is non-cracking if lead bending takes place.
Peter Bush currently working at the University of Baltimore has the reputation of taking more pictures of tin whiskers than anyone else. Whiskers are a solid crystal of metal with a tetragonal structure. Tin is anisotropic so the crystal walls oxidize at different rates. The whiskers grow from their base and often have greater than 10 grains worth of tin in them. The tin in grain boundaries and on the free surface have higher energy than the atoms within the bulk of a grain. Thus bright tin with the very small grains has a lot more high energy atoms that would favor becoming part of a bulk whisker.
Wan Zhang looked at several intermetalic compound growth patterns and found that the higher the IMC growth rate the less the whisker tendency. With 3.5 microns of IMC you win one year of whisker freedom.
George Galyon retook the microphone to speculate on what may be the story behind whiskers. First he conceded that it is a multi-factorial project and clean data on any factor has been hard to obtain so we will probably never know the real truth only develop time proven methods to mitigate the problem. His FIB views indicate it is not just stress but stress gradients that lead to the whisker growth. Although we think of compressive stress as the problem, tensile with a gradient toward compressive will grow a whisker. He sees Cu diffusing up a grain boundary to get near the surface of the Sn film as the main culprit. This starts the growing boundary at the base of a whisker. With enough Ni layer to cover the roughness of the Cu and treatment that lets the Sn diffuse into Ni rather than Cu into Sn there will be no whiskers. Remember Sn diffuses into Ni faster than Ni into Sn so the action is trapped below the Sn film. There is always CuSn at the root of a whisker, so this intermetalic growth is a clever way to keep the Cu away from the Sn.
Carol Handwerker was not able to show all the stress cantilever measurements made by her group at NIST. She did indicate that though thermodynamics can say what can and can't happen to a thin film, the kinetics driven by local conditions actually show which of the remaining possibilities actually happen. By taking stress measurements during deposition and annealing using cantilever beams NIST has been able to back up some of the rules of thumb that are used in industry. In the short term industry must try to really control all aspects of the Sn process so their results are constant. In addition, trying to keep the stress in the film tensile and keep the Cu from reaching the surface are prudent achievable goals.