Microvoids are numerous small voids at the interface of a solder joint. Also known as champagne voids, microvoids are less than 40 microns in diameter. They are typically found just above the intermetallic layers between the copper from the circuit board and the tin from the solder paste.

Like all solder voids, microvoids are troublesome when they exist in sufficient number to reduce the cross-sectional area of the joint. In extreme cases, a solder joint can have thousands of microvoids, causing the joint to fail physically and electrically.

Engineers have known about microvoids for years. However, the widespread use of high-resolution X-ray inspection equipment has raised awareness of the problem. Indeed, it's possible that microvoids have always been present in solder joints, and inspection methods were simply insufficient to detect them.

Another reason engineers are detecting more microvoids is related to today's high-density circuitry. Complex electronic assemblies are more difficult to reflow due to wide temperature variations across the PCB. Even with great care, temperature variation may exceed 20 C. To decrease that variation, engineers often reduce overall reflow temperatures, but prolong reflow time. However, in the effort to avoid overheating sensitive components, some areas of the PCB may not receive enough heat. The flux will volatilize, but not all the gas can escape the joint before the solder hardens.

The increase in microvoids is also related to the growing popularity of lead-free solder. The high temperatures needed to reflow lead-free solder exacerbate the problem of temperature variation across the PCB. Moreover, lead-free solder itself may encourage microvoid formation due to differences in surface tension and flux chemistry.

The transition to lead-free solder has brought with it new surface finishes for PCBs. One finish that is growing more popular is immersion silver. Could this finish be contributing to the problem of microvoids? We conducted a study to find out.

Initial Investigation

Our investigation of microvoids and immersion silver was prompted by the high-visibility failure of a ball grid array (BGA) on a PCB during postassembly power cycling. Microvoids were the culprit. But, what caused the microvoids? Several theories were put forth, including:

• insufficient temperature during reflow.
• outgassing of organic brighteners within the underlying copper deposit.
• Impure flux in the solder paste.
• Volatilization of water on the PCB or within the solder paste.
• Excess oxides on the silver deposit.
• Solder mask contamination on the copper.

At first, we thought the microvoids were entirely related to contamination and residue. The failed boards exhibited severe solder mask residue, which limited the available pad area for good soldering and weakened the joints. However, microvoids also appeared in X-ray images of other BGA joints that were not overly contaminated with solder mask.

This raised more questions. Does solder mask contribute to microvoids by providing a source of volatilized organics? Or, do fractures occur only where the solder joint is already weakened by excessive solder mask residue?

The first PCBs with microvoids were traced to a fabrication house with process control issues. Contamination on the boards included oils, developer foam, tin resist and films of sublimated solder mask volatiles. Cross-contamination from shared rinses, conveyors for organic solder preservative (OSP), and nearby immersion tin provided additional residue. Once the immersion silver process was improved, microvoids were nearly eliminated.

Unfortunately, microvoids were far more widespread than we initially believed. We found microvoids on PCBs regardless of which fabrication house or contract assembler they came from. Clearly, the problem could not be entirely related to contamination during fabrication.

One hypothesis was that microvoids were related to inordinately rough surfaces on the copper traces. Just as microvias can create large voids in solder joints, small entrapped air pockets on rough copper surfaces can produce microvoids.

Copper surfaces can be roughened in several ways. The deposition of copper itself can produce rough traces due to variations in current density. Tin stripping and solder mask preparation also can adversely affect the copper prior to surface finishing. Although the surface finishing process employs a cleaner and microetch agent, these relatively mild solutions are not enough to overcome extremely rough or dirty copper.

Reworking the silver finish can also adversely affect the copper surface. Operators often rework the silver finish by processing the PCB repeatedly through the plating bath. This can easily compromise the integrity of both the copper and silver deposits.

Testing With Fabrication Partners

In 2003, a group of companies representing chemical suppliers, board fabrication houses, contract assemblers and OEMs began projects to create microvoids experimentally. Early tests examined several variables at once, including silver thickness, solder pastes, cleaning processes and baking processes.

However, none of the experiments produced significant amounts of microvoids, and the data failed to resolve many of the hypotheses posed before the experiments. In particular, there was not enough data to verify or dispute theories on contamination, oxides, cleaning, chemical concentration and rinsing. The only clear trend was an apparent connection between silver thickness and microvoids.

The next efforts were less ambitious. In Europe, one of our fabrication partners tested theories relating the plating rate of immersion silver to plating quality and the creation of microvoids. Initial work supported the idea that microvoids might be related to efforts to control the plating rate by adjusting pre-dip acidity and silver bath conditions, such as temperature and concentration. This study led to further investigation of using a neutral pre-dip system to control the rate of silver deposition and, therefore, better control the plating thickness.

Next, MacDermid partnered with a U.S. fabricator to test a theory relating microvoids to the age of the silver bath. A simple test vehicle was assembled on production equipment with real components, but did not show any microvoids.

As a part of this study, samples were taken from the silver bath to link microvoids to the concentration of some chemical or contamination. Bath samples from one fabricator in particular were collected daily over a six-month period. However, no relationship was ever discovered. Additionally, samples were collected from chemical suppliers to discover if some unintentional manufacturing or raw material variation might explain the intermittent observation of microvoids. Once again, no such relationship was uncovered.

Back to the Lab

Unable to controllably create microvoids in a production setting, we decided to bring testing into the laboratory. We designed four-layer and six-layer test vehicles for surface-mount soldering of dummy BGAs. These parts could be inspected for microvoids by optical cross-section or X-ray. Ball-shear testing could also be done.

One initial test yielded promising results. A set of simple solderability coupons with BGA pad footprints were plated with increasing levels of silver, screened with solder paste, and reflowed. The thickest silver sample demonstrated a set of small voids just above the copper pad. This was the first time microvoids were observed by cross-section in simple experimentation.

Another leading suspect for producing microvoids was organic content within the circuit board assembly. Studies conducted by Universal Instruments Corp. (Binghamton, NY) linked joint reliability to the brightener content of the copper. The choice and content of organic materials within solder paste flux is well-known to influence voiding. In fact, many solder pastes are specifically formulated to prevent voids under various conditions. Organic material is present not only in the copper plating of outer layer conductors; it is also present in several surface finish deposits.

Several factors led us to suspect that the organic content of immersion silver might lead to the occurrence of microvoids. First, the failed circuit boards that first prompted this investigation demonstrated that solder mask residue contributed to weakness in the solder joints. In addition, reports in the literature linked organic material in gold deposits as a direct cause of microvoids. This organic content was the result of process control problems at the fabricator. Finally, the effect of flux and temperature on the formation of voids in BGAs has been well-documented.

Immersion silver deposits typically contain a small, almost undetectable amount of organic material, which is included to prevent tarnishing, refine grain structure, or stop dendritic growth. Could this minute amount of carbon be enough to outgas and form microvoids?

We put the question to test through the use of thermal gravimetric analysis and gas chromatography of heated samples. Any significant outgassing from the silver deposits would be detected as a weight loss or a chromatographic peak. No outgassing was detected. In contrast, when a silver-plated sample was contaminated with undeveloped solder mask residue, large peaks were observed when the sample was heated to 215 C.

Another characteristic of severe microvoids is their location in a plane just above the junction of the solder and the copper circuit traces. If the silver or organic material were responsible for embrittling this area of the joint, the silver should remain restricted to this area. To see if this was the case, a sample with microvoids was submitted for elemental mapping. The results showed that the silver was completely dispersed throughout the solder joint.

But, if the silver is completely dispersed, how does it produce microvoids? One theory proposed that the silver may cause the flux to exit the solder joint more slowly.

To study flux-silver interactions, solder paste was printed on a various PCBs and heated to 200 C. Escaping gases were trapped, and the gases were quantified with gas chromatography. The solder paste was identical to the one used in the initial assembly failures. A baseline measurement was taken of the volatiles evolved from the paste alone. Additional measurements were acquired with paste screened on silver, OSP and electroless nickel immersion gold (ENIG) surfaces. ENIG and OSP measurements were very similar to the baseline, but the volatiles were significantly lower for the silver samples.

In this experiment, a clear interaction was found between the thick silver deposits and the flux and paste material. Somehow, the silver slowed the formation of flux volatiles. One explanation proposed that catalysis or decomposition of the flux by the silver was enough to slow volatilization. In practice, slowed flux evolution would result in voiding. If a flux interaction occurred at the surface, the voids might be trapped at or near that surface, as seen in the microvoid samples.

With a better understanding of microvoids in the laboratory, screening tests continued at our fabrication partners. One study measured the impact of poor rinsing on microvoid formation. The hypothesis was that fabrication residues interfered with intermetallic formation and released gases during soldering. The experiment also looked at how long the boards were stored between the application of solder mask and the application of silver plating. The storage time may affect how much residue gets locked in from incomplete solder mask developing. The experiment found that poorly rinsed test vehicles had six times the ionic contamination of well-rinsed boards. However, X-ray analysis did not show widespread voiding.

Another experiment yielded one more factor in producing microvoids. Nitric and fluoboric strippers were used to partially strip tin plating before solder mask was applied. In previous experiments, only a very high thickness of silver (0.75 micron) yielded microvoids. In this experiment, a more moderate thickness of 0.4 micron showed microvoids. That sample was processed in the laboratory with nitric acid-based tin stripping. Laser profilometry showed that the surface demonstrated far more roughness than control samples.

One explanation proposed that a thin silver plating with a rough surface contained the same volume of silver as a thick silver plating with a smooth surface. Another important observation from this experiment was that microvoids were produced in the laboratory at low reflow temperatures. In the lab, microvoids were produced at peak reflow temperatures that were 10 to 20 C less than in a production setting.

The Perfect Storm

Our experiments indicated that several factors were necessary to produce microvoids. Intuitively, this seemed correct. A single variable would have been much easier to discover.

The next step was to confirm our theories by designing an experiment that examined all four factors identified with microvoids: flux chemistry, thick silver plating, rough copper traces, and cool reflow temperature. The experiment was dubbed the "perfect storm."

The experiment worked. A set of test vehicles and controls was run, and the samples graded on a scale of 1 to 10. A score of 10 indicated severe microvoids. The test vehicles yielded an average score of 7.1, while the controls produced an average score of 1.1.

Now that we knew how to produce microvoids on demand, we could conduct experiments to:

• rank the leading causes of microvoids.
• identify the significance of secondary factors.
• learn how to prevent microvoids.

These experiments indicated that the factors most associated with microvoids were a cool reflow temperature and a high silver thickness. Other significant factors included the type of copper pretreatment, ionic cleanliness, and the pH of the process chemistry. Very rough copper surfaces-those with a surface area ratio above 0.05-also promoted microvoids.

It should be noted that extreme conditions were needed to produce microvoids on demand. For example, soldering samples at a peak reflow temperature that was 20 C below the recommended value was the most effective way to produce microvoids. Similarly, a silver thickness of 1.5 microns was needed to produce microvoids in the laboratory, even though silver deposits on actual production boards would not normally exceed 0.6 micron.

Preventing Microvoids

In most experiments, very high thicknesses of silver were needed to cause problematic levels of microvoids. However, drastic reduction of immersion silver thickness is not a real option. A plating thicknesses below 0.13 micron can lead to increased tarnishing, migration of copper through pores, and possible loss in contact functions. As long as this lower limit is maintained, thickness can be reduced as much as practical.

A thin silver finish can be accomplished in several ways. First, thickness variation should be reduced by ensuring a uniformly microetched surface. Additionally, chemical suppliers can provide ways to control the chemical bath through pH, temperature and pre-dip to optimize the rate of deposition. Of course, dwell time in the silver bath is still the most direct way to control thickness.

The quality of the immersion silver bath has no meaning if other nearby processes are not well engineered and maintained. The copper must be free of organic and tin residues. The roughness of the incoming copper must not be highly variable. The microetching process immediately before silver deposition needs to produce copper with a clean, uniform surface. Rework should be prevented. If it must be done, rework should be conducted in strict accordance with supplier recommendations. Excessively rough silver surfaces will mimic extremely thick silver deposits. After silver deposition, parts need to be rinsed, handled and packaged with care, since our data show a link between ionic residue and microvoids. For quality acceptance, tarnish should be prevented, even though no link between tarnish and microvoids was discovered.

On the assembly line, a reflow profile with a medium preheat and low ramp rate will minimize microvoids. Low-voiding fluxes and better control of paste storage will also help prevent microvoids.

This article is based on a paper presented at the 2005 Electronic Circuits World Conference, which was held Feb. 22-24 in conjunction with APEX and the IPC Printed Circuits Expo.