The Materials Analyst, Part 17: Brittle parts, when pigment and polypropylene collide

In the past, we have spent a great deal of time discussing brittle parts. Usually the brittle behavior comes from some form of degradation that reduces molecular weight. This can easily be assessed with a variety of viscosity tests. Other cases have involved contamination. These are typically uncovered by techniques designed to evaluate composition such as DSC, TGA, or infrared spectroscopy. Sometimes, however, neither of these appears as a culprit, yet the parts do not perform as expected. It is especially troublesome when the problem is intermittent.

This was the situation with some thin-walled polypropylene parts that were submitted for analysis. The product had a nominal wall of .02 inch and was produced from a 4-melt homopolymer to which a white concentrate was added. Our client was unusually well prepared, and sent in raw material, concentrate, and three sets of parts from three manufacturing dates. Parts made in January were performing as expected, parts made in April were experiencing frequent brittle failures in the field, and parts from May were an unknown quantity. They had not been put into the application yet, and the client was looking for an assessment of these parts before they went out. We chose to use a combination of DSC and TGA to evaluate the composition of the samples and melt flow rate to check for degradation.

Testing Melt Flow
The melt flow test results showed a difference between good and bad product, but it was not enough to explain the catastrophic cracks that were occurring in the product. Table I shows the results of the melt flow tests. The virgin material was a 4-melt-flow product. With the concentrate added, the result was a slightly higher 4.25 melt flow because the base resin used as the color carrier was a much easier flowing material.

There is an obvious difference between the good and bad product, and based solely on these results, we would expect the third sample of product (which hadn't been used) would more closely resemble the bad product than the good. But are the differences significant? If we apply our rules for allowable melt flow shift, the change from parts to pellets is 18 percent for the good parts, 32 percent for the bad parts, and 30 percent for the parts of uncertain quality. None of these numbers approach the 40 percent level considered to be the line for good processing.

In addition, the only batch of parts that could be directly compared to the base resin came from the May production. The lots of raw material used to produce the January and April parts had been used up long ago, and as is so often the case, no retains were left. Without test values for the specific lots of raw material, we were just guessing about the melt flow rate shift in the other samples. The melt flow rate for this particular material can change from 3.4 to 4.6 on a lot-to-lot basis. With this level of uncertainty, these results alone were not sufficient to distinguish between the different lots of product.

The DSC results for each batch of molded parts were compared to a run on the virgin material. Figure 1 shows the initial melting process for all four samples. There was obviously no problem here. The tests were taken to a temperature high enough to detect any other semi-crystalline polymer; there was no obvious contamination.

Nucleation
In a situation like this, the concentrate is often the culprit. In order to reduce cost, the concentrate formulator will use a carrier with marginal compatibility. In this case, however, the concentrate also checked out as a polypropylene base. But when we cooled the resin from the melt back to the solid, the recrystallization temperatures did vary. Figure 2 shows this result. The virgin material without pigment reaches peak crystallization temperature at 107C. The molded parts display peaks at somewhat higher temperatures.

This effect is typical of nucleation, a process that speeds up the formation of crystals and usually increases the number of sites at which crystallization takes place. DSC cannot see the structural effects of such nucleation, but it can document its occurrence. The good parts peak at 110.5C, but the bad parts show a peak at 114.25C. The parts from recent production fell between these two extremes at 112.5C.

At this point, we had an emerging pattern, but no smoking gun. The good parts had a higher molecular weight and showed less of a departure from the virgin material in terms of crystallization. But we had no explanation for either.

The answers came from the TGA tests. White pigments are almost exclusively based on titanium dioxide, an inorganic material that does not decompose at the test temperatures used in the TGA. Figure 3 compares the weight loss behavior for the virgin material, an unfilled compound, and the concentrate. The residue constitutes more than 48 percent of the concentrate with the rest being polypropylene resin. At the intended ratio of 2 lb of concentrate per 100 lb of resin, a pigment loading in the finished parts of approximately 1 percent was expected.

Figure 4 shows TGA results for parts from January, April, and May. The good parts from January show a residue of .77 percent. A look at the residue under a microscope revealed that the residue was made up entirely of pigment. The April parts left a residue of 2.57 percent, more than three times higher. The May parts show an even higher loading of 4 percent. Because of the thin walls in the parts and the small samples used in the TGA tests, the runs were repeated three times and produced averages of .90, 2.20, and 3.90 percent for the January, April, and May parts.

Pigment Problems
While molded-in color is a distinct advantage in plastics, it is sometimes forgotten that the pigments used to produce the desired colors are, at best, contaminants. Even when properly matched to the resin system, they must be kept below a certain level in order to ensure the material delivers the expected properties. In thin-walled parts, this becomes even more important because the size of the pigment particles can begin to approach that of the nominal wall section. This can magnify the changes a pigment system can cause.

In this case, the additional pigment loading caused a change in the crystallization process in the polypropylene and ultimately reduced the impact resistance of the product. The higher letdown ratio of concentrate also helped to explain the higher melt flow rate of the bad parts. With as much as 10 lb of concentrate per 100 lb of resin in the molded parts, the high flow material used as the base for the concentrate became a significant factor in the final melt flow rate of the molded parts. This eliminated the possibility of degradation during processing and allowed the problem solving to focus on controlling the metering of color concentrate into the process.

Unfortunately, many processors, knowingly or not, use more color concentrate than they need to. In some cases, it is simply poor control, but in many instances, poor mixing caused by worn screws and barrels is covered up with additional colorant. Many of the products made in this way are vulnerable to the undesirable effects of high pigment levels.

With our client, a central question had been the fate of the May production. Based on our results, we recommended this product not be shipped. It looked like it was headed for the same problems as the April parts. Analysis of this kind can be costly but can lead to a better understanding of product failure and provide the processor with the tools to do it right the next time.