Improving the reliability of PU foam emission testing

The quality of air inside buildings, cars and other enclosed spaces is becoming important, but current methods give very variable results when emission tests for volatiles are carried out on polyurethane samples. Scientists at The Dow Chemical Company have developed a process that is designed to minimise the variability

By An Adams*, Esther Quintanilla, Stefan van Bloois and Adrian Birch

From the smell of a new car to the odour of a new bed-in-a-box mattress, polyurethane products are responsible for a number of emissions in the home and in automobiles.

The indoor air quality of cars, trucks and buildings is significantly affected by volatile organic compounds (VOCs), which are emitted from flexible polyurethane used inside these spaces.

However, there is a great deal of uncertainty when it comes to measuring the emissions from polyurethanes, despite many studies that have tried to improve on this uncertainty.

Often, these studies focus on developing and improving the analytical methods used to measure the emission of volatile substances from materials. They frequently neglect the history and pre-treatment of the samples that are being tested.

One of the most widely used tests is the German VDA 278 method, which is regularly specified to measure the level of volatile compounds in samples of polyurethane. Although it is well suited to the evaluation the emission potential of materials, there are often considerable lab-to-lab differences in tests of the same batch of foam.

The method yields two semi-quantitative values. The VOC value represents the total of the high-to-medium volatile substances. These are calculated as toluene equivalents. The test also produces a FOG value. This is the total of substances with low volatility, calculated as hexadecane equivalents.

Although specific instructions on sample handling and storage are provided in the VDA 278 standard, there is no comprehensive report on the impact of these parameters on the reliability of the VOC emission measurements.

Our approach focused on the effects sample storage and preparation had on VDA 278 test results. However, our findings are expected to be relevant for any selected method for the VOC emission measurement, including chamber, bag or headspace methods.

The VDA 278 test relies on common analytical methods based on thermal desorption, gas chromatography and mass spectrometry (TD-GC-MS). These are sound methods, but polyurethane foam is complex.

Table 1. Where variability comes from in VDA278 tests (relative standard deviation RSD, Test 1)	RSD VOC	RSD FOG
TD-GC-MS
Optimized in-house RSD		Not Rel
In-house, same day repeatability small sample
In-house, same day repeatability large sample	11%	12%
In-house, monthly repeatability large sample	23%	11%
Ring test variation specified in VDA-278 standard, 19 labs
• polyester film	31%	53%
• polyolefin film	36%	52%
Note: For TD-GC-MS analysis of Tenax tubes the VOC/FOG distinction is not relevant

At the scale of the tests, the polyurethane material is not homogeneous. This can lead to variability in results from samples taken across the surface of a foam batch. Additionally, the level of emissions is not stable, and usually falls with time as volatile by-products of the foaming reaction are evaporated. New volatiles can also be formed as a result of ageing processes. We found that it is critical to carefully control the time for which samples are stored and conditioned. We also found that sample preparation is critical to the reliability of VOC emission measurements. However, we found that neither storage temperature nor the quality of packaging in which the samples are stored influenced the final emission values from the tests. This work has enabled us to produce detailed foam preparation guidelines before VOC emission analysis. A summary of our findings is explained in the box on page 32.

Test methods

We carried out two studies. In the first, we tested to see if the level and composition of emissions were stable over a six-month period, and if these were consistent over the surface of a bigger piece of foam (a moulded backrest). In these stability studies, each system was designed to give a representative emission of volatile species from different components in the finished foam such as glycol ethers, volatile tertiary amine catalysts and silicone surfactants. These were not specifically designed to have low emissions. All foams were produced using commercially available materials. The results from this series of tests is shown in Table 1. In the second part, we designed a series of tests to measure the emissions from samples made with one foam formulation. These were taken and stored differently for different periods of time. We evaluated the influence of seven different storage/sampling conditions. Each of these was evaluated at a low concentration (Level 1) and a high concentration (Level 2), in different combinations. By performing a statistical design of experiments, the number of experiments was limited to 32, allowing the importance of all these seven conditions and possible interactions to be evaluated. The details of these factors are shown in Table 2.

Table 2.Factors covered in Test 2
Factor	Level 1	Level 2
Hours between foaming and packing	2	168
Days stored in analytical department	1	14
Storage temperature (°C)	4	20
Hours between unpacking and analysis	1	168
Sample distance from surface (cm)	zero	2
Did sample have skin	Yes	No
Packaging type	Al foil + PE bag	Open PE bag

We used a piece of commercially available statistical software to design the experiments in part two, and analyse the data that the tests produced. In both parts, the foams were analysed according to VDA-278 Thermal desorption analysis of organic emissions for the characterisation of non-metallic materials for automobiles, updated October 2011. The VDA tests themselves measured VOC and FOG levels from the samples of foam.

Test results

Different test results obtained on the same foam can be the result of a number of influences. These include sample history – what temperature and relative humidity were the samples were exposed to, and for how long? We will look at these in more detail later. The analytical method, sample characteristics and sample preparation can also affect test results. The VDA 278 method is difficult to calibrate because the test does not name a reference material. The lack of a suitable reference material with a stable and reproducible emission of volatiles complicates standardisation of VOC emission measurements, and can lead to variable test results between laboratories. We decided to use our own calibration process with a standard control solution. This was a methanol base, into which 17 different compounds specified in the VDA test were dissolved. Our results had a relative standard deviation of between 0.3% and 2% when we analysed the control solution. This gave us confidence in the calibration method, the standard Tenax TA tubes used to process the samples, and the TD-GC-MS analytical procedure. Relative standard deviation (RSD) is a measure of how spread out the results of a series of tests are from the average value. A low standard deviation means that the results are generally close to the average, with a few outliers. A high standard deviation means that the results are much more widely spread.

Variable samples

The samples used in the VDA 278 test are very small: just 15 mg is used in the TD-GC-MS procedure. We assessed the variability of the test by using nine different pieces of 15 mg from a backrest we had moulded. These produced results with a relative standard deviation of 11% for VOC value, and 13% for FOG. This is significantly higher than the analytical error of our standard control solution. The literature shows that, in some cases, the variability of the VOC emission levels from the sample itself exceeds the analytical uncertainty by an order of magnitude.

Table 3. Summary of VOC and FOG data Test 2
	VOC	FOG
Mean	184	640
SD	45	171
RSD, all 32 samples	24%	27%
RSD repetitions, 2 samples each treatment
Note: units are µg total volatiles/g sample.
SD: Standard Deviation, RSD Relative Standard Deviation

Additionally, the level of uncertainty depends on the characteristics of the sample and on the composition of each PU foam. The age of the sample could also have an effect on emissions detected in the test. We checked this hypothesis by repeatedly testing samples taken from a seat over a six-month period. Our tests showed a higher variation on the VOC value (23%) than the FOG value (11%). The VOC measurement looks at more volatile products of foaming than FOG, and so this should be more variable with time. We have summarised the different levels of variability we have identified so far in Table 1. The table shows that the test procedure is reliable, and does not affect the data generated by it. The results of the test are variable because polyurethane samples are not homogeneous and are not stable. Consequently, it is critical to control sample storage, handling, and preparation. These factors significantly influence the results of the test. The results of different packaging methods, conditioning times, storage times, storage temperatures and where in the foam the samples came from are shown in Tables 4, 5 and 6. By analysing our results statistically, we could identify which parameters had a strong impact on the VOC value, and which had not. In decreasing order of importance, storage time and the time between unpacking and analysis clearly showed the strongest contribution, followed by the place that the test sample was taken from in the moulding, the time between foaming and packaging the foam, and whether there was skin in the sample or not. While packaging is important to protect the sample, the type of packaging used and storage temperature of the samples have no significant effect on the VOC results. We found that tests on samples cut from 2 cm below the foam’s surface gave significantly more consistent results than surface samples. RSD for surface foam was 28%, and for deeper foam was 17%.

Table 4.VOC emissions data, differing storage and conditioning times
	All	Days storage			Days unpacked
	VOC	1	14		Zero	7
Mean	184	224	158		204	165
Stdev	45	45	24		49	31
RSD (%)	24	20	15		24	19
Note: units are µg total volatiles/g sample.
SD: Standard Deviation, RSD Relative Standard Deviation

Similarly, packing the sample airtight in aluminium foil (Al + PE) resulted in a considerably higher spread of data (RSD 31%) than a foam sample packed in an open PE bag (RSD 15%). Tight packaging limits initial outgassing, and causes irreproducible accumulation of some specific compounds. This shows that sample conditioning is a key feature of VOC emission measurement methods. Loose packaging allows highly volatile chemicals to escape from samples which may have been packed shortly after production. Consequently, it enables more stable and more representative levels of VOC emission to be measured. This has also been reported for other samples. In addition, carefully controlling the time between sample production and actual emission measurements (within the boundaries of the procedure guidelines) is the most important way to obtain reliable VOC data, helping limit variation in the test results. A slightly different combination of parameters had an effect on the FOG value. Here, the distance from the surface had a large impact as did the presence of skin on the sample. Additionally, storage time had a lower, but still significant effect on the FOG value. Again, storage temperature and packaging type were of low importance. Unlike VOC measurements, the time between unpacking and testing was unimportant, and the time between foaming and transport was also unimportant.

Table 5.FOG emissions data, differing storage and conditioning times and sample position
	All	Skin			Days storage
	FOG	No	Yes		1	14
Mean	640	554	777		681	615
SD	171	85	156		195	155
RSD (%)	27	15	20		29	25
Note: units are µg total volatiles/g sample.
SD: Standard Deviation, RSD Relative Standard Deviation

Time is a less critical factor for the emission of semi-volatile chemicals, which FOG measures, than the highly volatile VOCs. The FOG value was significantly higher for samples including skin than samples without skin (Table 5). This was a result of the accumulation of a tertiary amine catalyst in the skin. There was no significant difference in variability for the samples with and without skin. Comparing samples cut from within the block and those taken from the surface, material cut from within the block generated more consistent data. Emissions were also lower for this group. This was true for both VOC and FOG values. Half of the samples from the surface of the foam contained skin and half did not, and we expected this fact to affect the data. We therefore compared samples cut from the centre of the foam with surface samples that did not have skin. Results showed a more limited but still significant difference: samples cut from the centre showed a lower FOG value, with more repeatable results.

Table 6. VOC and FOG emissions from surface and bulk samples (no skin)
	All	All	Surface		Bulk
	VOC	FOG	VOC	FOG	VOC	FOG
Mean	178	557	188	646	173	512
SD	40	85	58	84	29	37
RSD (%)	22	15	31	13	17	7
Note: units are µg total volatiles/g sample.
SD: Standard Deviation, RSD Relative Standard Deviation

These results indicate that cutting a sample from below the surface of the foam sample is another useful way of to obtain more precise emission data. However, in practice, it makes more sense to sample the surface of a moulded foam (including the skin), and this is also required by most standard procedures. In conclusion, the variability of the data depends on the formulation of each specific sample. For samples that are prepared, stored and conditioned consistently as we have shown, the residual variance, which is unexplained by the seven factors considered, is about 4% for VOC and 2% for FOG. This noise cannot be controlled with the parameters identified here, but it is very low. Through this study, we have effectively explained how the variance in emission tests occurs, outlining ways to significantly improve the reliability of VOC emission data. If you are interested in more information about this article, please contact An Adams. [email protected] Tel: +31 115 671 547 Acknowledgements: Francois Graf, Swee-Teng Chin and Freddy Van Damme are acknowledged for their support to this work.