Thermogravimetry offers the opportunity to examine polymeric formulations by way of a weight-loss curve and also the first derivative of the weight loss. From experience in these laboratories the value of this derivative curve is difficult to overstate. Peak onset, position and end data are all of great value in interpreting the meaning of the weight-loss steps observed in the curve and are usually characteristic of particular formulations, provided that a number of conditions are met.

It is important that sample geometry and weight are kept consistent because a thin planar sample can lose volatiles more rapidly than a cubic one of the same mass, because the surface area available for volatile loss is greater and the mean path to the outside of the sample smaller for the planar sample. Similarly, a small sample can lose each weight-loss step more quickly than can a large sample, partly due to the effect of path length to the outside but also because a large sample requires more heat and is longer to reach the programme temperature. For filled polymers this is important because the samples are usually good thermal insulators. It therefore follows that adequate validation of temperature calibration is also attended to for the same reason. Generally, for most applications, a 10-mg sample, cut as close as possible to a cube, has proved satisfactory.

It has also been found important to carry out separate standard runs for materials which may be presented in both uncross-linked and cross-linked conditions. Cross-linked materials do not melt and flow, and so escaping gases cannot escape so rapidly as those from a fully molten uncross-linked sample. This leads to relative delays in the onset of each weight-loss step in the weight-loss curve and in the derivative peaks for the cross-linked sample as compared to the uncross-linked sample. This observation has proved useful for some formulations in verifying cross-linking or indeed its lack, where this is a specification requirement.

In general, the heating rate must also be carefully specified, since the thermal insulation effect of the sample will be exaggerated if the heating rate is increased. It follows, therefore, that standard reference runs should be carried out if it is desired to heat at high rates. In addition purge gas rate should be standardized, as purge rate will affect removal of the gases liberated from the test sample.

It has proved necessary on many occasions to specify the instrument for use in TGA, as test samples have given slightly different behaviour in instrument of different designs, so that derivative peak onsets, shapes and positions vary very slightly, although the quantitative analysis is consistent between instruments. These variations are presumably due to differences in gas flows and thermal profiles between the designs and do mean that comparison of data obtained on different designs of TGA should be approached with caution.

Many cable sheaths are manufactured from heavily filled polymers, and TGA has been found to be of immeasurable value in studies of such materials and some of their ingredients. The combination of the weight-loss curve, for quantitative analysis, with the derivative, for a qualitative interpretation, has been found to be particularly powerful for these applications particularly when taken together with data obtained by the other thermal methods already discussed but also with that obtained by FTIR spectroscopy. In addition the vinyl acetate content of ethylene–vinyl acetate (EVA) copolymer used in the preparation of some sheathing formulations has been measured and used as a confirmation of grade.

Figure 1 Weight-loss (solid) and derivative (dashed) curves for two EVA copolymers

Figure 1 shows the weight-loss curves and derivatives for a pair of EVA grades; they are readily distinguished. The first weight-loss step is due to loss of acetic acid from the vinyl acetate residues in the polymer and so the vinyl acetate concentration can be calculated from the stoichiometry of the decomposition reaction. The turning point (minimum) between the peaks in the derivative curve is used to define the end of the decomposition of the vinyl acetate for purposes of defining the weight-loss end.

It is also, of course, possible to observe mineral filler decompositions, such as those of aluminum and magnesium hydroxide, which are frequently used in cable sheathing as fire retardants (see Figure 2). Both materials quantitatively decompose to liberate water and their oxides; for example:

The data of Figures 1 and 2 were obtained with nitrogen purging at 20 cm3/min and a heating rate of 10 K/min.

Figure 2 Decomposition of aluminum hydroxide

Examination of Figure 3 shows that, although it is reasonable to conclude that there is much the same amount of aluminum hydroxide present in both samples, based on both decomposition and final ash content, the polymer matrix is different because the escape rate for the water liberated by the aluminum hydroxide is noticeably different in the weight-loss curve and the corresponding derivative curve peaks have different shapes. Furthermore, the inflection regions between the decompositions of the aluminum hydroxide and the polymer are also different for the two formulations, suggesting differences in copolymer. The actual polymer decompositions, whilst superficially similar, are different in detail, and finally curve 1 shows a response due to burning of some residual char when the purge gas is changed to air above 700◦C.

Figure 3 Thermogravimetric data for two alternative formulations of fire-retardant cable sheath. Heating rate 10 K/min and nitrogen purging at 20 cm3/min until 750◦C and then change to air.

Data from FTIR spectroscopy were used to corroborate the interpretation of the first weight-loss peak for each sample as due to aluminum hydroxide fire retardant. Data from both techniques were combined with crystalline melting point data obtained by DSC and TMA, as already discussed to form conclusions about the polymers present. Usually, such conclusions will also be drawn and will be consistent, with known formulation information or with knowledge of production plant conditions leading to the requirement for an investigation.

Thermogravimetry on polymer recovered by solvent extraction has also been found valuable for testing initial interpretations, although sample preparation can prove somewhat time consuming. A second approach is to prepare small experimental samples with which to test the analytical interpretation. The data from these samples should verify a correct analysis or provide further insight for a second iteration of interpretation. The effort to be given to these more laborious procedures will obviously depend on whether a simple survey, or a very detailed analysis, is deemed most fit for the purpose towards which the original analysis was directed.

For many years much cable for house wiring and domestic appliance flexible cords has been manufactured using plasticized and filled polyvinyl chloride (PVC). Calcium carbonate has commonly been used as a filler, with plasticizers chosen from a range to suit specific applications, phthalate esters being a frequently selected solution. PVC itself begins to dehydrochlorinate at just above 200◦C whereas calcium carbonate is stable up to about 600◦C after which it slowly liberates carbon dioxide and leaves a calcium oxide ash. Thus PVC insulation is amenable to examination by TGA, either as a quality check or for purposes of investigating process problems during blending of the ingredients prior to pelletizing ready for subsequent extrusion.

A suitable experiment has been found to consist of using nitrogen as a purge and a heating rate of 10 K/min together with a sample size of 10 mg. The data in Figure 4 are typical. The final step of weight loss in both curves is that of calcium carbonate decomposition, which is shown to vary in temperature and derivative peak shape from sample to sample, with concentration having an effect. This illustrates the point that data obtained from mixtures will not always exactly correspond to a proportional sum of curves for the various unmixed ingredients and that care should be taken to select appropriate reference mixtures for confirmation of peak identity. However, in this case a further effect is that arising from the insulation being a compressed cube, with a small surface area, and the dust having a much larger surface area per unit weight so that the weight-loss peaks occur at lower temperature.

Figure 4 Typical TGA data for PVC insulation (2) and dust found in part of the mixing plant (1).

For investigative work one can arrive at an approximate formulation (since the assumptions made ignore minor ingredients such as stabilizers) by assuming that the final derivative peak is due to filler decomposition and that the second weight-loss step is due to evaporation of the polymer backbone of the PVC. The first weight-loss step then consists of PVC dehydrochlorination and evaporation of plasticizer. It will be appreciated that it is then possible to calculate filler and PVC concentrations from the decomposition stoichiometries and to deduce the plasticizer concentration. For quality control it has been found to be preferable to utilize carefully weighed and mixed standards of known concentration so that a specification based on percentage mass loss for each step can be given and standard derivative curves retained for comparison since such standardization takes account of stabilizer and other additions as well as the major ingredients.

Using these strategies it would be possible to examine Figure 4 and confirm that the insulation was satisfactory, given reference data, but that the dust is an unrepresentative sample of the overall formulation since it is appreciably filler deficient. Such information can prove valuable to plant engineers in the course of investigations to rectify the process problem.

At first sight the idea of rapid heating over a wide temperature range for a TGA experiment would be expected to result in weight-loss steps shifted to higher temperature and broadening of derivative peaks with attendant loss of qualitative resolution. These objections can be overcome by use of standards to calibrate the range for each weight loss and derivative curve peak shape if one is required to validate many samples, and the writer has used this technique on occasions when validation of all polymeric components in a multicable system was required.

On another occasion it was desired to investigate the effect of changed heating rate on the data obtained from some fire-retardant material. Temperature lags for scans carried out at 100 K/min were found to be only of the order of 11◦C, when determined using Curie point standards, whereas the analytical samples gave greater lags, suggesting kinetic limitations to the decomposition or burning rate, depending on purge gas, of the samples.

Recently, this investigation also resulted in the realization that the rate of weight loss for a given step is much greater for a high heating rate than for a low one and so the sort of sensitivity improvement already discussed for high-speed DSC might also be accessible by TGA where analysis of very small samples has proved difficult at conventional heating rates, because signal to noise is poor.

The improvement in sensitivity in the derivative curve is considerable, although the quantitative data presented are rather inconsistent between the four samples. Variability of final ash content has not proved a problem with 10-mg samples and the inconsistency is believed to be a function of the small sample size. Baseline subtraction has been found valuable in allowing for this, as some buoyancy effects will always be inescapable in the TGA experiment. The qualitative data are a great improvement at the higher heating rate for the small samples and therefore leads to an opportunity to carry out sequential FTIR, TMA, DSC and TGA investigation of small difficult samples, which would otherwise be inaccessible.

The advantage of using several analytical methods to attack the identification of unknowns, particularly when small in size and contaminated by the matrix, has already been emphasized. A second possibility, that of investigating the mutual compatibility of the various layered components of cables, has also been exploited from time to time as follows. During a fire the components of a cable will variously melt, burn, deform and decompose at differing temperatures. The possibility of a molten layer interpenetrating a deformed or decomposed layer has, on occasion, been demonstrated in experimental designs, with resultant electrical failure under simulated fire conditions. Identification of fire-damaged material is often ambiguous using only a single technique and, on occasion, it has proved necessary to combine date from several thermal analyses with IR spectroscopy and light microscopy in order to understand the process taking place.

To avoid incompatible materials being placed together in costly prototypes and so avoid this mode of failure, a strategy of material screening was adopted in which interpretations of data from DSC, TMA and TGA experiments were plotted on a temperature axis so that overlaps of different domains, e.g. material A foams whilst material B melts, could be examined. This strategy led to design improvements for some products whilst achieving cost savings in terms of avoiding the production of non-viable prototypes.

The preceding discussion has covered the application of several thermal analysis techniques to a number of problems in polymer science, as it is concerned with electric cable manufacture, but many of these are readily generalized and should be transferable to other technologies. The strategy of attacking a problem with more than one technique, particularly for investigative work, has proved to be invaluable in some of the work described. Additional confidence is obtained from this approach in the investigation of unknowns. This is particularly true of very small, possibly irreplaceable, samples, where sequential analysis in order of increasing destructiveness has proved useful. With these very small samples the use of high heating rates has proved valuable in obtaining an analysis which would have proved impossible at conventional rates.