Infrared Photoacoustic Spectroscopy of Carbon Black Filled Rubber: Concentration Limits for Samples and Background
R. O . C A R T E R III ,* M . C . P A P U T A P E C K , M . A . S A M U S , a n d P . C . K I L L G O A R , J R . Research Staff, Ford Motor Company, Dearborn, Michigan 48121
Infrared spectroscopy studies of the cure chemistry, state of cure, and surface bloom on rubber materials have always been limited by the pres- ence of carbon black in samples. One of the modern methods for re- cording infrared spectra of solid samples is photoacoustie detection Fou- rier transform spectroscopy. It has been demonstrated in the past that surface-segregated species can be identified with this technique, but the results are complicated by the presence of carbon black, which limits the optical depth. As for the stady of bulk chemistry, photoacoustic detection does not require that t]he sample be infrared transparent, and the method can be used with saraples containing as much as a 15 wt % carbon. At loadings higher than 30 wt %, the material becomes a total absorber and can be used to record an instrument background spectrum. Index Headings: Analysis for polymers in the presence of carbon black; Infrared; PAS/FT-IR.
INTRODUCTION
In several recent reports we have used photoacoustic detection (PAS) FT-IR to track the weathering char- acteristics of organic coatings on different substrates ex- posed to natural and to accelerated weathering proto- cols. 1,2,3 These PAS infrared experiments were performed in the usual fashion with the use of carbon black powder to obtain normalized spectra. The effects of low density and increased acoustic cell volume, resulting from the voids in a powder (compared to a solid) have been re- ported. 4 In the course of the above work, it became ap- parent that carbon black: powder was not the best ma- terial to use when obtaining the background spectrum in order to normalize a sample spectrum. The most useful and convenient material was a solid rubber sample con- sisting of 35-40 wt % carbon. The thickness of the rubber sample was adjusted to match that of the samples under study, so that all of the volume constraints of low surface area samples could be matched in the reference material. A detailed theory of the process occurring in a gas-solid PAS cell has been presented by Rosencwaig 5 in an ex- tensive treatise on photoacoustics. This theory, referred to as the Rosencwaig-Gersho (R-G) theory, has been summarized by Griffiths and de Haseth, s and by Mc- Clelland. 7 The four parameters used to characterize the samples for infrared PAS examination are physical thick- ness, optical opacity, thermal thickness, and thermal dif- fusion length (TDL). The definitions of these variables are found in Rosencwaig's book) Carbon black powder is classified as a thermally thin optically opaque sample. 6 Nonpowder solids are generally thermally thick and are transparent if the radiation penetration is much greater
Received 2 May 1989. * Author to whom correspondence should be sent.
than the TDL, or opaque if the radiation penetration is much less than the TDL (i.e., this situation is referred to as photoacoustic saturation). This assumes that the physical thickness is greater than the TDL.
Carbon-filled polymer samples can be used in place of carbon black powder if the resulting material is totally absorbing under the condition of the experiment. There was no evidence of polymeric matrix in the reference spectra used in our earlier studies) ,4 even though the organic polymer network made up more than half the mass of the sample. Matching the volume in the cell while acquiring the background and sample is essential when the acoustic bandwidth approaches or includes the acoustic resonance of the cell. 8,9 When normalization re- quires careful volume control, neither powdered carbons 1° nor DTGS H-13 backgrounds can be used to ratio against, as has been common. The later practice has been re- ported to produce "anomalous bands, shoulders, band- widths, and intensities . . . . -14
The presence of carbon in samples, either as carbon black or graphite, has always presented an extreme dif- ficulty when infrared spectra of polymeric materials were being recorded. The quantity of carbon necessary to com- pletely extinguish evidence of the organic matrix is de- termined in order to establish the minimum amount of carbon needed in a sample for rapid-scan infrared PAS backgrounds and to determine how much carbon can be tolerated in a sample before vibrational information about the matrix is obscured. We have prepared and recorded the PAS-FT-IR spectra of samples ranging from neat elastomer to 25 wt % carbon black filled elastomer in order to track the effect of the composition on the re- sulting spectra. Spectra have also been recorded for car- bon black materials with different particle size and reac- tivity as well as for the rubbers compounded with the use of these carbons. Effects of interferometer mirror speed have also been documented. In addition, we have examined the effect of carbon black on the quantitation of the matrix components.
EXPERIMENTAL
The rubber formulations and technical specifications of the carbon blacks used are given in Table I. Elasto- meric materials were made in 200 g batches on a 200 × 400 mm two-roll laboratory mill at roller separations of 0.5 to 1 mm. The polymer was allowed to band on the slow roller, then activators and antioxidants were added and allowed to disperse. The carbon black was added in small portions over the course of ten minutes. The batch was cut and cross-mixed until uniform, according to spec-
ifications in ASTM D-3182. Curatives were added after the material was allowed to cool for approximately 30 rain. Standard ASTM test slabs were molded at 160°C to 100 % optimum cure as determined by oscillating disk rheometry.
Mid-infrared spectra were obtained with a Mattson FT-IR spectrometer with a water-cooled source. One hundred twenty-eight two-sided 8192-point interfero- grams were collected and transformed with triangular apodization without zero-filling. An EG&G-Princeton Applied Research photoacoustic cell and preamplifier were used. The preamplifier was modified to pass a band- width of 10 k to 6 Hz. A large excess in scan length was established prior to initiation of acquisition to minimize noise associated with mirror turn-around. At each ve- locity, the centerburst was examined and found to be completely stable with respect to position and intensity. Gas-chromatography-grade helium was used to purge the PAS cell for 5 min prior to each collection. The spec- trometer bench was generally purged with nitrogen.
Neat carbon black was contained in the sample cups provided with the cell. Rubber samples were 13-mm- diameter circles, 2-2.5 mm thick, cut from the slabs with a cork borer. These samples were placed on 13-mm-di- ameter steel platforms to produce a uniform sample height of 6 mm.
RESULTS
Six carbon blacks differing in particle size and reac- tivity were examined to determine whether the physical and chemical characteristics would be evident in the spectra of either the pure powders or the corresponding rubber material. Spectra obtained for the six neat carbon blacks are reproduced in Fig. 1. The general profiles of these single-beam spectra are all identical, even though these carbon blacks differ in many characteristics (see Table I). There is no evidence of chemical functionality in any of the spectra.
The negative-going features in these single-beam spec- tra are due to species in the optical path that attenuate
4~ otto
t~
t~
4000 3000 2000 1000 500
Wavenumber FIG. 1. Mid-infrared single-beam PAS spectra of carbon black pow- ders; labels correspond to ID in Table I.
the broad-band source radiation before it reaches the PAS detector/cell. The weak, negative features in the carbon-hydrogen stretch region (3100-2800 cm -1) orig- inate from small amounts of ubiquitous species on the mirrors and beamsplitter surfaces. As these remain con- stant once established, they are to be treated as part of the instrument background. The fine-structured fea- tured features at the high- and mid-frequency regions are water vapor in the spectrometer light path. The diminished downward-going water and carbon dioxide bands in Fig. 1F are a result of a better purge in the spectrometer. Ideally, the water vapor should be com- pletely purged from the spectrometer before data are acquired, but this is admittedly a time-consuming prop- osition.
These same carbon materials have been incorporated at 8.5 wt % in vulcanized natural rubber. The spectra of samples of these carbon-filled rubbers are reproduced in Fig. 2. The general characters of these six single-beam spectra are very similar. These rubber samples, however, are physically thick, about 2 mm, and can more easily dissipate the energy from infrared radiation into their mass with minimal temperature change than the corre- sponding carbon black powders can.
The detailed molecular features apparent in the six spectra in Fig. 2 indicate that, in addition to the polymer matrix, there are several different minor chemical species which reside in the near-surface region of the samples made from the six carbon blacks. These species may arise either from surface segregation, (i.e., "bloom") due to limited solubilities or from different network chemistry induced by the different carbon black structures. Further analysis of this issue will be reported in a future paper. It is clear, however, that, whatever the chemistry in the
APPLIED SPECTROSCOPY 1351
4000 3000 2000 1000 500
Fro. 2. Mid-infrared single-beam PAS spectra, recorded at 0.316 cm-l / s, of carbon-filled rubber with 8 wt % carbon black; labels correspond to ID in Table I.
polymer network is, it does not effect the overall infrared absorption by the carbon black.
A series of five samples were generated with carbon E in vulcanized natural rubber for which the carbon con- tent was varied from zero to 25 wt %. The PAS spectra obtained from these samples are reproduced in Fig. 3 along with the correspondi[ng spectrum obtained for pow- der E. This collection of spectra illustrates the gradual transformation from pure rubber to a carbon-filled com- posite and is representatJive of the results obtained for all six carbons.
The intensity of the polymer component of the spec- trum, Fig. 3F, diminishes rapidly as the carbon black is increased, until the polymer spectrum is negligible for the sample with 25 wt % carbon. At this sample com- position the resulting spectrum, Fig. 3B, looks similar to that of the carbon black powder, Fig. 3A.
The modulation frequency of the infrared radiation is a direct function of the interferometer mirror velocity, s The higher the modulation frequency, the shallower the region of the sample examined. 5 Thus, the faster the interferometer mirror is scanned, the shallower the re- gion sampled. By recording spectra at the extremes of instrument speed, one can obtain a differentiation of bulk and surface species. As will be apparent, this can only be true when the light penetration is greater than the thermal diffusion depth. The spectra reproduced in Fig. 4 demonstrate the spectral results obtained at four different velocities. The two slower velocities produce identical results. This is to be expected, since the acoustic bandwidth in these experiments is well removed and is
Fro. 3. Mid-infrared single-beam PAS spectra of carbon black powder and carbon-filled rubber using carbon E at a wt % carbon of 25% (B), 15% (C), 8% (D), 2% (E); neat polymer (F) and carbon powder (A), recorded at 0.316 cm-1/s.
below the Helmholtz resonance characteristic of the cell. 15 As the mirror speed is increased and the bandwidth opened to include first the wing and then the entire resonance, the effect becomes apparent. Thus, at the higher two velocities, the general profile is seen to change. The general profile changes are not associated with the sample (as it is unchanged) and are not associated with the spectrometer (as it functions extremely well with photoconductive detectors), but are the results of Helm- holtz resonances due to the cell design and sample vol- ume. The frequency of this resonance also depends on the composition of the transfer gas--helium in this case. This resonance is seen at a lower frequency when air is substituted for helium. 4 These results were found to be very reproducible but did depend on sample volume. Care in PAS cell purging and sample sizing must be exercised when spectra are being obtained at these higher speeds in order that spectral comparisons can be made.
DISCUSSION
From the results in Fig. 2 and from results obtained on samples with carbon loadings up to 25 wt %, we believe that there is little reason to differentiate any one of these carbon blacks from another as light absorbers for reference materials in PAS experiments.
When the R-G theory is being considered as a way to account for the PAS signal from a solid in a gas-micro- phone cell, the optical extinction in the sample is seen to effect the PAS signal via several terms in the overall expression of the signal amplitude and phase. 5 The pow- dered samples of carbon black are optically opaque and thermally thin samples. In the context of the R-G theory,
Wavenumber Fro. 4. Mid-infr~ed single-beam PAS spectra of 20 wt % carbon black E filled rubber, recorded at mirror velocities of 0.633 (A), 0.316 (B), 0.181 (C), and 0.115 (D) cm/s.
to be thermally thin, the actual material thickness of the sample must be less than the TDL. The carbon is opti- cally opaque at all wavelengths. The optical penetration into the sample is less than either the physical length or the TDL. As a result, all of the neat carbon black samples present an instrument profile spectrum (Fig. 1).
The use of neat carbon black as a reference material comes into question when the detector/cell volume makes a significant contribution to the resulting spectrum. 4 This is demonstrated with Fig. 4, where spectrum B contains an elevation of the 4000-cm -1 end of the spectrum, and spectrum A has a large hump in the middle of the spec- trum. These features indicate the presence of an acoustic resonance in A or the onset thereof in B. If the sample volume in the detector/cell is not matched when one is recording the spectrum of the reference and of the sam- ple, the ratioed spectrum will not be properly normalized. The low density of carbon powders, approximately 0.4 g/cm 3, is due to extensive voids in the agglomerated par- ticles. The voids add to the gas volume of the cell. In addition to the appropriate volume displacement, the ability of a solid reference material to dissipate heat absorbed from the source radiation is also better. Thus, the ratioed spectra for solid samples and references pre- sent reasonable and appropriate results at all instrument conditions. Thus, the highly carbon-filled rubber sam- ples are recommended for use to obtain reference spectra.
As a practical point, the highly carbon-filled rubber samples are neater and easier to use and produce more reproducible results than do powders. The gain settings required when one is using the highly carbon-filled rub- ber samples result in centerburst signals of the same magnitude as those for other solid samples. The gain needed for carbon black powder samples is much lower. If the gain used to obtain the background with the use of the carbon black powder is the same when the spectra of solid samples are recorded, the resulting centerburst
TABLE II. Relative intensity for the 1446-cm-' band from Fig. 3 and effective depth, in micrometers, in the presence of different weight per- cents of carbon black.
% Carbon Intensity Expected a Ratio (I/E) Depth (~m) b
, Expected calculated from mass proportion of rubber assuming carbon bulk density of 2 g/cm 3 and rubber density of 1 g/cm ~ with constant TDL.
b Depth calculated based on thermal diffusivity of 9.4 x 10 -4 cm2/s from Ref. 5.
signals will leave several bits unused in the analog-to- digital converter.
In a typical solid sample containing little or no carbon black, the sample can be assumed to be thermally thick and optically transparent or nearly so. That is to say, the sample thickness is greater than either the optical length or the TDL and the optical penetration depth is greater than the TDL. This is the R-G "case 2C, ''5 which includes assumptions indicating that the normalized PAS signal is proportional to the extinction in the sample.
In Fig. 3, the underlying carbon black absorption in- creases disproportionately, with increased carbon con- tent. Correspondingly, the weaker extinction from the polymer is diminished very rapidly. Thus we can assume that the presence of the carbon black undermines the proportionality of the signal to the extinction. In Table II a comparison is made between the relative intensity of the 1446-cm-1 band for the spectra of the solid samples found in Fig. 3B-F and the intensity anticipated on the basis of the reduction in polymer mass concentration and increased density but ignoring the extinction of the car- bon black. In all cases the observed spectral intensity is less than the anticipated, and the relative intensity re- duction increases as the carbon content increases.
As in the case of the diffuse reflection experiment, 16,17 the presence of a second absorber with a very large ab- sorption coefficient can have a dramatic effect on the quantification of the weaker species. The effective optical pathlength or scattering distance of radiation will change with the relative concentration of the two materials as well as with the absolute concentration of the two. In these PAS experiments with a concentration range of carbon, the optical penetration becomes shorter than the TDL. It is the TDL which generally determines the PAS response of the sample until all of the available radiation is gone before the TDL is reached. When the TDL is greater than the optical penetration in a carbon-filled rubber sample, the contribution of the weaker absorbing matrix must be drastically lessened. Typically, the TDL is a function of the matrix material which makes up the majority of a sample.
Thus, in the case of the spectra reported in Fig. 3F- B, the sample type in Rosencwaig's categories is changing from a thermally thick sample where the optical depth is greater than the TDL to a thermally thick sample with optical depth less than the TDL. The spectrum of the 25 wt % sample, Fig. 2B, is almost to the point of being an instrument profile spectrum independent of the ex- tinction in the polymer portion of the sample. This is
APPLIED SPECTROSCOPY 1353
expected for a thermal ly thick sample with T D L greater than the optical penet ra t ion depth. Therefore , the pres- ence of carbon black limits the dep th to which infor- mat ion can be gathered. An est imate of the dep th lim- i tat ion can be made from the relative reduct ion of the 1446-cm -1 band intensity. The es t imated T D L in the neat rubber is 5.7 ~m on the basis of the informat ion and formalism in Ref. 5 and t]he mirror velocity used. When one is considering only the reduct ion in rubber concen- t ra t ion and the densi ty of bo th the polymer and carbon, the sample with 8 wt % carbon would be expected to have an intensi ty of 0.96% tha t of the neat polymer. However, t ha t observed is only 80% of the expected intensity. Since the T D L should be very nearly tha t of the polymer alone, the observed intensi ty suggests a sam- ple dep th of 4.7 #m. Similar arguments are used to es- t imate the real dep th to 'which informat ion is obta ined for the other carbon black concentrat ions. These results are listed in Table II. In t]he 25 wt % sample, the sample dep th is es t imated to be only 0.7 #m, at 1446 cm -1. This analysis would indicate tha t there is no reason to scan strongly absorbing samples at anything other than rapid speeds in order to obtain the most scans in a given period of t ime and keep the scan-veloci ty-dependent T D L close to the optical depth. Thi[s suggestion is verified in the da ta in Fig. 4, which shows the 1446-cm -1 band at the same intensi ty at all scan speeds from 0.084 cm/s to 0.633 c m / s .
C O N C L U S I O N S
A reasonable sample to get an appropr ia te ins t rument profile for normalizing spectra of solid samples can be any carbon-filled material which presents no matr ix bands. At a mirror speed of 0.316 cm/s, 35 wt % carbon should suffice, bu t higher carbon loadings may be nec- essary at faster mirror velocities.
Careful calibration with samples of known carbon con- t en t are required, even to approach quant i ta t ion of mo- lecular bands for samples containing carbon black. This does not diminish the excellent quali tat ive results and
relative intensi ty studies possible in controlled experi- ments using FT-IR-PAS. The presence of strong ab- sorbers in samples can reduce the optical dep th in the sample to less than the TDL. In some cases a t low con- centra t ion of carbon black, this can be mit igated by in- creasing the ins t rument scan speed.
In order to distinguish a "b loom" from the underlying bulk polymer chemistry, spectra obta ined at different mirror velocities can be used to indicate which spectral features arise in the near-surface region and which are re la ted to the bulk. However, this ability to distinguish bulk f rom surface is diminished by the presence of carbon black in the sample since the optical dep th is l imited by this strong, universal absorber.
1. R. O. Carter III and D. R. Bauer, Proceedings of ACS Div. of Poly. Mat.: Sci. and Eng. 57, 875 (1987).
2. D. R. Bauer, M. C. Paputa Peck, and R. O. Carter III, J. Coating Tech. 59, 103 (1987).
3. R. 0. Carter III, M. C. Paputa Peck, and D. R. Bauer, Polymer Degradation and Stability 23, 121 (1989).
4. R. O. Carter III and M. C. Paputa Peck, Appl. Spectrosc. 43, 468 (1989).
5. A. Rosencwaig, Photoacoustics and Photoacoustic Spectroscopy (John Wiley & Sons, New York, 1980).
6. P. R. Griffiths and J. A. de Haseth, Fourier Transform Infrared Spectroscopy (John Wiley & Sons, New York, 1986).
7. J. F. McClelland, Anal. Chem. 55, 89A (1987). 8. N. Teramae, M. Hiroguchi, and S. Tanaka, Bull. Chem. Soc. Jap.
10. M. G. Rockley, Appl. Spectrosc. 34, 405 (1980); and M. G. Rockley and J. P. Devlin, Appl. Spectrosc. 34, 407 (1980).
11. J. A. Gardella, Jr., D.-Z. Jiang, W. P. McKinna, and E. M. Eyring, Appli. Surf. Sci. 15, 36 (1983).
12. J.A. Gardella, Jr., G. L. Grobe III, W. L. Hopson, and E. M. Eyring, Anal. Chem. 56, 1169 (1984).
13. D. A. Saucy, S. J. Simki, and R. W. Linton, Anal. Chem. 57, 871 (1985).
14. M. G. Rockley, A. E. Ratcliffe, D. M. Davis, and M. K. Woodard, Appl. Spectrosc. 38, 553 (1984).
15. N. C. Fernelius, Appl. Opt. 18, 1784 (1979); O. Nordhaus and J. Pelzl, Appl. Phys. 25, 221 (1981).
16. P. J. Brimmer and P. R. Grifliths, Anal. Chem. 58, 2179 (1986). 17. J. M. Olinger and P. R. Grifliths, Anal. Chem 60, 2427 (1988).
Deconvolution of Near-Infrared Spectra
T. K. N A D L E R , * S. T . M c D A N I E L , M . O. W E S T E R H A U S , and J . S. S H E N K Agronomy Department, Pennsylvania State University, University Park, Pennsylvania 16802 (T.K.N., M.O.W., J.S.S.); Accoustics Department, Pennsylvania State University, University Park, Pennsylvania 16802 (S.T.M.)
Deconvolution is a mathematical technique for improving the resolution of spectra. Deconvolution can be done in either signal space or frequency space. Both methods were attempted after special adaptations of the algorithms were carried out in order to make them specific for the near- infrared (NIR) spectrum. Signal-space deconvolution works well when a line is subtracted from the spectrum before a modified Jansson's al-
Received 17 February 1989. * Author to whom correspondence should be sent. New address: 214
Biochemistry, Purdue University, West Lafayette, IN 47907.
gorithm is applied. Frequency-space deconvolution requires the addition of a polynomial tail to the spectrum, the addition of a constant to the bandpass divisor, and a Gaussian apodization function.
Index Headings: Near-infrared; NIR; Deconvolution.
I N T R O D U C T I O N
The near- infrared (NIR) region is a quite complex are~ of the electromagnetic spectrum. The re are a great many
