Introduction

Fourier Transform Infrared (FTIR) spectroscopy is an established experimental technique for determining qualitative mineral identification, and is currently being developed for quantitative mineralogy. This report describes recent calculations of the mineral content of synthetically prepared mineral mixtures to test FTIR quantitative analysis methods. The weighed amounts ("actual") are compared with the calculated percentages of each mineral. A Non-Negative Least Squares (NNLS) method is used to calculate mineral composition. The ultimate goal is to use quantitative FTIR spectroscopy for data logging the minerals found in cores taken from oil wells.

It is not uncommon for FTIR spectra to exhibit backgrounds of as much as 0.2 absorbance units, a background intensity which varies with wavenumber, and weak absorption bands between 3000 and 1600 cm^-1, which are caused by inexact background subtraction or impurities. These cause little or no problem for routine qualitative research. However, this is not likely to be true for quantitative work. This report investigates the influence of these variables on the calculated mineral percentages.

Two types of spectral manipulation have been used to determine the influence of the variable background and the spurious bands upon the calculated percentages. The variable background has been omitted by subtraction to bring the baselines to zero. The spurious bands are removed by setting the baseline to zero between 3000 and 1600 cm^-1 or between 4000 and 1600 cm^-1. The original spectra (4000-400 cm^-1 with non-zero baseline) are first compared with spectra for which the background has been subtracted and the 3000 to 1600 cm^-1 region set to zero to eliminate spurious bands. The origin of the latter may be residual alcohol (from grinding). (Since there are no major mineral vibrational modes in this region, omission of it does not eliminate useful information.) Secondly we compare spectra for which the entire 4000 to 1600 cm^-1 region has been set to zero (1600-400 cm^-1). The latter removes spurious bands that are believed to originate from water adsorption (from the air).

Background

FTIR spectroscopy relies on detection of molecular vibrations. Mineral identification is possible because minerals have characteristic absorption bands in the mid-range of the infrared (4000 to 400 cm^-1). The concentration of a mineral in a sample can be extracted from the FTIR spectrum because the absorbance of the mixture is proportional to the concentration of each mineral. This is given by Beers Law (see equation below), where A is the absorbance of a mineral mixture at a given wavenumber, n is the number of mineral components, e _i is the absorptivity of component i, l is the absorption path length (pellet thickness), and c_i is the concentration of component i. All multicomponent analyses are based on Beers law, the absorbance at a specific wavenumber is the sum of the absorbance of all sample components that absorb at that wavenumber.

Experimental

Sample preparation for quantitative FTIR analysis is critical to the successful quantitative analysis. The typical sample to KBr ratio used is 1 mg sample/900 mg KBr. This ratio was chosen so that the absorption bands are in the linear region of Beers Law, and so that the signal-to-noise ratio is maximized. Reliable quantitative analysis of samples depends on precise and reproducible sample preparation.

For an accurate quantitative analysis by FTIR, the particle size should not exceed 2.5 mm. This limit coincides with the shortest wavelength radiation used in mid-IR analysis (2.5 mm). Particles that are comparable in size to the wavelength of the light interact with the light, and lead to broadened peak shapes and sloped baselines. When particle size is larger than or equal to the wavelength of the incident infrared radiation, in this case 2.5 to 25 mm, there will be a decrease of the intensity of the absorption band and an increased background. Particle size reduction and sample preparation, therefore, are very important.

Particle sizes of less than 2.5 mm for all samples, both standards and unknowns, were obtained by grinding the sample in a puck mill for 5 minutes. Precisely 1 mg of sample and 900 mg of KBr were weighed out, placed in a capsule, and mixed in a wiggle-bug mixer for 45 seconds. Three 300 mg aliquots of the mixture (sample and KBr) were weighed out, and pressed at 10,000 psi for 10 minutes, under vacuum, to produce 3 pellets

Absorbance FTIR spectra were collected with a Perkin-Elmer 1600 spectrometer. The spectra were collected over the 4000 to 400 cm^-1 wavenumber range, at a resolution of 4 cm^-1, and 48 scans were summed. KBr pellets were included in the reference scans.

A Non-Negative Least Squares (NNLS) fitting method is applied. This procedure is the same as that used in curve fitting. However, we are not fitting a line to a series of points; rather, a complicated curve (mixture spectrum) is fitted by summing a number of other curves (standard spectra). Figure 1 demonstrates how three mineral spectra (standards) can be added to give the spectra of a mixture. The NNLS procedure prohibits calculation of negative contributions to a mixture.

Results and Discussion

The mineral standards include 13 minerals: illite, kaolinite, quartz, opal-a, opal-ct, oligoclase, albite, microcline, chlorite, dolomite, calcite, gypsum, and barite. The absorption spectra of the thirteen mineral standards with non-zero baseline are plotted in Figure 2. The standards with baselines removed are plotted in Figure 3. The spectra of eight mixtures are plotted in Figures 4 - 11. The calculated percentages are listed and compared to the weighed amounts in Tables 1-8.

Weak vibrational bands in the 3000 - 2800 cm^-1 region and the 1800 - 1600 cm^-1 region (see Figures 2 and 4-11(top)) may correspond to C-H stretching vibrations of hydrocarbon film that remains on the sample after it has been ground in alcohol, or to polymer bands from the plastic wiggle bug capsule used to hold the sample during mixing with KBr. The calculated results for these spectra, containing the spurious bands and variable baselines (see Tables 1-8, before subtraction) are not in good agreement with weighed amounts ("actual"). These weak bands and the background variations may contribute to this reduced accuracy. This can be ascertained by eliminating the background and/or the region.

To determine how non-zero baseline regions influence the calculated percentages, we compare calculated compositions using spectra with a non-zero baseline with those calculated using spectra which have had the baseline removed (see Tables 1-8). In Figure 12 the dashed-lines demonstrate how a non-zero baseline is removed. For mixtures 1, 4, 7, 8 and 10, these calculated percentages (after subtraction) are in better agreement than for mixtures 2, 5 and 9. For mixtures 1, 4, 7, and 8 the results calculated using the two different regions with backgrounds subtracted are comparable. For mixture 10, ignoring the 4000 to 1600 cm^-1 region did improve the kaolinite results.

For mixtures 2, 5 and 9, the background subtraction only improved the results slightly. For mixtures 2 and 5, the calculation of the opal-ct and opal-a components are not satisfactory, although the sums of the calculated values approximate the empirical sums (0.60 and 0.72, respectively). This indicates that these two opal standards do not differ very much, as seen if one compares the two spectra (Figures 2K and 2L). Improved standards are being investigated and characterized. High or spurious kaolinite values, spurious gypsum, and low chlorite values typify these three mixtures. Examination of Figure 2D for chlorite shows that it exhibits low absorbance (< 0.6 units) and water vibrational bands in common with koalinite, albite, and gypsum (above 3200 cm^-1). The detailed understanding of the poor agreement observed for the three mixtures containing chlorite awaits further investigation.

Conclusions

For mixtures which do not contain chlorite, removal of background and spurious bands in the 3000-1600 cm^-1 region improves agreement with the weighed amounts. For these samples the calculated percentages are in satisfactory agreement (mixtures 1, 4, 7, 8 and 10). Baseline removal allows more accurate calculation of the mineral percentages. A quantitative measure is needed to evaluate how well a calculated mineral composition fits a given mixture and to quantitatively compare the goodness of fits between data sets using different baselines. Research is in progress to address this need.

.