Abstract: A simple identification and quantification of clay minerals, particularly those that are responsible for susceptibility of soils to expansion and shrinkage is a continuous focus of research in geotechnical engineering. The visible near infrared and short wave infrared spectral regions are well explored. However little is understood about spectral characteristics of such clay minerals in the wavelength beyond 2.5µm. This paper dealt with exploring potential of laboratory spectroscopy in the wavelength region from 2.5-14µm for characterizing clay minerals. Clay minerals that are established to be important indicators of soil expansion and shrinkage potential: montmorillonite, illite and kaolinite, were investigated. Characteristic spectral signatures and variations in spectral characteristics of these clay mineral mixtures were determined. Partial least squares (PLS) regressions in combination with continuum removal analyses were used to select spectral regions that best discriminate differences in mineralogical contents. Spectral contrast in the 3-5µm wavelength region was high, but generally low in the 8-14µm. Much of the variations in mineralogy were accounted for by the spectral indices (coefficients of correlations of >0.9). Overall, the study can contribute to comprehend spectral manifestations of such clay minerals in naturally occurring expansive soils.
Get your grade
or your money back
using our Essay Writing Service!
Key words: Clay minerals, montmorillonite, illite, kaolinite, spectral characteristics, PLS.
Montmorillonite, illite and kaolinite are the most common soil forming clay minerals (Fitzpatrick, 1980). In geotechnical engineering, they are important indicators of soil expansion and shrinkage potential (Al-Rawas, 1999; Chen, 1988; Fall and Sarr, 2007; Nelson and Miller, 1992; Thomas et al., 2000; Yong and Warkentin, 1975). These clay minerals are distinct in composition, structural arrangements and physiochemical characteristics (Brigatti et al., 2006; Yong and Warkentin, 1975). Montmorillonite has a 2:1 layer structure composed of a single octahedral sheet sandwiched between two tetrahedral sheets. Although the layers are continuous, the bonds between the layers provided by interlayer cations (Farmer, 1974) are very weak permitting water molecules to occupy the space between the layers (Gillot, 1987). Montmorillonite has high specific surface area and a negative layer charge (Brigatti et al., 2006; Gillot, 1987) thereby high affinity to water molecules (Ross, 1978; Skempton, 1984), and large susceptibility to substantial expansion (Al-Rawas, 1999; Karathanasis and Hajek, 1985; Seed et al., 1962; Thomas et al., 2000). Montmorillonite can be formed from parent materials with high levels of calcium ferro-magnesium oxides in low content of silica under favorable environmental conditions; poorly drained environment and seasonally moderate rainfall where evaporation exceeds precipitation (Fitzpatrick, 1980; Gray and Murphy, 2002). Illite has a similar 2:1 layer structure as montmorillonite (Brigatti et al., 2006). Bonding provided by Potassium ions are weak allowing water molecules to be absorbed in between the layers, which can cause this clay mineral to moderately expand (Al-Rawas, 1999; Gillot, 1987; Thomas et al., 2000). Illite has much lesser specific surface area than montmorillonite coupled with lower layer charge deficiency, hence lesser affinity to water molecules (Chen, 1988; Gillot, 1987; Seed et al., 1962; Skempton, 1984). Formation of illite is favored by alkaline to intermediate environmental conditions, in high levels of aluminum and potassium at the expense of calcium and sodium. Illite often occurs as mixed smectite-illite interstratified variety exhibiting a property in between the two clay minerals (Brigatti et al., 2006; Yong and Warkentin, 1975). Kaolinite has a 1:1 layer structure consisting of repetitions of one tetrahedral sheet and one octahedral sheet (Brigatti et al., 2006). Bonding provided by hydrogen molecules is strong which minimizes interlayer space for absorption of water molecules. Kaolinite has a much lesser specific surface area than illite and montmorillonite, and its layers are neutral (Brigatti et al., 2006; Yong and Warkentin, 1975). Thus kaolinite has little water affinity (Chen, 1988; Seed et al., 1962; Skempton, 1984; Yong and Warkentin, 1975) consequently minimal expansion rate (Chen, 1988; Thomas et al., 2000).
Soil expansiveness constitutes a significant challenge in geotechnical engineering (Al-Mukhtar et al., 2010; Al-Rawas, 1999; Kariuki et al., 2004; Morin, 1971; Seco et al., 2011; Shi et al., 2002; Snethen, 1975; Sridharan and Gurtug, 2004; Thomas et al., 2000); and is an inherent property caused by presence of active clay minerals in soils (Fityus and Buzzi, 2009; Seed et al., 1962; Skempton, 1984; Snethen, 1975). Detection of presence of such clay minerals is a key factor for differentiating among potentially expansive soils. Identification and quantification of their abundances on the other hand is essential for rating soil expansiveness. Conventional mineralogical analysis involves X-ray diffraction (XRD) analysis, scanning electron microscopy (SEM), transmission electron microscopy (TEM), differential thermal analysis (DTA), thermogravimetric analysis (TGA), and various chemical analysis methods. These methods are important in research laboratories for exploring the basic properties of clay minerals. But they are costly and necessitate sophisticated laboratory procedures, consequently are not commonly used in soil mechanics laboratories for routine analysis of soil geotechnical characteristics. A simple alternate identification of expanding clay minerals and determination of their quantities is thus a continuous focus of research in geotechnical engineering, among which application of remote sensing techniques.
Always on Time
Marked to Standard
Reflectance characteristics of clay minerals are subjected to intensive research in the visible near infrared (VNIR) and short wave infrared (SWIR) spectral regions (Bourguisnon et al., 2007; Chabrillat et al., 2002; Clark, 1999; Kariuki and Van der Meer, 2003; Kariuki et al., 2003; Kruse, 1991; Kruse et al., 1990; Rowan et al., 1977; Rowan et al., 2003; Van der Meer, 1999; Yitagesu et al., 2009). Particularly in the SWIR, clay minerals exhibit diagnostic absorption features (Clark, 1999; Kariuki et al., 2004; Mustard et al., 2008; Van der Meer, 1999) resulting from vibrational processes related with their structural water and hydroxyl molecules (Farmer, 1974). According to Farmer and Russell (1964) spectral responses of clay minerals generally result from vibrations of structural water and hydroxyl molecules, silicate ions, the octahedral and interlayer cations. Spectral parameters differ depending on chemical composition, structural arrangement and bonding characteristics (Clark, 1999; Van der Meer, 2004) providing a significant potential for discriminating clay minerals (Bourguisnon et al., 2007; Chabrillat et al., 2002; Frost et al., 2001; Goetz et al., 2001; Kariuki et al., 2004; Mustard et al., 2008; Roush et al., 1987; Yitagesu et al., 2009). Absorption feature analysis (Clark and Roush, 1984; Van der Meer, 2004) and multivariate regression analysis (Martens and Naes, 1989; Wold et al., 2001) are extensively used for estimating soil properties (Cloutis, 1996; Gomez et al., 2008; Rainey et al., 2003; Selige et al., 2006; Shepherd et al., 2005; Viscarra Rossel et al., 2006; Waiser et al., 2007) including soil geotechnical characteristics (Kariuki et al., 2003; Kariuki et al., 2004; Yitagesu et al., 2009). However, research in this respect is limited in the spectral region beyond 2.5µm. This spectral region is where molecules exhibit strong fundamental vibrations with high frequency (Arnold, 1991; Clark, 1999; Farmer and Russell, 1964; Ludwig et al., 2008) and is often termed as a fingerprint region (Griffiths and de Haseth, 2007). Farmer and Russell (1964) associated spectral sensitivity of clay minerals in this spectral region to their structural and compositional variations. Furthermore Famer et al., (1974) presented infrared transmission spectra and assigned wavelengths with vibrations of clay mineral constituent molecules. Frost et al., (2001) on their part studied absorbance spectra of sepiolites and palygorskites, and reported that changes in the spectral signatures of these minerals were related to differences in their structural arrangements and compositions. Although transmission and absorbance spectra contain similar spectral information (Michalski et al., 2006), they are not directly applicable in remote sensing where emission or reflectance rather than transmittance and absorbance are detected. Roush et al., (1987) discussed reflectance spectra of kaolinite, montmorillonite and palagonite in the 2.5-4.6µm spectral region; and spectrally discriminated those based on characteristic features ~3µm. Michalski et al., (2006) linked spectral emission features of clay minerals and clay mineral bearing rocks with crystal chemical properties; and reported detection of poorly crystalline clay mineral like minerals in the 6-25µm spectral region (from thermal emission spectrometer, TES data) on Mars that were previously not easily detectable in the VNIR and SWIR regions. Other workers (Cooper and Mustard, 1999; Hecker et al., 2010; Johnson et al., 1998; Salibury et al., 1994; Salisbury and D'Aria, 1992, 1994) demonstrated spectral behaviors of different minerals, and discussed various issues related with spectral acquisitions and interpretations.
The objectives in the current study were: to explore potential of laboratory spectroscopy in the 2.5-14µm spectral region for detection, identification and quantification of clay minerals; and determining characteristic spectral signatures and variations in spectral characteristics of clay mineral mixtures. Experimental investigations were carried out on the three clay minerals that are established to be important with respect to soil expansiveness. This study on pure clay minerals and known mixtures of these clay minerals aimed for future understanding of manifestations of such clay minerals in natural soils. Particular emphasis was given to the 3-5µm and 8-14 µm wavelength regions to give an outlook for a remote sensing implication.
2. Materials and Methods
2.1 Laboratory experiment
Experiments were conducted on pure clay minerals: montmorillonite, illite, kaolinite and their proportioned mixtures. The clay mineral specimens were commercially supplied by VWR international (https://www.vwrsp.com), all naturally occurring clay mineral powder (of particle size <2µm). Mixtures were prepared at a twenty percent by weight increment comprising in the end a hundred gram endmember. The specimens were weighted on a top loading precision balance (of ±1 gm precision, model: Mattler PE 360), then poured in a porcelain mortar bowl and stirred by a spatula for approximately five minutes to get homogenous mixtures. Spectral measurements were done on loosely packed, randomly oriented specimens. Spectra were acquired in the 2.5-14µm spectral region, using Bruker Vertex 70 Fourier Transform Infrared spectrometer (http://www.brukeroptics.com). The spectrometer was equipped with an integrating sphere coated inside with a diffusely reflecting gold surface attached to its external port. Hence enabled directional hemispherical spectral reflectance measurement from which Kirchhoff's law can be used to derive directional spectral emissivity (Johnson et al., 1998; Salisbury and D'Aria, 1994). The spectrometer was configured to provide spectral reflectance with a 4cm-1 spectral resolution with 512 scans per each measurement and eight measurements per individual specimen which were later averaged for a better signal to noise ratio. The spectrometer was continuously purged with nitrogen gas to remove any water vapor and carbon dioxide from the system. Liquid nitrogen was used for cooling the detector. OPUS spectroscopy software version 6.5 (Bruker Optik GmbH, 2007) in a desktop system which is integrated with the spectrometer was used for parameter setting and visualization of the acquired spectra.
2.2 Spectral analysis
This Essay is
a Student's Work
This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.Examples of our work
Two approaches were followed: 1) visual interpretations where trends in overall shapes and reflectivity (Bendor and Banin, 1995); types and number of absorption features; changes in absorption feature spectral indices such as depth intensity, position, width, area and asymmetry (Van der Meer, 1995, 2004) that appeared diagnostic relative to clay mineral proportions were examined. Selected absorption bands ~3.1µm and ~6.1µm were further analyzed quantitatively to establish relationships between absorption feature spectral indices and clay mineralogy. 2) using spectral characteristics as a function of wavelength (Shepherd et al., 2005; Viscarra Rossel et al., 2006; Waiser et al., 2007; Yitagesu et al., 2009) where a multivariate calibration technique partial least squares regression was employed to identify wavelengths relevant for estimating clay mineral contents from the mixtures. The multivariate quantitative analyses were limited in the spectral region 3-5µm and 8-14µm. The analyses were conducted on continuum removed spectra. A classical continuum removal modeled by straight line segments tangential to the spectra (Clark and Roush, 1984; Van der Meer, 2004) which is implemented in ENVI software (ITT Visual Information Solutions, 2009) was used. Van der Meer (2004) reported that continuum removal enhances all absorption features including noise, hence careful selection of spectral regions is important. Owing to differences in spectral contrast in the 3-5µm and 8-14µm spectral regions, continuum removal was done independently. Once spectral subset is limited to the region containing significant absorption features, ENVI uses an algorithm to detect sets of local maxima that maximizes the likelihood of identifying real absorption features (ITT Visual Information Solutions, 2009).
2.3 Partial least squares regression analysis
Partial least squares (PLS) regression analysis is important statistical tool for estimating soil properties from their reflectance spectra (Cloutis, 1996; Rainey et al., 2003; Shepherd et al., 2005; Viscarra Rossel et al., 2006; Waiser et al., 2007; Yitagesu et al., 2009). PLS1 (Martens and Naes, 1989) implemented in The Unscrambler software (CAMO Process AS., 2005) was used to establish relationships between spectral indices and clay mineral contents. The 3-5µm and 8-14µm spectral regions were treated separately. Prior to the PLS regression analyses, distributions of variables were checked and appropriate transformations were carried out on variables that showed skewed distribution. As described in Martens and Naes (1989) and (Wold et al., (2001) data were mean centered and scaled to unit variance prior to calibration, for enhancing variance in the explanatory data and removing any systematic bias. A full cross validation method based on a leave one out principle was used to calibrate and validate the prediction models. PLS component selections were based on evaluation of residual and explained variances and the associated root mean square errors; thus factors with low residual variances coupled with low root mean square errors were selected. The models performances were assessed by various statistics and graphical outputs as outlined in Martens and Naes (1989). Coefficients of correlations (R) and their squared, coefficients of determinations served to evaluate the goodness of fits. Expected prediction errors were assessed by the root mean square errors of predictions (RMSEP). Standard errors of performances (SEP) which were computed as the standard deviations of the residuals, indicated the precisions of the predictions over the whole samples. Bias showed interference errors, and were computed as average values of the variations that were not taken into account by the models. Offset indicated the point where the regression lines crossed the ordinate in the scatter plots summarizing the relationship between measured and predicted values of the response variables, thus showed deviations from the ideal one to one correspondences. Graphical outputs such as score plots, stability plots, scatter plots of X-Y relation outliers, and scatter plots of the measured versus predicted responses etc were used to examine sample distributions, identify samples causing perturbations, detect outliers and assess nature of relationships between predictors and responses (e.g. deviations from linearity), and quality of the regression model in terms of fitting the data respectively. Plots of the B-coefficients were used to visualize significant wavelengths for predicting clay minerals contents; the same were used to build models equations (weighted coefficients Bow and raw coefficients B were identical because no weighting were applied on the variables).
3.1 Spectral characteristics of the clay minerals
3.1.1 Montmorillonite, Illite, Kaolinite
Spectrum of montmorillonite (Figure 2A), illite (Figure 2B) and kaolinite (Figure 2C) contain well-resolved spectral features in the 2.5-14µm. All exhibited strong broad absorption bands in the 2.5-3.7µm wavelength regions. These absorption features are coupled with local absorption minima ~2.75µm and ~3.1µm in montmorillonite and illite, being generally broader in the illite spectrum. In kaolinite it is associated with strongly asymmetric and sharply defined absorption minima centered ~2.75µm.
The montmorillonite spectrum is further characterized by rounded troughs at ~5.1µm and ~5.4µm with the former being more rounded and less pronounced than the later; as well as a prominent absorption band centered at ~6.1µm, which according to Farmer (1974) is typical of water bearing clay minerals and as described by Frost et al., (2001) associated with the bending vibrations of lattice water molecules. Near 9.4µm, montmorillonite exhibited reflectance maxima following an absorption deep ~9µm.
Minor troughs at ~3.48µm and ~4.67µm; doublet feature with absorption minimums ~3.84µm and ~3.98µm respectively with the later being intense and sharp; and a narrow intense absorption band at ~5.56µm followed by a broader absorption band ~6.1µm are additional spectral characteristics in the illite spectrum.
A minor troughs ~3.69µm, and doublet feature with deeps at ~5.2µm and ~5.5µm respectively seem to be diagnostic of kaolinite. In addition the kaolinite spectrum contains minor troughs at ~4.7µm, ~8.6µm, ~9.8µm, ~10.6µm and ~12.4µm; and local reflectance peaks at ~6.3µm, ~9.4µm and ~11.7µm.
3.1.2 Mixtures of Montmorillonite and kaolinite
Overall shapes and reflectance intensities of spectra of montmorillonite, kaolinite and their mixtures exhibited characteristic differences (Figure 3). Depth, position, area, width, and asymmetry of absorption feature in the 2.5-3.7µm showed variations with changes in montmorillonite and kaolinite contents. The sharp asymmetric absorption minima ~2.75µm in kaolinite became less well defined with increasing montmorillonite content. The absorption minima ~3.1µm which is typical of water bearing clay minerals (Farmer and Russell, 1964; Roush et al., 1987) subtly shifted to lower wavelengths as kaolinite content increased. The minor trough ~3.69µm in kaolinite gradually disappeared with increments in montmorillonite content. Slopes of the spectra in the ~3-3.5µm increased as mineralogy changed from montmorillonite to kaolinite. Continuum removal enhanced variations in absorption feature spectral indices in the 3-5µm (Figure 4A) and 8-14µm (Figure 4B) spectral regions; and the changes seem to correspond with mineralogical differences in the clay mineral mixtures. While the depth of the feature centered ~3.1µm increased with increasing montmorillonite content, its position shifted to longer wavelengths. Similarly depth of the water molecule absorption feature ~6.1µm increased with increments in montmorillonite content in the mixtures. In a hundred percent kaolinite specimen spectrum the ~6.1µm absorption feature entirely disappeared. Its position on the other hand slightly shifted towards longer wavelengths with increments in kaolinite contents (Figure 4C). The doublet feature with deeps at ~5.2µm and ~5.5µm in kaolinite became rounded and less pronounced as mineralogy is dominated by montmorillonite. The rounded doublet in montmorillonite at ~5.1µm and ~5.4µm subtly disappeared in spectra of specimens where mineralogy is dominated by kaolinite. Additionally the sharp absorption feature ~9µm slightly varied in depth and ~11.6µm reflectance intensity increased with increasing kaolinite contents.
3.1.3 Mixtures of illite and kaolinite
Spectra of illite, kaolinite and their mixtures (Figure 5) differed in overall shapes and reflectance intensities. Depth, position, area, width, and asymmetry of the absorption feature in the 2.5-3.7µm varied with changes in mineralogy; with the sharp absorption minima ~2.75µm being less well defined with increasing illite contents. Slopes of the spectra in the ~3-3.5µm progressively increased with changes in mineralogy from illite to kaolinite. Plot of continuum removed spectra in the 3-5µm (Figure 6A) illustrated differences in depth and position of the feature ~3.1µm; while the depth increased with increasing illite content, its position shifted to longer wavelengths. On the other hand the minor trough ~3.48µm in illite disappeared in spectra where mineralogy dominated by kaolinite; and the slight deep in kaolinite ~3.69µm disappeared in spectra of specimen containing less than eighty percent kaolinite. Depth of the doublet feature with absorption minimums ~3.84µm and ~3.98µm respectively in illite spectrum lowered with increasing kaolinite. The intensity of the narrow deep absorption feature at ~5.56µm in illite also decreased with decreasing illite contents in the mixtures, missing only in the absence of illite. The doublet feature in kaolinite gradually disappeared as mineralogy is dominated by illite, missing in the spectra of specimens where illite is more than sixty percent. Depth intensity in the water molecule absorption feature ~6.1µm decreased with increasing kaolinite content, coupled with shifts in position towards longer wavelengths (Figure 6C).
3.1.4 Mixtures of montmorillonite and illite
Spectra of montmorillonite and illite exhibited some overlaps, particularly in the water absorption bands centered ~3.1µm and ~6.1µm. Positions of these absorption bands showed subtle variations with changes in concentrations of montmorillonite and illite. Both features tend to be broader with increasing illite contents (Figure 7 and 8A). Spectra of specimens containing illite showed additional slight deep ~3.48µm; and a narrow intense deep at ~5.56µm whose signature disappeared only in the absence of illite. Spectra of specimens containing montmorillonite showed rounded doublet at ~5.1µm and ~5.4µm. The doublet features with absorption minimums at ~3.84µm and ~3.98µm respectively, and the narrow intense absorption ~5.56µm that appeared in all spectra containing illite (lowered in intensity with decreasing illite content) were prominent features for differentiating illite and montmorillonite.
3.2 Clay mineral abundance estimation
Pair-wise correlation analyses showed that mean reflectance in the 2.5-14µm are strongly correlated with clay mineralogy at 0.01 significance levels. Montmorillonite content in the montmorillonite-kaolinite, illite content in the illite-kaolinite, and montmorillonite content in the montmorillonite-illite mixtures were all negatively related with mean reflectance at Pearson correlations of -0.99, -0.98 and -0.90 respectively. These indicated that linear models can best approximate the relationships between spectral reflectance and clay mineralogy. The magnitude of correlation in the montmorillonite-illite mixtures is lower than those obtained from the montmorillonite-kaolinite and illite-kaolinite mixtures, probably due to commonalities in the structure of the two water bearing clay minerals which might in turn influence their spectral characteristics.
Significant wavelengths for estimating the contents of montmorillonite and illite from spectra of montmorillonite-kaolinite (Figure 9A and B), illite-kaolinite (Figure 9C and D) and montmorillonite-illite (Figure 9E and F) mixtures are presented for the spectral regions 3-5µm and 8-14µm respectively. Model performance indices are summarized in Table 1 showing correlation coefficients, root mean square errors of the predictions and standard error of performances, biases and offsets. Much of the variations in clay mineralogy were accounted for by spectral indices. Better performances were achieved in the 3-5µm spectral region than in the 8-14µm particularly for the montmorillonite-illite mixture, where the coefficient of correlation is lower and the model error terms are higher (Table 1).
As in their compositional and structural differences, the clay minerals and their mixtures showed spectrally distinct features. Overall shapes of spectra and absorption bands (Bendor and Banin, 1995) and variations in absorption band spectral indices gave indications of clay mineral contents as described by Clark and Roush (1984) and Van der Meer (1995 and 2004) where these indices are effective measures of mineral abundances. The strong absorption band in the 2.5-3.7µm in the spectrum of montmorillonite (Figure 2A) exhibited local minima ~2.75µm and ~3.1µm showing a slight but clearly defined deep at the former. Hunt (1977) assigned absorption feature near 2.75µm to a fundamental vibration of hydroxyl molecules, ~2.9µm to an asymmetric stretching of hydroxyl molecules, ~3µm features to bending overtones of water molecules and the feature near 3.1µm to hydroxyl asymmetric stretching in water molecules. This feature due to overlapping vibrations of structural hydroxyl and water molecules (Farmer, 1974) appeared as a generally broader absorption in the illite spectrum (Figure 2B). In kaolinite the asymmetric stretching of hydroxyl molecule (Farmer, 1974; Farmer and Russell, 1964) showed a deep and narrow absorption band centered ~2.75µm (Figure 2C). Changes that took place when montmorillonite and kaolinite were heated have been discussed by Roush et al., (1987). They reported the ~2.75µm feature to become narrower in montmorillonite owing to loss of interlayer water molecules; but little change in kaolinite which they ascribed to a relative insensitivity of kaolinite to dehydration due to absence of lattice water. Spectral characteristics exhibited by the clay minerals in this spectral region is in agreement with Farmer and Russell (1964); where they reported clay minerals in which replacement of Al for Si is absent (as in kaolinite) or low (as in montmorillonite) to produce sharper absorption minima ~2.75µm. The generally wide absorption feature exhibited in the illite spectrum can thus be associated with the abundant substitution of Al for Si in this mineral (Yong and Warkentin, 1975).
Spectral features ~3.1µm and ~6.1µm were prominent features in the spectra of specimens containing montmorillonite and illite (Figures 3, 5, 7) owing to presence of interlayer water in their structures (Brigatti et al., 2006). Depth of both features decreased with increments in kaolinite contents among the spectra of montmorillonite-kaolinite (Figure 4A and C) and illite-kaolinite mixtures (Figure 6A and C). This strong negative relationship between kaolinite content and depth intensities in the two absorption bands is attributed to lack of interlayer water in kaolinite. In the PLS models, the significant opposite contributions of the wavelengths from 3-3.2µm in the montmorillonite-kaolinite (Figure 9A), montmorillonite-illite (9E) and illite-kaolinite mixtures (Figure 9C) are similarly related with differences in interlayer water. Absorption due to lattice water vary as a function of amount of water (Roush et al., 1987) which possibly explains the differences in depth intensities of the water absorption bands in the clay mineral mixtures. In the structurally bound hydroxyl molecules coordinated with the octahedral cations (Brigatti et al., 2006; Gillot, 1987), substitutions of cations can result in shifts in positions of absorption bands (Farmer, 1974) as a result of differences in energy levels required for metal-hydroxyl vibrations (Roush et al., 1987). This can probably describe shifts in positions of absorption bands in the clay mineral mixtures as suggested by van der Meer (1995) where shifts in positions are characteristics of cation substitutions at the exchangeable sites. Therefore shifts in positions, in the montmorillonite-illite mixture (Figure 7 and 8A) are probably related with relocation of the hydroxyl absorption band due to substitution of Fe and Mg for Al, or Al for Si (Farmer and Russell, 1964; Yong and Warkentin, 1975). The generally broad absorption feature centered ~3.1µm in the spectra of specimens containing illite might be further associated with distortion in structure to accommodate longer bonds resulting from replacement of Al for Si (Farmer and Russell, 1964; Yong and Warkentin, 1975). Illite exhibited distinct doublet feature with absorption deeps ~3.84µm and ~3.98µm. The unique presence of this doublet feature in all spectra of specimens containing illite (Figure 5, 6A, 7 and 8A) and the narrow intense absorption ~5.56µm which also appear in all spectra containing illite are deduced as diagnostic. These features and the minor trough ~3.48µm are significant PLS regression coefficients in estimating illite contents from mixtures of illite-kaolinite (Figure 9C) and illite-montmorillonite (Figure 9E) indicating their importance for detecting presence of illite.
As established by domain of experts, the 8-14µm spectral region is dominated by features resulting from SiO, hydroxyl and metal-hydroxyl vibrations (Clark, 1999; Farmer and Russell, 1964; Frost et al., 2001; Michalski et al., 2006). Michalski et al., (2006) reported that clay minerals exhibited absorptions centered in the 9-10µm due to SiO stretching, and assigned absorptions in the 10-14µm to metal-hydroxyl bending vibrations. Frost et al., (2001) on their part associated spectral features in the ~8-10µm with SiO stretching and those in the ~10-14µm with hydroxyl bending vibrations. Farmer and Russell (1964) similarly assigned spectral regions from 8-10µm to correspond with (Al, Si)-O stretching vibrations in the tetrahedral sheets and those from 10-14µm to the hydroxyl and metal-hydroxyl bending vibrations in the octahedral sheets. Spectra of the clay minerals investigated here show a sharp absorption feature near 9µm probably due to SiO stretching. But, this feature doesn't seem diagnostic for it showed overlaps in position, depth intensity, width, area and asymmetry particularly among spectra containing kaolinite and montmorillonite, while it appears shallower with its shoulder shifting to the longer wavelength in the illite spectrum. The appearance of this feature in illite might be influenced among other things by structural distortion related with substitution of Al for Si, where it is reported to cause shifts in positions of absorptions to longer wavelengths (Farmer and Russell, 1964; Michalski et al., 2006). Other features centered ~8.6µm, ~10.6µm and ~12.4µm show important spectral variations that seem to correspond with changes in mineralogy (Figures 4B, 6B and 8B); hence were useful in discriminating among the clay minerals as well as in estimating abundances of clay minerals from the mixtures (Figures 9B, 9D and 9F). The features ~10.6µm and ~12.4µm appeared strong and well structured in the presence of kaolinite (Figures 4B and 8B) probably due to higher abundance of aluminum relative to the other cations (Farmer and Russell, 1964) which was described to bring order in the hydroxyl absorption features of kaolinite (Kariuki et al., 2004). Generally apart from furthering the knowledge and understanding of optical properties of clay minerals, the approach demonstrated that laboratory spectroscopy in the 2.5-14µm is a useful technique for characterizing (detecting, identifying and quantifying abundances) clay minerals.
In this study potential of laboratory spectroscopy in the spectral region 2.5-14µm were explored for clay mineralogical discrimination and quantification from spectra of known clay mineral mixtures. The clay minerals exhibited distinct spectral characteristics. Overall shapes of the spectra; positions of the absorption features, their shapes, types and number of absorptions, depth intensities, width, area and asymmetry; differences in slopes of the spectral curves and variations in reflectance intensity were important qualitative parameters that showed variations among the spectra of the clay minerals. Reflectance and spectral contrast differed in the 3-5µm and 8-14µm wavelength regions being higher in the former, but generally low in the later.
Multivariate (PLS) quantitative abundance estimations from spectral parameters as a function of wavelengths were presented. Much of the variations in clay mineralogy were accounted for by the spectral features (Table 1). Significant wavelengths that appeared diagnostic of each clay mineral were identified. The 3-5µm spectral region performed better than the 8-14µm for all the mixtures (Table 1). The spectral region between 5-8µm can also provide important spectral differences among the clay minerals as well as the mixtures (Figures 3, 4C, 5, 6C and 7). This study demonstrated that variations in spectral characteristics of the clay mineral mixtures were mostly due to mineralogical variations. The presented laboratory spectral data on pure clay minerals and their mixtures can provide with essential background knowledge that can help in reasonable composition predictions of expansive soils and thereby making rational spectral-compositional correlations.