- © The Mineralogical Society
Coordinated visible/near-infrared reflectance/mid-infrared reflectance and emissivity spectra of four groups of phyllosilicates were undertaken to provide insights into the differences within and among groups of smectites, kaolinite-serpentines, chlorites and micas. Identification and characterization of phyllosilicates via remote sensing on Earth and Mars can be achieved using the OH combination bands in the 2.2–2.5 μm region and the tetrahedral SiO4 vibrations from ~8.8–12 μm (~1140–830 cm−1) and ~20–25 μm (500–400 cm−1). The sharp and well resolved OH combination bands in the 2.2–2.5 μm region provide unique fingerprints for specific minerals. Al-rich phyllosilicates exhibit OH combination bands near 2.2 μm, while these bands are observed near 2.29–2.31, 2.33–2.34 μm and near 2.35–2.37 μm for Fe3+-rich, Mg-rich and Fe2+-rich phyllosilicates, respectively. When a tetrahedral substitution of Al or Fe3+ for Si occurs, the position of the Si(Al,Fe)O4 stretching mode absorption shifts. Depending on the size of the cation, the Si(Al,Fe)O4 bending mode near 500 cm−1 is split into multiple bands that may be distinguished via hyperspectral remote sensing techniques. The tetrahedral SiO4 vibrations are also influenced by the octahedral cations, such that Al-rich, Fe-rich and Mg-rich phyllosilicates can be discriminated in reflectance and emissivity spectra based on diagnostic positions of the stretching and bending bands. Differences among formation conditions for these four groups of phyllosilicates are also discussed. Hyperspectral remote sensing can be used to identify specific phyllosilicates using electronic and vibrational features and thus provide constraints on the chemistry and formation conditions of soils.
The spectral properties of phyllosilicate minerals are diagnostic of chemical composition and crystal structure and, therefore, can be used in remote sensing for detection and identification of phyllosilicates. On this basis, spectral features attributed to vibrations of OH groups bound to Al, Fe or Mg in phyllosilicates have recently been identified on the planet Mars using hyperspectral near-infrared (near-IR) images acquired by the Observatoire pour la Minéralogie, L’Eau, les Glaces et l’Activité (OMEGA) on-board Mars Express (Poulet et al., 2005; Bibring et al., 2006) and by the Mars Reconnaissance Orbiter (MRO) Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) instrument (Mustard et al., 2007). However, detection of phyllosilicates in the Mars Global Surveyor (MGS) Thermal Emission Spectrometer (TES) data has remained elusive (Christensen et al., 2001; Michalski et al., 2005) either due to large spatial averaging or to the challenges in the thermal infrared region presented by fine particle-size components, such as phyllosilicates. The objective of this study is to characterize the spectral features of phyllosilicates from four groups (smectites, kaolinite-serpentines, chlorites and micas) in order to allow better identification of specific minerals and/or types of phyllosilicates on Mars.
Characterization of phyllosilicates on Mars is an integral component of understanding the formation and evolution of the planetary surface. We focused on the visible/near-infrared (VNIR) region from 0.4–4.0 μm measured by Mars Express/OMEGA (Bibring et al., 2005), MRO/CRISM (Murchie et al., 2007) and partially covered by Mars Exploration Rover/Pancam (Bell et al., 2006), as well as the mid-IR region (5–50 μm) covered by MGS/TES (Christensen et al., 2001). This study was carried out in conjunction with analyses of the Mössbauer parameters of many of these samples (Dyar et al., 2008). We dedicate this paper to our colleague Enver Murad who has contributed greatly to the fields of clay minerals and soil science through numerous studies of the spectral properties, chemistry and character of phyllosilicates, Fe oxide-bearing minerals and fine-grained soil components, many of which are likely to be present on Mars.
The structures of phyllosilicate minerals are well-known and have been summarized by Bailey (1980) and Deer et al. (1992). A brief summary of general structural patterns is given to provide a context for understanding the differences in spectral properties described here. All the minerals in this study are composed of octahedral (O) sheets bonded to tetrahedral (T) sheets. Kaolinite and chrysotile (serpentine) have a 1:1 ratio of tetrahedral to octahedral sheets, while all the others have a 2:1 ratio. All layer silicates share a common generalized formula of:
where I is the interlayer, which can be occupied by a cation such as K+, Na+, Ca2+, Cs+, NH4+, Rb+, Ba2+, H3O+, a vacancy or, in chlorite, a sheet of edge-sharing M2+(OH)6 octahedra; M is an octahedral cation such as Fe, Mg, Li, Mn, Zn, Al, Cr or Ti; □ is a vacancy in an octahedral site; T is the tetrahedral cation, commonly Si, Al, Fe3+, B and/or Be; A is F, Cl, OH, O and/or S in all cases except for chlorite (see below); and the subscripts 2–3 and 1–0 refer to variations in the cation occupancy of the octahedral sites.
There is a total of three M sites in the octahedral sheet, generally consisting of two types: one slightly larger M1 site, with two OH ions positioned on adjacent (cis) sides of the octahedra, and two slightly smaller M2 sites, with two OH groups residing on opposite (trans) sides of the octahedra. Due to differences in the geometries of these coordination polyhedra, trivalent cations generally prefer to occupy the M2 sites; therefore, in dioctahedral samples, the two M2 sites are generally filled with predominantly trivalent cations, leaving the M1 site vacant. In trioctahedral samples, all three M sites are occupied by predominantly divalent cations. Mineral structural diagrams are shown in Fig. 1⇓, with one example for each phyllosilicate group presented in this study. Figure 2⇓ shows a diagram of the OH-bending and stretching vibrations and how they are connected to the octahedral cations in the structures. Some phyllosilicates contain water bound to the cations in the interlayer region which is a necessary part of the structure and is distinct from the adsorbed water present on most mineral surfaces. For smectites, this water is more complex because the inner-sphere and outer-sphere water present in the interlayer region typically exhibits different vibrational energies due to differences in H-bonding, proximity to the interlayer cation and tetrahedral surface charge (e.g. Bishop et al., 1994).
Layer silicates are classified on the basis of three main criteria: (1) the ratio of tetrahedral to octahedral sheets; (2) the charge expressed at the interlayer site resulting from cation occupancies in the octahedral and tetrahedral sheets; and (3) the occupancy of the interlayer space.
Representing the kaolinite-serpentine group with one tetrahedral and one octahedral sheet (1:1), we studied a dioctahedral kaolinite and a trioctahedral chrysotile. The octahedral sites are dominated by Al3+ in kaolinite (restricted to the M2 sites) and by Mg in chrysotile (in both the M1 and M2 sites). These minerals do not need interlayer cations because they are already electrostatically neutral; as a result, they contain no molecular water in their mineral structures and have zero layer charge.
The 2:1 smectite group minerals studied here include a montmorillonite, nontronite and a ferruginous smectite (contains both Fe and Al in the octahedral sites). All are dioctahedral in composition, with the M2 site occupancy dominated by the trivalent cations Al3+ and Fe3+. Cation substitutions in the smectite group cause a slight negative charge which is balanced by a small number of cations, such as Mg, Ca and Na, in the interlayer site along with varying amounts of associated H2O. The resultant layer charge is ~0.2–0.6. This group of phyllosilicates is the only one studied here that contains molecular water in the mineral structure.
Micas are also 2:1 phyllosilicates, but have ~25% of the tetrahedral Si replaced by Al. This charge deficit is typically compensated by K, Na or Ca in the interlayer region. Due to their high layer charge (~0.6–1.0), micas are non-expanding and do not have interlayer water as in the case of smectites. Trioctahedral micas include biotite and zinnwaldite, the latter containing Li. Celadonite and glauconite are dioctahedral micas containing Fe3+; they differ in that glauconite has a large amount of Al-for-Si substitution in the tetrahedral sites, with extra octahedral cations and fewer interlayer cations.
Chlorite group minerals have edge-sharing M2+(OH)6 octahedra in the interlayer between each set of 2:1 layers; these are usually occupied by some combination of Mg and Fe2+. These OH groups in the interlayer sheet should have an effect on the IR spectra. Clinochlore and chamosite are the Mg- and Fe-rich trioctahedral species, respectively, in the chlorite group. Chlorites are also non-expanding and contain little or no water with variable layer charge.
Phyllosilicates with a variety of octahedral cation compositions were selected for this study in order to illustrate the effect of these cations on the spectral features used to identify phyllosilicates in hyperspectral remote-sensing studies. Samples were chosen from a large collection of previously-studied smectites, kaolinite-serpentines, chlorites and micas (O’Hanley & Dyar, 1993, 1998; Bishop et al., 2002a,b; Dyar, 2002) and care was taken to select pure samples. Except for the chrysotile fibres, all samples were drysieved to <45 or <125 μm prior to measurement of the reflectance and emission spectra. The particle size and structural formulae for these samples are summarized in Table 1⇓, with the sample chemistry given in Table 2⇓; further details can be found in Dyar et al. (2008).
The smectites used were contributed by the Institute of Inorganic Chemistry at the Slovak Academy of Sciences and had been prepared and characterized as part of previous studies (Novak & Číčel, 1978; Číčel & Komadel, 1994; Madejová et al., 1994). The SAz-1 montmorillonite and SWa-1 ferruginous smectite are from The Clay Minerals Society Source Clays Repository. The SAz-1 montmorillonite was found to be 99% smectite by X-ray diffraction (XRD) (Chipera & Bish, 2001). The nontronite is from Sampor, Slovakia. The samples were fractionated to a particle size of <2 μm and the interlayer cations were exchanged to Ca2+ in all three samples.
The kaolinite KGa-1 sample was characterized in detail as a Clay Minerals Society standard (Mermut & Cano, 2001). Chipera & Bish (2001) found that this contained ~96% kaolinite with ~3% anatase and ~1% crandallite/quartz impurities. The purity of the chrysotile sample was verified with XRD and electron microprobe (EMP) studies when the sample was characterized for earlier studies (O’Hanley & Dyar, 1993, 1998).
The glauconite is from Hurricane Mountain, which is one of the first non-marine, Li-rich glauconite localities to be described (Smith, 2005). This sample was collected by Carl Francis of the Harvard Mineralogical Museum from the White Mountain plutonic suite pegmatite associated with the Conway granite near North Conway, New Hampshire. The zinnwaldite and biotite mica-Fe are reference mica samples prepared by the Centre de Recherches Pétrographique et Géochemiques (Govindaraju, 1979; Govindaraju et al., 1994). The biotite was collected from a two-mica granite in the Massif Central, France, in the Massif of Saint-Sylvestre. Chemical data are summarized in Govindaraju (1979). Our particulate, <45 μm grain-size sample is similar to the ‘Fe-1’ sample of that study. Chemical data on the zinnwaldite sample are summarized in Govindaraju et al.(1994). The celadonite sample is from California and was prepared for an earlier study (Reid et al., 1988). The sample was purified by extracting carbonate and soluble salts, with the organic and the soil fractions separated using sediment fractionation. X-ray diffraction showed the sample to be primarily celadonite with minor admixtures of cristobalite, stilbite and trace smectite.
Visible/near-infrared (VNIR) and mid-IR reflectance spectra were measured at the Reflectance Experiment Laboratory (RELAB) at Brown University, USA. Bi-directional VNIR spectra were measured relative to Halon under ambient conditions, while biconical on-axis reflectance spectra were measured relative to a rough gold surface using a Nicolet 740, while a Nicolet Nexus Fourier transform infrared (FTIR) spectrometer was used to measure the off-axis reflectance spectra. Only small differences in the on- and off-axis spectra were observed and are not the focus of this study. The particulate samples were poured into dishes ~10 mm in diameter and 2 mm deep, tapped gently to settle the particles, then filled to an even level (the surface was not scraped). The samples were placed in a H2O- and CO2-purged chamber for at least 10 h prior to and during measurement in order to remove adsorbed water from the samples. Composite, absolute reflectance spectra were prepared by scaling the FTIR data to the bi-directional data near 1.2 μm. The spectral resolution is 5 nm for the bi-directional data and 2–8 cm−1 for the FTIR data.
In order to better characterize and compare the NIR-region OH combination bands, a continuum was removed over the region 1.8–2.6 μm using the MR PRISM program described in Brown & Storrie-Lombardi (2006). Each spectrum was subset to the desired range and automatically processed to remove the overall continuum, or spectral slope. A straight-line continuum was calculated between the extreme convex points on the spectrum and the original spectrum was divided by this continuum in order to obtain the ‘continuum-removed’ spectrum.
Emission spectra were measured at the Mars Space Flight Facility at Arizona State University using a Nicolet Nexus 670 E.S.P. FTIR spectrometer which was modified for emission measurements and equipped with a thermoelectrically stabilized deuterated triglycine sulphate (DTGS) detector and a CsI beam-splitter which allows the measurement of emitted radiation from heated samples over the mid-IR range of 2000 to 200 cm−1. To minimize the amount of water and maintain constant CO2 vapour inside the spectrometer and external sample chamber and glove box (and to minimize the degradation of the hydrophilic CsI beam-splitter) the entire system was scrubbed continuously using a compressed air and gas in-line filter. Spectra of loose particulate material were measured in a sample dish.
VNIR spectra of phyllosilicates
Laboratory reflectance spectra (in the range 0.3–3.3 μm) of the four groups of phyllosilicates in this study are shown in Fig. 3⇓ to illustrate the major spectral features observed in VNIR remote sensing. This region is dominated by strong water and OH-stretching vibrations from 2.7 to 3.1 μm and the H2O bending overtone near 3.1 μm. Additional features are observed from 0.4 to 1.2 μm due to Fe electronic transitions and from 1.35 to 2.5 μm due to overtones and combinations of the fundamental OH and H–O–H vibrations (e.g. Bishop et al., 1994).
In order to facilitate comparison of the combination bands in the 1.8–2.5 μm region, the spectral continuum was removed from 1.8 to 2.6 μm for all spectra using a straight-line continuum across the extreme convex points. This procedure is illustrated in Fig. 4⇓, where the downward slope at larger wavelengths caused by the strong H2O absorptions near 3 μm is removed. Shifting the positions of the tie points of the continuum changes the shape of the continuum-removed spectra, as observed by Milliken & Mustard (2007). Continuum-removed spectra are shown for each of the samples in Fig. 5⇓.
Spectral features are observed in the extended visible region (0.3–1.2 μm) due to electronic vibrations arising from crystal-field bands and charge-transfer bands in iron-bearing samples (e.g. Burns, 1993). Figure 6⇓ displays an expanded view of the extended visible-region spectra where these Fe bands occur. Nontronite reflectance spectra exhibit absorptions centred near 0.5, 0.65 and 0.95 μm. Chamosite, biotite and glauconite have strong absorption bands from 1.0 to 1.2 μm that correspond to the large amounts of Fe2+ in these samples (see Figs 3⇑ and 6⇓).
Overtones and combinations.
Sharper features between 1.35 and 2.5 μm in phyllosilicate reflectance spectra are due to combinations and overtones of the OH bound to octahedral cations and the H–O–H bound to interlayer regions or adsorbed on mineral surfaces (e.g. Clark et al., 1990; Bishop et al., 1994; Bishop et al., 2002a). The OH-stretching bands occur near 2.7–2.8 μm (~3570–3680 cm−1) and the OH-bending vibrations occur near 10.9–12.6 μm (790–915 cm−1). The symmetric and asymmetric H2O-stretching bands occur from 2.76–3.0 μm (~3340–3620 cm−1) and the H2O-bending bands occur near 6.0–6.2 μm (~1620–1650 cm−1).
As shown in Figs 3⇑ and 5⇑, the smectites predictably exhibit the strongest water overtones near 1.41 μm and water combination bands near 1.92 μm, while the kaolinite-serpentines exhibit the strongest OH overtones near 1.38–1.41 μm. The spectra of smectites and kaolinite-serpentines both have strong OH combination bands from 2.2–2.5 μm. These NIR combination bands have been investigated recently in terms of the individual OH stretching and bending modes as a function of variable octahedral cation composition (Bishop et al., 2002a; Bishop et al., 2002b; Gates, 2005). The OH combination bands observed in continuum-removed spectra in our study of several phyllosilicates are listed in Table 3⇓. In many cases these are sharp bands which are diagnostic of specific minerals. In general, the OH combination bands occur due to the two Al cations in the octahedral sites near 2.21 μm, while the OH combinations due to Fe and Mg cations occur near 2.29 μm and 2.32 μm, respectively.
Vibrations due to the OH overtone have been investigated recently by Bishop et al. (2002a,b). Near-infrared bands due to the Al2-OH stretching overtone are observed as a single band at 1.41 μm (7090 cm−1) for montmorillonite and as a triplet at 1.395, 1.405 and 1.415 μm (7160, 7120 and 7070 cm−1) for kaolinite. This overtone occurs at 1.43 μm (~6980 cm−1) for Fe2-OH in nontronite and at 1.38–1.39 μm (7230–7170 cm−1) for Mg3-OH in serpentines, with H–O–H stretching overtones also observed in this region for smectites: 1.41 μm for bound water and ~1.46 μm for adsorbed water. Related bands due to H–O–H combination stretching and bending vibrations occur near 1.92 and 1.97 μm, respectively. Weak H–O–H bands are observed near 1.92 μm in the kaolinite, serpentine and celadonite spectra, although H2O is not part of the mineral structure as in the case of smectites. This band is probably due to a small amount of H2O molecules trapped at grain boundaries or associated with impurities in these samples. Weak H–O–H combination bands are observed near 2 μm for adsorbed water in spectra of the serpentine and chlorites.
Near-infrared bands due to the OH stretching and bending combination vibrations in reflectance spectra of phyllosilicates have been characterized in a number of spectral studies (e.g. King & Clark, 1989; Post & Noble, 1993; Petit et al., 1999; Bishop et al., 2002a,b; Gates, 2005). They are observed for Al2-OH near 2.21 μm (~4530 cm−1) for montmorillonite and as a doublet at 2.17 and 2.21 μm (~4615 and 4530 cm−1) for kaolinite. The AlFe-OH and Fe2-OH combination bands in the nontronite spectra are observed at 2.24 and 2.29 μm (~4475 and 4375 cm−1) and related bands near 2.25, 2.35 and 2.40 μm (~4435, 4255 and 4175 cm−1) in biotite spectra. The Mg3-OH combination bands in serpentine spectra are observed at 2.32 and 2.34 μm (~4300 and 4270 cm−1) and in chlorite spectra near 2.33–2.36 μm (~4290–4240 cm−1). Chlorites also exhibit an additional band near 2.25–2.26 μm (~4450–4420 cm−1) due to either AlFe-OH or AlMg-OH.
When a continuum is removed from the reflectance spectra, more accurate band centres can be assigned, allowing resolution of additional bands near 2.4–2.5 μm. These bands are probably also due to OH stretching and bending combinations and may be associated with tetrahedral substitution of Al or Fe for Si or with bonding between the octahedral cation and the tetrahedral cation (Bishop et al., 2002a) and Gates (2005).
The OH-stretching bands occur in the range 2.7–2.85 μm (~3500–3700 cm−1) for phyllosilicates and are well documented in the literature for transmittance spectra (e.g. Farmer, 1974; Madejová & Komadel, 2001). They occur at ~2.75 μm (3630 cm−1) for montmorillonite, ~2.80 μm (3570 cm−1) for nontronite and near 2.72 μm (3680 cm−1) for Mg smectites. They occur near 2.69 μm (3710 cm−1) for Al-rich trioctahedral micas such as phlogopite and near 2.76 μm (3620 cm−1) for Al-rich dioctahedral micas such as muscovite (Beran, 2002). Four OH stretching vibrations are observed in well ordered kaolinites (Farmer, 1998): at 2.705, 2.725 and 2.738 μm (3697, 3670 and 3652 cm−1) for the inner OH groups and at 2.76 μm (3620 cm−1) for the outer OH groups. Fe-bearing kaolinites exhibit bands due to AlFeOH stretching at 2.772 μm (3607 cm−1) (Mendelovici et al., 1979). Serpentines exhibit a pair of OH-stretching vibrations near 2.74 μm (3644–3652 cm−1) and 2.71–2.72 μm (3682–3689 cm−1), plus weaker bands at lower energies (Bishop et al., 2002a). Chlorites have two broad bands near 2.79–2.81 μm (3560–3586 cm−1 ) and 2.91–2.94 μm (3400–3436 cm−1) attributed to the interlayer OH groups and a third weaker band near 2.72–2.76 μm (3620–3680 cm−1) assigned to the mica-layer OH groups (Hayashi & Oinuma, 1967).
The OH-stretching bands are observed in similar positions in reflectance spectra and are listed in Table 4⇓. The same trends observed in reflectance spectra are also found in the transmittance spectra. However, some of the bands are more difficult to assign in reflectance spectra without modelling or computational chemistry because of the overlapping H2O stretching bands in this region, especially for the smectite and chlorite groups. The OH-stretching band shifts from near 2.75 μm in montmorillonite to 2.80 μm in nontronite, and from 2.72 μm in clinochlore to 2.76 μm in chamosite. The kaolinite-serpentine and mica samples exhibit much weaker water vibrational bands near 2.8–3.0 μm and very weak water combination bands near 1.9–2.0 μm. For these minerals, multiple OH-stretching vibrations can be resolved.
Mid-infrared spectra of phyllosilicates
A comprehensive study of the transmittance spectra of phyllosilicates was performed by Farmer (1974). Reflectance and emission spectra differ from transmittance spectra in that only the imaginary component of the complex index of refraction (k) is expressed in transmittance, while both the real (n) and imaginary (k) components are expressed in reflectance and emission spectra, which are thus susceptible to the effects of surface scattering. The best method for comparisons of lab data with remote sensing data is to use emission spectra (e.g. TES and mini-TES), as this is the method in which mid-IR remote sensing data are measured. Reflectance spectra can often be used to approximate emission spectra using Kirchhoff’s Law (E = 1–R). However, this does not always hold true for small particle-sizes or if the surface is not a Lambertian scatterer, particularly for biconical reflectance (e.g. Salisbury et al., 1994). For finely particulate samples such as phyllosilicates, it is often easier to obtain high-quality reflectance spectra in the lab than emissivity spectra. We have measured both biconical reflectance and emissivity spectra in order to compare these techniques. Reflectance and emissivity spectra of the smectite group, kaolinite-serpentine and chlorite groups, and micas are shown in Figs 7⇓, 8⇓ and 9⇓, respectively. The mid-IR features observed in reflectance spectra from ~ 9–40 μm (250–1150 cm−1) are recorded in Table 5⇓.
As reported by Farmer (1974), the strongest transmittance IR bands observed for phyllosilicates are due to SiO4 and M2 or 3 –OH (M = Al,Mg,Fe) vibrations. The SiO4 stretching and bending vibrations occur near 9–10 μm (~1000–1100 cm−1) and ~20 μm (500 cm−1) in silicate minerals and are both influenced by substitution for Si in the tetrahedral sites (Lyon & Tuddenham, 1960; Farmer & Palmieri, 1975).
Si–O stretching vibrations.
The strongest bands for identification of phyllosilicates in thermal remote sensing result from the tetrahedral sheet Si(Al,Fe)O4 vibrations (Michalski et al., 2006). However, these SiO4 vibrations are not unique to phyllosilicates and must be coupled with other features for definitive identification of phyllosilicates. Michalski et al. (2005) observed a shift in the emissivity minimum near 1000 cm−1 towards smaller wavenumbers (larger wavelengths) with increasing amounts of Al and Fe and decreasing Si. This is consistent with the trends observed by Salisbury et al. (1991) for a large collection of silicate minerals and with observed transmittance spectra of phyllosilicates (Farmer, 1974). For example, Michalski et al. (2005) noted emissivity minima of 8.8 μm (1135 cm−1) for SWy-1 montmorillonite with an Si/O ratio of 0.399, and of 9.5 μm (1056 cm−1) for NAu-1 nontronite with an Si/O ratio of 0.349.
The Si(Al,Fe)O4 vibrations occur as reflectance peaks or emissivity minima in Figs 7⇑–10⇑⇑⇓. These are shown in plain (non-bold) text in Table 5⇑. As expected, a trend in the Si(Al,Fe)O4 stretching vibrations was observed. Shifts in the reflectance peaks were noted from 8.77 and 9.39 μm (1140 and 1065 cm−1) for SAz-1 montmorillonite to 8.84 and 9.57 μm (1130 and 1045 cm−1) for Sampor nontronite based on the Al and Fe substitution for Si and influence of the octahedral Al and Fe cations on the tetrahedral SiO bonds. For KGa-1, kaolinite reflectance peaks occur at 8.8, 9.3 and 10.2 μm (1135, 1070 and 980 cm−1), while the Mg-rich chrysotile-serpentine peaks are at 9.0, 9.9 and 11.4 μm (1112, 1010 and 880 cm−1). For the chlorites a shift was observed from 9.3 and 9.7 μm (1072 and 1030 cm−1) for the Mg-rich clinochlore to 9.5 and 9.9 μm (1055 and 1010 cm−1) for the Fe-rich chamosite. These bands are more complex for the micas, where the reflectance peaks near 9–10 μm (1000–1100 cm−1) are weaker and the transparency features near 10.5–11.8 μm (850–950 cm−1) are stronger.
Si–O bending vibrations.
The bending vibrations for tetrahedral SiO4 groups occur near 18–25 μm (400–550 cm−1) in reflectance and emissivity spectra when little or no tetrahedral substitution occurs (Salisbury et al., 1991; Michalski et al., 2005). Serpentines, chlorites and biotite consistently exhibit a strong SiO4 bending vibration centred near 20.6–21.3 μm (470–485 cm−1), although some structure is observed due to Si-O-Fe and Si-O-Mg modes. When there is substitution of Al or Fe for Si in the tetrahedral sites, the Si(Al,Fe)O4 bending vibrations are split into multiple bands. A strong doublet is observed in the smectites and in the micas zinnwaldite, celadonite and glauconite (all have tetrahedral substitution). Bending along the Si–O–M bonds from the tetrahedral Si–O into octahedral O–M bonds also occurs and is responsible for the weaker bands which often occur near 15–18 μm (550–670 cm−1) for Al and Fe cations and near 22–24 μm (400–450 cm−1) for Mg cations. The Si–O–Mg vibrations are particularly well resolved in the reflectance spectra of chrysotile, clinochlore and chamosite (Fig. 10⇑).
Bending vibrations of the OH bound to octahedral cations are observed from ~11–17 μm (590–950 cm−1) (Stubican & Roy, 1961; Farmer, 1974; Mendelovici et al., 1979; Beran, 2002; Gates, 2005). For smectites, these are found at ~10.9 μm (~920 cm−1) for Al2OH, ~11.4 μm (~880 cm−1) for AlFe3+OH, ~11.8 μm (~850 cm−1) for AlMgOH, ~12.2 μm (~820 cm−1) for Fe23+OH, ~12.6 μm (~790 cm−1) for MgFe3+OH and ~14–15 μm (~650–700 cm−1) for Mg3OH. The OH-bending vibrations occur near 11 μm (910–940 cm−1) in spectra of Al-rich dioctahedral micas, but may be as small as 16.9 μm (590 cm−1) in spectra of Al-rich trioctahedral micas. Kaolinite spectra exhibit Al2OH bands at 10.7 μm (936 cm−1) and 10.9 μm (915 cm−1) with AlFeOH bands occurring at 11.4–11.6 μm (865–875 cm−1). Serpentines have bands near 15.5 and 16.2 μm (618 and 646 cm−1) which are due to Mg3OH and some combination with Fe. Chlorite group spectra exhibit bands from 14.5–16.1 μm (620–692 cm−1) and 13.1–13.4 μm (744–765 cm−1) which vary with octahedral cation composition (Hayashi & Oinuma, 1965) and are probably due to OH-bending vibrations (Farmer, 1974).
These OH-bending bands are observed as absorption features or minima in the reflectance spectra of fine-grained materials and are shown in bold in Table 5⇑. For coarse-grained samples such as the chrysotile fibres in this study, the OH-bending vibrations are inverted and occur as reflectance peaks. This trend is observed in a study of transmittance and reflectance spectra of fine- and coarse-grained samples of several minerals by Salisbury et al. (1991) and is due to the dominance of specular (single-scattering) reflections in this wavelength region where the materials are strongly absorbing and the first-order reflections dominate the reflectance spectra.
Coordinating remote sensing and phyllosilicate occurrence
Phyllosilicates form under a variety of alteration conditions and are an important indicator of weathering processes (Jackson, 1959; Velde, 1985; Chamley, 1989; Deer et al., 1992; Nagy, 1995). The occurrences of the minerals in this study are summarized in these works. We hope that this information can be associated with spectral features important for the identification of phyllosilicates by hyperspectral remote sensing. The information may be useful and significant in establishing constraints for the geological conditions on Mars in regions where phyllosilicates are observed.
Weathering sequences for phyllosilicates suggest a trend of alteration from: (1) mica or chlorite to; (2) mixed-layer smectite and non-expandable minerals to; (3) smectite to; (4) kaolinite to; (5) gibbsite plus Fe oxides (e.g. Jackson, 1959). This pattern proceeds from lower temperature and less water to higher temperature and more water. Sedimentary micas such as muscovite and illite are likely to form under more highly-altered conditions, including those subsequent to smectite formation (e.g. Chamley, 1989). Dissolution rates are similar for most phyllosilicates in near-neutral environments and increase under lower and/or higher pH conditions (Nagy, 1995).
Cold and dry climates, such as those found at the Arctic, Antarctic and high elevations on Earth, as well as on Mars today, have limited liquid water and are dominated by physical weathering. In such environments, the phyllosilicates illite and chlorite are most common and are derived directly from the parent rocks. Cool and humid conditions may alter precursor rocks to form free silica and highly degraded rock-derived phyllosilicates. Temperate–humid climates favour the genesis of altered 2:1 layer phyllosilicates, typically irregular mixed-layers vermiculite ((Mg,Fe2+,Al)3(Al,Si)4O10(OH)2·4H2O) and poorly crystalline smectite, while warm sub-arid climates favour the formation of Fe-bearing smectites, especially in environments with seasonal variations in humidity. Alteration of volcanic material is characterized by the formation of Al-rich allophane group minerals such as allophane (Al2O3·SiO2) and imogolite (Al2SiO3(OH)4), followed by Fe-Al smectites, then halloysite (Al2Si2O5(OH)4). Hot and humid climates support active hydrolysis and ion leaching and often result in formation of kaolinite, goethite (Fe3+O(OH)), opal (SiO2·nH2O) and gibbsite (Al(OH)3).
Acidic conditions favour kaolinite formation under normal terrestrial temperatures and pressures on geological surfaces, while alkaline conditions favour smectites in the presence of Ca and illite in the presence of K. Kaolinite, the most aluminous phyllosilicate phase possible, is an important constituent of sediments, sedimentary rocks and hydrothermal alterations and is stable throughout most clay mineral environments, although it is highly susceptible to physical weathering. Acidic rocks such as granites and quartz diorites usually alter to form kaolinite. Sample KGa-1 studied here is from The Clay Minerals Society Source Clays Repository (CMS-SCR). It was collected near the top of the Buffalo Creek Kaolin member from the Buffalo China mine in Georgia. This kaolinite is believed to have formed by intense weathering in a deltaic environment followed by lateritic leaching (Pruett & Webb, 1993; Moll, 2001). Serpentines form in altered ultrabasic rocks containing forsterite and pyroxene through sedimentary processes or hydrothermal alteration. Migrating hydrothermal fluids at ⩽400°C are a common mechanism for serpentine formation in veins and along rock boundaries. For example, the C-200 sample studied here is from an asbestos vein in the Cassiar (alpine-type) serpentinite, formed from a harzburgite tectonite precursor. The chrysotile sample probably formed at a constant temperature of ~300°C (O’Hanley & Dyar, 1998).
The presence of abundant smectites indicates a moderately advanced weathering profile. More basic rocks tend to produce trioctahedral Mg-bearing smectites, while more acidic rocks and Fe-rich basaltic rocks tend to produce Al- and Fe-bearing dioctahedral smectites. Montmorillonite frequently forms by alteration of tuffs and volcanic ash in the presence of Mg. If the Mg is leached out, then kaolinite forms. Basic feldspathic rocks that are rich in Ca and Na typically alter to form montmorillonite. The SAz-1 montmorillonite was collected for the CMS-SCR at the Bidahochi Formation in northeastern Arizona. Vitric tuff and ash deposits were converted to montmorillonite through the action of lake waters which carried Mg, leached silica and alkalis (Moll, 2001). Other smectites such as saponite and nontronite are found associated with mineral veins or in pockets in altered basalt and basaltic glass. The ferruginous smectite in this study was collected from Grant County, Washington, USA (CMS-SCR); nontronites form along the Columbia River valley, including Grant County, from alteration of basaltic glass (Kornfeld, 1956). Smectites are commonly formed in ocean and sedimentary environments, but are less stable over time in aqueous environments. Smectites frequently convert to other phyllosilicates such as chlorite or glauconite in shallower waters where microbial activity provides a reducing environment, where wet/dry cycling occurs, or in the presence of Fe or high salinity.
Biotite forms in a much broader spectrum of geological environments than most other micas. It occurs commonly in a variety of rocks such as granites, granodiorites, diorites and norites and also with quartz. The Fe-mica sample studied here comes from a two-mica granite in the Massif-Central in France (Govindaraju, 1978). Zinnwaldite forms primarily in granite pegmatites and cassiterite-bearing veins and is typically associated with other Li-bearing minerals. The type-locality sample studied here is from an albitized zinnwaldite granite in the tin-mining district of Krusne hory, Czech Republic (Govindaraju et al., 1994). Glauconite forms primarily in marine sediments and is readily altered to biotite-rich deposits or layers containing kaolinite, Fe hydroxides and silica. The sample studied here, an unusual igneous occurrence, comes from the White Mountain Plutonic Suite pegmatite associated with the Conway granite, but has a composition closely resembling the diagenetic form. This glauconite occurs as the last mineral to fill arfvedsonite-bearing miarolitic cavities in metre-sized pegmatites in alkali granite (C.A. Francis, pers. comm., 2007). Celadonite is typically formed from volcanic materials in low-pressure and temperature environments and, once formed, is very stable, even under high temperatures. The celadonite studied here came from a hydrothermally-altered basalt near Red Rock Canyon (Mohave Desert), USA (Bowen et al., 1989).
Chlorites frequently form at sedimentary boundaries under reducing conditions through destabilization of smectite and kaolinite in sedimentary rocks and are typically considered unstable under oxidizing conditions. Contact, hydrothermal and regional metamorphism of biotite and other ferromagnesian minerals in igneous rocks is another formation mechanism of large chlorite deposits containing minerals such as clinochlore. Chamosite forms by metamorphism of iron deposits; it may also form in sedimentary environments containing large amounts of Al and Fe and is associated with other phyllosilicates and Fe oxides. The chamosite sample studied here comes from Ishpeming, Michigan, USA, and was probably collected from the iron ore mines there. Paragenetic context for the Quebec clinochlore is not available (these two samples were originally purchased from a mineral distributor); it probably originates from a large deposit associated with the Jeffrey Mine, where it coexists with serpentinites.
Coordinating NIR spectral features and geological setting.
By further analysis of the spectral features observed near 2.2–2.5 μm (Figs 3⇑ and 5⇑) we can group the phyllosilicates of this study by geological occurrence. The phyllosilicates studied here also contain a variety of Al, Fe and Mg; these are expressed relative to Si in Fig. 11⇓ in order to facilitate an understanding of the similarities and differences betweem the chemistries of these samples.
Even samples from completely different geological environments may have similar spectral absorption bands. For example, the montmorillonite and kaolinite in this study both have bands near 2.2 μm. Fortunately, these features are sufficiently distinct for discrimination, although they have similar band centres (Table 3⇑). Kaolinite is identified by a sharp doublet near 2.17 and 2.21 μm and no water band near 1.9 μm, while montmorillonite has a single band near 2.21–2.22 μm (which varies considerably with Al-Fe-Mg abundance; see Bishop et al., 2002b) and a strong water band at 1.92 μm. Kaolinite has the greatest Al, smallest Fe and smallest Mg concentrations of all the samples studied here (Table 2⇑ and Fig. 11⇑) and is associated with acidic conditions, while montmorillonite is associated with alkaline conditions and the presence of Ca (or Na) and Mg. Kaolinite plus amorphous silica and/or gibbsite are typically observed in highly leached and altered zones.
Al-rich micas such as zinnwaldite also exhibit an OH band associated with Al near 2.2 μm, but this can be distinguished from kaolinite and mont-morillonite by the additional band near 2.25 μm due to the OH bound to Al and either Fe or Mg. Micas are also generally formed under alkaline conditions as in the case of smectites; however, K is more likely to be present than Ca or Na, and some Mg usually occurs. The band near 2.24–2.26 μm is observed for Fe-smectite, clinochlore, chamosite, zinnwaldite, celadonite and biotite, all of which have a combination of Al, Fe and Mg present (Fig. 8⇑). The micas in this study can be distinguished from the smectites and chlorites by the additional bands near 2.33–2.34 μm due to OH bound to Mg and/or 2.35–2.37 μm due to OH bound to Fe2+, although many of them have a similar feature in the range 2.24–2.26 μm.
Finally, chlorites and serpentines both exhibit a Mg-OH band near 2.33–2.34 μm, but the chlorites can be readily identified by the presence of an additional band near 2.24–2.26 μm. Serpentines and chlorites both arise from diagenetic (e.g. sedimentary) or hydrothermal alteration of basic ferromagnesian rocks. Serpentines generally occur at higher temperatures or from rocks with a more ultrabasic chemistry.
Coordinating mid-IR spectral features and geological setting.
The reflectance minimum (emission maximum) at the primary silicate band, termed the Christiansen feature, has been correlated with Si abundance (Salisbury, 1993). Comparing the Si abundance (Fig. 11⇑) with the Christiansen feature position (Fig. 10⇑), an expected shift is observed. The Christiansen feature is found at ~8.1 μm (1230 cm−1) for montmorillonite, which contains ~60 wt.% SiO2, ~8.4 μm (1190 cm−1) for nontronite at ~47 wt.% SiO2, ~8.7 μm (1150 cm−1) for biotite and clinochlore at ~35 wt.% SiO2 and ~8.8 μm (1130 cm−1) for chamosite at ~22 wt.% SiO2. The band centre for this Si(Al,Fe,Mg)O4 stretching vibration does not correlate linearly with Si abundance and appears to depend on the amount of Al, Fe and Mg present as well.
Trends in chemistry are also observed for the Si(Al,Fe,Mg)O4 bending vibration. For the smectites, kaolinite and micas with ⩾45 wt.% SiO2, a distinct doublet is observed in the range 18–25 μm (400–550 cm−1). For the phyllosilicates with SiO2 <45 wt.% (chrysotile, clinochlore, chamosite, biotite) a broader single band is observed, in most cases superimposed by small peaks or shoulders. Most of the mid-IR features are more readily resolved in transmission spectra due to the surface-scattering effects for reflectance and emission spectra described earlier, thus transmission spectra are often easier to compare with phyllosilicate mineral structure and composition. However, remote sensing requires emission spectra, so it is important to correlate the resolvable emission features with mineral structure and composition as far as possible.
For thermal remote sensing of phyllosilicate-bearing regions on Mars, these features may contribute towards understanding the geological history of the surface. Phyllosilicates rich in Si and Al, such as montmorillonite, would have a Christiansen feature near 8.1 μm (1230 cm−1) and a doublet near 19 and 23 μm (525 and 430 cm−1). This would indicate an alkaline environment abundant in Ca or Na and with small amounts of K and Mg. In contrast, Fe-rich nontronite exhibits a Christiansen feature near 8.4 μm (1190 cm−1) and a doublet near 20 and 23.5 μm (500 and 425 cm−1). These features would indicate an alkaline environment richer in Fe and also abundant in Ca or Na but low in K and Mg. Fe-rich micas such as glauconite and biotite exhibit Christiansen features near 8.6–8.7 μm (1150–1160 cm−1) and bands near 20–23 and 20.6 μm (505–435 cm−1 and 485 cm−1), respectively. These features would also indicate an alkaline environment, but one with greater K and Fe and smaller Ca and Na abundances. Chlorites and serpentines both exhibit a Christiansen feature near 8.7–8.8 μm (1130–1150 cm−1) and a bending band near 21 μm (470–480 cm−1); however, differences in the shape of the Si-O-Al components of these features near 10 and 18 μm (1000 and 550 cm−1) may be useful in separating these two mineral classes. The presence of chlorites and serpentines implies alteration of ferromagnesian rocks.
The authors wish to thank C. Francis of the Harvard Mineralogical Museum for allowing us to use the Hurricane Mountain sample as part of this study, P. Komadel for contributing the purified SAz-1 montmorillonite, SWa-1 ferruginous smectite and Sampor nontronite samples, M. Jercinovic of the University of Massachusetts for new probe data on the glauconite, Mickey Gunter for crystal structure diagrams and T. Hiroi for assistance with the reflectance measurements. The paper was greatly improved by helpful editorial comments from J. Cuadros, W. Gates and R. Milliken. The RELAB spectroscopy laboratory is a multi-user facility operated under NASA grant NAG5-13609. We thank P. Christensen for the generous use of his emission spectrometer facility at the School of Earth and Space Exploration, Arizona State University. We are grateful for support from NASA grants NAG5-12687, NNG06G130G, NNX06AD88G, the NASA Astrobiology Institute and NSF grant EAR-0439161.
- Received April 24, 2007.
- Revision received December 18, 2007.