The Johns Hopkins University mineral library was published as a book by the Johns Hopkins University Press in 1991 (title and authors below). The Introduction to this book is summarized below to document how samples were selected, characterized and measured, as well as to explain the origins of spectral features for different kinds of minerals. Infrared (2.1 - 25 micrometers) Spectra of Minerals by John W. Salisbury The Johns Hopkins University Louis S. Walter Goddard Space Flight Center Norma Vergo U. S. Geological Survey and Dana M. D'Aria University of Maryland Dedication To our co-author, Norma Vergo, whose untimely death is mourned by us and all who knew her. Acknowledgments Most mineral samples were provided courtesy of John White, of the Smithsonian National Museum of Natural History, and Paul Powhat was extremely helpful in their selection. We are also grateful to Gene Jarosewich and Joseph Nelen for use of the Smithsonian microprobe. Ms. Jane Jellison, of the Goddard Space Flight Center, kindly prepared samples for microprobe analysis. This research was supported by the Solid Earth Science and Solar System Exploration Branches of the U.S. National Aeronautics and Space Administration. Introduction In her thorough review of the use of infrared spectroscopy in mineralogy, Estep-Barnes (1977) listed infrared spectral libraries for minerals, rocks, and inorganic compounds published prior to 1975. Only Hunt and Salisbury (1975, 1976), Lyon and Green (1975), Vincent et al. (1975), and Ferraro (1982) have been published subsequently. Most of these libraries consist of transmittance measurements, which have limited application to remote sensing problems because they do not include the effects of scattering (Salisbury et al., 1987a). Only the works of Lyon (1963, 1964), Lyon and Green (1975), Hunt and Salisbury (1974, 1975, 1976), and Vincent et al. (1975) contain reflectance and/or emittance data that can be used to predict the spectral behavior of minerals and rocks in a remote sensing situation. However, these data lack the potentially significant 3-5 micrometers region of the spectrum, do not document the effects of different particle sizes on the spectra of most samples, typically lack complete chemical and mineralogical sample characterization, and are not available in digital form. The present work seeks to remedy all of these shortcomings for 130 minerals. It is an expansion on an Open- File Report of the U. S. Geological Survey (Salisbury et al., 1987a) which, not being generally available to the public, was of limited usefulness to the remote sensing community. The present work also adds digital data for all of the spectra in ASCII format on an accompanying CD-ROM disk. Experimental Technique Acquisition and Preparation of Samples Most samples were acquired, often as single crystals, from the Smithsonian National Museum of Natural History. Other samples were obtained from the Hunt and Salisbury collection in Denver or from individuals. Most samples were crushed in a steel percussion mortar, after which steel particles were removed with a hand magnet. Crushed samples were hand-picked when impurities were present, ultrasonically cleaned, and then subjected, when necessary, to an acid wash to insure purity. Samples were ground in a sintered sapphire mortar under either acetone or (more commonly) alcohol, which facilitated grinding, avoided disordering of the sheet silicates, and prevented the finest particles from drifting away as aerosols. Particulate samples were dry-sieved to two different particle-size ranges (75-250 micrometers and 0-75 micrometers), and the coarser particle-size range was subsequently washed in acetone or alcohol to remove clinging fines. Clay samples were disaggregated ultrasonically, and the <2 micrometers size fraction was concentrated with a centrifuge. Characterization of Samples: Mineral specimens were initially examined as hand samples and, after grinding to a powder, under a petrographic microscope. This examination was typically combined with X-ray diffraction and microprobe analysis to identify those samples that were pure mineral phases and to establish their chemical and mineral compositions. When other approaches to characterization were employed, this is noted in the mineral description sheets. Unless otherwise noted on the description sheets, petrography was performed by Dana D'Aria, microprobe analyses by Louis Walter, and X-ray diffraction analysis by Norma Vergo. The 0-75 micrometers fraction (except where noted) of all samples was characterized by X-ray diffraction using a Siemens D-500 X-ray diffractometer with CuKa radiation. Samples were mounted using a side pack sample holder to achieve random orientation. There were several samples that appeared to be monomineralic when studied by optical methods, but that had "extra" peaks in the X-ray patterns compared to the Joint Committee on Powder Diffraction Standards (JCPDS) file cards. These patterns were modeled by calculating allowed d-spacings from input unit cell dimensions using the Appleman and Evans (1973) cell refinement program. The differences between the JCPDS and modeled patterns typically could be attributed to differences in ionic substitutions. Additional tests were done to characterize the clay minerals. The <2 micrometers fraction was oriented on a glass slide using a Millipore filter setup (Drever, 1973). These slides were then analyzed in the air-dried and glycol-solvated states. K-saturation was done on the vermiculite sample. The mixed-layer clay minerals were modeled using methods outlined in Reynolds (1980) and using the New Mod computer program written by R. C. Reynolds (Dartmouth College). Microprobe analyses were performed on 75-250 micrometer sample fractions. Through kind permission, the Applied Research Laboratory instrument at the Smithsonian Institution Division of Mineral Sciences was used. Six fixed-wavelength dispersive detectors were used to analyze for Si, Al, Fe, Mg, Ca, and K. Three movable, wavelength-dispersive detectors were used to analyze for Na, Ti, and Mn. Quartz was used as a background standard for Al; alumina was used as a background standard for all other elements. Standards employed were selected from those available at the Smithsonian: rhodonite for Mn and Kakanui hornblende for all other elements. The Bence-Albee correction procedure was used to convert intensity ratios to element/oxide abundances. As usual for microprobe analyses, ferric and ferrous iron were not distinguished and Fe is reported as FeO. In analyzing a mineral, points were accumulated for 10-second integration times. Normally, four or five randomly selected points were analyzed on one or two grains. Then, individual points were run on five or six additional grains. If these analyses were equivalent within statistical counting error, the material was assumed to be homogeneous. In some cases, small deviations from homogeneity were found and these are noted in the text. Estimated errors (as determined by replicate analyses) are: Oxide Coefficient of Variation (Rel. %) SiO2 1.7 Al2O3 2 FeO 2 MgO 2 CaO 2 K2O 3 Na2O 5 TiO2 2 MnO 1 These estimates apply to cases in which the oxide abundance is greater than 3%. The limit of reliable detection is generally about 0.5%. We found that all three different modes of sample characterization, microscopic, X-ray, and microprobe or other chemical analysis, were required to assure sample purity. Based on this sample characterization process, generally two out of every three mineral samples were found to be insufficiently pure to be included in this compilation. The results of the analyses are documented on each sample description sheet. Mineral names and ideal formulae used on these sheets were taken from Fleischer (1983). Acquisition of Spectra Spectra were acquired at 4 cm-1 resolution using either a Nicolet 5 DXB or a Nicolet System 510 interferometer spectrometer; both spectrometers have identical optical benches Transmittance and biconical reflectance spectral data were recorded from 2.08 to 25 micrometers with a triglycine sulfate (TGS) detector, and directional hemispherical reflectance data were recorded from 2.08 to 14 micrometers with a liquid helium-cooled, mercury-cadmium- telluride (MCT) detector. Transmittance was measured by passing the focused beam ( 6 mm diameter) of the interferometer through a KBr disk (commonly referred to as a pellet) containing sample material made in the manner of Stimson and O'Donnell (1952). Briefly, these pellets consist of 300 mg of KBr mixed with about 0.7 mg of sample ground under alcohol to a size range less than 2 micrometers. The pellets were pressed in a dye under vacuum for five minutes at 10,000 kg/cm2 pressure to produce transparent disks 13 mm in diameter and 1 mm thick. A reference pellet composed of pure KBr was used to record a background against which was ratioed the sample pellet transmittance. Biconical reflectance spectra were recorded using a Spectra Tech "Collector" diffuse reflectance attachment, which uses two 90¡-off- axis ellipsoids that act as 6x beam condensers; the focused beam diameter is thus reduced to 1 mm, and illumination and reflection occur at the specular angle over a solid angle that closely approaches p steradians. Thus, our bidirectional reflectance measurements actually record biconical reflectance over a large solid angle. An aluminum mirror was used as the background reference against which sample biconical reflectance was ratioed. A mirror was used as the reference instead of Halon, because the latter exhibits strong absorption bands in most of the spectral range measured here. Not only does the mirror provide a spectrally flat reference, but its high reflectance is required when measuring the reflectance of solid samples at the wavelengths of the fundamental molecular vibration bands. As described below, minerals have a mirror-like opacity at such wavelengths, and smooth solid samples may have reflectances close to that of a mirror. Use of biconical reflectance with a mirror reference makes the measured reflectance of a diffusely reflecting particulate sample roughly a factor of five lower than it would be for the integrating sphere measurements commonly made in the visible and near-infrared. The Nicolet 5DXB and System 510 can switch the infrared beam from the sample compartment to an external port, through which it exits in collimated form. An integrating sphere, coated inside with a diffusely reflecting gold surface, was attached to the instrument at this port. The sphere is 12.7 cm in diameter and has a 2.5 cm diameter entrance port in the top of the sphere at 10¡ off the vertical, through which the beam passes to fall on a 2.5 cm diameter sample/reference port in the bottom of the sphere. Beam size on the sample in that bottom port is 1.54 cm. A 2.5 cm detector port is placed at an angle of 90¡ to the principal plane in the side of the sphere, and the liquid-nitrogen-cooled MCT detector chip is baffled to eliminate direct viewing of either the sample or the specular "hot spot" on the sphere wall. A port at the specular angle was filled during the measurements reported below with a gold-coated plug having a surface curved to match the interior curvature of the sphere. The integrating sphere uses a Labsphere diffuse gold surface as a reference. Sphere performance was carefully calibrated to provide absolute reflectance by comparing measured reflectances of Halon, a front-surface aluminum mirror, water, and a black body cone with values obtained from the National Bureau of Standards or found in the literature (Salisbury and Milton, 1987). The usual sample holder for biconical reflectance measurements was 13 mm in diameter and 2 mm deep, although a "microreflectance" holder 3 mm x 2 mm could also be used for very small samples. The usual sample holder for hemispherical reflectance measurements was 2.5 cm in diameter and 3 mm deep. That samples were optically thick at a depth of 2 mm was verified by comparison of spectra measured in bare aluminum cups with spectra measured in cups painted black. Most particulate samples were sifted into the sample holders to attempt to achieve random orientation of grains. For some samples, including all of the clay minerals, solid samples were not available. In these cases, the fine powder samples were packed into a pseudo-solid sample so that the reflectance peaks of the fundamental molecular vibration (reststrahlen) bands could be seen. This was accomplished with <2 micrometers clays by placing them in a folded weighing paper and gently rolling a sample bottle over the outside surface. This produced cohesive flakes that could be readily transferred to the reflectance attachment and measured. Some samples such as the olivines, were not sufficiently fine-grained to produce flakes. To make packed sample measurements of these samples, they were simply pressed into a sample cup with the flat side of a spatula. Solid samples were fixed in a bed of moldable erasure to hold the surface being measured in a horizontal position. All samples could be raised or lowered by a micrometer screw mechanism so that the measurement surface was at the beam focus. The Nicolet scans an interferogram each second. Interferograms are then averaged to provide the desired signal-to-noise after deconvolution to a spectrum. This normally required 100 scans for transmittance measurements and 500 to 1000 scans for reflectance. Major Spectral Features of Minerals The spectral features of minerals in the wavelength range considered here are the result of vibrational processes. Their number, intensity and shape are dependent on atomic masses, interatomic force fields and, particularly, molecular geometry. One goal of the spectroscopist is to quantitatively describe the vibrational process so that the origin of each absorption band can be understood. Sophisticated calculations have been made that are consistent with observation, at least for the simpler minerals (e.g., Elcombe, 1967), although not necessarily correct or final. Even if a vibrational mode were understood precisely, it is virtually impossible to describe such a motion simply and concisely for such complex structures as silicates. Consequently, one must rely on some very general description, such as "Si-O symmetric stretch," to describe all those vibrations which predominantly involve the symmetric expansion and contraction of the silicon-oxygen bonds. Using such simplified visualizations, we can successfully generalize about the spectral behavior of minerals. For example, lighter atoms vibrate at higher frequencies (shorter wavelengths) than heavier atoms when substituted into the same structure (see olivines, p. 168-192). Higher bond strengths also result in higher frequencies of vibration, and this change in bonding in silicates is related to the degree of polymerization of the Si-O4 ion (Walter and Salisbury, 1989). This results in a systematic change in wavelengths of the fundamental vibration bands of silicates as the framework structure ultimately gives way to isolated tetrahedra. Finally, bond-stretching vibrations in covalent structures lie at higher frequencies than bending modes, and such internal molecular vibrations typically lie at higher frequencies than lattice modes (Farmer, 1974). The most prominent features in the infrared spectra of minerals can be understood in the context of the generalizations outlined above and are described below for different types of minerals. In particular, we point out those bands seen in reflectance or emittance that are not apparent in the transmittance spectra typically studied by others. This discussion of the origins of spectral features is not repeated in the text for each mineral, because it would prove highly repetitious for the relatively well understood bands, such as the fundamental internal molecular vibration bands of the minerals and associated water and hydroxyl. Those bands can easily be identified in the spectra of each mineral on the basis of the discussion below. The attribution of more complex features due to overtones and combination tones of the internal vibrations and lattice modes is a more speculative matter, even for the simplest of minerals. Such speculation would also be repetitious and is not the function of this work. Farmer (1974) discusses the spectral features of minerals at length, and Farmer and Palmieri (1975) provide an exhaustive list of references categorized by mineral. Estep-Barnes (1977) presented a good review of the major spectral features of minerals, accompanied by an extensive bibliography. The interested reader is referred to these works for the best information on the subject of detailed band assignments. Silicates: The most intense spectral features of silicates, occurring between 8 and 12 micrometers, are generally described as due simply to fundamental asymmetric Si-O-Si stretching vibrations, but Si-O-Al stretching vibrations may also contribute when aluminum is part of the crystal lattice (for the classic Si-O-Si stretching feature, see the 9.2 micrometers band in quartz). The appearance of these features typically changes in reflectance because of the role of the refractive index in scattering (see The Role of Surface and Volume Scattering, below). The weak side band near 8.5 micrometers in the transmittance spectrum of quartz, for example, becomes a well- defined lobe of a prominent reflectance doublet between 8 and 10 micrometers . The reflectance spectrum of a quartz glass displays a much weaker short-wavelength lobe, which in some cases of shocked quartz we have seen is reduced to a shoulder. This simplification of glass spectra of minerals is well known and is attributed generally to broadening of the bands (e.g., Farmer, 1974, p. 484). However, broadening would not appear to explain the reduced intensity of the 8.5 micrometers band in the spectrum of glass compared to that of crystalline quartz. An alternative explanation is that the short- wavelength lobe of the strong quartz reflectance doublet is not due entirely to internal molecular vibrations but depends to some extent on long-range order (Simon and McMahon, 1953). Whatever the details of their origin, these most intense features fall in the 8-14 micrometers atmospheric window, making them the most useful for terrestrial remote sensing of silicates (Kahle and Goetz, 1983; Walter and Salisbury, 1989). The second most intense silicate bands are broadly characterized as O-Si-O deformation or bending modes, which occur in the 18-25 micrometers region. Again, aluminum and, indeed, other cations may contribute additional band structure in this region (Farmer, 1974, p. 365). The relative intensities of the two quartz bands in this spectral region appear unchanged in reflectance compared to transmittance, but have been shifted about 1 micrometers to shorter wavelength by the interaction of absorption coefficient and refractive index on the scattered light, which is typical (Salisbury et al., 1987b). The weaker feature occurring at 18.3 micrometers in reflectance completely disappears in the spectrum of fused silica, indicating such a strong dependence on long-range order that it must be due to a lattice vibration. Weaker bands in quartz spectra between 12 and 15 micrometers have been attributed to symmetric Si-O-Si stretching vibrations (Farmer, 1974, p. 366). When some of the silicon atoms are replaced by aluminum, as in the feldspars, additional Si-O-Al stretching vibrations are added over a longer wavelength range. For example, albite displays eight highly characteristic bands in its spectrum between 12 and 20 micrometers. Again, such bands are greatly simplified or eliminated in the spectra of glasses (Nash and Salisbury, 1991). Additional weak bands are displayed as troughs between 3 and 7 micrometers. Such bands in silicate spectra have been largely ignored because they are usually too weak to be seen in transmittance spectra. However, they can be very useful in the spectral identification of fine particulate minerals and rocks, where they are quite prominent (Salisbury et al., 1987b; Salisbury and Walter, 1989; and Salisbury et al., 1991). Because they have not been assigned with any certainty, we refer to such bands simply as overtone/combination tone bands of internal and lattice modes. Carbonates: The strongest bands of carbonates are due primarily to fundamental internal molecular vibration bands of the CO3 ion, which are well understood (Farmer, 1974, p. 231). Carbonates typically display a strong band near 7 micrometers due to asymmetric C-O stretching vibrations and weaker bands near 11.4 and 14.3 micrometers due to bending modes, which can be seen in the spectrum of calcite. Very weak bands in the transmittance spectrum to shorter wavelength than 7 micrometers are strongly displayed as troughs in the reflectance spectrum of particulate calcite. Because of their relative visibility in transmittance spectra, these weak bands have been the subject of study and appear to be due to combination tones of internal and lattice modes (Farmer, 1974, p. 236). Sulfates: The sulfate ion displays a group of intense stretching fundamentals near 8.7 micrometers and two or more bending modes near 16 micrometers, as can be seen in spectra of gypsum and anhydrite. Again, the weaker features in transmittance spectra of sulfates at shorter wavelength than the strong stretching fundamental are strongly displayed in reflectance spectra of particulate samples. The complex feature near 4.6 micrometers appears to be a combination tone of the sulphate ion, perhaps accompanied by water combination tones (Hass and Sutherland, 1956). The features near 2.8 and 6.2 micrometers are due to water, the spectral features of which are discussed separately below. Oxides: The metal-oxygen stretching vibration bands in oxides occur at longer wavelength than the Si-O features (e.g., chromite). An interesting aspect of these features is that, because of the intense dipole oscillations induced by the vibrations of highly ionic oxides, their powder spectra are profoundly modified by the shape and size of the particles (Farmer, 1974, p. 183). Thus, it is sometimes uncertain whether variations in powder spectra given by different specimens of a given compound are due to real differences in purity or phase, or merely to shape and size. A case in point is provided by spectra of our two goethite samples. Sulfides: Most metal-sulphur vibration bands lie beyond our wavelength range in the far-infrared. We have included two examples (pyrite, and pyrrhotite) that do show bands within our wavelength range. Water and Hydroxyl: The most common vibration bands in minerals are due to water and hydroxyl, the spectral features of which have been thoroughly reviewed by Aines and Rossman (1984). When water is not fixed in a crystal lattice but is hydrogen-bonded to other water molecules, it results in a broad spectral feature centered near 2.9 micrometers due to O-H stretching vibrations and another near 6.1 micrometers due to H-O-H bending vibrations. Such water may be present in fluid inclusions, as interlayer water in sheet silicates, or as water of hydration. Water in a crystalline environment produces sharper O-H stretching absorption features than occur in the liquid water spectrum, which typically also occur at a shorter wavelength. Multiple O-H stretching vibrations can result when water is present at several sites in the crystal lattice. Beryl and cordierite, for example, contain water that resides at specific sites in channels parallel to the C axis (Aines and Rossman, 1984). These minerals are also interesting because they typically have CO2 trapped in these channels, which produces sharp bands near 4.3 micrometers. Minerals containing hydroxyl without water display O-H stretching features near 2.7 micrometers but lack the broad feature at 2.9 micrometers and the H-O-H bending mode at longer wavelength. A good example is kaolinite, which has no interlayer water. Most often, however, minerals display a combination of hydroxyl and molecular water bands, as in the case of antigorite. Many minerals contain a trace of OH and water, although this is not reflected in their chemical formulae. A good example is quartz, which typically displays multiple sharp O-H stretching features superimposed on a weak broad water band. The broad water band is probably due to a small amount of liquid water in fluid inclusions. The sharper hydroxyl features are associated with hydroxylated alkali metals that serve to balance charges when aluminum substitutes for silicon (Aines and Rossman, 1984). In addition to the fundamental O-H and H-O-H features commonly seen in the spectra of minerals in the 2-7 micrometers region, a variety of significant metal cation-OH bands can be found at longer wavelegth, especially in clay minerals (Stubican and Roy, 1964). Kaolinite, for example, displays a prominent Al-OH band near 11 micrometers. OH lattice vibrations are typically seen at still longer wavelengths (Farmer, 1974, p. 348), such as the 16-micrometers feature in the spectrum of antigorite. Water and hydroxyl bands are spectrally important because most silicate minerals capable of doing so have undergone incipient alteration to hydrous phases and/or contain fluid inclusions, even when appearing quite fresh, because of the ubiquity of water in the terrestrial environment. This is in marked contrast to other environments, such as that of the moon (Roedder, 1984). It should be pointed out that water and hydroxyl are usually not present in large amounts where they are not part of the mineral stoichiometry. However, spectral features due to water and hydroxyl may be very prominent, especially in reflectance spectra of fine particulate materials. This is due to the enhancement of such absorption bands by the increased scattering associated with fine particle size, which is discussed below. Thus, an estimate of the abundance of water and hydroxyl relative to other phases can best be obtained from transmittance spectra. Discussion of Spectra The Role of Surface and Volume Scattering: All major studies of the nature of mineral spectra have used transmittance data (Farmer, 1974; Farmer and Palmieri, 1975, and references therein). Transmittance spectra are simpler to interpret than reflectance or emittance spectra, because they depend solely on the absorption coefficient. Reflectance and emittance spectra involve both the absorption coefficient and the refractive index, which causes spectral features to change quite significantly, especially when scattering becomes important in particulate samples (Salisbury et al., 1987b). It is apparent from inspection of our mineral spectra that particle size has a very significant effect on reflectance spectra. A mathematical model of reflectance from particulate samples has been developed by Hapke (1981). We focus here on a qualitative physical model to provide a basic understanding of how and why spectral features change with particle size, addressing the role of surface and volume scattering. The radiation returned to the observer in reflectance from a particulate sample has been scattered by the particles. This scattering takes place by two processes: surface scattering, which involves rays that have reflected from the surfaces of grains without penetration; and volume scattering, which involves rays that have been refracted into grain interiors and then scattered or refracted back out. Which of these processes dominates returned radiation is determined primarily by the absorption coefficient and particle size (Vincent and Hunt, 1968). The wavelength variation in absorption coefficient can be determined from a transmittance spectrum. This can be illustrated with quartz transmission spectra, where the highest absorption coefficient is associated with the Si-O stretching vibration near 9.2 micrometers. A weaker band due to the bending mode near 21.5 micrometers is accompanied by the still weaker feature near 19.5 micrometers, and progressively weaker stretching features can be seen near 12.5, 12.7 and 14.5 micrometers. The effects of particle size on these spectral features of different intensity can be seen in our spectra of different particle size ranges of quartz. Reflectance from the smooth surface of a solid sample, of course, eliminates multiple scattering. As a result, the entire spectrum is dominated by simple Fresnel reflectance from the surface (Salisbury et al., 1987b), and the spectral features are reflectance peaks associated with the strongest molecular vibration bands. The correlation of reflectance peaks with very strong absorption bands is counterintuitive to those researchers accustomed to working in the visible and near-infrared. This effect is due to the very strong absorption coefficient associated with these bands, which induces a mirror-like opacity at these wavelengths. For the coarse particulate 75-250 micrometers particle-size range, the reflectance peaks, or "reststrahlen bands," discussed above are still apparent between 7.5 and 24 micrometers. However the appearance of the spectrum between 2.1 and 7.5 micrometers has changed significantly. Weak but distinct alkali metal-OH features can be seen between 2.8 and 3.2 micrometers, and a series of overtone/combination tone bands are visible between 3.5 and 7.5 micrometers. These spectral features are all expressed as troughs, not peaks, showing that this region of the spectrum is dominated by volume scattering. What has occurred is that the particle size has become small enough, on average, to allow passage of photons completely through the grains in this spectral region of relatively low absorption coefficient. That is, the grains are optically thin. More photons are absorbed in band centers than in band wings during volume scattering, and reflectance minima (troughs) occur. Thus, for particulate silicates, the spectral range documented in this library is generally dominated by volume scattering and absorption band troughs on the left side of each figure and by surface scattering and absorption band peaks on the right side. The dividing line is the sharp minimum in reflectance associated with the principal Christiansen frequency displayed for quartz at 7.4 micrometers. Changes in spectral contrast: Considering first the reflectance minima associated with volume scattering, band depth (Clark and Roush, 1984) is a measure of band intensity commonly referred to as spectral contrast. It is apparent in comparing spectra of coarse and fine particulate size ranges that this spectral contrast changes with changing particle size. However, the change is not consistent, spectral contrast sometimes increasing and sometimes decreasing with decreasing particle size. The explanation for this behavior lies in the relationship between mean optical path length and particle size, which can be illustrated with three cases. Mean optical path length (MOPL) is the mean distance photons will pass through a material before total absorption takes place (Clark and Roush, 1984). When the absorption coefficient is relatively high and the MOPL is much less than the mean grain diameter in a particulate sample, all photons entering the grains are absorbed. If the MOPL remains smaller than the mean particle diameter, even at a finer particle-size range, photon absorption remains essentially unchanged. Something close to this first case is seen in the centers of the overtone/combination tone bands of quartz near 5.4 and 6.2 micrometers. The reflectance in these band centers increases only slightly in the spectrum of the finer particle-size range compared to that of the coarser size range (i.e., the band remains nearly saturated). A better example is seen in the center of the hydroxyl stretching fundamental near 2.7 micrometers in the spectra of antigorite. The band center remains saturated at about 1% reflectance for both coarse and fine particle-size ranges. A second case occurs at intermediate absorption coefficient when the MOPL is initially on the order of the mean grain diameter. As the particle size decreases, many more photons survive passage through the grains and a significantly larger portion of the incoming radiation will be scattered back in the direction of the observer. This second case applies to the wings of the 5.4 and 6.2 micrometers bands of quartz and to the wings of the hydroxyl band of antigorite. Here reflectance increases with decreasing particle size. When the reflectance in the wings of a band rise significantly with decreasing particle size (Case 2), while the band center remains saturated or nearly so (Case 1), an increase in spectral contrast occurs. This effect causes absorption bands having the right ranges of absorption coefficient between band wings and band centers to display intense bands at fine particle size. This occurs not only for many of the silicate overtone/combination tone bands (especially for quartz and olivine) but also for carbonate and sulfate combination tones and the fundamental O-H and C-H stretching vibration bands. Thus, spectra of fine particulate mineral mixtures may be dominated out of proportion to their abundance by minerals that display such spectral behavior, as shown elsewhere for igneous rocks and meteorites (Salisbury et al., 1989; Salisbury et al., 1991). A third illustrative case occurs when the MOPL is already much greater than mean particle diameter at the coarser particle size. In this case, the bulk of potentially backscattered radiation is already returned to the observer at large particle size. Thus, a reduction in mean grain diameter has proportionally less effect than when MOPL starts out close to mean grain diameter. This third case applies to the wings of the weaker overtone/combinaton tone band of quartz near 4.5 micrometers. Here the wings rise relatively less than the band center, which is affected more by Case 2 scattering. Thus, this band decreases in spectral contrast with decreasing particle size, which is the norm for weak bands displayed in the visible and near- infrared. We turn now to the reststrahlen bands dominated by surface scattering. These absorption bands produce prominent reflectance peaks for solid samples and at coarse particle size, which are greatly reduced in intensity for the fine particle size range. In fact, the weakest reststrahlen reflectance peak of quartz at 14.5 micrometers turns into a trough at the finest particle size. While changing particle size may have resulted in changes in single particle albedo and scattering geometry (Conel, 1969), Salisbury and Eastes (1985) focused on a physical model depending on porosity to explain the decline in spectral contrast of the reststrahlen bands. They noted that when a very fine (<5 micrometers) powder is packed to reduce porosity, its spectrum displays reflectance peaks as prominent as those displayed by the coarser particle-size ranges. They suggested that the increased porosity associated with fine particle size resulted in formation of photon traps. That is, that the pores acted like small black bodies. A similar explanation was offered earlier by Lyon (1964) and Aronson et al. (1966). However, research in progress suggests that the change in shape and intensity of the reststrahlen bands is due more to the occurrence of substantial volume scattering at fine particle size rather than to photon trapping, which is essentially a surface scattering effect. Briefly, it appears that the role of porosity is to physically separate 1-5 micrometers diameter particles that are optically thin, even in the reststrahlen bands (Hunt and Logan, 1972). When such particles are separated by more than a wavelength, they scatter independently as optically thin, volume- scattering particles. When packed closely together, however, they scatter coherently as if they were large, optically thick, surface- scattering particles. Thus, the loss of spectral contrast of reststrahlen bands for materials of fine particle size appears to be due directly to particle size and only indirectly (but critically) to porosity. It is important to note in any discussion of the spectral contrast of reststrahlen bands that such bands do persist in spectra of fine materials, even if reduced in intensity and changed in shape. In addition, the reststrahlen bands of some minerals are more persistent than those of others. As shown by Salisbury and Walter (1989) and Salisbury et al. (1991), quartz, olivine, and pyroxene display such persistent reststrahlen bands. Thus, the reststrahlen region may suffer reduced spectral contrast but is by no means either featureless or useless in spectra of particulate mineral mixtures. Transparency peaks: We have referred to the region of the spectrum where the reststrahlen bands occur as dominated by surface scattering. This is obviously true for solid samples of silicates and for the coarse particle-size range, which exhibit prominent reflectance peaks. However, it can be seen in the transmittance spectrum of quartz that a region of relatively high transparency exists between the asymmetric and symmetric stretching vibrational features in the vicinity of 10.5 to 12 micrometers. The absorption coefficient here is low enough so that grains become optically thin and volume scattering of the Case 2 type (see above) occurs as the particle size range is reduced from 75-250 micrometers to 0-75 micrometers. Because the strong reststrahlen reflectance peaks are greatly diminished at fine particle size, the broad transparency peak becomes a prominent feature centered at about 11 micrometers. Such transparency peaks may be very prominent in the spectra of some minerals (e.g., antigorite). Transparency peaks were first noted without explanation in reflectance spectra of rocks by Hovis and Callahan (1966). Conel (1969) described and explained such features (expressed as troughs in emittance) in spectra of quartz. Salisbury and Walter (1989) showed that the wavelength of the transparency peak could be related to the composition of igneous rocks. Documented here is the spectral behavior of a wide range of rock-forming minerals, showing the nature and magnitude of transparency peaks that occur in reflectance of fine particulate samples. Note that the prominence of this feature is dependent on the degree to which the adjacent reflectance peaks associated with the reststrahlen bands are diminished in spectra of the fine particle-size range. As explained above, the loss of these peaks appears to be related to the increased porosity and volume scattering in a sample due to very fine particle size. When a sample is sifted into a sample cup, the finest particles tend to be suspended in air and drift away during sifting. Thus, a sample sifted and measured repeatedly will tend to increase in mean grain diameter and to display progressively stronger reststrahlen bands. We have attempted to avoid this with our samples, but there is some obvious variability of reststrahen- band spectral contrast (compare acmite1f and acmite2f). Also, a few samples were not prepared by us and vary in particle-size range as a result (olivine4, for example, is completely lacking in very fine particles). Despite the fact that some minerals fail to display transparency features, it is apparent from the spectra in this volume that such peaks (expressed as troughs in emittance) could be prominent spectral features in the 8-14 micrometers spectrum of a fine- particulate regolith. Salisbury et al. (1991) suggest that the distortion of a spectrum by the sharp thermal gradient associated with a vacuum environment may make such broad features difficult to detect, but further measurements of appropriate candidate materials in a simulated space environment are needed to confirm this. Christiansen feature: A final spectral feature that is prominent at fine particle size is associated with the principal Christiansen frequency. This is a reflectance minimum that occurs because the real part of the refractive index undergoes rapid changes (anomalous dispersion) at a slightly shorter wavelength than the most intense molecular vibration band. Consequently, the refractive index approaches 1, resulting in a minimum of scattering, at a wavelength where absorption is still relatively low. With little scattering and little absorption, infrared radiation can penetrate a sample relatively easily, resulting in a minimum in reflectance or a maximum in emittance. This feature can be seen in reflectance at all particle-size ranges, but it is one of the more easily recognized spectral features in reflectance of the finest particle-size range (e.g., near 7.3 micrometers for quartz). Conel (1969) first showed that the wavelength of this feature is a good indicator of mineralogy. Logan et al. (1973) showed that the wavelength of this feature can also be used to determine igneous rock type and demonstrated its utility for mapping compositional variations in the lunar regolith. The relationship between wavelength of the Christiansen feature and composition has been determined on a more quantitative basis for igneous rocks by Walter and Salisbury (1989) and for meteorites by Salisbury et al. (1991). The Effect of Crystallographic Orientation: Many minerals are birefringent and have a different spectral response depending on crystallographic orientation. This is illustrated in the solid sample spectra of orthoclase. The occurrence of such orientation effects is what makes it necessary to sift particulate samples, or at least their upper millimeter, into sample cups for measurement. Tapping a cup or using a knife edge to level a sample tends to orient the grains, especially if the mineral has a prominent cleavage. Some samples having an extremely asymmetric shape, such as micas, will tend to develop grain orientation despite sifting. Our laboratory studies of calcite show, however, that more equant grains are randomly oriented when sifted, despite very strongly developed cleavage. The Effect of Packing: As discussed above in the section on spectral contrast, it has been shown that packing a fine particulate sample to reduce its porosity will greatly increase the prominence of the fundamental molecular vibration bands. Thus, spectra of packed samples are presented for some minerals, such as clays, for which coarse-grained samples showing the fundamental molecular vibration bands are not available. Packing will tend to orient the grains parallel to their strongest cleavage or parting, so that these spectra are most similar to solid sample spectra obtained in that orientation. The Effect of Atmospheric Gases: A spectrometer must be vigorously purged with dry nitrogen (or dry air from which CO2 has been removed) in order to avoid incorporating band structure due to atmospheric H2O and CO2 in the spectra. Spectra of some samples were obtained without adequate purging and display extremely weak atmospheric bands (e.g. the solid sample spectrum of acmite1). These atmospheric bands are usually easy to identify and do not interfere with the major molecular vibration bands of the silicates, so these spectra were retained in the library. Some minerals, such as beryl, cordierite, and topaz, contain CO2 gas trapped in channels or in fluid inclusions within the minerals. The Effect of Impurities: Two kinds of impurities introduce spectral artifacts: those that are added to the sample during laboratory processing and those that were an original part of the sample. Impurities added include water (in KBr pellets) and hydrocarbon, while the most common impurities within samples are quartz, calcite, and products of exsolution and alteration. KBr pellets are pressed under vacuum to remove water, but the material is so hygroscopic that this process quickly begins to reverse. The method chosen to compensate for this was to ratio the sample pellet transmittance against that of a blank pellet made under identical circumstances. If the KBr for both pellets was initially ground to the same particle size so that the same surface area was available to atmospheric water vapor, held under vacuum for the same amount of time to remove water, and then exposed to humid room air for the same amount of time in transfer (as pellets) to the purged and dry environment of the spectrometer sample compartment, the KBr of both pellets should contain the same amount of water. Then a ratio of sample transmittance to that of the blank pellet should display water bands that are intrinsic to the sample only. However, in practice the sample transmittance spectrum sometimes displays more of a water band than the reflectance data suggest is reasonable. The transmittance spectrum of kyanite, for example, displays a strong molecular water band at 2.9 micrometers, but the lack of such a band in the spectrum of the 75-250 micrometers size range suggests that the KBr pellet is contaminated by introduced water. This water appears to be introduced as water absorbed on the very fine (<2 micrometers) particulate sample grains. We have tried to avoid this effect by grinding the samples under alcohol, but it remains a source of error in the spectral data. Thus, the broad water-absorption bands centered at 2.9 micrometers in transmittance should be used with caution to estimate sample water content and should be verified with reflectance data. It should also be noted that sample pellet and blank pellet sometimes do not contain an identical amount of KBr because of lost material squeezed past the die during pressing. This has caused a few transmittance spectra to exceed 100%, but it is a significant problem only at the short wavelength end of the spectrum, where the samples typically have very low absorbance. Sample spectra often display weak fundamental C-H stretching vibration bands (a triplet)near 3.3 micrometers. Solid samples typically have acquired fingerprints to account for their C-H bands, although some samples appear to have been contaminated by some oily lubricant used with a rock saw. Particulate samples may be contaminated by hydrocarbons in several ways. First, they were ground under alcohol or acetone as described above. Despite oven drying of these samples at 100¡ C, a trace of hydrocarbon may remain. We also find that particulate samples simply exposed to laboratory air will, over time, develop similar hydrocarbon bands due to adsorbed species. These spectral artifacts, like the water vapor bands, are usually so weak that they are difficult to detect at the scale at which spectra are presented in this book, but can be see, for example, in the fine particle size spectrum of grossular garnet (grossu2f). Samples contaminated with quartz were rejected, because quartz bands are so strong that even a small amount will introduce interfering bands. Samples contaminated with calcite (or dolomite) were treated with either acetic or hydrochloric acid to remove the contaminant. It was found that an exceedingly small amount of carbonate can usually be detected as a contaminant, because two of the carbonate overtone/combination band near 4 micrometers that gains spectral contrast at fine particle size occur where silicates tend to be transparent. The spectrum of the coarse 75-250 micrometers size range of epidote is a good illustration of this and also demonstrates that the infrared spectrum is more sensitive than X- ray analysis for detection of very small quantities of carbonate. Some minerals tend to contain exsolution products, and we have included several of these in our collection, despite their lack of perfect purity, because exsolution is so common. Microcline.1, for example, contains cryptoperthitic high albite, as does sanidine.3. Hypersthene.1 contains a small amount of exsolved clinopyroxene. As discussed above (see Major Spectral Features of Minerals), water is ubiquitous in terrestrial environments and terrestrial minerals subject to alteration have inevitably been slightly altered, even when quite fresh in appearance. Hydroxyl O-H stretching vibration bands tend to become more prominent (i.e., gain spectral contrast) as particle size decreases, as discussed above. At the same time, silicate minerals normally lacking hydroxyl tend to be quite transparent in the 2.6-3.0 micrometers region. Consequently, hydroxyl bands due to a very small amount of alteration product typically are very prominently expressed in reflectance spectra of the 0-75 micrometers particle-size range. Transmittance spectra give a better idea of the absolute amount of alteration product present. A good example can be seen in spectra of kyanite, which shows no hydroxyl bands superimposed on the broad water band in its transmittance spectrum near 2.7 micrometers, indicating the presence of very little hydroxyl, but does show very strong and distinctive kaolinite hydroxyl bands near 2.7 micrometers in the spectrum of the 0-75 micrometers particle-size range. Relating Laboratory Data to Remote Sensing Measurements: The crossover point (i.e., point of equal energy) for reflectance and emittance for most remote sensing targets in the solar system lies somewhere in the 2.5-5 micrometers region, depending on the albedo and temperature of the surface. Thus, most of the spectral range presented in this work is dominated by emittance. Yet, spectral data are presented in reflectance, primarily because reflectance is much easier to measure in the laboratory. Spectral emissivity is usually predicted from reflectance of opaque materials using Kirchhoff's Law, typically stated in its simplest form (without wavelength or directional subscripts) as E=1-R, where E and R are emissivity and reflectance (Nicodemus, 1965). Even in regions of relative transparency, where volume scattering dominates the powder spectra, the samples used in this study can be considered opaque because of their thickness. However, the direct application of Kirchhoff's Law to the spectral library presented here is not possible because measurements were made in biconical, rather than hemispherical, reflectance. This means that the library spectra do not provide a quantitative measure of the infrared radiation scattered in all directions, as do the relatively few directional hemispherical reflectance spectra described in the next section. However, laboratory and field measurements have repeatedly shown that biconical reflectance measurements may be used qualitatively to predict emissivity (Lyon, 1964; Hunt and Vincent, 1968, Bartholomew et al., 1989). That is, the shape of the spectral curves in our mineral library can be used to predict the shapes, but not the absolute intensities, of spectral curves in emittance. Thus, our library of mineral spectra may be used qualitatively in spectral searches of remote sensing data to identify unknown minerals. Quantitative prediction of the magnitude, as well as the kind, of spectral features to be seen in spectral emittance requires directional hemispherical reflectance measurements, such as those described in the next section. Such measurements capture both the forward and backscattered radiation, so that an absolute measurement is obtained of R. One caveat, however, is that Kirchhoff's Law holds true only under isothermal conditions. As shown by Logan et al. (1973), a vacuum environment fosters a thermal gradient in the uppermost few micrometers of a particulate sample that may exaggerate spectral contrast and distort spectral features. As shown by Salisbury and Walter (1989), however, systematic correction factors can be developed to predict the effect of a vacuum environment on spectral features. Fortunately, the presence of an atmosphere, even a thin martian-type atmosphere, leaves the thermal gradient and spectral contrast little changed from terrestrial conditions. Thus, it appears that the biconical reflectance spectra illustrated in this book may be used to qualitatively predict spectral emissivity in a variety of potential remote sensing environments. Mineral Mixing Models: Using quantitative directional hemispherical reflectance spectra, Thompson and Salisbury (1991) have recently shown that the spectra of silicate minerals in the reststrahlen region combine linearly to make up the spectra of mineral mixtures for particles large enough to be optically thick (e.g., the 75-250 micrometers size range). References Aines, R.D., and Rossman, G.R., 1984, Water in Minerals? A peak in the infrared: Jour. of Geophysical Research, v. 89, p. 4059-4071. Appleman, D.E., and Evans, H.T., Jr., 1973, Job 9214 Indexing and least-squares refinement of powder diffraction data: U. S. Department of Commerce, NTIS Document PB-216188. Aronson, J.R., Emslie, A.G., and McLinden, H.G., 1966, Infrared spectra from particulate surfaces: Science, v. 152, p. 345-346. Bailey, S.W., 1980, Structures of layer silicates: in Crystal Structures of Clay Minerals and Their X-ray Identification, G.W. Brindley and G. Brown, eds., Mineralogical Society, London, p. 1-124. Bartholomew, M.J., Kahle, A.B., and Hoover, G., 1989, Infrared spectroscopy (2.3-20 micrometers) for the geological interpretation of remotely-sensed multispectral thermal infrared data: International Jour. of Remote Sensing, v. 10, p. 529-544. Conel, J.E., 1969, Infrared emissivities of silicates: Experimental results and a cloudy atmosphere model of spectral emission from condensed particulate mediums: Jour. of Geophysical Research, v. 74, p. 1614-1634. Clark, R.N., and Roush, T.L., 1984, Reflectance spectroscopy: Quantitative analysis techniques for remote sensing applications: Jour. of Geophysical Research, v. 89, p. 6329-6340. Drever, J.I., 1973, The preparation of oriented clay mineral specimens for X-ray diffraction analysis by a filter membrane peel technique: American Mineralogist, v. 58, p. 553-554. Elcombe, M.M., 1967, Some aspects of the lattice dynamics of quartz: Proceedings of the Physics Society of London, v. 91, p. 947-958. Estep-Barnes, P.A., 1977, Infrared spectroscopy: Chapter 2 in Physical Methods in Determinative Mineralogy, 2nd edition, J. Zussman, ed., Academic Press, New York, p. 529-603. Farmer, V.C., ed., 1974, The Infrared Spectra of Minerals. Monograph No. 4, Mineralogical Society (London), 539 pp. Farmer, V.C., and Palmieri, F., 1975, The characterization of soil minerals by infrared spectroscopy: In Soil Components, v. 2 Inorganic Components, J. E. Gieseking, ed.,Springer-Verlag, New York, p. 573- 670. Ferraro, J.R., ed., 1982, The Sadtler Infrared Spectra Handbook of Mineral and Clays: Sadtler Research Laboratories, Philadelphia, 476 pp. Fleischer, Michael, 1983, Glossary of Mineral Species 1983: The Mineralogical Record, Inc., Tucson, 202 pp. Hapke, Bruce, 1981, Bidirectional reflectance spectroscopy: 1. Theory: Jour. of Geophysical Research, v. 86, p. 3039-3054. Harrington, P.B., and Isenhour, T.L., 1987, A quantitative measure of the reliability of searches of spectral libraries: Analytica Chimica Acta, v. 197, p. 105-119. Hass, M., and Sutherland, G.B.B.M., 1956, The infra-red spectrum and crystal structure of gypsum: Proceedings of the Royal Society, (London), v. 236, p. 427-445. Hoover, Gorden, and Kahle, A.B., 1986, A Portable Spectrometer for Use from 5 to 16 Micrometers: Jet Propulsion Laboratory Publications 86-19, California Institute of Technology, Pasadena, 53 pp. Hovis, W.A. and Callahan, W.R., 1966, Infrared reflectance spectra of igneous rocks, tuffs and red sandstone from 0.5 to 22 micrometers: Jour. of the Optical Society of America, v. 56, p. 639-643. Hunt, G.R., and Logan, L. M., 1972, Variation of single particle mid- infrared emission spectrum with particle size: Applied Optics, v. 11, p. 142-147. Hunt, G.R., and Salisbury, J.W., 1974, Mid-infrared Spectral Behavior of Igneous Rocks: Air Force Cambridge Research Laboratories Technical Report, TR-74-0625, 143 pp. Hunt, G.R., and Salisbury, J.W., 1975, Mid-infrared Spectral Behavior of Sedimentary Rocks: Air Force Cambridge Research Laboratories Technical Report, TR-75-0356, 49 pp. Hunt, G.R., and Salisbury, J.W., 1976, Mid-infrared Spectral Behavior of Metamorphic Rocks: Air Force Cambridge Research Laboratories Technical Report, TR-76-0003, 67 pp. Hunt, G.R., and Vincent, R.K., 1968, The behavior of spectral features in the infrared emission from particulate surfaces of various grain sizes: Jour. of Geophysical Research, v. 73, p. 6039-6046. Kahle, A.B., and Goetz, A.F.H., 1983, Mineralogic information from a new airborne thermal infrared multispectral scanner: Science, v. 222, p. 24-27. Logan, L.M., Hunt, G.R., and Salisbury, J.W., 1974, The use of mid- infrared spectroscopy in remote sensing of space targets: in Infrared and Raman Spectroscopy of Lunar and Terrestrial Minerals, C. Karr, ed., Academic Press, New York, p. 117-142. Logan, L.M., Hunt, G.R., Salisbury, J.W., and Balsamo, S.R., 1973, Compositional implications of Christiansen frequency maxima for infrared remote sensing applications: Jour. of Geophysical Research, v. 78, p. 4983-5003. Lowry, S.R., and Huppler, D.A., 1981, Infrared spectral search system for gas chromatography/Fourier transform infrared spectrometry: Analytical Chemistry, v. 53, p. 889-893. Lyon, R.J.P., 1963 Evaluation of Infrared Spectrophotometry for Compositional Analysis of Lunar and Planetary Soils: Stanford Research Institute, Final Report under Contract NASr-49(04) and NASA Technical Note D-1871, 139 pp. Lyon, R.J.P., 1964, Evaluation of Infrared Spectrophotometry for Compositional Analysis of Lunar and Planetary Soils. Part II: Rough and Powdered Surfaces: NASA Contractor Report CR-100, 262 pp. Lyon, R.J.P., and Green, A.A., 1975, Reflectance and emittance of terrain in the mid-infrared (6-25 micrometers) region: in Infrared and Raman Spectroscopy of Lunar and Terrestrial Minerals, C. Karr, ed., Academic Press, New York, p. 165-195. Nash, D.B., and Salisbury, J.W., 1991, Infrared reflectance spectra of plagioclase feldspars: Geophysical Research Letters, v. 18, p. 1151- 1154. Nicodemus, F.E., 1965, Directional reflectance and emissivity of an opaque surface: Applied Optics, v. 4, p. 767-773. Reynolds, R.C., 1980, Interstratified clay minerals: in Crystal Structures of Clay Minerals and their X-ray identification, G.W. Brindley and G. Brown, eds., Mineralogical Society ( London), p. 249- 303. Roedder, Edwin, 1984, Fluid Inclusions: Reviews in Mineralogy, v. 12, Mineralogical Society of America, Washington, 644 pp. Salisbury, J.W., and Eastes, J.W., 1985, The effect of particle size and porosity on spectral contrast in the mid-infrared: Icarus, v. 64, p. 586-588. Salisbury, J.W., and Walter, L.S., 1989, Thermal infrared (2.5 to 13.5 micrometers) spectroscopic remote sensing of igneous rock types on particulate planetary surfaces: Jour. of Geophysical Research, v. 94, p. 9192-9202. Salisbury, J. W., D'Aria, D. M., and Jarosewich, E. J., 1991, Midinfrared (2.5-13.5 micrometers) reflectance spectra of powdered stony meteorites: Icarus, v. 92, p. 280-297. Salisbury, J.W., Hapke, Bruce, and Eastes, J.W., 1987b, Usefulness of weak bands in mid-infrared remote sensing of particulate planetary surfaces: Jour. of Geophysical Research, v. 92, p. 702-710. Salisbury, J.W., Walter, L.S., and Vergo, Norma, 1987a, Mid-Infrared (2.1-25 micrometers) Spectra of Minerals: First Edition: U.S. Geological Survey Open-File Report 87-263, Reston, 386 pp. Simon, I., and McMahon, H. O., 1953, Study of the structure of quartz, cristobalite, and vitreous silica by reflection in the infrared: Journal of Chemical Physics, v. 21, p. 23-30. Stimson, M.M., and O'Donnell, M.J., 1952, The infrared and ultraviolet spectra of cystine and isocystine in the solid state: Jour. of the American Chemical Society, v. 74, p. 1805-1808. Stubican, V., and Roy, Ruston, 1964, A new approach to assignment of infra-red absorption bands in layer-structure silicates: Zeitschrift fur Kristallographie, v. 115, p. 200-214. Thompson, J. L., and Salisbury, J. W., 1991, The mid-infrared reflectance of mineral mixtures (7-14 micrometers): Remote Sensing of Environment, (in press). Vincent, R.K., and Hunt, G.R., 1968, Infrared reflectance from mat surfaces: Applied Optics, v. 7, p. 53-59. Vincent, R.K., Rowan, L.C., Gillespie, R.E., and Knapp, Charles, 1975, Thermal-infrared spectra and chemical analyses of twenty-six igneous rock samples: Remote Sensing of Environment, v. 4, p. 199- 209. Walter, L.S., and Salisbury, J.W., 1989, Spectral characterization of igneous rocks in the 8 to 12 micrometers region: Jour. of Geophysical Research, v. 94, p. 9203-9213.