The difficulty of uniquely determining mineralogy from chemical data alone (e.g., X-ray fluorescence or XRF data) is amply illustrated by the problems encountered in interpreting the Viking XRF data. The Viking XRF data provide multiple possible mineralogic fits to the chemistry, even after years of inquiry seeking to constrain the mineralogic interpretations with thermodynamic data (e.g., Clark and VanHart, 1981). This uncertainty in mineral identification is attributable largely to the presence of water, sulfur, and chlorine on the martian surface. With this mix of volatile and anionic species, the possible mineral forms that can occur dramatically increase in number. A simple normative calculation of mineralogy is therefore not possible. Indeed, the problem of determining extraterrestrial mineralogy from chemistry extends beyond Mars.

Recently, there has been wider recognition of the need for X-ray diffraction (XRD) mineral determinations in addition to the chemical data that can be obtained by XRF. For example, Fegley et al. (1992) cited our early CHEMIN concept as an approach to resolving mineralogic uncertainties for Venus. Any planetary, asteroidal, or cometary body with mineral constituents that contain water (including ices), sulfur, or halogens will require XRD in addition to chemical analysis (XRF) for adequate understanding of origin and evolution. The need for diffraction data in addition to chemistry led to the development of the combined XRD/XRF CHEMIN concept, in which the combined CHEmical and MINeralogical capability is reflected in the instrument name.

The origins and histories of planetary, asteroidal, and cometary bodies are reflected in their constituent minerals. The <100 naturally occurring elements combine together to form >3800 known minerals. This tremendous variety of mineralogy carries stories of pressure, temperature, oxygen fugacity, and solution chemistry, all intertwined with histories of sedimentation, igneous activity, metamorphism, impacts, and surface weathering. The science objectives of determining simultaneous chemistry and mineralogy thus span the full history of planetary, asteroidal, and cometary formation and evolution. We have had an active interest in the possibilities of miniaturized XRD/XRF instrumentation since 1990. Early concepts that we have pursued included both isotopic and tube-source instruments, as well as a variety of detector systems fitted to a transmission analytical geometry (Vaniman et al., 1991a,b; Blake et al., 1992). We have discussed elsewhere the applications and instrument designs of possible modern XRD/XRF instrumentation for planetary exploration.

CHEMIN is a miniaturized, CCD-based simultaneous X-ray diffraction/X-ray fluorescence (XRD/XRF) instrument. The device is designed to characterize elemental composition and mineralogy from small fine-grained or powder samples. The projected flight-instrument weight is <1 kg, the volume is 500 cc, and power requirement is 2 W. Usable diffraction data have been obtained from the prototype instrument in a few minutes, and flight-instrument data-collection times of 1-2 hours are predicted. Both diffraction and fluorescence data are obtained simultaneously by operating the CCD in single-photon counting mode. X-ray fluorescence data will be obtained for elements with 4<Z<92. Energy discrimination is used to distinguish between diffracted characteristic photons and fluorescence photons. Diffraction data are obtained in transmission mode, and resolution is presently sufficient to allow application of the Rietveld refinement method to the diffraction data. The Rietveld method is an advanced technique that allows interpretations of data by fitting a diffraction pattern, calculated from a knowledge of the component minerals, to the observed diffraction data. Such analyses provide a variety of chemical and crystal structural (mineralogic) information from a single diffraction pattern. In addition, because the CHEMIN concept combines diffraction and fluorescence data, simultaneous linear equations can be used to derive the most accurate, self-consistent mineral and chemical composition.

The prototype instrument currently in use at NASA Ames Laboratory is illustrated in Figure 1, and a schematic illustration of the CHEMIN flight instrument is provided in Figure 2. More detailed information is provided in Tables 1 and 2.

Figure 2 shows the spatial relationships between the three components that are critical to the operation of CHEMIN:

1. an X-ray tube source
2. a sample manipulation system
3. a CCD detector that discriminates both energy and position of X-rays

1). The X-ray source component is shown as a standard tube in Figure 2, reflecting one design option, but our preferred X-ray source is a sealed, high-vacuum envelope housing a micromachined field-emitter (MMFE) array, an electrostatic focusing element, and an anode, onto which a small focused beam of electrons is accelerated. The principle of micromachined field-emission sources was first described by Spindt (1971) and has been developed to a relatively advanced state for a number of uses such as flat-panel color displays (Spindt, 1992; Brodie and Schwoebel, 1994). In operation, a continuous electrostatic potential of 20 keV is applied between the anode and the MMFE array. An electron current is created by impressing a potential of ~50 V on the extraction anode of the MMFE array. The electron flow is turned on and off by impressing and removing this potential. The electron current leaving the MMFE is focused onto the anode using an electrostatic focusing system. The throughput of the system is limited by electron-beam heating of the target, which will be kept below the melting point of the metal. When electrons strike the target, characteristic and continuum X-rays are generated which propagate in all directions. Small fractions of these travel in the forward-scattered direction toward a 50-µm aperture. This aperture has the effect of creating a nearly parallel 50-µm beam of continuous and characteristic radiation from the anode.

2). The Sample Collection Disc or Carousel. The collimated X-ray beam strikes a thin-film substrate held in the second component of the system, a multi-position sample carousel. The carousel disc is the only moving part in the CHEMIN instrument. The carousel can be precisely and continuously rotated about 360 degrees by a stepper motor. The substrates are X-ray thin (~2 µm) mylar films made sticky on the top surface with a thin film of vacuum grease or the like. A protective cover is removed from the coated mylar as it advances into the sample collection port. A sample is dumped into the sample collection port and powder adheres to the sticky surface of the mylar, while excess powder and larger grains roll off the inclined substrate. The carousel is then rotated into position for analysis between the X-ray source and CCD detector array. While in the position for analysis, 50-µm diameter regions are illuminated by the collimated X-ray beam. Approximately 50 exposures are collected of each area, after which the motor steps ~100 µm so that a new area of sample material on the same substrate is analyzed. An "exposure" is made by providing a pulse of 50 volts to the extraction anode of the X-ray tube, flooding the CCD with diffracted and fluoresced X-ray photons.

3). CCD Detector Array. The third component of CHEMIN is the CCD detector array and detector electronics. The array consists of a matrix of 1024 X 1024 pixels, in a front-illuminated, thin polygate device. Salient characteristics of the device are listed in Table 1. In most applications CHEMIN can be operated in shaded or in local nighttime conditions, when the ambient temperature is sufficiently low to reduce background to an acceptable level.

Table 1. Characteristics of the proposed CHEMIN CCD

The CCD proposed for use in CHEMIN has already been used in X-ray astronomy applications (e.g., CUBIC). The principal characteristics of the CUBIC device are listed in Table 1. Three characteristics of this CCD are uniquely suited to the needs of a combined XRD/XRF instrument:

A thin-gate design is incorporated. A CCD consists of a layer of single-crystal, photon-sensitive silicon that is overlain by an insulator and polysilicon electrodes. These overlying layers (total thickness ~ 500 nm) absorb a portion of the incident energy, with absorption being most severe at energies less than 1 keV. For some instruments, enhanced soft X-ray sensitivity has been achieved by removing the thick, insensitive silicon substrate on which the CCD is fabricated and illuminating the detector from the backside. Although thinning and backside illumination do significantly improve low-energy sensitivity, they also produce significant drawbacks, including increased charge splitting between pixels and partial charge collection. The CCD to be used in the CHEMIN instrument is unthinned and front-side illuminated. A unique pixel design is employed to provide enhanced low-energy X-ray response .

Excellent charge transfer efficiency is maintained in the architecture. Individual X-ray photons generate very little charge (76 - 2050 electrons for the range of 277 - 7500 eV). Most CCDs do not transfer such small quantities of charge with sufficient efficiency to permit measurement of photon energy. The proposed CHEMIN CCD has been specifically designed (and has been demonstrated) to permit such measurement.

Table 2. Overall Instrument Parameters for CHEMIN

As described above, both diffraction and fluorescence data are obtained simultaneously by operating the CCD in single-photon counting mode. Figure 3 illustrates the flat-plate diffraction data obtained for Al₂O₃ (corundum) from CHEMIN, before radial integration. Further data processing, which involves integrating around each entire measured diffraction ring, produces a diffraction pattern like that shown in Figure 4, which illustrates a comparison between the CHEMIN diffraction pattern of quartz and the pattern obtained on a conventional laboratory diffractometer. Radial integration allows CHEMIN to obtain high-quality diffraction data from relatively coarse-grained samples.

Several instruments are available for remote chemical analysis of extraterrestrial surface samples, but there are none that determine mineral presence and abundance. Proposals for an XRD system with such capabilities have been advanced in the past , and instruments for collecting simultaneous XRD and XRF data have been suggested , but the systems envisioned then had limited capabilities. Both XRD and XRF technologies have advanced considerably in the past 30 years, allowing the possibility of a practical and effective combined and simultaneous XRD/XRF system for planetary exploration. Recent progress in X-ray detector technology allows the development of simultaneous XRD and XRF analysis in a system such as CHEMIN that is scaled down in size and power to the point where it can be mounted on fixed landers, soft penetrators, or small robotic rovers.

The quality of CHEMIN XRD data has now progressed to a point where trace mineral occurrences (<1%) can be identified and quantified. Such applications illustrate the high sensitivity of the prototype CHEMIN instrument in detecting very small quantities of diagnostic minerals. Contaminant detection that can distinguish between clastic, cemented clastic, and chemical sediments will be an important tool in remote petrologic and exobiologic studies. The ability to recognize and quantify minor minerals (present in only a few percent) will be important in these studies.

Table 3 compares two XRD analyses and a normative calculation for a terrestrial basalt sample. One of the XRD analyses was obtained with the NASA-Ames CHEMIN prototype instrument (Fig. 6); the other with a Siemens laboratory diffractometer. The basalt sample is a high-alumina basalt from the Southwest Nevada Volcanic Field. Although the CHEMIN instrument has poorer peak resolution, the data obtained with the CCD-based instrument are sufficient not only to determine the mineral phases present, but also to apply Rietveld analysis in estimating the relative proportions of mineral phases in the sample. Rietveld analysis involves calculating a model diffraction pattern based on the crystal structures of the known phases in a mixture. The quality of fit between the observed and calculated diffraction patterns is then improved via a least-squares process in which relative amounts of each phase and crystal structural parameters are varied. For minerals in which solid solution is possible, the atoms used to populate the cation sites can be varied to refine the fit, and unit-cell parameters determined by the analysis provide yet another key to solid solution variations. In addition, software is being developed to combine both the chemical XRF data and the XRD result to perform this operation more rigorously (Bish and Howard, 1988). The analysis represented in Fig. 6 uses only the XRD data. The goodness of fit for a Rietveld analysis of the XRD data is usually expressed by statistical parameters but can also be visualized as a difference pattern, shown at the bottom of Fig. 6.

The direct consideration of the ratios in which the calculated minerals are mixed to reproduce the observed pattern provides the weight proportions of minerals that best match the measured XRD data. These weight proportions are listed in Table 3, calculated for both the CHEMIN XRD pattern as shown in Fig. 6 and for XRD data of the same sample collected on a laboratory diffractometer. The errors listed are calculated instrumental errors of approximately two sigma. Note that the relative error varies not only with mineral abundance but also with mineral type. Clearly some minerals abundances can be more precisely measured than others; for example, the feldspars are complicated not only by ternary cation exchange (Ca, Na, K) and associated framework Al/Si ratio variations, but in high-alumina basalts such as this sample they also form as phenocryst and groundmass phases, with separate exsolution histories, resulting in superposed complex mixtures of high-temperature and low-temperature structural types. Although this makes the feldspar calculations more error-prone and, in this sample, of greater abundance than should be present (compare the calculated normative value in Table 3), in fact the complexity of the pattern indicates a wealth of unused information. Experience with other complex rocks (Bish and Post, 1993) indicates that advances within the next decade should allow the extraction of far more information from CHEMIN-quality data than currently obtainable. It should also be noted that in some respects, the Rietveld data are more accurate than a normative calculation. The normative calculation is exceptionally sensitive to SiO₂ content and the norm shown in column (3) of Table 3 reports less olivine and more oxide minerals than the sample petrography allow. This reflects the magnitude of the errors commonly associated with instrumental determinations of SiO₂ content in rock samples. In this instance, the Rietveld determinations of mineral abundance are closer to reality. The XRD-based analysis also identifies phlogopite, which is not recognized in the normative calculation, and does not assume spurious occurrences of ilmenite and hematite.

A prototype CHEMIN instrument has been in operation since July of 1996. This prototype has verified the principle of CCD-based simultaneous XRD and XRF, but the instrument is composed of surrogate components that do not illustrate the full capabilities that can be realized with an optimized instrument. The prototype CHEMIN instrument has demonstrated detection and quantification of minerals at abundances as low as 1%, as well as the ability to provide quantitative chemistry and mineralogy from complex mixtures (Table 3).

The CHEMIN instrument can be demonstrated as a series of subsystems:

1) Field-emission X-ray tube

2) CUBIC-quality CCD

3) Rock-powder preparation system

David Bish received a Ph.D. in Mineralogy and Petrology from Pennsylvania State University in 1977. After graduation, he completed three years of postdoctoral studies in the Department of Geological Sciences, Harvard University. Since 1980, he has been a research scientist in the Earth and Environmental Sciences Division of Los Alamos National Laboratory where he presently serves as Project Leader for Yucca Mountain Mineralogy-Petrology studies. He has authored or co-authored more than 75 peer-reviewed articles, three books and 150 abstracts covering all aspects of mineralogy and diffraction theory. His areas of specialization include clay and zeolite mineralogy, X-ray powder diffraction, and Rietveld refinement applications. He has received numerous honors and awards, and presently serves as president of the Clay Minerals Society and as associate editor for Zeitschrift für Kristallographie, an international crystallography journal.

David Blake received a Ph.D. in Geology and Mineralogy from the University of Michigan in 1983. After graduation, he completed two years of postdoctoral studies in the Department of Geological Sciences, University of Michigan and an additional two years as a National Research Council fellow at NASA Ames Research Center. Since 1988, he has been a research scientist in the Space Sciences Division of NASA Ames Research Center. His areas of specialization include the characterization of extraterrestrial ice analogs and interplanetary dust particles by electron microscopy, diffraction and X-ray microanalysis, and new analytical and diffraction techniques. He has authored or co-authored 25 peer-reviewed articles, and edited or written chapters for two books.

Steve Chipera received an M.Sc. in Geology from the University of North Dakota in 1985. Since 1985, he has been a research scientist in the Earth and Environmental Sciences Division of Los Alamos National Laboratory where he serves as member of the Yucca Mountain Mineralogy-Petrology team. He has authored or co-authored more than 21 peer-reviewed publications, 2 book chapters, 11 conference proceedings, over 50 abstracts, and numerous Los Alamos Laboratory reports. His areas of specialization include X-ray powder diffraction, water sorption/desorption reactions in hydrous minerals, and thermodynamic modeling of mineral equilibria.

Stewart A. Collins received a B.S. degree from Principia College in 1969 and has been employed by the Jet Propulsion Laboratory since that time. He was imaging science experiment representative for the Mariner Mars 1969 and the Voyager missions, and was a member of the Voyager Imaging Science Team. He is supervisor of the JPL Imaging Systems Section's Advanced Development Group, which is responsible for the development of imaging detectors for various NASA space-imaging systems. His experience includes performance analysis of scientific instruments, mission operations planning and implementation for imaging investigations, conceptual design of CCDs and of CCD instruments, and the management of CCD fabrication procurements and of CCD test activities. He received NASA's Exceptional Service Medal in 1981. Mr. Collins has authored or co-authored more than 30 peer-reviewed articles.

David Vaniman received a Ph.D. in Earth Sciences from the University of California, Santa Cruz in 1976. From 1976-1979, he concentrated on lunar regolith studies as a postdoctoral researcher at the State University of New York, Stony Brook. He has been employed from 1979 to the present as a research scientist in the Earth and Environmental Sciences Division of Los Alamos National Laboratory. In this capacity, he managed the mineralogy-petrology research group for the Yucca Mountain Project (investigation of a potential high-level nuclear waste repository in Nevada) from 1981-1988 and 1992-1993. Dr. Vaniman has authored or co-authored more than 60 peer-reviewed articles and co-edited Lunar Sourcebook: A User's Guide to the Moon. His research interests include the mineralogy and geochemistry of igneous and metamorphic rocks and their alteration assemblages, soil mineralogy, lunar and planetary petrology, and instrumentation for planetary mineralogic and geochemical studies.