3D Volumetric Analysis of Fluid Inclusions Using Confocal Microscopy
Fluid inclusions preserve valuable information regarding hydrothermal, metamorphic, and magmatic processes. The molar quantities of liquid and gaseous components in the inclusions can be estimated from their volumetric measurements at room temperatures combined with knowledge of the PVTX properties of the fluid and homogenization temperatures. Thus, accurate measurements of inclusion volumes and their two phase components are critical. One of the greatest advantages of the Laser Scanning Confocal Microscopy (LSCM) in application to fluid inclsion analsyis is that it is affordable for large numbers of samples, given the appropriate software analysis tools and methodology. Our present work is directed toward developing those tools and methods. For the last decade LSCM has been considered as a potential method for inclusion volume measurements. Nevertheless, the adequate and accurate measurement by LSCM has not yet been successful for fluid inclusions containing non-fluorescing fluids due to many technical challenges in image analysis despite the fact that the cost of collecting raw LSCM imagery has dramatically decreased in recent years. These problems mostly relate to image analysis methodology and software tools that are needed for pre-processing and image segmentation, which enable solid, liquid and gaseous components to be delineated. Other challenges involve image quality and contrast, which is controlled by fluorescence of the material (most aqueous fluid inclusions do not fluoresce at the appropriate laser wavelengths), material optical properties, and application of transmitted and/or reflected confocal illumination. In this work we have identified the key problems of image analysis and propose some potential solutions. For instance, we found that better contrast of pseudo-confocal transmitted light images could be overlayed with poor-contrast true-confocal reflected light images within the same stack of z-ordered slices. This approach allows one to narrow the interface boundaries between the phases before the application of segmentation routines. In turn, we found that an active contour segmentation technique works best for these types of geomaterials. The method was developed by adapting a medical software package implemented using the Insight Toolkit (ITK) set of algorithms developed for segmentation of anatomical structures. We have developed a manual analysis procedure with the potential of 2 micron resolution in 3D volume rendering that is specifically designed for application to fluid inclusion volume measurements.
X-Ray Tomography of Diamondiferous Eclogites: Clues to the Origin of Diamonds.
During the last decade or so, considerable new and significant data have been gathered concerning the origin of diamonds. This has come from the mantle xenoliths that are the rocks in which the diamonds originated, namely eclogites and peridotites, the host rocks for diamonds in the mantle. Upon rising through the crust to the surface in their kimberlite magmatic carriers and subsequent weathering on the surface, the weak olivine commonly alters, thereby reducing the crushing strength of the peridotite xenoliths. However, the eclogite xenoliths often retain enough toughness to resist total shattering after initial crushing during diamond recovery process. Subjecting these eclogite nodules to X-rays (e.g., 1.54 Å Cu K) can reveal the bright-blue fluorescence of any diamonds exposed at the surface of the xenoliths. Slow and careful extraction can result in recovery of large diamonds. Many of these unique rocks are the ones upon which we have performed High- Resolution X-ray Computed Tomography (HRXCT) at UT Austin. These data have formed the basis for further eclogite dissections and diamond polishing at UT Knoxville. The size of the diamondiferous eclogites that were scanned by HRXCT are from 20 g to 8.8 kg, all with many diamonds ranging from <1 mm to >1 cm, most octahedral, several with mineral inclusions. These diamondiferous eclogites have both textures and fabrics that provide evidence indicating the secondary formation of the diamonds. These include lineations of diamonds along zones of metasomatic alteration, former zones of enhanced permeability; the non-association of sulfides (po, pn, cpy) with the diamonds, versus sulfide-immiscible melt for the diamond origin; lack of any diamonds in direct contact with the primary garnets or clinopyroxenes; and the presence of some dodecahedral diamonds, indicative of resorption processes, typically attributed to the kimberlite melt. Indeed, these eclogites are not igneous in origin, but metamorphic products of their subducted oceanic crustal protoliths. In addition, detailed studies of the extracted diamonds supply more unexpected results. Polished diamonds examined with cathodoluminescence show evidence for a torturous life of cubic nucleation, growth, resorption, octahedral growth, resorption, and even plastic deformation. These are anomalous observations of diamonds supposedly formed along with the primary minerals in their eclogite hosts. The mineral inclusions in the diamonds reveal additional compelling clues of their origin. Multiple clinopyroxene inclusions can have different compositions within a single diamond, different between diamonds, and even different from that of the host. It is proposed that diamonds present in mantle eclogite xenoliths are secondary, having little to do with their hosts, formed by metasomatic fluids penetrating the eclogites along zones of permeability, and causing extensive secondary alteration and even partial melting of the primary garnets and omphacites.
New results achieved by high resolution neutron computed tomography
In recent years, considerable progress has been made in increasing the spatial resolution in Neutron Computed Tomography (N-CT), now achieving a resolution of 20-30 μm. This is a level comparable with standard X-ray CT methods. Unlike X-rays, cold neutrons provide good contrast for light elements and are capable of penetrating thicker layers of heavy elements more easily. It is hoped that the complimentary information provided by the NCT method can be combined with X-ray reconstructions to better characterise geomaterials in terms of crystal phases, glass, porosity and cracks. The study of syntectonic magmatic rocks provides important information about large-scale crustal processes, such as formation of deep-rooted faults/thrusts, and the compositional/structural re-organization of the continental crust in general. However, emplacement and crystallization of granitoid melts in regional stress fields, though very common processes, are not sufficiently understood. Magmatic structures are generally too coarse for thin-section analysis, too diffuse for precise conventional measurements, and often show irregular geometry which does not allow extrapolation from smaller to larger scales or from 2d to 3d. Neutron tomography offers, to date, the only sufficiently fast and precise method for 3d micrometer resolution of these magmatic structures (over sufficiently large scales: millimeter- to centimeter-scale). Such data sets form a strong and novel basis for quantification of rock fabrics. This is expected to lead to a deeper understanding of fabric-forming processes. A syntectonic tonalite from the fossil Hercynian lower crust of Calabria/Italy has been investigated. Because of good contrast of density, crystal structure, and composition between mafic (OH-bearing biotite, amphibole) and felsic (quartz, plagioclase) micrometer- to millimeter-sized phases, neutron tomography of 2x2x2 centimeter samples leads to 3d gray-shade patterns with clear contrasts and resolution on the 30- μm scale. These reconstructions were then analysed with the automated pattern quantification code, AMOCADO. This allowed us to quantitatively study the anisotropy in the marfic and felsic crystal phases. Once the NCT is done, the measured samples are milled down in 50 μm steps in order to obtain precise information about congruence between gray-shade patterns and crystal distribution patterns of various types of minerals. The resulting 400 2d patterns are digitized and transferred into mineral distribution patterns. These patterns are stacked and geostatistically interpolated, resulting in a voxel array. In addition, thin sections are prepared from defined positions within the samples. All relevant information taken from the thin sections allow comparison with the results of neutron tomography at high precision. The initial results from these comparative investigations are presented along with interpretations of pattern-forming processes and rock history.
X-ray microtomography at high pressure
X-ray microtomography at high pressure is now possible with the rotating anvil apparatus (RAA) on the 13-BM- D beamline at the Advanced Photon Source (Argonne National Lab). The high-pressure X-ray tomography microscope (HPXTM) can be used to determine densities of amorphous materials (glasses and melts) and in situ characterization of 3D microstructure of multiphase materials subject to temperature and shear deformation [1, 2]. Densities may be obtained directly by volume rendering or from X-ray absorption. The rotating anvil apparatus is compressed by a 250-ton hydraulic press between concentric thrust bearings. Toroidal and truncated cylindrical (Drickamer) anvils can be accommodated. The latter anvils perform well up to 11.5 GPa and 1873K, using boron epoxy/diamond epoxy gaskets and X-ray transparent aluminum or polytherimide plastic containment rings. Differential rotation allows for controlled sample deformation. Pressure is determined by energy dispersive diffraction of an internal standard by convenient switching from monochromatic and polychromatic radiation. In-situ calibrations of linear attenuation coefficient permit bracketing of natural basalt density to better than 1 percent relative, while  used volume rendering to determine the compressibility of magnesium silicate glasses and supercooled liquid. The utility of the RRA to characterize microstructural evolution will be discussed.  Wang et al., Rev. Sci. Instrum., 76, 073709, 2005.  Lesher et al., PEPI, in press, DOI: 10.1016/j.pepi.2008.10.023, 2009
Determining Crystal Size Distributions Over a Wide Range of Crystal Sizes From Multiple Tomographic Runs
X-ray tomography is a valuable method for determining crystal size distributions, in particular for pumiceous samples. One of its main limitations is a common trade off between sample size and image resolution: due to the characteristics of the experimental setup and/or time restrictions, image resolution in tomographic runs tends to be proportional to sample size. Consequently, large samples are generally imaged at low resolution, and high resolution is only possible with small samples. Such limitations make it a challenge to study a wide range of crystal sizes using x-ray tomography. In an effort to remedy this problem, we have developed a technique by which we combine multiple tomographic datasets at various resolutions to derive crystal size distributions that span a large range of crystal sizes. Cylinders of systematically varying size are cut from a single sample. Each cylinder is analyzed at a different resolution: the largest cylinders are analyzed at the lowest resolution, and the smallest cylinders at the highest resolution. A crystal size distribution is determined for each cylinder using image processing techniques, and the results are compared. An overall CSD is created by combining the individual CSDs, such that data for each bin size are obtained from the resolution that best quantifies crystals of that size. The result is a CSD that spans an extensive range of crystal sizes, up to two or three orders of magnitude. This technique was applied to a study of five pumice clasts of late-erupted Bishop Tuff. Five cylinders were cut from each pumice clast. The largest cylinders (A Runs) were 10 mm in diameter and analyzed at the lowest resolution (17 μm/voxel). Progressively smaller cylinders-B, C, and D Runs-were cut with systematic two-fold reductions in diameter (resulting in eight-fold reductions in volume), so as to allow imaging at increasingly higher resolution. The resultant sizes and resolutions were 5 mm at 8.5 μm/voxel, 2.5 mm at 4.25 μm/voxel, and 1 mm at 2.5 μm/voxel, respectively. The five individual CSDs for a given pumice clast were compared and, in each distribution, an obvious drop in the abundance of small crystals was apparent. In all cylinders, this limit occurred at a crystal size equivalent to four voxels. For example, A Run size distributions contained crystals up to 1000 μm, but few-to-no crystals <60 μm (ca. 4 voxels at 17 μm/voxel), while B Run size distributions revealed large numbers of crystals <60 μm. Clearly, crystals <60 μm existed but were not quantifiable at A Run resolution. In contrast, B, C, and D cylinders were small such that they were less likely or unable to contain larger crystals, and A Runs proved much superior for evaluating the abundance of large (i.e. >100 μm) crystals. Based on this information, the individual CSDs were combined to create an overall CSD, with information on crystals of sizes ∼10-1200 μm. Data for the largest bin sizes (70-1140 μm) were derived from A Runs, while smaller bin sizes of 35-70 μm, 17.5-35 μm, and 8.75-15.7 μm were taken from B, C, and D Runs, respectively. Combining multiple volumes of the same sample, imaged at various resolutions, is a powerful way to quantify crystal size distributions and other textural parameters over a large range of crystal sizes.
'tomo_display' and 'vol_tools': IDL VM Packages for Tomography Data Reconstruction, Processing, and Visualization
One of the challenges in tomography is the availability of suitable software for image processing and analysis in 3D. We present here 'tomo_display' and 'vol_tools', two packages created in IDL that enable reconstruction, processing, and visualization of tomographic data. They complement in many ways the capabilities offered by Blob3D (Ketcham 2005 - Geosphere, 1: 32-41, DOI: 10.1130/GES00001.1) and, in combination, allow users without programming knowledge to perform all steps necessary to obtain qualitative and quantitative information using tomographic data. The package 'tomo_display' was created and is maintained by Mark Rivers. It allows the user to: (1) preprocess and reconstruct parallel beam tomographic data, including removal of anomalous pixels, ring artifact reduction, and automated determination of the rotation center, (2) visualization of both raw and reconstructed data, either as individual frames, or as a series of sequential frames. The package 'vol_tools' consists of a series of small programs created and maintained by Guilherme Gualda to perform specific tasks not included in other packages. Existing modules include simple tools for cropping volumes, generating histograms of intensity, sample volume measurement (useful for porous samples like pumice), and computation of volume differences (for differential absorption tomography). The module 'vol_animate' can be used to generate 3D animations using rendered isosurfaces around objects. Both packages use the same NetCDF format '.volume' files created using code written by Mark Rivers. Currently, only 16-bit integer volumes are created and read by the packages, but floating point and 8-bit data can easily be stored in the NetCDF format as well. A simple GUI to convert sequences of tiffs into '.volume' files is available within 'vol_tools'. Both 'tomo_display' and 'vol_tools' include options to (1) generate onscreen output that allows for dynamic visualization in 3D, (2) save sequences of tiffs to disk, and (3) generate MPEG movies for inclusion in presentations, publications, websites, etc. Both are freely available as run-time ('.sav') versions that can be run using the free IDL Virtual Machine TM, available from ITT Visual Information Solutions: http://www.ittvis.com/ProductServices/IDL/VirtualMachine.aspx The run-time versions of 'tomo_display' and 'vol_tools' can be downloaded from: http://cars.uchicago.edu/software/idl/tomography.html http://sites.google.com/site/voltools/