- Copyright ©2012. The American Association of Petroleum Geologists. All rights reserved.
The microstructure of gas shale samples from nine different formations has been investigated using a combination of focused ion beam (FIB) milling and scanning electron microscopy (SEM). Backscattered electron (BSE) images of FIB cross sectioned shale surfaces show a complex microstructure with variations observed among the formations. Energy dispersive spectroscopy of the shale cross sections indicates that clay, carbonate, quartz, pyrite, and kerogen are the most prevalent components. In the BSE images, areas of kerogen are observed interspersed with the inorganic grains. Pores are observed in both the kerogen and inorganic matrix with the size, shape, and number of pores varying among the shale samples. By using FIB milling and SEM imaging sequentially and repetitively, three-dimensional (3-D) data sets of SEM images have been generated for each of the shale samples. Three-dimensional volumes of the shales are reconstructed from these images. By setting thresholds on the gray scale, the kerogen and pore networks are segmented out and visualized in the reconstructed shale volumes. Estimates of kerogen and pore volume percentages of the reconstructed shale volumes have been made and range from 0 to 90.0% for the kerogen and 0.2 to 2.3% for pores. Estimates of pore-size distributions suggest that although pores with radii of approximately 3 nm dominate in number, they do not necessarily dominate in total volumetric contribution. Scanning electron microscopy images and 3-D reconstructions reinforce the facts that shales are quite different and that their microstructures are highly variable and complex.
With increasing demand for secure and stable sources of natural gas, significant attention is being given to unconventional gas reservoirs, such as tight sands, coalbed methane, and gas shales. Gas shales have garnered particular interest in part because of the success of such shale plays as the Barnett, Marcellus, Haynesville, and Woodford. Shale is commonly only considered a source rock, in the case of organic-rich black shales, and seals for conventional gas reservoirs. However, for gas shales, the shale itself is the reservoir. Because gas shales have historically been ignored, the microstructure of these shales remains mostly unexplored. Advances in drilling and completion technology, specifically horizontal drilling and hydraulic fracturing, have made production economic. Understanding the microstructural controls on porosity and permeability has far-reaching implications. Porosity controls the amount of gas in place. In addition, as pore sizes approach the nanoscale, the surface area becomes significant and strongly affects the amount of gas stored as the free component versus the adsorbed component (Ambrose et al., 2010). The shapes of pores will directly affect the structural integrity and how the pore space responds to changes in stress. Connectivity of the pore space is crucial in shale's native unfractured ability to deliver gas to the borehole. Matrix permeability controls the spacing between fracture stages as well as deliverability.
Assessments of pore sizes are typically made by indirect petrophysical methods such as mercury injection capillary pressure (MICP) measurements and nuclear magnetic resonance (NMR) spectroscopy. These macroscopic averaging methods, which do not involve direct observation of individual pores, suggest that pore sizes for gas shales are on the order of a few to tens of nanometers (Howard, 1991; Sondergeld et al., 2010). Direct imaging of nanoscale pores using optical microscopy is not possible; therefore, electron microscopy is used. Scanning electron microscopy (SEM) has proven to be a useful tool to image shale microstructure (Chalmers et al., 2009; Loucks et al., 2009; Moncrieff, 2009; Wang and Reed, 2009; Ambrose et al., 2010; Curtis et al., 2010; Passey et al., 2010; Schieber, 2010; Sondergeld et al., 2010); however, to date, a comparative study of the microstructure of major gas shales has not been published. To begin to develop a better picture of the microstructure and thus understand its effect on gas production, sampling many different shale formations is important. High-quality preparation and imaging techniques must also be applied to preserve the microstructure and observe its delicate constituents with minimal artifacts. To best prepare and image the microstructure of gas shale samples, a focused ion beam (FIB) and SEM incorporated onto the same sample chamber are used. In addition, the ability to do in-situ elemental analysis with energy dispersive spectroscopy (EDS) on FIB cross sectioned surfaces allows the mineralogy of the surfaces to be examined.
In this article, we describe the microstructure of samples from nine different shale formations using FIB milling and SEM imaging in tandem. The two-dimensional (2-D) cross sectional images produced show a complex and varying shale microstructure, with pores on the order of a few nanometers to hundreds of nanometers in diameter found in the kerogen and inorganic matrix. Serial sectioning and imaging of the shales have been used to produce sets of sequential SEM images that permit a three-dimensional (3-D) visualization of kerogen and pore connectivity in the shale. In addition, estimates of the kerogen volume percentage, porosity, and pore-size distribution have been calculated from these 3-D reconstructions for the nine samples. Such observations are crucial in understanding and modeling gas shales.
Gas shale samples were taken from the Barnett, Eagle Ford, Fayetteville, Floyd, Haynesville, Horn River, Kimmeridge, Marcellus, and Woodford shales. The location and age of these shales are given in Table 1. Samples were taken from unpreserved core, except for the Fayetteville and Kimmeridge samples, which were taken from an outcrop. The samples that were mounted to SEM stubs were 0.125-in. (0.317-cm) core plugs extracted using air-cooled coring or simply broken off the core or outcrop. The broken samples were polished to create a level surface using dry emery paper. Samples were mounted to SEM stubs using carbon paste and coated with gold and palladium to provide a conductive surface layer. Each sample was inserted individually into an FEI Helios NanoLab™ 600 DualBeam™ FIB-SEM. In this system, a focused 30-kV beam of gallium ions mills the samples by sputtering away shale material via momentum transfer. The SEM images the newly milled shale surface in situ with a resolution of 2.5 nm at 1 kV accelerating voltage when the working distance is set to the coincidence point (∼4 mm) of electron and ion beams. Sites were selected and prepared in cross section (perpendicular to the bedding planes) by FIB milling for SEM imaging, as illustrated in Figure 1A. The ion beam and electron beam are at a 52° angle to each other. A strip of platinum was deposited over the area of interest to minimize curtaining artifacts on the milled shale surface. Bulk shale was removed in front of and on the sides of the platinum (Pt) strip using a 21-nA ion current at 30 kV accelerating voltage, leaving a section of shale protruding from the bulk. The protruding surface, which is protected by the Pt strip, was milled in cross section at a reduced ion current to expose the microstructure of the shale, as seen in Figure 1B. Imaging on the cross sectional surfaces was performed using backscattered electrons (BSE) for atomic number contrast. A 1-kV accelerating voltage with a 0.34-nA beam current was chosen for electron imaging. Operating at low kV in BSE mode minimized charging effects caused by the nonconductive surface produced by the ion milling. The technique gave good surface detail because of the shallower interaction volume and provided good contrast between pores, organic matter, and inorganic minerals in the shale. Serial sectioning of the shale cross section was performed using the FIB and SEM sequentially and repetitively with Slice & View G2 automation software from FEI. For serial sectioning, a fiducial reference mark was put on the shale surface next to the site of interest for registration of the FIB. The FIB was used to remove a 10-nm-thick slice of the shale cross section, exposing the fresh surface underneath. This was followed by SEM imaging in BSE mode of the cross sectional surface. Each individual image consists of 32 integrated frames with a resolution of 2048 × 1768 pixels and a 300 ns dwell time. This milling and imaging procedure was repeated anywhere from 250 to 500 times, resulting in a sequence of SEM images that constituted a 3-D data set of the internal microstructure of the shale. Each data set was then imported into Avizo® Fire 6.2 imaging software, and a 3-D rendering of each shale was generated. A Gaussian smoothing filter was applied to each of the images that results in some degradation of spatial resolution. Subsequent quantitative analyses of porosity, kerogen volume percentage, pore-size distribution, and pore connectivity were conducted on the rendered volumes in the software by assigning gray-scale values to the microstructural features of interest and setting thresholds on the gray scale to segment the features. The thresholds on the gray-scale values were assigned visually for each feature such as to minimize any overlap with other microstructural features of similar gray scale.
RESULTS AND DISCUSSION
Two-Dimensional Scanning Electron Microscopy Imaging
Figure 2 shows 2-D BSE images taken of the nine different gas shale samples. The horizontal field width of each image is 5.12 μm, and the images are perpendicular to the bedding plane. From the images, some striking differences and similarities can be seen. Note that these images represent only a small snapshot of a much larger reservoir. Such images allow for the exploration of the microstructural details, but generalizations about the reservoir on larger scales require more statistical sampling. From the images shown, kerogen (dark gray) can be seen dispersed within a matrix of inorganic minerals (light gray to white). Black arrows have been added to the images to identify some regions of kerogen. Because of the low atomic number of carbon compared with higher atomic number elements in the inorganic minerals (e.g., Si, Fe, Ca, and Al), the kerogen is discernible from the inorganic minerals by its darker gray-scale value when imaging with BSEs. The EDS performed in situ on the cross sectioned shale surfaces confirms that the dark-gray areas are organic kerogen, with the lighter gray inorganic matrix being primarily composed of varying amounts of quartz, pyrite, carbonate, and clays. The amount of the kerogen observed varies among the samples, with the Kimmeridge sample containing the largest proportion of kerogen. In contrast, the Haynesville, Eagle Ford, and Fayetteville images appear to contain very little kerogen. Large regions of kerogen can be seen in the Barnett, Woodford, and Horn River shale images. Samples such as the Marcellus and Floyd contain much smaller kerogen regions that follow a flowing pattern with the inorganic grain microstructure. Some images such as the Woodford and Horn River suggest connectivity of the kerogen. In the Barnett, Marcellus, and Floyd images, pyrite can be seen (white) with kerogen internal to groupings of individual grains of pyrite. Thin layers of kerogen can also be seen surrounding pyrite grains, such as in the Barnett and Floyd images.
Pores were also observed in the cross sectioned shale samples; some of which are indicated by white arrows in Figure 2. Major differences in the size, shape, and location of the porosity can be seen. Porosity was found to fall into three classes: (1) cracklike, (2) phyllosilicate, and (3) organophyllic. The image of the Woodford Shale sample displays a cracklike pore that can be seen on the left side of the image running horizontally toward the middle of the image. In the inorganic matrix, the crack appears to flow around the grain boundaries but can be seen to cut through a region of kerogen. In the images of the Barnett, Woodford, and Horn River, a large number of small round organophyllic pores on the order of a few nanometers to tens of nanometers in diameter can be seen within the kerogen. Such pores can also be seen in the Kimmeridge, Marcellus, and Eagle Ford images but are fewer in number. In the Barnett and Horn River images, larger organophyllic pores on the order of a couple of hundred nanometers in diameter can be seen in the kerogen. In addition, the walls of these large pores are perforated with smaller diameter pores. Organophyllic porosity in the shales could play an important function in the permeability of the shale (Wang and Reed, 2009). Kerogen porosities more than 50% have previously been reported (Sondergeld et al., 2010). Large connected networks of highly porous kerogen could reach a percolation threshold and produce a pathway for movement of gas within the shale. In addition, the organophyllic pores should have a different wettability than those found in the inorganic matrix, with the organic pores likely being hydrocarbon wet and the inorganic pores being water wet or mixed wet (Passey et al., 2010; Sondergeld et al., 2010). Furthermore, the high porosities observed in the kerogen imply that bulk kerogen densities are low and thus significantly change the estimation of bulk volume of kerogen from weight percent of kerogen (Passey et al., 2010; Sondergeld et al., 2010).
The small and large round organophyllic pores seen are in contrast to the phyllosilicate pores found to dominate the inorganic matrix of the Haynesville sample. A lower magnification image of a Haynesville sample is also shown in Figure 3. The pores in the image appear to conform to grain boundaries in the inorganic matrix. These phyllosilicate pores tend to fall into two shape classifications: triangular or linear. The linear pores in the image have one dimension that is much smaller compared with the other with the shorter dimension commonly perpendicular to the bedding plane; a few such pores are indicated with white arrows in Figure 3. Such a pore shape and orientation suggest that this type of pore may be subject to collapse under lithostatic pressure as gas is removed from the pore space. Linear pores can also be seen in the Eagle Ford sample, which contains both round organophyllic pores and linear phyllosilicate pores.
Significant porosity was found to be associated with pyrite and apatite within the shales sampled. Porosity can be seen in the pyrite of the Barnett image in Figure 2. Figure 4A shows an SEM image of a pyrite framboid from a Horn River sample. In between the pyrite crystals that constitute the framboid, kerogen can be seen that contains organophyllic porosity. Figure 4B shows large grains of pyrite from another Horn River sample, which do not belong to a framboid. Porous kerogen can be seen around the grains. In addition, a few small clay platelets can be seen within a large region of kerogen between some of the pyrite crystals. Small pores in the kerogen can be seen, along with larger diameter pores. In addition to pyrite, porosity was also observed in apatite. Figure 5 shows this highly porous apatite region in an Eagle Ford sample. A grain of quartz at the bottom of the region and some clay platelets on the right can be seen intruding into the apatite layer. The porosity of this layer is not like that seen in any of the previous examples. It is not associated with any kerogen and does not appear to be occurring along any grain boundary. Unlike the linear and triangular phyllosilicate porosity observed in the Haynesville sample shown in Figures 2 and 3 or the round organophyllic porosity observed in some of the samples with kerogen in Figure 2, the porosity in the apatite region is very irregular in shape. By setting thresholds on the gray scale of an image of this apatite layer and segmenting out the pore space, the porosity of the apatite was estimated to be approximately 29%.
Three-Dimensional Shale Microstructure
Three-dimensional data sets of SEM images for each of the formations shown in Figure 2 were used to reconstruct shale volumes for analysis. A 3-D rendering of the Horn River sample is shown in Figure 6 as an example. Figure 6A shows a volume of shale reconstructed from 500 BSE images, with each image representing a 10-nm slice of shale. The size of the most basic element of the volume (voxel) is 2.5 × 2.5 × 10 nm, with the 10-nm dimension representing the thickness of each shale slice. By visually setting thresholds of the gray-scale values of this volume, kerogen and pores were segmented out. A surface was then drawn around these thresholded gray-scale values to visualize the kerogen and pore networks. Panels B and C of Figure 6 show a rendering of the kerogen and pore networks, respectively. The pyrite contained within the volume is shown in Figure 6D. The kerogen network is complex, and connectivity can be seen across the volume. The pore network shows connectivity but not as well developed as the kerogen network. The percent volume of kerogen and porosity of this reconstructed shale volume were estimated and found to be 7.5 and 1.5%, respectively. Reconstructions were done on all the samples shown in Figure 2, and the resulting estimates for the kerogen volume percentage and porosity are shown in Table 2. Helium porosity and total organic carbon (TOC) results are also given for the samples from which the SEM specimens were taken in addition to porosity and TOC values for the formations found in literature.
The number of individual pore-body sizes was also estimated from the reconstructed volumes. A histogram of the pore-size distribution for the Horn River reconstruction is shown in Figure 7A, along with the cumulative percent of the distribution. The sizes of the pores have been given as the radii of spheres of equivalent volume to each pore. The histogram suggests that smaller pores dominate the distribution with the smallest and most numerous pores having a radius of approximately 3 to 6 nm. These estimated pore sizes, along with those observed in 2-D SEM images, are dimensionally consistent with those reported by MICP and NMR (Howard, 1991; Sondergeld et al., 2010). Figure 7B shows a histogram of the total volumetric contribution of the pore sizes. Although Figure 7A shows that the small pores dominate in number, Figure 7B shows that the greatest volumetric contribution comes from pores that are approximately 100 nm in radius.
Four of the reconstructed shale volumes were explored further by analyzing the connectivity of the pore networks. Connected pore networks were segmented out in the Avizo® Fire software by setting a threshold for connectivity of the networks. The threshold for pore connectivity was set to greater than or equal to 105 neighboring pore voxels. Pore voxels were considered neighboring if they contacted at a face, edge, or corner of each voxel. Figure 8 shows the result of this segmentation in the Horn River, Woodford, Kimmeridge, and Haynesville samples. The images in the left column (Figure 8A) are of all the pores segmented out by thresholding the gray scale. The corresponding images in the right column (Figure 8B) are the pores that are connected at the 105 voxel level or greater. The numbers of individual regions of connected pore space for the Horn River, Woodford, Kimmeridge, and Haynesville are 14, 6, 3, and 14, respectively, and the percentages of the total pore volume that is connected are 26.5, 67.0, 51.8, and 66.7%, respectively.
From the four samples analyzed, no single connected pore network was observed to completely span across a dimension of the reconstructed volume. The connected porosity in the Horn River sample appears bulbous and disconnected across the reconstructed volume. A more intricate and longer range connectivity was observed within the reconstructed volume of the Woodford Shale. Part of this connectivity is caused by a crack in the Woodford sample that can be seen on the left side of the 2-D Woodford image in Figure 2. The overall lack of connectivity in the porosity of the Kimmeridge agrees well with the vuggy appearance of the pores, which can also be seen in the 2-D Kimmeridge image in Figure 2. In the Haynesville reconstruction, some of the connected porosity appears sheetlike because of the linear pore structure, which can be observed in the 2-D Haynesville images in Figures 2 and 3.
Note that a typical 3-D reconstruction ideally represents an approximately 125-μm3 volume. Because of this small sampling size, the amount of kerogen and porosity contained within a single small volume sampled cannot be taken to represent that of a larger volume of the shale, much less the formation as a whole. In addition, being aware of the processes involved in segmenting the reconstructed shale volumes and how they affect the estimates based on the segmentation is important. Setting threshold levels on the gray scale of the 3-D renderings is subjective, and how these thresholds are set will have an effect on the segmentation of the volume. This is especially important in the case of the smallest pores that are represented by a single voxel. By setting the threshold too high on the gray scale, small pores represented by just a single voxel can be overestimated. Conversely, underestimation of the number of the smallest pores can occur as some of the pores may be below the resolution of the image. In the case of very large pores that are typically on the order of 100 nm in diameter or larger, an underestimation of their number and volume contribution can occur. This is a result of BSE from the inside walls of such pores escaping the pore and being collected by the detector. This will raise the gray-scale level in that area of the image, causing the region not to be segmented out as pore space. These large pores will not contribute to the porosity estimates or connectivity of the porosity, thus causing an underestimation of both.
The 3-D reconstructions that were generated and the values estimated from them represent individual very small volumes of very large formations. This raises the question of how to upscale the information for larger shale volumes. One method of upscaling is to sample larger volumes of shale with a resolution still sufficient enough to observe the microstructural detail. With the current state of milling and imaging technology, the time involved to acquire such data as well as the amount of data itself would be large. Scanning electron microscopy images show nonuniformity of porosity on the scale of a few microns. What constitutes a representative elementary volume for shale, which can vary among different shales, will have to be determined. Another possibility would be to combine FIB-SEM tomography with lower resolution but larger field of view imaging methods such as x-ray microtomography. In addition, upscaling will require the sampling of multiple volumes from regions of interest. In the case of sampling many small volumes of a larger volume of shale, thought will have to be given to what constitutes a statistically significant sampling of the shale.
The microstructure of shale samples from nine different shale formations has been imaged and analyzed. The results show a complex microstructure, with similarities and differences existing among the different shale samples imaged. Energy dispersive spectroscopy showed that the milled shale cross sections are primarily composed of varying amounts of clay, quartz, kerogen, carbonate, and pyrite. Kerogen content was found to vary between the different shale samples, with Kimmeridge, Horn River, and Woodford having the most kerogen. Porosity was observed in both the kerogen and inorganic matrix. Pores tend to fall into three classes, being cracklike, organophyllic, or phyllosilicate in nature. Some pores are on the order of a few nanometers in diameter, whereas others are several hundreds of nanometers in diameter. Significant organophyllic porosity was seen in the Barnett, Horn River, and Woodford samples, whereas the Haynesville sample showed mainly phyllosilicate porosity. The Eagle Ford sample showed a mixture of organophyllic and phyllosilicate porosity. Porosity was also found to be associated with pyrite and apatite in some of the shale samples. Three-dimensional renderings were generated from serial sectioning and imaging of the shale samples and permitted visualization of kerogen and pore connectivity across the volumes. The estimated pore-size distributions from the volumes show that smaller pores with radii approximately 3 to 6 nm dominate in number but do not necessarily dominate in total pore-volume contribution. The ability to image and analyze the microstructure of gas shales is important to understanding how the microstructure controls key aspects of gas shales, such as gas in place, mechanical properties, and fluid flow through the shales.
We thank Devon Energy for its generous support of this project.
The AAPG Editor thanks the following reviewers for their work on this paper: Katherine “Lee” Avery, David N. Awwiller, and J. Steven Davis.
Color versions of Figures 6 and 8 may be seen in the online version of this article.
- Manuscript receivedNovember 23, 2010.
- Revised manuscript receivedFebruary 28, 2011.
- Revised manuscript receivedJuly 11, 2011.
- Final acceptanceAugust 15, 2011.
Mark E. Curtis is a postdoctoral research fellow in the Mewbourne School of Petroleum and Geological Engineering at the University of Oklahoma. He holds a Ph.D. in physics from the University of Oklahoma. He has more than 8 yr of experience in electron microscopy of nanostructures. His current research interests are in the effects of shale microstructure on gas storage and deliverability.
Carl H. Sondergeld is currently a professor and the Curtis Mewbourne Chair at the Mewbourne School of Petroleum and Geological Engineering, University of Oklahoma. He earned a Ph.D. in geophysics from Cornell University and B.A. and M.A. degrees in geology from Queens College, City University of New York. He spent 19 yr at the Tulsa Research Center of Amoco Production Company and holds 14 U.S. patents.
Raymond J. Ambrose is a Ph.D. student in petroleum engineering at the University of Oklahoma and Director of Reservoir Engineering for Reliance. He holds a B.Sc. degree in chemical engineering and an M.Sc. degree in petroleum engineering from the University of Southern California. His current research interests are analytical solutions for shale productivity, scanning electron microscopy and pore-structure characterization, and storage mechanisms for shale.
Chandra S. Rai received his Ph.D. in geology and geophysics from the University of Hawaii. He worked at the Amoco Production Company Research Center in Tulsa from 1981 to 1999. Currently, he is the Director and Eberly Chair Professor in the Mewbourne School of Petroleum and Geological Engineering, University of Oklahoma. His research interests include rock physics, reservoir characterization, and petrophysics.