Permeability and Electrical Resistivity Measurements

The nine minicore samples (diameter = 2.54 cm; length = 2 cm) studied were highly variable in terms of their permeabilities, porosities, textures, and physical properties (Table T1). Some were soft and crushed or collapsed during testing, while others were quite competent. All samples were saturated with a 31 g/L sea salt solution at room temperature using "sea salt" from Sigma Chemical Co. to produce the brine. Samples were saturated by suspending them in a vacuum over a cup of brine for 1 hr, placing them into the brine, and releasing the vacuum. The samples were then equilibrated with 60 cm3 of the brine in individual polypropylene cups prior to testing. The differences between saturated and dry densities were used to infer porosity for each sample.

Permeability and electrical resistivity were measured using an AutoLab 1000 test system capable of automated hydraulic control of confining and pore pressures. The samples were placed in a dual electrical resistivity/permeability core holder, which allows permeability- and frequency-dependent complex impedance to be measured during the same experiment. A complex transient method was used to measure permeability using the equipment and technique described in Boitnott (1997). A single-frequency sinusoid was employed as the primary transient. Frequency was tuned to optimize signal to noise for each sample. An asymmetrical spike transient (Boitnott, 1997) was used as backup in case an increase in permeability due to sample crushing occurred, but it was not needed for this suite of samples.

The samples were first pressurized to ~5 MPa effective confining pressure (confining pressure of 10 MPa and pore pressure of 5 MPa) and allowed to equilibrate to that stress. Permeability measurements were then performed, and appropriate measurement settings were selected. An automated measurement script was then run to perform the tests, making measurements of permeability at various points during loading and unloading. After each step increase or decrease in pressure, the controlled pressures were maintained constant and the downstream pressure was monitored until the effects of time-dependent compaction and pore pressure equilibration had passed. The pressure history for each sample is provided in Table T1.

Frequency-dependent electrical properties were measured using a Z-Meter impedance analysis system. A four-electrode configuration was used, and data were collected at frequencies ranging from 0.1 to 10000 Hz. Reported formation factors were compensated for small changes in temperature (typically a few degrees Celsius) resulting from adiabatic heating and cooling of the confining oil due to changes in confining pressure by correcting the brine conductivity measured at standard temperature (25C) using the Arps (1953) empirical relationship for NaCl brines. Pore fluid temperature was monitored by a thermocouple located in the pore pressure tubing at the top of the sample.

X-Ray Computed Tomography Measurements

X-ray computed tomography (CT) produces images in which grayscale corresponds to the X-ray linear attenuation coefficient, which is primarily a function of density and atomic number. Individual CT images are referred to as slices because they correspond to what would be observed if an object were sliced open along the scan plane. By acquiring a contiguous set of slices, data for a complete volume can be obtained. The fundamental CT data unit is the voxel, or volume element, which corresponds to the volume bounded by the edges of a pixel and the thickness of the slice image. The CT value of a voxel is ideally a function of the attenuation coefficient of the material enclosed within it, although unavoidable blurring causes surrounding material to also have an influence.

If a voxel contains void space, its measured CT value will be a weighted average of the end-member values for air and the remaining material. If a fluid replaces the air, the net attenuation will rise due to the fluid being more attenuating than air; the increase in CT value is proportional to the percentage of void space. A map of the distribution of effective porosity can thus be created by imaging a sample twice, once each with pore space empty and filled, and subtracting the two data sets from each other.

All scanning was performed at the High-Resolution X-Ray CT facility at the University of Texas at Austin (UTCT), Texas (USA), and is described in detail by Ketcham and Carlson (2001). The X-ray source is a 200-kV FeinFocus model FXE200.20, which is capable of a <10-m focal spot size. The detector system is an image intensifier from which data are captured and digitized by a charge-coupled device (CCD) camera. Subsequent to the description by Ketcham and Carlson (2001), the camera has been upgraded from one with 512 pixel x 512 pixel resolution to one with 1024 pixel x 1024 pixel resolution. Additionally, new computer hardware and software enable collection of data for multiple slices during a single rotation by utilizing data off the true horizontal plane used for standard scanning. Commonly, there is a distortion with increasing distance from the horizontal plane, which is most readily observed as a misfit between adjacent sets of slices from different rotations. However, the distortion attained with the UTCT system is negligible if the amount of the off-plane data utilized is restricted to the 25–30 video lines surrounding horizontal.

In the two-stage scanning protocol, each specimen was first dried for at least 1 day in a warm (~38C) oven, weighed, and then scanned. The specimen was then infiltrated under vacuum with distilled water, weighed after excess water was removed with an absorbent tissue, and quickly scanned while still wet. The sample-mounting container was only marginally water-tight, and thus samples were wrapped in parafilm to restrict leakage. There was occasional slight drainage due to this protocol, and analyses would be improved by fabrication of water-tight sample mounts that precisely match the sample diameters. The maximum drainage inferred from mass loss of water during scanning was <0.1% of the fluid.

All the minicores were scanned with an X-ray peak energy of 150 kV and a current ranging between 0.21 and 0.25 mA. Each rotation consisted of 1000 views or angular positions, with an acquisition time of 0.133 s per view. Slices were two CCD camera pixel rows thick (57.8 m), and 13–15 slices were acquired per rotation. A 24-mm field of reconstruction was captured in a 512 pixel x 512 pixel image. Between 250 and 425 slices were gathered, depending on sample size. Total data acquisition times were on the order of 2 hr per scan.

An important aspect of the scanning protocol was the use of a wedge calibration (Ketcham and Carlson, 2001), which consists of a set of detector readings through a full rotation of a material with identical geometry and similar attenuation properties as the material being scanned. By ensuring that the hardness of X-rays measured during calibration is similar to the beam hardness encountered during scanning, an appropriate wedge calibration is an effective solution for both beam hardening and a number of related issues, such as ring artifacts that stem from utilizing a polychromatic X-ray source. The wedge material used in this study was a core of pure synthetic quartz of identical diameter to the minicores. It is also possible to derive a software beam-hardening correction to detector readings (e.g., Clausnitzer and Hopmans, 2000), although these may be less effective at eliminating ring artifacts.

Larger half-round samples corresponding to the minicores were also scanned, although their non-cylindrical geometry prevented the employment of a proper wedge calibration to reduce beam-hardening artifacts sufficiently for the two-stage protocol to be used. Accordingly, these images were primarily intended for general visualization and comparison with the minicore images and visualizations. These visualizations document the loss in resolution when scanning larger samples and increasing the imaged volume. Scanning conditions were similar to those employed for the minicores, except the slice thickness was 127 m and the field of reconstruction was 59 mm.

Three-dimensional (3-D) visualizations were created from the stacks of CT images using the software VG Studio Max (Volume Graphics GmbH, Germany). The software implements a visualization algorithm known as volume rendering, in which each voxel is assigned a color and opacity, allowing parts of the volume to be rendered fully or partially transparent. Three renderings were made for each sample scanned: one was made of the entire sample, one was made of the void space, and one was made of the high-density "bright" phases in the CT images. The images of the entire samples show the CT gray-level data rendered in 3-D. The void images highlight the larger pores, where CT gray levels are reliably below the levels for solid material. Large, open pores are opaque, whereas smaller pores that do not achieve as dark a gray level due to blurring with matrix are semi-transparent. For the minicores, the scans of the non-infiltrated samples were used, and microporosity detected by the two-stage scanning process was not included in these images so that they would be consistent with the visualizations of the half-rounds, for which such data were not available. In general, volume renderings of the partial-porosity data revealed by two-pass scanning were opaque and uninformative. The high-density phase visualizations often reveal at least two phases, with the higher-attenuation phase in yellow and the lower-attenuation phase in orange; however, in some cases the yellow phases have orange borders due to blurring or diffuse boundaries.

The color and opacity tables used to generate all of the renderings (Figs. F1, F2) are similar within each category but were manually adjusted using an interactive tool to improve the clarity of the visualization for each sample. The most significant adjustments made were to the opacity table where, if too much material is opaque, the visualization appears as a solid block, which is uninformative if the general purpose is to discern internal features. In all cases these images are meant to be illustrative, giving a sense of the distribution and character of the various components of interest. They are not quantitative, particularly in the case of void space, owing to the size range of the components of interest. More quantitative images of porosity are provided in a separate manuscript (Ketcham and Iturrino, in press).