Stress induced permeability anisotropy of Resedimented Boston Blue Clay


  • Amy L. Adams,

    Corresponding author
    1. Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
    • Corresponding author: A. L. Adams, Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Rm 1–353, 77 Massachusetts Ave., Cambridge, MA 02139, USA. (

    Search for more papers by this author
  • John T. Germaine,

    1. Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
    Search for more papers by this author
  • Peter B. Flemings,

    1. Jackson School of Geosciences, University of Texas at Austin, Austin, Texas, USA
    Search for more papers by this author
  • Ruarri J. Day-Stirrat

    1. Bureau of Economic Geology, University of Texas at Austin, Austin, Texas, USA
    2. Now at Shell International Exploration and Production, Projects and Technology, Houston, Texas, USA
    Search for more papers by this author


[1] In Resedimented Boston Blue Clay (RBBC), a low-plasticity glacio-marine illitic mudrock, the ratio of the horizontal to vertical permeability (the permeability anisotropy, rk) increases from 1.2 to 1.9 as the porosity decreases from 0.5 to 0.37 and the permeability decreases by more than 1 order of magnitude. Backscattered Scanning Electron Microscope (BSEM) images taken at formation stress levels reveal that particles rotate perpendicular to the axial loading direction by ∼22°, with larger particles rotating more significantly and achieving more uniform alignment than smaller particles. We show experimentally that preferred platy particle orientation can explain our permeability anisotropy measurements. The permeability anisotropy of mechanically compressed mudrocks is minimal, <2.5. We use a novel approach (cubic specimens) to measure the evolution of permeability anisotropy in different directions on the same specimen, unlike most other methods. Modified analytic techniques allow calculation of the permeability anisotropy for a specimen using directional constant head permeability methods. A better understanding of the evolution of permeability anisotropy during sediment burial is important for modeling subsurface transport processes, including hydrocarbon migration and contaminant transport, as well as estimating in situ conditions such as pore pressure, overpressure, and effective stress.

1. Introduction

[2] The evolution of mudrock permeability during burial has been studied for decades. Porosity can decrease from 0.9 to as little as 0.05 over a few kilometers depth with a corresponding decrease in permeability of up to 8 orders of magnitude [Neuzil, 1994]. The log of permeability generally decreases linearly with porosity, and at a given porosity mudrock permeability can vary by up to 3 orders of magnitude [Neuzil, 1994]. The permeability anisotropy of uniform (nonlayered) mudrocks, defined as the ratio of the horizontal to vertical permeability, typically increases with compression [Basak, 1972; Daigle and Dugan, 2011; Dewhurst et al., 1998; Yang and Aplin, 2007]. Changing mudrock porosity, and hence permeability, in a sedimentary basin directly affects fluid migration, consolidation rates, and overpressure generation [see Broichhausen et al., 2005; Bethke, 1989]. Numerical models are often used to simulate the complex behavior of a sedimentary basin. Some such models assume the permeability of a given stratum to be constant in all directions [e.g., Ungerer et al., 1990], while others include anisotropy at a constant value independent of mudrock porosity [e.g., Bekele et al., 2001]. While there is a mature approach to describing how bulk permeability varies during compression, the evolution of permeability anisotropy is not as well understood.

[3] There has been limited work to describe the evolution of permeability anisotropy of uniform, nonlayered mudrocks with compression. Clennell et al. [1999] showed that anisotropy increases from 1.1 to 3 for remolded pure clays, synthetic silty clays, and natural clays. Leroueil et al. [1990] studied marine clays and found the permeability anisotropy to range from 1.1 to 2.5, with some mudrocks exhibiting decreasing anisotropy with compression. Basak [1972] determined that the permeability anisotropy of kaolinites is dependent on soil structure and varies between 1 and 1.6. Finally, Yang and Aplin [1998, 2007] report permeability anisotropy measurements on intact mudrock cores varying between 1.15 and 11.8, but note that high anisotropy (>4) implies specimen heterogeneities such as layering. All of these studies suggest that in uniform mudrocks, the permeability anisotropy varies modestly between 1 and at most 3 or 4 in the first few kilometers below seafloor.

[4] It is generally inferred that the development of permeability anisotropy is due to the rotation of platy particles during uniaxial compression. Numerous models have been developed to describe the permeability anisotropy as a function of particle alignment. Daigle and Dugan [2011] derive one such model using geometry to compute the flow path tortuosity based on mean particle orientation, particle aspect ratio, and porosity. Their model predicts that the permeability anisotropy, given as the square of the tortuosity anisotropy, will not exceed ∼10 for 80% strain in a typical Illite–Kaolinite mudrock under uniaxial compression [Daigle and Dugan, 2011]. Both Daigle and Dugan [2011] and Arch and Maltman [1990] show that higher-permeability anisotropy (>10) in homogenous mudrocks is possible only when uniaxial compression is combined with simple shear. Similarly, Yang and Aplin [1998] derived a semiempirical model to capture the evolution of the pore geometry with compression. Their model predicts the permeability (horizontal or vertical) to within a factor of ±3; however, as result is unable to accurately predict the permeability anisotropy. Ultimately, theoretical advances in our understanding of permeability anisotropy are constrained by the limited availability of field and laboratory measurements.

[5] A key control on understanding permeability evolution is developing a quantitative understanding of how particle alignment develops during compression. Martin and Ladd [1975] reveal that 80% of the particle orientation at 10 MPa effective stress in slurry resedimented Kaolinites occurs below 0.1 MPa effective stress. A common approach is to model particle rotation with the March [1932] model which proposes that particle rotation begins at very small strains and continues asymptotically to very large strains. Yet, there is disagreement as to whether particle alignment occurs with increasing compression. Clennell et al. [1999] did not find any relationship between particle alignment and permeability anisotropy development for four different mechanically compressed remolded mudrocks up to 4 MPa. They identify clustering of clay particles around larger silt grains as well as nonuniformities in the micro fabric and suggest these as limiting factors for permeability anisotropy development.

[6] We experimentally investigate the evolution of permeability anisotropy with increasing stress and decreasing porosity for mechanically compressed mudrocks deformed in a uniaxial strain field. We first present a method to measure the permeability of Resedimented Boston Blue Clay (RBBC) using cubic specimens which allows for measurement of both the vertical and horizontal permeability of the same specimen. The permeability anisotropy is measured separately for each specimen. With this approach, we measure multiple specimens at different porosities to define the porosity versus log permeability trend for RBBC. Our new approach removes specimen variability from the permeability anisotropy measurement and contrasts other methods which require two specimens to compute the permeability anisotropy [e.g., Clennell et al., 1999; Leroueil et al., 1990]. We discuss the analytic methods used to determine the permeability anisotropy of a single specimen from directional constant head measurements. Image analysis of Backscattered Scanning Electron Microscope (BSEM) images taken at varying stress levels reveal that platy particles are initially randomly oriented at low stress, and become more oriented to the horizontal with increasing applied stress. Finally, we apply the Daigle and Dugan [2011] model to predict the permeability anisotropy of our specimens and find good agreement with our experimental measurements. We show experimentally that platy particle alignment can explain permeability anisotropy development in mudrocks, and that the March [1932] model does not capture the platy particle orientation we measure.

2. Materials

[7] RBBC is a mudrock that has been extensively studied at Massachusetts Institute of Technology (MIT) [Sheahan, 1991; Cauble, 1996; Santagata, 1994; Abdulhadi, 2009]. The Boston Blue Clay was deposited approximately 12,000 years ago in a glacio-marine environment in the region of Boston, MA [Barosh et al., 1989]. In this study, the fine fraction of BBC, defined as the fraction passing a standard US 10 sieve (<2 mm), is ground until it passes a 100 sieve (<0.152 mm), forming a powder. We currently use Series IV BBC powder, classified as a low-plasticity clay (CL) using the Unified Soil Classification System (USCS). The powdered mudrock contains 53% clay sized particles (<2 μm) and has a plastic limit, liquid limit, and plasticity index of 23%, 46%, and 23%, respectively. The specific gravity is 2.78. The maximum particle size as determined via sedimentation analysis [ASTM Standard D422, 2007] is less than 0.07 mm. RBBC is dominated by Illite, Illite-Smectite, and Trioctahedral-Mica, with lesser amounts of Chlorite, Hydro-Biotite, and Kaolinite [Schneider et al., 2011].

3. Specimen Preparation

[8] Resedimentation is the process of mixing a sediment slurry and then incrementally, and uniaxially, loading it to a desired axial effective stress. This procedure is undertaken to produce uniform, saturated soil specimens [Germaine, 1982; Abdulhadi, 2009; Schneider et al., 2011]. Ground RBBC powder is mixed at 100% water content with sea salt at 16 g/L, and the resultant slurry is deaired and poured into 7.6 cm internal diameter, 45 cm tall acrylic sedimentation columns. Specimens are incrementally loaded with two-way drainage in salt water with 16 g/L salinity. This method produces specimens with a pore fluid salinity of approximately 16 g/L. Specimens measured in this research were compressed to axial effective stresses ranging from 0.4 to 10 MPa.

[9] Once maximum effective stress is achieved, the specimen is unloaded to one quarter of the maximum vertical effective stress, giving it an Overconsolidation Ratio (OCR) of 4. An OCR of 4 produces a lateral stress ratio, equal to the ratio of the horizontal to the vertical stress, of 1 for BBC [Ladd, 1965]. This condition allows for stress removal, extrusion, and trimming with minimal disturbance and shear. The applied load is quickly removed, and the specimen is extruded from the sedimentation column using a manual hydraulic jack. A cube with 5 cm sides is trimmed with one axis aligned parallel to the axial loading direction. Faces are rough cut using a coarse saw and leveled and squared using a razor blade. Three flow directions are defined as Vertical (V), Horizontal 1 (H1), and Horizontal 2 (H2).

4. Permeability Anisotropy

4.1. Experimental Method

[10] The permeability of resedimented cubic specimens is measured using the constant head method in a flexible wall permeameter [ASTM Standard D5084, 2010] fitted with square-end adapters (Figure 1). A similar method was employed by Chan and Kenney [1973]. Salt water (16 g/L sea salt, equal to the in situ salinity for BBC) is flowed through the specimen under a constant head gradient in a temperature controlled enclosure stable to 26 ± 0.1°C. The directional permeability is measured by rotating the cubic specimen through three sequential setups measuring each of the V, H1, and H2 directions. Each sequential measurement requires disassembly and reassembly of the apparatus. Two-directional orders are applied: V, H1, H2 (Vertical order) and H1, H2, V (Horizontal order). Each measurement sequence is composed of the following steps: First, a cell pressure equal to the OCR 4 stress state is applied with the drainage lines closed. This allows the specimen to temperature equilibrate and come to a stable sampling effective stress. Second, the pore pressure is incrementally increased to between 0.35 and 0.4 MPa to back pressure the specimen while maintaining the sampling effective stress. Third, the specimen is recompressed to the hydrostatic effective stress corresponding to the OCR 4 stress state for permeability measurement. Fourth, a constant head gradient is applied by maintaining a differential pore pressure across the specimen, during which both the inflow and outflow volumes are measured over time. A permeability measurement is made once the inflow and the outflow rates are steady. A minimum of three constant head gradients are applied by increasing the inflow pressure and decreasing the outflow pressure, keeping the mean effective stress in the specimen constant. We applied constant head gradients ranging from 50 to 170 with lower gradients applied to higher-porosity (lower stress) specimens. Constant head gradients were applied in nonsequential order (neither increasing nor decreasing) and we do not identify any trend in measured permeability with hydraulic gradient. The measured permeability varied minimally (third significant digit) as the gradient was varied.

Figure 1.

Flexible Wall Permeameter apparatus fitted with square end adapters used for constant head permeability testing of cubic specimens [ASTM Standard D5084, 2010].

[11] Constant head permeability tests are known to cause volume change in mudrocks because high hydraulic gradients result in uneven pore pressure distributions across the specimen causing swelling at one end and compression at the other end of the specimen. However, we measure negligible (<0.1%) volume change during the application of hydraulic gradients for two reasons. First, our specimens are over consolidated (OCR 4), yielding a much stiffer volume change response than a normally consolidated mudrock. Second, the change in pore pressure at either end of the specimen is small (<10%) compared to the mean effective stress.

[12] The dimensions of each of the three axes of the cubic specimen are measured in four locations to the nearest 0.01 mm both prior to, and after, each directional permeability measurement. These dimensions are averaged for each axis and are used to compute the flow length and flow area for the permeability calculation according to D'Arcy's law. We measure negligible change in the dimensions (<5%, average 1.2%) before and after each permeability measurement. Most of this change is attributable to small (<1 mm) measurement errors stemming from slight imperfections in the shape of the cubic specimen itself.

[13] The wet mass and dimensions of the specimen are recorded at the beginning and end of each directional permeability measurement. The dry specimen mass is obtained following completion of all three permeability measurements. The specimen is dried in an oven set to 110°C for a minimum of 24 h. The porosity is computed using a mass-based approach assuming complete saturation and correcting for the presence of salt in the pore fluid:

display math(1)

where n is the porosity, Vv, Vw, Vsalt are the volumes of the voids, the water, and the salt, respectively, VT is the total specimen volume, Mw, Msalt, Ms are the masses of the water, the salt, and the solid grains, respectively, Gs is the specific gravity of the grains, ρw is the density of distilled water, and ρs is the density of pore fluid at the measured salt concentration. With this approach the specimen dimensions and specimen volume are not used to compute porosity, reducing error associated with imperfections in the shape of the cubic specimen.

[14] All mass measurements are taken after the specimen is removed from the permeameter. During disassembly of the apparatus, the cell pressure is released, inducing negative pore pressures in the specimen. To counteract this, a vacuum is applied to the specimen to prevent suction of free water found in the pore pressure lines. This procedure limits specimen volume change between permeability measurement and massing. The mass is recorded immediately upon removal from the device to prevent drying. Using this approach, errors in porosity measurement related to swelling or drying of the specimen are not likely.

[15] Negligible porosity change (≤0.01) is noted through mass-based porosity measurements (equation (1)) throughout the up to month-long duration of each specimen permeability anisotropy measurement. An average porosity is computed for each specimen using the three mass-based porosities corresponding to the end of each directional permeability measurement.

[16] The permeability anisotropy is computed directly as:

display math(2)

where rk is the permeability anisotropy, and kH and kV are the permeabilities in the horizontal and vertical directions, respectively. Equation (2) averages the two horizontal permeability measurements.

4.2. Permeability Results

[17] Permeability anisotropy was measured for 14 cubic RBBC specimens compressed to effective stresses ranging from 0.4 to 10 MPa. Each specimen was resedimented and measured separately. Figure 2 gives specimen porosity and permeability results.

Figure 2.

Measured compression and permeability behavior of Resedimented Boston Blue Clay (RBBC). (a) Comparison of the compression behavior of 14 individual, resedimented, cubic specimens of RBBC (solid circles) to the expected compression behavior measured using the Constant Rate of Strain (CRS) device (solid line). Resedimented specimens are rebounded to an Overconsolidation Ratio (OCR) of 4 prior to extrusion, trimming, and measurement. The OCR 4 line (dashed line) is computed by dividing the stress axis of the CRS virgin compression line (solid line) by 4. At high stresses, the cubic specimen porosity deviates from the expected OCR 4 line. Figure 2b compares the measured vertical (solid circles) and horizontal (open circles) permeability of the individual cubic specimens to the CRS vertical permeability (solid line). Permeability is a function of porosity, and not OCR, allowing direct comparison between the two data sets. Good agreement is obtained for the vertical permeability, with slight deviation for porosities below 0.43.

[18] The permeability and compressibility of RBBC was also measured separately using the Constant Rate of Strain (CRS) device [Wissa et al., 1971; ASTM Standard D4186, 2006] to provide a benchmark for comparison with the cubic specimen results. The CRS compression curve follows the virgin compression line (OCR = 1). We divide the stress axis of the virgin compression line (solid line, Figure 2a) by four to obtain the OCR 4 line (dashed line, Figure 2a). At stresses greater than 0.8 MPa, the cubic specimen results record a slightly greater stress at a given porosity than predicted by the OCR 4 line (solid circles versus dashed line, Figure 2a).

[19] We compare the cubic specimen permeability with the permeability of RBBC measured using the CRS device (Figure 2b). Unlike the compression curve, permeability is as function of porosity, and not OCR, allowing a direct comparison between the two measurement techniques. At porosities greater than ∼0.43 the porosity—vertical permeability behavior measured with our flexible wall permeameter closely matches the CRS results, while at lower porosities the CRS permeability is less than the flexible wall permeameter permeability for a given porosity. The slope of the vertical permeability versus porosity trend is shallower than the slope of the horizontal permeability versus porosity trend (closed versus open circles, Figure 2b). This indicates that the permeability anisotropy is increasing with decreasing porosity.

[20] The permeability anisotropy (equation (2)) increases from ∼1.2 to ∼1.9 as the porosity decreases from 0.5 to 0.37 (see dashed line, Figure 3). There is offset between specimens measured using Vertical Order (open circles) and Horizontal Order (open squares). Open symbols indicate measured results, and solid symbols differentiate results corrected for measurement sequence bias (section 4.1). Good repeatability was achieved for the two sets of duplicate measurements made at a porosity of 0.37, corresponding to a maximum applied stress of 10 MPa. At this porosity, the permeability anisotropy varies from 1.77 to 1.84 depending on the directional order of measurement. The H2 permeability is consistently lower than that in the H1 direction (Figures 4a and 4b).

Figure 3.

Permeability anisotropy of Resedimented Boston Blue Clay (RBBC). Individual specimen measurements are shown (open symbols) with differentiation between specimens measured using Vertical (circle) and Horizontal (square) Orders. Adjusted measurements (solid symbols) are corrected for measurement sequence bias.

Figure 4.

Illustration of measurement sequence bias. (a) The second (H2) horizontal permeability measurement is consistently measured to be less than the first (H1), shown as a deviation from the 1:1 line. The slope of the H2:H1 permeability line is the horizontal measurement sequence bias, xh, and can be used to correct the permeability data set. (b) An individual Vertical Order measurement sequence showing a predictable decrease in permeability as subsequent permeability measurements are performed on the same specimen. Extrapolation of the slope connecting two like directional measurements yields the expected permeability at measurement one, allowing for correction of the second and subsequent permeability measurements (similar to Chan and Kenney [1973]).

4.2.1. Porosity Deviation

[21] The cubic specimen porosity measurements show deviation from the expected OCR 4 line with increasing stress (solid circles versus dashed line, Figure 2a). Sidewall friction is a likely cause of the measured deviation. During resedimentation the specimen experiences sidewall friction because of the large contact area with the cylindrical resedimentation tube. Sidewall friction reduces the applied stress felt by the specimen. A specimen that is compressed to 10 MPa and then rebounded to 2.5 MPa during resedimentation has an OCR of 4. However, if sidewall friction supports 5% of the applied load, then the maximum applied stress becomes 9.5 MPa. The rebounded stress will increase, becoming 2.6 MPa. Overall, the specimen will be at OCR 3.65 (9.5 divided by 2.6) instead of OCR 4 as loaded. Sidewall friction results in a lower than expected OCR and can explain the noted deviation from the OCR 4 line (Figure 2a).

4.2.2. Measurement Sequence Bias

[22] Our measurements indicate a small but consistent difference between successive horizontal permeability measurements (H1 and H2) that is illustrated in a cross plot as a slope of less than 1 (Figure 4a). Repeated permeability measurements in the same direction are lower, regardless of the order of measurement (Figure 4). This is likely resultant from slight smearing of particles on the cubic faces during handling the specimen between permeability measurements. Chan and Kenney [1973] noted a similar smearing effect across layer interfaces in varved clays. Due to the low permeability of our specimens, the permeant invades the specimen by <2 mm depth during each directional permeability measurement; any slight smearing or alteration of the surface fabric structure will affect the permeability measurement. The decrease in permeability due to smearing is consistent and predictable (Figure 4a).

[23] We correct for the decrease in permeability in a given direction using the measurement sequence bias, x, defined as the slope of a cross plot of successive permeability measurements in the same direction (Figure 4a). Measurements from multiple specimens of the same mudrock are required to define this slope for each direction of interest. In Figure 4a, dividing the H2 permeability measurement by the slope x forces a 1:1 slope with the H1 measurement. For RBBC with maximum effective stresses ranging from 0.4 to 10 MPa, the horizontal measurement sequence bias, xh, was determined to be 0.975 (shown in Figure 4a).

[24] To estimate the true permeability, we apply a global correction:

display math(3)

where k1 is the adjusted permeability, kn is the measured permeability, x is the measurement sequence bias between successive measurements and n is the measurement number. The adjusted permeability, k1, is the permeability at measurement n = 1 corrected for permeability decreases associated with n permeability measurements. Equation (3) does not correct the first measurement, but corrects the second measurement (n = 2) for a permeability decrease associated with the first measurement, and corrects the third measurement (n = 3) for permeability decreases associated with the first two measurements.

[25] If only nonsuccessive measurements are available the successive test sequence bias x, required in equation (3), is derived using the number of intermediary measurements:

display math(4)

where x' is the measurement sequence bias for nonsuccessive permeability measurements, and Δn is the number of measurements separating the nonsuccessive permeability measurements. The parameter n must be constant for all measurements. We apply equation (4) to determine the vertical measurement sequence bias, xv, using measurement sequences where the vertical measurement was repeated at the end of all measurements: V1, H1, H2, V2 (Δn = 3). For RBBC over a maximum effective stress range of 0.4–10 MPa, xv was determined to be 0.967.

[26] An alternative approach, called the single specimen method, is better suited for single specimens or for small data sets. This approach plots directional permeability measurements versus the measurement number, n (Figure 4b). Sequential permeability measurements in the same direction are connected and the slope extended to determine the corrected permeability at measurement number 1 [Chan and Kenney, 1973]. We have applied both the global and single specimen approaches and found similar results.

[27] Our analysis shows that only the second directional permeability measurement requires correction. Therefore, the optimal measurement sequences are V, H1, H2 (Vertical Order) and H, V1, V2 (Horizontal Order).

[28] We correct our measured permeabilities using the global correction (equation (3)) and calculate the resultant anisotropy (equation (2)). We note that specimens measured using vertical order adjust to a higher anisotropy (see Figure 3, open circles move up to solid circles), and those measured using horizontal order adjust to a lower anisotropy (see Figure 3, open squares move down to solid squares). This is the expected result of substituting the adjusted, increased permeability value into equation (2). Though the test sequence bias correction does not significantly alter the trend of permeability anisotropy with porosity for RBBC (Figure 3), we find that it significantly reduces scatter in the directional permeability data set (given in Figure 2b).

5. Particle Orientation

5.1. Image Analysis Method

[29] We obtained back scattered (BSEM) images of three oven dried RBBC specimens at 0.1, 1, and 10 MPa (Figure 5) [Emmanuel and Day-Stirrat, 2012] and analyzed the particle orientation using JMicroVision [Roduit, 2013], a free image analysis software package. The specimen surfaces were cut perpendicular to the base of the specimen, parallel to the axial loading direction. Apparent particle angles relative to horizontal are measured in this plane. The oven dried surfaces were prepared using an Argon-ion beam milling technique [Loucks et al., 2009], which uses accelerated Argon ions to polish and smooth the surface, resulting in only minor topographic variations. The specimens were aligned in the microscope to produce an image where the base of the image corresponds to either the top or the bottom of the specimen; a vertical line drawn through the image parallels the line of axial loading. We first drew lines on all platy particles and then exported particle orientation and length data for analysis. We calibrate the software to convert pixel measurements to length using the image scale bar. Particle length is computed by JMicroVision using this calibration. We compute the average acute particle angle referenced to the base of the image. Particles of length ∼0.2–5 μm were considered. We could not see or delineate particles less than 0.2 μm, and we disregarded particles greater than 5 μm as unrepresentative given the field of view of ∼20 μm. Nonplaty particles (e.g., silts) with an indeterminable long axis (aspect ratio close to 1) were not included in the analysis. Three different images of each specimen, for a total of nine images, were analyzed to increase the statistical sample size. We determined the particle aspect ratio using the 2-D measurement feature by outlining the particles and exporting the particle aspect ratio computed by JMicroVision.

Figure 5.

BSEM images of RBBC at (a) 0.1 MPa, (b) 1.0 MPa, and (c) 10 MPa [Emmanuel and Day-Stirrat, 2012]. All images are of a vertical plane. Image analysis shows ∼22° change in mean particle orientation from 0.1 (inset A) to 10 MPa (inset C) (Table 1).

[30] We note two important sources of error in the image analysis results. First, oven drying is known to cause changes in porosity, pore size distribution and potentially particle orientation in mudrocks. Second, we measure the particle orientation and aspect ratio of 3-D particles using 2-D planes. We note that the values we measure are apparent in the plane of measurement. Despite this, we are able to compare the apparent orientations across different stress levels using this method and obtain a lower bound estimate of the particle aspect ratio.

5.2. Particle Orientation and Aspect Ratio Results

[31] We identified 5108 particles in nine images (three at each stress level) and computed the mean particle orientation at 0.1, 1, and 10 MPa (Table 1). The mean particle orientation shifts from 50° to the horizontal at 0.1 MPa to 28° at 10 MPa (Figure 6 and Table 1). These quantitative results are supported by visual analysis. At 0.1 MPa (Figure 5a), particularly the smaller platy particles have a slight vertical orientation. In contrast, by 10 MPa (Figure 5c) platy particles show a clear horizontal orientation. The mean particle size, given by the line length, does not vary significantly between specimens compressed to different stress levels (Table 1). This indicates that our orientation measurements are representative.

Table 1. Mean Particle Orientation for RBBC Via Image Analysis
Stress Level (MPa)PorosityStrainaOrientationb (°)Line Lengthb (μm)Measurements
  1. a

    Strain computed assuming a slurry starting porosity of 0.735.

  2. b

    Reported as the mean ±1 standard deviation.

0.10.570.3850 ± 240.97 ± 0.611845
10.450.5239 ± 241.11 ± 0.711211
100.350.5928 ± 210.97 ± 0.242052
Figure 6.

Comparison of measured mean RBBC particle orientation (Table 1) to March [1932] model predictions. The March [1932] model over predicts orientation at all porosities (strain levels) and under predicts rotation between 0.1 and 10 MPa.

[32] To determine the effect of size on orientation the data were divided into bins: particles 0.2–0.7 μm, 0.7–1.2 μm and >1.2 μm (Table 2). The bins were chosen so that at least 20% of the particles within any particular image sat within each bin. At high porosities there is little alignment distinction between large and small particles and all particle sizes show a wide orientation distribution revealed by a high standard deviation (Table 2). As the mudrock compresses larger particles undergo more rotation than smaller particles. At 10 MPa, the smallest particles between 0.2 and 0.7 μm have a mean orientation of 32°, while the particles >1.2 μm have rotated to a mean orientation of 24° (Table 2). Histograms of the particle orientation distribution (Figure 7) in 5° increments reveal that with increasing compression, larger particles become more uniformly oriented compared to smaller particles.

Table 2. Particle Orientation as a Function of Size for RBBC
Stress (MPa)PorosityStrainaParticle Orientation, θ (°)b
0.2–0.7 μm0.7–1.2 μm>1.2 μm
  1. a

    Strain computed assuming a slurry starting porosity of 0.735.

  2. b

    Reported as the mean ±1 standard deviation.

0.10.570.3850 ± 2450 ± 2451 ± 25
10.450.5240 ± 2438 ± 2437 ± 22
100.350.5932 ± 2227 ± 2224 ± 19
Figure 7.

Histograms of particle orientation as a function of size and stress: (a) 0.1 MPa, 0.2–0.7 μm, (b) 0.1 MPa, >1.2 μm, (c) 10 MPa, 0.1–0.7 μm, and (d) 10 MPa, >1.2 μm. As the stress increases the mean particle orientation decreases. Small particles maintain a wider particle orientation distribution with increasing stress whereas larger particles undergo significantly more rotation and become more uniformly oriented.

[33] Finally, we measured the aspect ratio of 3141 particles using six images (two at each stress level). Similar to line length, we do not note any change in aspect ratio with stress level. We compute a mean aspect ratio of 3.75. We follow a similar procedure to our particle orientation analysis and divide the measured aspect ratios into three bins: particles with aspect ratio 1–2.3, 2.3–4, and >4. The bins were chosen so that at least 20% of the particles within any particular image sat within each bin. We compute the mean aspect ratio and the area fraction of particles falling in each bin by combining data for all three stress levels (Table 3). These data are used as inputs for later analysis.

Table 3. Particle Aspect Ratio for RBBC
Aspect Ratio BinMeasurementsaMean Aspect RatiobTotal Particle Area (µm2)Particle Area Fraction
  1. a

    Combined total over six images, two each spanning three stress levels (0.1, 1 and 10 MPa).

  2. b

    Reported as the mean ±1 standard deviation.

1–2.39671.69 ± 0.3441338%
2.3–410933.11 ± 0.4929127%
>410816.23 ± 2.4637435%

6. Discussion

[34] We show that for Resedimented Boston Blue Clay (RBBC), the ratio of horizontal to vertical permeability increases monotonically from 1.2 to 1.9 as the porosity decreases from 0.5 to 0.37 over an effective stress range from 0.4 to 10 MPa (Figure 3). This compares well with permeability anisotropy ranges ∼1–3 reported by other researchers for similar materials [e.g., Clennell et al., 1999; Leroueil et al., 1990; Mondol et al., 2011].

[35] At 0.4 MPa particles are slightly horizontally aligned (∼43°, interpolated using power law relation from Figure 6) and with increasing stress they become more horizontally aligned, reaching a mean orientation ∼28° at 10 MPa (Figure 6 and Table 1). These results are supported by visual inspection of our images (Figure 5). We interpret that the rotation of platy particles drives the development of permeability anisotropy in homogeneous mudrocks.

[36] A common but simplistic conceptual view of mudrock evolution is that they are composed of randomly oriented platy particles at the seafloor and that with increasing burial, particle rotation occurs and a horizontally aligned fabric develops [e.g., O'Brien and Slatt, 1990]. The March [1932] model (equation (5)) [Daigle and Dugan, 2011] is commonly applied to simulate this behavior:

display math(5)

[37] Where εv is the volumetric (uniaxial) strain and θ is the mean particle orientation relative to the horizontal. We assume the particles are initially randomly oriented (i.e., the mean orientation is 45°) when we mix the slurry for resedimentation (porosity 0.73, strain 0) and that there is uniaxial strain.

[38] As the stress increases to 27 MPa, suggested as the transition point between mechanical and chemical processes in a generic basin (discussed later), the March Model predicts the particles will rotate from 45 to 22° to the horizontal (Figure 6, dashed line) as the porosity decreases from 0.73 to 0.33 and the strain increases from 0 to 0.60 for RBBC. However, our experimental results are significantly different from the March [1932] model prediction (Figure 6, compare closed circles to dashed line). At a porosity of 0.57 (strain 0.38) the specimen maintains a random fabric (Figures 6, 7a, and 7b). Thereafter, particles rotate to 28° to the horizontal as the strain increases to 0.59 and the porosity decreases to 0.35. Larger particles undergo more rotation, and become more uniformly oriented than smaller particles which maintain a more random distribution with strain (Figure 7 and Table 2). The difference between the modeled and measured particle orientation (Figure 6) suggests that the mechanical behavior during compression is more complex than described by the March model.

[39] Daigle and Dugan [2011] modeled the permeability anisotropy (kH/kV) of assemblages of uniform, circular particles as a function of particle angle (θ), particle aspect ratio (m), and porosity (n) (equation (6)). Particle aspect ratio is defined as the ratio of the particle diameter to particle thickness. They compute the flow path tortuosity in the horizontal and vertical directions, and relate the tortuosity to the permeability anisotropy [Witt and Brauns, 1983; Scholes et al., 2007]:

display math(6)

[40] For mudrocks comprised of particles with varying aspect ratio, Daigle and Dugan [2011] suggest that the equivalent aspect ratio (meq) (equation (7)) dominates anisotropy development:

display math(7)

[41] Where fi is the fraction of particles with aspect ratio mi. We compute an equivalent aspect ratio of 2.41 for RBBC using data in Table 3. This aspect ratio is interpreted as a lower bound value due to limitations of the 2-D image analysis.

[42] We model the permeability anisotropy using the mean particle angle measured at different porosities (Figure 6 and Table 1), interpolated using a power law trend (see Figure 6), and our computed equivalent aspect ratio of 2.41. We compare the model prediction using our measured particle orientation (solid line Figure 8) with our observed permeability anisotropy (symbols, Figure 8) and find that the model provides a lower bound prediction. We find that we are exactly able to replicate our permeability anisotropy measurements by inputting an aspect ratio of 6 (Figure 8), slightly higher than both our equivalent and mean aspect ratios (meq = 2.41, mavg = 3.75).

Figure 8.

Comparison of measured RBBC permeability anisotropy (solid circles) with Daigle and Dugan [2011] model prediction computed using measured particle orientation and (1) equivalent aspect ratio meq = 2.41 (solid black line) or (2) estimated aspect ratio, m = 6 (dashed black line). Model inputs are: (1) A power law interpretation of the mean measured particle orientation as a function of porosity (see solid circles, Figure 6) and (2) particle aspect ratio.

[43] The difference between our measured anisotropy results and the model prediction computed using our measured equivalent aspect may be related to a combination of the model assumptions and our measurement methods. The model assumes that permeability anisotropy development is related to the equivalent aspect ratio of an assemblage of circular disk-shaped particles. However, it is unknown whether equation (7) holds for assemblages of irregular-shaped particles with size and aspect ratio spanning multiple orders of magnitude. Further, as noted, we measure the aspect ratio of angled 3-D particles in a fixed 2-D plane in a limited field of view. We rely on sample size to reduce bias, resulting in a lower bound estimate for both the mean and equivalent aspect ratio. Given these limitations, the agreement between our measured data and the model prediction suggests that particle rotation drives permeability anisotropy development in fine grained homogeneous mudrocks comprised of platy particles (aspect ratio > 1).

[44] We note that both the model prediction and our experimental results suggest that the permeability anisotropy of RBBC is <2 for compressive stresses up to 10 MPa and porosity greater than 0.37. Despite observed platy particle rotation, RBBC does not develop significant permeability anisotropy.

[45] In sedimentary basins, effective stress ranges from 0.1 MPa at the near surface, to 5 MPa at ∼600 mbsf in overpressured offshore basins such as the Ursa basin [Long et al., 2011; Day-Stirrat et al., 2012], and can reach ∼27 MPa (2.4 km) before processes other than mechanical compression begin to dominate pore evolution [Day-Stirrat et al., 2008]. Naturally, local geothermal gradients will dictate the depth at which chemical compaction takes over from purely mechanical processes. However, extrapolation of our experimental data to 27 MPa, reflective of the position in a generic basin where mechanical process are succeeded by chemical processes, yields a mean particle orientation of 26° and permeability anisotropy ranging between 1.6 and 2.3 (dependent on assumed aspect ratio, see Figure 8) using the Daigle and Dugan model. This corresponds to porosity 0.33 and strain 0.61 for RBBC. Intense platy particle alignment is not likely to result from mechanical compression alone; the permeability anisotropy of homogeneous mechanically compressed mudrocks with platy particles similar to that of RBBC is limited to <2.5.

[46] Mudrock particle orientation is often measured using X-ray texture goniometry preferred orientation analysis instead of image analysis as we have applied here. The preferred orientation method results in values of the “multiples of a random distribution” (m.r.d) for specific minerals within the mudrock matrix that are assumed representative all particles in the mudrock.

[47] Day-Stirrat et al. [2011] measure the Mica preferred orientation using X-ray texture goniometry for similarly prepared RBBC and report an increase in orientation from 4.6 to 4.9 m.r.d over an applied stress range of 0.1–10 MPa. We convert these preferred orientation measurements to particle orientation in degrees by relating maximum pole density (ρmax) in m.r.d. to strain (εv) using equation (8) [Kanitpanyacharoen et al., 2011] and then computing orientation (θ) as a function of strain using March's [1932] theory (equation (5)):

display math(8)

[48] An m.r.d. (multiple of a random distribution, after Wenk [1985]) of 1 is equal to 45° orientation. For RBBC, preferred orientation results indicate that the mean particle orientation decreases from 24.9 to 24.3°, only ∼1° change, as the applied stress increases from 0.1 to 10 MPa. There is significant disagreement between particle orientations obtained via image analysis (∼22° particle rotation) and those computed using XRD preferred orientation (∼1° particle rotation).

[49] The micas present in RBBC are uncharacteristically large compared to the rest of the RBBC fabric. Our image analysis results indicate that large particles experience more rotation at lower effective stresses and become more aligned to the horizontal than smaller particles (Table 2 and Figure 7). The orientation of the large mica particles may not be representative of the mean particle orientation. We were unable to apply our image analysis technique to measure the mica orientation because the field of view of the available images is on the order of the size of a single large mica particle.

[50] Finally, we have shown that the analysis of resedimented mudrocks provides an approach to systematically explore the evolution of permeability and fabric during burial. Ultimately this approach has the potential to illuminate many other material properties. It allows us to control the specimen composition, the pore fluid salinity, and the stress history to produce repeatable mudrock specimens over a range of geologically realistic materials. We used cubic specimens which allowed us to make measurements of both the vertical and horizontal permeability on the same specimen. This approach is different from other more commonly applied approaches [e.g., Clennell et al., 1999; Leroueil et al., 1990] where specimens are incrementally consolidated with intermediary permeability measurements, but the vertical and radial permeability are measured using different specimens in differently configured oedometric apparatuses. The ability to make all measurements on the same specimen in the same apparatus reduces the effect of specimen variability.

7. Conclusions

[51] We have shown that for Resedimented Boston Blue Clay (RBBC), the ratio of the horizontal to the vertical permeability increases from 1.2 to 1.9 as the porosity decreases from 0.5 to 0.37 over an effective stress range of 0.4–10 MPa. Both the vertical and horizontal permeability decrease by more than 1 order of magnitude over this porosity range. BSEM images reveal that as mudrocks compress, platy particles rotate and align to the horizontal in a uniaxially compressed mudrock, producing flow path tortuosity and permeability anisotropy. Larger platy particles experience more rotation and become more uniformly aligned than smaller platy particles. The March [1932] model overestimates the mean platy particle orientation at any strain level but underestimates the trend in orientation with porosity between 0.1 and 10 MPa. We apply the Daigle and Dugan [2011] model to predict the permeability anisotropy using measured particle orientation and aspect ratio. We show experimentally that the increase in mean particle alignment with compression can explain permeability anisotropy development. The permeability anisotropy of mechanically compressed mudrocks with platy particles is typically in the range of ∼1–2.5.

[52] We used cubic specimens as an efficient means to measure both the horizontal and vertical permeability of the same specimen, eliminating potential specimen variability. Remeasurement of the same specimen causes subsequent decreases in the permeability measurement. As a result, measurement of, and correction for, measurement sequence bias is necessary. Permeability measurements can be easily corrected by repeating the second directional measurement and applying a global measurement sequence bias correction factor or extrapolating the decreasing permeability trend. Resedimented cubic specimens provide an efficient means to identify and understand the factors controlling permeability anisotropy development in mudrocks.


[53] This project was funded by UT GeoFluids consortium at University of Texas at Austin, (supported by 11 energy companies). This paper is University of Texas Institute for Geophysics contribution 2454 Publication authorized by the Director, Bureau of Economic Geology, University of Texas. We would like to thank Stephen Rudolph for his assistance in designing and building the required equipment, and for drawing Figure 1. Thanks also to Aiden Horan for providing CRS results, and to Mun Ngah Cheong and Keiron Durant for assisting with image analysis.