Int. J. Oil, Gas and Coal Technology, Vol. 5, Nos.

2/3, 2012 229

Copyright 2012 Inderscience Enterprises Ltd.











Three-dimensional pore networks and transport
properties of a shale gas formation determined from
focused ion beam serial imaging
Thomas A. Dewers*
Geomechanics Department,
Sandia National Laboratories,
Albuquerque, New Mexico, 87185, USA
E-mail: [email protected]
*Corresponding author
J ason Heath
Geophysics and Atmospheric Sciences Department,
Sandia National Laboratories,
Albuquerque, New Mexico, 87185, USA
E-mail: [email protected]
Russ Ewy and Luca Duranti
Chevron Energy Technology Company,
San Ramon, California, 94583, USA
E-mail: [email protected]
E-mail: [email protected]
Abstract: Three-dimensional pore network reconstructions of mudstone
properties are made using dual focused ion beam-scanning electron microscopy
(FIB-SEM). Samples of Jurassic Haynesville Formation mudstone are
examined with FIB-SEM and image analysis to determine pore properties,
topology, and tortuosity. Resolvable pore morphologies (>~10 nm) include
large slit-like pores between clay aggregates and smaller pores in strain
shadows surrounding larger clastic grains. Mercury injection capillary pressure
(MICP) data suggest a dominant 110 nm or less size of pores barely
resolvable by FIB-SEM imaging. Computational fluid dynamics modelling is
used to calculate single phase permeability of the larger pore networks on the
order of a few nanodarcys (which compare favourably with core-scale
permeability tests). This suggests a pore hierarchy wherein permeability may
be limited by connected networks of inter-aggregate pores larger than about
20 nm, while MICP results reflect smaller connected networks of pores residing
in the clay matrix. [Received: May 12, 2011; Accepted: September 14, 2011]
Keywords: shale gas; Haynesville formation; pore networks; focused ion
beam; FIB.
Reference to this paper should be made as follows: Dewers, T.A., Heath, J.,
Ewy, R. and Duranti, L. (2012) Three-dimensional pore networks and
transport properties of a shale gas formation determined from focused ion beam
serial imaging, Int. J. Oil, Gas and Coal Technology, Vol. 5, Nos. 2/3,
pp.229248.









230 T.A. Dewers et al.












Biographical notes: Thomas A. Dewers is a principal member of the technical
staff in the Geomechanics Department at Sandia National Laboratories in
Albuquerque, New Mexico. He has over 21 years experience in industry,
academia, and government relating to work in the geosciences, and holds a PhD
in Geology. His current research interests include nano-to-reservoir scale
inquiries into rock mechanics, multiphase flow and reactive transport.
Jason Heath is a senior member of the technical staff of the Geophysics and
Atmospheric Sciences Department, Sandia National Laboratories,
Albuquerque, New Mexico. He has a PhD and MS in Hydrology and Geology,
respectively. His research focuses on efficiently estimating well injectivity for
large-scale CO
2
storage projects, CO
2
-water-rock interactions, and caprock
sealing behaviour.
Russ Ewy is a Research Consultant with Chevron Energy Technology Co.,
where he performs research and technical service in the areas of reservoir and
wellbore geomechanics, and leads multidisciplinary research on shales. He
holds a BS and MEng in Mineral Engineering and a PhD in Rock Mechanics,
all from the U. of California at Berkeley. He has served on the board and the
executive committee of the American Rock Mechanics Association, and
currently serves as a Review Chair for the Society of Petroleum Engineers
Journal.
Luca Duranti is a Senior Research Geophysicist at Chevron Energy Technology
Company, in San Ramon, California. He has 15 years of industry experience in
industry and academia, primarily in research but including operational aspects,
and holds a PhD in Geophysics from Colorado School of Mines. His current
research interests are in experimental and theoretical poroelasticity, spanning
across the wide spectrum of granular media, their interaction with fluids in
depositional as well as production environments, and their detection in the
active and passive seismic signal.

1 Introduction
Shale gas plays around the world have generated interest in the variety and structure of
pore networks in mudstones. Typically, three dimensional imaging of these tiny pore
systems have been inaccessible via microCT and/or optical methods due to resolution
issues. SEM and TEM imaging methods can resolve useful two-dimensional information
for mudstone pore sizes down to nanometre (nm) sizes (Desbois et al., 2008, 2009;
Loucks et al., 2009). In conjunction with successive FIB milling of nanometre-scale
surfaces (Yao, 2007), image stacks are obtained that can be used to reconstruct three
dimensional digital pore and mineral structures (Heath et al., 2011). These are akin to
microCT and laser scanning confocal image sets for larger pore-sized sandstones.
Together, dual FIB-SEM methods can be a powerful tool in understanding pore types
(Holzer et al., 2004; Loucks et al., 2009; De Winter et al., 2009), pore networks (Heath
et al., 2011) and matrix flow properties (Tomutsa et al., 2007) in shale gas mudstones and
other rock types.
In this paper, samples of mudstone from the Upper Jurassic Haynesville Formation
(Tolson et al., 1983), a prominent new shale gas play underlying much of the US Gulf









Three-dimensional pore networks and transport properties 231












Coast, are examined using FIB-SEM methods. The Haynesville Formation represents an
Upper Jurassic post-rift transgressive deposit and encompasses an array of anhydrites,
limestones, mudstones and eolian, fluvial, and alluvial sandstones (Mancini et al., 1997).
It lies beneath the Smackover Formation ramp deposit carbonates and above the Cotton
Valley Group fluvial-deltaic sands (Dobson and Buffler, 1997).
Image stacks with 7.14 nm pixel size are obtained for Haynesville mudstone volumes
cubic microns in size. Pore properties including porosity, pore body volume and throat
size distributions, pore connectivity, and tortuosity are extracted from the reconstructed
three-dimensional digital pore structures using image analysis techniques and software.
FIB-SEM- derived pore throat distributions are compared to those estimated from
mercury porosimetry (MP), the latter emphasising the smaller pores in the distribution.
Single phase permeabilities are estimated from computational fluid dynamics (CFDs)
modelling to be in the range of 2 to 12 nanodarcies (nD), which are similar to values
obtained from core plug measurements (5 to 50 nD). These digital pore reconstructions
and modelling can aid in interpretation of core-scale geomechanical and single and
multiphase flow properties.
Figure 1 SEM image of 25 micron-wide trench in Haynesville formation mudstone sample, cut
using a Ga ion beam

Notes: The light grey material on top of the mudstone is a platinum coating, used to
minimise non-planar curtaining during ion beam sectioning. Imaging is done in a
pseudo-back scatter mode, so that grey levels on the flat mudstone surface vary
with mean atomic number and thus correlate with different mineral phases.









232 T.A. Dewers et al.












Figure 2 Four SEM photomicrographs of Haynesville fm. mudstone FIB slices, imaged at
7.14 nm pixel resolution


Notes: The up direction in all photos is the depositionalyounging direction. White spots
in all images are due to charging by secondary electrons and usually occur at
edges or interiors of pores.
(a) Close-up of FIB trench (note blurry platinum coating in foreground at
bottom). Mineral (quartz, q, pyrite, p, and clay matrix, cl) and organic (o)
phases are discernable by differing grey levels (due to variation in mean
atomic number).
(b) Types of pores, following the classification scheme of Heath et al. (2011) after
Desbois et al. (2009). Type I pores (a prime example is located just above the
label) are slit shaped and oriented perpendicular to depositional direction.
Type II pores (in the vicinity of the label) are crescent-shaped pores residing
in clay matrix. Type III pores (examples are located to the right of the left-
most label and to the left of the right-most label) are associated with the
compaction shadows of larger clastic (commonly quartz, q) grains. Light grey
rectangular grain is detrital mica (m).
(c) Micron-sized Type II and Type III pores in clay fabric.
(d) Open pore fabric of possible diagenetic clay (chlorite?) partially filled with
what islikely calcite (?, light grey) surrounded by tighter clay matrix.









Three-dimensional pore networks and transport properties 233












2 Methods
2.1 FIB-SEM imaging
One-inch core plugs of Hayneville Formation mudstones were provided by Chevron.
Pieces of these with known orientation were gold-palladium coated and a portion coated
with silver dag to minimise sample charging. Serial sectioning and imaging was done
using an FEI Helios 600 NanolabDualBeam
TM
instrument and SliceandView
TM
software.
A layer of platinum was deposited on top of the portion of sample to be sliced to
minimise the vertical striations or curtaining that often plague ion milling (Figure 1). Ion
milling was performed using a 30 kV Ga ion beam with current of 2.8 nA. Vertical slices
15 nm thick was obtained by successive ion milling. SEM imaging of slices 10s of
square microns in size was done in backscattered mode with a through-the-lense detector,
1 kV voltage and 1.4 nA current. This was done to best enable discernment of pores,
minerals, and organic matter, the latter two due to differences in mean atomic number
(Figure 2). Five sampling sites were selected based on visual inspection for
representativeness of the dominant mudstone microfacies. The data set presented here
was selected for further analysis based on quality of images and large size of the image
stack that was obtained. The volume of the final data set examined is a result of imaging
constraints by the FIB/SEM Helios Nanolab.
2.2 Image processing and analysis
The resulting image stacks required a number of image processing steps in order to obtain
three dimensional reconstructions. We used the ImageJ software (version 1.44p;
Rasband, 2010; Ferreira and Rasband, 2010) to align and crop the image stacks. The
ImageJ
TM
plugin TransformJ
TM
(Meijering, 2010) was used to interpolate between the
image stacks to obtain a final cubic voxel size of 7.14 nm. Thresholding or segmentation
of the image stack proved challenging due to the extensive charging of certain pore lining
phases and pore walls [Figure 2; Figure 3(a)]. To circumvent this, we used a two-step
segmentation where dark pores (observed in Figure 2 images) were first thresholded, and
then combined with a stack segmented for the bright white charging locations, which
always coincided with pores [Figure 3(b)]. The resulting series of two dimensional
images [e.g., Figure 3(c)] were assembled into a three dimensional stack and rendered as
a 3D surface using ScanIP
TM
software (Figure 4). Dimensions of the volume were
930 434 1,036 pixels, and with a pixel size of 7.14 nm, this results in a total
imaged size of 6.6 (in x) by 3.1 (in y) and 7.1 (in z) microns, or a volume of 152.2 cubic
microns.
The final segmented images were input into the 3DMA-Rock software (Lindquist,
1999; Lindquist and Venkatarangan, 1999; Lindquist et al., 2005) which was used to
calculate the medial axis (the middle axis of voxels running through pores, resulting in a
3D skeleton representation of the network of connected pores; Figure 5(a) and Figure
5(b). Medial axes also were used to determine cross-linked pathways (i.e., topology)
through the pore networks and calculation of tortuosity (Figure 6). Once the medial axis
was constructed, 3DMA-Rock algorithms were then used to define pore throats from pore
bodies, based on the wedge finding algorithm (Shin et al., 2005).










234 T.A. Dewers et al.












Figure 3 Sequence of images used in three dimensional pore reconstructions (see online version
for colours)



Notes: (a) Original image slice showing small dark pores and larger pores with charging
(bright white spots), as well lighter grey variations corresponding to mineral
phases.
(b) Combined segmented images into mineral (yellow), green (pores) and
charging regions (red). Segmenting was performed using the Otsu
thresholding algorithm (Rasband, 2010).
(c) Final data set segmented into pore (black) and solid (white), formed by
combining the two pore grey levels in (b).










Three-dimensional pore networks and transport properties 235












Figure 4 Visualisation of resolved pores for three dimensional portion of Haynesville formation
mudstone (see online version for colours)

Notes: Pores are shown as grey solid masses. Large open regions correspond to locations
of large quartz grains. Note what appears to be a large pore network rib running
from lower left to upper right in the xz (top) view. Dimensions are 930 434
1,036 pixels, and with a pixel size of 7.14 nm, this results in a total imaged size of
6.6 (in x) by 3.1 (in y) and 7.1 (in z) microns, or a volume of 152.2 cubic microns.













236 T.A. Dewers et al.












Figure 5 Medial axes and joint histogram of pore properties for Haynesville formation mudstone
sample (see online version for colours)


Notes: (a) and (b) Map and oblique views of pore network medial axes (same views as
Figure 4), obtained by connecting medial axis (MA) voxels into a 3D skeleton of
the porous framework. The pore network rib observed in Figure 4(a) is seen as a
higher density of MA voxels. Areas of little or no density of MA voxels
correspond to large clastic grains, likely quartz.
(c) Joint and marginal relative frequency histograms of pore body volume and
coordination number (Coor. no.). The maximum relative frequency of the joint
histogram is 0.12 (black). The marginal histogram for coordination number is
such that the bottom horizontal bin is for pore bodies of zero coordination number.
The second bin is for those with coordination number of 1, and so on in integer
values up to a coordination number of 10. This shows that the mean log10 (pore
volume) of ~4.5 are pores with the most frequent coordination number of 2.









Three-dimensional pore networks and transport properties 237












Figure 6 Shortest connected pathways and tortuosity of Haynesville formation sample
(see online version for colours)

Notes: (a) through (c). Shortest connected pathways through the reconstructed mudstone
volume in the x-, y-, and z-direction respectively. The x-pathways were chosen
from left to right as shown, y-pathways were chosen from top to bottom as shown,
and z-pathways were chosen from front to back as shown. (d) through (f).
Probability distributions (constructed from frequency histograms) of tortuosity
through the pore network in the x-, y-, and z-directions respectively, defined as
actual voxel x- y- or z-path length divided by the x-, y-, or z-dimension. Green
symbols show actual values for the shortest path lengths shown in (a)(c), while
blue symbols are all calculated paths. Blue and green lines show the cumulative
probability distribution for all and shortest paths respectively.
ScanIP
TM
algorthms were applied to the segmented image stack to visualise connected
pore networks by a floodfill algorithm. Two large domains of connected pores
(Figure 7) were then selected and used to generate tetrahedral finite element meshes with
ScanIP
TM
. Mesh size was selected to achieve at least si elements across the smaller pore
throats.
2.3 CFD modelling
CFDs modelling of incompressible Navier-Stokes flow were done using COMSOL
Multiphysics
TM
and the ScanIP
TM
-derived finite element meshes. No-slip boundary
conditions were applied to pore walls, and constant pressure was applied to inlet and
outlet portions of the pore network so as to apply a small pressure gradient (0.1 Pa/m)
across the z-direction, assuming a fluid (water) density and viscosity of 1,000 kg/m
3
and
0.001 Pa-s respectively. Permeability of the pore networks estimated from the applied
pressure gradient (outlet pressure minus inlet pressure divided by Euclidian distance
between them, L), inlet surface area A, fluid viscosity , and modelled flux Q using
Darcys Law expressed as









238 T.A. Dewers et al.












Ak Poutlet Pinlet
Q
L

=


(1)
and solving for k. This assumes that the pore networks contain a sufficient number of
pores to be a representative elementary volume (REV), so that the continuum concept of
Darcys Law is applicable. A detailed examination of what constitutes a REV in
mudstones is beyond the scope of what we attempt here, but FIB/SEM methods as
examined here may provide the necessary resolution to address this question.
Figure 7 Examples of connected networks of Haynesville formation pores (see online version for
colours)

Notes: (a) View of large swath of connected pores (light grey), which are not connected
to remaining pores in the sampled volume (dark brown), constructed using a
flood-fill algorithm.
(b) 2D FIB-SEM image and 3D view with cross section corresponding to this
image. Note the large quartz grain in the right of the image. This pore network
largely is adjacent to, and in the strain-shadow of, the large quartz grain.
(c) Connected network of slit pores, shown in light grey, which is disconnected
from remaining pores as determined using flood-fill algorithm.
(d) 2D FIB-SEM image and 3D view with cross section of slit pore network,
consisting of two large crack-like pores connected by a narrow throat. Note
what appears to be a faceted quartz grain in the lower right portion of the
photomicrograph.









Three-dimensional pore networks and transport properties 239












2.4 Mercury porosimetry
Mercury injection porosimetry (MIP) was performed on two plugs of proximal locality to
the FIB-SEM samples investigated here. The plugs were oven dried for 24 hours at 50C,
exposed to a vacuum, and then mercury was injected with step-wise pressure increase in
an omni-directional fashion. Raw volume-pressure data were subjected to a conformance
correction (Almon et al., 2008; Sigal, 2009), which sets the upper boundary for MIP pore
statistics.
2.5 Pulse decay permeability of cores
Permeability of mudstone samples were measured on 1 inch (25 mm) right cylindrical
core plugs with helium using a variation of the pulse decay technique as described by
Jones (1997). This involves creating a pressure difference between a tightly jacketed
sample and upstream and downstream reservoirs and collecting pressure pulse decay data
from upstream reservoir and downstream reservoirs. Core plugs were taken from the
same core but from a slightly different depth as the samples used for FIB/SEM analysis.
Measurements were done on core plugs in an as-received (nominally dry) condition, so
that the plugs probably contained water within the smallest pores.
3 Results
3.1 Haynesville formation mudstone composition and core-plug petrophysical
properties
X-ray diffraction of core samples near those used for imaging show a typical mudstone
mineralogical suite consisting of between an estimated 13%15% (by weight) quartz,
8%12% plagioclase, 15%21% carbonate (mostly calcite), 10%12% chlorite and
33%40% dioctahedral 2:1 layer clay which includes mixed layer illite-smectite,
discrete illite, smectite and detrital micas (determined using methods outlined by
Srodon et al., 2001). Together with the ~5% kerogen (from total organic carbon
measurements), this constitutes 43 to 52% by weight of clay and 48%57%
non-clay. RockEval measurements of total organic content for two samples are ~3.4% by
weight.
Porosity estimations vary by method of measurement. MICP, which probably
underestimates total porosity estimates for two measured samples range between 3.7 and
4.7%. Porosity (estimated from weight and volume of dried plug) -effective stress
measurements were performed on a dried plug and show a linear decrease of porosity
with effective stress (porosity values at 14 MPa, 27 MPa, and 41 MPa were 12%, 11%,
and 9% respectively). Extrapolation this linear relationship back to zero effective stress
gives a porosity value of 12.8%. Gas permeability from three samples ranged from 20 to
55 nanodarcy (2.0 to 5.4 10
20
m
2
) at 2,000 psi (13.8 MPa) confining pressure, and
dropped to ~5 nanodarcy (~ 5 10
21
m
2
) for all three samples at 6,000 psi (41.3 MPa).










240 T.A. Dewers et al.












3.2 Imaged mudstone pore types
The four photomicrographs in Figure 2 display some of the variety of mineral and pore
types observed by FIB-SEM in the Haynesville Formation mudstone samples. Figure 2(a)
shows quartz (q) grains suspended in a matrix of clay minerals (cl). Kerogen appears in
rare patches [labelled o in Figure 2(a)], and unlike Loucks et al. (2009) and Heath et al.
(2011, their Figure 7), we observed no distinct pore types in organic phases.
Energy-dispersive X-ray spectroscopic (EDS) chemical analysis was not done on these
samples, however based on XRD analysis the clay matrix is probably mostly
illite-smectite mixed layered clays. Figure 2(b) shows the range of observed pore types in
the imaged Haynesville samples. These are similar in morphology to the slit pores
(Type I), crescent-shaped pores in clay matrix (Type II), and pores adjacent to
clastic grains in so-called compaction- or strain-shadows (Type III) described by
Desbois et al. (2009) in tertiary boom clay (Belgium) and by Heath et al. (2011) in
continental and marine mudstones from a variety of depositional environments and
geologic ages (USA).
Slit pores are elongate and roughly perpendicular to the deposition direction (up in
the photos in Figure 2) and a good example occurs just above the label in Figure 2(b).
So-called Type II pores are crescent-shaped pores in folded clay fabrics, and are found
throughout the clay matrix [surrounding the label in Figure 2(b)]. Type III pores are
dominantly found in the compaction or strain shadows circumventing larger clastic quartz
and feldspar grains and are generally triangular in cross section. Good examples are
located adjacent to the centrally located labelled quartz grain in Figure 2(b). Many pore
types are in fact hybrids of these three distinct morphologies, as seem to populate the
sample imaged in Figure 2(c). In addition, pore types may originate from diagenetic
reactions. The middle cluster of pores in Figure 2(d) may be interpreted as resulting from
replacement of the illite-smectite matrix and is probably chlorite and associated calcite
in-fill [light grey phase in Figure 2(d)]. The authors have observed similar chlorite
rosette morphologies in other mudstones that have been verified compositionally by
EDS.
3.3 Pore network properties and tortuosity
Total porosity of the three-dimensional reconstruction of pore volumes in Figure 4 was
calculated by voxel count to be 7.5%, representing a pore volume of 11.4 cubic microns.
Ignoring pore voxels that touch the exterior of the imaged domain, this includes 97.8%
disconnected pore voxels (most just a few voxels in size) and 2.15% connected voxels.
Analysis of medial axes of connected pore voxels (i.e., the skeleton of connected pores
calculated using the 3DMA-Rock software package developed by Lindquist, 1999)
shown in Figure 5, discerns between pore bodies and pore throats. Open regions not
containing medial axis voxels [the clear regions in Figure 5(a) and Figure 5(b)] are
generally occupied by detrital grains of quartz or plagioclase. There appears to be
near-linear regions of connected pores, or regions with dense distributions of medial axis
voxels, running from the lower left to upper right region of Figure 5(a), which is
suggestive of localised pathways of fluid migration. A total of 9,546 pores are defined in
the imaged domain, including 8,674 pores within the interior of the imaged region, and
872 that touch the region boundaries.









Three-dimensional pore networks and transport properties 241












The correlation between pore body size and topology is investigated through use of
joint and marginal relative frequency histograms of pore body volume and coordination
number [Figure 5(c)]. The pore body volumes exhibit a log10 normal distribution (top
marginal histogram), apparently a typical distribution for pores in mudstones (Heath
et al., 2011). The coordination number (i.e., number of connected, neighbouring pores)
with the highest relative frequency is 2 (see the side marginal histogram), which
coincides with the peak of the pore body volume distribution. Thus, the most frequently
occurring pore body size has a coordination number of 2. This may represent the
touching of slit-shaped Type I or triangular shaped Type II and Type III pores at their
margins or tips. The up-and-to-the-right pattern within the joint histogram may reflect a
scaling relationship between pore body size and coordination number. The second most
frequent coordination number is zero, which corresponds with a range of pore body sizes,
due to the occurrence of disconnected pores of a variety of sizes (at FIB-SEM resolution).
The smallest imaged pores have coordination numbers of zero and 1, with 1 being the
more frequent coordination number showing that many small pores are part of
connected networks.
The medial axes can be used to determine lengths of connected pores across the
imaged domain and so be used to estimate tortuosity of possible flow paths. Figure 6(a),
Figure 6(b), and Figure 6(c) depict calculated shortest paths using the 3DMA software
and the medial axes voxels shown in Figure 5(a) and Figure 5(b). The x-pathways were
chosen from left to right as shown, y-pathways were chosen from top to bottom as
shown, and z-pathways were chosen from front to back as shown. In all cases many
initial dendritic paths neck down to a single flow path as one proceeds through the
imaged volume. Tortuosity is calculated by dividing the total distance between medial
axis voxels on opposing faces by the Euclidian distance between the same voxels
(Lindquist et al., 2005) and have been calculated for both the shortest paths shown in
Figure 6(a) to Figure 6(c) and all connected paths (e.g., there are ~9 million connected
paths in the y-direction). Binned probability histograms of tortuosity values in the
x-, y-, and z-direction are shown in Figure 6(d), Figure 6(e), and Figure 6(f) respectively.
In the x-direction, mean values include 2.1 for shortest paths and 2.2 for all paths with a
range from ~1.5 to ~4.5. Similar values are obtained for the y- and z-directions (~2 for
the y-direction and ~1.7 for the z-direction) which shows a surprising lack of anisotropy
in the distributions of connected pathways.
3.4 Single phase fluid flow and permeability in two pore network types
The transition from vicinal (or surface-bound) water to bulk water properties probably
occurs in pores of ~23 nm in size (Dewhurst et al., 1999) so that water in
FIB/SEM-resolved pores likely behaves as bulk water. This suggests that we could
apply a Stokes type of flow modelling to estimate flow properties of the imaged volumes.
It is debatable whether the small volumes investigated in this section constitute a REV,
but this is beyond the scope of this paper, requiring much more detailed characterisation
of heterogeneity at all scales of the mudstone samples than is attempted here. Large
portions of connected pores from the volume shown in Figure 4 can be isolated as
networks and further examined for flow properties. Using a floodfill algorithm in
ScanIP
TM
, two regions of discrete connectivity (i.e., are separate connected networks










242 T.A. Dewers et al.












isolated from remaining pores) have been selected from the same initial volume and
correspond to one network of Type II/Type III pores in between clastic grains [Figure
7(a) and Figure7(b)] and another corresponding to Type I or slit-like pores [Figure 7(c)
and Figure 7(d)].
Figure 8 CFD simulation of fluid flow through the connected network of Figure 7(a) and
Figure 7(b) (see online version for colours)

Notes: Units on axes of plots correspond to 10
6
metres (microns).
(a) Shows actual volume mesh and (b) shows streamlines through the volume.
Yellow areas are cross sectional slices through the pore network. Colours of
streamlines correspond to velocities, so the blue streamlines mostly correspond to
dead-end pores while the rainbow-coloured streamline, delineated by the red
arrow and corresponding to a portion of the shortest path in Figure 6(c), would be
the dominant flow path.









Three-dimensional pore networks and transport properties 243












Figure 9 CFD simulation of fluid flow through the slit-pore network of Figure 7(c) and
Figure 7(d) (see online version for colours)

Notes: (a) shows the actual volume while (b) shows streamlines. Unlike the case in
Figure 8, there is little dead-end volume indicated by the relatively uniform
streamline velocities and the associated colours. Units on axes correspond to 10
-6

metres (microns). Pore cross sectional areas are shown in blue. Flow direction is
from bottom to top.
Figure 7(a) depicts a network of many small Type II and III pores, corresponding to an
overall pore volume of 8.4 cubic microns. Figure 7(b) shows an image slice and cross
section for that slice through the network of pores. The mesh generating algorithm of
ScanIP
TM
was used to generate a finite element mesh of this pore network consisting of
~35 million nodes [Figure 8(a)] to the software package COMSOL Multiphysics
TM
for
CFDs simulation. The fluid dynamics module in COMSOLMultiphysics
TM
allows for









244 T.A. Dewers et al.












solution of the Navier-Stokes equations for modelling incompressible laminar flows in
arbitrary geometries, and was ideal for obtaining estimates of permeability, as described
in the Methods section. A small flux of 1.3 10
30
m
3
/s was imposed across the simulated
pore network with a small pressure gradient imposed at boundaries, which along with
fluid properties and cross sectional area and using Darcys Law [equation (1)] yields a
permeability of 2.7 nanodarcys (equal to 2.6 10
21
m
2
). Streamlines resulting from the
simulation are shown in Figure 8(b). It is notable that only one streamline shows a
velocity of any magnitude, labelled by the red arrow in Figure 8(b). The remaining
streamlines have negligible flow velocities and thus can be considered as dead-end
pores. The former, fast streamline corresponds to a portion of the shortest-path plotted
in Figure 6(c).
A network consisting of two micron-sized Type I slit pores connected by a narrow
pore throat is displayed in Figure 7(c), with Figure 7D showing a corresponding image
slice and cross section. This network is smaller, with pore volume of only 0.9 cubic
microns. A finite element mesh consisting of 580,000 nodes was used in a CFD
simulation of incompressible laminar flow [Figure 9(a)]. A small steady-state flux of
2.8 10
30
m
3
/s was imposed across the mesh with constant pressure boundary conditions
and corresponds to a calculated permeability of 11 nanodarcys (equal to 1.1 10
20
m
2
).
In contrast to the pore network modelled in Figure 8, the streamline velocities would
indicate that most of the pore volume in the slit-pore network is swept by flow, with little
dead-end volume.
3.5 A comparison of MICP-derived and FIB-SEM pore throat distributions
A number of studies use pore body and pore throat statistics determined from image data
sets to compare with properties discerned from MICP, with mixed results (e.g., Meyer
and Klobes, 1999). This in large part is due to the simple bundle of tubes models used to
infer pore throat size from mercury intrusion data, and the so-called ink-bottle effect
wherein small pore throats act as threshold capillary barriers to larger pores (Diamond,
2000). The larger pores, posing no capillary pressure barrier, would not be counted in the
mercury data compilation of pore throat sizes. Thus MICP data generally are thought to
overpredict the smaller pore throat sizes in a distribution. There is also the possiblity that
the large intrusion pressures used in MICP could cause partial pore collapse of
unsaturated smaller pores during an MICP test.
MICP capillary pressure curves for two core samples, corresponding to the samples
near those used in the image analysis portion of this study, are shown in Figure 10(a).
There is an almost linear increase in capillary pressure with non-wetting phase (i.e.,
mercury) saturation, which would correspond to a steady distribution of decreasing pore
sizes intruded by mercury, with no large threshold of pore sizes (this would result in a
plateau of the capillary pressure curve, as often seen in larger pore-sized rocks like
sandstones, Almon et al., 2008). Pore throat radii data, as a probability distribution, from
both MICP data and image data are plotted in Figure 10(b). It is clear that the mercury
data would predict a dominance of pore sizes between 1 nm and 5 nm, while the
FIB-SEM data predict a larger mean throat size, and in fact fall off below 10 nm (this is
due to resolution issues with voxel sizes of 7.14 nm). It thus appears that MICP and
image data sets are interrogating very different sets of pores.










Three-dimensional pore networks and transport properties 245












Figure 10 Comparison of porosimetry and imaging pore distributions and scaling relationships
(see online version for colours)


Notes: (a) Mercury intrusion capillary pressure (MICP) curves for two samples of the
Haynesville formation taken near the depth of the sampled FIB-SEM
volumes. The curves for two samples are essentially the same.
(b) Pore throat size distributions from MICP data compared to that determined
from FIB-SEM reconstruction. The MICP data (which really are indicative of
threshold dimensions) indicate a much smaller distribution of pore throats that
are below the FIB-SEM resolution.
(c) Same as (b) but in log-log space. The FIB-SEM data set is suggestive of a
bi-fractal relationship, with one power-law relation of smaller throats having
the same slope (or fractal dimension) as the MICP data, and one of larger
throats having a much larger slope, or fractal dimension, corresponding to
throat radii larger than about 20 nm.
This is further shown in a log-log plot of the pore throat size distributions [Figure 10(c)]
which indicate power-law scaling in pore throat sizes. Interestingly, the slopes of both
sets of MICP data (1.2 for the xx065 data set and 1.3 for the xx114 data set) are similar
to a power-law slope (1.2) for the smallest throats in the imaged data set distribution,
between about 8 nm and 20 nm. For throat sizes larger than 20 nm, a different power-law
scaling relationship appears to hold with slope equal to 4.6. There thus appears to be
two sets of scaling relations corresponding to two sets of pores. One set of pores less than
~20 nm is observed by both MICP and image data sets, and one set of larger pores is
observed only by the image data set. The former, which are barely resolvable as pores in
the clay matrix of Figure 2, for example, dominate the capillary pressure relations of
Figure 10(a). The latter form connected networks on the length scale of tens of microns









246 T.A. Dewers et al.












(our imaged volume) and may be responsible for the bulk measured absolute (single fluid
phase) permeability.
4 Conclusions
From the above imaging and analysis, albeit from a limited small volume of 152 cubic
microns, it appears that Haynesville Formation mudstones on the micron scale may
behave as a dual porosity medium. Pores 10 nm in size or larger are resolvable using
FIB-SEM methods (~7 nm pixel size), have mean volumes ~ 1 10
5
nm
3
connected to
~2 to 3 neighbours on average, and are bounded by throats with radii ~ 10 nm or more.
Qualitatively these pores are slit shaped or form connected networks that bound adjacent
clastic grains. Smaller pores 15 nm in size are indicated by MICP tests and likely reside
in the interstices of clay mineral packets. Although mudstones in the Haynesville
Formation are organic rich, we found few pores within or bound by masses of organic
matter in the samples we examined (i.e., as seen by Loucks et al., 2009 or as shown by
Heath et al., 2011, see their Figure 7).
Incompressible laminar modelling of flow in connected pore sub-volumes is used to
calculate permeabilities for networks of larger pores in the range of 2 to 12 nanodarcys.
While similar in magnitude to results of core-scale permeability tests using Haynesville
mudstone samples (550 nanodarcys), any conclusion of similarity needs to be applied
with caution given the small volume examined in this study. We make no claim as to
what constitutes a REV in mudstones, but suggest that the methods detailed here could be
used to discern REVs for different mudstone microfacies. It is nonetheless possible that
single phase fluid flow is limited by pore throats similar in size and topology to that
observed in the larger portion of the imaged pore throat distribution in Figure 10(c). The
slit-like pores should be especially stress-sensitive and may be responsible for the order
of magnitude decrease in measured core-plug permeability with factor of ~2 increase in
confining pressure.
These pore types are to be distinguished from smaller pores, the largest of which are
barely resolvable by FIB-SEM, with sizes at least as small as 1 nm determined by MICP.
These pores are part of the clay matrix fabric itself, and would contain bound water with
properties different than bulk water (Dewhurst et al., 1999). Such a dual-porosity model
is similar to the conceptual model of Sanchez et al. (2008), as used by them in developing
a mechanical constitutive model for partially saturated clays. FIB-SEM image data sets
such as the ones presented here could be used to better constrain conceptual, pore scale,
and continuum scale models for reactive transport, single and multiphase flow, and
geomechanical response of shale gas formations.
Acknowledgements
The authors wish to thank Joe Michael and Michael Rye for help in acquiring the image
slices and sample preparation. Brent Lindquist is thanked for providing the 3DMA-Rock
software. The comments of three anonymous reviewers were of great benefit in preparing
the manuscript and the authors thank them for their time and effort. The authors
gratefully acknowledge funding from the US Department of Energy Office of Basic
Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences.









Three-dimensional pore networks and transport properties 247












Sandia National Laboratories is a multi-program laboratory operated by Sandia
Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the US
Department of Energys National Nuclear Security Administration under contract
DE-AC04-94AL85000.
References
Almon, W.R., Dawson, W.C., Botero-Duque, F., Goggin, L.R. and Yun, J.W. (2008) Seal Analysis
Workshop: Short Course Notes held at the Joint Annual Meeting of the Geological Society of
America, Soil Science Society of America, American Society of Agronomy, Crop Science
Society of America, and the Gulf Coast Association of Geological Studies with the Gulf Coast
Section of SEPM, 59 October, Houston, Texas.
De Winter, D.A.M., Schneijdenberg, C., Lebbink, M.N., Lich, B., Verkleij, A.J., Drury, M.R. and
Humbel, B.M. (2009) Tomography of insulating biological and geological materials using
focused ion beam (FIB) sectioning and low-kV BSE imaging, Journal of Microscopy-Oxford,
Vol. 233, No. 3, pp.372383.
Desbois, G., Urai, J.L. and Kukla, P.A. (2009) Morphology of the pore space in
claystones evidence from BIB/FIB ion beam sectioning and cryo-SEM obersvations,
e-Earth Discuss., Vol. 4, pp.119, doi:10.5194/eed-4-1-2009.
Desbois, G., Urai, J.L., Burkhardt, C., Drury, M.R., Hayles, M. and Humbel, B. (2008) Cryogenic
vitrification and 3D serial sectioning using high resolution cryo-FIB SEM technology for
brine-filled grain boundaries in halite: first results, Geofluids, Vol. 8, pp.6072.
Dewhurst, D.N., Yang, Y. and Aplin, A. (1999) Permeability and fluid flow in natural mudstones,
in Applin, A.C., Fleet, A.J. and Macquaker, J.H.S. (Eds.): Muds and Mudstones: Physical and
Fluid Flow Properties, Vol. 158, pp.2343, Geological Society of London Special
Publications, Geological Society of London.
Diamond, S. (2000) Mercury porosimetry an inappropriate method for the measurement of pore
size distributions in cement-based materials, Cement and Concrete Research, Vol. 30, No. 10,
pp.15171525.
Dobson, L.M. and Buffler, R.T. (1997) Seismic stratigraphy and geologic history of Jurassi Rocks,
northeastern Gulf of Mexico, A.A.P.G. Bulletin, Vol. 81, No. 1, pp.100120.
Ferreira, T.A. and Rasband, W. (2010) The ImageJ User Guide Version 1.43, available at
http://rsbweb.nih.gov/ij/docs/user-guide.pdf (accessed on March 2011).
Heath, J.E., Dewers, T.A., McPherson, B.J.O.L., Petrusak, R., Chidsey, T.C. Jr., Rinehart, A. and
Mozley, P.S. (2011) Pore networks in continental and marine mudstones: characteristics and
controls on sealing behavior, Geosphere, Vol. 7, No. 5, pp.429454.
Holzer, L., Indutnyi, F., Gasser, P.H., Munch, B. and Wegmann, M. (2004) Three-dimensional
analysis of porous BaTiO3 ceramics using FIB nanotomography, Journal of
Microscopy-Oxford, Vol. 216, No. 1, pp.8495.
Jones, S.C. (1997) A technique for faster pulse-decay permeability measurements in tight rocks,
Society of Petroleum Engineers Journal, Vol. 12, No. 1, pp.1926.
Lindquist, W.B. (1999) Report No. SUSB-AMS-99-20, Dept. Applied Math. & Stat., SUNY
Stony Brook.
Lindquist, W.B. and Venkatarangan, A. (1999) Investigating 3D geometry of porous media from
high resolution images, Physics and Chemistry of the Earth Part A-Solid Earth and Geodesy,
Vol. 24, No. 7, pp.593599.
Lindquist, W.B., Lee, S.M., Oh., W., Venkatarangan, A.B., Shin, H. and Prodanovic, M. (2005)
3DMA-Rock: A Software Package for Automated Analysis of Rock Pore Structure in 3-D
Computed Microtomography Images, State University of New York at Stony Brook,
http://www.ams.sunysb.edu/~lindquis/3dma/3dma_rock/3dma_rock.html (accessed on March
2011).









248 T.A. Dewers et al.












Loucks, R.G., Reed, R.M., Ruppel, S.C. and Jarvie, D.M. (2009) Morphology, genesis, and
distribution of nanometer-scale pores in siliceous mudstones of the Mississippian Barnett
Shale, Journal of Sedimentary Research, Vol. 79, pp.848861.
Mancini, E.A., Epsman, M.L. and Stief, D.D. (1997) Characterization and evaluation of the Upper
Jurassic Frisco City sandstone reservoir in southwestern Alabama utilizing Fullbore Formation
MicroImager technology, Gulf Coast Association of Geological Societies Transactions,
Vol. 47, pp.329335.
Meijering, E. (2010) TransformJ A Java Package for Geometrical Image Transformation,
available at http://www.imagescience.org/meijering/software/transformj/ (accessed on March
2011).
Meyer, K. and Klobes, P. (1999) Comparison between different presentations of pore size
distribution in porous materials, Fresenius Journal of Analytical Chemistry, Vol. 363,
pp.174178.
Rasband, W.S. (2010) ImageJ, National Institutes of Health, Bethesda, Maryland, USA, available
at http://rsb.info.nih.gov/ij/ (accessed on March 2011).
Sanchez, M., Gens, A., Guimaraes, L. and Olivella, S. (2008) Implementation algorithm of a
generalized plasticity model for swelling clays, Computers and Geotechnics, Vol. 35, No. 6,
pp.860871.
Shin, S., Lindquist, W.B., Sahagian, D.L. and Song, S-R. (2005) Analysis of the vesicular
structure of basalts, Computers and Geosciences, Vol. 31, No. 4, pp.473487.
Sigal, R.F. (2009) A methodology for blank and conformance corrections for high pressure
mercury porosimetry, Measurement Science and Technology, Vol. 20, No. 4,
doi: 10.1088/0957-0233/20/4/045108.
Srodon, J., Drits, V.A., McCarty, D.K., Hsieh, J.C.C. and Eberl, D.D. (2001) Quantitative X-ray
diffraction analysis of clay-bearing rocks from random preparations, Clays and Clay
Minerals, Vol. 49, pp.514528.
Tolson, J.S., Copeland, C.W. and Bearden, B.L. (1983) Stratigraphic profiles of Jurassic strata in
the western part of the Alabama Coastal Plain, Geological Survey of Alabama Bulletin,
Vol. 122, p.425.
Tomutsa, L., Silin, D. and Radmilovic, V. (2007) Analysis of chalk petrophysical properties by
means of submicron-scale pore imaging and modeling, SPE Reservoir Evaluation and
Engineering, Vol. 10, No. 3, pp.285293.
Yao, N. (2007) Introduction to the focused ion beam system, in Yao, N. (Ed.): Focused Ion Beam
Systems: Basics and Applications, Cambridge University Press, Cambridge, UK.