Content uploaded by Andrew Fogden
Author content
All content in this area was uploaded by Andrew Fogden on Nov 06, 2017
Content may be subject to copyright.
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
1
COMBINING HIGH-RESOLUTION WITH LARGER VOLUME IMAGES
FOR IMPROVED CHARACTERIZATION OF MUDSTONE RESERVOIRS
Andrew Fogden, Alessio Arena, Christopher Zhang, Anna Carnerup, Eric Goergen, Terri Olson, FEI Oil & Gas;
Qianhao Cheng, Jill Middleton, Andrew Kingston, Department of Applied Mathematics, Australian National
University; Yulai Zhang, Ryan Armstrong, School of Petroleum Engineering, University of New South Wales
Copyright 2016, held jointly by the Society of Petrophysicists and Well Log
Analysts (SPWLA) and the submitting authors.
This paper was prepared for presentation at the SPWLA 57th Annual Logging
Symposium held in Reykjavik, Iceland June 25-29, 2016.
ABSTRACT
Optimization of production from shale reservoirs
requires understanding of rock properties over a range
of scales. Multiple imaging techniques can be combined
to determine the nature, connectivity, and wettability of
nano-scale pore systems as well as the underlying
mineralogy and organic textures that control reservoir
behavior and the propensity of the matrix to fail and to
contain expulsion cracks. The current study
demonstrates new capabilities in integrated multiscale
and time-resolved imaging and analysis workflows for
three organic-rich shale samples from two formations.
The spatial distributions of connected porosity, organic
matter, and microfractures within vertical sub-plugs
were quantified from micro-CT imaging, using X-ray
contrast enhancement strategies to detect their volume
contributions from sub-resolution features, together
with tomogram alignment and segmentation. These
registered 3D volume distributions comprising billions
of voxels showed that most of the porosity in these
three samples was hosted by organic matter and most of
the coring-induced fractures ran through laminations of
locally higher organic content. Dynamic micro-CT
imaging was also performed to directly visualize the
progress of liquid-liquid diffusion through the pore
space. The imaged concentration profiles were fitted to
models to estimate the average in-plane diffusivity
coefficient.
This tomographic analysis was validated and
complemented by automated high-resolution 2D back-
scattered SEM (BSEM) and SEM-EDS imaging and
mapping of pores, organic matter and mineralogy over
ion-milled sub-plug sections, and registration of these
image mosaics into the corresponding tomogram cross-
section. In this way, information on the fine scale of
individual features could be combined with statistics
over the more representative tomogram volumes. The
distribution of organic matter was characterized from
this 2D BSEM together with 3D FIBSEM imaging. The
majority of organic-hosted connected pores detected by
contrast-enhanced micro-CT lay below BSEM and
FIBSEM resolution. Secondary electron SEM images
(using FESEM) of raw broken surfaces revealed the
relatively homogeneous texture of the sub-10 nm pore
network permeating the fused aggregates of bitumen
nano-granules. Further, the same contrast technique
used to highlight bitumen in the tomograms was also
applied to ion-milled sections to extend the automated
BSEM imaging coupled with SEM-EDS mapping to
distinguish bitumen from kerogen at high resolution.
INTRODUCTION
The emergence of shale/mudstone as an important
reservoir rock type has led to development of new tools
and applications to characterize these challenging, fine-
grained, source rocks. This ability is critical to locating
production sweet spots, choosing the best target
intervals for horizontal wells, and optimizing
completion practices to maximize production while
controlling costs. High-resolution imaging with 2D
scanning electron microscopy (SEM) and 3D focused
ion beam (FIBSEM) has been the focus of many shale
characterization projects. Due to the trade-offs between
resolution, field of view and acquisition hardware
limitations, FIBSEM volumes over which connected
pore networks are sufficiently well resolved are
typically tiny compared to sample volumes used for all
other lab techniques. The context and representivity of
FIBSEM images is not well understood, mainly due to
the impracticalities of repeated imaging at many
locations.
3D imaging of core plugs and sub-plugs of
unconventionals with micro-CT (computed
tomography) provides this broader context as a basis for
determining heterogeneity and representative
elementary volumes (REV) at that scale and for
selecting appropriate sub-samples for higher resolution
SEM and FIBSEM imaging (Knackstedt et al., 2013).
Previous work had suggested that synchrotron-based
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
2
micro-CT was necessary to resolve the distribution of
low-density organic matter (OM) and the connected
pore network in shale samples (Peng et al., 2015).
However, supplementation of micro-CT with X-ray
contrast strategies specifically designed for shales,
combined with tomogram registration and segmentation
algorithms, enables determination of the volume
distributions of the key components in mudstone
reservoirs: connected porosity, OM, and microcracks
(Fogden et al., 2014a, 2014b). Significantly, their
contributions from sub-voxel features can be captured
since the contrast agent greatly amplifies the selective
signal. The spatial alignment of these phases, notably
porosity and OM, can thus be shown statistically for the
entire imaged volume or for subsets possessing specific
characteristics or representing different petrophysical
rock types. This provides an unbiased approach to
evaluating the proportion of porosity hosted by OM.
The REV for a given property can also be derived from
the micro-CT data and compared to the FIBSEM image
scales.
Another limitation of SEM-based methods is their static
nature. Their limitation to at most a low vacuum places
severe limitations on the scope for studying fluids in
place. Dynamic imaging provides a basis for
understanding transport properties by capturing the
changes in fluid occupancy over time and space. Micro-
CT images can be acquired in situ while the sample
occupancy is undergoing changes, e.g. in concentration
due to diffusion (Fogden et al., 2015a, 2015b). Again,
the use of contrast agents together with slow transport
in shales circumvents the necessity for synchrotron
facilities. Dynamic imaging is particularly important in
nano-porous systems which pose difficulties in
measurement using standard techniques and in
modeling due to insufficiently understood flow physics.
The current study extends previous work by using three
organic-rich shales from two formations to illustrate
and further develop this integrated multiscale workflow
for imaging and analysis, spanning from static and
dynamic micro-CT to SEM (using back-scattered and
secondary electrons) and SEM-EDS, to FIBSEM.
EXPERIMENTAL
Samples. Table 1 lists the three organic-rich shales
samples for study, together with their composition
measured by the suppliers. A2 is a Barnett shale from
the whole core section extensively characterized by the
University of Oklahoma and FEI (Curtis et al., 2014). It
is from the gas window of thermal maturity, with
vitrinite reflectance of 1.63 %Ro. In particular, the A2
sub-plug addressed here is the same as that in two
recent publications (Fogden et al., 2015a, 2015b). The
other two samples, A3 and A4, are calcareous shales
from the wet gas/condensate window in the same
formation, and share fairly similar compositions. The
A3 and A4 core plugs and/or their sub-plugs addressed
here were also the subject of recent publications
(Fogden et al., 2014b, 2015a).
Workflow. From each of the vertical core plugs, a
smaller sub-plug was cored perpendicular to bedding
using a manually-fed drill press with diamond coring
bit and air as lubricant. The sub-plug diameter was 5
mm for A2 and 12 mm for A3 and A4. Each sub-plug
then passed through all or some of the steps of the
workflow in Fig. 1. The first five Steps (a)-(e) involved
3D tomographic imaging of the full height of the sub-
plug, acquired by HeliScan micro-CT at the FEI-ANU
facility (Sheppard et al., 2014). All tomogram post-
processing was performed using MANGO software.
Each sub-plug was first cleaned by extended soaking in
toluene and methanol and dried under vacuum, after
which it was helically scanned in this Dry state (Step
(a)). It was then removed from its aluminum-tube
holder and saturated off-line with the X-ray dense
liquid diiodomethane (CH2I2) (Fogden et al., 2014a,
2014b), via vacuum infiltration for 1 day followed by
isostatic pressurization in CH2I2 at 55 MPa over 20
days. The sub-plug was returned to its holder for helical
scanning (under ambient conditions) in this CH2I2-
saturated state (Step (b)). After this and all following
Steps (c)-(e), the tomograms were corrected for beam
hardening artifacts and masked, and the new tomogram
was 3D spatially aligned to within one voxel with the
fixed Dry tomogram and its attenuation was linearly
rescaled to match, from which the difference tomogram
(New minus Dry) was generated (Sheppard et al.,
2014).
The CH2I2-saturated sub-plug was then replaced in its
holder, now filled with the miscible solvent toluene and
again sealed under ambient conditions. A section of
interest was circular scanned continuously to monitor
the diffusion of CH2I2 from the sub-plug into the
surrounding (X-ray transparent) toluene via the
corresponding loss of X-ray attenuation (Step (c) in Fig.
1). The sub-plug was then removed from its holder for
further soaking in toluene to removal all CH2I2 and
dried. The OM was later stained with the X-ray dense
molecule iodine (I2). This was performed by infiltrating
the sub-plug with 2.5 wt% I2 solution in toluene and
ambient soaking for 20 days, after which the sub-plug
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
3
was removed for ambient evaporation and outgassing of
toluene and unbound I2 for 20 days. The sub-plug was
then returned to its holder for helical scanning in this I2-
stained state (Step (d)).
A longitudinal section along the entire sub-plug was
prepared by broad-beam Ar ion milling. The sub-plug
was helically scanned in this Milled state (Step (e) in
Fig. 1) for subsequent application of a surface
projection algorithm (Dodd et al., 2014) to obtain a 2D
map of the milled face and its virtual location in the
sub-plug prior to milling. The milled face was back-
scattered SEM (BSEM) imaged (Helios, FEI) at 1 kV
without conductive coating, using MAPS (FEI)
software to automate acquisition (at 10 nm/pixel) and
stitching of a mosaic of tiles (Step (f)). The face was
then carbon coated for SEM-EDS imaging (Quanta,
FEI), using MAPS Mineralogy (FEI) software for
automatic acquisition (at 1 µm step size), stitching and
assignment of elemental and mineralogical
composition. This pair of 2D mosaic images was
spatially registered into the corresponding cross-section
through the sub-plug tomograms via the bridging Step
(e). A location on the milled face was chosen for
automated acquisition of a FIBSEM dataset (Helios,
FEI) (Step (g)), reconstructed, processed and analyzed
by PerGeos software (FEI). A raw surface was broken
off for ultra-high resolution FESEM imaging using
secondary electrons at 1 kV without conductive coating
(Step (h)).
The A2 sub-plug passed through all Steps (a)-(h) in Fig.
1, while A3 and A4 underwent only Steps (a)-(d),
although A4 omitted Step (c).
3D TOMOGRAPHIC MAPPING OF POROSITY
AND ORGANIC MATTER
Figs. 2-4 show the same longitudinal (perpendicular to
bedding) central slice from the tomograms of the
vertical sub-plugs A2-A4, registered between Steps (a),
(b) and (d) in Fig. 1. Image (a) in Figs. 2-4 is of the
cleaned Dry state, which serves to characterize the
overall distribution of dense (very bright) mineral
grains or aggregates (chiefly pyrite) together with the
resolved (bright) calcite grains prevalent in A4. The
Dry tomogram also shows the wider (dark) fractures
presumably induced by coring. Other resolved dark
features, such as intra-fossil cavities (principally within
foraminifera in A3 and A4) or more elongated, or
irregularly shaped, intergranular bodies, contain an
unknown combination of the two low-attenuating
phases, namely pore and OM. The sea of intermediate
grayscales surrounding these features reflect the mix of
sub-resolution mineral grains, pores and OM within
each voxel that together dictate the matrix transport
properties.
As explained in earlier publications (Fogden et al.,
2014a, 2014b), re-imaging of the sub-plug after its
saturation with the X-ray dense liquid CH2I2, combined
with registration to the Dry tomogram, allows
generation of the difference tomogram (CH2I2-saturated
minus Dry), shown in Image (b) of Figs. 2-4. Here the
solid fraction of all mineral grains and OM is subtracted
off (seen as black) and the brightness is solely due to
the connected porosity within each voxel, irrespective
of the pore sizes contributing to it. Wide fractures are
naturally brightest, while microfractures not apparent in
the Dry tomogram now become visible in the
difference, possessing intermediate brightness dictated
by their sub-voxel aperture. These bedding-plane
aligned micro-crack segments are prevalent in all three
sub-plugs, but less so in A4 due to its coarser grains.
The brightness of cavities resolved in the Dry
tomogram now represents their internal porosity,
decoupled from the surrounding organic or mineral
content. Similarly, the matrix grayscales in the
difference are directly proportional to the connected
sub-resolution porosity there.
After removal of all infiltrated CH2I2, the sub-plug was
later exposed to free iodine (I2), which infuses between
the molecules of OM and binds via charge
complexation to thus internally stain their solid fraction
(Fogden et al., 2014a, 2014b). All other association or
adsorption of I2 to minerals or OM is reversible, so only
the OM-bonded fraction remains on outgassing. The
same procedure of re-imaging of the sub-plug followed
by tomogram registration and differencing (I2-stained
minus Dry) results in Image (c) in Figs. 2-4. This
tomogram difference generally exhibits less contrast
than its analog in Image (b), since the attenuation
increase from I2 staining is less than from CH2I2
saturation, and the principal resolved features, namely
fractures and microfractures, are not OM-filled and
remain unhighlighted. The discernibly brighter local
features in Image (c), which become increasingly rare
from A2 to A3 to A4, correspond to the OM fraction of
cavities or of short slots lying along segments of
microfractures. Almost all of these were also
highlighted in Image (b) and thus correspond to OM-
hosted porosity. The brightness, i.e. organic content, of
the matrix in Image (c) varies considerably over the
sub-plug, and displays the clearest laminations in A3
(Fig. 3c). For all three samples, the coring-induced
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
4
fractures tend to run through organic-rich subregions,
while microfractures are present in these and in
organic-lean subregions.
The difference tomograms in Images (b) and (c) of
Figs. 2-4 then underwent inversion and systematic
segmentation for conversion to 3D maps of connected
porosity and OM volume fraction, respectively. Note
that porosity segmentation results for sub-plugs A2 and
A3 were recently reported (Fogden et al., 2015a); the
current study refined these estimates by pre-processing
using advanced algorithms to correct subtle imaging
artifacts that otherwise can have a substantial effect on
such relatively featureless shale samples. After
inversion of Image (b), the darkest voxels (within wide
fractures) and brightest voxels (within resolved mineral
grains) were used to seed an intensity-based
segmentation of 100% and 0% porosity subregions
using converging active contours (Sheppard et al.,
2014). Porosity was then ascribed to all remaining
voxels in linear relation to their grayscale variation
between these lower and upper thresholds. The porosity
value thus obtained, averaged over the entire sub-plug
tomogram, is listed in Table 2 and displays a good
agreement with the measured value in Table 1. In spite
of the prevalence of coring-induced fractures, their
contribution to Table 2 is very small.
An analogous segmentation procedure was applied to
Image (c), although the results should be regarded as
semi-quantitative owing to two limitations relative to
Image (b). First, instances of resolved features
comprising 100% solid OM are comparatively rare,
especially owing to the prevalence of OM-hosted
porosity in these three samples. Internal consistency
was enforced by constraining the sum of segmented
porosity and OM volume fraction to lie below 100%.
Second, different types of OM present within one sub-
plug will likely exhibit differing extents of I2 staining
per solid volume, since the stoichiometry of its binding
is dictated by the arrangement and condensation of
aromatic groups within the OM (Aronson et al. 1976).
In spite of these limitations, the overall averages
obtained in Table 2 are again in line with Table 1 if OM
solid density and shale bulk density are approximated
as 1.2 and 2.4 g/cm3, respectively (so that TOC in vol%
is double that in wt%).
Images (d) and (e) in Figs. 2-4 show these grayscale
segmentations of the inversions of Images (b) and (c),
respectively, colorized for illustration, such that the 5
p.u. interval bracketing the overall average value in
Table 2 is colored red, while the 5 p.u. interval above
this is green. (This interval was reduced to 3 p.u. for
porosity of A2, as its average value is lower than all
others.) Porosity and TOC values below the red interval
or above the green interval are given by the lighter or
darker grayscales, respectively. The smaller images in
Figs. 2-4 are digital zoom-ins of squares within Images
(d) and (e).
The A2 sub-plug (Fig. 2) contains a number of thin
bands rich in dense mineral and low in porosity and
OM. The laminations above the uppermost band and
below the lowermost band are noticeably leaner and
richer, respectively, in OM, while porosity is more
uniformly distributed on these millimeter scales.
Between these two bands, porosity and OM exhibit
significant colocation and trend with each other,
implying that one type of OM and porosity hosted by it
predominates over this large middle section. Although
the red interval brackets the mean porosity and TOC,
the mode is skewed towards the green due to the
significant population of light gray voxels of very low
porosity and (local) TOC (including the resolved solid
fraction in Table 2). The green levels of porosity and
TOC can thus be regarded as more representative of the
matrix, with red voxels often occurring as a sheath
between green and light gray due to partial volume
effects at the boundaries.
The A3 sub-plug (Fig. 3) displays the tightest unimodal
frequency distribution of porosity about its mean (i.e.
the greatest red fraction), owing to its lowest population
of resolved grains in Table 2 and to the paucity of high
porosity features, which are mainly limited to the
vicinity of fractures and microfractures and isolated
islands. On the other hand, the distribution of OM is
more heterogeneous and laminated, and with a bimodal
distribution peaking both at low (light gray) and high
(green) TOC. Accordingly, pathways of locally higher-
than-average TOC percolate through the matrix. High
porosity and high TOC subregions display some
colocation, principally due to coring-induced fractures
running through OM-rich laminations. However, the
spatial correlation of porosity and TOC is weaker or
more variable than in A2, suggesting that the sub-plug
comprises multiple OM types. In this case, differing I2
binding capacity of these types may add an extra source
of variability.
Although sub-plugs A3 and A4 are from the same
formation and are compositionally similar (Table 1),
their distributions in Figs. 3 and 4 show key
differences. Porosity displays a more heterogeneous,
laminated distribution than TOC over the A4 sub-plug,
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
5
while the opposite was true of A3. Both properties
display a bimodal frequency distribution since A4
possesses the highest fraction of resolvable grains
(calcite-cemented foraminifera) in Table 2. In common
with A3, the spatial correlation of porosity to OM does
not appear to follow a universal trend, again suggesting
variation in the type of OM or its porosity hosting over
the sub-plug.
3D TOMOGRAPHIC IMAGING AND
MODELING OF DIFFUSION
The contrast-enhanced, registered tomograms in Figs.
2-4 yield 3D maps of volume fraction of sub-resolution
connected porosity and OM per voxel. The CH2I2-
saturated state cannot directly provide information on
the size, shape and connectivity of individual nano-
pores contributing to this voxel signal. Thus preferred
pathways of transport created by locally larger nano-
pores or less tortuous networks cannot be inferred from
this tomogram difference. However, CH2I2 saturation
can be used as the initial state for dynamic micro-CT
imaging (Step (c) in Fig. 1) of the single-phase
diffusion of a miscible, but X-ray transparent, second
liquid, such as toluene (Fogden et al., 2015a, 2015b).
The progress of diffusion is monitored by the loss of
voxel attenuation within the sub-plug over time, as the
concentration of X-ray opaque CH2I2 in the pore-filling
liquid is reduced from its original 100% level by
toluene mixing. The CH2I2-saturated A2 and A3 sub-
plugs immersed in toluene were continuously fast-
scanned, in particular within the A3 sub-section marked
in Fig. 3a. The delay from sub-plug immersion to
acquisition commencement was around 15 min., and
each successive 360o rotation took 54 min. A total of 70
and 91 such successive circular scans (over a total
duration of 63 and 82 h) were acquired for A2 and A3,
respectively.
Fig. 5 shows a central longitudinal slice of the
registered tomograms of the A3 sub-plug for this
sequence of experiments. Fig. 5a gives for reference the
corresponding slice of the Dry tomogram (e.g. Fig. 3a)
within this sub-section, followed in Fig. 5b-g by a
selection of the dynamic tomogram differences
acquired over the 82 h of scanning. Fig. 5b is from the
first diffusion tomogram, and Fig. 5c-f correspond to
intermediate tomogram numbers increasing in a squared
progression, while Fig. 5g is the final member. The
difference in Fig. 5b-g is the CH2I2-saturated tomogram
minus the Diffusion tomogram, which serves to isolate
and highlight the local changes due to toluene ingress
via the progressive brightening. Fig. 5h gives the
corresponding slice of the difference tomogram CH2I2-
saturated minus Dry (e.g. Fig. 3b) within this sub-
section, displayed here using the same attenuation range
as in Fig. 5b-g. If the experiment had proceeded to
completion, the final Diffusion difference tomogram
would approach that in Fig. 5h, since toluene and air
have similarly low attenuation. As the experiment was
not continued to full equilibration, the final tomogram
difference in Fig. 5g remains darker than Fig. 5h,
especially within the sub-plug interior, although the
brightness in readily accessible regions does approach
the equilibrium level.
As expected, the wide coring-induced fractures leading
in from the sub-plug edges are the fastest routes of
toluene entry in the early stages (Fig. 5b). However,
subsequent diffusion transverse to these fractures
(perpendicular to bedding) is relatively slow, so radial
diffusion in from the sub-plug edge (parallel to
bedding) overtakes it (from Fig. 5c). At longer times
these two modes of ingress merge (from Fig. 5d), so at
the conclusion of scanning, only the central regions
farthest from the edge and fractures remain
substantially uninvaded and undiluted. No major
preferred or prohibited pathways through the matrix are
apparent. This is consistent with the relatively
homogeneous distribution of porosity segmented in Fig.
3d, which indicated limited percolation of high porosity
subregions transverse to fractures in A3.
Microfractures, which are the principal source of local
porosity variation, occasionally accelerate in-plane
ingress, particularly if they run close to coring-induced
fractures. However, many microfractures do not act as
preferential pathways and are only brightened in Fig. 5
once their matrix surrounds have admitted a substantial
concentration of toluene. The homogeneous progress of
diffusion suggests that, along with porosity, pore size
and tortuosity are fairly uniformly distributed over
these scales. It is difficult to assess the effect of the
more heterogeneous distribution of OM segmented in
Fig. 3e (from I2 staining in the later Step (d) of Fig. 1),
as diffusion through OM-rich laminations is biased by
the presence of coring-induced fractures.
Results of the diffusion experiment on the A2 sub-
plug were presented and discussed in earlier
publications (Fogden et al., 2015a, 2015b). The
observed behavior was analogous to that for A3 in
Fig. 5, although diffusion from the edges and
coring-induced fractures progressed closer to
completion, owing to the smaller sub-plug
diameter (5mm). The relatively uniform progress
of diffusion is again consistent with the
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
6
homogeneous distribution of porosity segmented
in Fig. 2 and the inference there that one type of
OM-hosted porosity predominates.
X-ray attenuation of voxels in tomograms and their
differences is linearly related to the local density (Akin
and Kovscek, 2003; Guerrero-Aconcha and Kantzas,
2009; Vega et al., 2013). Based on this link, the
concentration c (equivalent to density) of CH2I2 in the
pore space of the Diffusion tomograms is given by the
following quotient of differences:
Eq. 1
Here is the attenuation at location (x,y,z) of
the Diffusion tomogram after time t of toluene
immersion, is the corresponding value in
the CH2I2-saturated tomogram (Fig. 5h), and
is its value in the Dry tomogram (Fig. 5a).
The Diffusion tomogram differences in Fig. 5b-g are
thus the difference between numerator and denominator
in Eq. 1. The CH2I2 concentration per voxel is thus
quantified and represented on a [0, 1] scale.
To estimate the diffusion coefficient for the A3 sub-
plug, a curve-matching method was used, in which the
concentration profile generated from the tomograms via
Eq. 1 is matched to simulation, using Fick’s second law
in porous media:
Eq. 2
Here De is the effective diffusion coefficient and φ is
average porosity. The diffusion coefficient that
provides the best match is found by comparing the
squared difference between simulation and
experimental data. 1D concentration profiles were
generated along radial lines from the cylinder axis of
the concentration-scaled sub-plug tomograms (using
Eq. 1) to the perimeter at angles (θ) of 0o, 90o, 180o, and
270o. The theoretical concentration profiles were
obtained by solving Eq. 2 using different estimated
diffusion coefficients, after making a 1D model to
represent the sub-plug. The simulations were carried
out using MOOSE software from Idaho National
Laboratory (Gaston et al., 2009).
Assumptions of this approach include: (1) only
molecular diffusion occurs in the mass transport
process; (2) porosity variations are sufficiently small
that an averaged φ of 7% could be used throughout,
based on Fig. 3 and Table 2; (3) diffusive flux is
dominated by its radial component, based on Fig. 5. By
applying this approach, the average diffusion
coefficient of the voxels along a given line profile was
obtained. Fig. 6 presents for A3 the concentration
profiles from the tomograms and their best match curve
from solving Eq. 2. The best-fit De values are on the
order of 10-12 m2/s, similar to previously reported shale
diffusion data (Cavé et al., 2009). The diffusion
coefficient varies by more than a factor of two
depending on the radial angle at which it is measured,
although part of this variation may be due to the
contribution of fractures.
2D MICROSCOPY AND MINERALOGY OF
SECTIONS
Following Step (c) of Fig. 1, the A2 sub-plug was re-
cleaned, ion-milled longitudinally (perpendicular to
bedding) and re-scanned in this Milled state to generate
a 2D image of the planar-projected voxels at the milled
face (Step (e)). The high resolution BSEM image
mosaic and SEM-EDS mineral map subsequently
acquired on this milled section (Step (f)) were then
registered to this tomographic face projection, as shown
in Fig. 7. Further, the Milled tomogram was registered
to its Dry counterpart in Fig. 2a (Fogden et al., 2015b)
to align these BSEM and mineral maps with their
corresponding virtual location within the intact sub-
plug prior to milling. In this way these 2D images could
also be directly compared to the tomographic porosity
map (Fig. 2d) and OM map (Fig. 2e, from I2 staining in
Step (d) which was performed last).
Figs. 8 and 9 show digital zoom-ins of two subareas
from within the BSEM image and SEM-EDS mineral
map in Fig. 7, compared to the registered images,
within this same field of view, of the Dry and Milled
tomograms and the porosity and OM maps (now
without their colorizations in Fig. 2d-e). An exact
pointwise match, especially of fine features, cannot be
expected owing to tomogram resolution limits and
accumulated deviations from acquisition (e.g. electron
beam drift) and processing. However the overall match
is very good and serves to validate the tomographic
results in Fig. 2, while also shedding light on some of
the limitations of the workflow steps.
Fig. 8e reveals that the high organic content of A2 fills
virtually all space between mineral grains, in a manner
suggestive of secondary, migrated OM consistent with
mature bitumen. Virtually no intergranular or
intragranular porosity is apparent, and OM-hosted
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
7
porosity is largely below the 10 nm pixel size, aside
from sporadic cracks in OM bodies or at their mineral
interface. Most of the larger OM bodies correctly match
the darker (OM-rich) features in the OM tomographic
map of Fig. 8c. Most of these also colocate with the
darker (high porosity) features in the porosity map of
Fig. 8b. This illustrates a key advantage of the contrast-
enhanced micro-CT workflow, namely that connected
porosity can be resolved by its attenuation difference
signal even when the pores comprising it lie below the
BSEM spatial resolution. The black line in Fig. 8b is a
fracture, which correctly coincides with the same
feature in the Dry tomogram of Fig. 8a, but which
subsequently vanished from the Milled tomogram of
Fig. 8d due to filling by cutting debris. Thus the fine
grains along this location in Fig. 8e-f are a preparation
artifact. The mineral maps in Figs. 8-9 show that larger
grains of quartz and plagioclase are prevalent, along
with smaller grains of calcite with a little dolomite, plus
pyrite crystals and smaller framboids; the remainder is
mainly illite and smectite mixtures.
The band of migrated OM running near the bottom of
Fig. 9e also coincides with a dark band in the OM and
porosity maps of Fig. 9c and 9b, implying that it hosts
substantial porosity, much more than could be
accounted for by the cracks resolved in Fig. 9e. On the
other hand, the two large OM bodies towards the top of
Fig. 9e have particulate form suggestive of primary,
non-migrated OM, presumably kerogen. These are
picked up as dark (low attenuation) in the Dry
tomogram (red arrow in Fig. 9a), but do not clearly
coincide with dark features in the OM or porosity maps
of Fig. 9c and 9b. This different type of OM, which is
in the minority in A2, apparently does not host porosity,
and moreover does not bind I2. Thus the OM maps in
Figs. 2-4 probably should be more correctly regarded as
bitumen maps. This issue will be further investigated in
following sections.
3D FIBSEM
The power of the tomographic maps of porosity and
OM of Figs. 2-4 is that they provide the vehicle for
upscaling of pore-scale transport properties, derived
from much higher resolution images, to more
representative volumes within the sub-plug. For
example, Lattice-Boltzmann simulations on resolved
pore networks within small volumes can provide poro-
perm curves, or more generally, absolute permeability
plotted against both porosity and OM volume fraction
(since OM-hosted pores are fundamentally different
from mineral-hosted pores). The registered 3D maps of
these two volume fractions then become maps of local
permeability per voxel or grid cell for calculation of
overall permeability. Further, simulation of liquid-
liquid diffusion and its upscaling in this manner can be
directly compared to dynamic micro-CT imaging and
analysis as Figs. 5-6 for added validation.
FIBSEM is the most commonly used approach for 3D
imaging within small cubes at sufficiently high
resolution to simulate transport. A location on the
BSEM-imaged strip of A2 in Fig. 7 was selected for
acquisition of a sequence of FIBSEM slices (Step (g) in
Fig. 1), which were reconstructed into a 3D rectangular
prism for image processing, segmentation and analysis
within PerGeos software. Segmentation of
foreground and background regions took
advantage of the fact that charging artifacts in
pore-backs were in close proximity to pores. Low
intensity porous regions were first segmented via
thresholding. High intensity pore-backs were
similarly segmented, although high intensity
grains could not be excluded. A shape function
(essentially aspect ratio) was used to label and
separate these two, since pore-backs were
elongated whereas grains were more rounded. The
pore-back and pore segmentations were merged
and a 4-pixel kernel closing operation was used to
fill the low intensity pore-back regions between
the two. Finally, the OM regions were segmented
by capturing all low intensity regions and
removing the already segmented porous regions.
Visualizations of the segmented dataset are given in
Fig. 10. The resolved porosity was very low (0.4%),
mainly in the form of isolated cracks, as expected from
the BSEM mosaic. The FIBSEM-segmented bulk
volume fraction of OM (in the form of migrated
bitumen) was 16.5%, within which the vast majority of
sample porosity was hidden below resolution. The
12.1% estimate of the solid component of volume
fraction of OM (I2-stained bitumen) in Table 2 implies a
porosity of 27% within the sub-resolution OM-hosted
porosity, and an overall porosity for A2 of 4.8%,
consistent with Tables 1 and 2. While the lack of
percolating resolved porosity in Fig. 10 precluded
simulation of absolute permeability, an alternative
strategy for organic-rich shales is proposed in the
following section.
2D FESEM OF RAW SURFACES
Fig. 11 compares a digital zoom-in of bitumen from the
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
8
BSEM image strip of A2 in Fig. 7, in which the hosted
pores are faintly apparent as speckle but not resolvable,
with two ultra-high resolution FESEM micrographs of
bitumen at a raw broken surface of A2 (Step (h) in Fig.
1). Virtually all OM imaged over the surface possessed
this internal texture of grainy primary nanoparticles,
suggestive of bitumen, aggregated and fused to support
a network of pores of size around 10 nm, presumably
formed by production of mobile hydrocarbons. This
hosted nano-porosity was only seen to be absent within
the thin layer of OM in contact with mineral grains, the
plucking of which revealed this sealed skin (Fig. 11b).
It is possible that the broad ion beam locally seals OM-
hosted nano-pores in a manner analogous to their local
compaction at mineral interfaces.
Given that this fine texture pervades the bitumen in A2,
one future approach to property calculation would be to
first simulate transport, e.g. poro-perm, on a generic
ensemble of fused nano-granular aggregates, guided by
FESEM images. Note that for OM-hosted nano-
porosity, a Darcy treatment of absolute permeability is
not appropriate and wall slip effects must be included.
This effective medium could then be imported into
FIBSEM cubes to factor in the tortuous bitumen
pathways around grains. The resulting property relation
can then be further upscaled via the 3D sub-plug maps
of porosity and bitumen as mentioned above.
2D MICROSCOPY FOR TYPING OF ORGANIC
MATTER
Comparison in Fig. 9 of BSEM images of the ion-
milled section of A2, performed prior to I2 staining,
with its tomographic OM map after staining, suggested
that bitumen binds I2 while kerogen does not. To shed
further light on this possible distinction and its
significance for the Fig. 1 workflow, a sister sample of
A2 was ion-milled for a sequence of BSEM and SEM-
EDS analyses (without micro-CT), the results of which
are shown in Fig. 12 for two representative subareas.
The surface was first BSEM imaged (in the manner of
Figs. 7-9) in Fig. 12a and again after I2 exposure in Fig.
12b. As expected, the majority of the OM, appearing as
migrated bitumen, noticeably brightened due to its
increase in electron density by bound I2. However, OM
particles suggestive in form of kerogen (red arrows in
Fig. 12a) did not brighten to this same extent,
supporting the lower-resolution observations from Fig.
9. Note that some mineral surfaces also brightened from
Fig. 12a to 12b, suspected to be due to formation of
calcium iodide on the fresh ion-milled planes of calcite.
The sample was cleaned in methanol to remove this
contamination, after which it was carbon coated and re-
imaged by BSEM at higher kV (Fig. 12c), followed by
SEM-EDS mapping using MAPS Mineralogy software
(Fig. 12d) in the manner of Figs. 7-9. Further, the
software was used to automatically generate elemental
maps of carbon (Fig. 12e) and iodine (Fig. 12f). Fig.
12d confirms that calcite was indeed the mineral that
brightened on I2 exposure. The elemental maps in Fig.
12e and 12f largely coincide, and although the iodine
signal is somewhat weaker than carbon, a significant
exception is that the red-arrowed kerogen particles in
Fig. 12a are completely absent from the iodine map.
This provides further evidence that, compared to
kerogen, bitumen has a much greater capacity for I2
binding, due to its higher aromatic content (Aronson et
al. 1976). With appropriate calibration to OM
standards, I2 staining can expand the scope for
geochemical characterization by BSEM (using
before/after imaging) and/or by SEM-EDS (using
elemental mapping), in addition to its utility in
tomographic before/after imaging.
CONCLUSIONS
Contrast-enhanced micro-CT imaging and analysis of
shale samples provides 3D maps of connected porosity
and of OM (chiefly bitumen) for determination of REV
and for upscaling of pore-scale simulated transport
properties to these volumes. This micro-CT workflow
also allows for dynamic imaging of liquid-liquid
diffusion for additional anchoring of transport
simulations. The combination of micro-CT imaging
with BSEM imaging and SEM-EDS mapping of ion-
milled sections provides validation and complementary
insights. For organic-rich shales, ion milling for BSEM
or FIBSEM risks underestimation of OM-hosted
porosity and connectivity for input to simulations,
without support from FESEM imaging of broken
surfaces to visualize the nano-porous networks.
REFERENCES
Akin, S. and A.R. Kovscek, 2003, Computed
tomography in petroleum engineering research:
Geological Society, London, Special Publications, v.
215(1), p. 23-38.
Aronson, S., A. Schwebel, and G. Sinensky, 1976,
Absorption of iodine by coal and lignite: Carbon, v. 14,
p. 93-95.
Cavé, L., et al., 2009, A technique for estimating one-
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
9
dimensional diffusion coefficients in low-permeability
sedimentary rock using x-ray radiography: Comparison
with through-diffusion measurements: Journal of
Contaminant Hydrology, v. 103, p. 1-12.
Curtis, M. E., E. T. Goergen, J. D. Jernigen, C. H.
Sondergeld, R. J.Ambrose, and C. S. Rai, 2014,
Mapping of organic matter distribution on the
centimeter scale with nanometer resolution:
Unconventional Resources Technology Conference,
Denver, CO, Aug. 25-27, SPE-2014-1922757.
Dodd, N., R. Marathe, J. Middleton, A. Fogden, A.
Carnerup, M. Knackstedt, K. Mogensen, X. Marquez,
and S. Frank, 2014, Pore-scale imaging of oil and
wettability in native-state, mixed-wet reservoir
carbonates: International Petroleum Technology
Conference, Doha, Qatar, Jan. 20-22, IPTC 17696.
Fogden, A., S. Latham, T. McKay, R. Marathe, M. L.
Turner, A. Kingston and T. Senden, 2014a, Micro-CT
analysis of pores and organics in unconventionals using
novel contrast strategies: Unconventional Resources
Technology Conference, Denver, CO, Aug. 25-27,
SPE-2014-1922195.
Fogden, A., J. Middleton, T. McKay, S. Latham, R.
Marathe, M. L. Turner, A. Sheppard, J. J. Howard, and
F. D. Lane, 2014b, 3D mapping of pore and organic
matter distribution in unconventional reservoirs
utilizing a digital rocks approach: International
Symposium of the Society of Core Analysts, Avignon,
France, Sept. 8-11, SCA2014-27.
Fogden, A., T. Olson, Q. Cheng, J. Middleton, A.
Kingston, M. Turner, A. Sheppard, and R. Armstrong,
2015a, Dynamic micro-CT imaging of diffusion in
unconventionals: Unconventional Resources
Technology Conference, San Antonio, TX, July 20-22,
SPE-2015-2154822.
Fogden, A., E. Goergen, T. Olson, Q. Cheng, J.
Middleton, A. Kingston, M. Curtis, and J. Jernigen,
2015b, Applications of multi-scale imaging techniques
to unconventional reservoirs: SPE Asia Pacific
Unconventional Resources Conference and Exhibition,
Brisbane, Australia, November 9-11, SPE-176948.
Gaston, D., et al., 2009, MOOSE: A parallel
computational framework for coupled systems of
nonlinear equations: Nuclear Engineering and Design,
v. 239, p. 1768-1778.
Guerrero-Aconcha, U.E. and A. Kantzas, 2009,
Diffusion of hydrocarbon gases in heavy oil and
bitumen: Latin American and Caribbean Petroleum
Engineering Conference, Cartagena de Indias,
Colombia, 31 May-3 June, SPE-122783.
Knackstedt, M., A. Carnerup, A. Golab, R. Sok, B.
Young, and L. Riepe, 2013, Petrophysical
characterization of unconventional reservoir core at
multiple scales: Petrophysics, v. 54, p. 216- 223.
Peng, S., J. Yang, X. Xiao, B. Loucks, S. Ruppel, and
T. Zhang, 2015, An integrated method for upscaling
pore-network characterization and permeability
estimation: example from the Mississippian Barnett
Shale: Transport in Porous Media, v. 109, p. 359-376.
Sheppard, A., S. Latham, J. Middleton, A. Kingston, G.
Myers, T. Varslot, A. Fogden, T. Sawkins, R.
Cruikshank, M. Saadatfar, N. Francois, C. Arns, and T.
Senden, 2014, Techniques in helical scanning, dynamic
imaging and image segmentation for improved
quantitative analysis with X-ray micro-CT: Nuclear
Instruments and Methods in Physics Research B, v.
324, p. 49-56.
Vega, B., A. Dutta, and A.R. Kovscek, 2013, CT
imaging of low-permeability, dual-porosity systems
using high x-ray contrast gas: Transport in Porous
Media, v. 101, p. 81-97.
ABOUT THE AUTHOR
Andrew Fogden received his PhD from the Australian
National University and currently serves as Director of
FEI Australia. His research and development interests
are focused on digital imaging and analysis workflows
for conventionals and unconventionals and the role of
surface chemistry in oil recovery.
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
10
Table 1. Measured composition of the samples. Miss = Mississippian; Cret = Cretaceous.
Sample
Age
Porosity
(%)
TOC
(wt%)
Calcite
(wt%)
Illite/
Smectite
(wt%)
Illite/
Muscovite
(wt%)
Kaolinite
Chlorite
(wt%)
Quartz
Plagioclase
(wt%)
Siderite
(wt%)
Pyrite
(wt%)
A2
Miss
5.0
5.90
3
41
4
35
9
5
A3
Cret
7.6
5.55
54
16
6
4
18
0
2
A4
Cret
5.8
4.93
58
12
3
3
23
1
2
Table 2. Composition of the samples from tomogram segmentation.
Sample
Porosity
(%)
TOC
(vol %)
Resolved solid
(vol %)
A2
5.3
12.1
8.8
A3
7.4
10.8
4.6
A4
6.8
9.6
26.7
Fig. 1. Workflow employed in this study. The sample is scanned in its (a) clean and dry state, in which G, P and OM
denote grayscales of resolved solid grain, pore, and organic matter, and then in its (b) CH2I2-saturated state, which is
registered (blue arrow) to the dry-state tomogram. The sample is then (c) scanned during immersion in toluene to
image its diffusion into the CH2I2-saturated state. The re-cleaned sample is later scanned in its (d) iodine-stained
state. After ion milling, the sample is (e) scanned in this state to aid in registration of the subsequently acquired (f)
BSEM and EDS image mosaics of the milled face into the dry-state tomogram. Further milling into the face is
performed for (g) FIBSEM imaging, and a broken surface is (h) FESEM imaged.
(b) 3D Micro-CT: CH2I2saturated
(d) 3D Micro-CT: I2stained (g) 3D FIBSEM
(a) 3D Micro-CT: Dry
G P OM
(e) 3D Micro-CT: Dry Milled
(f) 2D BSEM and EDS
(c) 3D Micro-CT: Toluene Diffusion (h) 2D FESEM
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
11
Fig. 2. Central longitudinal slice of the registered tomograms of the 5 mm diameter A2 sub-plug for its (a) Dry
state, (b) difference CH2I2-saturated minus Dry, (c) difference I2-stained minus Dry, (d) porosity segmentation from
(b), and (e) OM volume fraction segmentation from (c). Red and green highlight voxels having lower or higher
ranges, respectively, of porosity or organic content. Field of view is 4.9 mm × 5.3 mm at 2.3 µm/voxel; (a) shows a
0.5 mm scale bar. The zoomed-in squares from the porosity (left) and OM (right) maps are 0.51 mm × 0.51 mm.
Volume fraction < v1
v1< Volume fraction < v2
v2< Volume fraction < v3
Volume fraction > v3
φ(%)
OM (vol%)
4710
10 15 20
v1v2v3
(a) (c)(b)
(e)
(d)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
12
Fig. 3. Central longitudinal slice of the registered tomograms of the 12 mm diameter A3 sub-plug for its (a) Dry
state, (b) difference CH2I2-saturated minus Dry, (c) difference I2-stained minus Dry, (d) porosity segmentation from
(b), and (e) OM volume fraction segmentation from (c). Red and green highlight voxels having lower or higher
ranges, respectively, of porosity or organic content. Field of view is 11.7 mm × 16.5 mm at 7.7 µm/voxel; (a) shows
a 1 mm scale bar and the field of view during diffusion. The zoomed-in squares from the porosity (left) and OM
(right) maps are 0.92 mm × 0.92 mm.
Volume fraction < v1
v1< Volume fraction < v2
v2< Volume fraction < v3
Volume fraction > v3
φ(%)
OM (vol%)
4914
914 19
v1v2v3
(a) (c)(b)
(e)
(d)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
13
Fig. 4. Central longitudinal slice of the registered tomograms of the 12 mm diameter A4 sub-plug for its (a) Dry
state, (b) difference CH2I2-saturated minus Dry, (c) difference I2-stained minus Dry, (d) porosity segmentation from
(b), and (e) OM volume fraction segmentation from (c). Red and green highlight voxels having lower or higher
ranges, respectively, of porosity or organic content. Field of view is 12.1 mm × 19.1 mm at 7.5 µm/voxel; (a) shows
a 1 mm scale bar. The zoomed-in squares from the porosity (left) and OM (right) maps are 0.92 mm × 0.92 mm.
(a)
(e)
(d)
(c)(b)
Volume fraction < v1
v1< Volume fraction < v2
v2< Volume fraction < v3
Volume fraction > v3
φ(%)
OM (vol%)
4 9 14
712 17
v1v2v3
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
14
Fig. 5. Central longitudinal slice of the registered tomograms of the 12 mm diameter A3 sub-plug for its (a) Dry
state, and a sequence of Diffusion difference tomograms (CH2I2-saturated minus Diffusion) corresponding to
tomogram numbers (b) 1, (c) 4, (d) 16, (e) 36, (f) 64, and (g) 91, with the equilibrium difference-tomogram (CH2I2-
saturated minus Dry) in (h). Field of view is 12.4 mm × 8.8 mm at 9.6 µm/voxel; (a) shows a 1 mm scale bar.
(a) (b)
(c)
(e)
(d)
(f)
(g) (h)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
15
Fig. 6. Radial profiles of CH2I2 concentration at four orthogonal angles calculated from the dynamic micro-CT
images of A3 sub-plug, compared to the best-fit curve and diffusion coefficient by solving the diffusion equation.
Fig. 7. 2D tomographic projection of the ion-milled face of the A2 5 mm diameter sub-plug, with 0.5 mm scale bar,
onto which the SEM-EDS map (4.80 mm × 2.08 mm) and the high resolution BSEM mosaic strip (3.66 mm x 0.23
mm) are overlain. The mineral color legend is slightly different from that in Fig. 12, with the muscovite class there
now subdivided into three subclasses.
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
16
Fig. 8. (a) Zoom-in of a subarea (156 µm × 212 µm, with 25 µm scale bar in (a)) of the high resolution BSEM strip
of A2 in Fig. 7, showing the corresponding registered subarea within the (a) Dry tomogram, (b) tomogram porosity
map (i.e. inversion of the difference tomogram CH2I2-saturated minus Dry), (c) tomogram OM volume fraction map
(i.e. inversion of the difference tomogram I2-stained minus Dry), and (d) Milled tomogram, compared to the (e)
BSEM image, and (f) SEM-EDS map (see the Fig. 7 legend).
(a)
(b)
(c)
(e)
(d)
(f)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
17
Fig. 9. (a) Zoom-in of a subarea (139 µm × 130 µm, with 25 µm scale bar in (a)) of the high resolution BSEM strip
of A2 in Fig. 7, showing the corresponding registered subarea within the (a) Dry tomogram, (b) tomogram porosity
map (i.e. inversion of the difference tomogram CH2I2-saturated minus Dry), (c) tomogram OM volume fraction map
(i.e. inversion of the difference tomogram I2-stained minus Dry), and (d) Milled tomogram, compared to the (e)
BSEM image, and (f) SEM-EDS map (see the Fig. 7 legend).
(a)
(b)
(c)
(e)
(d)
(f)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
18
Fig. 10. Visualizations of the FIBSEM-imaged rectangular prism (19.1 × 12.6 × 2.7 µm3), (a) cut-away, (b)
segmented OM (in red) within the left half, and (c) slice with segmented OM, pore (black) and pore backs showing
through (white).
(a)
(b)
(c)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
19
Fig. 11. (a) Subarea (4.7 µm × 1.6 µm) from the high resolution BSEM strip of ion-milled A2 in Fig. 7 with 10
nm/pixel, compared to FESEM images from broken surfaces of A2: (b) 200k micrograph with 0.7 nm/pixel, and (c)
×300 k micrograph with 0.5 nm/pixel, showing 100 nm scale bar for all.
(c)
(b)
(a)
SPWLA 57
th
Annual Logging Symposium, June 25-29, 2016
20
Fig. 12. Two subareas of an ion-milled surface of A2 for a sequence of 2D imaging modalities: (a) low energy
BSEM, uncoated, (b) low energy BSEM, uncoated after I2 staining, (c) high energy BSEM, coated after methanol
cleaning, (d) high energy SEM-EDS mineral map with color legend at right, (e) elemental carbon map from (d), and
(f) iodine map from (d). In (a), black scale bars are 10 µm, and red arrows point to kerogen particles.
(a)
(b)
(c)
(d)
(e)
(f)