GEOPHYSICS. VOL. 76, NO. 6 (NOVEMBER-DECEMBER 2011); P. E175–E186, 12 FIGS., 1 TABLE.

10.1190/GEO2010-0381.1
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Elastic and resistivity anisotropy of shale during compaction and

diagenesis: Joint effective medium modeling and field observations

Ran Bachrach1

ture, shale models frequently have many parameters that have spe-
ABSTRACT cific physical meaning, and are often difficult to calibrate.
Published laboratory measurements on clay properties and shales
Shales are anisotropic materials which have been ob- (Katahara, 1996; Vernik and Landis, 1996; Wang, 2002; Vanorio
served to have both electrical and elastic anisotropy. Shales et al., 2003) show large variation in the elastic properties of clay
also comprise most of the sedimentary column in clastic minerals and shales. Unlike sandstones, in which quartz elastic
basins, and thus, the ability to model shale response to elec- properties are well measured, most information published on clay
trical and seismic fields may improve our ability to better mineral stiffness is inferred from extrapolation of other measure-
resolve anisotropy by using different measurements. Com- ments and very few values have been measured directly. Also, be-
pacting shale-effective media modeling, which captures the cause water can be part of the mineralogical structure of clays, the
geological process of porosity reduction and geochemical relations between laboratory measurements on dry samples and
compaction, was formulated in this research in microstruc- natural shales as encountered in deep-water exploration are not well
tural parameters associated with particle alignment and pore- understood. Thus, effective media theories that have been success-
shape deformation. The state of compaction, captured by the fully tested for other materials, such as sandstones and carbonates,
porosity of the soft sediment, governs the pore aspect ratio are not widely used in exploration and development settings. The
and the amount of particle alignment. Diagenetic processes reason is partly because for a given theoretical shale model, the
are captured using temperature-dependent mineral elasticity parameters are often poorly constrained.
and conductivity. The model uses the same microstructural Recent advances in electromagnetic (EM) logging technology
parameters and equivalent effective medium approximation enabled not only measurement of the isotropic quantities associated
scheme to predict electrical and elastic anisotropy. In situ
with propagation of EM waves in subsurface formations, but also to
measurements of electrical anisotropy and seismic based es-
invert for the full electrical conductivity tensor. As the conductivity
timates of elastic anisotropy in shales in the same basin com-
in shale also depends on the microstructure of the sediments, joint
pare favorably with modeling results. The joint-modeling
analysis of both resistivity and elastic anisotropy may provide
formulation provides ways to study electrical resistivity
further insight to the rock-physics model and provide a new way to
and elastic-anisotropy jointly and explore the use of electri-
calibrate elastic models. Recently, some investigators (Brevik et al.,
cal resistivity anisotropy measurements to constrain elastic
2009; Dræge, 2010) attempted to model compaction with isotropic
anisotropy, and vice versa.
resistivity by using Archie’s law. Anisotropic effective media mod-
eling can help extending EM interpretation results to seismic and
vice versa.
Joint modeling of effective medium elastic and electrical ani-
INTRODUCTION sotropy was published in the material science community for
conductive material with electrical anisotropy (Sevostianov and
Seismic anisotropy of shales has been studied extensively during Kachanov, 2001, 2004) and were applied to geophysical applica-
the last few decades (Hornby et al., 1994; Sayers, 1994, 2005; tions by Kazatchenko et al. (2004, 2006).
Johansen et al., 2004; Dræge et al., 2006, Bandyopadhyay, In this article, I use the shale compaction and diagenesis to
2009). Because shales are complex in their fabric and microstruc- predict the anisotropy trends expected at different depths, rather

Manuscript received by the Editor 25 November 2010; revised manuscript received 9 November 2010; published online 22 December 2011.
1
Tel Aviv University, Department of Geophysics and Planetary Sciences, Tel Aviv, Israel; Schlumberger WesternGeco, Houston Texas, USA. E-mail:
[email protected]; [email protected].
© 2011 Society of Exploration Geophysicists. All rights reserved.

E175
 E176 Bachrach

than describing the response of a rock sample at a given fixed of simple compaction parameters and extend it to joint effective
porosity. I present a joint EM-elasticity formulation for effective medium modeling of elastic and electrical properties.
media approximation of both shale elastic and electrical anisotropy
for compacting shale that undergoes diagenesis. The model is based Compaction parameters
on the same microstructural parameters and provides an insight into
the relationship between electrical and elastic anisotropy in com- I define compaction as general reduction of porosity due to
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

pacting shale. The relations developed here can be used in various various, often complex, burial and diagenetic mechanisms (e.g.,
applications including joint interpretation of well logs, relating overburden stress, cementation effects, and various pore-pressure-
electrical anisotropy to seismic anisotropy, and joint EM-seismic related mechanisms).
inversion. The key factor in this modeling approach for shale is that the
porosity is the main controlling variable and can be measured
and evaluated from well logs and/or calibrated velocity/density
COMPACTING SHALE MODEL
transformation. Although, in general, a given porosity is not suffi-
Figure 1 (adopted from Avseth et al., 2003) represents the cient to determine the compaction process that is associated to the
conceptual compacting shale modeling framework. stress history, it is possible to define common compaction scenarios
The compacting clay model is based on the concept that, for com- for which porosity and other model parameters can be used to
pacting shale, the porosity reduction process controls the two main estimate microstructure parameters.
aspects described in the previous sections: (1) the general inclusion Representative pore-shape aspect ratio is one of the key param-
shape (or aspect ratio) and (2) the grain orientation function. A third eters associated with compaction. Based on the compaction scheme
important process is the mineralogy transition associated with the presented in Figure 1, it is reasonable to represent the general de-
temperature profile in the subsurface, which controls the smectite- pendency of the aspect ratio as a function of porosity. I define the
to-illite (S/I) transition, a process that is called geochemical com- relation between porosity and aspect ratio α as
paction (Storvoll and Brevik, 2008) and has been observed by
many authors in various sedimentary basins such as the Gulf of α ¼ ðϕ∕ϕ0 Þk ; (1)
Mexico (GoM) and the North Sea (Freed and Peacor, 1989;
where ϕ is the current porosity, ϕ0 is the critical porosity and k is
Peltonen et al., 2009). For a given compaction (porosity), the model
a model parameter. This relation states that, at zero porosity, the
uses ideas similar to those presented by Hornby et al. (1994) and
aspect ratio is zero, whereas at initial deposition when the porosity
Sayers (1995), which uses the orientation distribution function
is equal to the critical porosity ϕ0, the aspect ratio is one that sug-
and the anisotropic differential effective media theory to charac-
gests that the pore space has no preferential direction (on average).
terize the state of the sediment and its anisotropy for a given
The model parameter k governs the rate of pore space deformation.
porosity. However, I am interested in modeling the process of com-
I note that equation 1 is similar to the “compaction factor”
paction. Thus, I extend the theory given for a static process and
presented by Bandyopadhyay (2009). However, the compaction
apply it into the state of the sediment at any state of porosity
model in equation 1 has the flexibility to address different scenarios
reduction.
To model compaction, Johansen et al. (2004) introduces the con- of compaction as will be discussed below.
cept of compaction effects on shale fabric through modeling I also note that a fundamental parameter in the compacting shale
the orientation distribution function. Dræge et al. (2006) models model is the critical porosity (Nur et al., 1998), which is the point
the elastic response of compacting shale including diagenetical where the suspension of clay particles touches each other and a
effects. Bandyopadhyay (2009) also considers the effect of compac- phase transition occurs where shear modulus becomes finite. Cri-
tion on the orientation distribution function and was able to model tical porosity appears in the model in few places both for the state
some of the experimental laboratory measurements using compac- of alignment of particle and in the actual effective media calcula-
tion concepts. In this work, I revisit some of the concepts presented tions, as will be also discussed below.
by the authors mentioned here and reformulate the problem in terms Similarly, to the pore aspect ratio, the grain orientation
distribution function in terms of the state of compaction can be
related through the following relation (C. M. Sayers, private
communication)

W 200 ðϕÞ ¼ W max

200 ð1 − ϕ∕ϕ0 Þ
n200 ;

W 400 ðϕÞ ¼ W max

400 ð1 − ϕ∕ϕ0 Þ
m400
; (2)

where W 200 ðϕÞ, W 400 ðϕÞ are vertical transverse isotropic (VTI)
orientation distribution function which will be discussed later in the
paper; m400 , n200 are mode parameters, which will be also discussed
below. The above relations state that, whereas at deposition there is
no particular orientation, the parallel alignment of the particles will
occur at zero porosity. I note that equation 2 states that both second-
and fourth-order moments associated with the orientation distribu-
tion function are related to the state of compaction but the rate at
Figure 1. Conceptual compacting shale modeling (after Avseth which the alignment moments change with respect to each other is
et al., 2003). independent (i.e., different exponents are used). I will show below
 Elastic and EM anisotropy of shales E177

that this is a generalization of concepts introduced by Johansen

(2004) and Bandyopadhyay (2009).

Graphical and physical interpretation

of model parameters
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

In this model, the compaction is governed by three parameters:

m400 , n200 , and k. It is useful to associate a physical picture with the
model parameter.

Aspect ratio parameter k

Consider an ellipsoid pore shape with main axis a and c and
which is deforming from height a to a 0 (1D deformation) as pre-
sented in Figure 2. Recall that the aspect ratio in our case is defined
as α ¼ a∕c ¼ ðϕ∕ϕ0 Þk .
Figure 2. Graphical interpretation of aspect ratio exponent. See text
The pore volume change is given by a∕a 0 ¼ ðϕ∕ϕ0 Þ. for details.
For the case k ¼ 1 suggests that the aspect ratio deforms exactly
at the rate of porosity reduction. This is 1D consolidation. For the
case k > 1 suggests that aspect ratio is reduced faster than the rate of
porosity reduction k < 1 suggests that aspect ratio is reduced slower
than the rate of porosity reduction. All scenarios are shown in
Figure 2.

Orientation distribution function parameters m400 and n200

Orientation distribution functions have been reviewed exten-

sively in the literature. Detailed discussions can be found in Sayers
(1995), Johansen et al. (2004) and in Bandyopadhyay (2009). Here,
orientation distribution functions are briefly reviewed.
An orientation distribution function gives the probability density
for a particular orientation and must satisfy the normalization
condition ∫ WðΩÞdΩ ¼ 1 where Ω represents all three Eulerian
angles. The orientation distribution function satisfies the normal-
ization criteria 4π1 2 ∫ 1−1 WðξÞdξ ¼ 1, ξ ¼ cosðθÞ, and perfect align- Figure 3. Plot of normalized compacting orientation distribution
ment is given by WðξÞ ¼ δðξ − 1Þ, whereas no alignment at all function (WðξÞ∕A2 ) to demonstrate change in aspect ratio for com-
pacting orientation distribution function.
is given by WðξÞ ¼ 1∕8π 2. The two moments associated
with the VTI orientation distribution functions are given by
qffiffi
W 200 ¼ 52∫ 1−1 WðξÞP2 ðξÞdξ, where P2 ðξÞ ¼ 12 ð3ξ2 − 1Þ, the
Legendre polynomial of the second order, and
qffiffi
W 400 ¼ 2∫ −1 WðξÞP4 ðξÞdξ, where P4 ðξÞ ¼ 8 ð35ξ − 30ξ þ 3Þ,
9 1 1 4 2

the Legendre polynomial of the fourth order.

For the particular shape of the compacting orientation distribu-
tion function, Johansen et al. (2004) suggested that
1 A2
WðξÞ ¼ ; (3)
4π ðξ2 þ A2 ð1 − ξ2 ÞÞ3∕2
where A can be interpreted as the aspect ratio of the strain ellipse,
or in pore-shape aspect ratio as A ∝ 1∕α. In Figure 3, I plot the nor-
malized compacting orientation distribution function (WðξÞ∕A2 )
to demonstrate the effect of change in aspect ratio for compacting
orientation distribution function.
It is interesting to note that the compacting orientation distribu-
tion function or other statistical orientation distribution function
forces a certain relationship between the pore shape and the
alignment. In Figure 4, I show that the compacting orientation Figure 4. Plot of W 200 versus W 400 for different model parameter
distribution function (equation 3) is a special case of the general exponents and comparison to compaction orientation distribution
parameterization in the compacting shale model formulated in this function of Johansen et al. (2004).
 E178 Bachrach

article, and is equivalent for the special case where m400 ¼ 2 and rich mineral to illite-rich sediment as a function of the temperature.
n200 ¼ 1. However, additional relations can be modeled by the I choose the following equation to describe the fraction of illite
parameterization presented here. This is a major difference between PS ðTÞ as a function of temperature T:
this model and the one suggested by Johansen et al. (2004) and in  
1 1 T − TT
Bandyopadhyay (2009). Its impact on calibration of model param- PS ðTÞ ¼ þ tanh ; (4)
eters by using borehole seismic data will be demonstrated below. 2 2 2sT
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

As noted by Sayers (1994, 1995), the ratio of anellipticity, Here, T T is the temperature where 50% of the S/I reaction has ac-
which quantifies the deviation of the P-wave slowness curve from crued, and sT controls the spread of the reaction. This functional
an ellipse and is given by A ¼ ðC11 − C55 ÞðC33 − C55 Þ− relation can be calibrated to field observations. I note that equation 4
ðC13 þ C55 Þ2 where Cij is the 6 × 6 stiffness TI matrix, assuming is a phenomenological equation which assumes that the S/I transi-
Voigt notation is governed by the ratio of W 400 ðϕÞ∕W 200 ðϕÞ. In this tion is smooth. This is supported by field observations in the well.
model, this ratio is given by the expression W 400 ðϕÞ∕W 200 ðϕÞ ¼ Figure 5 shows an example of calibrating equation 4 to observed S/I
p3ffiffi ð1 − ϕ∕ϕ0 Þm400 −n200 . Thus, the parameters m400 and n200 are also ratios reported by Freed and Peacor (1989). In the model, the geo-
5
controlling the anellipticity of the system. When m400 ¼ n200 , the chemical compaction is captured by changing the mineral elastic
ratio is constant for all porosities. When n200 ≪ m400 , W 400 ðϕÞ ≪ stiffness tensor from smectite-rich to illite-rich sediments. Reuss an-
W 200 ðϕÞ and the anellipticity may be neglected (Sayers, 1994). isotropic average of the elastic tensors according to the temperature-
Thus, knowledge of anellipticity (which can be estimated using sur- dependent ratio defined by equation 4 is used to achieve represen-
face and borehole seismic data) can be used for model calibration. tative moduli for the transition.
One of the challenges associated with modeling S/I transition is
Geochemical compaction to assign material properties to smectite-rich or illite-rich sediment
and to properly capture the transition. In most places XRD analysis
I model the geochemical compaction (S/I transition) using a
shows that the actual shales are made from different minerals (e.g.,
logistic function that represents the transition between smectite-
Peltonen et al., 2009). Also the data from Freed and Peacor (1989)
show that the smectites and illites compose only 80% of the
mineral matrix. Thus, when we describe the material properties
for anisotropic model building we often refer to laboratory estimates
on shales which are actual averages of many minerals. In the Gulf
of Mexico, Freed and Peacor (1989) observe about 80% of the
rock composed of smectite which transform to about 80% illite-rich
sediments. Peltonen et al. (2009) shows examples from the North
Sea where smectite-rich shale has only about 35% smectite in it and
illite-rich shale has about 30% to 40% illite.
In Table 1, I present a summary of published typical modulus for
shales and shale minerals. In the literature, typically isotropic elastic
constant for clay mineral are given, and often mineral modulus are
derived by extrapolation of observation of shale to the zero porosity
point. For example, Mavko et al. (1998) report mineral modulus
for so called “gulf clays” to be in the range of 21 to 25 GPa
for bulk modulus and 7 to 9 GPa for shear modulus. Similar values
have been also measured in the laboratory using pure minerals
(Vanorio et al., 2003). Anisotropic measurements in shales
(e.g., Greenhorn Shale, Jones and Wang, 1981) show a similar range
Figure 5. Observed percent illite in wells and logistic fit to obser- of stiffness. Some of the published anisotropic data are within the
vation. Data from Freed and Peacor (1989).
range of modulus inferred for pure clays in the Gulf of Mexico.

Table 1. Properties of the Greenhorn (G) and Green River (GR) shales used as building blocks for smectite-rich sediments
(Table from Johansen et al., 2004). Illite properties are given by Wenk et al. (2007) and by Voltolini et al. (2009). Quartz
properties are from Mavko et al., (1998) and illite-rich mix represent an anisotropic mixture of 65% illite with 20%
softer clays (G) and 15% quartz. These minerals can be used as building blocks for the compacting shale model.
Material ρ (Kg∕m3 ) C11 (GPa) C13 (GPa) C33 (GPa) C44 (GPa) C66 (GPa) ε δ γ

G 2400 34.3 10.7 22.7 5.4 10.6 0.255 −0.05 0.48

GR 2075 31.3 3.4 22.5 5.4 6.5 0.2 −0.28 0.1
Illite 2007 2600 178 25.6 58.6 16.5 72 1.02 0 1.68
Illite 2009 2600 119 16.4 66.7 32.7 48.2 0.392 0.277 0.237
Silt (quartz) 2650 95.67 7.67 95.67 44 44 0 0 0
Illite-rich mix 2600 75.8 15.8 48.5 16.5 27.9 0.28 0.01 0.34
 Elastic and EM anisotropy of shales E179

Therefore, in this study I use the anisotropic measurement of Green- is modeled as a suspension of brine and clay particle. A Reuss aver-
horn Shale as reasonable anisotropic values for smectite-rich clay age is the exact solution for such mixture. Conventional DEM will
minerals. require the usage of shale and pure brine as pure end members when
In Table 1 I also list laboratory measurements anisotropic mineral the concentration of the phases vary from 0 to 1. However, because
stiffness estimates provided by Voltolini et al. (2009) and Wenk we are not interested in the case where the porosity is higher than
et al. (2007). I note that there is a very large difference in stiffness the critical porosity (a case we have identified as suspension), using
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

between the smectite-rich material and illite. a modified DEM with a critical porosity as the end-member pro-
Smectite-to-illite transition releaces both water and silica (note vides a realizable mixing law that is consistent at every state of com-
that quartz can cement the shale minerals in the form of microcrys- paction. It also ensures the model will be consistent with clay
talline quartz cement [Thyberg et al., 2009]). From an effective particles being initially deposited as particles in suspension when
medial modeling perspective, adding quartz to the effective medium porosity is at critical porosity.
building blocks can be done using volumetric averaging such as the
Reuss-Voigt-Hill average. Because quartz stiffness is high in com- Electrical and elastic DEM formulation
parison to the smectite-rich minerals (and even higher than C33 of
illite, as shown in Table 1), quartz cemented shales are expected to The DEM formulation for electrical anisotropy provides an
be stiffer as well. Quartz cement in the matrix will also lower the expression for a differential increment in electrical conductivity
building blocks anisotropy. In all cases, the geochemical alteration associated with ellipsoid inclusion and is given as (Sen et al.,
results in generally much stiffer sediment that is taken into account 1981; Dürr et al., 2002)
by the compaction shale model. In Table 1, I also list the anisotropic
illite-rich sediments that represent a mixture of 65% illite with 20% dνi
GoM shales and 15% Quartz. dσDEM ðν; αÞ ¼ ðσ − σDEM Þ×
1 − νi i
To summarize, the equations above provide us with a micro-
−1
structural characterization of the compacting shale that is governed ½I þ σ−1∕2 −1∕2
DEM ðν; αÞΛi ðαÞσDEM ðν; αÞðσi − σDEM ðν; αÞÞ
; (5)
by the pore-shape aspect ratio, the degree of particle alignment,
and with a control over the mineral moduli that is associated with
where σi is the electrical conductivity of the ellipsoidal inclusions
S/I transition. The two dominant factors needed to determine this
with aspect ratio α, and νi is the concentration of the ith
microstructural description are the porosity and temperature. The
role of the critical porosity is to define the state of compaction phase. The matrix Λi is related to the ith phase through the depo-
which is defined by ϕ∕ϕ0. Because the above relations are defining larization factors. For more details, I refer the reader to Dürr
microstructural parameters, they can be used for both elasticity and et al. (2002).
resistivity modeling, as will be shown in the following section. The equivalent elastic DEM equation for the elastic stiffness
tensor C (Hornby et al., 1994) is given by

JOINT EFFECTIVE MEDIUM APPROXIMATION

dνi
FOR COMPACTING SHALE ELASTIC AND dCDEM ðν; αÞ ¼ ðCi − CDEM ðν; αÞÞQi ðαÞ; (6)
ELECTRICAL ANISOTROPY 1 − νi

A recipe for elastic vertical transverse isotropy VTI effective

medium approximation for compacting shale is based on the where Ci is the anisotropic inclusions elastic tensor and Qi is a
Hornby et al. (1994) anisotropic effective medium formulation that shape factor that relates to the Eshelby tensor and are reviewed
models the shale using the anisotropic differential effective media by Hornby et al. (1994) and Jakobsen et al. (2000).
(DEM) scheme by first homogenizing the shale matrix and then
adding silt/sand as spherical inclusions. The shale is modeled first Particle alignment
using DEM and then particle alignment is addressed by averaging Morris (1969) presents an averaging procedure for fourth rank
over orientation distribution functions. The microstructural para- tensors that enables us to account for different particle orientations.
meters used for elastic modeling of shales are pore aspect ratio Because I would like to maintain mathematical consistency between
α and orientation distribution function parameters W 200 and W 400 .
the averaging of the fourth rank tensors for elasticity and the second
Some of the differences between the current scheme and the one
rank conductivity tensor, I follow the well-known approach that re-
presented by Hornby et al. (1994) and adopted by other investiga-
presents the second rank conductivity tensor in fourth rank notation.
tors in later studies are that the current scheme makes extensive use
Following Milton (2002), the constitutive relation between n
of a modified DEM (Mukerji et al., 1995) for both electrical
divergence-free vector fields J and n curl-free vector fields E
and elastic parameter estimation. In the following section, I will
through linear constitutive relation can be defined as
address some of the benefits of the scheme when modeling
composites.
J iq ðxÞ ¼ Siqjβ E jβ ; (7)
Critical porosity and modified DEM scheme
When performing effective media calculations for shale, it is where S is the material property tensor. Note that, here i and j are
useful to define the critical porosity explicitly as it is used for mod- space indices and q and β are field indices.
eling the state of compaction and alignment already. At the Then, the current electrical field constitutive relationship can be
critical porosity, the effective modulus of the clay-brine mixture written by using 6 × 6 matrix notation as
 E180 Bachrach

0 1 2 30 1
J1 σ 11 σ 12 σ 13 0 0 0 E1 orientation distribution function coefficient to average the clay
B J 2 C 6 σ 21 σ 22 σ 23 0 0 07 B E2 C particles’ effective electrical conductivity. In this case, following
B C 6 7B C
B J 3 C 6 σ 31 σ 32 σ 33 0 0 07 B C Sayers (1987), I obtain
B C¼6 7B E 3 C; (8)
B0C 6 0 0 0 0 0 0 7B
7 C
B C 6 B 0 C
@0A 4 0 0 0 0 0 0 5@ 0 A hSij i ¼ S 0ij þ Δij ðW 200 ðϕÞ; W 400 ðϕÞÞ; (9)
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

0 0 0 0 0 0 0 0

where σ ij is the second-rank electrical conductivity tensor. where the exact expressions for S 0ij and Δij are given explicitly
A general averaging of fourth rank tensors Akl (in Voigt notation) by Sayers (1987). In appendix A, I list the details associated with
was given by Morris (1969): equation 9. Note that, when averaging the conductivity tensor, I use
only expressions involving W 200 ðϕÞ because second-rank tensors,
hAij i ¼ T̄ ijkl Akl ; in general, can be fully expanded by orientation distribution
functions to the second order. For more details, I refer the reader
where T̄ ijkl defines the weight function tensor associated with the to Raymond et al. (1996).
fabric orientation. The representation of the conductivity tensor as The above equations can be used for both elastic and electrical
a fourth-rank tensor in Voigt notation enables us to use the same conductivity tensors associated with the Voigt (6 × 6) representa-
tion. The basic procedure I follow is

1) use anisotropic DEM scheme for a given porosity

2) apply fabric (orientation) correction to model sediment fabric

NUMERICAL RESULTS
In Figure 6, I first present the modeling of Greenhorn Shale
parameters by using the compacting shale model without modeling
the S/I transition. For reference, I also plot here the empirical C33
and C44 that are derived from Gardner’s (1974) velocity density
relationship and the Mudrock line (Castangna, 1993) by using equa-
tions given in Appendix B. I note that the Gardner-derived elastic
modulus can serve as simple check to determine if our modeling
results are reasonable as they have been derived empirically from
many observations in the GoM. Thus, a reasonably effective media
modeling for shales should be able to reproduce the basic empirical
relation widely used in many applications.
From close examination of Figure 6, one can clearly see that,
whereas the Gardner-derived C33 and mudrock-derived C44
can fit the model at high porosity, they clearly overpredicted the
modeled shale anisotropy at low porosity. This is easy to understand
if one looks at Table 1, because the Greenhorn Shale has mineral
modulus C33 of 22.7 GPa. In Figure 7, I plot the same data, but this
time, I model the mineral transition between smectite to illite, ac-
cording to equation 8. The temperature dependency was mapped
into the porosity range assuming a typical compaction curve for
the GoM (i.e., porosity reduction versus depth) and a simple linear
temperature gradient. It can be clearly seen that the model can
reproduce Gardner’s relations at all porosities reasonably well. I
note that, from this simple exercise, it is clear that the low-porosity
part associated with Gardner’s moduli-porosity curve must have
been calibrated on stiff sediment such as illite-rich sediment.
In Figure 8, I present the results of the joint effective medium
modeling of shale electrical anisotropy. The shale-brine mixture
uses a brine conductivity of 5000 mS/m and shale conductivity is
anisotropic with 400 mS/m horizontal conductivity and 100 mS/m
vertical conductivity. In this case, I did not have any laboratory mea-
Figure 6. Results of modeling compacting shale using Greenhorn surements associated with illite-rich mineral, and thus, I assumed
mineral values and not accounting for S/I transition. (a) Anisotropic
elastic stiffness. (b) Thomsen’s parameters. Note that the model that the electrical conductivity of illite is equal to that of smectite.
underpredicts Gardner’s C33 and C44 at low porosities. Here I use However, the model can accommodate different conductivities
m400 ¼ 2 and n200 ¼ 1 and k ¼ 0.25. using the same logistic transition function (equation 4). In this case
 Elastic and EM anisotropy of shales E181

the averaging of mineral resistivity is the Reuss average for 2nd (estimated by the petrophysical analysis) and extrapolate the values
rank tensors. using the trend in the borehole data. I demonstrate this method using
as an example the dual-water equation (Clavier and Coates, 1984)
Accounting for bounded water in elastic often used in petrophysical analysis. I note that here I will need to
and electrical DEM modeling be consistent with petrophysical concepts as explained below.
I show that the conventional petrophysical analysis can be inter-
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

When trying to estimate in situ mineral conductivity to better preted as the horizontal conductivity of the shales. No in situ
constrain model parameters, I face a shortage of laboratory mea- equivalent for vertical conductivity of the shale end-member can
surements associated with shale minerals and specifically with be estimated from conventional petrophysical analysis as will be
illite-rich clay. One way to overcome this lack of data is to use in discussed here.
situ estimates of electrical conductivity at a given porosity Bounded water plays an important role in the petrophysical ana-
lysis of shale conductivity. The dual-water equation mentioned
above can be considered an empirical extension to Archie’s law
for clays with bounded-water and is given by Clavier and Coates
(1984) as

    
1 a na Sbw S
σ t ¼ ϕm Sw 1− σ w þ bw σ bw ; (10)
a t Sw Sw

where σ t is the total conductivity (which can be measured), σ bw is

the bounded-water conductivity, Sbw ¼ αir V clay ∕ϕt is the saturation
of bounded water (i.e., how much bounded water is in the clay)
expressed in terms of irreducible clay porosity αir, volume fraction
of clay V clay, and total porosity ϕt, and is often calibrated during
petrophysical interpretation (Clavier and Coates, 1984). Sw is the
saturation of water in the formation; ma and na are Archie’s expo-
nents for dual-water clays. It is easy to see that the bounded-
water equation becomes the classical Archie’s equation for the
case of Sbw ¼ 0. Also, if the clay volume fraction (V clay ) is
100%, αir ¼ Sbw ϕt .

Figure 7. Results of modeling compacting shale accounting for S/I Figure 8. Results from compacting shale modeling for electrical
transition. Greenhorn mineral values are assumed to be the smec- conductivity anisotropy (in terms of horizontal [Sh] and vertical
tite-rich building and illite (Voltolini et al., 2009) are used as the [Sv] conductivities) derived with the same microstructural param-
illite-rich mineral building blocks. (a) Anisotropic elastic stiffness. eters as in Figure 4. Note that I present two versions of electrical
(b) Thomsen’s parameters. Note that the model underpredicts Gard- conductivities: one with mineral blocks derived by accounting for
ner’s C33 and C44 at low porosities. Here, I use m400 ¼ 2 and bounded-water in clay mineral conductivity and one using default
n200 ¼ 1, and k ¼ 0.25. Note Gardner’s C33 and C44 are now better values of 5 S/m and 0.4 S/m for horizontal and vertical mineral con-
predicted along the compaction path at all porosities. ductivities.
 E182 Bachrach

Effective media theory such as DEM provides a “mixing” or

homogenization recipe to estimate the effective conductivity of
the material made out of clay mineral and brine. As mentioned
earlier, one needs to know the two phases being mixed. I define
these as “end members” of the mixture. A new end-member for
clays with bounded-water can be estimated from the dual-water
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

equation and used within the modified DEM model as ex-

plained below.
Bounded-water are often considered in petrophysics to be part
of the sediment matrix. Typical values used in petrophysical analysis
are 16% for illite-rich minerals and 24% for smectite. When modeling
end members with bounded water we can consider the total porosity
of the end-member to be equal to Sbw . I then calculate the total con-
ductivity σ t using equation 9. This conductivity is the conductivity at
total porosity of the bounded-water. The modified DEM scheme uses
Figure 9. Comparison between conventional anisotropic electrical this conductivity as the “end-member” conductivity for ϕt ¼ Sbw.
conductivity modeling (presented in Figure 5) and modeling assum- Dual-water equation is based on modification of Archie’s law as de-
ing 16% of bounded water in the formation. scribed earlier. Because Archie’s law assumes the brine in the pore

Figure 10. VSP-based anisotropy estimates in a GoM basin (blue) and model prediction (green). Rock-physics-based predictions are within
the range of VSP observations.
 Elastic and EM anisotropy of shales E183

space is the connected phase (e.g., Guéguen, 1994) the electrical con- The check shot data are being used to map velocity to porosity. The
ductivity estimated using this method is the connected conductivity, anisotropic elastic model parameters are chosen to fit a single well.
which for VTI symmetry is the horizontal conductivity (i.e., the Based on this single well calibration I predict anisotropy in terms
connected phase assuming horizontal alignment for the mineral with of Thomsen’s ε, δ. Note that the rock model predictions in the
bounded water only). The vertical conductivity is not obtained by this four wells shown in Figure 4 are in agreement with the checkshot
method and must be assumed. estimates of the anisotropy parameters.
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure 9 gives a comparison of conventional electrical effective Figure 11a presents a short interval of triaxial induction borehole
media approximation for anisotropic compacting clays with one tool (RT scanner™) at a depth of ∼20; 000 ft (6000 m). The shale
that used estimated conductivity of the building block with 16% units have been logged with the RT scanner tool that provides the
bound water with the modified DEM scheme that accounts better electrical anisotropy. Unlike the case of VSP data, here I had only a
for bounded water in the clay matrix. In both cases, the mineral short interval of anisotropic conductivity estimates, so modeling the
point vertical to horizontal conductivity is assumed to be fixed. whole compaction is not possible. Figure 11b gives the 2D histo-
gram of all porosity values and the ratio between horizontal resis-
Model predictions and field observations tivity (Rh ) and vertical resistivity (Rv ). Within the short log interval,
it can be seen that the most probable porosity as measured by the
Figure 10 shows the results of applying the anisotropic compact-
histogram is a porosity of 24% and an the most probable Rh ∕Rv
ing shale model to several wells with where anisotropy have been ratio of 0.42. I estimate the electrical anisotropy in the shale for five
estimated independently using analysis of both surface seismic data GoM wells and plot the model prediction, the observed maximum
and checkshot. The wells are located near the observed anisotropy. likelihood porosity, and the Rh ∕Rv ratio on the plot (Figure 12).

Figure 11. (a) Gulf of Mexico (GoM) log data

with horizontal and vertical resistivity. (b) Two-
dimensional histogram of shale porosity and
Rh ∕Rv ratio associated with the GoM well log.
 E184 Bachrach

effective electrical anisotropy for shales. It is important to note that

the model is very flexible in terms of its ability to account for dif-
ferent values of anellipticity. The universality of the fabrics (or mod-
el parameters) is still an open research question. The initial tests we
performed on the GoM check shot data shows that in a given basin
we are able to predict the anisotropy trend using a single set
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

of model parameters.
The electrical resistivity modeling results shows that the observed
anisotropy is in agreement with the model parameters in the basin.
However, I would like to note that additional laboratory data on the
link between digenesis and electrical resistivity will enable better
validation of this important link between electrical and seismic
anisotropy.
The model assumes that a basin is representative by a typical
compaction behavior and that the model parameters and mineralogy
do not vary along the sedimentary column. In some basins this
assumption may be violated. Quantitative analysis of such cases
will enable us to better justify the use of this model, but it is beyond
the scope of this paper.
Figure 12. Model prediction and observation from six different The model results can be used to improve seismic anisotropy
anisotropic resistivity logs in the GoM. I present the model predic- estimates by constraining them to observed electrical anisotropy
tion using mineral end members and the dual-water equation that
handle the bounded-water structure of the clay mineral and contri- in well logs or perhaps in the future with surface-based EM
bute to the electrical anisotropy. inversion.

ACKNOWLEDGMENTS
As can be seen from the scatter in the data, for the porosity values
I thank Richa Richa from Stanford Rock-Physics project and
covered by the anisotropic logging tool, the electrical resistivity
modeling results are within the overall range of observed horizontal Alexander Kostin of Schlumberger DCS for assistance in calculat-
to vertical conductivity ratio. As the wells are within the same ing in situ mineral conductivity using extrapolation of RT scanner
region of the VSP data observation, I used the same microstructural observations. I also acknowledge Colin Sayers for many useful
parameters to predict electrical anisotropy. discussions and suggesting the functional dependency of the orien-
tation distribution function. Finally I would like to thank anon-
ymous reviewers whose comments helped in improving the clarity
DISCUSSION AND CONCLUSIONS
of this manuscript.
I presented a compacting shale model that can predict both elastic
and electrical anisotropy using the same conceptual compaction APPENDIX A
process and same microstructural parameters. The two dominant DETAILED FORMULAS FOR
factors controlling the fabric of the sediments are the porosity ANISOTROPIC AVERAGING
and temperature. The state of compactions defined by the ratio of
porosity to critical porosity (ϕ∕ϕ0 ). Although critical porosity for The Reuss average results for VTI material are given by Sayers
shales is a difficult parameters that may range anywhere between (1987); below is a summary of the results.
0.6–0.85, the state of compaction is a parameters associated with
the ratio of current porosity to initial depositional porosity. The
hSij i ¼ S 0ij þ Δij ; (A-1)
model uses microstructural parameters to define the topology of
the porous media. Therefore, same topology can be used for both
elasticity and resistivity modeling.
Additional parameters to the model are the end members stiffness 1
(or resistivity) for a given mineral composition. It is not easy to find S 011 ¼ S 022 ¼ S 033 ¼ ð8sa þ 3sa33 þ 4sa13 þ 2sa44 Þ;
15 11
a set of material mineral properties that can be considered “univer-
1
sal” in terms of their intrinsic elastic and electrical anisotropy. Often S 012 ¼ S 023 ¼ S 031 ¼ ðsa11 þ sa33 þ 5sa12 þ 8sa13 − sa44 Þ;
one needs to look at averaging properties of different minerals. 15
XRD measurements of mineral content in shales can help in such 2
S 044 ¼ S 055 ¼ S 066 ¼ ð7sa11 þ 2sa33 − 5sa12 − 4sa13 þ 3sa44 Þ;
calibration. 15
The compacting shale model takes into account both mechanical (A-2)
compaction and geochemical compaction. Comparison of model
predictions to field observations (in few different wells in a GoM
basin) shows that, in general, the model is in agreement with ob- where the component of the compliance matrix of the building
served anisotropy values and that model parameters calibrated to blocks is given by ðsaij Þ in Voigt notation, and the perturbation is
four VSP checkshots and seismic data can be used to predict given by (Sayers, 1987):
 Elastic and EM anisotropy of shales E185

 pffiffiffiffiffi pffiffiffi 
8 10 4 2
Δ11 ¼ π 2 a3 W 200 þ a1 W 400 ; C33 ¼ ρb V 2P ; C44 ¼ ρb V 2S ; (B-3)
105 35
 pffiffiffiffiffi pffiffiffi 
2 8 10 4 2
Δ22 ¼π a3 W 200 þ a1 W 400 ; where V S is estimated using the popular mudrock line (Castagna,
105 35
pffiffiffi   et al., 1985).
p ffiffi
ffi
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

16 2 2
Δ33 ¼ π − 5a3 W 200 þ 2a1 W 400 ;
105 REFERENCES
pffiffiffiffiffi pffiffiffi
8 10 2 4 2 2
Δ12 ¼− π ð7a2 − a3 ÞW 200 þ π a1 W 400 ; Avseth, P., H. Flesche, and A. van Wijngaarden, 2003, AVO classification of
lithology and pore fluids constrained by depth trends: The Leading Edge,
315 105
22, 1004–1011, doi: 10.1190/1.1623641.
Δ13 ¼ Δ23 Bandyopadhyay, K., 2009, Seismic anisotropy: geological causes and its
pffiffiffiffiffi pffiffiffi implications to reservoir geophysics. Ph.D. thesis, Stanford University.
4 10 2 16 2 2 Brevik, I., P. T. Gabrielsen, and J. P. Morten, 2009, The role of EM
¼ π ð7a2 − a3 ÞW 200 − π a1 W 400 ; rock physics and seismic data in integrated 3D CSEM data analysis:
315 105 79th Annual International Meeting, SEG, Expanded Abstracts, 835–838.
Castagna, J. P., 1993, AVO analysis — Tutorial and review, in
Δ44 ¼ Δ55 J. P. Castagna, and M. Backus, eds., Offset-dependent reflectivity —
pffiffiffiffiffi pffiffiffi Theory and practice of AVO analysis: SEG Investigations in Geophysics
8 10 2 64 2 2 8, 3–36.
¼− π ð7a2 þ 2a3 ÞW 200 − π a1 W 400 ; Clavier, C., G. Coates, and J. Dumanoir, 1984, Theoretical and experimental
315 105
pffiffiffiffiffi pffiffiffi bases for the dual-water model for interpretation of shaly sands: SPE
16 10 2 16 2 2 Journal, 24, 153–168.
Δ66 ¼ π ð7a2 þ 2a3 ÞW 200 þ π a1 W 400 ; (A-3) Dræge, A., 2010, Rock physics and resistivity depth trends, an integrated
315 105 approach: 70th Annual International Meeting, SEG, Expanded Abstracts,
711–715.
where Dræge, A., J. Morten, and T. A. Johansen, 2006, Rock physics modeling
of shale digenesis: Petroleum Geoscience, 12, 49–57, doi: 10.1144/
1354-079305-665.
a1 ¼ sa11 þ sa33 − 2sa13 − sa44 ; Dürr, O., W. Dieterich, P. Mass, and A. Nitzan, 2002, Effective medium
theory of conduction in stretched polymer electrolytes: Journal of
a2 ¼ sa11 − 3sa12 þ 2sa13 − sa44 ∕2; Physical Chemistry B, 106, 6149–6155, doi: 10.1021/jp013373a.
Freed, R. L., and D. R. Peacor, 1989, Variability in temperature of the smec-
a3 ¼ 4sa11 − 3sa33 − sa13 − sa44 ∕2: (A-4) tite/illite reaction in gulf coast sediments: Clay Minerals, 24, 171–180.
Gardner, G. H. F., L. W. Gardner, and A. R. Gregory, 1974, Formation
velocity and density — The diagnostic basics of stratigraphic traps:
Geophysics, 39, 770–780, doi: 10.1190/1.1440465.
Guéguen, Y., and V. Palciauskas, 1994, Introduction to the physics of rocks:
APPENDIX B Princeton University Press.
Hornby, B. E., L. M. Schwartz, and J. A. Hudson, 1994, Anisotropic effec-
tive-medium modeling of the elastic properties of shales: Geophysics, 59,
EQUATIONS USED TO MAP GARDNER’S 1570–1583, doi: 10.1190/1.1443546.
RELATIONS AND MUDROCK DATA INTO Jakobsen, M., J. A. Hudson, T. A. Minshull, and S. C. Singh, 2000, Elastic
properties of hydrate bearing sediments using effective medium theory,
EQUIVALENT POROSITY-STIFFNESS Journal of Geophysical Research Solid Earth: JGR, 105, 561–577, doi:
REPRESENTATION 10.1029/1999JB900190.
Johansen, T. A., B. O. Ruud, and M. Jakobsen, 2004, Effect of grain scale
The following equations are used to map Gardner relations to alignment on seismic anisotropy and reflectivity of shales: Geophysical
Prospecting, 52, 133–149, doi: 10.1046/j.1365-2478.2003.00405.x.
modulus-porosity response. Jones, L. E. A., and H. F. Wang, Ultrasonic velocities in Cretaceous shales
The Gardner equation is given by from the Williston basin. Geophysics, 46, 288–297, doi: 10.1190/
1.1441199.
Katahara, K. W., 1996, Clay mineral elastic properties: 66th Annual Inter-
b national Meeting, SEG, Expanded Abstracts, 1691–1694.
ρb ¼ ag V Pg ; (B-1) Kazatchenko, E., M. Markov, and A. Moustov, 2004, Joint modeling of
acoustic velocities and electrical conductivity from unified microstructure
of rocks: Journal of Geophysical Research, 109, B01202, doi: 10.1029/
where V P is the P-wave velocity and ag and bg are empirical coeffi- 2003JB002443.
cients, which, for GoM clays, have been calibrated and found to be Kazatchenko, E., M. Markov, and A. Moustov, 2006, Simulation of acoustic
velocities, electrical and thermal conductivities using unified pore-
ag ¼ 1.75, bg ¼ 0.265 (Mavko, et al., 1998). structure model of double porosity carbonate rocks: Journal of Applied
The density ρb is related to porosity ϕ as ρb ¼ ϕρf þ ð1 − ϕÞρm Geophysics, 59, 16–35, doi: 10.1016/j.jappgeo.2005.05.011.
Mavko, G., T. Mukerji, and J. Dvorkin, 1998, The rock physics handbook:
where ρf and ρm are the pore fluid and mineral density, respectively. Cambridge University Press.
Here, I use ρf ¼ 1.03 gr∕cm3 and ρm ¼ 2.65 gr∕cm3 . Therefore, I Milton, G. W., 2002, The theory of composites: Cambridge University Press.
can write the porosity as an explicit function of velocity and Gard- Morris, P. R., 1969, Averaging fourth-rank tensors with weight functions:
Journal of Applied Physics, 40, 447–448, doi: 10.1063/1.1657417.
ner’s parameters as Mukerji, T., J. G. Berryman, G. Mavko, and P. A. Berge, 1995, Differential
effective medium modeling of rock elastic moduli with critical porosity
constrains: Geophysical Research Letters, 22, 555–558, doi: 10.1029/
95GL00164.
b
ϕ ¼ ðag V Pg − ρm Þ∕ðρf − ρm Þ. (B-2) Nur, A., G. Mavko, J. Dvorkin, and D. Galmudi, 1998, Critical porosity: A
key to relating physical properties to porosity in rocks: The Leading Edge,
17, 357–3626, doi: 10.1190/1.1437977.
Peltonen, C., O. Marcussen, K. Bjorlykke, and J. Jahren, 2009, Clay mineral
diagenesis and quartz cementation in mudstones: The effects of smectite
It is assumed that the velocity is the vertical velocity and, thus, the to illite reaction on rock properties: Marine and Petroleum Geology, 26,
stiffness C33 and C 44 are given by 887–898, doi: 10.1016/j.marpetgeo.2008.01.021.
 E186 Bachrach

Raymond, O., L. Fuentes, and J. I. Gómez, 1996, Surface representation coating and their conductive and elastic properties: Materials Science
of polycrystal physical properties: All crystal classes, simple average and Engineering, A: Structural Materials: Properties, Microstructure
approximation: Textures and Microstructures, 28, 81–92, doi: 10.1155/ and Processing, A386, 164–174.
TSM.28.81. Storvoll, V., and I. Brevik, 2008, Identifying time, temperature and
Sayers, C. M., 1987, The elastic anisotropy of polycrystalline aggregates of mineralogical effects on chemical compaction in shales by rock physics
zirconium and its alloys: Journal of Nuclear Materials, 144, 211–213, doi: relations: The Leading Edge, 27, 750–756, doi: 10.1190/1.2944162.
10.1016/0022-3115(87)90033-X. Thyberg, B., J. Jahren, T. Winje, K. Bjørlykke, and J. I. Faleide, 2009, From
Sayers, C. M., 1994,. The elastic anisotropy of shales: Journal of Geophy- mud to shale: Rock stiffening by micro-quartz cementation: First Break,
Downloaded 01/15/20 to 170.38.23.24. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

sical Research, 99, 767–774, doi: 10.1029/93JB02579. 27, 27–33.

Sayers, C. M., 1995, Anisotropic velocity analysis: Geophysical Prospect- Vanorio, T., M. Prasad, and A. Nur, 2003, Elastic properties of dry
ing, 43, 541–568, doi: 10.1111/gpr.1995.43.issue-4. clay mineral aggregates, suspensions and sandstones: Geophysical
Sayers, C. M., 2005, Seismic anisotropy of shales: Geophysical Prospecting, Journal International, 155, 319–326, doi: 10.1046/j.1365-246X.2003
53, 667–676, doi: 10.1111/gpr.2005.53.issue-5. .02046.x.
Sen, P. N., C. Scala, and M. H. Cohen, 1981, A self-similar model Vernik, L., and C. Landis, 1996, Elastic anisotropy of source rocks: Impli-
for sedimentary rocks with application to the dielectric constant cations for hydrocarbon generation and primary migration: AAPG
of fused glass beads: Geophysics, 46, 781–795, doi: 10.1190/ Bulletin, 80, 531–544.
1.1441215. Voltolini, M., H. R. Wenk, N. H. Mondul, K. Bjorlykke, and J. Jahren, 2009,
Sevostianov, I., and M. Kachanov, 2001, Plasma-sprayed ceramic coating: Anisotropy of experimentally compressed kaolinite-illite quartz mixtures:
Anisotropic elastic and conductive properties in relation to the microstruc- Geophysics, 74, no. 1, D13–D23, doi: 10.1190/1.3002557.
ture; cross property correlations: Materials Science and Engineering, A: Wang, Z., 2002, Seismic anisotropy in sedimentary rocks, part 2: Laboratory
Structural Materials: Properties, Microstructure and Processing, A297, data. Geophysics, 67, 1423–1440, doi: 10.1190/1.1512743.
235–243, doi: 10.1016/S0921-5093(00)01022-4. Wenk, H. R., I. Lonardelli, H. Franz, K. Nihei, and S. Nakagawa, 2007,
Sevostianov, I., M. Kachanov, J. Ruud, P. Lorraine, and M. Dubois, 2004, Preferred orientation and elastic anisotropy of illite-rich shale: Geophy-
Quantitative characterization of microstructures of plasma sprayed sics, 72, no. 2, E69–E75, doi: 10.1190/1.2432263.