See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/249865204

A simple, effective-medium model for water saturation in porous rocks

Article  in  Geophysics · July 1995

DOI: 10.1190/1.1443835

CITATIONS READS
48 525

1 author:

Charles Berg
ResDip Systems
11 PUBLICATIONS   122 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The relationship between P10, P32, and size of natural fractures in boreholes. View project

Fold and Thrust Belt Techniques View project

All content following this page was uploaded by Charles Berg on 20 May 2016.

The user has requested enhancement of the downloaded file.

GEOPHYSICS, VOL. 60, NO. 4 (JULY-AUGUST 1995); P. 1070-1080, 7 FIGS., 1 TABLE.
A simple, effective-medium model for water
saturation in porous rocks
Charles R. Berg*
ABSTRACT
A general equation for water saturation is derived
from effective-medium theory. A simpler low-fre-
quency equation is also derived. Both equations are
directly solvable for water saturation. The model
should be applicable to any granular, water-wet for-
mation. Additional relationships are derived specifi-
cally for application to shaly sandstones, but the
model should be applicable to a wide range of rock
types, water conductivities, and tool frequencies. In
the derivation, hydrocarbons are included in the ma-
trix component of the equation and the combined
“matrix” elements are treated as resistors in parallel.
The low-frequency equation is compared to various
approaches to calculation of water saturation, such as
Dual-Water, Waxman-Smits, and Bussian. The gen-
eral equation is compared to three-component effec-
tive medium, porosity index, and complex refractive
index models (CRIM). The model is proven to work on
experimental data under a wide range of frequencies
and water conductivities. It is recommended that the
new saturation model be used for nearly all types of
electrical saturation calculation, whether the measure-
ments are from standard resistivity tools or from
dielectric tools.
INTRODUCTION
Water saturation (S,) in clean sandstones is easily calcu-
lated by the Archie (1942) equation. A problem arises,
however, when the rock is shaly or when the formation
water has low conductivity. A large number of models has
been generated for the shaly sand problem, most of which
are empirical or a combination of theoretical and empirical
approaches. In this paper “theoretical” means an approach
which is based on broad, proven principles, while “empiri-
cal” means an approach intended to fit a particular set of
Manuscript received by the Editor May 9, 1994; revised manuscript received October 17, 1994.
*20 Kearny Brook Pl., The Woodlands, TX 77381.
© 1995 Society of Exploration Geophysicists. All rights reserved.
1070
experimental or field data. By nature, an empirical model
can provide a close fit to the data set from which it is
derived. However, there are two ways in which empirical
approaches may lose accuracy: (1) conditions may extend
beyond the variability of the initial data set (for example, an
apparent, straight-line relationship may actually be curved
when the range of its variables is extended), or (2) properties
peculiar to the data set or to the experimental procedures
used may bias the models when applied to other data sets.
The desirability of theoretical approaches is that their gen-
erality should allow a wider range of application than that of
empirical approaches.
Some purely theoretical models for water saturation have
been based on the effective medium theory for concentrated
mixtures, specifically the Hanai-Bruggeman (HB) equation.
Effective medium theory is based on the Maxwell-Wagner
theory for calculating the conductivity of mixtures of dielec-
trics. Bruggeman (1935) developed the theory for concen-
trated mixtures of dispersed particles (disperse phase) in a
continuous medium, under the assumption that both dis-
perse phase and continuous medium were nonconducting.
Hanai (1960a, b) extended the Bruggeman theory by assum-
ing that both phases had some conductivity. Two examples
of effective-medium models are those of Feng and Sen (1985)
and Samstag (1992). Effective-medium methods are advan-
tageous for studying because (1) they can handle com-
plex conductivities or dielectric constants and thus correctly
model frequency dependence, (2) the HB equation upon
which they are based reduces to Archie’s law with high
formation water conductivity and low matrix conductivity,
and (3) they allow a conductive rock matrix. Until now,
effective medium theory was difficult to apply to log analysis
because the methods employed iterative techniques for
solutions that were difficult to apply. The general equation
derived here, although slightly more complex than some
approaches, is simple enough that it can easily be solved on
calculators or spreadsheets.
Conductivity measurements used to calculate have
traditionally been separated into high-frequency and low-
 Effective-medium Water Saturation
frequency techniques. Low-frequency logs, which include
standard “resistivity” logs such as induction logs and later-
ologs, have tools with operating frequencies at or below
about 20 KHz. High-frequency logs, also called dielectric
logs, use devices such as Schlumberger’s EPT and DPT
tools that have operating frequencies in the megahertz to
gigahertz range.
Measurement-while-drilling tools have recently been de-
veloped that operate at 2 MHz. These tools are generally
used to measure dc conductivity or resistivity by removing
of the dielectric (high-frequency) effects. The uncorrected
high-frequency data from these logs could be an additional
source of information for saturation calculations.
The model derived here should be applicable at all fre-
quencies because it is derived from an equation that is valid
for all frequencies: the HB equation.
THE HANAI-BRUGGEMAN EQUATION
The electrical properties of water-saturated rocks depend
on frequency. The frequency dependence of the conductiv-
ities used in the HB equation can be stated as
=a+
where a* = complex conductivity, a = real conductivity,
= angular frequency, = permittivity of free
space, and = dielectric constant. At low frequencies, the
right-hand, imaginary term of the equation is insignificant
and thus the real conductivity can be used. As frequency
increases, the imaginary part becomes significant, and thus
complex conductivity, permittivity, or dielectric constants
must be used. At extremely high frequencies in the gigahertz
range, real dielectric constants may sometimes be used. For
calculation of at low frequencies, the conductivity of
hydrocarbons is effectively zero, and thus a low-frequency
formulation does not require input of hydrocarbon conduc-
tivity. However, because hydrocarbon dielectric constants
are nonzero, the imaginary part of complex conductivity
[equation (l)] becomes significant at higher frequencies,
requiring a high-frequency saturation model to include hy-
drocarbon complex conductivity.
The HB equation for complex conductivity (Hanai 1960a,
b; Bussian 1983) of a two-component mixture of matrix and
water is as follows:
where = porosity, = complex water conductivity,
= complex mixture conductivity, = complex matrix
(disperse phase) conductivity, and = cementation expo-
nent or geometric factor. The asterisk (*) notation denotes
complex conductivities. Following this section, the notation
will be dropped, but at high frequency, conductivities are
assumed to be complex.
Two important aspects of the electrical properties of
water-saturated rocks are interfacial or induced polarization
at low frequencies and dispersion at high frequencies. Inter-
facial polarization in rocks is commonly related to anoma-
lous conductivity in clays, while dispersion is related to
transmission loss. Because the HB equation was formulated
to handle such effects using equation (1) with complex
1071
conductivity (Hanai, 1968), it is ideally suited for studying
conductivity in rocks at all frequencies.
Although effective-medium models should theoretically
work at all frequencies, their difficulty in application has
generally precluded them from standard log analysis. Be-
cause of the lack of an easily applied theoretical model,
calculation of has traditionally been separated into
low-frequency methods and high-frequency methods. For
low frequencies, especially in shaly sand analysis, a large
number of models have been developed, many of which have
a limited geographic and stratigraphic range. The shaly sand
problem is an example of how empirical models derived for
a given set of conditions will be certain to work only under
those conditions.
The HB equation at low frequency
At low frequency and low matrix conductivity the HB
equation (2) reduces to
where = water conductivity and = mixture conduc-
tivity. This is the empirical relationship commonly known as
Archie’s law. The relationship between Archie’s law and the
HB equation has been well documented (Sen et al., 1981;
Bussian, 1983).
Bussian (1983) demonstrated the suitability of the HB
equation to the study of conductivity in rocks at low fre-
quencies. In that paper he plotted the theoretical curves for
the variables against experimental values and showed the
usefulness and accuracy of the equation in the study of rock
conductivity.
The m of Archie’s law [equation (3)] translates roughly to
the m of the Hanai-Bruggeman [equation (2)] as shown in
Bussian (1983). This does not mean, however, that the
variables can be used interchangeably. One common method
of determining of m experimentally is to plot formation
resistivity factor, or log (F = versus log
(Figure 1). The variable is the slope of the best-fit line to
plotted core points. Often linear curve-fits of core data
dictate that Archie’s Law be stated in the form:
where a is they-intercept at = 1 (Figure 1). There has been
much dispute about whether the variable a exists or not, but
the problem can easily be explained by observing the
curve. When a linear fit is applied to core data, many times
there will be a positivey-intercept or a. The curve in the HB
relationship might explain this. For example, if the core data
had porosities ranging from 8 to 20% it is easy to see how
someone might pick an a of about 1.8 for this particular
formation (thin line). Note also that the slope of the
fit” line is not 2, which was the value of used in both
Archie’s law and the HB equation. Figure 1 supports
Bussian’s (1983) assertion that a should be equal to 1. Once
the differences between Archie’s and are demon-
strated, there still exists a problem of how to find for the
HB equation.
1072 Berg

and 4 are known, m can be found by the

HB equation. However, when analyzing experimental and
log data, both m and a, are unkowns. With a single core
sample, there is only a single porosity point and the above
method cannot be used. In this case, Archie’s m is usually
found by finding as the slope of versus a, and then
calculating m from 4 and F. A possible method for finding
HB m is to plot values of m calculated by Archie’s law
versus water resistivity (R, = l/u,) and then use the
y-intercept as m under the assumption that as ap-
proaches 0, Archie’s m should approach m because
the HB equation reduces to Archie’s law at low
(Figure 2).

THE PROPOSED MODEL

Initial assumptions

Assumptions shared with most other models.-All empiri-

cal models and many theoretical models use the following
assumptions, whether explicit or implied. (1) The rocks are
water-wet. One consequence of this assumption is that
below critical water saturation, which is usually around
5-15% and which should rarely occur under reservoir con-
ditions, models might become inaccurate because the water
loses its continuity, and rock conductivity decreases dramat-
ically. This is especially true with clayey rocks that lose their
anomalously high conductivities when dry, although in the
subsurface, clay minerals would resist giving up their water
because of capillary and adsorption effects. (2) For low-

FIG. 1. Formation resistivity factor versus for a

hypothetical formation. Ideal Archie prediction (dashed line)
and ideal HB prediction (solid curve). The input variables for
calculation were: = 10 S/m, a, = 0.02 S/m, and m = 2.
The slope of the Archie line determines Archie’s m, but
since the HB line is curved, the HB m cannot be determined
from this approach. When a linear fit is applied to core data,
many times there will be a positivey-intercept. The curve in
the HB relationship might explain this. For example, if the
core data had porosities ranging from 8 to 20% it is easy to
see how someone might pick an a [equation (4)] of about 1.8
for this particular formation (thin line). Note also that the
slope of the “best-fit” line is not 2, which was the value of m
used for both Archie’s law and the HB equation. This figure FIG. 2. A method for defining m for the HB equation at low
supports Bussian’s (1983) assertion that a should be equal to 1. frequencies. The oints calculated from core data with
Archie’s equation dots) show an a parent logarithmic- e
curve. When a logarithmic curve t is used, however, t e
results are poor. It is best to use an asymptotic (linear) fit to
the points of lowest The data used in this figure are from
sample 20 of Table 7 in Waxman and Smits (1968).
 Effective-medium Water Saturation
frequency relationships, the assumption is that the hydro-
carbons have zero conductivity (infinite resistivity).
Hydrocarbons associated with matrix.-One key assump-
tion of this model is that the hydrocarbons can be included
with the matrix or disperse phase and that the continuous
phase includes only the formation water. Following are
reasons for this assumption.
The original experiments carried out by Hanai (1960a, b)
to confirm his theoretical model were carried out on colloidal
suspensions of water in oil and oil in water. In some of the
experiments, the oil was the disperse phase (droplets) while
the water was the continuous phase, and in other experi-
ments the conditions were reversed with the water being the
disperse phase and the oil the continuous phase. In the
former case the oil was, in effect, the matrix. Consequently,
it is logical to include hydrocarbons along with matrix in a
water saturation model. Some might object to extending a
two-component system into three, but that has already been
done, in effect, because a rock such as shaly sand is already
a two-component system before the water is added. In fact,
in electrical properties, size, and shape, the hydrocarbons
are more similar to the sand than to the clay and, in turn,
more similar to the whole-rock matrix than to the water. For
water saturations above critical saturation (the point below
which the grains will start to become oil-wet, usually around
15%), hydrocarbons should maintain a configuration more
similar to that of the matrix than that of the water (Figure 3).
Interfacial tension between the hydrocarbons and the water
cause the hydrocarbons to assume outwardly curved shapes
that resemble the grains more than the water. There are
connections between the oil globules, and also connections
between the matrix grains, but the water is the effective
medium.
1073
effective medium theory is perfectly suited for handling
interfacial effects (i.e., interfacial polarization), the excess
conductivity can be treated as part of the matrix conductiv-
ity. The method of measuring the water-saturated matrix
conductivity is to use the HB equation solved for matrix
conductivity
Prior to discussion of clay conductivity, it is important to
define the concept of “bound” water. Because of their
extremely small pore spaces, clays give up their water with
great difficulty. This gave rise to the concept of “capillary
bound” water or simply “bound” water. The “clay water”
of Clavier et al. (1984) should not be confused with bound
water, because the “clay water” is a molecular-scale layer at
the surface of the micrometer-scale clay particles. It is
assumed that the only distinction between most of the bound
water and the water in the sand pores is location and not
composition because the water is “bound” only in the
mechanical sense, and it is actually connected with the water
in the sand pores.
Associating anomalous clay conductivity with matrix con-
ductivity allows the clay pore water to be treated identically
to the sand pore water. The thickness of the “diffuse” layer
of Clavier et al. (1984) is on the order of 6 or 6 x m
(The term “diffuse” layer is used to describe a so-called
“ion free” layer that extends from the clay-water interface).

Disperse-phase elements as resistors-in-parallel.-Related

to the assumption that hydrocarbons can be associated with
the matrix is the assumption that the matrix elements can be
treated as resistors in parallel in the derivation. Although
resistors-in-parallel is, in the strictest sense, an approxima-
tion, the approach may actually be more appropriate than
more exact models such as that in Feng and Sens (1985).
[See the calculations on the experimental data of Knight and
Endres (1990).] Until now, effective-medium models, either
explicitly or by default, assumed that within a three-compo-
nent system, the effective-medium mixing laws apply be-
tween the hydrocarbon and matrix phases. The resistors-in-
parallel approximation works because water, whether at low
or high frequencies, always has much greater absolute
conductivity than hydrocarbon. The use of resistors-in-
parallel to model the composite, disperse phase is the main
difference between this effective-medium model and previ-
ous ones. FIG. 3. Schematic diagram after Berg (1975) showing how
hydrocarbons are related to the matrix and water. In a
water-wet system such as this, surface tension will keep a
Association of excess clay conductivity with the matrix.- thin film of water (exaggerated here) between the oil and the
When shales are water saturated they have higher matrix grains. In an actual cross section, both grains and hydrocar-
conductivity than would be calculated using their dry matrix bon would appear to be “floating” particles in a continuous
conductivity. The higher matrix conductivity is generally water phase, while in 3-D space the hydrocarbon is like a
mesh of interconnected globules, especially at lower S,.
attributed to interfacial ions on or near the surface of clay This configuration is why resistors-in-parallel is a good
particles (Clavier et al., 1984, Knight and Endres, 1990). approximation of the conductivity of the mixture of hydro-
Because the excess conductivity is an interfacial effect and carbons and grains.
1074 Berg

Clay grains, on the other hand, might have a diameter of The general and low-frequency saturation equations.-
about 0.1 Assuming that the size of the pore spaces is on Association of hydrocarbons with the matrix changes the
the same order of magnitude as the clay grains, the volume HB equation (2) to
of “ion free” water should be approximately 1/100 of the
volume of the clay water if volume is equivalent to cross-
sectional area. The water outside of the “ion free” water is
not influenced electrically by the clay grains, and the ionic where a, = partially saturated mixture conductivity. Sub-
composition should be the same as adjacent sands under stitution of equation (5) into equation (6) yields
equilibrium conditions. Thus, nearly all of the “bound”
water can be considered the “far” water of Clavier et al.
(1984) since the new model associates the anomalous con-
ductivity of clay to the matrix and not the water. When
studying logs, this assumption will allow the log analyst to
use one water conductivity for both sandstone and shale
except under nonequilibrium conditions such as strong hy-
drodynamic flow or water floods.
Even with the limits imposed by the above assumptions, [Note that this equation reduces to the HB equation (2) at
the relationships derived below should be applicable to all 100% SW.] The quadratic with respect to SW is
water-wet, granular reservoir rocks.

Derivation

Disperse-phase conductivity.-Assuming resistors in paral-

lel, the disperse-phase conductivity can be stated

where = disperse phase conductivity, a, = matrix

conductivity, = hydrocarbon conductivity, = cross-
sectional area of the composite disperse phase, A, =
cross-sectional area of the rock matrix, and = cross- Solving for SW gives the general saturation equation

At low frequency, hydrocarbon conductivity approaches

zero, so equation (9) reduces to

sectional area of the hydrocarbons. Substituting volumetric
variables for cross-sectional areas gives
Solving for yields the disperse-phase conductivity
(5)
Grain conductivity (a,) in both equations (9) and (10) can
include clay grains if a suitable mixing formula can be used
to find the effective grain conductivity of the composite
matrix grains. Possibilities include the dispersed-clay formu-
lation described below or the models set forth by Lima and
Sharma (1990).
Equation (10) has been tried on about 30 low-frequency
shaly sand log examples so far and appears to be stable and
accurate. The most frequent calculation problem occurs
when the curve used to calculate clay content, usually the
 Effective-medium Water Saturation 1075

gamma log, does not reflect clay content accurately. Over- can be difficult. This is not a problem for the log analyst
estimation of clay content can result in anomalously low or when calculating using equation (10), but it is a problem
even negative saturations, not only with this model but with when analyzing the behavior of a, or a, with respect to S,.
other models as well. Similarly, underestimation of clay One commonly used method for solving the HB equation is
content can result in anomalously high saturations. the Newton-Raphson method, which is an iterative method
for finding roots to equations. This method can be used to
A dispersed-clay shaly sand model.-The following rela- solve equations (9) and (10) for a,. Figure 4a is a plot,
tionship will usually be used with the low-frequency generated by equation (10) using the above method in
equation (10). However, some conditions might warrant
using the general equation (9). For example, at intermediate
frequencies clay conductivity could still be significant while
at the same time the dielectric constant might also become
significant.
The dispersed-clay model, assuming resistors-in-parallel
for the matrix elements, starts with the equation

where = sand grain conductivity (small, but not insig-

nificant), = clay grain conductivity, A,, = cross-
sectional area of the sand matrix, and = cross-sectional
area of the clay matrix. Substituting volumetric variables for
cross-sectional areas gives

l
+ l l
where = sand porosity, = clay porosity, and =
clay volume ratio. [The sand porosity above is the
effective porosity divided by sand volume or, in other words,
the porosity attributable to sand relative to sand content.]
Substituting for a, (equation 8) yields
l
+ l l
Note that can be substituted directly into equation (10).
Another important aspect of mixing sand and clay prop-
erties is calculating an appropriate value for cementation
exponent (m). Cementation exponent can have a large effect
on calculated saturations, although it is common practice to
set m to 2. A typical clean sand might have m of 1.8 while
clay m can range from 2.5 to 3 based on the (1983)
and (1992) data. A linear estimate calculated from
clay volume is reasonable based on plots of a clay
volumetric variable, and m in the (1983) and
(1992) papers.
At frequencies, equation (11) is accurate as long as the
water conductivity is at least three times the clay grain
conductivity. In fresh or brackish water environments, this
may not ‘hold true and alternative mixing formulas, such as
Lima and Sharma’s (1990) may have to be used. Further
investigation will be required as to the appropriateness of the
mixing formulas at higher frequencies. This limitation of
resistors-in-parallel has very little effect on the saturation
equations (9) and (10) because hydrocarbon absolute con-
ductivities are, at all frequencies, much smaller than water FIG. 4. index” versus effective water
absolute conductivities. saturation calculated by equation (10) starting with
One difficult in working with the HB equation is that ranging from 0.0 to 0.4 for the dispersed-clay model (a) and
the laminated-shale model (b). The other variables were as
although it is fairly simple, it cannot be solved algebraically follows: 0.1 S/m, 0.01 S/m, = 100 S/m,
for a, or because of the cementation exponent For and 4 = 0.4. This figure demonstrates how much clay can
the same reason, solving equations (9) and (10) for a, or a, affect rock conductivity.
1076
conjunction with equation (11), of resistivity index
versus effective water saturation for five values of
ranging from 0 to 0.4. Figure 4 demonstrates how clay can
greatly increase the conductivity of partially saturated rocks.
A laminated shaly sand model.-Resistors-in-parallel for
whole-rock elements might be an appropriate model for
shaly sandstones having clay contained in laminated or
thinly bedded shale. This model holds true only for tools in
which the electrical field lines run parallel to bedding, in
other words, induction tools.
In contrast to the dispersed model, which puts the grain
conductivities in parallel, the laminated model puts the
whole-rock conductivities in parallel. The laminated model
starts with the relationship of conductivities and lengths as
resistors in parallel
where = length of rock mixture, = length of
sandstone, = length of shale, = partially-saturated
sand conductivity, and = shale conductivity. Substitut-
ing volumetric measures for length gives
Solution for yields
The variable is substituted into equations (7) and (10)
in place of The answer will be effective saturation (S,,)
and not whole-rock as in the parallel model. In case of
dispersed clay within the sandstone beds, a hybrid model
can be calculated using from equation (12) and other
input variables appropriate for the dispersed-clay model.
Figure 4b is a plot of the laminated shale model using the
same input parameters as Figure 4a.
COMPARISON TO OTHER SATURATION MODELS
Previous models in which hydrocarbons were associated
with matrix
An inherent assumption of many saturation models is that
the ratio of calculated porosity versus true porosity is equal
to the water saturation. Using such a ratio is a tacit admis-
sion that hydrocarbons are associated with the matrix. The
following are two examples of such models, one for low
frequency and the other for high frequency.
It is possible to derive Archie’s combined water saturation
equation from Archie’s law (1942) by assuming hydrocar-
bons can be associated with the matrix. Including hydrocar-
bon with the matrix in equation (2) gives
Rearranging gives
. the method because the clay conductivity in log analysis is
Berg
Accepting n equal to m and substituting Archie’s F for
yields Archie’s combined saturation equation:
= = = formation resistivity factor and
n = saturation exponent. The main difference between the
above derivation and the derivation for effective-medium
theory is that in Archie’s law there is no term for matrix
conductivity, hence there is no need to compensate for the
change in matrix conductivity that results from including the
hydrocarbons with the matrix.
Sherman (1987) described a high-frequency model based
on the HB equation. In the model he used the relationship:
where = the calculated porosity using the HB equation
and k = dielectric saturation exponent (approximately
equal to 1). Note the similarity of the left-hand sides of
equations (6), (13), and (14). Although Sherman’s relation-
ship implicitly assumes that the hydrocarbons are included
with the matrix, it does not account for the change in the new
“matrix” conductivity caused by the addition of the hydro-
carbons.
Low-frequency experimental data
A brief review of low-frequency methods.-There are many
low-frequency models, most of which are shaly sand meth-
ods. Only models in wide use or relevant to this paper will be
reviewed.
The Waxman-Smits (1968) or WS method is an empirical
shaly sand model based on experimental measurements of
rock conductivities and a chemical measurement of clay
conductivity The model is extremely accurate for
experimental data when the form of the equation uses the
resistivity index (WS, equation 29). However, when
the formation resistivity factor is introduced (WS, equa-
tion 22), the accuracy of the equation is reduced. Two
possible reasons for this change are: (1) the introduction of
two more variables (porosity and saturation exponent) in-
creases the amount of error entered into the equation, and
(2) the use of the resistivity index corrects for systematic
error in the measurement of rock conductivity. For a fair
comparison of models, only the WS equation (22) was used
because the other models to which it was compared used
and m in one form or another (F =
The dual-water or DW model, an empirical method intro-
duced by Clavier et al. (1984), is probably the most widely
used of the low-frequency methods. This model is based on
the assumption that the excess conductivity associated with
clays can be associated with their “bound” water. It must be
noted that the way in which Clavier et al. calculated and
m was unorthodox in that the chemically derived value for
was ignored and the values for and m were calculated
by obtaining the lowest variance in predicting for the
100% saturated samples. This, in itself, does not condemn
commonly derived directly from logs, but it does make
 Effective-medium Water Saturation 1077

comparison to other methods with experimental data some-
what problematic, especially since their variables and
are not actually used in log calculations. Another difference
between the theoretical DW method and the way in which it
is applied is that in log calculations all of the bound water is
treated as “near” water, which is not true.
The Bussian (1983) model, although accurate, has not
been widely used. This equation was developed from a
binomial expansion of the equation Bussian recognized
the similarity in form of the low-order terms of the expansion
to de Witte’s (1957) empirical model of shaly sand conduc-
tivity. The two new factors for de Witte’s empirical conduc-
tivity equation, “A” and “B” were then substituted into de
Witte’s saturation equation to derive Bussian’s saturation
equation. The Bussian equation is thus a hybrid theoretical-
empirical model. According to Bussian, the error resulting
from using only the low-order terms of the expansion re-
stricts the accuracy of the equation to conditions where the
matrix conductivity (a,) is less than about l/5 that of the
whole-rock, fully-saturated conductivity Another limi-
tation of this model, which is also true for the DW and WS
methods, is that it is valid only under low-frequency condi-
tions because it does not account for hydrocarbon conduc-
tivity (dielectric constant).
Defining m, and F.-To compare the above three
models to equation it was necessary to derive values for
m and a, that would be as objective as possible. The
variables m and a, used for equation (10) and the Bussian
equation were based on the method described in the section
titled “the HB equation at low frequency” (Figure 2).
Because all of the experiments were run at a of approx-
imately 1 ohm-m or less, m was found by linear regression of
all of the points for each water-saturated sample. Then
was calculated for each measured point using the HB equa-
tion solved for The median a, was then used in satura-
tion calculations. Note that was not needed at all because
matrix conductivity was derived from the fully-saturated
experimental data. All other variables used in calculating
were taken from their respective data sources.
Method for calculating and applying
The saturation exponent was found and applied for all
models in this section by the following equations. This
ensured uniformity for fair comparison. (Saturation expo-
nents were used on all calculations.) The variable was
calculated in the following manner:
where = calculated water saturation and =
measured water saturation. The following equation was used
to apply n:
= (16)
Results from equation (10) compared to DW, WS, and
Bussian models.- Clavier et al. (1984) published saturation
versus resistivity measurements taken by Waxman and
Thomas (1974), which also contained some measurements
taken by Hill and Milbum (1956). These measurements were
taken at water resistivities ranging from 0.0495 to
1.04 ohm-m with a wide variation in Qv. The data set
included 147 saturation measurements taken on 12 rock
samples. Although they did not report the results of their
saturation calculations, they reported the variance of their
equation when predicting and also reported m and
derived by a best-fit method to the 100% saturated data.
Using those calculated m and measurements, saturations
were calculated by their saturation equation. For WS
equation (22) (the equation), Clavier et (1984)
published values for chemically derived and Archie-type
m were used for the WS equation (Clavier et al.‘s WS1
group). Saturations were also calculated with equation (10),
the Bussian equation, and Sherman’s resistivity-index equa-
tion with values of m and calculated from the 100%
saturated measurements (Table 1).

Table 1. Standard error for five models on the data sets from Clavier et al. (1984), Sherman (1987), and Knight and Endres (1990).
The “n not applied” calculations are representative of the way in which saturations are usually calculated from logs.
conditions MODEL
Data set salinity frequency this paper Bussian DW WS Sherman CRIM

n not applied
Clavier, et al High-med. Low 4.2 4.8 5.6 4.1 5.5
Sherman High Very high 7.5 8.8 6.3
Knight & 0 High 12.4 14.2
Endres
n applied
Clavier, et al High-med. Low 3.0 2.4 3.3 3.7 3.9
Sherman High Very high 3.6 3.6 3.6
Knight & 0 High 8.4 7.8
Endres
1078 Berg
High-frequency experimental data stants may be substituted into equation (9) in place of their
complex conductivity counterparts.
The definitions for dielectric constants used in this section
are as follows. = complex dielectric constant of rock, High-water conductivity.-The study by Sherman (1987)
water and hydrocarbon mixture, = complex water was based on laboratory measurements of about 40 partially
dielectric constant, and = complex matrix dielectric saturated samples using water with high salinity. Measure-
constant. Prime notation (‘) denotes the real part of a ments were taken at 100% as well as at partial saturation
complex value and double prime notation denotes the to calculate a dielectric saturation exponent “k” and the
imaginary part. For example, is the real part of and depolarization factor (analogs to and m, respec-
is the imaginary part. These complex dielectric con- tively, of Archie). The experiments were run at 1.1 GHz, a

FIG. 5. Measured real dielectric constant (K’) versus for three rock samples at 105 kHz, 109 kHz, and 1.2 MHz from Knight
and Endres (1990) lotted with values calculated by equation (9) and CRIM. used for each model was calculated based on
an extrapolated, 1 00% saturated K’ . Calculations using Feng and Sen’s (1985) spherical model were nearly identical to the
CRIM results.
 Effective-medium Water Saturation
very high frequency. Results of calculations are shown on
Table 1. Equation (10) and Sherman’s model had an m of 1.5,
while for all three models was from the values given by
Sherman (1987).
Low-water conductivity.- Knight and Endres (1990) mea-
sured versus for three rock samples at 105 kHz and
1.2 MHz. They used deionized (fresh) water in their satura-
tion experiments. The complex refractive index model
(CRIM) of Wharton et al., (1980) was then compared to the
three-component, effective-medium model of Feng and Sen
(1985). To make the methods work better in predicting ,
the concept of wetted-rock was conceived. This resulted
in a fairly close fit to the data with the results of the two
methods being nearly identical.
To calculate for equation complex was calcu-
lated by the HB equation (2) at an extrapolated (if neces-
sary), 100% K’ was then calculated using an m of 1.5 for
all three samples. These calculations resulted in a close
match between the results of equation (9) and the data
(Figure 5). Use of wetted-rock K’ was not necessary for
equation (9) because the slope more closely matched that of
the experimental data than the other two models. For
FIG. 6. Modified Feng and Sen’s (1985) Figure 5b showing
their bimodal and spherical models. Results of comparable
calculations from equation (9) and CRIM are also shown.
The input variables were: = 0.05 S/m, = 0.22, = approaches 0, the values for
80, = 4.7, = 2.2, and frequency = 2.5 MHz. The
two curves for equation (9) used m values of 1.8, 1.5, and
1.2, while the bimodal model of Feng and Sen usedp of 0.05
and of 0.03 and are geometric factors similar to m and
Note that equation while being much simpler than
the spherical model, matches that model almost identically
with an m of 1.5, and that using an m of 1.2 results in a fairly
good approximation of the bimodal model.
1079
comparison, the CRIM model was used to calculate K’ using
a 100% CRIM-calculated (As with Knight and
Endres’ (1990) calculations, results for Feng and Sen’s
(1985) spherical model were nearly identical to the results for
CRIM, and hence are not shown.) Calculating in this
way, from the 100% saturated data, simulates the way in
which log analysts calculate in shaly sands. Listed in
Table 1 are calculations of on the three samples using
equation (9) and CRIM.
Comparison to the models of Feng and Sen and to CRIM
Feng and Sen (1985) derived spherical and bimodal effec-
tive-medium models for studying water saturation. The
spherical model, in essence, assumes an m of 1.5 while the
bimodal model allows the input of two geometric factors,
and 6. With m set to 1.5, equation (9) had results virtually
identical to the spherical model, and with m set to 1.2, the
results were fairly close to the bimodal model and CRIM
(Figure 6). Although their bimodal model allows two geo-
metrical factors to be input, they do not offer a relationship
of these two variables to the geometrical factors m and
Equation (9), on the other hand, uses a m similar to Archie’s
m, and n can also be applied in a similar manner. Although
FIG. 7. The percent difference in K’ between equation (9)
and Feng and Sen’s (1985) spherical model. The difference
approaches zero as ap roaches 1. This is to be expected
since both are based on t e HB equation, and both should
calculate the same value at 1. Note that as
diverge. At that point,
equation (9) approached a value of 4.150 and the spherical
model approac hed a value of 4.051. The former value is
consistent with resistors-in-parallel while the latter value
approximates an effective-medium solution with hydrocar-
bons dispersed within a continuous matrix. Note that the
equation (9) results were always larger than the Feng and
(1985) results, and that ius also true in figure 5 where
equation (9) has acloser fit with the experimental data.
1080
CRIM is the simplest of the models, it does not allow for
input of a geometric factor, which can have a great effect on
the dielectric constant, especially at lower frequencies.
When an m of 1.5 was used, equation (9) had results that
looked nearly identical to the spherical model in all of Feng
and Sen’s (1985) figures. Inspection of the actual values,
however, shows that there is a significant difference between
the models, especially as approaches 0 (Figure 7). As
approaches 0, the relationship for is a reduction of the
disperse-phase equation (5), in terms of dielectric constant,
to
which is reasonable for a dielectric constant of a mixture of
two continuous materials or resistors in parallel. Feng and
Sen’s model at
effective medium and the hydrocarbon-filled pores as the
disperse phase, which would imply that the hydrocarbons
assumption inherent in the HB equation that the pore spaces
are connected. [In Figure 7, as approached 0, equation
(9) had a of 4.150, Feng and Sen’s spherical model had a
of 4.051, while an effective medium of hydrocarbon in
matrix calculates as 4.056.] The relative merits of resistors-
in-parallel versus effective-medium for matrix and hydrocar-
bons are beyond the scope of this paper, but perhaps the
calculations on Knight and Endres’ (1990) data demonstrate
which model is more appropriate for rocks.
CONCLUSIONS
A theoretical model has been developed that is accurate
for saturation calculation for both standard resistivity and
dielectric measurements. The wide applicability of the
method proven by calculation of experimental saturations
under diverse conditions implies that it is robust and that it
should be the method of choice under all circumstances,
given that the empirical models now in use could be subject
to large error outside of the range of variables for which they
were derived. Use of a single equation for all types of
electric logs should ensure uniformity when comparing in-
vaded-zone saturation to uninvaded saturation, thus making
calculation of irreducible saturation and relative permeabil-
ity more accurate.
Because most of the experimental data tested were from
shaly sands, the model will be well-suited for the calculation
of in shaly sands and in rocks with low water conductiv-
ity. Another possible application might be the correction of
measurement-while-drilling dielectric readings to resistivity
as well as direct calculation of from the dielectric
readings. The general equation derived here could also have
applications outside of petroleum geology, such as in
groundwater investigations.
Direct solution for has many advantages. One of them
is that the equation is suitable for computation by calculator
or spreadsheet, especially for standard resistivity-log analy-
View publication stats
Berg
sis. Another advantage is that direct solution should make
computer algorithms much faster and more stable, and allow
real-time calculation of using an effective-medium ap-
proach.
ACKNOWLEDGMENTS
I would like to thank Robert R. Berg for introducing me to
the water saturation problem and for his thoughtful sugges-
tions and review of the manuscript. I also thank Alfred E.
Bussian for his insights into the problem and his suggestions
on organization and description of the physical processes. In
addition, I would like to thank Anthony F. Gangi and Marc
K. Croes for their review of the manuscript and Rosemary J.
Knight for access to her experimental data.
REFERENCES
Archie, G. E., 1942, The electrical resistivity log as an aid in
determining some reservoir characteristics: Petr. Tech., 1,55-62.
Berg, R. R., 1975, Capillary pressures in stratigraphic traps: AAPG
Bull., 59, 939-956.
Bruggeman, D. A., G., 1935, Berechnung verschiedener physikalis-
cher konstanten von heterogenen Substantzen: Ann. Physik, 24,
636664.
Bussian, A. E., 1983, Electrical conductance in a porous medium:
Geophysics, 48, 1258-1268.
Clavier, C., Coates, G., and Dumanoir, J., 1984, Theoretical and
experimental bases for the dual-water model for interpretation of
shaly sands: Soc. Petr. Eng. J., 24, 153-168.
de Witte, A. J., 1957, Saturation and porosity from electric logs in
shaly sands: Oil and Gas J., 55, 89-93.
Feng, S., and Sen, P. N., 1985, Geometrical model of conductive
and dielectric properties of partially saturated rocks: J. Appl.
Phys., 58, 3236-3243.
Hanai, T., 1960a, Theory of the dielectric dispersion due to the
interfacial polarization and its application to emulsions: Kolloid-
Zeitschrift, 171, 23-31.
- 1960b, A remark on “Theory of the dielectric dispersion due
to the interfacial polarization and its application to emulsions”:
Kolloid-Zeitschrift, 175, 61-62.
- 1968, Electrical properties of emulsions, in Sherman, P.,
Ed., Emulsion science: Academic Press, 354-378.
Hill, H. J., and Milburn, J. D., 1956, Effect of clay and water salinity
on electrochemical behavior of reservoir rocks: Soc. Petr. Eng.,
Trans. AIME, 207, 65-72.
Knight, R. J., and Endres, E., 1990, A new concept in modeling the
dielectric response of sandstones: Defining a wetted rock and bulk
water system: Geophysics, 55, 586-594.
Lima, 0. A. L., and Sharma, M. M., 1990, A grain conductivity
approach to shaly sandstones: Geophysics, 55, 1347-1356.
Samstag, 1992, F.J., An effective-medium model for complex con-
ductivity of shaly sands in the salinity, frequency, and saturation
domains: Texas A&M Univ. Dept. of Geophys. Rock Physics
Consortium, 4.
Sen, P. N., Scala, C., and Cohen, M. H., 1981, A self-similar model
for sedimentary rocks with application to the dielectric constant
of fused glass beads: Geophysics, 46, 781-795.
Sherman, M. M., 1987, A model for the determination of water
saturation from dielectric permittivity measurements: The Log
Analyst, 28, 282-288.
Waxman, M. H., and Smits, L. J. M., 1968, Electrical conductivities
in oil-bearing shaly sand: Soc. Petr. Eng. J., 8, 107-122.
Waxman, M. H., and Thomas, E. C., 1974, Electrical ronductivities
in shaly sands: I. The relation between hydrocarbon saturation
and resistivity index; II. The temperature coefficient of electrical
conductivity: J. Petr. Tech., Trans. AIME, 257, 213-225.
Wharton, R. P., Hazen, G. A., Rau, R. N., and Best, D. L., 1980,
Electromagnetic propogation logging: Advances in technique and
interpretation: Soc. Petr. Eng., AIME, paper 9267.