Thermal Conductivity in Porous Media: Percolation-Based Effective-Medium Approximation

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/294285181
Thermal conductivity in porous media: Percolation-based effective-medium

approximation
Article in Water Resources Research · December 2015

DOI: 10.1002/2015WR017236
CITATIONS READS
36 308
2 authors:
Behzad Ghanbarian Hugh Daigle

Kansas State University University of Texas at Austin
109 PUBLICATIONS 1,622 CITATIONS 117 PUBLICATIONS 1,069 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Scale-dependent pedotransfer functions View project
Microfluidic Study of Multi-Phase Flow in Hydrocarbon Reservoir View project
All content following this page was uploaded by Behzad Ghanbarian on 24 October 2017.
The user has requested enhancement of the downloaded file.

PUBLICATIONS
Water Resources Research
RESEARCH ARTICLE Thermal conductivity in porous media: Percolation-based
10.1002/2015WR017236
effective-medium approximation
Special Section: Behzad Ghanbarian1 and Hugh Daigle1
Applications of percolation
theory to porous media 1
Department of Petroleum and Geosystems Eng., University of Texas at Austin, Austin, Texas, USA
Key Points:
Novel porosity and saturation Abstract Knowledge of porosity and saturation-dependent thermal conductivities is necessary to investi-
dependent thermal conductivity gate heat and water transfer in natural porous media such as rocks and soils. Thermal conductivity in a porous
models are developed
medium is affected by the complicated relationship between the topology and geometry of the pore space
Theory matches both porosity and
saturation dependent experiments and the solid matrix. However, as water content increases from completely dry to fully saturated, the effect of
well the liquid phase on thermal conductivity may increase substantially. Although various methods have been
The percolation-based thermal
proposed to model the porosity and saturation dependence of thermal conductivity, most are empirical or
conductivity models follow
nonuniversal behavior quasiphysical. In this study, we present a theoretical upscaling framework from percolation theory and the
effective-medium approximation, which is called percolation-based effective-medium approximation (P-EMA).
Correspondence to: The proposed model predicts the thermal conductivity in porous media from endmember properties (e.g., air,
B. Ghanbarian, solid matrix, and saturating fluid thermal conductivities), a scaling exponent, and a percolation threshold. In
[email protected]
order to evaluate our porosity and saturation-dependent models, we compare our theory with 193 porosity-
dependent thermal conductivity measurements and 25 saturation-dependent thermal conductivity data sets
Citation:
and find excellent match. We also find values for the scaling exponent different than the universal value of 2,
Ghanbarian, B., and H. Daigle (2016),
Thermal conductivity in porous media: in insulator-conductor systems, and also different from 0.76, the exponent in conductor-superconductor mix-
Percolation-based effective-medium tures, in three dimensions. These results indicate that the thermal conductivity under fully and partially satu-
approximation, Water Resour. Res., 52,
rated conditions conforms to nonuniversal behavior. This means the value of the scaling exponent changes
295–314, doi:10.1002/2015WR017236.
from medium to medium and depends not only on structural and geometrical properties of the medium but
also characteristics (e.g., wetting or nonwetting) of the saturating fluid.
Received 13 MAR 2015
Accepted 16 DEC 2015
Accepted article online 18 DEC 2015
Published online 16 JAN 2016
1. Introduction
The thermal conductivity k (W m218C21) of a porous medium describes the medium’s ability to transmit heat.
There are three main mechanisms of heat flow through rocks and soils: (1) conduction, in which the heat
passes through the substance of the body itself, (2) convection, in which heat is transferred by relative motion
of portions of the heated body, and (3) radiation, in which heat is transferred directly between distant portions
of the body by electromagnetic radiation [Carslaw and Jaeger, 1986]. In liquids and gases, convection and radi-
ation are of paramount importance, but in solids convection is altogether absent and radiation typically negli-
gible [Carslaw and Jaeger, 1986]. Although conduction, convection, and radiation are mainly responsible for
heat flow through the connected solid particles and/or across pore space, at the soil-air interface transfer of
latent heat becomes important in terms of heat exchange [Bristow, 2012]. In practice, however, it is often diffi-
cult to separate thermal conductivity from convection, and therefore also from latent heat transfer. In addition,
thermal conductivity measurement is challenging, particularly under partially saturated conditions because
moisture tends to move and redistribute when a thermal gradient is imposed. Therefore, what is measured is
actually the apparent thermal conductivity. However, the experimental setups may be manipulated to bring
‘‘apparent’’ closer to ‘‘real,’’ for example by working at low overall temperatures using small thermal gradients
and short measurements times such as with heat pulse probes [see e.g., Bristow et al., 1995; Bristow, 1998].
The thermal conductivity of a composite medium such as soil or rock is not a simple function of the thermal
conductivities and volume fractions of the medium’s components. It is strongly affected by the composition,
shape, and mutual configuration of the various components in the medium, and particularly by water con-
tent [Shiozawa and Campbell, 1990]. Thus, the thermal conductivity of a porous medium under fully and/or
partially saturated conditions is an integrated combination of the thermal conductivity of the individual
C 2015. American Geophysical Union.
V phases e.g., air, solid particles, and water, the relative volumes and connectivity of each phase, and the
All Rights Reserved. microstructural geometry of the medium.
GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 295

Water Resources Research 10.1002/2015WR017236
Various models have been presented to describe the thermal conductivity of porous media. These models
include (1) components in series, (2) components in parallel, (3) fractals, and (4) effective-medium approxi-
mation (EMA). In the series model, the individual components of the medium are assumed to be aligned
perpendicular to heat flow, while in the parallel model, the individual components of the medium are
assumed to be aligned parallel to heat flow. The effective thermal conductivity ke in such parallel and series
structures are given by
ke 5fa ka 1fw kw 1fs ks (1)
and
1
ke 5 fa fw fs
; (2)
ka 1 kw 1 ks
where ka, kw, and ks are the thermal conductivities, and fa, fw, and fs are the volume fractions of the air,
water, and solid phases, respectively (fa 1 fw 1 fs 5 1). Note that if the medium is completely dry, then
fw 5 0. Equation (1), the weighted arithmetic mean, is the upper bound, and equation (2), the weighted har-
monic mean, is the lower bound for the effective thermal conductivity in composite materials [Bergman and
Stroud, 1992; Wang et al., 2006].
Besides the weighted arithmetic and harmonic means, other kinds of weighted means have been also
applied to model ke. de Vries [1963] assumed that the effective thermal conductivity of a partially saturated
porous medium ke can be expressed as the weighted sum of the thermal conductivities of water, air, and
solid particles as
ka fa ka 1kw fw kw 1ks fs ks
ke 5 : (3)
ka fa 1kw fw 1ks fs
The ks are constant coefficients determined empirically in the de Vries theory [Campbell et al., 1994]. Wood-
side and Messmer [1961] found that the de Vries and the weighted geometric mean (ke 5k/a k12/ s in which
/ 5 1 – fs is the total porosity) models predicted the porosity dependence of the dry thermal conductivity
more accurately than other models in unconsolidated (quartz sand, glass bead, and lead shot packs) and
consolidated (porous sandstones) media. The weighted geometric mean followed by Sass et al. [1971]
among others yields a reasonable estimate of thermal conductivity when the thermal conductivities of the
individual conducting phases (e.g., water and solid) do not contrast by more than one order of magnitude
[Woodside and Messmer, 1961]. The weighted geometric mean has also been used to estimate the thermal
conductivity of the solid phase ks when mineral composition is known [see e.g., Johansen, 1975; Co ^te and
Konrad, 2005].
Sepaskhah and Boersma [1979] evaluated the de Vries [1963] model using three soils (loamy sand, loam, and
silty clay) at two temperatures (25 and 458C). Their results indicated that although the predicted thermal
conductivity values were in good agreement with measurements, improvements were needed to account
for the enhancement of vapor transfer that occurred at 458C in medium to fine-textured soils. In addition to
vapor transfer, latent heat transfer might be also relevant since its value becomes a significant portion of
total heat transfer at relatively low water contents and high temperatures [Tarnawski and Leong, 2000].
Sepaskhah and Boersma [1979] also found that there is a water content threshold, below which thermal con-
ductivity becomes independent of water content that is a function of clay content and soil temperature.
The concept of the critical volume fraction is well known as the percolation threshold within percolation
theory, the effective-medium approximation, and the percolation-based effective medium approximation
(P-EMA) to be introduced in section 2.
Campbell et al. [1994] modified the de Vries [1963] model and proposed continuous functions for k applying
over the full range of water contents (see their equations (5)–(7)) by assuming particles are ellipsoidal with
one shape factor, which needs to be found empirically. Their new model describes thermal conductivity as
a function of bulk density, temperature, and water content, and has four parameters: thermal conductivity
of the solid phase ks, a threshold water content (at which water starts to affect thermal conductivity), an
exponent for the liquid flow function, and a shape factor. Campbell et al. [1994] found that their modified de
Vries model agreed well with measured data at low temperatures, and provided an upper boundary for
measurements at high temperatures.

Co^te and Konrad [2005] proposed the following equations to model thermal conductivity in dry (kdry), par-
tially (k(S)), and fully (ksat) saturated porous media
kdry 5v102g/ (4a)
kðSÞ2kdry jS
5 (4b)
ksat 2kdry 11ðj21ÞS
ksat 5k12/
s k/w (4c)
where v, g, and j are fitting parameters, / is porosity, and S is degree of saturation (5 h=/ in which h is vol-
^te and Konrad [2005] use the geometric mean for the water-saturated
umetric water content). Note that Co
condition, where the contrast between ks and kw is lower.
Beside the Co^te and Konrad [2005] model, Lu et al. [2007], and Lu et al. [2014] proposed exponential equa-
tions to describe the saturation dependence of thermal conductivity. Although the Lu et al. [2014] model is
empirical, Lu and her coworkers demonstrated that their model could predict the saturation-dependent
thermal conductivity at room temperature from kdry and two empirical parameters reasonably well. In their
approach, kdry is estimated from soil porosity and the empirical parameters are determined from sand and
clay contents and bulk density.
In addition to empirical and quasiphysical equations, theoretical models based on fractal approaches have
been also developed in the past decade to characterize the thermal conductivity. Yu and Cheng [2002] pro-
posed two fractal models for thermal conductivity modeling in bidisperse porous media. Based on a statisti-
cal self-similarity concept, Feng et al. [2004] proposed a fractal-based model for thermal conductivity by
assuming that a porous medium consists of two portions: (1) randomly distributed nontouching particles,
and (2) self-similarly distributed particles contacting each other with contact resistance. Feng et al. [2004]
found good agreement with experiments, but their model has many parameters (e.g., porosity, total cross-
sectional area of a unit cell, equivalent cross-sectional area having the same porosity as the nontouching
particles, ratio of solid thermal conductivity to fluid thermal conductivity, contact thermal resistance, and
two microstructural parameters).
Kou et al. [2009] proposed an expression for thermal conductivity based on a thermal-electrical analogy and
statistical self-similarity. They found that when the thermal conductivity of the solid ks and water kw phases
are greater than that of the air phase ka, the effective thermal conductivity of partially saturated fractal
porous media decreases with decreasing volumetric water content h (5 fw) and increasing fractal dimension
of pore space, tortuosity fractal dimension, and porosity. However, their model does not address the sharp
increase in thermal conductivity observed in experiments at small water contents (near the critical volume
fraction hc ). In addition, Kou et al. [2009] model produces a positive curvature in the k-h relationship (see
their Figures 3–5), in contrast to experimental evidence [Campbell et al., 1994; Singh and Devid, 2000; Singh
et al., 2001; Lu et al., 2007, 2014], and theoretical results [Ewing and Horton, 2007].
The effective-medium and differential effective-medium approximations have also been applied to model
thermal conductivity in porous materials [see e.g., Zimmerman, 1989]. There are two models based on
effective-medium theory initially used to determine effective transport coefficient e.g., conductivity of com-
posites and mixture: (1) Maxwell-Garnett [1904] self-consistent model, and (2) Bruggeman [1935] model.
However, both models include underlying assumptions. For example, the Maxwell-Garnett model is valid
under dilute conditions. In what follows, we demonstrate that the Bruggeman’s model is a special case of a
more generalized effective-medium approximation.
Regarding the practical application of the effective-medium approximation, Davis and Artz [1995] investi-
gated the thermal conductivity of metal-matrix composites using the effective-medium approximation and
finite-element technique and found good agreement between theory and simulations in two dimensions.
Wang et al. [2003] proposed a model for the thermal conductivity of nanofluids based on the effective-
medium approximation and fractal theory characterizing nanoparticle clusters and their distribution. They
found that their model predicted the thermal conductivity quite well for dilute suspensions of metallic
oxide nanoparticles. Gao and Zhou [2006] took into account the effects of physical anisotropy (originating
from the interfacial thermal resistance) and geometrical anisotropy (resulting from the large aspect ratio of

nanotubes) and presented a new differential effective-medium theory to estimate thermal conductivity in
nanofluids. Their theoretical results on the thermal conductivity of carbon nanotube/oil nanofluids matched
reasonably well with experiments. McLachlan [1987] proposed a generalized percolation-based effective-
medium approximation (P-EMA) reducing to the traditional effective-medium approximation model and
the Bruggeman model under special conditions as we discuss in section 2.3. However, McLachlan and his
coworkers [Deprez et al., 1988] evaluated their model for a series of sintered nickel samples, not natural rocks
and sediments.
To the authors’ knowledge, the general theoretical P-EMA approach proposed by McLachlan [1987] has
never been applied to natural porous media, such as sediments, rocks, and soils. Therefore, the main objec-
tives of this study are: (1) extending the P-EMA approach to describe the porosity and saturation depend-
ence of the thermal conductivity in rocks and soils, and (2) evaluating the proposed P-EMA based models
using experiments reported in the literature. The approach that we develop in this paper is, however, quite
general and applicable to any type of porous medium.
2. Theory
In this section, we first introduce concepts of percolation theory and the effective-medium approximation.
Then, we invoke the general P-EMA developed by McLachlan [1987] to model thermal conductivity under
fully and partially saturated conditions in porous media.
2.1. Percolation Theory

Percolation theory provides a theoretical approach to study conductivity in networks and porous media. It
may be mapped into two frameworks: insulator-conductor [see e.g., Stauffer and Aharony, 1994], and
conductor-superconductor [see e.g., Clerc et al., 1990; Bergman and Stroud, 1992]. In the former, the network
is a mixture of insulating and conducting phases with zero and finite conductivities. In the latter, the
conductor-superconductor mixture contains finite-conductivity and infinite-conductivity components.
For the sake of simplicity, consider a regular insulator-conductor lattice in which randomly occupied bonds
between neighbors are conductors (allowing flow) with probability p. A key feature in such a lattice is the
presence of a critical occupation probability pc at which a cluster spanning the network forms. Below the
threshold pc, all formed clusters are finite in size, and there is no spanning (infinite) cluster. Thus, the net-
work does not percolate, and there is no macroscopic conductivity. Alongside the many finite clusters, at pc
an infinite spanning cluster emerges, the mass of which scales as LDm , in which L is the lattice size and Dm is
the mass fractal dimension [Ben-Avraham and Havlin, 2000]. Above pc there are still finite clusters along
with the infinite cluster. If the lattice size L is less than the correlation length (the mean distance between
two sites on the same finite cluster), the lattice is heterogeneous, statistically self-similar, and thus fractal,
while for L greater than the correlation length the lattice is macroscopically homogeneous and its geometry
Euclidean [Ben-Avraham and Havlin, 2000]. Above but near pc, conductivity T is independent of the structure
and geometry of the network (or porous medium), a feature called universality, and conforms to the follow-
ing power law scaling
T / ðp2pc Þt ; p > pc (5)
where t is a scaling exponent whose universal value is 1.3 in 2-D, and 2 in 3-D [Stauffer and Aharony, 1994]
in networks constructed of one insulating phase and one conducting phase. The value t 5 2 incorporates
the effects of the connectivity and tortuosity [Ghanbarian et al., 2013]. Although equation (5) is theoretically
valid in the region above but near the percolation threshold pc, experimental results indicate that it is valid
over a much broader region above pc in irregular and complex lattices, such as porous media [Ghanbarian-
Alavijeh and Hunt, 2012; Hunt et al., 2014a, 2014b; Ghanbarian and Hunt, 2014; Ghanbarian et al., 2014; Ghan-
barian et al., 2015a, 2015b]. Note that p in equation (5) is the bond occupation probability in a lattice. One
can replace p and pc by volumetric water content h and critical volumetric water content hc in equation (5)
to describe saturation dependence of the conductivity T in a porous medium [Hunt et al., 2014b]. Within the
conductor-superconductor framework, the conductivity T would follow a power law similar to equation (5),
but the exponent (typically denoted by s) has a universal value of 0.76 in 3-D [Bergman and Stroud, 1992]
and may range between 0.37 and 1.30 [see e.g., Youngs, 2002; McLachlan et al., 2007; Chiteme et al., 2007;
and references in therein].

We caution that the value of t within continuum percolation theory might be nonuniversal and greater than
2, in particular when the conductance distribution (or equivalently the pore throat-size distribution for per-
meability and/or conductivity through pore space) is very broad [see e.g., Feng et al., 1987]. More recently,
Ewing and Hunt [2006] indicated that the exponent t for electrical conductivity is near the universal value 2
over the entire range of saturation. However, for media with very broad pore throat-size distributions, one
should expect t > 2. Based on critical path analysis, Ewing and Hunt [2006] proposed a method relating the
nonuniversal exponent t to pore space fractal dimension. They showed how the width of the conductance
distribution (or equivalently the pore throat-size distribution) relates to the transition from universal to non-
universal values of t. Sahimi [1994a] and Sahimi and Mukhopadhyay [1996] showed that long-range correla-
tion in the porous medium would result in t < 2.
Regarding the practical application of equation (5), Bonnet et al. [2007] experimentally investigated porosity
dependence of thermal conductivity of nanocomposites near the percolation threshold, and found t 5 1.96
(very close to the universal value) and pc 5 0.003. Foygel et al. [2005] studied thermal conductivity for ran-
domly oriented nanotubes with aspect ratio spanning more than two orders of magnitude using Monte
Carlo simulations. They found that t 2 for spherocylindrical tubes with aspect ratio near 1, but that t
decreases as the aspect ratio increases. For 3-D systems of randomly distributed and randomly oriented
nanotubes with aspect ratios of the order of 1022103, Foygel et al. [2005] reported t 5 1.2–1.6, which is near
1.3, the universal value for 2-D systems. We should point out that although Foygel et al.’s [2005] systems are
three-dimensional, the thermal conduction is essentially two-dimensional because of the very large aspect
ratios of spherocylindrical tubes. Similarly, Ghanbarian-Alavijeh et al. [2012] recently showed that two-
dimensional flow and invasion percolation were relevant to three-dimensional column with wall flow.
Saturation dependence of thermal conductivity is difficult to address within the context of percolation
theory [Hunt et al., 2014b], either in the insulator-conductor or the conductor-superconductor framework.
First, all three phases (air, water, and solid) have nonzero thermal conductivities, although the thermal con-
ductivity of air is typically negligible compared to that of the water and solid phases. Second, the ratio of
the thermal conductivity of water to solid (kw/ks) is not constant, but depends mainly on the mineralogy of
the solid particles [see Co^te and Konrad, 2005, Table 2], and thus varies from medium to medium. For exam-
ple, in organic soils the solid phase conductivity ks may be half of kw [Campbell et al., 1994; Hunt et al.,
2014b], while in pure quartz sandstone the solid phase conductivity may be as much as 15 times greater
than the water phase thermal conductivity [Hunt et al., 2014b]. Third, experimental measurements of
saturation-dependent thermal conductivity k(h) in porous materials [e.g., Singh and Devid, 2000; Singh et al.,
2001] support power law scaling i.e., equation (5) with t < 1. This results in a negative curvature of the k-h
relationship, which is not common within the percolation theory framework for random porous media in
which one expects t 2. We discuss t values less than unity in experimental results in the following as well
as the Results and Discussion sections.
More recently, Shao et al. [2008] investigated the percolation exponent t in insulator-conductor continuum
percolation systems. They assumed that the conductance g between two random nodes follows an expo-
nential equation (g5g0 e-xðMD =MB Þ , in which g0 is a constant, MD and MB are the mass of the dangling ends
(dead end parts of the sample-spanning cluster that carry no current) and of the backbone (the current-
carrying part of the sample-spanning cluster) belong to two randomly selected nodes, and x is a structure
parameter that decreases with increasing geometrical complexity of the conductor). They further assumed
that the distribution function of the dangling ends and the backbone follows an exponential trend
[f ðMD =MB Þ / e-gðMD =MB Þ in which g is hMD =MB i] and derived f ðgÞ / ½ln ðg=g0 Þgðg=x-1Þ . Shao et al. [2008]
replaced f ðgÞ by gðg=x-1Þ , which is valid for small g values, and used the Kogut and Straley [1979] model to
find theoretically t 5 2 1 ðx=g-1Þ 5 2 1 ðxhMD =MB i-1Þ, yielding t > 0 (see their Figure 1). The key step
within the Shao et al. [2008] derivation is assuming that g is very small, limiting their theoretical results to
some specific systems. Shao et al. [2008] fit equation (5) to DC (direct current) electrical conductivity experi-
ments in a Cu-Cu2O conductor-insulator medium and found t 5 0.87 and pc 5 0.23.
2.2. Effective-Medium Approximation

In contrast to percolation theory providing a general theoretical framework to calculate the effective con-
ductivity in both homogeneous and highly heterogeneous systems, the effective-medium approximation is
restricted to ordered and relatively disordered porous media with small local fluctuations. Furthermore, it is

well known in the literature that power laws from percolation theory are theoretically valid above and near
the percolation threshold, while the effective-medium approximation results are accurate far above the
threshold [Sahimi, 1994b, 2003, 2011]. The objective of the effective-medium approximation is to infer an
average conducting parameter for heterogeneous-disordered media from the statistics of local conducting
elements [David et al., 1990]. In the effective-medium approximation framework, a disordered medium is
replaced by a uniform network whose effective conductivity is equal to the true conductivity of the
medium. Replacing the heterogeneous medium with an equivalent homogeneous one causes local pertur-
bations due to the fact that the local conductivity is different from the effective equivalent conductivity.
The effective conductivity is calculated by forcing the average of these local perturbations to be equal to
zero over the medium [Kirkpatrick, 1973]:
ð Tmax
ðT2Te Þ
f ðTÞdT50 (6)
Tmin T1ðZ=221ÞTe
in which f(T) is the probability distribution of the local conductivity T, Te is the effective conductivity, Z is the
average coordination number [Jerauld et al., 1984; Lindquist et al., 2000], and Tmin and Tmax are the lower
and upper limits of the conductivity in the actual porous medium. As King [1989] states, the effective-
medium approximation results are reliable if the local fluctuations are small. Therefore, the effective-
medium approximation is not applicable to highly heterogeneous media with large local conductivity varia-
tion [Sahimi, 2003, 2011].
In a binary porous medium with low and high-conductivity components, equation (6) simplifies to
ðTl 2Te Þ ðTh 2Te Þ

fl 1ð12fl Þ 50 (7)
Tl 1ðZ=221ÞTe Th 1ðZ=221ÞTe
where fl is the volume fraction of the low-conductivity component, Tl is the conductivity of the low-
conductivity component, and Th the conductivity of the high-conductivity component. For a comprehensive
review of the applications of the effective-medium approximation, see Sahimi [2003, 2011].
2.3. Percolation-Based Effective-Medium Approximation (P-EMA)
By combining the concepts of percolation theory (equation (5)) and the two-component effective-medium
approximation (equation (7)), McLachlan [1987] proposed a general P-EMA as follows:
1=t 1=t 1=t 1=t

ðTl 2Te Þ ðTh 2Te Þ
fl 1=t 1=t
1ð12fl Þ 1=t 1=t
50 (8)
Tl 1½ð12fc Þ=fc Te Th 1½ð12fc Þ=fc Te
in which fc is the critical volume fraction below which Te is mainly controlled by the conductivity of the low-
conductivity component Tl. Above fc, it is, however, the high-conductivity component controlling Te.
This critical volume fraction can be conceptualized as a threshold above which the high-conductivity com-
ponent first forms a continuous percolation path across the medium [McLachlan, 1988]. As a consequence,
fc 51-fc’ in which fc’ is the critical volume fraction of the high-conductivity component (for further discussion
see dual lattice argument by Hunt et al. [2014b] p. 29). Note that 2/Z in the effective-medium approximation
equations (6) and (7) was replaced with fc in equation (8) because it is well known that 2/Z is not an accurate
estimate of the percolation threshold in three dimensions [Kirkpatrick, 1973; Koplik, 1981; Sahimi et al., 1983;
Sahimi, 1993]. Equation (8) reduces to equation (5) when the low-conductivity component coefficient is
zero (Tl 5 0), given that p 5 1 – fl and pc 5 fc’ . Equation (8) also reduces to Bruggeman’s symmetric-medium
equation when t 5 1 and fc 5 1/3.
Deprez et al. [1988] fit equation (8) to thermal conductivity data in a sintered nickel sample (ka 0 and
kNi 5 83 W m218C21) and found t 5 1.8 6 0.3 (near the theoretical universal value of 2) and
fc 5 0.939 6 0.06. Comparing with experimental measurements, we find t values different (less) than 2 for
porosity and saturation-dependent thermal conductivity in natural porous media, such as rocks and soils. In
the Results and Discussion sections, we report and discuss the calculated values and the results obtained
from natural rocks and soils in detail.

2.3.1. Porosity-Dependent Thermal Conductivity k(/)

When the medium is fully saturated, there exist two phases contributing to the thermal conductivity of
the medium: (1) fluid (e.g., water or air) filling all pores, and (2) solid matrix. Within the P-EMA framework,
following equation (8), the porosity dependence of the thermal conductivity k(/) is given by
1=tp
ðkf 2k1=tp ð/ÞÞ ðk1=t
s 2k
p 1=tp
ð/ÞÞ
/ 1=tp
1ð12/Þ 1=tp
50 (9a)
kf 1½ð12/c Þ=/c k1=tp ð/Þ k1=t
s 1½ð12/c Þ=/c k
p
ð/Þ
or equivalently
1=tp 1=tp 1=tp
½k ð/Þ2k1=t
s ½ð12/c Þkf
p
1/c k ð/Þ
/5 1=t 1=tp
(9b)
½kf p 2k1=t
s k
p
ð/Þ
where kf is the saturating fluid thermal conductivity, tp is the scaling exponent, and /c is the critical volume
fraction of the fluid phase (5 1- /c’ in which /’c is the critical volume fraction of the solid phase).
The percolation threshold /c not only depends on the microstructure and geometry of the medium, but
also depends on its topology and connectivity [e.g., Sahimi, 2011], the fluid saturating it [e.g., Ghanbarian
et al., 2015b], and the porous medium size (thickness) [e.g., Wilkinson and Willemsen, 1983]. Recently, Ghan-
barian et al. [2015b] demonstrated that the percolation threshold for diffusion processes in the same porous
medium was different for various saturating fluids. They stated that the percolation threshold for transport
in a water-saturated medium should be less than that of the same medium saturated by air. This means
that more volume fraction is required to form a pathway spanning the medium as contact angle changes
from nonwetting to wetting conditions. Therefore, one should expect that the percolation threshold
derived from equation (9) might be different for various fluids saturating the same porous medium.
2.3.2. Saturation-Dependent Thermal Conductivity k(h)
Thermal conductivity in partially saturated porous media k(h) is influenced by three conducting phases: air,
water, and solid. However, kdry(/) integrates the contributing effects of solid and air phases, and ksat(/)
solid and water ones. We therefore consider air and solid as a unified low-conductivity component and
water as the high-conductivity one. Following P-EMA, we propose
1=t
ðkdry s ð/Þ2k1=ts ðhÞÞ 1=t
ðksat s ð/Þ2k1=ts ðhÞÞ
ð/2hÞ 1=t
1h 1=t
50 (10a)
kdry s ð/Þ1½ð/2hc Þ=hc k1=ts ðhÞ ksat s ð/Þ1½ð/2hc Þ=hc k1=ts ðhÞ
or equivalently
1=t 1=t
½k1=ts ðhÞ2kdry s ð/Þ½hc ksat s ð/Þ1ð/2hc Þk1=ts ðhÞ
h5 1=t 1=t
(10b)
½ksat s ð/Þ2kdry s ð/Þk1=ts ðhÞ
where kdry(/) and ksat(/) are the thermal conductivity under fully saturated conditions when the saturating
fluid is air and water, respectively, ts is the scaling exponent, and hc is the critical water content at which the
high-conductivity component first forms a continuous percolation path. The concept of the critical water
content in our P-EMA is similar to the residual water content commonly used in soil physics and hydrology.
The main difference between the two is that below the residual water content there is no macroscopic
transport, while below the critical water content in our approach thermal conductivity still occurs but
mainly through the low-conductivity component.
For the case that t 5 1, equation (10) reduces to the weighted arithmetic mean when hc 5 0, to the weighted
harmonic mean when hc 5 /, and to the weighted geometric mean when hc 5 h 5 12 /.
3. Materials and Methods

To the authors’ knowledge, the P-EMA models, equations (9) and (10), have never been presented to
describe the porosity and saturation dependence of the thermal conductivity in natural porous media. In
the following, we therefore evaluate our proposed models (equations (9) and (10)) by fitting each equation
to experimental measurements, and report and discuss model parameters, such as the percolation thresh-
old and the scaling exponent.

^te and Konrad [2005] Database

3.1. Co
This database includes 119 measurements of porosity-dependent thermal conductivity in mineral soils,
and 74 in crushed rocks, collected from the literature. In mineral soils porosity ranges between 0.25 and
0.7, and measured dry thermal conductivity ranges from near 0.1 to 0.5 (W m218C21). In crushed rocks,
porosity ranges from near 0.1 to 0.6, and dry thermal conductivity varies between 0.1 and 1 (W
m218C21).
3.2. Woodside and Messmer [1961] Database

The Woodside and Messmer [1961] database consists of porosity-dependent thermal conductivities of six
sandstones ranging in porosity from 0.03 to 0.59 cm3 cm23, permeability from <0.1 to 1960 md, and quartz
content from 85 to 99%. Physical and hydraulic properties of each sample were presented in more detail by
Woodside and Messmer in their Table 1. Thermal conductivity was measured under completely dry as well as
fully saturated (using n-heptane and water) conditions on each sandstone.
3.3. Sepaskhah and Boersma [1979] Database

Sepaskhah and Boersma [1979] measured the saturation dependence of thermal conductivity in one loamy
sand, one loam, and one silty clay loam soil sample at two temperatures: 25 and 458C. Sand fraction ranged
from 9 to 90%, and bulk density from 1.2 to 1.7 g cm23. The samples were prepared by adding the amount
of water required to bring a packed and air dry soil sample to a desired water content [Sepaskhah and
Boersma, 1979]. Other physical properties of the sample were presented in Sepaskhah and Boersma [1979],
Table 2.
3.4. Lu et al. [2014] Database

The Lu et al. database includes saturation dependence of the thermal conductivity of 17 packed soils from
Lu et al. [2007, 2011, 2013]. It consists of a wide range of soil textures from clay loam to sand (see Table 2).
Sand fraction ranged from 7 to 94%, and bulk density 1.20–1.60 g cm23. The organic matter varied between
0.07 and 3.02%. Various soil water contents were obtained by adding a known amount of water to the air-
dried soil samples after grinding and passing through a 2 mm sieve [Lu et al., 2007, 2014].
3.5. Dong et al. [2015] Database

Dong et al. [2015] reported the saturation dependence of thermal conductivity for one sand, one silt, and
one clay soil. However, due to uncertainties in the measurements, we do not examine the sandy soil experi-
ment. The porosity of the silt was 0.39 and that of the clay was 0.5 (cm3 cm23). No information is available
whether thermal conductivity was measured under saturating or desaturating conditions.
For more information regarding each database, the interested reader is referred to the original published
articles. Equations (9b) and (10b), rearranged to derive porosity / and volumetric water content h as a func-
tion of other parameters, such as thermal conductivity, critical volume fraction, etc. were used to fit our
P-EMA models to the corresponding experimental data. The fitting process was performed using CurveFit-
ting toolbox of MATLAB [The MathWorks, Inc., 2013].
4. Results
Comparison with experiments showed that the theoretical models of the thermal conductivity (equations
(9) and (10)) match the measurements under fully and partially saturated conditions well. We first investi-
gate the sensitivity of the proposed models to their parameters, such as the threshold and the exponent t,
and then discuss the obtained results from comparison with the experimental data in the Results and Dis-
cussion sections.
Figure 1 shows the sensitivity of the porosity-dependent thermal conductivity model (equation (9)) to the
critical porosity /c and the exponent tp values. We set ka 5 0.024 W m218C21 (the air thermal conductivity
at 258C), ks 5 2.4 W m218C21 (a typical thermal conductivity value in soils) and varied 0 /c 0.9 and 0.05
tp 2. Our results presented in Figure 1a indicate that the critical porosity value may change the concav-
ity of the porosity dependence of the thermal conductivity, in agreement with the weighted arithmetic and
harmonic means yielding upper (convex) and lower (concave) bounds, respectively. We found that /c 5 0.1
was near the inflection point—separating the concave behavior from the convex one—and close to / 5 1,
and concave thermal conductivity behavior becomes dominant; in contrast, the thermal conductivity

Figure 1. Thermal conductivity as a function of porosity for different values of the critical porosity /c and the critical exponent tp in equa-
tion (9).
behavior for /c 5 0.9 is mainly convex. Therefore, as expected the critical porosity has an important effect
on the shape of the porosity-dependent thermal conductivity. The value of tp also controls the shape of the
curve. Figure 1b shows thermal conductivity as a function of porosity when tp varies between 0.05 and 2
(the universal value for 3-D systems). As tp decreases from 2 to 0.5, the inflection point and the critical
porosity become more distinct, particularly for tp < 1. We should caution that the inflection point might be
greater/less than or equal to /’c , depending on the value of tp. For example, for tp 5 0.05, /c 5 0.3,
ks 5 0.024 W m218C21 and ka 5 2.4 W m218C21 the volume fraction at the inflection point /inf 5 0.7 and
equal to /’c 5 0.7. However, increasing tp from 0.05 to 0.5 results in /inf 5 0.683, which is less than /’c .
We found similar results for the saturation dependence of the thermal conductivity (see Figure 2). We set
kdry 5 0.5 W m218C21 and ksat 5 2.5 W m218C21 and changed hc between 0 and 0.2 (cm3 cm23) and ts
between 0.05 and 2. We found that the saturation-dependent thermal conductivity is relatively sensitive to
hc , in particular at low water contents. As can be seen in Figure 2a, for ts 5 0.25, kdry 5 0.5 W m218C21, and
ksat 5 2.5 W m218C21 as hc increases the inflection point increases as well. At hc 5 0, there exist no inflection
point and the k2h curve becomes concave over the entire range of water content. For other hc values, we
found the water content at the inflection point hinf slightly greater than the critical water content hc . Similar
to the porosity-dependent results, the smaller ts values (ts < 1) yield more distinct inflection point and criti-
cal volumetric water content (see Figure 2b). As shown in Figure 2b, for kdry 5 0.5 W m218C21, ksat 5 2.5 W
m218C21, and hc 5 0.1, the green curve representing ts 5 2 does not display any inflection point, and the
curve is convex through the whole range of water content. We found the smaller the ts value, the sharper
the rise in the thermal conductivity at the critical water content hc (consistent with Figure 1b). We empha-
size that the water content at the inflection point hinf might be greater/less than or equal to the critical
water content hc , depending on the ts value.
4.1. Porosity-Dependent Results

^te and Konrad [2005] reported the porosity dependence of the thermal conductivity for 119 mineral soils
Co
and 74 crushed rocks collected from the literature. We reanalyzed these two databases and fit equation (9)
to determine ka, ks, /c , and tp values. As can be seen in Figure 3, the mineral soil thermal conductivity val-
ues are more scattered compared to the crushed rock ones. Simultaneous optimization of all four parame-
ters of equation (9), e.g., ka, ks, /c , and tp, resulted in an unrealistically large ks value. Setting ka 5 0.024 W
m218C21 (the nominal air thermal conductivity at 258C) and optimizing ks, /c , and tp concurrently also
yielded unrealistic ks and large tp values. Therefore, we set ks 5 2.9 W m218C21 (a typical value for most soil
minerals with the exception of quartz) [Lehmann et al., 2003] and optimized ka, /c , and tp values. Our results
presented in Figure 3 show equation (9) fit the measured mineral soils data relatively well (R2 5 0.84). The
scatter in the mineral soils data is due to the fact that soils’ mineralogical composition is typically very
heterogeneous. In particular, the presence of quartz may substantially affect the thermal conductivity in
soils since the thermal conductivity of quartz is considerably greater than that of other minerals. Fitting
equation (9) to the mineral soils yielded ka 5 0.083 W m218C21, /c 5 0.798, and tp 5 0.807.

Figure 2. Thermal conductivity as a function of water content for different values of the critical water content hc and the scaling exponent
ts in equation (10).
The calculated air thermal conductivity (0.083 W m218C21) is almost three times greater than the nomi-
nal air thermal conductivity at 258C (0.024 W m218C21). This discrepancy might be due to lack of meas-
urements at large porosities; the largest porosity value of the samples is 0.67. We note that Co ^te and
Konrad [2005] also reported thermal conductivity values in a peat soil with 96% porosity that were twice
that of air. They state, ‘‘This is probably caused by the existence of preferential heat flow along continu-
ous peat fibers, which contributes to maintaining the thermal conductivity at a fairly high level, given
that air occupies 96% of the total volume.’’ In addition to the Co^te and Konrad [2005] statement, a likely
explanation is that the thermal conductivity measurements may have not been performed under com-
pletely dry conditions. Note that a small amount of sorbed water that would be mobilized when a soil
sample is heated may augment the measured thermal conductivity by latent heat transfer. Another pos-
sibility is that measuring thermal conductivity by means of conductive currents may result in a higher
apparent conductivity.
For mineral soils, we found / 5 0.798 (/’ 5 0.212). This means at least 0.212 solid fraction is required to
c c
form a continuous percolation path for heat flow along the solid particles. The value of 0.212 is fairly close
to the continuum percolation threshold of 0.289 in overlapping, spatially uncorrelated spheres [Rintoul and
Torquato, 1997; Consiglio et al., 2003], and is within the range of 0.200 and 0.285, which are the continuum
percolation thresholds for randomly oriented, overlapping ellipsoidal particles with aspect ratio between
0.25 (oblate ellipsoid) and 4 (prolate ellipsoid) [Garboczi et al., 1995].
The value of tp 5 0.807 found for the mineral soils (see Figure 3) is substantially less than the universal value
t 5 2 (insulator-conductor mixture), but near 0.76 (conductor-superconductor mixture). Within the insulator-
Figure 3. Thermal conductivity as a function of porosity in completely dry mineral soils and crushed rocks (kf 5 ka in equation (9)). The
measured data are from C^
ote and Konrad [2005]. The solid line represents the fitted P-EMA model (equation (9)) developed in this study.

conductor percolation theory framework for all three percolation classes (site, bond, and continuum), only
one phase (conductor) of two (conductor and insulator) contributes to flow or transport. Given that the ratio
of the insulating phase conductivity to the conducting phase conductivity tends to zero, percolation theory
returns theoretically t 2 in uncorrelated and short-range correlated systems [see e.g., Sen et al., 1985; Feng
et al., 1987; Halperin, 1989]. In the conductor-superconductor framework, although both the conducting
and superconducting components conduct, the ratio of the conducting component conductivity to the
superconducting component conductivity tends to zero as well, similar to the insulator-conductor frame-
work. The universal scaling exponent is, however, 0.76 [Bergman and Stroud, 1992] and may vary between
0.37 and 1.3 [Youngs, 2002]. In the case of thermal conductivity in porous media under completely dry con-
ditions, two phases (air and solid) conduct heat, and the ratio of the air thermal conductivity to that of solid
particles is finite and nonzero. Therefore, the porosity-dependent thermal conductivity fits neither into the
insulator-conductor nor to the conductor-superconductor framework. As a result, one should expect tp to
depend on the ks/ka ratio. McLachlan [1987] also states that the value of tp in the P-EMA model may depend
on the shape of the particles and the structural orientation of the medium e.g., partially oriented or random.
Since the porosity-dependent thermal conductivity mainly occurs through the solid matrix of the porous
medium, the scaling exponent tp might be correlated with the fractal dimension of the grain-size distribu-
tion. Investigating these hypotheses requires accurately measured grain-size distributions that are not avail-
able for the experiments used in this study.
Figure 3 also shows the results found for the crushed rocks. By directly fitting equation (9) to the experi-
ments, we found ka 5 0.076 W m218C21, ks 5 1.528 W m218C21, /c 5 0.685, and tp 5 0.862 with R2 5 0.94.
Similar to the mineral soils, the calculated air thermal conductivity is three times greater than the nominal
air thermal conductivity at 258C (0.024 W m218C21). The reason we found /’c in the crushed rocks to be
greater than that in mineral soils (0.315 versus 0.212) might be because the latter consist of grains of more
variable sizes and shapes, e.g., finer platy clay and coarser spherical silt and sand particles. In contrast, the
grains in crushed rocks are less diverse and typically more angular. Given that finer particles can fit within
void space between larger particles and connect them to conduct heat through solid phase, one may
expect a smaller threshold for thermal conductivity in media with broader particle-size distribution. This,
however, has to be confirmed either experimentally or numerically. The fitted results are in agreement with
those reported by Garboczi et al. [1995] regarding the important effect of particle shape on the continuum
percolation threshold. In addition to the effect of particle shape, the mineralogy of crushed rocks is different
from that of mineral soils and therefore should result in different critical porosity values.
The value tp 5 0.862 derived from crushed rocks is not greatly different from that calculated from mineral
soils (tp 5 0.807), and still near the universal scaling exponent of 0.76 in three dimensions. As we discussed
earlier, porosity-dependent thermal conductivity does not fit perfectly into either insulator-conductor or
conductor-superconductor percolation theory framework, since more than one conducting phase contrib-
utes finitely to heat flow and thermal conductivity.
One should bear in mind that the fitted parameters tp, /c , and ks found in mineral soils and crushed rocks
databases are not applicable to all soils or rocks, particularly those with mineralogical characteristics sub-
stantially different than samples investigated here. Therefore, one should not expect equation (9) with tp,
/c , and ks values optimized in this study to characterize accurately the porosity dependence of thermal con-
ductivity in different porous media with various mineralogical properties and pore and solid structures.
4.2. Effect of Saturating Fluid on the P-EMA Model Parameters

Woodside and Messmer [1961] measured thermal conductivity under fully saturated (using air, n-heptane,
and water fluids) conditions in sandstones. We reanalyzed these experiments by fitting equation (9) to fully
saturated measurements (see Figure 4) and summarized the fitted results in Table 1.
For completely dry conditions in which air saturates the medium, theory (equation (9)) matched the experi-
ments excellently (R2 5 0.98). We found ka 5 0.179 W m218C21, ks 5 7.7 W m218C21, /c 5 0.789, and
tp 5 1.698. The air thermal conductivity ka 5 0.179 W m218C21 is still greater than its known value at 258C,
most probably due to lack of measurements at high porosity values (maximum measured / is 0.59). How-
ever, our fitted ks 5 7.71 W m218C21 is very close to the thermal conductivity of pure quartz (7.69 W
m218C21) [see Co ^te and Konrad, 2005, Table 2], which is the main component in sandstones. Our fitting
results indicated that /’c 5 0.211, compatible with the continuum percolation thresholds reported in the

literature [see e.g., Garboczi et al., 1995].

The value tp 5 1.698 is nearly twice as
large as the values we reported in mineral
soils (tp 5 0.807) and crushed rocks
(tp 5 0.862), and is much closer to univer-
sal scaling exponent t 5 2.
For fully saturated conditions when n-
heptane fills all pores, we found
kh 5 0.303 W m218C21 (n-heptane ther-
mal conductivity), ks 5 7.9 W m218C21,
/c 5 0.663, and tp 5 1.556 with R2 5 0.99.
In agreement with our recent results, the
determined thermal conductivity of n-
heptane (0.303 W m218C21) is almost
twice as large as the n-heptane thermal
conductivity at 258C (0.129 W m218C21).
Figure 4. Thermal conductivity as a function of porosity for sandstones from
Woodside and Messmer [1961]. The solid line represents the fitted P-EMA
The value ks 5 7.9 W m218C21 is, how-
model (equation (9)) developed in this study. The results of the fitting process ever, very close to the thermal conductiv-
are summarized in Table 1. ity of the mineral quartz. We found the
critical volume fraction for solid phase
/c’ 5 0.337. Comparing this value with that (0.211) determined under completely dry conditions (air is the
saturating fluid) confirmed the recent results of Ghanbarian et al. [2015b] who suggested that the percola-
tion threshold is not only a function of the microstructure of the medium but also the characteristics of the
fluid through which transport occurs.
For water-saturated condition, we found kw 5 1.252 W m218C21 (water thermal conductivity), ks 5 7.8 W
m218C21, /c 5 0.683, and tp 5 0.611 with R2 5 0.99 (see Table 1). Although more or less similar t values
were found for air and n-heptane (see Table 1), the value tp 5 0.611 determined for water experiments is
remarkably smaller than those found for air (1.698) and n-heptane (1.556). Our results summarized in Table
1 indicate that as the contact angle increases the percolation threshold /c decreases from 0.789 to 0.683
and the percolation exponent tp increases from 1.698 to 0.611. This is consistent with the results of Ghan-
barian et al. [2015b].
As we discussed earlier, the value of tp is affected by the ratio of the thermal conductivities of the low and
high-conductivity components. As Table 1 shows, the value of tp changes from 1.698 to 0.611 when kf/ks (kf
is the fluid e.g., air/heptane/water thermal conductivity) increases from 0.023 to 0.161. Note that in the
standard percolation theory problems consisting of conductor and insulator, only the conducting phase
contributes to flow because the resistivity of the insulating phase is infinitely large. Consequently, the ratio
of the insulating phase conductivity to the conducting phase conductivity tends to zero which results in t
values greater than 2 in uncorrelated and short-range correlated systems. Strictly speaking, one may expect
t value less than two in composite media in which two fluid and solid phases contribute to flow.
4.3. Saturation-Dependent Results

4.3.1. Lu et al. [2014] Database
We first analyze the Lu et al. [2014] database including saturation-dependent thermal conductivity of 17
soils. The optimized parameters of equation (10) the kdry, ksat, ts, and hc for each experiment, are presented
Table 1. Parameters of the Percolation-Based Effective-Medium Approximation Model (Equation (9)) for the Experiments of Woodside
and Messmer [1961]
Fluid kfa (W m218C21) ks (W m218C21) kf/ks tp /c /’c R2
Air 0.179 7.741 0.023 1.698 0.789 0.211 0.98
n-Heptane 0.303 7.916 0.039 1.556 0.663 0.337 0.99
Water 1.252 7.824 0.161 0.611 0.683 0.317 0.99
a
kf denotes the thermal conductivity of the saturating fluid. For example, for air kf 5 ka.

Table 2. Soil Physical Properties and Parameters of the Percolation-Based Effective-Medium Approximation Model (Equation (10)) for the Experiments of Lu et al. [2014]
Soil No. Texture Sand (%) Silt (%) Clay (%) BDa (g cm23) OM (%) kdry (W m218C21) ksat (W m218C21) ts hc R2 Reference
1 Sand 94 1 5 1.60 0.09 0.253 2.186 0.336 0 0.98 Lu et al. [2007]
2 Sandy loam 67 21 12 1.39 0.86 0.231 1.676 0.307 0.037 0.99 Lu et al. [2007]
3 Loam 40 49 11 1.20 0.49 0.215 1.352 0.329 0.032 0.97 Lu et al. [2007]
Loam 40 49 11 1.30 0.49 0.202 1.617 0.341 0.031 0.99 Lu et al. [2007]
Loam 40 49 11 1.40 0.49 0.203 1.610 0.340 0.032 0.99 Lu et al. [2007]
4 Silt loam 27 51 22 1.33 1.19 0.225 1.401 0.256 0.089 0.97 Lu et al. [2007]
5 Silty clay loam 19 54 27 1.20 0.39 0.227 1.366 0.369 0.090 0.99 Lu et al. [2007]
Silty clay loam 19 54 27 1.30 0.39 0.241 1.428 0.351 0.087 0.99 Lu et al. [2007]
Silty clay loam 19 54 27 1.40 0.39 0.305 1.466 0.254 0.102 0.99 Lu et al. [2007]
6 Silty clay loam 8 60 32 1.30 3.02 0.243 1.337 0.289 0.099 0.98 Lu et al. [2007]
7 Clay loam 32 38 30 1.29 0.27 0.222 1.367 0.279 0.101 0.99 Lu et al. [2007]
8 Sand 93 1 6 1.60 0.07 0.301 2.129 0.313 0.006 0.99 Lu et al. [2007]
9 Sand 94 1 5 1.60 0.09 0.272 1.735 0.337 0 0.99 Lu et al. [2013]
10 Sand 92 7 1 1.58 0.6 0.229 2.283 0.324 0.028 0.98 Lu et al. [2007]
11 Loam 50 41 9 1.38 0.25 0.225 1.769 0.296 0.036 0.99 Lu et al. [2007]
12 Silt loam 11 70 19 1.31 0.84 0.287 1.587 0.291 0.071 0.96 Lu et al. [2007]
13 Silty clay 7 50 43 1.29 2.09 0.163 1.001 0.225 0.135 0.97 Lu et al. [2011]
a
BD denotes bulk density, and OM is organic matter.
in Table 2. The high R2 values (R2 > 0.96) indicate excellent match between theory (equation (10)) and the
experiments. Figure 5 shows graphically equation (10) fit to six experiments (soils 8–13). We did not find
any specific trend between kdry and soil texture (e.g., clay, silt, or sand content). In addition, similar to empir-
ically derived relationships between kdry and bulk density [e.g., Johansen, 1975] or porosity [e.g., Co ^te and
Konrad, 2005; Lu et al., 2007, 2014], we found that kdry was weakly correlated with bulk density and/or
porosity (R2 5 0.22) for the Lu et al. [2014] database. This is mainly because mineralogical characteristics
might be greatly different from one soil to another. In fact, in addition to porosity there exist other factors
at play e.g., solid matrix mineralogy, connectivity in solid phase and pore space, temperature, small amount
of water adsorbed on solid particles even under dry conditions, etc. which affect kdry measurement. Note
that in our P-EMA framework, the porosity-dependent thermal conductivity kdry is not a simple function of
porosity (see equation (9)). It also depends on ka, ks, the critical porosity, and the exponent tp. Therefore, as
pointed out earlier, one should not expect equation (9) to describe the porosity dependence of the thermal
conductivity similarly in all types of soils or rocks, particularly those measured under various experimental
conditions such as different temperatures.
The fitted results (not shown) implied that ksat increases as sand content increases (with R2 5 0.75). This is
consistent with earlier postulation assuming that ksat can be estimated from ks and kw using the geometric
mean (see equation (4c)) because ks is mainly affected by quartz (or sand) content. A similar increasing
trend was also obtained between ksat and bulk density with R2 5 0.71.
For the saturation-dependent experiments in the Lu et al. [2014] database, we found 0.225 ts 0.369 (see
Table 2). The fitted ts values are indeed substantially smaller than those determined for the porosity-
dependent thermal conductivity (0.611 tp 1.698; see Table 1 and Figure 3) as well as the universal scal-
ing exponents of 2 (which incorporates the integrated effects of connectivity and tortuosity factors; Ghan-
barian et al. [2013]) and 0.76. The obtained results indicated that the scaling exponent ts value decreased
almost linearly as clay content and critical water content increased with correlation coefficient R2 5 0.30
and 0.36, respectively. We did not find any other remarkable trends between ts and other properties, such
as sand or silt content, bulk density, organic matter, and the ratio ksat/kdry. Since the saturation-dependent
thermal conductivity mainly occurs through the water in the pore space, the scaling exponent ts might be
related to the pore space fractal dimension. Its investigation, however, requires measured capillary pressure
curves, which were not available for these experiments.
The calculated critical water content values hc are also presented in Table 2. Generally speaking, as the clay
content increases, the value of hc increases as well. We found hc 5 0.0033 Clay(%) with correlation coeffi-
cient R2 5 0.93 (results not shown). Similar linear relationship was reported by Tarnawski and Leong [2000]
as well. Sepaskhah and Boersma [1979] also reported that the water content at which the apparent
saturation-dependent thermal conductivity begins to increase is a function of the clay content (see their

Figure 5. Thermal conductivity as a function of water content for six experiments from the Lu et al. [2014] database. The solid line repre-
sents equation (10) fitted to the measurements.
Figure 4). We caution that the critical water content determined by Sepaskhah and Boersma [1979] is not
equal to the critical water content hc within our P-EMA, but slightly smaller. In fact, hc in our model repre-
sents the critical volume fraction at which a continuous heat flow path through the high-conductivity com-
ponent forms. The range of hc presented in Table 2 is consistent with the range of critical volume fraction
(percolation threshold) reported for hydraulic conductivity [Hunt and Gee, 2003; Hunt, 2004; Ghanbarian
et al., 2015a], electrical conductivity [Ewing and Hunt, 2006; Ghanbarian et al., 2015a], air permeability [Hunt,
2005a; Ghanbarian-Alavijeh and Hunt, 2012; Ghanbarian et al., 2015a], and diffusion [Hunt et al., 2014a; Ghan-
barian and Hunt, 2015a, 2015b; Ghanbarian et al., 2015a,b]. We should point out that in a porous medium
with well-interconnected pore space one should expect a very low threshold for thermal conductivity, since
at low water contents heat may flow through adsorbed thin water films covering solid particles.

Figure 6. Thermal conductivity as a function of water content for two experiments from the Dong et al. [2015] database. The solid line rep-
resents equation (10) fitted to the measurements.
4.3.2. Dong et al. [2015] Database

We fit equation (10) to one clay and one silt loam soil sample from Dong et al.’s [2015] database to assess
the equation’s suitability and obtain fitted values for its parameters. In Figure 6, we summarize the calcu-
lated parameters, kdry, ksat, ts, and hc . The correlation coefficient R2 > 0.95 indicates excellent match
between theory (equation (10)) and the measurements in both experiments. The fitted ts values (0.192 and
0.181) less than unity are consistent with those found from the Lu et al. [2014] database (0.225 ts 0.369)
but slightly smaller. As demonstrated in Figure 6, the silt soil has critical water content (0.025 cm3 cm23)
substantially smaller than that of the clay soil (0.176 cm3 cm23). This large difference is because the amount
of water required to form a continuous heat flow path through the water phase depends on the surface
area of the solid particles, in particular their size and shape. The platy clay particles have a considerably
larger surface area than the silt and sand particles and thus need more water to produce a continuous heat
flow path. These results are analogous to those of Moldrup et al. [2001] who showed experimentally that
the threshold water content for solute diffusion was correlated with the specific surface area. One, however,
should not expect the thermal conductivity threshold be the same as the solute diffusion one, since they
are two different transport phenomena.
4.3.3. Effect of Temperature on the P-EMA Model Parameters
The Sepaskhah and Boersma [1979] database includes the saturation dependence of thermal conductivity in
three loam, silty clay loam, and loamy sand soils for two temperatures (e.g., 25 and 458C). Analyzing the
Sepaskhah and Boersma data sets rendered interesting results summarized in Figure 7 regarding the impor-
tant effect of temperature on our P-EMA model parameters. Generally speaking, comparing the thermal
conductivities measured at 258C with those at 458C for different soils (see also R2 values in Figure 7) indicate
more uncertainties and scatter at the higher temperature, consistent with the Campbell et al. [1994] results.
We found that in all three soils the critical water content for the saturation-dependent thermal conductivity
decreased as the temperature increased from 25 to 458C. The critical water contents were 0.215, 0.224, and
0.148 cm3 cm23 at 258C, which were twice as high as 0.112, 0.125, and 0.069 cm3 cm23 at 458C for the
loam, silty clay loam, and loamy sand, respectively. This might be because at higher temperatures viscosity
is lower and thus convection currents through liquid phase e.g., water can set up faster. However, the effect
of the latent heat transfer might also be nontrivial.
We did not find any specific trend between the value of ts and the temperature. By increasing temperature
from 25 to 458C, ts value decreased from 0.161 to 0.120 in silty clay loam, and increased from 0.215 to 0.263
and 0.335 to 0.470 in loam and sandy loam, respectively. More experiments are required to investigate the
possible effect of temperature on the exponent ts.
5. Discussion
In our porosity-dependent P-EMA framework (equation (9)), heat flows mainly through the solid compo-
nents, and thus tp should depend strongly on the characteristics of the solid matrix. In the saturation-

Figure 7. Thermal conductivity as a function of water content for six experiments from Sepaskhah and Boersma [1979]. The solid line repre-
sents equation (10) fitted to the measurements.
dependent thermal conductivity, however, heat flows mainly through the solid and water components, and
ts might be a function of both solid matrix and pore space properties. As a consequence, one may not
expect ts and tp be the same in a porous medium, since these parameters characterize thermal conductiv-
ities of different phases. Likewise, we do not expect /c be the same as hc in a porous medium.
As we pointed out earlier, it is challenging to fully address thermal conductivity within the percolation
theory framework. However, for the admittedly artificial conditions ka 5 0 and ks 5 kw (consistent with the
percolation theory concepts), Hunt et al. [2014b] postulated the approximate expression kðhÞ / ð12/1hÞ2
ða1chÞ0:25 in which a represents the contact area in the absence of water and c is a parameter whose value
must be zero for very small saturations and jump to a geometrically dependent nonzero value at some satu-
ration [Hunt et al., 2014b]. The first factor, conforming to universal scaling from percolation theory (equation
(5)), describes the thermal conductivity of a medium with zero percolation threshold in which the solid and

water phases contribute equally to the thermal conductivity. The second factor incorporates the effect of
capillary bridges demonstrated by Ewing and Horton [2007]. Hunt et al. [2014b] indicated that their expres-
sion only describes measurements for the thermal conductivity up to a fairly low water content and overes-
timates kðhÞ at higher saturations. Hunt et al. [2014b] stated, ‘‘The reason for this is, of course, that the
topological description near the percolation threshold cannot hold in the vicinity of p 5 1 [(h5/)].’’ They
demonstrated that a specific modification—reducing the power of the factor ð1-/1hÞ from 2 to
1—improved the thermal conductivity predictions at higher saturations. However, unrealistic assumptions
ka 5 0 and ks 5 kw may not be well supported in natural porous media in which 0.5 < ks/kw <15 [Hunt et al.,
2014b].
Within our P-EMA framework, we experimentally found tp and ts values less than the universal percolation
exponent (t 5 2) for both porosity and saturation-dependent thermal conductivity experiments. The fitted
values of ts less than unity are consistent with numerically calculated values reported in Ewing and Horton
[2007]. Considering liquid capillary bridges surrounding the solid-solid contacts in a cubic lattice of spheres,
Ewing and Horton [2007] simulated the saturation dependence of thermal conductivity. They stated, ‘‘Capil-
lary bridges can cause large increases in thermal conductivity, even at very low water contents [. . .].’’ Under
some specific conditions, such as up to a filling angle (angle between axis, sphere center, and point of
meniscus contact on the sphere’s surface) of p/4 and for a fixed conductivity ratio a 5 ks/kl 5 ks/kg in which
ks, kl, and kg are the thermal conductivity of the solid, liquid, and gas phases, respectively, they found a sim-
ple power law characterizing the thermal conductivity with an exponent whose value is a function of a (e.g.,
0.09 1 1.089/a). For example, 1 a 10 (one order of magnitude difference between ks and kl and/or kl
and kg), results in exponent values varying between 1.18 and 0.2. The results of Ewing and Horton [2007]
are, however, trivially applicable to natural porous media in which ks/kw 5 (assuming typical values
ks 5 2.9 W m218C21 and kw 5 0.6 W m218C21 at 258C) and kw/ka 24 (given that kw 5 0.6 W m218C21 and
ka 5 0.024 W m218C21 at 258C). Nonetheless, their results provided a promising interpretation for the rapid
rise in thermal conductivity caused by the addition of a small liquid volume. In order to evaluate Ewing and
Horton’s results experimentally, we suggest simultaneous measurement of the saturation dependence of
thermal conductivity and high-resolution 3-D images analysis. The latter would determine the volume of
water available as water films at low water contents and capillary bridges at intermediate ones.
Halperin et al. [1985] and Feng et al. [1987], among others, demonstrated that continuum percolation results
are nonuniversal meaning that the critical exponent t, scaling conductivity and permeability in three-
dimensional random continuum models, is greater than 2 and changes from one medium to another. Note
that within the continuum percolation framework, the nonuniversal t is mainly affected by the breadth of
the conductance distribution [see e.g., Feng et al., 1987]. It was, however, later acknowledged [Hunt, 2005b;
Ewing and Hunt, 2006; Hunt et al., 2014b; Ghanbarian et al., 2015a] that critical path analysis could provide a
theoretical framework to relate t to the fractal dimension of the pore space. This means the nonuniversal
exponents tp and ts might be determined from characteristics of the pore space and/or solid matrix. As a
consequence, one should expect that the nonuniversal scaling exponents tp and ts in our P-EMA might be
related to the fractal dimension of the solid matrix and/or pore space. Neither capillary pressure curves nor
grain-size distributions of the experiments used in this study were available. Thus, further investigation is
required to study how tp and ts values may change with the medium’s characteristics, such as pore space
and/or solid matrix fractal dimension.
A recent study of Likos [2014] indicated that the value of critical water content hc in our P-EMA model may
be estimated from the dry end (low water contents) of the capillary pressure curve. Likos [2014] demon-
strated that both thermal conductivity and matric suction as a function of water content show sharp
changes at low water contents. The measured water content at the dry end of capillary pressure curve has
been also applied by Ghanbarian-Alavijeh and Hunt [2012] to estimate the critical water content for water
flow to predict relative permeability using critical path analysis. Further numerical study is required to deter-
mine how the scaling exponent t changes with respect to the medium’s structure, the fluids’ characteristics,
and the grains’ sizes and shapes.
Thermal conductivity under partially saturated conditions also depends on saturation history [Bristow,
1998], which means that thermal conductivity measurement under drainage conditions might be different
than that determined under imbibition conditions. Further investigation is also required to address

hysteresis effects on the saturation dependence of the thermal conductivity in soils and our model parame-
ters ts and hc .
6. Conclusion
We developed porosity and saturation-dependent thermal conductivity models in porous media using
percolation-based effective-medium approximation (P-EMA). The proposed models are a function of the
endmember properties (e.g., air, solid matrix, and saturating fluid thermal conductivities), a scaling expo-
nent, and a percolation threshold. Comparison with experimental data indicated excellent match between
theory and the measurements. By analyzing the porosity-dependent thermal conductivity of 119 mineral
soils and 74 crushed rocks, we found the scaling exponent tp and the threshold porosity /c values for each
data set. However, one should bear in mind that the optimized tp and /c values reported in this study may
not represent all types of soils and rocks, particularly soils, which may contain diverse mineralogical compo-
nents substantially different from those analyzed here.
Generally speaking, we found that in the saturation-dependent experiments, the percolation threshold (crit-
ical water content) changes with clay content, with larger clay contents corresponding to greater percola-
tion thresholds. We also found that the scaling exponent ts is nonuniversal and depends on the
microstructure, fluid characteristics, and probably particles size and shape. Such nonuniversal results, which
have been more frequently reported in the physics literature, indicate that our approach depends not only
on topological properties of the medium, e.g., dimension and connectivity, but also geometrical and struc-
tural characteristics such as pore-size distribution and pore geometry.
The main difference between our theoretically developed models and other empirically based equations is
that in the latter, the parameters are generally speaking shape factors describing the dependence of ther-
mal conductivity on either porosity or saturation. These models are useful for describing qualitative behav-
ior of the porosity and saturation-dependence of thermal conductivity. However, the P-EMA approach
proposed in this study is based upon theoretical frameworks, i.e., percolation theory and the effective-
medium approximation, and includes scaling exponents (tp and ts) and percolation thresholds (/c and hc )
that are influenced by structural and geometrical characteristics of pore space and solid matrix, and are
thus nonuniversal.
Although we proposed theoretical frameworks to model porosity and saturation-dependent thermal con-
ductivity in natural porous media such as soils and rocks, further numerical and experimental investigations
are still required to examine factors influencing our P-EMA model parameters, such as the scaling exponent,
and the percolation threshold.
Acknowledgement References
The authors acknowledge David S.
Ben-Avraham, D., and S. Havlin (2000), Diffusion and Reactions in Fractals and Disordered Systems, Cambridge Univ. Press, N. Y.
McLachlan, University of the
Bergman, D. J., and D. Stroud (1992), Physical properties of macroscopically inhomogeneous media, Solid State Phys., 46, 147–269.
Witwatersrand, and Godfrey Sauti,
Bonnet, P., D. Sireude, B. Garnier, and O. Chauvet (2007), Thermal properties and percolation in carbon nanotube-polymer composites,
National Institute of Aerospace, for
Appl. Phys. Lett., 91, 201910, doi:10.1063/1.2813625.
their fruitful comments on the
Bristow, K. L. (1998), Measurement of thermal properties and water content of unsaturated sandy soil using dual-probe heat-pulse probes,
percolation-based effective-medium
approximation model fitting process, Agric. Forest Meteorol., 89(2), 75–84.
and Tusheng Ren, China Agricultural Bristow, K. L. (2012), Thermal conductivity, in Methods of Soil Analysis, Part. 4. Physical Methods, SSSA Book Ser. 5, pp. 1209–1226, Soil Sci-
University, and Yili Lu, Iowa State ence Society of America (SSSA), Madison, Wis.
University, for kindly sharing their Bristow, K. L., J. R. Bilskie, G. J. Kluitenberg, and R. Horton (1995), Comparison of techniques for extracting soil thermal properties from
experiments. We also thank the dual-probe heat-pulse data, Soil Sci., 160(1), 1–7.
Associate Editor, Lynn S. Bennethum Bruggeman, D. A. (1935), Berechnung verschiedener Physikalischer Konstanten von heterogenen Substanzen, Ann. Phys., LPZ 24, 636–679.
(Washington State University), two Campbell, G. S., J. D. Jungbauer, W. R. Bidlake, and R. D. Hungerford (1994), Predicting the effect of temperature on soil thermal conductiv-
anonymous reviewers, and Robert ity, Soil Sci., 158, 307–313.
P. Ewing (Iowa State University) for Carslaw, H. C., and J. C. Jaeger (1986), Conduction of Heat in Solids, Oxford University Press, Oxford, 510 pp.
their constructive comments, which Chiteme, C., D. S. McLachlan, and G. Sauti (2007), ac and dc percolative conductivity of magnetite-cellulose acetate composites, Phys. Rev.
improved the quality of this work. The B, 75, 094202.
digitized data used in this study are Clerc, J. P., G. Giraud, J. M. Laugier, and J. M. Luck (1990), The electrical conductivity of binary disordered systems, percolation clusters, frac-
available upon request to the tals and related models, Adv. Phys., 39(3), 191–309.
corresponding author. Consiglio, R., D. R. Baker, G. Paula, and H. E. Stanley (2003), Continuum percolation thresholds for mixtures of spheres of different sizes,
Physica A, 319, 49–55.
C^
ote, J., and J.-M. Konrad (2005), A generalized thermal conductivity model for soils and construction materials, Can. Geotech. J., 42,
443–458.

David, C., Y. Gueguen, and G. Pampoukis (1990), Effective medium theory and network theory applied to the transport properties of rock,
J. Geophys. Res., 95, 6993–7005.
Davis, L. C., and B. E. Artz (1995), Thermal conductivity of metalmatrix composites, J. Appl. Phys., 77, 4954–4960.
de Vries, D. A. (1963), Thermal properties of soils, in Physics of Plant Environment, edited by W. R. Van Wijk, pp. 210–235, North-Holland,
Amsterdam, Netherlands.
Deprez, N., D. S. McLachlan, and I. Sigalas (1988), The measurement and comparative analysis of the electrical and thermal conductivities,
permeability and Young’s modulus of sintered nickel, Solid State Commun, 66, 869–872.
Dong, Y., J. S. McCartney, and N. Lu (2015), Critical review of thermal conductivity models for unsaturated soils, Geotech. Geol. Eng., 33,
207–221.
Ewing, R. P., and R. Horton (2007), Thermal conductivity of a cubic lattice of spheres with capillary bridges, J. Phys. D Appl. Phys., 40,
4959–4965.
Ewing, R. P., and A. G. Hunt (2006), Dependence of the electrical conductivity on saturation in real porous media, Vadose Zone J., 5, 731–741.
Feng, S., B. I. Halperin, and P. N. Sen (1987), Transport properties of continuum systems near the percolation threshold, Phys. Rev. B, 35,
197–214.
Feng, Y, B. Yu, M. Zou, and D. Zhang (2004), A generalized model for the effective thermal conductivity of porous media based on self-sim-
ilarity, J. Phys. D Appl. Phys., 37, 3030–3040.
Foygel, M., R. D. Morris, D. Anez, S. French, and V. L. Sobolev (2005), Theoretical and computational studies of carbon nanotube composites
and suspensions: Electrical and thermal conductivity, Phys. Rev. B, 71, 104201.
Gao, L., and X. F. Zhou (2006), Differential effective medium theory for thermal conductivity in nanofluids, Phys. Lett. A, 348, 355–360.
Garboczi, E. J., K. A. Snyder, J. F. Douglas, and M. F. Thorpe (1995), Geometrical percolation threshold of overlapping ellipsoids, Phys. Rev. E,
52, 819–828.
Ghanbarian-Alavijeh, B., and A. G. Hunt (2012), Comparison of the predictions of universal scaling of the saturation dependence of the air
permeability with experiment, Water Resour. Res., 48, W08513, doi:10.1029/2011WR011758.
Ghanbarian-Alavijeh, B., T. E. Skinner, and A. G. Hunt (2012), Saturation dependence of dispersion in porous media, Phys. Rev. E, 86(6),
066316.
Ghanbarian, B., and A. G. Hunt (2014), Universal scaling of gas diffusion in porous media, Water Resour. Res., 50, 2242–2256, doi:10.1002/
2013WR014790.
Ghanbarian, B., A. G. Hunt, M. Sahimi, R. P. Ewing, and T. E. Skinner (2013), Percolation theory generates a physically based description of
tortuosity in saturated and unsaturated porous media, Soil Sci. Soc. Am. J., 77, 1920–1929.
Ghanbarian, B., A. G. Hunt, R. P. Ewing, and T. E. Skinner (2014), Universal scaling of the formation factor in porous media derived by com-
bining percolation and effective medium theories, Geophys. Res. Lett., 41, 3884–3890, doi:10.1002/2014GL060180.
Ghanbarian, B., A. G. Hunt, T. E. Skinner, and R. P. Ewing (2015a), Saturation dependence of transport in porous media predicted by percola-
tion and effective medium theories, Fractals, 23, 1540004.
Ghanbarian, B., H. Daigle, A. G. Hunt, R. P. Ewing, and M. Sahimi (2015b), Gas and solute diffusion in partially saturated porous media: Per-
colation theory and effective medium approximation compared with lattice Boltzmann simulations, J. Geophys. Res. Solid Earth, 120,
182–190, doi:10.1002/2014JB011645.
Halperin, B. I. (1989), Remarks on percolation and transport in networks with a wide range of bond strengths, Physica D, 38, 179–183.
Halperin, B. I., S. Feng, and P. N. Sen (1985), Differences between lattice and continuum percolation transport exponents, Phys. Rev. Lett.,
54, 2391–2394.
Hunt, A. G. (2004), Continuum percolation theory for water retention and hydraulic conductivity of fractal soils: Estimation of the critical
volume fraction for percolation, Adv. Water Resour., 27, 175–183.
Hunt, A. G. (2005a), Continuum percolation theory for saturation dependence of air permeability, Vadose Zone J., 4, 134–138.
Hunt, A. G. (2005b), Basic transport properties in natural porous media: Continuum percolation theory and fractal model, Complexity, 10,
22–37.
Hunt, A. G., and G. W. Gee (2003), Wet-end deviations from scaling of the water retention characteristics of fractal porous media, Vadose
Zone J., 2, 759–765.
Hunt, A., B. Ghanbarian, and R. P. Ewing (2014a), Saturation dependence of solute diffusion in porous media: Universal scaling compared
with experiments, Vadose Zone J., 13(4), doi:10.2136/vzj2013.08.0204.
Hunt, A., R. Ewing, and B. Ghanbarian (2014b), Percolation Theory for Flow in Porous Media, Lecture Notes in Physics, vol. 880, 3rd ed.,
Springer, Berlin.
Jerauld, G. R., J. C. Hatfield, L. E. Scriven, and H. T. Davis (1984), Percolation and conduction on Voronoi and triangular networks: A case
study in topological disorder, J. Phys. C Solid State Phys., 17, 1519–1529.
Johansen, O. (1975), Thermal conductivity of soils, PhD dissertation, Norwegian Univ. of Science and Technol., Trondheim. [CRREL Draft
Transl. 637, Cold Reg. Res. Eng. Lab., Hanover, N. H.]
King, P. R. (1989), The use of renormalization for calculating effective permeability, Transp. Porous Media, 4, 37–58.
Kirkpatrick, S. (1973), Percolation and conduction, Rev. Mod. Phys., 45, 574–588.
Kogut, P. M., and J. P. Straley (1979), Distribution-induced non-universality of the percolation conductivity exponents, J. Phys. C Solid State
Phys., 12, 2151–2159.
Koplik, J. (1981), On the effective medium theory of random linear networks, J. Phys. C Solid State Phys., 14(32), 4821–4837.
Kou, J., Y. Liu, F. Wu, J. Fan, H. Lu, and Y. Xu (2009), Fractal analysis of effective thermal conductivity for three-phase (unsaturated) porous
media, J. Appl. Phys., 106, 054905.
Lehmann, P., M. Stahli, A. Papritz, A. Gygi, and H. Fluhler (2003), A fractal approach to model soil structure and to calculate thermal conduc-
tivity of soils, Transp. Porous Media, 52, 313–332.
Likos, W. J. (2014), Pore-scale model for thermal conductivity of unsaturated sand, Geotech. Geol. Eng., 33(2), 179–192.
Lindquist, W. B., A. Venkatarangan, J. Dunsmuir, and T. F. Wong (2000), Pore and throat size distributions measured from synchrotron X-ray
tomographic images of Fontainbleau sandstones, J. Geophys. Res., 105, 21,509–21,527.
Lu, S., T. Ren, Y. Gong, and R. Horton (2007), An improved model for predicting soil thermal conductivity from water content room temper-
ature, Soil Sci. Soc. Am. J., 71, 8–14.
Lu, S., T. S. Ren, Z. R. Yu, and R. Horton (2011), A method to estimate the water vapour enhancement factor in soil, Eur. J. Soil Sci., 62,
498–504, doi:10.1111/j.1365-2389.2011.01359.x.
Lu, Y., Y. Wang, and T. Ren (2013), Using late time data improves the heat-pulse method for estimating soil thermal properties with the
pulsed infinite line source theory, Vadose Zone J., 12(4), doi:10.2136/vzj2013.01.0011.

Lu, Y., S. Lu, R. Horton, and T. Ren (2014), An empirical model for estimating soil thermal conductivity from texture, water content, and
bulk density, Soil Sci. Soc. Am. J., 78, 1859–1868.
Maxwell-Garnett, J. C. (1904), Colours in metal glasses, in metallic films and in metallic solutions: II, Philos. Trans. R. Soc. London A, 203, 385–420.
McLachlan, D. S. (1987), An equation for the conductivity of binary mixtures with anisotropic grain structures, J. Phys. C Solid State Phys., 20,
865–877.
McLachlan, D. S. (1988), Measurement and analysis of a model dual-conductivity medium using a generalised effective-medium theory, J.
Phys. C Solid State Phys., 21, 1521–1532.
McLachlan, D. S., G. Sauti, and C. Chiteme (2007), Static dielectric function and scaling of the ac conductivity for universal and nonuniversal
percolation systems, Phys. Rev. B, 76, 014201.
Moldrup, P., T. Olesen, T. Komatsu, P. Schjønning, and D. E. Rolston (2001), Tortuosity, diffusivity, and permeability in the soil liquid and gas-
eous phases, Soil Sci. Soc. Am. J., 65, 613–623.
Rintoul, M. D., and S. Torquato (1997), Precise determination of the critical threshold and exponents in a three-dimensional continuum
percolation model, J. Phys. A Math. Gen., 30, L585–L592.
Sahimi, M. (1993), Flow phenomena in rocks: From continuum models to fractals, percolation, cellular automata, and simulated annealing,
Rev. Mod. Phys., 65(4), 1393–1534.
Sahimi, M. (1994a), Long-range correlated percolation and flow and transport in heterogeneous porous media, J. Phys. I, 4, 1263–1268.
Sahimi, M. (1994b), Applications of Percolation Theory, 258 pp., Taylor and Francis, London, U. K.
Sahimi, M. (2003), Heterogeneous Materials I: Linear Transport and Optical Properties, pp. 691, Springer, Berlin, Germany.
Sahimi, M. (2011), Flow and Transport in Porous Media and Fractured Rock: From Classical Methods to Modern Approaches, 2nd ed., 709 pp.,
John Wiley, Weinheim, Germany.
Sahimi, M., and S. Mukhopadhyay (1996), Scaling properties of a percolation model with long-range correlations, Physical Rev. E, 54(4),
3870–3880.
Sahimi, M., B. D. Hughes, L. E. Scriven, and H. T. Davis (1983), Real-space renormalization and effective-medium approximation to the per-
colation conduction problem, Phys. Rev. B, 28, 307–311.
Sass, J. H., A. H. Lachenbruch, and R. J. Munroe (1971), Thermal conductivity of rocks from measurements on fragments and its application
to heat-flow determination, J. Geophys. Res., 76, 3391–3401.
Sen, P. N., J. N. Roberts, and B. I. Halperin (1985), Nonuniversal critical exponents for transport in percolating systems with a distribution of
bond strengths, Phys. Rev. B, 32, 3306–3308.
Sepaskhah, A. R., and L. Boersma (1979), Thermal conductivity of soils as a function of temperature and water content, Soil Sci. Soc. Am. J.,
43, 439–444.
Shao, W. Z., N. Xie, L. Zhen, and L. C. Feng (2008), Conductivity critical exponents lower than the universal value in continuum percolation
systems, J. Phys. Condens. Matter, 20, 395235.
Shiozawa, S., and G. S. Campbell (1990), Soil thermal conductivity, Remote Sens. Rev., 5, 301–310.
Singh, D. N., and K. Devid (2000), Generalized relationships for estimating soil thermal resistivity, Exp. Thermal Fluid Sci., 22, 133–143.
Singh, D. N., S. J. Kuriyan, and K. C. Manthena (2001), A generalized relationship between soil electrical and thermal resistivities, Exp. Ther-
mal Fluid Sci., 25, 175–181.
Stauffer, D., and A. Aharony (1994), Introduction to Percolation Theory, 2nd ed., Taylor and Francis, London, U. K.
Tarnawski, V. R., and W. H. Leong (2000), Thermal conductivity of soils at very low moisture content and moderate temperatures, Transp.
Porous Media, 41(2), 137–147.
The MathWorks Inc. (2013), MATLAB: The Language of Technical Computing, version 8.1., Mass.
Wang, B.-X., L.-P. Zhou, and X.-F. Peng (2003), A fractal model for predicting the effective thermal conductivity of liquid with suspension of
nanoparticles, Int. J. Heat Mass Transf., 46, 2665–2672.
Wang, J., J. K. Carson, M. F. North, and D. J. Cleland (2006), A new approach to modeling the effective thermal conductivity of heterogene-
ous materials, Int. J. Heat Mass Transf., 49, 3075–3083.
Wilkinson, D., and J. F. Willemsen (1983), Invasion percolation: A new form of percolation theory, J. Phys. A Math. Gen., 16, 3365–3376.
Woodside, W., and J.H. Messmer (1961), Thermal conductivity of porous media, J. Appl. Phys., 32, 1688–1706.
Youngs, I. J. (2002), Exploring the universal nature of electrical percolation exponents by genetic algorithm fitting with general effective
medium theory, J. Phys. D Appl. Phys., 35(23), 3127–3137.
Yu, B., and P. Cheng (2002), Fractal models for the effective thermal conductivity of bidispersed porous media, J. Thermophys. Heat Transf.,
16, 22–29.
Zimmerman, R. W. (1989), Thermal conductivity of fluid-saturated rocks, J. Petroleum Sci. Eng., 3, 219–227.
View publication stats

Thermal Conductivity in Porous Media: Percolation-Based Effective-Medium Approximation

Uploaded by

Copyright:

Available Formats

Thermal Conductivity in Porous Media: Percolation-Based Effective-Medium Approximation

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thermal Conductivity in Porous Media: Percolation-Based Effective-Medium Approximation

Uploaded by

Copyright:

Available Formats

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

Thermal conductivity in porous media: Percolation-based effective-medium

Article in Water Resources Research · December 2015

Behzad Ghanbarian Hugh Daigle

SEE PROFILE SEE PROFILE

Scale-dependent pedotransfer functions View project

Microfluidic Study of Multi-Phase Flow in Hydrocarbon Reservoir View project

The user has requested enhancement of the downloaded file.

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 295

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 296

kdry 5v102g/ (4a)

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 297

2.1. Percolation Theory

T / ðp2pc Þt ; p > pc (5)

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 298

2.2. Effective-Medium Approximation

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 299

ðTl 2Te Þ ðTh 2Te Þ

1=t 1=t 1=t 1=t

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 300

2.3.1. Porosity-Dependent Thermal Conductivity k(/)

3. Materials and Methods

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 301

^te and Konrad [2005] Database

3.2. Woodside and Messmer [1961] Database

3.3. Sepaskhah and Boersma [1979] Database

3.4. Lu et al. [2014] Database

3.5. Dong et al. [2015] Database

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 302

4.1. Porosity-Dependent Results

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 303

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 304

4.2. Effect of Saturating Fluid on the P-EMA Model Parameters

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 305

literature [see e.g., Garboczi et al., 1995].

4.3. Saturation-Dependent Results

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 306

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 307

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 308

4.3.2. Dong et al. [2015] Database

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 309

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 310

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 311

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 312

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 313

GHANBARIAN AND DAIGLE THERMAL CONDUCTIVITY IN POROUS MEDIA 314

View publication stats

You might also like