Applied Thermal Engineering 102 (2016) 972–979

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

A reconstruction of Maxwell model for effective thermal conductivity

of composite materials
J.Z. Xu a,1, B.Z. Gao a,b,c,⇑,1, F.Y. Kang a,b
a
City Key Laboratory of Thermal Management Engineering and Materials, Graduate School at Shenzhen, Tsinghua University, University Town, Shenzhen 518055, China
b
State Key Laboratory of New Ceramics and Fine Processing, School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China
c
Aerospace Materials and Technology Research Institute, Beijing, China

h i g h l i g h t s

Deficiencies were found in classical Maxwell model for effective thermal conductivity.

Maxwell model was reconstructed based on potential mean-field theory.

Reconstructed Maxwell model was extended with particle–particle contact resistance.

Predictions by reconstructed Maxwell model agree excellently with experimental data.

a r t i c l e i n f o a b s t r a c t

Article history: Composite materials consisting of high thermal conductive fillers and polymer matrix are often used as
Received 17 January 2016 thermal interface materials to dissipate heat generated from mechanical and electronic devices. The pre-
Accepted 29 March 2016 diction of effective thermal conductivity of composites remains as a critical issue due to its dependence
Available online 31 March 2016
on considerably factors. Most models for prediction are based on the analog between electric potential
and temperature that satisfy the Laplace equation under steady condition. Maxwell was the first to derive
Keywords: the effective electric resistivity of composites by examining the far-field spherical harmonic solution of
A. Composite materials
Laplace equation perturbed by a sphere of different resistivity, and his model was considered as classical.
B. Effective thermal conductivity
C. Potential theory
However, a close review of Maxwell’s derivation reveals that there exist several controversial issues (defi-
ciencies) inherent in his model. In this study, we reconstruct the Maxwell model based on a potential
mean-field theory to resolve these issues. For composites made of continuum matrix and particle fillers,
the contact resistance among particles was introduced in the reconstruction of Maxwell model. The
newly reconstructed Maxwell model with contact resistivity as a fitting parameter is shown to fit excel-
lently to experimental data over wide ranges of particle concentration and mean particle diameter. The
scope of applicability of the reconstructed Maxwell model is also discussed using the contact resistivity
as a parameter.
Ó 2016 Published by Elsevier Ltd.

1. Introduction
High thermal conductive fillers are often added to polymer
matrix to form composites as thermal interface materials to
improve thermal conductivity to dissipate heat generated from
mechanical and electronic devices [1,2]. Many kinds of fillers, such
as boron nitride, aluminum nitride, carbon nanotube, graphene,
graphite, aluminum, copper, aluminum oxide, and silver nanowires
⇑ Corresponding author at: City Key Laboratory of Thermal Management
Engineering and Materials, Graduate School at Shenzhen, Tsinghua University,
University Town, Shenzhen 518055, China.
E-mail address: gaobz2828@sina.com (B.Z. Gao).
1 thermal conductivity of composite materials. Most models were
These authors contributed equally to this work.
[3–11], having high thermal conductivity and low thermal expan-
sion, are widely used as high thermal conductive fillers. A number
of polymer materials, such as epoxy-resin, polyethylene, polyur-
ethane, polybenzoxazine, polypropylene, and silicone rubber
[12–16] are used as polymer matrix to improve mechanical prop-
erties and thermal stability of the composites. Apparently, the
effective thermal conductivity of composites depends not only on
volume fractions and thermal conductivities of the contributed
phases, but also on complicated geometry and morphologies of fil-
lers. As a result, the prediction of effective thermal conductivity of
composites remains as a critical task in the thermal management
industry.
Considerable efforts have been made on modeling the effective

http://dx.doi.org/10.1016/j.applthermaleng.2016.03.155
1359-4311/Ó 2016 Published by Elsevier Ltd.
 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979
Nomenclature
ke effective thermal conductivity of composite material
kf thermal conductivity of continuous first phase (polymer
matrix)
ks thermal conductivity of second phase (particle filler)
kse effective thermal conductivity of particles in contact
/ volume fraction of continuous first phase (polymer ma-
trix)
/s volume fraction of second phase (particle filler)
R1 electric resistivity of first phase
R2 electric resistivity of second phase
based on the analogy between electric and thermal conductions
under steady condition, where the electric potential and the tem-
perature both satisfy the Laplace equation. To evaluate the effec-
tive thermal conductivity, two approaches were generally
adopted. The first one is to regard the composite as a heteroge-
neous medium consisting of elements with different thermal con-
ductivities that are connected into a circuit network. The effective
thermal conductivity is then obtained by evaluating the overall
conductivity of the network. The second approach, pioneered by
Maxwell [17] for electric field and now generally considered as
classical, is to solve the Laplace equation and examine the far-
field spherical harmonic perturbed by a homogeneous spherical
particle with conductivity different from that of an infinitely
extended homogeneous ambient. The effective conductivity of a
large heterogeneous sphere partially filled with many smaller
spherical particles is then determined by equating the far-field
potential resulted from the superposition of smaller spherical par-
ticles to that from the large sphere having the effective conductiv-
ity. In the following, we should review briefly some models
obtained from the first approach and then focus on the Maxwell
model to demonstrate the salient features of the second approach.
However, this review also reveals several controversial issues
(deficiencies) inherent in the Maxwell model. These issues remain
unresolved with the extended Maxwell models by later research-
ers. The objective of this study is then to reconstruct the Maxwell
model to resolve these issues. This was achieved by firstly consid-
ering an infinitely extended heterogeneous medium that contains
smaller spherical particles and then regarding it as having an effec-
tive conductivity of a homogenous medium based on the potential
mean-field theory. The effective conductivity is then determined
by requiring that there will have no contribution to far-field poten-
tial if small spheres in a large sphere is removed and then refilled
through the superposition of small spheres. However, the superpo-
sition process implies that there has no resistivity among small
spheres, i.e., small spheres are in perfect contact, having the behav-
ior of a continuum. For composites made of continuum matrix and
particle fillers, the contact resistance among particles is taken into
consideration in the reconstruction of Maxwell model. The predic-
tions by the newly reconstructed Maxwell model are fitted to the
experimental results obtained recently by Gao et al. [18], which
shows an excellent agreement. Finally, the scope of applicability
of the reconstructed Maxwell model is discussed.
2. Models based on circuit network of conductors
Simplification of the complicated network was usually
employed to calculate effective thermal conductivity of compos-
ites. The pioneering work of Deissler and Boegli [19] assumed that
the two phases in composites can be disaggregated into two layers
973
Re effective electric resistivity of mesoscale sphere
containing n-spheres
R2e effective electric resistivity for phase-2 particles in
contact
Rc contact resistivity among spheres of second phase
Rc dimensionless contact resistivity of second phase,
Rc ð¼ Rc  ks Þ
a equivalent radius of mesoscale sphere containing
n-spheres
a2 radius of sphere of second phase (particle filler)
with the interfaces being either in parallel or in perpendicular to
the heat flux direction, namely layers-in-parallel or layers-in-
series. The effective thermal conductivities of composites were
obtained respectively as,
ke ¼ /kf þ /s ks ð1Þ
1 / /
¼ þ s ð2Þ
ke kf ks
The layers-in-parallel model as given by Eq. (1) has zero inter-
face thermal resistance, while the layers-in-series model by Eq.
(2) the most. Therefore, they represent respectively the upper
and lower limits of the effective thermal conductivity.
Combining layers in-parallel and in-series, Kunii and Smith [20]
derived expressions for predicting the effective thermal conductiv-
ity of porous beds, including the effect of thermal radiation. Zehner
and Schlunder [21] choose a cylindrical unit cell contained a
spherical particle in point contact with neighboring particles and
obtained a logarithmic expression for the effective thermal
conductivity of a packed bed. Comparison of predictions of
Zehner–Schlunder model with experimental data indicated that
at high solid/fluid conductivity ratios, the model under-predicted
considerably the effective thermal conductivity. Hsu et al. [22]
modified Zehner–Schlunder model by including the finite-area
contacts between spheres. The modified area-contact model of
Hsu et al. [22] was in much better agreement with experimental
data than the point-contact model of Zehner and Schlunder [21],
especially at high solid/fluid thermal conductivity ratios. Based
on the method with layers combined in-parallel and in-series, alge-
braic expressions for the effective thermal conductivity of three-
dimensional spatially periodic porous media were obtained by
Hsu et al. [23]. The model of Hsu et al. [23] was shown to fit excel-
lently well by Gao et al. [18] to their experimental data obtained
recently for the composites made of silicone rubber and spherical
Al2O3 particles.
3. Maxwell model and its extensions
Maxwell [17] was the first to derive the expression for effective
electric resistivity of a composite based on theory of electric poten-
tial that satisfies the Laplace equation. He firstly considered the
effect on the far-field potential by placing a sphere of radius a2
with electric resistivity R2 into an infinitely-extended homogenous
medium of electric resistivity R1 as shown in Fig. 1a. In terms of
spherical coordinate (r, w, h) with origin at the center of sphere,
the first mode of spherical harmonics of the electric potential at
a great distance is given by


1
V1 ¼ B0 r þ B1 2 cos h ð3Þ
r
974 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979
Fig. 1. Illustration of Maxwell’s modeling the effective electric resistivity of a heterogeneous sphere containing spherical particles.
where B0 is the coefficient of the background potential associated
with R1 and B1 is the coefficient of the potential induced by the
sphere with R2, which is given by
R2  R1 3
B1 ¼ a B0
2R2 þ R1 2
Maxwell then considered a large sphere of radius a filled with n
spheres of radius a2 as shown in Fig. 1b. From the linearity of
potential theory, Bn for the far field potential induced by n spheres
is simply the n-fold superposition of B1, i.e., Bn = nB1, which in term
of the particle volume fraction, /s ¼ na32 =a3 , is given by
R2  R1
Bn ¼ / a3 B0
2R2 þ R1 s
Now regarding the resistivity of the system of n spheres as
equivalently having an effective resistivity Re of a sphere with
radius a as shown in Fig. 1c, the expression for equivalent Bn is
the same as Eq. (4) with B1, R2 and a2 being replaced by Bn, Re
and a, respectively. By equating the modified Eq. (4) for Bn to Eq.
(5), the effective electric resistivity Re can be obtained as:
2R2 þ R1 þ /s ðR2  R1 Þ
Re ¼ R1
2R2 þ R1  2/s ðR2  R1 Þ
which is the equation given originally in the book of Maxwell [17].
From the analogy between the electric potential and the tem-
perature of steady heat conduction, Eq. (6) applies equally well
to the thermal resistivity. In terms of the thermal conductivity,
i.e., k1 = 1/R1, k2 = 1/R2 and ke = 1/Re, and denoting k1 and k2 by kf
and ks, respectively, where the subscripts f and s refer to the first
and second phases of composites, Eq. (6) becomes
2kf þ ks þ 2/s  ðks  kf Þ
ke ¼  kf
2kf þ ks  /s  ðks  kf Þ
which is the equation commonly used by researchers in heat trans-
fer community. Note that from Eq. (7), the effective thermal con-
ductivity depends solely on the thermal conductivities and
volume fractions of materials and contains no parameter to account
for the particle interfacial effect.
The above derivation of Eq. (6) by Maxwell [17] indeed repre-
sents a novel approach to the determination of the effective elec-
tric resistivity of heterogeneous materials. However, a close
examination of his derivation reveals that there are several issues
(deficiencies) associated with his model. These issues arise from
the use of the infinitively-extended homogeneous medium of R1
as the ambient and from the insensitiveness of far-field potential
to the detailed geometries and distributions of near-field medium
as implied by the potential theory. The use of R1 for the ambient
implies that there no spheres of radius a2 appeared outside the
sphere of radius a. Therefore, Eq. (6) may apply only for a compos-
ite of single heterogeneous sphere placed in the ambient of R1 as
shown in Fig. 1b, not for a continuously extended composite which
is of the main interest in engineering practice.
The insensitiveness of far-field potential to the detailed geome-
tries and distributions of near-field medium leads to the arbitrari-
ness of selecting n, a2 and a in determining the particle volume
fraction, /s ¼ na32 =a3 . If n and a2 are fixed, the arbitrary selection
ð4Þ of a will give an infinite number of (/s , Re) pair that have the same
far-field potential. In another word, the effective electric resistivity
in Eq. (6) is not uniquely determined from the far-field potential.
On the other end, if we fix the radius of sphere (a) and change n
and a2 while keeping the volume fraction of particles (/s ) at con-
stant, then, in the limit of n approaches infinity and a2 approaches
zero, the summation of 4pna32 =3 can be replaced by volume inte-
gral of a continuum second phase as shown in Fig. 2. Hence, the
ð5Þ Maxwell model is applicable to a continuum second phase, instead
of dispersed particles as normally interpreted by many researchers.
This is also apparent from the property of linear superposition,
which implies that there has no resistivity among particles of
radius a2, i.e., particles are perfectly in contact. For practical com-
posites with discrete particles with contact resistance among par-
ticles, lower effective thermal conductivities than that predicted by
Eq. (7) are anticipated.
Since the model was firstly proposed by Maxwell, there were
ð6Þ several extensions to his model. Rayleigh [24] considered the com-
posites with spherical particles of dilute concentration arranged
periodically in cubic array, which allows for the solution using a
cube as unit cell. The result of Rayleigh to the first order approxi-
mation of small particle concentration recovered Eq. (7) of Max-
well model (i.e., particles outside the unit cell have no effect) and
the higher order approximation occurred only at 10/3 power of
particle concentration.
Hasselman and Johnson [25] extended Maxwell model by
including the interfacial contact thermal resistance between the
ð7Þ two phases, but solved the Laplace equation directly for steady
heat conduction. Their modified equation for effective thermal
conductivity reduces to Eq. (7) when the contact thermal resis-
tance is set to zero. It should be noted that the inclusion of interfa-
cial contact thermal resistance will decrease the effective thermal
conductivity of the Maxwell model.
Recently, Shaker, Birgersson and Mujumdar [26] extended the
Fig. 2. The continuum second phase structure applicable by Maxwell model.
 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979 975

classical Maxwell model for the effective thermal conductivity of where B0 is the coefficient associated with Re.
nanofluids, by adding a new term to the Fourier’s law to account Let’s now remove the n particles from the control volume so
for nonlocal heat flux at nano-length-scale. Their modified-model that the sphere of radius a is purely filled with the phase-1 mate-
reduces to Maxwell model of Eq. (7) when the nonlocal effect is rial of electric resistivity R1 as shown in Fig. 3b. The first spherical
removed, although it was shown to be in good agreement with harmonic mode of the far field potential at a great distance is given
experimental results of alumina nanofluids at very low particle by:
concentrations.

1
Apparently, these later extensions basically recovered the result V1 ¼ B0 r þ B1 2 cos h ð9Þ
r
of Maxwell model to the first approximation and had not
addressed the issues (deficiencies) of Maxwell model. In the fol- which is of the same form as Eq. (3), but the coefficient B1 is now
lowing, we should reconstruct the Maxwell model to circumvent changed to:
these deficiencies.
R1  Re 3
B1 ¼ a B0 ð10Þ
4. Reconstruction of Maxwell model
2R1 þ Re
since the ambient electric resistivity is Re and the sphere of radius a
Consider an infinitively-extended heterogeneous composite has an electric resistivity R1.
consisting of spherical particles of phase 2 (with radius a2 and elec- The next step is to place the removed n spheres of radius a2 with
tric resistivity R2) in a continuum substrate of phase 1 (with elec- R2 back into the control volume. This will requires first the vacation
tric resistivity R1) as shown in Fig. 3a. Our tasks will be to of the space of n spheres of radius a2 with electric resistivity R1
determine the effective electric resistivity Re of the composite cor- from the sphere of radius a as shown in Fig. 3c. After placing back
responding to the mesoscale control volume of a sphere of radius a, the n spheres of radius a2 with R2, the end result as shown in
that contains n spheres of radius a2, as shown by the dashed circle Fig. 3d goes back to Fig. 3a. Employing the linear superposition,
in Fig. 3a. We note that Re is the background electric resistivity the contributions to the far field potential by the n spheres of
since the control volume can be placed at any position in the radius a2 with R1 and R2, respectively, are:
domain considered. By this the heterogeneous composite is
R1  Re
regarded as a continuous medium with resistivity Re based on DB1 ¼ / a3 B0 ð11Þ
the potential mean-field theory. The present consideration 2R1 þ Re s
apparently has circumvented the problem of non-uniqueness and
encountered in the Maxwell model since n paces with a
without changing /s , i.e., n will increase with increasing a such R2  Re
DB2 ¼ / a3 B0 ð12Þ
that /s ¼ na32 =a3 = constant, and the values of Re is uniquely 2R2 þ Re s
determined. similar to Eq. (5). Combining Eqs. (10)–(12), the far field electric
We will adopt the procedure of Maxwell by using the spherical potential becomes:
coordinate system (r, w, h) with origin located at the center of
sphere of radius a. Since the composite is regarded as a homoge- 1
V 1 ¼ B0 r cos h þ ðB1  DB1 þ DB2 Þ cos h ð13Þ
neous continuum with an effective electric resistivity Re from the r2
mean-field theory, the first spherical harmonic mode of the electric Since the composite of Fig. 3d is the same as that of Fig. 3a, the
background potential at a great distance is given by: far field electric potential given by Eq. (13) should be identical to
V 1 ¼ B0 r cos h ð8Þ the back ground potential given by Eq. (8). Consequently, we
require that:

Fig. 3. Illustrations to reconstruct Maxwell model for effective electric resistivity of heterogeneous composite with spherical particles.
976 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979

B1  DB1 þ DB2 ¼ 0 ð14Þ parameter to be determined experimentally. The consideration is

now given to a single element shown in Fig. 5b.
which is written equivalently into: The effective electric resistivity R2e for the phase-2 particles
under contact can now be expressed as:
R1  Re R2  Re R2e ¼ R2 þ Rc ð16Þ
/þ / ¼0 ð15Þ
2R1 þ Re 2R2 þ Re s
With R2e, the phase-2 can be regarded as continuum so that Eq.
where / ¼ 1  /s is the volume fraction of the phase one. (15) can be employed. After replacing R2 by R2e, Eq. (15) becomes:
From Eq. (15), it is seen that the value of Re is between R1 and R2, R1  Re R2e  Re
since the two terms have to be of the opposite signs, i.e., the con- /þ / ¼0 ð17Þ
2R1 þ Re 2R2e þ Re s
tributions to the far field potential by the two phases in the control
volume cancel. Eq. (15) remains the same if we interchange (R1, /) From the electrical and thermal analog under steady conduction
and (R2, /s), which implies that phases 1 and 2 are symmetric in and in terms of thermal conductivity, Eqs. (16) and (17) become:
nature. Therefore, we conclude that Eq. (15) is applicable to the 1 1
composite where both phases are continuum as that depicted in ¼ þ Rc ð18Þ
kse ks
Fig. 4. The use of n spheres to derive Eq. (15) based on Maxwell’s
approach does not imply that the phase-2 particles are dispersed. ke  kf ke  kse
The employment of superposition implies no contact resistance /þ / ¼0 ð19Þ
2ke þ kf 2ke þ kse s
among particles that renders the phase-2 being continuous. For
these reasons, we will anticipate that the effective electric resistiv- where the subscripts 1 and 2 are replaced by f and s respectively. Eq.
ity as predicted by Eq. (15) will be the lower limit of the composite, (19) is quadratic in ke which can be solved readily to give the pos-
i.e., upper limit for effective conductivity. For a composite consist- itive solution for ke as
ing of continuum substrate filled with discrete spherical particle, 1
higher effective resistivity (lower effective conductivity) is ke ¼ 3/ ðkse  kf Þ þ ð2kf  kse Þ
4q s i
expected. 2
To model the effective electric resistivity for the composite con- þ ½3/s ðkse  kf Þ þ ð2kf  kse Þ þ 8kse kf ð20Þ
sisting of continuum substrate (phase-1) filled with spherical par-
Eqs. (18) and (19), or (18) and (20), are the reconstructed Max-
ticles (phase-2), we introduce a contact resistivity Rc among
well model for the effective thermal conductivity of composites
particles to the model. In practice, there may be many contacts
with particle fillers, which has taken the thermal contact resistance
for a sphere. For simplicity, these contacts can be lumped statisti-
among particles into consideration. It is noted that when Rc = 0, the
cally into one as shown in Fig. 5a, with Rc being regarded as a
particles are in full contact which has a structure similar to that
given by Fig. 4, or to that of a sintered metal composite. This case
of Rc = 0 represents the upper limit of ke for the reconstructed Max-
well model. In practice, the lumped contact thermal resistivity Rc in
the newly reconstructed Maxwell model can only be determined
by fitting to experimental data.

5. Evaluations of the reconstructed Maxwell model

The effective thermal conductivities of composite materials

were systematically investigated experimentally and theoretically
by Gao et al. [18]. They obtained the composite samples by mixing
methyl vinyl silicone rubber matrix (kf = 0.15 W/m K) with spheri-
cal a-Al2O3 particles (ks = 30.0 W/m K) of four different mean
diameters over a wide range of volume fractions. The effective
thermal conductivities of composite samples were measured with
a thermal analyzer under steady conduction condition. They found
Fig. 4. Composites with symmetrical structure of two phases.
that for each composite at fixed mean particle diameter, the mea-

Fig. 5. Contact resistivity between two spherical particles.

 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979 977

sured effective thermal conductivity increases with the increase of Table 1

volume fraction of Al2O3 particles, but the rate of increase also The values of Rc , kse and Rc obtained from fitting experimental data of Gao et al. [18] to
the reconstructed Maxwell model.
increases with increasing volume fraction. For fixed volume frac-
tion of Al2O3 particles, the effective thermal conductivity is higher Average size (lm) Rc (W/m K) kse (W/m K) Rc
for larger mean particle diameter. 3 0.250 3.52 7.50
The experimental data of Gao et al. [18] are now used to evalu- 10 0.237 3.69 7.11
ate the reconstructed Maxwell model given by Eqs. (18) and (20). 35 0.198 4.33 5.94
75 0.166 5.01 4.98
Fig. 6 shows the comparisons of experimental data and model pre-
dictions with different fitted results of Rc corresponding to respec-
tive mean particle diameters. As shown in Fig. 6, excellent
agreements are found between model predictions and experimen- To give a complete picture of the reconstructed Maxwell model,
tal data. The fitted values of Rc as well as of kse are given in Table 1. the predictions by Eqs. (18) and (20) with different parametric val-
As seen in Table 1, the value of kse increase with increasing mean ues of Rc are plotted systematically as dimensionless ke =kf vs. /s in
particle diameter, and also are considerably lower than that of ks. Fig. 7 for fixed ks =kf ¼ 200 and vs. ks =kf in Fig. 8 for fixed /s ¼ 0:6.
The dimensionless values of Rc ð¼ Rc  ks Þ in last column of Table 1 In Fig. 7, Rc ¼ 0 represents the upper limit of the reconstructed
represent the ratio of contact thermal resistivity to particle thermal Maxwell model as mentioned earlier, while the experimental data
resistivity. of Gao et al. [18] fall in the range of Rc between 4.98 and 7.50. The
For comparisons, the predictions by the original Maxwell curve of Rc ¼ 17:4 fits barely well to the original Maxwell model.
model of Eq. (7) and by the layers-in-parallel and layers-in-series The lower limit of ke occurs when Rc is approximately equal to
models of Eqs. (1) and (2) respectively are also plotted in Fig. 6. 61.2, which corresponds to the best fit to the predictions by the
It is obvious that the original Maxwell model is only consistent layer-in-series model. Fig. 8 shows that ke =kf is mainly character-
with experimental data at very low Al2O3 volume concentration. ized by the contact parameter Rc . At very large ks =kf (say greater
The models of layers-in-parallel and layers-in-series indeed than 2000), the effective thermal conductivity increases linearly
provide the upper and lower limits of the effective thermal with the effective particle thermal conductivity (kse ).
conductivities.

Fig. 6. Comparison of effective thermal conductivities as predicted from various theoretical models with experimental data of Gao et al. [18] for different mean particle
diameters: (a) 3 lm, (b) 10 lm, (c) 35 lm, and (d) 70 lm.
978 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979
Fig. 7. Variation of ke =kf with /s based on reconstructed Maxwell model using Rc as
a parameter when ks =kf ¼ 200.
Fig. 8. Variation of ke =kf with ks =kf based on reconstructed Maxwell model using Rc
as a parameter when /s ¼ 0:6.
6. Conclusions
In this study, the effective thermal conductivity of composites
consisting of continuum matrix and particle fillers is investigated
theoretically by reconstructing the classical Maxwell model and
taking account of the thermal contact resistance among particles.
The modeling approach of Maxwell was first reviewed to demon-
strate the salient feature of his approach. However, a close exami-
nation of Maxwell’s approach reveals that there are several issues
(deficiencies) inherent in his modeling. These issues arise from the
misuse of the continuum matrix as an infinitely-extended ambient
and from the insensitiveness of far-field potential to the detailed
geometries of near-field particles. This leads to the issue of non-
uniqueness in the Maxwell model. The employment of linear
superposition by the potential theory also implies that there has
no contact resistance among particles, i.e., particle are in contact
perfectly as a continuum medium. Consequently, Maxwell model
is not applicable to composites with dispersed particles which
was commonly understood by many researchers. To resolve these
issues in Maxwell model, a new approach is adopted by consider-
ing an infinitely-extended heterogeneous composite which has an
effective conductivity based on potential mean-field theory. The
effective conductivity is then determined by requiring no contribu-
tion to far field potential after the spherical particles in a mesoscale
control sphere are removed and then replaced by linear superposi-
tion. The new model so constructed is only applicable for compos-
ites where both phases are continuum. For composites of
continuum matrix filled with discrete particles, contact resistivity
among particles is taken into consideration to reconstruct the Max-
well model.
The newly reconstructed Maxwell model with contact resistiv-
ity as fitting parameter is found to fit excellently to the experimen-
tal data of Gao et al. [18] over the whole range of particle volume
fraction, while the predictions from classical Maxwell model are
consistently much lower than the experimental data. It is noted
that the models of Hsu et al. [23] and Agari and Uno [27] were
found also to fit excellently well to the experimental data of Gao
et al. [18]. However, both models of Hsu et al. [23] and Agari and
Uno [27] are based on the circuit network approach, which is phys-
ically controversial as demonstrated in the recent paper of Xu et al.
[28]. In addition, two fitting parameters are needed in the models
of Hsu et al. [23] and Agari and Uno [27] to fit the experimental
data of Gao et al. [18], while only one is needed in the newly recon-
structed Maxwell model.
The scope of applicability of the newly reconstructed Maxwell
model was investigated by plotting the dimensionless effective
thermal conductivity ke =kf as a function of particle volume fraction
/s for fixed conductivity ratio at ks =kf ¼ 200 (Fig. 7), and as a func-
tion of ks =kf for fixed /s ¼ 0:6 (Fig. 8), using Rc , the ratio of contact
thermal resistivity to particle thermal resistivity, as a parameter.
The upper limit of the model is represented by Rc ¼ 0, while the
lower limit occurs near Rc ¼ 61:2, as the best fit to the layer-in-
series model. The experimental data of Gao et al. [18] fall in the
range of Rc between 4.98 and 7.50. When ks =kf is very large (say
greater than 2000), the effective thermal conductivity increases
linearly with the effective particle thermal conductivity (kse ).
Acknowledgements
We appreciated the financial support from Guangdong Province
Innovation R&D Team Plan for Energy and Environmental Materials
(No. 2009010025) and National Key Basic Research Program of
China (No. 2014CB932400).
References
[1] R.S. Prasher, Surface chemistry based model for the thermal contact resistance
of fluidic interstitial thermal interface materials, J. Heat Transfer 123 (2001)
969–975.
[2] D.D.L. Chung, Thermal interface materials, J. Mater. Eng. Perform. 10 (2001)
56–59.
[3] H. Ishida, S. Rimdusit, Very high thermal conductivity obtained by boron
nitride-filled polybenzoxazine, Thermochim. Acta 320 (1) (1998) 177–186.
[4] Y.S. Xu, D.D.L. Chung, C. Mroz, Thermally conducting aluminum nitride
polymer-matrix composites, Compos. A: Appl. Sci. Manuf. 32 (12) (2001)
1749–1757.
[5] C.W. Nan, G. Liu, Y.H. Lin, M. Li, Interface effect on thermal conductivity of
carbon nanotube composites, Appl. Phys. Lett. 85 (16) (2004) 3549–3551.
[6] K.M.F. Shahil, A.A. Balandin, Thermal properties of graphene and multilayer
graphene: applications in thermal interface materials, Solid State Commun.
152 (15) (2012) 1331–1340.
[7] M. Smalc, J. Norley, R.A. Reynolds, R. Pachuta, D.W. Krassowski, Advanced
thermal interface materials using natural graphite, in: International Electronic
Packaging Technical Conference and Exhibition, American Society of
Mechanical Engineers, 2003, pp. 253–261.
[8] I.H. Tavman, Thermal and mechanical properties of aluminum powder-filled
high-density polyethylene composites, J. Appl. Polym. Sci. 62 (12) (1996)
2161–2167.
[9] A. Boudenne, L. Ibos, M. Fois, J.C. Majesté, E. Géhin, Electrical and thermal
behavior of polypropylene filled with copper particles, Compos. A: Appl. Sci.
Manuf. 36 (11) (2005) 1545–1554.
 J.Z. Xu et al. / Applied Thermal Engineering 102 (2016) 972–979
[10] L.C. Sim, S.R. Ramanan, H. Ismail, K.N. Seetharamu, T.J. Goh, Thermal
characterization of Al2O3 and ZnO reinforced silicone rubber as thermal pads
for heat dissipation purposes, Thermochim. Acta 430 (1) (2005) 155–165.
[11] J. Xu, A. Munari, E. Dalton, A. Mathewson, K.M. Razeeb, Silver nanowire array-
polymer composite as thermal interface material, J. Appl. Phys. 106 (12) (2009)
124310.
[12] M.J. Adamson, Thermal expansion and swelling of cured epoxy resin used in
graphite/epoxy composite materials, J. Mater. Sci. 15 (7) (1980) 1736–1745.
[13] D. Kumlutas, I.H. Tavman, M.T. Coban, Thermal conductivity of particle filled
polyethylene composite materials, Compos. Sci. Technol. 63 (1) (2003) 113–
117.
[14] X. Lu, G. Xu, Thermally conductive polymer composites for electronic
packaging, J. Appl. Polym. Sci. 65 (13) (1997) 2733–2738.
[15] H. Ishida, Composition for Forming High Thermal Conductivity
Polybenzoxazine-based Material and Method, U.S. Patent, 1999-5-4,
5,900,447.
[16] Q.H. Mu, S.Y. Feng, G.Z. Diao, Thermal conductivity of silicone rubber filled
with ZnO, Polym. Compos. 28 (2) (2007) 125–130.
[17] J.C. Maxwell, A Treatise on Electricity and Magnetism, Clarendon Press, 1881.
[18] B.Z. Gao, J.Z. Xu, J.J. Peng, F.Y. Kang, H.D. Du, J. Li, S.W. Chiang, C.J. Xu, N. Hu, X.S.
Ning, Experimental and theoretical studies of effective thermal conductivity of
composites made of silicone rubber and Al2O3 particles, Thermochim. Acta 614
(2015) 1–8.
[19] R.G. Deissler, J.S. Boegli, An investigation of effective thermal conductivities of
powders in various gases, ASME Trans. 80 (1958) 1417–1425.
979
[20] D. Kunii, J.M. Smith, Heat transfer characteristics of porous rocks, AIChE J. 6
(1960) 71–78.
[21] P. Zehner, E.U. Schlunder, Thermal conductivity of granular materials at
moderate temperatures (in German), Chemie. Ingr.-Tech. 42 (1970) 933–941.
[22] C.T. Hsu, P. Cheng, K.W. Wong, Modified Zehner-Schlunder models for
stagnant thermal conductivity of porous media, Int. J. Heat Mass Transf. 37
(17) (1994) 2751–2759.
[23] C.T. Hsu, P. Cheng, K.W. Wong, A lumped-parameter model for stagnant
thermal conductivity of spatially periodic porous media, J. Heat Transfer 117
(2) (1995) 264–269.
[24] L. Rayleigh, On the influence of obstacles arranged in rectangular order upon
the properties of a medium, Lond., Edin., Dublin Philos. Mag. J. Sci. 34 (211)
(1892) 481–502.
[25] D.P.H. Hasselman, L.F. Johnson, Effective thermal conductivity of composites
with interfacial thermal barrier resistance, J. Compos. Mater. 21 (6) (1987)
508–515.
[26] M. Shaker, E. Birgersson, A.S. Mujumdar, Extended Maxwell model for the
thermal conductivity of nanofluids that accounts for nonlocal heat transfer,
Int. J. Therm. Sci. 84 (2014) 260–266.
[27] Y. Agari, T. Uno, Estimation on thermal conductivities of filled polymers, J.
Appl. Polym. Sci. 32 (7) (1986) 5705–5712.
[28] J.Z. Xu, B.Z. Gao, F.Y. Kang, H.D. Du, A statistical model for effective thermal
conductivity of composite materials, Int. J. Therm. Sci., 2016 (in press).