Three-Phase Lewis-Nielsen Model for the Thermal

Conductivity of Polymer Nanocomposites

R. Kochetov, A.V. Korobko, T. Andritsch, P.H.F. Morshuis, S.J. Picken, J.J. Smit
Delft University of Technology, the Netherlands
II. EXPERIMENTAL

Abstract- The thermal conductivity of polymer-based

microcomposites has been investigated for a long time. The
existing theories predict the thermal conductivity of the
polymers filled with conventional sized microparticles
quite well. Significant effort has been made in studying
and developing of the thermal conductivity behavior of a
relatively new class of composites, which contain nanosized filler. A considerable volume of the polymer matrix
has a modified structure with respect to the bulk polymer
due to polymer-filler bonding via a silane coupling agent
(SCA), which has been applied used to improve the
thermal contact between the individual components. The
thermal conductivity of a nanocomposite depends on the
interfacial layer, which can be defined as a transition layer
between a host material and incorporated nanofiller,
rather then the thermal conductivities of the constituents.
A composite material can be represented by composite
particles embedded into the polymer matrix. A composite
particle consists of a nanoparticle and the polymer close to
the particle surface, which is organized by the surface
modification. We propose a model for the thermal
conductivity of the polymer nanocomposites. This model is
intended to be used for systems which consist of a polymer
matrix, nanofiller and the interfacial layer around the
nanoparticles which has different properties than the bulk
polymer.

A. Materials
The base material chosen for this study was a commercially
available epoxy resin (ER) system, which consists of a
diepoxide-bisphenol-A Araldite CY231 and anhydride-type
hardener Aradur HY925. The fillers used in this study were
aluminum oxide (Al2O3), aluminum nitride (AlN) and
magnesium oxide (MgO).
The fillgrades were 0.5%, 2%, 5% and 10% wt. for all filler
types (+ 15 wt.% Al2O3). Unfilled epoxy samples were used
for comparison and reference measurement.
B. Sample Preparation
The samples were successfully fabricated using ex-situ
polymerization for surface functionalized nanoparticles. A
schematic representation is shown in Fig. 1.
Surface treatment of the particles was realized by
silanization, in order to improve the compatibility of the host
polymer and the ceramic filler, by organizing physical and
chemical bonding between the dissimilar materials [5, 6]. In
addition, it was done to obtain a uniform dispersion of
nanoparticles in the polymer matrix.
C. Characterization of the Samples
The crystalline structure of as-received particles was
examined by X-ray diffraction (XRD) spectrometry [7].

I. INTRODUCTION
Ethanol + SCA

The thermal conductivity value of composite containing

microparticles can be calculated by taking into account the
shape and volume of the incorporated particles, assuming
diffusive heat conduction in both phases. This approach
cannot be applied to a system with nanoparticles inside [1].
Various factors have to be taken into consideration for
nanocomposites, which can be disregarded when dealing with
microscale particles. Interface resistance and phonon
scattering become increasingly important in case of nano-scale
particles [2]. The incorporated surface modified nanoparticles
reorganize the structure and change the properties of a
polymer in the vicinity of filler [3]. The interfacial layer,
which can be defined as a transition layer between a host
material and incorporated filler, has different crystallinity,
glass transition temperature, crosslink density, permittivity,
thermal conductivity, etc. In some cases the physico-chemical
micro- and macro-properties of the interfacial layer play more
important role than the properties of the components [4].

Al2O3 / AlN / MgO nanoparticles

Epoxy resin
Ultrasonication

Ultrasonication

Mixing with high shear force

Weight monitoring Hardener

Degassing

Evaporation of ethanol

Mixing with high shear force

Pre-heating and treatment of Al molds
Degassing

Casting

Curing

Cleaning with alcohol

Specimen

Post curing

Fig. 1. Schematic representation of the nanocomposite preparation.

978-1-4577-0986-9/11/$26.00 2011 IEEE

338

the properties of this region are different compared to the bulk

ER [4]. The introduced surface modified nanoparticles
reorganize the polymer structure around them. The molecular
chains align perpendicular to the particles due to the effect of
silane coupling agent [3]. SCA connects to the inorganic
nanoparticle from one side and to the epoxy chain from other
side. The thermal conductivity of the layer is higher than that
of the bulk amorphous matrix, which is not affected by surface
functionalization of the particles.
Thus, a composite material can be represented by CPs
embedded into the polymer matrix. The CP has a
radius ( r + l ) , where r is the radius of a nanoparticle and l is

Morphological observations of as-purchased nanoparticles

and thin slices of created nanocomposites were realized by
means of transmission electron microscopy (TEM) [8, 9].
The thermal conductivity measurements were realized with
a Thin Heater Apparatus System (THASYS), produced by
Hukseflux Thermal Sensors. The accuracy of the
measurements is 6%. Each data point corresponds to an
average value of 4 measurements.
III. MODEL AND DISCUSSION
Lewis and Nielsen improved the Halpin-Tsai equation that
has been proposed for mechanical properties of composite
materials [10-12]. Originally aiming at the elastic moduli of
the composites, the Lewis-Nielsen model was adopted to
estimate the thermal conductivity of the composites. Using the
following formulas, one can do the basic estimations regarding
the thermal conductivity of the two-phase system, according to
Lewis-Nielsen model:

c = m

1 + f
1 f

4
4
3
CP = ( r + l ) n = r 3 n 1 + l r
3
3

and = 1 +

(1 M )
M2

= f (1 + ) ,

where n is the particle number per volume and = l

(1)

( )
where =
( + )
f

the thickness of the polymer layer that was affected due to

alignment of the polymer chains. The volume concentration
( CP ) of the CPs can be calculated as
(2)

is the

ratio of the interface layer thickness to the radius of the

nanoparticle [17].
A CP has a volume = f + l , where f is the volume of a

f .

particle and l is the volume of layer surrounding this

particle. The thermal conductivity of a CP ( F ) can be written
as follow:

c , m and f are the thermal conductivities of composite,

matrix and filler, respectively, f is the volume fraction of
filler content. The term is the shape factor, which takes into
account the shape of the filler, mainly the aspect ratio. The
term M is the maximum volumetric packaging factor of the
filler, which is sensitive to filler geometry. The relation
= ( f m ) ( f + m ) is coupling the conductivities of the

f
l
1
=
+
,
F f ( f + l ) l ( f + l )
where

components and the geometry of the filler. Finally the factor

was introduced to incorporate the maximum concentration
of particles possible to embed into the polymer matrix.
In a multiphase system there can be a strong scattering of
phonons, which occurs when the phonons propagate through a
boundary separating one phase from another. Therefore, the
large interfacial area plays a dominant role for the phonon
scattering mechanisms inside a polymer composite [13].
For nanocomposites, a thermal expansion and acoustic
mismatch and possible weak mechanical or chemical contact
at the interface, may lead to ineffective transport of phonons
through the interface. This is the so-called interfacial thermal
resistance (Kapitza resistance of an interphase boundary). It
provides a temperature discontinuity at the particle-polymer
interface, which vanishes when the particle length has a
dimension of about 100 nm or less [14-16].
In order to include the effect of Kapitza resistance, we
propose a composite particle (CP), which consists of a
particle and the polymer close to the particle surface, which is
structured by the surface modification. We defined the aligned
polymer layer around a nanoparticle as a separate phase since

+ l )

(3)

is the volume fraction of the filler in the

composite particle and

+ l )

is the volume fraction of

the interface layer around a nanoparticle. F can be written

via Kapitza resistance RK :

F =
1+

f
RK f

(4)
d

and RK = l l , where d is the average particle size and l is

the thermal conductivity of the interfacial layer [18]. In case of
very small particles the term RK / d converges to infinity,
RK / d . Therefore, the filler is not involved in the
thermal conductivity and the effective thermal conduction of
particle is zero, F = 0 . For large particles, the interfacial
resistance is not important, since RK / d 0 .
The thermal conductivities of neat ER and nanocomposites
filled with different types of nanoparticles are presented in

339

Table I and Fig. 2. As it was expected, the addition of the

inorganic filler raised the thermal conductivity of the epoxy
matrix, but not by much.
In our fitting analysis we assumed RK = 107 m2K/W for
all particles. Since the thermal conductivity of the ER is
known, one can calculate the Kapitza length, lK . Kapitza

Thermal conductivity, W/mK

0.21

length ( lK = 16 nm for our composites) is the length of the

aligned polymer layer in the CP [16, 19]. It is important to
notice that the values of the thermal conductivity for the filler
and the interfacial layer of the polymer are unknown.
Calculating the three-phase model with the interface thickness
equal to 16 nm for our experimental systems, we can derive
the effective thermal conductivity of the particle + interfacial
layer, i.e. CP. The unknown values of the thermal conductivity
of CP and shape factor are estimated by fitting experimental
data to the model. As shown in Table II, the thermal
conductivity F becomes smaller if the size of the CP
decreases, e.g. 0.36 W/mK for 22 nm MgO and 0.38 W/mK
for Al2O3 particles opposed to 1.11 W/mK for 60 nm AlN
particles. The thermal conductivity of neat ER might vary in
the range 0.170 0.02 W/mK. These variations are attributed
to the little changes in the ratio between epoxy and hardener
and time of polymerization between individual samples. The
shape factor ( ) appears to be different from 1.5, which
indicates the formation of aggregates. The values of are
obtained from fitting the model over all fractions of particles,
and therefore may reflect a value averaged over all
concentrations.
The experimental data obtained from thermal conductivity
measurements and calculated values using 3 phase Lewis and
Nielsen model as a function of filler loading are shown in
Figs. 3-5.

ER-AlN
ER-MgO
ER-Al2O3

0.20

0.19

0.18

0.17

0.00

0.01

0.02

0.03

0.04

0.05

0.06

Volume fraction

Fig. 2. Thermal conductivity of composites with different type of filler vs.

volume fraction.

.
0.21

Experimental nano-Al2O3

Thermal conductivity, W/mK

3-phase model fitting

0.20

0.19

0.18

0.17

0.00

0.01

0.02

0.03

0.04

0.05

Volume fraction

TABLE I
SPECIMENS INVESTIGATED AND THEIR THERMAL CONDUCTIVITIES
Specimen
Fillgrade, volume fraction
, W/mK
Neat ER
0.168
ER Al2O3 0.5
0.0015
0.173
ER Al2O3 2
0.0061
0.176
ER Al2O3 5
0.0157
0.182
ER Al2O3 10
0.0325
0.189
ER Al2O3 15
0.0506
0.203
ER AlN 0.5
0.0019
0.174
ER AlN 2
0.0075
0.179
ER AlN 5
0.0190
0.188
ER AlN 10
0.0393
0.205
ER MgO 0.5
0.0017
0.171
ER MgO 2
0.0068
0.175
ER MgO 5
0.0170
0.184
ER MgO 10
0.0359
0.200

Fig. 3. Experimental values of the thermal conductivity of ER-Al2O3

composites as a function of the filler loading (squares) fitted with the 3-phase
Lewis-Nielsen model (solid line).

0.21

Thermal conductivity, W/mK

Experimental nano-AlN
3-phase model fitting
0.20

0.19

0.18

0.17

TABLE II
THE FITTING PARAMETERS OF THE THREE-PHASE LEWIS-NIELSEN MODEL
ASSUMING 16 NM INTERFACIAL THICKNESS
Composite
m , W/mK

F , W/mK
( l + f ) f
ER-Al2O3 nano
ER-AlN nano
ER-MgO nano

0.171
0.171
0.169

2.5
3.1
4.3

0.38
1.11
0.36

0.00

0.01

0.02

0.03

0.04

Volume fraction

Fig. 4. Experimental values of the thermal conductivity of ER-AlN

composites as a function of the filler loading (squares) fitted with the 3-phase
Lewis-Nielsen model (solid line).

3.62
2.03
5.15

340

REFERENCES

Thermal conductivity, W/mK

0.20

[1] R. Krishnamoorti, R.A. Vaia, Polymer nanocomposites: synthesis,

characterization, and modelling, Washington: American Chemical Society,
2002.
[2] I.H. Tavman, Thermal conductivity of particle reinforced polymer
composites, International Communications in Heat and Mass Transfer 27(2),
pp. 253-261, 2000.
[3] T. Andritsch, Epoxy based nanocomposites for high voltage DC
applications. Synthesis, dielectric properties and space charge dynamics, PhD
thesis, TU Delft, 2010.
[4] P.C. Irwin, Y. Cao, A. Bansal, L.S. Schadler, Thermal and mechanical
properties of polyimide nanocomposites, IEEE Annual Report Conference on
Electrical Insulation and Dielectric Phenomena, pp. 120-123, 2003.
[5] E.P. Plueddemann, Silane coupling agents, Plenum press, 1982.
[6] Y-W. Mai, Z-Z. Yu, Polymer nanocomposites, Woodhead publishing
limited, 2006.
[7] R. Kochetov, T. Andritsch, U. Lafont, P.H.F. Morshuis, J.J. Smit, The
thermal conductivity in epoxy aluminum nitride and epoxy aluminum
oxide nanocomposite systems, Nordic Insulation Symposium (Nord-IS 09),
pp. 27-30, 2009.
[8] R. Kochetov, T. Andritsch, U. Lafont, P.H.F. Morshuis, J.J. Smit, Effects
of inorganic nanofillers and combinations of them on the complex permittivity
of epoxy-based composites, IEEE International Symposium on Electrical
Insulation (ISEI), pp. 340-344, 2010.
[9] T. Andritsch, R. Kochetov, B. Lennon, P.H.F. Morshuis, J.J. Smit, Space
charge behavior of magnesium oxide filled epoxy nanocomposites at different
temperatures and electric field strengths, IEEE Electrical Insulation
Conference (EIC), in press, 2011.
[10] L.E. Nielsen, Generalized equation fort he elastic moduli of composite
materials, Journal of Applied Physics 41(11), pp. 4626-4627, 1970.
[11] L.E. Nielsen, Thermal conductivity of particulate-filled polymers,
Journal of Applied Polymer Science 17(12), pp. 3819-3820, 1973.
[12] T.B. Lewis, L.E. Nielsen, Dynamic mechanical properties of particulatefilled composites, Journal of Applied Polymer Science 14(6), pp. 1449-1471,
1970.
[13] S.L. Shind, J.S. Goela, High thermal conductivity materials, Springer,
2006.
[14] C-W. Nan, R. Birringer, D.R. Clarke, H. Gleiter, Effective thermal
conductivity of particulate composites with interfacial thermal resistance,
Journal of Applied Physics 81(10), pp. 66926699, 1997.
[15] E.T. Swartz, R.O. Pohl, Thermal boundary resistance, Reviews of
Modern Physics 61(3), pp. 605-668, 1989.
[16] M. Hu, S. Shenogin, P. Keblinski, Molecular dynamics simulation of
interfacial thermal conductance between silicon and amorphous
polyethylene, Applied Physics Letters 91, 241910, 2007.
[17] W. Yu, S.U.S. Choi, The role of interfacial layers in the enhanced
thermal conductivity of nanofluids: a renovated Maxwell model, Journal of
Nanoparticle Research 5, pp. 167-171, 2003.
[18] W. Evans, R. Prasher, J. Fish, P. Meakin, P. Phelan, P. Keblinski, Effect
of aggregation and interfacial thermal resistance on thermal conductivity of
nanocomposites and colloidal nanofluids, International Journal of Heat and
Mass Transfer 51(5-6), pp. 1431-1438, 2008.
[19] A. Devpura, P.E. Phelan, R.S. Prasher, Size effects on the thermal
conductivity of polymers laden with highly conductive filler particles,
Microscale Thermophysical Engineering 5, pp. 177-189, 2001.
[20] D. Kumluta, I.H. Tavman, M.T. oban, Thermal conductivity of
particle filled polyethylene composite materials, Composite Science and
Technology 63(1), pp. 113-117, 2003.

Experimental nano-MgO
3-phase model fitting

0.19

0.18

0.17

0.00

0.01

0.02

0.03

0.04

Volume fraction

Fig. 5. Experimental values of the thermal conductivity of ER-MgO

composites as a function of the filler loading (squares) fitted with the 3-phase
Lewis-Nielsen model (solid line).

IV. CONCLUSIONS
The thermal conductivity of polymer systems containing a
small amount of surface modified nanoparticles is controlled
by the interfacial polymer layer, which acts as the main heat
conduction matter. A three-phase Lewis-Nielsen model was
proposed and used to fit the experimental data. The model fits
the experimental data accurately, but the obtained fitting
parameters are difficult to confirm experimentally, due to
complexity in determining the thermal conductivity of the
particles, the thickness of the interface layer and its thermal
conductivity. The main limitations of the three-phase model
come from these issues. The precise nature of the interfacial
layer between particle and polymer is not known. However,
the used interface thickness, which was found in literature,
satisfies the physical value of the thermal conductivity of a
nanocomposite. In our future work we need to quantify the
thickness of this layer and the exact values for the thermal
conductivity to make the model more concrete.
ACKNOWLEDGMENT
This work was performed for the nanoPOWER project,
which is sponsored by a Dutch government IOP-EMVT grant.
This work is part of the Research Programme of the Dutch
Polymer Institute (DPI), Eindhoven, the Netherlands, project
#623.

341