Progress in Polymer Science 61 (2016) 128

Contents lists available at ScienceDirect

Progress in Polymer Science

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

Review of thermal conductivity in composites: Mechanisms,

parameters and theory
N. Burger a,b, , A. Laachachi a , M. Ferriol b , M. Lutz c , V. Toniazzo a , D. Ruch a
a
Luxembourg Institute of Science and Technology, 5, avenue des Hauts-Fourneaux, L-4362 Esch/Alzette, Luxembourg
b
Universit de Lorraine, LMOPS E.A 4423, Metz F-57070, France
c
Thales Alenia Space, Mechanical & Thermal Technologies Dpt., 100 Boulevard du Midi, 06156 Cannes-La-Bocca, France

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

Article history: Thermal conductivity has become an important parameter for new technologies, especially
Received 8 January 2015 in aerospace and aeronautics. The advanced materials used in some electronic applications
Received in revised form tend to undergo an increase in temperature when they are used. Increasing the thermal
22 December 2015
conductivity of a material will help to diffuse heat faster to avoid substantial overheating
Accepted 3 May 2016
that could lead to premature degradation of the material.
Available online 12 May 2016
This article reviews theoretical and experimental aspects of thermal conductivity in com-
posites, from thermal energy generation to heat transfers. The fundamental mechanism of
Keywords:
thermal conduction, its mathematical aspects, and certain essential parameters to be con-
Thermal conductivity
Composites sidered in this study, such as crystallinity, phonon scattering, or ller/matrix interfaces
Enhancement are discussed in detail to examine their impact on thermal conductivity, complementing
Mechanisms several reviews in the eld.
2016 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Thermal conductivity: denition, mechanisms and parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Denition of thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. Basics of thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1. Theoretical mechanism in crystalline materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.2. Hardness and binding energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.3. Thermal conductivity in polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.4. Contribution of electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3. Mathematical aspect of thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.1. Thermal conductivity equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.2. General heat equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.3. Complexity of heterogeneous materials and models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Corresponding author at: Luxembourg Institute of Science and Technology, 5, avenue des Hauts-Fourneaux, L-4362 Esch/Alzette, Luxembourg.
Tel.: +352 42 59 91 4628.
E-mail address: nicolas.burger@list.lu (N. Burger).

http://dx.doi.org/10.1016/j.progpolymsci.2016.05.001
0079-6700/ 2016 Elsevier Ltd. All rights reserved.
2 N. Burger et al. / Progress in Polymer Science 61 (2016) 128

3. Thermal conductivity in composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3.1. Materials and methods used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2. Structures and defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.1. Defects in the crystalline structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.2. Phonon scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3. Incorporation of thermally conductive llers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.4. Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5. Inuence of dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.6. Fillers/matrix interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.7. Sizes and aspect ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.8. Filler alignment and networking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.9. Filler functionalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.9.1. Acid treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.9.2. Non-covalent functionalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.9.3. Covalent functionalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4. Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

Nomenclature
SWCNT single-walled carbon nanotube
thermal diffusivity t time
 wave length T temperature
 speed of a particle TiO2 titanium dioxide
 density UV ultraviolet light
 electrical conductivity V volume
c thermal diffusion characteristic time wt. percent weight (in % wt.)
heat ow rate x thickness
Ag silver
Al aluminum
AlN aluminum nitride 1. Introduction
BN boron nitride
C heat capacity In everyday life, heat and temperature are commonly
Cp specic heat capacity used and confused but these two fundamental concepts
Cal calorie should be distinguished from a scientic perspective: they
CB carbon black are directly related to thermal energy. Thermal energy
CNF carbon nanobers is dened by the microscopic vibrations of particles. The
CNT carbon nanotubes temperature, describing the state of a body, is a physical
Cu copper property quantifying those microscopic thermal vibrations
E energy of charged atomic oscillator of the particles. Heat, which is directly related to the ther-
h Planck constant mal conductivity, is dened as the thermal energy transfer
H2 SO4sulfuric acid from a particle to its adjacent particle(s), i.e., temperature
HNO3 nitric acid tells us how much particles are vibrating and heat evalu-
k thermal conductivity parameter ates how much of this energy is transferred, how fast and
ke electronic thermal conductivity (due to in what direction (Fig. 1)
electrons) These general concepts regarding heat transfer help us
kp phonic thermal conductivity (due to to introduce this review and are essential to understand
phonons) what is presented next. The fundamental and theoretical
l phonon mean free path mechanisms result from quantum physics and mechan-
Lc thermal diffusion characteristic length ics, with phenomena that are difcult to extrapolate at a
m mass macroscopic scale. This reviews objective is not to present
MWCNT multi-walled carbon nanotube thermal conductivity mechanisms at the quantum scale but
PCA pyrenecarboxylic acid to try to extrapolate and understand those mechanisms at a
Q thermal energy macroscopic scale. It complements the reviews by Han et al.
Q heat ux
S surface area
Si silicon

Fig. 1. Consequence of 2nd law of thermodynamics.

 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 2. Thermal conduction in matter by collision of particles.
[1] on the thermal conductivity of carbon nanotubes (CNTs)
and their nanocomposites or the reviews by Balandin et al.
[2] and Kumar et al. [3] addressing the exceptional thermal
properties of graphite/graphene. This review also updates
research that has been performed on thermal conductiv-
ity in advanced materials over the last decade [4,5], e.g.,
composites in aerospace, mainly concerning this problem.
Indeed, the metallic materials used, for instance in some
aerospace applications are being increasingly replaced
by high-performance polymers and composites, mainly
for the mass savings. Epoxy-based composites have been
developed, displaying some good mechanical and ther-
mal properties [6]. However, the thermal conductivity of
this type of material remains relatively low, compared to
metals. Thermosets, such as epoxy resins, are usually used
as thermally insulating materials; enhancing the thermal
conductivity of this type of material could be considered
unnatural. The challenge is then to improve their ther-
mal conductivity signicantly to achieve the properties
required for this type of application. This requirement
leads us to study and reconsider thermal conductivity
mechanisms in composite materials to understand how to
enhance this property.
2. Thermal conductivity: denition, mechanisms
and parameters
2.1. Denition of thermal conductivity
Heat transfer can occur through radiation, convection
or conduction [7,8]. It has been established that thermal
conduction is, in general, the main heat transfer mode
occurring in solid materials, especially conductors, even
though the optical depth plays a critical role in some insu-
lators plays a critical role. From a fundamental perspective,
thermal conduction corresponds to the transfer of a parti-
cles vibrational energy to adjacent particles without any
motion of the matter, mainly by collision (Fig. 2).
The complexity of the mechanism of thermal conduc-
tion through solids also arises from the many parameters
to be considered. Indeed, the materials nature is crucial for
thermal conductivity: for instance, plastics do not conduct
heat as well as metals. Size and proportions are naturally
important parameters: the thicker the material, the more
time for the heat to reach the opposite face. Morphologi-
cal aspects also play an important part in this mechanism,
and many other parameters must be studied to better
understand thermal conduction. In this section, some main
parameters are presented in detail, and their impact on
thermal conduction is examined.
3
2.2. Basics of thermal conductivity
2.2.1. Theoretical mechanism in crystalline materials
Before focusing on thermal conductivity in composites,
it is important to study and understand its fundamental
mechanism in crystalline materials. The most thermally-
conductive materials are highly crystalline materials, such
as diamond [9], graphene, metals, silicon and glass-
ceramics. The importance of crystallinity to thermal
conductivity has been well reviewed by Toberer et al.
[10], from a fundamental perspective. To elucidate the
fundamental aspects of thermal conduction [1113], we
illustrate the mechanism in a material with a crystalline
structure in Fig. 3.
This approach to the thermal conductivity mechanism
through a perfectly crystalline, ordered and rigid sample is
highly simplied, but nevertheless helps us to visualize the
phenomenon:
(a) Thermal energy is rst transmitted to surface atoms of
the sample.
(b) The surface atoms gain vibrational energy.
(c) The thermal energy of the surface atoms is then trans-
ferred to the adjacent atom(s) at the same speed, as in
a wave (modeled here in red lines).
(d) The thermal energy then diffuses in the sample.
(e) The thermal energy diffuses with a common vibrational
mode (or a phonon) to the whole crystal.
(f) When heat reaches the opposite surface of the sample, it
is then partially transferred by conduction or radiation
to the surroundings.
Fig. 4, adapted from The thermal properties of materials
by J. Ziman [14], clearly shows the vibrational and wave-
like aspects of the phonon transfer through a crystalline
structure. Thermal conductivity can also be theoretically
estimated by the Debye equation:
Cp vl
= (1)
3
where v is the average phonon velocity, and l is the phonon
mean free path.
The representation in Fig. 5 is closer to reality, with
the introduction of a temperature gradient, described by
the Fouriers law. The vibration of an atom leads to partial
vibrations of adjacent atom(s), propagating further into the
sample.
From these observations, some additional parameters
can already be deduced as directly impacting thermal con-
ductivity. Indeed, defects in a crystal, such as dislocations
4 N. Burger et al. / Progress in Polymer Science 61 (2016) 128

Fig. 3. Thermal conductivity mechanism in a crystalline material.

Fig. 4. Lattice vibration during heat transfer. [14], Copyright 2016. Reproduced with permission from Elsevier Ltd.

Fig. 5. Illustration of the temperature gradient in a crystalline material.

 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 6. Crystalline structure of diamond. [17], Copyright 2013. Reproduced
with permission from the author, SJ Sque.
or grain boundaries, would negatively inuence the ther-
mal conductivity. Others factors, such as hardness, atomic
density in the network, crystalline structure, and even
interatomic interactions, seem to be important parame-
ters for thermal conductivity, which could eventually be
calculated from the density of states and the BoseEinstein
equation [15].
2.2.2. Hardness and binding energy
Beyond the materials crystallinity, which is essential
for thermal conductivity, it is also important to consider
the hardness of the material, or more precisely the bind-
ing energy and stiffness of the structure, resulting from
the vibration freedom of the atoms. This rigid aspect is not
only complementary to the crystallinity, but probably even
more important for high thermal conductivity. Indeed,
a polymer can possess a crystalline structure (ordered
lamellae) that is very different that the crystal struc-
ture of diamond or a metal. As may be seen in Fig. 6
in the Bravais lattice of diamond, each carbon atom (in
the bulk) is totally blocked in 3 dimensions by the adja-
cent carbon atoms, which are themselves blocked by other
Fig. 7. Thermal conductivity mechanism in an amorphous polymer.
5
atoms, and so on. This situation partially explains why
the thermal conductivities of diamond (2000 W m1 K1 )
and a semi-crystalline polymer, e.g., 0.5 W m1 K1 for
semi-crystalline HDPE, are so different. Nevertheless, a
semi-crystalline polymer will display higher thermal con-
ductivity than its amorphous counterpart, but not on the
same order of magnitude as diamond or metals. Haggen-
mueller et al. [16] showed that HDPE (78% crystalline)
exhibits higher thermal conductivity than LDPE (33% crys-
talline).
2.2.3. Thermal conductivity in polymers
The heat diffusion mechanism is quite different for
amorphous and crystalline polymers, as shown simply in
Fig. 7.
Fig. 7 depicts the amorphous and disordered struc-
ture of a polymer is represented (no compact lattice, as
before). Chronologically, heat rst reaches a surface atom
or monomer, which is schematically the closest to the heat
source. The heat is then transferred to the adjacent atom,
then to the next one, and so on. The heat will not prop-
agate as a wave, as in the crystal (Fig. 3), but will diffuse
much more slowly in a polymer, causing disordered vibra-
tions and rotations of all the polymers atoms around their
equilibrium position, scattered to adjacent chains.
Fig. 8 is an image of the difference between a good and
bad thermal conductor, e.g., between a metal and a poly-
mer, as a function of their structure. To understand this
difference, these materials can be represented as a Newton
pendulum. Indeed, a classic ordered Newton pendulum can
symbolize a crystalline material, while a disordered pen-
dulum can symbolize an amorphous polymer. This gure
more precisely displays the vibrational aspect of particles
in both structures. An ordered structure (a) will diffuse the
initial vibration to the opposite face quite quickly, whereas
the initial kinetic energy is mostly spread into the atoms
of the disordered structure (b), causing vibrations of each
ball.
6 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 8. Schematic comparison by the Newton pendulum of thermal conductivity in crystalline materials (a) and polymers (b).
Regarding thermal conductivity, it could logically be
imagined that a polymers morphology is not ideal for
a good thermal transfer by conduction. The amorphous
aspect of polymers can be thermally compared to a combi-
nation of many defects, slowing the heat transfer. Polymer
chains are not organized as crystals, and thermal energy
will not be able to transfer quickly across the polymer
sample. Amorphous polymers, such as epoxy resins, are
not oriented according to a single direction, but more
randomly, contributing to thermal resistance. (The term
thermal resistance is used here to describe all phenomena
that decrease thermal conductivity, i.e., slow heat diffu-
sion.). In addition, some other fundamental parameters
such as anharmonicity, or different particle sizes, and bind-
ing energy can also explain the weak thermal conductivity
of polymers.
2.2.4. Contribution of electrons
Metals are particularly interesting materials in
terms of their thermal conductivity, especially copper
(k 380400 W/m K). Metals are widely used in appli-
cations requiring good thermal conductivity, such as
aluminum for satellite panels. Some studies have demon-
strated the proportionality between thermal and electrical
conductivities in metallic materials. It has therefore been
established that thermal conductivity in metals is mainly
due to electrons, according to the WiedermannFranz law:
ke = L0 e T
where the thermal conductivity due to electrons (ke )
is obtained by the product of the Lorentz constant
(L0 = 2.44 108  W K2 ), electrical conductivity ( e ) and
temperature (T).
More generally, according to this law, thermal conduc-
tivity can be decomposed into two mechanisms:
- Energized electron motions (ke )
- Atom interactions and collisions: phonons (kp )
Therefore [18]:
k = ke + kp
In the case of metals, thermal conductivity can be wholly
attributed to electron motions (electrical conductivity),
whereas the thermal conductivity due to phonons can be
considered as negligible, so:
k ke (4)
If we consider some carbon composites, such as epoxy-
based materials in our case, it would be interesting to
determine whether electrons contribute to heat transfer.
Considering the WiedermannFranz law (not particularly
adapted for these systems) and that the highest electri-
cal conductivity achieved by our systems is lower than
1 103 S/m, the resulting electronic thermal conductivity
would be estimated below 1 102 W m1 K1 (less than
2% of the apparent thermal conductivity of the material).
It is fairly difcult to estimate the electron contribution
during heat transfer in carbon composites, but it would
help to make this assumption that the electronic thermal
conductivity in composites can be considered negligible.
2.3. Mathematical aspect of thermal conductivity
2.3.1. Thermal conductivity equation
Thermal conduction is quantied by the thermal con-
ductivity parameter k in W/m K, dening how fast a
material can diffuse heat. The heat ow rate or Q (in
W or J/s) is dened as the energy fraction passing through
(2) the material in a certain amount of time dt:
dQ
= (5)
dt
The heat ux (in W/m2 ), corresponding to the heat
ow through a normalized surface dS, is as follows:
d
=

n (6)
dS
According to Fouriers law, the heat ux can also be
written as:
 = k grad(T )
(7)
where k is dened as a characteristic value of a mate-
(3) rial, i.e., having a constant positive value. The temperature
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128 7

gradient is a temperature function depending on time and The thermal diffusivity corresponds to the speed of the
the position in the sample. heat transfer in a material. In other words, thermal diffu-
Thus, sivity is considered as the propagation speed of diffusion in
the material. Alternatively, this parameter shows how fast
dQ dT (x, t)
=
= k grad(T ) = k 
n (8) a material would transfer heat from one face to the other.
Sdt dx
As seen in Eq. (15), the thermal diffusivity considers
k dT only the speed of the diffusion, not the amount of ther-
dQ = S dt (9)
dx mal energy considered in the heat transfer; this value is a
The heat ow is directly related to the difference in tem- transient-state property, usually determined by the Laser
perature between the ends of a sample at t = 0, by a constant Flash technique [19]. Characterization techniques are not
C, called the heat capacity (J K1 ) of the material, as follows: detailed in this review, but many references address this
eld [20,21]. Thermal diffusivity can also be mathemati-
dQ cally dened by characteristic length and time parameters,
C= (10)
dT respectively Lc and  c , as follows:
By denition, the heat capacity is the amount of heat
Lc = c (18)
that a material can store as a function of its volume or the
amount of heat required to increase its temperature by 1 C.
Hence, to diffuse twice as far, it takes four times longer.
In other words, each material requires a certain amount
Thermal diffusion is thus relatively low process at long dis-
of thermal energy to increase its temperature by 1 C. It
tances.
is also important to note that heat capacity is an exten-
sive value, directly proportional to the amount of matter.
The specic heat capacity (J kg1 K1 ) will then mainly be 2.3.2. General heat equation
used to characterize a material where heat capacity is nor- To conclude this mathematical section, it is also impor-
malized by the mass of the sample, usually measured at tant to understand the general heat equation. According to
constant pressure, as Cp : the 1st law of thermodynamics, the energy balance can be
written as:
C
Cp = (11)
m Qtotal = Qin + Qout + Qcreated (19)
The heat ow through a material is then related to heat
where the total energy in a system Qtotal results from the
capacity as follows:
sum of energy entering the system Qin (>0); the energy
dQ exiting the system, Qout (<0); and the energy released by
= m CP (12)
dT the system, Qreleased ( or
0). In detail,
Going back to the two Eqs. (9) and (12), we have:
- The stored energy of a system can be written as:
k dT
dQ = S dt = Cp m dT (13)   
dx
Qtotal = Cp dTdV (20)
Cp m dx dx v
k= (14)
S dx dt
- The energy released by the system (endothermic or
where: exothermic) is dened as:
dx dx (dx)
2   
= = (15)
dt dt Qreleased = dVs(t)dt (21)
m m V
= = (16)
S dx dV where s(x, y, z, t), is a heat source per volume unit (W m3 )
In Eq. (15), is dened as the ratio of an area and a - The energy balance between energies going in and out of
time. This characteristic value of a material is called the the system is:
thermal diffusivity in m2 s1 . This concept is described fur- 
ther hereafter. Moreover, Eq. (16) leads to the density of  dsdt =
Qin + Qout =  dVdt
div (22)
the material (kg m3 ). The thermal conductivity equation S
V
then becomes:

k = Cp  (17) or:

Therefore, thermal conductivity represents the capacity div  T = k 2 T

 = k div (23)
of a material to conduct heat. From a mathematical per-
spective, it is described as a thermal energy Q transferred Therefore, in Cartesian coordinates, the general heat
in a certain amount of time t, through the thickness x of a equation becomes:
material and perpendicularly to the surface area A, due to
a temperature difference
T between the opposite faces of T k 2 s
= T+ (24)
the sample. t Cp Cp
8 N. Burger et al. / Progress in Polymer Science 61 (2016) 128

The general heat equation is dened as a basic equa- In this case, by neglecting the external thermal source
tion for thermal diffusion, to dene the temperature at each (s) and energy created by the system, the equation
point as a function of time, using the following equation: becomes:
1
Tx,y,z,t = A + Berf (Fx,y,z ) (25) 2 T 2 T 2 T 2 T
t kxx 2
+ kyy 2 + kzz 2 + (kxy + kyx )
x y x xy
where the error function erf(u) is mathematically equiva-
lent to: 2 T 2 T T
+ (kyz + kzy ) + (kxz + kzx ) =  Cp (30)
x yz xz t
2
erf(x) = exp(u2 )du (26)

0

where: In anisotropic systems, such as graphite/epoxy sys-

1 tems, the heat equation solution requires numerical and
u = Fx,y,z
2.3.3. Complexity of heterogeneous materials and models
The general heat equation is unfortunately inapplica-
ble in the case of heterogeneous composites. Indeed, the
assumptions made to obtain Eq. (24) are not all satised for
heterogeneous materials, and some important parameters
need to be reevaluated:
1. Physical values from this equation are considered con-
stant with time. As the authors observed through their
work, thermal conductivity, density and heat capacity
are mostly functions of temperature. Theoretically, the
thermal conductivity of solids is described as
k = k0 (1 + aT )
where
- k0 is the materials thermal conductivity at absolute
zero.
(27) modeling methods [2325]. Usually, the nite element
modeling method is used to discretize the considered
system. Several models are proposed in the literature to
nd the best theoretical model, such as Finite Element,
Mori-Tanaka or molecular dynamics [26]. This partic-
ular model was developed by Mortazavi et al. [27], based
on the molecular dynamics-nite element model, to obtain
theoretical models as close as possible to the experimen-
tal results obtained in this study (e.g., for DGEBA/graphene
nanocomposites).
3. Thermal conductivity in composites
3.1. Materials and methods used
(28)
In this section, some of the results were obtained
experimentally by the authors. The epoxy resin used was
Bisphenol A Diglycidyl Ether (DGEBA) from SigmaAldrich
- a is a constant coefcient, characteristic of each mate-
rial.
Thus, a will be positive for insulating materials and
negative for thermally conductive materials. In other
words, when the temperature increases, an insulating
material partially loses its ability to isolate, and con-
versely, a thermally conductive material partially loses
its ability to conduct heat.
2. Heat is supposed to diffuse uniformly through the sam-
ples volume. However, the samples described here
are relatively heterogeneous composites, i.e., it can be
assumed that heat diffuses in a heterogeneous way
through these composites.
3. The materials structure is supposed to be isotropic and
homogeneous, with a constant thermal conductivity k.
As in the case of the previous assumption, the materi-
als studied here are epoxy-based composites, with llers
that can be anisotropic, such as carbon bers, or simply
having components with very different intrinsic ther-
mal conductivities. More precisely, in such a material,
the thermal conductivity would be dened as a tensor
[22] (with different values as a function of the position
in the sample and orientation of the llers):
kxx kxy kxz
kx,y,z,T = kyx kyy kyz (29)
kzx kzy kzz
Co. The hardener used was 4,4 -diaminodiphenylsulfone
from SigmaAldrich Co. Carbon nanobers (CNFs) were
obtained from Grupo Antolin S.A. Sulfuric and nitric acids
were obtained from Carl Roth Co. Non-covalent functional-
ization was performed using 1-pyrenecarboxylic acid from
SigmaAldrich Co.
Dispersion was performed by simple magnetic stirring,
except for the dispersion study, where a Bath Sonicator
Sonorex RK 103 H and a Hielscher Ultrasonic Processor
UP400 were also used. The thermal conductivity of the
samples was measured by Hot Disk TPS 2500 Standard
apparatus. A Quanta 200 FEG Scanning Electron Micro-
scope (SEM) was used for the microscopic observations of
the ller dispersion. The processing step of pressing was
performed under a Carver PLTN-I Press. Acid treatment
was performed using a sulfuric/nitric acid mixture (3:1,
v/v) at 80 C overnight in a 250 mL balloon. Samples were
than prepared with 5% wt. of CNF. The functionalization
was conrmed using a FTIR Spectroscopy Bruker Optics
Tensor 27 X-ray photoelectron spectroscope (XPS) with
a Hemispherical Energy Analyzer SPECS (PHOIBOS 150),
MALDI-TOF MS Bruker Daltonics Autoex and UV-Visible
Spectrometer PerkinElmer Lambda 35. The non-covalent
functionalization was performed using CNF/water and
pyrenecarboxylic/methanol mixtures at room temperature
for 2 h. The same analyses were performed as for acid treat-
ment to prove the success of the functionalization. Samples
were prepared with 10% wt. loading.
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 9. Defects in a carbon nanotube. [31], Copyright 2015. Reproduced
with permission from Kintech Laboratory.
3.2. Structures and defects
3.2.1. Defects in the crystalline structure
As explained in Section 2.2.1, crystallinity is an impor-
tant structural aspect of thermal conductivity. In a
crystalline material, the presence of an interface, a defect or
any discontinuity in the crystalline structure, as illustrated
in Fig. 9 with a carbon nanotube structure, will drastically
decrease the thermal conductivity. In the literature, this
thermal conductivity reduction at component interfaces is
usually called Kapitza resistance [28], quantifying this
thermal resistance [29]. The theory also species that
even with an interface where atoms of two identical com-
ponents are perfectly ordered and perfectly in contact with
each other, there would be a thermal resistance due to the
difference between vibration normal modes [30]. A normal
mode of an oscillating system is a particular frequency at
which an object can oscillate. In other words, quantum the-
ory tells us that only certain frequencies can make a body or
a particle vibrate, as a function of its structure, components,
and other factors. This anharmonicity in the vibration nor-
mal modes (different frequencies) is a fundamental cause
of thermal resistance, as described by Shind et al. [37].
Fig. 10 shows some classical defect examples found in
crystalline structures, such as point defects, dislocations or
grain boundaries. Many other types of defects can be found
in crystalline materials and will inuence thermal conduc-
tivity to a greater or lesser degree, but will always decrease
it. Kim et al. [32] also demonstrated that highly crystalline
CNTs possess a higher thermal conductivity than CNTs with
defects. In other words, the more defects are present in
Fig. 10. Phonon scattering in crystalline materials, due to various defects.
9
the crystalline structure, the lower the intrinsic thermal
conductivity of the crystal, as defects lead to phonon scat-
tering, i.e., phonons are reected, diffracted or refracted,
shortening the mean free path and hence reducing ther-
mal conductivity. Some articles address the improvement
of a llers crystallinity [33,34] to improve its properties.
Lim et al. [35] treated CNFs at 2800 C to reduce impurities
and increase crystallinity.
3.2.2. Phonon scattering
Any defect leads to some phonon scattering [36], i.e.,
phonons are no longer transferred as a single common
wave through the crystal. Without going any deeper into
the fundamental aspects of phonon scattering, this phe-
nomenon can appear through three main mechanisms
[37]:
- Phonon/defect scattering
This type of scattering can easily be imagined in an
imperfect crystal, as illustrated in Fig. 10. A defect in a
crystalline structure represents a barrier to phonon trans-
fer, leading to some scattering.
- Phonon/interface scattering
Because each sample has nite dimensions and thus
presents an interface or a disconnection, this type of scat-
tering is inevitable and depends on the sample geometry.
In composite materials, the many interfaces between
llers and matrix will display even more scattering.
- Phonon/phonon scattering
During heat transfer by conduction, atoms of the lattice
will vibrate to transfer thermal energy. In an ideal case,
all atoms are vibrating together at the same frequency,
i.e., harmonically. Actually, several frequencies or normal
modes are present in the material, i.e., atoms will have
anharmonic vibrations [38]. These multiple frequencies
explain how phonon/phonon scattering or Umklapp scat-
tering will occur in a sample. This type of phenomenon
can be represented by two initial rightward-moving
waves; when the two waves merge, the new wave
becomes leftward-moving. It has ipped its direction (or
Umklapp in German).
These phonon scattering mechanisms lead inevitably
to some sources of thermal resistance in a mate-
rial, but even more scattering phenomena contribute to
10 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
decreasing thermal conductivity, e.g., electron/defect scat-
tering, phonon/electron scattering, photon/photon scatter-
ing and others. A single atom modication in a crystalline
structure would lead to some scattering, as demonstrated
by Marchio et al., where 0.4% arsenic incorporated in copper
will lead to a 40% decrease in thermal conductivity [7]. In
the same way, Chen et al. [39] showed that even the addi-
tion of isotopic 13 C (50%) to a 12 C graphene sheet halved
the thermal conductivity. Singh et al. [40] also showed a
reduction in the intrinsic thermal conductivity of few-layer
graphene sheets compared to a single sheet. In addition,
other parameters will logically decrease thermal conduc-
tivity, such as porosity [41,42]. These parameters and their
inuence on thermal conductivity were modeled by Ran-
drianalisoa [43] from the Boltzmann Transport Equation,
using the analogy between the concepts of a photon and
phonon, and solved by the Discrete Ordinates method. It
is important to note here that the recent studies [4446]
on thermal conductivity at an atomic scale demonstrated
that each change in the perfect crystalline arrangement of
a material will inevitably lead to a decrease in the thermal
conductivity. It is now a scientic challenge to determine
how to use this knowledge to increase the thermal con-
ductivity of amorphous materials, such as polymers. For
instance, Kim et al. [47] achieved much higher thermal
conductivities with only cross-linked polymers and an
appropriately chosen linker structure, but for much higher
enhancement, the incorporation of thermally conductive
llers is required.
3.3. Incorporation of thermally conductive llers
To improve the thermal conductivity of non-conductor
materials, such as polymers, researchers have focused on
incorporating several types of llers or modifying the struc-
tural aspects of the material. Naturally, the improvement
rate of thermal conductivity depends on the llers, their
nature, their structure, and more. All type of thermally
conductive nanollers (mostly crystalline) were studied
[48,49] over the last years, leading to more or less com-
parable results: carbon-based llers [50] (graphite [51,52],
carbon nanotubes [53,54], carbon black [55], graphene
[5658]), metallic llers [59,60] (Ag, Cu [61], Al [62], TiO2 ,
AlN [63]), ceramic llers (BN [6466], Si [67], ZrB2 [68]),
vegetal bers [69], and so on. Using various ller combi-
nations [7074] and seeking some synergetic effect, many
authors showed the same improvement rates [75,76].
Since the attribution of a 3000 W m1 K1 theoreti-
cal thermal conductivity to carbon nanotubes, most of
the works involving thermal conductivity improvement
have focused on carbon nanotubes or even boron nitride
[77,78] for their exceptional intrinsic ability to conduct
heat. The review by Han et al. [1] perfectly summarizes the
worldwide research on CNT-nanocomposites to improve
thermal conductivity. After some rst pristine experimen-
tal results on the dispersion of simple carbon nanotubes
through a polymeric matrix, some excellent improvements
were observed in mechanical and electrical properties [79],
while the results obtained for thermal conductivity were
not as good as expected. Indeed, despite some improve-
ment, the results were not even close to the theoretical
value of the carbon nanotubes thermal conductivity. As
mentioned previously, the high intrinsic thermal conduc-
tivity of a carbon nanotube will be signicantly decreased
by simple incorporation into the matrix, mainly due to
defects and phonon scattering; Aliev et al. [80] describe
how bundling, tube-tube interconnections, defects or mis-
alignments decrease the effective thermal conductivity of
MWCNT from a theoretical 3000 W/m K to an actual value
of 50 W/m K. Currently, many methods are being studied
[81,82] to determine how to fully benet from the excep-
tional thermal conductivity of CNTs.
It can naturally be deduced that the higher the loading
of highly conductive llers, the higher the resulting ther-
mal conductivity would be. This phenomenon has been
demonstrated many times in the literature with several
articles. Nevertheless, despite extremely high loading lev-
els, thermal conductivity never reached the intrinsic values
of the ller, nor was even close. On average, 3050% wt.
loading is required to increase thermal conductivity by 10
times [83]. Choi et al. [84] combined Al and CNT particles
at 80% wt. to increase polyacrylates thermal conductiv-
ity by 10. Wattanakul et al. [85], for instance, obtained a
thermal conductivity of 1.97 W/m K with 28 vol% BN llers.
Lee et al. [86] needed 57 vol% AlN in epoxy matrix to
reach 3.4 W/m K. The best intrinsic heat-conductive ller
does not always lead to the highest thermal conductivity
in the nanocomposite. Weidenfeller et al. [87] reached a
2.5 W/m K thermal conductivity with 30 vol% talc in PP, and
only half with copper particles. Even in these cases, the
thermal conductivity of the composite does not evolve as
fast as expected, even with relatively long llers.
In this study, graphite nanollers were dispersed by the
authors in an epoxy matrix, for several loadings. The results
are displayed in Fig. 11.
As may be seen, the thermal conductivity increases with
graphite loading but does not reach an outstanding value.
Most studies display the same trend, while a few others
show some surprising results. Zhou et al. [88] discusses
a percolation threshold in thermal conductivity observed
with micro-SiC/epoxy composites at approximately 55%
wt. At this loading, the thermal conductivity increases
sharply. Evseeva et al. [89] obtained a decreasing trend in
the thermal conductivity with CNT loading over 3% wt. in
epoxy resin. Nevertheless, outstanding progress has been
made in enhancing thermal conductivity in composites by
incorporating conductive llers. Currently, with adapted
2D thermally conductive llers, such as graphene [9094],
the thermal conductivity of composites could reach simi-
lar values to metals, but additional work on the following
listed aspects is required to keep improving this property
and understand why a 2D conformation helped to consider-
ably enhance the thermal conductivity in composites. The
processing of the material is different in each case and can
impact the nal thermal conductivity.
3.4. Processing
The processing of a material is relatively important
for experimental studies, and more specically for indus-
tries. Indeed, defects at an atomic scale can have huge
consequences for thermal conductivity, and macroscopic
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128 11

Fig. 11. Experimental thermal conductivity of epoxy-nanocomposites lled with graphite nanollers, as a function of loading (performed by the authors).

defects will lead to even more serious consequences. For
instance, as briey introduced in Section 3.2, porosity is
a macroscopic defect that will denitely decrease ther-
mal conductivity. The microscopic voids present in the
polymeric structure lead to phonon scattering. There-
fore, logically should produce much larger voids to create
even wider phonon scattering. The authors conrmed this
assumption by comparing processing methods (pressing
and molding), as displayed in Fig. 12.
As may be seen, pressed samples show a higher thermal
conductivity than molded samples, but only above a certain
loading (here approximately 1012% wt.). Indeed, at this
loading, the samples show relatively high viscosity, which
is detrimental to good ller dispersion. More important,
the higher the viscosity, the larger the porosity of cured
samples. This problem is really minimized by the press-
ing process. Hence, pressed samples showed better thermal
conductivity [95]. Porosity is just one consequence of the
processing method used but is a good example to show
how the process can inuence the thermal conductivity
parameter.
Many other processing techniques were experimentally
compared to help improve the thermal conductivity. The
main processing techniques mentioned in the literature are

Fig. 12. Thermal conductivity comparison between molded and pressed samples (performed by the authors).
12 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 13. Experimental thermal conductivity vs. dispersion state of llers in nanocomposite (performed by the authors).
mechanical techniques, used rst to improve the dispersion
state of the llers into the matrix, such as sonication, roll-
mill [96,97], press [98], and mixer [99]. Chakraborty et al.
[100] showed that the roll-mill signicantly enhanced the
dispersion of CNTs into an epoxy matrix but also that the
CNTs re-agglomerate during the curing process. Ming et al.
[101] showed that thermal conductivity is two to four times
higher in directions perpendicular to the pressing direc-
tion for metal-organic frameworks loaded with graphite.
Chen et al. [102] modeled how the processing methods
impact the thermal conductivity, regarding the anisotropic
structure of the composite. The processing of the mate-
rial directly impacts the dispersion state of llers in the
matrix, which could inuence the thermal conductivity of
the resulting material.
3.5. Inuence of dispersion
In the literature, most of the outlooks proposed in recent
articles dealing with thermal conductivity improvements
mentioned the dispersion state as a major parameter to
be considered. Indeed, most of the llers incorporated into
a matrix will tend to agglomerate, especially during the
curing process [103]. Song et al. [104], for instance, showed
an improvement of the thermal conductivity with better
dispersion of CNT in an epoxy matrix. Therefore, through
several processes, the inuence of the dispersion state on
thermal conductivity was experimentally studied by the
authors (Fig. 13).
As shown by the authors, several dispersion states of
carbon nanobers (CNF) in an epoxy matrix were stud-
ied, with their respective effects on thermal conductivity.
Each sample was prepared with a 5% wt. loading, except
for the dispersion D, where approximately 7% wt. CNF was
used. Nevertheless, it may be demonstrated that dispersion
does not seem to be as important as claimed in most of the
articles. Indeed, with a better dispersion state than disper-
sion A, dispersion C, for instance, displays a lower thermal
conductivity. Moreover, even with the best dispersion state
and the highest loading, the dispersion D shows a ther-
mal conductivity barely higher than dispersion A. Wernik
et al. [105] demonstrated that the thermal contact resis-
tance between two CNTs rises sharply upon spacing them
out. Hence, creating more CNF/matrix interfaces is proba-
bly not the best way to enhance thermal conductivity. The
authors attributing the enhancement of thermal conduc-
tivity to the dispersion state usually focus on much lower
loadings. Karippal et al. [106] obtained higher thermal con-
ductivities with better dispersion of MWCNT in epoxy until
8 vol%. Dos Santos et al. [107] did not observe any change
in the CNT-nanocomposite thermomechanical properties
with enhanced dispersion. It is assumed that aggregated
CNFs form thermal paths and behave thermally as a larger
ller without any interfaces; enhancing dispersion only
adds interfaces in the nanocomposite. However, at low
loadings, isolated agglomerates in a poor dispersion are
not as effective as dispersed llers. Presumably, the large
mean inter-particle distances are more detrimental than
the interfaces caused by dispersion in this case.
Another study was performed by the authors, show-
ing the impact of aggregates on thermal conductivity. Two
epoxy-composites were prepared by incorporating 7% wt.
of expanded graphite (EG) and 3% wt. of carbon black
(CB) particles in different orders and through different dis-
persion processes. In Fig. 14, Sample 1 shows some huge
aggregates, up to 1.5 mm, entirely composed of CB, whereas
Sample 2 displays a much better dispersion of CB, with
aggregates below 500 m.
Compared to the pure epoxy resin (DGEBA/DDS) with
a thermal conductivity of 0.24 W m1 K1 , Sample 1 dis-
played a thermal conductivity of 0.56 W m1 K1 , whereas
Sample 2 reached only 0.51 W m1 K1 . This result conrms
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128 13

Fig. 14. EG-CB Composites Sample 1 (left) and Sample 2 (right) [performed by the authors].

Fig. 15. Theoretical representation of Sample 1 (left) and Sample 2 (right).

that a better dispersion does not necessarily imply a higher

thermal conductivity. As illustrated in Fig. 15, it appears
that the huge CB aggregates in Sample 1 can tend to tighten
the EG llers, whereas in Sample 2, the EG llers are bet-
ter dispersed. Because the EG llers are mainly responsible
for the increase in thermal conductivity (Fig. 11), squeez-
ing the EG llers reduces the mean inter-particle distance.
Hence, heat is transferred faster in Sample 1 than in Sample
2.
This result also conrms that with a physical percola-
tion threshold in the composite structure, no exceptional
enhancement in thermal conductivity is observed. Com-
pared to the usual electrical conductivity [108], the thermal Fig. 16. Representation of a nanocomposite loaded with llers.
conductivity property shows many differences, a simple
consequence of different fundamental mechanisms [109].
Additional studies should be performed, to determine how is created in the material. Thus a more or less irregular
the dispersion state inuences the thermal conductivity of ller/matrix alternation obtains along a given direction.
the resulting material. The mechanism of thermal conduc- In Fig. 16, several ller/matrix interfaces are encoun-
tance at the ller/matrix interfaces is one crucial issue to tered, following the x axis. Considering that this illustration
focus on in future work to better understand heat transfer is a small elementary representative volume of the whole
in composite materials. sample, many more interfaces can then be considered in the
nal macroscopic composite. Therefore, Fig. 17 can illus-
trate what happens at the interface between a thermally
conductive crystalline ller and the polymer matrix during
3.6. Fillers/matrix interfaces heat conduction.
During heat transfer by conduction through crystalline
As discussed above, when llers are dispersed in a llers (here in black), in a certain amount of time
t,
polymer matrix, e.g., an epoxy resin, some discontinuity thermal energy is theoretically considered to be totally
14 N. Burger et al. / Progress in Polymer Science 61 (2016) 128

Fig. 17. Schematic representation of thermal conductivity at ller-matrix interface.

diffused to the other side of the ller. In our system,

this extremity is partially in touch with a polymer chain
of the epoxy resin. Moyer et al. [110] were among the
rst to propose a theoretical and mathematical mecha-
nism of thermal conductivity and phonon scattering at
a metal/matrix interface. Wang et al. [111] showed that
incomplete contact at the interface leads to large phonon
scattering and thus thermal resistance. As explained pre-
viously, heat transfer through the polymer chain is much
less efcient than through the crystalline ller due to
the vibrations of the whole chain and phonon scattering.
Considering the same amount of time
t during thermal
conduction in the polymer chain, the distance traveled
by heat is much shorter than the equivalent path length Fig. 18. Drawing of a carbon nanotube surrounded by a polymer matrix.
through the ller, i.e., the polymers thermal conductivity [175], Copyright 2012. Reproduced with permission from Elsevier Ltd.
is much smaller than the llers. In addition, it seems that
the phonon scattering is likewise magnied by the many words, the thermal conduction of the nal composite is
ller/matrix contacts along a single ller, as illustrated in essentially governed by the intrinsic thermal conductivity
Fig. 18. of the most insulating component, in this case the polymer.
Furthermore, considering for instance that 10-fold more The ller/matrix interface represents a major parame-
time is required for heat to diffuse through the poly- ter to consider for improving thermal conductivity [112],
mer chain than through the crystalline ller (same path probably the most studied currently as summarized by
length) and that the sample is theoretically as represented McNamara et al. [113]. Gojny et al. [114], proposes that
in Fig. 17, i.e., half ller/half polymer, thermal energy would large interfaces (i.e., long distances between the ller and
mathematically require approximately 11
t to reach the the matrix) lead to strong phonon scattering. On one hand,
opposite face (10
t for polymer +
t for ller). In other the larger the interface, the more time the thermal energy
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 19. Schematic representation of composites with several short llers (a) and one single continuous ber (b); schematic representation of nanocom-
posites lled with spherical llers (c) and long llers (d).
of the polymer will take to be evacuated through ther-
mally conductive llers. On the other hand, the better the
compatibility at the interface, e.g., covalent bonds between
the llers and the matrix, the more thermal energy is
transferred from matrix to the conductive llers. Great
progress has been recently made regarding the impact of
the ller functionalization on thermal conductivity. Huang
et al. [115] found that the functionalization will not have
the same impact as a function of the ller concentration.
Above a critical concentration, covalent bonding results
in less effective improvement of thermal conductivity in
composites. Most papers addressing the reduction of the
Kapitza resistance by increasing heat transfer at the inter-
face [116] do not consider the ratio between the gain of
thermal conductivity at the interface and the loss of ther-
mal conductivity along the ller, which will be more or less
different according to the type of llers considered. This
competition between the two phenomena is further stud-
ied and explained in Section 3.9 and must be evaluated in
many different cases in future work to better understand
the thermal conductivity mechanism at the ller/matrix
interface. For example, much longer llers may be consid-
ered to reduce the number of ller/matrix interfaces.
3.7. Sizes and aspect ratios
The truly important point to highlight from Sections
3.5 and 3.6 is that the high intrinsic thermal conductivity
of the ller is not well used in nanocomposites. Indeed,
heat would quickly diffuse through the crystalline ller,
whereas it is considerably slowed by the polymer. We ben-
et from the high thermal conductivity of the ller, but
only along the ller. In other words, the longer the ller,
the higher the thermal conductivity of the material in the
llers direction (anisotropic property) [117,118]. In con-
trast, the shorter the ller, the lower is the enhancement of
thermal conductivity is [119]. It seems that nanoparticles
15
are probably not the best choice to obtain high thermal
conductivity composites [114,120].
Fig. 19 shows that the many short llers in situation (a)
present many ller/matrix interfaces. In this case, the more
interfaces, the more phonons will be scattered, leading to
thermal resistance. However, in situation (b), a single long
ller (equivalent in length) will not have these interface
problems: therefore, thermal conductivity will be higher.
Park et al. [121] showed that long MWCNTs (55 W/m K at
60% wt.) enhance thermal conductivity much better than
short MWCNTs (20 W/m K at 60% wt.) in epoxy composites.
Thus, the aspect ratio is also an important parame-
ter to consider [122125]. Comparison of the results with
spherical and one- or two-dimensional llers, may evaluate
how the aspect ratio will inuence thermal conductivity.
In Fig. 19, spherical llers (c) display more interfaces along
a single direction than long llers (d). In addition, since in
the same volume, spherical particles will always be shorter
than long llers long llers lead to higher thermal con-
ductivities than spherical llers [126,127]. This assumption
regarding aspect ratios was also demonstrated by Mor-
tazavi et al. [128], showing better enhancement of thermal
conductivity with platelet llers than spherical llers. Chir-
toc et al. [129] showed that longer expanded graphite with
higher aspect ratios is a better choice to enhance thermal
conductivity. Platelets seem to be particularly interesting
for thermal conductivity. One of the best improvements
obtained in the literature was obtained by Veca et al. [130]
with 25% GNPs (6.5 W/m K). Consequently, several mech-
anisms of heat transfer in composites can be proposed
(Fig. 20).
Epoxy resin, are thermally homogeneous, i.e., each poly-
mer chain or component will display similar low thermal
conductivity. Therefore, heat diffuses slowly and uniformly
through the resin, as a thermal wave. For a nanocompos-
ite, when long nanollers are well dispersed in the matrix,
heat diffusion is locally and very briey accelerated (along
16 N. Burger et al. / Progress in Polymer Science 61 (2016) 128

Fig. 20. Heat transfer mechanisms in composites, as function of size and aspect ratios of long llers.

the nanollers). However, at the end of the ller, it is con-
siderably slowed by the matrix. As explained in Section
3.6, the heat transfer is much more slowed by the polymer
matrix than sped up by the nanoparticles, which is why
the thermal conductivity enhancement is quite limited
with nanollers, and a large loading is required to reduce
the mean inter-particle distances and build better ther-
mal paths. Following this hypothesis, a microcomposite
with anisotropic microllers (aspect ratio = a), appears to
be more suitable to improve thermal conductivity. Consid-
ering the same loading as in the nanocomposite, better heat
transfer is expected due to far fewer ller/matrix interfaces.
As seen in Fig. 19, a long single ller is better adapted than
several short llers. Nevertheless, at the same loading, the
microcomposite will have much larger mean inter-particle
distances than the nanocomposite, which remains detri-
mental for good thermal transfer. Microllers with a much
higher aspect ratio (
a) can further reduce those distances mal conductivity. Indeed, Chu et al. [132] demonstrated
and build good thermal paths to achieve higher thermal
conductivities in the nal material.
In the same way, the authors compared the impact of
long and short expanded graphite particles on thermal
conductivity. More precisely, microparticles of expanded
graphite (below 10 m) were used in Sample A, particles
of expanded graphite (below 1 mm) were used to prepare
Sample B, and particles of larger expanded graphite (below
3 mm) were used to prepare Sample C (Fig. 21). All samples
were prepared in the same way, at 15% wt. in DGEBA/DDS
epoxy resin. The results are displayed in Fig. 22, and as
observed, the longer the ller, the higher the thermal con-
ductivity of the composite. With longer llers and higher
aspect ratios, there are fewer ller/matrix interfaces along
the llers direction. Therefore, longer llers will help to
reduce the resistance to heat transfer, and thus increasing
thermal conductivity.
Many other studies have also focused on the carbon
nanotubes shape [131], e.g., length, diameter, MWCNT
vs. SWCNT, and so on. The literature overview on this
ller is summarized in Section 3.3. It seems that any
change in the linear crystalline ller will decrease its ther-
that the waviness of graphene nanoplatelets signicantly
affects their intrinsic thermal conductivity. Xu et al. [133]
showed that the intrinsic thermal conductivity of graphene
is not only related to the length but also to the 2D
nature of graphene. Remarkable progress has been made
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128 17

Fig. 21. Top: Sample A; middle: Sample B; bottom: Sample C (performed by the authors).

Fig. 22. Thermal conductivity of epoxy-expanded graphite composites as a function of llers length (performed by the authors).
18 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
in understanding how these geometric parameters impact
the intrinsic thermal conductivity of the material, and
more studies are required for a complete understanding of
how these many phenomena combine to give the thermal
conductivity parameter. Modeling those phenomena effec-
tively would be very helpful in designing future materials
with extremely high thermal conductivities. Meanwhile,
research is focused on aligning ller arrays to increase
thermal conductivity, with minimum interfaces and linear
morphologies.
3.8. Filler alignment and networking
Because thermal conductivity is an anisotropic prop-
erty, a few recent studies have focused on improving
polymers thermal conductivity through a single direction
[134]. Chen et al. [135] demonstrated how intrinsic ther-
mal conductivity is also directly related to chiral angles.
Interfaces and short spherical llers, for instance, should
be avoided to achieve high thermal conductivity in a com-
posite. However, long continuous crystalline bers seem
to be an interesting approach to improving thermal con-
ductivity [136]. As a consequence, one possible way to take
advantage of the llers anisotropy is to align the ller along
the same axis [137], as shown in Fig. 23 to obtain a better
enhancement in the considered direction (quite difcult
experimentally) or through a network [138,139]. Shaikh
et al. [140] obtained a thermal conductivity of 8.3 W/m K for
an aligned CNT array on a glass substrate. Magnetic elds
have also been used several times in the literature to align
particles [141,142]. Choi et al. [143] focused on a 25 T mag-
netic eld to align CNTs in an epoxy matrix and improve
thermal conductivity. Marconnet et al. [144] grew aligned
MWCNTs by chemical vapor deposition and obtained an
exceptional axial thermal conductivity (4.87 W/m K) with
only 16 vol% MWCNT. Such woven composites manufac-
tured in certain advanced materials companies showed
huge improvements in thermal conductivity along the x
axis [145147]. Indeed, a high intrinsic thermal conduc-
tivity of the highly crystalline ber, a ber as long as the
Fig. 23. Schematic representation of an anisotropic composite (aligned
ber structure).
material itself with no ber/matrix interfaces along the x
axis, led to an exceptional thermal conductivity (with high
loading at 70% wt.)
Nevertheless, as shown in Fig. 18, the polymers chains
surround the carbon nanotube, leading to some phonon
scattering at the ller/matrix interface. The same phe-
nomenon will occur here at the ber/matrix interfaces,
where the probability is even higher considering that
macroscopic bers are much longer than microscopic
llers, such as nanotubes. Thus, continuous bers are a
good way to improve thermal conductivity along a single
direction, but the high intrinsic thermal conductivity of the
crystalline ller will inevitably be reduced just by incor-
porating them into a matrix. In contact with any particle,
the ber will scatter phonons at this interface, leading to
thermal resistance. On the other hand, along the y axis,
many ber/matrix interfaces can be observed, where the
bers thickness can be considered as a short ller itself
compared to its actual length. Consequently, thermal con-
ductivity along the y axis will not be enhanced as it is
along the x axis. Regarding the y axis, the morphological
structure can be compared to a dispersion of equivalent
microllers, traduced by a comparable thermal conductiv-
ity.
The resin can be compared to a river to cross. With-
out any ller, a man would have to swim to cross
it, requiring considerable time. If some llers are dis-
persed in the matrix, we can imagine that they are
similar to oating wooden boards, which the man can
use to cross the river faster by jumping from one board
to another, but it would still require a relatively long
time, as he would be slowed down by each jumping
step. It might naturally be thought that the best way to
cross a river is to build a bridge, e.g., a network sam-
ple [148]. However, a single bridge would generate a
trafc jam if many people want to cross the bridge; there-
fore, many bridges seem to be an ideal solution here,
e.g., a woven sample. Therefore, the thermal conduc-
tivity of a woven sample with highly conductive bers
shows a huge improvement. Kandare et al. [149] obtained
good synergistic interactions by incorporating 1 vol% of
graphene nano-platelets and silver-nanoinclusions into
CFRP composites. These llers increased the through-
thickness thermal conductivity of the CFRP composite by
40%.
Another work published by the authors [150] showed
the positive impact of structured or aligned llers in a
DGEBA/DDS resin, as represented by Fig. 24. Expanded
graphite llers were compressed at high temperature and
high pressure to obtain graphite foils. These foils were then
cut and structured into a 3D-network (Network sample) or
by aligning bers along the Z axis (Fibers sample). Ther-
mal conductivity was considerably increased in both cases,
respectively reaching 2.10 W m1 K1 and 1.84 W m1 K1 ,
whereas equivalent composites with equivalent expanded
graphite microparticles displayed thermal conductivities
less than half that value. In addition, the incorporation of
a DGEBA layer in the plane of the Fibers sample showed
a dramatic decrease in thermal conductivity, reaching
0.81 W m1 K1 with only two 4 m-thick DGEBA inter-
faces. This study showed the huge impact of structured and
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128 19

Fig. 24. Structured composites prepared with graphite foils. [150], Copyright 2015. Reproduced with permission from Elsevier Ltd.

aligned llers on thermal conductivity and the dramatic

effect of interfaces on heat transfer.
Alignment seems to be an interesting way to improve
thermal conductivity in a single direction. Beyond ller
alignment, the structural aspect and the morphology of
the nal composite will dene its thermal conductivity.
Most recent studies [151155] have demonstrated how the
geometrical aspect of the llers impacts their thermal con-
ductivity. The challenge is now to study the impact of the
geometrical aspect at a mesoscopic scale and determine
the best structuration(s) to reach higher thermal conduc-
tivities. Nevertheless, as seen in Section 3.6 and Fig. 18,
considerable phonon scattering occurs at the ller/matrix
interface, leading inevitably to thermal resistance; even
with aligned ller arrays, the thermal conductivity in the
llers direction is not the highest possible due to this
phonon scattering. Enhancing the compatibility between
the ller and the matrix at this interface will lead to bet-
ter heat transfer from one component to the other. Many
recent papers on thermal conductivity in nanocomposites
address ller functionalization to improve the compatibil-
ity with the matrix, as well as the thermal conductivity.
3.9. Filler functionalization
As repeated several times throughout this review,
ller/matrix interfaces seem to be among the major prob-
lems to solve to avoid thermal resistance. Over the past
decades, research has focused on ller functionalization
to obtain higher compatibility between the llers and
the matrix, consequently improving some of the compos-
ites properties, such as electrical properties [156159],
ame retardance [160], mechanical properties [161164],
or thermal properties [165]. However, alternative treat-
ments have also been considered, such as plasma treatment
[166,167], sol-gel methods [168170], coatings [171,172],
hybrid materials [173] or oxidizing methods [174]. Many
types of llers and functionalizations have been studied,
and studies on carbon nanotubes are a fairly good example
to illustrate our point. Some studies focused on improving
the dispersion state, alignment, size and aspect ratio of the
CNTs, but the results were always only slightly improved.
In addition, many types of functionalization were applied
to the carbon nanotubes [175] to study their impact on
thermal conductivity, as summarized by Sahoo et al. [176].
Fig. 25. Chemical representation of acid treatment on CNT.
3.9.1. Acid treatment
This type of functionalization on carbon nanotubes was
among the rst widely studied, as it is particularly effec-
tive. Its positive impact was demonstrated on mechanical
or electrical properties that are highly dependent on the
dispersion state [177]. Indeed, considering a nitric/sulfuric
acid mixture, the functionalization degree was shown to be
greatly increased by the grafting of some carboxylic and/or
hydroxyl functions onto the carbon nanotubes, as repre-
sented in Fig. 25. This result was conrmed in our study,
where carbon nanobers were functionalized by this acid
treatment. The functionalization degree was substantially
increased (by XPS, untreated CNF: 2% wt. oxygen; Acid
treated CNF: 14% wt. oxygen) and the dispersion state was
improved (SEM observations), thanks to better compatibil-
ity at the ller/matrix interface.
However, this type of functionalization is quite bru-
tal and aggressive toward the llers functionalized. As
shown in Fig. 25, acid treatment will lead to many grafted
functions, but at the expense of the carbon nanotubes
crystalline structure. Indeed, not only will functions be
grafted onto existing defects, but they also will continue
degrading the crystalline structure of the nanotube to graft
even more functions. Vesali Naseh [178] demonstrated the
destructive aspect of acid treatment (NH3 ) on carbon nano-
tubes. Talaei et al. [174] showed several other treatments,
such as KPS/KOH, leading to much better thermal con-
ductivity results than acid treatment. Hong et al. [179]
showed exceptional enhancement of thermal conductiv-
ity with CNTs in PMMA (3.44 W/m K with only 4% wt.), and
20 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 26. Experimental thermal conductivity of an epoxy-nanocomposite lled with acid treated CNF.
much lower results when treated with nitric/sulfuric acid.
This degradation is therefore detrimental to thermal con-
ductivity, as shown in Fig. 26. Instead of increasing with
acid treatment, thermal conductivity goes almost back to
the initial value of the reference epoxy resin. In conse-
quence, the acid treatment used here is absolutely not
helpful to increase thermal conductivity, although some
authors, such as Sun et al. [180] showed a 40% increase in
the thermal conductivity of graphite nanoplatelets in epoxy
matrix, treated with nitric acid.
3.9.2. Non-covalent functionalization
Non-covalent functionalization is an alternative way
to improve the dispersion state of carbon nanotubes in
a matrix [181183]. Fig. 27 shows how stacking
Fig. 27. Chemical representation of non-covalent functionalization on
CNT.
created by a non-covalent functionalization of pyrene-
based molecules on carbon nanotubes will help disperse
them into the matrix without any structural degrada-
tion. First, because no covalent bond will be generated
by this functionalization, no additional defect is produced.
The carbon nanotube keeps its initial crystalline structure.
However, stacking is a non-covalent interaction seen
in Fig. 27, if the radical R on the pyrene function is com-
patible with the matrix, then this type of functionalization
will help to disperse the llers into the matrix, as well as
to enhance the compatibility at the interface.
In this study, the illustrated non-covalent functionaliza-
tion based on pyrenecarboxylic acid (PCA) was performed
by the authors. The results showed some enhancement
in the dispersion state, while the thermal conductivity
decreased (Fig. 28), compared to untreated CNFs. The suc-
cess of the functionalization was conrmed by UVVisible
spectroscopy (PCA bands at 280 nm and 345 nm) and
MALDI-TOF (PCA function at m/z 245).
The pyrene function helps to compatibilize the ller and
the matrix, but this type of ller/matrix interface is not
really suitable for thermal conductivity. The non-covalent
functionalization will break CNF aggregates and maintain
their dispersion, leading to more CNF/matrix interfaces and
fewer CNF/CNF interfaces; hence a better dispersion. As
explained in Section 3.5, dispersion of this type of llers
does not involve higher thermal conductivity, but much
more CNF/matrix interfaces and thus much more phonon
scattering. Moreover, by looking closer to the ller/matrix
interface following the non-covalent functionalization, it
may be assumed that the stacking between CNF and
the pyrene function is not particularly helpful in transfer-
ring thermal vibrations from the ller to the matrix. Indeed,
molecules not directly bonded to llers will only be able to
transfer a small fraction of the thermal vibrations diffusing
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
Fig. 28. Experimental thermal conductivity of an epoxy-nanocomposite lled with non-covalently functionalized CNF.
through the ller. From the thermal conductivity perspec-
tive, this interface is quite comparable to Fig. 18. However,
the interesting aspect of non-covalent functionalization is
that the intrinsic thermal conductivity of the ller is not
as reduced, as for acid-treated CNFs. Lazarenko et al. [184]
discuss contact thermal resistance between the ller and
the polymer.
As a consequence, non-covalent functionalization is an
interesting way to enhance the dispersion of carbon nano-
tubes into a matrix or other properties such as electrical
conductivity [185], whereas thermal conductivity is not
improved but even decreased compared to untreated CNFs.
However, some authors showed improvement of the ther-
mal conductivity with non-covalent functionalization. Cao
et al. [186], for instance, performed a non-covalent func-
tionalization of graphene with gallic acid, considerably
increasing thermal conductivity, assuming the positive
impact of the dispersion state. In the same way, Yang
et al. [187] investigated the combination of MWCNTs and
graphene platelets in epoxy resin. A blend of 0.9% wt.
MGP and 0.1% wt. MWCNT showed the highest thermal
conductivity (46% enhancement). Teng et al. [188] also
showed the positive impact of non-covalent functionaliza-
tion of graphene by pyrene-PGMA on thermal conductivity,
assuming the importance of the dispersion parameter.
Nevertheless, non-covalent functionalization can be the-
oretically effective for some llers and not for others. For
example, it is assumed that non-covalent functionalization
is more efcient for longer llers with higher aspect ratios,
such as graphene sheets or other type of nanoplatelets
[189]. As we observed that size and aspect ratio are essen-
tial parameters for thermal conductivity, it follows that
each treatment will impact the thermal conductivity dif-
ferently, as a function of intrinsic parameters of the ller.
21
3.9.3. Covalent functionalization
Many chemical functional groups have been cova-
lently grafted onto carbon nanotubes in the literature,
including amines [190,191], silanes [175,192], and poly-
mers [193196], as summarized by Wu et al. [197] in
Fig. 29. Most of the covalent functionalizations showed
some enhancement of thermal conductivity, but the effect
was usually moderate. Ganguli et al. [198] improved exfo-
liated graphite/epoxy thermal conductivity (from 4.2 to
5.8 W/m K at 20% wt.) by a silane treatment of graphite
akes. In the same way, Yung et al. [199] functional-
ized BN particles with silane functional groups, showing
some enhancement in thermal conductivity (10% wt.
3 W/m K). However, Yuen et al. [200] and Hwang
et al. [191], respectively functionalizing CNT by silanes or
amines, did not show exceptional enhancement of thermal
conductivity (approximately 10%). While several processes
and synthesis were used to obtain functionalized carbon
nanotubes, the FriedelCrafts functionalization seems to
be one of the best ways to graft certain functions onto car-
bon nanotubes without degrading the crystalline structure
of the ller [201]. Indeed, contrary to the acid function-
alization, the FriedelCrafts synthesis will only occur on
existing defects of carbon nanotubes, while acid treatment
creates even more defects. In other words, through the
FriedelCrafts functionalization, grafted functional groups
will only slightly increase phonon scattering at existing
defects locations, while acid treatment increases phonon
scattering on the parts grafted but even more at the
defects created. Barikani et al. [202] obtained the best
enhancement of thermal conductivity with a FriedelCrafts
functionalization on MWCNT, compared to acid or amine
treatments. Therefore, each type of functionalization will
reduce intrinsic the thermal conductivity of the ller, but
22 N. Burger et al. / Progress in Polymer Science 61 (2016) 128

Fig. 29. Several covalent functionalizations on CNT studied in the literature. [197], Copyright 2010. Reproduced with permission from the Royal Society of
Chemistry.

not by the same proportion. The FriedelCrafts functional-

ization seems to be a good way to minimize this thermal
conductivity reduction compared to other types of func-
tionalization.
Furthermore, while the intrinsic thermal conductiv-
ity of the ller will decrease with functionalization, the
ller/matrix interface should logically be reduced and the
compatibility improved. This improvement would lead to
decreasing phonon scattering at this interface, i.e., reduc-
ing thermal resistance or increasing the heat transferred
from the ller to the matrix, and vice versa. Thermal con-
ductance is consequently increased at this interface, which
brings us again to some competition between these two
phenomena, i.e., thermal conductance improvement on the Fig. 30. Hypothetical representation of intrinsic thermal conductivity of
interface vs. intrinsic thermal conductivity reduction of the ller and thermal conductance at ller/matrix interface vs. function-
the ller. Gulotty et al. [203] proposed the same explana- alization rate.

tion by studying the impact of COOH functions on carbon

nanotubes. Liu et al. [204] observed the same effects of
chemical modications on carbon nanotube composites.
Fig. 30 helps to illustrate this situation.
As we try to demonstrate through Fig. 30, the intrin-
sic thermal conductivity of the ller will decrease with
functionalization due to heightened phonon scattering
at the grafted parts, whereas thermal conductance at
the ller/matrix interface increases due to a better
connection between the ller and the matrix. In this
theoretical representation, we can determine that at a
certain functionalization degree, the thermal conductivity
improvement at the interface is more important than the
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
intrinsic thermal conductivity reduction of the ller; hence,
the resulting composite should display a higher apparent
thermal conductivity. For acid functionalization, we can
imagine an opposite graph, where the red curve decreases
faster than the blue curve increases, i.e., the intrinsic ther-
mal conductivity of the ller decreases faster with the
functionalization rate than the thermal conductance at the
interface increases, leading to a decrease in the apparent
thermal conductivity of the composite considered. Never-
theless, some FriedelCraft functionalization, as presented
by Yang [205], can exert a positive impact on CNTs by
improving thermal conductivity from 0.13 to 0.96 W/m K
with 15 vol%.
The impact of ller functionalization on the thermal
conductivity of the bulk composite has not yet been well
dened. Some results show the positive impact of function-
alization by increasing the thermal conductivity of the nal
material, but the enhancement is generally quite small.
However, functionalization could help to improve some
other properties, such as mechanical or electrical prop-
erties. Dai et al. [206] presented some improvements in
thermal conductivity, as well as the electrical conductivity
of polyimide composites containing one-dimensional SiC
nanowires grown on graphene sheets. Some progress has
been made in using functionalization to enhance thermal
conductivity in composites, but the real question now is
to determine whether it is really the best way to achieve
higher values.
4. Summary and conclusion
This review addresses theoretical and fundamental
aspects of thermal conductivity in composite materials
and the progress made over the last decade. It has been
demonstrated that the thermal conductivity parameter
k is quite complex and difcult to model for composite
systems. Enhancing the thermal conductivity of polymer-
based composites by incorporating thermally conductive
llers requires a good understanding of the fundamental
mechanism and what parameters to focus on. Many ther-
mally conductive llers have been studied in the literature,
in efforts to improve the thermal conductivity of compos-
ites, but in most cases, to increase thermal conductivity by
10, a loading higher than 30% wt. is required. Currently, the
challenge is to further improve thermal conductivity with
much lower loadings.
In this review, several general parameters are revealed
to be essential and directly related to the thermal con-
ductivity parameter. First, crystallinity is one of the main
parameters to consider. Defects in the crystalline struc-
ture lead inevitably to phonon scattering, i.e., a decrease
in thermal conductivity. Actually, any change in the lin-
earity or regularity of the morphological aspect of the ller
will tend to decrease intrinsic thermal conductivity. Recent
studies at a quantum scale helped to understand the com-
plex mechanism of thermal conductivity a little better, but
even if some progress has been made in the comprehen-
sion of mechanism, it is now important to focus on how to
reduce the Kapitza resistance and fully benet from the
extraordinary intrinsic thermal conductivity of graphene
or CNTs.
23
Thermal conductivity is an anisotropic property; there-
fore, the aspect ratio, as well as the length, size, diameter,
and specic surface area of the ller, are particularly
important. From a general perspective, it seems that for
each of these parameters, the higher, the better. How-
ever, CNTs diameters remain a topic of discussion in the
literature regarding their impact on thermal conductiv-
ity. The dispersion state of nanoparticles does not appear
to be essential for thermal conductivity, as demonstrated
experimentally by the authors. The processing method is
also quite important, especially regarding the viscosity and
the resulting porosity of the sample. In the composite, the
alignment of anisotropic nanollers will directly impact
the thermal conductivity, increased in the ller direction.
Experimentally, this type of elaboration is relatively dif-
cult (magnetically or electrically), especially for industries.
Nevertheless, it will be crucial to focus in future work on the
structural and geometrical aspects of the materials, which
are essential parameters to increase thermal conductivity.
Finally, another way to improve the thermal conductiv-
ity of polymers is by reducing thermal resistance, mostly
due to the ller/matrix interfaces. Several types of function-
alization were performed over the years. Recent studies
demonstrated both the positive and the negative effects
of the functionalization of llers on thermal conductivity,
increasing thermal conductance at the ller/matrix inter-
face while decreasing the intrinsic thermal conductivity of
the ller. The challenge is now to determine how to benet
from functionalization at the interface without any loss in
intrinsic thermal conductivity of the ller.
To summarize, research on thermal conductivity
has reached an interesting point. Many intrinsic and
experimental parameters inuence the resulting ther-
mal conductivity of the material. In other words, it has
been illustrated through these examples how complex the
thermal conductivity parameter is and what compromise
should be made to improve this property in composites.
Enhancing the thermal conductivity of composites or poly-
mers requires naturally thermally conductive llers, but
to achieve much higher values of thermal conductivity,
it is crucial to focus on how to improve heat transfers
at the interfaces. Over the last decade, researchers were
able to reach thermal conductivities comparable to the val-
ues of metals, using organic-based composites. Enormous
progress has been made with carbon bers in composites,
but some encouraging results were also obtained using
micro- or nano-particles. Through this review, we present
the many important parameters for reaching higher ther-
mal conductivities. While some progress has been made for
composites, the challenges remain currently to determine
how to benet fully from the intrinsic thermal conductiv-
ity of highly conductive llers, such as graphene, CNTs, and
graphite and to achieve values as close as possible to the
theoretical ones.
Acknowledgements
The authors are grateful to Y. Jannot and A. Degiovanni
from the University of Lorraine for their characterizations
of the samples and their theoretical knowledge on ther-
mal conductivity characterization techniques. The authors
24 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
would also like to thank the FNR (Fonds National de la
Recherche Luxembourg) and Thales Alenia Space for nan-
cial support.
N. Burger is particularly grateful to the theoretical
physicist T. Schilling from University of Luxembourg for
the theoretical understanding of quantum physics.
References
[1] Han Z, Fina A. Thermal conductivity of carbon nanotubes
and their polymer nanocomposites: a review. Prog Polym Sci
2011;36:91444.
[2] Balandin AA. Thermal properties of graphene and nanostructured
carbon materials. Nat Mater 2011;10:56981.
[3] Sadasivuni KK, Ponnama D, Thomas S, Grohens Y. Evolution
from graphite to graphene elastomer composites. Prog Polym Sci
2014;39:74980.
[4] Anonymous. Livre Blanc La Recherche En Thermique Enjeux &
Perspectives En 2005. Paris: Socit Francaise de Thermique; 2005.
p. 146.
[5] Tong XC. Advanced materials for thermal management of electronic
packaging. New York: Springer; 2011. p. 607.
[6] White SR, Mather PT, Smith MJ. Characterization of the cure-state
of DGEBA-DDS epoxy using ultrasonic, dynamic mechanical, and
thermal probes. Prog Eng Sci 2002;42:5167.
[7] Marchio D, Reboux P. Introduction aux Transferts Thermiques.
Paris: Ecoles des Mines de Paris; 2008. p. 210.
[8] Lee JH, Lee SH, Choi CJ, Jang SP, Choi SUS. A review of thermal
conductivity data, mechanisms and models for nanouids. Int J
Micro-Nano Scale Transf 2010;1:269322.
[9] Monje IE, Louis E, Molina JM. Optimizing thermal conductivity in
gas-pressure inltrated aluminum/diamond composite by precise
processing control. Composites A 2013;48:914.
[10] Toberer ES, Baranowski LL, Dames C. Advances in thermal conduc-
tivity. Annu Rev Mater Res 2012;42:179209.
[11] Gerl M, Issi JP. Trait Des Matriaux 8 Physique des Matri-
aux. Lausanne: Presses Polytechniques et Universitaires Romandes;
1997. p. 704.
[12] Tritt TM. Thermal conductivity theory, properties & applications.
New York: Plenum Publishers; 2004. p. 285.
[13] Lienhard IV, Lienhard JHVJH. A heat transfer textbook. 4th ed.
Cambridge, MA: Phlogiston Press; 2012. p. 755.
[14] Callister WD. Materials science and engineering: an introduction.
7th edition New York: John Wiley & Sons Inc.; 2007. p. 832.
[15] Eliezer S, Ghatak A, Hora H, Teller E. Fundamentals of equations of
state. 1st ed. Singapore: World Scientic Pub Co Inc.; 2002. p. 394.
[16] Haggenmueller R, Guthy C, Lukes JR, Fische JE, Winey KI. Single
wall carbon nanotube/polyethylene nanocomposites: thermal and
electrical conductivity. Macromolecules 2007;40:241721.
[17] Sque SJ [PhD thesis] A First-Principles Study on Bulk And Transfer
Doping of Diamond. University of Exeter; 2005. p. 187.
[18] Ho CY, Powell RW, Liley PE. Thermal conductivity of the elements:
a comprehensive review. J Phys Chem Ref Data 1974;3(Suppl.
1):1810.
[19] Hay B, Davee G, Hameury J, Enouf O, Rongione L. Determination of
the thermal diffusivity and thermal conductivity of coatings at high
temperature by ash method. Rev Fr de Mtrol 2010;24:2333.
[20] Goodson KE, Flik MI. Solid layer thermal conductivity measurement
techniques. Appl Mech Rev 1994;47:10112.
[21] Paul G, Chopkar M, Manna I, Das PK. Techniques for measuring the
thermal conductivity of nanouids: a review. Renew Sustain Energy
Rev 2010;14:191324.
[22] Villire M, Lecointe D, Sobotka V, Boyard N, Delaunay D. Experimen-
tal determination and modeling of thermal conductivity tensor of
carbon/epoxy composites. Composites A 2013;46:608.
[23] Im H, Hwang Y, Moon JH, Lee SH, Kim J. The thermal conductivity of
Al(OH)3 covered MWCNT/epoxy terminated dimethyl polysiloxane
composite based on analytical Al(OH)3 covered MWCNT. Compos-
ites A 2013;54:15965.
[24] Agrawal A, Satapathy A. Mathematical model for evaluating effec-
tive thermal conductivity of polymer composites with hybrid
llers. Int J Therm Sci 2015;89:2039.
[25] Suplicz A, Hargita H, Kovacs JG. Methodology development for
through-plane thermal conductivity prediction of composites. Int J
Therm Sci 2016;100:549.
[26] Iwamoto N, Yuen MMF, Fan H. Molecular modeling and multiscal-
ing issues for electronic material applications. New York: Springer;
2012. p. 260.
[27] Mortazavi B, Benzerara O, Meyer H, Bardon J, Ahzi S. Com-
bined molecular dynamics-nite element multiscale modeling of
thermal conduction in graphene epoxy nanocomposites. Carbon
2013;60:35665.
[28] Giri A, Hopkins PE, Wessel JG, Duda JC. Kapitza resistance and
the thermal conductivity of amorphous superlattices. J Appl Phys
2015;118:165303, 16.
[29] McNamara AJ, Joshi Y, Zhang ZM. Thermal resistance of thermal
conductive adhesive anchored carbon nanotubes interface mate-
rial. Int J Therm Sci 2015;96:2216.
[30] Lu S, McGaughey AJH. Thermal conductance of superlattice junc-
tions. AIP Adv 2015;5:053205, 112.
[31] Anonymous. Kintech Lab. Calculation of thermal conductivity of
ideal and defective CNT; 2015, 1 pp. http://www.kintechlab.com/
solutions/nanotechnology/cnt-thermal-conductivity-calculation/.
[32] Kim YA, Kamio S, Tajiri T, Hayashi T, Song SM, Endo M, Ter-
rones M, Dresselhaus MS. Enhanced thermal conductivity of carbon
ber/phenolic resin composites by the introduction of carbon
nanotubes. Appl Phys Lett 2007;90:093125, 13.
[33] Song Z, Hou X, Zhang L, Wu S. Enhancing crystallinity and orien-
tation by hot-stretching to improve the mechanical properties of
electrospun partially aligned polyacrylonitrile (PAN) nanocompos-
ites. Materials 2011;4:62132.
[34] Weisenberger M, Martin-Gullon I, Vera-Agullo J, Varela-Rizo H,
Merino C, Andrews R, Qian D, Rantell T. The effect of graphitization
temperature on the structure of helical-ribbon carbon nanobers.
Carbon 2009;47:22118.
[35] Lim S, Yoon SH, Mochida I, Chi JH. Surface modication of car-
bon nanober with high degree of graphitization. J Phys Chem B
2004;108:15336.
[36] Uher C. Thermal conductivity of high-Tc superconductors. J Super-
cond 1990;3:33789.
[37] Shind SL, Goela JS. High thermal conductivity materials. New York:
Springer; 2006. p. 271.
[38] Dugdale JS, MacDonald DKC. Lattice thermal conductivity. Phys Rev
B 1955;98:17512.
[39] Chen S, Wu Q, Mishra C, Kang J, Zhang H, Cho K, Cai W, Baladin AA,
Ruoff RS. Thermal conductivity of isotopically modied graphene.
Nat Mater 2012;11:2037.
[40] Singh D, Murthy JY, Fisher TS. Mechanism of thermal conductiv-
ity reduction in few-layer graphene. J Appl Phys 2011;110:044317,
18.
[41] Dinwiddie RB, White MA, McElroy DL. Thermal conductivity 28
thermal expansion 16. In: 28th Int Therm Cond Conf. DEStech
Publications Inc.; 2005, 543 pp.
[42] Berryman JG. Thermal conductivity of porous media. Appl Phys Lett
2005;86:032905, 13.
[43] Randrianalisoa JH [PhD thesis] Transfert Thermique Par Ray-
onnement Et Conduction Dans Les Matriaux Poreux Micro Et
Nanostructurs. In: Analogie Transfert De Phonons Et De Photons.
Centre de Thermique de Lyon (CETHIL, INSA de Lyon); 2006. p. 315
pp.
[44] Yang L, Chen J, Yang N, Li B. Signicant reduction of graphene ther-
mal conductivity by phononic crystal structure. Int J Micro-Nano
Scale Transf 2015;91:42832.
[45] Wingert MC, Kwon S, Hu M, Poulikakos D, Xiang J. Sub-amorphous
thermal conductivity in ultrathin crystalline silicon nanotubes.
Nano Lett 2015;15:260511.
[46] Bouyrie Y, Candol C, Pailhe S, Koza MM, Malaman B, Dauscher
A, Tobola J, Boisron O, Saviot L, Lenoir B. From crystal to glass-like
thermal conductivity in crystalline minerals. Phys Chem Chem Phys
2015;17:197518.
[47] Kim GH, Lee D, Shanker A, Shao L, Kwon MS, Gidley D, Kim J, Pipe
KP. High thermal conductivity in amorphous polymer blends by
engineered interchain interactions. Nat Mater 2015;14:295300.
[48] Supov M, Martynkov GS, Barabaszov K. Effect of nanollers dis-
persion in polymer matrices: a review. Sci Adv Mater 2011;3:125.
[49] Tsekmes IA, Kochetov R, Morshuis PHF, Smit JJ. Thermal conduc-
tivity of polymeric composites: a review. IEEE Conf Solid Dielectr
2013:67881.
[50] Gaxiola DL, Keith JM, King JA, Johnson BA. Nielsen thermal con-
ductivity model for single ller carbon/polypropylene composites.
J Appl Polym Sci 2009;114:32617.
[51] Afanasov IM, Savchenko DV, Ionov SG, Rusakov DA, Seleznev AN,
Avdeev VV. Thermal conductivity and mechanical properties of
expanded graphite. Inorg Mater 2009;45:48690.
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
[52] Gu J, Yang X, Lv Z, Li N, Liang C, Zhang Q. Functionalized graphite
nanoplatelets/epoxy resin nanocomposites with high thermal con-
ductivity. Int J Micro-Nano Scale Transf 2016;92:1522.
[53] Yu A, Ramesh P, Sun X, Bekyarova E, Itkis ME, Haddon RC. Enhanced
thermal conductivity in a hybrid graphite nanoplatelet carbon
nanotube ller for epoxy composites. Adv Mater 2008;20:47404.
[54] Nanda J, Maranville C, Bollin SC, Sawall D, Ohtani H, Remil-
lard JT, Ginder JM. Thermal conductivity of single-wall carbon
nanotube dispersions: role of interfacial effects. J Phys Chem C
2008;112:6548.
[55] Hauser RA, King JA, Pagel RM, Keith JM. Effects of carbon llers
on the thermal conductivity of highly lled liquid-crystal polymer
based resins. J Appl Polym Sci 2008;109:214555.
[56] Balandin AA, Ghosh S, Bao W, Calizo I, Teweldebrhan D, Miao F, Lau
CN. Superior thermal conductivity of single-layer graphene. Nano
Lett 2008;8:9027.
[57] Martin-Gallego M, Verdejo R, Khayet M, Zarate J.M.O.d., Essalhi M,
Lopez-Manchado MA. Thermal conductivity of carbon nanotubes
and graphene in epoxy nanouids and nanocomposites. Nanoscale
Res Lett 2011;6:610, 17.
[58] Sun X, Sun H, Li H, Peng H. Developing polymer composite materi-
als: carbon nanotubes or graphene? Adv Mater 2013;25:515376.
[59] Chung DDL. Materials for thermal conduction. Appl Therm Eng
2001;21:1593605.
[60] Yao Y, Zeng X, Guo K, Sun R, Xu JB. The effect of interfacial state on
the thermal conductivity of functionalized Al2 O3 lled glass bers
reinforced polymer composites. Composites A 2015;69:4955.
[61] Boudenne A [PhD thesis] Etude Exprimentale Et Thorique Des
Proprits Thermophysiques de Matriaux Composites Matrice
Polymre. Universit Paris XII Val de Marne; 2003. p. 193.
[62] Hammerstroem DW, Burgers MA, Chung SW, Guliants EA, Bunker
CE, Wentz KM, Hayes SE, Buckner SW, Jelliss PA. Aluminum
nanoparticles capped by polymerization of alkyl-substituted epox-
ides: ratio-dependent stability and particle size. Inorg Chem
2011;50:50549.
[63] Ma A, Chen W, Hou Y, Zhang G. The preparation and cure kinetics
researches of thermal conductivity epoxy/AlN composites. Polym
Plast Technol 2010;49:3548.
[64] Lee GW, Park M, Kim J, Lee JI, Yoon HG. Enhanced thermal conduc-
tivity of polymer composites lled with hybrid ller. Composites A
2006;37:72734.
[65] Li TL, Hsu SLC. Enhanced thermal conductivity of polyimide lms
via a hybrid of micro- and nano-sized boron nitride. J Phys Chem B
2010;114:68259.
[66] Yuan FY, Zhang HB, Li X, Li XZ, Yu ZZ. Synergetic effect of boron
nitride akes and tetrapod-shaped ZnO whiskers on the thermal
conductivity of electrically insulating phenol formaldehyde com-
posites. Composites A 2013;53:13744.
[67] Firdaus SM, Mariatti M. Fabrication and characterization of nano
ller-lled epoxy composites for underll application. J Mater Sci
Mater Electron 2012;23:12939.
[68] Wu Y, Yu Z. Thermal conductivity of in situ epoxy composites lled
with ZrB2 particles. Compos Sci Technol 2015;107:616.
[69] Liu K, Takagi H, Osugi R, Yang Z. Effect of physiochemical structure
of natural ber on transverse thermal conductivity of unidirectional
abaca/bamboo ber composites. Composites A 2012;43:123441.
[70] Xu Y, Chung DDL. Increasing the thermal conductivity of boron
nitride and aluminum nitride particle epoxy-matrix composites by
particle surface treatments. Compos Interfaces 2000;7:24356.
[71] Ma AJ, Chen W, Hou Y. Enhanced thermal conductivity of epoxy
composites with MWCNTs/AlN hybrid ller. Polym Plast Technol
2012;51:157882.
[72] Hong JP, Yoon SW, Hwang T, Oh JS, Hong SC, Lee Y, Nam JD. High
thermal conductivity epoxy composites with bimodal distribution
of aluminum nitride and boron nitride llers. Thermochim Acta
2012;537:705.
[73] Im H, Kim J. Thermal conductivity of a graphene oxidecarbon
nanotube hybrid/epoxy composite. Carbon 2012;50:542940.
[74] Sanada K, Tada I, Shindo Y. Thermal conductivity of polymeric
composites with close-packed structure of nano and micro llers.
Composites A 2009;40:72430.
[75] Hwang Y, Kim M, Kim J. Enhancement of thermal and mechanical
properties of exible graphene oxide/carbon nanotube hybrid lms
though direct covalent bonding. J Mater Sci 2013;48:701121.
[76] Safdari M, Al-Haik MS. Synergistic electrical and thermal transport
properties of hybrid polymeric nanocomposites based on carbon
nanotubes and graphite nanoplatelets. Carbon 2013;64:11121.
[77] Song WL, Wang P, Cao L, Anderson A, Meziani MJ, Farr AJ, Sun
YP. Polymer/boron nitride nanocomposite materials for superior
25
thermal transport performance. Angew Chem Int Ed 2012;51:
6498501.
[78] Donnay M, Tzavalas S, Logakis E. Boron nitride lled epoxy with
improved thermal conductivity and dielectric breakdown strength.
Compos Sci Technol 2015;110:1528.
[79] Diez-Pascual AM, Guan J, Simard B, Gomez-Fatou MA.
Poly(phenylene sulde) and poly(ether ether ketone) composites
reinforced with single-walled carbon nanotube buckypaper: II
Mechanical properties, electrical and thermal conductivity.
Composites A 2012;43:100715.
[80] Aliev AE, Lima M, Silverman EM, Baughman RH. Thermal conduc-
tivity of multi-walled carbon nanotube sheets: radiation losses and
quenching of phonon modes. Nanotechnology 2010;21:035709,
111.
[81] Li Q, Duchemin I, Xiong S, Solomon GC, Donadio D. Mechanical
tuning of thermal transport in a molecular junction. J Phys Chem C
2015;119:2463642.
[82] Donadio D, Galli G. Thermal conductivity of isolated and inter-
acting carbon nanotubes: comparing results from molecular
dynamics and the Boltzmann transport equation. Phys Rev Lett
2007;99:255502, 14.
[83] Gaska K, Rybak A, Kapusta C, Sekula R, Siwek R. Enhanced thermal
conductivity of epoxymatrix composites with hybrid llers. Polym
Adv Technol 2015;26:2631.
[84] Choi SW, Yoon KH, Jeong SS. Morphology and thermal conductiv-
ity of polyacrylate composites containing aluminum/multi-walled
carbon nanotubes. Composites A 2013;45:15.
[85] Wattanakul K, Manuspiya H, Yanumet N. Thermal conductivity and
mechanical properties of BN-lled epoxy composite: effects of ller
content, mixing conditions, and BN agglomerate size. J Compos
Mater 2011;45:196780.
[86] Lee ES, Lee SM, Shaneeldz DJ, Cannon WR. Enhanced ther-
mal conductivity of polymer matrix composite via high solids
loading of aluminum nitride in epoxy resin. J Am Ceram Soc
2008;91:116974.
[87] Weidenfeller B, Hfer M, Schilling FR. Thermal conductivity,
thermal diffusivity, and specic heat capacity of particle lled
polypropylene. Composites A 2004;35:4239.
[88] Zhou T, Wang X, Liu X, Xiong D. Improved thermal conduc-
tivity of epoxy composites using a hybrid multi-walled carbon
nanotube/micro-SiC ller. Carbon 2010;48:11716.
[89] Evseeva LE, Tanaeva SA. Thermal conductivity of micro- and nanos-
tructural epoxy composites at low temperatures. Mech Compos
Mater 2008;44:11726.
[90] Shahil KMF, Balandin AA. Graphene multilayer graphene
nanocomposites as highly efcient thermal interface materials.
Nano Lett 2012;12:8617.
[91] Wang F, Drzal LT, Qin Y, Huang Z. Mechanical properties and ther-
mal conductivity of graphene nanoplatelet/epoxy composites. J
Mater Sci 2015;50:108293.
[92] Kim SY, Noh YJ, Yu J. Thermal conductivity of graphene
nanoplatelets lled composites fabricated by solvent-free
processing for the excellent ller dispersion and a theoretical
approach for the composites containing the geometrized llers.
Composites A 2015;69:21925.
[93] Zhao YH, Wu ZK, Bai SL. Study on thermal properties of graphene
foam/graphene sheets lled polymer composites. Composites A
2015;72:2006.
[94] Tang B, Hu G, Gao H, Hai L. Application of graphene as ller to
improve thermal transport property of epoxy resin for thermal
interface materials. Int J Micro-Nano Scale Transf 2015;85:4209.
[95] Filali M [PhD thesis] Conductivit Thermique Apparente des
Milieux Granulaires Soumis Des Contraintes Mcaniques: Mod-
lisation Et Mesures. Institut National Polytechnique de Toulouse;
2006. p. 211.
[96] Chandrasekaran S, Seidel C, Schulte K. Preparation and charac-
terization of graphite nano-platelet (GNP)/epoxy nanocomposites:
mechanical, electrical and thermal properties. Eur Polym J
2013;49:387888.
[97] Choi JR, Rhee KY, Park SJ. Inuence of electrolessly silver-plated
multi-walled carbon nanotubes on thermal conductivity of epoxy
matrix nanocomposites. Composites B 2015;80:37984.
[98] Gu J, Lv Z, Wub Y, Zhao R, Tian L, Zhang Q. Enhanced thermal
conductivity of SiCp/PS composites by electrospinninghot press
technique. Composites A 2015;79:813.
[99] Raza MA, Westwood AVK, Stirling C. Effect of processing tech-
nique on the transport and mechanical properties of graphite
nanoplatelet/rubbery epoxy composites for thermal interface
applications. Mater Chem Phys 2012;132:6373.
26 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
[100] Chakraborty AK, Plyhm T, Barbezat M, Necola A, Terrasi GP. Carbon
nanotube (CNT)-epoxy nanocomposites: a systematic investigation
of CNT dispersion. J Nanopart Res 2011;13:6493506.
[101] Ming Y, Chi H, Blaser R, Xu C, Yang J, Veenstra M, Gaab M, Mller
U, Uher C, Siegel DJ. Anisotropic thermal transport in MOF-5 com-
posites. Int J Micro-Nano Scale Transf 2015;82:2508.
[102] Chen L, Xu HF, Sun YY, Du XZ, He SJ, Lin J, Nazarenko S. Modeling of
anisotropic thermal conductivity of polymer composites contain-
ing aligned boron nitride platelets: effect of processing methods.
Polym Compos 2015, http://dx.doi.org/10.1002/pc.23863.
[103] Nguyen TMH [PhD thesis] Systmes Epoxy-Amine Incluant Un
Catalyseur Externe Phnolique: Cintique de Rticulation Vieil-
lissement Hydrolytique. Universit du Sud Toulon-Var; 2007. p. 103
https://tel.archives-ouvertes.fr/tel-00136381.
[104] Song YS, Youn JR. Inuence of dispersion states of carbon nano-
tubes on physical properties of epoxy nanocomposites. Carbon
2005;43:137885.
[105] Wernik JM, Meguid SA. Recent developments in multifunc-
tional nanocomposites using carbon nanotubes. Appl Mech Rev
2010;63:140.
[106] Karippal JJ, Murthy HNN, Rai KS, Krishna M, Sreejith M. Electrical
and thermal properties of twin-screw extruded multiwalled carbon
nanotube/epoxy composites. J Mater Eng Perform 2010;19:11439.
[107] Santos ASd, Leite TdON, Furtado CA, Welter C, Pardini LC, Silva
GG. Morphology, thermal expansion and electrical conductivity of
multiwalled carbon nanotube/epoxy composites. J Appl Polym Sci
2007;108:97984.
[108] Du Y, Shen SZ, Cai K, Casey PS. Research progress on polymer-
inorganic thermoelectric nanocomposites materials. Prog Polym
Sci 2012;37:82041.
[109] Baruch AE, Bielenki L, Regev O. Thermal conductivity improvement
of electrically nonconducting composite materials. Rev Chem Eng
2012;28:6171.
[110] Moyer CA. Effective conductivity for interface scattering in metal-
matrix composites. Phys Rev B 1993;47:1007982.
[111] Wang ZL, Mu HT, Liang JG, Tang DW. Thermal boundary resistance
and temperature dependent phonon conduction in CNT array mul-
tilayer structure. Int J Therm Sci 2013;74:5362.
[112] Nan CW, Liu G, Lin Y, Li M. Interface effect on thermal conductivity
of carbon nanotube composites. Appl Phys Lett 2004;85:354951.
[113] McNamara AJ, Joshi Y, Zhang ZM. Characterization of nanostruc-
tured thermal interface materials a review. Int J Therm Sci
2012;62:211.
[114] Gojny FH, Wichmann MHG, Fiedler B, Kinloch IA, Bauhofer W, Win-
dle AH, Schulte K. Evaluation and identication of electrical and
thermal conduction mechanisms in carbon nanotube/epoxy com-
posites. Polymer 2006;47:203645.
[115] Huang X, Iizuka T, Jiang P, Ohki Y, Tanaka T. Role of interface on the
thermal conductivity of highly lled dielectric epoxy/AlN compos-
ites. J Phys Chem C 2012;116:1362939.
[116] Goyal V, Balandin AA. Thermal properties of the hybrid graphene-
metal nano-micro-composites: applications in thermal interface
materials. Appl Phys Lett 2012;100:073113, 14.
[117] Chirtoc M, Horny N, Henry JF, Turgut A, Kkey I, Tavman I, Omas-
tov M. Photothermal characterization of nanocomposites based on
high density polyethylene (HDPE) lled with expanded graphite.
Int J Thermophys 2012;33:21107.
[118] Zhou W, Yu D, An Q. A novel polymeric coating with high thermal
conductivity. Polym Plast Technol 2009;48:12308.
[119] Kim HS, Jang JU, Yu J, Kim SY. Thermal conductivity of polymer
composites based on the length of multi-walled carbon nanotubes.
Composites B 2015;79:50512.
[120] Zhang P, Li Q, Xuan Y. Thermal contact resistance of epoxy compos-
ites incorporated with nano-copper particles and the multi-walled
carbon nanotubes. Composites A 2014;57:17.
[121] Park JG, Cheng Q, Lu J, Bao J, Li S, Tian Y, Liang Z, Zhang C, Wang B.
Thermal conductivity of MWCNT/epoxy composites: the effects of
length, alignment and functionalization. Carbon 2012;50:208390.
[122] Li J, Ma PC, Chow WS, To CK, Tang BZ, Kim JK. Correlations between
percolation threshold, dispersion state, and aspect ratio of carbon
nanotubes. Adv Funct Mater 2007;17:320715.
[123] Li B, Zhong WH. Review on polymer/graphite nanoplatelet
nanocomposites. J Mater Sci 2011;46:5595614.
[124] Deng F, Zheng QS, Wang LF, Nan CW. Effects of anisotropy,
aspect ratio, and nonstraightness of carbon nanotubes on ther-
mal conductivity of carbon nanotube composites. Appl Phys Lett
2007;90:021914, 14.
[125] Kapadia RS, Louie BM, Bandaru PR. The inuence of carbon
nanotube aspect ratio on thermal conductivity enhancement in
nanotubepolymer composites. J Heat Transf 2013;136:011303,
16.
[126] Su JF, Zhao YH, Wang XY, Dong H, Wang SB. Effect of interface
debonding on the thermal conductivity of microencapsulated-
parafn lled epoxy matrix composites. Composites A
2012;43:32532.
[127] Zhu BL, Wang J, Zheng H, Ma J, Wu J, Wu R. Investigation of thermal
conductivity and dielectric properties of LDPE-matrix composites
lled with hybrid ller of hollow glass microspheres and nitride
particles. Composites B 2015;69:496506.
[128] Mortazavi B, Baniassadi M, Bardon J, Ahzi S. Modeling of two-phase
random composite materials by nite element, MoriTanaka and
strong contrast methods. Composites B 2013;45:111725.
[129] Chirtoc M, Horny N, Tavman I, Turgut A, Kkey I, Omastov M.
Preparation and photothermal characterization of nanocomposites
based on high density polyethylene lled with expanded and unex-
panded graphite: particle size and shape effects. Int J Therm Sci
2012;62:505.
[130] Veca LM, Meziani MJ, Wang W, Wang X, Lu F, Zhang P, Lin Y, Fee R,
Connell JW, Sun YP. Carbon nanosheets for polymeric nanocompos-
ites with high thermal conductivity. Adv Mater 2009;21:208892.
[131] Smith DK, Pantoya ML. Effect of nanoller shape on effective ther-
mal conductivity of uoropolymer composites. Compos Sci Technol
2015;118:2516.
[132] Chu K, Li WS, Dong H. Role of graphene waviness on the
thermal conductivity of graphene composites. Appl Phys A
2013;111:2215.
[133] Xu X, Pereira LFC, Wang Y, Wu J, Zhang K, Zhao X, Bae S, Bui CT,
Xie R, Thong JTL, Hong BH, Loh KP, Donadio D, Li B, zyilmaz B.
Length-dependent thermal conductivity in suspended single-layer
graphene. Nat Commun 2014;5:3689, 16.
[134] Yu H, Heider D, Advani S. Role of in-plane stacking sequence on
transverse effective thermal conductivity of unidirectional com-
posite laminates. Int J Micro-Nano Scale Transf 2015;85:897903.
[135] Chen YC, Lee SC, Liu TH, Chang CC. Thermal conductivity of boron
nitride nanoribbons: anisotropic effects and boundary scattering.
Int J Therm Sci 2015;94:728.
[136] Ci L, Suhr J, Pushparaj V, Zhang X, Ajayan PM. Continuous carbon
nanotube reinforced composites. Nano Lett 2008;8:27626.
[137] Agarwal S, Khan MMK, Gupta RK. Thermal conductivity of poly-
mer nanocomposites made with carbon nanobers. Polym Eng Sci
2008;48:247481.
[138] Tavernier J. Sur quelques problmes gnraux de la conduction
thermique de rseau dans les solides. J Phys 1967;C1:4052.
[139] Yang K, Gu M. Enhanced thermal conductivity of epoxy nanocom-
posites lled with hybrid llers system of triethylenetetramine-
functionalized multi-walled carbon nanotube/silane-modied
nano-sized silicon carbide. Composites A 2010;41:21521.
[140] Shaikh S, Li L, Lafdi K, Huie J. Thermal conductivity of an aligned
carbon nanotube array. Carbon 2007;45:260813.
[141] Kim K, Kim J. Vertical ller alignment of boron nitride/epoxy com-
posite for thermal conductivity enhancement via external magnetic
eld. Int J Therm Sci 2016;100:2936.
[142] Gaska K, Kmita G, Rybak A, Sekula R, Goc K, Kapusta C. Magnetic-
aligned, magnetite-lled epoxy composites with enhanced thermal
conductivity. J Mater Sci 2015;50:25106.
[143] Choi ES, Brooks JS, Eaton DL, Al-Haik MS, Hussaini MY, Garmestani
H, Li D, Dahmen K. Enhancement of thermal and electrical prop-
erties of carbon nanotube polymer composites by magnetic eld
processing. J Appl Phys 2003;94:60349.
[144] Marconnet AM, Yamamoto N, Panzer MA, LWardle B, Good-
son KE. Thermal conduction in aligned carbon nanotube/polymer
nanocomposites with high packing density. ACS Nano 2011;5:
481825.
[145] Keith JM, King JA, Miller MG, Tomson AM. Thermal conductivity of
carbon ber/liquid crystal polymer composites. J Appl Polym Sci
2006;102:545662.
[146] Abot JL, Bardin G, Spriegel C, Song Y, Raghavan V, Govindaraju N.
Thermal conductivity of carbon nanotube array laminated compos-
ite materials. J Compos Mater 2011;43:32140.
[147] Ning QG, Chou TW. A general analytical model for predicting the
transverse effective thermal conductivities of woven fabric com-
posites. Composites A 1998;29:31522.
[148] Zeng J, Fu R, Shen Y, He H, Song X. High thermal conductive
epoxy molding compound with thermal conductive pathway. J Appl
Polym Sci 2009;113:211725.
[149] Kandare E, Khatibi AA, Yoo S, Wang R, Ma J, Olivier P. Improv-
ing the through-thickness thermal and electrical conductivity
of carbon bre/epoxy laminates by exploiting synergy between
 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
graphene and silver nano-inclusions. Composites A 2015;69:
7282.
[150] Burger N, Laachachi A, Ferriol M, Lutz M, Toniazzo V, Ruch D.
Alignments and network of graphite llers to improve thermal
conductivity of epoxy-based composites. Int J Heat Mass Transf
2015;89:50513.
[151] Chen L, Sun YY, Xu HF, He SJ, Wei GS. Analytic modeling for
the anisotropic thermal conductivity of polymer composites con-
taining aligned hexagonal boron nitride. Compos Sci Technol
2016;122:429.
[152] Mosanenzadeh SG, Naguib HE. Effect of ller arrangement and
networking of hexagonal boron nitride on the conductivity of
new thermal management polymeric composites. Composites B
2016;85:2430.
[153] Gori F, Corasaniti S. Effective thermal conductivity of composites.
Int J Micro-Nano Scale Transf 2014;77:65361.
[154] Ngo IL, Byon C. A generalized correlation for predicting the thermal
conductivity of composites with heterogeneous nanollers. Int J
Micro-Nano Scale Transf 2015;90:8949.
[155] Gharagozloo-Hubmann K, Boden A, Czempiel GKF, Firkowska I,
Reich S. Filler geometry and interface resistance of carbon nano-
bres: key parameters in thermally conductive polymer composites.
Appl Phys Lett 2013;102:213103, 14.
[156] Sandler JKW, Kirk JE, Kinloch IA, Shaffer MSP, Windle AH. Ultra-low
electrical percolation threshold in carbon-nanotube-epoxy com-
posites. Polymer 2003;44:58939.
[157] Arlen MJ, Wang D, Jacobs JD, Justice R, Trion A, Hsu JWP, Schaffer D,
Tan LS, Vaia RA. Thermal-electrical character of in situ synthesized
polyimide-grafted carbon nanober composites. Macromolecules
2008;41:805362.
[158] Ramasubramaniama R, Chenb J, Liu H. Homogeneous carbon nano-
tube/polymer composites for electrical applications. Appl Phys Lett
2003;83:292830.
[159] Morales G, Barrena MI, Salazar JMGd, Merino C. Conductive CNF-
doped laminates processing and characterization. J Compos Mater
2011;45:21138.
[160] Wang Z, Tang XZ, Yu ZZ, Guo P, Song HH, Du XS. Dispersion of
graphene oxide and its ame retardancy effect on epoxy nanocom-
posites. Chin J Polym Sci 2011;29:36876.
[161] Prolongo SG, Gude MR, Sanchez J, Urena A. Nanoreinforced epoxy
adhesives for aerospace industry. J Adhes 2009;85:18099.
[162] Palmeri MJ, Putz KW, Brinson LC. Sacricial bonds in stacked-cup
carbon nanobers: biomimetic toughening mechanisms for com-
posite systems. ACS Nano 2010;4:425664.
[163] Bortz DR, Merino C, Martin-Gullon I. Mechanical characterization
of hierarchical carbon ber/nanober composite laminates. Com-
posites A 2011;42:158491.
[164] Wang DH, Sihn S, Roy AK, Baek JB, Tan LS. Nanocomposites based on
vapor-grown carbon nanobers and an epoxy: functionalization,
preparation and characterization. Eur Polym J 2010;46:140416.
[165] Du F, Fischer JE, Winey KI. A coagulation method to prepare single-
walled carbon nanotube/PMMA composites and their modulus,
electrical conductivity, and thermal stability. J Polym Sci B Polym
Phys 2003;41:33338.
[166] Huang R, Chen Z, Chu X, Wu Z, Li L. Preparation and thermal prop-
erties of epoxy composites lled with negative thermal expansion
nanoparticles modied by a plasma treatment. J Compos Mater
2010;45:167582.
[167] Dongil AB, Bachiller-Baeza B, Guerrero-Ruiz A, Rodrguez-Ramos
I, Martnez-Alonso A, Tascn JMD. Surface chemical modica-
tions induced on high surface area graphite and carbon nanobers
using different oxidation and functionalization treatments. J Colloid
Interface Sci 2011;355:17989.
[168] Kim J, Im H, Kim JM, Kim J. Thermal and electrical conductivity
of Al(OH)3 covered graphene oxide nanosheet/epoxy composites. J
Mater Sci 2012;47:141826.
[169] Kim K, Kim M, Hwang Y, Kim J. Chemically modied boron nitride-
epoxy terminated dimethylsiloxane composite for improving the
thermal conductivity. Ceram Int B 2014;40:204756.
[170] Zhao J, Du F, Cui W, Zhu P, Zhou X, Xie X. Effect of silica coat-
ing thickness on the thermal conductivity of polyurethane/SiO2
coated multiwalled carbon nanotube composites. Composites A
2014;58:16.
[171] Guo J, Saha P, Liang J, Saha M, Grady BP. Multi-walled carbon
nanotubes coated by multi-layer silica for improving thermal
conductivity of polymer composites. J Therm Anal Calorim
2013;113:46774.
[172] Kim KT, Dao TD, Jeong HM, Anjanapura RV. Graphene coated with
alumina and its utilization as a thermal conductivity enhancer
27
for alumina sphere/thermoplastic polyurethane composite. Mater
Chem Phys 2015;153:291300.
[173] Bozlar M [PhD thesis] Modication de Surface des Nanotubes de
Carbone Par un Polymre Conducteur Electrognr Pour la Ralis-
ation de Nanocomposites Multifonctionnels. Ecole Centrale Paris;
2010, 208 pp.
[174] Talaei Z, Mahjoub AR, Rashidi AM, Amrollahi A, Meibodi ME. The
effect of functionalized group concentration on the stability and
thermal conductivity of carbon nanotube uid as heat transfer
media. Int Commun Heat Mass 2011;38:5137.
[175] Roy N, Sengupta R, Bhowmick AK. Modications of carbon
for polymer composites and nanocomposites. Prog Polym Sci
2012;37:781819.
[176] Sahoo NG, Rana S, Cho JW, Li L, Chan SH. Polymer nanocompos-
ites based on functionalized carbon nanotubes. Prog Polym Sci
2010;35:83767.
[177] Hatui G, Bhattacharya P, Sahoo S, Dhibar S, Das CK. Combined
effect of expanded graphite and multiwall carbon nanotubes on the
thermo mechanical, morphological as well as electrical conductiv-
ity of in situ bulk polymerized polystyrene composites. Composites
A 2014;56:18191.
[178] Naseh MV, Khodadadi AA, Mortazavi Y, Sahraei OA, Pourfayaz
F, Sedghi SM. Functionalization of carbon nanotubes using nitric
acid oxidation and DBD plasma. World Acad Sci Eng Technol
2009;49:1779.
[179] Hong WT, Tai NH. Investigations on the thermal conductivity of
composites reinforced with carbon nanotubes. Diam Relat Mater
2008;17:157781.
[180] Sun X, Yu A, Ramesh P, Bekyarova E, Itkis ME, Haddon RC. Oxidized
graphite nanoplatelets as an improved ller for thermally conduct-
ing epoxy-matrix composites. J Electron Packag 2011;133:16.
[181] Zhu J, Yudasaka M, Zhang M, Iijima S. Dispersing carbon nano-
tubes in water: a noncovalent and nonorganic way. J Phys Chem
B 2004;108:1131720.
[182] Chen RJ, Zhang Y, Wang D, Dai H. Noncovalent sidewall func-
tionalization of single-walled carbon nanotubes for protein
immobilization. J Am Chem Soc 2001;123:38389.
[183] Simmons TJ, Bult J, Hashim DP, Linhardt RJ, Ajayan PM. Noncova-
lent functionalization as an alternative to oxidative acid treatment
of single wall carbon nanotubes with applications for polymer com-
posites. ACS Nano 2009;3:86570.
[184] Lazarenko A, Vovchenko L, Prylutskyy Y, Matzuy L, Ritter
U, Scharff P. Mechanism of thermal and electrical conductiv-
ity in polymernanocarbon composites. Materwiss Werksttech
2009;40:26872.
[185] Liu L, Etika KC, Liao KS, Hess LA, Bergbreiter DE, Grunlan JC. Com-
parison of covalently and noncovalently functionalized carbon
nanotubes in epoxy. Macromol Rapid Commun 2009;30:62732.
[186] Cao L, Liu X, Na H, Wu Y, Zheng W, Zhu J. How a bio-based epoxy
monomer enhanced the properties of diglycidyl ether of bisphe-
nol A (DGEBA)/graphene composites. J Mater Chem A 2013;1:
50818.
[187] Yang SY, Lin WN, Huang YL, Tien HW, Wang JY, Ma CCM, Li
SM, Wang YS. Synergetic effects of graphene platelets and car-
bon nanotubes on the mechanical and thermal properties of epoxy
composites. Carbon 2011;49:793803.
[188] Teng CC, Ma CCM, Lu CH, Yang SY, Lee SH, Hsiao MC, Yen
MY, Chiou KC, Lee TM. Thermal conductivity and structure of
non-covalent functionalized graphene/epoxy composites. Carbon
2011;49:510716.
[189] Su J, Xiao Y, Ren M. Enhanced thermal conductivity in epoxy
nanocomposites with hybrid boron nitride nanotubes and
nanosheets. Phys Status Sol A 2013;210:2699705.
[190] Choi S, Im H, Kim J. Flexible and high thermal conductivity
thin lms based on polymers: aminated multi-walled carbon
nanotubes/micro-aluminum nitride hybrid composites. Compos-
ites A 2012;43:18608.
[191] Hwang Y, Kim M, Kim J. Improvement of the mechanical proper-
ties and thermal conductivity of poly(ether-ether-ketone) with the
addition of graphene oxide-carbon nanotube hybrid llers. Com-
posites A 2013;55:195202.
[192] Muratov DS, Kuznetsov DV, Ilinykh IA, Burmistrov IN, Mazov
IN. Thermal conductivity of polypropylene composites lled
with silane-modied hexagonal BN. Compos Sci Technol
2015;111:403.
[193] Tamburri E, Orlanducci S, Terranova ML, Valentini F, Palleschi G,
Curulli A, Brunetti F, Passeri D, Alippi A, Rossi M. Modulation of
electrical properties in single-walled carbon nanotube/conducting
polymer composites. Carbon 2005;43:121321.
28 N. Burger et al. / Progress in Polymer Science 61 (2016) 128
[194] Valentini L, Trentini M, Mengoni F, Alongi J, Armentano I, Ricco
L, Mariani A, Kenny JM. Synthesis and photoelectrical properties
of carbon nanotubedendritic porphyrin light harvesting molecule
systems. Diam Relat Mater 2007;16:65863.
[195] Yang K, Gu M. The effects of triethylenetetramine grafting of multi-
walled carbon nanotubes on its dispersion, ller-matrix interfacial
interaction and the thermal properties of epoxy nanocomposites.
Polym Eng Sci 2009;49:215867.
[196] Zhang WB, Xu XL, Yang JH, Huang T, Zhang N, Wang Y, Zhou
ZW. High thermal conductivity of poly(vinylidene uoride)/carbon
nanotubes nanocomposites achieved by adding polyvinylpyrroli-
done. Compos Sci Technol 2015;106:18.
[197] Wu HC, Chang X, Liu L, Zhao F, Zhao Y. Chemistry of carbon nano-
tubes in biomedical applications. J Mater Chem 2010;20:103652.
[198] Ganguli S, Roy AK, Anderson DP. Improved thermal conductivity for
chemically functionalized exfoliated graphite/epoxy composites.
Carbon 2008;46:80617.
[199] Yung KC, Liem H, Choy HS. Prerequisite for maximizing thermal
conductivity of epoxy laminate using ller. J Mater Sci Mater Elec-
tron 2013;24:1095104.
[200] Yuen SM, Ma CCM, Chiang CL, Chang JA, Huang SW, Chen SC, Chuang
CY, Yang CC, Wei MH. Silane-modied MWCNT-PMMA compos-
ite preparation, electrical resistivity, thermal conductivity and
thermal stability. Composites A 2007;38:252735.
[201] Lee HJ, Han SW, Kwon YD, Tan LS, Baek JB. Functionalization of
multi-walled carbon nanotubes with various 4-substituted benzoic
acids in mild polyphosphoric acid/phosphorous pentoxide. Carbon
2008;46:18509.
[202] Barikani M, Hasanzadeh I. Effect of different chemical modication
systems on thermal and electrical conductivity of functionalized
multiwall carbon nanotube/epoxy nanocomposites. Polym Plast
Technol 2013;52:86976.
[203] Gulotty R, Castellino M, Jagdale P, Tagliaferro A, Balandin AA.
Effects of functionalization on thermal properties of single-wall and
multi-wall carbon nanotube-polymer nanocomposites. ACS Nano
2013;7:511421.
[204] Liu CH, Fan SS. Effects of chemical modications on the ther-
mal conductivity of carbon nanotube composites. Appl Phys Lett
2005;86:123106, 14.
[205] Yang SY, Ma CCM, Teng CC, Huang YW, Liao SH, Huang YL, Tien
HW, Lee TM, Chiou KC. Effect of functionalized carbon nano-
tubes on the thermal conductivity of epoxy composites. Carbon
2010;48:592603.
[206] Dai W, Yu J, Liu Z, Wang Y, Song Y, Lyu J, Bai H. Enhanced ther-
mal conductivity and retained electrical insulation for polyimide
composites with SiC nanowires grown on graphene hybrid llers.
Composites A 2015;76:7381.