REVIEW

DOI: 10.1002/adma.200501851

Mechanical Reinforcement of Polymers

Using Carbon Nanotubes**
By Jonathan N. Coleman,* Umar Khan, and Yurii K. Gunko*

Owing to their unique mechanical properties, carbon nanotubes are considered to be ideal can-
didates for polymer reinforcement. However, a large amount of work must be done in order to
realize their full potential. Effective processing of nanotubes and polymers to fabricate new
ultra-strong composite materials is still a great challenge. This Review explores the progress that
has already been made in the area of mechanical reinforcement of polymers using carbon nano-
tubes. First, the mechanical properties of carbon nanotubes and the system requirements to
maximize reinforcement are discussed. Then, main methods described in the literature to pro-
duce and process polymernanotube composites are considered and analyzed. After that,
mechanical properties of various nanotubepolymer composites prepared by different tech-
niques are critically analyzed and compared. Finally, remaining problems, the achievements so
far, and the research that needs to be done in the future are discussed.

1. Introduction tics with exceedingly low percolation thresholds.[3] In another

Since their discovery in 1991, carbon nanotubes have gener-
ated huge activity in most areas of science and engineering
due to their unprecedented physical and chemical properties.
No previous material has displayed the combination of super-
lative mechanical, thermal, and electronic properties attribut-
ed to them. These properties make nanotubes ideal not only
for a wide range of applications[1] but as a test bed for funda-
mental science.[2]
In particular, this combination of properties makes them an
ideal candidate as an advanced filler material in nanocompos-
ites. Researchers have envisaged taking advantage of their
conductivity and high aspect ratio to produce conductive plas-
[*] Dr. J. N. Coleman, U. Khan
School of Physics
University of Dublin Trinity College
Dublin 2 (Ireland)
E-mail: colemaj@tcd.ie
Dr. Y. K. Gunko
School of Chemistry
University of Dublin Trinity College
Dublin 2 (Ireland)
E-mail: igounko@tcd.ie
[**] The authors thank Milo Shaffer and Brian Grady for helpful discus-
sions. In addition, financial help from IRCSET is appreciated.
area, it is thought that their massive thermal conductivity can
be exploited to make thermally conductive composites.[4]
However, probably the most promising area of nanocompos-
ites research involves the reinforcement of plastics using car-
bon nanotubes as a reinforcing filler.
The idea of using pseudo one-dimensional fillers as a rein-
forcing agent is nothing new: straw has been used to reinforce
mud bricks since about 4000 BC. In more recent times fibers
made from materials such as alumina, glass, boron, silicon car-
bide, and especially carbon have been used as fillers in com-
posites. However, these conventional fibers are typically on
the mesoscale with diameters of tens of micrometers and
lengths of order of millimeters. Their mechanical properties
are impressive, with carbon fibers typically displaying stiffness
and strength in the ranges 230725 GPa and 1.54.8 GPa,
respectively.[5] In recent years, carbon nanofibers have been
grown from the vapor phase with diameters on the order of
100 nm and lengths between 20 and 100 lm. These small
dimensions mean they have a much higher surface area per
unit mass than conventional carbon fibers, allowing much
greater interaction with composite matrices. They also tend to
have impressive mechanical properties, with Youngs moduli
in the range 1001000 GPa and strengths between 2.5 and
3.5 GPa.[6]
Nevertheless, the ultimate mechanical filler must be carbon
nanotubes. Nanotubes can have diameters ranging from

Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 689
 J. N. Coleman et al./Carbon NanotubePolymer Composites
REVIEW

1100 nm and lengths of up to millimeters.[7] Their densities
can be as low as 1.3 g cm3 and their Youngs moduli are su-
perior to all carbon fibers, with values greater than 1 TPa.[8]
On the other hand, their strength is what really sets them
apart. The highest measured strength for a carbon nanotube is
63 GPa.[9] This is an order of magnitude stronger than high-
strength carbon fibers. Even the weakest type of carbon nano-
tubes have strengths of several gigapascals.[10]
However, a large amount of work will have to be done
before we can really make the most of the exceptional me-
chanical properties of carbon nanotubes. In this paper, we will
explore the progress that has already been made to this end.
First, we will review the mechanical properties of carbon
nanotubes. This will lead us to a study of the system require-
ments in order to achieve reinforcement. The techniques used
in the literature to produce polymernanotube composites
will be reviewed before we look at what levels of reinforce-
ments have actually been achieved. Finally, we will discuss the
advances so far and work that needs to be done in the future.
2. Mechanical Properties of Nanotubes
The general properties of nanotubes have been discussed in
a number of publications.[1,11,12] In this work we will discuss
only their mechanical properties. From virtually the moment
nanotubes were discovered, it was expected that they would
display superlative mechanical properties by analogy with
graphite. It had long been known that graphite had an in-
plane modulus of 1.06 TPa,[13] and nanotubes were expected
to display similar stiffness. While the tensile strength of graph-
ite was not accurately known, Perepelkin had estimated it to
be as high as 130 GPa from the properties of CC bonds.[14] In
addition, Bacon had fabricated graphite whiskers in 1960 with
a yield strength of 20 GPa.[15] Thus, it was expected that car-
bon nanotubes would be in a class of their own in terms of
high strength and stiffness.
Long before large-scale production of nanotubes had com-
menced, a number of studies had used computer simulation to
investigate their properties. As early as 1993, Overney et al.[16]
calculated the rigidity of short single-walled carbon nanotubes
(SWNTs) using ab initio local density calculations to deter-
mine the parameters in a Keating potential. The calculated
Youngs modulus was 1500 GPa, similar to that of graphite.
This was followed by a range of papers predicting that the
Youngs modulus of nanotubes was approximately 1 TPa and
nanotube type and diameter independent.[17]
The first actual mechanical measurements were made on
multiwalled nanotubes (MWNTs) made by the arc-discharge
process. As only small amounts were available, early measure-
ments were carried out by transmission electron microscopy
(TEM). Treacy et al.[18] measured the amplitude of intrinsic
thermal vibrations observed by TEM. The used this to calcu-
late moduli of 0.41 to 4.15 TPa for a number of tubes. Three
years later, Poncharal et al. induced electromechanical reso-
nant vibrations, giving moduli values between 0.7 and

Dr. Jonathan Coleman received his Ph.D. from The University of Dublin, Trinity College in
Ireland in 1999. After his Ph.D., he was awarded a Higher Education Authority Research
Fellowship to continue his research into polymernanotube composites. In 2001, he became a
contract lecturer in the School of Physics, Trinity College Dublin. He was a visiting scientist at
the Nanotech Institute at the University of Texas at Dallas in 2002. He is now in the School of
Physics, Trinity College, and was recently elected to fellowship of the College. His main research
theme is the study of one-dimensional nanostructures, including carbon nanotubes and inorganic
nanowires. The main focus of this research is to understand the dispersion of these structures in
novel solvents and solutions. This knowledge can then be used to fabricate composites for electri-
cal and mechanical applications.

Yurii Gunko graduated from the Chemistry Department of Moscow State University in 1987.
He also received his Ph.D. degree in Inorganic Chemistry from Moscow State University in
1990. He then worked as a lecturer in Chemistry in Belorussian Institute of Technology
(Belarus). In 1994, he began a postdoctoral position in the group of Professor M. F. Lappert in
the University of Sussex (UK). In 1995, he was awarded an Alexander von Humboldt fellowship
and worked in the University of Magdeburg (Germany) with Professor F. T. Edelmann. He then
returned to the University of Sussex as a postdoctoral researcher. In 1999, Dr. Gunko moved to
the Chemistry Department of Trinity College Dublin (Ireland) to take up a position as a Lecturer
in Inorganic Chemistry. Currently, Dr. Gunko is a Senior Lecturer in the School of Chemistry.
His main research interests and activities are chemical functionalization of carbon nanotubes,
preparation of carbon-nanotubepolymer composites, magnetic nanoparticles, and magnetic
fluids, polyhedral silsesquioxanes, and metallasiloxanes.

690 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites
1.3 TPa.[19] In addition, Falvo et al. observed the reversible
bending of MWNTs with radii of curvature as low as 25 nm
indicating unprecedented flexibility.[20]
The first direct measurement was made by Wong et al. in
1997.[8] They used atomic force microscopy (AFM) to mea-
sure the stiffness constant of arc-MWNTs pinned at one end.
This gave an average Youngs modulus value of 1.28 TPa.
More importantly, they also managed to make the first
strength measurements, obtaining an average bending
strength of 14 GPa. Salvetat et al. used a slightly different
technique. They used AFM to bend an arc-MWNT that had
been pinned at each end over a hole,[21] obtaining an average
modulus value of 810 GPa (Fig. 1).
However, the ultimate measurements were carried out by
Yu et al. in 2000 when they managed to perform stressstrain
measurements on individual arc-MWNTs inside an electron
microscope.[9] For a range of tubes, they obtained modulus
values of 0.270.95 TPa. More interestingly, they showed
Figure 1. a,b) TEM images of typical nanotubes. c,d) AFM images of
nanotubes adhered on a polished ultrafiltration alumina membrane with
a portion bridging a pore of the membrane. a,c) For an arc-discharge
MWNT; b,d) for a catalytic MWNT. e) Cross-section profiles of the nano-
tube (A) and corresponding pore (B) depicted in (c). Reproduced from
[21].
Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
REVIEW
MWNT fracture at strains of up to 12 % and with strengths in
the range 1163 GPa. This allows the estimation of nanotube
toughness as 1240 J g1. In addition, failure was observed at
the outer tube with the inner walls telescoping out in a sword
and sheath mechanism.
Measurements on SWNTs took longer due to the difficul-
ties in handling them. The first measurement was performed
by Salvetat et al. using their AFM method.[22] They observed
a tensile modulus of 1 TPa for small-diameter SWNT bun-
dles by bending methods. However, the properties of larger-
diameter bundles were dominated by shear slippage of in-
dividual nanotubes within the bundle. Yu et al. were able to
measure the tensile properties of bundles by the same method
they used for their MWNT study. They saw moduli in the
range 0.321.47 TPa and strengths between 10 and 52 GPa.
Failure occurred at a maximum of 5.3 %, giving a toughness
of approximately 770 J g1. In addition, they observed that
failure occurred for the nanotubes on the perimeter of the
bundle only, with the rest of the tubes slipping apart.[23]
Intertube slippage within bundles presents a serious limita-
tion to their mechanical properties. The low shear modulus
means that effective moduli and strengths for bundles are far
below those expected for individual SWNTs. In general, it is
extremely difficult to de-bundle SWNTs. Forro and co-work-
ers showed that SWNTs could be fused together in bundles by
electron irradiation.[24,25] By fine-tuning the dose and irradia-
tion energy, they found that they could increase the bundle-
bending modulus to 750 GPa, close to that of individual
SWNTs.
The relatively high values of modulus and strength values
of 1 TPa and tens of gigapascals have been measured on
high-quality SWNTs and arc-discharge MWNTs. However,
nanotubes produced in large quantities by chemical vapor de-
position (CVD) are expected to display significantly reduced
values. The first measurements on CVD MWNTs were car-
ried out by Salvetat et al. using the AFM technique. They
measured Youngs moduli between 12 and 50 GPa.[21] Shortly
afterwards, Xie et al. performed stressstrain measurements
on bundles of CVD-MWNTs.[10] They measured a modulus of
0.45 TPa and values of tensile strength of 4 GPa. The much
larger variation in modulus for CVD-MWNTs compared to
arc-MWNTs strongly suggests that the modulus is very sensi-
tive to defect concentration and type.
3. System Requirements for Mechanical
Reinforcement
There are four main system requirements for effective rein-
forcement. These are a large aspect ratio, good dispersion,
alignment, and interfacial stress transfer. Aspect ratio must be
large to maximize the load transfer to the nanotubes.[5,26,27]
This is crucial in order to optimize composite strength and
stiffness. Dispersion is probably a more fundamental issue.
Nanotubes must be uniformly dispersed as isolated nanotubes
individually coated with polymer. This is imperative in order
www.advmat.de 691

REVIEW
to achieve efficient load transfer to the nanotube network.
This also results in a more uniform stress distribution and
minimizes the presence of stress-concentration centers.
Alignment is, in some ways, a less crucial issue. From geo-
metric considerations, the difference between random orien-
tation and perfect alignment is a factor of five in composite
modulus.[27] While alignment is necessary to maximize
strength and stiffness, it is not always beneficial. Aligned com-
posites have very anisotropic mechanical properties, which
may need to be avoided in bulk samples. In fibers, however,
alignment has no downside and is a good way to maximize
reinforcement.
Probably the most important requirement for a nanotube-
reinforced composite is that external stresses applied to the
composite as a whole are efficiently transferred to the nano-
tubes, allowing them to take a disproportionate share of the
load. The stress in the nanotube is proportional to the shear
stress in the polymer at the interface, which is proportional to
the applied stress at low strains.[28] At some critical applied
stress, the interface will fail. At this point, the shear stress in
the polymer at the interface is known as the interfacial shear
stress (IFSS), s. This parameter controls the maximum stress
transfer to the nanotube.
The interfacial shear strength is an important parameter for
any fiber-reinforced composite, and many studies have been
devoted to it. While the presence of stress transfer can be
demonstrated by Raman spectroscopy,[2932] the magnitude of
the IFSS is difficult to measure.
A number of computational[33,34] and experimental[3538]
studies have resulted in a wide range of values that appear to
cluster around 50100 MPa. However, it should be pointed
out that all these results are for non-covalently bonded com-
posites. Much higher values are expected when nanotubes are
covalently attached to the matrix.[33]
It should be pointed out that Wagner et al. have observed
stress-induced fragmentation of nanotubes in a polymer
matrix.[35] This can only occur for nanotubes longer than the
so-called critical length.[39] This results in the telescoping of
the inner layers while the outer layer remains fixed to the
matrix on either side of the break. The sword and sheath
mechanism has since been observed by a number of research-
ers.[40] This allowed the authors to estimate the IFSS to be
500 MPa. In order to explain this very high value, the authors
suggest the possibility that photoinduced reactions can occur,
resulting in covalent bonding of the matrix to the nanotubes.
Very strong polymernanotube interactions such as these can
result in a number of different nanotube failure mechanisms
such as buckling, collapse, and fragmentation.[41]
4. Polymer Composite Processing
4.1. Solution Processing of Composites
Perhaps the most common method for preparing polymer
nanotube composites has been to mix the nanotubes and poly-
mer in a suitable solvent before evaporating the solvent to
692 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. N. Coleman et al./Carbon NanotubePolymer Composites
form a composite film.[3,4247] One of the benefits of this meth-
od is that agitation of the nanotubes powder in a solvent facili-
tates nanotube de-aggregation and dispersion. Almost all
solution-processing methods are variations on a general
theme that can be summarized as: i) dispersion of nanotubes
in either a solvent or polymer solution by energetic agitation;
ii) mixing of nanotubes and polymer in solution by energetic
agitation; iii) controlled evaporation of solvent, leaving a
composite film.
In general, agitation is provided by magnetic stirring, shear
mixing, reflux, or, most commonly, ultrasonication. Sonication
can be provided in two forms: mild sonication in a bath or
high-power sonication using a tip or horn.
4.2. Melt Processing
While solution processing is a valuable technique for both
nanotube dispersion and composite formation, it is complete-
ly unsuitable for the many polymer types that are insoluble.
Melt processing is a common alternative that is particularly
useful for dealing with thermoplastic polymers.[4854] This
range of techniques makes use of the fact that thermoplastic
polymers soften when heated. Amorphous polymers can be
processed above their glass transition temperature while
semi-crystalline polymers need to be heated above their melt
temperature to induce sufficient softening. The advantages of
this technique are its speed and simplicity, not to mention its
compatibility with standard industrial techniques.[11,48]
Melt processing can be used for production of both bulk
polymer composites and composite fibers.
In general, melt processing involves the melting of polymer
pellets to form a viscous liquid. Any additives, such as carbon
nanotubes, can be mixed into the melt by shear mixing. Bulk
samples can then be fabricated by techniques such as com-
pression molding, injection molding, or extrusion. However, it
is important that processing conditions are optimized, not just
for different nanotube types, but for the whole range of poly-
mernanotube combinations. This is because nanotubes can
affect melt properties such as viscosity, resulting in unex-
pected polymer degradation under conditions of high shear
rates.[51]
For many applications, fibers are more suitable than bulk
materials. In addition, fiber-production techniques tend to be
suited to the alignment of nanotubes within the fiber. A num-
ber of studies have focused on production of composite fibers
by melt processing. Fiber processing is generally similar to
melt processing, but usually involves a process such as extru-
sion to produce an elongated sample which can then be drawn
into a fiber.[48,49,5456]
4.3. Processing of Composites Based on Thermosets
The most common thermosetting polymers used in the for-
mation of polymernanotube composites have been epoxy
Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites
resins. Generally, these are polymers that cure when mixed
with a catalyzing agent or hardener. In most cases, the epoxy
begins life in liquid form, facilitating nanotube dispersion by
the techniques described in Section 5. Curing is then carried
out to convert the liquid composite to the final solid state.
In the simplest cases, nanotubes are dispersed by sonication
in a liquid epoxy such as Shell EPON 828 epoxy.[29] This blend
can then be cured by the addition of hardener, such as trieth-
ylene tetramine, and the application of temperature or pres-
sure.[5759]
4.4. Chemical Processing of Carbon-Nanotube Composites
Chemical processing of carbon-nanotubepolymer compos-
ites involves in-situ polymerization and covalent functional-
ization of nanotubes with polymer molecules.
4.4.1. In-situ Polymerization Processing
Over the last five years, in-situ polymerization in the pres-
ence of carbon nanotubes has been intensively explored for
the preparation of polymer-grafted nanotubes and processing
of the corresponding polymer-composite materials. The main
advantage of this method is that it enables grafting of polymer
macromolecules onto the convex walls of carbon nanotubes.
This then provides a better nanotube dispersion and forma-
tion of a strong interface between the nanotube and the poly-
mer matrix. In addition, in-situ polymerization is a very con-
venient processing technique that allows the preparation of
composites with high nanotube loading and provides very
good miscibility with almost any polymer type. This technique
is particularly important for the preparation of insoluble and
thermally unstable polymers, which can not be processed by
solution or melt processing.
Initially, in-situ radical polymerization was applied for the
synthesis of poly(methyl methacrylate) (PMMA)MWNT
composites by Jia et al.[60] In this work, in-situ polymerization
was performed using the radical initiator 2,2-azobisisobutyro-
nitrile (AIBN). The authors believed that p-bonds in carbon
nanotubes were initiated by AIBN, and therefore nanotubes
could participate in PMMA polymerization to form a strong
interface between the MWNTs and the PMMA matrix. Since
then, Velasco-Santos et al.[61] and Putz et al.[62] have also used
AIBN as an initiator of in-situ radical polymerization to incor-
porate functionalized MWNTs and SWNTs into PMMA
matrices.
In-situ polymerization was also very useful for the prepara-
tion of polyamidecarbon-nanotube composites. For instance,
polyamide 6 (PA-6)MWNT composites have been prepared
by in-situ hydrolytic polymerization of e-caprolactam in the
presence of pristine and carboxylated MWNTs.[63] e-Caprolac-
tam monomer was found to form an electron-transfer com-
plex with MWNTs, giving a homogeneous, polymerizable
master solution, which facilitated the formation of composites
Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
REVIEW
with homogenously dispersed nanotubes. In another work,
Gao et al. reported a new and improved chemical-processing
technology that allows the continuous spinning of SWNT
PA-6 fibers by the in-situ ring-opening polymerization of
caprolactam in the presence of nanotubes (Fig. 2).[64] This
process results in a new hybrid material with characteristics of
both the fiber and the matrix, and with excellent compatibility
between the SWNTs and nylon 6.
In-situ epoxidation has also been used for the preparation
and processing of epoxy-polymer composites. For instance,
carboxyl- and fluorine-functionalized SWNTs have been inte-
grated into epoxy polymers via the formation of covalent
bonds by in-situ epoxy ring-opening esterification and amine-
curing chemical reactions.[65] The same group has further
developed a fully integrated nanotubeepoxy composite
material with direct covalent bonding between the matrix and
SWNTs. This uses functionalized SWNTs prepared via di-
amine reactions with alkylcarboxyl groups directly attached
to the SWNT sidewalls (Fig. 3).[59]
In-situ polymerization reactions have also been applied for
the preparation of other polymernanotube composites. For
example, Kumar et al. have synthesized new ultra-strong
poly(p-phenylene benzobisoxazole) (PBO) composites in the
presence of SWNTs in poly(phosphoric acid) (PPA) by in-situ
PBO polymerization.[66] After the polymerization, PBO/
SWNT composite fibers have been spun from the liquid-crys-
talline solutions using dry-jet spinning.
In general, in-situ polymerization can be used for the prepa-
ration of almost any polymer composite containing carbon
nanotubes that can be non-covalently or covalently bound to
the polymer matrix. The role of covalently functionalized and
polymer-grafted nanotubes is considered in more detail be-
low.
4.4.2. Functionalization of Carbon Nanotubes with Polymer
Molecules (Polymer Grafting)
Covalent functionalization and the surface chemistry of car-
bon nanotubes have been envisaged as very important factors
for nanotubes processing and applications.[67] Recently, many
efforts on polymer-composite reinforcement have been
focused on an integration of chemically modified nanotubes
containing different functional groups into a polymer matrix.
Covalent functionalization can be realized by either modifica-
tion of surface-bound carboxylic acid groups on the nanotubes
or by direct addition of reagents to the sidewalls of nanotubes.
Considered above, in-situ polymerization is also one of the
main approaches for the preparation of polymer-grafted
nanotubes. Therefore, work on the preparation of polymer-
grafted nanotubes frequently overlaps with in-situ polymer-
ization processing. Here we are going to discuss only recent
advances in the production of polymer-grafted nanotubes that
have been used to improve the mechanical properties of poly-
mer composites.
Two main strategies for the covalent grafting of polymers to
nanotubes have been reported: grafting from and grafting
www.advmat.de 693
 J. N. Coleman et al./Carbon NanotubePolymer Composites
REVIEW

Figure 2. Top: Scheme of the preparation of nylon 6SWNT composites by ring-opening polymerization of caprolactam in the presence of SWNTs. Bot-
tom: a) Schematic of the spinneret setup; b) photograph of the setup; c) photograph of the composite fiber; d) Scanning electron microscopy (SEM)
image of cross-sectional fracture of the composite fiber (scale bar: 100 lm). Reproduced with permission from [64]. Copyright 2005 American Chemi-
cal Society.

to. The grafting-from approach is based on the initial im-

a) mobilization of initiators onto the nanotube surface, followed
by the in-situ polymerization of appropriate monomers with
the formation of the polymer molecules bound to the nano-
tube. The advantage of this technique is that polymernano-
tube composites with quite high grafting density can be pre-
pared. However, this method requires strict control of the
amounts of initiator and substrate as well as control of condi-
b) tions for the polymerization reaction. The grafting-to
approach is based on attachment of pre-formed end-function-
alized polymer molecules to functional groups on the nano-
tube surface via chemical reactions. An advantage of this
method is that pre-formed commercial polymers of controlled
mass and distribution can be used. The main limitation of the
grafting-to technique is that initial binding of polymer
chains sterically prevents diffusion of additional macromole-
cules to the surface, leading to a low grafting density. Also,
only polymers containing reactive functional groups can be
used.
An example of the grafting-from strategy is a treatment
of SWNTs with sec-butyl lithium that generates carbanions on
Figure 3. a) Schematic presentation of the integration of nanotubes into
a polymer crosslinked structure. b) SEM image of functionalized nano-
the nanotube surface. These carbanions serve as styrene
tubes SWNTCH2CH2CONHC6H10CH2C6H10NH2. Reproduced from anionic-polymerization initiators for in-situ preparation of
[59]. polystyrene-grafted SWNTs.[68] This procedure permitted

694 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites

REVIEW
de-bundling of SWNTs and produced a homogeneous disper-
sion of nanotubes in solution.
The grafting-from technique was widely used for the
preparation of PMMA- and related-polymer-grafted nano-
tubes. For example, Qin et al. reported the preparation of
poly(n-butyl methacrylate)-grafted SWNTs by attaching n-bu-
tyl methacrylate (nBMA) to the ends and sidewalls of SWNTs
via atom transfer radical polymerization (ATRP) using meth-
yl 2-bromopropionate as a free initiator.[69] A similar ap-
proach was reported by Hwang et al. for the synthesis of
PMMA-grafted MWNTs by potassium persulfate initiated
emulsion polymerization reactions, who used the nanocom-
posites as a reinforcement for commercial PMMA by solution
casting.[40] Xia et al. used an ultrasonically initiated in-situ
emulsion polymerization to functionalize MWNTs with poly-
(butyl acrylate) (PBA) and PMMA polymers. Then, these
polymer-encapsulated carbon nanotubes were used to rein-
force a Nylon 6 matrix.[70]
Using the grafting-from approach, Tong et al. have
modified SWNTs with polyethylene (PE) by in-situ Ziegler
Natta polymerization. In this work, the surface of the SWNTs Figure 4. TEM images of MWNTs grafted with CPP. Reproduced with per-
was initially functionalized with a catalyst (MgCl2/TiCl4), and mission from [74]. Copyright 2004 American Chemical Society.
then ethylene was polymerized, giving PE-grafted SWNTs
that were then mixed with commercial PE by melt blend-
ing.[71]
The grafting-to method was used for the preparation of
composites with polymers containing reactive functional
groups. One of the first examples of the grafting-to ap-
proach was published by Fu et al. in 2001.[72] In this work, car-
boxylic acid groups on the nanotube surface were converted
to acyl chlorides by refluxing the samples in thionyl chloride.
Then, the acyl chloride functionalized carbon nanotubes were
reacted with the hydroxyl groups of dendritic poly(ethylene
glycol) polymers via esterification reactions.
Another original example of the grafting-to approach has
been reported by Lou et al.[73] In this work, MWNTs have
been grafted with polystyrene, poly(e-caprolactone), and their
block copolymers by addition reactions of the correspondent
alkoxyamine-terminated precursors.
A new organometallic modification of the grafting-to
strategy was recently introduced by Blake et al.[74] In this
work, MWNTs have been initially organometallically func-
tionalized using n-butyl lithium and then covalently bonded
to a chlorinated polypropylene (CPP) via a coupling reaction Figure 5. TEM images of MWNTs with attached ferritin molecules. Re-
with elimination of LiCl. The following addition of the CPP- produced with permission from [75]. Copyright 2005 American Institute
of Physics.
grafted nanotubes (Fig. 4) to the CPP polymer matrix resulted
in a significant increase of mechanical properties.
Recently, Bhattacharyya et al. have developed new fully instance, using the amidation reaction of octadecylamine with
integrated nanotubepoly(vinyl alcohol) (PVA) composite nanotubes, Yang et al. prepared functionalized soluble
materials through the functionalization of MWNTs by cova- MWNTs and mixed them in solution with copolymers of
lently attaching ferritin protein molecules onto the surface of methyl and ethyl methacrylate (poly(MMA-co-EMA)).[76]
nanotubes (Fig. 5).[75] Broza et al. prepared new poly(butylene terephthalate)
In addition to the grafting-from and grafting-to strate- (PBT)nanotube composites by introducing the oxidized
gies, there are also a few other techniques for incorporating SWNTs into the reaction mixture during a two-stage polycon-
functionalized carbon nanotubes into polymer matrices. For densation of butylene terephthalate in the molten state.[77]

Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.advmat.de 695

REVIEW
4.5. New Methods for PolymerNanotube Composite
Preparation
A number of novel composite preparation methods have
been reported that are distinct from the traditional methods
described above. While some of these methods, such as elec-
trospinning and coagulation spinning, have been used indus-
trially for years, we have included them in this section as they
are not commonly utilized in the field of polymernanotube
composites.
4.5.1. Composite Films
The simplest of these methods involves the infiltration of
polymer from solution into pre-existing nanotube networks.
This concept was first demonstrated by Coleman et al.[78] in
2003. They first made thin sheets of SWNTs by Buchner filtra-
tion (buckypaper). This buckypaper is extremely porous and
contains up to 70 % free volume. These sheets were then
soaked in polymer solutions for various times before rinsing.
In this way, polymer mass fractions of up to 30 wt.-% could
be incorporated.
A similar technique was demonstrated by Wang et al.[79]
They also fabricated buckypaper, but incorporated an epoxy-
hardener blend by Buchner filtration. To reduce the viscosity,
the blend was mixed with ethanol. Very good infiltration was
observed throughout the paper. The epoxy was cured by hot-
pressing 35 stacked sheets together at 177 C to form a thick
composite film.
An alternative method involves the growth of a nanotube
forest on patterned substrates by CVD.[80] Polydimethylsilox-
ane (PDMS) could then be spin-coated on top of the forest.
The polymer infiltrates the forest during spin-coating. In this
work, the polymer was cured after infiltration at room tem-
perature for 24 h. The composite film could then easily be
peeled from the substrate. The advantage of this method is
that the geometry of the nanotube network can be predeter-
mined by the growth conditions in the CVD chamber.
Another interesting method, first described by Mamedov
et al.[81] and subsequently refined by others,[82] is the layer-by-
layer (LbL) assembly method. This involves building up a
layered composite film by alternately dipping a substrate into
dispersions of SWNTs and polyelectrolyte solutions. In this
way, as many as forty layers are built up. In order to increase
the structural integrity of the film, crosslinking can be in-
duced. After every fifth deposition cycle, a layer of MWNTs
was replaced with a layer of polyacrylamide (PAA) to intro-
duce carboxyl functionalities for amide crosslinking between
polyelectrolytes. The film was then heated to 120 C for
30 min, resulting in amide bonds between PAA and polyethyl-
eneimine (PEI). In addition, covalent bonds also form be-
tween PEI and MWNTs at these temperatures. This method
has significant advantages. Film properties such as thickness
and the polymer/nanotube ratio can easily be controlled. In
addition, very high nanotube loading levels can be obtained.
696 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. N. Coleman et al./Carbon NanotubePolymer Composites
4.5.2. Composite Fibers
Composite fibers can be produced by solution-based meth-
ods as well as the traditional melt-spinning method. An ele-
gant method was demonstrated by Vigolo et al. in 2000.[83] In
this work, SWNTs were dispersed in water with the aid of a
surfactant. This dispersion was then injected into a rotating
bath of PVA dissolved in water such that the nanotubes and
PVA dispersions flowed in the same direction at the point of
injection. Polymer molecules then tended to replace surfac-
tant molecules on the nanotube surface, thus destabilizing the
nanotube dispersion, which collapsed to form a fiber. These
wet fibers could then be retrieved from the bath, rinsed, and
dried. Significant rinsing was used to remove both surfactant
and PVA. Shear forces during the flow led to nanotube align-
ment. However, the fibers could be rewetted and dried under
tension. This resulted in significantly enhanced alignment.
This method was further improved by Dalton et al.[84] They
injected the nanotube dispersion into the center of a co-flow-
ing PVA/water stream in a closed pipe. The wet fiber was then
allowed to flow through the pipe for approximately a meter
before being wound on a rotating mandrel. Flow in more con-
trollable and more uniform pipe conditions resulted in longer
( 100 m) and more stable fibers. Crucially, wet fibers were
not rinsed to remove PVA, although they were dried, produc-
ing fibers with final diameters of tens of micrometers. Further-
more, Miaudet et al.[85] have shown that these fibers can be
drawn at temperatures above the PVA glass-transition tem-
perature, resulting in improved nanotube alignment and poly-
mer crystallinity.
Another method recently used to form composite-based
fibers from solution is electrospinning. This technique has
been used to produce man-made fibers since 1934,[86] and in-
volves electrostatically driving a jet of polymer solution out of
a nozzle onto a metallic counter electrode. In 2003, Ko et al.
described using electrospinning to fabricate polymernano-
tube composite fibers and yarns (Fig. 6).[87] Composite disper-
sions of SWNTs and either poly(lactic acid) (PLA) or poly-
acrylonitrile (PAN) in dimethylformamide were initially
produced. The dispersion was then placed in a pipette with a
0.9 mm nozzle. A wire was placed in the pipette and con-
nected to a steel plate via a high-voltage power supply
(25 kV). The plate was 15 cm below the nozzle. When the
power supply was turned on, the composite solution became
charged, forcing it out of the nozzle and towards the counter
electrode. Charging of the solvent causes rapid evaporation,
resulting in the coalescence of the composite into a fiber
which could be collected from the steel plate. Fibers with
diameters between 10 nm and 1 lm can be produced in this
fashion. Yarns were also produced by collecting the fibers on
a rotating drum and twisting them.
In a similar study, Sen et al. formed fiber-based mem-
branes.[88] They spun from SWNTs dispersed in either polysty-
rene or polyurethane with a 3 cm needle-to-plate gap. In this
work, the solution was pumped slowly out of a needle under
application of 15 kV. Spinning was continued for 1 h until the
Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites
(a)
(b)
Figure 6. Electrospinning of SWNT-filled composites: a) schematic of
the electrospinning process; b) schematic of the continuous yarn-spin-
ning process; c) photograph of continuous yarn manufactured by the
electrospinning process. Reproduced from [87].
counter electrode was covered in a membrane built up from
the spun fibers.
5. Mechanical Properties of PolymerNanotube
Composites
In this section, we will discuss results from the literature on
mechanical properties of polymernanotube composites.
Many studies have been published, each with a different fo-
cus. However, the common theme seems to have been en-
hancement of Youngs moduli. In order to attempt to com-
pare different studies, we will use the rate of increase of
Youngs modulus with volume fraction dY/dVf as a yardstick
for reinforcement. In most papers this quantity is not quoted,
so we have attempted to evaluate it from the data presented.
In addition, as modulus does not always increase linearly with
volume fraction, the slope is calculated at a low volume frac-
tion. As this procedure potentially results in a large error, this
quantity is to be taken as a guide only.
In the interest of brevity, we have chosen to discuss only
publications that demonstrate reinforcement values above the
median value of dY/dVf for their processing type. It should be
Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
REVIEW
pointed out that early studies in this area suffered from poor
dispersion and/or stress transfer. Thus, the low reinforcement
values obtained in many cases are due to these issues and
illustrate the steepness of the initial learning curve in this
field.
5.1. Mechanical Properties of Solution-Processed Composites
Based on Thermoplastic Polymers
In this section, mechanical properties of solution-based
composites will be presented and discussed. While a number
of reports appeared in the literature before 2002,[31,46,8991] the
first significant increase in stiffness was observed by Cadek et
al. using PVA as a matrix material.[42] They carried out nano-
hardness tests on spun-cast films of arc-discharge MWNTs in
both PVA and polyvinylcarbazole (PVK). Increases in moduli
from 7 GPa to 12.6 GPa with 0.6 vol.-% MWNTs in PVA,
and from 2 GPa to 5.6 GPa with 4.8 vol.-% in PVK, were
observed. These increases are equivalent to reinforcement
values of dY/dVf = 990 GPa and dY/dVf = 75 GPa for PVA
and PVK, respectively. However, the value for PVA is proba-
bly artificially inflated, as modulus values for PVA are known
to be closer to 2 GPa.[5] This suggests that the reinforcement
value for PVA should be reduced to dY/dVf 280 GPa. Cru-
cially, Cadek et al. studied the morphology of the polymer
matrices using differential scanning calorimetry (DSC). They
observed that the PVA crystallinity increased linearly with
nanotube content. This suggests the nucleation of crystallinity
by the nanotubes. No such effect was observed for the PVK
samples. This suggests that the difference in reinforcement
may be related to the presence of a crystalline interface for
PVA composites, but an amorphous one for PVK materials.
This indicates the possibility that stress transfer may be maxi-
mized by the presence of an ordered interface.
It should be pointed out that Velasco-Santos et al.[119] mea-
sured reasonably large increases in modulus, from 0.71 GPa
for a poly(methyl methacrylate-co-ethyl methacrylate) co-
polymer to 2.34 GPa with 1 wt.-% arc-MWNTs. This corre-
sponds to a reinforcement of dY/dVf 272 GPa, which is on a
par with Cadeks value (scaled) for PVA composites. How-
ever, in this work no nucleation of crystallinity was observed.
This suggests that good stress transfer can be obtained at an
amorphous interface depending on the polymer.
While it has been shown that both CVD-MWNTs and arc-
MWNTs can be used to a greater or lesser degree to reinforce
polymers, it is not clear what is the optimum tube type. This
question was addressed by Cadek and co-workers in 2004.[43]
They fabricated composites using PVA as a matrix filled with
a range of different tube types: SWNTs, double-walled carbon
nanotubes (DWNTs), CVD-MWNTs, and arc-MWNTs. They
found that for all the nanotube types, except the SWNTs, the
reinforcement as represented by dY/dVf scaled with the
inverse nanotube diameter. This means that the composite
Youngs modulus scales directly with the total interfacial sur-
face in the film. However, for the SWNT composites the rein-
www.advmat.de 697
 J. N. Coleman et al./Carbon NanotubePolymer Composites
REVIEW

forcement was much lower than that which would be expected ites from PA-6, acrylonitrile butadiene styrene (ABS), and
by this trend. This was explained as being due to slippage of nanotubes. Here, the results were more modest, with increases
SWNTs in bundles. This work suggests that the optimum from 1.97 to 2.51 GPa at 7.5 wt.-%. This gives a slightly lower
nanotubes for composite modulus enhancement are small-di- reinforcement value of dY/dVf = 11 GPa. However, this can
ameter CVD-MWNTs. be explained partly by the fact that the nanotubes were ob-
This claim was put to the test shortly afterwards by Cole- served by TEM to reside only in the PA-6 phase. In the three
man et al.[44] They fabricated PVA-based composites using phase blends the ductility fell from 105 % to 40 %, a fall that
commercial low-diameter (D 15 nm) MWNTs as a filler. was accompanied by a decrease in impact strength by a factor
They observed significant reinforcement, with modulus en- of two.
hancement from 1.92 GPa to 7.04 GPa at 0.6 vol.-% nano- Zhang, Liu, and co-workers made composites of CVD-
tubes. This represents a reinforcement of dY/dVf = 754 GPa. MWNTs in PA-6.[95,96] They observed a threefold increase
Also, strength enhancement from 81 MPa to 348 MPa was in modulus from 0.4 GPa to 1.24 GPa on addition of only
observed. In addition, a microscopy study of composite 2 wt.-% nanotubes, corresponding to an impressive reinforce-
fracture surfaces showed nanotube pullout. However, the ment value of 64 GPa. In addition, a significant increase in
pullout diameter showed that a layer of polymer had re- yield strength from 18 to 47 MPa was observed, with similar
mained attached to the tube after fracture. The thickness of increases in ultimate tensile strength. Furthermore, no de-
this layer was very similar to the measured thickness of a crys- crease in toughness was observed, as the ductility only fell
talline polymer coating nucleated by the MWNTs. This slightly from 150 % to 110 %. These impressive results were
strongly suggests that the presence of interfacial crystallinity attributed to very good dispersion and adhesion, as observed
has a large bearing on the mechanical properties of compos- by microscopy measurements (Fig. 7).
ites. Finally, Manchado et al. blended small amounts of arc-
SWNTs into isotactic polypropylene.[97] They observed a mod-
ulus increase from 0.85 GPa to 1.19 GPa at 0.75 wt.-%. In
5.2. Mechanical Properties of Melt-Processed Bulk addition, the strength increased from 31 MPa to 36 MPa with
Composites 0.5 wt.-%. Both properties were observed to fall off at higher
loading levels. The ductility only dropped marginally, from
In this section, we discuss mechanical properties of bulk 493 % to 402 %, meaning that inclusion of nanotubes did not
composites fabricated by melt processing. While a number of result in any significant toughness reduction. This work corre-
other papers have appeared,[48,49,51,92,93] we will discuss only sponds to a reinforcement value of dY/dVf 68 GPa.
those with values of dY/dVf above the median quoted in
Table 1.
Gorga et al. increased the stiffness of PMMA from 2.7 to 5.3. Mechanical Properties of Melt-Processed Fibers
3.7 GPa on addition of 10 wt.-% of CVD-MWNTs (dY/
dVf 17 GPa). In addition, increases in strength from 64 to Some promising mechanical results[48,49,54] have been re-
80 MPa and a 170 % increase in toughness were observed.[94] ported for melt-processed composite fibers. Those with
However, it was in 2004 that a number of researchers began above-median (38 GPa) dY/dVf values will be reviewed in this
to report significant improvements. Meincke et al. fabricated section.
composites of CVD-MWNTs in PA-6.[50] This almost doubled In 2000, Haggenmueller et al. reported on melt-processed
the modulus, from 2.6 to 4.2 GPa at 12.5 wt.-%, correspond- fibers spun from PMMA and SWNTs for a wide range of draw
ing to a reinforcement value of dY/dVf 34 GPa. This was, ratios.[55] The maximum modulus observed was 7 GPa for a
however, accompanied by a significant reduction in ductility, draw ratio of 100. Fiber strength was observed to scale with
from 40 % to just 4 %. In addition, they made blend compos- the draw ratio to a maximum of 130 MPa at a high draw ratio
of 900. A typical value of reinforcement
was dY/dVf 57 GPa for a draw ratio of
Table 1. Summary and comparison of reinforcement of SWNT and MWNT composites fabri- 70. Nanotube alignment within the fibers
cated by various methods, where Y is the composite Youngs modulus, r is the composite
was measured by Raman spectroscopy. For
strength, and Vf is the nanotube volume fraction.
the higher draw ratios, significant alignment
Mean dY/dVf Median dY/dVf Max dY/dVf Max Y Max r was observed, with an orientation distribu-
[GPa] [GPa] [GPa] [GPa] [MPa] tion as low as 4 measured for a draw ratio
SWNTs MWNTs of 300.
Solution 309 128 112 1244 7 348
Fibers fabricated from polypropylene
Melt 23 11 68 64 4.5 80 (PP) and SWNTs were studied by Kearns
Melt (fiber) 128 38 530 36 9.8 1032 and Shambaugh[56] in 2002. The fibers were
Epoxy 231 18 94 330 4.5 41 post-drawn after spinning, which resulted in
In-situ polymerization 430 60150 960 150 167 4200 excellent mechanical properties. They saw
Functionalization 157 115 305 380 29 107
increases in modulus from 6.3 to 9.8 GPa

698 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites

REVIEW
Figure 7. A) SEM image showing an overall morphology of failure surface for PA-6 nanocomposite containing 0.5 % MWNTs. B) Enlarged morphology
of selected region in (A). C) SEM image showing microcracks linked by stretched nanotubes and their bundles in PA-6 nanocomposites. D) Enlarged
SEM image of the microcrack connected by MWNTs sheathed with polymer in the nanocomposites. Some MWNTs are sheathed by several small poly-
mer beads. Reproduced with permission from [95]. Copyright 2004 American Chemical Society.

on addition of 1 wt.-% SWNTs, which corresponds to a rein-
forcement value of dV/dVf 530 GPa. However, the strength
increased from an already impressive value of 709 MPa to
1027 MPa upon addition of 1 wt.-% SWNTs at a draw ratio of
8. This high strength is already good enough for a range of ap-
plications. In addition, the strain to break actually increased
with the introduction of nanotubes, from 19 to 27 %. This
means that the fibers actually toughen when nanotubes are
added.
In an extension of this work, the same group made fibers
containing SWNTs from two types of PP, a high melt-flow-
rate (HMFR) and a low melt-flow-rate (LMFR) polymer.[98]
In both cases, increases in modulus were observed, with rein-
forcements of 179 and 285 GPa for the LMFR and HMFR
PP, respectively. However, while increases in strength, tough-
ness, and ductility were observed for the LMFR polymer, the
opposite was true for the HMFR material. This is an impor-
tant point as it clearly shows that subtle differences in poly-
mer type can have a significant impact on the potential for
reinforcement.
PPSWNT fibers have also been fabricated by Chang et
al.[99] In this case, significant increases in modulus from 0.4 to
1.4 GPa (dY/dVf 38 GPa) were observed. In addition, the
strength tripled on the addition of 5 wt.-% SWNTs. Raman
measurements were used to show significant stress transfer
between polymer and tube. However, no increase in modulus
or orientation was observed as a function of draw ratio. This is
important, as it shows that a fuller understanding of post-
treatment processes is needed before results equivalent to
those of Kearns et al. are routine.
5.4. Mechanical Properties of Composites Based on
Thermosetting Polymers
Due to their wide range of industrial uses, thermosetting
polymers in general and epoxy resins in particular have been
widely studied as potential matrices for nanotube-based com-
posites.[100102] In these sections, all matrices were epoxy resins
unless otherwise stated. As in previous sections, only those

Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.advmat.de 699
 J. N. Coleman et al./Carbon NanotubePolymer Composites
REVIEW

papers with values of dY/dVf above the median (18 GPa) are been divided into subsections. The first deals with cases where
discussed. composite formation at least partially involved the in-situ
The first record of the formation of epoxynanotube com- polymerization of the polymer in the presence of nanotubes.
posites comes from a paper by Ajayan in 1994.[57] In this work, The second section describes composites where the nanotubes
nanotubes were aligned within the epoxy matrix by the shear were initially chemically modified. Composite formation was
forces induced by cutting with a diamond knife. However, no then typically by one of the methods described in Section 5.
quantitative mechanical measurements were made.
The first true mechanical study was made by Schadler et al.
5.5.1. Mechanical Properties of Composites Prepared
in 1998.[30] They measured the stressstrain properties of a
by In-situ Polymerization
MWNTepoxy composite in both tension and compression.
In tension, the modulus increased from 3.1 GPa to 3.71 GPa
In-situ polymerization methods are expected to achieve
on addition of 5 wt.-% nanotubes, a reinforcement of dY/
large interfacial shear strength because they can provide
dVf = 18 GPa. However, better results were seen in compres-
strong, frequently covalent, polymernanotube interactions.
sion, with an increase in modulus from 3.63 to 4.5 GPa, which
While early reports were promising,[106] a breakthrough
corresponds to a reinforcement of 26 GPa. No significant in-
came in a paper by Kumar et al.[66] They polymerized PBO in
creases in strength or toughness were observed. The differ-
the presence of SWNTs before wet-spinning fibers. The PBO
ence between tension and compression was explained by
modulus of 138 GPa was increased to 167 GPa on addition of
Raman studies, which showed significantly better stress trans-
10 wt.-% SWNTs, a reinforcement value of 550 GPa. In
fer to the nanotubes in compression than in tension. This can
addition, the fiber strength was improved from 2.6 GPa to
be explained by the fact that load transfer in compression can
4.2 GPa (Fig. 8). While this does not match the strength of
be thought of as a hydrostatic pressure effect, while load
commercial PBO fibers ( 6 GPa),[107] it is certainly an im-
transfer in tension relies on the matrixnanotube bond. How-
pressive start. Furthermore, incorporation of SWNTs actually
ever, it should be pointed out that later studies showed the
reverse to be true, i.e., load transfer in tension but none in
compression.[29] In further contrast, work by Wood et al. has
shown that mechanical response of SWNTs in tension and
compression are identical.[103]
Further work was carried out on CVD-MWNTepoxy com-
posites by Breton et al. Significant increases in modulus, from
2.75 GPa to 4.13 GPa, were observed on the addition of
6 wt.-% nanotubes, corresponding to dY/dVf 45 GPa. They
attributed the large reinforcement values to residual oxygen-
containing groups that had covalently bonded to the nano-
tubes during purification. The presence of these groups may
have improved interfacial bonding.[104] Bai observed a dou-
bling of Youngs modulus, from 1.2 to 2.4 GPa, on addition of
1 wt.-% CVD-MWNTs. This corresponds to a reinforcement
value of dY/dVf 330 GPa, which is the largest observed for
an epoxy-based composite as measured by tensile testing. In Figure 8. Typical stressstrain curves for PBO and PBO/SWNT (90/10)
addition, significant increases in strength, from 30 to 41 MPa, fibers. Reproduced with permission from [66]. Copyright 2002 American
were recorded. Excellent matrixnanotube adhesion was con- Chemical Society.
firmed by the observation of nanotube breakage during frac-
ture-surface studies.[105] increased the ductility, thereby improving the toughness,
Finally, Li et al.[101] studied SWNTepoxy composites by which is an important consideration for fiber materials.
nanohardness testing. They observed an increase in modulus In another study by Velasco-Santos et al., PMMA was poly-
from 4 to 7 GPa with 5 wt.-% SWNTs (dY/dVf 94 GPa). merized in the presence of arc-MWNTs.[61] This is important
Increases in hardness from 0.35 to 0.45 GPa were also ob- as they have higher intrinsic strength or stiffness than CVD-
served. MWNTs or bundles of SWNTs. Stiffness was increased from
1.5 to 2.5 GPa by addition of 1 wt.-% MWNTs, corresponding
to dY/dVf 150 GPa. In addition, the strength increased from
5.5. Mechanical Properties of Chemically Processed 30 to 50 MPa, with a proportional increase in toughness.
Carbon-Nanotube Composites Finally, Putz et al.[62] in another study on PMMA and
SWNTs observed an increase in stiffness as measured by
In this section, we will discuss composites where chemical dynamic mechanical analysis (DMA) from 0.3 to 0.38 GPa.
reactions have been used either in the formation of the com- However, this increase occurred at the extremely low volume
posite or as a method to treat the nanotubes. This section has fraction of 8 105. This corresponds to a reinforcement value

700 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites

REVIEW
of dY/dVf 960 GPa. This is close to the theoretical rule-of-
mixtures maximum for long, well-aligned, well-graphitized
nanotubes. However, it should be pointed out that this study
was carried out at 100 rad s1, which is a reasonably high fre-
quency. Timetemperature equivalence of mechanical proper-
ties for viscoelastic materials suggests that this may be an
overestimate, as compared was values obtained by pseudo-
static measurements.[108]

5.5.2. Mechanical Properties of Composites Based

on Polymer-Functionalized Nanotubes

Chemical functionalization of the nanotube surface is ex-

pected to maximize composite interfacial shear strength and
facilitate SWNT de-bundling. While it has been shown that
covalent chemical attachments may decrease the maximum
buckling force of nanotubes by about 15 %,[109] the fact that Figure 9. TEM images of a MWNT-containing PMMA thin film taken at
composites based on functionalized nanotubes are expected different times: a) t = 0, b) t = 4 min, and c) t = 10 min. Reproduced from
to have large interfacial shear strengths should more than [40].
compensate. Covalently grafted long-chain molecules entan-
gle with the polymer matrix, ensuring very good stress trans-
fer from matrix to nanotube. In addition, the functional
groups act to make the nanotubes more compatible both with
polymer hosts and solvents. This tends to dramatically im-
prove the nanotubes dispersion and hence further improve
composite properties.
While early reports showed promise,[59,74,76] only those
results above the median value for dY/dVf will de discussed.
Hwang et al. reinforced PMMA by the addition of PMMA-
grafted arc-MWNTs.[40] They observed an increase in modulus
as measured by DMA from 2.9 to 29 GPa on addition of
20 wt.-% nanotubes (dY/dVf 116 GPa). This is significant
for two reasons. First, they managed to successfully functiona- Figure 10. Stressstrain profiles of nylon 6SWNT composite fibers at
lize arc-MWNTs, which is challenging compared to func- different SWNT loadings. The curves are labeled with the percentage of
tionalization of CVD-MWNTs. This means that the fillers SWNTs in the polymer matrix. Reproduced with permission from [64].
Copyright 2005 American Chemical Society.
have a very high modulus. Second, good dispersion and a con-
tinued increase in modulus enhancement was observed up to
20 wt.-%, which is unprecedented. Failure of the nanotubes Finally, the protein, ferritin, was successfully grafted onto
by the sword and sheath method was observed, suggesting CVD-MWNTs.[75] This material was mixed with PVA. The
excellent interfacial stress transfer (Fig. 9). authors claimed that the ferritin was expected to form bonds
Melt-spun fibers from PA-6-grafted SWNTs in a PA-6 with the PVA. Indeed, a significant modulus enhancement
matrix were produced by Gao et al.[64] The fibers were stif- from 3.4 to 7.2 GPa was observed, corresponding to a rein-
fened from 440 MPa to 1200 MPa on addition of 1.5 wt.-% forcement of 380 GPa.
tubes (dY/dVf 120 GPa). The strength was doubled from 41
to 86 MPa (Fig. 10). Both strength and toughness were maxi-
mized at mass fractions of 0.2 %. 5.6. Mechanical Properties of Novel Composites
Liu et al. reinforced PVA with hydroxyl-functionalized
SWNTs.[110] The idea was that the OH groups would hydro- Here we will explore and compare the mechanical proper-
gen bond with the OH of the PVA. A reasonably good mod- ties of composites made using novel processing techniques.
ulus enhancement was observed with an increase from 2.4 to
4.3 GPa on addition of 0.8 wt.-% nanotubes, corresponding to
a reinforcement value of 305 GPa. A strength increase, from 5.6.1. Composites Prepared by Infiltration Methods
74 to 107 MPa, was also observed. These results were ex-
plained by the observation using Raman spectroscopy of good These techniques have focused on improving the mechani-
load transfer. cal properties of existing nanotube-based structures by infil-

Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.advmat.de 701
 J. N. Coleman et al./Carbon NanotubePolymer Composites
REVIEW

trating polymer from solution to act as a binding agent. Cole-

man et al. infiltrated PVA, PVP, and PS into buckypaper.[78]
They found increases in modulus, strength, toughness, and
strain to break for the infiltrated papers. At maximal loading
levels of approximately 30 wt.-% polymer, the modulus,
strength, and toughness increased by factors of 3, 9, and
28, respectively. This was attributed to improvements in
inter-bundle load transfer by polymer bridging. There were
slight differences in load transfer, with PS > PVP > PVA, in
agreement with their relative lipophilicities. In a further study,
Frizzell et al. showed that infiltration of high-molecular-
weight molecules was better for improving modulus and
strength, while lower-molecular-weight polymers were opti-
mal for the improvement of toughness. This was attributed to
variations in chain conformation related to the intercalation
process.[111] In a similar study, Wang et al. demonstrated se-
venfold improvements in buckypaper stiffness by intercala-
tion of epoxy, resulting in stiffnesses of up to 15 GPa.[79]
Lahiff et al.[112] infiltrated PDMS into a MWNT forest by
spin-coating. Increases in modulus by a factor of two were
observed, although the film was weaker and more brittle than
neat PDMS. The low level of reinforcement was attributed to
poor stress transfer.
It should be pointed out that these infiltration methods are
potentially important as they could be used to enhance the
mechanical properties of nanotube fibers[113,114] or sheets.[115]

5.6.2. Composites Prepared via Layer-by-Layer Deposition

Methods

Composites consisting of alternating layers of nanotubes

and polyelectrolyte have shown great promise as high-
strength, lightweight materials. Mamedov demonstrated such
materials fabricated from SWNTs and PEI.[81] Compared
with neat PEI, significant increases in modulus, from 0.3 GPa
to 11 GPa, were observed for composites containing
50 wt.-% SWNTs. Increases in strength, from 9 MPa to up
to 325 MPa, were observed. While the composites were less
ductile than the neat polymer (1 % versus 4 %), they still
showed a large increase in toughness by a factor of five to ten.
In a similar study, Olek et al.[82] fabricated LbL composites
using MWNTs and PEI. Two types of MWNT were used: hol-
low and bamboo (Fig. 11). The bamboo MWNT com-
posites displayed moduli and strengths of 4.5 GPa and
150 MPa, respectively. The hollow MWNT composites
showed slightly lower values of 2 GPa and 110 MPa, re-
spectively. To further emphasize the importance of tube type
and quality, hollow nanotubes were boiled in nitric acid
before composite fabrication. This resulted in much lower
modulus and strength values of 0.2 GPa and 35 MPa, re-
spectively.
5.6.3. Coagulation-Spun Fibers
In 2000, Vigolo et al. demonstrated that nanotube compos-
ite ribbons and fibers could be spun by coagulation spinning.
Figure 11. SEM images of a) hollow and b) bamboo (b) MWNTs. Re-
produced with permission from [82]. Copyright 2004 American Chemical
Society.
However, care was taken to remove the polymer post-spin-
ning. These fibers displayed reasonable moduli and strengths
of 915 GPa and 150 MPa, respectively.[83] They were ex-
tremely flexible and could easily be tied into knots (Fig. 12).
The mechanical properties of these fibers could be signifi-
cantly improved by stretching when wet. This tended to align
the nanotubes, resulting in moduli of 40 GPa and strengths
of up to 230 MPa.[116]
However, Dalton et al. demonstrated in 2003 that the prop-
erties could be markedly improved by leaving the polymer in
the fiber.[84] This resulted in large increases in Youngs modu-
lus and strength, to 80 GPa and1.8 GPa, respectively. In addi-
tion, these fibers could be stretched up to strains of 100 %,
resulting in massive toughness values of up to 570 J g1 which
is an order of magnitude larger than Kevlar. Miaudet et al.[85]
have recently shown that toughness values as high as 870 J g1
and 690 J g1 can be achieved for similar fibers spun from
SWNTs and MWNTs, respectively. In addition, by drawing
the fibers at elevated temperature, they have shown that
SWNT-based fibers with excellent all-around properties can

702 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites
Figure 12. Optical microphotographs of nanotube ribbons and fibers.
A) A single folded ribbon. B) A freestanding nanotube fiber between two
glass substrates (scale bar = 1 mm). C) Tying knots reveals the high flex-
ibility and the resistance to torsion of the nanotube microfibers. The fiber
shown in the picture is about 15 lm in radius. Reproduced with permis-
sion from [83]. Copyright 2000 American Association for the Advance-
ment of Science.
be obtained. These fibers had modulus, strength, and tough-
ness values of 45 GPa, 1.8 GPa, and 55 J g1, respectively.
Crucially, this toughness value was obtained at the low strain
value of 11 %.
Fibers made from SWNTs and PVA have also been spun
using DNA as a dispersant. These materials demonstrated
moduli in the range 1218 GPa and strengths of 90
110 MPa.[117]
5.6.4. Electrospun Fibers
Composite fibers have also been spun by electrospinning.
Ko et al. demonstrated stabilized PAN/SWNT fibers with sig-
nificantly increased Youngs moduli.[87] The fiber modulus
was 140 GPa at 4 wt.-% nanotubes, which compares well with
the PAN fiber modulus of 60 GPa. However, it should be
noted that these measurements were carried out in compres-
sion using an atomic force microscope. Sen et al. built up
membranes from electrospun fibers of SWNTs and polyure-
thane. These could then by characterized by tensile testing.
Polyurethane fibers had moduli and strengths of 7 MPa and
7 MPa, respectively. However, the membranes fabricated
from composite fibers demonstrated moduli and strengths of
up to 25 and 15 MPa.[88]
6. Conclusions and Future Outlook
Shown in Table 1 is a summary of the mechanical results
discussed in the previous sections for each composite prepara-
tion type. The values shown represent the mean and median
reinforcement values. In addition, the maximum value report-
ed in each section is reported for composites based on both
Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
REVIEW
SWNTs and MWNTs. Finally, the maximum reported values
for modulus and strength are also shown.
The most obvious result here is that composites based on
chemically modified nanotubes show the best results. This is
not surprising, as functionalization, for example, should signif-
icantly improve both dispersion and stress transfer. While the
results for solution-based composites are also impressive,
these results are biased by the high-crystallinity PVA-based
composites.[43,44] While these undoubtedly show significant
reinforcement, much of this is due to crystallinity nucleation,
which cannot be relied upon for all polymer matrices. How-
ever, details of the solution-based polymernanotube interac-
tion may result in a better interface. This fact needs clarifica-
tion.
A salient message is the low reinforcement observed for the
melt-processed composites. While the fibers display better
reinforcement than the bulk, the difference is consistent with
alignment effects. Of all the processing methods, the overall
results are worst for the melt-based systems. While the results
for the epoxy-based systems are slightly better, they are not
especially noteworthy.
It should be pointed out, though, that the best results
reported are actually very good. Both the rule of mixtures and
the HalpinTsai equations (taking Ym 1 GPa and the length/
diameter ratio L/D 1000) predict maximum dY/dVf values of
approximately 1 TPa. A number of the systems studied report
values in this range.[44,56,62]
We can also compare SWNT- versus MWNT-based compos-
ites. Table 1 shows that the best reinforcement occurred for
MWNT-based composites in half the categories. This shows it
is premature to assume that SWNTs are superior, as many
commentators have.
Another disappointing fact is that strength enhancement
has been poor except in the case of composite fibers. Given
that high strength is one area where nanotubes retain signifi-
cant advantages over carbon fibers, this is unexpected and
must be addressed.
These results suggest that there are three main areas where
significant improvement is urgently needed. As suggested pre-
viously, the problems with mechanical reinforcement of melt-
processed composites must be addressed. Any nanotube-rein-
forced composites produced at an industrial level are likely to
be produced by melt processing. It is likely that the problems
are based on dispersion or interfacial shear strength issues. If
this is the case, these may be addressed by functionalization
of the nanotubes. This will result in them being more compati-
ble with the polymer melt and so improve dispersion. In addi-
tion, IFSS will be improved for the reasons discussed above.
Another significant problem is the disappointing values for
composite strength, as discussed above. A more important is-
sue is probably a partial cause of this. In very few cases are
large volume fractions attained. One notable exception is the
work of Hwang et al., where the modulus increased up to
20 wt.-%. This resulted in by far the highest modulus for bulk
composites. It is imperative that high-volume-fraction com-
posites with good dispersion can routinely be made.
www.advmat.de 703

REVIEW
How do we address these issues? One possibility is to opti-
mize the nanotube type. The most obvious question is: SWNT
or MWNT? Here, the high-volume-fraction issue may pro-
vide an answer. Imagine that we could routinely produce
individual, isolated SWNTs dispersed in a composite. In that
situation, we can easily calculate that for Vf = 1 %, every poly-
mer strand would be within 5 nm of a nanotube. This separa-
tion is of order of the radius of gyration of typical polymers
and must surely represent an upper bound for the volume
fraction.[118] While much higher volume fractions have been
attained, all SWNTs exist in bundles under these conditions.
Under these circumstances, SWNTs lose their intrinsic advan-
tages.
In the case of MWNTs with diameters of 10 nm, however,
even at Vf = 10 %, the nanotubes are on average separated by
20 nm. This gives us much more room to attain high volume
fractions with MWNT. The volume fraction at which every
polymer strand is within 5 nm of a nanotube is a massive
25 vol.-%. Thus, in this ideal world, arc-MWNTs would be the
obvious choice as the ultimate filler. However, compared to
CVD-MWNTs, they are very expensive and difficult to func-
tionalize. A better option would be to look for some cheap
method to improve the graphitization of CVD-MWNTs.
Length and diameter can also be optimized. In order to
maximize composite strength and stiffness, long tubes are
required. Assuming an interfacial shear strength of 50 MPa
and a strength of 50 GPa for SWNTs and arc-MWNTs, and
10 GPa for CVD-MWNTs, we can calculate the so-called crit-
ical length in each case using the KellyTyson equation.[39]
These work out as 1 lm for arc-MWNTs, 100500 nm for
CVD-MWNTs, and 400 nm for SWNTs. In order to maxi-
mize reinforcement, we will need nanotubes with lengths sev-
eral times the critical length,[39] possibly > 5 lm. However,
long tubes are very hard to disperse properly, so we will be
faced with a trade-off. Similarly, small-diameter nanotubes
give rise to greater surface area, allowing the maximization of
interaction with the matrix. However, for reasons already dis-
cussed, if the diameter becomes too small, the maximum
nanotubes loading level will be compromised. For these rea-
sons, there must be an optimum diameter and length.
Once we have identified the optimum tube type, we must
maximize nanotube solubility, dispersion (and hence, volume
fraction), and stress transfer. Fortunately, all these factors can
be addressed as one by functionalization of the nanotubes.
Choice of the right functional group will optimize interactions
with either the solvent or the melt and aid stress transfer. In
order to perfect the functionalization scheme, there are three
aspects to consider. The molecular structure of the group must
be chosen. This may, for example, be a short-chain analogue
to the host polymer. In addition, the group length and surface
density would be expected to affect dispersion and stress
transfer.
Once the nanotubes have been fully optimized, we can
consider what mechanical properties are attainable. We can
calculate these properties for a typical thermoplastic with a
low intrinsic modulus and strength. Assuming that the
704 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. N. Coleman et al./Carbon NanotubePolymer Composites
optimum tube length is a few times the critical length,[39] we
can use the simple rules of mixtures for modulus and
strength.[5] We can assume that the upper limit for SWNT
loading is 1 vol.-% (as discussed) and that YSWNT = 1 TPa and
rSWNT = 50 GPa. Then, for an aligned composite, YC 10 GPa
and rC 0.5 GPa.
We can do significantly better for arc-MWNTs. We can as-
sume a conservative upper limit for a volume fraction as
20 vol.-%, as discussed above. Then, taking YMWNT = 1 TPa
and rMWNT = 50 GPa, as for SWNTs we obtain YC 200 GPa
and rC 10 GPa for an aligned composite. While these
numbers are significantly better than those for SWNTs, they
represent an upper limit. Real applications will almost
certainly use CVD-MWNTs, which have poorer mechanical
properties. If we assume for CVD-MWNTs YMWNT = 300 GPa
and rMWNT = 10 GPa, which seem reasonable, then we can
achieve YC 60 GPa and rC 2 GPa for an aligned compos-
ite. While these values are not momentous, they are respect-
able. In addition, they should be considered lower bounds. By
careful functionalization, it may be possible to push the maxi-
mum volume fraction beyond 20 vol.-%. In addition, the in-
trinsic properties of CVD-MWNTs can possibly be improved
by high-temperature annealing in analogy to vapor-grown car-
bon fibers. The challenge will be to find a cost-effective way
to achieve this.
Received: September 2, 2005
Final version: November 16, 2005
Published online: February 22, 2006
[1] R. H. Baughman, A. A. Zakhidov, W. A. de Heer, Science 2002,
297, 787.
[2] J. Cao, Q. Wang, M. Rolandi, H. Dai, Phys. Rev. Lett. 2004, 93,
216 803.
[3] B. E. Kilbride, J. N. Coleman, J. Fraysse, P. Fournet, M. Cadek,
A. Drury, S. Hutzler, S. Roth, W. J. Blau, J. Appl. Phys. 2002, 92,
4024.
[4] M. J. Biercuk, M. C. Llaguno, M. Radosavljevic, J. K. Hyun, A. T.
Johnson, Appl. Phys. Lett. 2002, 80, 2767.
[5] W. D. Callister, Jr., Materials Science and Engineering: An Introduc-
tion, 6th ed., Wiley, New York 2003.
[6] G. G. Tibbetts, C. P. Beetz, J. Phys. D: Appl. Phys. 1987, 20, 292.
[7] K. Hata, D. N. Futaba, K. Mizuno, T. Namai, M. Yumura, S. Iijima,
Science 2004, 306, 1362.
[8] E. W. Wong, P. E. Sheehan, C. M. Lieber, Science 1997, 277, 1971.
[9] M. Yu, O. Lourie, M. J. Dyer, T. F. Kelly, R. S. Ruoff, Science 2000,
287, 637.
[10] S. Xie, W. Li, Z. Pan, B. Chang, L. Sun, J. Phys. Chem. Solids 2000,
61, 1153.
[11] O. Breuer, U. Sundararaj, Polym. Compos. 2004, 25, 630.
[12] P. J. F. Harris, Int. Mater. Rev. 2004, 49, 31.
[13] B. T. Kelly, Physics of Graphite, Applied Science, London 1981.
[14] K. Perepelkin, Sov. Mater. Sci. 1972, 8, 198.
[15] R. Bacon, J. Appl. Phys. 1960, 31, 283.
[16] G. Overney, W. Zhong, D. Tomanek, Z. Phys. D: At., Mol. Clusters
1993, 27, 93.
[17] J. P. Lu, J. Phys. Chem. Solids 1997, 58, 1649.
[18] M. M. J. Treacy, T. W. Ebbesen, J. M. Gibson, Nature 1996, 381, 678.
[19] P. Poncharal, Z. L. Wang, D. Ugarte, W. A. de Heer, Science 1999,
283, 1513.
Adv. Mater. 2006, 18, 689706
J. N. Coleman et al./Carbon NanotubePolymer Composites
[20] M. R. Falvo, G. J. Clary, R. M. Taylor, V. Chi, F. P. Brooks, S. Wash-
burn, R. Superfine, Nature 1997, 389, 582.
[21] J. P. Salvetat, A. J. Kulik, J. M. Bonard, G. A. D. Briggs, T. Stockli,
K. Metenier, S. Bonnamy, F. Beguin, N. A. Burnham, L. Forro, Adv.
Mater. 1999, 11, 161.
[22] J. P. Salvetat, G. A. D. Briggs, J. M. Bonard, R. R. Bacsa, A. J. Ku-
lik, T. Stockli, N. A. Burnham, L. Forro, Phys. Rev. Lett. 1999, 82,
944.
[23] M. F. Yu, B. S. Files, S. Arepalli, R. S. Ruoff, Phys. Rev. Lett. 2000,
84, 5552.
[24] C. Miko, M. Milas, J. W. Seo, E. Couteau, N. Barisic, R. Gaal, L. For-
ro, Appl. Phys. Lett. 2003, 83, 4622.
[25] A. Kis, G. Csanyi, J. P. Salvetat, T. N. Lee, E. Couteau, A. J. Kulik,
W. Benoit, J. Brugger, L. Forro, Nat. Mater. 2004, 3, 153.
[26] H. L. Cox, Br. J. Appl. Phys. 1952, 3, 72.
[27] H. Krenchel, Fibre Reinforcement, Akademisk, Copenhagen, Den-
mark 1964.
[28] A. Kelly, N. H. Macmillan, Strong Solids, 3rd ed., Oxford University
Press, Oxford, UK 1986.
[29] P. M. Ajayan, L. S. Schadler, C. Giannaris, A. Rubio, Adv. Mater.
2000, 12, 750.
[30] L. S. Schadler, S. C. Giannaris, P. M. Ajayan, Appl. Phys. Lett. 1998,
73, 3842.
[31] S. L. Ruan, P. Gao, X. G. Yang, T. X. Yu, Polymer 2003, 44, 5643.
[32] C. A. Cooper, R. J. Young, M. Halsall, Composites, Part A 2001, 32,
401.
[33] S. J. V. Frankland, A. Caglar, D. W. Brenner, M. Griebel, J. Phys.
Chem. B 2002, 106, 3046.
[34] K. Liao, S. Li, Appl. Phys. Lett. 2001, 79, 4225.
[35] H. D. Wagner, O. Lourie, Y. Feldman, R. Tenne, Appl. Phys. Lett.
1998, 72, 188.
[36] C. A. Cooper, S. R. Cohen, A. H. Barber, H. D. Wagner, Appl.
Phys. Lett. 2002, 81, 3873.
[37] A. H. Barber, S. R. Cohen, H. D. Wagner, Appl. Phys. Lett. 2003, 82,
4140.
[38] A. H. Barber, S. R. Cohen, S. Kenig, H. D. Wagner, Compos. Sci.
Technol. 2004, 64, 2283.
[39] A. Kelly, W. R. Tyson, J. Mech. Phys. Solids 1965, 13, 329.
[40] G. L. Hwang, Y.-T. Shieh, K. C. Hwang, Adv. Funct. Mater. 2004, 14,
487.
[41] O. Lourie, D. M. Cox, H. D. Wagner, Phys. Rev. Lett. 1998, 81, 1638.
[42] M. Cadek, J. N. Coleman, V. Barron, K. Hedicke, W. J. Blau, Appl.
Phys. Lett. 2002, 81, 5123.
[43] M. Cadek, J. N. Coleman, K. P. Ryan, V. Nicolosi, G. Bister, A. Fon-
seca, J. B. Nagy, K. Szostak, F. Beguin, W. J. Blau, Nano Lett. 2004,
4, 353.
[44] J. N. Coleman, M. Cadek, R. Blake, V. Nicolosi, K. P. Ryan, C. Bel-
ton, A. Fonseca, J. B. Nagy, Y. K. Gunko, W. J. Blau, Adv. Funct.
Mater. 2004, 14, 791.
[45] F. Dalmas, L. Chazeau, C. Gauthier, K. Masenelli-Varlot, R. Dendie-
vel, J. Y. Cavaill, L. Forr, J. Polym. Sci., Part B: Polym. Phys. 2005,
43, 1186.
[46] A. Dufresne, M. Paillet, J. L. Putaux, R. Canet, F. Carmona, P. Del-
haes, S. Cui, J. Mater. Sci. 2002, 37, 3915.
[47] B. McCarthy, J. N. Coleman, R. Czerw, A. B. Dalton, M. I. H. Pan-
huis, A. Maiti, A. Drury, P. Bernier, J. B. Nagy, B. Lahr, H. J. Byrne,
D. L. Carroll, W. J. Blau, J. Phys. Chem. B 2002, 106, 2210.
[48] R. Andrews, D. Jacques, M. Minot, T. Rantell, Macromol. Mater.
Eng. 2002, 287, 395.
[49] R. Andrews, D. Jacques, D. L. Qian, T. Rantell, Acc. Chem. Res.
2002, 35, 1008.
[50] O. Meincke, D. Kaempfer, H. Weickmann, C. Friedrich, M. Vathauer,
H. Warth, Polymer 2004, 45, 739.
[51] P. Potschke, A. R. Bhattacharyya, A. Janke, H. Goering, Compos.
Interfaces 2003, 10, 389.
Adv. Mater. 2006, 18, 689706 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
REVIEW
[52] P. Potschke, T. D. Fornes, D. R. Paul, Polymer 2002, 43, 3247.
[53] J. Sandler, G. Broza, M. Nolte, K. Schulte, Y. M. Lam, M. S. P. Shaf-
fer, J. Macromol. Sci., Part B: Phys. 2003, 42, 479.
[54] J. K. W. Sandler, S. Pegel, M. Cadek, F. Gojny, M. van Es, J. Lohmar,
W. J. Blau, K. Schulte, A. H. Windle, M. S. P. Shaffer, Polymer 2004,
45, 2001.
[55] R. Haggenmueller, H. H. Gommans, A. G. Rinzler, J. E. Fischer,
K. I. Winey, Chem. Phys. Lett. 2000, 330, 219.
[56] J. C. Kearns, R. L. Shambaugh, J. Appl. Polym. Sci. 2002, 86, 2079.
[57] P. M. Ajayan, O. Stephan, C. Colliex, D. Trauth, Science 1994, 265,
1212.
[58] Q. Q. Li, M. Zaiser, V. Koutsos, Phys. Status Solidi A 2004, 201, R89.
[59] J. Zhu, H. Peng, F. Rodriguez-Macias, J. L. Margrave, V. N. Khaba-
shesku, A. M. Imam, K. Lozano, E. V. Barrera, Adv. Funct. Mater.
2004, 14, 643.
[60] Z. Jia, Z. Wang, C. Xu, J. Liang, B. Wei, D. Wu, S. Zhu, Mater. Sci.
Eng., A 1999, 271, 395.
[61] C. Velasco-Santos, A. L. Martinez-Hernandez, F. T. Fisher, R. Ruoff,
V. M. Castano, Chem. Mater. 2003, 15, 4470.
[62] K. W. Putz, C. A. Mitchell, R. Krishnamoorti, P. F. Green, J. Polym.
Sci., Part B: Polym. Phys. 2004, 42, 2286.
[63] C. Zhao, G. Hu, R. Justice, D. W. Schaefer, S. Zhang, M. Yang, C. C.
Han, Polymer 2005, 46, 5125.
[64] J. Gao, M. E. Itkis, A. Yu, E. Bekyarova, B. Zhao, R. C. Haddon,
J. Am. Chem. Soc. 2005, 127, 3847.
[65] J. Zhu, J. Kim, H. Peng, J. L. Margrave, V. N. Khabashesku, E. V.
Barrera, Nano Lett. 2003, 3, 1107.
[66] S. Kumar, T. D. Dang, F. E. Arnold, A. R. Bhattacharyya, B. G. Min,
X. Zhang, R. A. Vaia, C. Park, W. W. Adams, R. H. Hauge, R. E.
Smalley, S. Ramesh, P. A. Willis, Macromolecules 2002, 35, 9039.
[67] S. Banerjee, T. Hemraj-Benny, S. S. Wong, Adv. Mater. 2005, 17, 17.
[68] G. Viswanathan, N. Chakrapani, H. Yang, B. Wei, H. Chung, K. Cho,
C. Y. Ryu, P. M. Ajayan, J. Am. Chem. Soc. 2003, 125, 9258.
[69] S. Qin, D. Qin, W. T. Ford, D. E. Resasco, J. E. Herrera, Macromole-
cules 2004, 37, 752.
[70] H. Xia, Q. Wang, G. Qiu, Chem. Mater. 2003, 15, 3879.
[71] X. Tong, C. Liu, H.-M. Cheng, H. Zhao, F. Yang, X. Zhang, J. Appl.
Polym. Sci. 2004, 92, 3697.
[72] K. Fu, W. Huang, Y. Lin, L. A. Riddle, D. L. Carroll, Y. P. Sun, Nano
Lett. 2001, 1, 439.
[73] X. Lou, C. Detrembleur, V. Sciannamea, C. Pagnoulle, R. Jerome,
Polymer 2004, 45, 6097.
[74] R. Blake, Y. K. Gunko, J. Coleman, M. Cadek, A. Fonseca, J. B.
Nagy, W. J. Blau, J. Am. Chem. Soc. 2004, 126, 10 226.
[75] S. Bhattacharyya, C. Sinturel, J. P. Salvetat, M.-L. Saboungi, Appl.
Phys. Lett. 2005, 86, 113 104.
[76] J. Yang, J. Hu, C. Wang, Y. Qin, Z. Guo, Macromol. Mater. Eng.
2004, 289, 828.
[77] G. Broza, M. Kwiatkowska, Z. Roslaniec, K. Schulte, Polymer 2005,
46, 5860.
[78] J. N. Coleman, W. J. Blau, A. B. Dalton, E. Munoz, S. Collins, B. G.
Kim, J. Razal, M. Selvidge, G. Vieiro, R. H. Baughman, Appl. Phys.
Lett. 2003, 82, 1682.
[79] Z. Wang, Z. Liang, B. Wang, C. Zhang, L. Kramer, Composites, Part
A 2004, 35, 1225.
[80] E. Lahiff, C. Y. Ryu, S. Curran, A. I. Minett, W. J. Blau, P. M. Aja-
yan, Nano Lett. 2003, 3, 1333.
[81] A. A. Mamedov, N. A. Kotov, M. Prato, D. M. Guldi, J. P. Wicksted,
A. Hirsch, Nat. Mater. 2002, 1, 190.
[82] M. Olek, J. Ostrander, S. Jurga, H. Mohwald, N. Kotov, K. Kempa,
M. Giersig, Nano Lett. 2004, 4, 1889.
[83] B. Vigolo, A. Penicaud, C. Coulon, C. Sauder, R. Pailler, C. Journet,
P. Bernier, P. Poulin, Science 2000, 290, 1331.
[84] A. B. Dalton, S. Collins, E. Munoz, J. M. Razal, V. H. Ebron, J. P.
Ferraris, J. N. Coleman, B. G. Kim, R. H. Baughman, Nature 2003,
423, 703.
www.advmat.de 705

REVIEW
[85] P. Miaudet, S. Badaire, M. Maugey, A. Derre, V. Pichot, P. Launois,
P. Poulin, C. Zakri, Nano Lett. 2005, 5, 2212.
[86] A. Formhals, US Patent 1 975 504, 1934.
[87] F. Ko, Y. Gogotsi, A. Ali, N. Naguib, H. H. Ye, G. L. Yang, C. Li,
P. Willis, Adv. Mater. 2003, 15, 1161.
[88] R. Sen, B. Zhao, D. Perea, M. E. Itkis, H. Hu, J. Love, E. Bekyarova,
R. C. Haddon, Nano Lett. 2004, 4, 459.
[89] D. Qian, E. C. Dickey, R. Andrews, T. Rantell, Appl. Phys. Lett.
2000, 76, 2868.
[90] B. Safadi, R. Andrews, E. A. Grulke, J. Appl. Polym. Sci. 2002, 84,
2660.
[91] M. S. P. Shaffer, A. H. Windle, Adv. Mater. 1999, 11, 937.
[92] Z. Jin, K. Pramoda, G. Xu, S. H. Goh, Chem. Phys. Lett. 2001, 337,
43.
[93] E. T. Thostenson, T.-W. Chou, J. Phys. D: Appl. Phys. 2002, 35, L77.
[94] R. E. Gorga, R. E. Cohen, J. Polym. Sci., Part B: Polym. Phys. 2004,
42, 2690.
[95] T. X. Liu, I. Y. Phang, L. Shen, S. Y. Chow, W.-D. Zhang, Macromol-
ecules 2004, 37, 7214.
[96] W. D. Zhang, L. Shen, I. Y. Phang, T. X. Liu, Macromolecules 2004,
37, 256.
[97] M. A. L. Manchado, L. Valentini, J. Biagiotti, J. M. Kenny, Carbon
2005, 43, 1499.
[98] E. M. Moore, D. L. Ortiz, V. T. Marla, R. L. Shambaugh, B. P. Gra-
dy, J. Appl. Polym. Sci. 2004, 93, 2926.
[99] T. E. Chang, L. R. Jensen, A. Kisliuk, R. B. Pipes, R. Pyrz, A. P. So-
kolov, Polymer 2005, 46, 439.
[100] A. Allaoui, S. Bai, H. M. Cheng, J. B. Bai, Compos. Sci. Technol.
2002, 62, 1993.
[101] X. D. Li, H. S. Gao, W. A. Scrivens, D. L. Fei, X. Y. Xu, M. A. Sut-
ton, A. P. Reynolds, M. L. Myrick, Nanotechnology 2004, 15, 1416.
[102] T. Ogasawara, Y. Ishida, T. Ishikawa, R. Yokota, Composites, Part A
2004, 35, 67.
______________________
706 www.advmat.de 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. N. Coleman et al./Carbon NanotubePolymer Composites
[103] J. R. Wood, Q. Zhao, M. D. Frogley, E. R. Meurs, A. D. Prins,
T. Peijs, D. J. Dunstan, H. D. Wagner, Phys. Rev. B: Condens. Matter
Mater. Phys. 2000, 62, 7571.
[104] Y. Breton, G. Desarmot, J. P. Salvetat, S. Delpeux, C. Sinturel, F. Be-
guin, S. Bonnamy, Carbon 2004, 42, 1027.
[105] J. Bai, Carbon 2003, 41, 1325.
[106] C. Park, Z. Ounaies, K. A. Watson, R. E. Crooks, J. Smith, S. E.
Lowther, J. W. Connell, E. J. Siochi, J. S. Harrison, T. L. S. Clair,
Chem. Phys. Lett. 2002, 364, 303.
[107] See, for example, http://www.toyobo.co.jp/e/seihin/kc/pbo/ (accessed
February 2006).
[108] I. M. Ward, J. Sweeney, The Mechanical Properties of Solid Poly-
mers, 2nd ed., Wiley, New York 2004.
[109] A. Garg, S. B. Sinnott, Chem. Phys. Lett. 1998, 295, 273.
[110] L. Liu, A. H. Barber, S. Nuriel, H. D. Wagner, Adv. Funct. Mater.
2005, 15, 975.
[111] C. J. Frizzell, M. I. H. Panhuis, D. H. Countinho, K. J. Balkus, Jr.,
A. I. Minett, W. J. Blau, J. N. Coleman, Phys. Rev. B: Condens. Mat-
ter Mater. Phys. 2005, 72, 245 420.
[112] E. Lahiff, R. Leahy, J. N. Coleman, W. J. Blau, unpublished.
[113] Y.-L. Li, I. A. Kinloch, A. H. Windle, Science 2004, 304, 276.
[114] M. Zhang, K. R. Atkinson, R. H. Baughman, Science 2004, 306,
1358.
[115] M. Zhang, S. Fang, A. A. Zakhidov, S. B. Lee, A. E. Aliev, C. D.
Williams, K. R. Atkinson, R. H. Baughman, Science 2005, 309, 1215.
[116] B. Vigolo, P. Poulin, M. Lucas, P. Launois, P. Bernier, Appl. Phys.
Lett. 2002, 81, 1210.
[117] J. N. Barisci, M. Tahhan, G. G. Wallace, S. Badaire, T. Vaugien,
M. Maugey, P. Poulin, Adv. Funct. Mater. 2004, 14, 133.
[118] M. Shaffer, I. A. Kinloch, Compos. Sci. Technol. 2004, 64, 2281.
[119] C. Velasco-Santos, A. L. Martnez-Hernandez, F. Fischer, R. Ruoff,
V. M. Castano, J. Phys. D. 2003, 36, 1423.
Adv. Mater. 2006, 18, 689706