Materials and Design 69 (2015) 105113

Contents lists available at ScienceDirect

Materials and Design

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

Carbon nanotube network structure induced strain sensitivity and shape

memory behavior changes of thermoplastic polyurethane
Jie Chen, Zhi-xing Zhang, Wen-bin Huang, Jia-le Li, Jing-hui Yang, Yong Wang , Zuo-wan Zhou,
Ji-hong Zhang
Key Laboratory of Advanced Technologies of Materials (Ministry of Education), School of Materials Science & Engineering, Southwest Jiaotong University, Chengdu 610031, China

a r t i c l e

i n f o

Article history:
Received 17 October 2014
Accepted 26 December 2014
Available online 5 January 2015
Keywords:
Thermoplastic polyurethane/carbon
nanotube
Network structure
Mechanical properties
Strain sensitivity
Shape memory behavior

a b s t r a c t
Thermoplastic polyurethane (TPU) composites containing different carbon nanotube (CNT) contents
ranging from 0.5 to 5 wt% were prepared and the main attention was focused on investigating the relation between the network structure of CNTs and the macroscopic physical properties of the composites.
The network structure of CNTs in the TPU/CNT composites was demonstrated through morphological
characterization and rheological measurement. Tensile properties, electrical properties, strain sensitivity
and shape memory properties were comparatively investigated. The results showed that CNTs at content
above 2 wt% formed the percolated network structure in the TPU composites. High content of CNTs
exhibited excellent reinforcement effect. Adding a few CNTs greatly reduced the electrical resistivity of
the composites. The TPU/CNT composites exhibited good strain sensitivity, and the variation of electrical
resistivity during the stretching-restoring process was dependent upon CNT content. At room temperature, CNTs delayed the shape recovery behavior of the composites, especially when the percolated CNT
network structure was formed. However, with the aid of electrical actuation, largely accelerated shape
recovery behavior was observed for the TPU/CNT sample containing 5 wt% CNTs.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Polymer-based composites prepared by adding carbon nanotubes (CNTs) are the hottest issues in the past twenty years. Due
to the high strength and high modulus of CNTs and their large
aspect ratio, CNTs have been proved the most ideal reinforcement
agent for polymers [1,2]. Due to the similar chain-like structure to
polymer molecular chains and the large surface area, CNTs exhibit
excellent nucleation effect on the crystallization of semicrystalline
polymers [3]. Specically, the high electrical and thermal conductivity characteristics of CNTs are very favorable for preparing conductive polymer composites [46]. The interfacial interaction
between CNTs and polymer matrix and the dispersion state of
CNTs are the main factors determining the properties of CNT-based
composites.
When a certain concentration is achieved, CNTs tend to form a
percolated network structure in the matrix, which not only inuences the crystallization behavior of semicrystalline polymer but
also inuences the electrical and thermal conductivity of compos Corresponding author. Tel.: +86 28 87603042.
E-mail addresses: yongwang1976@163.com, yongwang1976@home.swjtu.edu.
cn (Y. Wang).
http://dx.doi.org/10.1016/j.matdes.2014.12.054
0261-3069/ 2015 Elsevier Ltd. All rights reserved.

ites. For example, the percolated CNT network structure reduces

the chain mobility of semicrystalline polymer, which results in
the decrease of spherulite growth rate [7]. However, once the percolated network structure is formed, polymer composites usually
exhibit largely enhanced electrical conductivity, and in this condition, the composites are changed from insulated material to conductive material [4,8,9]. The percolated CNT network structure
can be detected indirectly by measuring the rheological properties
[1013]. Generally, with the increase of CNT content, the increase
of storage modulus (G0 ) is much higher than that of loss modulus
(G00 ), and a plateau is present at low frequencies in the modulus
curves. The transition from liquid-like to solid-like viscoelastic
behavior at low frequency range is mainly related to the signicantly restrained long-range molecular chain motion induced by
the presence of CNT network structure [14].
The electrical response of the conductive polymer composites
under load condition is interesting and it has been one of the main
subjects of recent researches. Through recording the variation of
electrical properties, the deformation or damage in the material
can be diagnosed before the mechanical failure of the material.
Due to the intrinsic coupling of electrical properties and mechanical deformation ability, CNTs have been believed ideal candidates
for future multi-functional materials as sensors and actuators.

106

J. Chen et al. / Materials and Design 69 (2015) 105113

Therefore, much work has been carried out to prepare the strain
sensors with high sensitivity [1520]. Among those factors that
inuence the strain-sensitivity of the conductive composites, the
CNT network structure exhibits very important role and it has been
used to monitor the damage evolution in the material [2123].
Polyurethane (PU) attracts much attention since it has been
synthesized. Due to the excellent physical properties and good biocompatibility, PU has found many applications in the elds ranging
from automotives, rollers systems, and lms to biomedical products [24]. Recently, the application of PU as smart materials has
been a main subject of many researches due to its high tunable
microstructure, high strain recovery, easy processing, light weight
and low cost [25,26]. Different kinds of PU-based shape memory
materials have been developed through adding another polymer
[2730] or nanollers [3133]. Because adding CNTs greatly
improves the electrical conductivity of PU and the composites
can be used to prepare the electrically actuated shape memory
material, the PU/CNT composites with shape memory behavior
have been widely researched [3438]. However, it should be
stressed that in the literatures, most of researches only prove that
the PU/CNT composites have good shape memory behavior in a
certain condition and few work investigates the relation between
CNT microstructure and shape memory behavior of the composites. For example, in the work of Gu et al. [37], only 1 wt% CNTs
were used to prepare PU/CNT composites. Jung et al. [39] investigated the electrical actuated shape memory behavior of different
PU/CNT composites containing 4 wt% CNTs. Raja et al. [40] investigated the effect of CNT content on the mechanical properties, thermal and electrical conductivity of the PU/CNT composites.
However, when they analyzed the variation of shape memory
property of composites, only the sample containing 5 wt% CNTs
was investigated. Although Mahapatra et al. [38] investigated the
content of CNTs on the shape memory behavior of the PU/CNT
composites and they found that more than 98% shape recovery
and a rapid recovery time of 9 s in both thermal triggering and
electrical actuating shape memory behavior could be achieved, it
is worth noting that in their work, the thermal triggering shape
memory behavior was investigated at ambient temperature of
50 C, which was much higher than the glass transition temperature of PU. Similarly, enhanced shape recovery ability at 60 C
has been reported by Deka et al. [35]. Obviously, the effect of
CNT network structure on the shape memory behavior of PU/CNT
is still not clear. The PU/CNT shape memory materials have potential applications ranging from frozen food packaging to medical
apparatus and instruments.
In this work, we introduced different contents of CNTs into
thermoplastic polyurethane (TPU) and investigated the formation
of CNT network structure in the composites and the corresponding
variations of mechanical properties, strain sensitivity and shape
memory behaviors of the TPU/CNT composites. To the best of our
knowledge, this is the rst time to found direct relationship
between CNT network structure and the shape memory behavior
of TPU/CNT composites. We believe that through comparatively
investigating the changes of the macroscopic physical properties,
this work might provide a guide map for the design and preparation of the TPU/CNT smart materials.

CNTs (tradename of Flo TubesTM 9000) were purchased from

CNano Technology Co., Ltd. (Beijing, China). The purity of CNTs
was more than 95%, and the average diameter and length of a single CNT were ca. 11 nm and 10 lm, respectively.
2.2. Sample preparation
TPU was dried in an oven at 60 C for 10 h before it was used.
The melt compounding of TPU/CNT was conducted on an internal
mixer ZJL-300 (Changchun Zhineng Instrument, China) at a melt
temperature of 190 C. The screw speed was set at 100 rpm and
the mixing duration was 10 min. In this work, the content of CNTs
was varied from 0.5 wt% to 5 wt%. The sample notation was
dened as TPUxC, where x represented the content of CNTs in
the composites. After that, the composites were compressionmolded at a melt temperature of 190 C and a pressure of 5 MPa
to obtain the plates with a thickness of 1 mm.
2.3. Scanning electron microscopy (SEM)
The dispersion of CNTs in the TPU/CNT composites was investigated using a scanning electron microscope (SEM) Fei Inspect (FEI,
the Netherlands) with an accelerating voltage of 20 kV. The sample
was rst cryogenically fractured in liquid nitrogen. To characterize
the dispersion of CNTs in the TPU matrix, the cryo-fractured surface of the representative TPU2C sample was coated with a thin
layer of gold before crystallization. To characterize the network
structure of CNTs, all the cryo-fractured surfaces were also characterized without coating gold.
2.4. Rheological measurement
The rheological measurement was carried out using a sample
disk, which was cut from the compression-molded plate. The
diameter of the sample disk was 20 mm. The measurements were
conducted on a stress-controlled rheometer DHR-1 (TA Instrument, USA). During the measurement, the sample temperature
was set at 190 C and the measuring frequency was varied from
0.01 to 100 Hz. For all the measurements, the samples were measured within the linear viscoelastic strain range, which could be
estimated by an initial survey through a dynamic strain sweep
experiment at strains ranging from 0.01% to 100%. In addition, it
was stressed that the measurement was carried out in nitrogen
atmosphere.
2.5. Mechanical properties
Tensile properties of the TPU/CNT composite were measured on
a dumb-bell specimen that was cut from the previously compression-molded plate. The width and the thickness of the specimen
were 5.0 and 1 mm, respectively. The measurement was conducted
on a universal tensile machine AGS-J (SHIMADZU, Japan) according
to ASTM: D638 at room temperature (23 C). A cross-head speed of
100 mm/min was used.
2.6. Dynamic mechanical analysis (DMA)

2. Experimental details
2.1. Materials
Polyester-based TPU (tradename of WHT-1570) with a density
of 1.19 g/cm3 and a glass transition temperature of about 40 C
was purchased from Wanhua Polyurethane Co., Ltd. (Yantai,
China). It should be stressed that TPU was completely amorphous.

The dynamic mechanical analysis (DMA) measurement was carried out using a DMA Q800 analyzer (TA Instrument, USA). A rectangular cross-sectional bar (with a length of 25 mm and a width of
5 mm) was cut from the previously compression-molded plate.
During the DMA measurement, a tensile mode was selected. The
temperature was heated from 80 to 10 C at a heating rate of
3 C/min and the frequency was set at 1 Hz.

107

J. Chen et al. / Materials and Design 69 (2015) 105113

2.7. Electrical resistivity

the electrical resistivity measurement of sample through twopoint technique.

The electrical resistivity was measured using a universal meter

(DT9208, China). A rectangular cross-sectional bar (with a length of
30 mm and a width of 6 mm) was cut from the previously compression-molded plate. The two ends of the bar were coated with
a thin layer of conductive silver paint to erase the inuence of contact resistance, then the electrical resistivity was measured by a
two-point technique. The resistance change under the different
load conditions was also measured. Before stretching, the initial
electrical resistivity (R0) of sample was rst measured. Then, the
sample was stretched at a tensile speed of 100 mm/min. To understand the potential application of TPU/CNT composites as sensors
in monitoring the small deformation and/or damage of material
articles during the service period, a maximum stain during the
stretching process was xed at 20%. Once the strain of 20% was
achieved, the stretching was stopped and the electrical resistivity
(R) was measured again. After that, the load was released. When
the sample was recovered to the initial state, the electrical resistivity (R00 ) was also measured. Subsequently, the sample was
stretched again. The same operation was repeated for 15 cycles.
The schematic representations shown in Fig. 1a and b describe

Gauge

2.8. Shape memory property measurement

The shape memory property of sample was evaluated by
observing the evolution of deformed sample appearance with the
increase of time. For the common shape recovery without any
external stimulation, the deformed sample that was prepared at
room temperature was rst immersed into liquid nitrogen for
1 min, and then it was taken out and placed on a horizontal substrate. The shape recovery of the sample with increasing of time
was recorded by taken images. For the electrically actuated shape
recovery, a breeches-like sample was cut from the previously compression-molded plate. The two ends of the sample were then
coated with a thin layer of conductive silver paint to erase the
inuence of contact resistance. During the shape recovery process,
a constant voltage of 50 V was applied and the shape recovery was
then recorded. The schematic representation shown in Fig. 1c
describes the appearance of the breeches-like specimen that is
designed for measuring the electrically actuated shape memory
behavior. Namely, the specimen had the thickness of 1 mm, the
maximum width of 25 mm, and the maximum length of 70 mm.
The width and length for each leg were 10 and 25 mm, respectively. Furthermore, the length of part that was coated by conductive silver paint was 5 mm.
3. Results and discussion
3.1. Formation of CNTs network structure
First, the dispersion state of CNTs in the representative TPU2C
composite is shown in Fig. 2. In Fig. 2, the white dot represents
CNTs. It can be seen that CNTs exhibit relatively homogeneous dispersion in the TPU matrix, and there is no serious aggregation phenomenon. Furthermore, it can be deduced that there is strong
interfacial adhesion between CNTs and TPU because only the ends
of CNTs are exposed during the fracture process. To further investigate the contact between CNTs and the formation of CNT network
structure in TPU matrix, the cryo-fractured surfaces of different
samples were also characterized using SEM without coating gold
on the cryo-fractured surfaces. In this condition, only the conductive CNTs can be observed from the SEM image. From Fig. 3 one can
see that at CNT content of 0.5 wt%, some CNTs aggregate together

(a)
Specimen

Initial state

With strain of 20%

(b)
CNTs
25 mm

5 mm 10 mm

25 mm

5 mm
70 mm

(c)
Fig. 1. Schematic representation showing the measurement of electrical resistivity
under different conditions: (a) electrical resistivity measurement through twopoint technique and (b) electrical resistivity measurement of sample before and
after being stretched to a certain strain (20%). (c) Schematic representation showing
the sample appearance prepared for electrically actuated shape recovery
measurement.

Fig. 2. SEM image showing the dispersion of CNTs in the representative TPU2C
composite. Sample was cryogenically fractured in liquid nitrogen and the fractured
surface was coated with a thin layer of gold before SEM characterization.

108

J. Chen et al. / Materials and Design 69 (2015) 105113

(b)

(a)

(c)

Fig. 3. SEM images showing the dispersion of CNTs in the TPU/CNT composites with different CNT contents. Samples were cryogenically fractured in liquid nitrogen and the
fractured surfaces were directly characterized without coating gold. (a) 0.5 wt%, (b) 2 wt% and (c) 5 wt%.

at low frequency region, a noticeable change in G0 is observed for

the TPU/CNT composites. Addition of only 0.5 wt% CNTs induces
the apparent increase of G0 . The more the CNT content in the composites is, the more apparent the increase of G0 is. Specically, a
signicant plateau of the ow regime is observed at low frequency
range in which G0 is nearly invariant with frequency, indicating the
transition from liquid-like to solid-like viscoelastic behavior. Xu
et al. [14] suggested that the transition was mainly attributed to
the signicantly restrained long-range molecular chain motion,
which is induced by the presence of particle network structure.
This clearly indicates the formation of a three-dimensional network structure of CNTs. From Fig. 4b one can see that G00 increases
with increasing content of CNTs. Specically, a signicant plateau
of G00 is also observed when the content of CNTs is increased up
to 2 wt%. Furthermore, it is worth noting that G0 is apparently larger than G00 when the content of CNTs is above 2 wt%. This also conrms the existence of a dense CNT network structure.
The variation of g versus the content of CNTs is shown in
Fig. 4c. One can see that pure TPU exhibits nearly invariant g at

and form a conductive network-like structure. When CNT content

is increased up to 2 wt%, CNTs homogeneously disperse in the
whole matrix, and the conductive network structure becomes
more apparent. At CNT content of 5 wt%, the dispersion of CNTs
becomes denser and most of CNT contact each other, forming a
denser CNT network structure in the whole TPU matrix.
The percolated CNT network structure in the TPU/CNT composites were further investigated through rheological measurement.
Fig. 4 shows the variations of storage modulus (G0 ), loss modulus
(G00 ) and complex viscosity (g) of the TPU/CNT composites with
the increase of CNT content in the composites. From Fig. 4a one
can see that pure TPU exhibits a transition in G0 curve (shown by
arrow). As we know, TPU is a typical multi-block copolymer that
consists of alternating soft polyester and hard polyurethane-urea
segments. At an appropriate temperature, these segments experience micro-phase separation, forming hard and soft phases. As a
result, the elastic response of the melt under the shear condition
becomes more apparent, and a deviation in the terminal zone of
G0 curve is detected [41,42]. Compared with the pure TPU sample,

(a) 10

(b) 10

105

Loss modulus (G'',Pa)

Storage modulus (G',Pa)

104
3

10

102

TPU
TPU0.5C
TPU1C
TPU2C
TPU5C

101
100

0.01

0.1

10

105
104
103
102
101

100

TPU
TPU0.5C
TPU1C
TPU2C
TPU5C
0.01

0.1

Frequency (Hz)

10

100

Frequency (Hz)

Complex viscosity (
*,Pas)

(c)
105

TPU
TPU0.5C
TPU1C
TPU2C
TPU5C

104

103

102

0.01

0.1

10

100

Frequency (Hz)
Fig. 4. Rheological properties of the TPU/CNT composites. (a) Storage modulus, (b) loss modulus, and (c) complex viscosity.

J. Chen et al. / Materials and Design 69 (2015) 105113

low frequencies, indicating that the TPU melt has the feature of
Newtonian uid measured in this condition. However, dramatically increased g is observed for the TPU/CNT composites, and
all the composites exhibit the shear thinning behavior in all frequencies. Furthermore, it is worth noting that at relatively high
shear frequency, the difference in melt viscosity becomes smaller.
Obviously, the presence of CNTs greatly increases the melt viscosity, but the negative effect of increased viscosity on processing
becomes inconspicuous under the condition of high shear frequency. This is very signicant from a viewpoint of maintaining
the good process ability of the TPU/CNT composites.
3.2. Mechanical properties
Fig. 5 exhibits the representative engineering stressstrain
curves of neat TPU and the TPU/CNT composites. It can be seen that
neat TPU shows the typical tensile behavior of elastomer, namely,
stress increases gradually with the increase of strain. Since the
elongation at break of neat TPU is very large and it exceeds the
maximum testing range of the instrument, the sample cannot be
fractured under this condition. For the TPU0.5C and TPU1C samples, the stressstrain curves are very close to that of neat TPU. This
indicates that CNTs at low content exhibits inconspicuous reinforcement effect in TPU matrix. However, further increasing the
content of CNTs induces the apparent changes of the tensile behaviors of the composites. For the TPU2C and TPU5C sample, largely
increased tensile stress is achieved, especially at relatively large
strain. According to previous rheological measurement one can
see that when CNT content is above 2 wt%, CNTs form the percolated network structure in the TPU matrix. This indicates that the
reinforcement effect of CNTs in TPU is mainly related to the formation of the percolated CNT network structure. Furthermore, it is
interesting to observe that both TPU2C and TPU5C samples are
not fractured during the measurement. In other words, the percolated CNTs network structure does not inuence the tensile ductility of TPU apparently.
The dynamic mechanical properties were measured using DMA.
The results of representative samples are shown in Fig. 6. The corresponding parameters that reect the relaxation behaviors of
samples are shown in Table 1. Although the previous uniaxial tensile measurement show that CNTs at relatively low content
(61 wt%) do not exhibit apparent reinforcement effect, the storage
modulus shown in Fig. 6a clearly conrms the reinforcement effect
of CNTs, especially at temperature below glass transition temperature (Tg). The more the CNTs in the composites are, the more apparent the reinforcement effect is. From Fig. 6b and Table 1 one can
see that the presence of CNTs does not induce the apparent change

20

TPU
TPU0.5C
TPU1C
TPU2C
TPU5C

Stress (MPa)

15

10

0
0

100

200

300

400

500

Strain (%)
Fig. 5. The stressstrain curves of pure TPU and the TPU/CNT specimens obtained at
the tensile speed of 100 mm/min.

109

of Tg of TPU. However, both the intensity (tan dmax) and the area (IR)
of loss factor, which are direct indicators of the damping behavior
of a given material and represent the energy dissipation during the
measurement, are reduced at relatively high CNT content. For
example, the value of IR decreases gradually from 7.7 of neat TPU
sample to 7.6, 7.2 and 6.3 of TPU/CNT samples with 0.5 wt%,
2 wt% and 5 wt% CNTs, respectively. This is possibly attributed to
the formation of CNT network structure that prevents the viscous
ow of TPU molecular chains. In other words, less energy is dissipated and more energy can be stored for the TPU/CNT composites.
Obviously, the reduced energy dissipation during the deformation
process is very favorable for enhancing the shape recoverability of
the sample.
3.3. Electrical properties
Fig. 7 shows the variation of the volume electrical resistivity of
the TPU/CNT composites versus the content of CNTs. For neat TPU
sample, the volume electrical resistivity is bigger than 1011 X m.
Addition of only 0.5 wt% CNTs results in the great decrease of electrical resistivity and the value is decreased to 104 X m, indicating
the formation of a conductive network structure of CNTs. Further
increasing the content of CNTs induces the dramatic decrease of
electrical resistivity. It is worth noting that in the TPU2C and
TPU5C samples, in which CNTs form the rheological network structure and the samples still maintain excellent rubbery elasticity, the
electrical resistivity is decreased to 101 and 101 X m, respectively.
This indicates that the electrical conductivity of sample is dramatically improved.
To better understand the change of the electrical properties, the
conductivity of the TPU/CNT composites was rationalized in terms
of modied classical percolation scaling law [43]:

r r0 p  pc t

where r represents the conductivity of the composites, p is the ller

content and pc is the percolation threshold, r0 is a scaling factor and
t is a percolation exponent which generally reects the dimensionality of the system with values around 1.3 and 2 for two and threedimensions, respectively. The inserted graph shown in Fig. 7 represents a best t to experimental conductive data as a function of
p  pc. Consequently, a percolation threshold pc of 0.39 wt% and a
scaling exponent t of 2.76 are achieved. In other words, addition
of only 0.39 wt% CNTs, a conductive CNT network structure can
be formed in the TPU/CNT composites.
3.4. Strain sensitivity
The strain sensitivity of the TPU/CNT composites were investigated by measuring the volume electrical resistivity of sample
before and after being stretched to a certain strain of 20%. Comprehensively considering the variations of rubbery elasticity and electrical resistivity, the representative TPU2C and TPU5C samples
were investigated. The variation of electrical resistivity during
the repeated stretching process and the variation of (R  R0)/R0
with stretching cycles are shown in Fig. 8. From Fig. 8a one can
see that both TPU2C and TPU5C sample show the similar variation
trends of electrical resistivity during the repeated stretching process. Before stretching, the samples exhibit smallest electrical
resistivity. After being stretched, the electrical resistivity increases
greatly possibly due to that stretching destroys the initial conductive CNT network structure. After being stretched for several
cycles, i.e. 8 cycles for the TPU2C sample and 3 cycles for the TPU5C
sample, the variations of R and R00 become inconspicuous with further increasing stretching cycles. In this condition, the two samples
exhibit highly reversible currentvoltage characteristics when the

110

J. Chen et al. / Materials and Design 69 (2015) 105113

(b)

TPU
TPU0.5C
TPU2C
TPU5C

5000

Tg

TPU
TPU0.5C
TPU2C
TPU5C

4000

Tan

Storage Modulus (MPa)

(a) 6000

3000
2000
1000
0
-70

-60

-50

-40

-30

-20

-10

-70

-60

-50

-40

-30

-20

-10

Temperature ( C)

Temperature ( C)

Fig. 6. Dynamic mechanical properties of pure TPU and the representative TPU/CNT samples. (a) Storage modulus and (b) loss factor.

Table 1
Glass transition parameters of pure TPU and representative TPU/CNT composites
obtained via DMA measurement.
Sample

Tg (C)

tan dmax

IR

TPU
TPU0.5C
TPU2C
TPU5C

31.7
31.6
31.0
30.3

0.6
0.6
0.55
0.51

7.7
7.6
7.2
6.3

4
11

10
10

10

10

10

10

10

-1

10

-3

2
0

TPUxC
t =2.76
pc=0.39 wt%

-2
-4
-6
-8
-10

-2

-1

Content of CNTs (wt %)

Fig. 7. Variation of electrical resistivity of the TPU/CNT composites versus the
content of CNTs.

(b) 30

TPU2C-15cycles
TPU5C-15cycles

25

( R-R 0 ) /R 0

(a) 10

stress is applied and released. The similar variation of electrical

resistivity during stretching process has been already reported
elsewhere [44]. This indicates that the rearrangement of the conductive CNT network structure in the sample is nearly accomplished. Because the conductive CNT network structure in the
TPU5C sample is denser than that in the TPU2C sample, fewer
stretching cycles is required to accomplish the rearrangement of
CNT network structure. Furthermore, it is observed that the
stretched samples with strain of 20% exhibit smaller electrical
resistivity possibly due to that stretching promotes the deformation and orientation of conductive CNT network structure along
the tensile direction and increases the probability for the contact
of adjacent CNT network structure. From Fig. 8b it can be seen that
(R  R0)/R0 gradually increases with increasing stretching cycles.
Interestingly, it should be noticed that the TPU5C sample exhibits
smaller variations compared with the TPU2C sample. This also
conrms that the denser CNT network structure has greater ability to stabilize the electrical properties of the TPU/CNT composites under the load condition. However, if the TPU/CNT
composites are selected to prepare sensor that is used to monitor
the small deformation and/or damage of material articles, obviously, more apparent strain sensitivity must be required. In this
condition, relatively small content of CNTs, which endows the
TPU/CNT composites with loose network structure but high strain
sensitivity, is more appropriate. The similar results have also
been reported by Nanni et al. [45] In their work, aligned CNTs
were introduced into poly(ethylene terephthalate) (PET) and the
self-monitoring ability of the composites was investigated. They
found that too high CNT content decreased sensitivity at low

10

15
10
5

TPU2C

0
10
8

( R-R 0 ) /R 0

10

10

-1

10

20

0% 20%

0% 20%

Strain

0% 20%

TPU5C

6
4
2
0

10 11 12 13 14 15

Cycles

Fig. 8. (a) Variation of electrical resistivity of the representative TPU/CNT samples during the repeated stretching-restoring tests. The maximum tensile strain is 20%. (b)
Variation of (R  R0)/R0 versus the stretching-restoring cycles.

111

J. Chen et al. / Materials and Design 69 (2015) 105113

strain, whereas a minimum CNT content was required to insure

ohmic conductivity.
3.5. Shape memory behaviors
The shape memory behaviors of neat TPU and the representative TPU/CNT composites were qualitatively analyzed through
observing the shape recovery process of deformed samples. First,
the unconstrained bend recoverability of sample without electrical
actuation is shown in Fig. 9. The rectangular samples were
deformed into U shaped appearance at room temperature, and
then they were immersed into liquid nitrogen for 60 s to x the
appearance. It can be seen that for the deformed neat TPU sample,
once it is taken out from the liquid nitrogen, it experiences a fast
shape recovery process. For example, at time of 20 s, the shape
recovery is about 30% and after 60 s, about 75% structure is recovered. However, for the TPU/CNT composites, the rate of the shape
recovery is found to be dependent upon CNT content. The TPU0.5C
sample shows the recovery of 9.4% at time of 20 s, which is much
smaller than that of neat TPU sample. Further decreased recovery
of 5.0% and 2.8% are observed for the TPU2C and TPU5C samples,
respectively. This indicates that the shape recovery of composites
is greatly delayed. It is believed that the formation of the percolated CNT network structure, which restricts the chain motion of
TPU matrix, is the main reason for the delayed shape recovery
behavior. However, after being recovered for 60 s, the differences
in shape recovery among samples become smaller. For example,
the TPU2C and TPU5C samples show the recovery of about 71.6%
and 71.1%, respectively, which are slightly smaller than that of neat
TPU (75%). In other words, although the CNT network structure
inuences the initial rate of recovery, it does not inuence the nal
shape recovery of the composites apparently. It is well known that
the main factors that determine the shape memory behavior of
material are the presence of quantitative amount of unlocked oriented molecular chains and the value of room temperature modulus [35,46]. The former one is usually related to the crystallization
of soft segments during the deformation process. In this work, the
TPU matrix is completely amorphous and no crystallization occurs
during the deformation process. Furthermore, from previous DMA
measurement one can see that although the TPU/CNT composites

exhibit much higher modulus at low temperature (<Tg) compared

with neat TPU, they exhibit very similar room temperature modulus to that of neat TPU. Therefore, it is believed that the variation of
the shape memory behavior is mainly related to the formation of
CNT network structure.
The electrically actuated shape memory behaviors of different
samples are shown in Fig. 10. Here, the breeches-like sample was
also deformed into U shaped appearance and immersed into
liquid nitrogen for 1 min to x the appearance. When it was taken
out from liquid nitrogen, a constant voltage of 50 V was applied to
the sample and the evolution of shape recovery with time was
taken images. For neat TPU sample, after being recovered for
20 s, the recovery is about 4.4%. At time of 60 s, about 68.9% structure is recovered. According to the previous electrical property
measurement one can see that the sample is completely insulated.
Therefore, the role of voltage on shape recovery behavior of neat
TPU can be ignored. Interestingly, it is observed that the electrically
actuated shape recovery of the TPU/CNT composites is also dependent upon the CNT content, especially different variation trends
can be observed. For the TPU0.5C and TPU2C sample, the delayed
shape recovery behaviors are observed. For example, at time of
20 s, the shape recovery is nearly not actuated. Even if the time
is increased up to 60 s, the shape recovery of the TPU0.5C and
TPU2C sample is 60% and 40.6%, respectively. Specically, it is
found that the rate of the shape recovery of TPU2C is smaller than
that of TPU0.5C sample. This conrms that the formation of CNT
network structure delays the shape recovery of the composites.
However, largely accelerated shape recovery is observed for the
TPU5C sample. At 20 s, the recovery is increased up to 27.8%. When
the time is increased to 60 s, about 87.8% structure is recovered.
The electrically actuated shape recovery behavior is mainly
related to the temperature change in the sample when a certain
voltage is applied. Namely, a higher electrical conductivity is the
prerequisite for achieving higher efciency as a shape memory
material [47]. Mahapatra et al. [38] measured the variation in the
surface temperature of the composites and found that for the
hyperbranched polyurethane (HPU)/CNT composites with fast
shape recovery behavior, the surface temperature enhanced in a
very short time when a certain voltage was applied. In this work,
compared with neat TPU sample, the TPU2C sample exhibits

0%

0%

30%

0%

0%

9.4%

0%

0%

5%

51.7%

71.6%

0%

2.8%

50%

71.1%

0%

61.1%

54.4%

75%

72.2%

Fig. 9. Images demonstrating the unconstrained bend recoverability of pure TPU (A1A4), TPU0.5C (B1B4), TPU2C (C1C4) and TPU5C (D1D4) composites with increasing
of time.

112

J. Chen et al. / Materials and Design 69 (2015) 105113

0s

20 s

40 s

60 s

0%

4.4%

42.2%

68.9%

0%

0%

29.4%

60%

0%

0%

22.2%

40.6%

0%

27.8%

75.6%

87.8%

Fig. 10. Images demonstrating the unconstrained bend recoverability of pure TPU (A1A5), TPU0.5C (B1B5), TPU2C (C1C5) and TPU5C (D1D5) composites with increasing
of time. A constant voltage of 50 V was supplied during the measurement.

delayed shape recovery behavior while the TPU5C sample exhibits

largely accelerated shape recovery behavior. The former sample
has electrical resistivity 10 times higher than that of the latter
sample; therefore, at a certain voltage of 50 V, higher temperature
is provoked in the latter sample, which facilitates the shape recovery. On the other hand, from DMA measurements one can see that
the TPU5C sample exhibits less energy dissipation during the
deformation process compared with the TPU2C sample. The
reduced energy dissipation during deformation process also facilitates the enhancement of the shape recoverability when stress is
released.
Furthermore, it is worth noting that although conductive CNT
network structure can be formed at 0.5 wt% CNTs, the relatively
high electrical resistivity induces very small current at a certain
voltage of 50 V. Therefore, the heat induced by current can be
ignored. In this condition, the presence of CNTs prevents the chain
motion of TPU matrix, which delays the shape recovery of sample.
At 2 wt% CNTs, at which a rheological CNT network structure is
formed, the current in the sample is still small. However, the
restriction effect of CNT network structure on the chain motion
of TPU matrix becomes more apparent. Consequently, largely
delayed shape recovery behavior is observed for the TPU2C sample.
At 5 wt% CNTs, the positive role of temperature increase in the
sample exceeds the negative role of restricted chain motion on
the shape recovery of sample. Consequently, largely accelerated
shape recovery is achieved for the TPU5C sample. Specically,
the sample still maintains the excellent rubbery elasticity.
The electrically actuated shape recovery behaviors of the TPU/
CNT composites have been widely investigated in the literatures
[28,3439]. However, most researches only investigated the shape
recovery behavior of the composite with a certain composition. For
example, Raja et al. only reported that the modied CNTs at 10 wt%
[28] or 5 wt% [39] endowed TPU composites with remarkably
increased recoverability of their shape at lower applied voltages,
when compared to pure TPU or pristine CNTs loaded system.
Although Sahoo et al. [34] reported that a good shape recovery in
the 9095% range could be obtained for the TPU composite containing 2.5 wt% CNTs, it should be pointed out that they also introduced 2.5 wt% polypyrrole (PPY) into the composites, which
improved the electrical conductivity of the composites. Deka
et al. [35] reported that the shape recovery of hyperbranched

polyurethane (HBPU) composites increased with increasing CNT

content from 1 to 5 wt%, and a maximum shape recovery was
achieved at CNT content of 5 wt%. However, the addition of CNTs
also induced the crystallization of HBPU and the crystallinity
increased with increasing CNT content. This was also one of the
main factors that induced the improvement of the shape recovery
of the composites. In our work, because TPU matrix is amorphous,
the variation of the shape recovery behavior of the TPU/CNT composites is mainly related to the dispersion and the morphology of
the CNTs in the composites. Specically, the inuence of CNT network structure on the shape recovery behavior of TPU/CNT composites has been rst claried. This is very signicant in
developing new smart TPU materials.

4. Conclusions
In this work, the TPU/CNT composites with different CNT contents have been prepared. The rheological measurement shows
that CNTs form the percolated network structure in the TPU matrix
at CNT content above 2 wt%. Tensile properties measurement
shows that the percolated CNT network structure exhibits apparent reinforcement effect for TPU matrix and the composites still
exhibit excellent rubbery elasticity at CNT content of 2 and
5 wt%. Adding CNTs greatly reduces the electrical resistivity of
the sample and a percolation threshold of 0.39 wt% is achieved.
The TPU/CNT composites exhibit excellent strain sensitivity and
the variation of electrical resistivity of the sample during stretching-restoring process is dependent upon CNT content. The results
obtained from shape recovery behavior observation show that for
the shape recovery behavior without electrical actuation, the presence of CNTs delays the shape recovery of TPU, especially when the
dense CNT network structure is present in the composites. For the
electrically actuated samples, the shape recovery behavior is found
to be dependent upon CNT content. At relatively low content
(62 wt%), the composites still show the delayed shape recovery
behavior; However, at relatively high content (5 wt%), largely
accelerated shape recovery behavior is observed. This work demonstrates that the network structure of CNTs exhibits a dual role
in inuencing the mechanical properties, strain sensitivity and
shape memory behaviors of the composites. Through tuning the

J. Chen et al. / Materials and Design 69 (2015) 105113

density of the CNT network structure, the TPU/CNT composites

exhibit different potential applications.
Acknowledgements
Authors express their sincere thanks to the National Natural
Science Foundation of China (50973090), Distinguished Young
Scholars Foundation of Sichuan (2012JQ0057) and the Fundamental Research Funds for the Central Universities (SWJTU12CX010,
SWJTU11CX142, SWJTU11ZT10) for nancial support.
References
[1] Coleman JN, Khan U, Gunko YK. Mechanical reinforcement of polymers using
carbon nanotubes. Adv Mater 2006;18:118.
[2] Bose S, Khare RA, Moldenaers P. Assessing the strengths and weaknesses of
various types of pre-treatments of carbon nanotubes on the properties of
polymer/carbon nanotubes composites: a critical review. Polymer
2010;51:97593.
[3] Xu JZ, Zhong GJ, Hsiao BS, Fu Q, Li ZM. Low-dimensional carbonaceous
nanoller induced polymer crystallization. Prog Polym Sci 2014;39:55593.
[4] Spitalsky Z, Tasis D, Papagelis K, Galiotis C. Carbon nanotubepolymer
composites: chemistry, processing, mechanical and electrical properties. Prog
Polym Sci 2010;35:357401.
[5] Han ZD, Fina A. Thermal conductivity of carbon nanotubes and their polymer
nanocomposites: a review. Prog Polym Sci 2011;36:91444.
[6] Xie N, Jiao QJ, Zang CG, Wang CL, Liu YY. Study on dispersion and electrical
property of multi-walled carbon nanotubes/low-density polyethylene
nanocomposites. Mater Des 2010;31:167683.
[7] Xu DH, Wang ZG. Role of multi-wall carbon nanotube network in composites
to crystallization of isotactic polypropylene matrix. Polymer 2008;49:3308.
[8] Tiusanen J, Vlasveld D, Vuorinen J. Review on the effects of injection moulding
parameters on the electrical resistivity of carbon nanotube lled polymer
parts. Compos Sci Technol 2012;72:174152.
[9] Takeda T, Shindo Y, Kuronuma Y, Narita F. Modeling and characterization of
the electrical conductivity of carbon nanotube-based polymer composites.
Polymer 2011;52:38526.
[10] Ptschke P, Fornes TD, Paul DR. Rheological behavior of multiwalled carbon
nanotube/polycarbonate composites. Polymer 2002;43:324755.
[11] Sung YT, Han MS, Song KH, Jung JW, Lee HS, Kum CK, et al. Rheological
behavior of multiwalled carbon nanotube/polycarbonate composites. Polymer
2006;47:44349.
[12] McClory C, McNally T, Baxendale M, Ptschke P, Blau W, Ruether M. Electrical
and rheological percolation of PMMA/MWCNT nanocomposites as a function
of CNT geometry and functionality. Eur Polym J 2010;46:85468.
[13] Kim M, Mun SC, Lee CS, Lee MH, Son Y, Park OO. Electrical and rheological
properties of polyamide 6,6/c-ray irradiated multi-walled carbon nanotube
composites. Carbon 2011;49:402430.
[14] Xu ZH, Niu YH, Yang L, Xie WY, Li H, Gan ZH, et al. Morphology, rheology and
crystallization behavior of polylactide composites prepared through addition
of ve-armed star polylactide grafted multiwalled carbon nanotubes. Polymer
2010;51:7307.
[15] Li C, Thostenson ET, Chou TW. Sensors and actuators based on carbon
nanotubes and their composites: a review. Compos Sci Technol
2008;68:122749.
[16] Hu N, Karube Y, Yan C, Masuda Z, Fukunaga H. Tunneling effect in a polymer/
carbon nanotube nanocomposite strain sensor. Acta Mater 2008;56:292936.
[17] Hu N, Karube Y, Arai M, Watanabe T, Yan C, Li Y, et al. Investigation on
sensitivity of a polymer/carbon nanotube composite strain sensor. Carbon
2010;48:6807.
[18] Ji M, Deng H, Yan D, Li X, Duan L, Fu Q. Selective localization of multi-walled
carbon nanotubes in thermoplastic elastomer blends: an effective method for
tunable resistivitystrain sensing behavior. Compos Sci Technol
2014;92:1626.
[19] Lin L, Liu S, Zhang Q, Li X, Ji M, Deng H, et al. Towards tunable sensitivity of
electrical property to strain for conductive polymer composites based on
thermoplastic elastomer. ACS Appl Mater Interfaces 2013;5:581524.
[20] Witt N, Tang YH, Ye L, Feng LM. Silicone rubber nanocomposites containing a
small amount of hybrid llers with enhanced electrical sensitivity. Mater Des
2013;45:54854.
[21] Li C, Chou TW. Selective localization of multi-walled carbon nanotubes in
thermoplastic elastomer blends: an effective method for tunable resistivity
strain sensing behavior. Compos Sci Technol 2008;68:33739.

113

[22] Gao L, Chou TW, Thostenson ET, Zhang Z, Coulaud M. In situ sensing of impact
damage in epoxy/glass ber composites using percolating carbon nanotube
networks. Carbon 2011;49:33825.
[23] Pandey G, Wolters M, Thostenson ET, Heider D. Localized functionally
modied glass bers with carbon nanotube networks for crack sensing in
composites using time domain reectometry. Carbon 2012;50:381625.
[24] Petrovic ZS, Ferguson J. Polyurethane elastomers. Prog Polym Sci
1991;16:695836.
[25] Kim BK, Lee SY, Xu M. Polyurethanes having shape memory effects. Polymer
1996;37:578193.
[26] Huang WM, Yang B, Zhao Y, Ding Z. Thermo-moisture responsive polyurethane
shape-memory polymer and composites: a review. J Mater Chem
2010;20:336781.
[27] Wang WS, Ping P, Chen XS, Jing XB. Polylactide-based polyurethane and its
shape-memory behavior. Eur Polym J 2006;42:12409.
[28] Raja M, Ryu SH, Shanmugharaj AM. Thermal, mechanical and electroactive
shape memory properties of polyurethane (PU)/poly (lactic acid) (PLA)/CNT
nanocomposites. Eur Polym J 2013;49:3492500.
[29] Jeong HM, Ahn BK, Kim BK. Miscibility and shape memory effect of
thermoplastic polyurethane blends with phenoxy resin. Eur Polym J
2001;37:224552.
[30] Jeong HM, Song JH, Lee SY, Kim BK. Miscibility and shape memory property of
poly(vinyl chloride)/thermoplastic polyurethane blends. J Mater Sci
2001;36:545763.
[31] Razzaq MY, Anhalt M, Frormann L, Weidenfeller B. Thermal, electrical and
magnetic studies of magnetite lled polyurethane shape memory polymers.
Mater Sci Eng A 2007;444:22735.
[32] Kim JT, Kim BK, Kim EY, Park HC, Jeong HM. Synthesis and shape memory
performance of polyurethane/graphene nanocomposites. React Funct Polym
2014;74:1621.
[33] Saralegi A, Gonzalez ML, Valea A, Eceiza A, Corcuera MA. The role of cellulose
nanocrystals in the improvement of the shape-memory properties of castor
oil-based segmented thermoplastic polyurethanes. Compos Sci Technol
2014;92:2733.
[34] Sahoo NG, Jung YC, Yoo HJ, Cho JW. Inuence of carbon nanotubes and
polypyrrole on the thermal, mechanical and electroactive shape-memory
properties of polyurethane nanocomposites. Compos Sci Technol
2007;67:19209.
[35] Deka H, Karak N, Kalita RD, Buragohain AK. Biocompatible hyperbranched
polyurethane/multi-walled carbon nanotube composites as shape memory
materials. Carbon 2010;48:201322.
[36] Zhao JC, Du FP, Zhou XP, Cui W, Wang XM, Zhu H, et al. Thermal conductive
and electrical properties of polyurethane/hyperbranched poly(urea-urethane)grafted
multi-walled
carbon
nanotube
composites.
Compos
B
2011;42:21116.
[37] Gu S, Yan B, Liu L, Ren J. Carbon nanotubepolyurethane shape memory
nanocomposites with low trigger temperature. Eur Polym J 2013;49:386777.
[38] Mahapatra SS, Yadav SK, Yoo HJ, Ramasamy MS, Cho JW. Tailored and strong
electro-responsive shape memory actuation in carbon nanotube-reinforced
hyperbranched polyurethane composites. Sensor Actuat B Chem
2014;193:38490.
[39] Jung YC, Yoo HJ, Kim YA, Cho JW, Endo M. Electroactive shape memory
performance of polyurethane composite having homogeneously dispersed and
covalently crosslinked carbon nanotubes. Carbon 2010;48:1598603.
[40] Rajaa M, Shanmugharajb AM, Ryub SH, Subhac J. Inuence of metal
nanoparticle decorated CNTs on polyurethane based electro active shape
memory nanocomposite actuators. Mater Chem Phys 2011;129:92531.
[41] Velankar S, Cooper SL. Microphase separation and rheological properties of
polyurethane melts. 1. Effect of block length. Macromolecules
1998;31:918192.
[42] Madbouly SA, Otaigbe JU. Recent advances in synthesis, characterization and
rheological
properties
of
polyurethanes
and
POSS/polyurethane
nanocomposites dispersions and lms. Prog Polym Sci 2009;34:1283332.
[43] Stauffer D, Aharony A. Introduction to percolation theory. London: Taylor and
Francis; 1994.
[44] Chun KY, Kim SH, Shin MK, Kim YT, Spinks GM, Aliev AE, et al. Free-standing
nanocomposites with high conductivity and extensibility. Nanotechnology
2013;24:165401.
[45] Nanni F, Mayoral BL, Madau F, Montesperelli G, McNally T. Effect of MWCNT
alignment on mechanical and self-monitoring properties of extruded PET
MWCNT nanocomposites. Compos Sci Technol 2012;72:11406.
[46] Gunes S, Cao F, Jana SC. Evaluation of nanoparticulate llers for development
of shape memory polyurethane nanocomposites. Polymer 2008;49:222334.
[47] Liu Y, Lv H, Lan X, Leng J, Du S. Review of electro-active shape-memory
polymer composite. Compos Sci Technol 2009;69:20648.