Polymer 84 (2016) 209e222

Contents lists available at ScienceDirect

Polymer
journal homepage: www.elsevier.com/locate/polymer

Ultrasonic treatment of polycarbonate/carbon nanotubes composites

Xiang Gao, Avraam I. Isayev*, Chao Yi
Department of Polymer Engineering, The University of Akron, Akron, OH 44325-0301, USA

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

Article history: A twin-screw extruder having an ultrasonic treatment zone in the barrel was used to prepare poly-
Received 28 October 2015 carbonate (PC)/multi-walled carbon nanotubes (MWNTs) composites. The effect of molecular weight of
Received in revised form PC, carbon nanotubes (CNTs) concentration and ultrasonic amplitude on processing characteristics,
22 December 2015
electrical, morphological, rheological and mechanical properties of PC filled with 0.2e1.5 wt% CNTs was
Accepted 26 December 2015
investigated. Ultrasonic treatment showed significant effects on improving the dispersion and distri-
Available online 29 December 2015
bution of CNTs especially in low molecular weight PC (LPC). The electrical and rheological percolation
thresholds of low and high molecular weight PCs obtained by fitting the experimental data to classic
Keywords:
Polycarbonate
percolation theory were significantly decreased after ultrasonic treatment. Specifically, the electrical
Molecular weight percolation threshold of LPC/CNTs composites decreased from 0.176 vol% (0.30 wt%) for untreated
Carbon nanotubes samples to 0.088 vol% (0.15 wt%) for treated samples at an amplitude of 13 mm. To the best of our
Ultrasound knowledge, this obtained value is the lowest percolation threshold, so far, achieved in melt processed PC/
Percolation CNTs composites. The mechanism for the decrease of rheological and electrical percolation thresholds
after ultrasonic treatment was proposed. Also, some improvements in mechanical properties (including
the Young's modulus and elongation at break) of PC/CNTs composites after ultrasonic treatment at an
amplitude of 13 mm were observed. The ultrasonic extrusion technology is shown to be a suitable method
for manufacturing high quality products at lower CNTs concentration.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction composites. The percolation threshold is the critical concentration

of a conductive filler at which a conductive network is formed,
Since the discovery of CNTs in 1991 by Iijima [1], CNTs have exhibiting a sudden jump of the electrical conductivity [11]. After
drawn much attention due to their high aspect ratio, extraordinary the percolation threshold, the electrical conductivity will not
mechanical, electrical and thermal properties [2e6]. It has been change much. Therefore, reaching the electrical percolation
theoretically and experimentally proved that individual single-wall threshold at low filler concentration could provide the polymer/
carbon nanotubes (SWNTs) showed tensile modulus of 640 GPa to CNTs composites with electrical conductivity being sufficient to
1 TPa [7]. The tensile strength of multi-wall carbon nanotubes meet most of the applications, such as electromagnetic interface
(MWNTs) was found to be ~63 GPa [8]. The electrical conductivity (EMI) shielding and antistatic materials. Also, decreasing the
of CNTs can be as high as 9.2
104 S/m [9]. Also, the thermal percolation threshold can help to improve the processability of the
conductivity of CNTs can reach up to 6600 W/m$K at a room composites and maintain or even improve the mechanical prop-
temperature [10]. All these extraordinary properties make CNTs the erties of the composites, since high loading of CNTs may induce
best filler candidates to prepare high performance polymer com- more stress concentrators, diminishing the elongation and tensile
posites, especially conductive polymer composites. Though the stress of the composites. In order to decrease the electrical perco-
price of CNTs is steadily decreasing over years, it is still significantly lation, better dispersion of nanotubes in the polymer composites is
more expensive than traditional carbon fillers, like carbon black. required. However, due to the very large van der Waals force be-
Thus to lower the cost of products, it would be beneficial to reduce tween the nanotubes and large number of primary agglomerates
the electrical percolation threshold of conductive polymer existing in the raw CNTs, it is very difficult to disperse CNTs in
polymer composites, which makes the efficient transition of
nanotube properties to the polymer matrix challenging [12].
The commonly used methods to prepare polymer composites
* Corresponding author.
are in-situ polymerization [13e16], solution methods [17e19], and
E-mail address: aisayev@uakron.edu (A.I. Isayev).

http://dx.doi.org/10.1016/j.polymer.2015.12.051
0032-3861/© 2016 Elsevier Ltd. All rights reserved.
210 X. Gao et al. / Polymer 84 (2016) 209e222

melt processing [3e5,20e22]. Though good dispersion of CNTs can process were provided by Nanocyl S.A. (Belgium), under the trade
be achieved by in-situ polymerization, there are only limited name of NC7000. According to the producer, it has carbon purity of
numbers of polymers that can be used in in-situ polymerization. >90 wt%, average diameter of 9.5 nm, average length of 1.5 mm,
Solution methods are highly dependent on heavy use of chemicals surface area of 250e300 m2/g by BET method, bulk density of 51 g/
and it is not easy to scale up. Compared with these two methods, L, density of 2.04 g/cm3 by Helium Pycnometry. CNTs are used as
melt compounding is a promising method to disperse nanofillers in received in this study.
polymer composites. It is efficient and environmentally friendly.
Also, from a product commercialization perspective, it is most 2.2. Sample preparation
preferred because it can be easily scaled up. However, one draw-
back of melt processing method is that the dispersion of nanofillers PC easily absorbs moisture, which may influence the properties
in the products is not as good as using the other two methods. Over of PC after thermal processing. The allowable moisture content for
the past decade, high power ultrasound has been proven to be an molding a satisfactory part is 0.03% [31]. Therefore, prior to all
efficient method to improve the dispersion of nanofillers in poly- thermal processing (extrusion, compression molding, injection
mer matrix [4,20e25]. Swain and Isayev studied the high density molding), PC pellets were dried in a vacuum oven at 100  C over-
polyethylene (HDPE)/clay nanocomposites prepared by a single night to ensure low moisture content.
screw compounding extruder attached with an ultrasound die. It A co-rotating twin-screw extruder (PRISM USALAB 16, Thermo
was found that the intercalation of HDPE into lattice layers was Electron Corp. UK) with an ultrasonic treatment zone located in the
enhanced and the oxygen permeability of the nanocomposites was extruder barrel, as seen in Fig. 1(a), was used for composites
decreased after ultrasonic treatment [22]. Zhong et al. studied the preparation. The screw configuration used in this study is shown in
polypropylene (PP)/CNTs composites using ultrasonically aided Fig. 1(b). The screw consists of conveying sections, kneading sec-
twin-screw extruder. After ultrasonic treatment, better dispersion tions and an ultrasonic zone. The extruder has a screw diameter of
of CNTs was achieved and electrical, mechanical properties were 16 mm and an L/D of 25. Compounding was carried out at a screw
improved [4]. speed of 200 RPM, a feeding rate of 908 g/hr (2 lbs/hr), and a
Polycarbonate (PC) is a widely used engineering plastic exhib- temperature profile of 240, 260, 285, 285, 285, 285  C for HPC
iting high transparency, high impact-resistance and excellent composites, 210, 225, 240, 240, 240, 240  C for LPC composites from
flexibility over a wide temperature range. Preparation of PC/CNTs the hopper to the die. PC pellets were fed into the extruder by
composites using melt processing has been extensively studied feeder 1 and CNTs were fed downstream by feeder 2 located at
[26e28]. The effect of screw configuration of twin-screw extruder position where PC pellets were already melted. In this way, the
[26], polymer viscosity [27], and types of CNTs [28] were studied. It breakage of CNTs caused by solid PC particles can be reduced. After
was found that the aspect ratio of CNTs was critical to achieve lower feeding of CNTs, the composites material continued downstream
percolation [29]. Longer residence time and higher polymer vis- through a mixing section (KN2 in Fig. 1(b)) before it entered the
cosity in extrusion could break up more CNTs agglomerates, but the ultrasonic treatment zone. After ultrasonic treatment, the com-
nanotubes might be also broken up at the same time, resulting in posites were further mixed (KN3 in Fig. 1(b)) and then exited
reduction of their aspect ratio. Most of the electrical percolation through a circular die. The strand was subsequently water
thresholds of PC/CNTs composites in these works were reported to quenched and pelletized. The ultrasonic system (Branson Ultra-
lay between 0.3 wt% and 0.5 wt% [27,28]. As for the mechanical sonic Corp., CT, USA) works at a frequency of 40 kHz. Power con-
properties of PC/CNTs composites, little improvement was achieved sumption was measured using a Dataq data acquisition system
with the addition of CNTs in PC composites [28,30]. connected directly to the power supply. The torque of the extruder
In the present study, ultrasonically aided twin-screw extruder was recorded as the percentage of the maximum extruder torque
was used to prepare PC/CNTs composites. Two different molecular being 100%. Untreated and treated samples at ultrasonic ampli-
weight PCs were used to study the effect of molecular weight on the tudes of 7.5, 10, 13 mm with CNTs concentrations of 0, 0.2, 0.3, 0.4,
electrical, morphological and mechanical properties of ultrasoni- 0.8, 1.5 wt% were prepared. The samples at different concentrations
cally treated polymer composites. Rheological properties of the PC/ were denoted as PC-xCNT with x being the weight concentration of
CNTs composites were also investigated to explain the effect of CNTs.
ultrasonic treatment on the properties of composites. The electrical
and rheological percolation thresholds were obtained by fitting the 2.3. Electrical resistivity
experimental data to the classic percolation theory. High power
ultrasonic device placed in the barrel was found to be an easy and Samples for electrical resistivity test with a diameter of 90 mm
efficient method to improve the dispersion and distribution of CNTs and a thickness of 1 mm were prepared by compression molding at
and decrease the electrical percolation threshold of the PC/CNTs 250  C and 49 MPa using a compression molding machine (CARVER
composites to the level not seen in the available literature. The 4122, Wabash, IN, USA). A Keithley electrometer (Model 6517A,
mechanical properties of the PC/CNTs composites were also stud- Keithley Instruments, Cleveland, OH, USA) with an 8009 test fix-
ied. Finally, the mechanism of the decrease of percolation threshold tures was used to measure the volume resistivity of the composites
was proposed. in the range from 1018 to 107 Ohm cm in accordance with the ASTM
D257. Alternating polarity mode using a voltage of ±10 V was used
2. Experimental to eliminate the influence of background current. For samples with
volume resistivity lower than 107 Ohm cm, a Keithley Micro-
2.1. Materials ohmmeter (Model 580, Keithley Instruments, Cleveland, OH, USA)
was used for both the input and output signals in accordance with
Polycarbonate Lexan OQ3820 with high molecular weight the ASTM 4496.
(Mw ¼ 31,110 g/mol by GPC, denoted as HPC) and polycarbonate The Keithley electrometer 6517A with 8009 test fixtures was
Lexan OQ1030 with low molecular weight (Mw ¼ 17,610 g/mol by also used to measure the surface resistivity of the composites in the
GPC, denoted as LPC) were kindly provided by SABIC plastics. These range from 1015 to 107 Ohm/square. Van der Pauw four-point probe
two PCs have linear structure. The multi-walled carbon nanotubes method was used to characterize the surface resistivity of the
(CNTs) produced by catalytic carbon vapor deposition (CCVD) samples lower than 107 Ohm/square. Two Keithley 2400 source
 X. Gao et al. / Polymer 84 (2016) 209e222 211

Fig. 1. Schematic of the ultrasonic twin-screw extruder (a) and the screw configuration (b) used in this study.

meters were used during the test and four tungsten probes con- samples.
nected to the source meters were placed on top of the four silver
coated corners of the sample [5]. Two probes were used to apply 2.6. Tensile testing
the current and the other two probes were used to measure the
voltage. The data was collected by computer through the Labview The tensile test samples were injection molded with a Haake
softwares. MiniJet Injection Molding System (Thermo Scientific, USA). Due to
the difference in melt viscosity of HPC and LPC composites,
2.4. Rheological measurements different cylinder and mold temperatures were used. A cylinder
temperature of 330  C and a mold temperature of 130  C were used
An ARES G2 strain-controlled rotational rheometer (TA In- for HPC composites, and a cylinder temperature of 300  C and a
struments) with parallel plate geometry was used to study the mold temperature of 110  C were used for LPC composites. An in-
rheological properties of the composites. Circular disk samples jection pressure and packing pressure of 50 MPa was used for both
with a diameter of 25 mm and a thickness of 2 mm were prepared HPC and LPC composites. Tensile tests were conducted on an Ins-
by compression molding at 250  C and 49 MPa. These samples were tron Testing Machine (Model 5567, Instron Corp.) with 1 kN load
then loaded to the parallel plates in the rheometer and squeezed to cell at room temperature according to ASTM 638 (type V) without
1.8 mm. All the tests were conducted at strain amplitudes in the extensometer. A crosshead speed of 50 mm/min was used and the
linear viscoelastic region at a constant temperature of 230  C for reported results were averages of at least six samples. The Young's
HPC and 210  C for LPC composites. Different testing temperatures modulus, yield strength, elongation at break and tensile strength
were used for HPC and LPC in order to observe the terminal were measured.
behavior of the composites which was highly influenced by the
dispersion state of CNTs. The storage modulus, loss modulus and 3. Results and discussion
loss tangent were recorded in a frequency range from 0.05 to
200 rad/s. 3.1. Processing characteristics

2.5. Morphological study The power consumption of HPC and LPC composites during ul-
trasonic treatment as a function of CNTs concentration is plotted in
The dispersion of CNTs in the scale of several micrometers was Fig. 2. The measured power consumption is the total power con-
studied using an optical microscope (Olympus BX51 Microscope sumption during the treatment of composites. Part of it was dissi-
System). The sample was prepared by following procedures. Firstly, pated as heat and the rest was exerted to break the agglomerate of
a very small piece of composites was cut from a randomly chosen CNTs, improving the dispersion of CNTs in polymer melt. As seen
extruded pellets and put between two cover glasses. Then, the from Fig. 2, the power consumption increased with ultrasonic
glasses were put onto the lower hot face (250  C) of compression amplitude, because the increase of amplitude caused an increase of
molding machine and the upper cover glass was pressed slowly the acoustic pressure, resulting in the increase of power con-
with a tweezer, making the sample being hot pressed into a very sumption [32]. Higher power consumption indicated that more
thin film with a thickness of about 10 mm. The sample was then energy was transmitted to polymer melt. Especially, there was a
moved to optical microscopy for observation. Transmission elec- large increase of power consumption from 7.5 mm to 10 mm.
tron microscope (TEM) (JEOL JEM-1230, Tokyo, Japan) was used to Little difference of power consumption between samples at
investigate the nanoscale dispersion of CNTs in the composites. different CNTs concentration was observed. The average value of
TEM samples with a thickness of about 100 nm were cut from an power consumption of LPC at same CNTs concentration and ultra-
injection molded dumbbell sample with a Leica EM UC7 microtome sonic amplitude was higher than that of HPC composites. A theo-
machine. It needs to be mentioned that the TEM pictures were retical modeling of ultrasonic power consumption in a bubble filled
randomly taken in more than three specimens and therefore pic- polymer indicated that the power consumption increased with
tures shown here represent the dispersion state of CNTs in the sound velocity, hydrostatic pressure and longitudinal loss tangent
212 X. Gao et al. / Polymer 84 (2016) 209e222

Fig. 2. Ultrasonic power consumption of (a) HPC and (b) LPC composites.
Fig. 3. Torque of the twin-screw extruder of HPC (a) and LPC (b) composites.

of material at ultrasonic frequency [33]. The relatively higher power

while percolation threshold was reached. Ultrasonic treatment
consumption in LPC composites was possibly due to the relatively
showed a larger effect on decreasing the surface resistivity for low
higher value of loss tangent at the ultrasound frequency and dif-
CNTs concentration samples. The electrical percolation threshold
ference of hydrostatic pressure at the ultrasonic treatment zone.
continuously decreases with the increase of ultrasonic amplitude.
A continuous decrease of torque was observed with increasing
The volume resistivities of the composites are plotted in Fig. 5,
ultrasonic amplitude, as shown in Fig. 3. This torque decrease could
showing similar trend as the surface resistivity. Specifically, the
be caused by permanent effect including the degradation of poly-
volume resistivity continuously decreased with the increase of
mer under action of ultrasound and the thixotropic effects
CNTs concentration and the electrical percolation thresholds
including the shear thinning in the melt. The decrease of torque is
decreased with the increase of ultrasonic amplitude. Furthermore,
beneficial in polymer melt processing by improving the process-
it can be seen from Fig 5(a) that the electrical percolation threshold
ability of high viscosity polymers. Taking the error bar of the torque
for HPC composites treated at an amplitude of 13 mm was signifi-
data into consideration, there was little change of the torque at
cantly decreased compared with the untreated sample and samples
different CNTs concentrations. This was due to the thixotropic effect
treated at 7.5 mm and 10 mm. Also, as seen from Fig. 5(b), LPC
caused by ultrasonic treatment and the high shear rate in the
composites showed an even more obvious effect of ultrasound on
extruder. When comparing the torque of HPC and LPC composites,
the electrical percolation threshold. At 0.2 wt%, the samples treated
the torque of HPC at same CNTs concentration and ultrasonic
at 7.5 mm were already percolated and showed a large decrease of
amplitude was higher than that of LPC composites. This was caused
the volume resistivity compared to the untreated samples.
by the higher viscosity of HPC composites during extrusion.
In order to fit the volume resistivity data using classic percola-
tion theory, the volume resistivity r (Ohm$cm) was converted to
3.2. Electrical properties electrical conductivity s (S/cm) with following equation:

The surface resistivity of the PC/CNTs composites were

1
measured and plotted in Fig. 4. In general, the surface resistivity of s¼ (1)
r
the composites continuously decreased with the increase of CNTs
concentration. A sudden decrease of the resistivity can be observed The following equation was used to fit the electrical
 X. Gao et al. / Polymer 84 (2016) 209e222 213

Fig. 5. Volume resistivities of HPC (a) and LPC (b) composites.

Fig. 4. Surface resistivities of HPC (a) and LPC (b) composites.

Table 1
conductivity, s, of the composites in order to obtain the percolation The weight and volume fractions of CNT in PC/CNTs composites.
threshold of the composite at various ultrasonic amplitudes [34]:
wt% 0.2 0.3 0.4 0.8 1.5
n vol% 0.118 0.177 0.236 0.472 0.888
sð4Þ ¼ kð4  40 Þ (2)

where 4 is the volume concentration of CNTs, 40 is the electrical

percolation threshold, n is the exponent. According to the classic composites decreased from 0.176 vol% (0.30 wt%) for untreated
percolation theory, when the fillers forms a three dimensional samples to 0.088 vol% (0.15 wt%) for treated samples at an ampli-
network in the matrix, i.e. the percolated network in the PC/CNTs tude of 13 mm. When comparing data of HPC and LPC composites, it
composites, the value of n equals 2 [11,35]. Therefore, n ¼ 2 was can be observed that the electrical percolation threshold of un-
used in the further fitting. treated HPC composite was lower than that of untreated LPC
In order to fit the data to Eq. (2), the weight fractions were composite. The shear stress of HPC composite during extrusion was
converted to volume fractions as shown in Table 1, based on the higher than that of LPC composite (as indicated by higher torque
density of PC of 1.2 g/cm3 and the density of CNTs of 2.04 g/cm3 shown in Fig. 3), resulting in better CNTs dispersion (as indicated by
provided by manufacturer. the morphological results that will be discussed later). However,
Composites with CNTs concentration above the percolation the 40 of LPC composites decreased faster than HPC composites and
threshold were chosen to fit the electrical conductivity to Eq. (2). resulted in a lower percolation threshold for ultrasonically treated
The results are shown in Fig. 6 with the inset figure sample showing samples. This indicated that ultrasound showed larger effect on
fitting of the conductivity data for samples treated at 13 mm. The improving the electrical properties of LPC composite. To the best of
fitting parameters are shown in Table 2. The electrical percolation the authors' knowledge, the electrical percolation threshold of
thresholds in Table 2 are plotted in Fig. 7. 0.088 vol% (0.15 wt%) for treated LPC/CNTs composites at 13 mm is
According to Table 2, the electrical percolation threshold of HPC the lowest percolation threshold obtained in melt processed PC/
and LPC composites continuously decreased with the increase of MWNTs composites [11,28,30].
ultrasonic amplitude. The 40 of HPC composites decreased from
0.135 vol% (0.23 wt%) for untreated samples to 0.109 vol% (0.19 wt%)
for treated samples at an amplitude of 13 mm. The 40 of LPC
214 X. Gao et al. / Polymer 84 (2016) 209e222

Fig. 7. Electrical percolation threshold as a function of ultrasonic amplitude of HPC and

LPC composites.

significant differences between untreated and treated samples

were observed in HPC-1.5CNT and LPC-1.5CNT, respectively. A sig-
nificant improvement of CNTs distribution in LPC-1.5CNT after ul-
trasonic treatment could be observed by comparing Fig. 8(c) and
(d). Smaller sizes and less number of CNTs agglomerates were
observed in treated LPC-1.5CNT sample. Even for the samples with
highest CNTs concentration in this study (1.5 wt%), very few CNTs
agglomerates were found in the optical micrographs, showing that
CNTs can be dispersed very well in PC. This was also observed in the
study by Socher et al. [27]. Less difference was seen in the HPC and
LPC composites at lower CNTs concentrations from optical micro-
scopy (not shown here) due to its low magnification. In order to see
changes in the dispersion at higher magnification, TEM was used. In
fact, TEM was proven to be an efficient method to characterize the
dispersion of CNTs in the nanometer scale [36,37]. Accordingly,
TEM was used to characterize the dispersion of CNTs for PC-0.2CNT
and PC-0.8CNT samples untreated and treated at an amplitude of
13 mm, as shown in Figs. 9 and 10. It can be seen that in HPC-0.2CNT,
Fig. 6. Electrical conductivity of HPC (a) and LPC (b) composites as a function of vol- before ultrasonic treatment, there were some agglomerates in
ume concentration of CNTs at different ultrasonic amplitudes. Inset figure shows the Fig. 9(a). After ultrasonic treatment at an amplitude of 13 mm, the
fitting of PC/CNTs composites treated at an amplitude of 13 mm.

Table 2
Fitting parameters of Eq. (2) for HPC and LPC composites.

0 mm 7.5 mm 10 mm 13 mm

HPC k (S/m) 1.42E-04 1.96E-04 2.50E-04 2.28E-04

40 (vol%) 0.135 0.127 0.114 0.109
R2 0.9090 0.9020 0.9731 0.9273
LPC k (S/m) 6.23E-04 5.38E-04 5.90E-04 4.54E-04
40 (vol%) 0.176 0.110 0.104 0.088
R2 0.9086 0.9311 0.9732 0.9876

3.3. Morphology agglomerates of CNTs were broken up and very few agglomerates
can be seen in Fig. 9(b). This explains why the decrease of electrical
The decrease of the electrical percolation after ultrasonic resistivity occurred after ultrasonic treatment. With better disper-
treatment can be attributed to the improved CNTs dispersion and sion of CNTs, a more robust CNTs network was formed and thus the
distribution in the composites. Optical microscopy was used to electrical resistivity was decreased. At relatively high CNTs con-
characterize the dispersion of CNTs in the scale of micrometers. The centration of 0.8 wt%, very little difference in dispersion can be seen
optical micrographs of HPC-1.5CNT and LPC-1.5CNT samples un- before and after ultrasonic treatment from Fig. 9(c) and (d). As a
treated and treated at an amplitude of 13 mm are shown in Fig. 8, result, very limited change of the electrical resistivity occurred, as
indicating agglomerates present in composites. Little and can be seen from Figs. 4(a) and 5(a).
 X. Gao et al. / Polymer 84 (2016) 209e222 215

Fig. 8. The optical micrographs of HPC-1.5CNT sample untreated (a) and treated at an amplitude of 13 mm (b) and LPC-1.5CNT sample untreated (c) and treated at an amplitude of
13 mm (d). The circles highlight the agglomerates of CNTs.

The TEM micrographs of LPC-0.2CNTs and LPC-0.8CNT samples effects on the rheological behaviors, which suggested that CNTs did
untreated and treated at an amplitude of 13 mm are shown in Fig. 10. not significantly influence the short-range dynamic motion of
The results were similar to HPC composites, i.e. the dispersion of polymer chains. From Fig. 11(b), significant decrease of Tan d with
CNTs was obviously improved at 0.2 wt% after ultrasonic treatment, increasing CNTs concentration was observed at low frequencies,
while limited effect was observed in 0.8 wt% samples. This result which was also an indication of the effect of CNTs on the relaxation
was consistent with a lower electrical resistivity in 0.2 wt% samples behavior of polymer chains.
after ultrasonic treatment, as shown in Figs. 4(b) and 5(b). The influence of ultrasonic treatment on the rheological prop-
erties can also be seen from Fig. 11. In Fig. 11(a), the storage modulus
of untreated HPC was lower than that of virgin HPC, indicating that
3.4. Rheological properties
certain degradation of polymer happened during extrusion. With
the increase of ultrasonic amplitude, further decrease of G0 was
The rheological behavior of composites is considered as an in-
observed, which was caused by the polymer chain scission during
direct but efficient method to determine the dispersion state of
ultrasonic treatment. As can be seen from Fig. 11(a), for 0.2 wt%
nanofillers in polymer composites. Especially, the rheological
samples, the storage modulus at low frequencies continuously
behavior at low frequency range of small-amplitude oscillatory
increased with the increase of ultrasonic amplitude, indicating the
shear (SAOS) test is highly influenced by the fillerefiller and pol-
improvement of filler dispersion and formation of more robust
ymerefiller interactions in polymer-nanofillers composites [38,39].
CNTs networks in the composites after application of ultrasound.
Fig. 11 shows the effect of ultrasound on the storage modulus
The effect of ultrasound on G0 at low frequencies became less
and Tan d of HPC composites as a function of frequency at different
obvious for 0.3 wt% and higher CNTs concentrations samples. It
CNTs loadings. In Fig. 11(a), typical terminal behavior with the
should be noted that when ultrasound was applied during pro-
scaling properties of approximately G’~u2 was observed for pure
cessing, two effects occurred. One is the chain scission of the
HPC samples. With the increase of CNTs concentration, this ter-
polymer chains and the other is the breakup of the CNTs agglom-
minal behavior started to disappear at 0.2 wt%, which indicated
erates. Accordingly, the storage modulus was a result of the
that long-range motion of polymer chains was restrained by the
competition of these two effects. So it is reasonable to say that for
presence of CNTs. Additionally, with the increase of CNTs concen-
samples with CNTs concentration higher than 0.3 wt%, the effect of
tration, the storage modulus consistently increased and the slope of
ultrasound on improving the dispersion of CNTs was possibly
G0 at low frequencies decreased. This is a solid evidence of the
concealed by the degradation of polymer chains. It can be seen from
formation of CNTs network and the increasing interaction between
Fig 11(b) that Tan d at low frequencies decreased with the increase
polymer and CNTs. At high frequencies, CNTs showed limited
216 X. Gao et al. / Polymer 84 (2016) 209e222

Fig. 9. The TEM micrographs of HPC-0.2CNT untreated (a) and treated at an amplitude of 13 mm (b), HPC-0.8CNT untreated (c) and treated at an amplitude of 13 mm (d). The circles
highlight the agglomerates of CNTs.

of ultrasonic amplitude for samples at same CNTs concentration. dispersion of CNTs might be concealed by the degradation of
This was because the storage modulus increased more than the loss polymer matrix, so modified ColeeCole plot (i.e. log G00 vs log G0 )
modulus after ultrasonic treatment. The lower Tan d for ultrasoni- was used to reveal the structure variation in the composites while
cally treated samples indicated improved interaction between “hiding” the effect of molecular weight. The ColeeCole plot of
polymer and CNTs and more robust CNTs network formed with the polymers varying only in molecular weight can be reduced to a
increase of the ultrasonic amplitude. single master curve [40,41]. Therefore, it has been proved to be an
For pure LPC, there was no obvious decrease of the storage efficient method to reveal the change in structure of polymer or
modulus (as can be seen in Fig. 12(a)). The storage modulus dif- polymer composites that are caused by molecular weight distri-
ference between HPC and LPC after ultrasonic treatment was bution [42], molecular structure [43,44], and polymerefiller in-
probably due to lower processing temperature and lower shear teractions [38,45].
stress. This correlated with lower torque in Fig. 3(b) during the From the ColeeCole plots in Fig. 13, the data of pure HPC and LPC
extrusion of LPC composites. samples were reduced to a single curve, showing that there was no
LPC composites showed similar behavior of the storage modulus microstructure difference between untreated and treated PC resins.
and Tan d as HPC composites. The storage modulus G0 at low fre- For the PC/CNTs composites, at a given loss modulus G00 , the storage
quencies increased and the slope of G0 decreased with increasing modulus G0 increased significantly with increase of CNTs concen-
concentration of CNTs. At lower concentration of CNTs, e.g. 0.2 wt%, tration, indicating obvious changes of the microstructure of the
0.3 wt%, 0.4 wt%, the effect of ultrasound can be obviously observed composites. This was attributed to higher PCeCNTs interaction,
from the increase of the storage modulus with increasing ultrasonic making the composites at high CNTs concentration more elastic. In
amplitude. Especially, for the LPC/CNTs composites at 0.2 wt% CNTs, addition, for PC/CNTs samples at same CNTs concentrations, the
the storage modulus at the frequency of 0.05 rad/s showed an in- ColeeCole plot shifted towards higher storage modulus with
crease of almost one order of magnitude at an ultrasonic amplitude increasing ultrasonic amplitude. This was a solid evidence that the
of 7.5 mm. When comparing this result with that of HPC-0.2CNT ultrasonic treatment improved the dispersion of CNTs and thus
sample, ultrasound showed larger effect on LPC composites than increased the PC-CNTs interactions. Similar to results of the storage
HPC composites at 7.5 mm. From Tan d at low frequencies in modulus and Tan d shown in Figs. 11 and 12, ultrasound showed
Fig. 12(b), it was confirmed that ultrasound had larger effect on more pronounced effect at low CNTs concentrations and limited
LPC-0.2CNT on improving dispersion of CNTs and polymerefiller effect at high CNTs concentration.
interaction. As discussed above, with the addition of CNTs, the storage
As mentioned above, the effect of ultrasonic treatment on the modulus of the composites at low frequencies increased
 X. Gao et al. / Polymer 84 (2016) 209e222 217

Fig. 10. The TEM micrographs of LPC-0.2CNT untreated (a) and treated at an amplitude of 13 mm (b), LPC-0.8CNT untreated (c) and treated at an amplitude of 13 mm (d). The circle
highlights the agglomerates of CNTs.

dramatically. When a critical value of concentration was achieved, continuously decreased with the increase of ultrasonic amplitude.
the storage modulus showed a significant increase. This corre- The 40 of HPC composites decreased from 0.0619 vol% for untreated
sponded to the rheological percolation threshold showing the samples to 0.0439 vol% for treated samples at an amplitude of
onset of solid-like behavior. Above the percolation threshold, a 13 mm, and 40 of LPC composites decreased from 0.102 vol% for
powerelaw relationship between the storage modulus and CNTs untreated samples to 0.0550 vol% for treated samples at an
volume concentration could be observed [46]: amplitude of 13 mm. Ultrasound showed larger effect on improving
the dispersion of CNTs in LPC composite, which was indicated by
45% decrease of the rheological percolation threshold.
G0 ð4Þ ¼ kð4  40 Þn (3)
When comparing the rheological and electrical percolation
where 4 is the volume concentration of CNTs, 40 is the rheological threshold from Figs. 7 and 15, it is worthwhile to note that the
percolation threshold, n is the exponent, G0 is the storage modulus rheological percolation threshold is lower than electrical threshold
at the frequency of 0.05 rad/s in this experiment. n ¼ 2 was used in for the samples at the same conditions. For example, one can see
the fitting for the same reason as mentioned above. that for untreated HPC/CNTs composites, the rheological percola-
Composites with CNTs concentration above the percolation tion is 0.0619 vol%, which is only half of the electrical percolation
threshold, which is equal or above 0.118 vol%, were chosen to fit the (0.135 vol%). This can be explained by the different fillerefiller
storage modulus to Eq. (3). The storage modulus of both HPC and distances required for rheological and electrical percolation [46]. In
LPC composites as a function of the volume concentration were the PC/CNTs composites, electron hopping mechanism is applied to
fitted to Eq. (3) and the results are shown in Fig. 14. The single data the electrical conductivity, the required fillerefiller distance is less
point at the left bottom corner shows the storage modulus of virgin than 5 nm for the composites to be conductive. On the other hand,
PC. A large increase of G0 can be seen when percolation happens, as long as the fillerefiller distance is comparable to the diameter of
which indicated that a sudden structure change in the material random coils of PC chains, the CNTs network can effectively restrain
happened and CNTs formed a percolated network impeding the polymer motion. The average radius of gyration of PC chains can be
motion of polymer chains [46]. Samples treated at an ultrasonic calculated as:
amplitude of 13 mm were used as an example to show a power law
fit of Eq. (3) to the data in the inset figure. The fitting parameters are Rw ¼ bMw
a
(4)
summarized in Table 3. The rheological percolation thresholds in
Table 3 are plotted in Fig. 15. where Rw is the mean-squared radius of gyration with a unit of Å,
As shown in Fig. 15, the rheological percolation threshold Mw is the weight average molecular weight of PC, b ¼ 0.457, a ¼ 0.5
218 X. Gao et al. / Polymer 84 (2016) 209e222

Fig. 12. Storage modulus (a) and Tan d (b) of LPC composites.
Fig. 11. Storage modulus (a) and Tan d (b) of HPC composites.

melt to break the agglomerates of CNTs. Also due to the low vis-
according to [47]. Based on Eq. (4), the average diameter of the cosity of LPC during ultrasonic treatment, the cavitating bubble in
random coils 2Rw for HPC and LPC are 16.1 nm and 12.1 nm, the polymer melt can be easily expanded and contracted by ul-
respectively. Therefore, the required fillerefiller distance for rheo- trasound to break the agglomerates. Another reason is that the
logcial percolation is much larger than that for the electrical shear stress during the extrusion of LPC composites is lower than
percolation. As a result, more CNTs are required to reach the elec- that of HPC composites. Therefore, the scission of nanotubes
trical conductivity threshold in the composites. Also, it should be happened less in LPC composites and the aspect ratio of nanotubes
noted that CNTs could be metallic or non-metallic due to the dif- was generally higher than HPC composites [27]. As a result, elec-
ference in CNTs chirality determining the electrical behavior of trical and rheological percolations of LPC composites were more
CNTs [48]. Therefore, in the PC/CNTs composites non-metallic CNTs significantly decreased.
do not contribute to the electrically conductive network, though
they could restrict the motion of polymer chains. Furthermore, it
can be seen from Fig. 15 that the rheological percolation threshold 3.5. Mechanical properties
of HPC is lower than that of LPC. This can be explained by the larger
fillerefiller distance required for HPC composites to restrict the The Young's modulus, yield stress, elongation at break and
polymer chain motion than LPC composites. tensile strength of HPC composites untreated and treated at an
Overall, ultrasonic treatment showed larger effect in LPC com- amplitude of 13 mm as a function of CNTs concentration are shown
posites than HPC composites in the electrical and rheological in Fig. 16. The Young's modulus and yield stress generally increased
properties, which was mainly due to the better CNTs dispersion in with the increase of CNTs concentration. The significant improve-
LPC composites. One reason is that ultrasonic power consumption ment of Young's modulus at low CNTs concentration was possibly
during extrusion of LPC composites is higher than that of HPC due to the formation of the CNTs network and the immobilization
composites, indicating more energy was transmitted into polymer of HPC molecular chains on the surface of nanotubes. Ding et al.
 X. Gao et al. / Polymer 84 (2016) 209e222 219

Fig. 13. ColeeCole plot of HPC (a) and LPC (b) composites. Fig. 14. Storage modulus of HPC (a) and LPC (b) composites as a function of the volume
concentration of CNTs at different ultrasonic amplitudes. Inset figures show a power
law fit of Eq. (3) to the data of PC composites treated at an amplitude of 13 mm. The
[49] found that molecular chains of PC can form a sheathing layer single data point at the left bottom corner is the storage modulus of virgin PC.
on nanotubes, which can efficiently enhance the interaction and
load transfer between PC and CNTs. As a result, the Young's
in Fig. 17. Similar to the results of HPC composites, the Young's
modulus was improved after incorporation of CNTs. A drop of yield
modulus and yield stress of LPC composites generally increased
stress was observed in HPC-0.2CNT samples, which was probably
with CNTs concentration, while the elongation at break continu-
due to the incorporation of CNTs causing defects influencing the
ously decreased. This observation for LPC composites can be
yield behavior of the polymer. The elongation at break decreased
explained in the same way as that for HPC composites. Taking the
with the increase of CNTs concentration. Higher content of CNTs
error bar into consideration, there is almost no change in the tensile
might induce more defects in the samples leading to lower elon-
strength of LPC composites with CNTs concentration. This is
gation. The tensile strength showed similar decreasing trend as the
because the samples of LPC composites were failed in the necking
elongation at break with CNTs concentration. It should be noted
region. As a result, there is no large difference between samples at
that the samples generally broke up at the strain hardening region
different concentrations. The effect of ultrasonic treatment can be
after necking. As a result, the lower the elongation at break, the
clearly observed from the higher Young's modulus in treated
lower the tensile strength. Thus the tensile strength showed similar
samples.
trend as the elongation at break. The ultrasonically treated HPC/
When comparing HPC and LPC samples at the same CNTs con-
CNTs composites at an amplitude of 13 mm generally showed a
centrations and ultrasonic amplitudes, it can be seen that HPC
higher Young's modulus and elongation at break than untreated
samples showed higher yield stress, elongation at break and tensile
samples. The increase of these properties after ultrasonic treatment
strength. According to work by Gardner and Martin [50] and
was mainly caused by the improvement of CNTs dispersion and
Pitman et al. [51], the stress required for craze initiation decreases
thus more robust CNTs network and less defects.
as the molecular weight decreases. Therefore, LPC samples showed
The Young's modulus, yield stress, elongation at break and
a lower yield stress. It was also reported by Gardner and Martin [50]
tensile strength of LPC composites untreated and treated at an
that the elongation at break decreased dramatically as the
amplitude of 13 mm as a function of CNTs concentration are shown
220 X. Gao et al. / Polymer 84 (2016) 209e222

Table 3
Fitting parameters of Eq. (3) for HPC and LPC composites.

0 mm 7.5 mm 10 mm 13 mm

HPC k (Pa) 22,482 22,150 23,215 22,493

40 (vol%) 0.0619 0.0498 0.0502 0.0439
R2 0.99898 0.99178 0.99837 0.99836
LPC k (Pa) 27,531 30,091 32,744 25,672
40 (vol%) 0.102 0.0701 0.0612 0.0550
R2 0.99036 0.99589 0.99409 0.99874

lower elongation and tensile strength.

3.6. Mechanism of ultrasonic effect on electrical and rheological

percolations

From the above discussion, it can be seen that ultrasound

showed tremendous effect on the rheological and electrical prop-
erties of the PC composites at low CNTs concentration (0.2 wt %,
0.3 wt % and 0.4 wt %) but limited effect on higher CNTs concen-
tration composites. This was proven by the large increase of the
storage modulus at lower frequency, the significant decrease of
electrical surface and volume resistivity and the improvement of
mechanical properties. Also, the electrical and rheological perco-
lation threshold largely decreased with the increase of the ultra-
sonic amplitude. All these results indicated that a larger effect of
ultrasound can be easily observed at low CNTs concentration. Based
on these findings, the following mechanism for this effect was
proposed. This is schematically shown in Fig. 18. At low filler con-
Fig. 15. Rheological percolation threshold as a function of ultrasonic amplitude of HPC centration, CNTs agglomerates and bundles are loosely dispersed in
and LPC composites. the polymer matrix. In this case, the density of the fillers is not high

Fig. 16. Young's modulus (a), yield stress (b), elongation at break (c) and tensile strength (d) of HPC composites as a function of CNTs concentration. The data of samples treated at
13 mm were slightly shifted along abscissa to more clearly show the error bars.

molecular weight is reduced. Accordingly, the LPC samples showed enough to form a percolated network. However, after ultrasound
 X. Gao et al. / Polymer 84 (2016) 209e222 221

Fig. 17. Young's modulus (a), yield stress (b), elongation at break (c) and tensile strength (d) of LPC composites as a function of CNTs concentration. The data of samples treated at
13 mm were slightly shifted along abscissa to more clearly show the error bars.

Fig. 18. Schematic of the effect of ultrasonic treatment on low (a) and high (b) CNTs concentration composites.

treatment, the CNTs agglomerates are broken up by the high power torque of the extruder decreased. CNTs concentration showed little
ultrasound. With such an improvement of CNTs dispersion and effect on the power consumption and torque. The surface and
distribution, a percolated network is formed. On the other hand, for volume resistivity of the composites exhibited typical percolation
the composites with high CNTs concentration, a percolated behavior with the increase of CNTs concentration and the electrical
network is already formed, though most of the fillers are in the percolation threshold decreased tremendously with the increase of
state of agglomerates. After ultrasonic treatment, the agglomerates ultrasonic amplitude. This was found to be caused by the improved
are broken up and the dispersion of fillers are improved, but this CNTs dispersion and distribution in the composites as shown by the
improvement is not easy to observe from the rheological or elec- optical microscopy and TEM studies. The storage modulus at low
trical properties because the filler network is already percolated frequency increased with the increase of ultrasonic amplitude and
before the ultrasound treatment. the rheological percolation threshold for both HPC and LPC com-
posites significantly decreased with ultrasonic treatment. When
comparing the rheological and electrical percolation threshold, it
4. Conclusions was found that the rheological percolation threshold was lower
than the electrical threshold, which can be attributed to the shorter
Ultrasonically aided twin-extruder was employed to prepare fillerefiller distance required to restrain the motion of polymer
HPC/CNTs and LPC/CNTs composites without and with the impo- chains and that the nonmetallic part of the CNTs could not
sition of ultrasonic waves. With the increase of ultrasonic ampli- contribute to the electrical conductive path in the composites.
tude, the power consumption continuously increased and the
222 X. Gao et al. / Polymer 84 (2016) 209e222

Obvious improvement in mechanical properties after ultrasound [17] S. Kaida, J. Matsui, T. Sagae, Y. Hoshikawa, T. Kyotani, T. Miyashita, Carbon 59
treatment at an amplitude of 13 mm was observed. Finally, a (0) (2013) 503e511.
[18] V.C. Tung, L.-M. Chen, M.J. Allen, J.K. Wassei, K. Nelson, R.B. Kaner, Y. Yang,
mechanism was proposed to explain the decrease of the rheological Nano Lett. 9 (5) (2009) 1949e1955.
and electrical percolation. At low CNTs concentration, the originally [19] H. Wang, B. Hsieh, G. Jime nez-Ose s, P. Liu, C.J. Tassone, Y. Diao, T. Lei,
not percolated CNTs agglomerates were broken up after ultrasonic K.N. Houk, Z. Bao, Small 11 (1) (2015) 126e133.
[20] J. Choi, A.I. Isayev, Rubber Chem. Technol. 85 (1) (2012) 14e37.
treatment and formed a percolated CNTs network. On the other [21] A.I. Isayev, R. Kumar, T.M. Lewis, Polymer 50 (1) (2009) 250e260.
hand, this effect was not easy to be observed for high CNTs con- [22] S.K. Swain, A.I. Isayev, Polymer 48 (1) (2007) 281e289.
centration composites which already had a percolated CNTs [23] R. Kumar, A.I. Isayev, Polymer 51 (15) (2010) 3503e3511.
[24] S. Niknezhad, A.I. Isayev, J. Appl. Polym. Sci. 129 (1) (2013) 263e275.
network before ultrasonic treatment. [25] J. Zhong, A.I. Isayev, J. Appl. Polym. Sci. 132 (5) (2014), http://dx.doi.org/
10.1002/app.41397.
Acknowledgements [26] T. Villmow, B. Kretzschmar, P. Po € tschke, Compos. Sci. Technol. 70 (14) (2010)
2045e2055.
[27] R. Socher, B. Krause, M.T. Müller, R. Boldt, P. Po €tschke, Polymer 53 (2) (2012)
The authors wish to acknowledge the NSF Division of Engi- 495e504.
neering for financial support under grant CMMI-1131342 and SABIC [28] F.Y. Castillo, R. Socher, B. Krause, R. Headrick, B.P. Grady, R. Prada-Silvy,
P. Po€tschke, Polymer 52 (17) (2011) 3835e3845.
plastics for providing PC resins. The authors also thank Dr. Xiong [29] B. Krause, R. Boldt, L. H€ außler, P. Po€ tschke, Compos. Sci. Technol. 114 (2015)
Gong for the help in the surface resistivity test, Dr. Wei Sen Wong 119e125.
from Malvern Instruments for the help with GPC measurements, [30] J. Guo, Y. Liu, R. Prada-Silvy, Y. Tan, S. Azad, B. Krause, P. Po€tschke, B.P. Grady,
J. Polym. Sci. Part B Polym. Phys. 52 (1) (2014) 73e83.
Mr. Ming Xiao from The University of Akron and Mr. Yongsheng
[31] T.S. Long, R.J. Sokol, Polym. Eng. Sci. 14 (12) (1974) 817e822.
Zhao from Sichuan University for their help with TEM. [32] K. Huang, A.I. Isayev, Polymer 70 (2015) 290e306.
[33] V.V. Yashin, A.I. Isayev, Rubber Chem. Technol. 72 (4) (1999) 741e757.
References [34] J.N. Coleman, S. Curran, A.B. Dalton, A.P. Davey, B. McCarthy, W. Blau,
R.C. Barklie, Phys. Rev. B 58 (12) (1998) R7492eR7495.
[35] M.B. Isichenko, Rev. Mod. Phys. 64 (4) (1992) 961e1043.
[1] S. Iijima, Nature 354 (6348) (1991) 56e58. € tschke, D. Lellinger, T. Skipa, S. Pegel, G.R. Kasaliwal, T. Villmow,
[36] I. Alig, P. Po
[2] F. Mai, D. Pan, X. Gao, M. Yao, H. Deng, K. Wang, F. Chen, Q. Fu, Polym. Int. 60 Polymer 53 (1) (2012) 4e28.
(11) (2011) 1646e1654. [37] P. Po€tschke, A.R. Bhattacharyya, A. Janke, Eur. Polym. J. 40 (1) (2004) 137e148.
[3] X. Gao, S. Zhang, F. Mai, L. Lin, Y. Deng, H. Deng, Q. Fu, J. Mater. Chem. 21 (17) [38] P. Po€tschke, T.D. Fornes, D.R. Paul, Polymer 43 (11) (2002) 3247e3255.
(2011) 6401e6408. [39] P. Po€tschke, M. Abdel-Goad, I. Alig, S. Dudkin, D. Lellinger, Polymer 45 (26)
[4] J. Zhong, A.I. Isayev, K. Huang, Polymer 55 (7) (2014) 1745e1755. (2004) 8863e8870.
[5] M.T. Müller, P. Po €tschke, B. Voit, Polymer 66 (2015) 210e221.
[40] C.D. Han, M.S. Jhon, J. Appl. Polym. Sci. 32 (3) (1986) 3809e3840.
[6] J. Shui, M. Wang, F. Du, L. Dai, Sci. Adv. 1 (1) (2015). [41] E.R. Harrell, N. Nakajima, J. Appl. Polym. Sci. 29 (3) (1984) 995e1010.
[7] E.W. Wong, P.E. Sheehan, C.M. Lieber, Science 277 (5334) (1997) 1971e1975. [42] C.D. Han, J. Appl. Polym. Sci. 35 (1) (1988) 167e213.
[8] M.-F. Yu, O. Lourie, M.J. Dyer, K. Moloni, T.F. Kelly, R.S. Ruoff, Science 287 [43] J. Tian, W. Yu, C. Zhou, Polymer 47 (23) (2006) 7962e7969.
(5453) (2000) 637e640. [44] A.I. Isayev, T. Liang, T.M. Lewis, Rubber Chem. Technol. 87 (1) (2014) 86e102.
[9] K. Liu, Y. Sun, X. Lin, R. Zhou, J. Wang, S. Fan, K. Jiang, ACS Nano 4 (10) (2010) [45] J.C. Golba, J. Kunzelman, J.M. Spikowski, SPE ANTEC Paper, 2012.
5827e5834. [46] F. Du, R.C. Scogna, W. Zhou, S. Brand, J.E. Fischer, K.I. Winey, Macromolecules
[10] S. Berber, Y.-K. Kwon, D. Tom anek, Phys. Rev. Lett. 84 (20) (2000) 4613e4616. 37 (24) (2004) 9048e9055.
[11] W. Bauhofer, J.Z. Kovacs, Compos. Sci. Technol. 69 (10) (2009) 1486e1498. [47] D.G.H. Ballard, A.N. Burgess, P. Cheshire, E.W. Janke, A. Nevin, J. Schelten,
[12] M. Moniruzzaman, K.I. Winey, Macromolecules 39 (16) (2006) 5194e5205. Polymer 22 (10) (1981) 1353e1354.
[13] H.-J. Lee, S.-J. Oh, J.-Y. Choi, J.W. Kim, J. Han, L.-S. Tan, J.-B. Baek, Chem. Mater. [48] T.W. Ebbesen, H.J. Lezec, H. Hiura, J.W. Bennett, H.F. Ghaemi, T. Thio, Nature
17 (20) (2005) 5057e5064. 382 (6586) (1996) 54e56.
[14] S.J. Park, M.S. Cho, S.T. Lim, H.J. Choi, M.S. Jhon, Macromol. Rapid Commun. 24 [49] W. Ding, A. Eitan, F.T. Fisher, X. Chen, D.A. Dikin, R. Andrews, L.C. Brinson,
(18) (2003) 1070e1073. L.S. Schadler, R.S. Ruoff, Nano Lett. 3 (11) (2003) 1593e1597.
[15] K. Liu, L. Chen, Y. Chen, J. Wu, W. Zhang, F. Chen, Q. Fu, J. Mater. Chem. 21 (24) [50] R.J. Gardner, J.R. Martin, J. Appl. Polym. Sci. 24 (5) (1979) 1269e1280.
(2011) 8612e8617. [51] G.L. Pitman, I.M. Ward, R.A. Duckett, J. Mater. Sci. 13 (10) (1978) 2092e2104.
[16] H. Chen, C. Huang, W. Yu, C. Zhou, Polymer 54 (6) (2013) 1603e1611.