See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/286217836

Dispersion and stabilization of MWCNT in epoxy matrix using continuous

reactor

Article · January 2014

CITATIONS READS

0 152

4 authors, including:

Xiaobo Cheng Brian Greenhoe

Atlas Copco University of Southern Mississippi
137 PUBLICATIONS   1,190 CITATIONS    9 PUBLICATIONS   54 CITATIONS   

SEE PROFILE SEE PROFILE

Mohammad K. Hassan
Qatar University
107 PUBLICATIONS   1,389 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development of bionanocomposites based on PCL/PBS double crystalline blends and carbon nanotubes View project

Development of New Generation Electroless Ni-Based Nano-composite Coating for Oil and Gas Pipelines View project

All content following this page was uploaded by Mohammad K. Hassan on 28 January 2019.

The user has requested enhancement of the downloaded file.

 JOURNAL OF
POLYMER SCIENCE WWW.POLYMERPHYSICS.ORG
Universal Power Law Behavior of the AC Conductivity Versus
Frequency of Agglomerate Morphologies in Conductive Carbon
Nanotube-Reinforced Epoxy Networks
Brian M. Greenhoe,1 Mohammad K. Hassan,2 Jeffrey S. Wiggins,1 Kenneth A. Mauritz1
1
School of Polymers and High Performance Materials, the University of Southern Mississippi, Hattiesburg, Mississippi
2
Center for Advanced Materials, Qatar University, Doha, Qatar
Correspondence to: K. A. Mauritz (E - mail: kenneth.mauritz@usm.edu)
Received 5 April 2016; accepted 21 June 2016; published online 00 Month 2016
DOI: 10.1002/polb.24121
ABSTRACT: The Jonscher universal power law for ac conductivi-
ty versus frequency (f 5 x/2p) in the dispersion region was test-
ed for a multiwall carbon nanotube/epoxy nanocomposite. The
effect of changes in agglomerate morphology on the fitting
parameters A and n in the equation rac 5 Axn was investigat-
ed. Changing nanotube agglomerate morphology was tracked
by optical microscopy through curing. Evolving morphology
was compared alongside ac conductivity obtained via a broad-
band dielectric spectrometer to elucidate possible physical
meaning of the universal power law in the context of this sys-
tem. The 2logA/n was unaffected by changes in agglomerate
morphology affected during cure, yet connected with each oth-
er in their dependence on temperature. For this system, the
relationship between the fitting parameters in the universal
INTRODUCTION Electrical conduction through electrolytes is
characterized by log2log plots of r0 , the real part of the
complex conductivity (r* 5 r0 1 ir00 , where i 5 冑21) versus
frequency (f) as that illustrated in Figure 1 which shows
three distinct regions. In the low f range r0 decreases with
decreasing f. This is often attributed to the fluctuation of
interfacial polarization at the electrode–sample interface.1–4
In short, there is layered positive–negative charge separation
near the electrode which is primarily capacitive; conductivity
is limited in this region owing to the fact that the electric
field therein opposes the applied electric field. At higher f
there is a plateau of nearly frequency-independent conduc-
tivity which is referred to as the dc conductivity (rdc) which
is described later. The high frequency domain beyond this
plateau is the dispersive region characterized by frequency-
dependent conductivity which might reflect information on
the chemical environment within the material.5–9 The con-
ductivity proceeding from the plateau region and into this
dispersive region can be approximated by the empirical pow-
er law first described by Jonscher and referred to as the uni-
versal dynamic response as shown below:6,9
C 2016 Wiley Periodicals, Inc.
V
WWW.MATERIALSVIEWS.COM
FULL PAPERS
dynamic response equation remains empirical at this stage
with regard to biphasic “texture” or morphology within such a
network. Electrical conductivity r versus frequency x for a
composite consisting of agglomerated multiwalled carbon
nanotubes dispersed throughout a cured epoxy matrix was dis-
covered to follow the empirical universal dynamic response
equation of Jonscher. The frequency behavior of the exponent
n is discussed in terms of underlying morphology throughout
which charge carriers migrate. VC 2016 Wiley Periodicals, Inc. J.
Polym. Sci., Part B: Polym. Phys. 2016, 00, 000–000
KEYWORDS: universal conductivity power law; multi-walled car-
bon nanotubes; epoxy; morphology; broadband dielectric
spectroscopy
rac 5 Axn (1)
x 5 2pf is the angular frequency of the applied electric field
and A and n are fitting parameters of an empirical nature. n
is generally seen to fall between zero and unity although this
is not a requirement and examples for which n > 1 do
exist.10 n is also seen to be temperature dependent. Mauritz,
based on investigation of ion hopping in hydrated Nafion
membranes, suggested that n reflects the degree of connect-
edness of long range charge hopping pathways, or extent of
tortuosity for mobile charges.11
To date, no rigorous mathematical proof of eq 1 exists.
Equation 1 is seen to hold for different types of materials
including disordered semiconductors, conducting and semicon-
ducting polymers, conducting polymer composites, ceramics,
ion conducting glasses, and doped ionic crystals.
It has been considered as to whether there is a link between
A and n and whether these quantities have a mechanistic ori-
gin in terms of underlying disordered microscopic structure
on a given scale and a particular type of charge hopping
mechanism throughout this structure. Models have been
JOURNAL OF POLYMER SCIENCE, PART B: POLYMER PHYSICS 2016, 00, 000–000 1

FULL PAPERS WWW.POLYMERPHYSICS.ORG
FIGURE 1 Typical log–log real conductivity r0 versus frequency
relationship at a given fixed temperature for conductive disor-
dered materials.
proposed including the symmetric hopping model and the
macroscopic model, both discussed in a review by Dyre and
Schrøder.10 In general, the two models differ in the length
scale of the disorder relevant to conductivity. The details of
the mathematical analysis and underlying approximations
are too complicated and extensive to summarize here.
The symmetric hopping model assumes disorder of charge trans-
fer to be on a microscopic scale, presumably involving adjacent
moieties involved in short-ranged hopping of electrons or ions.
The vague macroscopic model assumes length scales on the
order of the correlation length of charge transfer moieties. This
model considers the distribution in the length of disordered
pathways accessible for electric charge flow which involves the
concept of percolation. Macroscopic conduction (from one end
of the sample to the other) will occur above a charge percola-
tion threshold. The size/shape of a percolation cluster is
dependent on morphology and affects nearest neighbor inter-
(and intra-, if possible) charge transfer contacts. There will be
dead ends on percolation clusters and poorly conducting
regions which impede conduction. Classical hopping over
potential energy barriers is considered, but also quantum
mechanical tunneling of delocalized electrons through these
barriers can occur as is likely the case of arrays of contacting
carbon nanotubes as those discussed here. Owing to consider-
able microstructural heterogeneity present in disordered solid
conductors, very broad distributions of jump/tunneling rates
and local charge mobility must be the case.
When f increases, sub-percolation clusters can become ac-
conductive as more rapid charge fluctuations, owing to the
shorter period of electric field reversals, can only allow
shorter overall charge displacements before the field
reverses. At increasingly lower f charge transport extends
over longer distances and direct current flow is approximat-
ed, accounting for the plateau on the graph in Figure 1.
Papathanassiou noted a relationship between A and n in that
2logA/n versus f was independent of disordered material
2 JOURNAL OF POLYMER SCIENCE, PART B: POLYMER PHYSICS 2016, 00, 000–000
JOURNAL OF
POLYMER SCIENCE
composition at all frequencies up to one terahertz.12 Further-
more, 2logA/n is nearly constant and consistent between a
wide range of materials.12,13 This finding contrasted with
theories which implied physical meaning behind these quan-
tities beyond their being empirical fitting parameters. There
was a subsequent study of an investigation of power law
behavior in a mixed crystal compound in which A and n
were sensitive to structural transitions within the crystal
upon increasing temperature. It was found that even through
multiple physical transitions the relationship between
2logA/n again remained constant.14 Interestingly, numerical
modeling of this power law suggests that this universal
response is just that of a random mixture of conductors and
capacitors in an equivalent circuit representation.15
To date, there has not been established an interconnection of
A and n with physical meaning at the microscopic level. The
purpose of this article is to not only test the application of
this power law to the curing of glassy epoxy/nanotube nano-
composites but also consider possible meaning that the rela-
tionship A between n may have with respect to microscopic
level disorder of charge motions in this particular system.
EXPERIMENTAL
The following materials were used as received: 4,40 -diamino-
diphenylsulfone (44DDS, Royce Chemical Corp, Mn 5 248.3g
mol21); tetraglycidyl-4,40 -diaminodiphenylmethane (TGDDM,
Huntsman, Mn 5 422.5g mol21); SMW200 purified MWCNTs
(Multi-Walled Carbon Nanotubes) from SouthWest NanoTech-
nologies (length 31lm, diameter 5 10 nm.) The density of
MWCNT after incorporation into epoxy matrices is approxi-
mately 1.75g cm23.16 The epoxy matrix for this work was
formulated at 1:1 stoichiometric equivalents of epoxide to
active amine hydrogen, and prepared using twin screw
extrusion described elsewhere.17
Nanocomposite Curing and Optical
Microscopy Characterization
Optical microscopy (OM) in transmission mode was per-
formed using a Nikon OPTIPHOT2-POL polarizing micro-
scope equipped with a METTLER FP82 HT hot stage (HS) to
track agglomerate morphology changes evolved throughout
cure. OM specimens, approximately 50 microns thick, were
compressed between two glass slides and inserted into the
HS. The HS was programmed with the same cure cycle that
was used on all the specimens in this study (25–180 8C at 1
8C min21, 180 8C isothermal hold for 3 h). An OLYMPUS
UC50 color camera recorded the evolution of nanotube
agglomerate dispersion states throughout cure. OLYMPUS
Stream Essentials [Version 1.7] image analysis software was
used to process the OM images.
Conductivity Measurements Using a Broadband
Dielectric Spectrometer
Complex conductivity versus f data were collected using a
Novocontrol GmbH Concept 80 Broadband Dielectric Spec-
trometer (BDS). Frequency sweeps were carried out from
0.01 Hz to 3 MHz over the temperature range of 25–180 8C
 JOURNAL OF
POLYMER SCIENCE WWW.POLYMERPHYSICS.ORG FULL PAPERS

FIGURE 2 Optical microscope (OM) images illustrating nanotube dispersion states (a) before and (b) after cure for a 0.2% wt/wt
44DDS-TGDDM nanocomposite cured at 25–180 8C at 1 8C min21 and 180 8C soak for 3 h.

using a nitrogen gas-controlled heating system (Quatro Cryo- The particular morphology observed in the cured state,
system Novocontrol). This particular temperature ramp was apparent in the 180 8C image, is considered to be a network
chosen specifically to mimic the cure program used in col- of MWCNTs. It is this network formation that drives the
lecting the OM images. Frequency sweeps were carried out increase in electrical conductivity observed as the emergence
when temperature stability of the instrument was within 6 of a frequency-independent conductivity plateau at 50 8C in
0.2 8 C. Samples were tested in a custom stainless steel liq- the r0 versus f plot and an extension of this plateau into
uid parallel plate sample cell equivalent to the Novocontrol higher and higher frequencies with increasing cure as illus-
BDS 1308. Samples were sandwiched between stainless steel trated in Figure 3.
electrodes of 2 cm diameter with a silica spacer to maintain
sample thickness at 0.9 mm. The data in Figures 3 and 4 were collected for consecutive
frequency sweeps conducted at different temperatures dur-
RESULTS AND DISCUSSION
ing a ramp of increasing temperature on a single sample.
The r0 and e00 were determined during the temperature ramp
Optical microscopic images of the nanocomposite morpholo- from 25 to 180 8C.
gy before (a) and after (b) cure are in Figure 2. In these
images the epoxy matrix is a light colored background while The curves show the same three distinct regimes exhibited by
the MWCNT agglomerates appear black. other disordered conductive solids as depicted in Figure 1: (1)
a low frequency upturn at lower temperatures, (2) a plateau
The before/after cure images illustrate the transition of mor- region, and (3) a “dispersion” region at the highest tested
phology from that of small, dispersed bundles of tubes rang-
ing in size from 50 lm down to individually dispersed tubes
(via transmission electron microscopy imaging), to a perco-
lated network visible at the micron length-scale. This
agglomeration is facilitated by a drop in matrix viscosity—
which allows bundles to come in contact with each other
affected by heating prior to gelation. Agglomeration of
MWCNTs during cure is widely reported and well under-
stood. Nanotubes agglomerate due to attractive London
forces and p-bonding. Starting from a dispersed state,
MWCNTs encounter each other while undergoing Brownian
motion in the low viscosity (high mobility) state at elevated
temperatures during epoxy cure. Afterward, these motions
are frozen out of relevant time-scales at gelation which for
this system occurs at 170 8C. Agglomerate morphology
evolves in a continuous and progressive manner during cure.
This percolated MWCNT network initially becomes apparent
during the 1 8C min21 ramp between 25 and 50 8C, a pro-
cess which can be observed with the OM with a hot stage
which provides temperature control through cure. FIGURE 3 r0 , the real part of the complex conductivity
(r* 5 r0 1 i r00 ; i 5 冑21), versus frequency through 1 8C min21
One can imagine conductive pathways along contacting ramp, on the same sample, to 180 8C for nanocomposite
MWCNT agglomerates that form throughout cure. This differ- epoxy. [Color figure can be viewed in the online issue, which
ence is most apparent in the images of 25 and 180 8C. is available at wileyonlinelibrary.com.]

WWW.MATERIALSVIEWS.COM JOURNAL OF POLYMER SCIENCE, PART B: POLYMER PHYSICS 2016, 00, 000–000 3
 JOURNAL OF
FULL PAPERS WWW.POLYMERPHYSICS.ORG POLYMER SCIENCE

nature; this is a common artifact owing to the presence of

ionic impurities, in dielectric spectra.1–4,19 Traces of the
NaOH catalyst used in the synthesis of TGDDM may contrib-
ute to this phenomenon to some extent.24

The crosslinking reaction is expected to be activated above

100 8C. Crosslinking would result in less chain segmental
mobility which would restrict charge hopping causing, in
turn, a lower degree of fluctuating positive–negative defect
charge polarization in the near-electrode regions.

The solid curves superimposed on r0 versus f data points are

R
fits of eq 1 using OriginV software. These fits are rather good
as indicated by high correlation coefficient (R2) values listed in
Table 1. Fitted rdc, A, and n values are listed in Table 1, as well.
As cure progresses there is an increase in rdc. The parameter n
decreases as conversion increases with temperature through
cure. The general expectation in a nondynamic (constant tem-
FIGURE 4 e00 versus frequency, for the same sample in Figure 3, perature) situation is that n ! 1.0 as T ! 0, 11 but the compli-
at different temperatures during ramp showing the b relaxation. cating influence of evolving morphology as well as electron
[Color figure can be viewed in the online issue, which is avail-
versus ion hopping must be factored in. The temperature
able at wileyonlinelibrary.com.]
dependence of A is not as well understood, but in the case of
the curing seen here, it is observed to increase directly with
frequencies. The frequency dependence of conductivity in the temperature and conversion.
dispersion region for the epoxy-matrix nanocomposite, at vari-
ous temperatures throughout cure, was compared against eq 1 As discussed in the Introduction, 2logA/n has been reported
which was fitted to the data. to be insensitive to material composition or structural transi-
tions that occur on increase in temperature. For the system
The frequency-independent plateau sections on the curves in reported here 2logA/n versus temperature is seen in Figure
Figure 3 correspond to a dc electrical conductivity, rdc.18 5 in which this ratio indeed stayspractically constant. This
The sections widen and elevate with increasing temperature occurs despite the shifts in both A and n throughout the tem-
indicating increasing conductivity over an increasingly perature ramp as seen in the inset of Figure 5. The tempera-
broader range of frequency. ture ramp causes the material to go through a process of
MWCNT re-agglomeration starting as an array of largely dis-
In Figure 3, at higher f beyond the dc range, are dispersion
connected small nanotube agglomerates seen in Figure 2(a)
regions that appear most distinct on the plots for 50 and 75
and evolving to the percolated disordered network of tubes/
8C. The corresponding e00 versus f plots in Figure 4 show a b
bundles seen in Figure 2(b). This re-agglomeration process is
relaxation, similar to that seen in BDS studies of similar well documented and is caused by a drop in viscosity during
materials, which shifts to the right and then moves off scale cure and more effective attractive forces between nanotubes.
with increasing temperature.19,20 The b relaxation, based on
the results of former dielectric studies of similar epoxide res- If n were linked to particular details of the biphasic mor-
ins with different curing agents, is assigned to local motions phology there should be an obvious trend in it as influenced
of dipoles, specifically the hydroxyl ether and amine func- by phase morphology change. In particular, one would expect
tionalities formed during the crosslinking reaction.21–23 n to move closer to unity as the material exceeds a
While these prior studies are relevant to those reported
here, the MWCNT/epoxy system is more complex in that TABLE 1 Parameters From Fit of r0 (x) 5 rdc 1 Axn to Experi-
conductivity effects overwhelm polymer sidechain relaxa- mental Data for 0.2% wt/wt Nanotube-Loaded 44DDS-TGDDM
tions into high f regions as seems obvious in Figure 4. Inter- Nanocomposite During Cure
nanotube electron transfer on a local level and degree of
MWCNT connectedness on a large scale would seem to domi- Temperature (8C) rdc (S cm23) n A R2
nate, and more so with increasing temperature. 25 4.46212 0.69 4.41215 0.946
29 214
This conductive behavior is in fact typical for disordered 50 1.54 0.60 1.89 0.977
materials and is consistent with the 75 2.2827 0.49 8.08212 0.973
100 2.9226 0.32 1.7627 1.000
Jonscher power law represented by eq 1. Small deviations in 125 1.41 25
0.30 4.0527
0.999
the curve fittings to the low f end in the r0 versus frequency
150 4.8925 0.27 7.1927 0.993
data might be attributed to sample/electrode interfacial
180 6.1425 0.34 8.1628 0.978
polarization effects which are capacitive (current-limiting) in

4 JOURNAL OF POLYMER SCIENCE, PART B: POLYMER PHYSICS 2016, 00, 000–000

 JOURNAL OF
POLYMER SCIENCE WWW.POLYMERPHYSICS.ORG
FIGURE 5 2logA/n versus temperature for 0.2% wt/wt 44DDS-
TGDDM nanocomposite cured at 25–180 8C at 1 8C min21, 180
8C soak for 3 h. Inset is—logA and n versus temperature.
percolation threshold between 25 and 50 8C. However, n, as
noted, is seen to decrease with increase in temperature,
excepting the value at 180 8C.
The 2logA versus n plot in Figure 6 illustrates the direct
tracking of these quantities such that the temperature evolu-
tion of logA is proportional to the temperature evolution of
n. While a straight line (R2 5 0.9825) has been fitted to the
data points, a sigmoidal curve might be considered but this
would add another parameter which would complicate the
analysis. Due to the fact that A and n show a steady linear
correlation, agglomerate morphology would appear to be
independent of trends or values of these two parameters.
This result reinforces the idea that this response is likely
just a general feature of the universal power law that is not
linked to particular substructure or changes within as stated
FIGURE 6 2log A versus n. Line was best-fit to experimental
data points.
WWW.MATERIALSVIEWS.COM
FULL PAPERS
by others.13,14 One obtains the same general relationship for
a system that does not undergo structural evolution.
CONCLUSIONS
The conductivity versus frequency behavior of a multi-
walled carbon nanotube-loaded 44DDS-TGDDM epoxy nano-
composite, during cure, was investigated in the 0.01 Hz–3
MHz range using broadband dielectric spectroscopy. Nano-
tube agglomerate morphology of these same samples was
imaged by optical microscopy before and after cure. Conduc-
tivity versus frequency in the dispersion region is well repre-
sented by Jonscher’s universal power law rac 5 Axn.
Moreover, it was determined that 2logA/n versus tempera-
ture remains essentially constant despite shifts in both A
and n throughout the temperature ramp to cure, during
which nanotube agglomerate morphology changes occur. Nei-
ther A nor n appear to be related to the biphasic morpholo-
gies of these up to and including formation of percolated
MWCNT networks during cure. It is still an open question as
to the link between A and n with microscopic quantities and,
if any, why the power law applies to different materials and
charge transport mechanisms.
ACKNOWLEDGMENTS
The authors gratefully acknowledge Joycelyn Harrison, AFOSR
Award Number FA9550-13-1-0103; and Gregg Bogucki and
Stephen Heinz from Boeing Research and Technology for their
kind financial support and collaboration.
REFERENCES AND NOTES
1 I. Stefanithis, K. A. Mauritz, Macromolecules 1990, 23, 2397.
2 H. Chen, M. K. Hassan, S. K. Peddini, K. A. Mauritz, Eur.
Polym. J. 2011, 47, 1936.
3 R. J. Klein, S. Zhang, S. Dou, B. H. Jones, R. H. Colby, J.
Runt, J. Chem. Phys. 2006, 124, 144903.
4 P. Atorngitjawat, J. Runt, Macromolecules 2007, 40, 991.
5 L. L. Hench, J. K. West, Principles of Electronic Ceramics,
Wiley: New York, 1990.
6 A. K. Joncher, Thin Films 1980, 11, 232.
7 A. K. Jonscher, J. Mater. Sci. 1981, 16, 2037.
8 M. D. Ingram, Phys. Chem. Glasses 1987, 28, 215.
9 A. K. Jonscher, Nature 1977, 267, 673.
10 J. C. Dyre, T. B. Schrøder, Rev. Mod. Phys. 2000, 72, 873.
11 K. A. Mauritz, Macromolecules 1989, 22, 4483.
12 A. Papathanassiou, J. Non-Cryst. Solids 2006, 352, 5444.
13 C. Cramer, S. Brunklaus, E. Ratai, Y. Gao, Phys. Rev. Lett.
2003, 91, 266601.
14 A. Papathanassiou, J. Phys. Chem. Solids 2005, 66, 1849.
15 D. P. Almond, C. Bowen, Phys. Rev. Lett. 2004, 92, 157601.
16 M. S. P. Shaffer, A. H. Windle, Adv. Mater. 1999, 11, 937.
17 B. M. Greenhoe, X. Cheng, A. Sharma. J. S. Wiggins, Car-
bon, in press.
18 W. Lee, B. Lim, J. Liu, A. Nowick, Solid State Ion. 1992, 53,
831.
JOURNAL OF POLYMER SCIENCE, PART B: POLYMER PHYSICS 2016, 00, 000–000 5

FULL PAPERS WWW.POLYMERPHYSICS.ORG
19 M. K. Hassan, S. J. Tucker, A. Abukmail, J. S. Wiggins, K. A.
Mauritz, Arabian J. Chem. 2016, 9, 305.
20 C. H. Childers, M. K. Hassan, K. A. Mauritz, J. S. Wiggins,
Arabian J. Chem. 2016, 9, 206.
21 M. Ochi, H. Iesako, M. Shimbo, J. Polym. Sci. Part B: Polym.
Phys. 1986, 24, 1271.
6 JOURNAL OF POLYMER SCIENCE, PART B: POLYMER PHYSICS 2016, 00, 000–000
View publication stats
JOURNAL OF
POLYMER SCIENCE
22 M. Shimbo, M. Ochi, H. Iesako, J. Polym. Sci. Polym. Phys.
Edition. 1984, 22, 1461.
23 M. Mangion, G. Johari, J. Polym. Sci. Part B: Polym. Phys.
1990, 28, 1621.
24 M. Fedtke, in Handbook of Polymer Synthesis: Part B; Krichel-
dorf, H. R., Ed.; CRC Press: Boca Raton, 1991; Ch. 23, p. 1492.