Materials Chemistry and Physics 191 (2017) 89e95

Contents lists available at ScienceDirect

Materials Chemistry and Physics

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

Modeling the dynamic percolation of carbon nanotubes and revisiting

critical exponents
M. Badard a, A. Combessis b, A. Allais b, L. Flandin a, *
a
Univ. Savoie Mont Blanc, LEPMI, F-73000 Chamb
ery, France
b
Nexans Research Center, 29 rue Pr

e Gaudry, 69353 Lyon, France

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Percolation threshold and critical

exponents are determined with an
unmatched accuracy.

The critical exponent strongly varies
during dynamic percolation of car-
bon nanotubes.

A correlation between critical expo-
nent and particles dispersion is
established.

A new model depicts conductivity as
a function of time and ller content.

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

Article history: In this paper, dynamic percolation of carbon nanotubes in silicone oil was probed by means of electrical
Received 30 May 2016 measurements. Carbon nanotubes were rst dispersed in silicone oils with ultrasounds. The obtained
Received in revised form suspension was then slowly diluted. After each dilution, the changes in electrical properties was
11 January 2017
monitored over time. This procedure enabled to collect a large number volume fraction and dene both
Accepted 14 January 2017
Available online 18 January 2017
the percolation threshold and the critical exponents with unmatched accuracy. The percolation threshold
was found to decrease with time as a consequence of the already described aggregation of conductive
nanoparticles. More surprisingly, the critical exponent also largely varied over time. This changes were
Keywords:
Carbon nanotubes
correlated to the quality of the dispersion of carbon nanotubes within the host matrix. Finally, statistic
Polymers percolation model was adapted into one that take into account for conductivity changes as a function of
Electrical properties time and ller content.
Modeling 2017 Elsevier B.V. All rights reserved.

1. Introduction orders of magnitude. In the case of 3D randomly dispersed systems,

The electric properties of dielectric materials lled with
conductive particles are known to depend strongly on the ller
fraction and their arrangement within the host matrix [1e6]. The
formation of a conductive path occurs at a critical ller content,
well known and is usually described as a percolation threshold
[7,8]. Above this amount of ller, the conductivity increases by
* Corresponding author.
E-mail address: lionel.andin@univ-smb.fr (L. Flandin).
the percolation transition was modeled with great success as a
power law by Kirkpatrick [9]:
s s0  V  Vc m (1)
with s the conductivity, s0 the theoretical macroscopic conduc-
tivity of the ller, Vc the critical ller fraction, V the ller content
and m the critical exponent. This model has been widely employed
by researchers to describe the conductivity changes of composite
materials with increased ller content.

http://dx.doi.org/10.1016/j.matchemphys.2017.01.043
0254-0584/ 2017 Elsevier B.V. All rights reserved.
90 M. Badard et al. / Materials Chemistry and Physics 191 (2017) 89e95
Most of experimental measurements and numerical simulations
led to a critical exponent m close to 2 for 3D random systems;
therefore assumed to represent a universal value [10]. However,
scattered values from 1 to 6 have also been reported for m in a wide
variety of materials [11,12]. These results led to countless discus-
sions on the universality of critical exponents. Several models and
theories were in particular proposed to explain non-universality in
polymer matrices mixed with carbonaceous llers [11e22]. Though
the origin of non-universality seems hard to determine, all agree
that a physical meaning is hidden behind the critical exponent.
Non universal values may be explained with the random resistor
network theory. Proposed by Kogut and Straley [13], this approach
consists in imagining the llers network as a distribution of re-
sistances in the percolating network. This induces critical expo-
nents higher than 2 especially if the distribution is broad. The
Random-Void (RV) model, developed by Halperin et al. [14], con-
sists in conducting media lled with void areas. Electron transport
is achieved through a path between insulating spaces. This model
predicts a critical exponent close to 2.5.
The non-universality might originate from anisotropy, but Cel-
zard and coworkers clearly showed that universality could be ob-
tained as well in anisotropic systems [16,17].
Non-universal values have been observed for angular particles
and porous media by McLachlan et al. [18e22].
Vionnet-Menot [12] observed that high critical exponent values
were usually found for composites with low percolation thresholds.
In these systems, it was assumed that there is no direct contact
between conductive particles. Electron transport is thus likely to
result from tunneling and appears to govern non-universal
values.
The application of thermal annealing on composites lled below
the percolation threshold may lead to a restructuration of isolated
particles into an innite path. This formation of a conductive
network with time is dened as dynamic percolation [23e29].
Recently, our research group [30e32] studied the changes in critical
exponent values of composites submitted to dynamic percolation.
A strong variation from 2 to 6 of the critical exponent during
annealing was reported. These results were ascribed to ller ag-
gregation processes in the melt. Then, non-universal values could
be attributed to aggregated systems, in which the dispersion and
distribution of conductive particles are not stochastic anymore.
Dynamic percolation was studied as a way to reduce the amount
of ller necessary to achieve a conductive material. Many in-
vestigations described the mechanisms responsible for this process,
however, only few scientists proposed quantitative models
[33e37]. The following models from the literature are all based on
Kirkpatrick's equation (cf. Eq. (1)), which was adapted with a ki-
netics term. In those models, the ller content V corresponds to the
concentration of conducting ller truly participating in the perco-
lating path. To describe the process of ller aggregation, kinetics
equations were chosen by the authors. V is then substituted in Eq.
(1) by an equation of the rst or second order. Cao et al. [34,35]
presented a model based on a rst order kinetics:
dNt
KNt (2)
dt
where Nt corresponds to the number of clusters over time and K is
the reaction rate. Integrating Eq. (2) leads to Eq. (3):
Nt cst  expKt (3)
Assuming that the fraction of clusters Vt in the composite is
proportional to the number of cluster ( Vt fnt ), the authors
adapted Eq. (3) into Eq. (4)


Vt V  V  V0  expKt (4)
V and V0 are respectively the fractions of percolated clusters at
long and initial time. Combined with Eq. (1) results in:
h

  im
st s0  V  V  V0  eKt  Vc (5)
As mentioned by Cao et al. [34], Vc should be substituted by
another term Vc , corresponding to the minimum ller content
required to achieve the percolation network. Alig and Skipa [33,36]
proposed a similar model based on a kinetics of aggregation of the
second order:
st s0
2 0 1 3m

 1
 V0 V V0  @1
4
A Vc 5
14kt V V0
(6)
Later, Kruckel [37] proposed a more detailed model based on
three distinct aggregation steps. In the rst step, particles are
dispersed in the matrix, the second one represents clusters not
connected and the third one represents the formation of a perco-
lation path. Calculations based on an equilibrium between three
kinetics equations led to Eq. (7). Details of calculation can be found
in the original paper [37]:
~
st s1  Aexpat ~
 Bexpk
2 t (7)
In all these models, the critical exponent is arbitrarily set to the
universal value. However, it was demonstrated [38] that the latter is
altered by the aggregation process. This could lead to some mis-
matches between the proposed models and experimental data.
In the present paper, a new experimental process allowed us to
accurately determine the critical exponent m. A detailed investiga-
tion of the standard deviation on this value was performed in order
to reduce the error and reveal meaningful differences over perco-
lation time. The method was then applied to scout changes in the
critical exponent over time. Finally, a model was proposed that
quantitatively describes the measured conductivity as a function of
ller content and time.
2. Experimental
2.1. Materials
The study was realized with a silicone oil, 47-V100, from Blue-
star Silicones. The matrix was lled with multiwall carbon nano-
tubes, N7000, from Nanocyl, synthetized by chemical vapor
deposition and characterized by a length of 1.5 mm and a diameter
of 10 nm (aspect ratio ~150).
2.2. Suspension preparation
Carbon nanotubes were dispersed in silicone oil by ultra-
sonication for 5 min. To do so, a 750 Watt ultrasonic probe from
Bioblock scientic was used. In order to prevent the suspension
from heating during sonication, the latter was cooled in an iced
bath during the process. The initial mother suspension contained
0.048 vol % of CNT, should be taken well above all the measured
percolation thresholds in these systems (0.01 vol %).
 M. Badard et al. / Materials Chemistry and Physics 191 (2017) 89e95
2.3. Suspension dilution
Percolation transitions are well known by experimentalists for
being difcult to probe, mainly because of the susceptibility of the
tested systems in the vicinity of the threshold, combined to the
challenge that represents processing sample series with a contin-
uum of ller fractions. In the case of liquid media, we developed a
way to gain signicant accuracy in the determination of both the
threshold and the critical exponents. Starting from a suspension
containing an amount of dispersed CNTs well above the percolation
threshold, a slow and controlled dilution allows to measure a large
number of volume fraction with extreme accuracy. The volume
fraction can be carefully varied by steps down to a single drop
(0.05 mL) over the total volume of the suspension (~tens of milli-
liters). This method essentially consists in dosing the percolation
transition.
A mother suspension containing 0.048 vol % of CNT was slowly
diluted at room temperature. The suspension was put in a beaker
with a 4 cm bar magnet. The beaker is positioned on a magnetic
stirrer RCT basic IKAMAG. During dilution (around 3 min), the
suspension was homogenized by mechanical stirring. Due to the
low viscosity of the matrix, a proper agitation is reached instan-
taneously. After the agitation is stopped, the change of conductivity
is recorded as a function of time for 750 s. A frequency sweep of 17 s
furnished the times interval between two measurements, a total of
44 data Fig. 1-A. Each ller content was measured at least three
times, with a homogenization step by mechanical stirring between
each measurement. Experimental measurements were dis-
continued after 700 s because llers aggregation reaches a
macroscopic scale, a rst step towards sedimentation [31]. Inter-
estingly the percolated network is settled after about 100 s, but the
llers continue to aggregate around. As a result m continues to
decrease for a much longer time (700 s and probably more).
2.4. Electrical characterization
Conductivity was recorded using a in-house cell constituted of 2
parallel electrodes (area: 4.5 cm2, gap: 2 mm) and connected to a
Modulab Materials Test System from Solartron Analytical. Elec-
trodes are immersed in the silicone oil, as depicted in Fig. 1-B. No
temperature variation has been noticed during measurement,
because the Joules effect could be neglected with these low cur-
rents and tensions. The set of data thus allowed to track the changes
in conductivity as a function of time and ller content.
Fig. 1. A- Overall pattern of the different steps performed in this paper. B- Apparatus
employed for the electrical measurement steps, the dilution steps and the mechanical
stirring steps.
91
3. Results
3.1. Determination of the critical exponent
The difculty to determine critical exponent m with condence
has rarely been discussed in the literature. A simple numerical
analysis however reveals that a slight error in the determination of
Vc induces important errors on the exponent value. This also might
explain some non-universal values reported in the literature. Let's
take a classic percolation plot (Fig. 2) with a critical exponent
determined through the least-squares tting method and let's
imagine only the percolation threshold is varied.
The correlation between the error on m and the error on Vc is
further depicted on Fig. 3. Experimental data were simulated with
threshold and critical exponent values respectively set to 0.5% and
2. Least-squares tting method was used with 3 values for Vc :
0.47%, 0.5% and 0.53%. This exercise demonstrates that a slight error
of 0.03% on Vc yields to major differences on the critical expo-
nents (which were in this case found at 2.5, 2 and 1.5 respectively).
An estimation of the error on critical exponent versus percola-
tion threshold is also plotted in Fig. 4. It results that to gain accuracy
on the critical exponent, reducing the error on Vc is mandatory. In
other words a large amount of data is required in the vicinity of Vc .
The process based on the dilution presented in this paper revealed
extreme efciency in the determination of Vc and m with a great
accuracy, because it provides numerous values in the transition
zone.
3.2. Change of critical exponent over percolation time
The mother suspension of CNT in silicone oil was diluted step by
step, and electrical properties were measured over time for each
corresponding ller content. The conductivity was found constant
over time with a constant agitation. Fig. 5 presents conductivity
versus ller content for three times: under agitation, 17 s after
agitation was stopped, and 700 s after agitation was stopped.
From Fig. 5, for ller contents from 0.02 to 0.0048 vol %, the
conductivity of the tested suspension increases over time by orders
of magnitude. This transition is attributed to dynamic percolation:
when agitation was stopped, carbon nanotubes started to
agglomerate. From a time value, which depends on the ller
Fig. 2. Least-squares tting method was applied to t Kirkpatrick's equation in order
to determine the percolation threshold in silicone oil lled with CNT, 17 s after stirring
was stopped. From classic percolation results, logarithm of conductivity is plotted
versus logarithm of V  Vc , the critical exponent corresponds to the slope (see inlet).
92 M. Badard et al. / Materials Chemistry and Physics 191 (2017) 89e95
Fig. 3. Least-squares tting method used with 3 Vc : 0.47%, 0.5% and 0.53%. By changing
the percolation threshold by only 6%, the critical exponents varied by 25% from 1.5 to
2.5.
Fig. 4. Standard error of exponent m versus standard error of Vc : a slight error on Vc
leads to an important error on the critical exponent. We consider that an error in the
range of 0.2 is acceptable on the value of the critical exponent as it would corresponds
to a threshold determined with an error below 2%.
content, a percolation path is created throughout the silicone oil.
Below 0.0048 vol %, there never was enough carbonaceous particles
to form a percolation path; above 0.02 vol %, the network is always
present, even under magnetic agitation.
Parameters used in Kirkpatrick's equation (Eq. (1)) to t the
experimental data are given in Table 1:
The changes in both percolation threshold and critical exponent
as a function of time are represented on Fig. 6. The decrease of the
percolation threshold results from dynamic percolation and could
be compared, to a certain extent, to what happens in polymer melts
[25,28,30].
The value of the critical exponent seems time dependent too:
under agitation low critical exponent values in the range of 0.5
were measured. The later gradually increased with time until
reaching 2.23, close the universal value, 17 s after agitation was
stopped. For longer times, the critical exponent decreased with
time to a nal value of 1.4 after 700 s.
An increase of m has already been observed in the literature for
Fig. 5. Conductivity versus CNT content measured for 3 time values after the agitation
was stopped. Experimental data (scatters) were adjusted with percolation equation
(lines).
dynamic percolation [30] of CNT in polyethylene held at 200  C for
24 h. This increase in the critical exponent is similar in trend to the
one observed for short times in silicone oil. The large difference
between the two time values may be explained by the orders of
magnitude of viscosity difference between a molten polymer and
silicone oil [39].
The decrease in Vc value over time is due to the ease of CNT to
aggregate in silicone oil. As illustrated in Fig. 7, under agitation the
critical exponent is measured below the universal value. This may
be explained by a non-statistic organization of CNT due to the kind
of the agitation (magnetic-stirring): llers could be centrifuged or
be oriented in the direction of the ow. Flow orientation of CNT
have already been observed in extruded polymer composites [28].
In these systems, the anisotropy was observed to vanish with
annealing time. After stopping the agitation, m increases from 0.5 to
2.7 in 35 s. The set of measured parameters could indicate that the
percolation prole obtained for short times after agitation corre-
sponds to a 3D random system. For short times after agitation is
stopped, llers are likely to get well dispersed and distributed in a
statistical and homogeneous fashion macroscopically and micro-
scopically: consequently the critical exponent tends to the univer-
sal value. For times longer than 35 s, critical exponent decreases
over time. We assume that the aggregation of CNT into clusters is
likely to break the random distribution. The time variation of m is
slower than that of the percolation threshold: the dynamic perco-
lation occurs in the rst 100 s and reaches quickly Vc, the min-
imum threshold for such system. However, while the percolated
network is settled, llers continue to aggregate around: this leads
to a decrease in mu for 700 s and more.
To sum up, CNT move from an articial distribution state gov-
erned by the agitation to an aggregated one, thermodynamically
more stable. The random distribution observed 17 s after the
agitation is stopped does not correspond to a stable conguration.
3.3. Modeling
Kirkpatrick's statistic percolation model (cf. Eq. (1)) was estab-
lished to describe the conductivity change versus ller content. In
the present case, percolation of our systems depends on time. We
thus tried and adapted Kirkpatrick's model implementing the time
dependence of Vc and m. In other words, our model would describe
 M. Badard et al. / Materials Chemistry and Physics 191 (2017) 89e95 93

Table 1 With Vc0 the value of Vc under agitation and Vc , the value of
Parameters used to t the results presented on Fig. 5. Vc at long time (~700 s). Vc variations are directly correlated to the
Agitation 17 s 700 s dynamic percolation rate t: after stopping the agitation, high ller
s0 (S.m1) contents (just below 0.02 vol %) form a percolation path in very
4:105 1:5:106 4:35:101 7:102 5:103 6:105
m 0.5 0.1 2.32 0.16 1.41 0.14 short time. Due to the shape ratio of CNT, only few contacts are
Vc vol % 0.02 0.0005 0.0087 0.0005 0.0048 0.0005 necessary to achieve percolation. In contrast, for low ller contents
(just above 0.0048 vol %), more time is needed by the nanoparticles
to self-organize into an innite cluster.
Substituting Eq. (8) into Kirkpatrick's model (Eq. (1)) yields to:
 


 tt mt
sV;t s0  V  Vc  Vc  Vc0  e (9)

This model is based on experimental measurements done on

silicone oil lled with CNT. Some parameters, such as Vc , Vc0
and t, would probably differ considering other ller-matrix
couples.
A good correlation was found between our model and experi-
mental data. Results are respectively displayed versus ller content
on Fig. 5 (dotted line) and versus time on Fig. 8.

3.4. Modeling and the time dependence of m

Fig. 6. Critical exponent m and percolation threshold versus time. A good correlation
was found between Vc data and Eq. (8) (dashed line) with the following parameters
Vc 0:0048 vol %, V0 0:02 vol % and t 15 s.
In this part, Eq. (9). is benchmarked with models of Cao [34] and
Alig [33,36], that are respectively based on rst and second order
kinetics of CNT aggregation (cf. Eq. (5) and Eq. (6)). In contrast with
previous studies, we decided to avoid the explicit time dependence
in the conductivity. In addition we made no assumption on the
overall amount of ller. It was kept constant. Following the
experimental evidences, we considered the percolation threshold
Vc as the time varying parameter.
Another major difference with former models lies in the critical
exponent value. In Cao's and Alig's papers, the latter was xed to
the universal value. In other words, the distribution of ller was
imagined random during the dynamic percolation process. This
assumption contradicts the observation of the aggregation process.
Concerning the critical exponent m, we assumed it reects a
dispersion state and may vary with time. This is directly supported
by the experimental evidences.
The conductivity change with time for a silicone oil lled with
0.0068 vol % of CNT was tted with all the models (Fig. 9). For all
models, the initially percolated ller V0 is 0 vol %. The percolated

Fig. 7. Illustration of critical exponent m and CNT structuration over time. Dynamic
percolation process can be divided in three phases. Under agitation, llers are isolated
and not in a statistic conguration. Consequently, m is different from universal value
and Vc is high. Right after agitation was stopped, CNT are randomly distributed and
ller start establishing connections between them: m increases until universal value
and Vc decreases by a factor 2. Beyond 35 s, aggregation continues to a larger scale.
Thus the random distribution of CNT is lost and the critical exponent m wander from
the universal value.

the conductivity dependence to time and to ller content.

As it was mentioned in the introduction, for Cao [34] ller ag-
gregation follows a rst order kinetics (Eq. (2)). Integrating Eq. (2)
leads to an exponential relationship (Eq. (3)).
The time dependence of the percolation threshold appears to
follow an exponential behavior (Fig. 6) and was thus compared to
(Eq. (8)).
  Fig. 8. Conductivity dependence to time plotted for 2 ller contents: 0.0068 vol % and

 tt 0.0082 vol %. Scatters correspond to experimental results, lines correspond in cf. Eq. (9)
Vct Vc  Vc  Vc0  e (8) with m time dependent and dotted line correspond to the same model with m set to
universal value.
94 M. Badard et al. / Materials Chemistry and Physics 191 (2017) 89e95
Fig. 9. Experimental conductivity (scatters) and simulated data (lines) versus time for silicone oil lled with 0.0068 vol % of CNT. Left: rst order kinetics model; Right: second order
kinetics model. For full lines models, critical exponent is time dependent whereas for dotted lines models, critical exponent is set to the universal value.
ller at innite time V is 0.0068 vol % (all CNT are participating to
conducting path). Vc is the lowest ller content necessary to
form a percolation path at innite time. Its value was determinated
to be 0.0048 vol % for this system (cf. Fig. 6).
In Figs. 8 and 9, we propose a comparison between two ts: for
the rst one, m is time dependent (full lines) whereas for the second
one, m is set to a universal value (dotted lines). As expected, a
constant m prevents description of the experimental data, Fig. 8.
4. Conclusion
Percolation of CNT in silicone oil was investigated by electrical
measurements. Changes in percolation curves over time were
thoroughly analyzed by monitoring the variations in both critical
fraction (Vc ) and critical exponent (m) obtained from tting Kirk-
patrick's equation. The experimental method we used - consisting
in progressively diluting a mother suspension - gave a unique op-
portunity to collect a large amount of experimental data in the close
vicinity of the transition, leading to outstanding resolution in the
determination of the critical fraction of ller Vc. This further
allowed to minimize the error on m.
An exponential decrease of Vc over time was observed. This
dynamic percolation process was attributed to a progressive ag-
gregation of CNT within silicone oil. More interestingly, the critical
exponent m was shown to vary with time.
This behavior may be correlated to the change in the distribu-
tion state of ller in the tested systems. Immediately after agitation
is stopped, llers turn from an oriented to a randomly distributed
state and m gets close to its universal value (2). For longer times CNT
aggregate and m diverges from the universal value.
An adaptation of Kirckpatrick's statistic percolation law was
proposed that takes into account the time dependence of Vc and m.
Good correlation between our model and the set of experimental
data was observed. According to us, the main scientic contribution
of the presented study is the importance of taking into account the
fact that m is correlated to the dispersion state. Thus it would be a
nonsense to consider its value as constant during dynamic perco-
lation process. The implementation of a time dependent value of m
in Alig's and Cao's models also yield to increased correlation be-
tween experimental data and modeling.
Acknowledgments

e des Pays de Savoie)
Financial support from the APS (Assemble
is gratefully acknowledged.
References
[1] F. Carmona, Conducting lled polymers, Phys. A 157 (1) (1989) 461e469.
[2] Ingo Alig, Petra Poetschke, Dirk Lellinger, Tetyana Skipa, Sven Pegel, Gaurav
R. Kasaliwal, Tobias Villmow, Establishment, morphology and properties of
carbon nanotube networks in polymer melts, Polymer 53 (1) (2012) 4e28.
[3] I. Balberg, A comprehensive picture of the electrical phenomena in carbon
black-polymer composites, Carbon 40 (2) (2002) 139e143 (Third Interna-
tional Conference on Carbon Black).
[4] F. Bueche, Electrical resistivity of conducting particles in an insulating matrix,
J. Appl. Phys. 43 (11) (1972) 4837e4838.
[5] J. Sandler, M.S.P. Shaffer, T. Prasse, W. Bauhofer, K. Schulte, A.H. Windle,
Development of a dispersion process for carbon nanotubes in an epoxy matrix
and the resulting electrical properties, Polymer 40 (21) (1999) 5967e5971.
[6] Erik T. Thostenson, Zhifeng Ren, Tsu-Wei Chou, Advances in the science and
technology of carbon nanotubes and their composites: a review, Compos. Sci.
Technol. 61 (13) (2001) 1899e1912.
[7] J.M. Hammersley, S.R. Broadbent, Percolation processes. crystals and mazes,
Proc. Camb. Philos. Soc. 53 (1957) 629e641.
[8] Dietrich Stauffer, Amnon Aharony, Introduction to Percolation Theory, Taylor
and Francis, 1991.
[9] Scott Kirkpatrick, Percolation and conduction, Rev. Mod. Phys. 45 (Oct 1973)
574e588.
[10] C.M. Chaves, Belita Koiller, Universality, thresholds and critical exponents in
correlated percolation, Phys. A 218 (3e4) (1995) 271e278.
[11] Wolfgang Bauhofer, Josef Z. Kovacs, A review and analysis of electrical
percolation in carbon nanotube polymer composites, Compos. Sci. Technol. 69
(10) (2009) 1486e1498. CNT-NET 07 Special Issue with regular papers.
[12] Sonia Vionnet-Menot, Claudio Grimaldi, Thomas Maeder, Sigfrid Str assler,
Peter Ryser, Tunneling-percolation origin of nonuniversality: theory and ex-
periments, Phys. Rev. B 71 (064201) (Feb 2005).
[13] P.M. Kogut, J.P. Straley, Distribution-induced non-universality of the perco-
lation conductivity exponents, J. Phys. C. Solid State Phys. 12 (11) (2015) 1979.
[14] B.I. Halperin, Shechao Feng, P.N. Sen, Differences between lattice and con-
tinuum percolation transport exponents, Phys. Rev. Lett. 54 (Jun 1985)
2391e2394.
[15] I. Balberg, Limits on the continuum-percolation transport exponents, Phys.
Rev. B Condens. Matter 57 (Jun 1998) 13351e13354.
[16] A. Celzard, J.F. Mareche, F. Payot, G. Furdin, Electrical conductivity of carbo-
naceous powders, Carbon 40 (15) (2002) 2801e2815.
[17] A. Celzard, J.F. Mareche, Non-universal conductivity critical exponents in
anisotropic percolating media: a new interpretation, Phys. A 317 (3e4) (2003)
305e312.
[18] Chiteme McLachlan, Ac and dc percolative conductivity of single wall carbon
nanotube polymer composites, J. Polym. Sci. Part B Polym. Phys. 43 (2005)
3273e3287.
[19] D.S. McLachlan, C. Chiteme, W.D. Heiss, Junjie Wu, Fitting the dc conductivity
and rst order ac conductivity results for continuum percolation media, using
percolation theory and a single phenomenological equation, Phys. B 338
(1e4) (2003) 261e265 (Proceedings of the Sixth International Conference on
Electrical Transport and Optical Properties of Inhomogeneous Media).
[20] C. Chiteme, D.S. Mclachlan, Measurements of universal and non-universal
percolation exponents in macroscopically similar systems, Phys. B 279
(1e3) (2000) 69e71.
[21] Junjie Wu, D.S. McLachlan, Percolation exponents and threshold in two nearly
ideal anisotropic continuum systems, Phys. A 241 (1e2) (1997) 360e366
(Proceedings of the Fourth International Conference on Electrical Transport
and Optical Properties of Inhomogeneoous Media).
[22] McMachlan, Godfrey Sauti, The ac and dc conductivity of nanocomposites
(mail de lf du 09/01/2012, kdo), J. Nanomater. 9 (2007) 2007.
[23] Ingo Alig, Dirk Lellinger, Martin Engel, Tetyana Skipa, Petra Poetschke,
Destruction and formation of a conductive carbon nanotube network in
 M. Badard et al. / Materials Chemistry and Physics 191 (2017) 89e95
polymer melts: in-line experiments, Polymer 49 (7) (2008) 1902e1909.
[24] Emiliano Bilotti, Han Zhang, Hua Deng, Rui Zhang, Qiang Fu, Ton Peijs, Con-
trolling the dynamic percolation of carbon nanotube based conductive poly-
mer composites by addition of secondary nanollers: the effect on electrical
conductivity and tuneable sensing behaviour, Compos. Sci. Technol. 74 (0)
(2013) 85e90.
[25] A. Combessis, N. Charvin, A. Allais, J. Fournier, L. Flandin, Understanding dy-
namic percolation mechanisms in carbonaceous polymer nanocomposites
through impedance spectroscopy: experiments and modeling, J. Appl. Phys.
116 (3) (2014), http://dx.doi.org/10.1063/1.4890342.
[26] Mercedes Fernandez, Maite Landa, Maria Eugenia Muuoz, Anton Santamaria,
Electrical conductivity of pur/mwcnt nanocomposites in the molten state,
during crystallization and in the solid state, Eur. Polym. J. 47 (11) (2011)
2078e2086.
[27] Mercio D. Lima, Menica J. Andrade, Viera Skakalova, Carlos P. Bergmann,
Siegmar Roth, Dynamic percolation of carbon nanotubes in liquid medium,
J. Mater. Chem. 17 (2007) 4846e4853.
[28] Bani H. Cipriano, Arun K. Kota, Alan L. Gershon, Conrad J. Laskowski,
Takashi Kashiwagi, Hugh A. Bruck, Srinivasa R. Raghavan, Conductivity
enhancement of carbon nanotube and nanober-based polymer nano-
composites by melt annealing, Polymer 49 (22) (2008) 4846e4851.
[29] Gabriella Faiella, Vincenza Antonucci, Samuel T. Buschhorn, Luis A.S.A. Prado,
Karl Schulte, Michele Giordano, Tailoring the electrical properties of mwcnt/
epoxy composites controlling processing conditions, Compos. Part A 43 (9)
(2012) 1441e1447.
[30] Anthony Combessis, Lorre ne Bayon, Lionel Flandin, Effect of ller auto-
assembly on percolation transition in carbon nanotube/polymer composites,
Appl. Phys. Lett. 102 (1) (2013), http://dx.doi.org/10.1063/1.4773994.
95
[31] Badard Mathieu, Combessis Anthony, Allais Arnaud, Flandin Lionel, Cnt ag-
gregation mechanisms probed by electrical and dielectric measurements,
J. Mater. Chem. C 3 (2015) 5769e5774.
[32] M. Badard, A. Combessis, A. Allais, L. Flandin, Electric eld as a tuning key to
process carbon nanotube suspensions with controlled conductivity, Polymer
82 (2016) 198e205.
[33] I. Alig, T. Skipa, M. Engel, D. Lellinger, S. Pegel, P. Poetschke, Electrical con-
ductivity recovery in carbon nanotube-polymer composites after transient
shear, Phys. Status Solidi B 244 (11) (2007) 4223e4226.
[34] Qing Cao, Yihu Song, Yeqiang Tan, Qiang Zheng, Thermal-induced percolation
in high-density polyethylene/carbon black composites, Polymer 50 (26)
(2009) 6350e6356.
[35] Qing Cao, Yihu Song, Yeqiang Tan, Qiang Zheng, Conductive and viscoelastic
behaviors of carbon black lled polystyrene during annealing, Carbon 48 (15)
(2010) 4268e4275.
[36] T. Skipa, D. Lellinger, M. Saphiannikova, I. Alig, Shear-stimulated formation of
multi-wall carbon nanotube networks in polymer melts, Phys. Status Solidi B
246 (11e12) (2009) 2453e2456.
[37] J. Krueckel, D.W. Schubert, Theoretical and experimental study of the dynamic
percolation behaviour of carbon black lled polymethylmethacrylate prior
and after shear a novel three phase approach, Eur. Polym. J. 53 (0) (2014)
50e57.
[38] Combessis, Apport des Nanotubes de Carbone a la Conduction Electrique de
Materiaux Organiques (PhD thesis), Universite de Grenoble, 2011.
[39] Guozhang Wu, Shigeo Asai, Masao Sumita, Carbon black as a self-diagnosing
probe to trace polymer dynamics in highly lled compositions, Macromole-
cules 35 (5) (2002) 1708e1713.