Sunteți pe pagina 1din 11

IEEE SENSORS JOURNAL, VOL. 16, NO.

1, JANUARY 1, 2016 77

Coupled Effects of Film Thickness and Filler


Length on Conductivity and Strain Sensitivity of
Carbon Nanotube/Polymer Composite Thin Films
Rubaiya Rahman, Saeid Soltanian, and Peyman Servati, Member, IEEE

Abstract— The coupled effects of varying composite film modulus of individual multi-walled nanotube (MWNT)
thicknesses and filler lengths on the conductivity and strain (0.27-0.95 TPa) [2] and single-walled nanotube (SWNT)
sensitivity of carbon nanotube (CNT)/polymer composite films (0.32-1.47 TPa) [3]. The high Young’s modulus of the
are investigated through modeling and experiments. Change
in average intertube distance is calculated statistically through CNTs shows its potential in the applications where composite
the Monte Carlo simulations for samples with different CNT reinforcement is necessary or repetitive strain/stress is exerted.
concentrations and film thicknesses for a given filler aspect CNTs when embedded in polymers have been reported to
ratio. The composite conductivity is then estimated from the enhance the piezoresistive property of the composites
intertube distance with a semi-analytical model based on a and show higher sensitivity than the conventional strain
tunneling current. The dependence of conductivity on mechanical
strain is investigated for varying film thickness for strain sensor sensors [4].
applications. A partial alignment of CNTs introduced at film Generally, in a conductor-insulator composite, if we
thicknesses less than the CNT length is observed to have a gradually increase the number of conductive fillers
significant influence on the composite conductivity and strain (e.g. CNTs), after reaching a certain filler volume fraction
sensitivity, specially at low CNT concentrations. The modeling the electrical conductivity of the composite takes a sharp
results can explain the observed experimental results of con-
ductivity for CNT composites, which illustrate a unique dip in upturn. At this filler concentration, known as percolation
conduction with increasing thickness. These results are important threshold, the electrical conduction effectively begins through
for understanding the composite characteristics with different the percolating conductive filler network. Several parameters
filler orientations and film thicknesses for a given filler length, including the type, concentration, aspect ratio, orientation
and useful for the design optimization of high performance order, synthesis method, surface modification of CNTs as
composite electronic films for applications in electronic skin and
sensors. well as polymer type and dispersion method influence the
percolation threshold, scaling law exponent and maximum
Index Terms— Nanocomposite, carbon nanotube, polymer, conductivity of the composite [5]. Most of the experimental
conductivity, modeling, film thickness, sensor.
results suggest exponential increase of film conductivity with
I. I NTRODUCTION the increase of CNT concentration [6], [7]. Larger aspect ratio
and random isotropic orientation of tubes are reported to lower
C ARBON nanotubes (CNTs) have attracted interest
of researchers working in nanoelectronics and
optoelectronics for the last few decades and recently in smart
the percolation threshold, and thus enhance the conductivity
of a CNT film [6], [8], [9]. From the recent studies it has
structure applications. CNTs possess exceptional electrical, been suggested that electron tunneling between neighboring
thermal and mechanical properties and high sensitivity that CNTs plays a major role in the electrical conductivity and
make them attractive candidates for multifarious application, piezoresistivity of a CNT/polymer composite [10], [11]. Since
i.e. transparent electrodes, photovoltaic devices, flexible the CNTs rarely make intimate contacts, in contrary to the
electronics, smart materials, etc. When they are introduced in assumption in percolation theory, the electrons tunnel through
an insulating polymer matrix even at a very low amount, the the polymer molecules present between neighboring CNTs,
CNTs can form an electrical percolating network that turns the given that the intertube gap is sufficiently small [12]. Thus the
insulating matrix to a conducting one [1]. The unique mechani- composite conductivity is mainly determined by the tunneling
cal strength of the CNTs can be understood from the Young’s conductivity specially in application where concentration of
CNTs is low, i.e. transparent conductors, strain sensors, etc.
Manuscript received August 10, 2015; accepted August 27, 2015. Date of The tunneling current in a CNT-CNT junction is determined
publication September 14, 2015; date of current version December 10, 2015.
This work was supported by the Natural Sciences and Engineering Research by different factors, i.e. the intertube distance d, energy barrier
Council of Canada. The associate editor coordinating the review of this paper height of the polymer λ, the type of CNT, etc. Again the
and approving it for publication was Dr. Stefan J. Rupitsch. intertube distance d in a composite film is influenced by the
The authors are with the Department of Electrical and Computer
Engineering, University of British Columbia, Vancouver, BC V6T 1Z4, CNT concentration, aspect ratio, alignment, film thickness, etc.
Canada (e-mail: rubaiya11@gmail.com; saeid@ece.ubc.ca; peymans@ece. Thus the film thickness can in turn influence the tunneling
ubc.ca). conductivity, which is a major component playing role in
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. strain sensitivity of composite films. Thin films of CNTs and
Digital Object Identifier 10.1109/JSEN.2015.2478447 CNT/polymer composite films have potential application in
1530-437X © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
78 IEEE SENSORS JOURNAL, VOL. 16, NO. 1, JANUARY 1, 2016

strain and pressure sensing for structural health monitoring,


damage or deformation detection, fluid flow sensing, artificial
skin, etc. [13]–[16]. To identify local structural damage and to
investigate structures at the component level, a large number
of sensors need to be installed. Thinner composite films can
help in cost reduction of the sensors and allow installation
of large number of sensors in a single monitoring system.
In a CNT/polymer composite film, when the film thickness is
reduced below the filler length, it introduces some alignment
of the fillers which can potentially affect the average intertube
distance and eventually the film conductivity and sensitivity.
Although there has been reports on the effect of film thickness
on the composite conductivity [17], [18], the comparative
effects of film thickness below and above the filler length have
not been reported yet. A thorough understanding of the relative
effect of film thickness and CNT length on the composite
conductivity can help to estimate the sensitivity of composite
films of different thickness. Hence a systematic study on this
coupled effect of film thickness and filler length can help in
design optimization of sensors, but still in need.
In this paper, we present our experimental and numerical
study of the relative effect of film thickness and filler length on
the electromechanical properties of CNT/polymer composites.
We numerically generated composite samples of different
thickness t, CNT volume fraction φ and alignment angle ψ,
then applied Monte Carlo based statistical method to calculate
the average intertube distance d. Then the composite film
conductivity and strain sensitivity was estimated using our
semi-analytical model. In order to verify the numerical results
on the behavior of film conductivity with the varying film
thickness and filler alignment, we carried out experiments with
SWNT/PMMA composite films and compared the experimen-
tal results with the numerical ones.
II. M ETHODOLOGY AND R ESULTS
In this section we first build a number of numerical com-
posite samples containing uniformly dispersed CNTs in order
to analyze the intertube distance. The number of CNTs are
varied to observe the effect of different CNT concentration on Fig. 1. Illustration of (a) a 3D cubic sample of CNT/polymer composite
the average intertube distance. The sample film thickness is and (b) a randomly oriented filler making orientation angles with X-Y and
changed below and above the length of CNTs to observe the X-Z planes.
effect of filler alignment on the intertube gaps introduced by
the sample thickness. Then the composite film conductivity is The concentration of CNTs are determined by the filler volume
obtained using our semi-analytical model. Finally, the strain fraction φ, which is the ratio of total volume of CNTs to the
sensitivity of the films with different thickness under applied volume of composite cell. The alignment order was determined
strain in longitudinal direction is estimated and analyzed. by two orientation angles-θ and ψ, the angles that the fillers
make with the X-Z and X-Y planes respectively, as shown
A. Construction of Numerical Samples of Nanocomposite in Fig. 1(b). We varied the CNT concentration by changing
A three dimensional microstructure sample of CNT/polymer the stick numbers in the unit samples. The cut-off angles θμ
composite is simulated where CNTs are assumed as multi- and ψμ are the maximum angle between an arbitrary filler
walled carbon nanotubes (MWNTs) and randomly dispersed in and the X-Z and X-Y planes respectively. The average value
the polymer matrix. Here, CNTs are considered as penetrable of intertube distance and other numerical data are obtained by
soft-core cylinders with length L = 5μm and diameter Monte-Carlo simulations over large number samples. We also
D = 50nm, hence aspect ratio AR of 100. The sample, varied the thickness of the samples by changing the length
as shown in Fig. 1(a), is simulated as a cubic unit cell of L z while keeping the L x = L y of same length. It is to
(L x × L y × L z ) with a length of 25μm on each side. note that the results of simulations can be influenced by the
The position and orientation of the stick-like fillers is size of the unit cell. Smaller cell size is helpful in reducing
randomly chosen following the method by Hu et al [11]. the computational cost but it generates very unstable results.
RAHMAN et al.: COUPLED EFFECTS OF FILM THICKNESS AND FILLER LENGTH ON CONDUCTIVITY AND STRAIN SENSITIVITY 79

Fig. 3. Intertube distance for varying CNT concentration for a random


3D network.

There can be multiple number of percolating clusters making


parallel conductive pathways across the sample.
If voltage is applied across a junction between two neigh-
boring tubes, the electron finds the shortest distance between
these two tubes and tunnels through it if the intertube gap
is below the maximum allowed tunneling gap. The energy
barrier between the tubes, due to the presence of polymer
layer, gets changed in shape under a potential difference and
causes the electrons to tunnel through the barrier [10]. Thus a
small current is generated at the junction. The current through
the entire sample is the summation of these small junction
currents through the conductive pathways. Fig. 2(b) shows a
pair of neighboring CNTs where electron tunneling occurs
between two tubes through the minimum intertube gap dt .
The resistance of the electron’s pathway is represented by a
Fig. 2. A schematic diagram of (a) top view of the CNT network with
electric field applied on it. The percolating cluster is shown within the dashed series resistance comprising the tunneling resistance Rt , and
line, and (b) a CNT pair with tunneling gap dt between them. The resistance CNT resistance Rc .
along the CNTs are denoted as Rc and the tunneling resistance as Rt . The composite conductivity is primarily influenced by
the tunneling currents. Since electron tunneling occurs only
between the nearest neighbor CNTs, the minimum distance
For this work we chose L x = L y = 5L and varying values
among the tubes plays significant role in determining the
of L z for 3D samples. This cell size is sufficiently large to
overall conductivity. In this work, we have used a Monte-Carlo
achieve numerical convergence and stability in the results as
procedure of large number of simulations to estimate the aver-
confirmed by literature [4].
age minimum distance between any arbitrary pair of tubes d
in a 3D CNT network for varying CNT concentration and film
B. Analysis of Intertube Distance thickness. For samples with thickness greater than the length
As the filler concentration in a composite sample increases, of the CNTs, the fillers can be oriented in any order inside
it decreases the inter-filler gaps. If an electric field is applied the 3D unit cells. Hence both the cut-off angles θμ and ψμ
across a CNT/polymer composite sample, the decreased are set at their highest value, i.e. 90°. and the fillers can
intertube distance at higher CNT concentration enhances the achieve complete randomness in orientation. Fig. 3 shows the
probability of electron transfer between neighboring tubes. simulation results of intertube distance for varying filler con-
In our simulation work, the direction of applied field and centration in a cubic unit cell with L x = L y = L z = 25 μm.
electrical conduction is assumed along X-axis, determined We can see as CNT volume fraction φ increases, the intertube
by the location of electrodes. The neighboring junctions distance d decreases in a power law manner, which is similar
transferring electrons can build up a conducting bridge from to the behavior observed in 2D composite samples [19].
the left electrode to the right one known as the percolating Apart from the filler concentration, the average intertube
cluster. It is shown within the dashed line in Fig. 2(a). distance can vary with the filler aspect ratio or orientation
80 IEEE SENSORS JOURNAL, VOL. 16, NO. 1, JANUARY 1, 2016

Fig. 4. Schematic diagram of CNT/polymer composite films with different


thicknesses.

Fig. 6. Intertube distance at varying sample film thickness for different CNT
concentration.

Fig. 5. Intertube distance at varying CNT concentration for CNT/polymer


composite films with different thickness.

which in turn can influence the film resistance [19]–[21].


If the sample film thickness is decreased below the filler
length, that will limit the randomness of the fillers and
introduce some alignment, thus reduce the cut-off angle ψμ
to a value below 90°. The schematic diagram in Fig. 4
demonstrates how the alignment order of the CNTs is changed Fig. 7. Intertube distance at varying alignment order of CNTs for different
when the film thickness is decreased below the CNT length, CNT volume fraction.
keeping the number of CNTs per unit volume unchanged. The
change in the CNT orientation affects the average distance decreasing t. This behavior opens up a question of the effect of
between tubes and thus can influence the conductivity and the partial alignment on d. Before t is decreased to 5μm, the
sensitivity of the composite film. fillers are allowed to have full randomness whereas right after
We studied the effect of film thickness on the intertube the t is reduced below 5μm, the fillers starts aligning to the
distance by numerical simulation which is shown in Fig. 5. X-Y plane gradually making ψμ less than 90°. Thus at
Since the CNTs used in our simulation have length of t = 50nm the composite film becomes almost a quasi-2D
L = 5μm and diameter of D = 50nm, we varied the film plane making ψμ ≈ 0°. We calculated the cut-off angle ψμ
thickness from 15μm to 50nm to study effect of film thickness for t ≤ 5μm and re-plotted the corresponding values of d
both below and above the CNT length. From the figure we in Fig. 7. It shows that the intertube distance reaches the lowest
can see for all film thickness t, the intertube distance d value at around ψμ = 45°. Therefore, the minimum intertube
decreases as the CNT concentration increases. At any CNT distance can be found at partial alignment, not in completely
concentration, the d has higher value at lower film thickness, random or completely aligned CNT network. This behavior
except when t is close to CNT length L = 5μm. Fig. 6 is consistently shown by samples of different concentrations,
demonstrates the change in d with thickness. Our simulation whereas the drop in the intertube gap at partial alignment
results show that when the film thickness t decreases from is more pronounced at lower concentration of fillers. For
15μm to 5μm the intertube distance increases gradually. But example, when ψμ is decreased from 90° to 53°, the intertube
as t drops slightly from 5μm, d starts decreasing, keeps distance decreases from 0.47nm to 0.05nm for φ = 7%
lower value for a while, then again starts increasing with the concentration, whereas for the same change in alignment, the
RAHMAN et al.: COUPLED EFFECTS OF FILM THICKNESS AND FILLER LENGTH ON CONDUCTIVITY AND STRAIN SENSITIVITY 81

drop in the intertube distance for φ = 1.42% is from 7.61nm film I f should be the accumulation of the currents passing
to 2.026nm. This behavior can be explained in context of through all the parallel pathways,
CNT-CNT junctions that build up the conductive pathway for

N 
N
Vf
percolation. In a highly concentrated random CNT network, If = Ii = . (3)
the number of nanotube junctions is high. When a filler is Ri
i=1 i=1
aligned slightly, it may lose contact with one neighbor filler.
Here, Ri is the resistance of the i th pathway consisting a
But since each filler is surrounded by many others, it can make
number of tunneling (Rt ) and CNT resistances (Rc ).
contact with another neighboring filler. Thus the average inter-
If the sample CNT/polymer composite film with length L f
tube distance does not get significantly affected by the partial
and cross-sectional area A f , has a total current I f flowing
alignment of fillers, and the percolation continues. However,
through it while the applied voltage across it is V f , we can
higher alignment causes loss of junctions and discontinu-
calculate the conductivity of the composite by
ity in the network, and results in increasing intertube gap.
In case of low CNT low concentration near percolation thresh- If L f
σcomp = (4)
old, slight alignment of the fillers allows less number of fillers Vf Af
to build the conducting bridges by reducing gaps between the
Lf 
N
neighboring fillers along the direction of conduction. However, = ( Ii ). (5)
at higher alignment, the low concentrated CNT network also A f Vf
i=1
experiences the loss of junctions like the highly concentrated Now, the current in the i th pathway Ii is mainly determined
networks and d increases. by the tunneling current at the junctions present in that
pathway. Hence, it should be proportional to the tunneling
C. Analysis of Conductivity current It , thus should have an exponential relationship with
The electrical conductivity of a composite is mainly the tunneling intertube gap. Again, the composite conductivity
governed by the tunneling current through the junctions. Hence has been reported to proportionally increase with the increase
the average intertube distance can be used as a key parameter of filler CNT conductivity [4]. Thus, considering the effect of
to estimate the composite conductivity. In order to estimate the intertube gap, CNT conductivity, CNT concentration, polymer
electrical conductivity of the composite samples of different energy barrier height, etc., we developed an analytical model
filler concentrations and film thickness, we used the simulated of composite conductivity [22] as follows,
data of intertube distance and applied our conductivity model σcomp = σC N T e−(sβd) + σmatri x . (6)
that we developed analytically in our previous work [22].
In a composite sample, there are usually several number Here, σmatri x is the conductivity of the polymer matrix
of percolating clusters making parallel pathways along without any filler present. The exponential term β includes
the sample. When a voltage is applied across the sample, the effect of the energy barrier height λ of the polymer by,

it generates potential difference between each tunneling β ∝ λ (7)
junction along the percolating pathways that exerts a driving √
= r λ. (8)
force on the electrons to tunnel through the barrier. The
resulting junction currents sums up to generate the total Here r = 9.7692nm −1eV −0.5 for the 3D composite model and
current in the composite sample. Considering a rectangular is independent of the polymer type. λ is dependent on the type
tunneling barrier, the tunneling current It through a single of polymer. For different polymers the energy barrier height
junction, with tunneling distance dt between the neighboring can vary within a wide range (e.g. for PMMA λ = 0.17 eV
CNTs, can be expressed as [23], and for epoxy λ = 1.5 eV [21], [24]). The average intertube
distance d is numerically determined for different sample film
It = cρ1 ρ2 e−βdt . (1)
thickness and CNT concentrations. The exponent s includes
Here, ρ1 and ρ2 are the density of states of the two neighboring the effect of filler concentration on the number of junctions
CNTs and function of their location, energy E with respect and the number of parallel pathways,
to their individual Fermi levels and applied voltage V . The s = mφ n . (9)
polymer dependent parameter β is given by
√ For an ideal 3D composite sample with uniform filler dis-
2π 2mλ tribution we obtained m = 91.7487 and n = 1.0915.
β= , (2)
h For conductivity estimation using our model, we considered
where, λ = V − E is the height of the energy barrier of the MWNTs with AR=100 and conductivity σcnt = 104 S/m
polymer. as the fillers. As the polymer we chose epoxy and PMMA,
In a conductive pathway, consisting of a number of tunnel- since they are widely used by the researchers as compos-
ing junctions, the current is limited by the junction with the ite materials for their transparency, stretchability, dispersion
largest tunneling gap. If the voltage applied across a composite property with CNTs, and biocompatibility [9], [11], [15], [20].
film is V f , and Ii is the current in the i th pathway, and We considered the conductivity of the polymer insulators of
there are N number of parallel conductive pathways along about 10−12 S/m [12]. We varied the CNT concentration by
the voltage bias direction, then the total current through the changing the number of fillers in the numerical samples and
82 IEEE SENSORS JOURNAL, VOL. 16, NO. 1, JANUARY 1, 2016

Fig. 8. MWNT/epoxy composite conductivity at varying CNT concentration.


Our model is compared with the literature data based on percolation model
by Hu et al. [20], numerical simulation by Hu et al. [21], and experimental
results by Hu et al. [11] and NCT Co. [25].

estimated the composite conductivity using Eq. (6). The result


for MWNT/epoxy composite is compared with the results
previously reported in the literature [11], [20], [21], [25] as
shown in Fig. 8. Our model shows a close fit to numerical
results while reducing the computational cost significantly.
It shows reasonable comparative values with the theoretical
and experimental results. Our model incorporates the tunneling
current through polymers, unlike the traditional percolation
model. It can estimate the composite conductivity both in sub-
percolation and high CNT concentration region, without know-
ing the percolation threshold from experiment beforehand.
This in effect is a new hierarchical method for estimation of
composite film properties that provides accuracy while having
enhanced ease of computation. However, since this model is Fig. 9. Conductivity of (a) MWNT/epoxy and (b) MWNT/PMMA compos-
ites at varying film thickness to filler length ratio for different CNT volume
based on the assumption that the CNTs as soft-core penetrable fraction.
cylinders, it can predict the electrical behavior of composite
films only under tensile strain, not when they are subjected conductivity achieved by film thickness variation. After t/L is
to compressive strain. To be able to deal with both the tensile decreased to a certain extent, σcomp starts dropping again with
and compressive strains, the model needs to consider hard-core lower film thickness. Fig. 10 shows the conductivity plot for
bendable structure of fillers so that the geometric change of the corresponding cut-off angles ψμ at different t/L both for
the fillers and the corresponding change in the junction areas CNT/epoxy and CNT/PMMA composites. It shows the highest
as an effect of mechanical compression can be included. σcomp is achieved at around ψμ of 45° to 50°. Thus our sim-
In the next stage the film thickness is varied for each ulation shows that for a given filler concentration and aspect
filler concentration to study the effect of film thickness on ratio the highest composite conductivity is obtained at slightly
composite conductivity. The result is shown in Fig. 9 for both aligned network rather than at isotropic network. Although this
MWNT/epoxy and MWNT/PMMA composites. We plotted result contradicts with many reports claiming that maximum
σcomp in terms of t/L to understand the relative effect of filler conductivity should occur at complete isotropy [26], [27],
length and film thickness. As the film thickness is reduced it agrees with the results by Du et al. [9]. This is demon-
from 15μm to 5μm, making t/L from 3 to 1, the conductivity strated by Fig. 10(b). We compared our simulation results
decreases. But just below t/L < 1 it increases sharply which for CNT/PMMA composites with the experimental results of
is due to the lowered d at partial alignment, as shown by our Du et al. Our results show similar pattern of the composite
results on intertube distance. Although the dip in conductivity conductivity with the partial alignment of fillers as found in
is observed near t/L ≈ 1 for both the epoxy and PMMA Du’s experiment, although it differs in the conductivity value.
polymer matrices, the extent of conductivity fall and rise vary The higher value of conductivity shown by the simula-
with the polymer. Hence, the polymer barrier height plays a tion results than that of the experimental ones may be
significant role here in determining the lowest and highest caused by the CNT agglomerations that we neglected in our
RAHMAN et al.: COUPLED EFFECTS OF FILM THICKNESS AND FILLER LENGTH ON CONDUCTIVITY AND STRAIN SENSITIVITY 83

Fig. 11. (a) Dropcasted samples of SWNT/DMF solution. (b) The SEM
image of a SWNT/DMF sample shows good dispersion.

were bought from Cheap Tubes Inc. The CNTs had purity
over 90% with average length of 30μm and outer diameter
of 1.5nm. As the polymer we used PMMA with molecular
weight of 35 × 104 . Dimethylformamide (DMF) was used as
the common solvent for both SWNT and PMMA.
At first, CNT/DMF dispersion of 0.5 mg/ml concentration
was prepared by continuous sonication of the dispersion
for 30 minutes using a Sonics VCX 750 ultrasonic
processor equipped with a 6 mm tip. We added
polyvinylpyrrolidone (PVP) as the surfactant in order to
obtain uniform dispersion. The PVP was added in 1:3 weight
ratio to CNTs. The CNTs were well dispersed in the DMF.
We made some dropcasted samples of the CNT/DMF solution
and used scanning electron microscope (SEM) to check
dispersion of CNTs. Fig. 11(a) shows some dropcasted
samples and Fig. 11(b) shows the SEM image of one of the
samples. We can see CNTs are well dispersed in the sample
and almost no agglomeration present there.
After preparing the CNT/DMF dispersion, PMMA powder
Fig. 10. Composite conductivity at varying alignment order of CNTs was added to it at different amounts in order to vary the CNT
for different CNT volume fraction. (a) Numerical estimation of conductiv- concentration in the mixture and final CNT/PMMA composite
ity of CNT/epoxy composite. (b) Numerical estimation of conductivity of
CNT/PMMA composite compared with experimental data of Du et al. [9]. films. The mixture was first stirred overnight at room temper-
ature in a closed container to obtain a uniform solution. Then
numerical procedure. Thus our semi-analytical model can the solution was again stirred in an open container at 80°C
explain and qualitatively confirm the result reported for few hours to evaporate the DMF and attain the desired
by Du et al. from the context of intertube distance, while viscosity. In the final CNT/PMMA/DMF composite solutions
other models could not do so. Our result indicates that with after evaporation, the PMMA concentration was maintained at
1
composite films slightly thinner than the filler length, causing 6 gm/ml to attain almost the same viscosity for all solutions
partial alignment of the fillers, we can attain significantly with varying CNT concentrations. Uniform viscosity helped to
higher conductivity in a composite sample with lower CNT control the film thickness while changing the spin speed for
concentration in comparison to a sample with higher CNT samples with varying CNT concentrations. In the next step,
concentration and random alignment. The rise in conductivity the CNT/PMMA/DMF composite solution was spincoated on
at partial alignment is more prominent in samples with low glass substrates. For each batch of composite solution with
CNT concentration. However, at higher alignment the con- different CNT concentrations, we varied the spin speed from
ductivity drops for all CNT concentration since the conducting 700 to 2000 rpms in order to vary the coated film thickness
pathway becomes discontinuous. within a range of 3 to 60μm. The spincoated samples were
then degassed in a vacuum chamber in order to make sure that
no air bubble stays inside the film. Finally the composite films
D. Experimental Verification of Thickness were dried in air at 50°C. In the final CNT/PMMA composite
Effect on Conductivity films, the CNT concentration varied from 0.24 to 1.5% in
To verify our numerical results, we conducted experiments weight fraction.
on CNT/polymer composite films with varying CNT concen- In order to examine the CNT distribution inside the polymer
tration and film thickness. Single walled carbon nanotubes matrix, the as prepared composite films were examined
84 IEEE SENSORS JOURNAL, VOL. 16, NO. 1, JANUARY 1, 2016

It was observed that for low CNT concentrations, i.e. CNT


weight fraction 0.24%, 0.26%, 0.28%, the film conductiv-
ity rises immediately after the film thickness is decreased
below 30μm, which is the average length of the SWNTs here.
But it drops after t is reduced further. Thus our experimental
results demonstrate similar pattern of conductivity change with
the varying film thickness. It is interesting to note that the film
with 0.26% CNT weight and 18.9μm thickness is showing
higher conductivity than the film with 0.28% CNT weight
and 34μm thickness despite having lower filler concentration
and lower film thickness. This result is significantly impor-
tant in cost reduction and design optimization of composite
devices based on thin films. However, similar increase in the
conductivity of the films with thickness around 30μm were
not observed for the higher CNT weight fraction. Thus our
numerical prediction of the relative influence of film thickness
and filler length on the composite conductivity is confirmed
by the experimental results.
In the real samples, the length and aspect ratio of the CNTs
are not constant. Hence we used weight fraction (m%) of
Fig. 12. (a) SWNT/PMMA composite film samples with different film thick-
ness having same CNT concentration. (b) The SEM image of a SWNT/PMMA CNTs as the CNT concentration parameter for experimental
film with its surface and edge. The CNTs look well distributed in the polymer data. However, for low CNT concentration, as used in the
matrix. experiment, the weight fraction and the volume fraction give
very close values. In general, we can convert the volume
fraction φ to weight fraction m by the following relation,
1
m= , (10)
1 + (ρ p /ρc ) × (1/φ − 1)
where, ρ p and ρc are the density of the polymer and CNT
respectively. Since during the experiments it is not practically
feasible to have samples with all the CNTs of same aspect
ratio, chirality, or uniformity in distribution, as assumed in
our simulation work, the measured values do not exactly
match those obtained by the numerical study. Nevertheless,
our experimental results follow almost the same trend as those
obtained by our numerical studies and qualitatively confirm
our numerical result, which is important for designing the
experiments before carrying them out.

E. Analysis of Strain Sensitivity


When tensile strain is applied on a CNT/polymer composite
Fig. 13. Experimental data of SWNT/PMMA composite conductivity at film, the change in the film dimension causes change in the
varying film thickness for different CNT weight fraction. orientation angle of most of the tubes in the film, which in turn
changes the intertube distance and the tunneling resistance.
Consequently, this affects the overall film resistance. The
by SEM. Fig. 12(a) shows three typical SWNT/PMMA change in resistance in a percolating network of a thin film can
composite film samples with different film thickness having be used to sense the applied strain which can be very helpful
same CNT concentration. From the SEM micrograph of a in many practical application, such as, early detection of
composite sample, as shown in Fig. 12(b), we can see the structural damage, stretch or deformation reading of artificial
SWNTs are almost uniformly distributed in the PMMA matrix. skin, etc. It is to note that, large deformation of the composite
After preparation, a large number of samples with different film due to large strain application can cause deformation of
thickness and CNT weight fraction were measured for con- the CNT fillers inside the matrix, leading to change in CNT
ductivity and thickness. The film thickness of the samples was properties. But at small strain, the physical change of the
measured using profilometer. The conductivity of the samples CNTs becomes negligible due to their much higher elastic
was measured with four point probe method. The measured modulus than the polymers. Hence the change in the filler
values of the conductivity of films with varying thickness and properties can be ignored. In this work, we have used small
different CNT weight fractions are shown in Fig. 13. strain, less than 5%, as most of the polymers deform plastically
RAHMAN et al.: COUPLED EFFECTS OF FILM THICKNESS AND FILLER LENGTH ON CONDUCTIVITY AND STRAIN SENSITIVITY 85

Fig. 15. Conductivity of MWNT/epoxy composite films (φ = 1.14%) at


Fig. 14. Effect on intertube distance at applied strain, for a CNT concentration applied strain for different film thickness.
φ = 1.14% and different film thickness.

beyond this range. We have numerically studied the effect of


uniaxial strain on the sensitivity of composite films of varying
thickness. The change in film dimension is determined by the
applied strain
and the Poisson’s ratio of the polymer ν. The
sensitivity of a thin film strain sensor is measured in terms of
the gauge factor G F,
R/R0 (R − R0 )/R0 R/R0
GF = = = . (11)
L f /L f 0 (L f − L f 0 )/L f 0

Here R0 and L f 0 are the initial resistance and length of the


film when no strain is applied, and R and L f are the resistance
and length of the film when strain is applied, respectively.
For a given strain
, higher resistance change ratio R/R0
indicates higher gauge factor and better sensitivity. Since we
observed from our results in the previous section that the film
thickness and filler length contribute to the intertube distance
and composite conductivity, the change in the film thickness
should also have impact on the gauge factor of the film. Fig. 16. Resistance of MWNT/epoxy composite films (φ = 1.14%) at applied
For different film thickness t and mechanical strain
, strain for different film thickness.
we numerically calculated the average intertube distance using
the fiber reorientation model [28]. For our numerical study, other curves. Again, with increasing strain, the values of d for
we considered epoxy as the polymer. We used the value of t = 5 and 7μm drop with a power law within this strain range.
ν = 0.4 as the Poisson’s ratio [19]. We then obtained the This exceptional trend of d shown by films with t/L ≈ 1
film conductivity applying our analytical model by Eq. 6. can be explained by the effect of partial alignment of fillers.
From the conductivity and film thickness we calculated the For films with t = 5 and 7μm, the CNTs in unstrained film
film resistance and resistance change ratio for varying strains. can be oriented with full randomness in all X, Y, and Z axes
The film thickness was varied from 1μm to 7μm. With directions. The application of small longitudinal strain aligns
changing t, the number of CNTs in the unstrained (
= 0) the fillers slightly towards X axis direction changing the cut-
composite sample was adjusted to keep the CNT concentration off angle ψμ from 90° to a lesser value. The partial alignment
at φ = 1.14%. The results are shown in Figs. 14-17. The brings the overall CNTs closer and reduces d. For lower film
change in the average intertube distance d with the increasing thicknesses, i.e., t < 5μm or t/L < 1, the CNTs are already
strain is shown in Fig. 14. As we can see, the intertube distance slightly aligned towards the X-Y plane in the unstrained films.
increases almost linearly with the strain for almost all film The application of strain aligns them further and causes the
thickness except when t takes value close to filler length L, intertube distance to increase.
i.e. for t = 5 and 7μm. At these film thicknesses we find The change in film conductivity due to the change in
that the initial intertube distance at
= 0 is higher than the intertube distance with applied strain is shown in Fig. 15.
initial d values for t = 4μm film, unlike the trend shown by We can see the conductivity towards the longitudinal
86 IEEE SENSORS JOURNAL, VOL. 16, NO. 1, JANUARY 1, 2016

III. C ONCLUSION
We have studied the coupled effect of film thickness and
filler length on the conductivity and strain sensitivity of
CNT/polymer composite films. The partial alignment of fillers
introduced by the film thickness lower than the filler length has
shown to have significant influence on the average intertube
distance of the fillers causing remarkable impact on film
conductivity and resistance change ratio. Our results on CNT
alignment effect show that higher conductivity in composites
can be obtained at low CNT concentration by aligning the
fillers slightly to the direction of conduction, rather than
by increasing filler concentration and keeping the network
random. It indicates that partial alignment of CNTs will allow
low cost composite fabrication with low CNT loading to
achieve the same conductivity offered by highly concentrated
and randomly distributed CNT composite samples. This find-
ing has significant importance for some particular applica-
Fig. 17. Resistance change ratio of MWNT/epoxy composite films tions. For example, transparent conductors can be made from
(φ = 1.14%) at applied strain for different film thickness. CNT/polymer composites for solar cell applications. Here low
filler concentration is desired to ensure higher transparency
and the conductivity then needs a trade-off. Our study on
direction decreases as the strain increases for most of the composite film thickness shows that conductivity works as
film thicknesses except for t = 5 and 7μm, as expected from a function of film thickness when the thickness goes below
the intertube distance pattern. However, for t = 5 and 7μm, filler length, and higher strain sensitivity can be obtained
the conductivity starts from a low value at unstrained films, with a thinner film keeping the same CNT concentration. The
then rises drastically with the application of strain smaller results on composite films with thickness close to the filler
than 5%. The corresponding film resistance were calculated length reveals an exceptional trend of conductivity with the
which is shown in Fig. 16. The films with t = 5 and 7μm, changing strain, that shows potential of this kind of films for
with sharp change in R with strain, show high potential stretchable switching applications. Our results presented here
in sensor application within smaller strain range. The is broadly applicable to conductive, stick-like filler networks
resistance drops from 1013 to 104 , that is, from almost in an insulating matrix.
insulator to conductor state, within strain application of 10%.
These results show potential of thin composite films with
R EFERENCES
t/L ≈ 1 for the application of stretchable switches. The
filler concentration can be optimized along with the thickness [1] J. K. W. Sandler, J. E. Kirk, I. A. Kinloch, M. S. P. Shaffer, and
A. H. Windle, “Ultra-low electrical percolation threshold in carbon-
of the film to obtain certain design window for specific nanotube-epoxy composites,” Polymer, vol. 44, no. 19, pp. 5893–5899,
switching requirements or polymer-filler combinations. 2003.
Finally we calculated the resistance change ratio R/R0 [2] M. F. Yu, O. Lourie, M. J. Dyer, K. Moloni, T. F. Kelly, and R. S. Ruoff,
“Strength and breaking mechanism of multiwalled carbon nanotubes
to estimate the strain sensitivity for different film thickness under tensile load,” Science, vol. 287, no. 5453, pp. 637–640, 2000.
which is shown in Fig. 17. We can see that, with increasing
[3] M.-F. Yu, B. S. Files, S. Arepalli, and R. S. Ruoff, “Tensile loading of
the film resistance increases resulting in higher absolute ropes of single wall carbon nanotubes and their mechanical properties,”
Phys. Rev. Lett., vol. 84, p. 5552, Jun. 2000.
value of resistance change ratio R/R0 . For films with [4] N. Hu, Z. Masuda, and H. Fukunaga, “Prediction of electrical conduc-
thickness t < L, the R/R0 values increase gradually and tivity of polymer filled by carbon nanotubes,” in Proc. 16th Int. Conf.
almost linearly with the strain, and are higher for thinner Comp. Mater., Kyoto, Japan, Jul. 2007, pp. 224333-1–224333-9.
[5] K. Mylvaganam and L. C. Zhang, “Fabrication and application of
films at any strain indicating higher gauge factor. However, polymer composites comprising carbon nanotubes,” Recent Patents
for t = 5 and 7μm, the decreasing resistance with applied Nanotechnol., vol. 1, no. 1, pp. 59–65, 2007.
strain results in negative values of R/R0 and drastically [6] D. Simien, J. A. Fagan, W. Luo, J. F. Douglas, K. Migler, and J. Obrzut,
“Influence of nanotube length on the optical and conductivity properties
reaches the minimum value of −100% within
≈ 1%. of thin single-wall carbon nanotube networks,” ACS Nano, vol. 2, no. 9,
So it can be said that, films with thickness t < L show good pp. 1879–1884, 2008.
sensitivity to a broader range of strain, whereas films with [7] H. E. Unalan, G. Fanchini, A. Kanwal, A. D. Pasquier, and
M. Chhowalla, “Design criteria for transparent single-wall carbon nan-
t ≈ L show high potential in sensor application within smaller otube thin-film transistors,” Nano Lett., vol. 6, no. 4, pp. 677–682, 2006.
strain range. In general, the sensitivity gets higher for thinner [8] T. Natsuki, M. Endo, and T. Takahashi, “Percolation study of orientated
films. However, since the conductivity goes lower as the film short-fiber composites by a continuum model,” Phys. A, Statist. Mech.
Appl., vol. 352, nos. 2–4, pp. 498–508, 2005.
gets thinner, a trade-off is needed to keep the resistance in [9] F. Du, J. E. Fischer, and K. I. Winey, “Effect of nanotube alignment on
measurable range while designing the sensors. Since thicker percolation conductivity in carbon nanotube/polymer composites,” Phys.
films have more robustness and higher potential for repeated Rev. B, vol. 72, p. 121404, Sep. 2005.
[10] C. Gau, C.-Y. Kuo, and H. S. Ko, “Electron tunneling in carbon nan-
strain application, sensor designing will need consideration of otube composites,” Nanotechnology, vol. 20, pp. 395705-1–395705-10,
all these parameters depending on the strain requirements. Sep. 2009.
RAHMAN et al.: COUPLED EFFECTS OF FILM THICKNESS AND FILLER LENGTH ON CONDUCTIVITY AND STRAIN SENSITIVITY 87

[11] N. Hu, Y. Karube, C. Yan, Z. Masuda, and H. Fukunaga, “Tunneling Rubaiya Rahman received the B.Sc. and M.Sc.
effect in a polymer/carbon nanotube nanocomposite strain sensor,” Acta degrees in electrical and electronic engineering
Mater., vol. 56, pp. 2929–2936, Aug. 2008. from the Bangladesh University of Engineering and
[12] C. Li, E. T. Thostenson, and T.-W. Chou, “Dominant role of tunnel- Technology, Dhaka, Bangladesh, in 2006 and 2008,
ing resistance in the electrical conductivity of carbon nanotube–based respectively. She received the Ph.D. degree in elec-
composites,” Appl. Phys. Lett., vol. 91, no. 22, p. 223114, 2007. trical and computer engineering from the University
[13] H. Cao et al., “Single-walled carbon nanotube network/poly compos- of British Columbia (UBC), Vancouver, BC, Canada,
ite thin film for flow sensor,” Microsyst. Technol., vol. 16, no. 6, in 2015. Her research interests include numerical
pp. 955–959, 2010. analysis of percolating network characteristics,
[14] C.-X. Liu and J.-W. Choi, “An embedded PDMS nanocomposite strain modeling of conductivity and sensitivity of carbon
sensor toward biomedical applications,” in Proc. Annu. Int. Conf. IEEE nanotube/polymer composite films and fibers,
Eng. Med. Biol. Soc. (EMBS), Minneapolis, MN, USA, Sep. 2009, and applications of nanocomposites in conductive films, sensors, and flexible
pp. 6391–6394. devices.
[15] R. Ormsby, T. McNally, P. O’Hare, G. Burke, C. Mitchell, and N. Dunne,
“Fatigue and biocompatibility properties of a poly(methyl methacrylate)
bone cement with multi-walled carbon nanotubes,” Acta Biomater.,
vol. 8, no. 3, pp. 1201–1212, 2012.
[16] A. K. Singh, “CNT based nanocomposite strain sensor for
structural health monitoring,” in Nanostructured Materials and
Nanotechnology VI, S. Mathur, S. S. Ray, and M. Halbig, Eds. Hoboken, Saeid Soltanian received the Ph.D. degree
NJ, USA: Wiley, 2012, ch. 9. from the University of Wollongong, Australia,
[17] D. Zhang et al., “Transparent, conductive, and flexible carbon nanotube in 2003. Before joining The University of British
films and their application in organic light-emitting diodes,” Nano Lett., Columbia (UBC), Vancouver, Canada, in 2010,
vol. 6, no. 9, pp. 1880–1886, 2006. he was an Associate Research Fellow with the
[18] M. Fu et al., “Significant influence of film thickness on the percola- University of Wollongong (2004), an Assistant
tion threshold of multiwall carbon nanotube/low density polyethylene Professor (2005–2008), and an Associate Professor
composite films,” Appl. Phys. Lett., vol. 94, no. 1, p. 012904, 2009. (2008–2010) with the University of Kurdistan,
[19] R. Rahman and P. Servati, “Effects of inter-tube distance and alignment Iran. He is currently a Research Associate with
on tunnelling resistance and strain sensitivity of nanotube/polymer the Department of Electrical and Computer
composite films,” Nanotechnology, vol. 23, pp. 055703-1–055703-11, Engineering, UBC, where he works with the
Jan. 2012. Flexible Electronics and Energy Laboratory. His recent research interests
[20] N. Hu, Z. Masuda, C. Yan, G. Yamamoto, H. Fukunaga, and T. Hashida, include fabrication and characterization of nanostructure materials and
“The electrical properties of polymer nanocomposites with carbon devices for sensing and energy applications.
nanotube fillers,” Nanotechnology, vol. 19, pp. 215701-1–215701-10,
Apr. 2008.
[21] N. Hu et al., “Investigation on sensitivity of a polymer/carbon nanotube
composite strain sensor,” Carbon, vol. 48, no. 3, pp. 680–687, 2010.
[22] R. Rahman and P. Servati, “Efficient analytical model of conductivity
of CNT/polymer composites for wireless gas sensors,” IEEE Trans. Peyman Servati (M’05) received the B.A.Sc. degree
Nanotechnol., vol. 14, no. 1, pp. 118–129, Jan. 2015. from the University of Tehran, Tehran, Iran, in 1998,
[23] D. W. Bonnell, Scanning Probe Microscopy and Spectroscopy. Theory, and the M.A.Sc. and Ph.D. degrees from the Electri-
Techniques, and Applications. New York, NY, USA: Wiley, 2001, p. 10. cal and Computer Engineering Department, Univer-
[24] J. Wang and T. Cui, “Design, simulation, fabrication, and characteriza- sity of Waterloo, Waterloo, ON, Canada, in 2000 and
tion of a PMMA tunneling sensor based on hot embossing technique,” 2004, respectively. Before joining The University of
Microsyst. Technol., vol. 11, no. 6, pp. 452–455, 2005. British Columbia (UBC), Vancouver, BC, Canada,
[25] Research Report, Nano Carbon Technol. Co., Ltd., Kawasaki, Japan, he was a Research Associate with the Centre for
2004. Advanced Photonics and Electronics, University of
[26] I. Balberg and N. Binenbaum, “Computer study of the percolation Cambridge, working on nanowire (NW) growth and
threshold in a two-dimensional anisotropic system of conducting sticks,” device engineering. Before joining the University of
Phys. Rev. B, vol. 28, p. 3799, Oct. 1983. Cambridge in 2005, he was a Senior Research Scientist with Ignis Innovation
[27] F. Du, J. E. Fischer, and K. I. Winey, “Coagulation method for prepar- Inc., a spin-off company of the University of Waterloo working on novel
ing single-walled carbon nanotube/poly(methyl methacrylate) compos- glass and plastic displays. He is currently an Associate Professor with the
ites and their modulus, electrical conductivity, and thermal stability,” Department of Electrical and Computer Engineering, UBC. His research
J. Polym. Sci. B, Polym. Phys., vol. 41, no. 24, pp. 3333–3338, 2003. interests include low-cost and flexible solar cells, flexible transistors and
[28] M. Taya, W. J. Kim, and K. Ono, “Piezoresistivity of a short electronics, growth and synthesis of semiconductor NWs and nanocomposites,
fiber/elastomer matrix composite,” Mech. Mater., vol. 28, nos. 1–4, electronic device modeling and engineering, and novel device engineering for
pp. 53–59, 1998. medical and energy applications.

S-ar putea să vă placă și