Thermal Conductivity of Polymer Nanocomposites

Made With Carbon Nanofibers

Sushant Agarwal,1 M. Masud K. Khan,2 Rakesh K. Gupta1

1
Department of Chemical Engineering, West Virginia University, Morgantown, West Virginia 26506

2
Faculty of Sciences, Engineering and Health, Central Queensland University, Rockhampton,
Qld-4702, Australia

An internal mixer was used to prepare polycarbonate in the aerospace, electronics, and energy storage industries
(PC)-based nanocomposites containing carbon fibers, [2]. They can be used as thermal interface materials or
carbon nanofibers (CNF), and mixtures of the two fill- heat sinks to dissipate heat from electronic devices such
ers. The influence of the filler volume fraction, the rela-
tive amounts of the two fillers, and the filler orientation as computers, flat panel displays, and cell phones [3]; in
relative to the direction of heat flow on the thermal some cases, it is the in-plane thermal conductivity that is
conductivity was examined. Filler orientation was important, while, in others, high values of the out-of-
obtained by the extrusion of strands of the nanocom- plane thermal conductivity are required. In principle,
posite. The thermal conductivity was measured using a proper alignment of the nanofillers can give the desired
steady-state heat conduction technique. The CNF were
fragile, and their aspect ratio could be decreased dur- directional properties, but it is not always easy to obtain
ing processing. In general, the composite thermal con- such alignment with the normal polymer processing meth-
ductivity increased with increasing filler content. Fiber ods of extrusion and injection molding.
alignment in the heat flux direction resulted in a sig- CNF are a form of multiwall carbon nanotubes
nificant increase in thermal conductivity. Mixing of (MWCNTs) which are synthesized by the catalytic
nanofibers with microfibers resulted contacts between
the microfibers. This, together with fiber alignment decomposition of aliphatic hydrocarbons. They differ
provided large increases in the thermal conductivity. from CNT in their diameter (dia.
100 nm) which is
POLYM. ENG. SCI., 48:2474–2481, 2008. ª 2008 Society of much larger than that of either single-wall nanotubes (dia.
Plastics Engineers
1 nm) or MWCNT (dia.
10 nm). In recent years, due
to advances in technology, CNFs have become readily
available in large quantities and at a price that is signifi-
INTRODUCTION cantly lower than that of CNTs. As a result, it is now pos-
sible to look for applications of polymer nanocomposites
Nanomaterials are increasingly being used as additives
made with CNFs. The thermal conductivity value of the
to polymers to simultaneously enhance a variety of pro-
CNFs is not precisely known, but it is thought to be very
perties without sacrificing any qualities of the base
high; its value should be bracketed by the thermal con-
polymer. Carbon nanotubes (CNT) and carbon nanofibers
ductivity of graphite (600 W/m K) and the corresponding
(CNF), for example, have very high stiffness values as
value of MWCNTs (3000 W/m K) [4]. Nonetheless, it is
well as very high thermal and electrical conductivities. As
expected that, by adding even a small amount of CNF to
a consequence, these nanofillers can be used to convert
polymers, a considerable enhancement in the thermal con-
thermally and electrically insulating plastics into conduc-
ductivity can be achieved. Thus, CNFs have been used to
tive materials [1] and to increase their mechanical proper-
make thermally-conducting nanocomposites with poly-
ties at the same time. This article is concerned with the
mers such as polypropylene, epoxy, polyimide, and ny-
thermal conductivity of polymer nanocomposites. Ther-
lon-11 [5–8]. However, it has been found that the thermal
mally conductive thermoplastics have many applications
conductivity of such nanocomposites is much lower than
what would be expected from the use of a mixture rule.
Correspondence to: Sushant Agarwal; e-mail: sushant.agarwal@mail. This mismatch is not surprising, and it results from the
wvu.edu very high interfacial resistance to heat transfer between
Contract grant sponsors: West Virginia EPSCoR Research Challenge the polymer matrix and the nanofibers [9]. It is for this
Grant Program, George and Carolyn Berry endowment.
DOI 10.1002/pen.21205
reason that the thermal conductivity of particulate compo-
Published online in Wiley InterScience (www.interscience.wiley.com). sites is not sensitive to the thermal conductivity value of
C 2008 Society of Plastics Engineers
V the filler, provided that the ratio of the filler to matrix

POLYMER ENGINEERING AND SCIENCE—-2008

thermal conductivity is large [10]. To increase the interfa- graphitic surface. These CNFs had a diameter in the range
cial interaction between the CNF and the polymer matrix, of 70–200 nm and lengths of tens of micrometers, and
acid treatment of the fibers has been attempted [7, 11]. their morphology was one of stacked conic sections. The
Results showed that the acid treatment promoted contacts fiber density is 1.99 g/cm3, and the fibers were highly
amongst the fibers and resulted in a higher thermal con- entangled with each other, forming aggregates. During
ductivity when compared with the untreated fibers. Other compounding, the CNFs had to be disentangled and
strategies to obtain higher thermal conductivities involve dispersed in order to obtain the best properties. In this
aligning the fibers in the direction of heat flow and the work, all these materials were used as obtained from the
use of a mixture of filler materials of different sizes and suppliers.
shapes. Kuriger and Alam [5] aligned pyrograf III carbon
fibers in a polypropylene matrix by extruding the compos-
ite through a die to obtain a strand and then compression Mixing
molded a bundle of these strands into an aligned sample.
A Thermo-Haake Polydrive 600 internal mixer was
Ghose et al. [7] used a polyimide resin and CNFs and
used to disperse CNFs, the carbon fibers and their mix-
extruded the composite in the form of a ribbon to align
tures in the PC matrix by melt mixing at 2508C. Before
the fibers. Cummins et al. [8] used nylon as the matrix
melt mixing, all the materials were dried overnight in an
and injection molding to align the fibers. These authors
oven at 1208C. To begin with, the required amount of
found that the thermal conductivity of the nanocomposites
polymer pellets was fed to the internal mixer, and, when
in the direction parallel to the fiber alignment was much
these pellets were completely melted, carbon fibers were
larger than that normal to it. Concerning the use of fillers
added and mixed for an additional 5 min at 60 rpm. After
of different sizes, Kim et al. [12] used a mixture of
mixing, the composites were removed from the mixer and
micron-sized carbon fibers and MWCNTs to make pheno-
stored in sealed plastic bags.
lic resin nanocomposites; the addition of 7 wt%
MWCNTs resulted in about a 57% increase in the thermal
conductivity. However, no fiber alignment was attempted.
Sample Preparation
It was postulated that the high aspect ratio nanofibers
formed a bridge between the microfillers, thus providing a Samples for thermal conductivity determination were
continuous path for heat conduction which resulted in the prepared in the form of circular disks of varying thick-
higher thermal conductivity. nesses, and these were used in the thermal conductivity
In this article, we studied the combined effect of fiber measurement cell described later. These samples con-
alignment and inter-fiber contacts through the use of a tained either one of the two carbon fibers or their mix-
hybrid mixture of microscopic carbon fibers and CNF. tures, and the sample composition was changed not only
Fiber alignment was achieved by extruding the strands of by varying the volume fraction of the filler(s) but also the
nanocomposite through a circular die and then compres- relative amounts of the two fillers. Samples were prepared
sion molding in the desired direction. in such a way that they contained fibers which were either
oriented isotropically or were aligned horizontally or ver-
tically with respect to the direction of heat flow. To
EXPERIMENTAL DETAILS obtain the isotropic samples, the composite, as obtained
from internal mixer, was dried and then molded into [1/
2]00 diameter disks of various thicknesses using a PHI hy-
Materials Used
draulic compression press at 2508C. To obtain fiber align-
Polycarbonate (PC) is commonly used in electronic ment, the isotropic composites were loaded into the barrel
applications, and this was used as the matrix polymer for of a Rosand-Bohlin capillary rheometer. A 2 mm circular
the nanocomposites. The specific material was Lexan 101, capillary die (aspect ratio ¼ 32) was installed and the
a general purpose PC that was purchased from GE Poly- temperature was set to 2808C. Strands were then extruded
merland. PC has a density of 1.2 g/cm3, a thermal con- into a water bath at a piston speed of 5 mm/min. This
ductivity of 0.29 W/m K, and it wets the surface of the was expected to result in the micro- or nanofibers being
carbon fibers easily. Two different carbon fillers were oriented along the axis of the strands. To make disks con-
used in the research. Thermalgraph DKD, Pitch-based taining fibers having a horizontal alignment, the solidified
carbon fibers, was obtained from Cytec Industries. As extruded strands were arranged flat on the surface of a
received, these fibers had a length of 200 lm and diame- mold and then compression molded. For samples with
ter of about 10 lm. Their thermal conductivity ranges vertically aligned fibers, the strands were held together
from 400 to 700 W/m K, and their density is 2.2 g/cm3. vertically and tightly packed in a vertical mold of circular
The CNF were Pyrograf III-PR19 PS, and they were pur- cross section and then compression molded into a speci-
chased from Applied Sciences. The term PS indicates that men of cylindrical shape; disks of various thicknesses
any deposited hydrocarbons on the surface of the fibers were then cut from this cylinder for thermal conductivity
had been stripped by pyrolytic stripping to expose the measurements. Figure 1 shows a schematic diagram of

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—-2008 2475

 the process of making samples having either horizontally
or vertically aligned carbon fibers.

Structure Characterization
Both transmission and scanning electron microscopy
were used to observe the morphology of the nano- and
microcomposites and to verify the expected fiber align-
ment. To prepare the samples for transmission electron
microscopy, a cryo-ultramicrotome with a diamond knife
was used to obtain 60–100 nm thick sections from the
compression molded sheets. These were then scanned at
100 kV accelerating voltage in a JEOL 100 CX transmis-
sion electron microscope (TEM). For scanning electron
microscopy, a Hitachi model S-4700 field emission scan-
ning electron microscope was used.

Thermal Conductivity Measurement

FIG. 1. Schematic diagram showing the preparation of horizontally and A steady-state heat-conduction technique similar to that
vertically aligned fiber samples. described in ASTM D5470 standard was used to measure

FIG. 2. Electron micrographs of polycarbonate-based nanocomposites: (a) SEM of top view of sample hav-
ing vertically aligned fibers. The sample has 30 vol% CF and 10 vol% CNF, (b) Same as (a), but sample has
horizontally aligned fibers, (c) TEM showing dispersion of CNFs in an isotropic sample containing 13.3
vol% CNF.

2476 POLYMER ENGINEERING AND SCIENCE—-2008 DOI 10.1002/pen

the thermal conductivities of the samples. The circular
disks of [1/2]00 diameter obtained according to the proce-
dure described earlier were sandwiched between two
copper rods. A temperature difference was imposed across
the assembly by contacting the end of the upper rod with
hot water at constant temperature, whereas the end of the
lower rod was brought into contact with cold water at a
constant temperature. The resulting (constant) temperature
gradient along the length of the copper rods was meas-
ured by the use of several thermocouples located at
known distances along the rods. A computer and data ac-
quisition system were used to record all the temperature
data. A knowledge of the thermal conductivity of copper,
in conjunction with the measured temperature gradient,
then gave the heat flux through the assembly using Four-
ier’s law. The thermal impedances was calculated and
plotted versus sample thickness. A straight line is fitted to
the data and the slope of the best-fit line was the recipro-
cal of the thermal conductivity of the sample.

RESULTS AND DISCUSSION

Figure 2 shows SEM and TEM pictures of the PC-
based nanocomposites. Figure 2a shows a low magnifica-
tion view of the cross section of a sample where vertical
fiber alignment was expected. This sample contains 30
vol% microfibers and 10 vol% nanofibers based on the
total sample volume. The carbon fibers are clearly visible,
and their cross section appears circular, indicating that
these fibers are aligned vertically. Similarly, Fig. 2b
shows horizontally aligned fibers demonstrating that our
technique of aligning the microfibers in a desired direc-
tion gives the expected results. Finally, a TEM of a sam-
ple containing 13.3 vol% of isotropically distributed
CNFs is shown in Fig. 2c, and one can see a variety of
fiber orientations. Note that white areas in this micrograph
represent voids that resulted from microtoming. Figure 3
shows TEM pictures of the top view of three vertically
aligned hybrid samples. One can see that the interstitial
space between microscopic carbon fibers is filled with dis-
persed CNF. Some of these nanofibers are in contact with
the carbon fibers, thus providing better contact between
the fibers. One also notices that the aspect ratio of the
nanofibers has been reduced during processing, and this
observation is consistent with results reported in the liter-
ature. From these electron micrographs, however, it can
be unequivocally concluded that the desired microfiber
alignment was successfully achieved with our sample
preparation technique, though CNF were not significantly
aligned. This is to be expected because, whereas individ- FIG. 3. TEM pictures of vertically aligned samples of (a)10 vol% CF
and 30 vol% CNF, (b) 20 vol% CF and 20 vol% CNF, and (c) 30 vol%
ual carbon fibers are straight circular cylinders, the CNF
CF and 10 vol% CNF. Pictures show that CNFs are dispersed in the
are irregularly curved along the length, and are highly areas between microscopic carbon fibers.
entangled with each other.
Figure 4 shows the experimentally measured values of about 40 vol% because, beyond this value, the composites
the thermal conductivity of the composites containing car- become very brittle. For isotropically distributed fibers, it
bon microfibers and CNF. The maximum loading level is can be seen that the thermal conductivity increased with

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—-2008 2477

 contacts being formed between the microfibers due to the
presence of the nanofibers. Figure 5 also suggests that, in
order to obtain large thermal conductivity values, it is
preferable to use equal volumes of the two fibers or even
a disproportionate amount of the nanofibers.

Model Predictions
Many theoretical models are available to predict the
thermal conductivity of nanocomposites. The simplest of
these are mixture rules such as series, parallel, and geo-
metric models, which are given by [15]:
FIG. 4. Thermal conductivity of polycarbonate-based microcomposites
and nanocomposites as a function of CF and CNF volume% for isotropi- Series model
cally and vertically aligned fibers.
kc ¼ ð1  fÞkm þ fkf ; (1)

parallel model
increasing concentration of both CF and CNF. Although a
very significant increase in the thermal conductivity of 1 1f f
¼ þ (2)
the PC can be achieved by adding 40 vol% of carbon kc km kf
fibers, the magnitude of the increase is similar to that
attained with the use of high thermal conductivity particu- and geometric mean model
lates [10]. It can also be seen that, at the lower concentra-
1f f
tion levels, isotropic nanocomposites have a higher ther- kc ¼ km kf : (3)
mal conductivity than isotropic microcomposites. At the
30 vol% filler level, however, the thermal conductivity of where, kc is the thermal conductivity of the composite
the nanocomposite began to decrease. A decrease in ther- and km and kf are thermal conductivities of the polymer
mal conductivity of nanocomposites with nanofiller con- matrix and filler material, respectively. However, the se-
centration has also been observed in other studies [12– ries model typically over predicts the thermal conductiv-
14]. We speculate that this was due to fiber breakage. ity, whereas the parallel model tends to under predict the
Figure 4 also shows thermal conductivity values for thermal conductivity of the nanocomposites. This is
microcomposites having fibers (vertically) aligned in the because these models are based on the amount of filler
direction of heat flow. It can be seen that these thermal loading and do not take into account the true geometry
conductivities are much higher than those for isotropic and the size of the filler particles. Other models such as
samples and fiber alignment resulted in a doubling of the the Hamilton-Crosser model [16] and the Lewis-Nielsen
thermal conductivity at each loading level. We also men- [17] model are based on particle size, geometry, and the
tion that Fig. 4 does not display data for nanocomposites
having vertically aligned fibers. Results did not seem to
be very sensitive to fiber orientation (see datum point at
30 vol% loading in Fig. 5), and it is suspected that with
CNF a high degree of alignment is difficult to achieve
due to irregular and entangled nature of the nanofibers as
evidenced by electron micrographs.
Thermal conductivities of the nanocomposites contain-
ing a mixture of microscale and nanoscale carbon fibers
are shown in Fig. 5. The relative amounts of CFs and
CNFs was varied, but the total filler concentration was
maintained at either 30 or 40 vol%. Results are presented
for isotropic and vertical alignments in each case, and for
horizontal alignment in some cases. In the case of vertical
fiber alignment, at both the 30 and 40 vol% total filler
levels, there was a dramatic increase in the nanocompo-
site thermal conductivity (relative to either the horizontal FIG. 5. Thermal conductivity as a function of filler concentration for
nanocomposites containing isotropic, horizontal, and vertically-aligned
or isotropic alignment) upon mixing the two kinds of fibers. X-axis shows percent volume fractions of CF and CNF in the
fibers. Also, the relative increase was greater at the higher nanocomposite (aCFbCNF represents a %v carbon fibers and b %v
loading level. We suspect that this is the consequence of carbon nanofibers).

2478 POLYMER ENGINEERING AND SCIENCE—-2008 DOI 10.1002/pen

manner of particle packing in the matrix. The expression
for the Hamilton-Crosser model is given by [16]:


kf þ ðn  1Þkm  ðn  1Þfðkm  kf Þ
kc ¼ km ; (4)
kf þ ðn  1Þkm þ fðkm  kf Þ

where kc, km, and kf are thermal conductivities of compos-

ite, matrix material, and fibers, respectively. f is the vol-
ume fraction of the filler material and n is a geometric
factor based on sphericity of the particles.
The Lewis-Nielsen model is given by [17]


1 þ ABf
kc ¼ km ; (5)
1  Bcf

where

kf=
km  1
B¼ and (6)
kf=
km þ A
!
1  fm
c¼1þ f: (7)
f2m

where A is a factor akin to Einstein’s coefficient, but it

also depends on the alignment of fibers with respect to
the heat flow and the aspect ratio. For example, for iso-
tropically oriented long fibers (l/d . 15) A is 8.38. Simi-
larly, for heat flow parallel to the fibers, A is 2l/d while
for heat flow perpendicular to fibers, it is 0.5. fm is the FIG. 6. Comparison of experimental values of isotropic thermal con-
ductivity of polycarbonate-based composites with theoretical predictions
maximum packing fraction for the particles in the matrix.
for (a) Carbon nanofiber nanocomposites, Prasher model 1 is for a ¼
For three dimensional random packing this quantity is 0.15 and Model 2 is for a ¼ 0.06, and (b) Carbon fiber microcompo-
0.52, whereas for uniaxial random packing it is 0.82 [17]. sites.
Figure 6 compares experimental data for both the
nanocomposites and the microcomposites made with ran-
domly oriented fibers in PC with a variety of theoretical of interfacial boundary thickness and the particle size.
predictions. For the microcomposite, the geometric and Their result is
Lewis-Nielsen models seemed to over predict the meas-
urements, whereas the parallel model under predicted the km
kc ¼ 3ð1aÞ
; (8)
results. The Hamilton-Crosser model appeared to provide ð1  fÞð1þ2aÞ
fairly close predictions when n was chosen to equal to 6,
which was for long fibers. For the nanocomposite sam-
ples, however, none of the models provided a satisfactory In which a is the Biot number defined as
prediction of thermal conductivity. This may be due to
the fact that the experimental data are themselves not Rb k m
internally consistent because the thermal conductivity is a¼ ; (9)
d
found to decrease with increasing filler concentration. It
has also been observed that nanomaterials, despite having where Rb is the interfacial resistance and d is the diameter
a very large thermal conductivity, do not, in general, of the nanoparticles. This equation shows that kc increases
impart the expected enhancement in thermal conductivity. with a decrease in the value of a, which happens if inter-
This behavior is attributed to their very large surface area facial resistance decreases or the particle size increases.
which results in a large interfacial resistance to heat flow Figure 6a shows theoretical predictions for a ¼ 0.15
between the nanoparticle surface and the matrix material (shown as model 1) and a ¼ 0.06 (shown as model 2),
due to incomplete bonding and phonon acoustic mismatch representing high and low thermal interfacial resistance.
[18, 19]. Prasher et al. [19] have proposed a model based This might be used to explain the observed decrease in
on Every’s model [20] that takes into account the effect the thermal conductivity at 30 vol% filler loading in the

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—-2008 2479

 As noted earlier, a very significant increase in the ther-
mal conductivity is observed if a hybrid mixture of CF and
CNF is used. To explain this increase in thermal conductiv-
ity of the hybrid nanocomposites, one may adapt the
Lewis-Nielsen model to the situation at hand to obtain [21]
" P2 #
1þ Ai Bi f i
kc ¼ km P2i¼1 : (10)
1 i¼1 Bi ci fi

In above expression A, B, /, and w are defined as ear-

lier in Eqs. 6 and 7 with components i ¼ 1 and 2 repre-
senting CF and CNF, respectively. Table 1 shows the
experimental values and the corresponding theoretical pre-
dictions for hybrid nanocomposites for isotropic, horizon-
FIG. 7. Comparison of experimental values of thermal conductivity of
polycarbonate-based microcomposites containing vertically aligned fibers tal, and vertically aligned fibers. For CF, 2l/d was chosen
with theoretical predictions of the Lewis-Nielsen Model. Experimental to be either 20 or 10 whereas for CNF, 2l/d was taken to
values are represented by solid circles in this graph. be either 50 or 100. These values were chosen from the
fiber size observed from the electron micrographs. In each
case, the total volume fraction of filler was kept constant
nanocomposites. Indeed, a leveling off in the thermal con-
at 0.4. It can be seen that for vertically aligned fibers, the
ductivity with nanoparticle loading or even a decrease has
measured values are close to or in between the theoretical
been observed by others as well [12–14, 19]. It has been
values for the small and large aspect ratios. However, for
suggested that, at higher concentration levels of nanofib-
isotropic and horizontal cases, the agreement was not very
ers, there is incomplete wetting of the nanofiber surfaces
good. Examining the theoretical values further, it should
leading to higher interfacial resistance and a lower ther-
be noticed that, for vertically aligned fibers, the theo-
mal conductivity [19].
retical thermal conductivity increases monotonically with
To conclude this section, Fig. 7 shows a comparison
increasing CNF fraction in the mixture whereas for the
between experimental values of thermal conductivities for
isotropic fibers it seems to go through a minimum. These
vertically aligned carbon fibers and theoretical predictions
trends do not always agree with the experimental results
of the Lewis-Nielsen model for aspect ratios (2l/d) of 50
and point to the need for better predictive equations.
and 10. These values are based on electron micrographs
of the samples. It can be seen that the predictions closely
track the experimental values for 2l/d ¼ 10. However, if
CONCLUSIONS
one applies the same model to nanocomposites having
vertically aligned fibers and containing 30 vol% CNF, the High thermal conductivity polymer nanocomposites
predicted value of the thermal conductivity is 1.66 W/m can be obtained by using a hybrid mixture of microscopic
K for 2l/d ¼ 10 and 6.5 W/m K for 2l/d ¼ 50; both these carbon fibers and CNF. In general, thermal conductivity
numbers are much higher than the observed experimental values increase as the filler volume fraction increases.
value of 0.88 W/m K. This again suggests that the nano- Alignment of the microfibers in the direction of the heat
fibers are not significantly aligned. flow increases the thermal conductivity. Nanocomposites

TABLE 1. Thermal conductivities of hybrid samples and comparison with theoretical predictions.

Thermal conductivity (W/m K)

Model vertical Model vertical
CNF CF Exp., (CF 2l/d ¼ 20, (CF 2l/d ¼ 10, Exp., Model Exp., Model
vol fr vol fr vertical CNF 2l/d ¼ 50) CNF 2l/d ¼ 100) isotropic isotropic horizontal horizontal

0.01 0.39 4.66 2.95 2.90 0.60

0.05 0.35 5.19 4.70 2.69 0.59
0.10 0.30 8.6 5.86 6.84 2.1 2.53 1.3 0.59
0.15 0.25 6.55 8.97 2.44 0.58
0.20 0.20 11.4 7.26 11.13 1.3 2.41 1.6 0.58
0.25 0.15 8.00 13.34 2.44 0.58
0.30 0.10 11.4 8.79 15.63 1.5 2.53 1.9 0.59
0.35 0.05 9.64 18.04 2.69 0.59
0.39 0.01 10.36 20.08 2.90 0.60
0.20 0.10 5.1 5.30 8.97 1.0 1.53 0.47

2480 POLYMER ENGINEERING AND SCIENCE—-2008 DOI 10.1002/pen

containing hybrid mixtures of two fillers show higher 8. A.T. Cummings, L. Shi, and J.H. Koo, Proceedings of
thermal conductivity due to increased contacts between IMECE2005, Nov 5–11, Orlando, FL, 773 (2005).
the microfibers that are promoted by the presence of the 9. C.W. Nan, R Birringer, D.R. Clarke, and H. Gleiter, J. Appl.
nanofibers, especially when microfibers are aligned. Phys., 81, 6692 (1997).
10. J. Rajaiah, G. Andrews, E. Ruckenstein, and R.K. Gupta,
Chem. Eng. Sci., 47, 3863 (1993).
ACKNOWLEDGMENTS
11. G.W. Lee, J.I. Lee, S.S. Lee, M. Park, and J. Kim, J. Mater.
Professor M.M.K. Khan spent two months as a Visiting Sci., 40, 1259 (2005).
Professor at West Virginia University at the initiation of 12. Y.A. Kim, S. Kamio, T. Tajiri, T. Hayashi, S.M. Song, M.
this work. Endo, M. Terrones, and M.S. Dresselhaus, Appl. Phys. Lett.,
90, 093125 (2007).
REFERENCES 13. C. Guthy, F. Du, S. Brand, K.I. Winey, and J.E. Fischer,
J. Heat Transfer, 129, 1096 (2007).
1. O. Breuer and U. Sundaraj, Polym. Comp., 25, 630 (2004). 14. O. Shenderova, T. Tyler, G. Cunningham, M. Ray, J. Walsh,
2. J. Njuguna and K. Pielichowski, Adv. Eng. Mat., 5, 769 S.H.M. Casulli, G. McGuire, V. Kuznetsov, and S. Lipa,
(2003). Diamond Relat. Mater., 16, 1213 (2007).
3. R. Prasher, J.Y. Chang, I. Sauciuc, S. Narsimhan, D. Chau, 15. R.C. Progelhof, J.L. Throne, and R.R. Ruetsch, Polym. Eng.
G. Chrysler, A. Myers, S. Prstic, and C. Hu, Intel Technol. Sci., 16, 615 (1976).
J., 9, 285 (2005). 16. R.L. Hamilton and O.K. Crosser, Ind. Eng. Chem. Fundam.,
4. P. Kim, L. Shi, A. Majumdar, and P.L. McEuen, Phys. Rev. 1, 187 (1962).
Lett., 87, 215502–1, (2001). 17. L.E. Nielsen, Ind. Eng. Chem. Fundam., 13, 17 (1974).
5. R.J. Kuriger and M.K. Alam, Exp. Heat Trans., 15, 19 18. C.W. Nan, G. Liu, Y. Lin, and M. Li, Appl. Phys. Lett., 85,
(2002). 3549 (2004).
6. M.J. Biercuk, M.C. Llaguno, M. Radosavljevic, J.K. Hyun, 19. R.S. Prasher, J. Shipley, S. Prstic, P. Koning, and J.L.
T.J. Johnson, and J.E. Fischer, Appl. Phys. Lett., 80, 2767 Wang, J. Heat Transfer, 125, 1170 (2003).
(2002). 20. A.G. Every, Y. Tzou, D.P.H. Hasselman, and R. Raj, Acta.
7. S. Ghose, D.C. Working, J.W. Connell, J.G. Smith Jr., K.A. Metall. Mater., 40, 123 (1992).
Watson, D.M. Delozier, Y.P. Sun, and Y. Lin, High Perf. 21. E.H. Weber, M.L. Clingerman, and J.A. King, J. Appl.
Polm., 18, 961 (2006). Polym. Sci., 88, 112 (2003).

DOI 10.1002/pen POLYMER ENGINEERING AND SCIENCE—-2008 2481