Content uploaded by Karen I Winey
Author content
All content in this area was uploaded by Karen I Winey on Jun 19, 2018
Content may be subject to copyright.
Flammability properties of polymer nanocomposites with single-walled
carbon nanotubes: effects of nanotube dispersion and concentration
*
Takashi Kashiwagi
a,
*, Fangming Du
b
, Karen I. Winey
c
, Katrina M. Groth
a
, John R. Shields
a
,
Severine P. Bellayer
a
, Hansoo Kim
c
, Jack F. Douglas
d
a
NIST, Building and Fire Research Laboratory, Fire Research Division, Gaithersburg, MD 20899, USA
b
University of Pennsylvania, Chemical and Biomolecular Engineering Department, Philadelphia, PA 10194, USA
c
University of Pennsylvania, Materials Science and Engineering Department, Philadelphia, PA 10194, USA
d
NIST, Materials Science and Engineering Laboratory, Polymers Division, Gaithersburg, MD 20899, USA
Received 9 September 2004; received in revised form 20 October 2004; accepted 21 October 2004
Available online 26 November 2004
Abstract
The effects of the dispersion and concentration of single walled carbon nanotube (SWNT) on the flammability of polymer/SWNT
nanocomposites were investigated. The polymer matrix was poly (methyl methacrylate) (PMMA) and the SWNT were dispersed using a
phase separation (‘coagulation’) method. Dispersion of SWNTs in these nanocomposites was characterized by optical microscopy on a
micrometer scale. Flammability properties were measured with a cone calorimeter in air and a gasification device in a nitrogen atmosphere.
In the case where the nanotubes were relatively well-dispersed, a nanotube containing network structured layer was formed without any
major cracks or openings during the burning tests and covered the entire sample surface of the nanocomposite. However, nanocomposites
having a poor nanotube dispersion or a low concentration of the nanotubes (0.2% by mass or less) formed numerous black discrete islands
with vigorous bubbling occurring between these islands. Importantly, the peak heat release rate of the nanocomposite that formed the
network layer is about a half of those, which formed the discrete islands. It is proposed that the formation of the discrete islands is due to
localized accumulation of the nanotubes as a result of fluid convection accompanying bubble formation and rise of the bubbles to the surface
through the molten sample layer and bursting of the bubbles at the surface. The network layer acts as a heat shield to slow the thermal
degradation of PMMA.
q2004 Elsevier Ltd. All rights reserved.
Keywords: Nanocomposite; Single walled carbon nanotube; Flammability
1. Introduction
There is a high level of interest in using filler particles
having at least one nano-dimensional scale (nanofiller) for
making polymeric nanocomposite materials with excep-
tional properties [1,2]. One of the promising applications
involves the improvement in flammability properties of
polymers with nanofillers because one weak aspect of
polymers is that they are combustible under certain
conditions. These filled systems are attractive as possible
alternatives to conventional flame retardants and further-
more they could simultaneously improve both physical and
flammability properties of polymers. At present, the most
common approach using nano scale particles is the use of
layered silicates having large aspect ratios. The flame
retardant (FR) effectiveness of clay-polymer nanocompo-
sites with various resins has been demonstrated and several
flame retardant mechanisms have been proposed [2–10].It
appears that the flammability properties of clay-polymer
nanocomposites are not significantly affected by whether
they are intercalated or exfoliated as long as they are
nanocomposites rather than microcomposites. Significant
reduction in heat release rate has been achieved with a clay
content of about 5% by mass.
0032-3861/$ - see front matter q2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.polymer.2004.10.087
Polymer 46 (2005) 471–481
www.elsevier.com/locate/polymer
This is a publication of the National Institute of Standards and
Technology (NIST), an agency of the US Government, and by statute is not
subject to copyright in the US.
* Corresponding author. Tel.: C1 301 9756699; fax: C1 301 9754052.
E-mail address: takaswhi.kashiwagi@nist.gov (T. Kashiwagi).
Carbon nanotubes are another candidate as a FR additive
because of their highly elongated shape (high aspect ratio).
This was demonstrated by using multi-walled carbon
nanotubes (MWNT) in polypropylene (PP) [11,12] and
also in poly(ethylene vinyl acetate) [13]. The in situ
formation of a continuous, network structured protective
layer from the tubes is critical for significant reduction in
heat release rate, because the layer thus acts as a thermal
shield from energy feedback from the flame [12]. Single-
walled nanotubes also have potential as flame retardants by
the same mechanism. Despite reports of the exceptional
physical properties of the nanocomposites with SWNT [14–
17], there are no published studies on the flammability of
SWNT polymer nanocomposites. The dispersion of SWNT
in polymers remains a challenge, so it is important to
determine the effects of the nanotube dispersion on
flammability properties. Thus, we investigate the effects of
small quantities of single-walled carbon nanotubes and their
dispersion in PMMA on the flammability properties of these
nanocomposites.
2. Experimental
2.1. Materials
The matrix polymer used in this paper is poly(methyl
methacrylate) (PMMA) (Polysciences,
1
M
w
: 100,000 g/
mol). SWNTs for the nanocomposites were synthesized by
the high-pressure carbon monoxide method (HiPCo)[18].
The metal residue in the SWNTs is less than 13% by mass.
The coagulation method was used to produce the SWNT/
PMMA nanocomposites [17]. In the coagulation method,
dimethylformamide (DMF) was chosen to dissolve the
PMMA and to permit dispersion of the SWNT by bath
sonication for 24 h. To obtain good nanotube dispersion, the
nanotube concentration in DMF is critical. Our atomic force
microscopy (AFM) results show that the average nanotube
bundle diameter increases with an increase of the nanotube
concentration in DMF, from w7 nm at the concentration of
0.2 mg (SWNT) per ml (DMF) to w13 nm at 0.4 mg/ml.
We can observe nanotube agglomerates at a concentration
higher than 0.4 mg/ml by the naked eye, while the 0.2 mg/
ml suspension is visually homogeneous. Therefore, we can
control the nanotube dispersion in the nanocomposites by
changing the nanotube concentration in DMF, assuming that
the state of nanotube dispersion is comparable in DMF
before coagulation and in the polymer matrix after
coagulation suspension [19]. Concentrations of 0.2 and
1 mg/ml were used to make nanocomposites with good and
poor dispersion, respectively. The content of the nanotubes
in the nanocomposites varied from 0.2% to 1% by mass. The
notation of the PMMA/SWNT(0.5%) means that the sample
contains 0.5% by mass of SWNT in PMMA.
All samples were compression molded at 210 8C under a
pressure of 6 metric tons to make 75 mm diameter by 4 mm
thick disks for the measurement of heat release rate and
8 mm thick disks for the gasification measurement in a
nitrogen atmosphere.
2.2. Sample characterization
The morphologies of the nanotubes in PMMA were
examined in transmission by optical microscopy. A hot
press was used to prepare nanocomposite films of w30 mm
in thickness, which were examined by an optical micro-
scope (Olympus, BH-2) with a magnification of 200 to study
the global dispersion of SWNTs. The morphology of the
nanocomposites residues after the nitrogen gasification tests
was investigated using scanning electron microscopy
(SEM) (JEOL 6300FV, at 5 kV). The transmission electron
microscopy (TEM), JEOL 2010 operated at 200 kV, was
used for original SWNT and collected residues after the
gasification tests. The collected residue was gently crushed
with a mortar and dispersed in ethanol with a 15 min
sonication. The bright field (BF) TEM images were
recorded with a charge-coupled device (CCD) with the
resolution of 1024!1024 pixels.
Thermal gravimetric analyses (TGA) were conducted
using a TA Instruments TGA Q 500 at 5 8C/min from 90 to
500 8C in nitrogen (flow rate of 60 cm
3
/min) for the original
nanocomposite samples (w8 mg) in a platinum pan and
from 90 to 900 8C in nitrogen for the SWNTs and for the
residues collected after nitrogen gasification tests. The
standard uncertainty in sample mass measurement is G1%.
The complex viscosity of the sample was measured using a
Paar Physica UDS 200 Rheometer. The sample was located
between a stationary and an oscillating plate at 0.1 rad/s
(low shear simulating burning condition of the sample) from
190 to 280 8C at an a heating rate of 1 8C/min in nitrogen.
2.3. Flammability property measurement
A cone calorimeter built by NIST was used to measure
ignition characteristics, heat release rate, and sample mass
loss rate according to ASME E1354/ISO 5660. An external
radiant heat flux of 50 kW/m
2
was applied. All of the
samples were measured in the horizontal position and
wrapped with thin aluminum foil except for the irradiated
sample surface. The standard uncertainty of the measured
heat release rate was G10%.
A radiant gasification apparatus, somewhat similar to a
cone calorimeter, was designed and constructed at NIST to
study the gasification processes of samples by measuring
mass loss rate and temperatures of the sample exposed to a
fire-like heat flux in a nitrogen atmosphere (no burning).
The apparatus consists of a stainless-steel cylindrical
1
Certain commercial equipment, instruments, materials, services or
companies are identified in this paper in order to specify adequately the
experimental procedure. This in no way implies endorsement or
recommendation by NIST.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481472
chamber that is 1.70 m tall and 0.61 m in diameter. In order
to maintain a negligible background heat flux, the interior
walls of the chamber are painted black and the chamber
walls are water-cooled to 25 8C. All experiments were
conducted at 50 kW/m
2
. The unique nature of this device is
threefold: (1) observation and results obtained from it are
only based on the condensed phase processes due to the
absence of any gas phase oxidation reactions and processes;
(2) it enables visual observations of gasification behavior of
a sample using a video camera under a radiant flux similar to
that of a fire without any interference from a flame; (3) the
external flux to the sample surface is well-defined and
nearly constant over the duration of an entire experiment
(and over the spatial extent of the sample surface) due to the
absence of heat feedback from a flame. A more detailed
discussion of the apparatus is given in our previous study
[20]; the standard uncertainty of the measured mass loss rate
is G10%.
3. Results
3.1. Sample morphology
The distribution of the nanotubes in
PMMA/SWNT(0.5%) was examined by optical microscopy
to globally observe the dispersion of the nanotubes, as
shown in Fig. 1.Fig. 1(a) indicates that the nanotubes are
relatively uniformly distributed within the polymer matrix
on a micrometer scale. By using a higher concentration of
SWNT in the DMF suspension, the sample in Fig. 1(b)
shows regions of nanotube aggregation. In this study, the
former sample is designated as having ‘good dispersion’ and
the latter sample is designated as having ‘poor dispersion’.
The TEM image of the purified original SWNT shows many
nanotube bundles with a small amount of amorphous carbon
and of large carbon fullerenes with many iron particles in
the nanotubes from the residual catalyst, as shown in Fig.
1(c). The SEM image shown in the previous study indicates
that approximately 20 nm diameter nanotubes bundles such
as shown in Fig. 1(c) are uniformly distributed on a sub-
micrometer scale [17].
3.2. Thermal stability
Thermal gravimetric analysis was conducted in nitrogen.
Although previous studies did not conclusively exclude the
effects of oxygen in surrounding air on thermal degradation
of polymeric materials during burning of the polymers,
oxidation reactions of the polymers appear to be insignif-
icant (oxygen is mainly consumed by gas phase reactions
i.e., the flame). Exception is the case in which the flame does
not cover the entire burning surface or the burning/pyrolysis
rate is extremely low [21,22]. Either addition of SWNTs to
PMMA or the distribution of nanotubes in PMMA does not
show significant effects on the thermal stability of PMMA,
as shown in Fig. 2. On the other hand, results previously
published for the acrylonitrile–butadiene–styrene (ABS)/
SWNT samples showed that the addition of SWNT reduced
the thermal stability of ABS at higher mass fractions of
nanotubes than those used in this study [23].
3.3. Complex melt viscosity
During a nitrogen gasification tests, both liquid-like
behavior and solid-like behavior of the nanocomposite
samples were observed as described later in this paper.
Thus, complex viscosities of the PMMA/SWNT samples
were measured as a function of temperature (at a heating
rate of 1 8C/min in nitrogen) at a constant frequency rate of
0.1 rad/s which was selected as representing the low shear
rates during burning. The results are shown in Fig. 3.
Complex viscosity significantly increases with an increase
of SWNT content especially at higher temperatures. With
1% by mass of SWNT, complex viscosity of the
nanocomposite increases at least one order above that of
pure PMMA. The slope of the relationship between complex
viscosity and temperature tends to gradually decrease with
an increase of SWNT content in the nanocomposite.
Furthermore, at the composition of 0.5% by mass the
nanocomposite with poor nanotube dispersion exhibits
much lower complex viscosity than that with good nanotube
dispersion at high temperatures.
3.4. Effects of dispersion of SWNT on flammability
properties
Heat release rate histories of the three different samples,
PMMA, PMMA/SWNT(0.5%) and PMMA/SWNT(0.5%,
poor dispersion), were measured at an external radiant flux
of 50 kW/m
2
; the results are shown in Fig. 4. The heat
release rate of the sample with good nanotube dispersion of
the nanotubes is much lower than those of pure PMMA and
of the sample with poor dispersion. The heat release rate of
the sample with poor dispersion is not appreciably reduced
from that of PMMA. However, the total heat releases of all
samples (integrated value of heat release rate with respect to
time) are comparable. This indicates that the sample with
relatively good nanotube dispersion burns much slower than
that with poor nanotube dispersion but both samples
eventually burn almost completely.
It is important to understand how the difference in
dispersion of the nanotubes affects heat release rate of the
nanocomposite. In order to answer this question, the
behavior of the two samples during the gasification test in
a nitrogen atmosphere was observed by taking video
movies. Selected pictures from the video images are
shown in Fig. 5. For the sample with good nanotube
dispersion, numerous small bubbles formed initially and
their bursting was observed at the surface. This was shortly
followed by a solid-like behavior with no overt fluid motion.
A slightly wavy solid surface and gradual receding of the
T. Kashiwagi et al. / Polymer 46 (2005) 471–481 473
sample surface were observed over the test period. The
sample with poor nanotube dispersion formed initially many
small bubbles and their bursting at the surface was followed
by the formation of many small black discrete islands.
Vigorous bubbling was subsequently observed between the
islands. At a later time, the islands coalesced into a
connected structure and their size gradually increased
during the course of the test. The mass loss rate curves of
samples with good and poor nanotube dispersion are plotted
in Fig. 6 along with that of pure PMMA for comparison. The
results clearly show that the dispersion of the nanotubes is
critical for reduction in mass loss rate. This trend is the same
as that observed for the heat release rate curves, as shown in
Fig. 4. This confirms that the effects of the dispersion of the
nanotubes on flammability properties are based on a
physical or chemical process in the condensed phase.
The residues of PMMA/SWNT(0.5%) were collected
after the completion of the gasification tests (8 mm thick
sample) and also of cone calorimeter (burning) tests (4 mm
thick sample). The pictures of the residues are shown in Fig.
7. Both residues of the samples with good dispersion of
nanotubes were covered by a continuous dark layer
compared to many black islands for the residues of the
samples with poor nanotube dispersion. Glowing combus-
tion of the residues shortly after the burning tests was
observed. The pictures of the burned residues show faint
orange color and a small amount of ash. Similar observation
was made for the PP/MWNT nanocomposites and glowing
Fig. 1. Optical micrographs of PMMA/SWNT(0.5%) with two different dispersion of nanotubes, (a) ‘good dispersion’ and (b) ‘poor dispersion’ with numerous
agglomerates. (c) TEM image of original SWNT.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481474
combustion was due to oxidation of MWNTs with
surrounding oxygen in air (no more flaming at the end of
the test), which was enhanced by iron from the residual
catalyst for synthesizing the nanotubes [12]. The orange
color of the burned residues was caused by the formation of
iron oxide. It is assumed that observed glowing combustion
in this study is also due to the oxidation enhanced by the
residual iron in the original SWNT as seen in Fig. 1(c).
Smaller islands in the burned residue (Fig. 7(d)) that those in
the gasified residue in nitrogen (Fig. 7(c)) were due to
glowing combustion at the end of the burn test, which
consumed some of the residue during cooling after
removing the residue from the cone calorimeter.
3.5. Effects of SWNT concentration on flammability
properties
Heat release rate curves of various PMMA/SWNT
nanocomposites having good dispersion of the nanotubes
were measured; the results are shown in Fig. 8. The addition
of 0.1% by mass of SWNT did not significantly reduce the
Fig. 2. Derivative thermogravimetric mass loss rates of selected samples at
heating rate of 5 8C/min in nitrogen.
Fig. 3. Effects of the addition of SWNT on complex viscosity of PMMA at
0.1 rad/s in nitrogen.
Fig. 4. Effect of SWNT dispersion on heat release rate of
PMMA/SWNT(0.5%) nanocomposite at external radiant flux of 50 kW/m
2
.
Fig. 5. Selected video images of PMMA/SWNT(0.5%) during gasification
tests at 50 kW/m
2
in nitrogen; (a) with good nanotube dispersion and (b)
with poor nanotube dispersion.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481 475
heat release rate of PMMA. A roughly 25% reduction in the
peak heat release rate was achieved by the addition of 0.2%
by mass of the nanotubes. The most reduction in heat release
rate was achieved by 0.5% by mass. All nanocomposite
samples ignited earlier than pure PMMA and this trend is
due to the increase in surface absorptivity of the
nanocomposites compared to lower surface absorptivity of
PMMA (by absorption bands) with respect to broad
emission spectra (grey body) of the cone calorimeter heater
[12].
The sample behavior was observed during the gasifica-
tion test in a nitrogen atmosphere to understand the effects
of the concentration of the nanotubes on mass loss rate
curve. As expected, the PMMA sample melted and behaved
like a liquid accompanied by numerous bubbles and their
bursting at the sample surface, as shown in Fig. 9(a). At the
end of the test, no residue was left behind as shown in Fig.
10(a). For the nanocomposite sample with 0.2% mass
fraction of the nanotubes, many small, black discrete islands
were formed after initial numerous small bubbles and their
bursting at the surface. Bubbling was observed between
islands and it appeared that bubbling pushed nanotubes to
the islands and the size of islands gradually became larger
and eventually some of the islands were connected to each
other. The connected black islands were left behind at the
end of the test, as shown in Fig. 10(b). This picture is very
similar to the one for PMMA/SWNT(0.5%, poor dis-
persion), as shown in Fig. 7(c). For the samples with 0.5%
by mass and 1% by mass, both samples appeared to be solid-
like; a network layer covered the sample surface during the
entire test period and was left behind without any major
cracks at the end of the test, as shown in Figs. 10(c) and (d).
The measured mass loss rate curves of all samples tested
in a nitrogen atmosphere are shown in Fig. 11. The trend of
Fig. 6. Effect of SWNT dispersion on mass loss rate of PMMA/SWNT at
50 kW/m
2
in a nitrogen atmosphere.
Fig. 7. Effects of dispersion of nanotubes on pictures of residues of PMMA/SWNT(0.5%) (note 8 mm thick sample for the gasification test and 4 mm thick for
the burning test); (a) nitrogen gasification residue, good dispersion sample (b) burned residue, good dispersion sample (c) nitrogen gasification residue, poor
dispersion sample (d) burned residue, poor dispersion sample.
Fig. 8. Effects of SWNT concentration on heat release rate curve of
PMMW/SWNT at 50 kW/m
2
.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481476
the curves is similar to that of the heat release rate curves
shown in Fig. 8. This clearly indicates that the observed
effects of the concentration of SWNT on the flammability
properties are based on physical processes in the condensed
phase (thermal stability results of the nanocomposites show
little effect due to the concentration of the tubes, as shown in
Fig. 2).
4. Discussion
The formation of a protective network layer covering the
entire surface without any cracks or openings is critical for
reducing the heat release rate and mass loss rate of the
nanocomposites. Therefore, it is important to understand
how the black discrete islands are formed instead of the
formation of the continuous layer and how to avoid them. In
the early stage of the gasification test, the upper part of the
sample is heated and starts melting. When the temperature
of the sample becomes high enough, degradation starts to
generate methyl methacrylate (MMA) as the main degra-
dation product [24]. Because the degradation temperature of
PMMA (shown in Fig. 2) is much higher than the boiling
temperature of MMA (100 8C), MMA is superheated and
nucleates, forming bubbles in the melt layer. With the
addition of the nanotubes, it is quite possible that
heterogeneous nucleation initiated by numerous nanotubes
accelerates the formation of bubbles. The bubbles rapidly
rise (and expand) to the sample surface if the surrounding
layer is a melt with low viscosity. The melt viscosities of the
sample containing SWNT at 0.1% by mass and 0.2% by
mass are low (particularly at high temperatures). Therefore,
the gasification of both PMMA/SWNT(0.1%) and PMMA/
SWNT(0.2%) samples behaved like a liquid, with vigorous
bubbling. Then, the bubbles induce convective movement
through the molten layer as they rise to the surface where
they burst. This disrupts any accumulating layer consisting
of the nanotubes (even a nanotube network is formed in the
initial sample). As bubbles rise and burst at the surface, the
nanotubes are pushed away from bubble areas and
accumulate to form small discrete islands. This proposed
mechanism of the formation of many small islands is
illustrated in Fig. 12(a). Because the regions between the
islands were exposed to an undiminished external radiant
Fig. 9. Selected video images of PMMA/SWNT during gasification tests at external radiant flux of 50 kW/m
2
in a nitrogen atmosphere, (a) PMMA, (b)
PMMA/SWNT(0.2%), PMMA/SWNT(0.5%), and (d) PMMA/SWNT(1%).
Fig. 10. Pictures of the residues of PMMA/SWNT after the gasification tests in a nitrogen atmosphere at external radiant flux of 50 kW/m
2
. (a) PMMA, (b)
PMMA/SWNT(0.2%), (c) PMMA/SWNT(0.5%), (d) PMMA/SWNT(1%).
T. Kashiwagi et al. / Polymer 46 (2005) 471–481 477
flux, vigorous bubbling and bursting occurred preferentially
between the islands. This is the reason why the mass loss
rate and heat release rate of the samples with 0.1% and 0.2%
by mass of SWNT are much higher than those of samples
containing higher content of nanotubes. A similar process
also occurred with the PMMA/SWNT(0.5%, poor dis-
persion) sample due to low melt viscosity (in particular at
high temperatures) even with the same amount of the
nanotubes in the sample.
The melt viscosities of the samples with 0.5% and 1%
mass fractions of the nanotubes are high enough to behave
like a solid material during the gasification test. The bubbles
remain small in the high viscosity layer and their transport
to the surface tends not to disrupt the structured layer, such
that the layer is preserved during gasification and burning.
(Other possibilities are; (1) bubble size is related to viscosity
and low viscosity allows for bubble coalescence while high
viscosity does not, (2) the nanocomposites tend to generate
smaller bubbles by nucleation from well-dispersed, high
content of the nanotubes and the nanotube network inhibits
the formation of large bubbles.) This proposed mechanism
for the formation of the network structured layer is
described for the sample containing higher contents of the
nanotubes in Fig. 12(b). To form this structured layer, both a
sufficient amount of nanotubes and their good, micrometer
scale dispersion in a polymer are required in order to form
such layer.
The importance of the formation of the protective layer
on the reduction in heat release rate of the nanocomposite
has been clearly demonstrated. However, the characteriz-
ation of the protective layer is needed to understand how it
reduces heat release rate. The SEM image of the residue of
PMMA/SWNT(1%) shows a network structure consisting
of bundled, inter-wined carbon nanotubes, as shown in Fig.
13. The residue was strong enough to be readily handled
without breaking it. However, the sample studied in this
work shrank during the gasification test and the thickness of
the residue was about 3 mm for PMMA/SWNT(0.5%) and
about 3.5 mm for PMMA/SWNT(1%). This thickness is
much less than the, approximately, 7 mm thickness of the
residue of PP/MWNT(1%). Possible reasons for this
difference could be that the aspect ratio of the MWNT
(large diameter, long tubes) was larger than that of the
SWNT (very small, relatively short tubes) or the size
Fig. 11. Effects of SWNT concentration on mass lossrate of PMMA/SWNT
at external radiant flux of 50 kW/m
2
in a nitrogen atmosphere.
Fig. 12. Schematic illustration of the formation of islands (a) and of a network structured layer (b). Light color represents a melt layer. Circles are bubbles.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481478
distribution of the former was very large compared to
relatively narrow size distribution of the latter. The network
structured layer consisted of the large size distribution and
high aspect ratios of the tubes might yield a physically
stronger layer than that consisted of the nanotubes having
the narrow size distribution and small aspect ratios.
The amount of each residue collected after the gasifica-
tion test was measured and is listed in Table 1. The results
indicate that the addition of the nanotubes slightly increases
the amount of the residue from PMMA. The sample with
poor nanotube dispersion generates less residue compared to
that with good nanotube dispersion. This could be due to
more confinement of PMMA and MMA by the network
layer than by the discrete islands. Thermal analysis of the
residue was conducted to determine the thermal stability of
the top layer of the residue and the results were compared
with that of the bottom layer of the residue (the top layer
appeared to be a continuous, pasty looking material
compared to a more porous, granular structure for the
bottom layer). Furthermore, the results might indicate any
hydrocarbons in the residue assuming that the nanotubes
themselves are thermally stable (no mass loss) in the
temperature range used in the TGA in nitrogen. The TGA
results of the original PMMA/SWNT(0.5%), original
SWNT, and the top layer and bottom layer of the residue
of PMMA/SWNT(0.5%) collected after the nitrogen
gasification test are shown in Fig. 14. Each test was
conducted using ultra high purity nitrogen at a heating rate
of 5 8C/min from 90 to 900 8C after holding 90 min at 90 8C
for removing any moisture trapped in the sample (Note that
nitrogen purity was critical, because oxygen impurities were
found to approximately double the mass loss in the SWNT
and residues). Surprisingly, about 21% mass of the original
SWNT was lost by 900 8C. This could be caused by thermal
degradation of amorphous carbon and large carbon full-
erenes as impurities as observed in Fig. 1(c). The top layer
lost about 13% mass of the original residue and the bottom
layer lost about 20% mass of the original residue. The
residue tends to be more thermally stable than the original
SWNT. This is due to ‘cleaning’ of the nanotubes in the
sample by heating during the gasification test. It is estimated
that the temperature of the top layer during the gasification
experiment reached at most 650 8C (assuming that the
radiant source temperature of the top layer emits a radiant
flux equal to the flux difference from the incident flux minus
the transmitted flux of about 12 kW/m
2
through the residue
layer as described later). By 650 8C, this figure shows about
9% mass loss of the original SWNT. It appears that SWNT
in the top layer of the residue lost about 9% of the original
SWNT mass during the gasification test (The heating
duration in the gasification test was at most 15 min. The
TGA results shown in Fig. 14 indicates little mass loss of
SWNT from 650 to 725 8C (temperature increase during
15 min at 5 8C/min heating rate).). Then, the nanotubes in
the top layer of the residue contain about 12% of thermally
degradable impurities by 900 8C. The mass loss from the top
residue is about 13% by 900 8C. Therefore, the top residue
consists of about 99% of its mass of the nanotubes with
about 1% by mass of newly formed residue. This indicates
that the top layer of the residue consists almost entirely of
the nanotubes. The estimated temperature of the bottom
layer is a little over 400 8C (assuming that the radiant source
Fig. 13. SEM image of the residue of PMMA/SWNT(1%) collected after
nitrogen gasification indicating a randomly interlaced structure.
Table 1
Mass fractions of the residues normalized by the original sample mass collected after the gasification test
Mass fraction of
SWNT(%)
0.0 0.2 0.5
a
0.5 1.0
Residue mass/original
mass (%)
0.0 0.068G0.05 0.76G0.05 0.99G0.05 1.81G0.05
a
Poor dispersion.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481 479
temperature emits 12 kW/m
2
of the flux) which is high
enough to degrade PMMA. In this temperature range, the
nanotubes loose about 5% by mass. The bottom layer of the
residue looses about 20% of by mass by 900 8C and it
consists of about 4% by mass of newly formed residue (20%
-(21%–5%)) and about 96% by mass of the nanotubes. TEM
images of the bottom layer of the residue show bundles of
the nanotubes coated with amorphous carbon and residual
iron particles (Fig. 15).
The thermal characteristics of the network structured
layer are important in determining the flame retardant
effectiveness of the PMMA/SWNT nanocomposites. A test
was conducted to measure the transmission of a broadband
external radiant flux and also the thermal insulation
performance of this layer. The test was conducted in the
gasification device in a nitrogen atmosphere to avoid any
exothermic glowing combustion of the layer in air. At first,
the external radiant heat source was turned on with the
closed water cooled shutter over the residue until the heat
source reached a steady temperature emitting a steady flux
of about 51 kW/m
2
. Then, the shutter was opened by a
pneumatic piston and the residual layer was exposed to the
external radiant flux. The layer was directly mounted on
(and in contact with) a water cooled Gardon type flux gauge
(diameter of 15 mm) which recorded heat flux through the
layer. The recorded transmitted flux of about 12 kW/m
2
through the residual layer of the PMMA/SWNT(0.5%)
sample is shown in Fig. 16. The flux gage sees a
combination of transmitted external radiant flux plus a
part of re-emission from the hot layers. The results show
that the gauge detected the steady-state value of transmitted
flux almost instantly within 2–3 s from the start of opening
of the shutter, (full opening took about 1 s, the response of
the gauge was about 1 s, and data were taken every 1 s).
Another important aspect of the results is that the
transmitted flux remained constant during a 6 min period.
This means that thermal conduction through the network
Fig. 14. TGA results of PMMA/SWNT(0.5%), SWNT, and the residue of
PMMA/SWNT(0.5%) collected after the gasification test, top is top layer
and bottom is bottom layer of the residue, at 5 C/min in nitrogen.
Fig. 15. TEM image of the bottom layer of the residue of
PMMA/SWNT(0.5%) collected after the gasification test.
Fig. 16. Transmission characteristics of the residue of
PMMA/SWNT(0.5%) collected from the gasification test at 51 kW/m
2
in
nitrogen. A shutter was closed shortly after 300 s with the residue.
T. Kashiwagi et al. / Polymer 46 (2005) 471–481480
layer appears to be negligible compared to radiative
transfer. The external radiant flux of 51 kW/m
2
was
absorbed at the top layer of the residue and heated the
layer nearly instantaneously due to its low density (about
0.03 g/cm
3
for the residue of PMMA/SWNT(0.5%)). The
hot top layer re-emitted radiation to the gas phase as a heat
loss and also to the inside of the residue. Since the heat-up
time of the layer was almost instantaneous due to its low
density, achievement of steady-state radiative transfer
through the residue was very quick. The network structured
layer acts as a thermal shield to reduce the exposure of the
polymer resin in the nanocomposite to an external radiant
flux or to heat feedback from a flame.
5. Conclusion
PMMA/SWNT nanotube nanocomposites were prepared
by the coagulation method and the effects of nanotube
dispersion and concentration (up to 1% by mass) on the
flammability properties of these nanocomposite were
determined. A nanotube-containing network structured
layer without any major cracks or openings was formed
during the burning tests and covered the entire sample
surface of the nanocomposite having good nanotube
dispersion. However, the nanocomposite having poor
nanotube dispersion or a low content of the nanotubes
(0.2% by mass or less) formed numerous black discrete
islands and vigorous bubbling was observed between the
islands. The peak heat release rate of the nanocomposite
which formed the network structured layer is about a half
less than those which formed the islands. It is proposed that
the formation of the islands is due to localized accumulation
of the nanotubes as a result of bubble bursting at he surface
and bubble-induced flow from inside the sample to the
surface through the molten sample layer. Bubbles are
formed from nucleation of the degradation product (methyl
methacrylate) of PMMA. The network structured layer
consists of mainly the nanotubes with a small amount of
hydrocarbons and amorphous carbon. The layer acts as a
heat shield to slow the thermal degradation of PMMA.
Acknowledgements
We thank Carbon Nanotechnologies Incorporated and
Foster Miller Company for providing SWNTs, Mr Richard
Harris for preparing the sample disks and Ms Caitlin Baum
for making Fig. 12. F. Du and K. I. Winey acknowledge
funding from the Office of Naval Research.
References
[1] Giannelis E. Adv Mater 1996;8(1):29–35.
[2] Gilman JW, Kashiwagi T. SAMPE J 1997;33(44):40–6.
[3] Zhu J, Morgan AB, Lamelas J, Wilkie CA. Chem Mater 2001;13:
3774–80.
[4] Zanetti M, Camino G, Mulhaupt R. Polym Degrad Stab 2001;74:
413–7.
[5] Gilman JW, Jackson CL, Morgan AB, Harris Jr RH, Manias E,
Giannelis EP, Wuthernow M, Hilton D, Phillips SH. Chem Mater
2000;12:1866–73.
[6] Zhu J, Uhl FM, Morgan AB, Wilkie CA. Chem Mater 2001;13:
4649–54.
[7] Alexandre M, Beyer G, Henrist C, Cloots R, Rulmont A, Jerome R,
Dubois P. Macromol Rapid Commun 2001;22:943–6.
[8] Zhu J, Start P, Mauritz A, Wilkie CA. Polym Degrad Stab 2002;77:
253–8.
[9] Morgan AB, Harris RH, Kashiwagi Jr T, Chyall LJ, Gilman JW. Fire
Mater 2002;26:247–53.
[10] Kashiwagi T, Harris Jr RH, Zhang X, Briber RM, Cipriano BH,
Raghavan SR, Awad WH, Shields JR. Polymer 2004;45:881–91.
[11] Kashiwagi T, Grulke E, Hilding J, Harris R, Awad W, Douglas J.
Macromol Rapid Commun 2002;23:761–5.
[12] Kashiwagi T, Grulke E, Hilding J, Groth K, Harris R, Butler K,
Shields J, Kharchenko S, Douglas J. Polymer 2004;45:4227–39.
[13] Beyer G. Fire Mater 2002;26:291–3.
[14] Haggermueller R, Gommans HH, Rinzler AG, Fischer JE, Winey KI.
Chem Phys Lett 2000;330:219.
[15] Ajayan PM, Schadler LS, Giannaris C, Rubio A. Adv Mater 2000;12:
750–3.
[16] Mamedov AA, Kotov NA, Prato M, Guldi DM, Wicksted JP,
Hirsch A. Nature Mater 2002;1:190–4.
[17] Du F, Fischer JE, Winey KI. J Polym Sci: Part B, Polym Phys 2003;
41:3333–8.
[18] Nikolaev P, Bronikowski MJ, Bradley RK, Rohmund FR, Colbert DT,
Smith KA, Smalley RE. Chem Phys Lett 1999;313:91.
[19] Du F, Scogna RC, Zhou W, Brand S, Fischer JE, Winey KI.
Macromolecules 2004 [in press].
[20] Austin PJ, Buch RR, Kashiwagi T. Fire Mater 1998;22:221–37.
[21] Brauman SK. J Polym Sci Chem Ed 1975;26:1159–71.
[22] Kashiwagi T, Ohlemiller TJ. Proc Combust Inst 1982;19:815–23.
[23] Yang S, Castilleja JR, Barrera EV, Lozano K. Polym Degrad Stabil
2004;83:383–8.
[24] Madorsky SL. Thermal degradation of organic polymers. New York:
Interscience; 1964 [Chapter 8].
T. Kashiwagi et al. / Polymer 46 (2005) 471–481 481