real space and wave-number space studies of the phase structure and morphology of ipp/peoc blends...
TRANSCRIPT
Full Paper
516
Real Space and Wave-Number Space Studiesof the Phase Structure and Morphology ofiPP/PEOc Blends Using Scanning ElectronMicroscopy
Lin Zhu, Na Song, Xinhua Xu*
The influence of blend composition on the phase structure and morphology of poly(propy-lene)/poly(ethylene-co-octene) blends was studied using SEM. A diameter dg was defined andcalculated in real space to discuss the phase structure andmorphology of iPP/PEOc blends. Thefigure-estimation method was introduced to determine the distribution width of dg. It wasshown that the distribution of dg obeys a log-normal distribution and the distribution width s
of dg was calculated. In wave-number (h) space,the correlation distance, ac, was defined by apply-ing light scattering theory to power spectrumimages obtained by 2D Fourier transformation.Moreover, a fractal dimension, Dc, was introducedto describe the uniformity of the spatial distri-bution.
Introduction
Polymer blending has received much attention both in
theoretical research and practical applications, since it has
been proven to be an excellent way of developing new
materials exhibiting combinations of properties that
cannot be obtained from any one polymer. There has been
an extraordinary growth in the field of polymer blends
during the last few decades compared to homopolymers
and copolymers.[1]
It is well established that the ultimate properties of
polymer blends are determined by the type of phase
morphology and the phase dimensions. Therefore, to
control the blend properties, the final morphology, which
depends largely on the morphological evolution during a
L. Zhu, N. Song, X. XuSchool of Materials Science and Engineering, Tianjin University,Tianjin 300072, ChinaFax: þ86 22 2740 6127; E-mail: [email protected]
Macromol. Mater. Eng. 2009, 294, 516–524
� 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
melt-blending process carried out with a twin-screw
extruder or a batch mixer, should be understood. The
microscopic phase structure and morphology of polymer
blends can be directly observed by some microscopy
methods, e.g., scanning electron microscopy (SEM), trans-
mission electron micrograph (TEM), and the microphoto-
graphs represent the dimensional distribution and the
orientation of dispersed phase. Furthermore, thanks to
improvements in the coupling of microelectronics and
computer science, automatic image treatment and analysis
systems can be designed. Digital image analysis (DIA) can
provide a quantitative approach to the microphotographs
based on mathematical concepts.[2–4] Using DIA, not only
various information, e.g., the dimension and the shape of
the dispersed phase, but also the power spectrum image
obtained by two-dimensional Fourier transformation
(2DFT) of an original image can be obtained. Since the power
spectrum image in the wave-number (h) space directly
corresponds to an image obtained by small-angle light
scattering (SALS).[4] Therefore, theeffectivewaytostudythe
DOI: 10.1002/mame.200900068
Real Space and Wave-Number Space Studies . . .
phase structure and morphology can be realized by the
analysis of the microphotographs using DIA. On the other
hand, after the scaling theorywas introduced into polymer
field, the understanding of the polymer systems has been
much extended.[5–7] In fact, the phase structure and
morphology may have some relations with the fractal
behavior.[8] However, the potential fractal behavior in the
phase structure and morphology of polymer blends has
been paid rare attention up to now.
Becauseof its easyprocessing, lowcost, andgoodthermal
and mechanical properties, isotactic poly(propylene) (iPP)
has been extensively used as a commodity polymer.
However, its application is somewhat limited due to its
highshrinkage rate, andrelativelypoor impact resistanceat
room or low temperatures. Therefore, how to improve the
impact toughness of PP resin has been extensively paid
attention. It is generally known that the toughness of
poly(propylene) (PP) can be improved by the addition of
elastomer such as the ethylene/propylene copolymer
(EPR)[9,10] and ethylene/propylene/diene terpolymer
(EPDM).[11,12] Among the elastomers used, poly(ethylene-
co-octene) (PEOc) being narrowmolecular weight distribu-
tion produced via metallocene technology has been
commercialized and have distinctive properties due to
long chain branching. So it has started displacing the
traditional impact modifiers, and many literatures have
reported about them in recent years.[13–17]
In this paper, the phase structure andmorphology of iPP/
PEOc blends were investigated both in the real space and
the wave-number (h) space based on the DIA of SEM
micrographs. The average sizes of dispersed phase particles
and their distribution width were calculated. Moreover,
the fractal dimension, Dc, was obtained to describe the
phase structure and morphology of iPP/PEOc blends using
the correlation function.
Experimental Part
Materials
The basicmaterials used in this studywere a commercial grade iPP
(1 300) with density of 0.90 g � cm�3 and melt flow index of
1.10 g � (10min)�1 (measured under 2.16 kg at 230 8C), supplied by
Beijing Yanshan Petrifaction Co. and a commercial grade PEOc
(Engage 8150), a metallocene catalyzed copolymer with 25wt.-%
of comonomer, provided by Dow Elastomers with density of
0.87 g � cm�3 and melt flow index of 0.5 g � (10min)�1 (measured
under 2.16 kg at 190 8C).
Blend Preparation
The blendswere prepared in an internalmixer (XXS-30mixerwith
rotordiameterof35mmandtotal volumeof50 cm3,China). Prior to
processing, allmaterialswere dried for 12hunder vacuumat 50 8C.
Macromol. Mater. Eng. 2009, 294, 516–524
� 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
The virgin constituent polymers and blends with different
compositions [iPP/PEOc¼60:40, 70:30, 75:25, 80:20, 85:15, 90:10,
95:5 by volume concentration]were prepared. In order to study the
effects of the blend composition on structure and morphology of
the PEOc toughened PP, the blend experiments were performed in
the internal mixer for 10min at constant rotor speed of 40 rpm at
200 8C.The following equation was used to calculate the PEOc volume
concentration:
VPEOc %ð Þ ¼ WPEOc=rPEOcWPP=rPP þWPEOc=rPEOc
� 100% (1)
where V, W, and r are volume, weight, and density of
corresponding components in the subscripts, respectively.
Morphological Characterization
The phase structure and morphology of the blends were
investigated using the scanning electron microscope (Philips
XL-30 ESEM). Because it was difficult to extract the iPP phase
without affecting the PEOc phase using the solvent, all samples
werecryogenically fractured in liquidnitrogenforat least10min to
make sure that the fracture was sufficiently brittle and they were
etched in heptane at 60 8C for 10min to extract the elastomeric
PEOc phase. The samples were dried for a period of 72h and were
coated with gold prior to SEM examination. Considering that
the quality and resolution of SEM images are strongly affected by
the thickness of the plated gold, the sputter time was strictly
controlled to be identical for each sample. The microscope
operating at 25 kV was used to view the specimens, and several
SEM microphotographs were taken for each sample.
Image Analysis
The SEMmicrographswere binarizedusing our self-made software
EMPP, and the particles of the dispersed phase were chosen as our
studied object. The number of pixels embodied by every particle
was calculated using the software. Then the software created a
circle that enclosed the same number of pixels and defined the
diameter of the circle as that of each particle; thus, the equivalent
diameter, dg, of each particle can be obtained. Consequently, the
average size of the dispersed phase over all particles could be
calculated by averaging the dg using Equation (2),
dg ¼P1
i ni dg� �
iP1i ni
(2)
where ni is the number of particles. To obtain more reliable data,
about 100 particles were considered to calculate this structure
parameter for each sample.
In order to compare dg with other diameters, the number-
average diameters and volume average diameters were also
obtained using the following equations, respectively:
www.mme-journal.de 517
L. Zhu, N. Song, X. Xu
518
The number-average diameter is
Macrom
� 2009
Dn ¼P
NiDiPNi
(3)
The volume average diameter is
Dv ¼P
NiD4iP
NiD3i
(4)
where Di is the diameter of each droplet and Ni is the number of
droplet with a diameter Di.
The polydispersity (d) was characterized by means of the ratio:
d ¼ Dv
�Dn (5)
Results and Discussion
Studies in Real Space
The SEM micrographs of iPP/PEOc blends with different
compositions prepared at the rotor speed of 40 rpm for
10min at 200 8C are depicted in Figure 1(a–g). The light area
of the SEM represented the iPP phase and the black for the
PEOc phase which was etched off with black holes left. For
the whole sample studied, PEOc was distributed in the
matrix phase in the form of spherical particles, forming the
dispersed phase and iPP was the continuous phase, i.e., a
droplet dispersion type of morphology could be observed
and this type of morphology was observed for many other
blends. It also could be seen that, as the concentration of
the dispersed phase increased, the dispersed phase size
increased and the structure and shape of the particles
became more and more complex (Figure 1).
Although these micrographs showed a vivid phase
structure and morphology, it was hard to quantitatively
determine the variation of phase size just based on
these micrographs. For this reason, the SEM micrographs
were transformed by DIA software designed by our group.
These results, i.e., the binarymicrographs are obtained and
shown inFigure1(h–n).Using the software, thedomain size
was calculated in the form of dg to quantitatively describe
the phase size. In addition, the number-average diameters
and volume average diameters were also calculated and
obtained to compare them with dg. Actually, when the
samples were fractured, the section might go through or
around the particle, thus, a difference existed between our
statisticalvalueandtherealvalueofdiameter.But, sincewe
just wanted to find the changing rule of the diameter with
blend composition rather than their absolute value and the
methodwasbasedon the statistics. Sowherever the section
was, the influence on the result was negligible.
ol. Mater. Eng. 2009, 294, 516–524
WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
The variations of dg, Dn, and Dv as the function of
concentration of PEOc are shown in Figure 2. When the
concentration of the dispersed phase was low, the size of
dispersed particles was small. As the concentration of
dispersedphase increased, thesizeof thedispersedparticles
increased. However, this incrementwas not largewhen the
volume fraction of PEOc was lower than 30%. This
phenomenon indicated that the increase in PEOc concen-
trationfirst raised thenumber of domainsbutnot their size.
Only when the volume fraction of PEOc was higher than
30%, the number of domains approached to saturation and
the size of these domains increased greatly (Figure 2). A
similar result has also been found by others. Nair et al.
studied the relation between phase dimensions and
compositions of nylon/PS blends and found that when
the volume fraction of the dispersed phase was small, the
diameterof thedispersedphaseparticles increased linearly,
but when the concentration of the dispersed phase was
relatively high, the size of particles increased more rapidly
due to the coalescence of the particles and the fibrillation of
nylon.[18] Besides, thevariationofdgwas inaccordwith that
ofDn andDv,whichmeantdgwas an effective parameter to
study the sizes of the dispersed phase particles.
However, itwashardtodeterminethedistributionwidth
of sizes quantitatively based on these images only. In this
paper, the graph-estimation method was introduced to
study the distribution of dg in detail.[19]
Detailed calculations leading to the figure-estimation
theory were given in the Appendix. Based on the figure-
estimation theory, the cumulative distribution of particle
size of different iPP/PEOc blends as the function of
concentration of PEOc is calculated and shown in
Figure 3 (the corresponding curves were omitted for
clarity). Thedistributions ofdgwithdifferent concentration
of PEOc could be transformed into ascending lines,
indicating that the distribution of dg obeyed a log-normal
distribution and therefore, the corresponding parameter s
can be easily calculated to describe the size distribution
(Figure 3).
Figure 4 shows the variations of distributionwidth s and
the polydispersity d as the function of concentration of
PEOc.When the content of the dispersedphasewas low, the
distribution width of dg was narrow. As the content of the
dispersedphase increased, thesizeof thedispersedparticles
and the distribution width of dg increased, which could be
ascribed to increase in the number of domains. In this case,
the density of domains increased and these domains were
distributed rather than clustering together. So the distribu-
tion width of dg and the polydispersity d was small
(Figure 4). However, the number of domains was saturated
when the concentration of PEOc increased further. There-
fore, some neighboring domains would aggregate into a
large and irregular one, which destroyed the uniformity of
the distribution of the particles. So the distributionwidth s
DOI: 10.1002/mame.200900068
Real Space and Wave-Number Space Studies . . .
Figure 1. (a–g) SEM micrographs, (h–n) their corresponding binary micrographs, and (o–u) their corresponding 2DFT transformed patterns ofPP/PEOc blends with different compositions.
which had a similar variation with the polydispersity d
increased further.
Studies in Wave-Number (h) Space
The size and distribution of dispersed phase in polymer
blends cannot only be defined in the real space, but also in
Macromol. Mater. Eng. 2009, 294, 516–524
� 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
the wave-number (h) space. The images in the h space can
be gained by applying 2DFT to the image in the real space.
Furthermore, there existed a corresponding relationship
between the 2DFT images and SALS images as mentioned
by Tanaka.[4]
The observed intensity of an image by a scanning
electron microscope I(r) (r is the position vector), is
www.mme-journal.de 519
L. Zhu, N. Song, X. Xu
Figure 2. The variation of dg as the function of PEOc concen-tration.
Figure 4. The variation of distribution width s as the function ofPEOc concentration.
520
proportional to the difference of the refractive index dn(r) to
good approximation, i.e.,
Figdiff
Macrom
� 2009
I hð Þ / dn rð Þ (6)
The power spectrum of 2DFT of an original image P(h) is
expressed as
P hð Þ ¼ F hð Þj j2¼ FF� (7)
where F(h) is the Fourier transform of I(r) expressed by
F hð Þ ¼ZV
I rð Þ exp �jhrð Þdr with j ¼ffiffiffiffiffiffiffi�1
p(8)
ure 3. The cumulative distribution of dg of PP/PEOc blends witherent composition.
ol. Mater. Eng. 2009, 294, 516–524
WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
From the convolution theorem,
FF� ¼ZV
I rð ÞI 0ð Þh i exp �jhrð Þdr (9)
Therefore, P(h) is generally given by
P hð Þ ¼ZV
I rð ÞI 0ð Þh i exp �jhrð Þdr
/ZV
n rð Þn 0ð Þh i exp �jhrð Þdr (10)
On the other hand, the scattering intensity S(h) obtained
by a light scattering experiment is given by
S hð Þ /ZV
n rð Þn 0ð Þh i exp �jhrð Þdr (11)
By comparing Equation (10) with Equation (11), it was
found that thepower spectrumof2DFTof anoriginal image
is equivalent to the scattering intensity S(h). Therefore,
the phase structure and morphology of polymer blends
couldbedescribed in termsof light scattering theory,which
was introduced in the Appendix section.
The corresponding power spectrum micrographs are
shown in Figure 1(o–u). It could be seen that the facula on
thepower spectrummicrographs became smaller, brighter,
andbasically roundas the concentration of dispersedphase
increased.When the concentration of dispersed phaseswas
higher, especially at 60/40 vol.-%, the facula was elliptical,
because of the distortion of particles. Besides, the direction
of the longer axis of the ellipsewas perpendicular to that of
DOI: 10.1002/mame.200900068
Real Space and Wave-Number Space Studies . . .
Figure 5. The variation of ac1 and ac2 as the function of PEOcconcentration.
tropismof particles (Figure 1). The resultwas in accordwith
that of Izumitani and Hashimoto.[20]
Bymeansof theSALS theory, the correlationdistancesac1and ac2 could be calculated as functions of the PEOc
concentrations (shown in Figure 5). It could be shown that
the value of ac2 increased, while the value of ac1 decreased
as the concentration of PEOc increased. Furthermore, the
value of ac2 had a similar variation with dg, Dn, and Dv
(Figure 2), which indicated that it was valid to apply
corresponding light scattering theory to 2DFTmicrographs
to study the phase structure and morphology.
Besides the sizes of dispersed phase particles and their
distribution, the uniformity of the spatial distribution of
particles was another aspect which should be paid more
attention. In this paper, the fractal dimension, Dc, was
introduced todescribe thisuniformityusing the correlation
function.[21] The correlation function was a fundamental
statistical quantity, thus the fractal dimension can be
obtained using it.
Let r�(x) be the density at a position x of a set of points,
which are randomly distributed through space. Then the
correlation function c(r) is defined as
Macrom
� 2009
c rð Þ �< r� xð Þr� xþ rð Þ > (12)
Figure 6. The plot of lg [I(h)] versus lg h for PP/PEOc (80:20) blend.
Here c rð Þ �< � >denotes an average. If the distribution is
uniformand isotropic, the correlation function is a function
only of the distance, r, between the two points.
In theoretical models, we often assume that the
exponential function exp r=r0ð Þ or the Gaussian
exp �r2�r20
� �is the correlation function. But these functions
do not possess the fractal property because both of them
have a r0. Any pair of pointswhose distance is less than r0 is
strongly related to each other, but the correlation decays
rapidly for r� r0.
ol. Mater. Eng. 2009, 294, 516–524
WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
On the other hand, when the distribution is fractal, the
correlation function followsapower law. Then there isno r0and the rate of decrease in the correlation is always at the
same rate. For instance, if the correlation is something like
c rð Þ / r�a (13)
then the correlation becomes 2�a times smaller as the
distance of the two points becomes 2 times longer. The
relation between the exponent a and the fractal dimension
Dc is simply
Dc ¼ d� a (14)
where d denotes the dimension of the space.
If the correlation function c(r) scales as in Equation (13),
its Fourier-transform I(h), which is called the power
spectrum, also follows a power law. Indeed, when
0<d – Dc< 1,
I hð Þ ¼ 4
Z 1
0dr cos 2phrð Þ c rð Þ / hd�Dc�1 (15)
Using this relation, the fractal dimension fromthepower
spectrum can be estimated. Here, we should note the
meaning of Dc. Dc means the dimension of a spatial
distribution, i.e., if the points cluster together, they have an
intimate relation andDc is larger, the scattering uniformity
is imperfect.
The range of the dimensionless region can be confirmed
by the linear part of the curve shown inFigure 6. The chosen
component of theblendwas randomly selected fromall the
iPP/PEOc blends systems. Figure 7 shows the variation ofDc
as the function of PEOc concentration. As the concentration
of the dispersed phase increased, the value of Dc increased,
www.mme-journal.de 521
L. Zhu, N. Song, X. Xu
Figure 7. The variation of Dc as the function of PEOc concen-tration.
522
which indicated that the uniformity of the spatial
distribution became more imperfect (Figure 7). Further-
more, the value of Dc had a similar variation with
distribution width s and the polydispersity d, which also
indicated that it was valid to apply corresponding light
scattering theory to 2DFT micrographs to study the phase
structure and morphology.
Conclusion
The influence of blend composition on the phase structure
andmorphology of iPP/PEOcblendswas studiedusing SEM.
The results showed that the diameter (dg, Dn, and Dv)
increased with the increase in the volume fraction of the
dispersed phase. In addition, the distribution of dg obeys a
log-normal distribution, and the distribution width of dgand the polydispersity d increased as the content of the
dispersed phase increased. Besides, the phase structure and
morphology of iPP/PEOc blends were also studied in the
wave-number (h) space applying light-scattering theory to
2DFT images, and the results were in accord with those in
the real space. Furthermore, a fractal dimension, Dc, was
introduced to describe the uniformity of the spatial
distribution. The result showed that Dc was an effective
parameter to study the distribution of particles of the
dispersed phase.
Appendix
Figure Estimation Theory of the Log-NormalDistribution
In the present work, the figure-estimation theory was
introduced to judge whether the distribution of a variable
was the log-normal distribution.[19]
Macromol. Mater. Eng. 2009, 294, 516–524
� 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Supposing t is a positive random variable, then if ln
t¼N(m, s2), t obeys a log-normal distribution and can be
markedas t¼ ln(m,s2).Here,m is theexpectationands is the
standard variance. The s reflects the distribution range of t,
the larger s, thewider the distribution range.We introduce
the graph-estimation method to judge whether the
distribution of dg obeys the log-normal distribution.
Ordinarily, the logarithm in a log-normal distribution is a
common logarithm, and the function of a log-normal
distribution is given by
F tð Þ ¼ log effiffiffiffiffiffi2p
pst
Z t
0exp � 1
2
log t � m
s
� �2" #
dt
for t 2 0;þ1ð Þ(A1)
This function is a continuously increasing curvebutnot a
straight line. It can be described by the standard normal
distribution function as follows:
F tð Þ ¼Z log t�mð Þ=s
0
1ffiffiffiffiffiffi2p
p e�x2=2dx
¼ Flog t � m
s
� �for t 2 0;þ1ð Þ (A2)
Since the standard normal distribution function is a
monotonously increasing function, its inverse function
exists. The inverse function can be defined as
F�1 F tð Þ½ � ¼ log t � m
s(A3)
If we mark F�1[F(t)] as Y and log t as X, then the inverse
function can be changed to be
Y ¼ 1
sX � m
s(A4)
This equation corresponds to a straight line in the X-Y
reference frame whose slope is 1/s and intercept is �m/s.
Thus, the relation between a log-normal distribution
functionandan increasing straight line in theX-Y reference
frame is found, which is described as the graph-estimation
method. Basedon thismethod, it canbe judgedwhether the
distribution of the value dg of obeys the log-normal
distribution and determine the distribution width of dg.
Light Scattering Theory
To obtain further information on the phase structure and
morphology of polymer blends, the light scattering theory
is introduced in this section. Several parameters, suchas the
diameter and chord length, could be used to describe the
DOI: 10.1002/mame.200900068
Real Space and Wave-Number Space Studies . . .
particle sizes of dispersed phase. In this paper, the light
scattering theory is introduced to compute the correlation
distance, ac1 and ac2, which is used to describe the particles
sizes of dispersed phase.
It is well known that the dispersed phase scatters in the
matrix in the form of particles. Therefore, the size of
particles can be computed using a correlation function. For
this purpose, the modification of Debye and Bueche,[22]
whichdescribes the scattering fromrandomheterogeneous
media, is used; it is given by
Macrom
� 2009
Ih ¼ 4pKVhh2Z 1
0g rð Þ sin hrð Þ
hrr2 dr (B1)
Figure A1. The sketch map of ac on curves of I(h)�1/2 versus h2.
where K is a proportionality constant and h¼ (4p/l) �sin(u/2). h2 is the mean square fluctuation and h is the
fluctuation in scattering power of the system, which for
SALS is equal to the deviation in polarization from itsmean
value at position r. g(r) is the correlation function
corresponding to fluctuation of medium.
General for systems not having a apparently defined
structure, g(r) often decreases monotonically with r and
may be represented by an empirical equation such as
g rð Þ ¼ exp �g=acð Þ (B2)
where the parameter ac is known as correlation distance
and can be used to describe the size of the heterogeneity.
For discrete particles in dilute solution, ac is related to the
particle size. For more concentrated systems, ac is not
simply related to the size of the structural unit but depends
upon both interparticle and intraparticle distances. It may
be considered as an average wavelength of the h(r)
fluctuations whereas h2 is a mean-square fluctuation.
If the second equation is substituted into the first
equation, one can obtain
Figure A2. The physical signification of the correlation distancesac1 and ac2.
I hð Þ ¼ K00h2a3c 1þ h2a2c
� ��2(B3)
Upon rearrangement, it gives
I hð Þ½ ��1=2
¼ K 00h2a3c
� ��1=21þ h2a2
c
� �(B4)
Consequently, aplot of I(h)�1/2 againsth2 should lead toa
straight line having a ratio of slope to intercept of ac. But a
plot of I(h)�1/2 against h2 can lead to two straight lines (see
FigureA1) for SALSwhere u is small, the correspondingac1 is
due to scattering from large particles. As for SALS, when
u!1, the corresponding ac2 is due to scattering fromsmall
particles. Crugnola and Deanin[23] suggested the dimen-
sions of the ac1 and ac2 parameters, measured by the light
scattering, have the physical significance shown in
Figure A2.
ol. Mater. Eng. 2009, 294, 516–524
WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Acknowledgements: The authors gratefully acknowledge thefinancial support of the National Natural Science Foundation ofChina (no. 20490220).
Received: March 3, 2009; Revised: May 18, 2009; Accepted: May19, 2009; DOI: 10.1002/mame.200900068
Keywords: microstructure; morphology; polymer blends; scan-ning electron microscopy (SEM)
[1] M. J. Folkes, P. S. Hope, ‘‘Polymer Blends andAlloys’’, Blackwell,Glasgow 1993.
[2] L. Averous, J. C. Quantin, A. Crespy, Compos. Sci. Technol. 1998,58, 377.
[3] T. Hashimoto, M. Takenaka, T. Lzumitani, J. Chem. Phys. 1992,97, 679.
[4] H. Tanaka, T. Hayashi, T. Nishi, J. Appl. Phys. 1986, 59, 3627.
www.mme-journal.de 523
L. Zhu, N. Song, X. Xu
524
[5] W. Wang, T. Shiwaku, T. Hashimoto, Macromolecules 2003,36, 8088.
[6] A. Nakai, W. Wang, T. Hashimoto, A. Blumstein, Macromol-ecules 1996, 29, 5288.
[7] G. Schmidt, A. I. Nakatani, P. D. Butler, A. Karim, C. C. Han,Macromolecules 2000, 33, 7219.
[8] A. Nakai, T. Shiwaku, W. Wang, T. Hashimoto, Macromol-ecules 1996, 29, 5990.
[9] M. Seki, H. Nakano, S. Yamauchi, Macromolecules 1999, 32,3227.
[10] P. Doshev, R. Lach, G. Lohse, A. Heuvelsland, W. Grellmann,H. J. Radusch, Polymer 2005, 46, 9411.
[11] Q. Fu, Y. Wang, Q. J. Li, G. Zhang,Macromol. Mater. Eng. 2002,287, 391.
[12] W. Jiang, S. C. Tjong, R. K. Y. Li, Polymer 2000, 41, 3479.[13] X. L. Yan, X. H. Xu, L. Zhu, J. Mater. Sci. 2007, 42, 8645.[14] X. H. Xu, T. B. Zhu, L. Zhu, N. Song, J. Sheng, Polym. Int. 2008,
57, 488.
Macromol. Mater. Eng. 2009, 294, 516–524
� 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
[15] X. L. Yan, X. H. Xu, T. B. Zhu, C. H. Zhang, N. Song, L. Zhu,Mater.Sci. Eng., A 2008, 476, 120.
[16] A. L. N. Da Silva, M. C. G. Rocha, F. M. B. Coutinho, Polym. Test.2002, 21, 289.
[17] L. Zhu, X. H. Xu, F. J. Wang, N. Song, J. Sheng,Mater. Sci. Eng., A2008, 494, 449.
[18] S. V. Nair, Z. Oommen, S. Thomas, J. Appl. Polym. Sci. 2002, 86,3537.
[19] K. T. Fang, J. L. Xu, ‘‘Statistics Distribution’’, China SciencePress, Beijing 1987, pp. 136–158.
[20] T. Izumitani, T. Hashimoto, J. Chem. Phys. 1985, 83, 3694.[21] H. Takayasu, ‘‘Fractals in the Physical Sciences’’, Manchester
University Press, Manchester 1990.[22] P. Debye, A. M. Bueche, J. Appl. Phys. 1949, 20, 518.[23] A. M. Crugnola, R. D. Deanin, ‘‘Toughness and Brittleness of
Plastics’’, American Chemical Society, Washington 1979,p. 286.
DOI: 10.1002/mame.200900068