Professional Documents
Culture Documents
6 Nmrimagingofpolymers
6 Nmrimagingofpolymers
J. L. KOENIG
The J. Donnell Institute Professor, Departments of Macromolecular Science and
Chemistry, Case Western Reserve University, University Circle, Cleveland, OH
44102-7202, USA
6.1 INTRODUCTION
Nuclear magnetic resonance imaging (NMRI) is a technique for measuring
spatially resolved features of inhomogeneous samples. The technique has found
particular utility in the medical field, where it is used for diagnosis based on the
fact that the mobility of water in diseased tissue is different from that in normal
tissue. However, in recent years, NMRI has found its way into the field of
materials, particularly polymers.
NMRI has the capability of measuring inhomogeneities in finished articles by
a noninvasive and nondestructive method. Defect or nonuniform areas of the
polymeric materials will be clearly shown in the NMR image. NMRI may be
considered as a type of chemical microscope and, as such, the concept transcends
any other methodology for generating images.
6.1.1 BASIS O F N M R I M A G I N G
Nuclear magnetic resonance (NMR) [1] is based on the fact that many atomic
nuclei oscillate like tiny gyroscopes when in a magnetic field. In NMR, a sample is
placed in a magnetic field which forces the nuclei into alignment. The sample is
then bombarded with a radio wave. As the nuclei absorb the radio wave, they
topple out of alignment with the magnetic field. As they lose the absorbed energy
from the radio wave, they line up again. By measuring the specific radio
frequencies that are emitted by the nuclei and the rate at which the realignment
occurs, spectroscopists can obtain detailed information about the molecular
structure and motion of the sample they are studying.
Conventional NMR spectroscopy is used to determine chemical structure, as is
described in Chapter 4, but cannot locate the position of the stimulated nuclei.
NMRI is a method where the stimulating signal is spatially encoded so that an
image can be reconstructed showing the distribution of nuclei in the sample.
Polymer Spectroscopy. Edited by Allan H. Fawcett
© 19% John Wiley & Sons Ltd
Other than spatially encoding the signal, imaging works on the same principles as
standard NMR.
The NMRI technique relies on the interaction of nuclei in only a small and
controllable region of the sample by placing the sample in a spatially in-
homogeneous magnetic field whose nuclear resonance frequency is matched to
the r.f. signal in only that region. NMR imaging is involved in obtaining the
spatial distribution of all parameters that NMR can detect. The NMR signals
inherently depend on nuclear relaxation time constants T1 and T2, which in turn
reflect the structural environment of the emitting nucleus. NMR is capable of
providing information about molecular structure and motion; consequently,
NMR imaging can provide a variety of structural factors measured in situ.
There are several ways of spatially encoding the NMR signal [ I ] . One is to
apply a linear magnetic field to the original static field (Figure 6.1). The purpose of
the nonuniform field is to label, or encode, different regions of the sample linearly
with different NMR frequencies. As the magnetic field is varied in a known
manner at specific positions within the sample, the frequency of the NMR signal
indicates the spatial position of the resonating nuclei (Figure 6.1). In one
dimension (D), the position of the sample is related to a frequency by the rela-
tionship
Ao2 = O)2- (O0 = y G2Z9
where the magnetic field gradient G = dBJdz. A tailored r.f. pulse with a narrow
frequency range is used to excite only those nuclei at corresponding positions in
the z dimension. The amplitude of the NMR signal received from the z axis line is
a measure of the number of resonant nuclei on that line, and so the NMR
spectrum represents a graph of spin density versus distance (neglecting relaxation
effects). The field gradient is described by a tensor with nine components but, for
large B, we need only be concerned with the three components Ga = BBJBOL,
where a = x, y, z.
Field Strength
Sample Position
The spin densities and the molecular environments of the nuclei are reflected in
the time variation of the amplitude of the measured r.f. signal, and hence are
reflected in the intensity of each voxel in the image. (A voxel is the smallest volume
that the imaging process recognizes and presents.) When the values of T1 and T2
are different in the voxels of a heterogeneous sample, these differences can be
exploited to develop contrast in the NMR images. The pulse sequence that is
usually used to measure the T2 relaxation phenomena in images is called multiple
spin-echo. At a given repetition time TR, the NMR signal is measured at several
different echo times TE. These echoes provide a measure of the T2 relaxation. By
repeating the process at different TR values, the T1 relaxation can also be
measured.
Because differences in relaxation times and spin densities determine image
contrast, data on relaxation times are important in the selection of the optimal r.f.
pulse sequence for imaging a selected sample. Relaxation times can be measured
at any point on an image. The ability accurately to quantify relaxation rates is
important in understanding and optimizing image contrast. Spin density, T1 and
T2 images can be computed from measurements using pulse sequences with
predetermined variations [2]. These fundamental images represent the inherent
data in the system, and can be recombined to reconstitute computed images for
a given pulse sequence.
Contrast in NMRI depends on both material-specific and operator-selected
parameters. The material-specific parameters include the spin density and the
relaxation times T1 and T2. The operator-selected parameters include the pulse
sequence (inversion recovery, spin-echo, etc.) and the pulse delay and repetition
times (timing parameters). For a given imaging system and pulse sequence, it is
the delay and repetition times in conjunction with the intrinsic material par-
ameters which dictate the appearance of the final image. If the correct pulse
sequence is employed and the relaxation times of the two materials are known, it
is possible to calculate the delay and/or repetition times that will produce the
maximum difference in signal intensity between those materials.
Figure 6.2 The timing diagram for a spin-echo imaging pulse sequence using a selective
90° pulse
The spin-echo (SE) technique is the most common pulse sequence applied in
MRI today [I]. Images are constructed by acquiring a multitude of projections
(typically 256 per image) each with an identical setting of a readout gradient
during which the sequence is samples. Each projection is differentiated from the
others by a phase difference, which is produced by advancing the phase encoding
gradient.
As shown in Figure 6.2, the spin-echo method consists of a series of r.f. pulses
which are repeated many times in order to achieve a sufficient signal-to-noise
ratio. Each projection is produced by a 90° pulse, followed by a 180° pulse for
induction of the spin-echo. The 90° r.f. pulse tips the magnetization into the
xy plane, where it begins dephasing. The 180° r.f. pulse is applied after a time
t, and forces the magnetization to refocus at a time It (also known as the echo
time TE) after the 90° r.f. pulse, at which time the data is collected. The frequency
encoding gradient Gx causes the spins to precess at different frequencies de-
pending on their position in the static magnetic field. The phase encoding
gradient Gy is orthogonal to Gx. Varying the intensity of Gy causes the spins
to dephase at different rates, providing the second dimension of a two-dimen-
sional image. The slice selection gradient G2, and the Gaussian-shaped 90°
r.f. pulse determine the position and thickness of the region of interest. The
data is Fourier transformed in two dimensions to produce the image of the
selected slice. The time delay between the observation pulse and the observation
is called the "echo time" (TE). The time between two consecutive pulse sequences
is labelled as the "repetition time" (TR), and usually ranges from 250 to
2500 ms.
Spin-echo techniques have a unique position in NMR applications. The main
problem with NMR imaging is the long data collection time, due mainly to the
spin-lattice relaxation time T1. Each measurement necessitates a time period of
the order of T1 (which is %0.5 s for aqueous systems) for the system to return to
equilibrium magnetization. By using spin-echo repetition, a large number of
spin-echoes can be repeated within a T1 or T2 decay period.
6.1.3 R E S O L U T I O N I N N M R I M A G I N G
6.1.4 U T I L I T Y O F N M R I
6.3 APPLICATIONSOFNMRITOPOLYMERS
6.3.1 DETECTION OF VOIDS IN COMPOSITES
The void content of pultruded composite rods have been studied using NMRI
[15]. Glass fiber-reinforced nylon rods with fiber contents of 51% by volume
were first mixed with different catalyst contents following the reaction injec-
tion molding (RIM) process, and then pultruded with a pulling speed of 18 inches
per minute. Approximately 4% of the catalyst mixture, containing sodium
hydride and phenyl isocyanate, was used. The diameter of the die in the pultruder
was 0.90 cm. The rods were then soaked in water at 80 0 C for 25 weeks before
imaging. The uptake in water was 3.7%, as measured by the increase in weight
of the composite rods. The images were recorded on a Bruker MSL 300 spectro-
meter using a spine-echo pulse sequence. The slice thickness was 1.0 mm and
the slices were taken transverse to the fiber axis. With the rods standing in
1.5 cm diameter vials containing water, the two images shown in Figure 6.4
are taken 0.5 cm apart through the pultruded rod. The light areas in the image
represent void areas filled with water. The marker in the upper left hand portion
of the image is 1 mm in diameter. Comparison of the sizes of the voids in the
pultruded rod intidcates that some of the voids approach the magnitude of
the marker. A comparison of the corresponding edge-enchanced images show
that some of the voids in the images occur in the same location, which indi-
cates that the voids are connected or tubular in shape. Thus, a channel-like
void region is suggested over a length of 0.5 cm. From the computer conparison
of the two images taken 0.5 cm apart, it is possible to identify a tubular shaped
void running from one image to the other within the nylon rod. Such a void
could be obtained if an air bubble was trapped in the matrix during the pul-
trusion process. It appears that water diffuses by following the fabers in the
composite.
Image of water in pultruded nylon Contour plot of image showing
rodsreinforcedwith glass fibers presence of tubular void
Figure 6.4 Image of water in pultruded nylon rods reinforced with glass fibers and a contour plot of image showing the presence of
a tubular void
6.3.2 DETECTION OF N O N U N I F O R M DISPERSION
O F FILLER
The improvement in mechanical properties by inorganic fillers is considerably
reduced if there is a nonuniform dispersion of particles in the polymer matrix by
formation of agglomerates. NMRI can produce visual pictures of the spatial
variation of the organic phase distribution. This is accomplished by observing the
proton images of the elastomers as a function of proton density and spin-spin, T2,
relaxation times. These NMR parameters provide a measure of the molecular
mobility, which in turn is related to the spatial variation of the polymer and the
filler in the sample.
Samples of poly(dimethylsiloxane) (PDMS) which were reinforced by in situ
precipitated silica were examined by NMRI [16]. The images were obtained with
a spin-echo technique, with a slice thickness of 500 |im and a digital resolution of
185 urn, and required a time of 25.6 min. A dark rim was observed around the
sample which indicated a reduced mobility of the network chains compared with
the sample core. This difference arises from the high concentration OfSiO2 in this
region.
NMR imaging techniques has been used for the study of sorption and diffusion
and of the desorption of multiple chemical substances in polymeric materials
[18-28]. NMR imaging can directly provide the diffusion coefficients as a char-
acteristic quantity of a liquid component in a sample, making it possible to map
molecular migration on a microscopic scale. NMR imaging also provides
additional information on the microdynamic and structural properties of hetero-
geneous systems, such as subregion diameters, exchange times, and phase
boundary resistances [29, 30].
The principal advantage of NMR imaging is the possibility of making spatially
localized diffusion measurements [30]. One can examine by NMR imaging the
concentration and location of a permeating liquid in a solid sample. A true
diffusion parameter image is obtained, where calculated diffusion coefficients are
encoded into an intensity scale.
One of the obvious advantages of NMR imaging for the study of diffusion is the
visual presentation of the data in the form of images. Such a presentation allows
one to view directly the concentration and location of the penetrant and to ignore
extraneous factors influencing the diffusion. Another advantage of NMR imaging
is that it allows the study of samples of virtually any shape, and allows the
detection of initial imperfections in the sample being studied. It is generally
difficult to interpret liquid sorption measurements in solids because the samples
being examined are not perfect, that is, they initially contain cracks and voids
which increase both the diffusion and the uptake of the liquid. Also, the induced
volumetric changes, though small, can cause microcracking or void formation.
The NMR imaging technique also allows the system to be studied dynamically,
as measurements can be made on the solid sample immersed in the penetrant. The
measurements are rapid. Using the FLASH techniques [22], an image can be
obtained in a few minutes. In this fashion, it is possible to study the dynamics of
the diffusion process. Of course, the sample-penetrant system can be studied
under isothermal conditions.
Finally, a primary advantage of NMR imaging is the fact that all of the NMR
parameters of the sample can be measured and used to interpret the diffusion or
sorption process [30]. Images obtained utilizing different pulse sequences and
interrupt times can be used to calculate the spin-lattice T1 and spin-spin T2
relaxation times and the spin density. These additional parameters relate to the
bonding and environment of the penetrant in the polymer system. These types of
measurements have been useful in understanding the morphological changes
which are observed [22].
NMR imaging techniques have been used for the study of sorption, diffusion
and chemical reactions as well as the desorption of chemical substances in
polymeric materials [23]. NMR imaging can directly provide the diffusion
coefficient as a characteristic quantity of the fluidity of a component in a sample,
making it possible to map molecular migration on a microscopic scale.
The diffusion coefficients have been quantitatively evaluated from a series of
images recorded with different gradient field strengths [22]. Analysis involved the
simulation of the effects of diffusion using the dynamic magnetization equations
to calculate the magnetization for each pixel, which ultimately yielded an image
whose intensities represented the spatially resolved diffusion coefficients. Finally,
a true diffusion constant image was obtained in which the calculated diffusion
coefficients were encoded into an intensity scale [22]. In this scale, high intensities
corresponded to fast diffusion. In this manner, the spatial diffusion of a liquid into
a solid material was characterized in a quantitative fashion.
NMR imaging has been used for methanol diffusing into PMMA [22].
Figure 6.5 shows the image from the PMMA in methanol after 48 h. The diffusion
coefficient can be calculated by measuring the thickness of the sorbed layer as
a function of time. With data processing techniques, it is possible to simplify the
measurements by giving the images a three-level gray scale and then drawing
a profile across the sample as shown in Figure 6.6. The results are shown in Figure
6.7, where the thickness of the layer is plotted versus time. The linearity of this
plot with time confirms that case II diffusion is occurring. The nonzero intercept
at time zero is indicative of an initial Fickian diffusion process followed by case II
diffusion. The constant level of methanol in the penetrant front is also reflective of
case II diffusion.
NMR relaxation parameters are useful probes of molecular motions in poly-
mers. Each correlation time represents the average value of the system, with some
distribution around that average. NMR imaging permits the determination of the
spatial distribution of NMR relaxation times. This distribution provides infor-
mation concerning the local motions of the system. In this case, the polymer is
partially swollen with solvent, and the spatial distributions of relaxation times
reveal the interactions between the solvent and the polymer in the diffusion
process.
Figure 6.5 The proton NMR image of a 30 mm PMMA sphere initially submersed in
methanol (left) and the image taken after 48 h of exposure to methanol at 300C. Reprinted
with permission from John Wiley & Sons Inc. Journal of Polymer Science 1989
Figure 6.6 The intensity profile of the image on the right side of Figure 6.5. The features
are: (a) bulk methanol; (b) equilibrium methanol volume in PMMA; (c) sharp concentra-
tion front; and (d) glassy PMMA. Reprinted with permission from John Wiley & Sons Inc.
Journal of Polymer Science 1989
Acetone swells PMMA to a greater extent than methanol, and the self-
diffusion coefficients of the system are about two orders of magnitude greater
than those of the methanol/PMMA system. This is apparently due to the
increased volume available to the acetone molecules. The self-diffusion coeffi-
cients decrease by 35% from equilibrium in the outer regions to the region near
the glassy core. The decreasing motions of the polymer chains as the core is
Diffusion distance (mm)
Time (hours)
Figure 6.7 The plot of diffusion distance measured with NMR imaging vs. exposure time
for the PMMA shere in methanol at 300C. Reprinted with permission from John Wiley
& Sons Inc. Journal of Polymer Science 1989
last image being only 50% of that of the initial image. The diameter of the rod
decreases by 816 ±68 mm over the 10Oh period as measured from the images.
A Td image calculated from 20 images taken at 5 h intervals over 100 h is shown
in Figure 6.8. In this image, the light portion represents the largest Td and dark
represents the smallest T6, which corresponds to the slowest and the fastest
intensity decreases, respectively. This image shows that the faster intensity
decreases are near the surface of the rod and the slower intensity decreases are
near the glassy PMMA core. The Td at the surface is 58 h and that at the glassy core
is 450 h, as determined from this image. This agrees with the Fickian characteris-
tics, and indicates that imbibed solvent near the surface desorbs quickly, with the
desorption rate decreasing toward the sample core. However, for this system
there is no evidence of a glassy skin developing on the polymer surface [23].
6.3.6 MULTICOMPONENT DIFFUSION AS STUDIED
BY NMRI
It is possible to use NMRI to study multicomponent diffusion in a number of
different ways, utilizing differences in chemical shifts, relaxation times or by
isotopic labeling. We have chosen the last method [26]. We studied the simulta-
neous diffusion of acetone and methanol into polycarbonate by completely
deuterating one of the components and recording the images of the other
component in the PC. Using perdeuterated methanol (MeOH-J4) and acetone
(AC), the FLASH images show the movement of the acetone diffusion front
within the PC rod. These images show that the 75:25 ACrMeOH-J4 mixture
diffuses more rapidly than the 50:50 AC:MeOH-J4.
Perdeuterated acetone :hydroxy-deuterated methanol (AC-J6: MeOD) mix-
tures of 75:25,65:35 and 50:50 volume ratios AC-J6: MeOD were also studied.
From these images we can monitor the solvent front movement of acetone and
methanol into the PC rods. The movement of the solvent front of acetone into PC
versus the square root of time is shown in Figure 6.9(a) for the 75:25,65:35 and
50:50 AC .MeOH-J4 mixtures. The diffusion into PC increases as the acetone
content of the mixture increases. In addition, there is a linear dependence between
the solvent front movement and the square root of time. This indicates that the
diffusion of AC: MeOH-J4 into PC is Fickian [26]. This result was anticipated, as
the diffusion of both pure methanol' and pure acetone has previously been
determined to be Fickian [33]. (Note, however, that the diffusion of pure acetone
was found to be Fickian after an initial period in which the diffusion was reported
to be anomalous [33]). This may seem contrary to what is usually found for the
diffusion of solvent into a glassy polymer which is below its glass transition
temperature Tr Instead, either case II or anomalous diffusion is most often
encountered in this temperature region. This is because, below the Tg, the polymer
chains are in the glassy state, and are therefore not mobile enough to accommo-
date the solvent. In contrast, Fickian diffusion is prevalent above the 7^ because
the polymer chains are in the rubbery state and can therefore accommodate the
solvent more readily. Although the T% of dry PC is 149 0 C, acetone and methanol
diffuse into PC in a Fickian manner because the 7^ is reduced to « —9 0 C by the
solvent ingress [31]. In this way, PC is able to relax and accommodate the solvent
almost immediately, allowing the solvent movement not to be inhibited by the
rate of polymer relaxation.
The solvent front was also monitored for the AC-J6: MeOD mixtures. For
the mixtures of 75:25,65:35 and 50:50 AC-J6: MeOD into PC, similar behavior
was also seen in Figure 6.9(b), where the methyl resonance of methanol was
monitored. There is an increase in the rate of the front movement as the ace-
tone content increases, and also a linear dependence of the front movement
with the square root of time. This linear dependence indicates Fickian dif-
fusion.
Millimeters
Figure 6.9 Solvent front movement of (a) acetone into PC for the 75:25,65:35, and 50:50
v/o AC/MeOH-^ mixtures and (b) methanol into PC for the 75:25,65:35, and 50:50 v/o
AC-</6\MeOD mixtures versus the square root of time. Reprinted with permission from
[26]. Copyright 1992 American Chemical Society
We were interested in determining whether the two solvent fronts move at the
same rates for a particular acetone to methanol ratio. Comparison of the rates of
the solvent front movements of the 65:35 AC:MeOH-^4 and the 65:35 AC-
d6:MeOD mixtures revealed that they were equal within experimental error.
There is an overlapping of the front movement when plotted as a function of
square root of time. This indicates that acetone and methanol are diffusing jointly
through the PC rod in a Fickian manner and do not appear to separate. Similar
behavior is also found when comparing the 50:50 AC:MeOH-^4 and the 50:50
AC-d 6 :MeOD mixtures.
It may be proposed that this joint diffusion through PC is the result of the
formation of a weak "complex", possibly a weak hydrogen bonding interaction
[22] between acetone and methanol. The 65:35 v/o AC:MeOH mixture has
approximately a 1:1 molar ratio, and thereby all the acetone may be complexed
with methanol. This would account for the overlap of the front movement plots
within error. In the 50:50 v/o AC:MeOH mixture, the molar ratio is « 35:65
AC: MeOH. If all the acetone complexes with methanol, that leaves approximate-
ly equal amounts of ACMeOH complex and free methanol which would diffuse
together. As shown by the 65:35 v/o AC:MeOH mixture diffusing at a greater
rate than the 50:50 v/o AC:MeOH mixture, the "complex" may have a greater
interaction with PC than the free methanol would in the 50:50 v/o mixture. The
complex may be able to reduce the T% further than the free methanol, allowing an
increased rate of penetration of slovent.
6.4 ACKNOWLEDGEMENTS
The author acknowledges the support of ALCOM (Advanced Liquid Crystalline
Optical Materials) DMR 8920147 and the students whose work was responsible
for the citations in this paper.
6.5 REFERENCES
[1] P. Mansfield and P.G. Morris, NMR Imaging in Biomedicine, Academic Press, New
York, 1982.
[2] J. Liu, A.O. Nieminen and J.L. Koenig, J. Magn. Reson., 1989,85, 95.
[3] D.G. Cory, Solid State Imaging, Annu. Rep. NMR, 1992, 25.
[4] J.L. Koenig, Spectroscopy of Polymers, ACS, Washington, DC, 1992, Chapter 11.
[5] A.O.K. Nieminen and J.L. Koenig, J. Adhes., 1989,30,47.
[6] W.T. Dixon, Radiology, 1984,153,189.
[7] L.D. Hall, S. Sukumar and S.L. Talagala, J. Magn. Reson., 1984,56, 275.
[8] L.D. Hall and V. Rajanayagam, J. Magn. Reson., 1987,74,139.
[9] A. VoIk, B. Tiffon, J. Mispelter and J.M. Lhoste, J. Magn. Reson., 1987,71, 168.
[10] J. Liu, A.O.K. Nieminen and J.L. Koenig, Appl. Spectrosc, 1989,43, 1260.
[11] D.G. Cory, A.M. Reichwein and W.S. Veeman, J. Magn. Reson., 1988,80, 259.
[12] P. Bornert, W. Creher, A. Gossler, G. Klee, R. Peter and W. Schneider, J. Magn.
Reson., 1989,81, 167.
[13] L. Garrido and J.E. Mark, Polym. Repr., 1989,30, 217.
[14] A.O.K. Nieminen and J.L. Koenig, Appl. Spectrosc, 1989,43, 153.
[15] K.P. Hoh, B. Perry, G. Rotter, H. Ishida and J.L. Koenig, J. Adhes., 1989, 27, 245.
[16] L. Garrido, J.E. Mark, CC. Sun, J.L. Ackerman and C. Chang, Macromolecules,
1991,24,4067.
[17] P. Bliimler and B. Blumich, Macromolecules, 1991,24, 2183.
[18] S. Blackband and P. Mansfield, Solid State Phys., 1986,19, L49.
[19] W.P. Rothwell and P.P. Gentempo, Bruker Rep., 1985,1,46.
[20] J.L. Koenig, Appl Spectrosc., 1989, 43,1117.
[21] C. Chang and R.A. Komoroski, Macromolecules, 1989, 22, 6000.
[22] L.A. Weisenberger and J.L. Koenig, Macromolecules, 1990, 23, 2445.
[23] L.A. Weisenberger and J.L. Koenig, Macromolecules, 1990, 23, 2454.
[24] B. Blumich, P. Bliimler, E. Gunther, and G. Schauss, Bruker Rep., 1990, 2, 22.
[25] S.R. Smith and J.L. Koenig, Macromolecules, 1991,24, 3496.
[26] R.A. Grinsted and J.L. Koenig, Macromolecules, 1992, 25,1229.
[27] R.A. Grinsted and J.L. Koenig, Macromolecules, 1992, 25,1235.
[28] P. Maffei, L. Kiene and D. Canet, Macromolecules, 1992, 25, 7114.
[29] B. Bluemich and W. Kuhn (Eds.), Magnetic Resonance Microscopy, VCH, Basel,
1992.
[30] P. Callaghan, Principles of Nuclear Magnetic Resonance Microscopy, Claredon
Press, Oxford, 1991.
[31] R.A. Ware, S. Tirtowidjojo and C. Cohen, J. Appl Polym. ScL, 1981, 26, 2975.
[32] E.E. LaBarre and D.T. Turner, J. Polym. Sci.f Polym. Phys. Ed., 1982, 20, 557.
[33] A.R. Barens and H.B. Hopfenberg, J. Polym. ScL, Polym. Phys. Ed., 1979,17, 1757.
[34] L. Mandelkern and F.A. Long, J. Polym. ScL, 1951, 6, 456.
[35] N.L. Thomas and A.H. Windle, Polymer, 1978,19, 255.