You are on page 1of 16

Chapter 4

Structure Formation in PET During


the Induction Period of
Crystallization
K. Kaji
1. Introduction
It is well known that crystalline polymers form spherulites when crystal-
lized under quiescent conditions or shish-kebab structures when crystallized
under flow [1-4]. However, the mechanisms of formation of such complex
higher order structures (hierarchic structures) are not sufficiently under-
stood: for instance, how and where the nuclei of spherulites are formed in
the case of homogeneous crystallization. In the case of crystallization from
the glass, called glass crystallization, a great number of small spherulites
are produced, while in the case of crystallization from the melt, called melt
crystallization, a small number of large spherulites are formed. What is
the reason for this? According to van Krevelen [5], in these two cases the
number densities of such spherulites differ by six orders of magnitude.
A more fundamental problem is how the long polymer molecules trans-
form from the more or less entangled amorphous state into the parallel
oriented closest packing state or crystalline state. In 1949, Flory [6] theo-
retically predicted that polymer chains in the molten state or in the frozen
glassy state assume Gaussian or random flight conformation (ideal chains)
which corresponds to the state in dilute solutions. This prediction was
confirmed in a vast amount of studies, mainly by a small-angle neutron scat-
tering (SANS) technique in the 1970s [7-1O]; polymer chains are densely
Handbook of Thermoplastic Polymers: Homopolymers, Copolymers, Blends, and Composites
Edited by Stoyko Fakirov
Copyright 2002 WILEY-VCH Verlag GmbH, Weinheim
ISBN: 3-527-30113-5
226 K. Kaji
entangled with one another and hence they cannot diffuse easily in the
melt, as shown by a reptation model of de Gennes [11,12] and by the tube
model of Doi-Edwards [13,14]. On the other hand, the crystallization rate
is much faster than the diffusion rate under the usual crystallization con-
ditions [15]. Hence, the polymer chain molecules cannot easily escape from
the "trap", consisting of their entangled neighboring chain molecules, be-
fore extending to the rigid (crystalline) conformations as a whole. Instead,
they partially crystallize to form the so-called long period structure; the
entanglement points are excluded from the crystalline lamellae to cause
amorphous regions on the lamellar surfaces. This is also the reason why
polymer molecules cannot completely crystallize under the usual condi-
tions.
The above assumption, however, does not directly explain how the pri-
mary crystal nuclei are formed. To solve this problem, one has to investigate
structure formation processes during the induction period prior to crystal-
lization. The first clue for this was brought by the author's group [16,17].
Our completely new finding was that a peak, differing from the so-called
long period peak, appeared in the small-angle X-ray scattering (SAXS)
pattern prior to crystallization. Further detailed analyses revealed the sur-
prising fact that the time evolution of this new peak follows the mechanism
of a spinodal decomposition (SD) type microphase separation. This finding
was based on studies of the glass crystallization of poly(ethylene terephtha-
late) (PET) just above the glass transition temperature, T
9
. Later, similar
experiments were performed for the case of melt crystallization just below
the melting temperature, T
m
, but such a new peak was not observed, al-
though the SAXS intensities at very small scattering angles increase with
crystallization time. This means that either the characteristic wavelengths
of SD strongly differ in the cases of glass and melt crystallization or the
mechanisms of microphase separation are completely different, as shown by
Olmsted et al. [18]. The SAXS studies on the glass crystallization of other
polymers, such as poly(ether ketone ketone) [19] and poly(ethylene naph-
thalene 2,6-dicarboxylate) (PEN) [20], supported the SD in the induction
period.
The mechanism of this surprising phenomenon has been studied in de-
tail. The first step of crystallization is the change in molecular chain confor-
mation from amorphous to rigid (crystalline), which was studied by SANS
on PET [21,22] and by Fourier transform infrared spectroscopy (FT-IR)
on syndiotactic and isotactic polystyrenes (sPS, iPS) [23,24]. The second
step is the parallel orientation process of the rigid segments, involving the
SD-type microphase separation, that was confirmed by depolarized light
scattering (DPLS) for the glass crystallization of PET [25], iPS [23], and
sPS [24], as theoretically predicted by Doi et al. [14,26].
In this chapter, an attempt is made to visualize the structural forma-
tion mechanisms prior to crystallization, concentrating on the experimental
results with PET.
Structure Formation in PET During the Induction Period of Crystallization 227
2. Discovery of the spinodal decomposition (glass crystallization
case)
Yeh and Geil [27,28] found a spherical or nodular structure with diameter
about 75 A in a glass rapidly quenched from a PET melt, using an electron
microscope, and proposed a chain-folded fringed micelle particle model,
ie., that the amorphous structure consists of ordered domains (more dense
regions of loosely folded molecular chains) and disordered domains (less
dense regions of random coils) in between. This model of the amorphous
structure contradicted Flory's random coil model for the same structure
[6], and many experimental data, mainly SANS data, finally confirmed
that Flory's model is relevant. Unfortunately, the nodular structure itself
was also inevitably considered to be due to a misinterpretation of either
surface artifacts or "ghost" effects due to out-of-focus electron micrographs
[7] (see also Chapter 3).
However, it may be assumed that the nodule-like structure actually
exists because it is deformed by drawing and the nodule size depends on
the annealing temperature [29]. Furthermore, such a nodular structure can
be considered to be caused by a precursory phenomenon of crystallization
because it emerges clearly from crystalline polymers only, such as PET
and isotactic polypropylene (iPP) [3O]. In amorphous polymers, such as
atactic polystyrene, some structure is seen, but it is very ambiguous [31].
This concept was our motivation in the studies of the structure formation
of PET prior to crystallization [32]. Below, this precursory phenomenon is
considered in detail on the basis of the glass crystallization of PET; SD
actually occurs during the induction period of crystallization.
2.1. Determination of the induction period
As a first step, the induction period of crystallization should be deter-
mined. Figure 1 shows the time dependence of the crystallization isotherm
and density measured by differential scanning calorimetry (DSC) and with
a density gradient column, respectively, when a PET glass was crystallized
at 115
0
C, ie., 4O
0
C above the glass transition temperature T
9
= 75
0
C; af-
ter the first minute of annealing, a steep exotherm appears. Similarly, the
amorphous density of 1.333g/cm
3
remains unchanged for lmi n, but then
it increases abruptly to 1.363g/cm
3
, which corresponds to a degree of crys-
tallinity of 24.6%, and levels off after annealing longer than 10 min. These
data suggest that the induction period is ca. l min. This is clearly confirmed
by wide-angle X-ray scattering (WAXS) measurements. Figure 2 shows the
WAXS profiles of the samples as a function of annealing time and their
distance distribution functions, P(A), obtained by inverse Fourier trans-
form. During the first minute, only short-range order exists, but thereafter
long-range order due to crystals suddenly appears. Again, one can con-
clude that the induction period is about l min. However, it is too short to
228 K. KaJi
0.8
.2
^0.6
0.2
(a)
0.1
1.39

Q
1.37
1.35
1.33
(b)
1.0 10 100 1000
0.1 1.0 10 100 1000
Annealing time (min)
Figure 1. Annealing time dependence of crystallization isotherm (a) and density
(b) for PET, isothermally crystallized at 115
0
C from the glass [16]
(a)
10 20 30 40
Q (nm-
1
)
50 O 1.0 2.0 3.0 4.0 5.0
R (nm)
Figure 2. (a) WAXS profiles and (b) their distance distribution functions, P(R),
as dependent on annealing time for PET, isothermally crystallized at 115
0
C from
the glass [16]; M.Q.: melt-quenched glassy sample
Structure Formation in PET During the Induction Period of Crystallization 229
_ o
0.6
I
4
S
0.2
O
0.1 1.0 10 100
Annealing time (min)
1000
Figure 3. Crystallization isotherm for PET, isothermally crystallized at 8O
0
C
from the glass [17]
perform quantitative experiments to study the structural formation pro-
cesses. Therefore, for detailed experiments it is necessary to adopt a lower
crystallization temperature, T
0
. In the case T
0
= 8O
0
C, or 5
0
C above T
p
,
the induction period was determined to be about 120 min from the DSC
isotherm, as shown in Figure 3, ie., it is sufficient to measure the time
dependence of the SAXS profiles quantitatively. The crystallization tem-
perature T
0
= 8O
0
C was employed in the following experiments.
2.2. Observation of a new peak in SAXS
Figure 4 shows the time-resolved SAXS intensity 1(Q) in log-log scales as
a function of the magnitude of scattering vector, Q = (480)/, and
being the half of the scattering angle and the X-ray wavelength, respec-
tively [17]. It is seen that the intensity for the melt-quenched glassy sample
increases monotonously with decreasing Q. This excess intensity may be
considered to correspond to the tail of the light scattering intensity which
is generally observed in glass-forming materials, and the correlation length
of such density fluctuation is in the range of more than several hundred
nm and is independent of other correlations appearing in the SAXS range
[33,34]. However, the origin of this excess intensity is not well understood.
On the other hand, the SAXS curves of the samples annealed for 243 and
313 min indicate an intense broad peak of the well known long period at
around Q = O.onm"
1
, which is due to the alternation of crystalline and
amorphous layers while at the very initial stage of annealing; a new peak
differing from the long period peak appears at around Q = O^nm"
1
and
increases in intensity with time. Though this peak looks weak and broad on
the logarithmic scale, it does exist and can be seen more clearly in the linear
expression. In Figure 5, the intensity differences are plotted vs. Q for dif-
ferent annealing times, after the intensity of the melt-quenched sample was
230 K. Kaji
I
+3
3
a
Long period
Annealing time (min)
Induction Cryst.
period stage
115 313
51 + 243
23 *187
3 O 147
O Melt-quenched
New scattering peak
0.1 10
Q (nm
Figure 4. Annealing time dependence of the SAXS curve for PET, isothermally
crystallized at 8O
0
C from the glass [17]
subtracted from the scattering intensities of the annealed samples because
it may be considered independent of the concerned intensities, as described
above. Figure 5a, corresponding to the induction period of crystallization
(< 120min), shows the development of the new peak, which increases in
intensity with time while its position shifts from Q = 0.4Um"
1
towards
lower values. As seen in Figure 5b, this peak continues to grow even after
entering the crystallization stage (> 120 min) and disappears outside the
resolution of the SAXS camera. On the other hand, the so-called long pe-
riod peak starts to appear near the initiation of crystallization and increases
in intensity, but its position hardly shifts from Q = O.Gnm"
1
.
2.3. Spinodal decomposition
As a next step, the annealing time dependence of the new peak was exam-
ined quantitatively. Though the result is not shown here, the time evolution
of the logarithmic intensity of the new peak at several Q values gave two re-
gions, suggesting the existence of a second stage after ca. 20 min; in the first
stage the intensity difference increased exponentially while in the second
stage it increased more slowly. Figure 6 shows the time dependence of both
the position Q
7n
and the intensity I
7n
of the new scattering peak. Here, it
is also noticed that the induction period can be divided into two stages;
Q
7n
remains constant within the first 20 min and then decreases, following
a power law. At the same time, the increase in I
7n
is first exponential while
after about 20 min it obeys another power law. These behaviors resemble a
Structure Formation in PET During the Induction Period of Crystallization 231
OQ
1U
4J
'S
2
fc
5
8
S-I
+3

Jl
QJ
O

; 4
CD
J2

CD
+= _
O
Annealing time
- (a) 3 min
23 min
+ 62 min
&

115 min

"


+ -
4- ++

- +-h
-
+
+ "^
1 K .'- A
* ^^^
~ S5$?3 WPf^P^^
0.3 0.9 1.2 1.5
137 min
187 min
+ 233 min
313 min
O
Figure 5. Time evolution of the SAXS intensity difference for PET, isothermally
crystallized at 8O
0
C from the glass: (a) induction stage; (b) crystallization stage
[17]
spinodal decomposition type of phase separation. Thus, the process of SD
is divided into at least an early stage and a late stage. According to the
linearized theory of Cahn and Hilliard [35], in the early stage the wave-
length of density fluctuations, the so-called characteristic wavelength, A,
remains constant and only its amplitude increases. In other words, in the
early stage the peak position does not change, but the peak intensity in-
creases exponentially with time. These features are in good agreement with
the observations described above. On the other hand, the late stage, which
is described by the theory of Binder and Stauffer [36] and the scaling theory
of Furukawa [37], reflects the saturation of the amplitude of density fluctu-
ations to an equilibrium value, but the characteristic wavelength, A, grows
with time. In this case, the density fluctuations are caused by the forma-
232 K. Kaji
*3
1.0
0.1
(a)
Im~t
(b)
-0.2
10
Time (min)
100
Figure 6. Time evolution of (a) the maximum intensity, /
m
, and (b) the maximum
position, Q
m
, of the new SAXS peak for PET, isothermally crystallized at 8O
0
C
from the glass [17]
tion of clusters of higher density (the cluster regime) and hence A is about
twice as large as the average diameter, R
7n
(I], of the clusters, A ~IR
7n
(L).
According to Binder and Stauffer, the time dependence of R
7n
(L) can be
assumed to be R
7n
(L) ~t
a
, a = l/(3 +d), d being the dimensionality. In
this case, the peak position Q
7n
becomes
If the structure grows with a self-similarity owing to the diffusion and
reactions of clusters, the size distribution function of the clusters can be
scaled with a single parameter R
7n
(t) and hence the scattering function can
also be scaled by Q
7n
(I). Furukawa derived such a function for the cluster
regime as
where = QfQ
7n
(L) and 5(1) = 1. This function is time-independent
and called a universal scaling function. Therefore, the scattering intensity
/(Q, t) is given by
/(Q, (3)
Structure Formation in PET During the Induction Period of Crystallization 233
Annealing time
_L .O r\
42 mm
62 min
81 min
129 min

. *"

.

.
1.0
0.5
O 0.4 0.8 1.2 1.6 2.0

Figure 7. Universal scaling functions, (#), calculated from the SAXS intensities
of PET (isothermally crystallized from the glass) in the late stage of spinodal
decompostion. The solid line indicates a theoretical curve by Furukawa [17]
because it is proportional to the mass or the volume R^
n
(Y) of a cluster.
The peak intensity /
m
(t) is obtained by substituting = 1 into Eq. (3)
Im(t) ~Q-
3
(i)~
30
t
b
(4)
where b = 3. Hence in three dimensions (d = 3) one gets a = 1/6, b = 1/2,
and the ratio b/a = 3.
In the following, let us examine the theoretical predictions described
above for the late stage, using the experimental data. As seen in Figure 6,
the experimental data give the relations Q
7n
~t~
2
and I
7n
~
6
(hence
= 0.2, b = 0.6) after 20 min. These exponents roughly agree with the the-
oretical values for d = 3 [36], = 1/6 and b = 1/2, respectively, within the
experimental error. Furthermore, the universal scaling function S(x) was
calculated using Eqs. (2) and (3), and the results are shown in Figure 7;
S(x) is independent of the annealing time within the experimental error,
indicating that the structure grows, keeping self- similarity in the late stage,
in accordance with Furukawa's theory. The quantitative examinations de-
scribed so far suggest that some kind of spinodal decomposition actually
occurs during the induction period prior to crystallization.
234 K. Kaji
2.4. Structural sizes in crystallization processes
Several sizes, originating from microphase separation occurring in the in-
duction period, as well as from crystallization, were estimated in order
to establish the way crystallization proceeds [38]. For the process of mi-
crophase separation, the characteristic wavelength and the average size of
dense domains with partially parallel orientation of segments or nematic
structure were determined as a function of annealing time. The character-
istic wavelength = 2/(5
7
was obtained from the peak position, Q, of
SAXS. The dense domain size DI for the early stage of SD was calculated
from the spatial density correlation function 7(7) defined by Debye and
Bueche [39]:
7
(r) = Mr
1
)TXr
2
)V(T
7
V (5)
where 77(7*1) and 77(7*2) are the local density fluctuations at points r and r 2
with r = r2\ from the average value (
2
}^. The function 7(7) is given
by inverse Fourier transform of SAXS intensity /(Q), and DI is obtained
from the minimum value of r at 7(7) = O. The dense domain size D
2
for
the late stage of SD is obtained from a method by Komura et al [4O]. Thus,
the Fourier transform F(y) of the universal function (Eq. (2)) is given by
OO
F(y) = (1/2
2
) / x
2
S(x) sin xy/xydx (6)

where y = rQ
7n
, and the minimum value yi of y at F(y) = O gives D
2
.
According to this calculation, y\ = 2.517 and hence D
2
= 2.571/Qm- After
crystallization, the long period, L, due to the alternation of crystalline and
amorphous layers, was determined from the usual SAXS peak appearing
at a higher Q. Furthermore, the values D
0
of the crystalline lamellar thick-
ness were estimated from the one-dimensional density correlation function,
based on a method by Strobl et al. [41]. The results are shown in Figure 8.
In the early stage of spinodal decomposition, A remains 15 nm, and in
the late stage it increases up to 29 nm just before crystallization while the
dense domain size D (Di and D
2
for the early and late stage, respectively)
increases from 6.0 to 8.5nm during this period; the increase in D is quite
small, as compared with A. The cause for this observation is not clearly
understood, but one possible explanation is as follows. In the late stage the
clusters grow with time by diffusion and collision. However, at very low tem-
peratures only slightly above T
9
, the clusters may not be fused completely
because a high concentration of entanglements on the cluster surfaces can-
not diffuse out to the outer surfaces of larger fused clusters because of
their low molecular mobility. Thus, the larger clusters may be considered
containing amorphous regions with entanglements in the nematic liquid
crystalline domains. Such clusters cause the so-called long period structure
Structure Formation in PET During the Induction Period of Crystallization 235
30
20
a
O
10
_L
100 200
Annealing time (min)
300 400
Figure 8. Annealing time depencence of various parameters of the crystallization
process of PET, isothermally crystallized at 8O
0
C from the glass: A: characteristic
wavelength of spinodal decomposition, D: dense domain size, L: long period, b
p
:
persistence length, D
0
: lamellar stem length
inside themselves, probably with an alternation of smectic layers and thick
amorphous layers, before crystallization, and then the smectic regions con-
vert into lamellar crystals. This possibility is supported by the fact that the
long spacing (8.6nm) just after the start of crystallization is almost equal
to the dense domain size D just before crystallization; the alternation of
the smectic phase and the amorphous phase may be reflected by the long
period structure.
The crystalline lamellar thickness obtained by the Strobl method is
initially 1.4nm and grows to about 2.0nm which is roughly equal to the
value of 2.8nm estimated from WAXS. Interestingly, the persistence length
measured by SANS just after crystallization [21,22], b
p
= 1.32nm, almost
equals the crystal thickness.
3. Structure formation during melt crystallization
In this section, the possibility of spinodal decomposition by crystallization
of the polymer directly from the molten state, is discussed.
236 K. KaJi
3.1. SAXS measurements
Figure 9 shows the time evolution of the SAXS intensity difference observed
in situ when a PET sample was crystallized by cooling down to 244
0
C from
the melt at 29O
0
C (T
m
~267
0
C) in the camera [32]; in this case, the in-
duction period was about 110 min. As seen in Figure 9a, the intensities at
low Q values increase with time, but no peak is observed. After crystalliza-
tion, a well known long period peak appears at around O.Snm"
1
and shifts
to 0.4Um"
1
with increasing intensity (Figure 9b). For what reason is the
characteristic peak of SD invisible during melt crystallization? As pointed
out in the Introduction, either the characteristic wavelengths of SD greatly
differ in the cases of glass and melt crystallization or the mechanism of
microphase separation is completely different, as theoretically predicted by
Olmsted et al. [IS]. Subsequent investigations suggested that both possibil-
ities exist. Thus, a critical temperature, namely the spinodal temperature
-B
'3
O
il
X

T3
t
1.4
1.2
1.0
0.8
0.6
0.4
0.2
n n
V
9
V Annealing time
x
X 121 min
D
D 101 min
- -
1
o'o O
O
&&
74 min
50 min
29 min
20 min
11 min
-
0.0 0.1 0.2 0.3 0.4
0.1 0.2 0.3 0.4
Q (nm-
1
)
0.5
Figure 9. Time evolution of the SAXS intensity difference for PET, isothermally
crystallized at 244
0
C from the melt [32]
Structure Formation in PET During the Induction Period of Crystallization 237
T
5
actually exists; above T
5
, the phase separation type drastically changes
to the nucleation and growth mechanism. The crystallization temperature
244
0
C should be above T
8
. In order to observe SD in melt crystallization,
the crystallization temperature must be lower than T
3
. Furthermore, the
SD characteristic wavelength was presumed much larger in the case of melt
crystallization. A preliminary result from confocal laser microscopic obser-
vations suggests that it is of the order of [42]. In this case one may
expect that the SD pattern can be directly observed with an optical mi-
croscope, as shown in the next subsection.
Aside from this, another interesting problem is that the long period
peak after crystallization slightly shifts to higher Q with time (also seen
in Figure 9b). This leads to a seemingly strange result that the long pe-
riod decreases with time. Gehrke et al. [43], who were the first to deal
with this problem, explained it in the following way. Initially formed crys-
talline lamellae have wavy surfaces, but as the crystallization proceeds,
these become smooth, resulting in a decrease in the long period. Based on
our results, we propose another, more plausible explanation. A late crys-
tallization occurs even in the less dense regions, and the thickness of the
respective lamellae is smaller because of the higher density of entangle-
ments and gives shorter long periods. These thinner lamellae would give
additional scattering intensities at higher Q values, shifting Q
7n
to a wider
angle.
3.2. Direct observation of spinodal decomposition
As described in the previous subsection, we could not observe the char-
acteristic wavelength of SD for the melt crystallization by SAXS, but we
showed a possibility of observing SD patterns by light microscopy. After
considerable efforts, mainly owing to Nishida, we have finally succeeded
in observing such SD patterns [44,45]. Figure 10 shows such light micro-
graphs as a function of time, for PET quenched from the melt (29O
0
C)
to 13O
0
C. Ten seconds after quenching, the SD pattern is not yet seen,
but within 15-2Os a faint SD pattern having a characteristic wavelength
A ~ 4 starts to appear, and, after crystallization begins at around 40 s,
the contrast of SD becomes more intense, increasing with time to attain a
maximum between 40 and 60 s. However, after ca. 70 s, the SD pattern be-
comes weaker. This phenomenon is in agreement with a time-resolved light
scattering observation in the early stage of crystallization of iPP by Okada
et al. [46]. The reason may be that the density increase in the less dense
domains, due to later crystallization, diminishes the contrast between the
dense and the less dense domains. Here, it should be noted that A hardly
increases with time, and crystalline entities, that may not be spherulites,
cannot grow over the size of the dense domains at this temperature.
On the other hand, when the crystallization temperature is sufficiently
high (22O
0
C), spherulites can grow over the dense domain size even after
238 K. Kaji
s ,;> '
1
^
s
\' N ; ^ ^' n^^A^MV*' ^;^*\,
Figure 10. Light micrographs indicating time evolution of the sructure of PET,
isothermally crystallized at 13O
0
C immediately after quenching from the melt at
29O
0
C
Structure Formation in PET During the Induction Period of Crystallization 239
Figure 11. Polarized light micrograph indicating the initial growth of spherulites
of PET, crystallized at about 22O
0
C from the melt. The distribution of spherulites
shows a spinodal-like pattern [32]
crystallization. In this case, a spinodal decomposition-like pattern can be
observed by polarized light microscopy (PLM), as shown in Figure 11; the
dense and dilute domains of spherulites can be clearly distinguished just
after crystallization and the interdomain spacing between them inereases
to several tens of microns with time.
3.3. Comparison of the characteristic wavelengths of glass and
melt crystallization
As shown in the previous subsection, the characteristic wavelengths A of the
glass and melt crystallization differ surprisingly by two orders of magnitude;
they amount to a few tens of nanometers and microns, respectively. At
the moment, the cause for this observation is unclear. However, here a
discussion is attempted on the basis of recent additional data of ours. In
the case of melt crystallization, A follows the relation of van Aartsen [47]
(2

-T)]^
2
(7)
where / is a range of molecular interaction and T
8
is the spinodal tem-
perature; for PET, these are ca. 0.24 and 213
0
C, respectively. On the
other hand, Eq. (7) does not work for glass crystallization. When the PET
glass was crystallized between 120 and 24O
0
C, A of about 2 was ob-
tained, regardless of the crystallization temperature. At 12O
0
C, this value
240 K. Kaji
agrees with the prediction of Eq. (7) but, surprisingly, it remains almost
constant at higher temperatures. This strange phenomenon might be re-
lated to a memory effect of the glass. On the basis of these observations,
the great difference between melt and glass crystallization in number den-
sity of spherulites, shown by van Krevelen [5], might be explained by the
assumption that crystal nuclei are formed within A\ for the former case,
A may become infinite as the temperature approaches T
8
. Furthermore, A
is of the order of tens of nanometers in the case of crystallization from
the glass below 115
0
C. Hence, there might exist a critical crystallization
temperature between 115 and 12O
0
C. Again, the cause for this is unknown,
but it might be related to the range of molecular interaction of Eq. (7).
4. Cause of the spinodal decomposition process
The SD is a phase separation process, usually occurring in systems con-
sisting of more than two components, such as solutions and blends. In the
present case, however, the pure PET system employed is composed of one
component. Then, what triggers such an SD-type phase separation? Doi et
al [14,26] proposed a dynamic theory of the isotropic-nematic phase tran-
sition for liquid crystalline polymers and showed that the orientation of
rod-like molecules causes an SD-type phase separation. Our experimental
finding of SD may be due to such orientational fluctuations. In 1956, Flory
[48] proposed a concept of two-step crystallization parallel orientation
of the rigid segments and their closest packing. We believe our finding is
direct evidence of his concept, though Flory did not predict the SD.
Here, we first elucidate the Doi theory in some detail and then show our
experimental results to clarify the cause of SD, based on the assumption
that the partial extension of polymer chains triggers the parallel orientation
of the stiff segments during the induction period of crystallization.
4.1. Doi's theory
Figure 12 is a schematic diagram, explaining Doi's theory [14,26]. It is well
known that polymer molecules assume a random coil (Gaussian) confor-
mation and are entangled with one another. When the sample is quenched
down to a temperature at which the polymer can crystallize, the polymer
chains tend to assume a crystalline conformation which is energetically the
most stable. However, this is a difficult process because of two resistant
factors, z.e., chain entanglements and the increase of excluded volumes of
crystalline sequences. The entanglements greatly delay the process and the
extension of the crystalline sequences makes the system unstable because
of the rapid increase in their excluded volumes. The latter relation is given
by Doi et al [14]
Ve
x
Ci = 2dL
2
\ sin \ (S)

You might also like