Chemorheology of Polymers From Fundamental Principles To Reactive Processing PDF
Chemorheology of Polymers From Fundamental Principles To Reactive Processing PDF
Chemorheology of Polymers From Fundamental Principles To Reactive Processing PDF
Chemorheology of Polymers
From Fundamental Principles
to Reactive Processing
PETER J. HALLEY
University of Queensland
GRAEME A. GEORGE
Queensland University of Technology
ISBN-13
978-0-511-53984-8
eBook (EBL)
ISBN-13
978-0-521-80719-7
hardback
Contents
Preface
1
page ix
1
1.1
1
2
5
8
11
23
24
36
59
85
99
105
106
111
126
127
128
139
149
162
169
2.1
2.2
169
169
169
170
175
176
177
179
180
181
181
2.3
2.4
Chapter rationale
Polymer physics and dynamics
2.2.1 Polymer physics and motion early models
2.2.2 Theories of polymer dynamics
Introduction to the physics of reactive polymers
2.3.1 Network polymers
2.3.2 Reactively modied polymers
Physical transitions in curing systems
2.4.1 Gelation and vitrication
2.4.2 Phase separation
2.4.3 Timetemperature-transformation (TTT) diagrams
vi
Contents
186
186
187
191
192
195
3.1
3.2
195
196
196
197
202
203
206
207
208
208
209
213
2.5
216
222
235
240
244
255
259
259
263
271
272
275
280
282
282
283
287
292
305
311
312
321
4.1
4.2
321
321
321
Introduction
Chemorheology
4.2.1 Fundamental chemorheology
Contents
vii
4.3
Chemoviscosity proles
4.3.1 Chemoviscosity
4.3.2 Gel effects
4.4 Chemorheological techniques
4.4.1 Standards
4.4.2 Chemoviscosity proles shear-rate effects, gs gs(c, T)
4.4.3 Chemoviscosity proles cure effects, gc gc(a, T)
4.4.4 Filler effects on viscosity: gsr(F) and gc(F)
4.4.5 Chemoviscosity proles combined effects, gall gall(c, a, T)
4.4.6 Process parameters
4.5 Gelation techniques
References
327
327
336
336
338
338
342
343
344
344
345
347
351
5.1
5.2
Introduction
Chemoviscosity and chemorheological models
5.2.1 Neat systems
5.2.2 Filled systems
5.2.3 Reactive-extrusion systems and elastomer/rubber-processing systems
5.3 Chemorheological models and process modelling
References
351
351
351
357
370
370
371
375
6.1
6.2
6.3
6.4
6.5
6.6
Introduction
Casting
6.2.1 Process diagram and description
6.2.2 Quality-control tests and important process variables
6.2.3 Typical systems
6.2.4 Chemorheological and process modelling
Potting, encapsulation, sealing and foaming
6.3.1 Process diagram and description
6.3.2 Quality-control tests and important process variables
6.3.3 Typical systems
6.3.4 Chemorheological and process modelling
Thermoset extrusion
6.4.1 Extrusion
6.4.2 Pultrusion
Reactive extrusion
6.5.1 Process diagram and description
6.5.2 Quality-control tests and important process variables
6.5.3 Typical systems
6.5.4 Chemorheological and process modelling
Moulding processes
6.6.1 Open-mould processes
6.6.2 Resin-transfer moulding
375
375
375
375
376
376
378
378
379
379
380
380
380
382
385
385
387
388
389
391
391
393
viii
Contents
395
397
400
403
405
406
407
407
409
413
413
415
416
417
420
420
424
426
426
429
431
435
440
Preface
Plastics are the most diverse materials in use in our society and the way that they are
processed controls their structure and properties. The increasing reliance on plastics for
high-value and high-performance applications necessitates the investment in new ways
of manufacturing polymers. One way of achieving this is through reactive processing.
However, the dynamics of reactive processes places new demands on characterization,
monitoring the systems and controlling the complete manufacturing process.
This book provides an in-depth examination of reactive polymers and processing, rstly
by examining the necessary fundamentals of polymer chemistry and physics. Polymer
characterization tools related to reactive polymer systems are then presented in detail with
emphasis on techniques that can be adapted to real-time process monitoring. The core of the
book then focuses on understanding and modelling of the ow behaviour of reactive
polymers (chemorheology). Chemorheology is complex because it involves the changing
chemistry, rheology and physical properties of reactive polymers and the complex interplay
among these properties. The nal chapter then examines a range of industrial reactive
polymer processes, and gives an insight into current chemorheological models and tools
used to describe and control each process.
This book differs from many other texts on reactive polymers due to its
The purpose of this chapter is to provide the background principles from polymer physics
and chemistry which are essential to understanding the role which chemorheology plays in
guiding the design and production of novel thermoplastic polymers as well as the complex
changes which occur during processing. The focus is on high-molar-mass synthetic polymers and their modication through chemical reaction and blending, as well as degradation
reactions. While some consideration is given to the chemistry of multifunctional systems,
Chapter 2 focuses on the physical changes and timetemperature-transformation properties
of network polymers and thermosets that are formed by reactions during processing.
The attention paid to the polymer solid state is minimized in favour of the melt and in
this chapter the static properties of the polymer are considered, i.e. properties in the
absence of an external stress as is required for a consideration of the rheological properties. This is addressed in detail in Chapter 3. The treatment of the melt as the basic
system for processing introduces a simplication both in the physics and in the chemistry
of the system. In the treatment of melts, the polymer chain experiences a mean eld of
other nearby chains. This is not the situation in dilute or semi-dilute solutions, where
density uctuations in expanded chains must be addressed. In a similar way the chemical
reactions which occur on processing in the melt may be treated through a set of homogeneous reactions, unlike the highly heterogeneous and diffusion-controlled chemical
reactions in the solid state.
Where detailed analyses of statistical mechanics and stochastic processes assist in the
understanding of the underlying principles, reference is made to appropriate treatises, since
the purpose here is to connect the chemistry with the processing physics and engineering of
the system for a practical outcome rather than provide a rigorous discourse.
1.1
H H
H H
2.54
1.54
109.5
H
H
H H
H H
H H
Figure 1.1. The carboncarbon backbone of a polyethylene chain in its extended planar
(all-trans) conformation (a) and its kinked, out-of-plane (cis) conformation (b).
l
Rrms
Figure 1.2. A schematic diagram of a freely jointed polymer chain with n segments of length l
(in this case n 14) showing the end-to-end distance, Rrms.
sp3-hybridized tetravalent site of carbon that denes the angles and distances between the
atoms along the backbone is shown in Figure 1.1 in
(a). an all-trans conformation with a planar CC backbone and
(b). with the introduction of a cis conformation (as occurs in a cyclic six-membered hydrocarbon, cyclohexane), which allows the chain to kink out of the plane and change direction.
In the following we will initially consider the simpler concept of a freely jointed chain in
which none of these constraints are present.
1.1.1
1:1
1:2
2
Rrms
6Rg 2 :
1:3
Thus,
From mechanics, the radius of gyration, Rg, is the average value of the rst moment of all
segments of the chain with respect to the centre of mass of the chain. If the chain is fully
extended with no constraints regarding bond angles, i.e. a fully jointed chain as dened by
Kuhn, then the maximum value of Rrms becomes
Rmax nl:
1:4
2
The ratio (Rrms
/Rmax) is a measure of the stiffness of the chain and is termed the Kuhn
length. Thus, if there is a hypothetical freely jointed polyethylene that has 1000 carbon
atoms separated by 1.54 A then Rmax 1540 A, Rrms 49 A and Rg 20 A.
The limitations of the random-ight model when applied to real polymer chains arise from
1:5
Experimental values of this parameter are given in Table 1.1 and it may be seen that the actual
end-to-end distance of a polyethylene molecule with 1000 carbon atoms (degree of poly
merization DP of 500) is 126 A from Equation (1.5) (i.e. C1/2n1/2l) rather than 49 A from the
Kuhn model, Equation (1.1). Data for several polymers in addition to polyethylene are given,
including a rigid-rod aromatic nylon polymer, poly(p-phenylene terephthalamide) (Kevlar),
as well as the aliphatic nylon polymer poly(hexamethylene adipamide) (nylon-6,6).
Comparison of the values of C for the polymers with a exible CC or SiOSi
backbone (as occurs in siloxane polymers) of about 610 with the value for the rigid-rod
polymer of 125 demonstrates the fundamental difference in the solution properties of the
latter polymer which has a highly extended conformation characteristic of liquid-crystal
polymers. Equation (1.5) also shows that for a real chain the value of R0 would be
expected to increase as the half power of the number of repeat units, i.e. the degree of
polymerization, DP1/2.
Table 1.1. Experimental values of the characteristic ratio, C, for Equation (1.5)
Polymer
Characteristic ratio, C
Poly(ethylene)
Poly(styrene)
Poly(hexamethylene adipamide), Nylon-6,6
Poly(p-phenylene terephthalamide), Kevlar
6.7
10.2
5.9
125
1:6
The magnitude of a depends on the forces of interaction between the solvent and the
polymer chain. Thus, if the polymer is polar, when it dissolves in a polar good solvent, it
will expand and a is large. The converse is true for poor (eg. non-polar) solvents and the
chain will contract to lower than the unperturbed dimensions and, in the limit, the polymer
may precipitate from solution. When a combination of solvent and temperature is found
that is neither good nor poor, i.e. a 1, then the chainsolvent and polymerpolymer
interactions balance and R0 is the unperturbed dimension of the chain. For a particular
solvent, the temperature at which this occurs is the h-temperature.
An interesting calculation is that of the volume occupied by the segments themselves
compared with the total volume that the chain occupies. The diameter of a sphere within
which the chain spends 95% of the time is about 5R0. Since the chain segments occupy only
about 0.02% of this volume, the remaining space must be occupied by other chains of
different molecules both when the polymer is under h-conditions and in the presence of
solvent molecules when it is expanded. Thus, except in very dilute solutions, polymer
molecules interpenetrate one anothers domains so that intermolecular forces between
chains are signicant.
CHAIN IN MELT
1.0
CHAIN IN SOLUTION
Local Monomer
Concentration
cm 0.5
chain
centre
0
x
Figure 1.3. Comparison of the change in local monomer concentration with distance from the chain
centre for a random chain in solution and in the melt. Adapted from Strobl (1996).
wm ve cm kT:
1:7
This produces a net force for all non-uniform density distributions so that for the bell-shaped
distribution in Figure 1.3 there will be a net force of expansion of the chain. When the melt
is considered, every chain is surrounded by a chain of the same type, so the concentration cm
is constant in all directions (the dotted line in Figure 1.3). No distinction is drawn between
repeat units on the same or different chains. (As noted above, there will be interpenetration
of chains in all but dilute solutions.) The result is that there is no gradient in potential and
there are no forces of expansion. In effect, the polymer chain in the melt behaves as if the
forces of expansion due to excluded volume were screened from each chain and the
dimensions are those for the unperturbed chain.
This result may, by a similar argument, be extended to the interpenetration of chains
as random coils in the amorphous solid state. These results will be of importance when
the rheological properties of the melt through to the developing solid are considered in
Chapters 2 and 3.
1.1.2
(a)
H
+120
H
+120
H
Gauche+ (G+)
Trans (T)
H H
(b)
Gauche (G)
H H
H
H
H
H
Eclipsed H
(E)
Eclipsed Chain
(E)
+60
+60
+60
Eclipsed H
(E)
+60
+60
+60
Gauche+
+
(G )
H
H
H
H
Gauche
(G )
H
H
Trans (T)
Trans (T)
Energy
kJ/mol
E
60
120
180
240
300
360
Figure 1.4. (a) Conformations adopted by a segment of a polymer chain by successive rotation about
a CC bond. The balls represent the carbon atoms from the continuing chain (initially in an
all-trans extended-chain conformation). (b) Changes in conformational energy on successive
rotation of an all-trans extended-chain conformation by 60 about a CC axis.
This shows the successive rotation by 120 about the central CC bond of the adjacent
methylene group. Initially all carbon bonds lie in a plane and then after each rotation a
hydrogen atom lies in the initial plane. A detailed analysis may be made of the rotational
isomeric states of model compounds progressively from ethane, butane and pentane to
determine the energy states of the conformers and, from the Boltzmann distribution, their
populations (Boyd and Phillips, 1993).
The depiction of the conformers is facilitated by a simple schematic approach in which the
atoms in Figure 1.4(a) are viewed along the central CC bond initially in the trans (T)
conformation and then rotation of the groups clockwise by 60 occurs in succession about this
axis (Figure 1.4(b)). Analysis of these conformations identies the energy maxima (eclipsed,
E, conformations) and the energy minima (trans and gauche conformations) separated by up to
21 and 18 kJ/mol, respectively, as shown in the energy prole in Figure 1.4(b).
For polyethylene, the actual bond rotation from the trans (T) position to the other stable
conformers (the gauche positions, G and G, respectively) is slightly less than 120 due to
unsymmetrical repulsions (Flory et al., 1982). There are situations in which the repulsion
due to steric crowding results in further deviations. For example, the sequence TGG will
produce a structure with a sharp fold where the steric repulsion between methylene groups
no longer allows an energy minimum. This is accommodated by a change in the angle of
rotation giving an angle closer to that for the trans position (the so-called pentane effect)
(Boyd and Phillips, 1993, Strobl, 1996).
When other groups are introduced into the polymer chain, such as oxygen in poly(oxy
methylene) [CH2O]n, the most stable conformation is no longer the all-trans chain but the
all-gauche conformation GGG, etc. This means that the chain is no longer planar but
instead is helical. The stability of the gauche conformation over trans is linked in part to the
electrostatic interactions due to the polar oxygen atom in the chain (Boyd and Phillips, 1993).
These conformations translate to the most stable structure expected at low temperatures.
However, the low energy barriers between isomeric states mean that in the melt a large
number of conformations is possible, as indicated in the previous section where the melt is
seen to reproduce the properties of an ensemble of ideal random interpenetrating coils.
H Me
H H
H Me
H H
Meso diad
Me H
H Me
H H
H H
Racemic diad
Figure 1.5. A schematic diagram illustrating meso (m) and racemic (r) diads.
As will be discussed later, special synthetic techniques are required to achieve isotactic
and syndiotactic structures, and polypropylene, the example above, achieved commercial
success only through the discovery of stereoregular polymerization to achieve the isotactic
structure. The measurement of the degree of tacticity of a polymer is achieved through 13C
NMR studies of the polymer in solution (Koenig, 1999).
Isotactic polypropylene will adopt a conformation very different from the extended chain of
polyethylene. In the early part of Section 1.1.2 it was noted that the minimum-energy conformations were considered to be attained by rotation about the CC backbone and this
introduced the possibility of gauche conformers as alternative energy minima. This can now be
performed on the meso dyad in isotactic polypropylene by considering rotations about the two
CC bonds that will minimize the interactions between the pendant methyl groups. The starting
point in this analysis is the nine near trans and gauche conformers since these dene the local
minima in energy of the backbone in the absence of the methyl groups. Introduction of the
steric repulsion by the methyl groups in a TT conformation (Figure 1.5) suggests that this is not
going to be a likely conformation and the conformers which are able to minimize the repulsion
due to methyl groups in a meso dyad are limited to TG and GT. Just as a helix was generated
when gauche conformers were accessible minima in poly(oxymethylene), so too we have two
possible helices if the chain consists of m-dyads as in isotactic polypropylene. For TG it will
be right-handed and for GT it will be left-handed. This helix will have three repeat units in
one turn of the helix, i.e. a 3/1 helix, and this is the form which crystallizes.
In syndiotactic polypropylene, the methyl groups are well separated and the TT
form is favoured, but there are other energy minima among the gauche conformations and
TT/GG and TT/GG sequences can generate left- and right-handed helices, respectively,
where the repulsions are minimized (Boyd and Phillips, 1993). The chains may crystallize
both in the TT and in the TTGG form, so syndiotactic polypropylene is polymorphic.
1.1.3
The addition polymerization reactions, discussed later in Section 1.2, result in the growth
of polymer chains that consist of chemically identical repeat units arising from addition
reactions of the original monomer, terminated by groups that will be chemically different
from the repeat unit due to the chemistry of the reaction, the starting materials (e.g. initiators, catalyst residues), which may be attached to the chain, and impurities. Since these are
generally only a very small fraction of the total polymer mass, the effect of the chemistry of
the end groups can be ignored to a rst approximation, although their quantitative analysis
provides a method for estimating the number average molar mass as discussed below.
Particular defects such as chain branching, must be taken into account when the molar
massproperty relationships are developed since the chain is no longer linear.
The mass of the linear polymer chain is thus related directly to the number of monomer
units incorporated into the chain (DP) and will be M0 DP, where M0 (g/mol) is the molar
mass of the monomeric repeat unit. Thus, if all chains grew to exactly the same DP, then
M0 DP, would be the molar mass of the polymer. If the end groups on the chain can be
readily and uniquely analysed, then an average molar mass, Mn, or number-average
molecular weight (as discussed in the next section) can be immediately determined since,
if there are a mol/g of end group A and b mol/g of end group B then
Mn 2=a b g=mol:
1:8
The conformation, end-to-end distance and radius of gyration of the polymer would be
described by the simple considerations in Section 1.1.1. In the real polymer, the length of
the polymer chain is controlled by the statistics of the chemical process of polymerization, so
the distribution of chain lengths will depend on the reaction chemistry and conditions. The
distribution is discontinuous since the simple linear chain can increase only in integral
values of the molar mass of the repeat unit, M0. The chain mass also includes that of the end
groups Me, so the rst peak appears at M0 Me, and then increments by DP M0 as shown in
Figure 1.6(a). When the molar mass is low, as in oligomers, the individual polymer chains may
be separated by chromatographic or mass-spectroscopic techniques and a distribution such as
that shown in Figure 1.6(a) is obtained. For the large molar masses encountered in vinyl
polymers (>105 g/mol) the increment in molar mass for each increase in DP is small and the
end-group mass is negligible compared with the total mass of the chain. The distribution then
appears to be continuous and sophisticated analytical methods such as MALDI-MS are
required to resolve the individual chains (Scamporrino and Vitalini, 1999).
Size-exclusion chromatography (SEC) has become the technique of choice in measuring the
molar-mass distributions of polymers that are soluble in easily handled solvents (Dawkins,
1989). The technique as widely practised is not an absolute method and a typical SEC system
must be calibrated using chemically identical polymers of known molar mass with a narrow
distribution unless a combined detector system (viscosity, light scattering and refractive index) is
employed.
The effect of the chemical reactions during polymer synthesis on the molar-mass distribution
is discussed in Section 1.2, but prior to this it is important to consider the various averages
and the possible distributions of molar mass that may be encountered. It is then possible to
examine the experimental methods available for measuring the distributions and the averages
which are of value for rationalizing dependence of properties on the length of the polymer chain.
10
(a)
wi
Molecular Weight
[M0+M e] [2M0+Me] etc................................... [15M0+Me ]
(b)
Mn
Mw
Mz
wi
The simple averages that are used for propertymolar-mass relations of importance in
chemorheology as used in this book are the following:
X
X
X
wi Mi
Ni Mi
Ni ;
Mn 1
1:9
P
where wi is the weight fraction of species i (i.e. i wi 1) and Ni is the number of
molecules with molar mass Mi;
(b). the weight average, Mw:
X
X
X
2
Mw
wi Mi
Ni Mi
Ni Mi ;
1:10
i
X
i
wi Mi2
X
i
w i Mi
X
i
Ni Mi3
X
Ni Mi2 :
1:11
Thus, when considering what these different averages describe regarding the actual molarmass distribution, it is useful to consider the continuous distribution as shown in Figure 1.6(b).
11
It is seen that the averages correspond to the rst, second and third moments of the
distribution and the ratio of any two is useful as a way of dening the breadth of
the distribution. Thus the polydispersity is given by the ratio Mw/Mn. A normal or
Gaussian distribution of chain lengths would lead to Mw/Mn 2.
The experimental measurement of these averages has largely been performed on polymers in solution (Hunt and James, 1999). Since Mn depends on the measurement of the
number of polymer chains present in a given mass, colligative properties such as vapourpressure depression DP (measured by vapour-phase osmometry) and osmotic pressure
(measured by membrane osmometry) relative to the pure solvent, can in principle provide
the molar mass through an equation of the form
DP Kc=Mn ;
1:12
where c is the concentration of the polymer in solution and K is a constant for the colligative
property and the solvent. As noted before, end-group analysis also provides a measure of
Mn. All of these techniques lose precision at high values of molar mass because the change
in property becomes extremely small. As may be seen from Figure 1.6(b), the numberaverage molar mass is biased to low molar mass.
The weight-average molar mass, Mw, may be obtained by light scattering (Berry and
Cotts, 1999). An analysis of the Rayleigh scattering of a dilute solution at various angles
and concentrations as well as the difference in the refractive index between solution and
solvent for these concentrations allows the measurement of the weight-average molar mass,
as well as the mean square radius of gyration. The technique is extremely sensitive to any
scattering impurity or particle and any aggregation that may occur.
Ultracentrifugation is less sensitive to these effects and also enables a value of Mw to be
obtained, but because of the specialized nature of the equipment required is not as widely
used for the study of synthetic commercial polymers as light scattering. It is possible to
determine a value for Mz through measurements of sedimentation at various rotor speeds
(Budd, 1989). As noted in Figure 1.6(b), the weight-average molar mass is biased to higher
molar mass on the distribution.
Viscosity studies of dilute solutions provide a convenient relative measure of the molar
mass and the resultant average, Mv, will lie closer to Mw than to Mn. The behaviour of
polymer melts will be discussed in detail later, but it is noted that the melt viscosity is a
strong function of the weight-average molar mass since the parameter m in the relation
g kMwm
1:13
changes from 1.0 to 3.4 when the critical molar mass for entanglements is reached. This
value varies with the chemical composition of the polymer.
1.1.4
12
Tc
1.25
Tm
Tg
Specific
Volume
cm3/g
1.0
150 120
110 150
Temperature C
Figure 1.7. Changes in specic volume as a polymer melt (e.g. polyethylene) is cooled from above
of the polyolens is their semi-crystalline nature, i.e. the solid polymer contains both
amorphous and crystalline material.
If we consider the process of cooling molten polyethylene, there will be a progressive
decrease in the volume that the chains occupy. This specic volume, Vs, is the reciprocal of
the density and this is shown in Figure 1.7 for the case on cooling the polymer from 150 C
to 150 C. It is seen that there is a linear decrease in Vs with decreasing temperature, which
is consistent with the coefcient of thermal expansion.
As the temperature of crystallization, Tc, is approached, there will be a sudden decrease in
Vs and an exothermic process corresponding to a rst-order phase transition. Thermodynamically this corresponds to a discontinuity in the rst derivative of the free energy, G,
of the system with respect to a state variable, i.e. in this case a discontinuity in volume:
@G=@PT V :
1:14
The extent to which this occurs in the polymer is controlled by a number of factors, and the
volume fraction of crystalline material, uc, is related to the density, q, by
c q qa =qc qa ;
1:15
where qc and qa are the densities of the crystalline and amorphous phases of the polymer,
respectively. These values may be obtained from X-ray diffraction and scattering data. The
actual structure of this crystalline material is considered in the following section.
The process of crystallization from the melt takes a considerable period of time,
in contrast to the situation with a low-molecular-mass hydrocarbon (e.g. C44H90)
(Mandelkern, 1989) that will crystallize over a temparature range of less than 0.25 C.
This results in the curvature in Figure 1.7, since signicant undercooling, of up to 20 C, is
required in order for the crystallinity to develop. The detailed curve prole and the degree
of crystallinity, uc, depend both on the degree of polymerization and on the molar-mass
distribution (Mandelkern, 1989). These results highlight the reason for the undercooling,
13
and this is the difculty of extracting ordered sequences of the polymer chain and forming
a thermodynamically stable structure. As discussed in the next section, this is even more
difcult for polymers that have a more complex conformation than polyethylene. Thus a
semi-crystalline structure will always result and the detailed morphology will depend on
the cooling rate.
Further cooling of the polymer below Tc results in a further decrease in Vs, which again
follows the coefcient of thermal expansion of the solid polymer, until there is a change in
slope of the plot at the glass-transition temperature, Tg. This is a second-order transition (in
contrast to melting, which is a rst-order transition) since there is a discontinuity in the
second derivative of the free energy with respect to temperature and pressure, i.e.
@ 2 G=@P2 T @V =@PT jV ;
1:16
1:17
Thus there is a step change in heat capacity Cp at the glass transition, which is most
conveniently studied by differential scanning calorimetry (Section 3.2).
It is also seen for the coefcient of thermal expansion,
@=@T@G=@PT P @V =@TP a V :
1:18
As shown in Figure 1.7 there is a decrease in the coefcient of thermal expansion, a, to values
lower than extrapolated from the polymer melt (dotted line). This results in the polymer
having a lower density than would be predicted and there is thus a measurable free volume,
Vf, which has an important bearing on the properties of the amorphous region of the polymer.
This and the detailed analysis of the glass transition are considered after the molecular
requirements for polymer crystallization and the structure of the crystalline region.
Polymer crystallinity
It was noted in Section 1.1.3 that, when one moves to polymers more complex than
polyethylene, the likelihood of the polymer being able to crystallize depends on the
chemical composition, in particular whether the repeat unit has an asymmetric centre. When
it does, then the ability to crystallize rapidly diminishes with the amount of atactic material
in the polymer. Fully atactic polymers will generally be amorphous and the properties of the
glass resulting from the cooling of an atactic polymer from the melt are discussed in the
following section.
As the simplest example of a linear polymer, the crystallinity of polyethylene has
been most widely studied. A consideration of the thermodynamics of melting of a series of
n-alkanes provides the starting point for extension to oligomers of differing chain length. The
equilibrium melting temperature of a perfect polymer crystal composed of innitely long
chains cannot be determined, but it may be approached by extrapolation from calculations for
chains of nite length. This enables the melting point to be calculated and compared with the
experimental values, which reach an asymptotic value above n 300 of 145 C, compared
with the observed value for high-density polyethylene of 138.5 C. It was noted earlier that
crystallization requires undercooling by 20 C compared with Tm, so it is clearly a nonequilibrium process. Nucleation of crystallization is thus important and the process of creating
a crystal analogous to an n-alkane from a melt that consists of highly entangled polymer
14
Repeat Unit
H2C
CH2
c = 2.55
b = 4.95
a = 7.42
Figure 1.8. The orthorhombic crystal structure of polyethylene. Adapted from Strobl (1996).
Melt
SOLID
Melt
Figure 1.9. A schematic diagram of the development of the solid, semi-crystalline polymer by
growth of crystalline lamellae in the direction r from the melt on cooling. Adapted from
Strobl (1996).
chains is impossible. The process is believed to follow a two-step process whereby initially a
partly crystalline phase separates and acts as the nucleation site. X-ray diffraction allows an
orthorhombic unit cell to be associated with the growing polyethylene crystals as shown in
Figure 1.8. The chain is in the all-trans planar zigzag, which, as discussed in Section 1.1.2, is
the lowest-energy conformation of the chain.
The polymer crystal continues to grow by nucleation at the phase boundary as those
sequences of the hydrocarbon chain that are untangled can add to the layer being formed.
This results in a lamellar structure as shown in Figure 1.9 in which chains are attached over
15
a length corresponding to the alkane chain which has formed the original crystallite face. It
is this which determines the crystallite thickness, which is typically only 100 A. Figure 1.9
shows many important feature of the crystal region of the polymer.
The chain, being of fully extended length on the order of micrometres, must pass through
the crystalline layer many times.
The chain will not necessarily enter immediately at the end of the crystal, since
this requires higher-energy conformations, and it may traverse the amorphous region
before re-entry.
Growth in the chain direction, rather than the lateral direction, is suppressed because of
the folds and entanglements of the amorphous region at the upper and lower boundaries
of the crystallite.
Pure single crystals are observed only by slow growth from solution, and these show
crystalline lamellae of layer thickness 100 A and have a higher incidence of tight folding
(Keller, 2000) than is possible from the melt, with re-entry of chains occurring within
three lattice sites of exit. Observation of the melt-cooled, partially-crystalline polymer
under the polarizing microscope shows another level of structure, namely the formation of
spherulites, the sizes of which are sensitive to the thermal history of the cooling melt.
Diameters may range up to hundreds of micrometres and occasionally reach centimetres
for slow cooling of polymers such as poly(ethylene oxide) melts. Spherulites are always
depicted in two dimensions as in Figure 1.10 because of their appearance under polarized
light, but are overlapping spheres formed by aggregations of lamellae each of thickness
100 A as they grow. This means that the chains forming as shown in Figure 1.9 are
aggregating normal to the radius vector r. As shown in the micrograph in Figure 1.10(a)
and schematically in Figure 1.10(b), the spherulites nucleate and grow radially, becoming
distorted as they meet other growing centres. The amorphous material (the entangled
chains in the solid in Figure 1.9) may be seen as that rejected by the growing lamellar
brils and lying between the lamellae in the spherulite. This will be highly entangled, and
there are tie molecules that traverse more than one lamella. These have a major role in the
achievement of the mechanical properties of the solid polymer since they effectively
couple the separate lamellae.
16
(a)
(b)
Tie Molecule
Nucleus
Amorphous
Lamellar fibril
Spherulite
Figure 1.10. (a) The spherulitic habit of a semi-crystalline polymer on cooling from the melt.
(A micrograph of polyethylene glycol viewed under crossed polarizers). (b) A schematic
diagram of lamellar brils that have nucleated from the points shown and grow to the
spherulite boundaries. Tie molecules connecting lamellae are shown.
growth rate is sensitive to temperature and is maximum between Tg and Tm. This results
from a balance between the ease of nucleation at lower temperature (at which the polymer
is more likely to be found in its lowest-energy conformation, which favours crystallization) and the increase in viscosity such that untangling and transport of chains to the
growing crystal face becomes less probable at lower temperature and drops to zero at Tg.
When the melt temperature is approached, chains are lost from the growing crystal surface
faster than they can be attached. It has been proposed that the transition from the
entangled melt to the partially crystalline state occurs through a state of lower order that
forms the primary site, and these blocks are found, by atomic-force microscopy (AFM), to
fuse into homogeneous lamellae (Heck et al., 2000).
Crystallization can also be induced by the chain extension which occurs in an orientational ow. This is of obvious importance in rheology and processing, and it has been
17
found (Somani et al., 2002) that shearing of isotactic polypropylene resulted in oriented
structures that did not relax even after 2 hours at temperatures 13 C above the melt
temperature. These structures were non-crystalline; their evolution was found to follow
the Avrami equation; and they were proposed as the primary nucleation sites. It is known
that in polyethylene the crystallization under extensional ow results in a shish-kebab
structure (Vaughan and Bassett, 1989). This consists of a chain-folded morphology
that has nucleated on a core with an elongated structure due to chains with an extended
all-trans conformation. Use of AFM has permitted this process to be followed at the
molecular level, and it has been found that the rates of lamella growth vary widely, both
spatially and with time (Hobbs et al., 2001).
The thickness of the lamellae which form by secondary growth on the primary lattice
sites may be understood by invoking the thermodynamics and kinetics of crystallization
(Painter and Coleman, 1994). This analysis shows that
the thermodynamics of the system allows the conclusion that the lowering of the melt
temperature of the polymer relative to the equilibrium melting temperature, Tm Tm,
depends inversely on the thickness, l, of the polymer crystal, as given by the rst term in
the equation below;
the crystal thickness is kinetically controlled and represents the thickness that allows the
growing crystal to be stable,
l 2re =DhTmo =Tmo Tm dl;
1:20
where re is the free energy per unit area of the crystal face where growth occurs; Dh is
the enthalpy of fusion and dl is the thickness of the primary crystallite, which is of the
order of 1040 A.
18
V0l
(a)
Glass
V 0g
V0g
Vf
Rubber/Liquid
V
Specific
Volume
V 0l
Temperature
Tg
Poly(styrene)
(b)
Glass
log E
Rubber
Melt
Temperature
Tg
Figure 1.11. (a) The change in specic volume of an amorphous polymer as it is cooled from the melt
through to the glassy state (contrast this with Figure 1.7). (b) Changes in the modulus (E) of
amorphous poly(styrene) as it is heated from low temperature to the melt and the modes of
energy dissipation accompanying the relaxations shown.
the temperature is decreased. At very low temperatures, the only modes available to
polystyrene are rotations about side groups (c-relaxation at 100 C) and the backbone
as a crankshaft (b-relaxation at 0 C). The effect of these motions on the modulus is
modest compared with the change at Tg. Understanding the nature of the molecular
relaxations at Tg requires an appreciation of the theories for the glass transition
(McKenna, 1989). The glass transition is a kinetic effect and shifts with the frequency of
observation. It is also sensitive to the molar mass of the polymer and the presence
of crosslinks between chains. Any theory for the glass transition must accommodate
these results.
In the rheology and processing of polymers the kinetic aspects of the glass transition are
of particular interest since the achievement of thermodynamic equilibrium in the amorphous, high-molar-mass polymer is beyond the time frame of the dynamic environment of
processing. One way of viewing the glass transition (Stachurski, 1987) is to consider the
19
T = T 1 - T0
V 0 (T>Tg)
1: T small (T1 >> Tg)
Ve
2: T large (T1 ~T g)
Ve
Ve
Time
Figure 1.12. The change in free volume with time after applying a temperature drop, DT, to a
polymer that is above, (1), close to, (2), or below, (3), Tg. Adapted from Stachurski (1987).
polymer at some point T0 above Tg in Figure 1.12, where the volume is Ve(T0). From
this point the temperature is suddenly decreased to T1 at an innitely fast rate. At the
molecular level, the changes to the chain can be considered in the framework discussed in
Sections 1.1.1 and 1.1.2 and consist of
a decrease in thermal motion leading to a decrease in volume;
an increase in the intermolecular forces as the chains adopt the conformations of lower
energy; and
a consequent decrease in the end-to-end distance, R0, of the chain, which, due to
entanglements, will be time-dependent.
The changes in conformation require co-operative motion to achieve the bond rotation
needed to attain the gauche- and trans-conformers which are the lower-energy states of the
chain. This will require motion of various segments of the chain and these transitions may
be viewed as having particular relaxation times, si. At a time after the application of the
temperature change that is very much longer than the largest value of si, the system will
again be in equilibrium, so the volume is now Ve(T1).
The behaviour of the systems volume can be considered for the cases of T1 being above,
below and at the glass-transition temperature when T0 Tg.
(1). If T1 Tg and T1 T0 is small, then Ve(T1) will be attained instantaneously
since there will be no change in the available conformations or relaxation times,
so the new equilibrium volume is achieved within the time frame of the temperature drop.
(2). When T1 Tg, i.e. T1 T0 is large, Ve(T1) is attained slowly because adoption of the
new equilibrium conformation for that temperature is controlled by the relaxation
times, si.
(3). When T1 < Tg, i.e. T1 T0 is large, Ve(T1) is attained rapidly but corresponds to a nonequilibrium glass since at T1 < Tg the relaxation times are innitely long compared with
the time frame of the experiment.
20
1:21
where aT is the ratio of the value of the property at temperature, T, to that at the glass
transition, Tg. From this it has been concluded that the value of the fractional free volume
Vf/(V0 Vf) at the glass transition is 0.025.
21
(a)
A
V
X
Y
B
T
(b)
Cp
Tg
Tgapparent
Figure 1.13. (a) A volumetemperature plot of a polymer sample on vitrication and (b) the effect
of cooling history on the heat-capacity (Cp) anomaly observed as an apparent peak by DSC
when the rate of heating exceeds the original rate of cooling for vitrication. Adapted from
Chynoweth (1987).
22
Table 1.2. Glass-transition temperature (Tg) and correlation with polymer structural features
Polymer
Tg (oC)
Structural features
Poly(ethylene)
Poly(propylene)
100
0
Poly(vinyl chloride)
Poly(styrene)
Poly(p-phenylene
terephthalamide)
80
100
345
Flexible CC backbone
Hindered CC backbone due to pendant methyl
groups
Strong dipolar intermolecular forces
Chain stiffening due to pendant phenyl groups
In-chain stiffening from p-phenylene groups together
with amide hydrogen bonding
result in the chains sweeping out a large volume, so Vf is large and Tg is low. Similarly,
if the side groups are exible there will be a higher free volume and thus a lower Tg
compared with a rigid and bulky side group (contrast poly(styrene), Tg 100 C, with
poly(1-butene), Tg 45 C). Another example of the importance of free volume
may be seen in the effect of the number-average molar mass of a homopolymer on Tg.
The following FoxFlory relation holds well for poly(styrene) polymer and oligomers
with B 105:
Tg Tg1 B=Mn :
1:22
This reects the effect of the greater number of chain ends at lower molar mass resulting in
a larger local free volume and thus a lower Tg. While this relation hold well for linear
polymers, there are exceptions linked to the nature of the end groups, e.g. when they are
ionic or hydroxyl groups, and also when cyclics are studied. McKenna (1989) considers in
more detail these and other factors which may affect Tg. The effect of crosslinking, which
is important for reactive processing of both elastomers and three-dimensional networks, is
considered in a later section.
23
Deform
Heat
Figure 1.14. A schematic diagram of a crosslinked elastomer network and the changes in the (circled)
section of the network when it is deformed in the direction shown by the arrows and then
either is allowed to relax or is heated.
1.2
24
Linear
Hyperbranched
Cyclic
Branched
Short
Long
Network
Dendrimer
1
2
Generation Functionality
0
3
1
6
2
12
Figure 1.15. A schematic diagram illustrating the various types of molecular architecture of polymers
1.2.1
Stepwise polymerization
This polymerization method is one of the rst that was established commercially for
polymer fabrication directly by reactive processing, i.e. the phenolformaldehyde
25
Polyesterification
Ester formation is a standard organic reaction between an alcohol and a carboxylic acid,
which is an equilibrium reaction that has been shown to occur under catalysis by either acid
or base. In polyesterication involving an organic acid, the substrate is itself the catalyst. It
has been noted (Pilati, 1989) that, among the many reaction mechanisms, Scheme 1.1
is the most likely for acid-catalysed esterication, with the second reaction being the
rate-determining step.
RCO2H + H3O+
[RC(OH)2 ] + + H2O
[RC(OH) 2 ]+ + ROH
[RC(OH) 2 (ROH)]+
26
The reaction rate will depend on the factors which favour the rate-determining step
moving to the right, i.e. the concentration of the free-ion species from the rst reaction
step. The removal of the water ensures that the reaction favours formation of the ester,
RCO2R0 ; otherwise the reverse reaction becomes signicant.
In the absence of added catalyst, the reaction is formally of third order, since it will be
second order in the acid RCO2H which is both reagent and catalyst:
Rate kRCO2 H2 ROH:
1:23
This may be extended to polyesterication by replacing the alcohol and acid with a diol and
a diacid. Depending on the polarity of the medium, the reaction mechanism may involve
different reactive intermediates, since the formation of charged species will be less probable
in media of low dielectric constant as may occur in the polyesterication at high extents of
reaction. The overall experimental kinetic order is the same as for simple esterication.
On considering the net reaction for esterication, it is seen that the reaction product will
contain a terminal acid group and an alcohol:
HOR0 OH HO2 CRCO2 H $ HOR0 OCORCO2 H H2 O
dimer
This new dimer species has end groups that may undergo three possible reactions, as shown
in Scheme 1.2.
The formation of dimers, trimers and tetramers will consume the reagents early in the
reaction, and the attainment of a high-molar-mass polymer requires the subsequent reactions of
these oligomers, e.g. trimer dimer ! pentamer, or in general terms n-mer m-mer to give
(n m)-mer. The degree of polymerization of the resulting oligomer, DP, is (n m)/2. The
analysis of the kinetics of this system is necessary to determine the molar mass of the polymer
and also the factors that control the rate of polyesterication. This analysis is simplied by the
observation from model compounds that the rate of esterication of a series of acids and
alcohols is independent of the alkyl chain length, n, for both acid and alcohol after n > 3.
1. Reaction with another dimer to give a tetramer with the same end groups:
2HOROCORCO2H
HOROCORCOOROCORCO2H + H2O
tetramer
2. Reaction of the dimer with further diol to give a dialcohol-terminated trimer:
HOROCORCO2H + HOROH
HOROCORCOOROH + H2O
trimer
3. Reaction of further diacid with the dimer to give a dicarboxyl-terminated trimer:
HO2CRCO2H + HOROCORCO2H
HO2 CRCOOROCORCO 2 H + H2O
trimer
Overall:
H(ORO 2 CRCO)nOH + (2n1)H2O
nHO2CRCO2 H + nHOROH
Scheme 1.2. Formation of oligomers during the early stage of polyesterication and the overall
reaction.
27
The molar mass of the resultant polymer may be conveniently determined by the titration
of the acid end groups of the separated polymer. Thus, if C is the number of moles of acid
groups per gram of polymer, then Mn 1/C g/mol (since for a stoichiometric reaction the
polymer would be expected to have an equal number of acid and alcohol end groups, so that
each chain, on average, has one acid end group).
1:24
and, since [RC(OH)2] will be given by the equilibrium value from the rst reaction of
Scheme 1.1,
RCOH2 KRCO2 HH3 O ;
1:25
1:26
Thus, depending on whether the acid catalyst is an added strong acid, or is derived by
ionization of the carboxylic acid reagent, the overall experimental order in the carboxyl may
be second or rst order, respectively.
Detailed studies of systems with no added acid and added strong acid have been performed, and lead to the following kinetic relationships (Manaresi and Munari, 1989).
Self-catalysed polymerization
The concentrations of the starting reagents in step polymerization are usually chosen to be
the same, so that
dR0 OH=dt kRCO2 H2 R0 OH kR0 OH :
3
1:27
It is convenient to introduce the extent of reaction, p, as the fraction of the reagent that has
reacted at a time, t, so that
p R0 OH0 R0 OH=R0 OH0 ;
1:28
1:29
28
This equation has been tested (Manaresi and Munari, 1989) for simple self-catalysed
aliphatic polyester formation and marked deviations have been found at low and high
conversions. This is attributed to the differences in polarity of the medium that may
result in different reaction mechanisms since these are also seen for non-polymerizing
systems. At high conversions there has been the removal of signicant amounts of water
and there has been a change in the reaction volume together with a large increase in
viscosity. The ability for the reaction to proceed to completion depends on precise
stoichiometry since side reactions and loss of volatiles may result in an apparent
decrease in reaction rate.
The molar mass of the polymer is related to the extent of reaction, since this is linked to
the end-group concentration. Then Mn M0/(1 p), where M0 is the molar mass of the
polymer repeat unit. The development of the polymer chain length, the degree of polymerization, DP, is thus
DP 1=1 p 2kR0 OH2 t 11=2 :
1:30
Thus, in an uncatalysed reaction the polymer chain grows with the square root of the time of
reaction, a direct consequence of third-order kinetics. Consequently, in any practical
application, the use of an external catalyst is necessary in order to achieve high-molar-mass
polymer in the shortest possible time.
1:31
k R OH :
Solving this second-order equation gives
1=R0 OH 1=R0 OH0 k 0 t;
1:32
1:33
This shows the advantage of using the external catalyst, since the chain length of the
polymer now increases linearly with time. A detailed experimental study of aliphatic
polyesterication (Manaresi and Munari, 1989) shows that this relation is followed over a
wide range of conversion, but with deviation at short time. This may again be linked to
polarity effects, as for the uncatalysed reaction and simple ester formation.
Equation (1.33) does show that high-molar-mass polymer does not form until high extents
of reaction at long time and shows the importance of the stoichiometry (and thus purity) of
the starting materials, as given in Table 1.3.
Figure 1.16 shows the chain-length distribution of the polymer at increasing extents of
conversion. At short times the polymer has a narrow molar-mass distribution since the
reaction mixture contains only low-molar-mass oligomers. As the reaction proceeds it
29
1p
DP
Relative time
0.9
0.99
0.999
0.1
0.01
0.001
10
100
1000
1
11
111
1.5
p = 0.96
1.0
% Weight
Fraction
p = 0.9875
0.5
p = 0.995
0
0
100
200
300
400
500
DP
Figure 1.16. Changes in the chain-length distribution with conversion, p, for a typical stepwise
1:34
1:35
1:36
For example, if there is 3% deviation in the concentration of, say, the acid, then the limiting
DP is 65. While high values of DP, e.g. over 200, may be required for high-strength
polymers such as PET, for liquid polyesters (as used in unsaturated polyester composite
fabrication where the resin impregnates the reinforcing bre) a value of 1020 is required
and the stoichiometry may be used to control this. In the practical polyesterication reaction, the decision on the reagents is often based on the ease with which the starting material
30
may be puried and the extent to which it will not participate in side reactions at the high
temperature necessary to drive the reaction to completion. Thus, in the important reaction of
formation of poly(ethylene terephthalate) (PET), because of the lack of high-purity terephthalic acid, alcoholysis of the methyl ester was used for many years. This reaction is
related to the ester exchange reaction encountered in reactive processing of other thermoplastics (Lambla, 1992).
31
H2O
H-(OR)-COOH
(OR)COO
However, when the number of methylene groups in R is above seven, then linear polymers are again able to be formed due to the lower probability of cyclization when initially
the ring becomes crowded and then, above 15 carbon atoms, the probability of the ends
meeting tends to zero.
Ring opening of lactones in the presence of a suitable catalyst provides an alternative
route to polyesters. The importance of ring-opening polymerization in reactive processing
requires further discussion of this process for polymers such as polyamides and epoxides in
addition to polyesters.
nHO
C
CH3
CH3
CH3
OH + nCl
Cl
C
CH3
O
OC
* + 2nHCl
32
33
The versatility of urethane chemistry is seen in the wide range of properties that result
from changes to the diol or the di-isocyanate. Thus, urethanes can range from soft
elastomers and bres to tough coatings and rigid foams. In one-shot formulations, the
short-chain diol, polyol, isocyanates, chain extenders and additives (catalysts, plasticizers,
llers) are rapidly mixed and injected into the mould. These formulations are generally
prepared as two separate parts, with the isocyanate being separate from the other components that are then rapidly mixed. The use of multifunctional diols or isocyanates allows
crosslinking in the mould so that reactive processing such as reaction injection moulding
may be employed (Frisch and Klempner, 1989). These are considered later in further
detail.
Reaction with difunctional amines (see the next paragraph) allows the formation of poly
(urethane-co-urea) and further extends the versatility of the segment architecture. For further control of the structure, a prepolymer is formed. The reaction, such as the polyurethane
reaction shown above, is carried out with excess di-isocyanate so that an isocyanateterminated prepolymer is obtained. The isocyanates used are typically aromatic, such as
toluene di-isocyanate. This prepolymer is then reacted with a short-chain diol or diamine
(for a polyurea) to form the nal polymer.
Polyureas are analogues of the polyurethanes but with the diol being replaced by a
diamine (Ryan and Stanford, 1989):
nH2 NRNH2 nOCNRNCO ! HNRNHCONHR0 NHCOn:
While side reactions such as that of the urea with isocyanate are possible, they are generally
limited by the high reaction rate for the isocyanate with the primary amine, provided that
stoichiometric amounts of reagent are used. Polyureas form an industrially important class
of polymers and, while they may be formed by the reaction of amines with carboxy
compounds, the reaction route through the isocyanate has the advantage of there beiny no
evolution of condensation by-products.
The properties of the polyurethane and polyurea homopolymers and copolymers depend
on the chemical structures of the R and R0 groups. Thus, if R is a relatively long-chain
polyester or a similarly exible group then it will form a soft segment in the structure.
Similarly, aromatic groups and other short-chain units will form a hard segment. The
properties depend on the balance of these soft and hard segments as well as the degree of
crosslinking.
(CH2)n
HN
O
+ H2O
H2N(CH2)nCOOH
34
H
C
CH2 + R
NH2
H
C
CH2
NH
OH
H
C
CH2 + R
H
N
H
C
CH2
N
R
OH
Etherification:
R
H
C
CH2 + ROH
O
H
C
CH2
OH
OR
Scheme 1.3. The principal reactions between an epoxy resin and an amine.
This reaction is industrially important since it is used in batch and continuous industrial
processes but, because of the slow reaction rate, is inappropriate for carrying out in an
extruder. (The anionic polymerization has, however, been demonstrated in a twin-screw
extruder, and is discussed later.) The reaction takes place in the presence of 5%10% water
and, to increase the rate of initiation, a small of amount of e-amino caproic acid (the product
of ring-opening of the monomer) is added to provide the starting material without the need
for hydrolytic ring opening, which may be slow. Since the nal DP depends on the equilibrium water concentration, this is driven off at high conversion and the resulting polymer
will contain about 8% residual monomer and 2% cyclic oligomers. The oligomers up to
n 9 are seen , but only in small amounts (Odian, 1991).
Other step-polymerization reactions of large cyclic monomers by entropically driven
ring-opening polymerization have been reviewed (Hall and Hodge, 1999). An important
example is the heating of cyclic bisphenol-A carbonates over a titanium isopropoxide
transesterication catalyst to yield a polycarbonate with a relative molar mass of about
250 000. This has also been performed under conditions of reaction injection moulding
(Hall and Hodge, 1999).
Perhaps the most widely exploited cyclic monomer in reactive processing of composite
materials via a stepwise reaction is the oxirane or epoxy group (Hodd, 1989). Epoxy resins
are principally used to form three-dimensional networks, but linear polymerization is
possible. The main linear polyaddition reactions involve catalysed ring-opening in an ionic
chain reaction. However, it is appropriate to consider the chemistry of the oxirane group in
its reaction with nucleophilic reagents, principally amines, at this point so that the range of
possible reactions may be introduced.
Epoxyamine reactions
The epoxy group is characterized by high reactivity to both nucleophilic and electrophilic
reagents, and reactions with a primary amine are shown in Scheme 1.3. These reactions
35
O
H2C
CH3
HC
H2CO
CH3
H
OCH2 C
C
CH3
CH2O
OH
OCH2
CH
CH2
CH3
DGEBA
O
OCH2
CH
CH2
O
H2C
HC
CH2 O
OCH2
C
H
CH
CH2
TACTIX
O
H2C
O
HC
H2C
CH2
N
HC
CH2
CH2
CH
CH2
CH
N
H
CH2
CH2
TGDDM
Scheme 1.4. Structures of typical difunctional, trifunctional and tetrafunctional epoxy resins.
have been elucidated by the use of model compounds such as phenyl glycidyl ether and
aniline for which R and R0 are the phenyl group (Mijovic and Andjelic, 1995).
The rst reaction results in the formation of a secondary amine and an alcohol. Each of
these may react with another mole of epoxide, so further chain extension may occur. If the
epoxide is difunctional as in the common commercial epoxy resin DGEBA (diglycidyl ether
of bisphenol-A) then a three-dimensional network will rapidly form (Dusek, 1986). This is
discussed later. The viscosity of the resin depends on the extent of oligomerization (n) that
occurs during synthesis and, in the fabrication of bre-reinforced epoxy-resin composites,
this has an important effect on the rheology of the system during bre impregnation and in
the cure kinetics of the system and thus the chemorheology during cure as the network
forms, since the OH groups participate in the chemistry of the cure.
It has been noted that, when dealing with bulky epoxy resins such as DGEBA or the even
bulkier TGDDM, rather than model systems, the rate coefcients for each of the three
reactions are different (St John and George, 1994). Thus the rate of primary amine reaction
is much greater than that for the secondary amine (reaction 2 in Scheme 1.3), so the rst
stage of the reaction is to form a linear polymer by a stepwise addition. It has also been
found that the reaction between primary amine and epoxide will be catalysed by hydroxyl
groups either present on the backbone of the polymer or present as impurities in the
commercial resin (St John et al., 1993). The effect of the OH is believed to arise by virtue
of hydrogen bonding to the oxirane as shown in Scheme 1.5.
There are several hydrogen-bonding congurations that may be drawn that also involve
another molecule of amine, so the system may exhibit self-catalysis. A detailed study of the
36
CH2
CH
O
H
NH2
H
R
CH2
CH
O
H
O
CH
CH2
H
H
O
R
CH
CH2
NH
R + R
OH
OH
Scheme 1.5. Alcohol catalysis of the epoxy-ring-opening reaction by amine groups (the OH may
arise from impurities in the resin, the resin itself, e.g. DGEBA, Scheme 1.4, or following
ring-opening of another epoxy group, Scheme 1.3). Reproduced with permission from
St John (1993).
kinetics of model epoxyamine systems has shown that self-catalysis is important, and it is
often difcult to separate this from impurity effects in actual epoxy systems (Kozielski
et al., 1994). The etherication reaction (i.e. epoxyhydroxyl reaction) has a high activation
energy and is important at high temperatures.
It is also important to recognise that the structure of the amine nucleophile will have a
major effect on the reactivity of the system. Thus aromatic amines such as 4,40 -diaminodiphenyl sulfone (DDS), shown in Scheme 1.6, will be much weaker bases and less reactive
than aliphatic amines, so high temperatures are required for reaction to occur in the absence
of added catalysts.
In many catalysed systems the reaction mechanism will be a competition between
stepwise reactions and addition, so the resulting network may be very complex. This is
addressed in more detail when considering the chemorheological development of the threedimensional network.
1.2.2
Cyclization
Cyclization may occur in a number of types of polymerization reaction where there is a
nite probability that the reactive chain end may react with the other end of the same chain
rather than with a different molecule. The result is a polymer with no chain ends and with a
set of conformation-dependent properties different from those of the linear precursor. For
37
O
H2N
NH2
DDS
H
H2N
NH2
DDM
Scheme 1.6. Aromatic amine curing agents: DDS, 4,40 -diamino diphenyl sulfone; and DDM,
example, a cyclic polymer with the same molar mass as a linear polymer will have one half
the radius of gyration (Rg) (Richards, 1989).
As discussed earlier regarding polyesterication, simple condensation polymerization
reactions operate by reactions of the type AA BB ! ABAB and, as this continues, there
will be successive formation of dimers, tetramers etc. as the chain builds in a stepwise
fashion. The rule governing these reactions is that the end group A may react only with end
group B and, while at low extents of reaction the probability of an A group and a B group
from the same chain reacting will be low, this probability will increase as the chain length
increases. This will be governed in part by the thermodynamic stability of the resulting ring.
Thus it is well known that for a simple cycle of methylene groups (CH2)n there is a
minimum in ring strain energy at n 6, after which it increases again, until beyond n 14 it
is vanishingly small. The components that make up this strain are angle strain and conformational strain (the most important forms of these being torsional strain, due to eclipsed
conformations, and transannular strain, due to ring crowding). Similar concepts apply for
other cyclic systems such as polydimethyl siloxanes (PDMSs), (Si(CH3)2O)n and it has
been reported that, at n 22, the ring is planar (Richards, 1989). Siloxanes are of interest
also since they are synthesized by the ring-opening reaction of the cyclic precursor with
n 4, commonly known as D4, and the polymerization is an equilibrium reaction between
the linear and cyclic forms.
In addition to the above thermodynamic conditions for ring formation, the kinetics of the
reactions must be considered. Thus, for a reaction to take place the two ends of the polymer
chain must be in the correct conformation for sufcient time for the new bond to form. The
kinetic factor for cyclization is proportional to Rg5/2, so the net effect of the thermodynamic
and kinetic factors is that rings are not favoured between n 8 and n 11. Suter (1989) has
considered the theoretical approaches of Jacobsen and Stockmayer and compared theoretical
and experimental values for macrocyclization equilibrium constants. This has also been
performed for Monte Carlo as well as rotational-isomeric-state calculations for the statistical
conformations of cyclic esters (decamethylene fumarates and maleates) and agreement with
experimental molar cyclization equilibrium constants found (Heath et al., 2000).
38
Step copolymers
The earlier example of the formation of a poly(urethane-co-urea) showed the change in
properties made possible by including a comonomer (in that case a difunctional amine)
together with the usual diol for reaction with a di-isocyanate. This can be extended to a wide
range of step polymerizations where an additional reactant is added. Examples could be the
use of two AB-type monomers (e.g. amino acids) or two AA (e.g. diacids) to react with one
BB (a diamine) to form co-polyamides. Several features of step polymerization help in
understanding the resultant copolymer. For example, since high-molar-mass polymer is
formed only late in the reaction, the composition of the copolymer will be that of the feed
ratio of the monomers.
39
>60oC
O
+ HOCH2CH2OH
HOCCH CHCOCH2CH2OH
O
>160oC
n HOCCH
O
CHCOCH2CH2OH
CCH
HO
CHCOCH2CH2O
H
n
+ n H2O
Scheme 1.7. The reaction scheme for maleic anhydride condensation with ethylene glycol, showing
fumarate group formation in unsaturated polyester resin. Adapted from Hepburn (1982).
CH
COOH
CH
COOH
Maleic acid
COOH
H+
CH
CH
COOH
Fumaric acid
Scheme 1.9. Chain branching in unsaturated polyester formation due to addition of diol to a double
bond.
40
Polyester or Polyether
HO
OH
POLYOL
+
OCN
Aromatic
OCN
DIISOCYANATE
NCO
Aromatic
NCO
URETHANE PRE-POLYMER
O
OCN
Aromatic
H
N
O
Polyester or Polyether
DIOL
POLYURETHANE
H
N
Aromatic
NCO
DIAMINE
POLYURETHANE UREA
Scheme 1.10. Formation of urethane prepolymer and subsequent reaction with a diol or diamine
extender to give a polyurethane or a polyurethanepolyurea. Adapted from
Hepburn (1982).
The reaction of the double bond to form crosslinked networks and the chemorheology of
these systems are considered later. Different acid and diol starting materials result in
polyester resins with superior properties. For example, the use of iso-phthalic acid instead of
phthalic anhydride (that produces ortho-phthalate units) results in improved water resistance
of the resins, but at increased cost.
Block copolymers
The formation of step copolymers that are forming blocks is best illustrated through
considering polyurethanes that are formed by prepolymers. This was discussed earlier in
outline for the polyurethane co-polyureas. In the following example, a polyester-urethane
is formed by reacting an OH-terminated polyester with an NCO-terminated polyurethane. Each of these is formed by the appropriate stepwise polymerization and then
reacted to give a larger NCO-terminated prepolymer:
HOpolyesterOH NCOpolyurethaneOCN
! OCNfpolyurethaneNHCOOpolyesterOCONHpolyurethanegNCO:
n
This prepolymer (Scheme 1.10) is then reacted with a chain-extender diol or diamine
to give the nal polyurethane or polyurethanepolyurea block copolymer, respectively.
The diversity in properties that the nal urethane block copolymers exhibit is governed by
the nature of the three building blocks: the diol-terminated polyester, the di-isocyanateterminated polyurethane and the chain extender (Hepburn, 1982). As shown in Figure 1.17,
these form exible blocks (from the polyester) that undergo phase segregation as structures of
size 10002000 nm.
41
SOFT
SOFT
HARD
HARD
HARD
HARD
HARD
HARD
SOFT
SOFT
> 1000 nm
150 nm
42
(b)
(a)
BABABAB
BABABAB
B
A
B
A
B
A
B
B
B
B
B BB
B
B
B
B
B
B B
B
B
Figure 1.18. Branching during step polymerization. (a) Addition of a trifunctional agent (e.g. a triol,
BBB) to a self-condensing difunctional system (e.g. a hydroxy acid, BA). (b) Hyperbranching
from a system AB2. In both cases A may react only with B.
1:37
i.e. it should approach 1.33 for the system discussed. This is to be contrasted with the
system without branching, where Mw/Mn approaches 2 as the reaction reaches completion
(Equation (1.35)).
Crosslinking will occur when a multifunctional reagent, Bf, that can react with the second
group is added, since then the branches may connect the chains. Alternatively, it can occur
when either Af or Bf is added to a difunctional system AA plus BB. The importance of this
in network formation and chemorheology is considered in more detail in Section 1.2.4. At
this stage it is appropriate to note the conditions under which gelation may occur in order to
contrast network formation with hyperbranching.
The condition for gelation was demonstrated by Flory (1953) to occur, for a system
consisting of a triol (B3) added to a difunctional monomer (A2), at the critical conversion,
ac, of 0.5 (i.e. ac 1/( f 1), where f is 3 for the triol).
43
HO
OH
HO
HO
HO
O
2-ethyl-2-(hydroxymethyl)1,3-propanediol
OH O
PTSA/140oC
O
OH
HO
OH
O
CORE
BRANCH
O
HO
OH
O
HO
OH
2,2-bis(hydroxymethyl)propionic acid
OH
Scheme 1.11. Early stages of the synthesis of a hyperbranched polyester. After Jikei and
Kakimoto (2001).
Since the conversion, a, at any time is related to the conversion of A2, pA, for a mixture
with stoichiometric ratio (r [B3]/[A2]) by the relation
a p2A =r;
1:38
a calculation of the extent of conversion for the stoichiometric amounts of the polyol and
polyurethane (r 3/2) considered before shows that this will occur at conversions of the
isocyanate groups of 0.87. Thus gelation will result before full conversion of isocyanate.
In the special case of hyperbranched polymers, as shown in Figure 1.18(b), the branching
is controlled so that novel molecular architectures are achieved without gelation. If, instead
of a system consisting of difunctional and trifunctional reagents, A2 and B3, a single reactant
of structure AB2 is employed, the conversion with time is given by
a pA = f 1 ac :
1:39
44
B
B
B
S
S
B
S
G=3
S
B
C
S
B
S
S
B
B
B
B
S
S
S
S
S
S
1: DENDRIMER (Generation 3)
S
S
S
S
3: CORE
S
S
B
F
2: DENDRON
Figure 1.19. Schematic of a Dendrimer with generations (G) indicated and the building blocks B.
At generation 3 this dendrimer has 32 active sites (S), which may continue reaction to
generation 4. Nomenclature is explained in the text. After Frechet and Hawker (1996).
groups (labelled S) on the total dendrimer of generation 4 will be 4 24 64. The number of
building blocks B is 4 (23 22 21 20) 60. Examination of the dendrimer structure
shown in Figure 1.19 reveals the unique feature of the large number of end groups compared
with a linear polymer of the same molar mass. The globular structure is also achieved
without the chain entanglements that would be required of a linear polymer. These features
result in unusual rheological and reaction properties that are still being elucidated.
The feature that distinguishes dendrimers from hyperbranched polymers (Figure 1.18(b))
is the much more well-controlled architecture in which there are no linear components in the
structure. Two broad strategies have been employed to achieve dendrimers, namely
divergent and convergent growth, which differ in terms of the direction that the synthesis
follows with respect to the core (Frechet and Hawker, 1996). In both cases it is important
that the sample be puried after each generation to remove any by-products of side reactions
(e.g. cyclization) that may affect the growth of the next generation. The divergent approach
has allowed very large amino-functionalized dendrimers based on a trifunctional core and
up to generation 10 to be prepared in commercial quantities, such as poly(amidoamine)
45
(PAMAM) or Starburst dendrimer. This has 3 210 3072 end groups. It has been noted
(Frechet and Hawker, 1996) that the difculty in achieving a perfect structure increases with
each generation. The convergent approach which builds individual dendrons commencing
from the reactive sites, S, and proceeding inwards has fewer unwanted side reactions.
However, it involves the challenge of coupling the large and bulky dendron fragments to the
core, C, before the dendrimer structure is achieved.
The divergent approach has been shown (Bosman et al., 1999) in examples of a poly
(propylene imine) dendrimer to produce only 23% dendritic purity by the fth generation, which has 64 end groups. This is a simple consequence of the statistics of the
process used, since incomplete reaction of amino groups or intramolecular cyclization
between arms of the growing dendrimer will produce occasional defect groups. Since it
requires 248 reactions to reach generation 5 commencing from a tetrafunctional core, an
incidence of defects per step of only 0.5% will result in the above level of dendritic
purity. If it is possible to purify each generation (as happens in the convergent
approach) then low levels of impurities may be achieved and polymers can appear with
a single mass, as demonstrated by MALDI-MS (Bosman et al., 1999). The actual
conformation that the dendrimer may adopt, the hydrodynamic radius and the molarmass dependence of the viscosity are discussed later when these systems are compared
with linear polymers. The main point noted here is that, for the same molar mass,
dendrimers differ from linear polymers by having low entanglements and a large
dependence of properties on the end groups. This is most marked in the glass-transition
temperature (Tg) (Bosman et al., 1999).
Hyperbranched polymers are synthesized by a one-pot strategy, in contrast to the
sequential approach for dendrimers, and are characterized by high levels of impurity
structures, due to the large number of possible congurations that may arise from the
statistics of reaction of an AB2 system (Kim, 1998). Also the polydispersities of hyperbranched polymers are very large since, unless special care is taken, the larger molecules
grow at the expense of the smaller molecules since they have a large number of available
reactive sites and statistically are more likely to react. The degree of branching, DB, is a
convenient measure in terms of which to differentiate hyperbranched polymers. An ideal
degree of branching is that seen for the pure dendron, where every reaction point is a
branch, i.e. DB 1. This obviously reduces for each reaction site that results in a linear unit
L, rather than a dendritic unit, D. There will be terminal units, T, both for dendrons and for
hyperbranched polymers, and the denition is
DB ND NT =ND NT NL ;
1:40
where ND is the number of dendritic units, and so on. The other extreme is the linear
polymer, where DB 0, so the hyperbranched systems DB will always fall between these
extremes and, for a one-pot synthesis of AB2, will statistically approach 0.5. Ways of
achieving a higher degree of branching (Jikei and Kakimoto, 2001) are
use preformed dendron fragments, e.g. by pre-reacting three AB2 units to give AB4; the
theoretical DB then becomes 0.75;
polymerize in the presence of a core unit (if a core unit of functionality B3 is used then a
DB of 0.8 is possible);
enhance the further reactivity of any linear units that are formed
These techniques will also result in a decrease in the polydispersity of the system.
46
NH2
H2C
MeOOC
H
HC
H
N
H
CH2
Me
CH2
H2 H2
C C
HN
NH2CH2CH2NH2
CH2
MeOOC
H2
C
CH2
H2 C
H2N
COOMe
H
N
C
H2
CH2
C
CH2
H2 H 2
C C
CH2
N
CH2
MeOOC
CH2 H2C
H2C
CH2
H2C
HN
MeOOC
CH2
H 2C
N
H 2C
O
H2
C
HN
H2C
HC
CH2
Me
CH2
H2C
C
CH2
NH2
H2N
NH2
H2N
CH2
NH2
HN
CH2
H2 H 2
C C
CH2
CH2
MeOOC
H2 C
COOMe
H2C
C
N
O
H2N
N
NH
CH2
H2C
H2
H2
N
C
C
COOMe
MeOOC C
C
H2
H2
H2N
NH2
N
H2N
NH2
H2N
N
NH2
NH2
PAMAM Dendrimer
47
HO
R
K 2CO3
+
Br
OH
OH
R
O
OH
PPh 3 (Activation)
CBr4
HO
K 2 CO3
(Coupling)
OH
OH
R
O
R
Repeat
O
O
DENDRON
(Activation/
Coupling)
OH
R
Figure 1.21. Synthesis of a dendron, precursor to a dendrimer, by a convergent route. Adapted from
Frechet and Hawker (1996)
Convergent synthesis has been illustrated through the formation of polyethers based on
3,5-dihydroxybenzyl alcohol built from the surface functional group by reaction with benzyl
bromide (Figure 1.21). Alkylation occurs at the two phenyl OH groups so that the
hydroxymethyl group is left available for the next step, which is regeneration of a benzyl
bromide group that is able to repeat the reaction with 3,5-dihydroxybenzyl alcohol. This
sequence was chosen because it proceeds to high yield under mild conditions and has no
signicant side reactions. This two-step procedure may then be repeated to build dendrons
of the desired generation prior to coupling to a core molecule. This is shown for a trifunctional core of 1,1,1-tris(40 -hydroxyphenyl)ethane. Because of the purication procedure
at each step, the most structurally pure dendrimers are produced by the convergent approach,
but the number of generations achievable will be limited by steric considerations in the nal
coupling reaction to the core. Results from MALDI-MS conrm that the convergent dendrimers are monodisperse, with the theoretical and experimental molar masses in agreement.
48
group A is added to B, the reactivity of the second B group will be reduced for steric
reasons. One of the commercially available hyperbranched polymers used in applications
such as thermoset toughening is an aliphatic polyester (Boogh et al., 1999), and this is an
appropriate example to choose to illustrate the polymerization reactions.
Polycondensation of 2,2-bis(hydroxymethyl)propionic acid has been carried out in
the melt at 140 C using p-toluenesulfonic acid (PTSA) as the catalyst. A core molecule,
2-ethyl-2-(hydroxymethyl)-1,3-propanediol was used, as shown below, to achieve DB
values in the range 0.830.96 (Jikei and Kakimoto, 2001). The early stage of the reaction is
shown in Scheme 1.11, where the core molecule has all OH groups reacted and one branch
has extended by reaction of the two OH groups.
Self-polycondensation would be expected to compete with the reaction with the core, but
it has been found that it is possible to drive the reaction to achieve the high degree of
branching required. The resulting polymer has predominantly OH terminal groups.
Similar considerations apply for the aromatic polyesters such as those prepared from
3,5-dihydroxy benzoic acid or the trimethylsililated adduct (Jikei and Kakimoto, 2001). In the
use of the alternative monomer 3,5-bis(trimethylsiloxy)benzoyl chloride the monomer was
required to be of high purity in order to avoid the occurrence of side reactions and formation
of an insoluble product. The terminal hydroxy groups are regenerated by hydrolysis of
the trimethylsiloxy (TMS) groups in methanol on completion of the reaction and the DB was
0.550.60. It is noted that the Tg value for the aromatic ester was 197 C, compared with 40 C for
the aliphatic polyester, although both have thermal stabilities of over 300 C. If the aromatic OH
groups were esteried with adipic acid, Tg fell to 6 C, showing that the properties of hyperbranched polymers, like those of dendrimers, are very dependent on the end-group composition.
Phenolic resins
Phenolic resins are generally prepared commercially by the crosslinking reaction of prepolymers. The crosslinking is normally produced by the elimination of water at high
49
OH
OH
+
phenol
H2C
OH
H2
C OH
OH
formaldehyde
CH2
H2C
HO
H2C
H2
C
OH
H2
C
HO
OH
OH
H2
C
OH
H2C
H2
C
OH
HO
HO
CH2
CH2
HO
H2
C
OH
trimethylol phenol
temperature (>160 C), but can also be made to occur at lower temperature (60 C) by the
use of strong acid catalysts. In this case the water phase separates and remains in the resin.
There are many forms of phenolic resins, but the general categories are resoles and
novolacs. The chemistry of resole formation is illustrated here (Knop et al., 1989).
Commencing from phenol and formaldehyde, the products of base-catalysed condensation reaction at T < 60 C are a series of mono-, di- and tri-methylol-substituted phenols
(Scheme 1.12).
The reaction mechanism probably involves the formation in base of phenolate anion
(PhO), followed by the electrophilic attack of the formaldehyde preferentially in the ortho
and para positions. Depending on the molar ratio of formaldehyde to phenol, the further
methylolation reactions may occur. This reaction is of second order, i.e.
Rate kPhO CH2 O:
1:41
Resole prepolymer formation is then made to occur by heating above 100 C to evolve
water. Above 130150 C a methylene bridge (to give a novolac by the evolution of
formaldehyde, as shown in Scheme 1.13) is formed, rather than an ether link of the
resole.
The prepolymers are soluble in common solvents (e.g. ethanol), to give varnishes, or may
be emulsied to impregnate cellulosic reinforcement, to give a composite. The crosslinking
50
OH
H2
C
H2
C
HO
H2
C
OH HO
OH
H2
C OH
>130 oC
HO
CH2
H2
C
>100 C
OH
H2
C
HO
HO
H2
C
H2
C
CH2
OH
OH
HO
H2
C
OH
HO
CH2
CH2
HO
H2
C OH
CH2
Novolac
+ H2O
HO
OH
H2
C
+
HO
CH2
H2C
Resole
Scheme 1.13. The reaction scheme for resole and novolac formation.
OH
H2
C OH
O
CH2
-H2O
quinone methide
Scheme 1.14. The reaction scheme for quinone methide formation.
reaction at 200 C drives off water and results in a network with very high crosslink density
and Tg, so without reinforcement the fully cured resins are brittle.
A particular feature of the high-temperature-crosslinked resin is the darkening due to the
formation of quinone methide (Scheme 1.14). Since this reaction is important only above
160 C, by using acid catalysts at T < 80 C darkening can be avoided. In principle the acidcatalysed reaction will occur at room temperature, but for the best properties a post-cure
reaction up to 80 C is required.
Cyclic products from the reaction of p-substituted phenols with formaldehyde may be
formed during the alkaline condensation reaction. These calix(4)arenes contain both
methylene and dimethylene ether links and form even-numbered cycles ranging from n 4
to n 8 (Knop et al., 1989).
While novolac links may form at elevated temperature during resole synthesis
(Scheme 1.12), the preferred synthesis of novolac resins takes place under acid conditions
(pH from 1 to 4) and with an excess of phenol. In this case the methylol intermediates
cannot be isolated since they react rapidly to give methyene-bridged structures with relative
molar mass 2000, as shown in Scheme 1.15.
The absence of the methylol groups means that the systems cannot crosslink on heating
but require an added curing agent such as hexamethylene tetramine.
51
OH
OH
H2
C OH
CH2
+ H+
OH
+ H2O
OH
OH
H2
C
CH2
+ H+
+
OH
Novolac
Scheme 1.15. The reaction scheme for formation of novolac.
NH2
NH2
O
NH2
H2N
Urea
H
N
+ H2C
NH2
NH2
Melamine
NH2
O
C
NH2
C OH
H2
H2C
H
N
O
O
C
N
H
C OH
H2
C OH
H2
52
(1)
(2)
(3)
(4)
where R = -CH2NHCO-.
Scheme 1.17. Possible crosslinking reactions of ureaformaldehyde resins.
Self-condensation of the methylol groups to give a dimethylene ether linkage (reaction (2))
may, at high temperatures, also lose formaldehyde to generate the methylene bridge
(reaction (3)). Tertiary-amine sites may also be produced by the reaction of a methylol
group with the amido nitrogen (reaction (4)).
The methylol group may then undergo further crosslinking reactions. While the completion of crosslinking should in principle give methylene linkages and tertiary-amine sites,
there will also be dimethylene ether linkages and free methylol groups as well as imino
groups from incomplete reaction with formaldehyde.
The reactions of melamine (the structure of which is shown in Scheme 1.16) are similar to
those of urea except that there are three amino groups per molecule, so it is hexafunctional
with respect to formaldehyde and it is possible to separate the different reaction products
with formaldehyde up to hexamethylol melamine. The crosslinking reactions are identical to
those of ureaformaldehyde resins.
The lower curing temperature of these resins (100 C under acidic conditions) avoids the
discolouration of phenolics but there is still the evolution of water (and formaldehyde) on
crosslinking that requires special procedures during autoclave curing of composites containing these resins. These problems do not arise on crosslinking of epoxy resins, in which
there is no small molecule evolved, but this is still a step-growth polymerization with the
appropriate choice of curing agent.
53
CH3
Bisphenol A
CH3
HO
OH
OH
CH3
CH3
O
H
C
+ 2 H2C
+ H2O
CH3
H2
C Cl
Cl
H2 H
C C
CH3
CH3
O
O
CH3
OH
H
C
H2
C Cl
OH
Epichlorohydrin
DGEBA
H
C
H2C
H2
C
CH3
O
CH3
O
O
CH3
H2
C Cl
O
H2CO
C
CH3
CH3
CH3
HC
H
C
H2C
CH2
O
H2C
+ 2 HCl
CH3
O
CH3
H2 H
C C
CH2
O
CH3
Cl
H2 H
C C
H
OCH2 C
CH2O
OH
OCH2
CH
CH2
n CH3
EPOXY RESIN
It is noted that the value of the degree of polymerization in the nal product may range
from n 0.15, giving a molar mass of 380 g/mol (as in the commercial resins Shell Epon
828 or Ciba Araldite 6005 and Dow DER 331 (Bauer)), up to n 13, giving a molar mass
of 3750 g/mol in Shell Epon 1009. The actual DGEBA resin as shown (n 0) is not
commercially used. The viscosity also increases with the value of n, with the resin
becoming semi-solid at n 1 and the Tg progressively increasing from 40 to 90 C.
Normally the resins are difunctional, but, if ring closure of the chlorohydrin group, CH
(OH)CH2Cl, is incomplete, then the group will remain as an impurity and the resin will
have a small amount of a monofunctional epoxy. The value of n depends on the extent of
reaction of the formed epoxy resin (with n 0) with more epichlorohydrin to give the
oligomers n 1, 2, . . . and thus is linked to the original stoichiometry. Chromatographic
(HPLC and SEC) analysis of the commercial Shell Epon 828 shows that the majority of
the resin (83%) is n 0, with the amounts of other oligomers decreasing to be vanishingly
small by n 3. The total difunctional resin content is about 94%. As n increases the molar
mass between crosslinks will also increase, so the resulting networks will have a lower
crosslink density (as discussed in more detail later).
Other high-performance epoxy resins have been synthesized with the goal of achieving
higher-temperature performance. Initially epoxy novolacs, which have a more rigid aromatic backbone and a higher epoxide functionality, were prepared.
More recently, trifunctional and tetrafunctional resins (Tactix 742, namely the triglycidyl
ether of tris(hydroxyphenyl) methane, and Shell 1153, namely a diaryl-methylene-bridged
DGEBA) have been trialled, but the alternative approach of curing glycidyl amine resins
with aromatic amines has proven to be more widely employed to achieve high Tg. These
resins were shown earlier in Scheme 1.4.
54
NH2(CH2)2NH(CH2)2NH2
NH2(CH2)2NH(CH2)2NH(CH2)2NH2
CH3
NH2
Isophorone diamine (IPD)
H3C
H3C
C NH2
H2
H2
C
H2N
NH2
NH2
m-phenylene diamine (mPDA)
NH2
H2N
H2
C
NH2
Scheme 1.19. Selected aliphatic, cycloaliphatic and aromatic curing agents for epoxy resins.
The reaction mechanism of DGEBA with a range of aliphatic amines and polyamines has
widely been studied by thermal analysis as well as chemical analysis of the extent of
reaction. Some of the many aliphatic and aromatic difunctional and polyfunctional amines
that have been employed are shown in Scheme 1.19 (Pascault et al., 2002).
Instrumental analytical methods including HPLC, NMR and FT-IR have enabled the
course of the reaction to be delineated by analysing the sol and gel fractions over time. In
Section 1.2.1 the individual amineepoxy reactions were presented, since the rst stage of
the reaction with a primary amine involves chain extension. This reaction competes with
crosslinking since the reaction of the primary amine with epoxide is much faster than the
reaction of the secondary amine. It is the latter reaction that results in branching of the chain
and thus the formation of the rst crosslinks.
These reactions are of importance in the chemorheology of the network formation since it
is possible to link the chemical changes with the viscosity changes that signal the development of an innite network. The statistical analysis of the network and the relation to
rheological properties are discussed in detail in Chapters 2 and 3. At this stage it is useful to
consider the chemistry of the formation of an innite network and the role that the changes
in the viscosity of the system play in the chemistry. There is a feedback process between the
physical changes occurring to the system and the rate of the reactions as they become
diffusion-controlled.
As an example of the crosslinking reaction of multifunctional epoxy resins with amines
through a step-growth polymerization, we will examine the details of the reaction of
TGDDM (N,N,N0 ,N0 -tetraglycidyl-4,40 -diamino diphenyl methane) with aromatic amines
(St John and George, 1994). The kinetic approach developed has also been employed for
glycidyl ether resins (Kozielski et al., 1994). The main reason for the development of the
55
current level of understanding of these reactions is the wide use of the resins in aerospace
systems and the importance of ensuring reliable cure to achieve the desired balance of
thermal and mechanical properties.
The understanding of the reaction mechanisms and structures of likely products of the
chain extension and crosslinking reactions of TGDDM with DDS has been based initially on
the reaction of model compounds such as diglycidyl aniline (representing the epoxy reactive
sites) and various secondary amines representing the structure formed after one mole of
epoxide has reacted. The aim has been to elucidate the feature that distinguishes TGDDM
systems from DGEBA systems: the steric crowding about the epoxy groups versus the likely
catalytic effects of hydroxyl groups adjacent to the second epoxide group after the rst has
reacted (Matejka and Dusek, 1989, Matejka et al., 1991). St John (1993) has summarized
the results from these and other model-compound studies regarding the features of the
reaction as follows.
1.. Reaction of a primary amine with epoxide is catalysed by species that may undergo
hydrogen bonding to it, including primary amines and the hydroxyl groups produced
following other epoxy-ring-opening events.
2.. Reactions of secondary amines with epoxide occur in a similar way to those of primary
amines, though often at a signicantly slower rate due to steric hindrance.
3.. Etherication (hydroxyl reaction with epoxide) has a high activation energy, so it occurs
only at elevated temperatures and is catalysed by teriary amines.
4.. N-glycidylanilino compounds, on reaction with aromatic amines, give rise to several
cyclic structures, including an eight-membered ring, through an intramolecular
secondary-amineepoxy addition reaction, morpholine and seven-membered ether rings
as well as a 1,2,3,4-tetrahydro-3-hydroxyquinoline ring.
5.. N-diglycidylanilino compounds, after one epoxy group has reacted, exhibit enhanced
reactivity of the remaining epoxy group, presumably through catalysis by the hydroxyl
group formed in the rst reaction.
6.. In N-glycidylanilino compounds there occurs little chain growth due to etherication or
an anionic growth mechanism, especially when hydroxyl groups are present.
When these results are applied to the TGDDM/DDS system the likely chain extension,
crosslinking and competing reactions (e.g. cyclization from the intramolecular reaction) are
those shown in Scheme 1.20 (St John, 1993).
These reactions assume a TGDDM molecule of perfect composition. A detailed
analysis of the synthesis of TGDDM shows that a range of impurities can result from
incomplete reaction, of the type shown earlier for the synthesis of DGEBA but exacerbated by the steric crowding once a single epoxide ring has closed (St John and George,
1994). Many of these have been shown to have an important bearing on the shelf life of
the epoxy resin (Podzimek et al., 1992). The resin in its pure form (>90% TGDDM) is
very stable and should remain constant for over one year at room temperature. However
the impurities B, C, D and E in Scheme 1.21 have been determined to play a role in
reducing the shelf life; all impurities reduce the functionality of the resin and lower the
crosslink density.
The effect that these resin impurities have on the chemorheology is pronounced insofar as
it has been shown that the gel time of the resin is extremely sensitive to the impurities,
particularly the hydroxyl-containing species (St John et al., 1993). This is discussed in more
detail later.
56
CH2
CH
CH2
N
CH2
R
OH
O
CH2
CH
CH2
S
O
CH2
N
H
CH2
H
+
CH
CH2
R
CH2
OH
O
CH2
CH
CH2
CH2
N
R
Intramolecular:
OH
O
S
O
CH2
CH
CH2
N
H
CH2
CH2
CH2
CH
O
OH
O
S
O
CH2
CH
CH2
CH
CH2
CH2
CH2
OH
Scheme 1.20. The possible reaction of TGDDM with DDS, showing the possible intramolecular
reaction leading to cyclization instead of crosslinking. Reproduced with permission from
St John (1993).
57
O
H2C
O
HC
CH2
N
H2C
HC
CH2
CH2
H
O
H2C
O
HC
CH2
N
H2C
CH2
CH
HC
CH2
CH2
CH
CH2
CH2
CH
CH2
OH
Cl
N
H
B
O
H2C
H2C Cl
HC
CH2
N
H2C
HC
CH2
CH2
CH
CH2
CH CH2
OH
O
H2C
O
HC
CH2
N
H2C
HC
CH2
CH
CH
CH2
CH2
Cl
OH
O
HC
CH2
N
H2C
CH2
CH2
N
H2C
HC
CH2
CH2
CH
CH2
N
H
OH
Scheme 1.21. Possible impurities in commercial TGDDM resins. Reproduced with permission
from St John (1993).
58
together with thermal analysis (Section 3.2.2) have allowed a quantitative kinetic model to
be developed. For the application of this in real-time process control, bre-optic techniques
(Section 3.4.2) as well as a suitable processing model (Chapter 6) are required.
In kinetic models of the development of rheology during the step-growth crosslinking of an
epoxy resin with a nucleophile (e.g. an aromatic primary amine) the equations derived must
be not only empirically useful for predicting reaction behaviour but also mechanistically
sensible. In many empirical equations derived, for example, from DSC results, the mechanistic information is necessarily limited and certain assumptions must be made (e.g. the relative
rate coefcients of reactions of primary and secondary amines) (Cole et al., 1991). This results
in a set of equations that t the thermal-analysis data and enable reactive-group proles to be
constructed. However, these proles predict total consumption of secondary amine when, in
fact, residual groups are present on completion of cure. This is important because it affects the
subsequent moisture uptake of the resin (since water can undergo hydrogen bonding with the
secondary-amine groups). Similarly (but more seriously), in the development of a thermochemical model for the cure of a commercial epoxyamine formulation under autoclave
conditions (Ciriscioli and Springer, 1990) an empirical set of equations was developed from
DSC results. These included a negative pre-exponential factor for part of the conversion, but
no comment was made regarding the physical meaning of this result.
In the following kinetic analysis of the cure of TGDDM with DDS at elevated temperatures,
the equations are developed from both DSC and NIR data (in some circumstances collected
simultaneously (de Bakker et al., 1993)), and enable quantitative estimation of the reactivegroup prole at any time during isothermal (and in some cases temperature-ramped) cure.
1. Primary-amine reaction
dPA=dt k1 PA2 EP k2 E0 E1 OH k3 PAE1 OH:
1:42
The rst term represents the self-catalysis of the reaction by primary amine that may
undergo hydrogen bonding with the epoxide. The presence of the [OH] term recognizes
both the auto-catalytic nature of the reaction, since OH groups are produced by epoxyring-opening, and the catalytic effect that pre-existing hydroxyl-group-containing impurities
have on the reaction rate. (The chemical structures of these were shown at the end of the
previous section.) This rate equation also recognizes that the reaction rate of an epoxy group
(E1) after an adjacent epoxide has reacted differs from that before one has reacted (E0)
(Scheme 1.22). This requires a statistical approach in determining the relative concentration
of the two species (St John, 1993).
E0
N
OH
O
CH2
CH
CH2
CH2
CH
CH2
CH2
CH
CH2
N
CH2
CH
CH2
E0
E1
Scheme 1.22. The naming convention for epoxy groups of TGDDM before (E0) and after (E1)
reaction with amine has occurred. Reproduced with permission from St John (1993).
59
1:43
The rate coefcients for the secondary-amine reactions were found to be only 17% of those
for the primary-amine reaction, thus explaining the residual secondary amine found at the
end of cure. This equation was found to explain the development of the main crosslinking
site, namely the tertiary-amine site formed on the DDS, corresponding to network interconnection. However, overlaid with the rate equation for chemical conversion that implies
that all reagents are accessible to one another is the effect of the development of the network
so that the reactions become diffusion-controlled. This is of interest since this means that the
rate coefcients now reect the chemorheology of the system, not just the chemistry. Thus,
if kd and kc represent the rate coefcients for diffusion and chemical control, the measured
rate coefcient, ke, will be given by (Cole et al., 1991)
1=ke 1=kc 1=kd :
1:44
When this analysis is applied to this system it is found that the two rate coefcients are
sensitive to different phases formed during cure. Thus k4 is sensitive to gelation whereas k5
is not. This follows since the intramolecular reaction to form a cyclic structure is of higher
probability for k5, so this reaction should not be affected by the diffusion of the reagents
together as is k4.
3. Etherification reactions
dET=dt k6 EPOH k7 EPOHTA:
1:45
This reaction is important at elevated temperatures because of the high activation energy
of the reaction. It is important to consider the products of the reaction since it is also a route
to formation of cyclic species that lowers the total crosslink density. The possible etherication reactions are shown in Scheme 1.23, and, in view of the different possibilities, the
rate equation is relatively straightforward. The inclusion of the term for catalysis by tertiary
amine reects the importance of catalysis in etherication (Cole et al., 1991).
The tertiary-amine catalysis will favour the intramolecular cyclization reactions which
will lower the crosslink density of the network in comparison with the situation if only
intermolecular reactions took place.
1.2.3
Addition polymerization
Addition or chain-growth polymerization is the most important industrial process for the
production of polymers. Polyethylene, polypropylene and polystyrene are all formed
60
Etherification
Intermolecular:
O
CH
CH2
CH2
R OH
R
R
OH
CH2
CH
CH2
CH2
CH
CH2
OH
OR
Intramolecular:
H
O
O
CH2
CH
CH2
N
R
CH2
CH2
CH
CH2
O
O
S
O
CH2
N
R
CH
O
CH2
N
CH
HO
CH2
OR
CH2
CH2
O
S
O
CH
CH2
N
CH2
N
O
R
CH2
CH2
CH2
CH
OH
Scheme 1.23. Possible etherication reactions during the cure of TGDDM with DDS. Reproduced
61
Step growth
The nature of the reactive centre denes the chemistry of the polymerization, the rate and
conditions under which high polymer may form, and particular features of the polymer
architecture (such as the tacticity; see Section 1.1.2). The nature of the reactive centre and
the monomer may also control the side reactions (such as branching) and defect groups that
may be introduced, which may affect the subsequent performance of the polymer. In the
following, we will consider the most common types of addition polymerization since this
may dene the properties of the polymer that then control the chemorheology. Certain of
these reactions are more important than others in reactive processing, and the particular
examples of reactions that occur in forming networks as well as modication of polymers
will be considered in more detail.
Some of the types of polymerization that come into this category include
free-radical polymerization
anionic polymerization
cationic polymerization
co-ordination polymerization
ring-opening polymerization
living polymerization
In some cases, the monomer may be polymerized by more than one method, and in ringopening polymerization the reaction may also occur through step-growth polymerization
(Section 1.2.1) in addition to the initiated chain-growth polymerization.
The chemistry of chain-growth polymerization may be readily distinguished from
step-growth polymerization by the features in Table 1.4.
Free-radical polymerization
The starting point for the free-radical polymerization is to choose a monomer (M) that will
react with a free radical R0 to add to the radical and also create another radical centre R0 M of
high reactivity that may add another monomer molecule. The source of the free radical to
initiate the process may be the monomer itself or an added initiator (I) chosen to produce two
free radicals, R0 , per molecule cleanly and efciently at a temperature suitable for the
addition of monomer to occur. In the case of self-initiation, heat or radiation must be supplied
in order to fragment the monomer or otherwise create the radical to initiate the chain
polymerization. Table 1.5 gives examples of typical vinyl monomers that undergo free-radical
polymerization, and Table 1.6 shows some initiators. Provided that certain thermodynamic
62
Monomer
Structure
Ethylene
(ethene)
H2 C
Styrene
(phenyl ethene)
Vinyl chloride
(chloro ethene)
HC
HC
Polymer
CH2
CH 2
CH 2
Poly(ethylene),
( CH2
CH2 )n
Poly(styrene),
( CH2
PS
CH2 )n
Cl
PE
CH2 )n
Cl
Methyl methacrylate
(2-(methoxycarbonyl)
-1-propene) or
(2-methyl 2-propenoic
acid, methyl ester)
Poly(methyl
methacrylate),
CH 3
C
CH 2
CH 3
H2
C
CH 3
CH 3
PMMA
considerations are met (in particular that, at the temperature of reaction, the free-energy
difference, DG, between monomer and polymer is negative), the reaction of monomer, M,
with the radical R0 M may propagate through m steps, and high polymer rapidly forms. At
each step of monomer addition, the radical centre at the chain end is recreated: R0 (M)m1M.
This reaction will continue, consuming monomer steadily, until the radical centre undergoes
a reaction to terminate the chain polymerization. One way for this to occur is by two
chains colliding to give bimolecular termination of the radicals, so creating inert polymer:
R0 (M)2mR0 . The degree of polymerization, DP, is thus 2m. These processes and other competing
reactions such as chain transfer are considered in detail below.
63
Table 1.6. Chemical structures of initiators and temperatures at which each initiator has a
half-life of one hour
Initiator
Dilauryl peroxide
T(C) for
t1/2 = 1 h
Structure
CH 3 (CH 2 )10 C
O
Dibenzoyl peroxide
CH 3
CH 3
C OOC
CH 3
CH 3
CH 3
H3C
CH 3
2,2-Azo bis-(2,4dimethyl valeronitrile)
91
C OOC
Dicumyl peroxide
HC
CH 3
137
CH 3
N
CH 3
H2
C C N
C
79
2,2-Azo bis(isobutyronitrile)
C(CH 2 )10 CH 3
CH 3
CH 3
N
C
C
82
H2
C
CH 3
CH
64
CH 3
obtain an expression for rp and mp in terms of the rate coefcients for the elementary
reactions and the concentration of monomer and any other reagents (e.g. chain-transfer
agents) added to modify the polymerization.
In Table 1.7, the simplest situation is considered.
The initiator decomposes thermally to give the initiating radicals R0 with an efciency f.
This is typically between 0.1 and 0.8, depending on the type of initiator and the viscosity
of the polymerizing medium, since it is reduced by radical recombination. The initiators
such as AIBN shown have a higher efciency than, say, dialkyl peroxides, since they
liberate nitrogen to produce two radicals that are more widely separated. The rate
coefcient, kd, will be described by an Arrhenius temperature dependence.
The propagation reaction is described in terms of an average rate coefcient, kp, and the
polymer radical concentration, [R]. The value of kp is very large, 102104 dm3 mol1
s1, and is independent of chain length. This is the reaction that consumes monomer and
grows the polymer chain.
The termination reactions shown are bimolecular; these may occur either by recombination to give one chain, or, depending on the monomer, disproportionation to give two
polymer chains (where DP is half that for recombination). The elementary reaction rate is
64
Table 1.7. Rate coefficients and reaction rates for elementary reactions for free-radical addition polymerization
Process
Elementary reaction
Rate coefcient
Rate of reaction
Initiation
Propagation
I ! 2R0
R0 M ! R0 M
R0 M M ! R0 MM
R0 M(m1) M ! R0 Mm
ri 2f kd[I]
rp kp[M][R]
(setting:
P 0
[R Mm ] [R])
Termination
R0 Mm R0 Mm ! R0 M2mR
R0 Mm R0 Mm ! 2R0 Mm
kd
kp1
kp2
kpm
(kp1 kp2 kpm kp)
kt
rt 2kt[R]2
the same for both cases. A simplication has been made, namely that all chains are of the
same length, m, whereas in fact there will be a statistical distribution of chain lengths,
leading to a distribution of molar mass.
The features of a linear chain reaction may be recognized in the above table. The initiation
step creates free radicals and the termination step destroys them. The rate-determining step
will be the propagation reaction, which is the product-forming step and has the characteristic
feature of a linear chain reaction, viz. one radical is formed for every radical that reacts.
Noting these features, we have
Rate of polymerization rp dM=dt kp MR :
1:46
The value of [R] is obtained by applying the steady-state approximation that d[R]/dt 0,
i.e. the rates of initiation, ri, and termination, rt, must be equal since these are the radicalcreation and -destruction steps:
ri rt :
1:47
2fkd I 2kt R 2 ;
1:48
1:49
1:50
Thus
so then
A more general expression that allows initiation by other means (e.g. radiation) in addition
to thermal initiation by an added initiator is
rp dM=dt kp ri =2kt 1=2 M keff M:
1:51
From this relation it can be seen that the free-radical polymerization will be of rst order in
monomer, M, and the effective rate coefcient, keff, will have a temperature dependence that
will depend on the activation energies Ei, Ep and Et of the elementary reactions:
keff A expEeff =RT;
1:52
65
where
Eeff Ep
1
1
Ei Et :
2
2
1:53
Since the termination reaction is a radical-recombination reaction the activation energy, Et, is
very low, 10 kJ/mol. A typical activation energy for a peroxide initiator is Ei 130 kJ/mol,
and for a vinyl monomer Ep is in the range 2040 kJ/mol. Thus Eeff, a typical activation
energy for a thermally initiated vinyl polymerization, is between 80 and 90 kJ/mol. The
dominant factor in this calculation is the activation energy for initiation, and it is seen that, if
this can be lowered, then the reaction rate will be higher at the same temperature. This may be
achieved in a number of ways.
Initiating polymerization by absorption of radiation by a photoinitiator that produces
radicals with a quantum efciency of u. If Ia is the absorbed intensity in quanta dm3 s1
then ri 2 u Ia, Ei 0 and Eeff 1535 kJ/mol.
Adding a catalyst that will lower the activation energy for initiation. An example is the
use of a transition-metal ion to act as a redox agent to decompose peroxides and
hydroperoxides. Typical values are Ei 40 kJ/mol and Eeff 3555 kJ/mol, so that the
activation energy is approximately half that of the uncatalysed reaction.
1:54
1:55
1
1
Et Ei :
2
2
1:56
Thus, since Ep is less than 12 Ei Et ; a typical value being 45 kJ/mol, the chain length
will decrease with temperature for thermally initiated polymerization. In contrast, if
photochemical initiation is used, for which Ei 0, then Em is positive (a typical value being
20 kJ/mol) and the molar mass will then increase with increasing temperature. For the
redox-catalysed, peroxide-initiated polymerization discussed above, Em will be close to zero,
so the chain length will not be sensitive to temperature.
66
The relationship between the chain length, mp, and the degree of polymerization, DP, will
depend on the mechanism of termination. If recombination occurs, DP 2mp, whereas if
disproportionation is dominant, DP mp. Values for DP can range up to many thousands
depending on the monomer.
1:57
1:58
s 1=2kt ri 1=2 :
1:59
we have
It is noted that the kinetic chain length may be written in terms of the radical lifetime as
mp kp M=2kt ri 1=2 kp Ms:
1:60
Thus, if for any reason the average lifetime of the growing radical increases, then there will
be an increase in the polymer chain length. An often-encountered example of this is the
Trommsdorff or gel effect that occurs in the polymerization of solutions of high monomer
concentration when the viscosity, g, increases and, after a certain extent of conversion, there
is a rapid acceleration in the rate of polymerization. This is interpreted as an indication of
the decrease in the rate of termination as this reaction becomes diffusion-controlled. A
feature of diffusion-controlled reactions is that the rate coefcient kt is not chemically
controlled but depends on the rate at which the terminating radicals can collide. This is most
simply given by the diffusion-controlled rate coefcient, kd, in the Debye equation:
kt kd 8RT=3g:
1:61
A practical example of this is seen (Odian, 1991) in the polymerization of methyl methacrylate, in which the term [kp/(2kt)]1/2 (that reects the rate of polymerization) increases sixfold between 10% and 40% conversion. There is some compensation for the decline in
translational diffusion by an increase in segmental diffusion, but inevitably auto-acceleration
occurs. Over the conversion range from 10% to 80% the radical lifetime increases from 1 s
to over 200 s. The full understanding of kinetics at high conversion is further complicated
by the effect of the viscosity on the initiator efciency, f, as well as a decline in kp. The
nal stage may be reached when the glass-transition temperature, Tg, exceeds the
reaction temperature and the reaction ceases because the system vitries. Radicals may
be trapped in the glassy state and have very long lifetimes because there are no termination
reactions.
67
Table 1.8. Effects of addition of an inhibitor, S, or chain-transfer agent, XA, on the elementary reactions
and their rates in free-radical polymerization
Process
Elementary reaction
Inhibition
Chain transfer
Re-initiation
R0 M S ! R0 MS
R0 M XA ! R0 MX A
A M ! AM
kst (kp)
ktr
ka
kst[R0 M][S]
ktr[R0 M][XA]
ka [A][M]
68
occur to all species that are present in the polymerization solution: the polymer, monomer,
solvent and initiator. The effect that this has on the polymerization depends on the reactivity
of the radical that is formed. Certain solvents such as chlorinated hydrocarbons are powerful
chain-transfer agents. A chain-transfer coefcient can be dened for all species as the ratio
of the rate coefcient for the transfer reaction to that for propagation: ktr/ kp. Chain transfer
to polymer will result in the formation of branches and broadening of the molar-mass
distribution. The role of chain transfer to polymer that results in branching cannot be
realized by examination of DP since it does not change the number of polymer molecules in
the system (Barson, 1989).
The case of chain transfer to initiator may be exploited in special forms of controlled freeradical polymerization, e.g. with sulfur-containing initiators that are discussed later within
the topic of living free-radical polymerizations.
The practical case industrially is when radical transfer to the added chain-transfer agent,
XA, is the dominant reaction. The effect on DP may be determined by noting that in
the presence of XA the rate of formation of polymer has increased above that produced by
the normal termination reaction. The kinetic chain length, mp, is the ratio of the rate of
propagation to the rate of formation of chain ends:
DP 2mp 2rp =rt ktr R0 M XA:
1:62
If the degree of polymerization in the absence of the chain transfer agent is DP0, then, using
Equations (1.46) and (1.47), the following form of the Mayo equation is obtained:
1=DP 1=DP0 ktr XA=kp M:
1:63
The ratio ktr/kp is the chain-transfer coefcient, CA. Similar coefcients, CS, CM and CI, may
be dened for chain transfer to solvent, monomer and initiator, respectively. In commercial
systems, the value of CA is chosen to be >1 so that only small concentrations of XA are
required in order to have a marked effect on the value of DP. For example, if 0.1% of XA is
added with respect to monomer and DP0 is 1000, then, if CA is 1, the DP is reduced to 500.
Also, since ktr kp, the agent is consumed slowly and there is no effect on the rate of
polymerization.
Conversely it is possible to produce low-molar-mass oligomers or telomers by deliberately choosing an agent with a large value of CA (e.g. methyl mercaptan, CA 2 105 in
styrene), so that DP is reduced to 5 for a concentration of 0.001%. Further particular
examples of chain transfer (e.g. to polymer to form branches) will be discussed later,
together with the use of reversible-addition fragmentation transfer (RAFT) and other
radical-mediated synthetic strategies.
1:64
These changes are for converting one mole of monomer to one mole of polymer repeat
unit. There is an obvious loss of entropy on taking a monomer molecule with a high
69
Table 1.9. Enthalpy and entropy changes for conversion of monomer to polymer, and the
corresponding ceiling temperature
Monomer
Styrene
a-Methyl styrene
Methyl methacrylate
Ethylene
Propylene
DH0 (kJ/mol)
Tc (oC)
70
35
55
93
84
105
104
117
155
116
394
66
198
327
451
1:65
Thus the values shown in Table 1.8 are for standard conditions and represent just one of
a series of ceiling temperatures for various monomer concentrations above which
polymer formation is not favoured. Thus, in a bulk polymerization reaction the ceiling
temperature may change with conversion in such a way that complete conversion is not
achieved. For example, if methyl methacrylate is polymerized at 110 C the value of
[M]c calculated from the above equation is 0.139 M and this will be the monomer
concentration in equilibrium with the polymer. The polymer, when removed from the
monomer, will have the expected ceiling temperature as given in Table 1.8 and will
depolymerize only if there is a source of free radicals to initiate the depolymerization
(Section 1.4.1)
Anionic polymerization
Polymerization by ionic initiation is much more limited than that by free-radical initiation
with vinyl monomers, but there are monomers such as carbonyl compounds that may be
polymerized ionically but not through free radicals because of the high polarity. The
polymerization is much more sensitive to trace impurities, especially water, and proceeds
rapidly at low temperature to give polymers of narrow molar-mass distribution. The chain
grows in a living way and, unlike in the case of free-radical polymerization, is generally
terminated not by recombination but rather by trace impurities, solvent or, rarely, the
initiators counter-ion (Fontanille, 1989).
70
Initiation:
n-BuLi + CH2 =CHR n-Bu-CH2 -(HR)C - Li +
Propagation:
n-Bu-CH2 -(H)(R)C- Li + + CH 2 =CHR n-Bu-(CH 2 CHR)nCH2 -(H)(R)C- Li +
Termination by chain transfer to water (or alcohol):
---CH2 -(H)(R)C- + H2O ---CH2 -(R)CH2 + OHScheme 1.24. The sequence of reactions for anionic polymerization of an electron-withdrawing
monomer (e.g. acrylonitrile) initiated by n-butyl lithium.
The range of vinyl monomers that may be polymerized via an anionic route is restricted
to those with strong electron-withdrawing groups attached to the vinyl group so that attack
by the negatively charged (nucleophilic) anionic initiator is favoured. They also then stabilize the propagating anionic species, as shown in Scheme 1.24 for the polymerization by
n-butyl lithium of a monomer such as acrylonitrile. The propagation step is insertion of
monomer at the anion and the initiator is recreated so that the chain grows until the
monomer is exhausted or there is chain transfer to the solvent. The living chain end may
spontaneously decay through an elimination reaction (e.g. hydride elimination from the
polystyryl carbanion (Odian, 1991)). Usually the chain is terminated by the addition of a
chain-transfer agent such as water or alcohol.
The effectiveness of water as a chain-transfer agent is seen from its value of Cs of 10 at
25 C in the polymerization of styrene. The reactions are often carried out at low temperature in order to limit unwanted chain transfer and side reactions. The long life of the
carbanion is seen by virtue of the maintenance of its bright colour (e.g. red for the styryl
carbanion).
One feature of anionic polymerization is that it is possible to have polymerization
occurring at both ends of an initiating dimer formed by electron transfer from a metal aryl
compound such as sodium naphthalenide. This electron transfer results, in the case of
styrene monomer, in the formation of a dianion:
71
Table 1.10. Rate coefficients and reaction rates for elementary reactions for anionic addition polymerization
Rate
coefcient
Process
Elementary reaction
Initiation
ki
MtA ! Mt A
ki
A M ! AM
kp
kp
AM M ! kAMM
p
AMn1 M !AMn
ktr
AMn H2O ! AMH OH ktr
Propagation
Terminationa
a
Rate of reaction
ri ki[A][M]
kiK[MtA][M]/[Mt]
rp kp[M][M]
rtr ktr[M][H2O]
all propagating species and termination is taken to be chain transfer to water present as an
impurity, then solving using the steady-state approximation applied to [M] gives
rp kp ki KM2 MtA=ktr H2 OMt :
1:66
1:67
M=CtrW H2 O;
1:68
noting that the coefcient for chain transfer to water CtrW ktr/kp.
In many systems where the initiator decomposes completely (so the anion concentration
is that of the initiator, MtA) and there is no chain transfer, i.e. the chain length is governed
by the consumption of monomer, on complete reaction
DP Monomer consumed=number of chain ends
M=MtA:
1:69
The striking feature of anionic polymerization of having a narrow molar-mass distribution results from the fact that initiation occurs rapidly and totally before propagation, so
every anion represents a growing polymer chain and the termination processes that may
lead to a distribution of chain lengths are limited. A feature of such living polymerization is that at the end of the reaction, with complete consumption of monomer,
the reactive centres are still available for further reaction and, if monomer is added, the
reaction proceeds with the molar mass continuing to increase linearly with conversion.
General aspects of living polymerization, particularly those which may occur at
higher temperatures and with other types of initiation (including the use of free-radical
initiators) are discussed later.
72
polyacetal.
The polymer has a relatively low ceiling temperature (119 C), but this is the highest of
all the formaldehyde polymers. The lack of success at polymerizing other monomers was
due to the low ceiling temperatures (e.g. 39 C for acetaldehyde) (Odian, 1991). Aldol
condensation can be a side reaction competing with polymerization for these monomers.
There are other routes, such as cationic ring-opening polymerization of cyclic acetals, to
achieve the same polymer (Penczek and Kubisa, 1989).
Cationic polymerization
Cationic polymerization shares several of the features discussed above for anionic polymerization, but the requirements in terms of the monomer structure will be opposite, i.e. the
vinyl monomers should have an electron-donating side group since the propagating centre
carries a positive charge. Alternatively, the presence of an aromatic ring can stabilize the
cation by making available resonance structures. Examples of these are alkyl vinyl ethers,
isobutylene, isoprene, styrene and a-methyl styrene (Sauvet and Sigwalt, 1989).
Initiation of cationic polymerization may be achieved by any proton-donating species, but
the limitation in, for example, protonic acids is the nucleophilic nature of many of the
counter-ions, e.g. halide, such that hydrogen halides are ineffective initiators since they
produce only the addition product with vinyl compounds. Lewis acids such as BF3 and
AlCl3 are more widely used but require a proton donor, e.g. water, and it is this proton donor
that is the actual initiator and the counter-ion is a product of the resulting reaction: e.g. BF3
initiation of the polymerization of isobutylene (Odian, 1991) shown in Scheme 1.26.
In recognition of the fact that the initiating species is not BF3 (since under anhydrous
conditions this polymerization could not occur) but rather the proton donor (or protogen),
the BF3 is described as the co-initiator. The initiating system in Scheme 1.26 is thus the
initiatorco-initiator complex. The donor may be a carbocation rather than a proton, and the
initiator is then a cationogen, with an example being AlCl3 and t-alkyl chlorides. Selfinitiation by a self-ionization process may also occur, but the reaction rate is much lower
than that for the initiatorco-initiator system. The details of carbocation generation leading
to different initiation routes are discussed in detail by Sauvet and Sigwald (1989).
Propagation occurs through the addition of monomer at the cation by insertion between it
and the counter-ion. A rapid succession of additions to the cation then produces high
polymer. The temperature must be kept sufciently low to enable this reaction to compete
with chain transfer to monomer (shown below) that terminates the chain and results in
re-initiation, but prevents the formation of high-molar-mass polymer.
Isomerization may occur to compete with propagation, particularly in the 1-alkenes, other
than isobutylene (the illustration used here). This is linked to the possible rearrangements of
the propagating cation to more stable cations. Monomers such as styrene and vinyl ethers
are not affected since they do not have structures available for rearrangement. An example
73
Initiation:
BF3 + H2O BF3OH2
BF3OH2 + (CH3)2C=CH2
Propagation:
(CH3)3C+(BF3OH) + n(CH3)2C=CH2
H-[CH2C(CH3)2]n CH2C+(CH3)2(BF3OH)
Chain transfer to monomer:
H[CH2C(CH3)2]n CH2C+(CH3)2(BF3OH) + (CH3)2C=CH2
H[CH2C(CH3)2]n CH2C(CH3)=CH2 + (CH3)3C+.... (BF3OH)
Termination:
H[CH2C(CH3)2]n CH2C+(CH3)2(BF3OH)
H[CH2C(CH3)2]n CH2C(CH3)2OH + BF3
Scheme 1.26. Elementary reactions in the cationic polymerization of isobutylene initiated by boron
triuoride.
74
Table 1.11. Elementary reactions, rate coefficients and reaction rates for a simple cationic polymerization
Process
Elementary reaction
Initiation
I RX ! RIX
RIX$ R IX
ki
RIX M ! RM IX
Propagation
Termination
RMIX nM ! RMnMIX
kt
RMnMIX ! RMn1 X I
kp
Rate
coefcient
Rate of reaction
ki
ri kiK[I][RX][M]
kp
kt
rp kp[RMIX][M]
rt kt[RMnMIX]
summarised (Table 1.11). In the above relations, the average concentration of all reactive
species is assumed. In the steady state, the rates of initiation and termination are equal,
so the concentration of the carbocation reactive intermediate may be determined and
substituted into the propagation rate equation:
rp kp RM IX M kp ki K=kt IRXM2 :
1:70
1:71
It was noted above that chain transfer to monomer was a more frequent chain-breaking
reaction than termination, but the chain-transfer reaction resulted in re-initiation of the
chain. The degree of polymerization in this case is reduced from DP0 to DP depending on
the value of the chain-transfer coefcient, CM ktr/kp, hence
DP rp =rt rtr :
1:72
1=DP 1=DP0 CM :
1:73
Thus
It has been noted (Odian, 1991) that the steady-state approximation cannot be applied in many
cationic polymerizations because of the extreme rate of reaction preventing the attainment of
a steady-state concentration of the reactive intermediates. This places limitations on the
usefulness of the rate expressions but those for the degree of polymerization rely on ratios of
reaction rates and should be generally applicable. The molar-mass distribution would be
expected to be very narrow and approach that for a living polymerization (with a polydispersity index of unity), but this is rarely achieved, due to the chain-transfer and termination
reactions discussed above. Values closer to 2 are more likely.
The temperature of reaction for cationic polymerization is usually kept very low. The rate
of initiation is largely insensitive to temperature, so the rates of propagation and termination
alone determine the temperature dependence of the polymerization. Thus the observed
activation energy, Eobs, will be just the difference between propagation and termination
activation energies:
Eobs Ep Et :
1:74
Generally Et is greater than Ep, so the observed activation energy is negative (so that high
reaction rates occur at low temperature). The same conclusion is reached for the degree
75
Co-ordination polymerization
The use of co-ordination polymerization is best known in the ZieglerNatta polymerization
of olens containing an asymmetric centre (e.g. propylene) to give stereoregular polymers
of industrial signicance (Tait, 1989). The importance of the polymer tacticity in achieving
conformations that allow crystallization and thus useful mechanical properties was discussed earlier (Section 1.1.2). For monomers without an asymmetric centre, such as
ethylene, ZieglerNatta catalysts allow polymerization at low pressures and at room temperature to give polymers with higher density and lower degrees of branching than those
achievable by free-radical polymerization.
Modern, heterogeneous ZieglerNatta catalysts are prepared from a (restricted) combination of the following species:
(1). a group IVVIII transition-metal compound, e.g. TiCl3
(2). an organometallic compound of group IIV, e.g. Al(C2H5)3 or Al(C2H5)2Cl
(3). a donor such as a Lewis base or an ester, e.g. butyl benzoate
(4). a support such as MgCl2
The goal in catalyst development is to achieve a high activity (dened as kilograms of
polymer produced per gram of initiator) together with high stereospecicity. Thus a modern
catalyst for isotactic polypropylene may have an activity of 2000 kg/g Ti and an isospecicity of 98% (Odian, 1991). The support has an important role to play in achieving
high activity since the number of available sites on Ti may be less than 1% of the theoretical
maximum if no support is used. In some cases the catalyst may covalently bond to the
support. The catalyst remains in the polymer after polymerization, and this may become an
issue affecting the long-term stability of the polymer (Section 1.4). Many catalysts based on
soluble metallocenes have been developed in the attempt to achieve higher iso-specicity
with high yield.
Mechanism of stereopolymerization
There is no single mechanism for ZieglerNatta polymerization because of the variety of
catalyst and co-catalyst systems as well as the different phases in which the reaction may
take place. The process of stereoregular polymerization can be understood from the
mechanism of initiation and propagation as the monomer is inserted at the polymerization
site on the catalyst surface. The detailed mechanism for achieving stereospecicity is an
active area of research (Corradini and Busico, 1989, Tait and Watkins, 1989), but some
general principles may be learned from the simple ZieglerNatta catalysts (Allcock and
Lampe, 1981).
The mechanism involves the formation of a transient co-ordination complex between the
olen and the transition metal (Ti) because of a vacant orbital on the metal as shown in
Scheme 1.27. The aluminium complex (not shown) also participates by constraining the
direction of addition of the monomer. This may also involve alkylation of the titanium as
76
Polypropylene chain
CH3
CH
Cl
Cl
CH2
Cl
Ti
CH3
+ H2C
CH
CH2
Cl
CH
Cl
Ti
CH2
Cl
Cl
CH3
propylene monomer
Cl
Cl
Cl
CH2
Cl CH2
Ti
Cl
CH3
CH2
Cl
CH
Cl
Ti C
H2
CH3
Cl
Cl
CH2
Cl
Ti
Cl
Scheme 1.27. A simple reaction scheme for polymerization of isotactic polypropylene using a
TiCl4 ZieglerNatta catalyst. After Boor (1979).
well as stabilizing the reaction site. The reaction will occur without the aluminium complex,
but the stereospecicity and yield are lower. The above references provide the detail of the
mechanisms that have been postulated, for the purposes relevant here we note only that
reaction takes place via an insertion followed by migration of an alkyl group through a fourcentred transition state (Boor, 1979).
The reason for stereospecic polymerization of the olen is the combined effect of the
steric and electrostatic repulsions between the substituent (e.g. methyl group on propylene)
and the ligands of the transition metal. These sites occur on the edges of the titanium crystal,
so the efciency is often related to the physical structure of the catalyst. This can change on
proceeding through the polymerization because the pressure from the growing polymer can
shatter larger crystals, and the rate progressively increases until a steady rate corresponding
to the smallest achievable particle size has been reached.
The rate of polymerization is thus
rp kp Ti
M;
1:75
where [Ti*] is the concentration of active sites, which may be between 102% and 10% of
the actual [Ti]; [M] is governed by the adsorption isotherm for the monomer on the catalyst
surface under the conditions of the polymerization. To determine the expected molar mass
of the polymer the termination reactions need to be known. The lifetime of the growing
chain is governed by the chain-transfer reactions that may occur. It is well known that
polyethylene polymerized through a ZieglerNatta route contains a range of unsaturated end
groups. These may arise by the mechanisms shown in Scheme 1.28.
These reactions account for terminal vinyl-group formation, but there are other species,
viz. trans-vinylene and vinylidene, that cannot be accounted for in this way. It is noted that
these do not terminate the chains since re-initiation at the Ti site may occur.
The polydispersity with heterogeneous ZieglerNatta polymerization may range from 5
to 30. High-density polyethylene (HDPE) produced by co-ordination polymerization has
a degree of branching of about 15 per 1000 monomer units, compared with 2550 for
77
low-density (free-radical, high-pressure polymerized) polyethylene (LDPE). It is consequently higher melting (135 C versus 110 C) and has a degree of crystallinity of >70%,
compared with 50% for LDPE. The molar mass is typically 50 000250 000 g/mol, but
there are applications (e.g. in biomedical orthopaedic implants) where ultra-high molar
masses of 1.5 106 and above are needed. In this case the chain entanglements persist well
above the melt temperature and the materials have very high wear resistance.
Ring-opening polymerization
Ring-opening polymerization of a cyclic monomer to yield a linear polymer is considered a
separate class of polymerization even though it may involve an initiation system that is either
anionic or cationic and thus have some features in common with those systems discussed
before. Not all ring-opening polymerizations are addition polymerizations; an example of a
stepwise reaction is the industrially-important ring-opening of e-caprolactam to give nylon-6.
The main difference arises in the thermodynamic requirements for polymerization. In the
case of the polymerization of a vinyl monomer, there is a large enthalpic difference between
the monomer and polymer that overcomes the loss in entropy that accompanies the constraining of the monomer to form a linear chain. In the case of conversion of a cyclic
molecule to a polymer, there is little change in the enthalpy per repeat unit between the
monomer and the polymer, just a loss of the ring strain energy.
Considering a hypothetical system such as the polymerization of cyclic alkanes to give
polymethylene, if the ring is unstrained (as in six-membered cyclics) then polymer formation
is not favoured thermodynamically since DG is positive for this system. The actual polymerization of the smaller and larger rings, for which DG is negative, is generally not achieved
since there is no reactive site for ring-opening and thus no mechanistic pathway to make the
reaction kinetically feasible. Introduction of a hetero-atom or a functional group such as an
ester or amide immediately makes the reaction feasible by offering a reaction pathway.
These reactions are all commercially signicant and the reactions may in some cases take
place in a reactive extruder. The mechanism of the reaction depends on whether the reaction
is anionically or cationically initiated, and the effects of reaction conditions are those as
discussed before.
Nylon-6
The example in Scheme 1.29 of the ring-opening polymerization of e-caprolactam is given
since it has also been exploited in reactive processing (Brown, 1992). In the example given
the initiator is 0.2%0.3% sodium lactamate. The anionic polymerization (such as that in
Scheme 1.29) is faster than cationic or hydrolytic polymerization.
78
Na
N
H
N
O
C
sodium lactamate
O
H
N
O
H
N
C
n
caprolactam
Polycaprolactam (Nylon 6)
Scheme 1.29. Polycaprolactam polymerization from ring-opening of caprolactam. After
Brown (1992).
BF3OH +CH2CH[OH]
Scheme 1.30. Ring-opening polymerization of epoxides initiated by BF3.
Epoxides
It was noted earlier that, among the reactions of diglycidyl epoxide resins, the linear
polymerization with an initiator such as BF3 was feasible. This is often exploited in
reactive processing of carbon bre pre-impregnated with epoxy resin that contains a
latent curing agent that is a complex that does not initiate at room temperature such
as BF3MEA. Whether the polymerization occurs through a cationic or an anionic
mechanism depends on the detailed composition of the catalyst (Hodd, 1989). For Lewis
acids and other protonic reagents the mechanism involves a co-catalyst such as water
(Scheme 1.30).
The bulky anion then stabilizes the intermediate adduct from protonation of the epoxy
group and then facilitates insertion of epoxide at the cationic propagation site. Rapid
polymerization can then occur. Cationic photopolymerization of epoxides often involves the
photo-generation of acid from an initiator such as diaryliodonium or triaryl sulfonium salts
(Crivello, 1999). The anions are important in controlling the addition at the cationic site and
are typically BF4 and PF6. The reactivity of the system depends also on the structure of
the epoxide.
Polydimethyl siloxanes
There are several inorganic polymers that are formed by a ring-opening polymerization,
among which polydimethylsiloxanes (PDMSs) are most relevant to studies in reactive
processing. The starting material is the cyclic monomer, D4, [Si(CH3)2O]4, as shown in
Figure 1.22. The anionic polymerization is initiated by base such as KOH to give the
anionic centre shown in Scheme 1.31.
Propagation occurs by addition and ring-opening of D4 at the anionic site. Water
may act as a chain-transfer agent or be used to terminate the polymerization, as
shown above. Alternatively, to create the more familiar trimethyl-terminated PDMS,
hexamethyl disiloxane is added to end-cap the chain, as shown in the last reaction in
Scheme 1.31. This may then initiate further chains at the new anion site. Consequently
PDMS oligomers may have a range of trimethyl and hydroxyl terminating groups.
Thus, while PDMS is the most common silicone polymer, a range of other groups, such
79
OCH2CH2
)n
ethylene oxide
O
( OCH2 )n
O
O
Poly (oxy methylene)
or Polyacetal
trioxane
O
O
C
O
-caprolactone
H3C
CH3
CH3
Si
Si
H3C
Poly (caprolactone)
CH3
Si
O
H 3C
C
n
O
O
CH3
Si
H3C
Si
CH3
CH3
octamethylcyclotetrasiloxane or D4
Figure 1.22. Examples of cyclic monomers and their polymers from ring-opening polymerization.
Initiation:
KOH + D4 HO-Si(CH3)2-O----Si(CH3)2O- K+
Propagation:
HO-Si(CH3)2-O----Si(CH3)2O- K+ + nD4 HO-[Si(CH3)2-O]4n+4- K+
Termination:
HO-[Si(CH3)2-O]4n+4- K+ + H2O HO-[Si(CH3)2-O]4n+4H + KOH
Chain Transfer:
----Si(CH3)2O- K+ + (CH3)3Si-O-Si(CH3)3
----Si(CH3)2O Si(CH3)3 + Si(CH3)3O- K+
Scheme 1.31. The reaction mechanism for ring-opening polymerization of the cyclic siloxane D4 to
80
as vinyl and phenyl, may be introduced to achieve different physical and chemical
properties.
It has been noted that this polymerization is thermodynamically unusual in that the driving
force for the polymerization is the increase in entropy, DS, of 6.7 J mol1 K1 on moving from
the constrained cycle of D4 to the extremely exible polydimethylsiloxane backbone, which
has many available conformations (Odian, 1991). Since DH is close to zero the only reason
why the polymerization proceeds is the increase in entropy on polymerization.
Cyclic oligomers
The above example of the entropically driven ring-opening polymerization of D4 to yield
PDMS may be extended to other cyclic oligomers (Hall and Hodge, 1999). It was noted
earlier that cyclic oligomers offered a route to high-molar-mass polycarbonates by the ringopening reaction using a transesterication catalyst. Similar reactions have been observed
under conditions for addition polymerization. The equilibrium between the cyclic monomer
and polymer is driven in the direction of polymer at the high concentration used in the
polymerization. Among the attractive features of the use of cyclic oligomers are the absence
of heat of reaction and the control of molar mass achieved by controlling the number of
end groups.
It is noted that these reactions need not always proceed in the same way as for the
non-cyclic monomer. For example, in the thermal polymerization of a macrocyclic ethylene
terephthalate dimer, it has been noted (Nagahata et al., 2001) that, from the molar-mass
distribution, the polymerization cannot be occurring through a stepwise polymerization but
must rather go through an ionic addition reaction. It was suggested that an alkoxide ion
generated through thermal scission of an ester linkage initiated the reaction by ring-opening
at the ester carbonyl of other macrocyclics.
81
Chain
Molar
Mass
Living
Step-growth
20
40
60
80
100
% Conversion
Figure 1.23. The relation between polymer molar mass (molecular weight) and conversion for chain,
achieve high-molar-mass polymer since the activation energy for propagation is much lower
than that for chain transfer (often to monomer) or termination by adventitious impurities.
Unlike with anionic polymerization, the presence of these reactions results in a broader
molar-mass distribution.
Since high-molar-mass polymer is produced from the onset of polymerization, neither of
these reactions, when performed with the initiator system and conditions described so far,
can be regarded as a living polymerization. However, it has been noted that, if the
necessary low concentration of the reactive intermediate could be achieved by having a
dormant species in equilibrium with the propagating species (Matyjaszewski, 1999), then
it is possible to control the extent of the polymerization more precisely, so that it increases
linearly with conversion. Whether this can be truly classed as a living polymerization
has been debated (Darling et al., 2000), but there is agreement that features of living
polymerization occur when propagation and reversible termination are faster than any
irreversible termination. A further feature is that living polymers can be re-initiated by
supplying further monomer. In the following, just a few of the many examples of
approaches that have been employed for living free-radical and cationic polymerizations are
considered, which may be added to the anionic living polymerizations discussed earlier.
Some of these systems have been applied in the area of reactive processing and, since these
are principally based on free-radical systems, only these are considered here.
Nitroxide-mediated polymerization
The key feature of the use of a dormant species may be seen in the following general
scheme (Scheme 1.32) that involves complexation of the propagating species by
means of a stable nitroxide radical (Hawker et al., 2001). The PnO bond of the alkoxy
amine PnONR is thermally labile at the polymerization temperature, so this becomes
the site for the insertion of monomer. Propagation then occurs at a rate that is much
slower than for a simple free-radical addition reaction since the propagating radical
concentration (which is governed by the position of the equilibrium with the alkoxy amine
82
Pn-O-NR Pn + O-NR
+M
H3C
H3C
H3C
H3C
N
N
H3C
130oC
O
H3C
H3C
CH3
O
CH
H3C
CH3
HC CH2
n
H3C
H3C
H3C
H
C
H2
C HC
N
O
H3C
CH3
Scheme 1.33. Nitroxide-mediated living radical polymerization of styrene using TEMPO. After
Hawker et al. (2001).
at the reaction temperature) is very low. It is this low concentration of reactive chain
ends that minimizes irreversible termination reactions that occur in the normal freeradical polymerization. A further living feature is that DP increases linearly with
conversion.
The initiation of the polymerization shown above may take place from a preformed
alkoxy amine with an end group commensurate with the monomer, M. This is shown in
Scheme 1.33 for styrene polymerization (Hawker et al., 2001).
While the initial alkoxy amines were based on the nitroxyl radical TEMPO (2,2,5,6tetramethyl piperidinyloxy) as shown above, this system is of limited applicability
because of the high temperatures (>125 C) and long polymerization times (>24 hours)
83
84
ADDITION
Pn
S
C
Pn
S
R
Pn
+ R
EQUILIBRATION
RE-INITIATION
mM + R
FRAGMENTATION
S
Pm +
M
S
C
Z
S
Pn
Pm
Pn
S
Pm
C
Z
+ Pn
M
Scheme 1.35. The principle of living radical polymerization using a RAFT agent. After Mayadunne
et al. (1999).
1:76
This is seen in Figure 1.24 (Mayadunne et al., 1999), in which two potential RAFT agents,
at a concentration of 2.97 102 M, are assessed in the bulk polymerization of styrene
at 100 C. It is found that the rst agent, S-benzyl N, N-diethyldithiocarbamate; (i.e. Z is N,
N-diethyl-) produces polymer after 6 hours that has Mn 317 100 g/mol and Mw/Mn 1.86
(GPC trace A). The theoretical value after this time is 4 700 g/mol and it is apparent
that RAFT polymerization is not occurring for this compound and normal free-radical
polymerization of the bulk monomer is occurring, with a high polydispersity. In contrast,
if S-benzyl N-pyrrolocarbodithioate (Z is N-pyrrolo-) is used, then trace B is obtained
after 6 hours (Mn 6480 g/mol) and trace C (Mn 15 600 g/mol) after 30 hours. The
85
trace C
trace B
trace A
107
10
105
104
1000
calculated value of Mn in the latter case is 18 200 g/mol and the measured polydispersity
is Mw/Mn 1.20. It is clear that this compound is an effective RAFT agent.
The kinetics and mechanisms of these systems are being elucidated from studies of model
polymerRAFT-agent adducts. It has been shown that a high value of the exchange rate
coefcient is required in order to ensure that low-polydispersity polymer is formed early in
the polymerization (Goto et al., 2001). The usefulness of RAFT, and indeed all living
polymerization systems such as those discussed above, lies in the existence of a living
site that is available, after the exhaustion of monomer, for re-activation by the addition of a
second monomer to give a block copolymer of low polydispersity. The range of molecular
architecture that may readily be formed in living polymer systems is discussed in the
next section.
1.2.4
86
Figure 1.25. A GPC trace for the polymer from the polymerization of D4 with CF3SO3H, showing
products ranging from cyclics through to linear (LP) and high polymer (HP). Reproduced with
permission from Kendrick, Parbhoo et al. (1989). Copyright 1989, Pergamon Press.
Cyclization
The same thermodynamic and ring-stability considerations for the formation of large cyclic
polymer molecules as discussed in Section 1.2.2 will apply to polymers formed through
addition polymerization. It has been noted that in siloxane polymers cyclic oligomers may
constitute a signicant component of the nal polymer. This depends on the polymerization
mechanism and, in the case of the base-and acid-catalysed ring-opening polymerization of
the cyclotetrasiloxane D4 (by anionic and cationic reactions, respectively), there is a continuous population of cyclic oligomers in equilibrium with the linear polymer up to D400
(Kendrick et al., 1989). The isolation and identication of oligomers beyond D25 is challenging, but GPC and mass-spectral methods such as MALDI-TOF-MS enable these to be
precisely identied (Hunt and George, 2000). In the cationic polymerization of D3 and D4
with CF3SO3H macrocyles with molar mass up to 10 000 g/mol are formed, as seen from the
resolution of these macrocycles from the high linear polymer at one end of the distribution
and the small cycles at the other end in the GPC trace (Figure 1.25).
It is of interest to determine whether these macrocycles form by addition of the small
cyclic species, e.g. D3, directly to the growing macrocycle or the reaction proceeds by
closure of the linear chain after addition (Kendrick et al., 1989). Although the latter seems
mechanistically improbable for large cycles, the actual root mean separation of the ends
(Rrms, Section 1.1.1) of the polymer increases as the square root of the degree of polymerization, so it has been noted that the end groups for a 50-siloxane-unit chain will be only
4 nm apart (Kendrick et al., 1989). This suggests that a signicant fraction will be able to
react, so cyclization competes directly with linear propagation. Analysis of the conformational entropy of the cyclic oligomers by Stockmayer and Jacobson led to the prediction that
the concentration, [Dn], of the n-oligomer varied as n5/2. The kinetic analysis gave a
prediction of n3/2. In either case the end-closure mechanism is favoured over the ringexpansion mechanism for the production of the siloxane macrocycles.
The importance of cyclization resides in the difference between the properties of linear
and cyclic oligomers of the same molar mass. The cyclics have a lower viscosity and a
higher diffusion coefcient and, in applications of siloxane polymers, the role that
87
the highly mobile fraction of oligomers may play in determining the properties,
particularly of the surface, of even a crosslinked PDMS elastomer should be considered
(Hunt et al., 2002).
Copolymerization
Many of the polymers of industrial importance are copolymers consisting of units of
more than one type of monomer in the same polymer chain, and addition polymerization
is one of the major routes for copolymer formation. The properties that result depend
on how the monomers are distributed along the polymer chain as well as on their
concentration.
Considering two monomers M1 and M2, the possibilities are shown below.
1.. Random or statistical copolymer: M1M1M1M2M2M2M1M2M1M1M1
In this case the distribution of M1 and M2 along the copolymer chain cannot be predicted
or controlled and arises when the propagating radicals (or ions) of the different monomers have a nite probability of reaction with either monomer.
2.. Alternating copolymer: M1M2M1M2M1M2M1M2M1M2
In this case, the propagating radical of one monomer will react only with the second
monomer, not with its own monomer. After initiation, the system will develop an
alternating composition regardless of the initial ratio of the two monomer concentrations.
3.. Block copolymer: (M1M1M1M1M1M1)(M2M2M2M2M2M2)
The block copolymer cannot readily be synthesized from an initial mixture of the two
monomers, but requires a strategy such as the reaction of preformed oligomers with
functional end groups or the polymerization of a sequence of M1 followed by a sequence
of M2. The living polymerization methods described above in Section 1.2.3 are a useful
route to such systems.
4. Graft copolymer:
M1
M1
M1
M1 M2 M2 M2 M2 M2 M2 M2
M1
M1
M2M2 M2 M2 M2 M2M1
M1
In graft copolymerization, there are also sequences of M1 and sequences of M2, but they are
formed at branch points, so all graft copolymers are, by denition, branched. These are also
formed in sequential reactions, and graft copolymerization is frequently employed in reactive
processing.
The composition of a copolymer formed in an addition-polymerization reaction will not
simply be the composition of the feed M1/M2 since the reactivity of the two monomers to
the initiating and propagating species (whether free radical, anion or cation) may differ.
The kinetics of copolymerization is a suitable route for the introduction of the concepts
since this then allows the composition of the copolymers to be described systematically.
Free-radical reactions are those mostly encountered in reactive processing and are
considered below.
88
M1
M2
M2
M1
Rate coefcient
!
!
!
!
M1M1
M1M2
M2M2
M2M1
k11
k12
k22
k21
Reactivity ratio
r1 k11/k12
r2 k22/k21
1:77
1:78
By simply dividing one of these two equations by the other, it is possible to obtain an
expression for the instantaneous composition of the copolymer:
dM1 k11 M1 M1 k21 M1 M2
:
dM2 k12 M2 M1 k22 M2 M2
1:79
To solve these equations and eliminate the reactive intermediates, we note the stationarystate approximation and the denition of the reactivity ratios r1 and r2 above. Then
dM1 M1 r1 M1 M2
:
dM2 M2 r2 M2 M1
1:80
This is the copolymer equation, which may also be derived by statistical means without
invoking the stationary-state approximation (Odian, 1991).
Inspection of the copolymer equation shows the importance of the reactivity ratios in
determining the type of copolymerization reaction that will occur. Thus, we consider the
following possibilities.
(a). r1 r2 0. This means that k11 k12 and k22 k21, i.e. each monomer will react
preferentially with the other monomer radical rather than homopolymerize. This will
result in an alternating copolymer, i.e. d[M1]/d[M2] 1.
(b). r1 r2 1. The propagating radicals have an equal reactivity with each of the monomers. This will result in a random copolymer with composition identical to that of the
feed, i.e. d[M1]/d[M2] [M1]/[M2].
(c). r1 1/r2. This is termed an ideal copolymer since each of the propagating species has
an equal probability of reaction with each of the monomers (for the same concentration
of each). The composition is then
dM1 =dM2 r1 M1 =M2 :
89
Table 1.13. Copolymer composition for the ideal copolymerization, for which r1 1/r2; f1 and f2 are the
molar feed ratios of monomers, M1 and M2, and F1 and F2 are the mole fractions of the repeat units in the
copolymer (Equation (1.82))
r1
r2
f1
f2
F1
F2
10
0.1
0.2
0.5
0.9
0.5
0.1
0.9
0.5
0.1
0.9
0.5
0.1
0.9
0.5
0.1
0.1
0.5
0.9
0.1
0.5
0.9
0.1
0.5
0.9
0.1
0.5
0.9
0.99
0.91
0.53
0.98
0.83
0.36
0.9
0.5
0.1
0.82
0.33
0.05
0.01
0.09
0.47
0.02
0.17
0.64
0.1
0.5
0.9
0.18
0.67
0.95
1:81
1:82
when r1 1/r2.
Note that Table 1.13 also includes the special case r1 r2 1, for which F1 f1 and
F2 f2. Another condition for such azeotropic copolymerization in which the copolymer composition is equal to the feed composition is when f1/f2 (r2 1)/(r1 1). The
desirable range of r1 to produce copolymers with a reasonable range of both monomers
present is from 0.5 to 2.
(d). 0 < r1, r2 < 1. In this case, each radical will react with the other to give a random copolymer
where the composition in the copolymer is given by the copolymer equation.
(e). r1, r2 > 1. In this case, for both monomers there is a greater tendency to homopolymerize than to copolymerize and the resulting copolymer will have extended blocks
of the individual monomer units. This is not a controlled block copolymer of the type
that may be prepared using a living polymerization system as discussed in Section 1.2.3.
90
Table 1.14. Sequence-length distributions (N1)x for copolymerization of monomers M1 and M2 with reactivity ratios r1
and r2 and feed compositions f1 and f2
r1
r2
f1
f2
(N1)1
(N1)2
(N1)3
(N1)4
(N1)5
(N1)6
1
1
5
5
0.1
0.1
1
1
0.2
0.2
0.1
0.1
0.5
0.75
0.5
0.75
0.5
0.75
0.5
0.25
0.5
0.25
0.5
0.25
0.5
0.75
0.17
0.375
0.91
0.97
0.25
0.19
0.14
0.23
0.08
0.03
0.13
0.05
0.12
0.15
0.01
0.06
0.01
0.10
0.09
0.03
0.02
0.08
0.06
0.07
0.04
r1 r2 0, when the monomer units alternate along the backbone. Of interest is the distribution of the lengths of the extended sequences of M1 and M2. The probabilities of a sequence
of two monomer units, (M1M1) or (M2M2), occurring are P11 and P22, respectively, and that for
the termination of the sequence is for the cross-reaction, i.e. P12 or P21. These may be related to
the reactivity ratios and concentrations in the feed by (Hamielic et al., 1989, Tirrell, 1989):
P11 r1 =r1 M2 =M1 ;
1:83
1:84
1:85
1:86
The number fraction (N1)x of sequences of M1 of length x is given by the product of the
probabilities:
N1 x P11 x1 P12 :
1:87
Similarly,
1:88
Using these equations, it is thus possible to draw up Table 1.14 that gives sequence length
distributions for various compositions and reactivity ratios. It is noted that when r1 and r2
are small (0.10) the system has a narrow distribution and approaches the alternating
sequence since (N1)1 is dominant in the distribution. In contrast, the sequence length distribution for r1 1/r2 5 is very broad, with sequences up to and beyond the hexad, (N1)6,
for both of the feed compositions shown.
Carbon-13 NMR spectroscopy may be used for detailed studies of the microstructure to test
the predictions of the sequence lengths from the simple copolymer equation. Renements
to the equation include the consideration of the effect of the penultimate group in the chain on
the reactivity of the terminal radical with the monomers (Tirrell, 1989). These studies do not
provide information regarding the molecule-to-molecule variation in composition that may
occur, since again these sequence distributions are the average over the whole population.
Some typical values of reactivity ratios for monomers that will copolymerize via a freeradical route are shown in Table 1.15 (Tirrell, 1989, Allcock and Lampe, 1981). These
have been selected in order to span the range of values shown in the preceding tables. It may
be noted that the copolymerization of vinyl monomers with maleic anhydride will invariably lead to alternating copolymers since r1 r2 0. The mechanism of alternating
91
M2
r1
r2
Vinyl acetate
Vinyl acetate
Styrene
Styrene
methacrylate
Styrene
Methyl
methacrylate
Methyl
methacrylate
Maleic anhydride
Maleic anhydride
Vinyl chloride
Acrylonitrile
Acrylonitrile
Methyl
0.240.98
0.040.07
0.290.55
0.520.59
1.032.3
4.056.0
0.00.17
0.460.54
Vinyl chloride
Vinyl chloride
12.425
8.99
0.0050.16
0.07
Acrylonitrile
1.32
0.14
Acrylonitrile
Styrene
0.0
0.00.02
6.0
0.00.097
M 1 + M2 [M1+M 2-] ----(M 1M 2)n----Scheme 1.36. Formation of a donoracceptor complex that then homopolymerizes as a route to
formation of an alternating copolymer.
copolymerization has been studied in detail (Cowie, 1989b) and the best-characterized
example, and the one that seems most applicable for the examples shown above, is through
a 1 : 1 donoracceptor complex. Thus styrene and acrylonitrile are electron donors, which,
when combined with maleic anhydride as an electron acceptor, form a charge-transfer
complex with a characteristic UVvisible absorption spectrum. It has been suggested
(Cowie, 1989b) that this complex may then homopolymerize as shown in Scheme 1.36.
This mechanism accounts for several features of the polymerization of monomers that
form alternating copolymers (Odian, 1991) but in particular the absence of any sequences
of M1M1 or M2M2. If the mechanism of polymerization may be changed, then a pair of
monomers that might otherwise form random copolymers may form alternating copolymers.
Thus, from Table 1.14, methyl methacrylate and acrylonitrile should form a random
copolymer with 25% diads (M1M1) and 14% triads (M1M1M1). However, if the monomers
are polymerized in the presence of a Lewis acid such as zinc chloride, then there is a
tendency towards alternation, which suggests that complexation between the electron-donor
monomer (methyl methacrylate) and the electron-acceptor monomer (acrylonitrile) is
facilitated by the Lewis acid. This will often be restricted to a limited temperature and
composition range, outside which the two mechanisms may occur simultaneously.
Block copolymers
Block copolymers were dened above, and it was noted that there was a tendency for blocks
to form when both of the reactivity ratios, r1 and r2, were greater than unity. This approach
results in a random copolymer with extended blocks of M1 and M2 rather than a pure block
copolymer with well-dened block length and composition. More precise control can be
achieved through living polymerization, since on exhaustion of the rst monomer added, M1,
92
the reactive sites are still able to initiate polymerization. Thus addition of M2 will allow
continued chain growth to a DP given by the ratio [M2]/[initiator]. These sequential methods
may be anionic or living free-radical techniques such as ATRP and RAFT (Section 1.2.3).
The limitations to living block copolymerization arise when the anionic addition site is
unable to initiate the polymerization of the second monomer. An example is the methacrylate anion formed with lithium, which is unable to initiate the polymerization of styrene.
This contrasts with the reverse situation in which the polymerization of methyl methacrylate
may be initiated by polystyryl lithium (Cowie, 1989a). Thus the synthesis of a PSPMMA
block copolymer by an anionic route requires first the polymerization of styrene and then
addition of methyl methacrylate.
If both monomers have high electrophilicity there will be mutual initiation at the anionic
sites. An example is butadiene and styrene, which form industrially important diblock and
triblock polymers by sequential addition of monomer. It has been noted that, because of the
difference between the reactivity ratios of styrene and butadiene in anionic polymerization, if
a mixture of the two monomers is initiated, there is rapid reaction of butadiene so that it is
almost totally polymerized before styrene adds to the chain. The chain sequence then consists
of a copolymer of styrene and butadiene until the latter has been exhausted and then the
styrene homopolymerizes, to give a styrene block. The resulting tapered copolymer has
blocks of polybutadiene and polystyrene separated by a segment with a progressive change in
composition. In another example, the styrenebutadienestyrene triblock copolymer, which
has short segments of styrene and a long segment of butadiene, forms a thermoplastic
elastomer. In this case the high-Tg styrene blocks act as physical crosslinks at ambient
temperature, separating the exible butadiene segment so elastomeric behaviour is observed.
On heating above Tg of polystyrene, the entire system softens, elastomeric behaviour is lost
and the polymer may be melt-processed. The importance of phase separation in this and
related examples is discussed later.
Alternative approaches involve the reaction together of preformed blocks, such as in
telechelic polymers and coupling reactions. The latter allow highly branched architectures
such as star polymers to be formed by linking living anionic chains to a multifunctional core
such as SiCl4 (Cowie, 1989a).
Telechelic polymers are end-capped oligomers in which the capping group may react
with other end groups to form a block copolymer. They have the general formula Y(M)nZ
and are formed by the polymerization of the monomer, M, with a telogen, YZ, so that the
telogen both reacts with the initiator and acts as a chain-transfer agent (Boutevin, 2000).
The resulting oligomer has a low DP (1 < n < 10) and, if the end groups, are vinyl groups, is
referred to as a macromonomer. These may function in crosslinking reactions, as well as
undergoing polymerization.
Living free-radical polymerization methods provide a versatile approach to block
copolymer formation such that polymers unachievable by other routes have been synthesized (Hawker et al., 2001). Using nitroxide-mediated polymerization, it is possible to
prepare a range of block copolymers, two examples of which are shown in Scheme 1.37.
Here (A) shows an example of a sequential polymerization of two different monomers using
the nitroxide IN (as in living anionic polymerization discussed above) and (B) shows
formation of a macro-initiator by using an alkoxy amine XN to terminate the polymerization of a monomer that cannot react by a living free-radical route. This is then reacted by
a living free-radical route using monomer M2 and allows a functional group, X, to be
introduced at the junction between the two blocks.
93
(A):
I N + n M1 I(M 1)n N
+ m M 2 I (M 1)n (M 2)m N
(B):
R + n M1 R (M 1)n
+ X N R (M 1)n X N
m M 2 R (M 1)n X(M 2)m N
94
O
O
H3C
OH
O
S
HO
CH3
DCC
H3C
Ph
CH3
H2C
S
C
Ph
CH
Ph
O
O
H3C
S
O
n
CH3
Ph n
S
C
Ph
Scheme 1.38. Use of RAFT agent for (styreneblock PEO) copolymer formation starting with
poly(ethylene oxide).
M1 + M1 + M 2 M1M2M2M2M2M2
Incomplete conversion of all homopolymer means that the resulting polymer will be a
blend of homopolymer of M1 and block copolymer (M1)n(M2)m.
If two homopolymers are masticated together then terminal macro-radicals (M1) and
(M2) are formed that can cross-terminate to give a block copolymer. In this case there
will be a blend of two homopolymers as well, since formation of the cross-termination
product will compete with recombination of the macro-radicals. There will also be
incomplete chain scission of the homopolymers. This strategy has also been employed to
produce compatibilization and enhanced interfacial adhesion of immiscible homopolymers since the block copolymer will be soluble in each phase and thus able to bridge
the phase boundary. This and other topics concerning polymer blends are discussed in
Section 1.3.
Graft copolymers
Graft copolymerization has some of the features of block-copolymer formation except that
the radical centres are not at the end of the chain but in-chain. The polymer to which the
graft is attached is the backbone polymer. Cowie (1989a) has recognized three different
ways in which graft copolymers may be formed:
grafting from the backbone, in which a backbone reactive site is created and this initiates
polymerization, as described above;
grafting onto the backbone, in which a separately initiated polymer chain reacts with a
site on the backbone polymer and becomes grafted to it; and
coupling of a macromonomer that has a known chain length to a single reactive group that
can then react with the active sites on the backbone to create branches of precise length.
95
Of these, the rst is the most common and the graft copolymer is formed on the backbone
polymer by creating the in-chain radical by means of
Ionic routes such as formation of the polymer anion by reaction with a strong base or the
direct reaction of a polyamide with sodium are less likely to be used in reactive processing
than is free-radical initiation. The process of self-graft polymerization by chain transfer to
polymer, when it occurs in a single monomer/polymer system during polymerization, is an
example of chain branching that is discussed in the next section.
96
CH3
CH3
C O O C
CH3
CH3
CH3
C O
CH3
2 CH3 + 2
CH3
C O
Scheme 1.39. Chemical structure and decomposition of DCP to form either cumyloxy radicals
or methyl radicals and acetophenone.
of the alkoxy radical to undergo b-scission to form the methyl radical at higher temperature (Moad, 1999). Other reactions of the alkoxy radical that compete with backbone
radical formation are addition to monomer and hydrogen-atom abstraction from monomer,
but it has been noted that these are less important at elevated temperatures and at the low
concentrations of monomer present than are the reactions with the substrate (Russell,
2002). Other factors noted in the choice of a free-radical initiator (Moad, 1999) are the
following:
the solubility in the polymer melt, and the partition coefcient in the different phases if it
is a multicomponent blend;
the physical form and vapour pressure at the reaction temperature;
the dependence of the free-radical escape efciency ( f ) on the viscosity; and
the tendency towards (radical) induced decomposition and other side reactions that lower
the radical yield.
The formation of the backbone radical site depends on the nature of the polyolen.
The presence of tertiary carbon atoms as in polypropylene and branch points in lowdensity polyethylene, LDPE, would be expected to provide the preferred site for
hydrogen-atom abstraction. However, it has been noted (Russell, 2002) that the difference between the reactivity of tertiary and secondary carbon atoms decreases at elevated
temperatures.
Addition of monomer to backbone radical. The addition of the monomer to the backbone
radical R may be considered as equivalent to the copolymerization of M2 with the radical
from monomer M1. Thus, in the terminology of the copolymerization equation the reaction
rate coefcient of importance is k12 and the estimates of the magnitude of this place it
sufciently high to ensure that the monomer will add to the radical site in polyethylene
(Russell, 2002). The monomer will add preferentially to tertiary sites over secondary sites in
ethylene copolymers with a high concentration of propylene, but in LDPE there seems to be
little preference for branch points.
Propagation. The reaction of the radical RM2 with further monomer M2 will depend
on the availability of monomer versus other competing termination or chain-transfer
reactions. The kinetic chain length of the graft depends on the ratio of these rates,
with the most signicant process at the higher graft temperatures being chain transfer to
another polymer backbone site. This may restrict the graft chain length to a few monomer
repeat units before either intermolecular or intramolecular chain transfer occurs.
The availability of intramolecular sites tends to favour these reactions unless bulky
substituents inhibit the process. Another feature of a high-temperature reaction is
the likely importance of depropagation limiting the graft length. Thus when the graft
polymer is near the ceiling temperature, the equilibrium will favour monomer over
polymer. This has an important application in the grafting of maleic anhydride (MA) to
polyolens, which is an industrially important reactive processing operation. The ceiling
97
R + R R-R
RM2 + R RM2R
RM2 + RM2 RM2M2R
Scheme 1.40. Various possible crosslinking reactions during grafting.
temperature of the polymer is 100 C so the graft reaction will lead to molecular addition
and there is no homopolymer formed. This is an advantage in ensuring that there are no
homopolymer side reactions, which can be signicant in grafting styrene for example
(Russell, 2002).
Termination. The presence of small reactive radical species such as the methyl radical
produced by fragmentation of the DCP initiator means that there may be a range of termination
routes available. The bimolecular termination of graft radicals on different chains may result
in a crosslink unless they disproportionate. The termination of the primary backbone radicals,
R, will be competitive with monomer addition at low concentrations of M2 and again produce
a crosslink. The possible competing crosslinking reactions are shown in Scheme 1.40.
98
Back-biting
PE chain
H
H
CH2
CH
CH3
H2C
CH2
CH
CH3
n-butyl branch
99
Ph
S
Ph
S
Ph
Ph
Ph
S
Ph
Ph
Ph
S
Ph
S
S
S
Ph
Scheme 1.42. Multi-arm RAFT agents for the formation of star polymers. After Chong (1999).
S
CH
Ph
Ph
CH
Ph
Ph
Ph
Ph
Ph
H2C
CH
CH
Ph
Ph
Ph
Ph
Ph
Scheme 1.43. Branching from RAFT polymerization with a polymerizable thio-ester. After
Wang (2003).
that contained a dithiobenzoate C( S)S moiety at each branch point was obtained
(Scheme 1.43). After cleavage of the branches with amine, the molar mass of the branches
was narrow and increased linearly with monomer conversion, which is consistent with
living free-radical polymerization.
It has been noted (Guan et al., 1999) that polyethylene is not just branched (as in
linear low-density polyethylene, where branching is controlled by the copolymerization of
1-hexene, or in low-density polyethylene, where it is uncontrolled due to back-biting) but
may, under certain circumstances, also be hyperbranched. The mechanism for hyperbranching is to create the branch point by controlled isomerization of the active site by an
appropriate choice of both co-ordination catalyst and pressure.
1.2.5
100
Difunctional crosslinkers
The most widely studied addition-polymerization reaction for crosslinking is the freeradical polymerization of difunctional monomers (such as divinyl benzene and ethylene
glycol dimethacrylate) in the presence of the corresponding monofunctional monomer.
This may be thermally or photochemically initiated, and the latter application is widely
used for coatings and dental composites. This is shown in Figure 1.26.
The polymerization reaction results in the formation of an innite network so
the system gels and then with further reaction vitries as the glass-transition temperature, Tg, of the network reaches the reaction temperature. The polymerizing system,
which had become diffusion-controlled between gelation and vitrication, then ceases
chemical reaction until it is heated further to T > Tg. The rates of gelation and vitrication depend both on the reactivity and on the concentration of the difunctional
monomer.
The system may be modelled initially by considering a system in which all
the functional groups are the same type chemically and so have the same reactivity. Some
typical systems would be monomer styrene (PhCHCH2) and crosslinker divinyl benzene
(CH2CHPhCHCH2), and monomer methyl methacrylate (CH2CMeCOOMe)
and crosslinker ethylene glycol dimethacrylate (CH2CMeCOOCH2CH2OOCMeCCH2).
Thus, labelling the monomer M1 and the difunctional crosslinker MM and taking
the reactivity ratios r1 r2, the number of crosslinks is the number of molecules of
MM in which both double bonds have reacted. The difunctional monomer may
also homopolymerize as well as participate in crosslinking, i.e. only one of the
M groups reacts. The respective concentrations of reactive groups are [M1] and
[M2] 2[MM].
To determine the extent of reaction of the monomers at gelation, pc, it is necessary to
determine the critical concentrations of M1 and M2 when the number of crosslinks per chain
is 0.5. If the polymer had not contained the difunctional reagent, it would have attained a
weight-average degree of polymerization DPw under the conditions of polymerization. The
gelation condition is then
pc M1 M2 =M2 DPw :
1:89
Even for very low concentrations of crosslinker [MM] 5 103 mol/dm3 the gel point
occurs at extents of conversion, pc, of 12%20%. The process of network formation is
shown in Figure 1.26.
101
CH2
H3C
C
O
H2 H 2
C C O
CH2
CH2
C
CH2
HC
CH3
CH
O
divinyl benzene
DVB
Figure 1.26. A schematic diagram for the formation of a crosslinked network by the polymerization
Other cases with different reactivity ratios as well as other functional group reactivities
have been considered (Odian, 1991).
102
H2C
HC
O
O
C
O
C
H
C
H
H2C
CH
O
C
CH2
CH2
H2 H
C C
HC
CH
Ph n C
Ph
O
O
H2C
H2C
CH
O
HC
O
H2 H
C C
CH2
CH
Ph n C
Scheme 1.44. Crosslinking reaction of styrene with fumarate group in an unsaturated polyester resin.
synthesis of these resins was shown earlier in Scheme 1.7 of Section 1.2.2.) The
crosslinking reaction differs from the system that uses a polyfunctional reagent (e.g. DVB in
Figure 1.26) in that the crosslinker is a monofunctional monomer, styrene, that copolymerizes with the in-chain fumarate groups that were previously step-copolymerized in a
polyester oligomer, as shown in Scheme 1.44.
The properties depend on the crosslink density, the length of the crosslink as well as the
molecular structure of the crosslinker. When styrene is used, the degree of polymerization,
n, at a crosslink is low (about 12), so the network is stiff. The low DP reects the low value
of the reactivity ratio of styrene with the fumarate group r1 r2 0, so the system tends to
alternation and thus can form only a short crosslink (Odian, 1991). The nal modulus also
depends on the number of fumarate groups along the polyester backbone between the
saturated aromatic acid groups, since this controls the crosslink density. It is noted (Odian,
1991) that, if methyl methacrylate is used as the crosslinking monomer, then longer chains
are formed, so the network has a lower modulus.
103
Crosslinked polyethylene
The crosslinking of polyethylene is of great practical importance in order to increase the
softening temperature and maintain dimensional stability at elevated temperature. There are
various ways of achieving this through using chemical initiators as well as radiation to
create macro-radicals that may recombine intermolecularly or undergo reaction with end
groups in the polyolen to create a crosslink. Scheme 1.45 shows a sequence of radical
reactions that lead to H-type crosslinks in polyethylene. However, the available free
volume for such a reaction to take place is limited and the actual crosslink need not be of
H-type but may instead involve reaction of functional groups such as vinyl at chain ends.
This produces a Y-type crosslink as proposed for c-radiation-initiated crosslinking in
UHMWPE (Bracco et al., 2005), as shown in the third reaction of Scheme 1.45.
An alternative approach is to graft the polyethylene with a reactive monomer that may
then undergo a separate reaction to crosslink the chains. An example is the hydrolytic
crosslinking of triethoxy silane groups that had previously been grafted onto PE during
reactive processing. Scheme 1.46(a) shows the free-radical-initiated grafting reaction of
vinyl trimethoxy silane (VTMS) which is performed in the extruder, for example in the
coating of a metal cable. This is followed by the slow reaction with water, Scheme 1.46(b),
in a separate step that is diffusion-controlled because reaction is occurring in the rubbery
state (unlike the grafting reaction in the molten polymer). The condensation reaction
to give the crosslinks may be catalysed by an organotin compound such as dibutyltin
dilaurate.
104
H
C
R
H
C
H
C
C
H
C
H
H-crosslink
CH2
CH
Y-crosslink
Terminal vinyl
(a) Grafting
Me O
Me O Si
Me O
C
H
CH2
Me O
Me O Si
Me O
C
H2
VTMS
H2O
(b) Hydrolysis and crosslinking
C Si
H2
OH
OH
OH
HO
HO Si
HO
-H2O
OH
Si
OH
O
OH
Si
OH
C
H2
+ MeOH
105
In this case the nal crosslinking reaction is not an addition polymerization reaction but
the hydrolysis and condensation reaction of alkoxy silanes incorporated through a separate
step of addition graft copolymerization.
1.3
106
Two-phase
LCST
T
One-phase
One-phase
UCST
Two-phase
Composition ()
Figure 1.28. A phase diagram for a (hypothetical) polymersolvent or polymerpolymer system
showing both lower (LCST) and upper (UCST) critical solution temperatures. The boundary
may be determined by locating the cloud point as a function of temperature for a xed composition
as the system moves from being a single-phase system to being a two-phase
system and vice versa.
1.3.1
Miscibility of polymers
The understanding of the formation of miscible and immiscible polymer blends requires the
application of the principles of phase chemistry. A miscible blend may be regarded as a
solution of one polymer in the other. The thermodynamic criteria for the miscibility of
liquids are well known and may be applied to polymers as a rst approximation. The added
complexity comes from the long-chain nature of polymers. In addition to the entropic
factors there are kinetic factors to be considered. Since in reactive processing the reactions
are occurring within a short time, they will very often be a long way from equilibrium.
107
solvent
A
B B B
solvent
A A A
polymer
B B B
solvent
B A
B B
B B
polymer
polymer
Figure 1.29. Illustration of the difference in mixing of two solvents, A and B. Mixing and
solvation of a polymer, [AA]n by a low-molar-mass solvent, B, and mixing of two polymers
[AA]n and [BB]n. The segments that are not solvated are highlighted. Adapted from
Olabesi et al. (1979).
either a UCST or an LCST, and it has been noted (Sperling, 2001) that, because of the effect
of polymer chain length on the entropy of mixing of the components, a UCST is seen with
shorter chains and an LCST with higher-molar-mass polymer blends.
It is noted that ideal solutions occur rarely even for two liquids, since this requires that there
be no volume change (and thus no enthalpy change) on mixing. The dissolution process arises
solely due to the entropy change arising from the increased degree of freedom on mixing. A
schematic way of comparing the entropy change on mixing two liquid solvents, a polymer and
a solvent, and two polymers is shown in Figure 1.29 (Olabisi et al., 1979). This rstly shows
the situation when the segments A and B are uncorrelated, i.e. low molar mass. The unrestricted mixing process of the liquids ensures that every A segment has a B segment alongside
it, but this becomes progressively more difcult as rstly the A and then the B segments are
covalently linked in the polymer A solvent B (second case) and the polymer A polymer B
(third case) systems as depicted. The segments that are isolated from the other component are
circled and clearly are greatest in the polymerpolymer example.
This combinatorial entropy of mixing for low-molar-mass species is given by
DS M kB N1 N2 x1 ln x1 x2 ln x2 ;
1:90
1:91
108
where the mole fraction xi has been replaced by the volume fraction ui. Here i 1 for the
solvent and 2 for the polymer.
The enthalpy of mixing DHM is also calculated for the solventpolymer interactions
taking into account the occupancy of the lattice sites by the polymer. If w is the energy
of interaction between solvent and polymer and z is a co-ordination number, then (Olabisi
et al., 1979) the free energy per unit volume is
DGM =V kB Tu1 =V1 ln u1 u2 =V2 ln u2 u1 u2 =V1 ;
1:92
109
Polymer
Solubility parameter, d
(MPa)1/2
Poly(ethylene)
Poly(vinyl chloride)
Poly(ethylene terephthalate)
Poly(vinylidene chloride)
Poly(hexamethylene adipamide) (nylon-6,6)
16.2
19.9
21.9
25.1
28.0
electron-donor groups charge transfer can occur. This is often seen as the formation
of a coloured complex, indicating the formation of a new absorption band due to the
new species.
The extent to which these polymersolvent interactions may control the miscibility of a
system has been described for some time in terms of the solubility parameter, d. In simple
liquids, this has a direct link to the cohesive energy density (DE/V) of the material:
d DE=V 1=2 ;
1:93
where DE is the molar energy of vaporization of the solvent and is thus a measure of the
forces of interaction between the molecules. The temperature dependence of the vapour
pressure enables this to be determined for a low-molar-mass solvent. The units of the
solubility parameter will be (J/cm3)1/2 or MPa1/2.
In order to extend the concept to the polymer component, an indirect method of measurement is required. This is done by dissolving the polymer in a range of solvents and
measuring the viscosity. Since the maximum expansion of the chain on dissolution will
occur with a solvent having the same value of d as the polymer, and thus give the highest
viscosity for a particular concentration, this is the value taken for the solubility parameter of
the polymer. This approach may also be applied to the segments of a crosslinked polymer
that may be swollen by the solvent. Some values for polymers determined in this way are
given in Table 1.16. In order to gain an understanding of the different forces that make up
the cohesive energy density, d2, the separate components may be considered:
d2total d2nonpolar d2polar d2hydrogen bonding :
1:94
Considering Table 1.16, only the rst polymer, polyethylene, has non-polar contributions
alone; the next three have also polar components and the last, nylon-6,6, has contributions
from all three forces. The largest solubility parameter for this polymer also corresponds to
the highest melting point and stiffness, reecting the importance of cohesive energy density
as a measure of intermolecular forces.
Solubility parameters may also be calculated by a molar-additivity approach if the
chemical composition is known. Tables of group molar attraction constants, G, obtained
from molar heats of vaporization (Sperling, 2001) may be used. The solubility parameter, d,
P
is then given by summing these for all species ( G) in the monomer repeat unit of molar
mass M and density q:
X
dq
G =M:
1:95
110
2G/2 = 0
G( )
[at T0]
1
spinodal
Two Phases
T0
One Phase
0
Figure 1.30. A plot of the change in Gibbs free energy, G, as a function of composition, u,
for a system that a LCST that shows spinodal as well as binodal decomposition on phase separation.
Adapted from Kwei and Wang (1978).
1:96
where b1 is the lattice constant of entropic origin and is 0, V1 is the volume fraction of the
solvent and R is the gas constant. Typical values of for non-polar polymersolvent
combinations at low concentration of polymer are within the range 0.30.4 (Sperling, 2001).
If 0.5 the solvent is a h-solvent at that temperature.
While the concept of solubility parameter has a physical basis, it is essentially an
empirical parameter that may be related to the FloryHuggins parameter. As such it has
value in understanding polymer blends in that the conditions for phase separation may be
rationalized.
More detailed understanding of the process of mixing of polymer chains has
been developed from theories of equations of state. While these are founded in the
equations of state for gases (starting with the fundamental equation pV nRT and then
applying corrections for molecular volume and forces of interaction as discussed
earlier in this section), the major difference is in the inclusion of free volume in the
system. Detailed discussions of this approach may be found in Olabisi et al. (1979) and
Sperling (2001).
1.3.2
111
Phase-separation phenomena
The separation of two, initially miscible, polymers will occur under conditions of
temperature and composition given in a phase diagram such as Figure 1.30. It may be
recognized that the process of passing from one phase to another when the temperature
change moves the system across a phase boundary cannot occur instantaneously. The most
common example is the growth of the solid phase from the liquid. In this case nucleation of
the crystalline state is followed by growth of the phase in three dimensions. The formation
of spherical domains provides the minimization of the surface energy in the new system. In
liquidliquid transitions the process of spinodal decomposition leads to phase separation.
This is also seen in polymer blends.
Figure 1.30 shows a phase diagram for such a system that has a lower critical
solution temperature; the change in free energy as a function of composition is also
shown (Kwei and Wang, 1978). This allows the condition for spinodal decomposition
to be identied as the points of inection on the free energy (G)composition (u)
diagram. The condition for this inection is @ 2 G=@u2 0: Between these spinodal points
the value of the second derivative is negative and the system is unstable. In this region,
small uctuations in concentration lead to phase separation by a spontaneous and continuous process. The line corresponding to @ 2 G=@u2 0 may be constructed on the temperaturecomposition phase diagram for all temperatures and is shown as the dotted line.
Between this spinodal line and the binodal line of the phase diagram, the system is metastable. It requires larger uctuations in composition to effect phase separation by nucleation
and growth.
In spinodal decomposition, phase separation does not arise due to nucleation and growth
of spherical domains, but rather arises from uctuations in concentration such that phase
separation arises in interconnected regions. Sperling (2001) has highlighted the differences
between the two methods of phase separation in terms of
activation energy: nucleation is activated; spinodal decomposition is not;
diffusion coefficient: nucleation is positive (migration towards the nucleation site);
spinodal decomposition is negative (migration away from the domain); and
connectivity: nucleation produces discrete, spherical domains; spinodal decomposition
produces (initially) interconnected domains with no sharp interface between domains
until later stages of phase separation.
Figure 1.31 contrasts the two processes and the local changes in concentration as phase
separation takes place. Also the different phase structures for nucleation and growth
(approximately spherical domains) and spinodal decomposition (interconnected phases) are
shown. The difference in the sign of the diffusion coefcient, D, is fundamental to the two
processes.
In the case of nucleation, there is a small nucleus that then grows, fed by a positive
concentration gradient, so D is positive. In spinodal decomposition the domains form by
small-amplitude modulation of the concentration, which, because the energy barrier is low,
leads to formation of the new phase. The sign of the diffusion coefcient is negative since
the growth occurs against the concentration gradient created by the small uctuations. The
formal theory of spinodal kinetics is well developed (Kwei and Wang, 1978) and, since the
actual morphology may be affected by external factors such as applied shear, this is the only
unambiguous test of spinodal decomposition.
112
Figure 1.31. The development of the characteristic morphology of polymer blends that have
phase-separated by nucleation and growth compared with the case of spinodal
decomposition. The local uctuations in concentration leading to phase separation are also
shown. Adapted from Olabisi (1979).
The further development of the separate phases from spinodal decomposition occurs by a
viscous-ow process that may lead to layered structures or coarsening, giving other partially
dispersed arrangements of the phases. The simple phase diagram such as Figure 1.28 with
an LCST has been modied by inclusion of the separate regions for spinodal decomposition
and nucleation (Figure 1.30) since these lead to different morphologies. These morphologies
may readily be studied by electron microscopy.
113
114
Interface
PHASE A
PHASE B
Interface
Figure 1.32. A schematic diagram showing the phase-separated structure of a block copolymer
comprising chemically distinct blocks and a (theoretical) sharp interface between the phases.
On changing the length of the two blocks, the morphology changes from spheres to
cylinders to lamellae as follows:
spheres: short blocks up to u 0.25 separate as spheres in the continuous phase of the
longer block
cylinders: unequal-length blocks with u 0.250.4 separate as cylinders of the minor
component in the longer block as continuous phase
lamellae: equal-length blocks with u 0.5 form alternating layers with the interface
separating the layers
In addition to these morphologies, perforated layers, bicontinuous layers and other
unstable morphologies have been described (Sperling, 2001). It has also been noted that,
unlike polymer blends that become less miscible at higher temperatures, block copolymers become more miscible due to the temperature dependence of the FloryHuggins
parameter, .
The importance of these block copolymers is realized when they are used as minor
components with normally immiscible homopolymers (each having the composition of one
of the blocks). In the absence of the block copolymer (compatibilizer), the resulting twophase system might not be of practical usefulness. The often high interfacial tension
between the two phases results in poor dispersion of the minor phase in the other, continuous phase. On processing, this will lead to macroscopic separation, so the result is a
crumbly, useless material. The presence of a minor third component, the block copolymer,
can enhance adhesion between the two components and so stabilize the morphology of the
system. This process is discussed in more detail later.
This compatibilization relies on the chemical bonding between the two blocks, so
alternative routes such as grafting may be used to achieve the same effect. Thus the usual
route to HIPS is to graft the rubber toughener with styrene and the resulting graft
copolymer is then blended with polystyrene. The resulting morphology as revealed by
TEM is shown in Figure 1.33 (Sperling, 2006). This shows that the morphology is quite
complex, with the phase-separated domains containing polystyrene (PS, light zones)
115
Figure 1.33. Transmission electron micrograph of HIPS after staining with OsO4 to enable
differentiation of the PB domains from the PS. Micrograph courtesy of Bob (J. S.)
Vastenhout of Dow Benelux B.V., The Netherlands.
grafted to the elastomer (dark zones of polybutadiene, PB, from OsO4 staining of
unsaturation) and the particles adhering strongly to the continuous polystyrene phase. This
has been described as a phase-within-a-phase-within-a-phase organization (Sperling,
2006), and in HIPS the continuous phase is poly(styrene). Development of this structure is
controlled by mixing during polymerization, and it has been noted (Paul, 1978) that the
role of the compatibilizer is multifunctional in controlling the dispersion of the minor
phase and the adhesion between the phases. The comparison of the impact properties
shows that there is little gain in toughness of the ungrafted blend, compared with
toughening by a factor of 45 for the HIPS containing well-dispersed rubber particles with
a diameter of 0.5 lm.
In many systems, such as epoxy resins, the rubber toughener may be soluble in the other
phase, such as epoxy resin, so the phase separation must be achieved during cure. This is an
example of phase separation during reactive processing.
116
O
HO
H2
C C
H
C
H
H2
C
H2
C CH
C
y
x
C
OH
CTBN
O
R
O
OH
R3N
NHR3
O
R
HC
CH2
O
R
H2 H
C C
O
NHR3
Scheme 1.47. The structure of CTBN elastomer and reaction of the terminal carboxyl groups with
epoxy in the presence of amine curing agent.
crosslinking and three-dimensional network formation. In the latter process it has been
recognized that there may be three possibilities (Pascault et al., 2002):
a dispersion of elastomer particles in the matrix material
a phase-inverted dispersion of the matrix as particles in an elastomer-rich continuous
phase
two co-continuous phases
The separate phases will be rich in one component but may have the other present as a
minor component. In order to control compatibility the elastomer may have reactive end
groups to enhance interfacial adhesion. A common example in epoxy-resin technology is
the carboxy-terminated butadieneacrylonitrile copolymer (CTBN). The structure is shown
in Scheme 1.47. In this resin the solubility in the epoxy resin is conferred by the acrylonitrile group, and an increase in the fraction present decreases the upper critical solution
temperature, with 26% acrylonitrile conferring total miscibility of CTBN with a DGEBAbased epoxy resin (Pascault et al., 2002).
The process of phase separation during cure arises from the change in the phase
diagram as the cure reaction of the epoxy resin progresses. This is shown schematically in
Figure 1.34 (Pascault et al., 2002) for a system with an upper critical solution temperature
(UCST) in which the lower curve represents the system miscible at room temperature, with
the fraction uR0 of elastomer corresponding to the initial composition of the rubberepoxyresin system before any cure reaction has taken place.
The material is raised to the cure temperature Tcure and reaction commences. There is an
increase in the molar mass of the epoxy-resin network as the oligomers react, so that in the
equation for the entropy of mixing there will be a decrease in the number of molecules, N2,
and thus the entropic component (DSM) to the free energy of mixing (DGM). The resultant
decrease in the solubility of the rubber in the network produces phase separation before the
resin has undergone gelation (characterized by the formation of an innite network, i.e.
117
Rubber dissolved
in uncured resin
Crosslinking
shifts phase
boundary
Tcure
pCP
pCP
p=0
p=0
R0
Figure 1.34. Change in the phase boundary (UCST) during the cure reaction for a system of a rubber
(e.g. CTBN) dissolved in an epoxy resin. Toughening requires that phase separation be achieved
during the cure cycle. After Pascualt et al. (2002).
Mw ! 1). This new phase boundary for the cloud point of the rubber is shown as the upper
curve in Figure 1.34 and will correspond to an extent of conversion pCP < pgel. There will
now be the formation of the rubber (b-)phase dispersed in the thermoset (a-)phase and, as
shown in Figure 1.34, the theoretical composition of the b-phase will be given by the
position on the cloud-point boundary as shown. (It has been noted that this is not strictly
correct due to polydispersity effects and the actual composition lies outside this line
(Pascault et al., 2002)).
This is a simplication of the process occurring in a curing resin-hardener system and a
detailed discussion may be found in Pascault et al. (2002), Williams et al. (1997) and
Inoue (1995). The main parameter that it is important to control in the reactive phase
separation is the diameter of the elastomer particle. This is because the toughness of the
resulting network is controlled by the energy-absorbing mechanisms such as particle
cavitation and rubber bridging of cracks. Also of importance is the limitation of the effect
of the rubber dispersed phase on the critical properties of the cured epoxy resin such as the
stiffness and Tg. This will be affected by the extent to which the rubber dissolves in the
matrix-rich phase.
The composition of the a- and b-phases will continue to change with conversion beyond
pCP, with them becoming richer in their main components. However, there will be lowmolar-mass species of the other phase present in each, which may lead to secondary phase
separation, e.g. the presence of a thermoset phase with a lower extent of conversion inside
the rubber particles. Such sub-structures may be seen by TEM of sections of the cured
thermoset. The nal morphology that develops will then depend on the temperature prole
used in cure and post-cure, but is largely controlled by that achieved by the onset of gelation
of the a-phase.
If the cloud-point curve is plotted as a function of conversion at constant cure temperature, the possible morphologies may be recognized and the composition of the original
118
pgel
Spinodal
Cloud point
curve
Figure 1.35. Phase separation of the rubber phase from the epoxy phase plotted as conversion, p,
mixture necessary to achieve a b-phase rich in rubber readily determined. This is seen in
Figure 1.35 for an (epoxy resin rubber) system such as that which was shown in
Figure 1.34 in temperaturecomposition co-ordinates for just two conversions (p 0 and
p pCP) now plotted in conversioncomposition co-ordinates for a single cure temperature.
The phase boundaries may be recognized that
dene the stable region for the dissolution of the two components in each other (below
the binodal line);
dene a metastable region between the binodal and the spinodal line corresponding to
nucleation and growth of spherical domains (b-phase) dispersed in the predominant
phase (a-phase) (where the composition of the elastomer is below the critical
composition, point C, e.g. point A, the morphology will be approximately spherical
rubber particles in the thermoset; in contrast, when the elastomer composition is above
point C, e.g. point B, the composition will be phase-inverted, i.e. spherical thermoset
particles dispersed in an elastomer continuous phase); and
dene an unstable region within the spinodal curve (this is directly accessible at point C,
the critical point, and will lead to a co-continuous morphology due to spinodal demixing
(Section 1.3.2)).
All of the above morphologies may be recognized by SEM of the resulting networks after
reactive processing. For a toughened network the morphology accessed from point A is the
one desired, and this is achieved by ensuring that a low composition of the elastomer (below
point C in Figure 1.35) is added to the epoxy resin before cure.
An alternative approach to using a rubber such as CTBN as the disperse phase is to use an
engineering thermoplastic such as polyether sulfone or a polyimide in which the mechanical
properties and Tg are more closely matched to those of the resin. These systems exhibit
LCST behaviour, so the cure temperature Tcure is located below the initial miscibility gap
119
and the phase boundary shifts to lower temperature with conversion until the cloud-point
conversion is reached. It is noted that a phase-inverted structure may occur in the course of
phase separation of thermoplastics such as polyether sulfone (PES) from an epoxy resin
during the early stages of cure because of the high molar mass of PES compared with that of
the elastomers such as CTBN (Williams et al., 1997). This can contribute to improved
fracture toughness, provided that the nal phase-separated structure with the dispersed
thermoplastic is achieved.
The thermodynamics of the phase-separation process depends both on changes in the
FloryHuggins parameter, , as the chemistry changes and on the entropic effects as the
molar mass increases. Consequently, any functionalization of the original thermoplastic (or
elastomer) in order to enhance interfacial adhesion between the phases will affect the phaseseparation trajectory as given in Figure 1.35 and thus the nal morphology. At the extreme
of not achieving particles of the correct size distribution, the efciency of the system in
toughening the epoxy-resin phase may be compromised. It has been noted (Williams et al.,
1997) that the phase-separation process described is thermodynamic in origin and does not
depend on any chemical reactivity between the functional groups of the dispersed and
continuous phases. The kinetics of cure, however, have a role to play insofar as it is
important that the rate of phase separation exceeds the rate of the crosslinking reaction that
leads to gelation. If the reaction rate is too high, the trajectory will pass to the spinodal
region and a co-continuous structure will be obtained.
In a fundamental study of the factors affecting the growth of the rubber domains in a
CTBN-toughened DGEBA epoxy resin cured by piperidine (Manzione and Gillham, 1981)
the kinetics of phase separation were linked to the diffusivity, DAB, of the rubber (A)
dissolved in the epoxy resin (B). The relevant dependence on the molar volume of the
rubber (proportional to the radius RA of the rubber molecules when dissolved in the epoxy
resin) at the viscosity gB for the temperature T of reaction is given by the StokesEinstein
equation:
DAB kB T=6pRA gB :
1:97
From this, the time scale for diffusion of the rubber particles to coalesce to form a particle
may be estimated from the SEM data showing the distance, L, between phase-separated
particles. This was found to be of the order of 10 lm in the DGEBACTBN system
(Manzione and Gillham, 1981). Since
tdiff L2 =2DAB ;
1:98
tdiff is of the order of minutes, which indicates a lower bound for the processing window to
allow the kinetics of phase separation to take place. The processing window may be
regarded as tgel tcl: the difference between the time to cloud point (tcl) and the gel time
(tgel), which in the system studied was >100 min at the cure temperature. This is a very large
processing window and in industrial processing this may need to be reduced by at least an
order of magnitude to achieve efcient production (bearing in mind, however, that slow
reaction times generally produce better control of the cure exotherm and thus minimize
shrinkage stresses).
From the above discussion, the choice of both composition and cure schedule is essential
in the successful reactive processing of a thermoset toughened by the separation of dispersed elastomeric or thermoplastic (spherical) domains during the curing reaction. The
120
cure kinetics and thus tgel may be controlled through the choice of curing agent as well as
the stoichiometry in addition to the temperaturetime prole. The effect of the phase
separation of a thermoplastic polyimide toughener on the cure rate of a DGEBA resin cured
with an aromatic diamine has been studied (Bonnet et al., 1999), and the main effect is the
change in the local concentration of the reacting components as phase separation occurs. At
concentrations of polyimide above 30 wt.% (which is used in some compositions) there is a
sudden acceleration in the global reaction rate. This is attributed to the formation of an
epoxyamine-rich phase that has a higher reaction rate. Parallel studies of the rheology of
the phase-separating system (Bonnet, 1999) showed that the viscosity prole reected the
nature of the phase-separation process for compositions corresponding to A, B and C on
Figure 1.35. For low concentrations of thermoplastic (point A) there was a decrease in
viscosity on phase separation, followed by a rise to gelation. For the critical composition, C,
there was a sharp increase followed by a decline and then a steady increase to gelation
attributed to the percolation of the thermoplastic-rich phase in the co-continuous structure.
At composition B, where phase inversion is possible, there was a steady increase in viscosity to gelation and then vitrication. For DGEBA cured with an aromatic diamine the
compositions for A, C and B correspond to 10%, 20% and 33% thermoplastic polyetherimide, respectively.
Hyperbranched polymers (HBPs) provide an approach intermediate between thermoplastics and CTBN elastomers as toughening agents (Boogh et al., 1999) insofar as they
have a large increase in fracture toughness for a low additive level (about 5 wt.%). The
branched structure (see Figure 1.18 and also Section 1.2.4) results in a high surface functionality and a low viscosity (Aerts, 1998) because of the compact molecular architecture
(Lederer et al., 2002). In terms of the above equations, the low values of both RA and gB
result in efcient phase separation. It has also been noted that it is this unique chemical
functionality of the shell that enables a wide processing window to be achieved without
compromising the ultimate properties. The near-spherical architecture of the hyperbranched
molecule is a factor in achieving optimum nucleation and growth rates of the particles and
avoiding spinodal decomposition. The good toughening properties at low levels of addition
were attributed to the high connectivity in the phase-separated particles (Boogh et al.,
1999). It has been shown that there is a low concentration of the continuous (epoxy) phase
dissolved in the HBP particle, but this appears to be graded in composition, producing an
effective interface with a gradation in modulus, rather than a step function. In a related study
the improvement afforded by HBPs over the comparable linear polyester was questioned
(Wu et al., 1999) and the only advantage was that noted above, viz. the ability to tailor the
interfacial behaviour.
Even when phase separation of a rubber such as CTBN or a HBP is optimized to produce
an effective material, there will be the small fraction of the disperse phase that is soluble
in the continuous phase, and trapped there at gelation. As noted earlier, this can affect the Tg
if an elastomer is used, but is not a problem with the high-performance thermoplastics
such as PES.
One way of overcoming this is to have the second phase already separated before cure,
i.e. the toughener is insoluble in the epoxy resin before cure. Examples are found in
toughened thermoplastics, where coreshell particles that optimize adhesion and compatibility and have a particle-size distribution to maximize toughness have been synthesized.
These particles can be added at the desired volume fraction to achieve toughness without
compromising performance rather than relying on the phase trajectory to achieve the desired
121
SHELL
SHELL
A
SHELL
CORE
Figure 1.36. A schematic diagram of an ABA block copolymer and the coreshell morphology of
toughening particles preformed for dispersion in an epoxy resin.
volume fraction and size distribution. Figure 1.36 illustrates the morphology and the
chemical composition of some coreshell particles. Typical systems are ABA block
copolymers that have A as poly(styrenebutadiene) and B as poly(methyl methacrylate).
The A block forms a particle that is insoluble in the resin and the B block is soluble at all
extents of conversion, so giving compatibility.
One disadvantage is the need to ensure that there is a uniform dispersion of the particles
in the epoxy resin up to gelation and the corresponding much higher viscosity that the twophase system will have even before cure. The difculty in processing is heightened if
agglomeration occurs. The compatibility and thus high interfacial adhesion between the
phase-separated materials is vital in the application of the nal material since the efcient
transfer of stress between the toughener and the more brittle thermoset matrix is vital in the
achievement of a high value for fracture toughness, for example. The process of compatibilization has become an important area in the reactive processing of polymers. This is
considered in the next section before the effect of shear on the process of phase separation is
discussed.
122
is complex, involving a range of energy-dissipation mechanisms that can occur at the crack tip,
such as cavitation in the rubber or in the rubbermatrix interface, as well as localized shear
yielding in the matrix due to stress concentration in the rubbery particles (Kinloch, 1985).
The actual reaction between the carboxyl group on the CTBN and the epoxy resin in an
amine-cured epoxy system is complex due to the low concentration of carboxyl groups
compared with that of amine groups (a ratio of about 1 : 10), so the rst step in the reaction
is the formation of a salt with the amine which then proceeds to react with the epoxy groups
(Scheme 1.47). As noted earlier, there will be solubility of the rubber in the resin, so there
will be a copolymerization reaction as the matrix cures and the conditions for phase separation are achieved (Bucknall, 1977). It was recognized that this interfacial-reaction
chemistry did not control the process of phase separation since this was governed by
thermodynamic considerations (Pascault et al., 2002).
An extension of the use of CTBN has been noted in the use of it as a compatibilizer.
In this case it is used as a minor component that is soluble in the thermoplastic toughener
(e.g. poly(ether sulfone)) and is able to react at the interface with the epoxy resin,
thereby promoting interfacial adhesion (Pascault et al., 2002). It has been recognized
that when a small amount of such a compatibilizer is used, there may be the possibility
of creating a separate interphase at the interface that is not a separate thermodynamic
phase but is a region with properties intermediate between the two phases, thereby
promoting adhesion and other properties that are desirable in the end application of the
material. This has been studied in detail in composite materials science, where the use of
coupling agents is mandatory in order to modify the surface of glass bres and control
both the mechanical properties (particularly toughness) and the environmental durability
of the composite by resisting debonding due to the ingress of moisture (Ghotra and
Pritchard, 1984). It has been suggested that this interphase extends for several hundred
nanometres from the bre into the resin and will have a gradation in composition and
properties that is important in order to avoid having a sharp stress boundary due to the
abrupt change in properties.
The denition of compatibility has been differentiated from miscibility since it is concerned with phase-separated polymers and is approached through the attainment of optimum
properties for the blend (Bonner and Hope, 1993). Two of the main technologies used to
achieve it are the addition of a third component (as discussed above) and reactive blending.
The target in using a compatibilizer is the control of the interfacial tension between the
components in the melt, translating to interfacial adhesion in the blend after processing.
The reduction of the interfacial tension in the melt results in the formation of lowerdiameter particles as phase separation occurs. These are also stabilized by the compatibilizer and resist growth during subsequent thermal treatment such as annealing. Thirdly,
the enhanced interfacial adhesion results in mechanical continuity such as stress transfer
between the phases. This is a fundamental requirement for a useful polymer blend.
123
+
CH3
CH3
CH3
PP
MAH
R
n R
CH3
CH3
CH3
CH3
CH3
CH3
Scheme 1.48. Grafting of maleic anhydride to polypropylene (PP) to give either in-chain isolated
succinic anhydride groups or a terminal grafted chain of poly(succinic anhydride).
124
125
(Higgins et al., 2000). It is recognized that a owing mixture is a steady-state system, not an
equilibrium thermodynamic system, so it must be treated as a perturbation and only a
qualitative explanation is possible.
A real-time study of the process of shear-induced mixing and de-mixing was performed using on-line Raman spectroscopy followed by microscopic examination of the
extrudate (Adebayo et al., 2003). Shear ows were performed through a slit conduit
starting in the one-phase and two-phase regions of blends of PMMA and poly(a-methyl
styrene-co-acrylonitrile) that exhibit an LCST. For a high melt viscosity sample near the
(equilibrium) cloud point, the shear-heating effects over-ride all other effects. In contrast, for a lower-viscosity system, shear-induced de-mixing is observed at low shear
rates, while with increased shear rate the system underwent mixing prior to the onset
of shear heating that resulted in the system entering the two-phase region and
optically observable phase-separated domains corresponding to the maxima in shear
heating across the extrudate. Processing a system without detailed understanding of the
phase properties and the shear heating will result in complex shear-dependent
morphologies.
The above examples do not involve reactive processing (or degradation, even when
carried out in a twin-screw extruder (Papathanasiou et al., 1998)), and the introduction of
the possibility of interfacial reaction for in situ compatibilization will have an effect on the
morphology. One approach to the study of the dynamics of these processes involves the
freezing of the reaction in the twin-screw extruder and the analysis of the resultant
morphology (Van Duin et al., 2001). This has provided a method for studying several of the
most common reactive-processing operations:
processing of polyolens with peroxides either to crosslink or to create branch and graft
sites in PE, or to degrade PP to alter the melt-ow properties
grafting of maleic anhydride (MA) onto polyolens to create polar sites for compatibility
reactive blending of nylon-6 with these MA-grafted polyolens, such as an ethylene
propylene rubber (EPR)
The technique involves the uses of rotatable sampling ports that enable the polymer to
be removed and quenched as it passes particular zones (corresponding to different values of
L/D) as shown in Figure 1.37 (Van Duin et al., 2001). In this way 2 g of polymer can be
collected for off-line analysis in about 35 s. For the above polymers, the most common
prole for product formation along the axis of the extruder was one that followed the prole
for the decomposition of the peroxide with time (and thus distance along the extruder).
Reaction of MA was very rapid, and there was an exponential growth in the amount of
maleated polyolen as the MA was consumed.
Of particular relevance here was the study of the reactive blending of the MA-grafted
rubber with the polyamide and the morphology as a function of distance along the barrel.
In the melt zone a graft copolymer formed between the nylon-6 and the EPR-g-MA
that consumed the MA sites and formed a layer around the nylon-6 pellets at the rst
sampling point. As the nylon-6 then melted, by the second sampling port the quenched
polymer blend had EPR domains a few micrometres in diameter. These decreased in
diameter on kneading and exhibited a narrower particle-size distribution. No phase
inversion was observed, although the EPR particles did contain inclusions of nylon-6,
presumably from the graft copolymer that had acted as an in situ compatibilizer to give a
stable morphology to the blend. One consequence of the grafting reaction between
126
(a)
Barrel-wall aperture
Sampling cavity
Cylindrical valve
Sampling cavity
(b)
Vacuum
L/D = 8
L/D = 9
L/D = 10
L/D = 11
L/D = 16
L/D = 21
L /D = 29
Figure 1.37. (a) Detail of the sampling valve and (b) location at various L/D positions in a twin-screw
reactive extruder to enable monitoring of the chemical and morphological changes during
reaction in the extruder barrel. Reproduced with permission from Van Duin et al. (2001). Copyright
Wiley-VCH Verlag GmbH & Co. KGaA.
amide and anhydride that formed the compatibilizing copolymer was the liberation
of water that at the extrusion temperature led to hydrolysis and a decrease in molar mass
of the nylon-6. This example illustrates well the complexity of reactions and side
reactions that can occur during reactive modication and blending in a reactive extruder
(Van Duin et al., 2001).
1.3.3
Interpenetrating networks
It has been noted above that phase separation in thermoplastics is a common occurrence
when two or more polymers are mixed and that miscibility is the uncommon event.
This is exploited in toughening of thermosets by elastomers when phase separation
occurs during the reaction that leads to three-dimensional network formation. If
macroscopic phase separation is not desired then it is possible to achieve a different
microscopic morphology and in some cases maintain some features of miscibility
127
IPN
Both networks crosslinked
Semi-IPN
One network crosslinked
1.4
128
These conditions, if uncontrolled, will lead to changes in the chemical structure of the
polymer, so the understanding of this process is essential both to accurate chemorheology
and to reactive processing. In many cases the degradation or crosslinking of the polymer is
desirable in order to modify the nal molar-mass distribution or intermolecular interactions
in the material, and the goal in this section is to indicate the strategies that may be employed
to control degradation.
In this section the fundamental process underlying the changes to molar mass and its
distribution through either chain scission or crosslinking, namely the formation of volatile
products, either by exceeding the ceiling temperature or by chemical reaction, will be
considered by using the polyolens as the starting point and then extending the study to
more complex polymers. Following the attainment of an understanding of reaction mechanisms, the tool which has been employed to quantify the oxidation of polymers has been
chemical kinetics. The theory of free-radical chain-reaction kinetics in the gas phase or
well-mixed liquid state has long been established. While molten polymers differ from lowmolar-mass compounds in solution because of persistent entanglements and signicant
viscosity effects, the theory for oxidation and stabilization of the polymer melt during
processing is consistent with homogeneous chemical kinetics. In this system, the concentration of the reactants may be averaged over the volume to produce a representative value
for kinetic modelling.
1.4.1
129
C CH2
H2
(1)
H2 H2
C C C C
H2
H2
H
C C
H2
H2
C C
H2
-scission
H
C C
H2
H2
C C
H2
C C
H2 H
CH2 + H2C
C CH3
H2
C
H2
(e.g. 77200 K) (Sohma, 1989a, 1989b). The temperature dependence of the species formed
also assisted in determining the mechanism of radical formation. Scheme 1.49 illustrates the
parent free radical (1) formed on cryogenic milling of polyethylene and the subsequent
abstraction to give an in-chain radical that will compete with radical recombination.
The further reaction of this radical may also result in a decrease in molar mass through
b-scission and formation of terminal unsaturation, as well as regenerating a terminal
radical equivalent to (1) as shown in Scheme 1.49.
Extension of these studies to the polymer in solution and the measurement of the
attendant decrease in molar mass showed that it was the chain length which was important,
not the solid-state properties. In particular the solid-state studies showed that there was a
critical molar mass before mechanochemical effects could be observed (for polyethylene
(PE), a degree of polymerization of about 100). In solution a similar effect was observed
since the degradation of poly(methyl methacrylate) (PMMA) could not continue when the
chain length reached about 4000 repeat units (Sohma, 1989a, 1989b). This was shown by a
slow decrease in molar mass to the asymptotic value as shown in Figure 1.39.
This difference between the behaviour of the solid PE and the solution of PMMA was
attributed to the fundamental mechanism of scission, such that in the solid state there was
amplication of the effect of crack growth from the initial site of chain scission whereas in
PMMA it was chain entanglement that produced the mechano-scission.
In order to rationalize this result for PE, a simple calculation of the energy required
for scission of main-chain bonds (ECC) with the frictional dissipation between chains
due to mechanical action was performed. The latter was calculated from the activation
energy for viscosity of the monomer repeat unit, Eg, determined from a low-molar-mass
analogue of polyethylene. This was then used to determine the number of repeat units, n,
necessary to exceed the CC bond energy of 349 kJ/mol. From the value of 4.22 kJ/mol for
Eg, an estimate for n of 83 was made, which was considered a good approximation to the
experimental result of 100 (Sohma, 1989b). However, such a simple approach is not
amenable to extension to systems of structural complexity such as chain scission during
the mastication of rubber.
Casale and Porter (1978) brought together much of the early work on the theory of
mechanical generation of free radicals. Much of the theory for radical formation in the
rubbery state has developed from the ideas of Bueche regarding the scission of entangled
chains. The entangled chain does not permit the rotation of the entire chain to dissipate
energy when shear is applied and, furthermore, tensions reach a maximum at the centre of
the chain. The number of scissions will decline exponentially with distance from the centre
of the chain such that at link q from the centre the ratio of the number of scissions, nq, to that
at the centre, n0, is
130
10
Mn
(105)
10
Time/hrs
Figure 1.39. Change in molar mass of PMMA due to shear in solution for times up to 11 hours.
Adapted from Sohma (1989).
1:99
where F0 is the force acting on the bond, d is the bond elongation before breaking, and n is
the degree of polymerization. A typical value for F0d/(kBT) is 20 so that, on substituting
that into this equation, the number of chains that break in the centre is 150 times the
number which break a quarter of the way from the chain end. This relationship also shows
that the longer chains have a greater probability of undergoing chain scission. This
conclusion has been conrmed experimentally from studies on a number of polymers, e.g.
polybutadiene and cis-polyisoprene. These studies do not provide any details of the
kinetics of the process, and it is necessary to measure the rate of change of molar mass
under mechanical shear.
As noted earlier, on mechanical degradation of chains with an initial number-average
molar mass, MIn, there is a limiting molar mass MLim, below which no further mechanical
chain scission can occur (as shown by the asymptotic value in Figure 1.39). The molar mass
at time, t, Mt, is then
Mt MLim MIn MLim expkt;
1:100
where k is a rate coefcient that is unique for the polymer type. This may be extended to
determine the number of radicals, Z, produced per gram of polymer as
Z 2N 1=MLim 1=MIn ;
1:101
6
131
1:102
where E is the bond energy in the absence of the applied force F0 to the bond which has an
elongation to breakage of d.
Other equations have been proposed to relate the experimental observations of the
decrease in molar mass and the observations of free radicals to the input of mechanical
energy into the system. The complexity in energy dissipation in many systems ensures that
the relations are often specic for a particular polymer and set of conditions, since these
dene MLim.
The important results are that mechanoradicals are formed, but their reactivity ensures
that they are affected by the temperature of the system as well as by the presence of trace
amounts of radical-scavenging chemicals. One of the most efcient radical scavengers is
molecular oxygen, which will react with alkyl radicals at the diffusion-controlled rate for
the partial pressure of oxygen in the system. The scavenging of radicals produced either
mechanically or thermally is considered in the following sections.
132
H
Ph
C
H2C
HC
CH
Ph
CH2
CH
Ph
H2C
C
Ph
H2
C CH2
Ph
Ph
Scheme 1.50. Oligomer formation from poly(styrene) scission by back-biting.
133
(a)
(b)
2
12
320 C
10
300 C
294 C
310 C
1
300 C
4
6
Time/hr
287 C
10
287 C
280 C
4
6
Time/hr
10
Figure 1.40. Chain scission (given as number of scissions, S, per number-average molecule)
during thermal degradation of poly(styrene) polymerized anionically (a), Mn 2.3 105, or by freeradical initiation (b), Mn 1.5 106. Adapted from McNeill (1989).
information from these analyses because of the sensitivity of the reactions to sample size.
This may arise from the diffusion of volatiles and the self-catalysis of degradation, e.g.
dehydrochlorination and polyene formation in PVC is catalysed by evolved hydrogen
chloride as discussed later. Thus the ow rate of inert gas can have an effect on the rate of
removal of synergists. Another mass effect will occur when the degradation reaction may
be highly exothermic, so the heat rise in the sample will depend directly on the mass
and inversely on its surface area. These considerations are of greatest importance in
kinetic studies.
The number of chain-scission events during thermal degradation is most readily
determined by the change in degree of polymerization as a function of time. The number
of chain scissions S(t) at any time t per number-average molecule Mn0 is given by the
relation
St Mn0 =Mn t 1;
1:103
where Mn(t) is the relative molar mass at time t. This is plotted in Figure 1.40 for polystyrene polymerized by two different methods, (a) anionically, with Mn0 of 229 000, and (b)
free-radically with Mn0 of 1 490 000 (McNeill, 1989).
There are several features immediately apparent.
1.. The absolute number of scissions, S, is much lower for the anionically polymerized
polymer and increases linearly with time.
2.. The higher-molar-mass polymer polymerized by a free-radical route is much less stable,
e.g. the number of scissions at 300 C is greater by a factor of seven.
3.. The plot for the free-radical-produced polymer shows an intercept at zero time of 2
scissions per molecule. This indicates that there are weak links in this polymer that
degrade very rapidly at all temperatures studied. These are believed to be peroxide
groups in the polymer chain, formed by the rapid scavenging of the styryl radical by
oxygen during polymerization, and are primary initiation sites for thermal degradation.
134
Table 1.17. Elementary reactions and the related rate coefficients and rates for the thermal degradation of a
polyolefin, Pn
Process
Elementary reaction
Rate coefcient
Rate of reaction
Initiation
Propagation
intermolecular
intramolecular
Termination by
disproportionation
Pn ! Rr Rnr
ki
ri 2ki[Pn]
Pn Rr ! Pr Rj Pnj
Rn ! Rnr Pr
Rn Rr ! Pn Pr
k1
k2
kt
rp k1[Pn][Rr]
r2 k2[Rn]
rt 2kt[Rn][Rr] ~ 2kt[Rr]2
4.. The plot for the anionically polymerized polystyrene has a zero intercept indicating that
there are no potential weak links. The initiation in this case must involve homolysis of
the backbone followed by disproportionation.
The role of weak links has been considered in all thermal-initiation possibilities (Scott,
1995), and the tendency of many monomers to copolymerise with trace amounts of oxygen
in free-radical polymerization (in the case of styrene, to form a 1 : 1 copolymer) until it has
all been scavenged is well known.
The polyolens, particularly polyethylene (PE) and polypropylene (PP), have been studied
in detail because of their importance both in processing and in recycling and energy recovery.
Low-density PE is highly branched, which this has a bearing on the thermal degradation since
these sites constitute weak links, resulting in scission at lower temperatures. Polypropylene is
extremely sensitive to the presence of oxygen but even in the absence of air, volatiles are
evolved at temperatures 100 C lower than for PE. The products are a series of alkenes due to
intramolecular reactions of the secondary alkyl radical following random chain scission. In a
study of the effect of multiple extrusion cycles on the evolution of volatile organic compounds, it was found for PP that the levels and rate of generation were closer to mild thermooxidation rather than pure thermal degradation, indicating the difculty of achieving an
oxygen-free environment under processing conditions. It was found that the changes in
composition and evolution of volatiles were consistent with changes in rheological data
generated at the same time (Xiang et al., 2002).
The kinetics of the degradation of PP have been studied by following the molar-mass
distribution rather than just the chain scissions as a function of degradation time for temperatures from 275 to 315 C (Chan and Balke, 1997b). A simplied model was established
to account for these changes and the minimum number of parameters was chosen to t the
data. Initiation is assumed to be by random scission into two free radicals, and the only
radical reactions are scission and chain transfer. Termination is assumed to be by disproportionation, not combination (i.e. there is no increase in molar mass through the formation of crosslinks or other combination reactions). The mechanistic equations are
summarized in Table 1.17. It is assumed that a stationary state is reached and also that the
reaction sequence may be further simplied by assuming that the probability of intramolecular transfer is very low compared with that of intermolecular transfer, i.e. k2 0. The
system becomes a simple free-radical kinetic solution for a single initiation, propagation and
termination step, in which the rate of polymer degradation in the initiation step is equal to
that for recovery in the termination step (which is a consequence of the stationary-state
assumption that d[Rr]/dt 0) and thus the stationary-state concentration of radicals Rr is
135
1:104
and on substituting into the intermolecular propagation reaction, the rate of degradation of
the polymer is given by
dPn =dt k1 ki =2k t 1=2 Pn 3=2 :
1:105
It is possible to determine the change to the molar-mass distribution by integrating this relation over
time. Also the overall apparent activation energy for the degradation will be given by the following
relation for the activation energies for the propagation, initiation and termination reactions,
respectively: Eapp E1 12 Ei 12 Et , which is a form common to many free-radical reactions
(Dainton, 1956). The model provided a good t to the molar-mass-distribution data over the
temperature range from 250 to 375 C (Chan and Balke, 1997a). The simplications of this
model should be noted, however. For example, the assumption that homolysis of the macroalkyl radical (rate coefcient k2, above) is unimportant compared with the intermolecular
propagation reaction neglects the importance of b-scission as a route in PP. This reaction is
one of the reasons why PP undergoes negligible crosslinking in comparison with scission
since the radical is too short-lived to react with another radical to form a crosslink before it
undergoes scission to give the products CH3CCH2 and CCH3CH2.
1:106
with an activation energy of 247 19 kJ/mol, which is similar to that for the degradation
of simple esters. Copolymers with b-hydroxy pentanoic acid (valeric acid) at levels of
5%25% to give PHB-co-VA have a lower crystallinity and can be melt-processed at
lower temperatures (<170 C) without degradation. It should be noted that these elimination reactions are not free-radical in origin, so the stabilization strategies discussed
later will not be applicable.
The cyclization reaction of poly(acrylonitrile) results in the formation of an extended
conjugated backbone through the intramolecular imidization reaction which occurs without
the elimination of any product. Reactions of this type are favoured by the chain conformation
136
(a)
H
C C
H2
C
O
O
H
C
CH3
CH3
H3C
H
C C
H2
C
O
CH
H
C C
H2
C
+ H2C
O
CH
CH3
OH
C
H2
(b)
O
H
C C O
H2
CH3
H3C CH
H2 H
C C
C
O
CH3
CH
O
H3C CH
O
H
CH
PHB
O
H
H3C
H3C
unsaturated chain end
CH
HO
CH
+
C
H
C
O
HO
+
C
H
C
O
crotonic acid
Scheme 1.51. (a) Elimination of ester side groups through a six-membered transition state: propylene
from thermolysis of poly(isopropyl acetate). (b) PHB elimination of crotonic acid and
accompanying chain scission.
producing adjacent nitrile groups. This reaction is exploited in the production of carbon bres
by the cyclized product being pyrolysed at high temperature in an oxidizing environment to
eliminate hydrogen from the backbone, so producing aromatization, and, if the PAN precursor
is a bre, the result is a carbon (graphite) bre (Bansal and Donnet 1989) (Scheme 1.52).
These reactions when studied by FT-IR emission spectroscopy gave little production of
oxidation products such as carbonyls under the graphitization conditions (Celina et al. 1997).
Intermolecular reactions between carbon-centred radicals on adjacent chains will produce
crosslinking, which will compete with the chain-scission and intramolecular reactions. The
effect on the molar-mass distribution will depend on the competition between these reactions. For example the poly(acrylates) will preferentially crosslink rather than degrade and
the recovery of monomer from these polymers is less than 1%, compared with close to
100% for poly(methacrylates). Poly(vinyl chloride) (PVC) is also a polymer that undergoes
crosslinking in preference to scission, but the major observation of importance is the rapid
elimination of hydrogen chloride by an auto-catalysed reaction as shown in Scheme 1.53.
This is discussed in more detail later.
137
C
N
-H2
C
PAN
C
N
C
N
C
N
aromatization
cyclization; imidization
H2
C
H2
C
CH
CH
CH
CH
Cl
Cl
Cl
Cl
- HCl
C
H
H2
C
H2
C
H
C
CH
CH
CH
Cl
Cl
Cl
- HCl
PVC
H
C
C
H
H2
C
H
C
C
H
polyene
CH
CH
Cl
Cl
Scheme 1.53. Auto-catalysed elimination of HCl from PVC to give a polyene sequence.
in the degradative pathway, this will not be the case and combination of the macro-alkyl
radicals is an important process leading to the formation of a crosslink.
The control of this process is technologically important in the formation of elastomeric
networks, the rst example of which was the vulcanization of rubber (Brydson 1988). Thus, in
natural rubber, the two processes of mastication to lower the molar mass of the cis-poly
(isoprene) and then the use of sulfur or peroxides to form a network of effectively innite
molar mass are exploited. Crosslinking is one of the commonly performed reactive processing
operations and the control of the chemorheology underpins successful elastomer production.
The process for crosslinking of polyethylene has been described earlier (Scheme 1.45) and the
use of specic additives (e.g. silane grafting agents prior to hydrolytic crosslinking) was
discussed (Scheme 1.46) and is also considered later. Here we are concerned with the process
only in the presence of a free-radical generator. Thus the starting point is the polymer free
radical formed by the abstraction of a hydrogen atom from the backbone by an initiator, which
may be deliberately added, such as the free radical, I, from the decomposition of a peroxide,
or a macro-alkyl radical from a chain-scission event, Scheme 1.54.
The radical Rn may be a backbone radical, not a terminal radical as results from scission,
so if there is reaction with another chain having a backbone radical, Rr, the crosslink will
form at random between two chains. If there is a vinyl group on the other chain then an
addition reaction is possible and the nature of the crosslink will depend on the location of the
unsaturation (Scheme 1.45). Abstraction of hydrogen allyl to the double bond will also compete
with this, but the resultant main-chain macro-alkyl radical may then also undergo crosslinking.
When crosslinking predominates over disproportionation or scission, the molar-mass
distribution will broaden and Mw will shift to higher values. The process of crosslinking
may be followed by extracting the polymer to determine the residual soluble (sol) fraction,
S, i.e., the amount of material not yet incorporated into the innite network. The equation
describing this is the modied CharlesbyPinner equation (Boyd and Phillips 1993):
S S 1=2 p0 =q0 1=q0 DPn r;
1:107
where p0 and q0 are the fractions of monomer units involved in a crosslink or chainscission event, respectively, per free radical and r is the radical yield (proportional to the
initiator concentration); DPn is the number average-degree of polymerization of the original polymer. Thus, in an application of this equation, a plot would be made of (S S1/2)
138
RnH + I Rn + HI
Rn + Rr Pn--Pr
Scheme 1.54. Free-radical formation and crosslinking initiated by a radical I from an external
source (radiation or a labile molecule, e.g. peroxide).
against the inverse of the initiator concentration, 1/[I], and the ratio of the intercept to the
slope of this linear plot will be (p0DPn) so that both p0 and q0 may be determined. There
are several assumptions in this equation, and an alternative approach is via measurement
of equilibrium swelling, but then the polymersolvent-interaction parameter must be
known (Boyd and Phillips 1993).
The technology of controlled crosslinking during reactive extrusion has been surveyed by
Brown (1992) and there have been many reviews of the technology of elastomer vulcanization (Brydson, 1988, Morrison and Porter, 1989).
1.4.2
139
140
Initiation:
Polymer (defect)
(1)
2 POOH
PO2 + PO + H2O
(1)
+ O2
POO
(2)
POO + PH
POOH
PO
Propagation:
P
(3)
OH
(4)
Chain branching:
POOH
PH
OH
H2O
(5)
PH
PO
POH
(6)
chain scission
PO
(7)
POO
(8)
POO +
POOP
(9)
PP
(10)
POO
H
C C C
H2
H2
CH3
C C C
H2
H2
CH3
+ POOH
(11)
C C
H2 H
+ POOH
(12)
PP
POO
C C
H2 H2
PE
Scheme 1.56. Reactions of the chain-carrying peroxy radical with polypropylene (11) or
polyethylene (12) to give tertiary- or secondary-alkyl radicals, respectively.
Hydrogen abstraction is known to occur from secondary carbon atoms in polyethylene (12)
and may also occur in polypropylene, but with lower reaction rates. For polypropylene it was
shown that intramolecular hydrogen abstraction in a six-ring favourable stereochemical
arrangement will preferentially lead to the formation of sequences of hydroperoxides in close
proximity (Scheirs et al., 1995b, Chien et al., 1968, Mayo, 1978). Infrared studies of polypropylene hydroperoxides showed that more than 90% of these groups were intramolecularly
141
CH3
C C
H2
O
CH3
C C
H2
O
CH3
H2
C C
H
CH3
H2
C C
H
CH3
CH3
C C
H2
O
C C
H2
O
+ H2C
CH3
H2
C C
H
(13)
C
H
CH3
(14)
Scheme 1.57. Chain scission (13) or elimination (14) reactions of alkoxy radicals in polypropylene
resulting in terminal (13) or in-chain (14) ketone groups as well as alkyl radicals that may
react further with oxygen.
hydrogen-bonded. The results indicated that the peroxide groups were preferentially present in
sequences of two or more, which supported intramolecular hydrogen abstraction during the
oxidation (Chien et al., 1968). It was also concluded that the cleavage reaction accompanied
the oxidation propagation rather than the termination reaction.
As shown in Scheme 1.55, chain branching by thermolysis or photolysis, reaction (4),
of polymer hydroperoxides (POOH) results in the formation of very reactive polymer
alkoxy radicals (PO) and hydroxyl radicals (OH). The highly mobile hydroxyl and
polymer alkoxy radical can abstract hydrogen atoms from the same or a nearby polymer
chain by reactions (5) and (6), respectively (Zweifel, 1998). These polymer oxy radicals
can react further to result in b-scission, by reaction (13), or the formation of in-chain
ketones by reaction (14), as shown in Scheme 1.57, or can be involved in termination
reactions.
The termination of polymer radicals occurs by various bimolecular recombination reactions. When the oxygen supply is sufcient the termination is almost exclusively via reaction
(8) of Scheme 1.55. At low oxygen pressure other termination reactions may take place
(Zweifel, 1998). The recombination is inuenced by cage effects, steric control, mutual
diffusion and the molecular dynamics of the polymer matrix. In melts the recombination of
polymer peroxy radicals (POO) is efcient due to the high rate of their encounter.
142
C CH
H2
POO
O2
C C
H2
(15)
C C
H2
C C
H2
OO
(16)
O
OO + PH
C C
H2
C C
H2
(17)
+ OH
(18)
(19)
OOH +
C C
H2
OOH
C C
H2
O
C C
H2
POOH
O
O
+ PH
C C
H2
OH +
Scheme 1.58. Formation and reactions of peracids ultimately producing carboxylic acids during the
oxidation of polypropylene.
CH3
CH3
C C
H2 H
CH2
O2
PH
CH2 OOH +
(20)
H2O
(21)
CH3
CH3
C C
H2 H
C C
H2 H
CH2 OOH
O2
C C
H2 H
C OOH
In another related mechanism, Scheme 1.59, the primary alkyl radicals produced via
b-scission (13) in Scheme 1.57 will form the hydroperoxide by reaction (20), which may
be oxidized further to yield peracids in reaction (21) (Gijsman et al., 1993).
Solubility measurements indicated that the peracids were of macromolecular nature,
being formed by the oxidation of macroaldehydes, presumably via radical attack on methyl
hydrogens, subsequent peroxidation and disproportionation (Zahradnikova et al., 1991).
The difculty of analysis of these reactive species has meant that unequivocal quantitation
of the reactions has not been achieved.
143
1:108
1:109
A more complex relationship was presented with Equation (1.110) below, obtained with the
assumption that the kinetic chain is not too short and that k92 k10 k8 (and with the initiation
rate ri k10 [POOH]2 from (10 )):
1=2
1=2
1=2
1:110
144
constant and the oxidation demonstrates auto-acceleration. Such reactions have long been
studied in the high-temperature gas-phase oxidation of hydrocarbons, and the formation of
hydroperoxides as a degenerate branching agent results in the observation of cool ames
in which the chain reaction is linear until a critical concentration of hydroperoxide is
reached (Dainton, 1956). It then becomes branched, an exponentially increasing chain
reaction (ame or explosion) ensues, which consumes the branching agent, and the system
then becomes a linear chain again. This alternating process continues until the fuel is
exhausted. By altering the conditions of pressure and temperature, the termination rate
may be altered and the system may be held in the explosion (branched) regime or in the
linear regime in which hydroperoxide accumulates.
Obviously, under the conditions of processing and chemorheology, the oxidation of the
polymeric hydrocarbon cannot reach the above critical condition, and homogeneous kinetic
treatments of the oxidation of polypropylene involve perturbations of the steady-state
approximation. The application of models involving degenerate branching has been a
particular feature of kinetic treatments of the oxidation of polypropylene by many Russian
authors (Emanuel and Buchachenko, 1987). In one such study, it was suggested that, under
conditions for degenerate branching, the maximum rate of oxygen consumption may be
expressed by the following formula, where d is the probability of degenerate chain
branching, which was derived again by neglecting reactions (9) and (10) at high oxygen
pressure and assuming the oxidation chain to be sufciently long:
dO2 =dtmax dk3 RH2 =2k8 1=2 :
1:111
It is noted that, in many of these studies, the basic kinetic information is obtained from an
oxygen-uptake curve. If the consumption of oxygen has been measured by a pressure
transducer (Scheirs et al., 1995b) then one limitation is the lack of specicity of the
measurement. The total pressure in the system will depend on the partial pressure of the
products (for example carbon dioxide and water) as well as on that of oxygen. This may be
overcome by undertaking a total analysis of the gases present in the apparatus.
From the slope of the oxygen-uptake curve in the linear region, a steady-state rate of
O2 consumption has often been used to compare the oxidation of various polymer
samples (Dainton, 1956). The determination of rate coefcients and kinetic information
has been repeatedly obtained from studies of the oxidation of model hydrocarbons in
solution at relatively low conversions. Studies of model substances, such as 2,4-dimethylpentane and 2,4,6-trimethylheptane, as well as of polypropylene enabled the evaluation of rate constants for propagation and termination (Chien and Wang, 1975). Any
attempt to establish precise rate constants for the initiation, propagation and termination
in the auto-oxidation of PP is, however, complicated by the fact, as pointed out by Mayo
(1972), that intermediate oxidation products of a saturated hydrocarbon are 10100 times
more susceptible to oxidation than is the starting material. Rate constants as measured
may be those for a complex co-oxidation of a highly degraded polymer and its oxidation
products (Mayo, 1972). These include the range of by-products outlined in Schemes 1.57,
1.58 and 1.59, viz. carboxylic acids, peracids and peresters, from the oxidation of the
aldehyde produced as a primary product of the free-radical oxidation of the polymer.
The diversity of products can be signicant in determining the subsequent performance of
the polymer since the chemical products will be photoinitiators of oxidation, and this is
one of the reasons for the incorporation of stabilizers in the melt during processing.
145
CH2
P or
PO
H2
C C
H2
C
CH2
C
CH2
High [O2]
OO
C
H
C
H2
CH2
H
C
+ P
C
H2
Low [O2]
C
H
H
C
C
H
CH2
C
H2
CH2
C
H
+ P
C
H2
CH2
CH2
C
H
C
H2
CROSSLINKS
Low [O2]
PO
-scission
C
H2
CH2
OOH
C
H
C
C
H2
O
C
C
H
C
H2
CH2
H2C
CHAIN SCISSION
Scheme 1.60 Competitive chain scission and crosslinking with various [O2] in thermo-oxidation
Polyethylene
In polyethylene, the tertiary carbon atom, which dominated the chemistry of the oxidative
degradation of PP, is present only at branch points. This suggests that there may be a
difference among LDPE, LLDPE and HDPE in terms of the expected rates of oxidation.
This is complicated further by the presence of catalyst residues from the ZieglerNatta
polymerization of HDPE that may be potential free-radical initiators. The polymers also
have differences in degree of crystallinity, but these should not impinge on the melt
properties at other than low temperatures at which residual structure may prevail in the melt.
Also of signicance is residual unsaturation such as in-chain trans-vinylene and vinylidene
as well as terminal vinyl, which are defects in the idealized PE structure.
The reaction sequence (Scheme 1.60) shows the competition between chain scission and
crosslinking reactions as well as the participation of unsaturation in the thermo-oxidation
of LDPE in the melt (Al-Malaika, 1989). The process of formation of H and Y crosslinks
involving terminal unsaturation (Scheme 1.45) and the formation of terminal unsaturation during
b-scission (Scheme 1.49) in the absence of oxygen have been discussed previously.
It has been shown that the type of catalyst used in HDPE polymerization affects the ow
behaviour and that crosslinking is favoured at higher temperatures (Zweifel, 1998). As in
146
H2
C C
H2
Shear
C
CH2
CH2
H2C
CH2
C
CH2
H
C
C
H2
H2
C C
H2
CH2
Scheme 1.61 Initiation of the mechano-oxidation of PE and subsequent in-chain radical (P)
formation by abstraction by a terminal alkyl radical after the initial chain-scission event.
After Scott (1993b).
PP, the polymer hydroperoxide is the key intermediate which controls the oxidative
performance in the melt. In an analysis of the published data on the processing performance
of PE, Gugumus (1999) noted the following.
1.. The induction period for the oxidation was inversely proportional to the vinyl content of
the polymer, suggesting the importance of initiation by primary attack by oxygen of the
allylic CH bond.
2.. The differentiation between PP and HDPE in the processing behaviour was linked to steric
effects that limited direct reaction of a macro-alkyl (or alkoxy or peroxy) radical with the
alkene CH2C(CH3) formed in PP, in contrast to the unhindered reaction of the radical
with the CC group in HDPE. Thus crosslinking was favoured in HDPE over PP.
3.. The terminal vinyl groups appeared to be those responsible for crosslinking, rather than the
trans-vinylene or vinylidene groups known to be present in LLDPE. This conclusion has
been supported by results from later studies on c-initiated crosslinking (Bracco et al., 2005).
In his further studies Gugumus (2000) showed that, for PE, the contribution of direct thermal
production of free radicals during mechanochemistry in the melt was negligible in comparison with mechanical effects such as shear. This was based on the dependence of the
initial rate of hydroperoxide formation being sensitive to the melt viscosity as well as the
high activation energy for pure thermal initiation. Scott (1993b) also reached this conclusion,
and proposed a scheme similar to that in Scheme 1.60 in terms of competition between chain
scission and crosslinking for mechano-oxidation in the limits of high and low oxygen concentration. The major difference is that the initiating step in mechano-oxidation is the scission
of the main chain to produce two terminal primary alkyl radicals (Scheme 1.61) rather than
an in-chain secondary-alkyl radical (P in Scheme 1.60) as in thermo-oxidation. The subsequent free-radical chemistry and the scission and crosslinking reactions are identical for
thermo- and mechano-degradation since the terminal primary radicals then abstract protons
from the PE chain to produce secondary-alkyl radicals.
An interesting result was the way in which the viscosity, g, had another effect, namely to
control the diffusion of the free radicals out of the initial cage in which they were formed.
Geminate recombination of radicals is therefore important and the formation of chain-carrying
radicals was dependent on g 1/3. Also of interest is the observation that the decomposition
reactions at higher temperatures (170200 C) are different from those at lower processing
temperatures (150160 C). This may reect the importance of clusters of associated
hydroperoxides, which decompose at the higher temperatures by a bimolecular reaction
(Gugumus, 2002a, 2002b).
147
H2
C
CH
H2
C
CH
Cl
H2
C
CH
Cl
H2
C
Shear
CH
CH
Cl
Cl
Cl
CH
CH2
H2
C
low [O2]
CH
Cl
Cl
O2 + (PVC)
CH2
C
H
CH
Cl
low [O2]
Shear
H2
C
HC
H2
C
HOO
+ Cl
Cl
CH
CH
Cl
Cl
+ HCl
H
C
C
H
C
H
CH2
HCl
H2
CH2
C
+ HCl
CH C
H
Cl
H2
C
CH
CH
Cl
Cl
+ Cl + H2O
polyene
H2
C
O
terminal
chlorocarbonyl
CH
Cl
Cl
Scheme 1.62. The reaction sequence for PVC oxidation during mechanical working.
Poly(vinyl chloride)
While the polyene formation characteristic of PVC degradation will occur both in
an oxidizing and in an inert atmosphere as described above, the presence of oxygen leads to
oxidation of the polyenes to both cyclic peroxides and hydroperoxides (Tudos et al., 1979).
This is seen as the bleaching of the polyenes in thermally degraded PVC after the admission
of oxygen because of the loss in conjugation. The delocalized nature of the polyenyl radical
leads to a large number of possible products, as shown in Scheme 1.62.
The products formed may be rationalized as a competition between polyene formation
by dehydrochlorination and hydroperoxide formation through direct radical reaction as
well as oxidation of the formed polyenes. Unlike the case for PE or PP, the decomposition of the hydroperoxides is catalysed by the hydrogen chloride present, so alkoxyradical formation is enhanced. This is seen in the reaction sequence on the right of
Scheme 1.62. The alkoxy radical, RO, rearranges to form a terminal chlorocarbonyl
and the chlorine atom abstracts hydrogen to form hydrogen chloride and continue the
cycle. The simultaneous crosslinking and chain scission has been analysed by use of a
CharlesbyPinner plot (Equation (1.107)) and the ratio of scission to crosslinking
was found to range from 0.4 to 0.7 at 180 C in oxygen, depending on the type of PVC
(Tudos et al., 1979).
Aliphatic polyamides
Most of the polymers discussed so far have been addition polymers but there are important
condensation polymers that have relevant mechanochemistry. This may involve reactions in
addition to free-radical oxidation (Levchik et al., 1999), and it has been shown that the
148
O
N
H
H
N
O
H
N
C
O
carboxy-terminated Nylon 6
O
C
+
C
OH
N
H
OH
-caprolactam
thermal and thermo-oxidative degradation of aliphatic nylons include several parallel and
consecutive reactions that are sensitive to the temperature and conditions, particularly the
presence of impurities. The aliphatic nylons of commercial signicance are nylon-6, synthesized by the ring-opening of caprolactam, and nylon-6,6, which is synthesized by the
condensation of hexamethylene diamine and adipic acid. This difference in synthesis carries
over to the difference in degradation pathways.
Nylon-6 will undergo re-equilibration with the cyclic monomer as well as with larger
cyclics at elevated temperature. This is the reverse of the polymerization process, which
occurs at 200 C and takes place through an intermediate carboxy-terminated hydrolysis
fragment that undergoes intramolecular (or intermolecular) reaction to generate the cyclic
monomer e-caprolactam as shown below in Scheme 1.63.
This may also occur through the amino end group if this can be formed via an aminolysis
reaction. Caprolactam formation is favoured by the presence of amino end groups. This is
an example of a more general reaction for all polylactams that are in equilibrium with the
cyclic monomer and the larger oligomers. There are many secondary reactions that may
occur and one of key importance is the formation of precursors to crosslinks. This would
include secondary amines from the condensation of amine end groups to eliminate ammonia
and form an in-chain amine, which can then condense with an acid end group to form a
crosslink. While the detailed chemistry of these reactions has been questioned, gelation
from reactions such as this can occur on prolonged heating at 300 C. It is known that
residual catalyst from anionic polymerization of caprolactam will result in depolymerization. Base-catalysed decomposition is the more efcient route for recovery of caprolactam from nylon-6.
Nylon-6,6 undergoes gelation through crosslinking more readily than do other aliphatic
nylons, through a mechanism similar to that for nylon-6. Other reactions are also observed,
and different crosslinks have been proposed (Levchik et al., 1999). In the presence of
oxygen the polymer undergoes yellowing and discolouration through a free-radical chain
reaction involving the macroradical below formed by abstraction from the N-vicinal
methylene.This may arise from the rearrangement and reaction of the radicals from primary
main-chain scission of the amide bond due to shear, as shown in Scheme 1.64. The acyl
radical (R) and amino radical (R0 ) will abstract the hydrogen of the methylene group a to
the amide group in another chain of nylon-6. The resulting alkyl radical is rapidly scavenged
by oxygen to form a peroxy radical and then a hydroperoxide. Decomposition of this to the
alkoxy radical results in an a-hydroxy amide, which is thermally unstable and results in
chain scission.
It has been found that the end groups play a major role in the rate of oxidation since
carboxylic acid groups will catalyse decomposition of hydroperoxide and lower the temperature for the maximum rate of degradation by 60 C. In contrast polyamides terminated
with amine end groups are more stable.
149
C C
H2
O
C C
H2
O
H
N
H
N
Shear
C
H2
R
R'
C
H2
C C
H2
O
R
HN
C
H2
R'
H
N
C C
H2
O
C
H
+ RH, R'H
O2, Nylon 6
O
H
N
C C
H2
O
C C
H2
O
OH
C C
H2
O
C
H
H
N
C
H
chain scission
NH2
HC
O
H
N
C
H2
OH
C C
H2
O
H
N
C
H
Scheme 1.64. The reaction sequence for shear-initiated oxidation and chain scission of nylon-6.
1.4.3
150
Free-radical scavenging
The strategy to be employed in free-radical scavenging is to intercept the radical species
responsible for the propagation step in the degradation, since this reaction is the main route
to produce products (in this case a degraded polymer through chain scission or crosslinking).
For all free-radical reactions, the kinetic chain length, v, of the reaction is dened as the
number of products produced per initiating event or the rate of the propagation reaction rp
divided by the rate of initiation ri. For a simple linear chain reaction of a hydrocarbon
polymer of concentration [PH] as shown earlier in Table 1.17, Equation (1.105) may be
written in a form that does not explicitly identify the initiation reaction so that
m rp =ri kp PH=kt ri 1=2 ;
1:112
where kp and kt are the rate coefcients for the propagation and termination reactions,
respectively. The kinetics of free-radical oxidation reaction as described in Equations (1.108)
and (1.109), Section 1.4.2, were sensitive to the partial pressure of oxygen since this dened
the nature of the propagation reaction as well as the most likely termination steps
Depending on the conditions of the reaction, m may range from slightly greater than
unity to several thousand. For example, in the initiated oxidation of the model hydrocarbon, tetralin (tetrahydronaphthalene) at 30 C the kinetic chain length, m, is calculated
from published values (Kamiya and Niki, 1978) to be about 2500. If this reaction is to be
inhibited, then the value of m must be reduced to <1. To achieve this level of inhibition,
an additive that can compete with the propagation reaction by providing alternative
termination routes for the radicals other than the usual recombination reaction is required.
This problem is routinely met in the inhibition of polymerization of a liquid monomer
such as styrene, and in that case a free-radical scavenger such as hydroquinone or the
hindered phenol BHT is used. The purpose of a free-radical-scavenging stabilizer is to
reduce the kinetic chain length by providing a more readily abstracted hydrogen atom and
producing a free-radical product that is stable at the temperature of the reaction.
In the inhibition of the oxidation of polymer melts, the types of antioxidant may be
characterized by their mode of action.
O
Benzoquinone
151
(a)
[Inhibitor]: C3 > C2 > C1
C2
C1
C3
Oxygen
Uptake
3
Time
(b)
0 + C/r i
Induction time
0
C 0
C [Inhibitor]
Figure 1.41. (a) Oxidation induction time, s, for various concentrations of oxidation inhibitor
(antioxidant) as determined from oxygen uptake (or carbonyl index) plots. (b) Estimation of
intrinsic inhibitor concentration, C0, from a plot of s versus [Inhibitor]. The inverse of the slope of
the plot reects the rate of initiation.
These stabilizers react stoichiometrically with the alkyl radicals and are thus consumed in
the process, unless there is a subsequent reaction to liberate the quinone again. In the absence
of this, the kinetic relation for the CB-A stabilizer may be determined by simply replacing the
usual termination reaction of the free-radical reaction scheme with that for the reaction of the
inhibitor with the alkyl radical, R, which would otherwise be propagating the chain. Thus, if
reaction with the inhibitor can occur at the diffusion-controlled rate, then the rate of termination, rt is dominated by the reaction with inhibitor of concentration [Inh] and becomes, in the
steady state,
1:113
rt kinh InhR ri :
The kinetic chain length, m, is then reduced to
m rp =ri rp =rt kp RHR =kinh InhR
kp RH=kinh Inh:
1:114
Since the reaction of Inh with the radicals has a low activation energy, kinh is a diffusioncontrolled rate constant for the reaction (about 109 l mol1s1, depending on the viscosity
of the melt) and thus, even for low concentrations of inhibitor, m 1. The oxidation rate,
given by rp, will be very low until [Inh] < kp[RH]/ kinh and thereafter the oxidation will
proceed as if there were no additive. This is shown schematically in Figure 1.41(a), and it
may be seen that the induction period increases with the concentration, C, of inhibitor. If
152
the reaction of the inhibitor is efcient and destroys one radical at each encounter, then the
rate of consumption of inhibitor must be a measure of the rate of formation of free radicals
in the system. In this case the induction period, s, of the chain reaction is given by
s Inh=ri :
1:115
If the inhibitor is able to destroy more than one radical (as described below for chain-breaking
redox antioxidants) then the efciency of inhibition will be increased by this ratio.
A plot of the induction period against the concentration of inhibitor should be a straight
line passing through the origin and with a slope depending on the efciency of the antioxidant at that temperature. If the polymer already contains a free-radical scavenger or has
intrinsic radical-scavenging capability, then the intercept on the induction period axis is
non-zero and the system contains an equivalent inhibitor concentration given by the
negative intercept on the concentration axis, as shown in Figure 1.41(a).
t-Bu
t-Bu
OH
t-Bu
t-Bu
PO2
POOH
t-Bu
O
x2
Me
t-Bu
t-Bu
Me
BHT
t-Bu
H
C C
H
t-Bu
+ BHT
t-Bu
t-Bu
t-Bu
t-Bu
CH2
H
C
Galvinoxyl radical
t-Bu
Scheme 1.66. Some BHT stabilization reactions with peroxy radicals. After Scott (1993b).
t-Bu
153
reactive intermediate formed from the coupling together of quinone methide fragments
from the rearrangement of the phenoxy radical after the initial donor reaction.
The resulting galvinoxyl radical was found to be uctuating in concentration throughout
the processing of polypropylene (PP) in a closed mixer, and also the addition of quinone
methide oxidation products of BHT was more effective than the antioxidant itself in stabilizing PP. The quinone products operate as CB-A antioxidants as discussed in the previous
section. The mechanism involves a cycle between the phenoxyl radical and its reduced
form, which is formed by accepting a hydrogen atom in the reaction with an alkyl radical.
This feature common to many additives, namely that they can function as two different
types of stabilizer by cycling through the oxidized and reduced forms, is discussed below.
This is the reason why a stabilizer that would appear to be effective only as a CB-D
antioxidant against oxy radicals (and therefore not apparently able to scavenge alkyl radicals) is an effective processing stabilizer.
The molecules able to be efcient H-atom donors to alkyl radicals are not restricted to
hindered phenols. It has long been noted in the rubber industry that secondary aromatic
amines are effective radical scavengers, and study of these materials has shown that they
produce a range of coupling products that function in a way similar to the phenols. The
problem of the intense colour of these by-products is limited when they are used in carbonblack-lled rubber but a major limitation in other polymers (Zweifel, 1998). The exceptions
are the polyamides, in which aromatic amines have a greater stabilization efciency than
do hindered phenols. This may be related to the general observation in polyamides (e.g.
nylon-6) that the amine-terminated polymer is more stable than the carboxylic acid-or
methyl- terminated polymer.
t-Bu
H
C
t-Bu
t-Bu
+ H (+ e-)
O
t-Bu
H
C
OH
-H(-e)
t-Bu
t-Bu
t-Bu
t-Bu
Scheme 1.67. The redox reaction of galvinoxyl radical, G, and the parent phenol, GH. After
Al-Malaika (1989).
154
An interesting stabilizer system that may involve a redox couple is the use of copper
iodide in the stabilization of polyamides. This is often the melt stabilizer of choice, and
the reaction (Scheme 1.68) to remove the main chain-carrying radical is proposed,
followed by a redox reaction with peroxy radicals to regenerate the cuprous ion.
It was noted that the efciency is increased in the presence of amines (Scott, 1993b), so
the effectiveness in conjunction with aromatic amine free-radical (CB-D) stabilizers or the
amine end groups of nylon would be expected. It is also noted that cuprous salts are often
used as the iodide, and the effective alkyl-scavenging ability of iodo compounds has been
demonstrated (Henman, 1979). Copper salts are also effective in the melt stabilization of
polyesters, but in this case the effectiveness is improved by using a hindered phenol antioxidant. This performance is in contrast to the strong prodegradant effect of copper and
other transition metals on the polyolens as discussed later.
Hydroperoxide decomposers
The efcient decomposition of hydroperoxides by a non-radical pathway can greatly
increase the stabilizing efciency of a chain-breaking antioxidant. This generally occurs by
an ionic reaction mechanism. Typical additives are sulfur compounds and phosphite esters.
These are able to compete with the decomposition reactions (either unimolecular or
bimolecular) that produce the reactive alkoxy, hydroxy and peroxy radicals and reduce the
peroxide to the alcohol. This is shown in the rst reaction in Scheme 1.69 for the behaviour
of a triaryl phosphite, P(OAr)3 in reducing ROOH to ROH while itself being oxidized to the
phosphate.
The structure of the aryl groups is designed to increase the hydrolytic stability of the
compound. The sulfur compounds which are the most efcient hydroperoxide decomposers
155
Acid scavengers
In the degradation of PVC during processing, the most important factor is the liberation of
HCl. This results in the formation of polyenes and thus secondary oxidation involving
hydroperoxides, and then catalyses the decomposition of these peroxides to form further
free radicals. The removal of HCl is the target of the main group of stabilizers in PVC
technology, namely the metal bases and inorganic salts. Some of the more effective of
these, such as lead and cadmium stearates, are toxic and their use is no longer recommended. Zinc stearate is less effective, but non-toxic, so it is used in combination with
other stabilizers. While it has been suggested that there may be reactions of zinc stearate
with allylic chlorine to inhibit the initiation of dehydrochlorination, there is no evidence
for this being signicant compared with the direct scavenging reaction to produce zinc
chloride and stearic acid.
There have been many different compounds used as acid scavengers, but the products must
be stable and bind the chlorine at the temperature of processing. Such compounds include
epoxides (as in epoxy resins), which will undergo ring-opening to give the chlorohydrin;
and
dibutyltin maleate, which, as well as scavenging HCl, is considered also to undergo a
DielsAlder reaction with the unsaturation on the PVC following dehydrochlorination
(Scott, 1993b).
156
30
20
Weight
Loss %
10
: 0.3% melamine
: AO + Ca Stearate (0.3% each)
0
0
20
40
60
80
100
120
140
Other polymers undergo acid-catalysed depolymerization during melt processing and can
be processed only in the presence of an acid scavenger. Thus calcium stearate and aromatic bases such as melamine are essential during the melt processing of polyacetal,
which has a relatively low ceiling temperature. The effectiveness of stabilization of
polyacetal against weight loss due to liberation of formaldehyde in the melt by these
additives (when combined with a hindered phenol antioxidant, Irganox 245) is shown in
Figure 1.42 (Zweifel, 1998).
Free-radical generation
The addition of peroxides, R0 OOR00 , during reactive processing has become a routine aspect
of reactive extrusion technology for polypropylene and polyethylene (Xanthos, 1992). In
particular, the dialkyl peroxides at a level ranging from 0.001 to 1 wt.% have been used
and have half-lives ranging from 1 to 3 minutes at 180 C, so 99% consumption will occur
in 720 minutes at that temperature. The reaction steps depend on whether oxygen is present
157
(a)
(b)
R
Ph
SH
POOH
Ph
POOH
Ph
PhSO
PH
O
Ph
Ph + Ph
Ph
PO +
Ph
H2O
+ POH
POOH
SO3
PhS
SO2 + POH
Antioxidants
PhSH + P
Pro-oxidant
O
Scheme 1.70. Pro-oxidant activity of alkyl mercaptans (a) and both pro-oxidant and delayed
antioxidant activity of diphenyl disuldes (b) during melt processing of polymer PH
containing hydroperoxide groups POOH. After Scott (1993b).
and will follow the radical reaction schemes in Section 1.4.1 for an inert environment and
Section 1.4.2 for low and high oxygen partial pressures. The rate-determining step in the
reaction is the decomposition of the peroxide to the active alkoxy radical, which is a rstorder reaction with a characteristic half-life. The kinetic schemes will apply with a higher
rate of initiation and a lower chain length of oxidation (since m rp/ri). Loss of volatile
peroxides during processing can affect the degradation or crosslinking rate, as can the
presence of free-radical-scavenging stabilizers.
In the subsequent use of the polymer it is important to recognize that many of the
products of processing degradation will be photoinitiators (as will any residual peroxide),
and the incorporation of suitable stabilizers that do not affect the controlled degradation
may be necessary.
Transition-Metal ions
It was noted (Scheme 1.68) that the redox cycle of Cu I/Cu II was an effective way to ensure
the continued decomposition of alkyl radicals in the inhibition of the oxidation of aliphatic
polyamides. The converse is observed with certain systems such as polypropylene, where
158
the metal ion can decompose hydroperoxides catalytically and is a pro-oxidant. This is a
problem in service when polyolens are used in contact with copper, and it has been
demonstrated that the contact with metal causes degradation remote from the metal, suggesting that there is migration of the ions in the solid or the production of a volatile prooxidant that can spread the oxidation. In the molten state it has been recognized that the
extrusion of polyolens through dies containing various levels of copper will affect the
extensional rheology and the morphology of the polymer when it cools. This has been
linked to the change in the molar-mass distribution of the polymer. The role of transitionmetal catalyst, notably titanium residues from the ZieglerNatta polymerization, in the
initiation of free-radical reactions and the spreading of infectious oxidation in the solid
polymer has also been recognized (George and Celina, 2000). In this case the redox
chemistry involves Ti III/Ti IV.
The dual role of transition metals as either prodegradant or stabilizer is rationalized by the
observation that the free-radical chemistry which dominates depends on the concentration
of the metal ion. At low concentrations Cu I/Cu II may be a pro-oxidant, whereas at high
concentrations it may be a stabilizer. The explanation lies in the complexation of hydroperoxides with transition metals (Black, 1978). Thus, taking the example of Co II/Co III, the
reactions in Scheme 1.71 are recognized (Black, 1978), which together give the usual
reaction for the metal-catalysed bimolecular decomposition of hydroperoxides to alkyl
peroxy and alkoxy radicals.
When the reaction is written in this shortened form it neglects the role of the complexes
(ROOHCo)2 and (ROOHCo)3 in the reactions since, depending on the concentration
of hydroperoxide, the free ions available for the radical-scavenging redox reactions in
Scheme 1.72 will be limited.
Co 2+ + ROOH (ROOHCo) 2+
(ROOHCo) 2+
Co2+ + ROOH
(ROOHCo)2+
RO + Co3+ + OH-
Co3+ + ROOH
(ROOHCo)3+
RO2 + Co2+ + H+
3+
/ Co
2 ROOH Co
RO + RO2 + H2O
159
Thus, only when the transition-metal-ion concentration is very much greater than that
of the hydroperoxide, ROOH, will there be a stabilizing effect. Therefore, at the low
levels obtained from a metal due to trace acid attack for example, there will be a pronounced pro-oxidant effect with the bimolecular reaction above. If there is a need to
control this reaction, then it is possible to add complexing agents such as oxanilide, which
form metal complexes of greater stability than the hydroperoxides (Zweifel, 1998). The
efciency of the above reactions differs from metal to metal, depending on the stability
constants for the complexes and the detailed kinetics of the reactions. For example, in the
case of Cu I/Cu II it has been shown that the inhibitor is Cu II, which must then compete
with oxygen for alkyl-radical scavenging and stabilization. This has a lower rate coefcient than has cobalt, so the pro-degradant behaviour is greater in non-polar media (Black,
1978). The solubility of oxygen in the polyamides is much less than that in polyolens,
which fact and this may have a bearing on the differentce in performance of copper salts in
the two polymers.
The metal-catalysed decomposition of hydroperoxides is exploited in the formation of
crosslinked polymer networks in composite fabrication, where cobalt octoate and cobalt
naphthenate are used to catalyse the decomposition of cumene hydroperoxide and methyl ethyl
ketone hydroperoxide at ambient temperature to provide the initiating radicals for the styrene
crosslinking reaction of unsaturated polyester networks. The reaction sheme for this process
was given earlier in Scheme 1.44. The use of drying oils as surface coatings involves the
addition of metal ions to catalyse the decomposition of hydroperoxides formed on atmospheric
oxidation of the unsaturated fatty acid esters (triglycerides) and so facilitate crosslinking and
network formation (Scott, 1993a). The same reactions as shown above are applicable and show
the generality of the use of metal ions as either prodegradants or stabilizers.
160
H2 H2
C C O
PET
H+/ H2O
O
O
C
OH
+ HO
H2 H2
C C O
Scheme 1.73. Acid hydrolysis of PET producing chain scission and generation of acid and alcohol
end groups. The carboxylic acid catalyses further hydrolysis.
H+
N
H
N
H
C
OH
C
O
H2O
NH2
H2
N C
O
C
OH
chain scission
Scheme 1.74. The mechanism for acid hydrolysis of polyamide leading to chain scission and
amine and carboxylic acid end groups. After Zaikov (1985).
It has been noted (Scheirs, 2000) that this leads to a dramatic decrease in viscosity that
renders the polymer unprocessable or, at the best, results in defects such as haze due to
crystallites that nucleate more readily from the lower-molar-mass, degraded polymer. The
rate of loss of properties due to hydrolysis is orders of magnitude faster than oxidative
degradation at the same temperature. To avoid these effects, the moisture level in an
aromatic polyester such as PET must be kept below 0.02%.
The sensitivity of the polymer to processing with an ambient level of moisture may be
determined by monitoring the change in the melt viscosity at the processing temperature
(Seo and Cloyd, 1991). This is of importance in assessing the effect of repeated recycling
on the performance of a polymer. It has been noted that the combined hydrolytic and
thermomechanical chain scission result in a dramatic decline in the properties of the nal
PET product after three repeated extrusions, but this extent of reduction in properties is
not apparent after further reprocessing (La Mantia and Vinci, 1994). An important
observation in PET blends with ABS has been that the blend undergoes degradation
during processing more rapidly than does the PET alone, and this has been attributed to
the presence of catalytic impurities in the ABS that accelerate chain scission of the PET
(Cook et al., 1996).
Polyamides, particularly nylon-6, are very susceptible to hydrolysis during both processing and service, so the resulting mechanical properties of tensile strength and modulus
are compromised. The mechanism of this process at low concentrations of acid is believed
to involve protonation of the nitrogen of the amide followed by a bimolecular reaction with
water leading to chain scission and generating amine and acid end groups, Scheme 1.74
(Zaikov, 1985)
161
The properties of polyamides are particularly sensitive to agents that may disrupt the
hydrogen bonding between chains and this includes trace-metal (such as zinc and cobalt)
salts as well as water.
Polycarbonates are processed at very elevated temperatures (>300 C) so that hydrolysis
of the carbonate linkage to yield carbon dioxide and a phenol end group can occur if the
polymer is not rigorously dried (residual water <0.02 wt.%). Complex chemistry can follow
(Scheirs, 2000), resulting in the formation of low-molar-mass species and darkening of the
polymer in addition to chain scission. The degradation is very sensitive to trace acidic or
basic impurities, and it has been noted that in blends (e.g. PC/ABS) there may be initiation
of the degradation by impurities in the ABS (Cook et al., 1996).
162
amide, liberating water. The sensitivity of the polymer to hydrolysis in the presence of acid
catalyst was noted earlier. This interfacial effect was superimposed on the degradation of
the nylon-6 itself that occurred under the reactive-extrusion conditions (which may have
been linked to residual moisture).
Many commercial polymer formulations contain pigments, stabilizers, processing aids etc.
and the chemistry that may occur during high-temperature processing is complex. There has
been recent interest in the use of clay nanollers to produce thermoplastic nanocomposites in
which there is rstly intercalation and then exfoliation of the clay by the polymer. This has
resulted in polyamides with greatly elevated heat-distortion temperatures. These may be
formed either by in situ polymerization of caprolactam (Kojima et al., 1993) or by melt mixing
of the polyamide with a clay that has been modied with an organic reagent such as an alkyl
quaternary ammonium salt or long-chain fatty acid to assist in intercalation. It has been
observed that when injection moulding of these nanocomposites occurred at 295 C there was a
decrease in the molar mass of the nylon-6 (Davis, 2003). From a consideration of the degradation route for polyamides under various conditions (Levchik et al., 1999) it was concluded
that the main reason for degradation was scission of the amide bond due to the release of water
from the surface of the clay and/or the polymer when held for long times (12.5 min) at about
300 C. This water could also come from dehydroxylation of the aluminosilicate clay. The
catalytic role that the clay may play has been hypothesized as involving sites such as Fe2 and
Fe3 or, alternatively, the cation sites that had not exchanged with the organic modier that
facilitates intercalation (Davis, 2003). No conclusive evidence for a mechanism was provided.
Nylon-6 without added montmorillonite did not degrade under the same conditions, and it
was unambiguously the case that the clay had a particular role to play in promoting the
hydrolytic degradation through the different product distribution. In the absence of the clay
(and nucleophile) only thermal degradation took place, with caprolactam being the principal
product above 200 C and larger cyclics forming only above 390 C. The latter temperatures
are unrealistic for reactive processing. These studies focussed on conditions of injection
moulding.
In a related study of nylon-6 nanocomposite processing with a twin-screw extruder, it was
found that the level of degradation was related both to the type of nylon and to the surfactant
used in the treatment of the clay (Fornes et al., 2003). The higher rate of degradation and
discolouration in higher-molar-mass polymer was linked to the increase in clay-platelet
exfoliation. The temperature employed (240 C) was somewhat lower than in the above
example of injection moulding and this may be responsible for the elimination of hydrolysis
as a major contributor to degradation and attention being focussed on the reactions of the
surfactant used for exfoliation. Interestingly, it was found that these degradation reactions did
not appreciably affect the morphology and physical properties of the nanocomposite. In an
extension of this study to polycarbonate nanocomposites processed at 260 C, it was also
found that the presence of unsaturation or hydroxy-ethyl groups in the surfactant used in the
clay increased the darkening of the polymer. The presence of iron sites in several of the clays
was also linked to this darkening.
References
Adebayo, G., Koombhongse, R. & Cakmak, M. (2003) Int. Polym. Processing, 18, 260272.
Aerts, J. (1998) Comput. Theor. Polym. Sci., 8, 4954.
Al-Malaika, S. (1989) Effects of anti-oxidants and stabilizers, in Allen, G. (Ed.) Comprehensive
Polymer Science, Oxford: Pergamon.
References
163
164
Chong, Y. K., Le, T. P. T., Moad, G., Rizzardo, E. & Thang, S. H. (1999) Macromolecules,
32, 20712074.
Chynoweth, K. (1989) Glass transition and crystallization, in Cook, W. D. & Guise, G. B. (Eds.)
Polymer Update: Science and Engineering, Melbourne: Royal Australian Chemical Institute.
Ciriscioli, P. R. & Springer, G. S. (1990) Smart Autoclave Cure of Composites, Lancaster, PA:
Technomic.
Clagett, D. C. & Shafer, S. J. (1989) Polycarbonates, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
Cole, K. C., Hechler, A. A. & Noel, D. (1991) Macromolecules, 24, 30983110.
Commerce, S., Vaillant, D., Phillippart, J. L., Lacoste, J. & Carlsson, D. J. (1997) Polym. Deg.
Stab., 57, 175182.
Cook, W. D., Moad, G., Fox, B., et al. (1996) J. Appl. Polym. Sci., 62, 17091714.
Corradini, P. & Busico, V. (1989) Monoalkene polymerization: stereospecity, in Allen, G. (Ed.)
Comprehensive Polymer Science, Oxford: Pergamon.
Cowie, I. A. (1989a) Block and graft copolymers, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
Cowie, J. M. G. (1989b) Alternating copolymerization, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
Crivello, J. V. (1999) J. Polym. Sci. A: Polym. Chem., 37, 42414254.
Dainton, F. R. (1956) Chain Reactions, London: Methuen.
Darling, T. R., Davis, T. P., Fryd, M. et al. (2000) J. Polym. Sci. A: Polym. Chem., 38, 17061708.
Davis, R., Gilman, G.J., Van Der Hart, D.I. (2003) Polym. Deg. Stab., 79, 111121.
Dawkins, J. V. (1989) Size exclusion chromatography, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
De Bakker, C. J., St John, N. A. & George, G. A. (1993) Polymer, 34, 716726.
Doi, M. & Edwards, S. F. (1986) The Theory of Polymer Dynamics, Oxford: Clarendon Press.
Dong, J., Fredericks, P. M. & George, G. A. (1997) Polym. Deg. Stab., 58, 159169.
Dusek, K. (1986) Network formation in curing of epoxy resins, in Dusek, K. (Ed.) Advances in
Polymer Science 78: Epoxy Resins and Composites III, Berlin: Springer-Verlag.
Eldarov, E. G., Mamedov, F. V., Goldberg, V. M. & Zaikov, G. E. (1996) Polym. Deg. Stab.,
51, 271279.
Emanuel, N. M. & Buchachenko, A. L. (1987) Chemical Physics of Polymer Degradation and
Stabilization, Utrecht: VNU Science.
Fischer, H. (2001) Chem. Rev., 101, 35813610.
Flory, P. J. (1953) Principles of Polymer Chemistry, New York: Cornell University Press.
Flory, P. J., Allcock, H. R., Eisenberg, H., Feldman, R. J. & Yoon, D. Y. (1982) The science of
macromolecules, in Seitz, F. (Ed.) Outlook for Science and Technology: The Next Five Years,
San Francisco, CA: W. H. Freeman.
Fontanille, M. (1989) Carbanionic polymerization: general aspects and initiation, in Allen, G. (Ed.)
Comprehensive Polymer Science, Oxford: Pergamon.
Fornes, T. D., Yoon, P. J. & Paul, D. R. (2003) Polymer, 44, 75457556.
Fradet, A. & Arlaud, P. (1989) Unsaturated polyesters, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
Frechet, J. M. & Tomalia, D. A. (2001) Introduction to the dendritic state, in Frechet, J. M. &
Tomalia, D. A. (Eds.) Dendrimers and other dendritic polymers, Chichester: John Wiley and Sons.
Frechet, M. J. & Hawker, C. J. (1996) Synthesis and properties of dendrimers and hyperbranched
molecules, in Allen, G. (Ed.) Comprehensive Polymer Science, Oxford: Pergamon.
Frisch, K. C. & Klempner, D. (1989) Polyurethanes, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
References
165
166
Knop, A., Bohmer, V. & Pilato, L. A. (1989) Phenolformaldehyde polymers, in Allen, G. (Ed.)
Comprehensive Polymer Science, Oxford: Pergamon.
Koenig, J. L. (1999) Spectroscopy of Polymers.
Kojima, Y., Yusuki, A., Kawasumi, M. et al. (1993) J. Polym. Sci.: Polym. Chem., 31, 17551758.
Kozielski, K. A., George, G. A., St John, N. A. & Billingham, N. C. (1994) High Perform. Polym.,
6, 263286.
Kwei, T. K. & Wang, T. T. (1978) Phase separation behavior of polymerpolymer mixtures, in
Paul, D. A. N. (Ed.) Polymer Blends, New York: Academic Press.
La Mantia, F. P. & Vinci, M. (1994) Polym. Deg. Stab., 45, 121125.
Lambla, M. (1992) Reactive processing of thermoplastic polymers, in Allen, G. (Ed.) Comprehensive Polymer Science, Oxford: Pergamon.
Lederer, A., Voigt, A., Clausnitzer, C. & Voit, B. (2002) J. Chromatogr. A, 976, 171179.
Levchik, S. V., Weil, E. D. & Lewin, M. (1999) Polym. Int., 48, 532557.
Manaresi, P. & Munari, A. (1989) Step polymerization general aspects, in Allen, G. (Ed.)
Comprehensive Polymer Science, Oxford: Pergamon.
Mandelkern, L. (1989) Crystallization and melting, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
Manzione, L. T. & Gillham, J. K. (1981) J. Appl. Polym. Sci., 26, 889905.
Matejka, L. & Dusek, K. (1989) Macromolecules, 22, 29112917.
Matejka, L., Spacek, P. & Dusek, K. (1991) Polymer, 32, 31903194.
Matyjaszewski, K. (1999) Similarities and discrepancies between controlled cationic and radical
polymerizations, in Puskas, J. E. (Ed.) Ionic Polymerizations and Related Processes, Dordrecht:
Kluwer Academic.
Mayadunne, R. T. A., Rizzardo, E. et al. (1999) Macromolecules, 32, 69776980.
Mayadunne, R. T. A., Rizzardo, E. et al. (2000) Macromolecules, 33, 243245.
Mayo, F. R. (1972) J. Polym. Sci. Polym. Lett. Edn., 10, 921923.
Mayo, F. R. (1978) Macromolecules, 11, 942946.
McKenna, G. B. (1989) Glass formation and glassy behaviour, in Allen, G. (Ed.) Comprehensive
Polymer Science, Oxford: Pergamon.
McNeill, I. C. (1989) Thermal degradation, in Allen, G. (Ed.) Comprehensive Polymer Science,
Oxford: Pergamon.
Mijovic, J. & Andjelic, S. (1995) Macromolecules, 28, 27872796.
Moad, G. (1999) Progress in Polymer Science, 24, 81142.
Morrison, N. J. & Porter, M. (1989) Crosslinking of rubbers, in Allen, G. (Ed.) Comprehensive
Polymer Science, Oxford: Pergamon.
Muller, A. H. E. (1989) Carbanionic polymerization: kinetics and thermodynamics, in Allen, G.
(Ed.) Comprehensive Polymer Science, Oxford: Pergamon.
Nagahata, R., Sugiyama, J. J., Goyal, M. et al. (2001) Polymer, 42, 12751279.
Nuyken, O. & Pask, S. D. (1989) Carbocationic polymerization: alkenes and dienes, in Allen, G.
(Ed.) Comprehensive Polymer Science, Oxford: Pergamon.
Odian, G. (1991) Principles of Polymerization, New York: John Wiley and Sons.
Oh, S. J., Lee, S. K. & Park, S. Y. (2006) Vib. Spectrosc., 42, 273277.
Olabisi, O., Robeson, L. M. & Shaw, M. T. (1979) PolymerPolymer Miscibility, New York:
Academic Press.
Painter, P. C. & Coleman, M. M. (1994) Fundamentals of Polymer Science, Lancaster, PA:
Technomic.
Papathanasiou, T. D., Higgins, J. S. & Soontaranum, W. (1998) Polym. Polym. Composites, 6, 223227.
Pascault, J.-P., Sauterau, H., Verdu, J. & Williams, R. (2002) Thermosetting Polymers, New York:
Marcel Dekker.
References
167
168
2.1
Chapter rationale
This chapter focusses on the physical properties and models of network and reactively
modied polymers. Understanding changes in physical properties during curing, in tandem
with changes in chemical properties (Chapter 1), and chemorheological properties
(Chapter 4), is essential to fully characterizing network and reactively modied polymer
systems. This chapter will rst give a brief introduction to polymer physics and dynamics
before focussing on redening network and reactively modied polymer systems. Then it will
focus on dening the key changes in physical properties during cure. Finally this chapter will
focus on key experimental techniques for describing changes in physical properties during cure.
2.2
2.2.1
170
2:1
r 1 2=wwvRT=2k2 1=k
2:2
for the James and Guth model, where r is the stress, w is the functionality of the crosslink, m is
the number of crosslinks, R is the gas constant, T is the temperature and k is the deformation.
Modications to these classical statistical models can also be made, such as by the
incorporation of loose chains (non-load-bearing chains), physical crosslinks (temporary or
permanent) and intramolecular crosslinks (loops). At higher deformations (strains) or
increasing crosslink densities, it may be necessary to use a non-Gaussian statistical treatment that considers the nite extensibility of the chain. Non-Gaussian models are reviewed
extensively by Treloar (1975).
In reality, of course, polymer systems are varied and consist of polymer solutions,
polymer melts, and polymer networks and gels, and the polymer dynamics of these systems
should be examined in detail.
2.2.2
Introduction to rheology
Rheology is the study of uid ow and deformation, and provides the specic science to
characterize polymer dynamics. There are many types of ow, but generally they are
classied into shear or elongational ow. In a localized sense shear ow involves shearing
of molecules or polymer chains over one another, whereas elongation ow involves
localized stretching of molecules or polymer chains. In shear or elongational rheology one
can further subclassify systems into steady (time-independent and steady-state), dynamic
(time-independent and oscilliatory) and transient (time-dependent) ows.
Steady-state shear rheology typically involves characterizing the polymers response to
steady shearing ows in terms of the steady shear viscosity (g), which is dened by the ratio
of shear stress (r) to shearing rate (c 0 ). The steady shear viscosity is thus a measure
of resistance to steady shearing deformation. Other characteristics such as normal stresses
(N1 and N2) and yield stresses (rY) are discussed in further detail in Chapter 3.
171
Dynamic shear rheology involves measuring the resistance to dynamic oscillatory ows.
Dynamic moduli such as the storage (or solid-like) modulus (G0 ), the loss (or uid-like)
modulus (G 00 ), the loss tangent (tan d G 00 /G0 ) and the complex viscosity (g*) can all be
used to characterize deformation resistance to dynamic oscillation of a sinusoidally imposed
deformation with a characteristic frequency of oscillation (x).
Transient shear ows involve examining the shear stress and viscosity response to a timedependent shear. The stress build up at the start of steady ow (r) and at the cessation of
steady ow (r) and the stress decay (r(t)) after a dynamic instantaneous impulse of
deformation strain (c) can be used to characterize transient rheological behaviour.
Elongational behaviour under steady and transient conditions can also be characterized,
and more information is provided in Chapter 3. This short introduction to rheology should
provide enough background for further interpretation of polymer-dynamic theories of
suspensions, polymer solutions and polymer melts.
Suspensions
A suspension is a dispersion of particles within a solvent (usually a low-molar-mass liquid).
Thermodynamics (Brownian motion and collisions) favours the clumping of small particles,
and this can be increased by ow. However, particles over 1 lm tend to settle under gravity,
unless stability measures have been considered (matching the density of the particle to that
of the medium, increasing the Brownian/gravitational force ratio, electrostatic stabilization,
steric stabilization). Other complications can occur in the dynamics of suspensions, such as
particle migration across streamlines, particle inertial effects and wall slip (Larson, 1999).
Early work (Einstein, 1906) showed that the shear viscosity, g, of a suspension at low
volume fractions (u < 0.03) can be described by
g gs 1 2:5u;
2:3
2:4
2:5
lim u!0
g gs =ugs :
2:6
At very high concentrations, however, these equations fail to account for the overgrowing,
and one should use the empirical KreigerDougherty equation (Kreiger and Dougherty,
1959)
g gs 1 u=um gum ;
2:7
172
Figure 2.1. The dependence of the critical stress for shear thinning on the volume fraction of
sterically stabilized hard spheres. From Larson (1999, Figure 6.6) by permission of Oxford
University Press, Inc.
2:8
where gr g/gs is the reduced viscosity, gr0 is the low-shear-rate viscosity, gr1 is the highshear-rate viscosity, rr ra3 /(kBbT) is the reduced shear stress and b is a tting parameter.
Deviations from this equation at high shear rates can occur due to shear thickening (Larson,
1999). Also a critical stress for shear thinning to occur, rc, has been found (Larson, 1999)
and is shown in Figure 2.1. Note that the effect of high concentrations is to reduce the
critical stress to shear thinning.
An excellent discussion in Larson (1999) examines additional effects such as steric
stabilization, particle-size distributions, shear thickening, and deformable particles and
droplets. It is interesting to examine the equation for deformable droplets,
5M 2
g gs 1
u ;
2:9
2M 1
where M gd/gs is the ratio of the viscosity of the droplet to the viscosity of the solution.
One can see that, in the limit as M approaches 1 (hard sphere), the equation reverts back to
Equation (2.3) (Einsteins equation). An excellent section follows in Larson (1999) about
droplet break up and coalescence and the application of these equations to foams, emulsions
and immiscible liquids.
Polymer solutions
A polymer solution is a mixture of a solvent (usually a low-molar-mass liquid) and a
solute (usually a high-molar-mass polymer). For dilute polymer solutions (from suspension theory and [g] 2.5u) it is simple to show from geometric considerations that the
viscosity varies as
g k Mw0:5 :
2:10
173
2:11
2:12
where, n0 is the number of dumbells per unit volume, kB is the Boltzmann constant, T is the
temperature and k is the relaxation time. Note that no shear thinning is predicted by this
theory.
The Zimm model (Zimm, 1956) extends the spring model by considering intermolecular
forces such as hydrodynamic forces (perturbations of the velocity eld near beads by other
beads), reduced excluded-volume effects (coil expansion and reduced contacts), non-linear
spring forces (nitely extendable springs) and internal viscosities (coil sluggishness). One can
obtain the following expression for the viscosity from the Zimm model:
g gs n0 kB Tk=1 k2 x2 ;
2:13
where x is the dynamic shear rate (or frequency). Note that this expression does account for
shear thinning.
The Rouse model (Rouse, 1953) extends these theories to multiple beads and springs (or
multiple-relaxation modes). Here the expression for the viscosity becomes
g
N
X
gs n0 kB Tki =1 k2i x2 :
2:14
i1
For more concentrated solutions one must begin to consider interactions with neighbouring
polymers. The expression for viscosity for concentrated solutions is
g gs 1 gc k 0 g2 c2 :
2:15
For the coil-overlap region (where polymer coils overlap) the viscosity is
g gs =gs c g expkcg;
2:16
where coil overlap occurs when c[g] ( cMw ) 110. Figure 2.2 further characterizes
concentrated systems. The dilute region is dened as that with concentration less than the
critical overlap concentration (c*), the semi-dilute region is that with concentration above
the overlap concentration but below the critical packing concentration (c) and the
concentrated region is that with concentration above the critical packing concentration.
Physically what is happening in these regions is shown in Figure 2.3.
Here at low concentration (c < c*) the chain dimension, r is unaffected by its neighbours,
but at higher concentrations (c > c*) the chain dimension is reduced. Eventually, at very
high concentrations (c > c), the polymers are well entangled and no further reduction in
spacing can occur. Also, as shown in Figure 2.4, the viscosity behaviour becomes different
above a critical concentration times molecular weight (cMw).
a
174
Figure 2.2. Concentration regimes in good solvents. From Macosko (1994, Figure 11.3.2).
Copyright (1994). Reprinted with permission of John Wiley and Sons, Inc.
Figure 2.3. Chain dimension versus concentration. Adapted from Macosko (1994, Figure 11.3.5).
Copyright (1994). Reprinted with permission of John Wiley and Sons, Inc.
Polymer melts
In some respects the dynamics of melts is simpler than the dynamics of polymer solutions.
In melts, for example, the excluded-volume effect of solutions and hydrodynamic interactions of molecules within the same chain are screened by surrounding chains (hydrodynamic screening). However, the complication of melts is that the motion of each chain is
affected by the entanglements in the surrounding chain. Grassley (1982) noted that there
was a co-operative relaxation process in polymer melts that was characterized by a relaxation time (k). De Gennes (1971) noted that the entanglements of surrounding chains
restricted the motion of a polymer chain, and described a polymer chain as surrounded by an
effective tube, where the polymer chain would have to undergo reptation or snake-like
motion to relax (or get out of tube). The time taken to reptate out of the tube was
175
Figure 2.4. Typical viscositymolar-mass dependence for molten polymers. From Macosko (1994,
Figure 11.5.4). Copyright (1994). Reprinted with permission of John Wiley and Sons, Inc.
characterized by a reptation relaxation time (krep). It can be shown from this theory that
the viscosity should scale with the weight-average molar mass as g Mw3; however,
experimentally this relationship has been shown to be g Mw3.4 (Berry and Fox, 1968).
Doi and Edwards (1978a, 1978b, 1978c) extended the work of de Gennes by incorporating non-linear deformation at higher rates. Shear thinning and elasticity are predicted with
the DoiEdwards model, but overpredictions of shear thinning can occur at high shear rates.
Extensions to this work include incorporation of primitive path uctuations, constraint
release, tube stretching, polydispersity, star molecules and long-chain branching, and are
well summarized in Larson (1999).
2.3
176
Figure 2.5. A typical polymer gel network. From Larson (1999, Figure 5.1) by permission of Oxford
University Press, Inc.
2.3.1
Network polymers
The structure of a network polymer is sketched in Figure 2.5 (de Gennes, 1979).
Network polymers may be formed via chemical or physical interactions. There are three
types of chemical reactions that produce chemical gels (de Gennes, 1979).
stepwise (including condensation) polymerization (Section 1.2.1)
addition polymerization (Section 1.2.4)
vulcanization (Section 1.4.1)
These reactions have been explained in detail in the sections of Chapter 1 indicated and
they are all identied by permanent, covalent bonds between polymeric chains. These
chemical-network polymer systems are an important sub-class of network polymers and are
sometimes referred to as thermoset polymers. Thermoset polymers cannot be re-melted or
undergo ow under the inuence of heat once they have been polymerized (i.e. they have a
thermoset morphology, unlike thermoplastic polymers, which can be re-melted and ow
under application of heat). Owing to their excellent properties at high temperatures (and
under extreme environmental conditions) thermoset polymers are suitable for a wide variety
of high-performance applications such as electronic packaging, automotive panelling and
177
physical processes
chemical processes
polymerization
phase transition
cross-linking
entanglements
(multifunctional)
curing
vulcanization
liquid-liquid
irradiation
combination crystallization
(L L) segregation
(L C)
associations
polar
complex
colloid
Figure 2.6. Processes causing network junctions. Reproduced by permission of the Royal Society of
Chemistry from Figure 1 (Keller, 1995).
aerospace applications. Thermoset polymers have a wide range of applications because their
nal physical properties may be tailored by control of the initial monomer chemistry,
control of cure conditions, addition of llers and control of the reactive processing.
For physical gelation de Gennes (1979) described the following pathways:
local helical structures
microcrystallite junctions
nodular domains of end groups or parts of chains
Typical intermolecular forces or associations in physical gels are van der Waals forces,
electrostatic forces and hydrogen bonding, all of which are characterized by weak,
reversible links.
An instructive sketch by Keller (1995) is shown in Figure 2.6. This incorporates a broader
denition of network polymers (termed processes creating junctions) that incorporates other
forms of reactions, but essentially follows the same lines as the physical and chemical
classication of de Gennes.
An important parameter in determining the network structure in a network polymer
system is the functionality of the monomers. The functionality ( f ) of a molecule is the
number of bonds it can form with other molecules (Larson, 1999), thus the higher the
functionality the more likely it is that the molecules may react to develop a network system.
Of course, during cure as molecules interact the functionality of the growing intermediate
chains may increase, further accelerating cure.
2.3.2
178
Explanation
Bulk polymerization
Grafting reaction
Inter-chain
copolymerization
Coupling or crosslinking
reactions
Controlled degradation
Functionalization
Example systems
reactive moulding
irradiation processing
Reactive extrusion is well surveyed by a recent text (Xanthos, 1992), and has been
referred to as the deliberate performance of chemical reactions during continuous extrusion
of polymers and/or polymerizable monomers (Xanthos, 1992). This usually represents
chemical reactions conducted during a continuous extrusion process at relatively short
residence times. Examples of types of reactions and example polymer systems used in
reactive-extrusion processes are given in Table 2.1.
Clearly reactive extrusion is attractive for production in that it provides chemical
modication of a polymer system in a continuous process. The disadvantages of this process
can be lack of process control and the inability to understand fully the degree of modication during the process.
Reactive batch compounding involves the use of batch mixers or kneaders, which are
usually employed as long-residence-time chemical reactors in the compounding industry.
Reactive compounding can be further categorized, like conventional polymer compounding,
in terms of intensive (break-up of agglomerates or droplets) or extensive (reducing the nonuniformity of distribution without disturbing the initial scale of dispersant) mixing. Note
that reactive compounding, as opposed to conventional polymer compounding, in addition
involves a chemical reaction in parallel with the mixing process. Example systems
179
Type of polymer
Temperature
crosslinked wires (PE, PVC, PE-EVA, PU) for increased heat and abrasion resistance
radiation-crosslinked rubber tyres
heat-shrinkable (polyolen and uoropolymer) tubing and sheets
crosslinked hydrogels for contact lenses and wound dressings
medical-supply sterilization (PP syringes)
The aim of this book is not to discuss each of these reactive modication techniques in
detail, but rather to show how chemorheology can be a useful tool from fundamental
understandings to understanding processing, process control and optimization.
In this chapter we are interested in understanding the physical properties of a curing
system. We shall consider this by rst describing the important transitions (gelation, vitrication and phase transitions) that occur during curing and then examining techniques
used to characterize changes in physical properties in curing systems. We shall address the
detailed chemorheological behaviour in Chapter 4.
2.4
180
Equilibrium properties
Liquid
Solid
Ge
pc
Conversion p
Figure 2.7. The dependence of zero-shear viscosity g0 and equilibrium modulus G0 on conversion
p for a crosslinking system. Reprinted from Figure 1 p. 344 (Winter 1990). Copyright (1990).
Reprinted with permission of John Wiley and Sons, Inc.
important milestones to measure. The chemical reactions underlying these processes are
described in Section 1.2.2.
2.4.1
181
material is fully crosslinked and there is no further reaction (Enns and Gillham, 1983).
However, other papers (Huang and Williams, 1994, Barral et al., 1995) indicate that further
curing is possible after the vitrication point. Of course, vitrication tends to be a transition
region rather than a single transition point due to the fact that there is a distribution of
polymer chain lengths in the curing polymer system. That is, vitrication begins when
higher-molar-mass species enters the glassy regime (or cannot diffuse) and ends when the
lower-molar-mass species reaches the glassy state (or cannot diffuse).
2.4.2
Phase separation
During cure phase separation whereby the solid phase separates from the unreacted liquid
phase may also occur (de Gennes, 1979). Thermoplastic-modied thermoset systems are
useful as toughened thermosets that have improved ductility over that of conventional
thermosets. Reactive-phase separation is thus another important transition in curing systems, and this process is controlled by the interplay between the thermodynamics of phase
separation and the cure. For example, other work (Girard-Reydet et al., 1998) has shown
that important factors in controlling phase separation in polyetherimide-modied epoxyresin systems are the proximity of the initial modier concentration to the critical concentration and the ratio of the phase-separation rate to the polymerization rate. It is known,
for phase-separation processes, that for materials near the critical concentration (as determined by mixing theory for reactive blends (Clarke et al., 1995, Inoue, 1995) and based on
critical miscibility) phase separation occurs unstably via spinodal de-mixing (as mentioned
in Section 1.3.2), whereas for off-critical concentrations, phase separation proceeded via a
stable nucleation/growth mechanism. Further excellent studies (Bonnet et al., 1999a, 1999b)
have shown the effects of the phase separation on the cure kinetics and rheology of polyetherimide-modied DGEBA/MCDEA epoxies, highlighting again the interdependence
of phase separation and curing in determining epoxy-resin properties. Additionally, both
cure and post-cure temperatures may control the extent of phase separation due to vitrication halting the evolution of morphologies. Clearly, to optimize the nal morphology of
phase-separated thermoset systems one needs to monitor carefully the interdependence of
phase-separation and cure properties.
Gelation, vitrication and phase-separation transitions in curing systems are very well
described in the work of Gillham (1986) via the use of timetemperature-transformation
diagrams.
2.4.3
ll c
Fu
GEL
RUBBER
e
ur
Vitr
SOL/GEL
RUBBER
Tg
CHAR
ifica
tion
Devitrifi
cation
GEL G
Temperature, Tcure
182
LASS
Ge
Tg = T
lat
Ph
as
ion V
itr
if
ica
se
tio
pa
ra
t
SOL/GEL GLASS
io
LIQUID
Tg,gel
Tg = T
g,gel
Tgo
SOL GLASS
Log time
Figure 2.8. A timetemperature-transformation (TTT) isothermal cure diagram for a thermosetting
temperature of the initial uncured system, so for any Trxn below this temperature the
system will remain unreacted. Tg,gel is the temperature at which gelation and vitrication
coincide. At Trxn between Tg0 and Tg,gel the system is initially a liquid, but will vitrify
before gelling. Tg1 is the maximum Tg of the cured system. At any Trxn between Tg,gel and
Tg1 the system is initially a liquid and will reach gelation (or the formation of two phases:
a nite-molar-mass solvent and an innite-molar-mass gel). Eventually the Tg of the
system will approach the Trxn and vitrication will occur. The full cure line represents the
time required at any Trxn for Tg to equal Tg1, and is important for allowing comparative
analyses of systems. Above Tg1 the material will gel but will not vitrify until degradation
(charring).
It should be noted that in Gillhams work he utilized a torsional braid analyser (TBA),
which determines the gelation and vitrication points via the change in resistance to
torsional motion of a xed braid support in an epoxy-resin matrix (more details are given
later in Chapter 3). Thus gelation and vitrication were denoted by two peaks in the
damping function the rst peak was gelation and the later peak was vitrication. Note
that Stutz and Mertes (1989) found for their epoxy-resin systems that the rst peak was
not caused by gelation of the resin, but occurred at approximately the same viscosity and
was observed only when a bre braid was used as support in a TBA. Subsequently
Cadenato et al. (1997) developed TTT diagrams based on TMA and DMTA tests that
reveal discrepancies between the methods for determining gel and vitrication times.
Mijovic et al. (1996) utilized FT-IR, dielectric and rheological measurements to determine the kinetics and transition data for a TTT diagram of a trifunctional epoxyaniline
system, and found good agreement among results obtained with FT-IR, dielectric and
rheometric methods to determine transitions. Overall it is recommended that any work on
183
developing TTT diagrams should utilize several different techniques to conrm kinetics,
gelation and vitrication, and these methods will be discussed in more detail later in this
chapter and in Chapters 3 and 4.
The TTT diagram has been utilized for many reactive systems. Wissanrakkit and Gillham
(1990) show the kinetics and TTT diagram for a stoichiometrically balanced DGEBA/
TMAB epoxyamine system, whilst Simon and Gillham (1992) show the TTT diagram for
an off-stochiometric DGEBA/TMAB epoxyamine system. Pang and Gillham (1989) and
Wang and Gillham (1993a) show the usefulness of using a TTT diagram to describe the
effects of physical ageing of high-Tg epoxyamine systems. Pang and Gillham (1989) also
show that the density of an epoxyamine system has a maximum at gelation with respect
to conversion. This is stated to be due to the effects of physical ageing prior to gelation,
which increases density, competing against the chemical ageing after gelation (which
decreases density). Vendetti et al. (1995a) conrmed that there is a specic volume increase
(density decrease) with increasing conversion after gelation, and this was subsequently
conrmed by PALS testing in a later paper (Vendetti et al., 1995b).
Such TTT diagrams have also been useful in describing the cure of polyimide systems, as
shown in Palmese (1987), which shows the TTT diagram of a polyamicacid/polyimide
system. The TTT diagram of a polycyanurate system is developed by Simon (1993) on
the basis of FT-IR, DSC and torsional braid measurements. Kim et al. (1993) developed a
TTT diagram for a thermosetthermoplastic blend, specically a tetrafunctional epoxyresin/poly(ether sulfone)/dicyandiamide thermosetthermoplastic blend (Figure 2.9).
Note that Figure 2.9 shows the transitions in terms of (1) onset of phase separation,
(2) gelation, (3) xation of the dimension of phase-separated structure, (4) the end of phase
separation and (5) vitrication. Lu and Pizzi (1998) investigated the curing of phenol
and urea formaldehyde wood adhesives in a TTT diagram. Special mention is made of the
effect of the wood substrate on the cure of this composite system.
Other useful diagrams have been developed from the TTT diagram. Wang and Gillham
(1993b) show the use of a Tgtemperatureproperty (TgTP) diagram for an epoxyamine
system as illustrated in Figure 2.10.
In this diagram the x-axis is the extent of cure (Tg) and the y-axis is the temperature. The
properties of the curing system may be separated into distinct regions of the glass transition
(Tg), the beta-transition (b-Tg) , the end of the glass transition (e; Tg) and gelation (Tg,gel).
Properties may also be implied on the third axis out of the plane of the paper. Thus
the physical properties (behaviour) of a system with respect to conversion are determined by
its temperature and Tg.
Wang and Gillham (1993b) and Simon and Gillham (1994) show the use of a continuous-heating (CHT) diagram as illustrated in Figure 2.11. Essentially the diagram
includes the basic events (vitrication, gelation, devitrication) of a TTT diagram;
however, the key additions to the diagram are the constant-heating-rate curves (where
rate 1 > rate 2 etc.). The heating-rate curves are used to show transitions passed through
under non-isothermal heating rates. Simon and Gillham (1994) also incorporate lines of
constant Tg (iso-Tg curves) into a TTT diagram to produce the iso-TgTTT diagram (refer
to Figure 2.12). Essentially these diagrams allow one to determine the time taken to reach
a specic Tg at a given isothermal curing temperature.
The conversion-temperature diagram (CTT) is described in Figure 2.13 (Prime, 1997).
Here conversion is plotted against temperature and various regions are identied. The gel
conversion (agel) is noted on the y-axis as a constant with respect to temperature (but may
thermoplastic blend system. Reprinted from Figure 12 from (Kim et al., 1993). Copyright (1993),
with permission from Elsevier.
Temperature
184
SOLGEL RUBBER
REGION
GLASS-TRANSITION
ion REGION
t
a
c
i
vitrif
Tg
Tg0
Tg
UNGELLED GLASS
REFION
on
-transit
T, gel
GELLED GLASS REGION II
gelation
Tg, gel
T0
Tg0
Tg
Figure 2.10. A Tgtemperatureproperty (TgTP) diagram. Adapted from Figure 1 from (Wang
and Gillham, 1993b). Copyright (1993), with permission from Elsevier.
185
200
Tg = 180 C
180
160
140
120
100
80
60
Vitrification
Gelation
Iso-Tg Contour
(for Tg = 0 to 180 C, every 20 C) Tg0 = 9 C
40
20
0
20
1
SOL/GEL
RUBBER
GEL
RUBBER
GELATION
VITRIFICATION
Tg
GEL GLASS
FULL CURE
TC
INCREASING
CONVERTION
SOL/GEL
GLASS
LIQUID
Tg,gel
Tg0
SOL GLASS
LOG TIME
Figure 2.12. An iso-Tg-TTT diagram. Adapted from Figure 7 from (Simon and Gillham, 1994).
Copyright (1994), with permission from Elsevier.
change depending on variations of chemical reactions with temperature). The x-axis has
Tg0,gel Tg,gel, Tg1 and Td (the decomposition temperature) noted. The Tgconversion curve
is also plotted in the diagram.
186
RUBBER
gel
UNGELLED
GLASS
DECOMPOSITION
CONVERSION
g V
S.
C
O
NV
ER
SI
O
N
GELLED
GLASS
LIQUID
0
Tg0
Tg, gel
Tg
Td
TEMPERATURE
Figure 2.13. A conversiontemperature-transformation (CTT) diagram. Reprinted from
(Prime, 1997). Copyright (1997), with permission from Elsevier.
Also Osinski (1993) develops the aTT or TTaT diagram, on which the transitions may
be expressed as a function of conversion (a) instead of time (Figure 2.14).
Obviously these curves are exible in that they may be used for isothermal and
non-isothermal cure.
2.4.4
2.5
187
30
60
90
12
01
1
50
80
0
21 20 40
60 8
0 100
1
20 14
0 16
0 180
200
1.0
1.0
F CURE
DEGREE O
0.8
0.8
0.6
0.6
0.4
0.4
0.2
18
21
0.2
in)
(m
12
ME
90
60
TI
40
60
TEM 80 1
PER 00 12
ATU 0 14
RE 0 16
(c) 0 18
02
30
20
15
0.0
00
from Figure 7 (Osinski, 1993). Copyright (1993), with permission from Elsevier.
2.5.1
Network models
Dusek (1986a) characterizes network-formation models into the following categories:
spatially independent and spatially dependent models. Of the spatially independent models,
there are statistical models (in which network structure is developed from various interacting monomer units) and kinetic models (in which each concentration of species is
modelled by a kinetic differential equation).
Statistical network models were rst developed by Flory (Flory and Rehner, 1943,
Flory, 1953) and Stockmayer (1943, 1944), who developed a gelation theory (sometimes
referred to as mean-eld theory of network formation) that is used to determine the gelpoint conversions in systems with relatively low crosslink densities, by the use of probability to determine network parameters. They developed their classical theory of network
development by considering the build-up of thermoset networks following this random,
percolation theory.
Percolation is the process of network formation by random lling of bonds
between sites on a lattice. If one increases the fraction of bonds (p) formed, then larger
clusters of bonds form until an innite lattice-spanning cluster (at the percolation
threshold, p pc) is formed. Figure 2.15 (Stanley, 1985) shows the growth of the
network corresponding to values of the fraction p of (a) 0.2, (b) 0.4, (c) 0.6 (p pc)
and (d) 0.8.
188
Figure 2.15. The phenomenon of bond percolation from a fraction of links, p, from 0.2 to 0.8.
Reprinted from Figure 16 (Stanley, 1985). Reprinted with permission of John Wiley and Sons, Inc.
Copyright (1985).
The gel point in this classical theory is the percolation threshold (pc) given by
pc 1= f 1;
2:17
where f is the co-ordination number of the tree or the number of bonds that can form at
each site (Flory, 1953).
Figure 2.16 (Stanley, 1985) shows the analogy between (a) the gelation threshold and
(b) the percolation thershold, and R(a) is the probability that there exists an innite
cluster.
It should be noted that this theory neglects loops or cyclic formations, and this affects the
size distribution and other cluster properties. Some of these properties (and their relationships at gelation) are highlighted in Table 2.3 (Larson, 1999), as are their experimental
values compared with the classical and three-dimensional-percolation theoretical values at
gelation.
Note that power-law behaviour is prevalent at gelation. This has been proposed to be due
to a fractal or self-similar character of the gel. Note that the exponent Df is termed the
fractal dimension. For any three-dimensional structure (D 3) the exponent Df 3 (where
Df < 3 indicates an open structure and Df 3 indicates a dense structure). Also Muthukumar (Muthukumar and Winter, 1986, Muthukumar, 1989) and Takahashi et al. (1994)
show explicitly the relationship between fractal dimension (Df) and power-law index of
viscoelastic behaviour (n). Interestingly, more recent work (Altmann, 2002) has also shown
a direct relationship between the power-law behaviour and the mobility of chain relaxations,
which will be discussed further in Chapter 6.
189
Table 2.3. Scaling exponents for classical and percolation theories of gelation
Exponent
Relation
Classical
Three-dimensional
percolation
Experimental
k
r
c
t
Df
b
N(m) ~ mk
Mz ~ e1/r
Mw ~ ec
Rz ~ et
RDf ~ M
P ~ eb
5/2
1/2
1
1
4
1
2.20
0.45
1.76
0.89
2.5
0.39
2.182.3
1.02.7
1.98
RZ
G ~ et
f ~ ef
g ~ ek
t
f
kft
3
3
0
2.7
4.02.7
01.35
EN
1.94
2.6
0.75
1.93.5
3.9
0.751.5
(a)
R (a)
Sol phase
ac
a
Gel phase
(b)
b=
R (p)
pc
b finite
Finite clusters
pc
Finite clusters
and infinite clusters
p
Figure 2.16. The analogy between (a) the gelation threshold and (b) the percolation-theory
threshold. Reprinted from Figure 17 (Stanley, 1985). Reprinted with permission of
John Wiley and Sons, Inc. Copyright (1985).
190
Macosko and Miller (1976) and Scranton and Peppas (1990) also developed a recursive
statistical theory of network formation whereby polymer structures evolve through the
probability of bond formation between monomer units; this theory includes substitution
effects of adjacent monomer groups. These statistical models have been used successfully in
step-growth polymerizations of amine-cured epoxies (Dusek, 1986a) and urethanes (Dusek
et al., 1990). This method enables calculation of the molar mass and mechanical properties,
but appears to predict heterogeneous and chain-growth polymerization poorly.
Kinetic models determine the state of cure by predicting the concentration of reacting
species from the solution of differential equations for each reacting species. Mikos et al.
(1986) and Tobita and Hamielec (1989) have developed kinetic models for vinyl and freeradical network systems. Chain-growth polymerization has also been modelled through
kinetic simulations by Okay (1994).
Lattice (Stauffer et al., 1981) and off-lattice (Leung and Eichinger, 1984) percolation
models are typical examples of spatially dependent network models and are usually used for
fast, diffusion-controlled reactions. Anseth and Bowman (1994), Manneville and de Seze
(1981), Boots and Pandey (1985) and Simon et al. (1989) use another spatially dependent
network theory called kinetic gelation modelling, which involves simulation of the network
structure in space using percolation simulations. Application of a kinetic gelation model to
free-radical polymerization of divinyl monomers (Kloosterboer, 1988) indicates that the
structure is well modelled when it begins with regions of inhomogeneity, but becomes more
homogeneous at higher conversions.
Overall it is considered that the spatially independent network models, whilst simpler,
should strictly be used for lightly crosslinked, homogeneous networks, whereas the spatially
dependent models, although computationally intensive and limited by pre-dened lattice
structure, provide a better understanding of network heterogeneities in highly crosslinked
systems.
191
Pgel 2/favg
Flory
Pgel 1/( f 1)
Flory
2
1/[r( fa 1)( fb 1)]
Pgel
Flory
Chain
Flory
Pgel 1/(Dpwav 1) f
P
Pgel is the extent of reaction at gelation: favg i(ni fi)/ni
is the average functionality, where ni is the number of
molecules of monomer I with functionality fi. Useful for
systems with equivalent numbers of reactive groups
(Carothers, 1936).
Neglects unequal reactivities of functional groups,
polydisperse oligomers (Flory, 1941).
fa and fb are the weight-average functionalities of
molecules with reactive groups of types a and b
(Macosko and Miller, 1976).
DPw0 is the initial weight-average functionality of
starting polymer chains; DPn0 is the initial numberaverage functionality of starting polymer chains
(Miller et al., 1979).
Dpwav is the weight-average degree of polymerization, f is
the fraction of double bonds in the monomer; use only
when double bonds have equal reactivity and the
copolymer is random (Flory, 1953).
acid esterglycidyl ether system. Gupta and Macosko (1990) describes a combined model
incorporating non-random processes (in this case, initiated living chain-type polymerization)
and a random branching model (that recognizes that some processes are statistically random
(a naive statistical model)). Results are compared with those derived from existing naive
statistical models (Riccardi and Williams, 1986, Bidstrup et al., 1986, Tsou and Peppas, 1986)
and an approximate combined model (Dusek, 1986a), and show the importance of including
the non-random processes when considering the polymerization of an epoxy-resin system.
Recent work (Altmann, 2002) has focussed on combining a dynamic Monte Carlo
percolation-grid simulation for reaction kinetics and an enthalpy-based group-interaction
viscoelastic model to develop a model for the chemorheological and network properties of
reactive systems. More emphasis will be placed on this model in Chapter 6.
2.5.2
192
References
Adabbo, H. & Williams, R. (1982) J. Appl. Polym. Sci., 29, 13271334.
Altmann, N. (2002) A Model for the Chemorheological Behaviour of Thermoset Polymers,
Brisbane: University of Queensland.
Anseth, K. & Bowman, C. N. (1994) Chem. Eng. Sci., 49, 2207.
Barral, L., Can, J., Lopez, A. J., Nogueira, P., & Ramirez, C. (1995) Polym. Int., 38, 353356.
Barton, J., Deazle, A., Hamerton, I., Howlin, B. & Jones, J. (1997) Polymer, 38, 43054310.
Bauer, J. & Bauer, M. (1990) J. Macromol. Sci. Chem. A, 27, 97116.
Berry, G. & Fox, T. (1968) Adv. Polym. Sci., 5, 261.
Bidstrup, W., Sheppard, N. & Senturia, S. (1986) Polym. Eng. Sci. 26, 359361.
Bonnet, A., Pascault, J., Sautereau, H. & Camberlin, Y. (1999a) Macromolecules,
32, 85248530.
Bonnet, A., Pascault, J., Sautereau, H., Taha, M. & Camberlin, Y. (1999b) Macromolecules,
32, 85178523.
Boots, H. & Pandey, R. (1985) Polym. Bull., 11, 219223.
Brown, M., Coates, P. & Johnson, A. (1994) Reactive processing of polymers, in RAPRA (Ed.)
RAPRA Review Reports, Report 73. vol. 7, no. 1.
Cadenato, A., Salla, J., Ramis, X., et al. (1997) J. Thermal Anal. 49, 269279.
Carothers, W. (1936) Trans. Faraday Soc., 32, 39.
Clarke, N., McGleish, T. & Jenkins, S. (1995) Macromolecules, 28, 4650.
Couchman, P. (1978) Macromolecules, 11, 117119.
De Gennes, P. (1971) J. Chem. Phys., 55, 572.
Degennes, P. (1979) Scaling Concepts in Polymer Physics, New York: Cornell University Press.
Di Benedetto, A. (1987) J. Polym. Sci. B: Polym. Phys., 25, 19491969.
Di Marzio, E. (1964) J. Res. Nat. Bureal Standards A, 68, 611617.
References
193
Doi, M. & Edwards, S. (1978a) J. Chem. Soc. Faraday Trans. II, 74, 1789.
Doi, M. & Edwards, S. (1978b) J. Chem. Soc. Faraday Trans. II, 74, 1802.
Doi, M. & Edwards, S. (1978c) J. Chem. Soc. Faraday Trans. II, 74, 1818.
Dusek, K. (1986a) Adv. Polym. Sci., 78, 1.
Dusek, K. & Demjanenko, M. (1986b) Radiat. Phys. Chem., 28, 479486.
Dusek, K., Spirikova, M. & Havlicek, I. (1990) Macromolecules, 23, 1774.
Einstein, A. (1906) Ann. Phys., 19, 289.
Enns, E. & Gillham, J. (1976) J. Appl. Polym. Sci., 28, 2562.
Enns, J. & Gillham, J. (1983) J. Appl. Polym. Sci., 28, 25672591.
Flory, P. (1941) J. Amer. Chem. Soc., 63, 3083.
Flory, P. (1953) Principles of Polymer Chemistry, Ithaca, NY: Cornell University Press.
Flory, P. & Rehner, J. (1943) J. Chem. Phys., 11, 512.
Fox, T. & Loshaek, S. (1955) J. Polym. Sci., 15, 371390.
Gan, S., Seferis, J. & Prime, R. (1991) J. Thermal Anal., 37, 463478.
Gedde, E. (1995) Polymer Physics, London: Chapman and Hall.
Gillham, J. (1986) Polym. Eng. Sci., 26, 14291434.
Girard-Reydet, E., Sautereau, H., Pascualt, J. P. et al. (1998) Polymer, 39, 22692280.
Grassley, W. (1982) Adv. Polym. Sci., 47, 68.
Gupta, A., Macosko, C. W. (1990) J. Polym. Sci. Part B, Polym. Phys., 28, 25852606.
Hale, A., Macosko, C. & Bair, H. (1991) Macromolecules, 24, 26102621.
Huang, M. & Williams, J. (1994) Macromolecules, 27, 74237428.
Inoue, T. (1995) Prog. Polym. Sci., 20, 119.
James, H. & Guth, E. (1943) J. Chem. Phys. 11, 455.
Janssen, L. (1998) Polym. Eng. Sci., 38, 2010.
Keller, A. (1995) Faraday Disc., 101, 149.
Kim, B., Chiba, T. & Inoue, T. (1993) Polymer, 34, 28092815.
Kloosterboer, J. (1988) Adv. Polym. Sci., 84, 1.
Kreiger, I. (1972) Adv. Colloid Interfacial Sci., 3, 111.
Kreiger, I. & Dougherty, T. (1959) Trans. Soc. Rheology, 3, 137.
Kuhn, P. (1934) Kolloid Z., 68, 2.
Labana, S. (1985) in Mark, H. (Ed.) Encyclopedia of Polymer Science and Engineering, 2nd edn,
New York: Wiley.
Larson, R. (1999) The Structure and Properties of Complex Fluids, New York: Oxford University Press.
Leung, Y. & Eichinger, B. (1984) J. Chem. Phys., 80, 3877.
Lu, X. & Pizzi, A. (1998) Holz Roh- Werkst., 56, 339346.
Macosko, C. & Miller, D. (1976) Macromolecules, 9, 199206.
Manneville, P. & De Seze, L. (1981) Numerical methods in the study of critical phenomena, in
Della Dora, J., Demongeot, J. & Lacolle, B. (Eds.) Numerical Methods in the Study of Critical
Phenomena, Proceedings of a Colloquium, Carry-le-Rouet, France, June 24, 1980, Berlin:
Springer-Verlag.
Mijovic, J., Andjelic, S., Fitz, B. (1996) J. Polym. Sci., Part B: Polym. Phys., 34, 379388.
Mikos, A., Takoudis, C. & Peppas, N. (1986) Macromolecules, 19, 2174.
Miller, D., Valles, E. & Macosko, C. (1979) Polym. Eng. Sci., 19, 272.
Moad, G. (1999) Prog. Polym. Sci., 24, 81142.
Muthukumar, M. (1989) Macromolecules, 22, 46564658.
Muthukumar, M. & Winter, H. (1986) Macromolecules, 19, 12841285.
Okay, O. (1994) Polymer, 35, 1994.
Osinski, B. (1993) Polymer, 34, 752758.
Palmese, G. & Gillham, J. (1987) J. Appl. Polym. Sci., 34, 19251939.
194
3.1
196
reactions competing, controlled initially by the concentration and kinetic parameters of the
species and then by the physical state of the system as the viscosity increases rapidly on
passage to the gel and glass. In the practical application of these reactions, the polymerization may be occurring in a mould (compression moulding; reaction injection moulding
(RIM)), or in an autoclave (bre-reinforced-composite curing from prepreg), so the possibility of measuring the chemorheological changes by conventional techniques is extremely
limited. To this end, novel methods employing combined techniques with bre-optic probes
have been developed and are discussed in a later chapter concerned with processing and
quality control (Chapter 6). In this chapter the fundamental analytical techniques and how
they may be used under laboratory conditions to generate the instantaneous chemical
structure of the polymerizing system are described. This is then extended to the dual
physicochemical analysis methods developed for probing the link between the physics and
the chemistry of the reacting systems.
3.2
3.2.1
3:1
197
R
sensor array
heater
heater
Figure 3.1. A schematic diagram of a differential scanning calorimeter (DSC) showing the sample (S)
In a power-compensated DSC (Figure 3.1) the sample and reference calorimeters are
provided with power necessary to maintain them at the same temperature while the polymer
passes through a transition or undergoes a chemical reaction. The difference in power, DP,
supplied to the sample and reference calorimeters is directly related to the heat-capacity
difference at that temperature. As the temperature is scanned, DP will be either positive or
negative depending on whether there is an endotherm, or exotherm respectively. Then the
direct conversion to heat ow from power means that
DP dQs dt dQr =dt dDQ=dt dH=dt Cp :
3:2
There are several different types of instrument covered under the term DSC, which have
evolved from differential thermal analysis (DTA) and measure the temperature difference
between sample and reference pans located in the same furnace. This is then converted to
heat ow using a calibration factor. A detailed analysis of DSC requires consideration of the
various sources of heat loss, and these are generally captured in the calibration routine for
the instrument. Absolute temperature calibration is achieved through the use of pure indium
(156.6 C) and tin (231.9 C) melting-point standards. A comprehensive analysis of the
theory of DSC contrasted with DTA may be found in several reference works (Richardson,
1989, Gallagher, 1997).
A more recent development is temperature-modulated DSC (Jones et al., 1997), which is a
particular example of the broader area of modulated calorimetry (Gmelin, 1997) and enables
deconvolution of kinetic and thermodynamic processes during the reactive curing of polymer
networks (Van Assche et al., 1997). This is discussed in more detail later (Section 3.2.3).
3.2.2
198
Tc
EXO
O2 admitted
Tg
Cp
Oxidation
Tm
Melt
ENDO
Temperature
Figure 3.2. Typical output from a DSC when a semi-crystalline polymer is heated through the glass
(Tg), crystallization (Tc) and melt (Tm) transitions. The exotherm on oxidation is also shown.
or oxygen, in which the time to the onset of the exotherm is measured. Thus the same
information as from oxygen uptake to measure OIT (Figure 1.41) may be obtained by
measuring the time after oxygen is admitted to the onset of the exotherm. The limitation
of OIT for applications at lower than processing temperatures has been emphasized
(Billingham et al., 1981), although for many practical purposes it must also be recognized
that a DSC oven with 10 mg of material cannot totally simulate the environment of a
reactive extruder in which the shear is responsible for mechanoradical formation as
described below (Section 3.3.2). In one example, the contribution to free-radical formation
by thermal oxidation was less than that by shear-induced chain scission (Sohma, 1989a), but
this depends on the availability of oxygen in the reacting system. The technique is well
suited to the comparative ranking of processing stabilizers.
An example of monitoring of reaction rate by DSC is the cure reaction of an epoxy resin.
Figure 3.3 shows the DSC curve for the heating of an epoxy resin with its curing agent
through the temperature range corresponding to the chemical reactions leading to threedimensional network formation.
A broad exotherm is obtained, corresponding to a typical heat of reaction (shown here as
DH0) of an epoxy resin of 107 4 kJ/mol epoxide (Prime, 1997b). Also shown is the curve
obtained by heating the system as mixed in an oven at 60 C for increasing times and then
running the DSC from room temperature to 150 C. Several features may be noted.
(i). There is a decrease in the area under the curve with time of heating at 60 C.
(ii). After some reaction has occurred, there is a shift in the baseline indicating an
endotherm, which corresponds to the appearance of the Tg of the developing network
(shown as the expansion of the early section of the cure curve in Figure 3.3).
(iii). The value of Tg, as given by the midpoint of the change in baseline (12 DCp), increases
with time of cure.
(iv). Isothermal cure of the resin for even long times does not result in complete removal of
the residual exotherm on scanning.
These DSC results may be further analysed to give the extent of reaction as a function of
cure time by using the assumption that the total area under the exotherm represents
the integrated heat of reaction, DHr, for the reaction of epoxide with amine (typically 107 4
kJ/mol epoxide as indicated above). Thus the reduction in this area on reacting the system for a
199
H0
dH/dt
Cure time /min:
H20
0
20
H40
40
100
H100
Temperature
Tg(0)
dH/dt
Tg(20)
Tg(40)
Temperature
Figure 3.3. Some DSC traces from the exothermic reaction of an epoxy resin after various times of
cure, showing the progressive reduction in reaction enthalpy, DH. The expanded region shows the
change to the baseline due to the enthalpy change at the glass transition. Note the progressive
increase in Tg after partial cure.
time t is a measure of the concentration of the epoxide consumed in the curing reaction of the
resin up to this time,
p DHr Ht =DHr ;
3:3
3:4
Thus, at the cure temperature of 60 C, the reaction rate may be obtained either from the
heat-ow rate in the DSC trace for the previously unheated resin in Figure 3.3 or from the
slope of the extent-of-reaction curve (Figure 3.4) at zero time.
As noted above, several signicant features of the network-forming system, which are
important for reactive processing, may be learned from this simple analysis. Firstly it may
200
1.0
p 0.5
0
0
50
Cure time/min
100
Figure 3.4. The time development of the extent of reaction, p, from the reaction-enthalpy results of
Figure 3.3, as analysed by use of Equation (3.3). Note that the system does not reach full cure
(p 1) because Tg of the resin reaches the cure temperature and reaction ceases.
be seen that, as the Tg of the resin approaches the cure temperature, the reaction rate
becomes vanishingly small, so the nal processed resin will have residual reactivity if it is
ever heated to above the processing temperature. This is the process of post-curing which
results in an increase of Tg and consumption of all reactive species. If this temperature is
chosen too high, the network will start to degrade, and this is accompanied by a further
exotherm if air is present since oxidation reactions may then dominate.
As an alternative to curing the resin for pre-set times and then measuring the heat ow on
ramping the temperature, the system could be cured (in this case at 60 C) in the DSC in the
isothermal mode. The disadvantage of this approach for moderate reaction temperatures that
are well below the maximum in the heat-ow curve (which for this resin occurs at 120 C) is
the low heat ow and thus difculty in achieving a good signal-to-noise ratio (SNR) in the
analysis. If the isothermal experiment is run at much higher temperatures, there is the
problem of achieving equilibration in the calorimeter before a signicant part of the reaction
has taken place. This then introduces uncertainty into the data for kinetic analysis.
Figure 3.5 shows an isothermal DSC curve for the curing of TGDDM with the aromatic
amine DDS (de Bakker et al., 1993b). Also shown is the NIR analysis of the instantaneous
rate of reaction of epoxide. The difference arises due to the change in heat capacity on
gelation and vitrication that affects the DSC baseline during cure, as discussed below.
The isothermal DSC curve is a plot of dH/dt against time so that, from Equation (3.4), it
may be integrated to determine the extent of reaction and so analyse the kinetics of the
system. There is still the necessity to perform a scanning DSC experiment in order to
determine the residual exotherm from the sample due to the cessation of reaction at the cure
temperature as the Tg of the resin reaches the isothermal cure temperature. This just
becomes a correction in the form of a scaling factor for the entire curve. A further
requirement in this analysis of the isothermal cure exotherm is the determination of the form
of the baseline for the integration. In many cases the simple assumption is to choose a at
baseline. However, this ignores the fact that the heat capacity, Cp, of the system will
201
Figure 3.5. Comparison of heat ow as measured by the DSC and calculated heat from the extent of
epoxide reaction (measured by NIR Spectroscopy) for TGDDM with 27% DDS at 177 C for the
times indicated. The difference arises from the change in heat capacity change during cure which
requires a baseline correction to the DSC trace. Adapted from de Bakker et al. (1993b).
decrease as the curing epoxy-resin passes from a liquid through a rubber to a glass. Prior to
the advent of modulated DSC, the baseline could only be estimated, and proles were
suggested (Barton, 1985, Gallagher, 1997). An example of the effect of the change in heat
capacity on the cure exotherm may be seen by considering simultaneous FT-IR analysis in
the NIR of the concentration of epoxide together with the heat-ow data obtained by DSC
(de Bakker et al., 1993b). This is discussed in more detail in Section 3.3.6. As shown in
Figure 3.5, there is a signicant difference between the reaction rate obtained by direct
analysis of epoxide and that derived from the DSC heat-ow curve which is attributable to
the change in heat capacity on cure. While this simultaneous method gave a clear indication
of how the baseline changed during cure, it could not unequivocally show that this change
arose solely from the change in Cp. This may be demonstrated more readily by modulatedtemperature DSC as discussed in Section 3.2.3.
There is another particular application of isothermal DSC that requires modication of a
traditional DSC, and this is in rapid photopolymerization (Scott et al., 2002) such as of
multifunctional acrylates (Peinado et al., 2002). In a typical photo-DSC experiment, the
202
sample and reference pans are operated in an open mode and the system is tted with an
ultraviolet illumination system that irradiates both, to ensure that heat ow results solely
from the polymerization exotherm. The response time of the system may be improved and
quantitative characterization achieved by replacing the DSC with a calorimeter that provides a dynamic range better by an order of magnitude than that of photo-DSC (Roper,
2005). The response time is reduced to 1 s for 90% of maximum heat ow, compared with
4.8 s for photo-DSC. This has enabled the curing of thin lms of thickness 163 nm to be
measured (Roper et al., 2005), and the results obtained are comparable to those from realtime FT-IR monitoring as discussed in Section 3.3.5.
In the study of the chemorheology of photopolymerizable systems, the use of a traditional
rheometer is particularly difcult and the development of improved time resolution from
several seconds to better than 1 ms has been noted, together with an increase in the intensity
of initiating radiation that may be employed (Schmidt et al., 2005). The challenge is to
gather data at the best possible SNR and then use the appropriate algorithm for recovery and
analysis of the phase information and intensity to enable calculation of G* and tan d
(Schmidt et al., 2005, Chiou and Khan, 1997). These studies are often augmented by
simultaneous real-time spectroscopic studies on the sample in the rheometer in order to
provide conversion data as discussed in Section 3.3.
3.2.3
3:6
203
0.04
2.4
0.03
2.1
t
0.02
Non-reversing
Heat Flow
(W/g)
Cp
1.8
exo
0.01
1.5
0
0
100
200
300
400
500
1.2
600
Time (min)
Figure 3.6. Non-reversing heat ow and change in heat capacity (Cp) during cure of an epoxy resin
cured with an anhydride at 80 C. The cure time needed to reach a halving of Cp is marked as t.
Adapted from Van Assche et al. (1997).
component (the reaction exotherm) to be separated from the reversing component (the
change in Cp with time). Therefore, provided that the cure reaction is not too rapid and there
are many modulations over the total cure time, Cp is readily obtained as a function of
temperature. Figure 3.6 shows a separation of the non-reversing heat ow from the heat
capacity for the cure of an epoxy resin with an anhydride at 80 C (Van Assche et al., 1997).
The time for half-decrease in Cp is shown and, in a separate experiment, the resin at this
point was found to have a Tg of 80 C, conrming that the change in Cp arises predominantly due to vitrication of the resin during cure. If cure occurs at a temperature above Tg1
then the change in Cp is small. Figure 3.7 shows the change in Cp as a function of the time of
cure of an epoxy resin with an aromatic amine at three different temperatures in a MDSC
(Lange et al., 2000).
The three curves correspond to cure below, near and above Tg1, and it may be seen that
only in the rst case is there a clear step-change in heat capacity. The processing of the resin
above Tg1 (150 C) results in complete chemical reaction since there is no vitrication
during cure.
3.2.4
204
80 oC
1.5
Cp
1.3
1.1
0
4000
12000
8000
Cure time/s
140 oC
2.2
Cp
2.0
1.8
0
2000
4000
Cure time/s
6000
8000
170 oC
1.6
Cp
1.4
1.2
0
2000
4000
6000
Cure time/s
Figure 3.7 The change in heat capacity, Cp, during isothermal cure of an epoxy resin with an
aromatic amine at temperatures below, near and above Tg. Adapted from Lange et al. (2000).
vulcanization of rubber (Brazier, 1981) the factors affecting DHr were determined for
various formulations of crosslinker (mostly sulfur) and accelerators (e.g. thiazoles) and
llers (mostly carbon black). Degradation reactions may also be studied so that overcure
may be avoided in processing. While DSC is a favoured techniques for providing cure data
on a wide range of elastomers (through comparison of DHr values), it has been noted
(Sircar, 1997) that the success of the enthalpy method depends on satisfaction of the
following criteria.
1.. The heat of vulcanization must be large enough to reduce error in any measurement of
residual heat of vulcanization.
205
2.. The residual reactants must not decompose after press cure.
3.. The evolution of heat must be proportional to crosslinking.
Since the instantaneous heat ow is (with some assumptions) representative of the rate of
reaction, there have been many attempts to extract kinetic and mechanistic information from
scanning DSC studies. These have severe limitations (as discussed later) but at this stage it
is useful to review the methods of analysis. The ultimate test of a scanning method is
whether it is able to predict the same phenomena as observed in a comparable isothermal
experiment.
The starting point for the use of scanning DSC for chemorheology of systems such as
cure of an epoxy resin is the kinetic model which links the variables of temperature and
reactant concentration to the rate of reaction. This is chosen to be a simple Arrhenius
relation governed by the pre-exponential factor, A, and the activation energy, Ea,
dp=dt A expEa =RT f p;
3:8
where the form of f (p) is chosen according to the chemistry of the system and the appropriate
mechanistic scheme or experimental knowledge of the rate law. The most usual starting
point is nth-order chemical kinetics, f (p) (1 p)n, so that in linear form Equation (3.8)
becomes
lndp=dt=1 pn ln A Ea =RT :
3:9
This is the basis of the Borchardt and Daniels method, which was originally developed
for DTA (Borchardt and Daniels, 1957) and is now offered in almost all available commercial DSC software. The DSC heat ow, dH/dt, is treated as equivalent to the reaction
rate dp/dt (as in an isothermal experiment) and plots of Equation (3.9) are made for various
values of n, the so-called reaction order. The value yielding the best straight line is taken as
the order of the reaction and the values of Ea and ln A are then calculated from the slope
and intercept of the plot (Barrett, 1967). The attractiveness of the method is that it yields
apparent kinetic data from a single DSC trace, but it has been shown for many polymerizing systems that they have no kinetic meaning, and results that are at variance with those
obtained by isothermal methods have been derived for many epoxy-resin and other
reactive-network systems (Prime, 1997b). Of some concern is that several of the commercial packages offer the calculation from a single DSC scan of isothermal parameters
such as gel time and the time to peak exotherm, both of which are valuable for reactive
processing. It has been shown, and will be discussed later, that there are some limitations to
this method (George and St John, 1993), but the most obvious is the over-simplication of
the kinetics and mechanism of an often auto-accelerating reaction. In general it is
inappropriate to apply the Borchardt and Daniels and the related Freeman and Carroll
(Freeman and Carroll, 1958) methods to complex reacting systems since they assume
simple nth-order kinetics.
In an extension of scanning techniques, heating methods that involve no assumptions
regarding the order of reaction have been developed. The most widely used (and available
also as standard software in commercial DSC systems) are the Kissinger and Ozawa
methods of analysis. Both of these are based on the time taken to reach the same extent of
conversion at various temperature-scanning rates and have an underlying assumption that
the activation energy is independent of temperature. Since it has been found for many
reacting systems that the extent of conversion, pm, at the maximum of the cure exotherm,
206
Tm, is constant, this is used as the isoconversion point for the Kissinger analysis (Kissinger,
1957) of the DSC exotherms at various scan rates, b,
dln b Tm2 d1=Tm Ea =R:
3:10
The Ozawa method involves a renement of this equation, namely
dln b=d1=Tm 1:052Ea =R;
3:11
and is based on an empirical solution to the integral of expEa =RT from the onset
temperature (T0) to the maximum (Tm). This has been shown to give apparent activation
energies with an accuracy better than 10% when compared with isothermal methods for
the curing of DGEBA epoxy resin with a polyamide curing agent (Prime, 1997b). This has
been adapted to ASTM method E69879, although there are clear recommendations that
any use of the data should be supported by an isothermal experiment (such as half-life
measurement) to conrm the predicted behaviour. There are several concerns with these
methods for complex reactions, especially when the DSC reaction proles are compared
with independent chemical information such as from NIR spectroscopy (Section 3.3.6).
From these studies it has been seen that there is a signicant scan-rate dependence of the
coincidence of the exotherm maximum and the true reaction-rate maximum (George and St
John, 1993). This heat-lag effect is most pronounced at and beyond Tm, where it distorts the
DSC prole and, since the effect is non-linear with heating rate, will affect both the
Kissinger method and the Ozawa method of analysis.
It has been noted (Richardson, 1989) that the reaction path may often change with
temperature, and the parameters derived from the scanned DSC curve analysed by using
either Equation (3.9) or Equation (3.10) represent some nonexistent average reaction. So,
even though the values of Ea and A will have no fundamental meaning, they may allow
reasonable estimates to be made for quality-assurance purposes or for some baseline data for
empirical modelling of the processing of resins as in the autoclave cure of composite
materials (Ciriscioli and Springer, 1990). Of particular importance in the industrial process
is the reaction which occurs during temperature proles, such as ramping, and it is
important to be able to transform the generated DSC data to a form that may be used for
process control.
3.2.5
207
chemical kinetics, the Arrhenius relation is applicable and the shift factor is the empirical
activation energy, Ea. Of interest in prepreg cure in an autoclave that may have a large
thermal lag is the equivalent isothermal time, s, which may be attributed to a sample that
has experienced a complex thermal prole. An example of importance is the time taken to
reach gelation, tgel, for the epoxy-resin prepreg, since this will decide the time at which to
apply pressure in order to consolidate the composite. The calculation involves breaking up
the prole into a series of isothermal steps of duration ti at temperatures, Ti, with the
accumulation up to the nal temperature, Tr. Then it is possible to compute the accumulated
cure time, s, as
X
X
s
si
ti expEa Ti Tr =RTi Tr :
3:12
i
In this the way the effect of the rate of temperature ramping may be taken into account in
the total cure prole. In such studies the empirical Ea data may be adequate, but for detailed
kinetic modelling it is generally recognized that isothermal experiments are appropriate, and
even they may have limitations, as discussed later.
3.2.6
3:13
where k1 and k2 are rate constants representing the uncatalysed and catalysed reactions,
respectively. This has been generalized to the empirical equation (Barton, 1985)
dp=dt k1 k2 pm 1 pn ;
3:14
where m and n are adjustable parameters representative of the order of the individual terms.
Barton (1985) obtained a t using m n 1 over the rst 30% of conversion, implying that
the dominant reaction mechanism is bimolecular. However, the small extent of conversion
over which the equation applied limits the applicability in process control. As an illustration
of the very empirical approach often used, the cure of the tetraglycidyl amine resin system,
TGDDM, with 4,40 -diaminodiphenylsulfone (DDS) was shown (Mijovic and Wang, 1986)
to t Equation (3.14), but the values of m and n obtained were temperature-dependent.
208
Thus, while there is a t, the equation can have no physical meaning regarding the kinetics
of the system. This approach has been modied by introducing a separate exponent, l, for
the uncatalysed reaction:
dp=dt k1 1 pl k2 pm 1 pn :
3:15
For TGDDM/DDS the values l 2 and m n 2 enabled a good t for DSC data for up
to 70% conversion, after which diffusion control of the reaction rate became dominant.
The difculty with this type of approach is that for each different resin it is necessary
to determine not only new pseudo-rate constants but also new values for the exponents (Lee
et al., 1992, Mijovic and Wang, 1986). This supports the view that the exponents are more
akin to curve-tting parameters rather than being a true reection of the underlying reaction
mechanisms involved.
A different approach to using isothermal DSC data has been to derive rate expressions
that encompass all the expected mechanistic features of an epoxyamine cure, including a
substitution effect on amine reactivity and etherication reactions (Chiao, 1990, Cole et al.,
1991) as discussed in Section 1.2.1 (Schemes 1.3 and 1.23). However, as more parameters
are introduced, the possibility of achieving a t increases, but often with a reduction in the
mechanistic meaning. Also, as with other DSC studies, the assumption of equal reaction
enthalpies for all epoxy reactions had to be made resulting in an underlying uncertainty in
the conclusions. Nonetheless, the methodology described represented a major advance in
the use of isothermal DSC in the study of cure kinetics, and it is only through the direct
measurement of the absolute concentration of reacting species, such as the use of spectroscopic techniques, particularly NIR, that further advances in the understanding of
reactive systems such as epoxy-resin networks has been made. These are described in the
following sections.
3.3
3.3.1
209
3.3.2
3:16
where H is the applied magnetic eld, b is the Bohr magneton and g is a constant, the
spectroscopic splitting factor, which for a free electron has the value 2.0023. Since
the population difference between the two states is small at normal temperatures (being
governed by the Boltzmann distribution) spectra are collected by sweeping the applied eld
and noting the resonance condition. Structural information regarding the free radical is
obtained by observing the splitting of electron-spin energy states due to hyperne interaction with the nuclear spin of the protons or nitrogen atoms in the radical. (Since the
backbone carbon atoms have no nuclear spin, the interactions are greatly simplied in ESR.)
The g factor will shift from that for the free electron. Thus an alkyl macro-radical from the
scission of a polymer with a methylene bridging unit will give a three-line spectrum with a
relative intensity of 1 : 2 : 1 (Sohma, 1989b). In the presence of oxygen, the spectrum of
210
polyolens that have been oxidized collapses to a broad singlet characteristic of the peroxy
radical which is the main chain carrier in the auto-oxidation of the polymers.
The radicals must be trapped in order to measure the spectrum, and a steady-state
population of 1013 spins is required in the analytical volume, so it is impossible to observe
free radicals by ESR during reactive processing. In principle it could be possible to trap
the free radicals off-line by quenching the melt in liquid nitrogen without exposure to air,
but this presents experimental challenges. A simple device that incorporates quenching by
liquid nitrogen for near-real-time sampling of the polymer in the barrel of a reactive
extruder has been described (Machado et al., 1999). This could be applied to collecting and
maintaining a sample for ESR study of the free-radical species present in the extruder at any
stage of the process. Chemical trapping techniques offer greater opportunity for demonstrating the formation of free radicals due to the shear forces during extrusion (Sohma,
1989b). Spin traps such as nitroxyl radicals are frequently used in ESR spectroscopy to
capture transient species, and a variation on this in which a hindered amine (the commercial
HALS light stabilizer, Tinuvin 770, Figure 3.8 (Sohma, 1989a)) is included with PP during
processing in an extruder has been described.
Following extrusion at 230 C with a residence time of 30 s, the characteristic ESR
spectrum of a triplet due to nitrogen coupling was observed from the sample (Sohma,
1989a). The results in Figure 3.8 show that the trapped free-radical concentration increased
with the concentration of HALS and that the main contribution was the shear degradation of
the polymer, not thermal scission since negligible radical formation occurred at 230 C
alone with in the same time frame.
In other reactive systems in which the network rst gels and then vitries, free-radical
species and radical cations may be stabilized physically as well as chemically and then
analysed. An example is the vitrication of difunctional methacrylate monomers such as
ethylene glycol dimethacrylate (EGDMA). The free radicals formed during the polymerization and crosslinking process are trapped in the glassy network and may be analysed by
ESR. The difference in free-radical concentration between a system of methyl methacrylate
(MMA) containing EGDMA and a neat MMA polymerization may be seen in Figure 3.9.
It is noted (Carswell et al., 1996) that the concentration of radicals almost reaches
millimolar levels in the crosslinking system, i.e. 40 times that for the uncrosslinked polymer. Species other than those responsible for network formation may also be observed, if
they are stable, and then used to monitor crosslinking. The oxidation of amine-containing
epoxy resins may occur during cure or on UV irradiation, and ESR has been used to identify
the cation radical species formed (Figure 3.10) (St John, 1993, Fulton et al., 1998).
By irradiating the curing system and measuring the concentration of this species by ESR
it is possible to observe the considerable change in the network structure that occurs
between gelation and vitrication and leads to an increasing population of trapped radical
cations (Figure 3.10). In another use of ESR, the lifetime of the free radicals formed after
c-irradiation has been used to monitor the state of the network formed on cure. The
occurrence of two distinct decay processes suggested that inhomogeneities in the network
were being probed (Kent et al., 1983).
The major limitations to the use of ESR other than for fundamental studies of the radical
and other trapped species formed during reactive processing are the experimental requirements of the apparatus. There has been success in using ESR to monitor the concentration
of the propagating free radicals during the emulsion batch polymerization of methyl
methacrylate (Parker et al., 1996) by using a time-sweep method for data acquisition and
211
Me
Me
Me
HN
(CH2)8
Me
NH
Me
Me
Me
HALS: Tinuvin 770
230oC; 30 s in Extruder
2
[HALS]
(Relative)
[HALS] Wt%
Figure 3.8. Radical formation due to mechanochemical degradation measured by using Tinuvin 770
as a spin trap and detecting the alkyl radicals via the change in nitrogen-centred radical
concentration by ESR. Adapted from Sohma (1989a).
400
10
MMA
8
300
MMA+ EGDMA
106[R]
mol dm-3
6
106[R]
mol dm-3
200
MMA+ EGDMA
4
MMA
100
2
0
0
200
100
0
300
Time / min.
Figure 3.9. Changes in free-radical population [R] during polymerization of MMA and after
addition of the crosslinker EGDMA (note the different scales). Adapted from Carswell et al. (1996).
212
N
R
TGDDM resin
Radical cation
x10-14/g
+ e-
[SPINS]
VITRIFICATION
GELATION
2
0
0
50
100
150
with 26.5% DDS at 160 C. Brief UV irradiation is used to create the radical cation species
analysed. Reproduced with permission from St John (1993).
213
It has been demonstrated that on-line monitoring in a laboratory environment of a coppermediated living radical polymerization provides increased precision in kinetic analysis
because of the greater number of data points obtained (Perrier and Haddleton, 2003). One
major limitation in both ESR and NMR spectroscopy is the difculty in applying the
technique to either on-line or in-line analysis in real time in an industrial environment,
although it can be expected that further advances in technology may lead to this possibility.
An example of this is the high-temperature surface NMR probe that has been able to be
adapted to operation at the die of an extruder (Gottwald and Scheler, 2005). While it was
demonstrated that the system could operate in the aggressive environment, no interpretation
of the relaxation time of PVC during extrusion was performed. The NMR technique is one
of the most powerful off-line analytical tools since the signal is sensitive to the chemical,
conformational and macroscopic environment and it has been used successfully for in situ
study of ZieglerNatta polymerization of propylene (Mori et al., 1999), but much further
development is needed in order to provide an on-line monitoring technique. Of some
interest is the demonstration of rheo-NMR, in which magnetic resonance imaging is used to
visualize polymer melts under shearing and extensional ows (Callaghan, 1999).
Since many products of reactive processing are solids, the advent of solid-state (magicangle-spinning (MAS)) NMR spectroscopy has meant a greater understanding of networks
after gelation and vitrication. An excellent example is with elastomers, and Koenig (1999)
gives a detailed analysis of the information which solid-state NMR has provided regarding
the mechanism of vulcanization, the effect of ller (e.g. silica and carbon black) on cure
chemistry and also the thermal oxidation of the network. In relation to oxidation, Clough
(1999) has demonstrated the extensive information which may be generated by using 17O2
for oxidation studies and measuring the NMR spectrum of the resulting oxidation products.
While this demonstrates the power of NMR, and there are further developments in magnetic
resonance imaging that are having a major impact on polymer science, the direct application
to chemorheology is not apparent except in assisting other techniques through establishing
mechanisms, identifying the effects of impurities on the reaction route etc.
An example of this relevant to chemorheology of advanced epoxy-resin networks is the
detailed studies of model compounds to identify the possible structural units from the
reaction of N, N-diglycidylanilino groups with aromatic amines as models for the TGDDM/
DDS epoxy-resin system (Laupretre, 1990) and then the use of 13C CP/MAS NMR to
characterize fully cured TGDDM/DDS networks (Attias et al., 1990). From this it was
concluded that the network structure of TGDDM/DDS had a considerably lower crosslink
density than expected from the high functionality of the resin and hardener. Other
researchers (Grenier-Loustalot et al., 1990) found that the aliphatic-carbon peak could be
resolved into primary- and secondary-alcohol, ether and epoxy bands, and that, for TGDDM
and MY721, about 20% of epoxy groups formed ethers when cured at up to 185 C. The
nal glass-transition temperature (Tg) decreased with increasing ether and hydroxyl group
concentration. As is discussed later, these results are of value in interpreting data from NIR
and other vibrational spectroscopic techniques that can be used for real-time reaction
monitoring.
3.3.3
214
typically from 400 to 4000 cm1. Infrared spectroscopy generally involves the measurement of the wavelength (or energy) and the amount of radiation absorbed by the polymer
from a beam of infrared radiation. The analysis of the spectrum therefore provides
information about the composition of the system as well as the amount present. The
vibrational spectra of small molecules such as water and carbon dioxide in the vapour
state appear as many peaks in the MIR region due to the excitation of rotational states of
the molecules, and these may be assigned to specic vibration and rotation modes of the
molecules. This gives a characteristic band envelope for water and carbon dioxide, which
is often seen in FT-IR transmission spectra because of the presence of water and carbon
dioxide in the optical pathlength when polymer samples are analysed (Koenig, 1999).
These rotations cannot be resolved in the condensed phase and only a broadened single
band may be seen.
The formal requirement for these transitions to occur is that the molecular vibration
should produce a change in dipole moment. Thus, in a molecule such as carbon dioxide, the
symmetrical stretching vibration (m1) will be infrared inactive and the bending (m2) and
asymmetric stretch (m3) modes will be active. The fundamental frequency at which a particular vibration will occur is given by the classical formula for a diatomic harmonic
oscillator:
v 1=2pk=l1=2
3:17
where m is the frequency of the fundamental vibration, k is the force constant for the bond
undergoing vibration and l is the reduced mass of the atoms in the molecule.
For a bending vibration, k will be lower than for stretching a bond, such that in CO2 this
frequency is lower (667 cm1) than that for a stretching vibration (2349 cm1).
The observed spectral bands correspond to allowed transitions between quantized energy
levels of energy:
1
Ev v hv;
3:18
2
where v is the vibrational quantum number, which may take values 0, 1, 2 . . .
The selection rule for allowed transitions is Dv 1, so the energy separation between
the levels is hm, which is the observed spectral frequency. The fundamental band observed in
the spectrum then corresponds to the transition from v 0 to v 1, since at normal temperatures most molecules will occupy the ground vibrational state.
These relationships are based on a pure harmonic oscillator, but real molecular vibrations
are often anharmonic. Consequently the solution for the quantized harmonic oscillator
(Equation (3.18)) is replaced by
1
1 2
Ev v hv v
hvv;
3:19
2
2
where is the anharmonicity constant, which is small and positive. The effect of anharmonicity is to relax the selection rule to Dv 1, 2, 3, . . . so allowing transitions
corresponding to the rst and second overtones as well as combination bands to be seen.
Thus, in the spectrum of carbon dioxide, in addition to the two allowed fundamental
vibrations, weaker bands corresponding to overtone bands (multiples of the fundamental)
and combination bands (sums or differences of two fundamentals) are seen in the MIR. Both
215
of these result from the anharmonicity of the vibrations of real molecules, which is the rule,
rather than the exception, when studying polymers.
The combinations and overtones of many high-energy vibrations are at energies beyond
the MIR spectral region and are therefore seen in the near infrared (NIR). The extent
to which this occurs depends on the anharmonicity of the particular vibration. The
most anharmonic vibrations are those involving light atoms attached to heavy atoms, such
as CH, NH and OH ( is large in Equation (3.19)), and it is found that these combinations and overtones are the most intense in the NIR spectra of organic molecules and
polymers. As will be discussed in Section 3.3.6, these are still weaker than the fundamental
modes by two orders of magnitude (Colthup et al., 1990), so the pathlength needed to study
the spectra at a reasonable absorbance value is of the order of millimetres rather than
micrometres as in the MIR. One advantage of the NIR region of the spectrum is that the
effects of atmospheric absorption bands are much less signicant than in the MIR and the
optical components may be of quartz or glass and are readily adapted to optical bres for
remote spectroscopy of reacting systems.
Classically, in a polyatomic molecule, a normal mode of vibration occurs when each
nucleus undergoes simple harmonic motion with the same frequency as and in phase with the
other nuclei. The possible normal modes of vibrational of a non-linear polyatomic molecule
with N atoms are 3N 6 and are categorized as arising from stretching, bending, torsional and
non-bonded interactions. Whether a particular mode is IR active depends on whether the
symmetry of the vibration corresponds to one of the components of the dipole moment.
Information regarding the normal modes of a polyatomic molecule, which are not IR
active, may often be obtained from the Raman spectrum. Raman spectroscopy is an
inelastic-scattering technique rather than requiring the absorption or emission of radiation of
a particular energy. The selection rule differs from the IR in that it is required that the
incident electric eld of the radiation can induce a changing dipole moment of the molecule.
This results in a different symmetry requirement for the normal modes of vibration to be
Raman active, since it now depends on the polarizability of the molecule.
The scattered radiation then contains information about the vibrational states of the
molecule encoded as a band shift Dm, either to the high or low energy of the elastically
scattered exciting laser radiation,
Dv v0
v;
3:20
where m0 is the frequency of the incident radiation and m is a vibrational frequency that
satises the Raman selection rule.
The exciting laser radiation may be in the UV, visible or NIR region of the spectrum, so
there is considerable exibility in choosing the appropriate radiation source to minimize
effects that mask the Raman spectrum, such as uorescence from the polymer or impurities.
However, since the Raman effect is inversely proportional to the fourth power of the
wavelength of the exciting radiation, the scattering with NIR laser excitation will be weaker,
although the interference from uorescence is decreased.
As shown in Equation (3.20), analysis of the Raman spectrum may be performed to the
lower (Stokes shift) or higher (anti-Stokes shift) energy of the exciting light, and the
vibrational modes that may be resolved depend on the efciency of removing the Rayleigh
scattering. As with NIR, the technique may be used with little atmospheric interference
and quartz or glass optical components. These and other experimental requirements are
discussed in Section 3.3.7.
216
Table 3.1. Components of an interferometer (Figure 3.11) for optimum FT-IR spectral performance
Region
Source
Beam splitter
Detector
Mid-IR
Near-IR
Glowbar
Tungstenhalogen
KBr
Quartz
Detector
Interferogram
Beam
Splitter
Moving Mirror
Source
l
Fixed Mirror
Figure 3.11. A schematic diagram of the operating principle of an FT-IR spectrometer, showing the
Michelson interferometer.
3.3.4
217
this is usually not a major factor, but in on-line analysis the SNR achieved may limit the
detection limit of species and the analytical precision. A factor that also limits the resolution
of the spectrometer is the nite distance the mirror travels in the interferometer. The FFT is
taken between 1 so, the result is the presence of a spectral artefact (spectral side lobes)
that is automatically removed by multiplication by an apodization function before the
spectrum is displayed. This correction produces spectral line broadening, but this may be
removed during Fourier self-deconvolution (with return of the side lobes) to assist in
determining the presence of overlapping bands. This is just one of many data-analysis
routines available to assist spectral interpretation, but there are limitations to quantitative
analysis, which are discussed below.
Transmission spectra
In the simplest spectral-acquisition method, the sample to be analysed lies between the
source and the interferometer. Since FT-IR is a single-beam spectral technique, the spectrum of the source is obtained in the absence of the sample and then obtained again with it in
the beam. The ratio of the two generates a transmittance spectrum, which is then converted
to an absorbance spectrum to enable spectral manipulation such as subtraction. An
advantage of an FT-IR system is that spectra are acquired digitally with high wavenumber
accuracy so that bi-component polymer lms may be analysed by spectral subtraction if one
of the components is known. An example is shown in Figure 3.12 for a sample containing
Figure 3.12. Transmission IR spectra plotted in absorbance of a two-polymer laminate (A), one
component of which is able to be measured (B) whereas the other is unknown. Identication is
possible by spectral subtraction to give spectrum (C), which may be compared with a library
spectrum (D). Spectrum courtesy of Dr L. Rintoul.
218
two polymers. The rst spectrum (A) is that of the bi-component system. The second
spectrum (B) is that of polypropylene which was obtained by attenuated total reection
(ATR) on one surface of the system (see the next section). The third spectrum (C) is the
result of subtracting the second from the rst and shows features that are characteristic of a
polyamide. This is conrmed by comparison with a spectrum of nylon-6 from the digital
FT-IR library as shown in the fourth spectrum (D) in Figure 3.12. It may be seen by
comparing this with the subtraction spectrum that there are minor differences due to artefacts introduced by the subtraction process, in which spectra may differ very slightly in
wavenumber. This may produce derivative-like spectra such as is seen around 2950 cm1.
There are also negative bands at 17001750 cm1 due to oxidation bands present in the
polypropylene that are over-subtracted out. Provided that these minor alterations are
recognized, this method does offer analytical capabilities that may otherwise be difcult to
obtain, such as when there is a laminate with one component surrounded top and bottom by
another polymer so that the external polymer may be measured by ATR (see below) and the
total laminate by transmission spectra.
Such analytical processes rely on the additivity of absorbance given in Beers law:
Av log10 I0 =I e1 c1 b1 e2 c2 b2 ei ci bi ;
3:21
where I0 and I are the intensities of radiation without and with the sample present,
respectively, A(m) is the total sample absorbance at a frequency corresponding to absorption
by all components and e, c and b are the molar absorptivity (dm3 mol1 cm1), concentration (mol/dm3) and pathlength (cm), respectively, of each component i at that frequency.
In many systems, such as homogeneous mixtures, b will be a constant for each component. In the MIR region of the spectrum, the molar absorptivities are typically
10100 dm3 mol1 cm1 so that, for solid polymers, a pathlength of about 10 lm is required.
For this reason, sampling methods involving reectance spectroscopy are employed for
thick samples (Section 3.3.4). Beers law will hold for systems in which there are no
interactions between the components, such that spectral subtraction should enable quantitative analysis of the components, if all but one are known and standard spectra are
available. These may be from one of the many libraries of polymer spectra available in
digital form with commercial instrumentation. Where species are unknown, then methods
such as least-squares regression analysis and principal-component analysis may be
employed to determine the number of species present in a mixture as well as the principal
components of each spectrum (Haaland et al., 1985, Koenig, 1984, 1999). The interactions
which may occur between components are important in blends, and small spectral differences are of importance in understanding the enthalpic contributions to miscibility and the
origin of the intermolecular forces controlling the structure (Garton, 1992). Specic
examples are considered in Section 3.3.5.
One of the features of transmission FT-IR spectra in the MIR is the appearance of
atmospheric bands as either positive or negative spectra. This may often limit the analysis
of oxidation products since the water-vapour bands lie in the same spectral region as those
of some conjugated and carbonyl species. While atmospheric effects may be subtracted out,
purging of the full optical path is the most reliable way of eliminating these interferences. It
is also important to remember that analyses are more reliable at low absolute absorbance.
The subtraction of two bands with high absorbance will heighten effects due to non-uniform
sample thickness (Koenig, 1984). The reverse effect occurs when samples are of an
extremely uniform thickness, so interference fringes are observed, which may swamp the
219
Reflection spectra
The transmission (T) of radiation by a sample is always lower due to reection (R) losses,
and, as shown in Figure 3.13(a), there is also a third component to be considered, namely
emission (E), as discussed in the next section. In general T R E 1.
Internal-reection spectroscopy (IRS), also known as attenuated total-reection (ATR)
spectroscopy, is one of the most widely used techniques to generate spectra from intensely
absorbing samples. The principle of operation is shown in Figure 3.13(b).
The radiation from the IR source is transferred through mirrors to enter normal to the
edge of a high-refractive-index optical element. This edge is cut at an angle to enable total
internal reection of the radiation through the element and then emergence through the
opposite edge. Other mirrors gather the radiation and direct it to the interferometer. If a
material of lower refractive index is in contact with the element along one or both sides,
then, as the radiation reects at or near the Brewster angle, there will be attenuation of the
radiation due to absorption of the evanescent wave by the sample. This wave penetrates a
distance, d, into the sample dependent on the refractive indices of the internal reection
element (IRE), n1, and the sample, n2, the angle of reection, h, and the incident wavelength, k, as given by
1=2 o
n
d k 2pn1 sin2 h n2 n21
:
3:22
The wavenumber dependence of the penetration depth for a KRS-5 optical element
(a mixed salt of thallium bromide and iodide) (n1 2.5) and a sample with n2 1.5 is shown
in Figure 3.13(c).
It may be seen that the total pathlength for a multi-reection element with, say, six
reections will range from 13 lm at 800 cm1 to 3 lm at 3000 cm1. Therefore ATR
spectra are dependent on the experimental arrangements and apparatus. Other materials for
ATR elements are ZnSe, Ge and diamond, with the choice being governed by the refractive
index compared with those of the samples being studied.
As discussed later, in Section 3.4, this technique is particularly well adapted to coupling
to optical bres, so a probe that can be immersed in a polymer reaction stream may be
fabricated to enable spectra to be collected in real time. The principles of total internal
reection and its application in bre-optics are discussed in more detail in Section 3.4.1.
Another widely used reection technique for polymer samples, particularly powders, is
diffuse reectance, which, when used with an FT-IR Spectrometer, is given the acronym
DRIFT. The principle of operation requires a renement of the scheme in Figure 3.13(a) to
ensure that no specularly reected radiation is collected. It has been shown that this may be
achieved by ensuring that all radiation collected has passed through the particles rather
than having been reected from the surface by covering the material being analysed with a
layer of nely ground potassium bromide. Under these circumstances the KubelkaMonk
220
(a)
E
Polymer
I0
I
T=I/I 0
R
T+R+E = 1
(b)
IRE
n1
Polymer n2
(c)
4.0
3.0
2.0
d /m
1.0
0
0
1000
2000
3000
4000
5000
Wavenumber/ cm 1
Figure 3.13 (a) A schematic diagram of light transmission (T), reection (R) and emission (E) by a
polymer lm; I0 and I are the incident and transmitted beam intensities. (b) A schematic diagram of
single reection with an internal reection element (IRE) of refractive index, n1, in contact with a
polymer lm, of refractive index n2. The penetration depth, d, of the evanescent wave depends on
the angle of incidence h and wavelength, k, of the radiation as given in (c). (c) A plot of Equation
(3.22), showing the wavenumber (1/k) dependence of penetration depth, d, of the evanescent wave
for a KRS-5 IRE of refractive index 2.5 in contact with a polymer lm of refractive index 1.5.
221
2.5
Graphite
Emissivity
2
Pt + sample
1.5
Platinum
Corrected Sample
0.5
0
2500
2000
1500
1000
500
Wavenumber (cm1)
Figure 3.14. Emission spectra required for the generation of an equivalent absorbance spectrum from
a heated sample (Equations (3.23) and (3.24)). Reproduced with permission from Blakey (2001).
correction (a routine built into FT-IR software for DRIFT) applies and quantitative analysis
is feasible (Koenig, 1999).
Emission spectra
The contribution of emission to the infrared radiation detected by the FT-IR detector may be
neglected under the normal conditions of measuring MIR spectra by any of the techniques
described above. In most experiments, the sample is at room temperature, so there will be a
negligible population of excited vibrational states created by processes other than absorption of the infrared radiation. There will be emission of radiation from those states populated
by absorption, but this amounts to a negligible contribution since it is generated in all
directions and only a small fraction will reach the interferometer, compared with the collimated beam from the light source.
However, if the sample is at an elevated temperature, above 150 C, as in many
processing operations, then signicant emission in the MIR occurs. If the light source in the
FT-IR system is replaced with a thin polymer lm on a platinum hotplate, and the resulting
single-beam spectrum referenced to a black body (graphite) at the same temperature, then
an emissivity spectrum qualitatively equivalent to an absorbance spectrum is obtained.
Figure 3.14 shows the sequence of measurements necessary to generate the emissivity
spectrum.
Since the spectrometer will itself be emitting infrared radiation to the detector (at 77 K),
determination of the emissivity requires the following measurements:
ev ISa IPt = Ibb ISp ;
3:23
where ISa, IPt, Ibb and ISp represent the intensities of infrared emission at each frequency, m,
from the sample, platinum sample holder, black body and spectrometer, respectively. The
nal corrected emissivity spectrum of the sample as a function of wavenumber is shown in
Figure 3.14 and may be scale-expanded for detailed analysis.
222
A simple analysis (George et al., 1995) gives the relation between absorbance A(m) and
emissivity e(m) as
1
Av ln1 ev:
2
3:24
This relation holds only for low values of emissivity, and, in order to ensure that there is
no loss of spectral denition due to reabsorption and to prevent the creation of thermal
gradients in the polymer, a thickness less than 10 lm is required. The technique, like all
emission methods, is extremely sensitive (Chase, 1981), and it has been applied to study the
cure chemistry of thermoset resins on metal substrates, and thus is appropriate as a technique for monitoring reactive processing such as coating deposition (Pell et al., 1991) and
epoxy-resin prepreg crosslinking (George et al., 1996). An example of the use of emissivity
to study the oxidation of polypropylene in real time is shown later in Figure 3.22.
Excitation spectra
Excitation spectra are generated when the wavelength of radiation absorbed by a sample is
varied and the intensity of emitted radiation is measured. The resulting spectrum of
emission intensity against excitation wavelength is analogous to the absorption spectrum of
the sample. While such measurements are usually associated with luminescence processes
(Section 3.3.8), the analogous technique in the MIR is photo-acoustic spectroscopy (PAS).
In this technique, the extent to which the sample is heated by the incident radiation is
measured by the pressure change to gas contained in an enclosed chamber. Since the FT-IR
radiation is modulated at the frequency of the moving mirror, this will appear in the audio
range of 12 kHz. The pressure sensor is a microphone that detects a photo-acoustic
equivalent of the interferogram, which therefore contains the absorption information for the
sample. An advantage of the PAS technique is that reected or scattered radiation does not
affect the modulated signal, so it is particularly effective with highly lled samples, such as
those with carbon black (Koenig, 1984). Depth proling to one-half the thermal diffusion
length from the surface may be performed by varying the modulation frequency of the
incident radiation by altering the mirror velocity. However, PAS spectra are generally of
low SNR, so a large number of scans needs to be co-added and thick samples are prone to
saturation effects
It has been reported that PAS can be used to monitor reactive processing of network
polymers (Taramae et al., 1982), but it is noted that at high temperatures it may be
necessary to account for emission from the sample during the PAS experiment (McGovern
et al., 1985).
3.3.5
223
0.8
Absorbance
0.7
24
Time/min
48
0.6
72
920
900
880
860
840
Wavenumber (cm1)
Figure 3.15. A real-time FT-IR waterfall plot of the change in the absorbance of the CH2 wag (at
887 cm1) during the polymerization of isobutylene. Reprinted with permission from Storey (1998).
Copyright 1998 American Chemical Society.
Addition polymerization
Vinyl polymerization reactions may be studied by transmission FT-IR, but, because of the
short pathlength (e.g. 50 lm) required in the MIR and the need for uniform heating of
the monomer in the absence of air, alternative approaches such as ATR are employed. The
traditional, laborious gravimetric approach to the study of kinetics of the living cationic
polymerization of isobutylene has been replaced with a real-time MIR analysis of the
consumption of monomer by monitoring the disappearance of the CH2 wagging band at
887 cm1 (Storey et al., 1998). In Figure 3.15, a stack plot of successive spectra obtained by
ATR using a composite diamond insertion probe is shown (after subtraction of solvent) for a
reaction time of over one hour.
Kinetic analysis of peak height and peak area gave identical proles and enabled temperature uctuations due to the high reaction exotherm to be highlighted. In a further study
with this system, which uses a light-conduit pipe so that the reaction is remote from the
spectrometer, the block copolymerization of styrene to living polyisobutylene chain ends
was monitored (Storey et al., 1999). It was shown that the real-time analysis for styrene
content agreed with the off-line analysis by 1H NMR. A similar study of the kinetics of
carbocationic polymerization of styrene and isobutylene using an ATR bre-optic probe
produced similar agreement between the MIR method and gravimetry (Puskas et al., 1998).
Fibre-optic probes are available with MIR transmitting optics as accessories for all brands of
FT-IR spectrometers for laboratory studies of polymerization, oxidation and functionalization reactions. The availability of commercial laboratory-scale polymerization reaction
systems incorporating ATR-IR probes (e.g. Mettler-Toledo React-IR) has allowed
qualitative and mechanistic studies of metallocene-catalysed copolymerization, polycondensation and polyaddition reactions (Sahre et al., 2006) as well as high-output polymer
screening (HOPS) to determine catalyst productivities for copolymerization by using kinetic
measurements in real time (Tuchbreiter, 2004). These systems and further applications are
discussed in the section on remote spectroscopy.
224
An industrial system based on transmission that is able to analyse molten polymer from a
reactor stream by using a variable-pathlength cell and a heated transfer line has been
reported (Fidler et al., 1991). The system is designed to tolerate pressures up to 300 bar and
temperatures up to 400 C, this requires optical components in the sample area to be made
of diamond or ZnSe. The application reported involved the determination of vinyl acetate
concentration in polyethylene, and good quantication was obtained by using the overtone of
the ester band rather than the fundamental. This enabled a longer pathlength to be used and so
avoid the high back-pressure from the narrow pathlength necessary to obtain quantication
of an intense band such as that of the ester. As will be discussed later, the use of overtone
and combination bands and longer pathlengths is a feature of NIR spectroscopy, which is
the technique of choice for applications such as reactive extrusion (Dumoulin et al., 1996).
Photopolymerization of acrylic monomers and related systems is a challenging area for
the application of real-time MIR spectroscopy since the reactions are often completed
within seconds, so there needs to be an excellent SNR in the spectra if reliable kinetic data
are to be generated. The temporal resolution of an FT-IR spectrometer at a resolution of
16 cm1 is 11 ms, so 100 spectra can be gathered in the time frame of the fastest reaction,
but without multiplexing to improve the SNR. Using a diamond ATR element of area 4 mm2
that could also be heated to 200 C, uniform irradiation was possible and the cure reactions
of lms of thicknesses ranging from 1 to 20 lm could be followed (Scherzer and Decker,
1999).
Network polymerization
The determination of the chemical changes in the development of a three-dimensional
network as the curing polymer moves from a liquid through a gel to a glass may be readily
studied by IR techniques. Systems that have been well studied include polyurethanes
(Koenig, 1999), cyanate esters (Bauer and Bauer, 1994) and epoxy resins (St John and
George, 1994). Because it attains a high Tg on curing with aromatic amines such as diaminodiphenylsulfone (DDS) at temperatures from 160 to 180 C, tetraglycidyl diamino
diphenyl methane (TGDDM) has widely been used as an aerospace epoxy resin for composite materials. It has therefore been studied widely by a range of techniques (St John and
George, 1994) including MIR spectroscopy. While IR cannot directly provide rheological
information, it may enable the factors which control the kinetics of the network formation to
be determined. In the study of reacting systems it is important, because of the overlapping
bands in the MIR, to have a detailed assignment of all peaks in the spectrum so that the
changes which occur on mixing and then subsequent reaction may be understood.
Figures 3.16(a) and (b) show the MIR spectra in two spectral regions for the TGDDM
resin (Ciba MY721) and the aromatic amine hardener DDS as well as the mixture containing 27% DDS, which is typical of commercial formulations used in composite materials.
The assignments of the signicant bands in the spectra are given in Tables 3.2 for the
TGDDM (MY721 resin) and 3.3 for DDS.
The changes in the MIR spectra on curing may be seen in Figure 3.17, which shows
spectra of the TGDDM 27% DDS mixture before and after curing for three hours and the
difference spectra (uncured cured). The bands which have become negative represent the
functional groups which have participated in the reaction.
From a comparison of the rst spectrum (Figure 3.17(a)) with the assignment tables, it is
seen that the intensities of all NH bands have decreased, as have those of the two epoxy
CH stretching bands at 3049 and 2993 cm1.
225
(a)
Absorbance
DDS
0
MY721
3600
3400
3200
Wavenumber (cm
3000
2800
1)
(b)
DDS
Absorbance
0
MY721
1600
1400
1200
1000
Wavenumber (cm1)
800
600
Figure 3.16 Mid-IR spectra for resin, hardener and their mixture, (a) in the hydrogenic region and (b)
in the ngerprint region. Reproduced with permission from St John (1993).
At the same time there is an increase in intensity for the methylene groups at 2918 and
2844 cm1 and a broad, hydrogen-bonded OH band. All of these observations are a result
of the ring-opening reaction of the epoxide with primary amine, the major chain-extension
reaction occurring up to gelation of the resin.
The changes in the ngerprint region, Figure 3.17(b), are more complex due to the
overlapping bands, some of which broaden and shift due to the increase in polarity of the
network as curing takes place. In spite of this it is possible to discern two amine-related
226
3623, 3508
3049, 2993
2918, 2844
1613
1516
1384
1257
1191
906
829
797
585
bands (at 1630 and 544 cm1), which are suitable for monitoring in addition to the main
epoxide band at 906 cm1.
Such detailed spectral analysis was rst reported by Morgan and Mones (1987), who
showed, through transmission MIR, the difference between the rates of the primary- and
secondary-amine reactions with epoxy and the etherication reaction, which requires higher
cure temperatures. A detailed knowledge of the mechanism of cure can enable, in certain
circumstances, a rationalization of the differences among gel times for resins of various
purities. In a study of the cure of tetraglycidyl diamino diphenyl methane (TGDDM) epoxy
resins of various purities (St John et al., 1993) it was shown that the major factor controlling
the gel time was the catalysis of the cure reaction by residual hydroxyl groups present as
impurities (e.g. glycols) in the resin.
227
(a)
Uncured
Absorbance
0
Cured
Cured Uncured
1
3800
3600
3400
3000
3200
2800
2600
Wavenumber (cm1)
(b)
Uncured
Absorbance
Cured
Cured Uncured
0
1600
1400
1200
1000
Wavenumber (cm1)
800
600
Figure 3.17. Changes in the MIR spectra in (a) the hydrogenic region and (b) the ngerprint region of
MY721 cured with 27% DDS for 3 hours at 160 C. Spectral subtraction (cured uncured) shows
the (negative) bands that change on cure. Reproduced with permission from St John (1993).
Figure 3.18(a) shows the MIR spectrum in the OH stretch region of three TGDDM
resins, namely commercial resins MY720 and MY721 and a puried resin from the ash
chromatography of MY721 (labelled TGDDM). It can be seen that there is a progressive
decrease in the total band area with increasing purity of the resin.
Absorbance
0.15
MY721
0.1
TGDDM
0.05
0
3600
3400
Wavenumber (cm1)
3200
(b)
0.1
Absorbance
228
1
0.05
3700
3600
3500
3400
3300
Wavenumber (cm1)
Figure 3.18. (a) Mid-IR spectra in the OH stretch region of MY720, MY721 and TGDDM resins.
Note the decrease in absorbance and band shift with increasing resin purity. (b) Curve tting of the
OH band for MY721 from (a) as three components. Note the close t of the sum of the bands to
the original spectrum. (St John et al., 1993).
The overlapping bands may be resolved and assigned to free and hydrogen-bonded
hydroxyl groups of impurities in the resin as shown in Figure 3.18(b) for the MY721
commercial TGDDM. The bands are assigned to 1, free OH (36503590 cm1); 2, intramolecularly hydrogen-bonded OH (35703450 cm1); and 3, a polymeric strongly bound
229
95
90
85
80
75
Gel
Time 70
(min.)
65
60
55
50
0
0.1
0.2
0.3
[OH0] (mol/kg)
0.4
0.5
Figure 3.19. A plot of the gel time of TGDDM resins of various levels of purity as a function of
[OH]0 as determined by MIR. The curve is calculated from a model of OH group catalysis of
epoxy ring-opening (St John et al., 1993).
type of hydrogen-bonding (below 3450 cm1). The most highly puried sample still contains hydroxyl groups, but there are equal amounts of the three types. As the resin increases
in hydroxyl content there is an increase in the fraction of intramolecular and strongly bound
hydroxyl groups. This has the effect of increasing the catalysis of the ring-opening of the
epoxy by the amine nucleophile, with a resultant shortening of the gel time of the resin for
the same concentration of amine.
This is shown in Figure 3.19 through the relation between the gel time of the resin and
the initial OH concentration in the resin (Oh et al., 2006, St John et al., 1993). The
experimentally measured values are shown and the effect is an increase of up to 50% in gel
time as the resin is puried.
The reasons behind the observed differences in rheological behaviour of these TGDDM
resins could be determined by analysis of the kinetics of the system and the quantitative
analysis of the hydroxyl content by MIR or other methods of chemical analysis. This is a
clear example of the effect of chemical purity of the resin on rheological properties.
The above studies were performed by using transmission MIR. In a study of a similar
resin system on a composite prepreg, it has been shown (Cole et al., 1988) that in situ ATR
using a Ge optical element was quantitatively more reliable than extracting the resin from
the prepreg and analysing by transmission IR. The error arose due to variability in sample
thickness of the cast lm and thus uncertainty in the absorbance values which were used
for calculation of the concentration of components. The applicability of DRIFT was very
limited for the in situ sample due to the high specular component from the glossy carbonbre prepreg surface. As was noted in Section 3.3.4, the KubelkaMonk function will
generate reliable absorbance spectra only if all components are diffusely reected. The KBroverlayer method of Koenig (1999) would avoid this problem and should produce reliable
DRIFT spectra.
The problem of reliably analysing the epoxy-resin component of a carbon-bre
epoxy prepreg has also been addressed by using FT-IR emission spectroscopy (FTIES)
0.8
0.7
0.6
0.5
0.4
Emissivity
230
0.3
0.2
2000
1500
1000
500
NH
nd
co
Se
NH
0.1
0
2000
1800
1600
1400
1200
1000
800
600
Wavenumber (cm1)
Figure 3.20. Successive FT-IR emission spectra from a thin lm of resin from a carbonepoxy-resin
prepreg during cure. Bands that change during cure are marked (George et al., 2006).
(George et al., 1996) as described above in Section 3.3.3. Instead of removing the resin from
the bres by extraction, a small amount was transferred to the hotplate of the emission oven
by contact, and the spectra were collected as the thin lm of epoxy resin and curing agent
was heated through its cure cycle. High-quality spectra were generated (Figure 3.20) and
the amine and epoxy bands identied in MIR transmission (Figure 3.16 and 3.17) were
identied and found to change quantitatively as cure proceeded.
One band for amine and one for epoxy suitable for such analysis are marked. Figure 3.21
shows an expansion of the region highlighted with a * in Figure 3.20, which corresponds to
the phenyl-ring quadrant bend mode of DDS at 552 cm1 (George et al., 2006).
There is a small shift of this band to higher wavenumber in emission compared with MIR
absorption (Table 3.3), but of most importance is that it is seen to decrease with cure, and
this decrease is accompanied by an increase in intensity of the band that is initially a
shoulder at 575 cm1. There is an isobestic point (marked) that is invariant in emissivity as
the two spectral regions change in intensity. This is an indication that there is a smooth
inter-conversion of the reactant to product and that Beers law is obeyed throughout cure.
This means that these bands may be used for quantitative analysis of the cure reaction. The
FTIES experiment, carried out on a vanishingly small amount of resin transferred from the
surface of the prepreg, was thus able to generate the cure kinetics of the resin and identify
the components. Further, FTIES was found to be able to follow the ageing of the prepreg
and so provide a rapid quality-assurance technique. This technique has also been applied to
the thermal cure kinetics of photocatalysed dicyanate esters and directly compared with
MIR transmission spectra (Liu and George, 2000). It has been shown to be reliable to high
extents of conversion for quantitative analysis by applying Equation (3.24) to the determination of absorbance, A, calculated for each wavelength from the emissivity, e(m),
231
0.40
Emissivity
0.35
0.30
0.25
ISOBESTIC POINT
AS PRIMARY AMINE REACTS
0.20
0.15
0.10
620
600
580
560
Wavenumber (cm1)
540
520
Figure 3.21. Detail from Figure 3.20 marked with * showing the decrease and growth of emission
bands during cure of thin lm of epoxy-resin prepreg.
determined from separate emission spectra using Equation (3.23). Emission spectra offer
the prospect of being able to monitor the reactive processing of thin lms of thermosetting
resins on metallic substrates (Pell et al., 1991).
Oxidation reactions
Among the earliest uses of MIR transmission and ATR spectroscopy has been the study of
the oxidation of polymers. The principal oxidation products from polyolens, polyamides
and elastomers such as EPDM are carbonyl-containing species including aldehydes,
ketones, esters, lactones, acids and peracids. The reaction scheme for their formation has
been determined from studies of model compounds and polymers over the past 20 years,
and a typical scheme is shown in Scheme 3.1 (Blakey, 2001).
All of these functional groups have a distinctive absorption band in the region, and a
Fourier self-deconvolution routine is often used to attempt to resolve the often broad
resulting absorbance band to give the best t of the various species to the composite
spectrum.
In determining this t, model compounds as well as specic reactions of the separate
components to produce adducts with distinctive spectra are used to identify the individual
spectra of the components. The latter strategy is also used when, under the processing
conditions, the components have short lifetimes and are reactive intermediates such as
hydroperoxides.
An example of this is the range of derivatization reactions devised by several authors
(Lacoste et al., 1991, Carlsson et al., 1987, Delor et al., 1998). This has enabled the
identication of p-, s- and t- hydroperoxides and also peracids in the oxidation of polypropylene. In the study of oxidation reactions during processing it has been shown that, in
the absence of deliberate addition of oxygen or pro-oxidants such as transition-metal salts,
chain scission is not accompanied by oxidation (Gonzalez-Gonzalez et al., 1998). It has
232
H2C
OH
C
H2C
CH2
n
tio
ac
str
b
HA
CH3
CH3
-Scission
H2C
CH2
CH3
+ H2C
CH 3
CH 3
CH 2
CH3
OH
H2C
H2C
CH2
H
H2
C
CH 2
CH3
O2
O
CH3
O
CH 3
O
H2C
H Abstraction
Thermolysis
CH
H2C
CH2
H
H2
C
CH3
H Abstraction
O2 / H Abstraction
CH 3
O
C
CH2
CH
OH
+ H
CH2
H2C
CH3
OH
H2
C
C
H
H2
CH 2
CH 3
CH3
O
C
CH
CH2
OH
O
CH2
CO2
+
OH
CH 3 +
HC
CH 3
CH3
H3C
C
CH
CH2
+ 2H2O
CH2
H
HO
RO
H
CH2
CH3
O
C
H3C
HC
CH
CH2
RO
CH
+ H2O
Scheme 3.1. Formation of oxidation products in the degradation of polypropylene through the
tertiary hydroperoxide intermediate that may be identied by MIR spectroscopy (often combined
with derivatization). After Blakey (2001).
233
.1
Emissivity
.15
6000
4000
2000
atio
n Ti
me
(s)
Oxid
.05
0
2000
1800
1600
1400
1200
1000
800
Wavenumber (cm1)
Figure 3.22. Time-resolved FT-IR emission spectra (every 13 s) showing the increase in emissivity in
the region 16501800 cm1 due to formation of oxidation products. Reproduced with permission
from Blakey (2001).
Figure 3.23. Detail of emission spectra (Figure 3.22) showing expanded the region 16501850 cm1
been usual to determine the products off-line, so real-time monitoring has proved difcult,
but IR emission offers an alternative approach.
Laboratory studies of FT-IR emission (George et al., 1995, Celina et al., 1997) have
demonstrated that it may provide real-time oxidation proles. Figure 3.22 shows successive
emission spectra from a single particle of polypropylene at 150 C in the presence of
oxygen.
These were collected every 13 s and the development of the oxidation bands with time is
apparent, while the remaining bands are unaltered. Figure 3.23 shows a detail of ve of
these spectra over the carbonyl spectral region (1850 cm1 to 1650 cm1) and the formation
234
of the range of oxidation products as predicted by the scheme in Figure 3.20 may be seen, of
which the lactone and terminal carbonyl are highlighted.
This should be adaptable to on-line analysis by using the transient IR emission methodology developed by (Jones and McClelland (1990, 1993) for other applications. In this a
hot gas jet is used to bring about transient heating of a thin layer of the sample so that
reabsorption effects from thick samples (Section 3.3.4) are avoided. Having a moving
stream of sample means that only a thin layer of emitting polymer is seen by the spectrometer, so a continuous monitor of composition is obtained during reactive processing.
This may be adapted to real-time monitoring using an appropriate MIR-transmitting optical
bre (George et al., 2006) as discussed later.
3:25
While this works well for polymers of moderate molar mass (<5 10 mol/g) and welldened end-group composition, it does require the use of difcult-to-handle solvent mixtures. It would be simpler if the titration could be performed spectroscopically and in real
time so that a continuous record could be obtained during the course of reaction without the
need to dissolve the polymer. Nuclear magnetic resonance (NMR) provides one method,
provided that the terminal protons may be resolved and integrated as discussed in Section
3.3.2. An infrared method has been illustrated for polyesters, for which a similar relation to
Equation (3.25) is obtained, where hydroxyl end groups replace the amine end groups of
the polyamide. The analysis of poly(butylene terephthalate) end groups in the MIR (Kosky
et al., 1985) to measure [COOH] was achieved at 3290 cm1 and at 3535 cm1 (Oh et al.,
2006). Because the end groups are isolated, model compounds may be used to determine the
molar absorptivities necessary to convert absorbance to concentration, Equation (3.21). The
analysis of hydroxyl groups is complicated spectroscopically because of the strong
hydrogen bonding (shown above in Figure 3.18(a) for TGDDM-based epoxy resins) which
affects spectral line width as well as molar absorptivity, so careful calibration against
titration data is necessary to establish quantication.
There is a further complication in the analysis of polyesters (and several other polymers)
and this is the presence of cyclic oligomers during the polymerization process, which are not
removed from the commercial polymer. These oligomers will affect the rheological and
mechanical properties but, since they have no end groups, cannot be detected by end-group
4
235
0.06
0.05
0.04
Area 841
0.03
0.02
0.01
0.00
0
20
40
60
80
100
% Cyclic silicone
Figure 3.24. Relative area of the band at 841 cm1 (due to trimethylsilyl end groups) in the MIR
analysis (either spectral or chemical). This effect is also seen with polydimethylsiloxanes
(PDMSs), which are polymerized by the ring-opening reaction of a cyclic precursor so that
a typical level of residual cyclic material is 10%. The end-group analysis of PDMS in
the MIR region may be most readily performed at 841 cm1. Study of model compounds
(Noll, 1968) enables assignment of the absorption of the dimethyl (in-chain) group
[(Me)2SiO] to the strong bands between 790 and 830 cm1, compared with the trimethyl
end groups [(Me)3SiO] as two weak bands at 841 and 750 cm1. The effect of the
presence of cyclic oligomers on the absorbance of the band at 841 cm1 (and thus the
apparent molar mass of the sample) may be seen in Figure 3.24, which shows the decrease
in the band area (A841) as a function of cyclic-silicone content. At low levels, the effect on
calculated Mn is about 15%.
In spite of the limitations, in suitable polymers, MIR analysis offers a rapid technique for
studying changes in molar mass that occur during processing. The greatest limitation is in
the adaptation of the technique to on-line analysis and the requirement for thin samples due
to the high molar absorptivities. The near-infrared region offers many experimental
advantages, as discussed below, but there are severe restrictions on the functional groups
which may be analysed at the levels encountered in chemorheological applications, particularly those involving controlled oxidation.
3.3.6
236
the light H atom, bands such as CO modes that are strong in the MIR are very difcult to
observe in the NIR, since the anharmonicity is small. The molar absorptivities of the
overtones and combination bands in NH, CH and OH are lower by one to two orders of
magnitude than those of the fundamental vibrations, so pathlengths of millimetres are
typically used, again simplifying sample handling. Consequently, this spectral region has
been the most heavily studied for process control, often involving the use of bre-optics
(as discussed later).
237
Network polymerization
The analysis of the cure reactions of networks has been facilitated by the use of NIR
spectroscopy because of the rapid development of the crosslinked system which renders
solution spectroscopy impossible. The advantage of use of the NIR region over the MIR
studies, previously discussed in Section 3.3.3, is the ability to analyse thicker samples
(typically of thickness 15 mm) and thus obtain results more representative of the bulk
curing reactions. The quantitative features of NIR spectral analysis were recognized with
one of the earliest analyses of the epoxide and hydroxyl-group content of a commercial
DGEBA epoxy resin before and during cure (Dannenberg, 1963). The advent of highperformance composite materials, and the need for a deep understanding of the chemistry of
cure and post-cure of networks and the correlation with mechanical and environmental
performance, meant that there was a need for spectral techniques that would complement
the DSC and DMA methods being used for cure analysis (Gupta et al., 1983, Cole et al.,
1988). Most of these studies were in the MIR and the limitations of this spectral region for
kinetic analysis were discussed earlier (Section 3.3.5). There have been several studies of
DGEBA-based epoxy resins cured with a range of hardeners in which the NIR studies
provide quantication and the MIR studies, which are less reliable quantitatively, are able to
assist in mechanistic understanding (Poisson et al., 1996, Strehmel and Scherzer, 1994). For
example, in the case of DGEBA cured with dicyandiamide and accelerated by diphenyl
dimethyl urea (Diuron) it was found that, while the heterogeneous system resulted in light
scattering, this did not affect quantication in the NIR, but the MIR results showed poor
correlation with NIR results and those from other methods such as SEC and chemical
titration (Poisson et al., 1996). The use of simultaneous analytical methods together with
NIR offers the advantage that artefacts can be determined and a deeper understanding of
subtle changes in analytical results can be gained, for example by NIRdielectric methods
(Kortaberria, 2003).
The use of NIR for monitoring the reactive processing of high-temperature-curing, highperformance resins such as TGDDM (e.g. the Ciba resin MY721) cured with DDS was rst
reported in an application that used bre-optics to couple the sample to the spectrometer
(George et al., 1990, 1991). In Figure 3.25 is shown a NIR spectrum for a sample of
TGDDM (MY721) containing 27% DDS.
The assignment of the bands in the spectrum to combination and overtone bands of the
fundamental vibrations assigned earlier for both TGDDM and DDS (Tables 3.2 and 3.3) is
shown in Table 3.4 for both the resin and the amine.
4519
4557
4804
5072
5645, 5774
5887, 5987
5873, 6057
6575
6684
7000
8627, 8837
9801
10191
combination
bands
Absorbance
238
CH
1st overtone
bands
1.5
NH
1st overtone
bands
OH & NH
2nd overtone
bands
10000
CH
2nd overtone
bands
9000
OH
1st overtone
bands
7000
8000
Wavenumber (cm1)
6000
5000
Figure 3.25. Near-IR spectra and region assignments for the epoxy resin MY721 with 27% DDS
hardener before cure reaction at 160 C. Reproduced with permission from St John (1993).
It should be noted immediately that there are many fewer bands in the NIR; also they are
broader and overlapping of bands is common. In some instances, Fourier self-deconvolution
(Section 3.3.4) is required in order to assign bands, but the original bands should be used
when possible for quantitative analysis since Beers law is generally obeyed and quantication is straightforward. When the sample in Figure 3.25 was heated at 160 C for three hours
for the cure reaction to occur, the spectral changes shown in Figure 3.26 were observed.
The bands that have become negative correspond to consumption of the amine and
epoxide, and it is the detailed analysis of these bands which permits a quantitative analysis of
the cure kinetics, which may then be applied to real-time monitoring of reactive processing.
239
Absorbance
Uncured
Cured
0
Cured Uncured
9000
8000
7000
6000
Wavenumber (cm1)
5000
Figure 3.26. Near-IR spectra of MY721 plus 27% DDS before and after cure at 160 C for 3 hours
together with the difference spectrum to highlight bands that change during cure. Reproduced with
permission from St John (1993).
There are many commercially signicant networks that are based on free-radical polymerization of multifunctional monomers. Obvious examples include crosslinked surface
coatings (Paul, 1989), styrene-crosslinked unsaturated polyesters (Melby and Castro, 1989)
and crosslinked acrylates and related systems for optical transparencies (Carswell et al.,
1996), all of which require process control as the network is formed in order to achieve
optimum performance. These are often copolymers with monofunctional monomers in
which polymerization is initiated either by UV radiation and a photoinitiator or by a thermal
free-radical initiator such as a peroxide, so the polymerization is often rapid and results in a
heterogeneous network (Dusek, 1996). The kinetics for network formation are those for
free-radical polymerization, so, in contrast to the stepwise polymerization of epoxy resins,
during which the molar-mass increases slowly, there is the immediate formation of highmolar mass polymer, which is initially soluble in the monomer, but, because of the multifunctionality, will soon crosslink and become gel particles. This gel may form at extents of
conversion as low as 10%15%, so subsequent reaction will occur in the gel particles, but
macroscopic gelation may be delayed due to the formation of intramolecular crosslinks and
cyclization (Dusek, 1996). Near-infrared spectroscopy may be applied in these systems in
the way that it is applied to other free-radical polymerizations (Yamada et al., 2000), and it
has been shown (Rey et al., 2000) that, by establishing a molar absorption coefcient for
each of the reacting species, both methacrylate and vinyl conversions could be obtained
240
after spectral deconvolution. The effects of inhibition by oxygen and onset of diffusion
control were demonstrated and are of vital importance in monitoring of reactive processing.
3.3.7
3:26
241
3400
3000
2600
2200
1800
1400
1011
902
845
798
752
641
1384
1259
1191
3071
3002
2919
1614
1000
600
200
monitoring of emulsion polymerization (Charmot et al., 1999, Bauer et al., 2000) has successfully been demonstrated using a ruggedized version of the laboratory spectrometer.
In a study of ATRP of halogen-free amino-functionalized poly(methyl methacrylate), the
solvent (anisole) was used as the internal standard that enabled the on-line consumption of
monomer as given by the decrease in the Raman band at 1641 cm1 to be followed with a
time resolution of 2 min. Close agreement was found with off-line analysis by 1H NMR, but
it was noted that the immersion probe detects a thin layer (thickness 500 lm) on the
surface, which may adhere at high conversion when the viscosity is high, and then material
exchange with the polymerizing bulk might not occur (Fischer et al., 2006). For these
reasons, non-contact analysis is preferred, but might not always be practical.
In the area of monitoring of the cure of high-performance epoxy resins for composite
materials the relative merits of FT-Raman have been compared with those of FT-IR in the
NIR region (as discussed in Section 3.6.2) (de Bakker et al., 1993a).
Figure 3.27 shows the FT-Raman spectrum of TGDDM (MY721), and the assignment of
the prominent bands is made in Table 3.5.
The Raman-spectral data from the aromatic amine curing agent DDS are also shown in
Figure 3.28, and assignments are given in Table 3.6.
It is of interest to compare these with the MIR spectra discussed earlier (Figures 3.16 and
3.17 and Tables 3.2 and 3.3) and note the differences in the spectra and, in particular, band
intensities arising from the different selection rules. Of particular interest are the differences
in the analytical band used in the MIR for epoxide analysis (904 2 cm1) due to an
asymmetric ring vibration, which is very weak in the Raman, where the preferred analytical
band is the symmetrical ring vibration (at 845 cm1) that is not active in the MIR.
For complete kinetic analysis and to give a cross-check when using spectral methods for
monitoring reactive processing, it is important to monitor several bands due to groups that
participate in the chemical reactions for network formation and at least one band that does
242
3071
3002
2919
1614
1440, 1385
1260
1191, 1010
902
845
798
642
3367
3330
3065
1629
1597
1500
1280
1140
1107
1073
822
634
482
289
not change (an internal standard). For the TGDDM/DDS system, the primary amine is a
sensitive index of the early reaction up to gelation, so the vibrations of NH2 have been
monitored both in MIR (Morgan and Mones, 1987) and in NIR (St John and George, 1992)
analyses of cure. Unfortunately, in the Raman, all of the bands that are attributable to this
group are weak, so this immediately limits the power of the Raman analysis compared with
the NIR, where the NH2 combination band at 5072 cm1 (Table 3.4) has been shown to
give an excellent SNR and is ideal for monitoring the epoxideprimary-amine reaction (St.
John and George, 1992). A comparison of the potential application of the two methods to
monitoring the cure of TGDDM/DDS may be seen in Figure 3.29.
In this plot the Raman data are referenced to the peak at 2919 cm1 which is due to a
stretching mode of the methylene group in TGDDM and therefore should not vary
during cure because this group does not participate in the reaction. It may be seen that the
poorer SNR for the Raman spectra results in a high uncertainty in the degree of conversion
243
3400 3000
1400
1000
Raman
290
191
482
634
1073
1278
1498
3368
1629
837
3068
1140
1592
600
200
Shift / cm1
Figure 3.28. The FT-Raman spectrum of the aromatic hardener DDS (de Bakker et al., 1993a).
1.0
0.8
0.6
Conversion
p
0.4
0.2
0
0
Raman
NIR
10
20
30
40
50
60
Time/min
Figure 3.29. Comparison of the degree of conversion, p, as a function of cure time for TGDDM cured
with DDS as measured by NIR and Raman methods. Adapted from de Bakker et al. (1993a).
of NH2 at any time, which is emphasized by the long time between data points due to the
large number of interferograms which need to be co-added to generate the spectrum. This is
contrasted with the large number of data points that establish the cure prole from the NIR
curve. It was concluded that NIR was superior to Raman spectroscopy for real-time
monitoring of the resin cure, but an interesting insight into a competing side reaction
244
involving cyclization of a chlorohydrin impurity in the commercial resin was obtained from
an analysis of the Raman spectrum (de Bakker et al., 1993a). This could not be detected by
other spectral techniques. The strong uorescence of these resin systems prevents the use of
grating systems, and FT-Raman systems will not have the superior SNR and time resolution
seen in charge-coupled-device (CCD)-based detector systems with which spectra may be
collected every 50 s (Fischer et al., 2006).
Degradation reactions
There have been several studies of the degradation of polymers, which may be adapted to
the study of reactions leading to changes in properties (including viscosity during processing) and thus are relevant to chemorheology. One of the most obvious changes which
may be monitored by FT-Raman is the dehydrochlorination of poly(vinyl chloride) (PVC).
The mechanism of dehydrochlorination and stabilization of PVC has been studied for some
time (Braun, 1981, Starnes, 1981), and the main reason for the rapid production of hydrogen
chloride is known to be the facile unzipping reaction which is auto-catalysed. This results in
the formation of extended sequences of polyenes (as shown below), which may be readily
detected by Raman spectroscopy:
CH2 CHCln ! CHCHm CH2 CHClnm mHCl:
3:27
dramatic changes in the FT-Raman spectrum occur on heating PVC and its copolymer
with poly(N-vinyl-2-pyrrolidone) (PVP) (Dong et al., 1997) due to polyene formation.
Figure 3.30 shows intense bands around 1500, 1100, 610690 and 1430 cm1 corresponding
to (CC)n, (CC )n and CCl stretching vibrations and CH2 deformation bands,
respectively. The extent of dehydrochlorination and polyene formation of 90% PVC/10%
PVP blends with respect to time at 120 C can be seen from 13 consecutive spectra presented in stacked form. Each spectrum was recorded at 5-minute intervals over 50 scans,
while keeping the temperature constant at 120 C (Kaynak et al., 2001). The PVP is able to
undergo hydrogen bonding to PVC, and increasing the amount of PVP in the PVC/PVP
blends decreased the dehydrochlorination temperature and time and resulted in an earlier
decrease in the mean polyene length. The sensitivity of Raman Spectroscopy for detecting
polyenes means that as little as 104% dehydrochlorination of PVC and its copolymers can
be detected (Hillemans et al., 1993).
While successful in monitoring the formation of highly conjugated products on degradation, Raman spectroscopy is much less sensitive than IR for the detection of oxidation
products such as carbonyls, so that any measurements made to monitor changes during
oxidation involve secondary effects such as the loss of aliphatic CH compared with aromatic CH during the oxidation of an acrylonitrileEPDMstyrene terpolymer as a way of
locating the site of degradation (Mailhot and Gardette, 1996) rather than direct measurement
of oxidation bands.
3.3.8
245
Figure 3.30. Successive Raman spectra collected during the thermal degradation of a PVC : PVP blend
(90 : 10) at 120 C, showing the rapid increase in the bands at 1100 and 1500 cm1 due to polyene
formation. Adapted from Kaynak et al. (2001).
to determine the reaction pathway and develop strategies for controlling the reaction rate.
Thus, for the chemorheology of photoinitiated reactions, detailed knowledge of the excited
states and energy-transfer routes to enable sensitization of the reaction and/or the possible
quenching of excited states is invaluable (Guillet, 1987). By exploiting this fundamental
knowledge and identifying particular spectral parameters that are sensitive to viscosity
changes, it may be possible to achieve non-invasive monitoring of the viscosity and phase
changes which accompany network formation.
Figure 3.31 and Table 3.7 describe the processes of emission and energy relaxation which
may take place following absorption of radiation of intensity Ia einsteins dm3 s1
expressed in terms of the rates of the process (George, 1989b). These relaxation processes
may involve either the singlet excited states, leading to uorescence from the rst excited
singlet state, S1, or radiationless processes of internal conversion (shown as wavy lines); or
the triplet states which may be populated by inter-system crossing and then decay by
emission of phosphorescence from the rst excited triplet state, T1, or radiationless intersystem crossing to the ground (singlet) state, S0.
In an isolated molecule these processes will be unimolecular (as shown by the rate
expressions in Table 3.7), but in solution or the molten state there will be processes of
energy transfer that may take the form of quenching of either the singlet or the triplet state,
leading to a decrease in the intensity of emitted uorescence or phosphorescence. These are
bimolecular processes as shown in Table 3.7. Because direct population of the triplet state
by absorption from the ground singlet state (T1 S0) is formally forbidden, the lifetime of
this state is long (from milliseconds to several seconds in the absence of bimolecular
processes). This long lifetime is the reason why the triplet state is photochemically active
and also why phosphorescence may be detected only in the absence of quenching molecules
such as ground-state oxygen. In contrast, the transition to the rst singlet state (S1 S0) is
allowed, so the lifetime of the rst singlet state is very short (nanoseconds in the absence of
246
Table 3.7. Photophysical processes and their rates after absorption of light, intensity I0,
in terms of the rate coefficients and populations of excited states shown in Figure 3.31
Transition or reaction
Rate
Unimolecular process
Absorption
Internal conversion
Internal conversion
Inter-system crossing
Fluorescence
Phosphorescence
Inter-system crossing
Sv1
S0
Sv1 ! S1
S1 ! S0
S1 ! Tv1
S1 ! S0 ( hm)
T1! S0 ( hm 0 )
T1! S0
Ia I0(110ecd)
kIC[Sv1]
kIC[S1]
kISC[S1]
kF[S1]
kP[T1]
k 0ISC[T1]
Bimolecular process
Singlet quenching
Triplet quenching
Excimer formation
Excimer emission
S1 Q ! S0 Q*
T1 Q ! T0 Q*
S1 S0 ! (S1S0)*
(S1S0)* ! 2S0 ( hm00 )
kQ[S1][Q]
k0Q[T1][Q]
kDM[S1][S0]
kFD[(S1S0)*]
Singlet states
Triplet states
Higher excited states
S2
kIC
T2
v=3
v = 2 Intersystem crossing
v=1
v=0
kISC
S1
Absorption
kIC
kIC
T1
kF
kP
S0
v=3
v=2
v=1
v=0
Ground state
v=n
v=1
v=0
k'ISC
S0
Figure 3.31. An energy-level diagram for an aromatic molecule or similar chromophore that may
absorb radiation in the visible or UV region of the electromagnetic spectrum (George, 1989).
quenching) and the more intense the absorption, the shorter the lifetime. However, bimolecular processes do occur and an interesting example is excimer emission through the
formation of an excited state complex (S1S0)* between an excited molecule (S1) and a
ground state molecule (S0). These bimolecular processes are relevant in this particular
application since the rate constants for quenching (kQ) and excimer formation (kDM) as
247
shown in Table 3.7 will be dependent on the encounter probability of molecules in solution
and therefore related to the viscosity.
In the situation described as SternVolmer quenching, the uorescence quantum yield UoF
is reduced to UF in the presence of Q, a quenching molecule of concentration [Q],
U0F UF 1 kQ sF Q;
3:28
where sF is the lifetime of the rst singlet excited state in the absence of any quenching,
i.e. sF 1/(kF kIC kISC). The rate coefcients for these competing processes are shown in
Table 3.7.
In solution the quenching rate coefcient will be the diffusion-controlled rate coefcient
for encounter, kd, which is related to the viscosity g by the Debye equation:
kQ kd 8RT=3g:
3:29
Thus there is the potential for the quenching of excited states to be used as a monitor of
the viscosity of the system containing a uorescent molecule with an excited-state lifetime
appropriate for a measurable change in uorescence intensity (i.e. from UoF to UF) to occur
over the range of viscosity of interest. The quenching molecule also must meet the energetic
requirements of having an energy level lower than the singlet excited state S1 of the
chromophore and also having a rapid loss of the excitation energy through radiationless
decay (kIC is large) so that emission is not observed from the quencher.
Excited-state quenching is one example of excitation energy transfer from a donor
(excited) molecule, D*, to a previously unexcited acceptor molecule, A, having a lowerlying singlet energy level at which the acceptor does not itself uoresce. In many cases, A*
may emit radiation and then there will be an emission spectrum, which, depending on the
concentration of A, will feature the spectra of both D and A and, in the case of total energy
transfer, emission from A alone. In certain circumstances the rate of dipoledipole energy
transfer can be so efcient that it may occur at greater than encounter distances (e.g. 30
100 A) such that the rate of energy transfer exceeds the diffusion controlled rate and
becomes independent of viscosity. In this regime, the uorescence is unsuitable as a monitor
for reactive processing. However, there has been a particular example of these systems in
which the D and A are electron-donor and -acceptor species and are located on the one
molecule to form intramolecular charge-transfer uorescence probes (Song and Neckers,
1996). These exhibit spectral shifts during polymerization or network formation that are due
to the effect of microviscosity on the accessibility of the twisted charge-transfer state.
Excimer emission is a broad structureless emission red-shifted from the normal uorescence and is observed at high chromophore concentration because it requires close
encounter of the excited and the unexcited molecule to form the encounter complex (S1S0)*
with a separation of 3 A. This will depend on the encounter occurring during the lifetime
of the excited state so that, as the viscosity increases, there will be a decrease in the intensity
of excimer emission. This is discussed and an example shown in Section 3.3.8.
248
cure chemistry without affecting the overall course of network formation (Sung et al., 1986,
Yu and Sung, 1990). The reaction of the probe mimics that of the curing agent and, through
studies of a series of ve model compounds corresponding to progressive reaction of the
amine with the epoxide, to give secondary or tertiary adducts that are either monofunctional
or difunctional, the corresponding red shifts in the p* p absorption spectra were recorded.
These were then deconvoluted as cure proceeded to yield the relative contributions to the
different reactions of the amine. This molecule is thus not a viscosity probe but a reactivity
probe, and its use requires that the reactivity of the probe with the resin has the same kinetic
parameters as for the hardener. The reactivity of aromatic amines varies with the base
strength of the amine, which is inuenced by the bridging group, e.g. DDS (with a sulfone
bridging group) is a weaker base than DDM (which has a methylene bridging group) and
requires a higher reaction temperature. Indeed, the authors note that DAA reacts faster than
DDS and hence corrections need to be made (Sung et al., 1986). This is a severe restriction
to the general use of this approach in real-time monitoring of the reaction, but it can provide
H2N
NH2
DAA
Me
Me
Me
Me
NO2
CN
H
CN
DANS
DABM
BPP
Me
Me
N
Me
N
Me
Me
DASPI
QB
Scheme 3.2. Molecular probes used for UV absorption and uorescence monitoring of the reactions
249
interesting fundamental information that can complement that obtained from the other
approaches such as NIR spectroscopy, which was described above in Section 3.3.6. However, the advantage of NIR and MIR spectroscopies is that they can follow the actual
reactive species, so no assumptions need to be made regarding the relative reactivity of the
probe and the resin components. It is also an advantage if the spectral properties of the probe
are sensitive to the viscosity of the medium rather than the reactivity.
It has been suggested that the DDS itself will show a progressive red shift of up to 24 nm
in the p* p absorption spectra with cure, and this is most readily seen from the uorescence excitation spectra (Paik and Sung, 1994), which are equivalent to absorption
spectra. It was shown that three techniques, namely UV reection, UV transmission and
uorescence excitation, gave similar spectral shifts as the cure reaction took place at 160 C
(Yu and Sung, 1995). Excitation spectra are generally more sensitive at low concentrations
of analyte but may saturate and become unrepresentative of the absorption properties at high
concentrations. This was illustrated with a DGEBA-based epoxy resin and, while there was
a linear relation between the spectral shift and the extent of reaction as determined by IR, it
was noted that there was a temperature dependence of the shift for the DDS. This was
accommodated by using a rhodamine dye as an internal standard, which also allowed for
any variations in lamp intensity, alignment etc. Thus even this method needed to have a
probe added, which does result in some limitations in a processing environment.
excimer-forming probes
intramolecular charge-transfer probes
twisted intramolecular charge-transfer or TICT probes
Fluorescent salts
In many cases the differences among the mechanisms of action of the last three categories
are subtle and are related to whether there is a shift in the uorescence spectrum with
changing polarity of the medium and the actual nature of the excited-state relaxation that is
environment-sensitive. Since there are now many hundreds of organic molecules available
as uorescence probes (Song and Neckers, 1996, Vatanparast, 2001, Bosch et al., 2001,
Bosch, 2005, Loutfy, 1981, 1982, 1986, Wroblewski et al., 1999, Rettig, 1986, Quirin and
Torkelson, 2003), only a few representative examples are considered. In choosing a probe
issues to be considered include the likely reactivity of the probe under the conditions of
reactive processing and the ease of extracting the data of interest (Mikes et al., 2002a,
2002b).
The p, p0 -diamino azobenzene (DAA) reactivity probe, described in the previous section,
has been demonstrated to form reaction products with DGEBA-based epoxy resins that have
a greater uorescence intensity than does DAA itself. By examining model adducts with
phenyl glycidyl ether, the intensity enhancement varied from 1400 times to 9 times
depending on whether the DAA reacted at a crosslink or a chain end, respectively
250
(Sung et al., 1985). Part of this enhancement is due to a greater molar absorptivity, but there
is also a contribution from an enhanced uorescence quantum yield. When incorporated into
a curing epoxy-resin system at a level of 0.10.3 wt.%, the increase in intensity with time is
sigmoidal, with the point of inection identied with the gelation of the resin. The levelling
off in intensity was identied with vitrication of the resin, but it was recognized that this
was due not to the rapid increase in viscosity, but rather to a loss of reactivity consistent
with DAA being a reactivity probe, not a viscosity monitor.
The possibility of using a probe molecule possessing intrinsic uorescence properties
that change with the viscosity of the medium has been considered for some time (Tredwell
and Osborne, 1980). This change in the intrinsic unimolecular rate processes for uorescence is to be contrasted with the bimolecular quenching of uorescence which may follow
SternVolmer kinetics, as discussed above in Section 3.3.8. By using a solvent series that
increased in viscosity, it was shown that the uorescence lifetime (sF) of the rhodamine dye
N-(2-tolyl)-N 0 -(2-tolyl-5-sulfonate) rhodamine, Fast Acid Violet 2R (or FAV) increased
from 10 to 160 ps as a result of viscosity-dependent ultrafast internal conversions (kIC, Table
3.7). The relation followed was
sF Cg2=3 :
3:30
However, it was the microviscosity, rather than the bulk viscosity which was probed by
using picosecond time-resolved uorescence methods, since the value of C was constant
within a homologous series of solvents but was different for monofunctional and difunctional alcohols. The probe molecule has a natural (radiative) lifetime, soF of 4.6 ns, so the
reduction in lifetime results from the competition between uorescence (kF 2 108 s1)
and the high rate of internal conversion (kIC 1011 s1). This was attributed to rotation of
the N-tolyl group of the dye, which then has to sweep through the solvent and hence probes
the local environment.
This concept was extended by Loutfy (Law and Loutfy, 1981, Loutfy, 1982, 1986), who
studied a series of p-(N,N-dialkyl amino)benzylidene malononitriles. These molecular
rotors have a large side group attached to the double bond of the malononitrile, as shown
for the structure DABM in Scheme 3.2. Since the molecule loses excitation energy by
rotation about the double bond, if this is inhibited, there will be a lowering of kIC and thus an
increase in the uorescence intensity. When these probes were incorporated into a free
radical polymerization of styrene, acrylates or methacrylates, there was a sigmoidal increase
in emission intensity with time. When this was converted to viscosity, Equation (3.30) was
followed over the range from 0.2 to 100 cP, indicating that the macroscopic and microscopic
viscosities are both probed by the molecular rotor. However, as the macroscopic viscosity
increased even further as the glassy state was approached, the microviscosity remained
constant and there was no change in uorescence intensity. This probe therefore is sensitive
to gelation, not to vitrification.
The relative sensitivity of reactive and non-reactive monosubstituted nitro-stilbene probe
molecules for the photopolymerization of diacrylates and dimethacrylates has been studied
by real-time uorescence (Jager et al., 2001). It was found that the acrylate-terminated
nitrostilbenes were more sensitive than the non-reactive probes at the early stages of the
reaction. This was attributed to the higher rate of incorporation of the reactive probes into
the highly crosslinked regions which form during early stages of the polymerization reaction
(Jager and Vanden Berg, 2001). There has been detailed study of the non-reactive DANS
251
Increasing
Cure Time
Intensity
(arb. units)
500
550
600
650
probe, 4-(N,N-dimethylamino)-40 -nitrostilbene (Scheme 3.2) which shows polarity sensitivity and thus a shift in the uorescence spectrum as cure of an epoxy resin takes place
(Quirin and Torkelson, 2003) (Figure 3.32). This sensitivity arises since the molecule
contains electron-donor and -acceptor moieties, which results in signicant charge separation in the excited state. The probe is self-referencing and insensitive, to temperature,
so non-isothermal cure reactions could be followed (Quirin and Torkelson, 2003).
Figure 3.33 shows the effect of conversion of epoxy (i.e. increasing viscosity) on the
wavelength of the maximum emission for excitation at 475 nm and the ratio of emission
intensity at 588 and 564 nm. There is a sharp increase in emission intensity but no spectral
shift at the conversion corresponding to the Tg at the cure temperature, indicating that the
probe is sensitive to both conversion and vitrification.
It has been noted (Strehmel et al., 1999) that the uorescent dyes 1,10 -dimethyl-2,20 carbocyanine iodide (QB) and p-N,N-dimethylamino-styryl-2-ethylpyridinium salt (DASPI)
may be used both to monitor the crosslinking process and to detect Tg. These dyes have
twisted intramolecular charge-transfer (TICT) states that are accessible from the rst
excited singlet state. Such TICT states may be considered to arise from intramolecular
donoracceptor electron transfer whereby the new state is stabilized in a geometry perpendicular to the original conformation (Figure 3.34).
Rotation occurs about the double bond in a manner similar to what occurs in the malononitrile derivatives described by Loutfy (1982, 1986). However, instead of measuring the
steady-state emission intensity, the rate coefcient for the formation of the TICT state was
obtained by picosecond laser uorescence lifetime measurements (Strehmel et al., 1992,
1999, Younes et al., 1994).
If kAB is the rate coefcient for formation of the TICT state, the measured lifetime is
sF 1=kF kIC kAB :
3:31
252
660
4
640
max (nm)
600
I 588 /I 644
620
580
0
20
40
80
60
% Conversion
Figure 3.33. The uorescence emission intensity ratio and peak wavelength of DANS
(see Figure 3.32) as a function of epoxy-resin conversion at 80 C. Adapted from Quirin and
Torkelson (2003).
eD
kAB
D+
Figure 3.34. A schematic diagram showing movement occurring in the case of a TICT mechanism in
a molecular probe such as DASPI and the reaction volume probed. After Strehmel et al. (1999).
The reference lifetime in the absence of TICT formation, sF0 , is obtained by measuring
the uorescence decay in a rigid solvent at 77 K, at which temperature it may be assumed
that kAB 0. Then,
kAB 1=sF 1=s0F 1=sAB :
3:32
Plots of ln sAB against 1/T (K) of DASPI in a DGEBA resin show that the probe is able to
detect Tg of the resin as a change of slope. Interestingly, when the cure of the resin was
studied with these two probes, different proles were obtained, with DASPI having an onset
for the increase of sAB at the gel point, whereas for QB the gel point is the inection in the
sAB versus time curve. Thus QB behaves exactly as did the molecular rotor in sensing
gelation, whereas DASPI monitors the network formation beyond gelation and into the
glass. From a study of the cure of a range of resins and curing agents with DASPI, these
results are rationalized in terms of the available free volume that the probe is able to sense
(Strehmel et al., 1999).
In a study of seven different dyes containing electron-donor and -acceptor groups, it was
noted that the TICT states are sensitive to the polarity of the matrix as well as to the
microviscosity (Vatanparast et al., 2001). From a study of the curing of a DGEBA epoxy
253
resin with methyl nadic anhydride, signicant blue-shifts in uorescence spectra were
observed during cure. These shifts were linked to polarity changes, so they were not reliable
indicators of the degree of cure. It was also shown that various probes exhibited different
behaviours in the same curing environment, with uorscence some of them even decreasing
in intensity.
The choice of viscosity probes for polymerization is not restricted to dye molecules, and
certain organometallic complexes show changes in luminescence during thermal and
photochemical polymerization (Lees, 1998b). In particular, it was noted that metal carbonyl
derivatives containing lowest-energy metal-to-ligand charge-transfer transitions are sensitive to viscosity changes of over ve orders of magnitude when incorporated into epoxy and
cyanate ester thermosets and acrylate and epoxy photo-curing resins (Lees, 1998a).
It was noted in Section 3.3 that the bimolecular processes, which compete with uorescence, such as energy transfer, quenching and excimer formation, should offer a way to
monitor the changes in the viscosity of the medium. In contrast to the probes discussed
before, the intensity of uorescence generally will decrease with an increase in viscosity as
the resin cures. In a number of studies, Wang et al. (1985, 1986a, 1986b) investigated the
changes in the uorescence properties of excimer-forming dyes dissolved in epoxy resins
and other polymerizing systems. The uorescence of the probe 1,3-(bis-1-pyrene) propane
(BPP in Scheme 3.2) shows all of the features of excimer emission outlined at the end of
Section 3.3.8, viz. a structured free-molecule or monomer emission and a broad redshifted excimer spectrum. Pyrene has been the most-studied molecule that readily forms a
sandwich-type excimer as the planes of the molecule come together (Guillet, 1987). In
solution, the intensity of excimer emission (IE) of pyrene increases compared with the
monomer emission, IM, with increasing concentration, as expected for a bimolecular process, but for BPP the excimer may form intramolecularly, regardless of the concentration.
The sensitivity of the intensity of excimer emission to change in viscosity arises since the
excited pyrene group may adopt a large number of conformations, only a few of which will
allow it to overlap sufciently with an unexcited group on the same molecule to achieve the
minimum in energy to stabilize the excimer. Thus, as viscosity increases, the probability of
achieving this conformation within the lifetime of the excited state decreases, so decreasing
the excimer emission intensity (Fanconi et al., 1986). This has been shown for the amine
cure of DGEBA (Wang et al., 1985), for which, after an initial increase in IE/IM, due to the
decrease in viscosity as the temperature was increased from 22 to 60 C, there was an eightfold decrease as the resin gelled. Thus a plot of the inverse intensity ratio, IM/IE, allows a
comparison with the known viscosity prole on curing of an epoxy resin (Apicella, 1986).
There have been studies in which pyrene alone has been used as the uorescent probe for
in situ polymerization (Pekcan et al., 1997, 2001, Pekcan and Kaya, 2001). It is assumed
that the observed enhancement of uorescence intensity or increase in singlet-state lifetime
of the monomeric pyrene as the network forms arises from a decrease in bimolecular
quenching events as the viscosity increases and kQ decreases (Equations (3.28) and (3.29)).
It is noted that the spectra showed no evidence of pyrene excimer formation, suggesting that
the pyrene is isolated in the network at the concentrations employed.
An alternative approach to obtaining a photophysical measurement of viscosity that may
be applied to monitoring polymer cure is to measure the polarization properties of the
uorescence of an added probe (Scarlata and Ors, 1986). This again measures a property of
the molecule that depends on its rotation in the polymerizing medium. The uorescence is
excited with polarized light and the extent of depolarization of the emission will depend on
254
the rotation of the probe following excitation, in addition to the depolarization due to the
angle between the absorbing and emitting dipoles of the molecule. The emission anisotropy,
r, is given by
r I I? =I 2I? ;
3:33
where I and I? are the luminescence intensities when measured through analysing polarizers
which are oriented parallel and perpendicular, respectively, to the plane of polarization of
the exciting light. The limiting anisotropy, r0, applies to the value obtained if the probe is
motionless for the excited-state lifetime, s. The link between anisotropy and viscosity is
given by the Perrin equation:
r0 =r 1 kB s=V T=g;
3:34
255
species with the lowest-lying energy levels dominating the emission spectrum. Commercial
polymers will also contain pigments and additives such as antioxidants and UV absorbers,
so these will also add their spectral features.
It has been noted in Section 3.3.7 that the rapid formation of polyenes was a feature of
thermal degradation of PVC and that these could be most readily detected in the Raman
spectrum. The polyenes also have a characteristic uorescence spectrum that may be used to
study the spatial as well as temporal development of degradation (Remillard et al., 1998).
Other polymers for which the detection of conjugated oxidation products or a change in the
extent of conjugation of the backbone or side groups is most readily performed by luminescence include polyethylene terephthalate (Edge et al., 1995), polycarbonate (Chipalkatti
and Laski, 1991) and the polyolens (Allen and Owen, 1989a, 1989b, Jacques and Poller,
1993). In PET the appearance of excimer uorescence at the expense of molecular
(monomer) emission on heating at melt temperatures is interpreted as due to main-chain
breakdown and aggregation of the terephthalate fragments, while the longer-wavelength
species are attributed to hydroxylation of the aromatic ring. More extensive oxidation
results in quinone and stilbene quinone species, which are responsible for the yellowing of
the polymer. Bisphenol-A polycarbonate at processing temperatures forms phenyl o-phenoxy benzoate, which is uorescent. The progress of degradation was followed by monitoring emission intensities at 330 and 347 nm (Chipalkatti and Laski, 1991).
The changes in the luminescence properties of polyolens with degradation have been
attributed to the formation of a, b-unsaturated ketones (or ene-ones) (Jacques and Poller,
1993, Allen and Owen, 1989b), which may have a pro-oxidant effect. While these species
emit strong phosphorescence, the detection of these and other conjugated sequences by
uorescence has been suggested (Jacques and Poller, 1993, Allen and Owen, 1989b).
3.3.9
3:35
where R is the rate of the reaction which produces the emitting species in the excited state.
In spite of the low value of UCL in polymer oxidation, reactions with rates as low as
1012 mol dm3 s1 may be studied, which correspond to free-radical concentrations of
1013 radicals dm3, which in an analytical volume of 1 ml gives a sensitivity 1000 times
that of ESR (Section 3.3.2).
256
In oxidizing polyolens such as polypropylene, the oxidation in the melt may be treated
as a homogeneous system kinetically, in contrast to the situation in the solid state (George
and Celina, 2000), such that the rate of oxidation is dependent on the rate of reaction of the
chain-carrying alkyl peroxy radical with the polymer. In the Russell mechanism for CL, the
energy for the excitation of CL is given by the termination reaction of two peroxy radicals
(Blakey et al., 2006). Formally this requires one of the radicals to be either primary or
secondary, so that the chemistry occurring in the light-producing reaction(s) may change
with the extent of oxidation (George, 1989a). However, this may be simplied kinetically if
steady-state conditions prevail, and the rates of initiation (by hydroperoxide thermolysis)
and termination (by peroxy-radical recombination) are equal:
ICL UCL kt RO2 2 UCL kd ROOHn ;
3:36
where kt is the rate coefcient for bimolecular termination of the alkyl peroxy radicals of
steady-state concentration [RO2], kd is the rate coefcient for the decomposition of the
initiating hydroperoxides, ROOH, and n is the order of this reaction. In an auto-oxidation at
elevated temperatures, [ROOH] will change continuously, passing through a maximum and
then decreasing. This is most clearly seen in the oxidation of PVC during processing (Scott,
1993). At lower temperatures and in the absence of transition-metal salts, which catalyse the
decomposition through a redox cycle, there may be an accumulation of hydroperoxide as an
oxidation product, but in the temperature regime for processing this will not occur.
In principle, the CL from polyolens could be used as a monitor of the extent of oxidation
during reactive extrusion in the presence of peroxides to decrease the molar mass (Brown,
1992), but this has not been reported. Attention has been paid to the effect that change in
viscosity of a polymerizing network has on the CL prole, as described in the following section.
257
60
50
40
30
20
10
0
10
20
30
40
50
60
maximum in chemiluminescence (CL) intensity over the temperature range 140200 C for the
following epoxy-resin/hardener systems: *, Tactix 742/27% DDS; D, Shell 1153/22% DDS;
and Shell 1071/27% DDS. Adapted from Kozielski et al. (1995).
TGDDM, the uorescence intensity and lifetime increased with viscosity (Equation (3.30))
since it has a structure too similar to that of the molecular rotors (Levy, 1984). Thus, on
combining the two relations, it would be expected that
ICL C 0 RTg0:33 RO2 2 ;
3:37
so that the intensity of CL should decrease with time as the viscosity increases. In principle,
the change in [RO2] with time should follow the prole for the oxidation of the neat resin
since the viscosity should be constant, but this will hold only if the oxidation reaction does
not become diffusion-controlled in the curing resin. The complexity of the CL prole from
epoxy resins was demonstrated when systems other than TGDDM were studied (Kozielski
et al., 1995). It was found that all systems exhibited a maximum or deection in CL with
time of cure that correlated with the gel times as measured by the crossover of G0 and G00 in
separate measurements (Figure 3.35).
However, it was also shown that the CLtime prole was sensitive to the purity of the
resin and the amine, DDS, which was found to contain a quenching agent, so the quantum
yield of emission was reduced in the uid state (Equation (3.28)). It is therefore concluded
that simple CL experiments are not generally applicable as a method for monitoring the cure
of epoxy resins. In addition to the self-probe uorescence of TGDDM, the CL measurements rely on absolute emission intensities, which, in the absence of an internal standard for
calibration, is unreliable.
It was noted above that measurements of the lifetime of emission provide an alternative
way of accessing the photophysical parameters of a polymer. The analogous approach in CL
is to subject the oxidizing polymer to an external perturbation and then observe the change
in CL as the system returns to the steady state (George, 1981). This will generally take the
form of a short period of UV irradiation after which there is a burst of CL followed by
decay, which may be analysed according to an assumed kinetic model, such as second-order
258
kinetics for the CL arising from the bimolecular termination of peroxy radicals. Such an
analysis leads to the following relation for the change in CL intensity, I, with time:
1=2
I 1=2 I0
kt =UCL 1=2 t:
1/2
3:38
1/2
Thus a plot of I
against time should be linear and have a slope of (kt/UCL) , while
the lifetime of the decay will depend on the initial peroxy radical concentration:
1=2
3:39
It was shown that these equations could be used to analyse the CL from thin lms of
epoxy resins during cure (Schweinsberg and George, 1986) and there was a change in the
parameter (kt/UCL)1/2 as cure passed through gelation. This was also shown to be applicable
to monitoring the cure of a carbon-breepoxy-resin prepreg (George, 1986), but the change
in the above parameter as the resin transformed to a gel was only ten-fold and this, together
with the time taken to measure the decay properties, was considered a severe limitation to
the usefulness of the CL techniques for studying the early stages of network formation.
An interesting observation was that the emission intensity increased rapidly as the resin
vitried, becaming insensitive to the presence of oxygen. The lifetime of the emission also
increased and it was concluded that the emission arose from charge-recombination luminescence (Billingham et al., 1989).
Charge-recombination luminescence
The origin of charge-recombination luminescence (CRL) during the cure of epoxy resins
has been studied in detail (St John, 1993, Sewell et al., 2000, Billingham et al., 1989). From
the ESR and UVvisible spectra it was found that the photo-irradiation of glassy TGDDM
produced a radical cation centred on the tertiary amine site (Figure 3.10), analogous to the
Wursters Blue cation (St John, 1993). While photo-ionization may occur in the uid state,
the photo-electron must be trapped before electronion recombination can occur, so the
system must both have a trapping site and be sufciently rigid to separate the ion pair for a
sufcient time to allow the decay to be resolved over times ranging from seconds to hours.
This has been exploited in studying the vitrication during cure of TGDDM, and it has been
determined that it is the hydroxyl groups which are more important than the amine sites in
determining the stability of the ion pair (St John, 1993, Sewell et al., 2000). The rapid
increase in CRL was found to occur always at the same extent of reaction regardless of the
cure temperature. In a study of various epoxy-resin and diacrylate networks, it was found
that, although the CRL was weaker and shorter-lived than that from TGDDM, the vitrication of these resins too could be monitored (Lange et al., 1998, 1999). In the case of the
diacrylate network, there was evidence presented for micro-vitrication since the CRL was
detected at low extents of conversion, which is a further example of the extreme heterogeneity of these networks (Dusek, 1996).
Experimentally the technique is simple since the sample may be irradiated from time to
time and the emission intensity measured by a photomultiplier over a period of tens of
seconds (Lange et al., 1998). However, these studies conrmed the limited usefulness of
CRL alone for real-time monitoring since the emission is signicant only in the glassy state,
so this technique cannot detect the gel point for the majority of networks. When combined
with other luminescence methods it has potential use because it is independent of the
atmosphere and is readily adaptable to bre-optics.
3.4
259
Remote spectroscopy
The spectroscopic techniques described in Section 3.3 have all been developed for application in an analytical laboratory and generally require polymer samples to be taken to the
spectrometer. For the monitoring of chemorheological processes and reactive processing, it
is necessary to take the spectrometer to the reacting system or the rheometer. For systems
such as NMR that require large magnetic elds to operate, this is clearly impractical, so
magnetic-resonance spectroscopy has not been assessed for real-time remote monitoring inline other than in the preliminary studies such as of the high-temperature surface NMR
probe that has been adapted to operation at the die of an extruder (Gottwald and Scheler,
2005). There is greater potential for NMR and MRI techniques in off-line monitoring and
also for providing fundamental information such as imaging of ow (Callaghan, 1999). In
contrast, vibrational and electronic spectroscopy, both in absorption mode and in emission
mode, lend themselves to remote sampling by coupling with bre-optics. Some examples of
laboratory-based optical-bre systems were given in Section 3.3.5. In the following, specic
examples are presented from the viewpoint of sampling and spectroscopic considerations
such as analytical volume. The use of remote sensors for process control has been
researched for the past 30 years, and the expansion of bre-optic telecommunication networks and the decreasing cost of low-loss bre-optic cable have provided an impetus for
research in this area (Krohn, 1988, Boisde and Harmer, 1996). Developments in optoelectronics and photonics have been responsible for sophisticated bres ranging from single-mode ultrathin monolament bres and bundles transmitting a single wavelength to
large-diameter multimode bres that transmit a wide range of wavelengths and are therefore
suitable for remote spectroscopy. A detailed review of the application of optical-bre
sensors to the process monitoring of reinforced organic matrix composites (Fernando and
Degamber, 2006) includes a comprehensive survey of the fundamental principles of optical
bres and experimental congurations for monitoring of thermosetting reactions. Hardware
available for monitoring of conversion, copolymer composition and particle size by techniques in addition to bre-optic spectroscopy has been reviewed comprehensively (Kammona et al., 1999).
3.4.1
Principles of fibre-optics
The fundamental requirement of all radiation-transfer techniques in remote spectroscopy is
that the radiation be transferred from the spectrometer to the sample, probe the reactions or
transformations of interest and then return the modied beam of radiation to the spectrometer for the measurement of intensity at each wavelength in the spectral region of
interest, all without any contributions from the transfer medium and with little loss in
energy. While bre-optics meets these requirements under favourable circumstances, there
are potential artefacts and limitations, which may be understood from the principles of
operation.
The key principle in the operation of bre-optics, and also in several of the sampling
methods, is total internal reection. This is illustrated in Figure 3.36, which shows three
cases for a ray of light travelling in a medium of refractive index n1, incident at three
different angles, h, to the interface with a medium of lower refractive index, n2.
It is noted that only for values of h greater than hc will there be no refracted ray such that
light is totally internally reected at the interface and propagates within the medium of
260
Refracted
ray
r
n2
n1
n1
Critical
ray
n2
n2
n1
Total
reflection
and n2 below, at and above the critical angle, hc (n1 > n2).
n2 cladding
Beam in
n0
c
lim
n1
core
1
n2
Beam out
n0
cladding
Figure 3.37. A schematic diagram of an optical bre showing propagation of light beams dened
by the numerical aperture of the bre and the refractive indices n1 and n2 of the core and
cladding materials; n0 is the refractive index of air. Adapted from Boisde and Harmer (1996).
refractive index n2. Thus, if the light is launched into a bre at angles less than sin1(n2/n1),
there is propagation along the length of the bre and the attenuation of the energy depends
on the following.
1.. The intrinsic absorption properties of the bre material at the wavelength, k, of the
radiation. This is usually expressed in optical power loss, ak, in dB/m, per unit length of
bre, L, which is related to the spectrophotometric absorbance of the material, A
(Equation (3.21)), by the relation
ak 10A=L:
3:40
2.. The numerical aperture (NA) of the bre. This is shown in Figure 3.37, which is a simple
ray diagram for the entry of radiation into the bre under the condition for total internal
reection (i.e. h < hc) and thus propagation along the length. This denes the limiting
angle of incidence, hlim, at the end of the bre in contact with the air (refractive index n0),
and then
1=2
NA n0 sin hlim n21 n22
:
3:41
Thus the numerical aperture, NA, depends on the materials of construction of the bre and
the larger the difference between the refractive indices of the cladding and the bre core,
the greater the range of angles for rays that may propagate along the bre. This is relevant
for matching the numerical apertures of spectroscopic and other optical components for
analysis of the beam in order to optimize the SNR.
3.. The absorption of the evanescent wave. The evanescent wave is the component of the
totally internally reected wave that penetrates into the medium of lower refractive index
(i.e. the cladding) by a distance that depends on the angle of incidence. This is illustrated in
261
x
Cladding
Cladding
n2
A
B
C
Core n1
Evanescent
wave
d
E
A
B
C
B
Core
Figure 3.38. The penetration depth, d, for the evanescent wave from the core into the cladding for
Figure 3.38, which schematically illustrates three beams A, B and C at different angles of
incidence to the cladding and the depth of penetration of the evanescent wave for each. E is
the electric-eld intensity and it can be seen that this decays exponentially into the
cladding.
The wavelength and angle dependences of the depth of penetration, d, were given earlier
(Equation (3.22)), and this principle is used in the MIR for the sampling technique of ATR
spectroscopy. Figure 3.13(c) illustrated this, and, for typical refractive indices of bre core
and sample of 2.5 and 1.5, respectively, and an angle of incidence of 45 , the depth of
penetration of the evanescent wave is about 0.15k. The effect of the evanescent wave on
bre-optic spectroscopy may be illustrated by the example of plastic-clad silica (PCS) bre
optics for remote spectroscopy in the NIR spectral region, as discussed below.
Examples of the effect of these factors on remote spectroscopy may be seen by considering the use of PCS bre in a simple capillary transmission cell for the measurement of
NIR absorption spectra (George et al., 1991). The energy from a source such as a tungsten
lamp or glow-bar emitting in the NIR is focussed on a polished end of the PCS bre by
using a quartz lens or a parabolic mirror. The optics are designed to optimize transfer to the
bre by matching the NA of optics and bre. The various methods of coupling the bre to
the reacting system are discussed in the next section, but the simplest method is to cut and
polish the bre and place the sample between the two ends at a distance as short as possible.
To maximize transfer from the launching bre to the gathering bre, collimating lenses
(again matched to the NA) may be used, or, if pathlengths are short, the energy losses for
simply mounting the sample in a metal capillary cell may be tolerable. The gathering bre is
then returned to the detector of the spectrometer and the signal recorded. For high-quality
spectra it is necessary to subtract the spectrum of the bre itself from that of the sample plus
bre. Figure 3.39 gives the NIR spectrum of a 2-m length of silicone-clad silica bre-optic
monolament (George et al., 1991) and several features may be noted.
(i). The absorbance of 3 for the bre material at 7100 cm1, which would correspond to an
attenuation of 15 dB/m at this wavelength. This arises from the rst overtone of the OH
vibration of silica at 3540 cm1 and, if spectral information is required in this region, it is
necessary to use a bre core with a low silica OH concentration. Other bands attributed
to OH are at 8000 and 10 500 cm1. When used in a capillary cell, this results in a dead
band (Figure 3.40) with a total bre length of only 2 m. Another dead band corresponding
262
3.00
A
b
s
2.63
2.25
1.88
r
b
1.50
a
n
c
e
1.13
0.75
0.38
0.00
10675 10025
9375
6125
5475
4825
Figure 3.39. The NIR absorbance of a 2-m length of PCS optical bre, showing the spectral artefacts
that may be introduced by species present in core and cladding (George et al., 1991).
1.00
A
b
s
0.88
0.75
0.63
r
b
0.50
0.38
n
c
e
0.25
DEAD BAND
DEAD BAND
0.13
0.00
10675 10025
9375
6125
5475
4825
Figure 3.40. Dead bands and negative bands arising in a transmission cell used for remote
spectroscopy due to the core and cladding bands shown in Figure 3.39 (George et al., 1991).
to the cut-off of the transmission of silica in the IR is seen at 4800 cm1. At wavelengths
away from the OH bands the attenuation is of the order of 310 dB/km.
(ii). The presence of weaker, but sharper, bands in the range 54006000 cm1. These correspond to the rst and second overtone and combination bands of the methyl group of
poly(dimethyl siloxane) used for the cladding. That these arise from the evanescent wave
may be calculated from the measured absorbance of the methyl band of 0.15 at
5950 cm1. Using a reported molar absorptivity of 0.1 dm3 cm1 mol1 (Goddu and
Delker, 1960), an equivalent pathlength of 500 lm may be calculated from Beers law
(Equation (3.21)). By using the attenuated total reection relation (Equation (3.22)) it is
found that this pathlength corresponds to the attenuation of the evanescent wave for a
263
launch angle of 75.1 , which is close to the limiting angle of 76 (Equation (3.40)) of the
bre-optics, which has a numerical aperture of 0.4. The effect of these cladding bands on
the use of the capillary cell for spectroscopy may be seen in Figure 3.40. Negative bands
may arise due to a mismatch between the sample and reference pathlengths. Bending of
the bre may contribute, since this will change the angle at the interface and thus the
depth of penetration into the silicone cladding. The effect of these cladding bands on
quantitative bre-optic spectroscopy is severe with certain polymers such as epoxy
resins, and signicant interference of analytical bands for the monitoring of the reactive
processing has been reported (George et al., 1991). This may be overcome by moving to
silica-clad silica, which has a low OH content to minimize attenuation. In this way it is
possible to obtain spectra in a remote capillary cell that are of good SNR and free of
interfering negative bands.
3.4.2
Beam splitter
He-Cd laser
and optics
Figure 3.41. A system for remote bre-optic uorescence monitoring of viscosity changes during
processing of a carbon-breepoxy-resin laminate. Adapted from Levy and Schwab (1987).
264
3:42
This analysis would predict that the reection coefcient, R, would decrease with cure,
since n2 will progressively increase, whereas R was found to increase systematically and
produce a prole that followed that of a DSC conversiontime curve for an epoxy resin.
265
(a)
(b)
ex
em
em
ex
(c)
(d)
em
ex
em ex em
em
Figure 3.42. The arrangement of excitation (kex) and emission (kem) beams from optical bres for
remote uorescence or Raman spectroscopy. The resulting analytical volume is shaded.
Reproduced with permission from Boisde and Harmer (1996). Copyright 1996, Artech House.
This suggests that the process of energy transfer to the bre is more complex and that
changes to the NA of the bre may be important.
An alternative to the illumination system of the type shown in Figure 3.41 is to use
separate excitation and gathering bres, either as separate bundles, held side by side, or as a
mixed bundle with a central excitation core surrounded by gathering bres. This avoids the
use of a beam splitter and efciently transfers the emitted radiation to the detector in the
second arm of the bifurcated bundle. This geometry is equivalent to the front-face illumination mode of uorescence spectroscopy and has been used by several researchers when
studying the monitoring of reactive processing using molecular probes that are sensitive to
viscosity (Vatanparast et al., 2001) or using energy-transfer efciency as a measure of blend
morphology when the bre-optics is coupled to a rheometer (Vorobyova and Winnik, 2001).
The analytical volume in this and the single-bre mode has been considered for several
different possibilities shown in Figure 3.42 (Boisde and Harmer, 1996).
These considerations apply to remote Raman spectroscopy using bre-optics (Section 3.4.2)
as well as uorescence. While a single-bre system offers convenience, especially in conned
spaces and with small samples as may be encountered in an on-line sampling port or a reactor,
the double-bre system has superior SNR (Louch and Ingle, 1988). This is especially so when
the uorescence intensity from the sample is low and high excitation intensities must be
employed. In this case the uorescence from the single bre will also be detected, severely
limiting the length of bre that can be used. As may be seen qualitatively from Figure 3.42,
the volume probed and thus sensitivity increase if the excitation and emission bres are at an
angle of 15 30 to one another. In general the excitation and gathering efciencies are
governed by the NA of the bres. Commercial systems are available with congurations
266
Reaction volume
Evanescent volume probed
Mirrored end
Clad fibre
Bared fibre
Figure 3.43. Use of a bared optical bre as a reaction probe by using the evanescent wave to sample a
volume controlled by the refractive index of the bre and reactant. Adapted from Park and
Sung (1994).
(c) and (d), but such geometries are expensive and are not always suited to the rugged environment of reactive processing. They do, however, overcome the limitations of the uorescence
of single bres, which can become dominant when the optical bre is bent (Myrick et al., 1990).
An alternative geometry is the optical-bre-facing conguration in which the excitation and
gathering bres are in-line, and lters are used to remove spurious radiation, but this requires a
sampling arrangement similar to that for transmission spectroscopy, which might not always
be compatible with the available space in the reaction zone. The transfer efciency may be
increased by using graded-refractive-index lenses bonded to the ends of the bres as well as
collimating optics such as Winston cones (Myrick et al., 1990). The use of bre-optics in
conjunction with other cure-monitoring techniques to obtain simultaneous spectral and physical information on reactive processing is discussed later in this chapter.
In one common method used with multimode bres (Paik and Sung, 1994, Powell et al.,
1995, 1998), the cladding is stripped from the optical bre and the polymerizing network is
sensed by the evanescent wave (Figure 3.43).
As the refractive index of the resin approaches that of the bre, the energy is lost since the
condition for total internal reection (Section 3.4.1) is no longer met. In a variation on this
method, a short section of the optical bre is replaced with fully cured resin of the same
dimensions, before immersion in the curing resin system. Initially, the energy will be
transmitted by total internal reection along the resin section, but, when the refractive index
of the curing resin equals that of the resin section, energy will again be lost. The system is
also able to monitor the progressive change in refractive index as well as the fully cured
state, since there will be a change in the critical angle, sin1(nuncured/ncured), as the network
develops, and the detector will see a progressive change in signal (Afromowitz, 1988, Lam
and Afromowitz, 1995a, 1995b). One advantage of using a length of bre with the same
refractive index as that of the fully cured resin as the sensor is that the sensitivity can be
controlled by merely increasing the length of the sensing bre, but the penetration of the
evanescent wave is limited at shorter wavelengths, so the sensor length may need to become
prohibitively large in order to achieve an acceptable SNR in the visible region. However,
this small depth of penetration can enable the interphase region in composite materials to be
studied (Paik and Sung, 1994, Woerdeman and Parnas, 2001). (The interphase is the region
of resin adjacent to the reinforcing bre in which the properties of the resin differ from the
bulk due to the reactions between the resin and the coupling agents deposited on the bre
surface and the effect of the activity of the surface on the local concentrations of both resin
and hardener.) This may extend several tens of nanometres from the surface. When properties closer to those of the bulk polymerization are to be monitored, the limited penetration
depth of the evanescent wave may be overcome by operating in the NIR and MIR regions.
In this case, the penetration depth increases from 50 nm in the visible region to 0.1 mm in
267
Cool air in
Sapphire
window
To and from
spectrometer
Optical probe
Air out
Figure 3.44. A bre-optic probe with air cooling and a sapphire window suitable for use in a reactive
the NIR and 1 lm in the MIR. The precise depth of penetration for a particular wavelength
may be calculated from Equation (3.22).
268
interchange of probes without affecting the process stream. An analysis of the materials of
construction for windows in the high-pressure zones (Reshadat et al., 1999) shows that
sapphire has the best combination of thermal and mechanical properties and has a reection
loss of between 7.0% and 7.5% in the NIR, depending on wavelength.
The limitation of performance in the NIR due to the materials of construction of the bre
has been noted (Dallin, 1997), and the small residual OH functionality in most commercial
bres will absorb energy and result in an upper spectral range of 2250 nm. This has been
shown in the example in Section 3.4.1 (Figure 3.39), and the dynamic range is generally
reduced from an upper absorbance limit of about 3.03.5 in NIR spectroscopy in the
instrument to about 1.52.0 when monitoring is performed using bre-optics (Dallin, 1997).
However, the major contribution to this limit is the actual loss of energy at each interface in
the optical path rather than the intrinsic energy losses in the bre alone. One way of
minimizing bre energy losses is to shorten the path or, in transmission measurements,
place either the light source (Calvert et al., 1996) or the detector (Dallin, 1997) adjacent to
the sample, with the main function of the bre being to enable the spectrometer to be
operated remotely from the process line, which is usually a hostile environment for the
optical system of a grating or interferometer. It has been shown that a total optical-bre
length of 2 km may be used with a transmission cell for the real-time analysis of a process
stream in the 12 lm spectral range by using a pair of multimode silica bres that have very
low OH absorption (Mackison et al., 1992).
Quartz bres generally have a refractive index that is too low for evanescent-wave
spectroscopy of curing epoxy resins (although, as shown earlier, in Figure 3.39, the NIR
absorption spectrum of the lower-refractive-index silicone cladding is recorded in the
spectrum of the PCS quartz bre-optics). However, if a section is fabricated with a material
of higher refractive index (e.g. tungsten oxide or sapphire) then ATR spectra of the substrate
in the NIR region may be obtained. A dip probe may be constructed by coating the end of
the element with gold so that the effective pathlength is doubled because also the return
beam probes the substrate. The length of the element is governed by the molar absorptivity
of the bands to be analysed, but, for the curing of composite materials in an autoclave,
sufcient attenuation of the evanescent wave is obtained for one to be able to monitor the
chemical changes as the network polymerizes (Druy et al., 1988b). In the particular
application, the bre must be surrounded by resin in the article being fabricated and the
effect that the permanently embedded bre (which will be at least ten times the diameter of
a reinforcing bre in the composite) will have on the mechanical properties and integrity of
the product must be considered. Alternatively, the sensing bre may be located in a test
piece accompanying the part in the autoclave. In spite of the limitations, the interphase and
in particular the reactions occurring between the bre surface, the adsorbed silane coupling
agent and either epoxy resins or unsaturated polyester resins have been monitored using
evanescent-wave spectroscopy in the NIR (Johnson et al., 2000). In this application, a
bundle of specially fabricated bres having a int-glass core of diameter 25 mm and a sodalime glass cladding of thickness 1 mm approximated the thickness of typical E-glass
reinforcing bres. A further advantage of the system was that the cladding composition
approximated that of E-glass so that the interphase around the special bre bundle was
considered to approach that around the reinforcing bres. The length of the sensing bre is a
denite limitation to the use of ATR in the NIR, so more applications have been researched
in the MIR spectral region.
269
270
271
and Suhonen, 2001) an absorptive CdTe lter replaced the usual holographic notch lter for
laser line rejection at 830-nm excitation with no loss in efciency. Another factor affecting
overall sensitivity is the volume of the sample from which the scattered radiation is
gathered. An analysis of the sampling depth for an FT-Raman spectrometer revealed that
this was generally limited by the scattering properties of the sample rather than a limitation
in depth of eld of the spectrometer and optics system (Barrera and Sommer, 1998). In some
low-absorbing and optically transparent samples, the depth of analysis approached 10 mm.
When applied to the real-time monitoring of processing this ensures that Raman spectra are
measuring the bulk reactivity of a system and are not sensitive to surface effects and
reactions at the interphase as may occur with evanescent-wave techniques (Section 3.4.2).
A feature of Raman spectra is the temperature sensitivity of the ratio of anti-Stokes
scattering to Stokes scattering. As noted in Section 3.3.3 the anti-Stokes scattering lies to
higher energy than the Rayleigh line and so arises from the molecules which are in the
higher-lying vibrational state. The population in this state and thus the intensity will depend
on the Boltzmann distribution, so the ratio of intensities, IAS/IS is given by
IAS =IS kS =kAS 4 exphv=kB T ;
3:43
where kS and kAS are the Stokes- and anti-Stokes-band wavelengths, respectively, and m is
the frequency of the exciting laser radiation. If either the sample or the bre shows a Raman
band that does not change with the reaction occurring, then the ratio IAS/IS may be used to
determine the temperature (Dakin, 1989). In the case of the bre being the sensor, by using a
pulsed laser and gating the Raman spectra, it is possible to determine temperatures at
different points (e.g. 10 m apart) along lengths of bre. This is not expected to be applicable
in real-time processing because of the shorter distances involved, in particular since the
capability of this method is considered to be 1 K or less per metre of bre (Dakin, 1989). If,
however, the sample has a band that is invariant with conversion and also shows strong
Raman scattering, then the technique has an obvious advantage for measuring temperature
at the same point as that at which the concentration of reactive species is measured. This has
been demonstrated for the measurement of both local temperature and conversion of a
commercial cyanate ester polymer (AroCy L-10) (Cooper et al., 1996).
A remote monitoring probe for in situ Raman spectroscopy based on a single bre
terminating in a Teon capillary, so enhancing the Raman signal 15-fold compared with the
bre-optics alone, has been described (Aust et al., 1996). When immersed in an epoxy resin,
the lower refractive index of the tube ensures that it acts as a waveguide, so enhancing the
signal. A further feature of this simple system is that the multivariate data-analysis routine
(see Section 3.5) not only allowed the epoxy-resin conversion to be followed in the CH
region of the spectrum, but also showed the effect of temperature uctuations. This was
achieved by using a peak at 3069 cm1 that remained unchanged during cure, but was
temperature-sensitive. Unlike the earlier system (Cooper et al., 1996) which ratios the antiStokes and Stokes shifts (Equation (3.43)), this approach is based solely on the Stokesshifted band and so does not yield a direct measurement of temperature without calibration.
3.5
272
3.5.1
273
Consider the set of spectra generated from a polymer sample as it undergoes chemical
reaction during processing. At each time interval, a new set of spectra is collected and
averaged to give a spectrum for that particular time. These are digitized and stored. As the
reaction takes place, some spectral bands will remain unchanged and others will change in
absorbance at a particular wavelength due to the chemical reaction changing either the
concentration of the species or the chemical nature of the species. If there is a change in
chemical species then there may be changes to all of the variables in the spectrum, viz.
frequency, bandwidth and absorbance.
The challenge is to determine which components are changing and their concentration at
any point in time. Collecting a large number of spectra corresponding to the various
compositions generates a large data set that can allow this to be achieved. Modern FT-IR
spectrometers come with chemometric routines for this data analysis, and stand-alone
systems that can perform these computations in real time as spectra are collected are also
available. The main issue is the quality of the spectral data when quantitative analysis is
required.
274
6000
0 MINUTES
2.4 MINUTES
4000
2000
(a)
0.4
0.2
(b)
PC1
0.0
1100
1200
1300
1400
1500
Wavenumber (cm1)
1600
Figure 3.45. (a) Raman spectra from an epoxy resin (DER332/T403) before and after 2.4 minutes of
microwave curing and (b) a loading plot of PC1 versus Raman shift. From Stellman et al. (1995).
Copyright 1995 by Society for Applied Spectroscopy; reproduced with permission of Society for
Applied Spectroscopy.
8000
6000
4000
2000
PC1
Score
0
-2000
-4000
0.5
1.0
1.5
Microwave time (min.)
2.0
Figure 3.46. A score plot of PC1 versus cure time for the system shown in Figure 3.45. From
Stellman et al. (1995). Copyright 1995 by Society for Applied Spectroscopy; reproduced with
permission of Society for Applied Spectroscopy.
multiplication of the score at each time by the loading should regenerate the spectrum at
that time.
This loading plot shows features similar to the sigmoidal curve generated by the integrated
heat ow during a DSC run for the analysis of epoxy-resin cure (Figure 3.5, Section 3.2.2)
that measures the extent of reaction as a function of cure time.
275
100
80
60
% cure
40
20
0
: Intensity of CN band
: PC1 Scores
100
300
200
time (minutes)
400
Figure 3.47. A comparison of cure reaction monitoring by a univariate method (CN band intensities)
and a multivariate method (PC1 scores) for a cyanate ester resin. Adapted from Cooper (1999).
The system has thus been reduced to a univariate problem and conventional analysis may
be employed for subsequent quantication if required. In a closely related study of the
curing of a cyanate ester rather than an epoxy resin (Cooper, 1999) a similar prole to
Figure 3.46 was obtained after PCA (Figure 3.47), and a comparison with the univariate
analysis of single peak intensities that are the major peaks in the rst principal component
after mean centring was made.
This provides an interesting comparison of the univariate and multivariate approaches.
By virtue of using all of the spectral data in PCA, the SNR achieved is much higher than for
any single-wavelength analysis, providing a more precise plot of extent of reaction over
time. However, there is limited mechanistic information since the subtle changes due to
reaction intermediates may be lost in lower-order principal components that, because they
do not contribute greatly to the overall variance from the average spectrum, are ignored due
to their low statistical weight. A detailed analysis of a single band enables these subtleties to
be recognized.
It is seen that PCA may answer the rst question posed in Section 3.5 (i.e. how many
spectral components are changing during the processing reaction?) as well as, in part,
question (iii) regarding the relation to the chemorheological changes on cure. However, the
technique does not address the remaining, more quantitative, questions since there has been
no calibration of the spectral data. This would require regression of the data against a
calibration set that has been constructed under similar conditions.
3.5.2
Multivariate calibration
The data generated from a NIR or Raman spectrum do not immediately provide the concentrations of the species at any time, so there is no predictive capability. Construction of a
calibration set requires an independent measure of the property, e.g. by HPLC or by NIR of
known mixtures of the components. Two such methods are principal-component regression
(PCR) and partial least squares (PLS). As soon as quantitative analysis is considered, the
question of noise and reproducibility of the data set becomes important. It is therefore
necessary to treat the raw data to remove the drift in baseline etc. which will occur over a
long period of spectral acquisition.
Absorbance
(a)
1200
1400
1600
1800
2000
2200
Wavelength (nm)
(b)
Absorbance
276
1200
1400
1600
1800
Wavelength (nm)
2000
2200
Figure 3.48. (a) The NIR spectra of 12 kinds of EVA copolymers before Multiplicative Scatter
Correction (MSC) treatment (Shimoyama et al., 1998). (b) The effect of MSC on the spectra shown
in (a). Copyright 1998 John Wiley and Sons, Inc.; reproduced with permission.
277
It is apparent that there has been signicant offset of the baseline between samples and
hence that the data are unsuitable for quantitative analysis in the raw form. Three different
methods of correction of these data have been compared (Shimoyama et al., 1998):
multiplicative scatter correction (MSC)
taking the rst derivative
taking the second derivative
The rst approach, MSC, reduces the differences due to light scattering by using the
assumption that the wavelength dependence of light scattering is different from that of the
NIR absorption (Fischer et al., 1997). An average spectrum of the set is used, and all spectra
are then regressed against this average spectrum and the slope and intercept determined.
After ratioing the slope to the ideal spectrum, an offset value is generated, which is then
subtracted from the original spectrum to give the MSC-corrected spectrum. This processing
retains the chemical information in the spectra while removing the non-chemical differences
such as scattering.
The effect of MSC on the spectra of EVA copolymers in Figure 3.48(a) is shown in
Figure 3.48(b) (Shimoyama et al., 1998). The changes occurring in the region of 1200 nm
and, more particularly, between 2100 and 2200 nm, are now apparent.
However, there are many instances where regression against a physical property such as
viscosity is required and it is difcult to make assumptions about the nature of the correction. Derivative spectra have the advantage that they are sensing the change in slope of
the spectrum so that, at the baseline, the derivative is zero and thus the offset in the intensity
or absorbance scale due to baseline drift is eliminated. There is also an enhancement of
shoulders in the spectrum an obvious advantage when dealing with heavily overlapped
spectra as in the NIR. However, as in all pre-processing there is an increase in noise in the
system. Figure 3.49 shows the original and rst-derivative spectra from another study of
EVA copolymers (Hansen and Vedula, 1998), and the systematic change in the derivative
spectra with change in composition may readily be seen.
The second derivative enables the location of the peaks in the spectra and thus assists in
assignment, since it will be negative if the original spectrum has a local maximum, positive
if it has a local minimum and a point of inection when it is zero. Again the baseline drift is
eliminated (Shimoyama et al., 1998) and either the rst or the second derivative may be
used in subsequent data analysis by one of the regression methods. In a comparison of the
three correction methods specied above, it was found that MSC was best for the discrimination of the EVA copolymers due to their chemical similarity (Shimoyama et al.,
1998), while in a comparison of rheological properties (Vedula and Hansen, 1998) the rstderivative spectra were satisfactory.
In spectral data there will be noise from both multiplicative scatter (MS) and additive
scatter (AS). Normalization works well to limit the effects of MS and the derivative spectra
remove AS, but only the MSC method can minimize the effects of both (Shimoyama et al.,
1997).
(a)
0.7
0.6
Increasing
VA Content
Absorbance
0.5
0.4
0.3
0.2
0.1
0
0.1
1640
1680
1720
1760
1800
1840
Wavelength (nm)
(b)
Absorbance First Derivative
278
0.04
0.02
Increasing
VA Content
0.02
Increasing
VA Content
0.04
1640
1680
1720
1760
Wavelength (nm)
1800
1840
Figure 3.49. Overlaid NIR absorbance spectra of EVA copolymers and the rst-derivative plots
(lower traces) that result in better spectral resolution and remove baseline offsets (Hansen and
Vedula, 1998). Copyright 1998 John Wiley and Sons, Inc.; reproduced with permission.
robust data set against which unknown samples may be quantitated. An example of this is
the extension of the PCA analysis shown earlier in Figures 3.45 and 3.46 to a precise
determination of the percentage cure of an epoxy resin as a function of time by calibrating
the Raman data against DSC data. The available data for both Raman Spectroscopy and
DSC were divided into two groups to give calibration and validation sets. Attention was
focussed on high extents of cure, and Figure 3.50 shows the relationship between the
percentage cure and PC1 score from 93% to 99% cure at two different concentrations of
hardener. This study also highlighted the problems of operating with bre-optics when a
stress develops in the bre. The multivariate analysis immediately shows this as an increase
in standard error of prediction (SEP), indicating that spectral artefacts have been introduced,
rendering the calibration invalid (Stellman et al., 1995).
279
Figure 3.50. The relationship between PC1 score and percentage cure as determined by DSC for the
epoxy-resin/hardener system DER 332/Jeffamine T-403 in the following mixing ratios: *, 100/45;
and X, 100/55. Aust et al. (1997). Copyright 1997 by Society for Applied Spectroscopy; reproduced
with permission of Society for Applied Spectroscopy.
280
1000
300
100
Predicted 30
MI
(g/10 min) 10
3
1
0.3
0.1
0.1
0.3
Calibration
10
30
3
Measured MI (g/10 min)
100
300
1000
Figure 3.51. Use of NIR to predict the melt index (MI) of a series of EVA copolymers with differing
VA contents. A three-factor model was used to predict ln(MI) using the PRESS and the F-statistic
criterion (Hansen and Vedula, 1998). Copyright 1998 John Wiley and Sons, Inc.; reproduced with
permission.
which one sample was left out of the set and then used to predict the value from the
calibration based on the remaining samples. In addition to PRESS, the standard errors of
calibration (SEC) and of prediction (SEP) were calculated. This leave-a-sample (LAS)
method was then used for each of the samples in the set, in turn. Figure 3.51 shows the plot
which links the NIR predictions for MI using the three principal components and the PRESS
criterion.
The values of SEC and SEP are 0.14 and 0.46, respectively. The validity of a calibration
set when working with materials that may be variable in properties must be continually
questioned and, if the results deviate by greater than SEP the calibration might no longer
be valid. In a study of the in-line monitoring of the composition of PE/PP blends by NIR
spectra, the effect of outliers on the PLS analysis was seen by noting prediction results that
consistently exceeded the measured data (Rohe et al., 1998). In a renement of the data
prior to PLS analysis it was found that spectral averaging, smoothing and data reduction
(Rohe et al., 1999) allowed the SEP to be reduced to 0.25%, in comparison with 2% for the
untreated data. This reiterates the importance of pre-treatment of data before multivariate
regression.
3.5.3
281
(a)
0.5
Absorbance
original
deconvoluted
6200
6000
5800
Wavenumber (cm1)
5600
5400
Figure 3.52. The NIR spectrum of MY721 27% DDS (a) before and (b) after Fourier self-
deconvolution (K 2). This shows the complete resolution of the epoxy band at 6057 cm1 from
the other CH overtone bands. Reproduced with permission from St John (1993).
to the overlap in the NIR, a good estimate of the number of components may be made. As
for other multivariate techniques, pre-treatment of the data is required, and in this application MSC (Section 3.5.2) was effective at generating spectral data that, when analysed
over narrow wavelength regions, gave meaningful reaction proles.
It is important to note that among the techniques for pre-treatment of data there are
information-recovery processes including Fourier self-deconvolution (FSD). This is particularly relevant in FT-IR since the spectra as obtained have been apodized. This involves
the removal of an artefact appearing in the Fourier transformation due to the nite travel of
an interferometer so that the limits of the transformation are not innite as required
mathematically. The effect on the spectrum is to produce side-lobe artefacts, which, in order
to aid spectral interpretation, are removed by multiplication by an apodization function.
This improvement in cosmetic quality of the spectrum comes at the price of band broadening, so the purpose of FSD is to reduce the spectral bandwidth by removing the apodization and further narrowing the spectrum by multiplying it by the appropriate function.
Since this will introduce noise into the system, the spectra must have a high SNR before
FSD is performed. The resulting improvement in resolution often assists peak assignment,
but quantication must now involve band areas instead of peak heights (Koenig, 1999).
Figure 3.52 shows the effect of Fourier self-deconvolution (FSD) on NIR spectra from an
epoxy-resin system (St John, 1993), providing the resolution of the epoxy band and so
enabling quantitation (through band area) of the consumption kinetics during reactive
processing. This would not have been possible before the application of FSD. A chemometric
analysis will also produce the same result for the cure prole and may be regarded as more
accurate, since the entire spectrum of the resin is involved, but, if absolute concentrations
are required (for example to determine rate coefcients), then a band-area analysis after
FSD is unambiguous.
282
(b)
Absorbance
0.5
0
6200
6100
6000
5900
5800
5700
5600
Wavenumber (cm1)
Figure 3.52. (cont.)
3.6
3.6.1
283
101
102
100
101
101
100
102
101
103
2
Log Time (min)
Log Decrement
102
Figure 3.53. Isothermal cure of an epoxy-resin using torsional braid analysis (TBA). Reprinted from
substrates have been used, which offer the advantages of the solid substrate, without the
experimental difculties of applying a thick, even coating. Prime et al., (1988) showed that
measurements obtained using the wire-mesh supports were signicantly more reproducible
than were those from tests using glass braids. Semi-quantitative measurements of G0 and G00
have also been developed for the wire-mesh tests (Dillman et al., 1987), which provide a more
in-depth analysis of viscoelasticity changes for the curing systems.
3.6.2
Mechanical properties
Mechanical analysis is a common approach for evaluating polymer properties. For reactive
systems we shall limit ourselves to the thermal mechanical analyser (TMA) and the
dynamic mechanical thermal analyser (DMTA).
The TMA essentially operates by measuring the dimensional change of, or the penetration into, a material at a specied time, temperature and (typically non-oscillating) load
stress. The important variables to measure for reactive systems are the glass-transition
temperature, softening temperature and heat-deection temperature, the linear coefcient of
thermal expansion and physical ageing effects on these variables. For example, Figure 3.54
shows the use of a knife-edge probe to determine the softening points both of the inner and
of the outer coatings of a thermoset-coated motor winding wire (Prime, 1997a).
Much more detailed procedures are described in Prime (1997a). Ramis et al., (2003) use
shrinkage measurements on a TMA to determine the gelation point of polyester-based
thermoset coatings. As shown in Figure 3.55, the gel point is dened as the point at which
the rate of change of shrinkage abruptly slows down.
Alternatively, Ramis and Salla (1997) dene gelation of a polyester resin via Figure 3.56
as the point at which a submerged TMA probe becomes embedded (stuck) in the resin.
Figure 3.54. Typical TMA curves for a motor-winding wire. Reprinted from Figure 42 (Prime, 1997).
Copyright (1997), with permission from Elsevier.
0
0.00
20
0.04
TMA gelation
0.08
40
0.12
dL/dt (m/s)
Shrinkage, L (m)
284
60
0.16
0.20
80
0
10
20
30
Time (min)
Figure 3.55. Gelation determined from shrinkage and shrinkage rate versus time data from TMA.
Reprinted from Figure 3 (Ramis et al. 2003). Copyright (2003), with permission from Elsevier.
Garcia and Pizzi (1998) presented a detailed paper on the use of TMA for monitoring the
gelation of phenolformaldehye resins. This paper noted that there were two peaks of the
modulus in TMA curves during cure; the rst peak is related to critical entanglements
during cure and the second peak is related to the crosslinking network. They suggest that the
initial rise of the second peak should be considered as the gel point in these resin systems
wherein entanglements are important. Cadenato et al., (1997) highlighted differences when
measuring gel and vitrication times via TMA and DMTA of an epoxyamine system.
285
Figure 3.56. Gelation determined by damping of oscillations from TMA. Reprinted from Figure 1
(Ramis and Salla 1997). Copyright (1997), with permission from Elsevier.
These differences were removed when one presented the data as a function of conversion,
instead of time, indicating that there are differences in curing between DMTA and TMA.
The DMTA measures the complex modulus, compliance and viscosity in various modes
of deformation (shear, exural, tensile, bending, compression), typically at a range of
applied oscillatory deformation frequencies. The glass-transition temperature, gelation and
vitrication, secondary transitions, creep, mechanical properties and stress relaxation may
all be determined using this technique. Also isothermal and non-isothermal temperature
proles may be used. The DMTA is relatively insensitive to events prior to gelation, but
becomes an essential tool for monitoring transitions and mechanical properties after gelation.
In dynamic mechanical analysis for thermosetting systems the principal use is to monitor
the glass-transition temperature (Tg). In a DMTA Tg is dened as the maximum in loss
modulus, loss compliance or loss tangent. For example, Figure 3.57 (Prime, 1997a) shows
the various transitions for a cured magnetic ink coating that had previously been cured
under air and N2.
Figure 3.58 (Labana, 1985) generically shows the effects of increased crosslinking on the
Tg of curing systems.
Gelation and vitrication may also be observed using a DMTA. Generally gelation is
described by a peak in loss modulus, whereas vitrication is described as a peak in loss
modulus or loss tangent. Figure 3.59 (Prime, 1997a) shows the progression of the storage
and loss moduli (under shear) of a curing DGEBADDS system. This work shows the
gelation and vitrication as peaks in the loss modulus.
The DMTA has also been used to monitor the changes in dynamic viscoelastic properties
during cure as highlighted by Dillman and Seferis (1989) for TGDDM/DDS epoxy-resinbased systems. Hofmann and Glasser (1990) developed a DMTA system for DGEBA/mPDA epoxies impregnated onto glass braids. They dene gelation and vitrication by
(a)
28
240
243 C
Frequency (Hz)
Air cure
20
160
16
120
12
80
N2 cure
8
4
0
0.0
Temperature (C)
200
24
40
0
Gel point
20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.0 220.0
Time (min)
(b) 50
700
40
N2 cure
500
20
400
10
300
, air
Damping (mV)
600
Air cure
30
Frequency (Hz)
286
200
10
100
20
120 80
40
40
80
120
160
200
240
280
320
Temperature (C)
Figure 3.57. Examples of (a) gelation and (b) thermal transitions of curing coatings as determined by
DMTA. Reprinted from Figure 46 (Prime, 1997). Copyright (1997), with permission from Elsevier.
maxima in the loss tangent and also monitor the progression of cure conversion by DSC.
They compile a TTT diagram from the DMTA and DSC data, Figure 3.60, clearly showing
gelation, vitrication, maximum-conversion-rate and end-of-conversion points at a range of
isothermal curing temperatures.
A fundamental property that determines the state of a reacting system is its extent of cure
or chemical conversion (a). Several papers have shown that there is a unique relationship
between the glass-transition temperature (Tg) and a that is independent of cure temperature
and thermal history. This may imply that molecular structures of materials cured with
different histories are the same or that the changes in molecular structure do not affect Tg.
There are generally accepted to be two approaches to modelling glass-transitionconversion
relationships, namely thermodynamic and viscoelastic approaches. These are summarized in
Table 3.8.
287
(a)
M%
Molar mass
Mw
Mn
Mz
M0
20
40
60
Reaction, %
80
100
Modulus
(b)
c
b
a
Tg1
Tg2
Tg3
Temperature
Figure 3.58. An example showing (a) the variation in molar mass with extent of reaction and (b) the
variation of modulus versus temperature for polymers with three different crosslink densities, as
determined by DMTA. Adapted from Figures 1 and 4 (Labana, 1985). Copyright (1985). Reprinted
with permission of John Wiley and Sons, Inc.
3.6.3
Dielectric properties
Measurements of dielectric properties of reactive materials constitute an attractive method
for determining physical properties during curing, especially because of their easy application on-line to processing equipment. Dielectric analysis involves the measurement of
electrical polarization and conduction properties of a sample subject to a time-varying
electric eld. Dielectric properties such as the permittivity, loss factor and conductivity may
be measured and related to the glass transition, secondary transitions, rheological phenomena, vitrication, segmental mobility and dipolar relaxations.
8.6
1.5
9.0
8.0
7.5
0.5
8.4
8.2
Log [G (Pa)]
1.0
VITRIFICATION
8.5
Log [tan d]
Log [G (Pa)]
288
0.0
8.0
GELATION
7.0
0.5
6.5
7.8
0
50
150
100
Time (Minutes)
200
250
Figure 3.59. Determination by DMTA of changes in storage and loss moduli during cure. Reprinted
from Figure 59 (Prime, 1997). Copyright (1997), with permission from Elsevier.
Figure 3.60. A DMTA-derived TTT diagram. Reprinted from Figure 6 (Hofmann and Glasser, 1990)
289
Equation
Features
Thermodynamic
Hale (Hale et al.,
1991)
1
1
ka
Tgu Tg0
1987)
Tgu
1 1 C1 =C0 X C is the segment mobility. This gives a good
t for many systems.
Modied
Dibennedetto,
(Adabbo and
Williams, 1982)
Couchman
(Couchman, 1978)
Viscoelastic
Gan (Gan et al.,
1991)
Tg0
1 1 C1 =C0 a
Tg Tg0
ka
Tg1 Tg0 1 1 ka
1 1 Tg0 ka
Tg
1 k2 X =1 lX 2
ln Tg 1
aln Tg0
DCp1 DCp0 a ln Tg1
1 a DCp1 DCp0 a
Er
Tg
R lnC1 1 au C2
290
Figure 3.61. The voltage response for dielectric materials. Reprinted from (Prime, 1997). Copyright
In dielectric analysis a sample is placed between two electrodes and a sinusoidal voltage
(at specic frequencies of interest) is applied to one electrode. This applied voltage
establishes an electric eld in the samples. In response to this the sample can become
electrically polarized (dielectric polarization) and may also conduct a net charge from one
electrode to the other (ionic conduction). Both the dielectric polarization and ionic conduction give rise to a current, the amplitude of which is dependent on the frequency of the
applied voltage, the temperature and the structural properties of the sample. Figure 3.61
(Prime, 1997a) shows input voltage and output current waveforms.
Note that the amplitude and phase shift are recorded and may be related to the dielectric
properties of the medium (as long as plate geometries are also included in the assessment).
The dielectric permittivity of a medium (relative to the permittivity of free space, e0
8.85 1012 F/m) is given by e0 and measures the polarization of the medium per unit applied
electric eld. The dielectric loss factor arises from energy loss during time-dependent
polarization and bulk conduction. The loss factor is written as e00 . The loss tangent or dissipation of the medium, tan d is dened by e00 /e0 . The orientation of molecular dipoles has a
characteristic time sd. Typically sd is short early in the cure but grows large at the end of the
cure.
The simplest model of dipole orientation is that given by Debye (1912), in which one
assumes a single relaxation time for all species. This leads to the following expressions for
e0 and e00
.h
i
e0 eu er eu 1 xsd 2 ;
3:44
e00 r=xe0 er eu xsd
1 xs2d ;
3:45
where eu is the unrelaxed permittivity (or baseline permittivity) and er is the relaxed permittivity equal to the bulk permittivity when molecular dipoles align with the electric eld
to the maximum extent.
Reiterating, there are generally two effects considered in curing systems, namely conduction of ionic resin impurities and molecular-dipole orientation. The ionic conductivity
r Ri qiNili, where li is the ion mobility, Ni is the number of ions per unit volume and qi is
the charge, for species i. Qualitative relationships between conductivity and properties of
291
Figure 3.62. Peaks in permittivity and loss factor denoting Tg for a dielectric test. Reprinted from
Figure 73 (Prime, 1997). Copyright (1997), with permission from Elsevier.
the resin can be established with reference to simplications. For example, Stokes law
states that li qi/(6pgri) where ri is the sphere radius and g is the viscosity of the medium.
In this simple model the mobility and hence conductivity varies with 1/g.
Both the dipole-relaxation time and the ionic conductivity are related to the glass-transition temperature Tg. As a material is heated through its glass-transition temperature, static
dipoles gain mobility and start to oscillate in an electric eld. This causes an increase in
permittivity and a loss-factor peak is noted. Obviously this motion is affected by frequency
(lower frequencies have greater effects). This effect is shown in Figure 3.62 (Prime, 1997a),
which shows the peaks in permittivity and loss factor at Tg.
Note the frequency dependence of the peaks. The low-frequency peaks are known to
correspond well to other thermal measures of Tg.
There are differing opinions on whether gelation can be measured by dielectric analysis.
Ionic conductivity is related to viscosity, in that conductivity decreases with increasing
292
viscosity (increasingly restricted mobility). Some authors attribute a rapid decrease (Acitelli
et al., 1971) or an inection (Boiteux et al., 1993) in conductivity with gelation. However,
Senturia and Sheppard (1986) argue that auto-catalysis and the proximity to vitrication can
also contribute to large decreases in conductivity.
There is also a good relationship between dielectric properties and chemical kinetics. The
relaxed permittivity is related to chemical composition because it is a strong function of the
concentration of polar molecules. Also dipolar mobilities and ionic mobilities depend on the
extent of reaction and the changing value of Tg during cure. The relaxed permittivity
(maximum dipolar alignment at a given temperature and chemical composition) can be
qualitatively related to chemical changes, but is generally not a useful quantitative measure
due to the occurrence of complex inter-relationships.
The average dipole mobility at a given temperature and conversion (as given by the
frequency of the maximum in loss factor, or 1/frequency sd) and the distribution of
relaxation times (as evaluated from measurements of the frequency dependences of the
permittivity and loss factor) are also important in determining transition properties. Early in
cure, the dipole relaxation time correlates well with viscosity. The dipole loss peak has
also been shown to have a relationship to the vitrication peak. Viscosity and resistivity
(1/conductivity) are correlated before gelation but diverge at gelation. Attempts to correlate
conductivity, chemical conversion and Tg have been made. The sensitivity of conductivity
to Tg changes means that conductivity can be used as a good measure for Tg.
Bidstrup (Bidstrup, 1986, Bidstrup and Senturia, 1987, Bidstrup et al., 1989) highlights
microdilatometry (combined dielectric and rheological measurements on a small parallelplate sample) as a useful technique with potential for on-line and combined dielectric
rheological data generation. Lane (Lane et al., 1986), Kranbuehl (Kranbuehl et al., 1987,
1990), Gorto (Gorto and Yandrastis, 1987, 1989) and Boiteux (Boiteux et al., 1993) all
highlight in detail the benets of using combined dielectric and rheological methods to
determine the gelation transition and properties of epoxy-resin systems from microdilatometry. Kenny et al. (1991), Sanjana (1986) and Hsieh and Ho (1999) extend the
application of dielectrics to thermoset polymer composite materials.
3.6.4
Rheology
A brief introduction to rheology
Rheology is the study of ow and deformation of matter. It is used for many applications,
including
classication of materials,
characterization of ow properties,
troubleshooting processing problems,
modelling, controlling and optimizing processing and
understanding ow phenomena in complex uids.
From these uses it is evident that rheology is both an excellent diagnostic for ow
behaviour and extremely useful in assisting optimization and prediction of processing.
There are many excellent and complete texts on rheology (Macosko, 1994, Larson, 1999);
however, there are relatively few that describe experimental rheology in sufcient depth to
allow one to begin experiments. Thus it is the aim of the next section to provide a background to experimental rheology in order to understand chemorheology better.
293
Fy
W
Area=A
Fx
x
z
Fz
Figure 3.65 Ideal steady-shear ow.
Steady-shear rheology
In steady-shear rheology continuous steady shearing deformation is applied to the material,
and steady-shear properties are measured. For the simplest case of an innite parallel-plate
system, Figure 3.65, one can dene relevant properties as follows;
The shear rate (c0 ) is dened as the change in uid velocity (V) over the gap between the
innitely parallel plates (H) which is perpendicular to ow,
c 0 V =H:
3:46
The shear stress (r) is dened as the shear force induced by the uid on the plate in the
direction of ow (F) divided by the area normal to ow (A) upon which the force is acting,
r F=A:
3:47
h01
log Viscosity
294
h0
h03
3
h0 = K2(M )3.4
Figure 3.66 Steady-shear viscosity for a polymer system at three different molar masses.
The viscosity may then be dened as the stress divided by the shear rate,
g r=c 0 FH=AV :
3:48
The normal stresses (N1s or N2s) (which are induced in ow of viscoelastic uids) are
dened as the force normal (Fx or Fz) to ow divided by the area in that direction,
N1s Fx =A;
3:49a
N2s Fz =A:
3:49b
The normal-stress difference (NSD) may also be dened as the difference between the
two normal stresses perpendicular to ow,
NSD N1s N2s
3:50
More detailed analyses have been given for analysis of more complex geometries that are
used in practical rheometers and for extensions to three-dimensional ow for general
modelling applications, and the reader is recommended to refer to Macosko (1994) for those
applications.
Figures 3.66 and 3.67 show typical steady shear ow results for polymer systems.
Figure 3.66 shows the steady-shear viscosity for a polymer system at three molar masses.
Note the plateau in viscosity at low shear rates (or the zero-shear viscosity). Also note how
the zero-shear viscosity scales with Mw to the power 3.4. (This is predicted by Rouse theory
(Rouse, 1953).) Figure 3.67 shows the viscosity and rst normal-stress difference for a highdensity polyethylene at 200 C. Note the decrease in steady-shear viscosity with increasing
shear rate. This is termed shear-thinning behaviour and is typical of polymer-melt ow, in
which it is believed to be due to the polymer chain orientation and non-afne motion of
polymer chains. Note also that the normal-stress difference increases with shear rate. This is
also common for polymer melts, and is related to an increase in elasticity as the polymer
chain motion becomes more restricted normal to ow at higher shearing rates.
295
Figure 3.67. Viscosity and rst normal-stress difference versus shear rate for a HDPE melt. Adapted
from Figure 6.3.4 (Macosko, 1994). Copyright (1994). Reprinted with permission of John Wiley
and Sons, Inc.
Fy
V = dx /dt
V = dx /dt
Area = A
Fx
x
z
x = sin(vt)
Fz
Figure 3.68. Ideal dynamic-shear ow.
Dynamic-shear rheology
In dynamic-shear rheology an oscillating or dynamic shearing deformation is applied to the
material, and dynamic-shear properties are measured. For the simplest case of an innite
parallel-plate system, Figure 3.68, one can dene relevant properties as follows.
The shear rate (c0 ) is dened as the change in the uids oscillatory velocity (V ax
cos (xt)) over the gap between innite parallel plates that is normal to the ow,
c 0 v=H ax=H cosxt c0 x cosxt;
3:51
3:52
Now, for a solid material (entirely elastic), one would expect the force (F) or stress (r) to
be in phase with or follow the displacement (or strain, c). That is,
F Fm sinxt
3:53
r Fm =Asinxt r0 sinxt;
3:54
or
where r0 Fm/A.
296
For a Newtonian liquid (entirely viscous) one would expect the force or stress to follow
the rate of motion,
r gc0 gc0 x cosxt:
3:55
Also you can reach this conclusion if you expect the force to lag relative to the displacement by 90 ,
F Fm sinxt 90 Fm cosxt
3:56
r Fm =Acosxt r0 cosxt:
3:57
or
However, for a viscoelastic uid the stress must arise from a contribution of in-phase and
out-of-phase components,
r r0 A sinxt B cosxt;
3:58
where r0A sin(xt) represents the in-phase or solid-like stress response and (r0B cos(xt)
represents the out-of-phase or liquid-like stress response.
Rewriting this as
r c0 G0 sinxt G00 cosxt;
3:59
we dene G0 as a storage (or solid-like) modulus and G0 as the loss (or liquid-like) modulus.
One may also dene the loss tangent (tan d) by
tan d G00 =G0 ;
3:60
p 0 2
2
g
g g00 :
3:61
Interestingly, it has been shown experimentally for many systems (Cox, 1958) that the
prole of the complex viscosity (g*) as a function of the dynamic frequency (x) is
equivalent to the prole of the steady-shear viscosity (g) with respect to the shear rate (c 0 )
for the same system. That is,
g*x gc 0
3:62
for a range of polymer systems. Milner (1996) has begun to develop a theoretical rationale
behind this empirical rule.
Typical dynamic-shear proles are shown in Figures 3.69 and 3.70. Figure 3.69 shows a
typical set of proles of elastic modulus, loss modulus and dynamic viscosity versus applied
frequency for a polymer melt. This example (Rheometrics-Scientic, 1990) is a polyethylene melt at 160 C. Initially the loss modulus is higher than the storage modulus
because the melt behaves more like a viscous uid. This can generally be said to be due to
the fact that the imposed frequency is low enough that most of the molecules in the system
have enough time to relax after the perturbation, and the uid dissipates the stresses due to
the imposed shear. At higher frequencies the storage modulus grows larger than the loss
297
Figure 3.69. Typical proles of elastic modulus, loss modulus and dynamic viscosity versus applied
frequency for a polymer melt.
Figure 3.70. Dynamic proles for a range of polymer systems from (a) concentrated to (b) dilute
systems. Adapted from Figure 3.3.5 (Macosko, 1994). Copyright (1994). Reprinted with permission
of John Wiley and Sons, Inc.
modulus, and the system is said to behave more like a solid. This may be explained by the
fact that larger molecules become trapped by the higher frequencies of imposed shear and
do not have enough time to relax. The viscosity prole also starts from a plateau at low
frequencies where the behavior is pseudo-Newtonian (or uid-like) and then decreases with
increasing frequency (or becomes more pseudo-plastic or shear-thinning). Figure 3.70,
298
adapted from Macosko 1994, shows deviations from a concentrated polymer solution. Note
in Figure 3.70 how the prole of a concentrated polymer solution is similar to that for the
polymer-melt system described in Figure 3.69. Here note that initially at low frequencies the
storage modulus scales with frequency to the power 2 (since G0 x2) and the loss modulus
scales with frequency to the power 1 (since G00 x1). This can be predicted for Newtonian
systems, and further describes the initial low-frequency response as Newtonian or uid-like.
Next note that Figure 3.70 shows the response of a rubber system and note how the storage
modulus now dominates the loss modulus (G0 G00 ; since the system is more solid-like)
and the storage modulus is independent of frequency (G 0 x0; which is also typical of a
pure solid).
Timetemperature superposition
An interesting extension of dynamic rheology is timetemperature superposition (TTS)
(Ferry, 1980). Essentially TTS recognizes for some systems (known as thermorheologically
simple systems) the equivalence of the effects of time (or imposed frequency) and temperature throughout certain regions of behaviour. A corollary of this is to consider that, for a
thermorheologically simple uid, the response to a low frequency at a low temperature
would be equivalent to the response to a high frequency imposed at a high temperature. In
effect this allows one to determine the response of a system, at a standard (reference)
temperature, over a larger range of frequencies than could be determined experimentally.
Dependences of dynamic properties are given by Ferry (1980):
lngTr qr =gr Tq lnaT ;
3:63
where g is the viscosity, T is the temperature, q is the density, with subscript r denoting the
reference state, and aT is the shift factor.
One can obtain responses reduced to a specic reference temperature by use of the
following expressions:
Gr0 G0 Tr qr =Tq xr xaT ;
3:64
Gr00 G00 Tr qr Tqxr xaT ;
3:65
3:66
xr xaT :
Figure 3.71 shows how dynamic proles for the loss modulus versus frequency may be
reduced to a single reference temperature.
The overlapping of curves from a range of temperatures to a single curve at the reference
temperature denes its thermorheological nature.
Transient shear
Transient shear is dened as when a material is subject to an instantaneous change in
deformation and the response as a function of time is measured. For example, Figure 3.72
shows an instantaneously applied step-strain test.
The imposed strain is denoted in Figure 3.72(a) and the measured stress relaxation is
shown in Figure 3.72(b) (where G, the relaxation modulus, is dened as r(t)/c0; where r(t)
is the stress at time t after the step strain and c0 is the magnitude of the step strain.)
299
10
105
106
105
104
155 C
125 C
135 C
145 C
165 C
175 C
103
102
101
103
125 C
135 C
145 C
155 C (ref)
165 C
175 C
104
103
102
104 103
103 104
105
Figure 3.71. An example of timetemperature superposition (TTS) of rheological data. Adapted from
Figure 11.1 (Ferry, 1980). Copyright (1980). Reprinted with permission of John Wiley and
Sons, Inc.
G(t )
t
Figure 3.72. A transient rheological test showing (a) instantaneously applied strain and (b) the
associated stress response.
t
-(t)
+(t)
Strain
Rate g
Figure 3.73. A transient rheological test showing (a) the start up of an imposed strain rate and the
subsequent measurement of the stress build-up, and (b) the cessation of an imposed shear rate and
the subsequent stress decay.
g0
0
0
High g0
Stress t
300
Lowg0
0
Time
Figure 3.74. An example of a typical stress build-up during start up of shearing for a polymer melt.
Note that, in the higher-shear-rate test, the polymer shows a large stress overshoot. This
overshoot is common for viscoelastic uids and is believed to be related to hindrance to
chain disentanglement and conformational alterations.
Elongation rheology
Elongation rheology is dened as when the material is subject to stretching (elongational)
deformation and elongation-rheology properties are measured. Figure 3.75 shows ideal
elongational ow.
The elongation rate (ce0 ) may be dened as the rate of deformation per unit length (dv/dx)
in the direction of ow,
301
F1
F2
V1
V2
A2
x
A1
z
t=t1
t=t2
c0e dv=dx:
3:67
The normal stress (rn) may be dened as the force (F(t)) divided by the area (A(t)) in the
direction of ow,
rn F tAt:
3:68
3:70
Typically this value ranges from 3.0 for Newtonian systems to up to 1000 for nonNewtonian systems.
There are many types of elongation ow, which are shown in Figure 3.76. Simple or
uniaxial extension, equibiaxial extension and planar extension will provide different elongational properties specic to their type of deformation. The type of elongation ow
selected will depend on the type of elongation ow that is encountered during the specic
polymer process to be investigated. An example of the effect of the type of elongation ow
on the properties of a polymer solution is shown in Figure 3.77.
Here uniaxial and biaxial (compression) elongation viscosity of a 1% polyacrylamide in a
glycerinewater system is shown. Note that the material shows a reasonably constant
resistance to biaxial (compression) elongation, but quite an increase in uniaxial extension
with increasing extension rate. This has implications for processing of this uid using
predominantly uniaxial or biaxial ows.
Rheological classifications
Simple classications of uids can be made on the basis of their rheological proles.
Figure 3.78 shows the (a) shear stress and (b) viscosity proles for various systems. From
Figure 3.78 one may dene the following systems. Newtonian systems have a constant
viscosity with respect to shear rate. Dilatant (or shear-thickening) systems have a viscosity
that increases with respect to shear rate. Pseudo-plastic (or shear-thinning) systems have a
viscosity that decreases with respect to shear rate. Yield-stress materials are materials that
have an initial structure that requires a nite stress before deformation can occur. The stress
that initiates deformation is dened as the yield stress.
302
Simple/Uniaxial Extension
Equibiaxial Extension
Planar Extension
Figure 3.79 shows the viscosity proles for time-dependent uids. From Figure 3.79 one
may dene the following systems. Time-independent uids obviously undergo no change in
viscosity with respect to time. Rheopectic uids show an increase in viscosity with respect
to time. Thixotropic uids show a decrease in viscosity with respect to time.
Table 3.9 shows many various rheological models used to categorize and model uid
systems. It is written in terms of the three-dimensional form (where terms are discussed in
Macosko (1994)) and the simple two-dimensional shear-ow relationship (where terms
have been dened here).
Typical uid viscosities for various materials and typical shear-rate ranges for their
processing operations are shown in Tables 3.10 and 3.11 (Macosko, 1994).
Rheometers
Many commercial rheometers are available both for shear and for elongation rheology. For
an intensive review the reader is referred to Macosko (1994). In summary, Figure 3.80
303
100
Viscosity (poise)
Extension
10
Compression
Shear
0.1
1
10
100
1000
1)
Strain rate (s
Figure 3.77. An example of elongational and shear viscosities for a polyacrylamide solution. From
Figure 7.7.9 (Macosko, 1994). Copyright (1994). Reprinted with permission of John Wiley
and Sons, Inc.
Newtonian
Yield-stress
fluid
Dilatant
Pseudoplastic
Shear
stress
Shear Rate
Dilatant
Yield-stress
fluid
Viscosity
Newtonian
Pseudo-plastic
Shear Rate
Figure 3.78. Classications of uids on the basis of dependences of shear stress and shear viscosity
on shear rate.
304
Form
In simple shear
Viscous models
Newtonian
s g(2D)
s f (2D)
Power law
s m 2D(n1)/2(2D)
s12 gc 0 12
g constant
s12 gc 0 12
g g(c 0 )
n
s12 mc 0 12
0 n1
g mc 12
(g g_)/(g g0) 1/[1 (Kc 0 )(1 n)/2]
g/(g g0) 1/[1 (Kc 0 )(1 n)/2]
Cross
Ellis
Plastic models
Bingham
s Gc or c 0 for s < sy
s gc 0 sy for s > sy
P
i
Gi k i g 0
g g0 and u1 2g0k
R
s12 c M(t)h(c)dt
(Various models based on different
expressions for fc and fd)
g gs n0kBTk and u1 2n0kBTk2
Gt 13 n0 kB T expt=k
Gt 13 n0 kB T
expt=ki
Pom-pom model
Gt Gc 1 a
1 xa expt=s xdx
Rheopectic
Viscosity
Newtonian
Thixotropic
Shear Rate
Figure 3.79. Classication of uids on the basis of the time dependence of shear viscosity.
305
Glass
Molten glass (500 C)
Asphalt
Molten polymers
Heavy syrup
Honey
Glycerine
Olive oil
Light oil
Water
Air
1040
1012
108
103
102
101
100
101
102
103
105
Process
Sedimentation of ne powders
in a suspending liquid
Levelling due to surface tension
Draining under gravity
Screw extruders
Chewing and swallowing
Dip coating
Mixing and stirring
Pipe ow
Spraying and brushing
Rubbing
Injection-moulding gate
Milling pigments in uid bases
Blade coating
Lubrication
Typical range of
shear rates (s1)
Application
106104
Medicines, paints
102101
101101
100102
101102
101102
101103
100103
103104
104105
104105
103105
105106
103107
(shear) and Figure 3.81(elongation) give a good review of various commercial rheometers.
Figure 3.82 shows the appropriate shear- and elongation-rate ranges for these rheometers.
The advantages of various types of shear and elongation rheometers are discussed in
Table 3.12 (shear) and Table 3.13 (elongation).
3.6.5
Other techniques
SAXS, XRD
X-ray diffraction tests are based on the fact that incident X-ray radiation is diffracted at
certain angles by regularly spaced inclusions. The two main techniques used in reactive
polymer systems are wide-angle X-ray diffraction (WAXD) and small-angle x-ray
306
Drag Flows:
Chapter 5
(section number)
Coordinates
x1
x2
x3
z
x
r
Pressure Flows:
Chapter 6
(section number)
Capillary (6.2)
(Poiseuille flow)
y
x
x
r
Figure 3.80. Shear-ow rheometer geometries. (Adapted from Figure 5.1.2 (Macosko, 1994)).
Copyright (1994). Reprinted with permission of John Wiley and Sons, Inc.
scattering (SAXS); WAXD (using angles from 5 to 120 ) is able to monitor scales of order
from about 1 to 50A, whereas SAXS (using angles from 1 to 5 ) is able to measure
structures of the order of 70500A.
A large amount of literature recently has focussed on the use of XRD to examine
structural order in epoxy-resin nanocomposite systems. Lan et al. (1996) examined the
initial separation (intercalation) of clay layers in self-polymerization of Epon 828 epoxy
resin with onium-ion-exchanged clays. Much XRD work on epoxy-resin and thermoset
nanocomposites has continued since this work, up until and including recent work by
Becker et al. (2003a, 2003b) examining the clay structure of a range of epoxy-resin systems
with a range of different functionalities and curing conditions.
307
Geometry
(section number)
Co-ordinates
x1
x2
x3
x
x
r
y
r
x
Extension (7.2)
or
Compression (7.3)
x
r
xz
Figure 3.81. Extensional-ow rheometer geometries. Adapted from Figure 7.1.2 (Macosko, 1994).
Copyright (1994). Reprinted with permission of John Wiley and Sons, Inc.
Much work has been reported on studying the structure of thermoset resins via SAXS,
especially focussing on interpenetrating network polymers (IPNs), thermoset nanocomposites, rubber-modied thermosets and thermosetthermoplastic blends. Most recently Guo
et al., (2003) have examined the use of SAXS to monitor the nanostructure and crystalline
phase structure of epoxypoly(ethyleneethylene oxide) thermosetthermoplastic blends.
This work proposes novel controlled crystallization due to nanoscale connements.
SANS
Small-angle neutron scattering (SANS) is also used for characterizing structure in reactive
systems. With SANS (using angles <2 ) one can examine polymer morphology and
chain dimensions, typically enhanced by deuteratation (which enhances differences in
308
MFI
CAPILLARY
RDA DYNAMIC
RMS STEADY
RSR CREEP
TIME TEMPERATURE SUPERPOSITION
SHEAR RATE (SEC1 RAD/S)
108
106
104
102
102
104
106
108
108
106
104
102
102
104
106
108
YEARS
DAYS
MOLECULAR STRUCTURE
MICROSEC
IMPACT
PERFORMANCE
INJECTION MOULDING
SURFACE ROUGHNESS
BLOW MOULDING
EXTRUSION
THERMOFORMING
DIE SWELL
ROTATIONAL MOULDING
LONG-TERM CREEP
PHYSICAL AGEING
FATIGUE
Figure 3.82. Appropriate shear and elongation-rate ranges for rheometers.
neutron-scattering length versus undeuterated chains). Once again the most common systems
investigated in recent literature are thermosetthermoplastic blends (due to the ability to
control structure during cure of these systems). Elliniadis et al. (1997) examined the structure
of a thermosetthermoplastic blend of epoxypolyarylsulfone. They were able to use SANS
to determine an effective FloryHuggins interaction parameter and incorporated this into a
recent model for linearbranched polymer blends to predict the evolution of phase diagrams
under isothermal cure. Carter et al. (2003) examined the phase structure during cure of
low-temperature-cured epoxythermoplastic blends for aerospace composites using SANS.
PALS
Positron-annihilation spectroscopy (PALS) provides a measure of the free volume of
polymers via the lifetime of orthopositroniums (oPs). The typical lifetime of 13 ns corresponds to free-volume cavities of 0.30.7 nm. For reactive polymers, PALS can be linked
to Tg, gas diffusion rates, crosslink density, network formation and ageing and post-cure
effects. There have been some very instructive reviews (Hill, 1995, Simon, 1997) on PALS
that also incorporate examples involving reactive polymers. Suziki (2001) used PALS to
monitor Tg (in terms of the temperature dependence of the positronium (Ps) lifetime) and
gelation (in terms of an increase in the intensity of Ps and a sharp decrease in the lifetime) of
a curing epoxy cresol/phenol novolac system. Wang et al. (2003) also examined physical
ageing of epoxy-resin systems using PALS. They found that the side group plays an
important role in determining free-volume properties of epoxy cresol resins, i.e. the stronger
the crosslinking of curing agent, the smaller the free volume. However, during ageing the
309
Advantages
Disadvantages
Sliding plates
Simple design
Homogeneous
Linear motion
high g, G(t, y)
t 103 s
Falling ball
Very simple
Needle better
Sealed rheometer
High T, p
Concentric cylinders
(Couette ow)
Low g, high c0
Homogeneous if R/R 0.95
Good for suspension settling
End correction
High-g uids are difcult to clean
N1 impractical
Contained bobs
Sealed
(Brabender, Mooney) Process simulator
Indexers
Friction limits range
Capillary (Poiseuille
ow)
High c0
Sealed
Process simulation
gext from Dpent
Wide range with L
Slit ow
Axial annular ow
Squeeze ow
Simple
Process simulation
g(c0 ) at long times
free volume diminished as a result of the formation of the local order structure due to the
chain-segmental rearrangement.
GPC
The molecular weight or, more rigorously, molar mass (M) and molecular-weight distribution (MWD) of a reactive system (shown in Figure 1.6 and discussed in Section 1.1.3) can
be monitored via gel permeation chromatography (GPC). The sample is introduced to the
310
Advantages
Disadvantages
Tension
Homogeneous
Clean data
Wind-up is easiest to generate
Lubricated compression
Need lubricant
High g
gb < ge
e2
Low g
Process simulator
Sample preparation easy
Entrance condition
Corrections; g, FD
Photo
Bubble collapse
Stagnation
Entrance ows
Simplest
Wide g range
Process simulation
Complex ow
system in a solvent. As the material is passed through the GPC column, the material is
selectively retained on the basis of on its size (or M). From the time the material fractions
spend in the GPC column, a distribution of molecular weights can be obtained. For network
systems this method is limited as the M approaches innity, the MWD becomes innitely
broad and the material becomes insoluble at gelation. However, for reactively modied
polymers GPC is widely used to examine changes in M of modied polymers. Japon et al.
(2001) highlight the use of GPC in determining molecular structural changes in reactively
modied epoxidized-PET. Chen and Chiu (2001) examine the use of GPC to validate a
Monte Carlo model of the increase in M of the sol part of a curing epoxy-resin network.
SEM/TEM
Scanning electron microscopy (SEM) involves scanning an electron beam (510 nm) across
a surface and then detecting the scattered electrons. Literature abounds, with work focussing
on the use of SEM in the fracture and failure of epoxy resins and other thermoset polymers.
Also work on multiphase thermosets (thermosetthermoplastic blends, thermoset nanocomposites, interpenetrating network (IPN) polymers) is abundant.
AFM
Atomic-force microscopy (AFM) is a technique that examines the surface topography of
polymer samples. The probe moves across the surface and a piezo-electric detector can
311
monitor the topography in tapping or contact modes. Shaffer et al. (1995) examined the
interphase regions in rubber-modied epoxies, nding that AFM assessed interphase
dimensions more accurately than did SEM. Karger-Kocsis et al. (2003) examined the
morphology of an epoxyvinyl ester IPN using AFM, nding interesting two-phase structures, and could also monitor post-cure modications of this structure.
3.6.6
Spectroscopyrheology
A review of rheo-optical techniques by Sherman et al. (1996) notes that there has been an
increase in the use of rheo-optic set-ups both for FT-IR dichroism and for dynamic IR
dichroism spectroscopies for polymer melts and polymer blends. Skytt et al. (1996) highlight the use of simultaneous measurement of the transient or steady-state rheological
properties and IR dichroism to characterize orientation in polymer melts. However, there is
little reference to dual spectroscopicrheological techniques for reactive polymer systems in
the literature.
The authors have also previously proposed (George and Halley, 2000) a novel system in
which a NIR spectrometer is coupled through bre-optics to a Rheometrics ARES rheometer to provide a unique capability for thermosetting systems. The particular advantages
of working in the NIR region of the spectrum have been discussed in Section 3.3.6,
including the adaptation of bre-optics for real-time reaction monitoring (Sections 3.4.1 and
3.4.2). Two approaches can be employed for coupling the NIR bre-optic system to a
rheometer. The rst uses single-lament bres and micro-positioners above and below the
rheometer plates in order to transfer and collect the NIR radiation. The second uses a
dedicated NIR source above the plates and an optical bre for return of the radiation to the
spectrometer. The rst is expected to have a poorer SNR than the second, but has a betterdened pathlength, which is essential for quantitative analysis. Kinetic data acquisition
routines can be developed to synchronize rheological and spectral data. In principle the two
methods should provide the absolute viscosity at any point for comparison with the absolute
rate coefcient (since NIR measures concentration in absolute units). At the rst level this is
able to calibrate the bre-optic probe as a pseudo-rheometer for this system. In addition to
the above, the availability of rheological and conversion data as well as the absolute rate
parameters provides an opportunity to examine the theories for the onset of diffusion control
as the crosslinking network develops and so unambiguously connect the chemical and
rheological changes.
Chemoviscosity proles may also be correlated to chemical rate parameters. Simultaneous acquisition of the NIR spectrum and low-shear dynamic rheology may also be
undertaken from prior to gelation to beyond vitrication for various epoxy-resin systems for
selected isothermal proles to provide network information from rheological power-law
indices, fractal dimensions and network rheological structure parameters.
312
Recently Benali et al. (2004) reported the development of a new laboratory-made system
that allows the combination of rheometric (Alessi et al., 2005) and spectroscopic (Fouriertransform-near infrared (FT-NIR) spectroscopy) measurements for reactive polyurethane
materials along the lines of these principles.
Rheo-dielectrics
The development of simultaneous rheometry and dielectric spectroscopy is more mature
than is the rheo-optics of reactive systems. Initial work by Senturia et al. (1982, Senturia
and Sheppard, 1986) highlighted the use of microdielectrometry to monitor the complex
dielectric constant of reactive polymer materials during or after cure. Gorto and Yandrastis
(1989) extended this work by examining the simultaneous use of microdielectrometry and
rheometry to monitor the dielectric loss factor and the dynamic viscosity of thermosetting
bismaleimides and epoxies, by mounting a microdielectric sensor on the bottom plate of a
parallel-plate rheometer. Dipole peaks were correlated to vitrication phenomena, and the
maximum in the loss factor followed the same heating rate dependence as the minimum in
the viscosity. Further work on rheo-dielectrics for reactive thermosets was carried out by
Maffezzoli et al. (1994), Simpson and Bidstrup (1995) and Mijovic et al., (1995), who obtained
detailed correlations between dielectric-spectroscopic data and rheological properties.
References
Acitelli, M., Prime, R. & Sacher, E. (1971) Polymer, 12, 333343.
Adams, M. (1995) Chemometrics in Analytical Spectroscopy, Cambridge: Royal Society of
Chemistry.
Afromowitz, M. A. (1988) J. Lightwave Technol., 6, 15911594.
Agbenyega, J. K., Ellis, G., Hendra, P. J. et al. (1990) Spectrochim. Acta A, 46, 197216.
Aldridge, P., Kelly, J. J., Callis, J. B. & Burns, D. H. (1993) Anal. Chem., 65, 35813585.
Alessi, S., Calderaro, E., Fuochi, P. et al. (2005) Nucl. Instrum. Methods Phys. Res. B, 236, 5560.
Allen, N. S. & Owen, E. D. (1989a) Luminescence studies of the photooxidation of polymers, in
Zlatkevitch, L. (Ed.) Luminescence Techniques in Solid State Polymer Research, New York:
Marcel Dekker.
Allen, N. S. & Owen, E. D. (1989b) Polymer analysis by luminescence spectroscopy, in Zlatkevitch, L.
(Ed.) Luminescence Techniques in Solid State Polymer Research, New York: Marcel Dekker.
References
313
314
References
315
Durand, A., Hassi, L., Lachenal, G. et al. (2006) J. Near Infrared Spectrosc, 14, 161166.
Dusek, K. (1986) Network formation in curing of epoxy resins, in Dusek, K. (Ed.) Advances in
Polymer Science 78: Epoxy Resins and Composites III, Berlin: Springer-Verlag.
Dusek, K. (1996) Angew. Makromol. Chem., 240, 115.
Edge, M., Allen, N. S., Wiles, R., McDonald, W. & Mortlock, S. V. (1995) Polymer, 36, 227234.
Elliniadis, S., Higgins, J., Clarke, N. et al. (1997) Polymer, 38, 48554862.
Fanconi, B. M., Wang, F. W., Hunston, D. & Mopsik, F. (1986) Material characterization for
systems performance and reliability, In McCauley, J. W. & Weiss, V. (Eds.) Materials Characterization for Systems Performance and Reliability, New York: Plenum.
Farrington, P. J., Hill, D., ODonnell, J. H. & Pomery, P. J. (1990) Appl. Spectrosc., 44, 25432545.
Feng, L. & Ng, K. Y. S. (1990) Macromolecules, 23, 10481053.
Fernando, G. F. & Degamber, B. (2006) Int. Mater. Rev., 51, 65106.
Ferry, J. (1980) Viscoelasticity of Polymers, New York: Wiley.
Fidler, R. A., Rowe, D. & Weis, G. (1991) On-line infrared process control of molten polymers via
a high pressure, high temperature ow cell, in Proceedings of SPE ANTEC 91, pp. 850855.
Fischer, D., Bayer, T., Eichhorn, K.-J. & Otto, M. (1997) Fresenius J. Anal. Chem., 359, 7477.
Fischer, D., Sahre, K., Abdelrhim, M. et al. (2006) C. R. Chim., 9, 14191424.
Fontoura, J. M. R., Santos, A. F. et al. (2003) J. Appl. Polym. Sci., 90, 12731289.
Freeman, E. S. & Carroll, B. (1958) J. Phys. Chem., 62, 394397.
Fulton, M. I., Pomery, P. J., St. John, N. A. & George, G. A. (1998) Polym. Adv. Technol., 9, 7583.
Gallagher, P. K. (1997) Thermoanalytical instrumentation, techniques and methodology, in Turi,
E. A. (Ed.) Thermal Characterization of Polymeric Materials, San Diego, CA: Academic Press.
Garcia, R. & Pizzi, A. (1998) J. Appl. Polym. Sci., 70, 11111119.
Garton, A. (1992) Infrared Spectroscopy of Polymer Blends, Composites and Surfaces, Munich:
Hanser-Verlag.
George, G., Hynard, N., Cash, G., Rintoul, L. & OShea, M. (2006) C. R. Chim., 9, 14331443.
George, G. A. (1981) Use of chemiluminescence to study the kinetics of oxidation of solid polymers, in Grassie, N. (Ed.) Developments in Polymer Degradation 3. London: Applied Science.
George, G. A. (1986) Mater. Forum, 9, 224236.
George, G. A. (1989a) Chemiluminescence of polymers at nearly ambient conditions, in Zlatkevitch,
L. (Ed.) Luminescence Techniques in Solid State Polymer Research, New York: Marcel Dekker.
George, G. A. (1989b) Luminescence in the solid state: general requirements and mechanisms, in
Zlatkevitch, L. (Ed.) Luminescence Techniques in Solid State Polymer Research, New York,
Marcel Dekker.
George, G. A., Cash, G. A. & Rintoul, L. (1996) Polym. Int., 41, 169182.
George, G. A. & Celina, M. (2000) Homogeneous and heterogeneous oxidation of polypropylene, in
Halim Hamid, S. (Ed.) Handbook of Polymer Degradation, 2nd edn, New York: Marcel Dekker.
George, G. A., Celina, M., Vassallo, A. M. & Cole-Clarke, P. A. (1995) Polym. Deg. Stab, 48, 199210.
George, G. A., Cole-Clarke, P. A. & St John, N. A. (1990) Mater. Forum, 14, 203209.
George, G. A., Cole-Clarke, P. A., St John, N. A. & Friend, G. (1991) J. Appl. Polym. Sci., 42, 643657.
George, G. A. & Schweinsberg, D. P. (1987) J. Appl. Polym. Sci., 33, 22812292.
George, G. A. & St John, N. A. (1993) Chem. Australia, 60, 654.
Gmelin, E. (1997) Thermochim Acta, 304/5, 126.
Goddu, R. F. & Delker, D. A. (1960) Anal. Chem., 32, 140.
Gonzalez-Gonzalez, V. A., Neira-Velazquez, G. & Angulo-Sanchez, J. L. (1998) Polym. Deg. Stab.,
60, 3342.
Gorto, J. & Yandrastis, M. (1987) Proceedings of Society of Plastics Engineers ANTEC 87,
pp. 10391042.
316
References
317
Kranbuehl, D., Hood, D., Wang, Y. et al. (1990) Polym. Adv. Technol., 8, 9399.
Krohn, D. A. (1988) Fiber Optic Sensors Fundamentals and Applications, Research Triangle
Park, NC: Instrument Society of America.
Kupper, L., Heise, H. M. & Butvina, L. N. (2001) J. Molec. Struct., 5634, 173181.
Labana, S. (1985) Cross-linking, in Mark, H. (Ed.) Encyclopedia of Polymer Science and Engineering, 2nd edn, New York: Wiley.
Lachenal, G. (1995) Vib. Spectrosc., 9, 93100.
Lacoste, J., Carlsson, D. J., Falicki, S. & Wiles, D. M. (1991) Polym. Deg. Stab., 55, 34.
Lam, K.-Y. & Afromowitz, M. A. (1995a) Appl. Opt., 34, 56355638.
Lam, K.-Y. & Afromowitz, M. A. (1995b) Appl. Opt., 34, 56395643.
Lan, T., Kaviratna, D. & Pinnavaia, T. (1996) J. Phys. Chem. Solids, 57, 10051010.
Lane, J., Seferis, J. & Bachmann, M. (1986) Polym. Eng. Sci., 26, 346353.
Lange, J., Altmann, N., Kelly, C. T. & Halley, P. J. (2000) Polymer, 41, 49495955.
Lange, J., Ekelof, R. & George, G. A. (1998) Polymer, 40, 149155.
Lange, J., Ekelof, R. & George, G. A. (1999) Polymer, 40, 35953598.
Larson, R. (1999) The Structure and Properties of Complex Fluids, New York: Oxford University
Press.
Lau, W. & Westmoreland, D. G. (1992) Macromolecules, 25, 44484449.
Laupretre, F. J. (1990) Prog. Polym. Sci., 15, 425474.
Lavine, B. K. (2000) Chemometrics. Anal. Chem., 72, 91R97R.
Law, K. Y. & Loutfy, R. O. (1981) Macromolecules, 14, 587591.
Lee, S.-N., Chiu, M. & Lin, H.-S. (1992) Polym. Mater. Sci. Eng, 32, 10371046.
Lees, A. J. (1998a) Polym. Polym. Composites, 6, 121131.
Lees, A. J. (1998b) Coord. Chem. Rev., 177, 335.
Levy, R. L. (1984) Polym. Mater. Sci. Eng., 50, 124126.
Levy, R. L. (1986) Polym. Mater. Sci. Eng., 54, 321324.
Levy, R. L. & Schwab, S. D. (1987) Polym. Mater. Sci. Eng., 56, 169174.
Liu, H. & George, G. A. (2000) Polym. Int., 49, 15011512.
Louch, J. & Ingle, J. D. (1988) Anal. Chem., 60, 26372640.
Loutfy, R. O. (1981) Macromolecules, 14, 270275.
Loutfy, R. O. (1982) J. Polym. Sci. Polym. Phys. Edn, 20, 825835.
Loutfy, R. O. (1986) Pure Appl. Chem., 58, 12391248.
Machado, A. V., Covas, J. A. & Van Duin, M. (1999) J. Appl. Polym. Sci., 71, 135141.
Mackison, R., Brinkworth, S. J., Belchamber, R. M. et al. (1992) Appl. Spectrosc., 46, 10201024.
Macosko, C. (1994) Rheology: Principles, Measurements, and Applications, New York:, VCH.
Maffezzoli, A., Trivisano, A., Opalicki, M., Mijovic, J. & Kenny, J. M. (1994) J. Mater. Sci., 29,
8008.
Mailhot, B. & Gardette, J.-L. (1996) Vib. Spectrosc., 11, 6978.
Malinowski, E. R. & Howery, D. G. (1980) Factor Analysis in Chemistry, New York: John Wiley.
Mcgovern, S., Royce, B. S. H. & Benziger, J. (1985) J. Appl. Opt., 24, 15121514.
Melby, E. G. & Castro, J. M. (1989) Glass-reinforced thermosetting polyester molding: materials
and processing, in Allen, G. (Ed.) Comprehensive Polymer Science, Oxford: Pergamon.
Mijovic, J., Bellucci, F. & Nicolais, L. (1995) J. Electrochem. Soc., 142, 11761182.
Mijovic, J. & Wang, H. T. (1986) J. Appl. Polym. Sci., 37, 26612673.
Mikes, F., Baselga, J. & Paz-Abuin, S. (2002a). Eur. Polym. J., 38, 23932404.
Mikes, F., Gonzalez-Benito, F., Serrano, B., Bravo, J., Baselga, J. (2002b) Polymer, 43, 43314339.
Miller, C. E. (1991) Appl. Spectrosc. Rev., 26, 277339.
Milner, S. (1996) J. Rheology, 40, 303315.
Morgan, R. J. & Mones, E. T. (1987) J. Appl. Polym. Sci., 33, 9991020.
318
Mori, H., Kono, H., Terano, M., Nosov, A. & Zakharov, V. A. (1999) Macromol. Rapid Commun.,
20, 536540.
Myrick, M. L., Angel, S. M. & Desiderio, R. (1990) Appl. Opt., 29, 13331344.
Niemela, P. & Suhonen, J. (2001) Appl. Spectrosc., 55, 13371340.
Noll, W. (1968) Chemistry and Technology of Silicones, New York: Academic Press.
Donnell, J. H. & OSullivan, P. W. (1981) Polym. Bull., 5, 103110.
Oh, S. J., Lee, S. K. & Park, S. Y. (2006) Vib. Spectrosc., 42, 273277.
Ozpozan, T., Schrader, B. & Keller, S. (1997) Spectrochim. Acta A, 53, 17.
Paik, H.-J. & Sung, N.-H. (1994) Polym. Eng. Sci., 34, 10251032.
Palmese, G. & Gillham, J. (1987) J. Appl. Polym. Sci., 34, 19251939.
Parker, H.-Y., Westmoreland, D. G. & Chang, H.-R. (1996) Macromolecules, 29, 51195127.
Paul, S. (1989) Crosslinking: chemistry of surface coatings, in Allen, G. (Ed.) Comprehensive
Polymer Science, Oxford: Pergamon.
Peinado, C., Alonso, A., Salvador, E. F., Baselga, J. & Catalina, F. (2002) Polymer, 43, 53555361.
Pekcan, O. & Kaya, D. (2001) Polymer, 42, 78657871.
Pekcan, O., Kaya, D. & Erdogan, M. (2001) Polymer, 42, 645650.
Pekcan, O., Yilmaz, Y. & Okay, O. (1997) Polymer, 38, 16931698.
Pelikan, P., Ceppan, M. & Liska, M. (1994) Applications of Numerical Methods in Molecular
Spectroscopy, Boca Raton, FL: CRC Press.
Pell, R. J., Callis, J. B. & Kowalski, B. R. (1991) Appl. Spectrosc., 45, 808818.
Pelletier, M. J. (2003) Appl. Spectrosc., 57, 20A39A.
Perrier, S. & Haddleton, D. M. (2003) In situ NMR monitoring of living radical polymerization, in
Puskas, J., Long, T. E., Storey, R. (Eds.) In situ Spectroscopy of Monomer and Polymer Synthesis, New York: Kluwer Academic.
Poisson, N., Lachenal, G. & Sauterau, H. (1996) Vib. Spectrosc., 12, 237247.
Powell, G. R., Crosby, P. A., Fernando, G. F. et al. (1995) Proc. SPIE, 2444, 386395.
Powell, G. R., Crosby, P. A., Waters, D. N. et al. (1998) Smart Mater. Structures, 7, 55768.
Prime, R. (1997) Thermosets, in Turi, E. (Ed.) Thermal Characterization of Polymeric Materials,
2nd edn, New York: Academic Press.
Prime, R., Burns, J., Karmin, M., Moy, C. & Tu, H. (1988) J. Coatings Technol., 60, 5560.
Puskas, J. E., Lanzendorfer, M. G. & Pattern, W. E. (1998) Polym. Bull., 40, 5561.
Quinn, A. C., Gemperline, P. J., Baker, B., Zhu, M. & Walker, D. S. (1999) Chemom. Intell. Lab.
Syst., 45, 199214.
Quirin, J. C. & Torkelson, J. M. (2003) Polymer, 44, 423432.
Rakicioglu, Y., Schulman, J. M. & Schulman, S. G. (2001) Applications of chemiluminescence in
organic analysis, in Garcia-Campana, A. M. & Baeyens, W. R. G. (Eds.) Chemiluminescence in
Analytical Chemistry, New York: Marcel Dekker.
Ramis, X., Cadenato, A., Morancho, J. & Salla, J. (2003) Polymer, 44, 20672079.
Ramis, X. & Salla, J. (1997) J. Polym. Sci. B, Polym. Phys., 35, 371388.
Reis, M. M., Araujo, P. H. H., Sayer, C. & Giudici, R. (2004) Ind. Eng. Chem. Res., 43, 72437250.
Remillard, J. T., Jones, J. R., Poindexter, B. D., Helms, J. H. & Weber, W. H. (1998) Appl. Spectrosc., 52, 13691376.
Reshadat, R., Desa, S., Joseph, S. et al. (1999) Appl. Spectrosc., 53, 14121418.
Rettig, W. (1986) Angew. Chem. Int. Edn. Engl., 25, 971988.
Rey, L., Galy, J., Sauterau, H. et al. (2000) Appl. Spectrosc., 54, 3943.
Rheometrics-Scientic (1990) Understanding Rheological Testing Thermoplastics, Piscataway,
NJ: Rheometrics Inc.
Richardson, M. J. (1989) Thermal analysis, in Allen, G. (Ed.) Comprehensive Polymer Science,
Oxford: Pergamon.
References
319
Rohe, T., Becker, W., Kolle, S., Eisenreich, N. & Eyerer, P. (1999) Talanta, 50, 283290.
Rohe, T., Becker, W., Krey, A. et al. (1998) J. Near Infrared Spectrosc., 6, 325332.
Roper, T. M., Guymon, C.A., Hoyle, C.E. (2005) Rev. Sci. Instrum., 76, 054102-1054102-8.
Roper, T. M., Lee, T. Y., Guymon, C. A. & Hoyle, C. E. (2005) Macromolecules, 38, 10 10910 116.
Rouse, P. (1953) J. Chem. Phys., 21, 1272.
Rozenberg, B. A. (1986) Kinetics, thermodynamics and mechanism of reactions of epoxy oligomers
with amines, in Dusek, K. (Ed.) Advances in Polymer Science 75: Epoxy Resins and Composites:
II, Berlin: Springer-Verlag.
Sahre, K., Schulze, U., Hoffmann, T. et al. (2006) J. Appl. Polym. Sci., 101, 13741380.
Sanghera, J. S. & Aggarwal, I. D. (1999) J. Non-Cryst. Solids, 256257, 616.
Sanjana, Z. (1986) Polym. Eng. Sci. 26, 373379.
Santos, A. F., Silva, F.M., Lenzi, M.K., Pinto, J.C. (2005) Polym. Plast. Technol. Eng., 44, 161.
Sasic, S., Kita, Y., Furukawa, T. et al. (2000) Analyst, 125, 23152321.
Scarlata, S. F. & Ors, J. A. (1986) Polym. Commun., 27, 4142.
Scherzer, T. & Decker, U. (1999) Vib. Spectrosc., 19, 385398.
Schmidt, L. E., Leterrier, Y., Vesin, J.-M., Wilhelm, M. & Manson, J.-A. E. (2005) Macromol.
Mater. Eng., 290, 11151124.
Schweinsberg, D. P. & George, G. A. (1986) Corros. Sci., 26, 331340.
Scott, G. (1993) Oxidation and stabilization of polymers during processing, in Scott, G. (Ed.)
Atmospheric Oxidation and Anti-oxidants, Amsterdam: Elsevier.
Scott, T. F., Cook, W. D. & Forsythe, J. S. (2002) Polymer, 43, 58395845.
Senturia, S. & Sheppard, N. (1986) Adv. Polym. Sci., 80, 343.
Senturia, S., Sheppard, N., Lee, H. & Day, D. (1982) J. Adhesion, 15, 6990.
Sewell, G. J., Billingham, N. C., Kozielski, K. A. & George, G. A. (2000) Polymer, 41, 21132120.
Shaffer, O., Bagheri, R., Qian, J. et al. (1995) J. Appl. Polym. Sci., 58, 465484.
Sherman, B., Neaffer, R. & Galiatsatos, V. (1996) Trends Polym. Sci., 4, 7273.
Shim, M. G. & Wilson, B. C. (1997) J. Raman Spectrosc., 28, 131142.
Shimoyama, M., Hayano, M., Matsukawa, K. et al. (1998) J. Polym. Sci. B: Polym. Phys., 36, 1529
1537.
Shimoyama, M., Maeda, H., Matsukawa, K. et al. (1997) Vib. Spectrosc., 14, 253259.
Simon, G. (1997) Trends Polym. Sci., 5, 394400.
Simpson, J. & Bidstrup, S. (1995) J. Polym. Sci. B: Polym. Phys., 33, 5562.
Sircar, A. K. (1997) Elastomers, in Turi, E. A. (Ed.) Thermal Characterization of Polymeric
Materials, San Diego, CA: Academic Press.
Skytt, M.-L., Faernert, G., Jansson, J.-F. & Gedde, U. W. (1996) Polym. Eng. Sci., 36, 17371744.
Sohma, J. (1989a) Mechanochemical degradation, in Allen, G. (Ed.) Comprehensive Polymer
Science, Oxford: Pergamon.
Sohma, J. (1989b) Prog. Polym. Sci., 14, 451596.
Song, J. C. & Neckers, D. C. (1996) Polym. Eng. Sci., 36, 394402.
Sourour, S. & Kamal, M. R. (1976) Thermochim. Acta, 14, 4159.
Spragg, R. A. (1984) Perkin-Elmer Infrared Bull., 100, 7.
St John, N. A. (1993) Spectroscopic Studies of the Cure and Structure of an Epoxy Amine Network,
Thesis, Department of Chemistry, St Lucia, The University of Queensland.
St John, N. A. & George, G. A. (1992) Polymer, 33, 26792688.
St John, N. A. & George, G. A. (1994) Prog. Polym. Sci., 19, 755795.
St John, N. A., George, G. A., Cole-Clarke, P. A., Mackay, M. E. & Halley, P. J. (1993) High
Perform. Polym., 5, 212236.
Starnes, W. H. (1981) Mechanistic aspects of the degradation and stabilization of poly(vinyl chloride), in Grassie, N. (Ed.) Developments in Polymer Degradation 3, London: Applied Science.
320
Stellman, C. M., Aust, J. F. & Myrick, M. L. (1995) Appl. Spectrosc., 49, 392394.
Storey, R. F., Donnalley, A. B. & Maggio, T. L. (1998) Macromolecules, 31, 15231526.
Storey, R. F., Maggio, T. L. & Brister, L. B. (1999) ACS Polym. Preprints, 40, 964965.
Strehmel, B., Strehmel, V., Timpe, H.-J. & Urban, K. (1992) Eur. Polym. J., 28, 525533.
Strehmel, B., Strehmel, V. & Younes, M. (1999) J. Polym. Sci. B: Polym. Phys., 37, 13671386.
Strehmel, V. & Scherzer, T. (1994) Eur. Polym. J., 30, 361368.
Stutz, H. & Mertes, J. (1989) J. Appl. Polym. Sci., 38, 781787.
Sung, C. S. P., Chin, I.-J. & Yu, W. C. (1985) Macromolecules, 18, 15101512.
Sung, C. S. P., Pyun, E. & Sun, H.-L. (1986) Macromolecules, 19, 29222932.
Taramae, N., Hiroguchi, M. & Tanaka, S. (1982) Bull. Chem. Soc. Japan, 55, 20972100.
Tonelli, A. E. & Srinivasarao, M. (2001) Polymers from the Inside Out; An Introduction to
Macromolecules, New York: Wiley-Interscience.
Tredwell, C. J. & Osborne, A. D. (1980) J. Chem. Soc. Faraday II, 76, 16271637.
Tuchbreiter, A., Marquardt, J., Kappler, B., Honerkamp, J., Mulhaupt, R. (2004) Macromol. Symp.,
213, 327333.
Van Assche, G., Van Hemelrijck, A., Rahier, H. & Van Mele, B. (1997) Thermochim. Acta, 304/
305, 317334.
Vatanparast, R., Li, S., Hakala, K. & Lemmetyinen, H. (2001) J. Appl. Polym. Sci., 82, 26072615.
Vatanparast, R., Li, S., Hakala, K., Lemmetyinen, H. (2002) J. Appl. Polym. Sci., 83, 17731780.
Vatanparast, R., Li, S., Lemmetyinen, H. (2001) J. Appl. Polym. Sci., 82, 26072615.
Vedula, S. & Hansen, M. G. (1998) J. Appl. Polym. Sci., 68, 873889.
Vieira, R. A. M., Sayer, C., Lima, E. L. & Pinto, J. C. (2001) Polymer, 42, 89018906.
Vorobyova, O. & Winnik, M. A. (2001) J. Polym. Sci. B: Polym. Phys., 39, 23022316.
Wang, B., Gong, W., Liu, W. et al. (2003) Polymer, 44, 40474052.
Wang, F. W., Lowry, R. E. & Cavanagh, R. R. (1986a) Polymer, 26, 16571661.
Wang, F. W., Lowry, R. E. & Fanconi, B. M. (1985) Polym. Mater. Sci. Eng., 53, 180184.
Wang, F. W., Lowry, R. E. & Fanconi, B. M. (1986b) Polym. Preprints, 27, 306307.
Wetton, R. (1986) Dynamic mechanical thermal analysis of polymers and related materials, in
Dawkins, J. (Ed.) Developments in Polymer Characterization, London: Elsevier.
Wetton, R. & Duncan, J. (1993) Society of Plastic Engineers ANTEC Proceedings ANTEC 1993,
pp. 29912995.
Woerdeman, D. L. & Parnas, R. S. (2001) Appl. Spectrosc., 55, 331337.
Wroblewski, S., Trzebiatowska, K., Jedrzejewska, B. et al. (1999) J. Chem. Soc., Perkin Trans. 2,
19091917.
Wu, C., Danielson, J. D. S., Callis, J. B., Eaton, M. & Ricker, N. L. (1996) Process Control and
Quality, 8, 123.
Xanthos, M. (Ed.) (1992) Reactive Extrusion: Principles and Practice, Munich:, Hanser-Verlag.
Yamada, B., Azukizawa, M., Yamazoe, H., Hill, D. J. T. & Pomery, P. J. (2000) Polymer, 41, 5611
5618.
Younes, M., Wartewig, S., Lellinger, D., Strehmel, B. & Strehmel, V. (1994) Polymer, 35, 5269
5278.
Young, P. R., Druy, M. A., Stevenson, W. A. & Compton, D. A. C. (1989) SAMPE J., 25, 1116.
Yu, W. C. & Sung, C. S. P. (1990) Macromolecules, 23, 386390.
Yu, W. C. & Sung, C. S. P. (1995) Macromolecules, 28, 25062511.
Zhang, X., Garcia-Campana, A. M. & Baeyens, W. R. G. (2001) Application of chemiluminescence
in inorganic analysis, in Garcia-Campana, A. M. & Baeyens, W. R. G. (Eds.) Chemiluminescence
in Analytical Chemistry, New York: Marcel Dekker.
Zweifel, H. (1998) Stabilization of Polymeric Materials, Berlin: Springer-Verlag.
4.1
Introduction
This chapter highlights the importance of chemorheology in determining cure properties of
reactive systems. A brief introduction to experimental rheology has been provided in
Chapter 3 to provide a baseline knowledge of experimental rheology. In this chapter we
examine a description of chemorheology in terms of basic chemorheology, chemoviscosity,
gelation and vitrication transitions and ultimate properties. Finally, examples of chemorheological analysis will be discussed. (We will briey summarize chemorheological
data and models in this chapter, but only for reference to chemorheological testing. A more
extensive examination of chemorheology and modelling of systems will be presented in
Chapter 5.)
4.2
Chemorheology
The denition of chemorheology (in this text) is the study of the deformation properties of
reactive polymer systems. Figure 4.1 shows a schematic representation of the structural
development during thermoset cure.
Step (a) shows unreacted monomers, and cure proceeds to step (b), at which there is the
formation of some branched molecules. By step (c) the cure has progressed to the gel point,
such that an innite network is formed across the whole structure. Further cure can occur to
point (d), at which the material becomes fully cured and vitrication is reached.
The essential elements of a chemorheological study are
fundamental chemorheology
chemoviscosity proles
gelation
vitrication
ultimate chemorheological properties
modelling
We shall focus on modelling in Chapter 5. This chapter will focus on the other chemorheological properties and use examples from the literature to clarify key points.
4.2.1
Fundamental chemorheology
Many chemorheological studies omit what we have termed fundamental chemorheological
behaviour, by which we mean the basic tests that can assess the fundamental rheological
322
(a)
(b)
(c)
(d)
Figure 4.1. A schematic diagram of structural development during thermoset cure from (a) unreacted
monomers, (b) formation of initial links, (c) the gel point or a path of covalent links across the
sample and (d) the cured crosslinked system with some unreacted groups.
323
4.2 Chemorheology
105
v = 1 rad/s
v = 10 rad/s
104
v = 100 rad/s
103
0.1
10
Strain g (%)
Figure 4.2. Dynamic properties as functions of strain in a dynamic step-strain test to determine the
linear viscoelastic region for a highly lled epoxy-novolac moulding sample used for computerchip encapsulation.
response of materials is related to the work by Giacomin and Dealy (1993) that examined
the non-linear response of thermoplastic polymer melts to very large strains. This work
focusses on the ability to generate effective G0 and G00 data from non-linear responses.
It should be noted that the linear viscoelastic behaviour needs to be examined as a
function of temperature and cure level during the reaction in order to describe fully the
viscoelastic behaviour of the system.
Yield stress
There are various forms of yield-stress test that can assess the yield stress (rY) of a material,
and they are typically separated into indirect and direct methods. Indirect methods monitor
properties that may in turn be related to a yield stress, whereas direct methods measure the
yield stress directly. Various techniques are described in more detail in Gupta (2000). Here
we shall describe two methods that have been used successfully in our laboratory.
An example of an indirect method is the use of a steady-shear-rate sweep test, which
examines the steady-shear viscosity (g) and shear stress (r) as functions of shear rate (c0 ).
The results, once obtained, may be plotted in terms of steady-shear viscosity (g) versus
shear stress (r). The existence of an innite-viscosity asymptote at low levels of shear stress
is indicative of a yield stress, and the asymptotic stress is dened as the yield stress (rY). If
there is no asymptote, then the material is deemed to have no measurable yield stress. This
is shown diagrammatically in Figure 4.3.
An example of a direct method is the vane technique (Collyer and Clegg, 1988). Here a
vane rheometer is used (see Figure 4.4), in such a way that the vane attachment is used to
monitor the shear response of a rotating uid bath.
1.2 105
NOT
1.0 105
324
8.0 104
6.0 104
4.0 104
2.0 104
0.0 104
0
2000
4000
6000
8000
10000
stress for a highly lled epoxy-novolac moulding sample used for computer-chip encapsulation.
Dv
Figure 4.4. The vane-rheometer conguration used for direct yield-stress measurements. Reprinted
The advantage of the vane technique is that material is allowed in between the prongs of
the vane and in effect shears the material upon itself (rather than inducing shearing between
a mechanical xture and the material, as would be seen if a standard cup-and-bob or Couette
rheometer were used). This eliminates the potential for slippage between a mechanical
xture and the material. A typical response is given in Figure 4.5.
325
4.2 Chemorheology
Figure 4.5. Typical response from a Vane rheometer. Reprinted from Figure 9.10 (Collyer and
Clegg, 1988).
The yield stress may then be calculated on the basis of a simple force-balance calculation
from the following equation;
Tm pDv 3 =2Hv =Dv 1=3ry ;
4:1
where Tm is the maximum in torque in a steady time test (in which the torque (T) is
monitored as a function of time (t) at a xed shear rate (c 0 )), Dv is the external diameter of
the vane, Hv is the height of the vane and ry is the yield stress. Once again, with reactive
systems the effects of temperature and cure on the yield stress should be considered in order
to obtain a full chemorheological description of the material.
Wall slip
There is also a variety of wall-slip techniques used in rheological analysis (Gupta, 2000).
The methods described here are ow visualization, capillary ow and torsional ow.
Flow visualization essentially determines the slip velocity at the wall via particle-tracking
software and data capture, and is typically used only for model systems and surfaces (Piau
et al., 1995). However, some work on visualization of wall slip in highly lled reactive
epoxy-resin moulding compounds has been reported (Manzione and Weld, 1994, Manzione,
1995).
Capillary rheometry and parallel-plate rheometry use the fact that wall slip will manifest
itself as a geometry-dependent phenomenon. That is, wall slip will appear as a geometric
effect on apparent rheological properties. In the capillary-rheometer technique, slip will
manifest itself as an effect of capillary diameter (D) on the shear stress (sw). Wall slip in
capillary rheology can be calculated from an analysis that involves the following:
(a). determining the wall shear stress (sw DP/[4(L/D)]) versus apparent shear rate (c0 app
32Q/(pD3)) curves for a variety of diameters, where DP is the pressure drop across the
capillary, L is the capillary length, D is the capillary diameter and Q is the ow rate;
(b). plotting the apparent shear rate (c0 app) versus 1/D for various values of the shear stress
(sw); and
326
(c). calculating the slip velocity from the slope of the curves in (b) or via the expression
8Vs dc 0app =d1=D;
4:2
4:3
Where c0 RH1 and c0 RH2 are the shear rates at the plate edge for plates of gaps H1 and H2,
respectively.
Figure 4.6 shows that all wall-slip techniques can be used successfully to determine the
wall slip of a material over a wide range of shear stresses. Once again, for full characterization of wall slip in reactive systems the effects of temperature and extent of cure on wall
slip should be determined.
Figure 4.6. Wall-slip tests with a range of rheometer systems. Reprinted from Figure 9.11 (Collyer
327
10 6
Viscosity h or h* (Pa s)
STEADY VISCOSITY
MOD. COXMERZ
COXMERZ RULE
10 5
10 4
10 3
10 2
10 1
10 0
Figure 4.7 shows the steady- and dynamic-viscosity proles as functions of shear rate for a
lled reactive epoxy-resin moulding compound. Here, interestingly, the CoxMerz rule
provides a better correlation than does the modied CoxMerz rule.
Overall these fundamental chemorheological tests will provide essential characterization
of the typically complex reactive systems under investigation. As mentioned previously, to
achieve full characterization we are next interested in chemoviscosity proles.
4.3
Chemoviscosity profiles
4.3.1
Chemoviscosity
The chemoviscosity of thermosetting resins is affected by many variables. In a major
review, Ryan 1984) expressed the chemoviscosity (g) as a function of pressure (P), temperature, time, shear rate (c 0 ) and ller properties (F ), as shown by the following general
equation:
g gT; P; c0 ; t; F:
4:4
The effects of each variable on the chemoviscosity are usually examined by separate tests
such as of
cure effects,
gc gc T; t;
4:5
shear-rate effects,
gsr gsr c0 ; T;
4:6
gf nf F;
4:7
gf gc F; T; t;
4:8
gf gsr F; c0 ; t:
4:9
ller effects,
or
or
Recent work has included tests that examine the effects of shear rate and cure simultaneously (which will also be discussed below). The models derived from these tests are
recombined to provide an overall chemoviscosity model to be used in processing applications. The effect of pressure on chemoviscosity has not been studied extensively; however,
system pressure may be relevant to high-pressure injection-moulding and transfer-moulding
processes.
The cure effects on the chemoviscosity are two-fold; the viscosity will initially decrease
due to the increase in thermal effects but will eventually increase due to formation of the
crosslinked network via the curing reaction. This is shown schematically for the chemoviscosity of a polyester resin during injection moulding in Figure 4.8.
Initially, during plasticization, the viscosity of the resin decreases due to shear heating
and thermal effects (stage I). As the temperature (or time) increases, the curing reaction
begins and the decrease in viscosity due to heating is compensated by the increase in
viscosity due to the curing process. At the point of minimum viscosity the polyester is
injected into the mould (stage II). Finally, the viscosity of the resin increases as the material
is transformed into an infusible solid (stage III). The cure effects on viscosity (gc) may be
determined by measuring the activation energies of the reaction and of viscous ow.
Viscosity
328
II
III
Time
Figure 4.8. The viscosity of a thermoset polymer during processing.
329
The effects of temperature and time on the chemoviscosity can also be described
explicitly in terms of the extent of cure (a) from knowledge of the kinetics of the cure
(i.e. (T, t)) and temperature by the following equation:
gc gc T; a:
4:10
Extensive work in the literature has focussed on the determination of the cure effects on the
chemoviscosity of thermosetting resin, and a summary of the effects of cure on various
chemorheological models is given in Table 4.1.
These models, which examined the effects of cure on the chemoviscosity, range from simple
empirical models (Malkin and Kulichikin, 1991, Lane, 1987, Mussati and Macosko, 1973,
Kamal and Sourour, 1973, Roller, 1976, Dusi et al., 1982) (used for epoxy-resin and polyurethane systems), probability-based and molecular models (Vinogradov and Malkin, 1977,
Lipshitz and Macosko, 1976, 1977, Mijovic et al., 1996) (used for polyurethane and epoxyresin systems), gelation models (Malkin and Kulichikin, 1991, Castro and Macosko, 1982,
Yang and Suspene, 1991) (used for polyester, epoxy-resin and melamine systems) and
Arrhenius models (Martin et al., 1989, Kamal, 1974, Kojima et al., 1986, Tungare et al., 1986,
Dusi et al., 1983) (used for many thermosets, including epoxies and polyimides) to detailed
models based on free-volume analyses (Hale et al., 1989, Chiou and Letton, 1992) (used for
epoxy-resin systems). Further reviews of curing-effect models are given by Roller (1976) and
Ryan (1984). Great accuracy in determination of the cure effects on viscosity is required, since
they have a large effect on the processing of thermosets, especially during the nal curing
stage. Accurate data are also needed for determining the parameters in some of the models.
Of course, it should be noted that cure conversion or kinetic models themselves should be
accurately determined, because they must be used in parallel with cure models of chemoviscosity. There are essentially two forms of kinetic model used to describe thermoset
curing reactions, namely empirical and mechanistic models. Empirical models assume an
overall reaction order and t this model to the kinetic data. This type of model provides
no information on the kinetic mechanisms of the reaction, and is predominantly used to
provide models for industrial samples. Mechanistic models are derived from an analysis of
the individual reactions involved during curing, which requires detailed measurements of
the concentrations of reactants, intermediates and products. Essentially, mechanistic models
are intrinsically more complex than empirical models; however, they are not restricted by
compositional changes, as are empirical models. Typical kinetic models used in the analysis
of thermosetting chemical reactions are listed in Table 4.2.
These models include simplistic empirical models (Malkin and Kulichikin, 1991, Martin
et al., 1989, Ryan, 1973, 1984, Arrelano et al., 1989, Han and Lem, 1984, Hale et al., 1989,
Dutta and Ryan, 1979, Lane, 1987) (used for epoxy-resin and polyester systems) and
complex mechanistic models derived from chemical analysis and probability theories (Yang
and Suspene, 1991, Gupta, 1990, Riccardi and Williams, 1986, Batch and Macosko, 1987,
Pannone and Macosko, 1988, Mussati and Macosko, 1973) (used for epoxy-resin, polyester
and polyurethane systems). Determination of the most appropriate kinetic model for an
application will depend on the type of system and the accuracy and form of results required.
Evaluation of the appropriate kinetic model parameter from kinetic measurement denes
the degree of conversion as a function of time and temperature, a(T, t). This relationship is
then used together with chemoviscosity and gel effects to dene the thermoset process.
The shear-rate effect on the viscosity of thermosetting resins, gsr(c), is also essential to the
determination of the chemoviscosity, g. The exact relationship will depend on the type of
Combined
Self-inhibition
Self-acceleration
Mechanistic
Autocatalytic, 2
Autocatalytic, 1
Polynomial
Series of nth-order
reactions
nth order
da
k1 a1 Ca1 ea
dt
da
k1 k2 a1 aB a
dt
a
f concentration
agel
da
k1 a1 Ca
dt
da
k1 a1 ea
dt
da
Ea
a0 a1 a a2 a2
k exp
RT
dt
da
k1 k2 am 1 an
dt
k1 k0 expE1 =RT
k2 k0 expE2 =RT
da X Ai
Ei
gi
exp
1 ai ni
b
RT
dt
i
da
k1 a
dt
k A expE=RT
da
k1 a2
dt
k A expE=RT
da
k1 an
dt
k A expE=RT
First order
Second order
System
Expression
Model
e, constant
C, constant
B, stoichiometry factor
n, m, reaction orders
k1, k2, rate constants
E1, E2, activation energies
gi, factor
Ai, coefcient
b, heating rate
ni, reaction order
a0, a1, a2, constants
n, reaction order
k, rate constant
Ea, activation energy
A, coefcient
Notes
gc gTexpfkt
gc kM a
First-order, nonisothermal
Empirical, 3
Empirical, 2
Empirical, 1
Molecular, 2
g0
1 kt a
1 t=t
log gc log gv
a a3 for tp < t
Ev
ka
RT
gc Kta
a a1 for t < tc
a a2 for tc < t < tp
a 1 for M < M c
f f a
Polyurethane (Gonzalez-Romero
and Casillas, 1989)
gc g0 expht
First-order, isothermal
Molecular, 1
1 C expE =RT
a
Macosko
t1
System
Expression
Model
Homogeneous reaction up
to termination of reaction
No diffusion limitations
or gel effects considered
a, K, constants
tc, tp, characteristic times
k, constant
Ev, viscous activation energy
gv, initial viscosity
Reaction and gel effects
with no phase separation.
k, constant
A, D, C, S, constants
M 0 , initial molar mass
Comments
Modied Arrhenius
rst order
nth order
(non-isothermal)
Arrhenius rst-order
(isothermal)
Microgel
Microgel, 1
Gel model
Empirical, 4
Model
Ev
kk
RT
Ev
kk
RT
Ek
dt
RT
1 an exp
exp
Ev
Ek
tkk exp
RT
RT
ln g ln gv
ln g ln gtv
ln g ln gv
log gs A B
t
g
gs c 1
go
gc a
ABa
g0
a* a
g
t
b
1
g0
t
a 1 for t < tc
a 3:5 for t > tc
Ek
dt
RT
Expression
t
g
1
b log
log c
a log
gm
t
1 t=t
Polyurethane (Gonzalez-Romero
and Casillas, 1989)
n, reaction order
Comments
System
Modied WLF, 2
Modied WLF, 1
WLF
1
Mw
fa =B a1 =BT Tg
E
C1 T Tg
RT C2 T Tg
Mw Mw a
Tg Tg a
C1 aT Tg a
gc T
ln
C2 a T Tg a
gc Tg
Mw Mw a
Tg Tg a
a at
gc g0 exp
ln gc ln ga ln M w
334
Form
Systems
Notes
Power law
A, constant
n, index
Bueche
Harding
Carreau
gSR
1
g0
1 kx3=4
gSR
1 kx2 n1=2
g0
k, relaxation time
g0, zero-time
viscosity
Cross
1 kc0 2 n1=2
gSR
1
g0
1 kx1n
1
1 g0 c0 =s
1n
Ellis
gSR
1
1n
g0
g=g0 kx n
1
1 s=s1=2
WLF
gSR
at
;
A0 A1 xat A2
C1 T T0
at exp
C2 T T0
system used. For example, Sundstrom and Burkett (1981) and Hartley and Williams (1981)
found that polyesters and polyurethane exhibit essentially Newtonian behaviour, whereas
epoxies and phenolics show marked shear thinning. More detailed rheological models used to
characterize the effect of shear rate on chemoviscosity are displayed in Table 4.3.
The power-law model is the most extensively used shear-rate model for thermosets and
has been used for unlled (Ryan and Kamal, 1976, Kascaval et al., 1993, Riccardi and
Vazquez, 1989) and lled (Ryan and Kamal, 1976, Knauder et al., 1991) epoxy-resin
systems. Sundstrom and Burkett (1981) showed that there was a good t of the viscosity of
diallyl phthalate to the Cross model. The viscosity of polyesters has been modelled by Yang
and Suspene (1991) using a Newtonian model. The WLF model has been used by Pahl and
Hesekamp (1993) for a moderately lled epoxy-resin system. Rydes (1993) showed that the
viscosity of DMC polyesters followed a power-law relationship at high shear rate.
The filler effects on the chemoviscosity of thermosetting resins have not been studied
extensively, but are vital to understanding the rheology of lled thermosets. For example,
the effects of ller concentration on viscosity can be used in process control to monitor
batch-to-batch variations or to provide essential information for research into alternative
ller/resin batches. Ng and Manas-Zloczower (1993) examined an epoxy-resin system with
silica ller and established that the elastic modulus of the resin can be expressed in terms of
335
Expression
Power law/WLF
g g0 exp
System
fg
B
a
B T Tg
!
Mw
Mw f a
Tg Tg a
Power law/WLF/
Conversion
WLF/Arrhenius
Power law/
Arrhenius/
molecular, 1
Power law/
Arrhenius/
molecular, 2
Carreau
Mould ow
g0 Acn1
A3 A4 a
ag
C1 T Tg
g A1 c0A2
exp
C2 T Tg
ag a
Z
DEk
ln g* ln g*T; x k a
exp
dt
RT
aT T
g*T; x
A0 A1 xaT TA2
0 n1
c
dc
gl 0
expbT T0 aPMxm
c0
dr
Mw f a
g Ac0B exp
C
ag DEa
ag a
T
"
0 2 #n1=2
g ga
g c
1 0
g0 ga
s*
ag BCa
E
where g0 A exp
T ag a
ln g A1 A2 ln c0 A3 T A4 ln c02
ag
A5 lnc0 T A6 T 2 A7 A8 X ln
ag a
Epoxy resin
(Castro and Macosko, 1982)
Filled epoxy resin
(Pahl and Hesekamp, 1993)
Assumes CoxMerz rule is
valid
Epoxy resin (Epoxy/Diamine)
(Riccardi and Vazquez,
1989)
Epoxy resin (Castro and
Macosko, 1982, Nguyen,
1993)
Epoxy resin (Peters et al.,
1993)
ller concentration and the modulus of the matrix, and that the gel time decreases with
greater ller concentration. Dutta and Ryan (1979) examined the effects of llers on the
kinetics of an epoxydiamine reaction, for which they found that the type of ller affected
the reaction kinetics. They noted that carbon-black llers increase the reaction rate via preexponential terms and that silica llers affect the reaction rate through activation energies.
Han and Lem (1983a, 1983b, 1983c, 1984, Lem and Han, 1983a, 1983b, 1983c) noted the
effects of various llers on the kinetics and chemorheology of polyester resins. They found
that increasing the concentration of clay or glass bre exaggerated the degree of shear
thinning at low shear rates (a behaviour similar to that of concentrated suspensions in
Newtonian uids), whereas the addition of CaCO3 at all concentrations produced Newtonian behaviour. Han and Lem (1984) found an interaction among llers, thickeners and
low-prole additives that produced complex rheological behavior. Kalyon and Yilmazer
(1990) and Metzner (1985) stated that the effect of llers on the viscosity of highly lled
polymeric suspensions induces wall slip, ow instabilities, yield stresses and dilatancy.
Kalyon and Yilmazer (1990) also noted that ller concentration, surface interactions,
orientation in the ow eld, particle shape, particle diameter and resin properties will
inuence the effect of llers on the overall viscosity.
336
By combining the effects of cure, shear rate and ller, a complete model for the chemoviscosity can be established. Examples of these combined models are shown in Table 4.4.
It should be noted that the effects of llers may be incorporated into the cure and shearrate effects. The main forms of combined-effects model consist of WLF, power-law or
Carreau shear effects, Arrhenius or WLF thermal effects and molecular, conversion or
empirical cure effects. Nguyen (1993) and Peters et al. (1993) used a modied CoxMerz
relationship to propose a modied power-law model for highly lled epoxy-resin systems.
Nguyen (1993) also questions the validity of the separability of thermal and cure effects in
the derivation of combined models.
Together with the chemoviscosity modelling, other variables are important to the processing of thermosets. These include
gmin the minimum viscosity (as a function of temperature) for injection or transfer
moulding, and
dT/dt the optimum heating rate to ensure that the minimum-viscosity the and gelation
times are optimized.
The measurement of these properties, together with an accurate chemoviscosity model,
enables prediction of optimum ow conditions.
4.3.2
Gel effects
Various viscosity models have implicitly included the effects of gelation on the chemoviscosity, and these were reviewed in Table 4.2 incorporating gelationconversion and
glass-transition-temperature effects implicitly in the cure effects on chemoviscosity. Explicit
models for the expression of gel time versus temperature and time are sparse, with empirical
measurements mainly being used.
Thus, with a knowledge of the effects of gelation and an accurate chemoviscosity model,
the optimum ow and cure conditions can be established. The measurement techniques
for chemoviscosity and gelation effects will be reviewed next.
4.4
Chemorheological techniques
Testing procedures in the measurement of the chemorheology of thermosetting resins
usually incorporate the following techniques:
reference to international or national standards (these, in some cases, use techniques that
mimic true rheological characterization given below)
measuring the effect of shear rate and temperature on chemoviscosity: gsr gsr(c 0 , T)
determining the effect of curing on the chemoviscosity: gc gc(T, t) gc(T, a)
measuring the effect of llers on the chemoviscosity: gsr(F) and gc(F)
measuring the effect of temperature and llers on the gel time: tgel(T, F)
measuring the combined effects of shear rate and curing: gsrc gsrc(T, a)
Determining optimum process properties: minimum viscosity, gmin, and optimum heating
rate, (dT/dt)opt.
Each of these techniques will be discussed in detail below.
337
Cup ow
Spiral ow
Impregnated
resin ow
Roller mixer
Designation
Notes
BS2782-86
Method 720 B
Slow-curing thermosets
Measure time to ll
Set closing speed, force, amount of ash and temperature
ASTM D731-84 Measure time to ll given mould
Set closing speed, force, amount of ash and various temperatures
ASTM D3123-83 Measure length of ow along spiral die at end of test
Set pressure and temperature from transfer press
ASTM D3795-70 Measure ow from ash or central piece in plate press
Method 105 D
Set temperature and pressure for epoxies, phenolics and silicones
ASTM D3795-90 Measure torque over time in roller mixer
Measure initial maximum, minimum and nal maximum torques
and corresponding times
Set temperatures, roller speed and charge volume
Viscosity tests
Capillary rheometer ASTM D3835-90 Measure shear effects on viscosity at high rates
Given sample conditioning (ASTM D618), given pressures or
ow rates, determine viscosity via various corrections
Determine dwell time that does not inuence results (due to curing)
Dynamic properties ASTM D4065-90 Viscoelastic properties
ASTM D4473-90 Cure effects on viscosity
Dynamic properties of supported and unsupported resins
Isothermal and non-isothermal tests
Viscosity of epoxy ASTM D2393-86 Measure pre-cure viscosity of lled and unlled epoxy liquids
resins
Viscosity versus spindle speed at desired temperature
Gel-point tests
Rotating rod
ISO 2535-74
Measure gel time at 50 Pa s of polyesters
Rotating rod at given speed, temperature and rod dimensions
Hand probe
ASTM D2471-79 Measure gel time and exotherm temperature
Wooden hand-stirred probe at given temperature and rod
dimensions
Glass rod
BS 2782
Measure gel time of phenolics
(M:835A-80)
Stir resin with glass rod until rubbery at given temperatures
Test tube
(M:835B-80)
Measure gel time of polyesters
Resin test tube at given temperature and volume
Gel timer
(M:835C-80)
Measure gel time from restricted motion of plunger for epoxies
and polyesters
Knife
(M:835D-80)
Vertical reciprocating disc and plunger is stopped when material
gels at given temperature
Press
(M:835E-80)
Measure gel time
Stir sample on hot plate with knife
Measure gel time for impregnated resin
Exude material from between plates with wire at given
temperature and size of plates
Dynamic properties ASTM 4065-90
Viscoelastic properties
ASTM 4473-90
Gel point from crossover tests and time for dynamic viscosity
to reach 100 Pa s
For supported and unsupported resins
Isothermal and non-isothermal tests
338
4.4.1
Standards
Standards for the chemoviscosity of reactive uids are shown in Table 4.5. Clearly these
tests mainly are typical of rapid, quality-control measurements; however, in contrast the
standard proposed by Rydes (1993) for using capillary and slit-die rheometry is based on
true rheological analysis. Alternative rheometers used for reactive systems are summarized
in detail in Table 4.6.
4.4.2
339
Measurement
Advantages/Limitations
Advantages
Most common geometry
Small sample size
Easy to clean
Various plate designs
Limitations
Non-uniform shear rate
Low to moderate shear rates
Advantages
Small sample size
Uniform shear rate
Limitations
Difcult to measure slip
Edge-fracture problem
Limited to low shear rate
Eccentric disc (Collyer and Clegg, Isothermal dynamic-shear sweep Advantages
1988, Sundstrom and Burkett, g (c0 , T) low c0 from
Small sample size
1981)
CoxMerz rule
Limitations
Only one deformation mode
Sample exposure
Capillary (Ryan, 1984, Ryan and Isothermal steady-shear sweep Advantages
g g(c0 , T) high c0
Kamal, 1976, Rydes, 1993,
Large shear rates
Blyler et al., 1986)
slip
Limitations
yield stress
Large sample size
Cannot monitor time effects
Difcult to clean
End effects
Slit (Collyer and Clegg, 1988,
Isothermal steady-shear sweep Advantages
Han and Wang, 1994)
g g(c0 , T) high c0
Sample ow
slip
Large shear rates
yield stress
Limitations
Cannot monitor time effects
Difcult to clean
Advantages
Squeezing ow (Tungare et al.,
Isothermal steady time test
Simple ow
1986)
g g(T, t) at low c0
Non-isothermal steady time test Limitations
Limited time
g g(T, t) at low c0
Low shear rates
340
Measurement
Advantages/Limitations
341
Measurement
Advantages/Limitations
Isothermal tests
g g(T, t, c) from inverse
problem solution
Advantages
Good for fast-curing systems
Isothermal damping
tgel (T, t) from damping of a
xed oscillation
Limitations
Requires complex numerical analysis
Advantages
Good for fast-curing systems
Limitations
Requires complex numerical analysis
342
4.4.3
343
examined the relationship between curing complex viscosity and dielectric properties
through temperature-ramp tests. A standard procedure for dynamic temperature ramps tests
is given in ASTM 4440-84/90, where ramp rates of 35 C/min are recommended, in order
to allow the sample to equilibrate at a given temperature. Although most industrial ramp
rates are non-linear, this linear rate is representative of typical curing rates used.
An inherent assumption when using the above dynamic techniques is that the complex
viscosity gives a good representation of the steady-shear viscosity during the curing reaction. This has been validated for many systems. However, care should be taken when
relating the effects of cure on complex viscosity to the processing viscosity; in other words
the CoxMerz rule or a similar relationship must be validated.
Steady-shear temperature ramp tests are used to monitor the change in steady-shear
viscosity at a given shear rate during a linear or non-linear temperature ramp. Tungare
(1986) used a squeezing ow viscometer to measure the change in viscosity during a linear
temperature ramp of an embedded epoxy resin and developed a model for the curing
viscosity using an optimization technique.
The most common geometry used for the determination of the cure viscosity is that of the
parallel-plate rheometer (Martin et al. 1989, Hale et al. 1989, Sundstrom and Burkett, 1981,
Kojima et al. 1986, Pahl and Hesekamp, 1993, Ryan and Kamal, 1976, Ng and ManasZloczower, 1993, Nguyen, 1993, Peters et al. 1993, Kenny et al. 1991, Choi and Lin, 1991,
Halley et al. 1994). This rheometer may be operated at low shear rates under isothermal and
non-isothermal conditions. Alternative measurements have been used to measure the effects
of cure on the viscosity, and these are summarized in Table 4.6.
4.4.4
344
It is evident that strict control over the composition of a lled thermoset system is vital to
establishing the chemoviscosity of the system. Also, the measurement of samples with
various ller loadings, ller types, surface treatments and ller dimensions would aid in
fully describing the effect of llers and enable better characterization of lled samples.
4.4.5
4.4.6
Process parameters
Optimum process parameters such as the minimum processing viscosity (gmin) and the
optimum heating rate can be determined by performing a series of non-isothermal dynamic
temperature ramps. Using various heating rates, the optimum minimum processing viscosity
at the desired heating rate can be established. The optimum heating rate is determined by
optimizing the time needed to reach the minimum viscosity and the time over which this
minimum viscosity extends. It should be noted that increasing the heating rate by too much
can cause large temperature differences in the moulded part. The time at which the viscosity
is at or near its minimum value is known as the processing window. Non-isothermal
dynamic temperature tests are generally performed on a parallel-plate rheometer at xed
frequencies and strains to determine this window.
This review of chemorheological techniques highlights the numerous possibilities for
monitoring chemoviscosity and gel-point measurements. Determination of the appropriate
techniques must be based on the following considerations:
The selection of the optimum chemorheological techniques can then be used to inuence the
choice of rheometer for chemorheological measurements. Final selection of the rheometer
design must be determined by the considerations given in the previous section, as well as the
following:
cost
ease of use
reproducibility of results
versatility
345
4.5
Gelation tests
One of the key transitions that also needs to be characterized is gelation. The physical nature
and models of gelation have been described in Chapter 2, so here we wish to examine the
methods used to determine gelation. The determination of the gel point of thermosets may
be monitored by the following methods:
In isothermal dynamic time tests the storage modulus (G0 ) and loss modulus (G00 ) are
monitored at a given frequency (x) and strain (c) over time. The gel point, or the initial
formation of an innite network, has been reported in the literature (Winter and Chambon,
1986) as the time at which G0 and G00 cross over. However, Winter (1987) notes that the
crossover point is valid only for stoichiometrically balanced systems or those with excess
hardener. For these systems the gel point is equivalently dened as the point where the loss
tangent (tan d G00 /G0 ) reaches unity. Standard ASTM 4473-90 also describes the technique determining for the crossover-point analysis via isothermal dynamic time tests for
thermosetting resins. Scanlan and Winter (1991) and Winter et al. (1994) expand on
Winters initial work by analysing effects of gelation on the gel constitutive equation
(i.e. the constitutive equation at gelation) and its key parameters of gel strength (S) and
power law index (n). They note the characterization of network structure via analysis of
S and n and the effects of Mw on these parameters. Arbabi and Sahimi (1990), Lairez et al.
(1992) and Eloundou et al. (1996a, 1996b) extend the characterization of gelation to
analysis of critical viscoelastic properties near gelation and correlate the results to known
molecular theories.
Arrelano et al. (1989) evaluated the gel points of various DGEBA-based epoxy-resin
systems via the crossover method. They also dened the vitrication point as the maximum
in G00 in an isothermal dynamic time test. In all these gel-point measurements the frequency
must be chosen such that the relaxation of the network is enabled during data sampling. This
is represented as
texp
2p
> trelax ;
x
4:11
where texp is the experiment time and trelax is the characteristic relaxation time of the system.
Thus, a low enough frequency must be used to enable equilibrium values to be obtained.
However, as DeRosa and Winter (1993) explain, the time must not be too large (frequency
must not be too low), in order that one does get gel effects during the experiment. Harran
and Laudouard (1986) also used isothermal dynamic time tests to measure the gelation and
vitrication of DGEBA epoxy-resin systems. The gelation point in their study was, however, measured as the change in slope of G00 with time (due to the slowing of the reaction
rate). Harran and Laudouard (1986) also noted that there was a minimum temperature for
this inection-point method that was not related to vitrication. The vitrication point is
determined by the maximum in tan d (or G00 ), as was done by Lairez et al. (1992). However,
Lange et al. (1996) prefer to dene vitrication as a transition process rather than a single
346
transition point, and dene its initiation as where the dynamic modulus (G*) becomes nonzero and its end when the dynamic modulus plateaus.
Isothermal steady time tests are used to determine the gel point of a thermoset system as
the point at which the shear viscosity tends towards innity. In these tests the viscosity is
measured as a function of time at a constant shear rate. This method has the following major
disadvantages. Firstly, the innite viscosity can never be measured due to equipment
limitations and thus the gel time must be obtained by extrapolation. Secondly, shear ow
may destroy or delay network formation. Finally, gelation may be confused with vitrication or phase separation since both these processes lead to an innite viscosity (St John
et al., 1993). However, some work by Matejka (1991) and Halley et al. (1994) has shown
that extrapolation to zero values of reciprocal viscosity or normal stress (i.e. extrapolation to
innite viscosity and normal stress) can be used with some success.
An isothermal multiwave test is a recent technique that uses a Fourier-transformation
response to a multiple-frequency wave to determine the gel point. This test is used to
measure the dynamic response of a uid to simultaneously applied sine waves of various
dynamic frequencies, each with the same maximum strain. The gel point is determined as
the point at which the loss tangent becomes independent of frequency. Winter (1987)
describes this method as useful for determining the gel point of any resin system. De Rosa
and Winter (1993) note that for entangled polymers a frequency window must be chosen
such that the system is in the terminal zone. The lower limit of frequency is expressed by
Equation (4.11) as previously mentioned, and the upper limit is determined by the crossover
of G0 and G00 with respect to frequency. Matejka (1991) nds the gel point of an epoxy-resin
system via the independence of the loss tangent with respect to frequency, but uses a series
of isothermal dynamic rate sweeps. The advantage of the multiwave test is that a series of
frequencies can be imposed simultaneously, thereby eliminating the need for performing
multiple isothermal dynamic time tests, which may suffer from test-time effects (Lairez
et al. 1992).
Winter (Winter and Chambon, 1986, Winter, 1987) also described an isothermal dynamic
relaxation test to measure the gel point. He noted that the gel point coincides with a powerlaw relationship between the relaxation modulus (G) and relaxation time (t) (G Stn,
where S and n are constants. An isothermal step strain test measures the relaxation modulus
as a function of time after an instantaneously applied strain. The gel time can be measured
as the point at which the prole of the relaxation modulus can be expressed by this power
law. This model is equivalent to tan d being independent of frequency.
Dynamic temperature-ramp tests are used to measure the storage and loss modulis of a
system as a function of time during a given temperature ramp. The gel time is expressed as
the time at which the loss tangent reaches a maximum. The cure time has also been dened
by Choi and Lin (1991) as the intersection of the increasing storage modulus and the plateau
modulus, determined using a dynamic temperature-ramp test. Nguyen (1993) also used this
test to measure the gel time of a lled epoxy-resin system. Peters et al. (1993) expressed
reservations about the validity of this type of test for highly lled epoxy resins due to
frequency dependences. Standard ASTM 4473-90 also describes an experimental procedure
using the temperature-ramp test to determine the gel point of thermosets via the maximum
in tan d.
An extensive review of the use of various gelation tests (G0 /G00 crossover, independence
of tan d with respect to frequency, innite viscosity and innite normal stress, gel ltration
and DSC theoretical gel conversion) is given by Halley et al. (1994) for unlled epoxy-resin
References
347
References
Aranguren, M., Mora, E., Degroot, J. & Macosko, C. (1992) J. Rheology, 36, 11651182.
Arbabi, S. & Sahimi, M. (1990) Phys. Rev. Lett., 65, 725728.
Arrelano, M., Velasquez, P. & Gonzalez-Romero, V. (1989) Society of Plastics Engineers Annual
Technical Conference ANTEC 1989, p. 838.
Barnes, H. & Carnali, J. (1990) J. Rheology, 34, 841.
Batch, G. & Macosko, C. (1987) Society of Plastics Engineers Annual Technical Conference
ANTEC 1987, p. 974.
Bidstrup, W., Sheppard, N. & Senturia, S. (1986) Polym. Eng. Sci., 26, 359361.
Blyler, L. Jr, Bair, H., Hubbauer, P. et al. (1986) Polym. Eng. Sci., 26, 1399.
Carriere, C., Bank, D. & Christenson, C. (1992) Polym. Eng. Sci., 32, 426.
Castro, J. & Macosko, C. (1982) Polym. Eng. Sci., 28, 250.
Chiou, P. & Letton, A. (1992) Polymer, 33, 3925.
Choi, J. & Lin, C. (1991) Society of Plastics Engineers Annual Technical Conference ANTEC 1991,
p. 1746.
Chu, B. & Hilker, R. (1989) Rev. Sci. Instrum., 60, 3828.
Collyer, A. & Clegg, D. (1988) Rheological Measurements, Essex: Elsevier Applied Science.
Cox, W. (1958) J. Polym. Sci., 28, 619.
De Rosa, M. & Winter, H. (1993) Society of Plastics Engineers Annual Technical Conference
ANTEC 1993, p. 2620.
Doraiswamy, D., Mujumdar, A., Tsao, I. et al. (1991) J. Rheology, 35, 647685.
Dusi, M., May, C. & Seferis, J. (1982) ACS Org. Coat. Appl. Polym. Sci. Proc., 47, 635.
Dusi, M., May, C. & Seferis, J. (1983) ACS Symp. Series, 227, 301318.
Dutta, A. & Ryan, M. (1979) J. Appl. Polym. Sci., 24, 635.
348
References
349
Lane, J., Seferis, J. & Bachmann, M. (1986) Polym. Engin. Sci. 26, 346353.
Lange, J., Altmann, N., Kelly, C. & Halley, P. (2000) Polymer, 41, 59495955.
Lange, J., Manson, J. & A, H. (1996) Polymer, 37, 58595868.
Lee, L. (1981) Polym. Eng. Sci., 21, 483.
Lem, K. & Han, C. (1983a) J. Appl. Polym. Sci., 28, 779.
Lem, K. & Han, C. (1983b) J. Appl. Polym. Sci., 28, 3185.
Lem, K. & Han, C. (1983c) J. Appl. Polym. Sci., 28, 3207.
Lipshitz, S. & Macosko, C. (1976) Polym. Eng. Sci., 16, 503.
Lipshitz, S. & Macosko, C. (1977) J. Appl. Polym. Sci., 21, 2029.
Maazouz, A., Dupuy, J. & Seytre, G. (2000) Polym. Eng. Sci., 40, 690701.
Maier, P. & Goritz, D. (1996) Kaut. Gummi Kunststoffe, 49, 1821.
Malkin, A. & Kulichikin, S. (1991) Adv. Polym. Sci., 101, 218254.
Manzione, L. (1995) Processing epoxy molding compounds, in Thermoset Polymer Meeting for
J. K Gillhams 60th birthday, Princeton, MA: Princeton University.
Manzione, L. & Weld, J. (1994) Society of Plastics Engineers Annual Technical Conference
ANTEC 1994, pp. 13711376.
Martin, G., Tungare, A., Fuller, B. & Gorto, J. (1989) Society of Plastics Engineers Annual Technical
Conference ANTEC 1989, p. 1079.
Matejka, L. (1991) Polym. Bull., 26, 109.
Metzner, A. (1985) J. Rheology, 29, 739.
Mijovic, J., Andjelic, S., Fitz, B. et al. (1996) J. Polym. Sci., B: Polym. Phys., 34, 379388.
Mijovic, J. & Lee, C. (1989) J. Appl. Polym. Sci., 38, 2155.
Mussati, F. & Macosko, C. (1973) Polym. Eng. Sci., 13, 236.
Nass, K. A. & Seferis, J. C. (1989) Polym. Eng. Sci., 29, 315.
Ng, H. & Manas-Zloczower, I. (1989) Polym. Eng. Sci., 29, 1097.
Ng, H. & Manas-Zloczower, I. (1993) Polym. Eng. Sci., 33, 211216.
Nguyen, L. (1993) Proceedings from the 43rd IEEE Electronic Component and Technology
Conference, Buena Vista, FL, pp. 110.
Pahl, M. & Hesekamp, D. (1993) Appl. Rheology, 70.
Pannone, M. & Macosko, C. (1988) Polym. Eng. Sci., 28, 660.
Payne, A. (1962) J. Appl. Polym. Sci., 6, 5763.
Peters, G., Spoelstra, A., Meuwissen, M., Corbey, R. & Meijer, H. (1993) Rheology and rheomerty
for highly lled reactive materials, in Dijksman, J., Nieuwstadt, F. T. M. (Eds.) Topics in Applied
Mechanics, Dordrecht: Kluwer Academic Publishers.
Pethrick, R. (1993) in Collyer, A. (Ed.) Techniques of Rheological Measurements, London:
Chapman & Hall.
Petrie, C. & Denn, M. (1976) AIChE J., 22, 209.
Piau, J., Kissi, N. & Mezghani, A. (1995) Non-Newtonian Fluid Mechanics, 59, 1130.
Pichaud, S., Duteurtre, X., Fit, A. et al. (1999) Polym. Int., 48, 12051218.
Prime, R. (1973) Polym. Eng. Sci., 13, 365.
Rabbinowitsch, B. (1929) Z. Physik. Chem. (Leipzig) A 145, 126.
Riccardi, C. & Vazquez, A. (1989) Polym. Eng. Sci., 29, 120.
Riccardi, C. & Williams, R. (1986) Polymer, 27, 913.
Roller, M. (1976) Polym. Eng. Sci., 16, 687.
Ryan, M. (1973) Thesis, Montreal, McGill University.
Ryan, M. (1984) Polym. Eng. Sci., 24, 698.
Ryan, M. & Kamal, M. (1976) Proceedings of the VII International Congress on Rheology,
pp. 289290.
350
Rydes, M. (1993) Aspects of the Rheology of Unsaturated Polyester Dough Molding Compounds,
London: University of West London and National Physical Laboratories.
Scanlan, J. & Winter, H. (1991) Macromolecules, 24, 4754.
Sourour, S. & Kamal, M. (1976) Thermochim. Acta, 14, 41.
St John, N., George, G., Cole-Clarke, P., Mackay, M. & Halley, P. (1993) High Perform. Polym., 5, 21.
Sundstrom, D. & Burkett, S. (1981) Polym. Eng. Sci., 21, 1108.
Tajima, Y. & Crozier, D. (1983) Polym. Eng. & Sci., 23, 186.
Tungare, A., Martin, G. & Gorto, J. (1986) Society of Plastics Engineers Annual Technical
Conference ANTEC 1986, p. 330.
Utracki, L. & Fisa, B. (1982) Polym. Composites, 3, 193210.
Vinogradov, G. & Malkin, A. (1977) Rheology of Polymers, Moscow: Mir.
Wilhelm, M., Reinheimer, P., Ortseifer, M., Neidhofer, T. & Spiess, H. (2000) Rheologica Acta, 39,
241246.
Winter, H. (1987) Polym. Eng. Sci., 27, 1698.
Winter, H. & Chambon, F. (1986) J. Rheology, 30, 367382.
Winter, H., Izuka, A. & Derosa, M. (1994) Polym. Gels Networks, 2, 239245.
Yang, Y. & Suspene, L. (1991) Polym. Eng. & Sci., 31, 321.
Yoshimura, A. & Prudhomme, R. (1988) J. Rheology, 32, 5367.
Yoshimura, A. & Prudhomme, R. (1988) J. Rheology, 32, 575.
5.1
Introduction
Chapters 3 and 4 presented chemical, physical and chemorheological techniques useful
for characterizing various reactive polymer systems. This chapter will now focus on a
review of chemorheological analyses for a variety of polymer systems, including detailed
experimental ndings and chemorheological modeling.
5.2
5.2.1
Neat systems
Chemorheological models for neat (unlled) curing systems can be grouped into the
following categories:
352
Chemorheological modelling
g Kta ;
5:1
with a a1 for t < tc, a a2 for tc < t < tp, and a a3 for tp > t, where g is the viscosity,
K, a1, a2 and a3 are tting parameters, tc and tp are characteristic times and t is time.
For reacting polyurethane systems with gel effects Malkin and Kulichikin (1991) also
showed that the chemoviscosity could be described by
g=g0 1 kt=1 t=t
a ;
5:2
5:3
with a 1 for t < tc and a 3.5 for t > tc, where gm is the viscosity of the mediums, a and b
are constants, tc is a characteristic time and the sign changes as t > tc.
Lane and Khattack (1987) showed also for an epoxy resin that the chemoviscosity could
be described by
ln g ln g0 Ev =RT Ka;
5:4
5:5
where t1 is the time taken to reach a specic viscosity g1, C is a constant and E is the
activation energy for this reaction up to t1.
First-order isothermal (Kamal and Sourour, 1973) and non-isothermal (Dusi et al., 1982,
1983) chemoviscosity proles were presented for epoxy-resin systems in the forms:
g g0 expht
5:6
5:7
5:8
where a is the conversion, a* is the conversion at the gel point and A and B are constants.
Yang and Suspene (1991) and Malkin and Kulichikin (1991) showed that for melamine
formaldehyde and polyesterstyrene systems prone to microgel formation the following
chemoviscosity models were relevant:
g=g0 1 t=t
b ;
5:9
353
5:10
from Malkin and Kulichikin (1991), where A, B and b are constants and t* is the gel time.
Here distinct regions for microgelation, transition and macrogelation are observed and
modelled.
Arrhenius models
Dusi et al. (1987) examined the chemorheology of Fiberite976 resin (TGDDM/DDS). The
chemoviscosity is described by a simple Arrhenius relationship as shown by the following
equation:
g g1 expU=RT ka;
5:11
where g is the viscosity, g1 and k are constants, U is the activation energy for viscosity, R is
the universal gas constant, a is the conversion and T is the temperature.
Martin et al. (1989) and Kojima et al. (1986) showed the validity of the following
Arrhenius rst-order model for the chemoviscosity of epoxy-resin and polyamide systems:
Z t
ln gt; T ln g0 Ev =RT Kk
expEk =RTdt;
5:12
0
where g(t, T) is the viscosity as a function of time (t) and temperature (T), g0 is the zero-time
viscosity, Kk is the apparent rate constant, Ev and Ek are the activation energies for viscosity
and kinetics, respectively, and R is the gas constant.
A modied Arrenhius rst-order reaction is given by Dusi et al. (1983),
ln gt; T ln g0 f expEk =RTt;
5:13
5:15
354
Chemorheological modelling
There is reasonable agreement with the viscosity data except that some deviations occur
at long times. Use of this equation with mass- and heat-transfer equations for prepreg
production is useful for optimizing processing conditions.
Cheng et al. (1994) examine the chemorheology of a DGEBA/2,4,6-tris(dimethylaminomethyl)phenol (TDAP) system using the dual Arrhenius model
Z tc
ln gt; T ln g1 DEg =RTt
k11 expDEk1 =RTdt
0
5:17
Z t
expEk =RTdt;
5:18
where n is the order of the reaction and f is related to the degree of chain entanglement.
The results give a good t to an accelerated (by BF3-MEA) TGDDM/DDS system;
however, they do not predict results for the unaccelerated system well at longer times. The
introduction of the order of reaction greatly improved ts.
Roller (1986) reviews the rheology of curing thermosets, describing both isothermal
(Equation (5.19)) and non-isothermal (Equation (5.20)) models for chemoviscosity:
ln gt ln g1 Eg =RT tk1 expEk =RT;
Z
ln gt; T ln g1 Eg =RT
k1 expEk =RTdt:
5:19
5:20
355
5:21
where g is the viscosity, f is a friction factor (such as that derived by branching theory) and F
is a structure factor (such as the Mw or Mwa). For example, Berry and Fox (1968) showed that
f f0 expfB= fg a1 T Tg g;
5:22
where f0 is the initial friction term, fg is the friction factor at the glass-transition temperature,
Tg, and B and a1 are constants.
Vinogradov and Malkin (1977) showed that for linear polymer polymerizations
g kMwa ;
5:23
where a 1 for Mw < Mc and a 3.5 for Mw > Mc, k is a constant, Mw is the molecular
weight (molar mass) and Mc is a critical molar mass.
Richter and Macsoko (1980) describe the chemoviscosity of a reactive polyurethane
(4,40 -diphenyl methane diisocyanate, polyester triol and dibutyltin dilaurate catalyst) reasonably well by the following model:
g A expD=RTMw =Mw0 C=RTS ;
5:24
where A, D, C and S are constants, Mw is the molecular weight of the system and Mw0 is the
initial Mw.
Pichaud et al. (1999) highlight the chemorheology and dielectrics of the cure of DGEBA
with isophorone diamine (IPD). The kinetics are well described by an autocatalytic model
and the chemorheology is well described by the Macosko model
gT=g0 T agel =agel aABa ;
5:25
with g0(T) Ag exp[Eg/(RT)], where agel is the conversion at gelation, a is the conversion,
A and B are parameters, g0(T) is the viscosity of the uncured material at a 0, Ag is the
pre-exponential factor, Eg is the activation energy for viscosity and R is the gas constant.
Here dielectric data could also be related to kinetics and chemoviscosity via experimental
relationships.
Free-volume models
Chiou and Letton (1992) examine the chemorheology of a complex industrial resin, Hercules 3501-6, containing a major epoxide (TGDDM), two minor epoxides (alicyclic diepoxy
carboxylate and epoxy cresol novolac), a hardener (diaminodiphenyl sulfone, DDS) and a
Lewis-base catalyst (boron triuroride monoethylamine complex). They found that both the
conventional WLF model (Equation (5.26)) (Ferry, 1980) and a modied WLF model
(Equation (5.27) were appropriate to describe the chemorheology:
gT
ln
gTg
ln
gT
gTg0
C1 aT Tg a
;
C2 a T Tg a
5:26
C1 T Tg a
;
C2 a T Tg a
5:27
356
Chemorheological modelling
where g is the viscosity, g(Tg) is the viscosity at the glass-transition temperature Tg (or a
constant), g(Tg0) is the reference viscosity at Tg, C1 and C2 are parameters as a function of
conversion (a), and T is the temperature.
The modication represents the case in which an iso-free-volume assumption (or C1
constant) applies. This work also highlighted the use of three independent nth order reaction
rates for understanding the kinetics of the complex system.
Hesekamp and Pahl (1996) describe the chemoviscosity of an epoxyamine (DGEBA/
TMAB) adhesive using a modied WLF model shown as
b1 T Tg X
b1 T Tg X 0
gT; X
C1 T T0
;
5:28
ln
g
0
C2 T T0 b2 T Tg X b2 T Tg X 0
where b1 and b2 are new parameters. The rst term is the temperature dependence, the
second term is the crosslinking effect and the nal term is for normalization or to cancel out
the crosslinking effect in uncured systems. The isothermal and non-isothermal chemoviscosity results are well tted by this model.
Hou and Bai (1988, 1989) have investigated the chemorheology of Hercules 3501-6 and
found good agreement with a modied WLF chemoviscosity model. The equation is
summarized as
!
C1 T Tg
gT
ln
;
5:29
gTg
C2 T Tg
with Tg A(T)a(t) B(T) and a a(T, t).
Kenny (Kenny et al., 1989a, 1989b, Kenny and Opalicki, 1993) examined the kinetics
and chemorheology of epoxy-resin-based composite laminates for which rst the effects of
degree of cure and temperature on resin viscosity were characterized and then, given
thermal proles from kinetic, mass- and heat-transfer calculations, the viscosity proles
during processing were determined. Here they describe the changes in viscosity of their
TGDDMDDS system by
gT; a
Mw a 3:4 expC1 Tr Tg0 =C2 Tr Tg0
g
;
5:30
gT0
Mw0
expC1 T Tg a=C2 T Tg a
where g(T, a) is the viscosity at temperature T and conversion a, g(T0) is the viscosity at
reference temperature Tr, g is the ratio of the radius of gyration of a branched chain to that of
a linear chain at the same molecular weight (molar mass), Mw(a) is the molecular weight at
conversion a, Mw0 is the molecular weight at the initial T0, C1 and C2 are WLF constants,
Tg0 is the glass-transition temperature at T0 and Tg(a) is the glass transition temperature at a
conversion of a. This equation is shown to work well for the epoxy-resin laminate proles
examined under selected non-isothermal processing conditions.
Shear effects
Simple chemoviscosity models have been used to examine the effects of shear rate on the
chemoviscosity g g(c, T) as shown in Table 4.3. For example, Sundstrom and Burkett
(1981) and Hartley and Williams (1981) found that polyesters and polyurethanes exhibit
essentially Newtonian behaviour, whereas epoxies and phenolics show marked shear
thinning. The power-law model is the most extensively used shear-rate model for thermosets and has been used for unlled (Ryan and Kamal, 1976, Kascaval et al., 1993,
357
Riccardi and Vazquez, 1989) and lled (Ryan and Kamal, 1976, Knauder et al., 1991)
epoxy-resin systems. Sundstrom and Burkett (1981) also obtained a good t of the viscosity
of diallylphthalate to the Cross model. The viscosity of polyesters has been modelled by
Yang and Suspene (1991) using a Newtonian model. The WLF model has been used by Pahl
and Hesekamp (1993) for a moderately lled epoxy-resin system. Rydes (1993) showed
that the viscosity of DMC polyesters followed a power-law relationship at high shear rate.
5:31
where g is the viscosity, c0 is the shear rate, n is a power-law index, Eg is the activation
energy for ow and A is a parameter (depending on the cure).
Macosko (1985) describes the importance of models that combine chemical kinetics,
branching theory and composition and processing conditions (temperature, shear rates) in
chemorheological models for structureproperty relations.
Kamal (1974) summarizes the chemorheology data of various epoxy-resin moulding
compounds and describes the relationship in general as
g gc 0 ; T; a;
5:32
where g is the viscosity, c 0 is the shear rate, T is the temperature and a is the conversion.
Han and Lem (1983) reviewed the cure kinetics, chemorheology and properties of curing
polyester resins.
5.2.2
Filled systems
It is necessary, before we summarize the effects of llers on the chemoviscosity of reactive
systems, to highlight a review of the effects of llers on rheology in general. (A more
introductory discussion of the rheology of lled systems (suspensions) is given in Chapter 2.)
358
Chemorheological modelling
Hardness (mohs)
Aspect ratio
CAS number
Calcium carbonate
Coarse-ground
Fine-ground
Precipitated
Silica
Ground
Precipitated
Fumed
Gel
Kaolin
Air-classied
Water-washed
Calcined
Mica
Talc
Wollastonite
Feldspar/nepheline
2.7
2.53.5
Low
471-34-1
6.57
Low
763186-9
2.53
Moderate
1332-58-7
2.53
1
4.55.5
5.56.5
High
Moderate
High
Low
12001-26-2
14807-96-6
14567-51-2
68476-25-5/
37244-96-5
34
Low
Barium sulfate
Barytes
Blanc xe
Carbon black
Alumina trihydrate
Glass bre
Organic
Wood our
Organic bre
Nutshell our
Starch
Corncob
Rice hull
Beads
Glass
Glass, hollow
a
b
1040
110
0.72.0
2.0
1.92.2
2.2
1.3
2.6
130
0.020.01a
<0.1a
0.010.02a
2.8
2.8
2.9
2.6
0.210
0.210
0.83
2500
175
310
515
4.5
4.4
12.3
2.4
2.5
120
0.25
0.010.1a
0.420
0.7
0.71.6
1.4
1.2
1
40325b
30100b
325b
2150
100125b
2.5
0.2
5300
1200
2.53.5
Low
Moderate
High
13462-86-7
7727-43-7
1333-86-4
14762-49-3
Low
High
Low
High
Low
5.5
Silica and carbon black form aggregates 10100 times the size of the primary particles.
Mesh size.
Aranguren et al. (1992) examined the rheology of PDMS lled with untreated and silanetreated fused silica. It was found that the dynamic moduli increase with increasing concentration of silanol groups (as present on the untreated silica) on the surface, due to the
increase in interactions with the PDMS. Direct interactions and indirect entanglement
interactions are thought to be the types of interactions occurring.
Gupta and Seshadri (1986) discussed calculations of the maximum loading level of lled
liquids. They found that a geometric method for calculating the maximum packing fraction
works well for polydisperse spherical suspensions.
359
Name
Silane-based
c-Chloropropyl
trimethoxy silane
Vinyl triethoxy silane
Vinyl trimethoxy silane
Vinyl tris(b-methoxyethoxy silane)
c-Methacryloxy propyl
trimethoxy silane
b-(3,4-Epoxycylohexyl
ethyltrimethoxy silane)
c-Glycidoxypropyl
trimethoxy silane
c-Mercaptopropyl
trimethoxy silane
c-Aminopropyl triethoxy
silane
N-b-(Aminoethyl)-caminopropyl methoxy
silane
Chemical formula
Minimum
covered
Solubility Solvent for
area (m2/g) in water modication
ClC3H6Si(OCH3)3
394
None
CH2CHSi(OC2H5)3
CH2CHSi(OCH3)3
CH2CHSi(OC2H4OCH3)3
411
526
279
None
None
<5%
CH2CCH3COOC3H6Si(OCH3)3 316
None
318
None
Water/ethanol 3 : 1
CH2OCHCH2OC3H6Si
(OCH3)3
HSC3H6Si(OCH3)3
332
<5%
Water
399
None
NH2C3H6Si(OC2H5)3
354
<5%
Water
NH2C2H4NHC3H6Si(OCH3)3
353
<5%
Water
C2H4Si(OCH3)3
Name
Chemical formula
Titanate-based
O
CH3
CH3CHOTi[OCC17H35]3
CH3
C12H25]3
CH3CHOTi[OS
CH3CHOTi[OPOP(OC8H17)2]3
OH
CH3
[CH3CHO]4 Ti[P(OC8H17)2OH]2
(C8H17O)4 Ti[P(OC13H27)2OH]2
(CH2OCH2CH=CH2)2
Tetraisopropyl di(diocitylphosphito)titanate
CH3
Isopropyl tri(dioctylpyrophosphato)titanate
[C2H5CCH2O]4Ti[P(OC13H27)2OH]2
Ti[OPOP(OC8H17)2]2
OH
CH2O
Di(dioctylpyrophosphato)ethylene titanate
CH2O
CH2O
CO
Ti[OPOP(OC8H17)2]2
OH
360
Chemorheological modelling
Lepez et al. (1990) examined the rheology of glass-bead-lled HDPE and PS. They found
that a Cross model describes the viscosityshear-rate relationship, a Quemada model
describes the concentration dependence of the viscosity, and a compensation model applies
for the temperature dependence of the viscosity. This model is expressed as
g* x; f; T
1
P
5:33
361
Parthasarathy and Klingenberg (1995a, 1995b) examined the steady shear and microstructure of electro-rheological suspensions. They note there is a highly non-linear rheological response and that the non-linearity is related to rearrangement of structure rather
than gross deformation of structure.
Gahleitner et al. (1994) describe the rheology of talc-lled PP. They note the effects of
ller concentration, particle size and dispersion on linear viscoelastic properties.
Utracki and Fisa (1982) and Metzner (1985) review the rheology of (asymmetric) breand ake-lled plastics, noting the importance of the llerpolymer interface, llerller
interactions, ller concentration and ller-particle properties in determining rheological
phenomena such as yield-stress, normal-stress and viscosity proles (thixotropy and rheopexy, dilatancy and shear thinning).
Quemada (1978a, 1978b) examined the rheology and modelling of concentrated dispersions and described simple viscosity models that incorporate the effects of shear rate and
concentration of ller and separate effects of Brownian motion (or aggregation at low shear)
and particle orientation and deformation (at high shear). The ratio of structure-build-up and
-breakdown rates is an important parameter that is inuenced by the ratio of the shear rate to
the particle diffusion. A simple form of viscosity relation is given here:
2
1
gr 1
1 Kf ;
5:34
2
with K (k0 k1cP)(1 cp), where gr is the reduced viscosity, f is the particle concentration, k0 is the intrinsic viscosity at c0 0, k1 is the intrinsic viscosity at c0 1, P is a
shear-thinning parameter and c0 is the shear rate.
Rucker and Bike (1995) examined the rheological properties of silica-lled PMMA. They
showed the existence of a yield stress and a poor t of viscosity data to existing ller models.
Seymour et al. (1982) highlighted the rheology of linear and gelled random styrene
acrylate resins lled with carbon black for toner usage. The melt rheology was strongly
affected by the initial structure in the resin and by the dispersion of the carbon black.
Van der Werff and de Kruif (1989) examined the scaling of rheological properties of a
hard-sphere silica dispersion (sterically stable monodisperse silica in cyclohexane) with
particle size, volume fraction and shear rate. The shear-thinning behaviour was found to
scale with the Peclet number (Pe 6pgsa3c/(kBT), or the ratio of shear time to structurebuild-up time, where a is the particle radius, gs is the viscosity of the solution, c is the shear
rate, kB is the Boltzmann constant and T is the temperature) and at higher volume fractions
the shear thinning transition shifts to lower Pe.
Doraiswamy et al. (1991) developed a non-linear rheological model combining elastic,
viscous and yielding phenomena for lled polymers. The model predicts a modied Cox
Merz relationship for lled melts:
gc 0 g*c0 x;
5:35
where g(c0 ) is the steady viscosity as a function of the shear rate (c0 ) and g* is the dynamic
viscosity as a function of the maximum strain (c0) and frequency (x). The model gives a
good prediction of dynamic- and steady-shear properties of a highly silica-lled PE.
Bicerano et al. (1999) provide a simplied scaling viscosity model for particle dispersions
that states the importance of the shear conditions, the viscosity prole of the dispersing
uid, the particle volume fraction and the morphology of the ller in terms of its aspect ratio,
the length of the longest axis and the minimum radius of curvature induced by exibility.
362
Chemorheological modelling
363
Tungare et al. (1988) examined the chemorheology of an epoxy-resin (FR4) and glasscloth assembly via squeeze-ow geometry and a Newtonian-ow assumption. The results
tted well to the Arrhenius model and the sensitivity of the Arrhenius model parameters
is discussed.
Walberer and McHugh (2000) investigated the use of squeeze-ow rheometry to examine
the relaxation and elongation rheological behaviour of very highly lled glass-beadPDMS
and reactive-glass-beadPVA resins. The rheological data presented could not be collected
for such systems using conventional shear rheometers.
Boutlin et al. (1992) examined the chemorheology of highly lled epoxy-resin moulding
compounds. They found that the crossover method is not a good procedure and that the
independence of tan d is a better diagnostic for gelation prediction (as dened in Table 4.5).
Also they found that steady-shear results can disturb the gel structure and that dynamic
data are more reliable. Furthermore, they found that the Macosko model gives a better
prediction of the chemorheological data than does the double Arrhenius model (as dened
in Tables 4.2 and 4.4).
Chang et al. (1998) examined the chemorheology of a highly lled epoxy-resin moulding
compound. They found that a coupled power law and the Macosko model adequately
describe the chemorheology of the system. This equation is
ga; T; c0 expEa a=RTcna g0 a;
5:36
where
Ea a C1 C2 a C3 a2 C4 a3 ;
na C5 C6 a C7 a2 C8 a3 ;
g0 a expC9 C10 a C11 a2 C12 a3 :
The authors also note the application of the modied CoxMerz rule relating dynamic and
steady viscosities, g(c0 ) g*(cmx).
Han et al. (1997) examined the chemorheology of a highly lled epoxy-resin moulding
compound that is characterized by a modifed slit rheometer. Results show that a modied
CoxMerz rule relating dynamic and steady viscosities is established, g(c 0 ) g*(cmx).
Also the material was shown to exhibit a yield stress at low shear rates and power-law
behaviour at higher shear rates. The temperature dependence of the viscosity is well predicted by a WLF model, and the cure effects are described by the Macosko relation.
Kuroki et al. (1999) examined the chemorheology of an epoxy resin with 90 wt.% silica.
Gelation was predicted by independence of tan d, power-law behaviour in dynamic moduli
and GPC. A modied power-lawArrhenius chemorheology model tted the data well.
Petti (1994) showed the importance of silica-particle shape and maximum packing volume fractions on the chemoviscosity and spiral ow of highly lled epoxy-resin moulding
compounds. Ultrahigh concentrations of silica could be used using the correct ller.
Saeki and Kaneda (1988) examined the apparent mean viscosity, minimum viscosity and
apparent gel time of an epoxy-resin moulding compound during the transfer-moulding
operation by using an instrumented transfer mould.
Spoelstra et al. (1996) examined the chemorheology of a highly lled epoxyresin moulding compound and showed that the modied CoxMerz rule (as dened in
364
Chemorheological modelling
5:37
They also found that neither the crossover of dynamic moduli nor the independence of the
loss tangent was a good diagnostic to measure the gel point.
Baikerikar and Scranton (2000) examined novel photopolymerizable liquid encapsulants
(epoxy novolacvinyl ester) lled with silica llers. They found that the desired viscosity
reduction occurred due to a blending of large and small particles, and that there was an
optimum size distribution. A silane coating was also desirable, in order to reduce viscosity.
5:38
365
and crack deection (for larger particles) (Sue, 1991), interfacial effects on solubility and
dispersion (Sue et al., 1996) and intermediate/strong interfacial bonding (Nakamura et al.,
1986). However, although toughness was improved in CTBN and coreshell rubbermodied epoxies, this was typically accompanied at the cost of losses in modulus, glasstransition temperature and solvent resistance (Sue et al., 1996).
Ratna (2001) examined the cure and phase separation of carboxy-terminated poly(2-ethyl
hexyl acrylate) (CTPEHA) liquid modier in DGEBA/diethyltoluene diamine (DETD).
It was found that the CTPEHA liquid rubber causes a delay in the polymerization of the
epoxy-resin matrix due to chain extension during pre-reaction and the viscosity effect. The
Tg decreases with increasing rubber concentration.
Di Pasquale et al. (1997) examined the effects of addition of a thermoplastic amine
functionalized poly(arylene ether sulfone) on the cure of a DGEBA/DDS system. The
fracture toughness increased with addition of the reactive thermoplastic due to the effects of
the thermoplastic as a curing agent, an interfacial bonding agent, reducing the crosslink
density and inducing phase separation.
Girard-Reydet et al. (1998) examined the reaction-induced phase separation in polyetherimide (PEI)DGEBADDS or MCDEA systems via in situ SAXS, light-scattering and
TEM methods. The phase separation was dependent on the initial PEI concentration and the
ratio of phase separation to polymerization rates. For concentrations near the critical concentration, the system was put into the unstable region and phase separation proceeded by
spinodal demixing. This was independent of the polymerization rate. For off-critical concentrations, the homogeneous solution de-mixed slowly via a nucleation- and- growth
mechanism, and the cure temperature effects vitrication of the thermoplastic-rich phase
and hence controls morphologies.
Ho et al. (1996) examined polyol or polysiloxane thermoplastic polyurethanes (TPUs) as
modiers in cresolformaldehye novolac epoxy resins cured with phenolic novolac resin for
computer-chip encapsulation. A stable seaisland dispersion of TPU particles was achieved
by the epoxy ring-opening with isocyanate groups of the urethane prepolymer to form an
oxazolidone. The exural modulus was reduced by addition of TPU and also the Tg was
increased due to the rigid oxazolidone structure. Mayadunne et al. (1999) extended this
work to a series of phenol- and naphthol-based aralkyl epoxy resins.
Kim and Kim (1994, 1995) examined the cure and vitrication of ATBN-modied
DGEBA/TETA epoxy resins. The presence of the ATBN rubber had only a small effect on
kinetics up to 20 wt.%. Reaction injection-moulding experiments and simulations were also
conducted on this system.
Kim et al. (1993) examined the cure of a polyethersulfonedicyandiamideTGDDM
system and established a timetemperature-transformation cure diagram for the system.
During the early stage of curing the system was in the single-phase regime, and after an
induction period phase separation occurred by spinodal decomposition and the domain
spacing increased until gelation. Phase separation was characterized by the time variation of
light scattering. Gelation and vitrication were monitored via rheological methods.
Kim et al. (1999) examined the morphology and cure of semi-IPN epoxy resin or
dicyanatepolyimide/polysulfonecarbon-bre lms. Polyimide or polysulfone lms were
inserted into the curing epoxydicyanate monomers to form semi-IPNs with seaisland
morphology at the thermosetthermoplastic interface. The nal carbon-bre-thermoplasticdicyanate lms had fracture toughness three to ve times higher than that of unmodied
carbon-bre-dicyanate composites.
366
Chemorheological modelling
Martinez et al. (2000) examined the chemorheology and phase separation of polysulfonemodied DGEBADDM epoxy-resin mixtures. They found a delay in polymerization due
to dilution and viscosity effects. The nal morphologies are controlled by curing temperature
due to the effect of temperature on phase separation. The fracture toughness increased with
increasing immiscibility and was at its maximum for a bicontinuous morphology.
Poncet et al. (1999) monitored frequency-dependent dielectric measurements to examine
the phase-separation process in poly(2,6-dimethyl-1,4-phenylene ether) (PPE) in a
DGEBAMCDEA resin. Dielectric measurements measured the build up in Tg both in the
PPE-rich continuous phase and in the epoxy-rich occluded phases for 3060-wt.% PPE
mixtures. In the 30% PPE mixture, the rate of reaction of the thermoset phase is equivalent
to that of the neat system due to two opposing effects, namely a slower reaction rate due to
dilution and a low level of conversion at vitrication due to the presence of high-Tg PPE. In
the 60-wt.% mixture the dilution effect of the PPE has a large effect of decreasing the reaction
rate. The continuous thermoplastic-rich phase vitries rst, followed by the thermoset
occluded phase. The nal morphology (size of occluded particles and composition of continuous phase) is affected by kinetics, diffusion and viscosity during phase separation.
Woo and Hseih (1998) examined the design of micro-epoxy particles with a thermoplastic polymer shell. As cure progressed, the phenoxy component (polyhydroxyl ether of
bisphenol-A) was expelled from the epoxy-resin (TGDDM or DGEBADDS) phase during
phase separation. At later stages of cure, etherication occurred for high temperatures of cure
in the TGDDMDDSphenoxy systems, between the pendant hydroxyl group of the phenoxy
shell and the residual epoxide of the crosslinked epoxy core particles. This led to a chemical
link between the thermoplastic and the epoxy-resin. This phenomenon was not seen for
DGEBADDSphenoxy systems, possibly due to the lack of residual epoxide groups.
Bonnet et al. (1999a, 1999b) examined the reaction kinetics and rheological behaviour of
PEIDGEBAMCDEA systems through phase transitions. Initially systems are homogeneous, with no effects of PEI on the kinetics, other than dilution. At a conversion from 0.2 to
0.4, phase separation and a sudden increase in reaction rate occur. This is due to the
formation of an epoxyamine-rich phase at phase separation. The chemorheological
behaviour was highly dependent on the concentration of PEI: for low concentrations (lower
that 10%15%) phase separation induced a decrease in viscosity; for concentrations close to
the phase-inversion composition (20%) a bicontinous dependence of viscosity on frequency
was observed; and at PEI concentrations above 30 wt.% phase separation led to an increase
in viscosity. Chemorheological modelling based on miscible polymer blends, lled polymer
and chemorheological theories gave good agreement with experiments.
Kim and Char (2000) examine the rheology of a PESDGEBADDM system during
isothermal cure. They found a uctuation in viscosity at phase separation that could be
simulated by a two-phase suspension model that incorporated chemoviscosity effects.
Tillie et al. (1998) examined the effect of the bre/matrix interface on the cure of glassbre-lled epoxy-resin systems. They found that the introduction of a lower-Tg interphase
based on hydroxylated PDMS oligomers allowed an increase in toughness without a
reduction in modulus or Tg. This was due to a modication of the stress eld under load due
to the elastomeric interphase.
Pearson and Lee (1991) examined the effects of particle-size and particle-distribution
effects on rubber-toughened epoxy resins. They examined a variety of CTBN liquid rubbers
and a methacrylated butadiene styrene coreshell particle in a DGEBApiperidine system.
They found that the toughening mechanism for small particles was internal cavitation of the
367
rubber particles and subsequent formation of shear bands. Subtle toughening mechanisms
such as crack deection and particle bridging were noted for large particle sizes. No synergistic effects were seen for mixtures of particle sizes.
Qian et al. (1997) examined the fracture toughness of structural coreshell particles in an
epoxy-resin matrix. They focussed on varying the shell composition of polybutadiene-costyrene (PBS)-corepolymethylmethacrylate (PMMA)-shell particles by incorporating
acrylonitrile (AN) comonomer into the PMMA shell at various AN : MMA ratios and by
crosslinking of the shell. The degree of particle dispersion was controlled precisely by the
AN content and by crosslinking. Microclustered dispersion provided higher toughness than
did a uniform particle distribution.
Wilkinson et al. (1993) examined the effects of thermoplastic additives on the cure of
commercial bismaleimide. A range of thermoplastics based on polyether sulfones, polyether
phosphine oxide and polyimides were used in commercial Matrimid bismaleimide resin.
Incorporation of a low-Mw polyethersulfone thermoplastic brought about an increase in
fracture toughness without increasing viscosity. Reactive end groups contributed greatly to
the improvement in fracture toughness by improving linkages between continuous and
discrete phases, especially for phase-inverted structures. Improvements in fracture toughness in carbon-bre prepregs with the toughened bismaleimides were also shown.
Ving-Tung et al. (1996) examined the rheological monitoring of phase separation on a
high-Tg thermoplastic in a TGDDM/aromatic diamine epoxy resin. The chemorheological
behaviour is affected by the vitrication of the thermoplastic-rich phase, gelation of the
epoxy-resin-rich phase and the morphological effects of the occluded phase. It is shown that
complex viscosity is a sensitive measure of the onset of phase separation and also is
sensitive to the type of phase separation that occurs (that is related to the concentration of
the thermoplastic phase). Three types of morphology based on the three types of phase
separation are found, namely nucleation and growth for 510 wt.% of the thermoplastic
phase (the epoxy-resin-rich phase is the continuous phase), spinodal decomposition for
1220 of the wt.% thermoplastic phase (co-continuous phases) and nucleation and growth
with phase inversion for above 20 wt.% of the thermoplastic phase (the epoxy-resin phase
is the occluded phase); and the chemoviscosity proles are correlated with these phases.
The effects of shear strain have an effect on nal morphologies.
Vlassopoulos et al. (1998) examined the gelation of three epoxy-rubber thermoset blends
(based on TGDDM/DDS/(acrylonitrile/butadiene rubber/methacrylic acid copolymer) of the
same chemistry but different pre-cure treatments. The pre-treatments used heat and catalysts
to promote epoxycarboxyl reactions, and there was some evidence of a decrease in gelation
time and an effect on pre-gel rheology with these treatments.
More recently a large body of research has focussed on thermoplastic-modied thermoset
matrices (Pearson and Yee, 1992, Ohnaga et al., 1994, Hay et al., 1996, Girard-Reydet
et al., 1997, Hodgkin et al., 1998). Typically, high-performance high-temperature-resistant
thermoplastics such as polysulfones, polyphenylene ethers and polyimides are used.
Although a good improvement in toughness can be achieved in these systems without loss in
terms of other thermomechanical properties, the viscosity of the system increases (Hodgkin
et al., 1998) and thus the processibility decreases. Therefore, there exists a need to develop a
system that improves the toughness without deleterious effects on other properties such as
strength, thermal and solvent resistance and processibility.
There has been much work on characterizing the phase separation and network development for thermoplastic-modied epoxy-resin systems. Initial work (Yamanaka and Inoue,
368
Chemorheological modelling
Epoxyhyperbranched-polymer systems
Recently the development of dendritic and hyperbranched polymers (HBPs) has attracted
much attention (Tomalia, 1985, Newkome et al., 1985, Webster, 1991, Chu and Hawker,
1993, Wooley et al., 1994, Feast and Stanton, 1995, Malmstrom et al., 1995, Kim, 1998).
The key features of the macromolecular architecture of dendrimers and HBPs are given in
Section 1.2, and their synthesis by stepwise polymerization is discussed in Section 1.2.1.
Dendrimers and HBPs are globular macromolecules that have a highly branched structure
with multiple reactive chain ends (shell), which converge to a central focal point (core); see
Figure 5.1, where I is the core, II is the structure and III is the shell.
Although the terms dendrimer and HBP are often interchanged, they are different and can
have quite different properties (Frechet et al., 1996). Strictly dendrimers have a precise endgroup multiplicity and functionality, but are very tedious to synthesize, whereas HBPs
mimic (but do not achieve) regular dendritic growth, but are easily synthesized. Many HBPs
based on polystyrene, polyesters, polyethers, polyphenylene, polyamides and various
engineering polymers have been synthesized (Kim, 1998). Key attributes of HBPs are their
high potential for crosslinking and blending (due to the large number of reactive end groups)
369
III
II
and their low viscosities (lower than those of linear polymers of equivalent molar mass
(Wooley et al., 1994) due to the elimination of intermolecular entanglements). It is these
properties that are of interest regarding the use of HBPs as tougheners in epoxy-resin systems.
There have been very few reports on epoxy-hyperbranched toughened polymers, but what
has been done has shown both the enormous potential and the conicting nature of results
on these systems. Recent work (Boogh et al., 1999) has shown that epoxy-functionalized
HBPs can greatly improve toughness (a 250% increase in K1c) without affecting Youngs
modulus, the glass-transition temperature and processing viscosity of a model DGEBF/
isophorondiamine epoxy-resin matrix at very low levels of HBP (5 wt.%). In this study it
was noted that the shell chemistry and polarity affected the initial miscibility and kinetics of
phase separation, whilst subsequent cure conditions and phase separation controlled processibility and structure. Unfortunately, details of the shell chemistry and cure proles were
not disclosed, thus detailed relationships between phase separation and shell chemistry/cure
conditions could not be determined. However, other recent work (Wu et al., 1999) has
shown on the contrary that very slight improvements in toughness were obtained for
hydroxy- and acetyl-terminated HBPs added to DGEBA/DDS epoxy-resin systems in
comparison with the system that contained a linear analogue of the HBP. However, this
paper (Wu et al., 1999) does note that the use of an epoxy-terminated end group may
promote better interfacial adhesion and thus better toughness. It was also noted that better
control of phase separation (via varying cure conditions) may be necessary in order to
optimize the nal structure and properties.
It is postulated that the role of the HBP in toughening will be to act similarily to a core
shell particle; that is, the core of the HBP will act to cavitate and promote shear yielding,
and the shell will be able to be tailored to control aggregation and interactivity with the
epoxy-resin matrix. Increases in core Mw should promote cavitation, and shell-chemistry
functionalization should increase dissolution and reactivity with the epoxy resin. However,
unlike with the coreshell particles, the inherently greater number of shell sites and low
viscosity of the HBP will enable the toughening to occur without deleterious effects on
other properties.
370
Chemorheological modelling
Epoxynanocomposite systems
Epoxynanocomposite systems are relatively novel systems examined rst by Pinavaia and
Gianellis (Wang and Pinnavaia, 1994, Messersmith and Giannelis, 1994). A recent review
of epoxy nanocomposites is provided by Becker and Simon (2005). In terms of chemorheology of epoxynanocomposite systems, there are relatively few papers, however.
Le Pluart et al. (2004) and Becker et al. (2003a) showed the effects of nanoclays on
dynamic rheology and steady viscosity prior to cure and post-cure. There was a clear
increase in low-frequency elastic modulus and low shear viscosity before cure, and a clear
increase in fully cured mechanical properties post-cure with increasing clay content.
Becker et al. (2003b) examined the inuence of an organically modied clay on the
curing behavior of three epoxy-resin systems widely used in the aerospace industry and the
effects of various structures and functionalities. Diglycidyl ether of bisphenol-A (DGEBA),
triglycidyl p-amino phenol (TGAP) and tetraglycidyl diamino diphenylmethane (TGDDM)
were mixed with an octadecylammonium-ion-modied organoclay and cured with diethyltoluene diamine (DETDA). As the nanoller concentration increased, the gelation time
decreased. It was found that this was due to the fact that the rate of reaction is increased
by the addition of the organoclay.
Dean et al. (2005) examined epoxy layer silicate nanocomposites and showed that
the interlayer spacing increased with the temperature of cure, resulting in intercalated
morphologies with varying degrees of interlayer expansion, depending on the cure temperature used. Chemorheological studies of the curing process indicate that inter-gallery
diffusion and catalysis occur before curing and are essential for exfoliation, before the
morphology is frozen in by gelation and vitrication.
Sun et al. (2006) examined the use of novel silica nanollers in underll for ip-chip
applications, and showed that pre-cure rheology and post-cure values of Tg are effected by
nanosilica surface treatment.
Hadjistamov (1999) examined the effect of nanoscale silica on the rheology of silicone
oil and uncured epoxy-resin (araldite) systems. Shear thickening and yield-stress-like
behaviour were observed and found to be due to a build-up of network structure associated
with the nanocomposite phase.
5.2.3
5.3
References
371
References
Aranguren, M., Mora, E., Degroot, J. & Macosko, C. (1992) J. Rheology, 36, 11651182.
Argarwal, S. & Salovey, R. (1995) Polym. Eng. Sci., 35, 12411251.
Baikerikar, K. & Scranton, A. (2000) Polym. Composites, 21, 297304.
Barnes, H. (1989) J. Rheology, 33, 329366.
Bascom, W., Cottingham, R., Jones, R. & Peyser, P. (1975) J. Appl. Polym. Sci., 19, 2545.
Becker, O., Cheng, Y., Varley, R. & Simon, G. (2003b) Macromolecules, 36, 16161625.
Becker, O. & Simon, G. (2005) Adv. Polym. Sci., 179, 2982.
Becker, O., Simon, G., Varley, R. & Halley, P. (2003a) Polym. Eng. Sci., 43, 850862.
Berry, G. & Fox, T. (1968) Adv. Polym. Sci., 5, 261.
Bicerano, J., Douglas, J. & Brune, D. (1999) J. Mater. Sci. Rev. Macromol. Chem. Phys. C, 39,
561642.
Bidstrup, S. & Macosko, C. (1990) J. Polym. Sci. B: Polym. Phys., 28, 691709.
Boersma, W., Laven, J. & Stein, H. (1995) J. Rheology, 39, 841860.
Bonnet, A., Pascault, J., Sautereau, H. & Camberlin, Y. (1999b) Macromolecules, 32, 85248530.
Bonnet, A., Pascault, J., Sautereau, H., Taha, M. & Camberlin, Y. (1999a) Macromolecules, 32,
85178523.
Boogh, L., Pettersson, B. & Manson, J. (1999) Polymer, 40, 2249.
Boutin, L., Ajji, A. & Choplin, L. (1992) Proceedings of the XIth International Congress on
Rheology, Belgium, pp. 327329.
Castro, J. & Macosko, C. (1982) AIChE J., 28, 251260.
Chang, R., Lin, Y., Lin, F., Yang, W. & Hsu, C. (1998) Society of Plastic Engineers ANTEC
Proceedings ANTEC 1998, pp. 13971401.
Cheng, K., Chiu, W., Hsieh, K. & Ma, C. (1994) J. Mater. Sci., 29, 721727.
Chiou, P. & Letton, A. (1992) Polymer, 33, 3925.
Chu, F. & Hawker, C. (1993) Polym. Bull., 30, 265.
Clarke, N., McGleish, T. & Jenkins, S. (1995) Macromolecules, 28, 4650.
Couchman, P. (1978) Macromolecules, 11, 117119.
Dean, D., Walker, R., Theodore, M., Hampton, E. & Nyairo, E. (2005) Polymer, 46, 30143021.
De Miranda, M., Tomedi, C., Bica, C. & Samios, D. (1997) Polymer, 38, 10171020.
Di Pasquale, G., Motto, O., Recca, A. et al. (1997) Polymer, 38, 43454348.
Dickie, R. & Newman, S. (1973) J. Appl. Polym. Sci., 17, 65.
Doraiswamy, D., Mujumdar, A., Tsao, I. et al. (1991) J. Rheology, 35, 647685.
Dusi, M., Lee, W., Ciriscoli, P. & Springer, G. (1987) J. Comp. Mater., 21, 243261.
372
Chemorheological modelling
Dusi, M., May, C. & Seferis, J. (1982) ACS Org. Coat. Appl. Polym. Sci. Proc., 47, 635.
Dusi, M., May, C. & Seferis, J. (1983) ACS Symp. Series, 227, 301318.
Dutta, A. & Ryan, M. (1979) J. Appl. Polym. Sci., 24, 635.
Eley, R. (1983) ACS Symp. Series, 227, 281299.
Feast, W. & Stanton, N. (1995) J. Mater. Chem., 5, 405.
Ferry, J. (1980) Viscoelasticity of Polymers, New York: Wiley.
Frechet, J., Hawker, C., Gitsov, I. & Leon, J. (1996) J. Macromol. Sci. Pure Appl. Chem. A, 33, 1399.
Gahleitner, M., Bernreitner, K. & Neissl, W. (1994) J. Appl. Polym. Sci., 53, 283289.
Gandhi, K., Park, M., Sun, L. et al. (1990) J. Polym. Sci. B: Polym Phys., 28, 27072714.
Girard-Reydet, E., Sautereau, H., Pascualt, J. P. et al. (1998) Polymer, 39, 22692280.
Girard-Reydet, E., Vicard, V., Pascault, J. & Sautereau, H. (1997) J. Appl. Polym. Sci., 65, 2433.
Gupta, R. & Seshadri, S. (1986) J. Rheology, 30, 503508.
Hadjistamov, D. (1999) Appl. Rheology, SepOct, 212218.
Han, C. & Lee, D. (1987) J. Appl. Polym. Sci., 33, 28592876.
Han, C. & Lem, K. (1983a) J. Appl. Polym. Sci., 28, 743.
Han, C. & Lem, K. (1983b) J. Appl. Polym. Sci., 28, 763.
Han, C. & Lem, K. (1983c) J. Appl. Polym. Sci., 28, 3155.
Han, C. & Lem, K. (1984) Polym. Eng. and Sci., 24, 473.
Han, S., Wang, K., Hieber, C. & Cohen, C. (1997) J. Rheology, 41, 177195.
Hartley, M. & Williams, H. (1981) Polym. Eng. Sci., 21, 135.
Hay, J., Woodne, B. & Davies, M. (1996) High Perform. Polym., 8, 35.
Hesekamp, D. & Pahl, M. (1996) Rheologica Acta, 35, 321328.
Ho, T., Wang, J. & Wang, C. (1996) J. Appl. Polym. Sci., 60, 10971107.
Hodgkin, J., Simon, G. & Varley, R. (1998) Polym. Adv. Technol., 9, 3.
Hou, T. & Bai, J. (1988) Chemoviscosity Modelling for Thermoset Resins, Nasa contractor report
181 718.
Hou, T. & Bai, J. (1989) Chemoviscosity Modelling for Thermoset Resins, Nasa contractor report
181 807.
Inoue, T. (1995) Prog. Polym. Sci., 20, 119.
Jeyaseelan, R. & Giacomin, A. (1995) Polym. Gels Networks, 3, 117133.
Kalyon, D. & Yilmazer, U. (1990) In Collyer, A., Utracki L. A. (Eds.) Polymer Rheology and
Processing, London: Elsevier.
Kamal, M. (1974) Polym. Eng. Sci., 14, 231239.
Kamal, M. & Sourour, S. (1973) Polym. Eng. Sci., 13, 5964.
Kascaval, C., Musata, F. & Rosu, D. (1993) Angew. Makromol. Chem., 209, 157.
Kenny, J., Apicella, A. & Nicolais, L. (1989a) Polym. Eng. Sci., 29, 973983.
Kenny, J., Apicella, A. & Nicolais, L. (1989b) Makromol. Chem. Macromol. Symp., 25, 4554.
Kenny, J. & Opalicki, M. (1993) Makromol. Chem. Macromol. Symp., 68, 4156.
Kenny, J., Opalicki, M. & Molina, G. (1995) Society of Plastics Engineers Annual Proceedings
ANTEC 1995, pp. 27822789.
Kim, B., Chiba, T. & Inoue, T. (1993) Polymer, 34, 28092815.
Kim, D. & Kim, S. (1994) Polym. Eng. Sci., 34, 625631.
Kim, D. & Kim, S. (1995) Polym. Eng. Sci., 35, 564576.
Kim, H. & Char, K. (2000) Kor-Aust Rheol. J., 12, 7781.
Kim, Y. (1998) J. Polym. Sci. A: Polym. Chem., 36, 1685.
Kim, Y., Min, H. & Kim, S. (1999) Polymer Processing Society Meeting, Den Bosch (proceedings
on CD).
Knauder, E., Kubla, C. & Poll, D. (1991) Kunstoffe German Plastics, 81, 39.
Kogan, E., Kutseba, S. & Kulichikhin, V. (1988) Khimicheskie Volokna 3, 3637.
References
373
Kojima, C., Hushower, M. & Morris, V. (1986) Society of Plastics Engineers Annual Technical
Conference ANTEC 1986, p. 344.
Kunz-Douglas, S., Beaumont, P. & Ashby, M. (1981) J. Mater. Sci., 16, 2657.
Kuroki, M., Takahashi, H. & Ishimuro, Y. (1999) Nihon Reoroji Gakkaishi, 27, 235241.
Lane, J. & Khattack, R. (1987) Society of Plastics Engineers Annual Technical Conference ANTEC
1987, p. 982.
Lee, D. & Han, C. (1987) Polym. Eng. Sci., 27, 955963.
Lee, L., Marker, L. & Grifh, R. (1981) Polym. Composites, 2, 209.
Lem, K. & Han, C. (1983a) J. Appl. Polym. Sci., 28, 779.
Lem, K. & Han, C. (1983b) J. Appl. Polym. Sci., 28, 3185.
Lem, K. & Han, C. (1983c) J. Appl. Polym. Sci., 28, 3207.
Lepez, O., Chopin, L. & Tanguy, P. (1990) Polym. Eng. Sci., 30, 821828.
Le Pluart, L., Duchet, J., Sautereau, H., Halley, P. & Gerard, J. (2004) Appl. Clay Sci., 25, 207219.
Macosko, C. (1985) Brit. Polym. J., 17, 239245.
Macosko, C. & Miller, D. (1976) Macromolecules, 9, 199206.
Malkin, A. (1990) Adv. Polym. Sci., 96, 7097.
Malkin, A. & Kulichikin, S. (1991) Adv. Polym. Sci., 101, 218254.
Malmstrom, E., Johansson, M. & Hult, A. (1995) Macromolecules, 28, 1698.
Martin, G., Tungare, A., Fuller, B. & Gorto, J. (1989) Society of Plastics Engineers Annual
Technical Conference ANTEC 1989, p. 1079.
Martin, G., Tungare, A. & Gorto, J. (1989b) Polym. Eng. Sci., 29, 12791285.
Martinez, I., Martin, M., Eceiza, A., Oyanguren, P. & Mondragon, I. (2000) Polymer, 41, 10271035.
Maurer, F. (1990) Interphase effects on viscoelastic properties of polymer composites, in
Ishida, H. E. (Ed.) Controlled Interphases in Composite Materials, Amsterdam: Elsevier.
Mayadunne, R. T. A., Rizzardo, E. et al. (1999) Macromolecules, 32, 69776980.
McGee, S. (1982) Polym. Eng. Sci., 22, 484491.
Messersmith, P. & Giannelis, E. (1994) Chem. Mater., 6, 17191725.
Metzner, A. (1985) J. Rheology, 29, 739.
Mitsuishi, M. & Kawasaki, H. (1996) Fillers surface modication, in CRC Handbook on Polymer
Science, Boca Raton, FL: CRC Press.
Mussati, F. & Macosko, C. (1973) Polym. Eng. Sci., 13, 236.
Nakamura, Y., Tabata, H. & Suzuki, H. (1986) J. Appl. Polym. Sci., 32, 4865.
Nakamura, Y., Tabata, H. & Suzuki, H. (1987) J. Appl. Polym. Sci., 33, 885.
Newkome, G., Yao, Z., Baker, G. & Gupta, V. (1985) J. Org. Chem., 50, 2003.
Ng, H. & Manas-Zloczower, I. (1993) Polym. Eng. Sci., 33, 211216.
Ohnaga, T., Chen, W. & Inoue, T. (1994) Polymer, 35, 3774.
Pahl, M. & Hesekamp, D. (1993) Appl. Rheology, 70.
Parthasarathy, M. & Klingenberg, D. (1995a) Rheologica Acta, 34, 417429.
Parthasarathy, M. & Klingenberg, D. (1995b) Rheologica Acta, 34, 430439.
Pearson, R. & Lee, A. (1991) J. Mater. Sci., 26, 38283844.
Pearson, R. & Yee, A. (1989) J. Mater. Sci., 24, 2571.
Pearson, R. & Yee, A. (1992) Polymer, 34, 1379.
Peng, W. & Riedl, B. (1994) Polymer, 35, 12801286.
Petti, M. (1994) 7th International SAMPE Electronics Conference, pp. 526540.
Phillips, R., Brady, J. & Bossis, G. (1988) Phys. Fluids, 31, 34623472.
Pichaud, S., Duteurtre, X., Fit, A. et al. (1999) Polym. Int., 48, 12051218.
Poncet, S., Boiteux, G., Pascault, J. et al. (1999) Polymer, 40, 68116820.
Qian, J., Pearson, R., Dimone, V., Shaffer, O. & El-Aasser, M. (1997) Polymer, 38, 2130.
Quemada, D. (1978a) Rheologica Acta, 17, 632642.
374
Chemorheological modelling
Industrial technologies,
chemorheological modelling and
process modelling for processing
reactive polymers
6.1
Introduction
The aim of this chapter is to describe a range of industrial processing technologies for
reactive polymer systems, and specically to
. characterise the process and highlight important processing-quality-control tests, process
variables and typical systems used,
. highlight applications of chemorheology in the process and
. examine the use of chemorheology in modelling of the production process.
In this way we will be bringing the concepts and understanding from all subsequent chapters
into practical processing applications in order to aid acquisition of deeper understanding of
these processes.
6.2
Casting
6.2.1
6.2.2
376
Industrial technologies
Casting material
Mould
Figure 6.1. A schematic diagram of the casting process. Figure 8.1 (Lee and Neville, 1957)
Cost is typically reduced by minimizing the amount of resin used (either via incorporation
of llers or modiers, or by foaming). Viscosity is critical for casting operations, in which
requirements concerning ease of processibility and large loadings of llers need to be
optimized. Exotherms typically pose processing problems for large casting masses since the
cure reaction evolves >100 kJ/mol (Section 3.2.2), which may be reduced by incorporation
of llers, modiers and the correct curing agent at an optimum concentration. Shrinkage,
although not typically a problem with epoxy-resin systems, may be reduced by addition of
a ller or modier. Pot life (or the lifetime of resin in storage prior to use) can be extended
by addition of llers, but this will also increase cure times or involve higher cure temperatures, and thus must be optimized.
6.2.3
Typical systems
Typically epoxy resins are used for castings due to their low cost, low shrinkage, ease of
rapid curing and good dimensional stability. Cast-foamed epoxies are also used in the
aerospace industry, where light weight and high strength are required.
6.2.4
6:1a
k1 A1 expE1 =RT;
6:1b
377
6.2 Casting
Figure 6.2. Flow-front advancement in the casting process experimental and model prediction.
Reproduced from Mitani and Hamada (2005). Copyright (2005). Reprinted with permission of John
Wiley and Sons, Inc. Figure 13.
k2 A2 expE2 =RT;
0n
g Bc
d1 d2 a
agel
expTb =T
agel a
6:1c
6:2
where a1, a2, E1, E2, A1, A2, B, n, Tb, agel, d1 and d2 are material constants, k1 and k2 are
reaction rate constants, g is the viscosity, c 0 is the shear rate, T is the temperature and a is
the conversion. These empirical models predicted the kinetic and chemorheological data
well over the processing conditions to be examined (Mitani and Hamada, 2005), but are not
based on reaction mechanisms and are valid only insofar as they provide a t to the data and
have a predictive capability.
The model predicts ow-front advancement and temperature proles during nonisothermal and reactive casting ow, giving good agreement with experimental results.
Specically, this is shown in Figure 6.2, where the ow-front advancement predicted by the
simulation agrees well with the experimental verication.
Similar agreement between simulated and experimental temperature proles is also seen
in Mitani and Hamada (2005).
378
Industrial technologies
Figure 6.3. A schematic diagram of the potting process. Figure 8.1 (Lee and Neville, 1957)
Figure 6.4. A schematic diagram of the encapsulation process. Figure 8.1 (Lee and Neville, 1957)
reproduced with permission of the McGraw-Hill Companies.
6.3
6.3.1
379
Figure 6.5. A schematic diagram of the sealing process. Figure 8.1 (Lee and Neville, 1957)
6.3.2
6.3.3
Typical systems
Owing to their barrier resistance, durability and low viscosities, epoxy resins are widely
used in the potting industry in applications such as electrical coils for transformers and
electrical motors. Epoxy pottings are able to withstand extreme environments and have
shown resistance to fungal growth and environmental resistance in arctic, desert and tropical
conditions (Lee and Neville, 1957).
Lightweight epoxy-resin foams have found applications as sandwich laminates in the
aerospace industry, and can be produced by using chemical foaming agents (which liberate
gas as a reaction by-product) or physical foaming agents (use of which involves including
llers containing a gas such as nitrogen).
380
Industrial technologies
6.3.4
6.4
Thermoset extrusion
6.4.1
Extrusion
Process diagram and description
Extrusion involves mixing and forcing molten material via a rotating screw, which is heated
by external barrel heaters and via friction, through a die to impart nal prole shape. The
screw (shown in Figure 6.6) ts snugly (with sufcient clearance to allow rotation yet
minimize back ow) into a cylindrical barrel.
The polymer is fed into the feed hopper, which regulates ow into the screw. As the
screw rotates, feed is drawn into the feed section and ows in the feed zone in the helical
ow channel between the rotating screw and the barrel wall. Typically this channel depth
decreases in the compression zone, which promotes increasing pressure in the extruder to
promote mixing and expel air voids. In the nal metering zone the channel depth is constant
again, and the material is homogenized and supplied to the die zone at constant temperature
and pressure to ensure consistent production. The die zone imparts the cross-section to the
prole, screens out impurities and distributes the material evenly across the cross-section.
An interesting extension of thermoset extrusion is the use of reactive thermosetting
solvents to process intractable engineering polymers. Many engineering polymers have
very high glass-transition temperatures (200250 C) and thus require high processing
381
Typical systems
Typical systems are unsaturated polyesters, usually lled with mineral llers and/or bres.
Newer systems such as thermosetthermoplastics using reactive solvents for processing
high-performance engineering polymers are being researched.
Process modelling
There have not been many papers directly focussing on extrusion of thermosetting polymers, since extrusion is usually used since a primary mixing procedure or a conveying stage
to other processes such as thermoset injection moulding and other secondary shaping
processes. However, there have been papers focussing on modelling the extrusion of
thermosetthermoplastic materials. As mentioned above, one of the key drivers of this work
is the desire to utilize the thermoset phase as a reactive solvent to allow processing of
intractable high-performance polymers and also to control phase separation and possibly
phase inversion in order to achieve controlled thermosetthermoplastic morphologies.
Riccardi et al. (1996) focussed on the modelling of phase separation in reactive
thermosetsolventhigh-performance-polymer mixtures by characterizing the interaction
parameter as a function of conversion in a thermodynamic model. The reaction-induced
382
Industrial technologies
phase-separation process (Section 1.3.2) is caused by the increase in the average molar mass
of the polymer, superimposed on possible variations of the FloryHuggins interaction
parameter, , with conversion a. This secondary effect may favour mixing or de-mixing,
depending on whether the interaction parameter decreases or increases, respectively, with
conversion.
Girard-Reydet et al. (1998) highlight the use of small-angle X-ray scattering (SAXS),
light transmission (LT) and light scattering (LS) to model phase structures in thermoset
thermoplastic phase-separation processes, and characterize the phase separation in terms of
nucleation and growth or spinodal decomposition.
6.4.2
Pultrusion
Process diagram and description of process
Pultrusion is a continuous process involving pulling a collection of bres on a creel
system in the form of a roving, tow, mat or fabric through a resin bath (for impregnation)
and then through a heated die to cure the resin and impart a constant cross-section to the
product, Figure 6.7.
The die entrance is typically characterized by a tapered shape to remove the excess resin
and make the ingress of material easier. A water cooling channel is placed in the rst part of
the die, to prevent premature solidication. The heat for polymerization is provided by
heating platens placed on the top and bottom surfaces of the die. Outside the die, the cured
composite material is pulled by a continuous pulling system (caterpillar or reciprocating
pullers) and then travelling cut-off saws are programmed to cut the product into desired
lengths. Since the predominant orientation direction is longitudinal, typical products are
strong and stiff in tension and bending, but relatively poor in transverse properties (unless
mats or fabrics with transverse reinforcement are employed).
Typical systems
Pultrusion-grade resin matrices are available in a variety of systems such as polyester,
epoxy, vinyl ester and phenolic resins. Unsaturated polyester resins are most commonly
used because of the low heat input required with faster gelation than with other resin
systems, although more recently carbon/epoxy-resin and breglass/epoxy-resin composite
systems have found numerous applications in low-weight, high-strength aerospace and
industrial applications. Almost all types of reinforcing materials can be used, including
rovings, tows, mats, cloth and any hybrid of these. The most widely used reinforcing
material is glass bre, e.g. E-glass and S-glass bres.
To produce pultruded products with consistent and high quality, it is important to tailor
and control the pultrusion process. To achieve a uniform degree of cure in the cross-section
of a product, control of the temperature prole inside the pultrusion die is an essential
aspect. Also, to achieve consistent bre wetting, controlled ow and pressure build-up must
383
Resin
Control
Preform
Plates
Puller
Mechanism
Heated
Die
Cut-off
Saw
End
Product
Heating Zone 1
Heating Zone 2
Heating Zone 3
FIBRE
REINFORCEMENT
RESIN
IMPREGNATION
TANK
(a)
L1
Computational Outflow
Boundary
Rs
Rp
Fibre/Resin system
Intersection point
Liquid Resin Backflow
Computational Inflow Boundary
Preform Plate
(b)
Tapered Boundary
Top Boundary
Computational
Domain Outlet
Computational
Domain Inlet
Centreline
384
Industrial technologies
be achieved in the die. It is therefore important to develop process models to simulate these
key variables and the pultrusion process, and to optimize the process.
6:3
where a is the cure conversion, af is the nal conversion, n is a pseudo reaction order, I0 is
the initial concentration of initiator, I is the instantaneous concentration of initiator, f is the
initiator efciency, Ap is the pre-exponential factor, Ep is the activation energy for the
propagation reaction, R is the gas constant and T is the temperature. By combining this
model with an energy balance of the pultrusion process, temperature and conversion proles
during pultrusion are predicted.
Initial modelling on predicting velocity proles in pultrusion dies was carried out
by Gorthala et al. (1994). Here a two-dimensional mathematical model in cylindrical
co-ordinates with a control-volume-based nite-difference method was developed for
resin ow, cure and heat transfer associated with the pultrusion process. Raper et al.
(1999) and Gadam et al. (2000) highlight process models for pultrusion built on mass
balances and Darcys law for ow through porous media, to predict velocity and pressure
elds in pultrusion dies. The pressure rise in the die inlet contributes to a major extent by
enhancing bre wet-out and suppressing void formation in the manufactured composite.
In this work, permeability, bre porosity and the chemoviscosity of the reactive resin must
be coupled together to predict pressure and velocity proles. Interestingly, Gadam et al.
(2000) use an empirical chemoviscosity equation incorporating effects of temperature
and cure level:
gx g1 E=RTx kax;
6:4
where g(x) is the viscosity as a function of distance (x) in the die, g1, E, R and k are
constants, T is the temperature and a is the cure level. In this model increases in the pull
speed, bre volume fraction, viscosity and preform plate area ratio all resulted in increased
pressure rises, as expected from experimental results.
Three-dimensional analyses of heat transfer and cure in pultrusion of epoxy-resin composites have been examined by Chachad et al. (1995, 1996) and Liu et al. (2000). Carlone
et al. (2006) review nite-difference and nite-element process models used for predicting
heat transfer and cure in pultrusion. In this work they recommend the following empirical
nth-order cure model for predicting cure kinetics of epoxy-resin composites, which is then
coupled to the systems energy balance to predict thermal properties and cure conversion:
da=dt k0 expE=RT 1 an ;
6:5
where k0, E and n are constants, a is the cure conversion and T is the temperature. This
model assumes simple nth-order kinetics and does not incorporate factors such as autocatalysis and similar reaction-dependent rates that are well known for epoxy resins from
both thermal (Section 3.2.6) and spectroscopic (Sections 3.3.5 and 3.3.6) analysis. (Note
that in this analysis it is also presumed that the resin fully wets the bres and does not
ow during the process namely that is there is no chemoviscosity term in the analysis.)
385
1
0.9
Degree of cure
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
200
400
600
800
Distance from die entrance (mm)
1000
This simplied kinetic model has also been used in other pultrusion modelling studies by
Liu et al. (2000) and Valliappan et al. (1996).
Carlone et al. (2006) examined using nite-difference and nite-element discretization
methods to model the process with respect to cure and thermal properties, and found very
good agreement between these models and experimental data from the literature as found in
Figure 6.8.
On closer inspection, relatively more accurate values for the temperature peak and mean
degree of cure have been found using the nite-element model, which has provided an
overestimated value of the standard degree of cure. The nite-difference model seems
to provide underestimated values for all parameters considered; however, both analyses
produce good agreement with experimental values.
6.5
Reactive extrusion
6.5.1
386
Industrial technologies
Figure 6.9. Interactions during reactive extrusion. Reproduced from Figure 3 (Janssen, 1998).
Copyright (1994). Reprinted with permission of John Wiley and Sons, Inc.
387
grafting and the chemical processes were described in Section 1.2.4. Polymer grafting
via extrusion has the advantages of there being no use of solvents, continuous processing
and rapid preparation, and limitations arising from the high temperatures required,
intimate mixing and side reactions of degradation, crosslinking and discolouration.
The factors affecting polymer grafting are the mixing efciency in the extruder, freeradical concentration, backbone-chain chemistry, extrusion processing conditions
(geometry, pressure, temperature proles, residence times) and by-product removal.
Systems can be developed from monomerpolymer grafting or polymerpolymer
grafting. Typical monomers include styrene and acrylates, and the nal grafted polymers
are useful for grafted block copolymers, compatibilizers, adhesion promoters,
surface modiers, nucleating agents, coupling agents between llers and matrices and
dispersing agents.
Reactive compatibilization processes in reactive extrusion generally involve reduction of
interfacial tension, developing ne and stable morphological structure, stable reactions
across phase boundaries and reactions that occur rapidly enough during the extrusion
residence time. Typical systems are polymer alloys and blends that have morphologies
optimized via reactive compatibilization with added llers, interfacial agents, emulsiers,
coupling agents or block copolymers.
Controlled-rheology reactive extrusion typically involves developing high-melt-ow-rate
(low-viscosity) polymers through controlled-molar-mass breakdown or degradation
during extrusion. The chemistry of degradation reactions and their control are discussed in
Section 1.4. This controlled-viscosity degradation involves addition of free-radical-generating
agents (hydroperoxides or peroxides) or degradation catalysts, temperature-prole
control and shear control. In practice masterbatching has been used in production to
facilitate mixing of the peroxide (typically with an absorbant) with the polymer at high
levels, and this masterbatch is then subsequently added to the base polymer. Of course, to
maintain control over the viscosity drop, typically scavengers (such as nitroxyl radicals) are
also added to control degradation effects. (The range of stabilizer strategies that can be
used is given in Section 1.4.) Chain-branching chemicals have also been used to produce
long-chain branched polyolens. The most abundantly used systems for controlledrheology reactive extrusion are high-melt-ow polypropylene polymers and their blends
(Moad, 1999b).
6.5.2
388
Industrial technologies
6.5.3
Typical systems
A large range of polymers can be obtained by polymerization in reactive extrusion, and the
techniques involved and polymers produced include the following (Fink, 2005):
radical polymerization, poly(styrene), poly(butyl methacrylate);
ring-opening polymerization, poly(lactide);
anionic polymerization, poly(styrene), styrenebutadiene copolymer, polyamide-12;
metathesis polymerization, poly(octenylene);
grafting; poly(e-caprolactone)-grafted starch, starch and poly(acrylamide), poly(propylene)
wood-our composites, poly(styrene)-grafted starch.
In reactive polymerization it is important to control extrusion conditions (extruder proles,
temperature proles, residence times) as well as having a good understanding of the
polymerization kinetics, thermodynamics and properties (i.e. reaction rates, conversion
times, heats of polymerization, ceiling termperatures of the polymerization reactions, viscosity (Mw) changes). All of the processes discussed in Section 1.2 are brought to bear in the
environment of the extruder.
Moad (1999a, 1999b) reviewed the chemistry of polyolen graft polymerization. The
following classications were used:
free-radical-induced grafting of unsaturated monomers
end-functional polyolens via the ene reaction
hydrosilation
carbene insertion
transformation of pending functional groups (transesterifcation, alcoholysis)
Moad also notes that the most common grafting modications made to polyolens are via
maleic anhydride, maleate esters, styrene, maleimides, acrylates and their esters, and vinyl
silanes. Other polymer systems (Fink, 2005) that undergo grafting are polystyrene/maleic
anhydride (useful for PA6/PS blends), PVC/butylmethacrylate (for improved processibility), PET/nadic anhydride, starch/vinyl acetate and starch/methyl acrylate (for improved
water resistance).
Reactive compatibilization of blends and lled systems is common, and Table 6.1
highlights typical systems and their compatibilizers.
Here there are many compatibilized systems, which can be loosely grouped into the
following categories:
compatibilization by additives (such as inorganic llers, modied llers, modied
nanolays, modied rubber particles)
block copolymers
coupling agents (such as peroxides, diamines, epoxies, anhydrides)
ionomeric compatibilizers (as in PEVOH and PETG blends)
end-functional polymers (as produced via the iniferter (initiator-transfer-termination)
method, reversible chain termination and atom-transfer radical polymerization]
Common peroxide degradants and stabilizers used in controlled-rheology reactive extrusion
are shown in Table 17.2 and 17.8 of Fink (2005).
The predominant system for controlle-rheology reactive extrusion is polypropylene (PP)
with systems such as ultrahigh-melt-ow polypropylene, high-impact polypropylene blends
389
Compatibilizer
LDPE/PA6
LLDPE/PET
LDPE/starch
HDPE/HIPS
HDPE/PA12
PP/PS
PP/PA6
PA6/PPO
PBT/EVA
PC/PVDF
PE/wood our
PET/PA6
6.5.4
where Cc is the average conversion, Cc(t) is the conversion at time t (from batch kinetics)
and E(t) is the residence-time distribution (from mass and force balances on the extruder).
Chen et al. (1996) used the same process model as Equation (6.6) to model the conversion
390
Industrial technologies
System
Model
Outputs
Methyl methacrylate
Controlled degradation
of PP
Chemical modication
of polyethylene
terephthalate (PET)
Polymerization of
caprolactone
Polymerization of
caprolactone
Polymerization of
poly-e-caprolactone
Semsarzadeh et al.
(2004), Puaux et al.
(2006)
Poly(urethaneisocyanurate) (PUIR)
Polypropylene and
polystyrene reactiveextrusion systems
Polymerization model
Pressure, temperature,
(second-order thermal
residence time, lling
self-initiation) coupled
ratio, number-average,
with ow and energy
and weight-average
equations
molar masses (Mn and
Mw) along the extruder
length can be
calculated
One-dimensional global Temperature, pressure,
shear rate, viscosity,
model of polymer ow
residence time
in co-rotating, selfwiping, twin-screw
extruders, coupled with
degradation reaction
kinetics
Effects of stoichiometry,
Kinetic model and
temperature, rate of
chemorheological
shear, chemical
model coupled with
composition and molar
qualitative process
mass (Mw) on torque
model
changes, residual
carboxyl content and
insoluble content.
Review of one- and three- Predictions of mixing
dimensional process
mechanism, heat
models coupled with
generation and heat
kinetics and
loss are better from
chemorheological
three-dimensional
models
model
A one-dimensional
Predicts pressure, lling
dynamic model of a
ratio, temperature,
twin-screw reactive
molar conversion
extrusion
proles and residencetime distributions
under various operating
conditions
Reaction conversion and
Engineering model
linking residence-timeMw from extruder via
model, shear-effects
correlating specic
model and reaction
mechanical energy
conversion during
and Mw
extrusion
Coupled kinetic model
Conversion and
with axial dispersion
residence-time
residence-timedistributions are
distribution model.
predicted
Two-dimensional
Concentration proles at
isothermal singlevarious axial locations
screw process model
in the extruder are
combining kinetics
presented for various
391
System
Model
Outputs
Degradation of
polypropylene
values of power-law
index, ow rate and the
homogeneous and
heterogeneous reaction
rate parameters
Final degree of
Kinetic modeling of PP
degradation
degradation in reactive
extrusion
6.6
Moulding processes
6.6.1
Open-mould processes
Process diagram and description
Open-mould processes are thermoset moulding processes with open moulds in the male or
female form. In this process the mould is made from a pattern and the reinforced thermoset
is applied to the mould to reproduce the shape and nish of the mould. Typical examples
would be boat hulls made from breglass-reinforced gel coats. Figure 6.10 shows the basic
contact-moulding process.
Here the mould surface is polished and treated with mould-release agent to promote
release of the nished part from the mould. Then a thermoset resin (usually termed a gel
coat and typically a thermosetting polyester resin with pigments) is applied via brush or
spray to the mould surface, typically to thicknesses of about 0.51.0 mm. This coating
process controls surface features of the nal product. When the gel coating layer has set, a
layer of ne glass surfacing tissue is applied, followed by layers of chopped strand mat
(CSM). Additional layers of gel coat resin and woven bre mat can be applied sequentially, to ensure that the mat is wetted out and impregnated by the resin and allowed to gel
(but not cure fully) before the next layer is applied. When the nal layer is applied the
whole laminate is fully cured (typically at room temperature).
392
Industrial technologies
BACK SURFACE
GEL COAT
PRIME
SURFACE
393
Typical systems
Typical systems include room-temperature-curable polyester resins, epoxy resins and vinyl
ester resins, and glass-bre and mat reinforcement.
6.6.2
Resin-transfer moulding
Process diagram and description
In resin-transfer moulding (RTM) there is typically a two-part mould, which forms a
cavity corresponding to the nal product shape. In the process, as shown in Figure 6.11,
reinforcement, typically being a woven roving, cloth or mat, is placed (dry) in the female
cavity, and then the male half is placed into position.
The liquid resin is then injected into the mould, typically from the bottom mould, assisted
by pressure and vacuum to aid in providing an even resin distribution and achieving the
removal of voids. A key processing consideration here is that the resin ow should match
the wetting out and impregnation of the reinforcement.
394
Industrial technologies
VACUUM APPLIED
RISER TUBES
6:7
6:8
where a is the conversion, n is the reaction order, k is the reaction rate, A is the preexponential factor, Ea is the activation energy and T is the absolute temperature. The
chemorheological model (on extending from Macoskos equations, (5.8) and (5.25)) is given by
g
B expTb =T
1 g0 c0 =s
1n
ag =ag aC1 C2 a ;
6:9
6:10
395
k1 A1 expB1 =RT;
6:11
k2 A2 expB2 =RT;
6:12
where a is the conversion, T is the temperature, R, A1, A2, B1 and B2 are constants, and m
and n are reaction orders.
Process modelling of RTM polyesters was examined by Kenny et al. (1990), who
identied the following (auto-catalytic) cure kinetic and (empirical CastroMacosko (Castro
and Macosko 1982) model, as highlighted in (Equation 5.8)) chemorheological models for
use in their process simulation:
da=dt kaf an am ;
g Kg expEg =RTaf =af aABa ;
6:13
6:14
where a is the cure conversion, t is time, k is a constant, af is the nal conversion, m and n
are constants, Kg is the viscosity constant, Eg is the activation energy for viscosity, R is
the gas constant, T is temperature, and A and B are constants. In this work these models
are incorporated into a ow model to predict degree-of-cure proles and experimental
temperature proles in various moulds.
Of course, in RTM process modelling one must combine the above kinetic and
chemoviscosity models into mass, momentum and energy balances within a ow simulation. Specically, the momentum balance must combine any ows induced by
pressure and any ows into porous media (as characterized by Darcys law). Simple onedimensional RTM ow modelling and two- and three-dimensional RTM simulations
have been summarized by Rudd et al. (1997) and show the importance of kinetic,
rheological and permeability coefcients to the simulation of pressure-prole and owfront predictions.
6.6.3
396
Industrial technologies
CAVITY
Typical systems
Typical systems are polyester and phenolic resins lled with glass bres and mineral llers.
397
excellent papers by Han and Lem (Han and Lem, 1983a, 1983b, 1983c, Lem and Han,
1983a, 1983b, 1983c), who used the following expressions for the cure kinetics (autocatalytic equation) and chemorheology (shear-thickening) proles:
da=dt k1 k2 am 1 an ;
6:15
N1 Arn1 ;
6:16
where a is the cure conversion, k1 and k2 are rate constants, m and n are rate orders, N1 is the
normal-stress difference, A and n1 are power-law contants and r is the shear stress. Note that
these equations presume minimal cure effects on viscosity proles during shearing, and,
although effects of llers, additives and ller surface treatments of process ows were
examined, no formalization of a process model was made. However, in later work Han and
Lee (Han and Lee, 1987, Lee and Han, 1987) extended their chemorheological modelling
(as rst presented in Equation (5.29)) to incorporate cure effects by writing the following
WLF-based equation:
ln g ln gTg aT Tg =C1 T Tg ;
6:17
6:18
g g1 g0 g1 1 kc2 n1=2 ;
6:19
and
where ge is the elongation viscosity, e is the elongation rate, k is a relaxation time, g is the
shear viscosity, g1 is the innite-shear viscosity, g0 is the zero-shear-rate viscosity, c is
the shear rate and n is a power-law index. It is also assumed that g0, k and n obey
an Arrhenius relationship with temperature; however, it is assumed that there is no
cure during ow. In this work the predominant ow is shown to be biaxial elongational
ow when lubricated moulds are used. Non-isothermal and isothermal ows are modelled.
6.6.4
Transfer moulding
Process diagram and description
Transfer moulding is a variant of compression moulding in which the injection of the resin
is controlled by a transfer ram, as shown in Figure 6.13.
The process involves a mould cavity and a transfer cavity. The lled charge is initially
placed in the transfer cavity, and then heated until it has softened. Then pressure is applied
to the ram, causing the charge to ow through the transfer port to the mould cavity. Excess
charge is employed to allow a hold pressure to be applied to the mould cavity and to
accommodate sample shrinkage during cure. The pressure is maintained during cure and
released when the material has gelled.
398
Industrial technologies
TRANSFER
CAVITY
RAM
399
Plunger
Gate
Cavity
Heating
rods
uld
Mo ttom
bo
Leadframe with
wire-bonded parts
prior to transfer of the lled epoxy-resin encapsulant. The lled epoxy-resin is heated
(to about 8090 C) before being injected into the mould. This injection is conducted in such a
way that the viscosity is at a minimum value (e.g. the temperature is high enough to melt
the lled resin and reduce the viscosity due to thermal effects, but not high enough to induce
high viscosities from cure effects) such that minimal damage is done to the integrated circuitry.
6:20
where g is the viscosity, g1 is a constant, Eg is the activation energy of ow, R is the gas
constant, T is the temperature, c is the shear rate, n is a power-law index, agel is the gel
conversion and a is the conversion. This has also been highlighted in Table 4.4 and extends
from the general form of the Ryan model in Equation (5.31).
Nguyen et al. (1992, 1993) highlighted a full process model for transfer moulding of
highly lled epoxies in integrated-circuit encapsulation. This model used the following
kinetic and chemorheological models:
da=dt k1 k2 am1 1 am2 ;
6:21
400
Industrial technologies
6:22
where g(T, c, a) is the viscosity as a function of temperature (T), shear rate (c) and cure
conversion (a), g0(T ) is the zero-shear viscosity (which is a (usually Arrhenius-type)
function of temperature), s* is a critical shear stress, agel is the gel conversion, a is the
conversion and C1 and C2 are constants.
These semi-empirical kinetic and chemorheological models are combined with a niteelement ow model (incorporating mass, momentum and energy balances) to predict
ow and process problems such as wire sweep (breakage of the intricate integratedcircuit wiring in the transfer mould during processing due to high ow viscosities). A
typical process simulation of lling a multiple-cavity mould is shown in Figure 6.15
Turng and Wang (1993) also highlighted the simulation of microelectronics encapsulation using epoxy-resin moulding compounds using a semi-empirical kinetic model (as
previously used in Equation (6.21)) and the following chemorheological model:
g B expTb =Tagel =agel aC1 C2 a ;
6:23
where g is the viscosity, B and Tb are constants, agel is the gel conversion, a is the conversion and C1 and C2 are constants. Similarity to the Macosko model (Equations (5.8)
and (5.25)) is evident here.
6.6.5
401
streams are delivered, in correct proportions, to the mould. The runner system is also
designed to ensure that the two components impinge and mix in a series of turbulent
mixing chambers prior to ow into the mould cavity. The materials then cure rapidly at
room temperature
(
g
g0 aT bT 1 kc 0 2 n1 1=2
g0 a T b T
n1 1 c 0
6:24
6:25
where
aT exp
c1 T T0
c2 T T0
6:25a
Industrial technologies
(a)
107
20
pressure (Pa)
2.5
2
1.5
1
0.5
0
(b)
709
258
10
20
30 40
time (s)
50
60
70
60
70
106
4
pressure (Pa)
402
709
3
2
1
0
258
10
20
30 40
time (s)
50
Figure 6.17. Agreement between simulation and experiment for RIM mould lling of an insulator.
BT expaT bT ;
6:25b
g0 g0 a and
6:25c
n1 n1 a;
where a is the conversion, t is the time, Ea is the activation energy of cure, R is the gas
constant, T is the temperature, n is the reaction order, g is the viscosity, g0 is the zero-shear
viscosity, aT and bT are thermal parameters, k is a relaxation time, c 0 is the shear rate, n1 is
the power-law coefcient and C1, C2 and b are constants. These kinetic and chemorheological models are used in a ow-lling simulation to predict ow-front ow, velocity
proles and pressure proles for a reactive EVA copolymer. Here very good agreement with
simulation and experiment for mould lling of an insulator is found, and shown in terms of
pressure proles in Figure 6.17.
Macosko (1989) examined process modelling for RIM processes. Interestingly, the
kinetic models examined here are pseudo-mechanistic, like the following for polyurethane
RIM processing:
dNCO=dt A1 expE1 =TCa NCOb Hc
A2 expE2 =THd ;
6:26
where [NCO] is the isocyanate concentration, [C] is the catalyst concentration, [H] is the
hydrogen concentration, t is the time, A1, E1, A2 and E2 are energy parameters, and a, b, c
403
and d are constants. Macosko (1989) highlights the use of the following general
chemorheological models in a RIM mould-lling simulation:
g g0 ag =ag aC1 C2 a
6:27
g g0 Mw =Mwa0 ;
6:28
or
with g0 An expEg =RT; where g is the viscosity, Ag and Eg are parameters, R is the gas
constant, T is the temperature, Mw is the molecular weight (molar mass) at conversion a,
Mw0 is the original molecular weight, ag is the gel conversion and a is the conversion.
6.6.6
404
Industrial technologies
6:29
k Ae n1 ;
6:30
where g is the shear viscosity, k is the elongation viscosity and n is the power-law
index. Note that A and B are constants for isothermal ow or Arrhenius functions (i.e.
A A0 exp[H/(RT)] and B B0 exp[H/(RT)]) for non-isothermal ow. Interestingly
here, even though thermal and shear rates are considered, no effect of cure is considered.
Kamal and Ryan (1980) presented a review of early kinetic and chemorheological models
of thermoset materials for injection moulding. They highlight the following model:
da=dt k1 k2 am 1 an ;
g gT; c 0 ; a;
6:31
6:32
where a is the cure conversion, k1 and k2 are constants, m and n are reaction orders, g is the
chemoviscosity and g(T, c, a) is a generic expression for g as a function of temperature T,
shear rate c 0 and conversion a. Examples of simple ow lling models for injection
moulding were also presented.
Knauder et al. (1991) highlighted the following kinetic and chemorheolgical model for
injection-moulded epoxy resins:
da=dt k0 expEa =RT1 an ;
ln g ln g1 Eal =RT k1
R t*
0
6:33
1 an expEa2 =RTdt;
6:34
where a is the conversion, t is the time, t* is the end-of-reaction time, k0 is the rate constant,
Ea is the activation time, R is the gas constant, T is the temperature, n is the reaction order, g
is the viscosity and g1, Ea1, Ea2 and k1 are constants. Here the reaction is of nth order and
the chemoviscosity prole combines an Arrhenius model with a cure model (but no shear
rate effects are considered) as has been discussed in Section 5.2.1. The model provides local
and time-dependent temperature proles, shear rates and shear-stress and pressure proles
in the mould.
405
6:35
with Tg Tg(a), C1 C1(T) and C2 C2(T), where g is the viscosity, a is the conversion,
T is the temperature, gg is the viscosity at Tg, Tg is the glass-transition temperature at a, and
C1 and C2 are functions of temperature; or
ga; c 0 ; T g0 =1 g0 c 0 =s
1n ;
6:36
with n n(a) and s* s*(a), where g is the viscosity, a is the conversion, T is the
temperature, g0 is the zero shear viscosity, c0 is the shear rate, s* is a yield-stress term and
n is the Cross-model index.
6.6.7
6:37
with
g0 g1 expEg =RT ;
k k1 expEk =RT;
6:38
6:39
406
Industrial technologies
BLEEDER PACK
PREPREG PACK
CORK DAM
MEMBRANE
SEAL
FAN
MOULD
HEATERS
PRESSURE
VACUUM
where g is the viscosity, g0 is the zero-time viscosity, k is the viscosity rate constant, Eg is
the activation energy for viscosity and Ek is the activation energy for viscous rate rise. The
similarity to the Arrhenius models in Section 5.2.1 is evident.
Overall the model combines mass and force balances incorporating the chemorheological
model (above), considerations for viscous ow and ow through bre bundles, to simulate
loss of resin during prepreging, the bre volume fraction, the pressure distribution, temperature effects and part dimensions. Verication experiments show that the model predicts
experimental results to within 8%.
6.6.8
407
6:41
6.7
6.7.1
408
Industrial technologies
Figure 6.20. A schematic diagram of an internal rubber mixer. Reprinted from Figure 1
(Cherimisinoff, 1986). Copyright (1986), with permission from Elsevier.
There are two basic designs of rotors in internal mixers, non-intermeshing (Banbury type)
and intermeshing (intermix type). Intermeshing rotors generally provide more consistent heat
transfer, and thus are better for long mixing of heat-sensitive materials, although in general
rotor design has proven to be relatively unimportant to mixing efciency, given the large
elongation forces seen in mixers, irrespective of rotor design (Cherimisinoff, 1986).
Mixing mills are open two-roller mills consisting of two parallel horizontal rollers
rotating in opposite directions, that pull the ingredients through the gap (or nip) between the
rollers. The back roller usually rotates faster than the front roller, and most of the work is
done on the slower front roller. Rollers can be heated centrally.
Mixing extruders are automated-feed internal mixers in which the feed is conveyed from
the feed hopper into the mixing section (note that the extrusion section is principally used
for conveying, not mixing). The mixing section is similar to that of an internal mixer where
material is intensively sheared between the rotor and the wall.
409
Figure 6.21. Mixing-time zones in rubber mixing. Figure 1 (Campanelli et al., 2004). Reprinted
with permission of John Wiley and Sons, Inc.
Process modelling
Campanelli et al. (2004) discuss a model of rubber mixing in an internal mixer based on
kinetic, thermodynamic and rheological equations that is used to determine the extent of
dispersion, batch temperature and relative batch viscosity over time. The chemoviscosity
model is complex in that a model must be developed for the compaction zone, the incorporation zone and the size-reduction zone, as dened by the mixing-time zones in Figure 6.21.
For example a combined chemoviscosity model, combining the effects of temperature
and llers as shown in the following equation is useful for the incorporation zone:
g gref expAT Tref =B T Tref 1 A1 fB1 ;
6:42
6:43
where g is the viscosity, gi is the initial batch viscosity, ff is the fragment volume fraction,
fp is the parent-ller volume fraction, ff0, fp0, C1 and C2 are constants and kf and kp are
breakdown rate constants. Useful predictive model correlations to experimental mixing
temperatures and torques are provided for rubbercarbon-black mixtures.
Schwartz (2001) and Haberstroh and Linhart (2004) note that rubber mixing processes
are inherently extremely complex and use articial neural networks to evaluate rheometric
properties and process models for mixing of rubber compounds.
6.7.2
Rubber processing
Process diagram and description
Rubber extrusion is used to convert feed into a continuous nished product such as rods,
tubes or proles. Feed material is introduced into a hopper and conveyed forwards by a
rotating screw (which is driven by an electric motor). This is similar to the process shown in
Figure 6.18 for thermoset extrusion.
410
Industrial technologies
(a)
(b)
Feed
(c)
Coated
fabric
Slow
Feed
Feed
Fast
Fabric
Fabric
Slow
The external barrel is heated, which, together with frictional shear heating, softens the
feed material. The material has a reduced viscosity due to the process and is quite
deformable by the end of the screw where the die imparts the nal shape to the prole. The
screw has three sections the feed section (which conveys feed material and helps expel
voidage gases), the transition section (which heats and mixes the materials) and the compression section (which homogenizes the material and builds up the pressure necessary to
convey the material consistently through the die).
Rubber calendering involves two rollers rotating in opposite directions. The gap between
the rollers is the nip, through which the material is processed to form sheets or laminates.
Common calendering operations include sheeting, fractioning onto a fabric and double
coating (which are shown in Figure 6.22).
Sheeting is essentially the forming of sheet products via two- or three-roller calendering
and a controlled take-up reel. Frictioning involves wiping an elastomer into a textile substrate in a three-roller calendar system that operates at uneven roller speeds to promote
wiping. The coating operation is similar to fractioning except that the rollers are controlled
at even speeds and the elastomer is merely pressed against a substrate.
Rubber moulding processes are similar to thermoplastic and thermoset moulding processes,
except that generally the elastomer has a relatively high initial viscosity. Fortunately, most
elastomers have a shear-thinning viscosity prole with increasing shear rate, so they can be
processed at high shear rates. Compression moulding is done by pressing elastomeric feed
between two metal plates at shear rates in the range 110 s1 (Figure 6.23). Since this is a
relatively low shear rate, the viscosities are high and processing times are longer.
Transfer moulding is a variation of compression moulding, where by a rubber preform is
prepared, heated in the transfer chamber and injected via a plunger into a mould; refer to
Figures 6.24(a) and (b).
Injection moulding is completed by conveying the feed materials via a single-screw extruder
into a shot volume and then injecting the shot volume into the mould cavity (Figure 6.25).
Shear-rate ranges for injection moulding are 100010 000 s1, thus processing
viscosities are low and cycle times are reduced. Of course, higher pressures and more
complex control systems are encountered in injection-moulding processes, constituting a
more continuous processing method (continuous feed system) than compression or
transfer moulding.
411
FLASH GROOVE
LAND
TOP PLATE
SLOT
DOWEL
PINS
CAVITY PLATE
BOTTOM PLATE
MOULD CAVITY
RUBBER PREFORM
Figure 6.23. A schematic diagram of rubber compression moulding. Adapted from Figure 17
(Cherimisinoff, 1986). Copyright (1986), with permission from Elsevier.
Typical systems
Typical elastomer systems have been mentioned in the previous section, Section 6.7.1.
Process modelling
Isayev and Wan (1998) and Deng and Isayev (1991) examined the simulation of a naturalrubber injection-moulding process. A shear-rate- and temperature-dependent viscosity
function based on the modied Cross model was tted to experimentally measured viscosity
data, and a non-isothermal vulcanization model with non-isothermal induction time was
tted to the non-isothermal curing kinetic data obtained by differential scanning calorimetry
(DSC). These data were utilized in the simulation of the cavity-lling and curing stages to
predict pressure traces during lling and curing conversions.
412
Industrial technologies
PRESS HEAD
(a)
DEFLECTION
INSULATION
MOULD
SHOULDER
INSULATION
PLATENS
RAM
RAM BOLSTER
DEFLECTION
POST
GLAND
CYLINDER
(b)
PLUNGER
RUBBER PREFORM
TRANSFER POT
SPRUE
CAVITY PLATE
MOULD CAVITY
(c)
PLUNGER
TRANSFER POT
FLASH PAD
MOULDED RUBBER
PART
Figure 6.24. A schematic diagram of rubber transfer moulding. Adapted from Figures 19
and 20 (Cherimisinoff, 1986). Copyright (1986), with permission from Elsevier.
413
THROAT
RAM
NOZZLE
BARREL
MOULD
INJECTION CHAMBER
HEATING
FLUID
Figure 6.25. A schematic diagram of rubber injection moulding. Adopted from Figure 22
(Cherimisinoff, 1986). Copyright (1986), with permission from Elsevier.
Wan and Isayev (1996) examined a hybrid approach of control-volume nite-element and
nite-difference modelling of injection moulding of rubber compounds. The effect of vulcanization on viscosity and yield stress during cavity lling is reported. On comparing two
versions of the modied Cross viscosity models with and without the effect of cure the use
of a viscosity model that accounts for the cure was found to improve the accuracy of the cavitypressure-prediction models. When the modied Cross model was further extended to include
the yield stress and was implemented in the simulation program a signicant improvement in
the prediction of cavity pressure was obtained in the case of low injection speed.
Haberstroh et al. (2002) modelled the injection moulding of liquid silicone rubber via the
use of curing kinetics, ow models and the pressurevolumetemperature behaviour.
Nichetti (2003, 2004) examined the use of a cure model to predict mechanical properties
of isothermally vulcanized moulded items.
6.8
High-energy processing
6.8.1
Microwave processing
Process diagram and description
Clark and Sutton (1996) reviewed the theory of microwavematerial interactions, diagnostics, process control, sterilization, medical treatments, analytical characterization,
equipment design, modelling, dielectric-property measurements, materials synthesis,
sintering, drying, melting, curing, waste remediation, special microwave effects and
process scale-up for microwave processing. A signicant increase in research into
microwave processing of ceramics and polymers began in the 1980s, most probably as the
result of two factors: (a) many laboratories were equipped with inexpensive home
microwave ovens (operating at 2.45 GHz) that could easily be modied for high-temperature research; and (b) researchers were reporting successes and unusual effects with
microwave processing. As a consequence of the rst factor, most of the studies have been,
and still are being, performed at 2.45 GHz.
Microwaves (0.3300 GHz) lie between radiowave (RF) and infrared (IR) frequencies in
the electromagnetic radiation spectrum. Microwaves can be reected, absorbed and/or
transmitted by materials. Microwaves can interact with materials through either polarization
or conduction processes. Polarization involves the short-range displacement of charge through
414
Industrial technologies
the formation and rotation of electric dipoles, whereas conduction requires the long-range
(compared with rotation) transport of charge (Clark and Sutton, 1996). During interaction,
energy in the form of heat is generated in the material primarily through absorption. At room
temperature, many ceramics and polymers do not absorb appreciably at 2.45 GHz. However,
their absorption can be increased by increasing the temperature (e.g. once a material has been
heated to a critical temperature, Tc, microwave absorption becomes sufcient to cause selfheating), adding microwave-absorbing additives (e.g. SiC, carbon, binders), altering the
microstructure or defect structure, changing their form (e.g. bulk versus powder), or changing
the frequency of the incident radiation (Clark and Sutton, 1996).
Microwave processing has been used to process thermoset polymers and polymer
composites, including polyesters, polyurethanes polyimides and epoxies, and in most
studies it has been concluded that the curing speed is faster using microwave energy
(Clark and Sutton, 1996). The effects of continuous and pulsed microwave irradiation on
the polymerization rate and nal properties also have been studied, and it was demonstrated that, for certain epoxies, a pulsed microwave cure resulted in improvements in
mechanical properties, better temperature uniformity and a faster polymerization rate
(Thuillier and Jullien, 1989).
Wei et al. (1995) showed for an epoxyamine system that
increased reaction rates were observed in the microwave cure as compared with thermal
cure at the same temperature;
the primary-amineepoxy reactions are more dominant in the microwave cure than in the
thermal cure relative to the secondary-amineepoxy and etherication reactions; and
there is a difference in the crosslinked network structure between the microwave and
thermally cured DGEBA/DDS.
Microwave processing has also been suggested to be able to aid pultrusion and autoclave
processing (Clark and Sutton, 1996).
Process modelling
Hedreul et al. (1998) examined a model of the cure kinetics of a thermally and microwavecured rubber-modied epoxy-resin formulation. The phenomenological cure kinetic model
used was
da=dt k1 k2 am 1 an ;
6:44
415
k1 A1 expE1 =T;
k2 A2 expE2 =T;
where a is the cure conversion, t is the time, k1 and k2 are cure parameters, A1, A2, E1 and E2
are constants, m and n are reaction orders and T is temperature. Interestingly, the cure
kinetics from thermal and microwave processing were similar.
Liu et al. (2004, 2005) examined a three-dimensional non-linear coupled auto-catalytic
cure kinetic model and transient-heat-transfer model solved by nite-element methods to
simulate the microwave cure process for underll materials. Temperature and conversion
inside the underll during a microwave cure process were evaluated by solving the nonlinear anisotropic heat-conduction equation including internal heat generation produced by
exothermic chemical reactions.
Zhao et al. (2001) examined a model for computing the electromagnetic eld and powerdensity distribution in a cavity, and their effects on cure of a thin epoxy-resin layer during a
novel microwave rapid-prototyping process.
6.8.2
Ultraviolet processing
Process diagram and description
One of the most efcient methods for rapid generation of highly crosslinked polymers is by
exposing multifunctional monomers to UV radiation in the presence of a photoinitiator. A
liquid resin can be transformed within a fraction of a second into a solid polymer that is
very resistant to chemical, heat and mechanical treatments. A review of UV curing of
polymers is presented by Decker (1989) .
Ultroviolet curing is a process that transforms a multifunctional monomer and a photoinitiator (since most monomers do not produce initiating species with sufciently high yield
upon exposure, it is often necessary to introduce a photosensitive initiator that will make the
polymerization start at illumination) into a crosslinked macromolecule by a chain reaction
initiated by reactive species generated by UV irradiation. A unique advantage of UV
initiation, compared with thermal polymerization, is that very high concentrations of radical
species can be attained simply by increasing the light intensity, and the depth of cure can
be nely controlled through the photoinitiator concentration which directly affects the light
absorbance (Decker, 1994).
416
Industrial technologies
acrylate and methacrylate systems (which nd wide uses), epoxides (microcircuits) and
vinyl ether systems.
Process modelling
Ultraviolet processing has been restricted to lab-scale studies; however, many authors have
examined kinetic cure modelling of these systems during the development of new UV-cure
materials.
Lee and Cho (2005) developed a cure model for UV nano-imprinting. It was shown
via bre-optic FTIR techniques (as developed by Decker and Moussa (1990)) that
the degree of cure had an exponential relation to the UV-irradiation time, power and
temperature.
Cho and Hong (2005) used photodifferential scanning calorimetry to investigate the
photocuring kinetics of UV-initiated cationic photopolymerization of 1,4-cyclohexane
dimethanol divinyl ether (CHVE) monomer with and without a photosensitizer, 2,4diethylthioxanthone (DETX) in the presence of a diaryliodonium-salt photoinitiator.
Two kinetic parameters, the rate constant (k) and the order of the initiation reaction (m),
were determined for the CHVE system using an auto-catalytic kinetics model as shown in
the following equation:
da=dt kam 1 an ;
6:45
where a is the cure conversion, k is the cure constant, and m and n are reaction orders. The
photosensitized system gave much higher k and m values than did the non-photosensitized
one (due to the photosensitization effect).
Abadie et al. (2002) showed that the following auto-catalytic kinetic model was useful for
modelling the UV cure of epoxy-resin systems:
da=dt k1 k2 am 1 an ;
6:46
where a is the cure conversion, k1 and k2 are cure constants (and Arrhenius exponential
functions of temperature), and m and n are reaction orders.
6.8.3
Gamma-irradiation processing
Process diagram and description
Gamma (c) irradiation rst found major utilization in microbiocidal applications, owing to
its ability to inactivate many microorganisms. The application to polymer reactions, particularly grafting, requires knowledge of the free-radical chain reactions (Section 1.2.3) as
well as degradation reactions (Section 1.4) that can lead to loss of mechanical properties of
the substrate. Gamma radiation is made up of pulses of electromagnetic energy (c-ray
photons) from a radioactive source such as cobalt-60. Reactors have common features such
as shielding, a radioactive source and an efcient mechanism for inserting and removing the
material to be irradiated by the source.
417
curing temperature
the materials cure kinetics and
the materials chemoviscosity
Process modelling
Little process modelling has been developed due to the low industrial usage of c-irradiation
processing for reactive polymers. However, many researchers are examining cure models
for promising materials.
Alessi et al. (2005) examined the c-irradiation and electron-beam processing of an
epoxy-resin system in the presence of a photoinitiator. They showed that increasing the
irradiation dose frequency and photoinitiator concentration greatly increases the temperature reached by the samples. The increase in temperature of the system during
irradiation is related to the balance between the heat-evolution rate, due to both polymerization reaction and radiation absorption, and the rate of heat release towards the
environment. High dose rates and high photoinitiator concentrations increase the reaction
rate and the difference between the heat produced and the heat released to the environment (Alessi et al. (2005).
Nho et al. (2004) also examined c-irradiation and electron-beam irradiation of epoxyresin systems. They showed that the gel fraction of two epoxy resins increased markedly
with increasing dose, and that the gel fraction of the epoxy resin irradiated by c-irradiation
was higher than that of the epoxy resin irradiated by electron-beam irradiation (possibly due
to the greater depth of cure with c-irradiation).The degree of curing of epoxy resins
irradiated in nitrogen was higher than that of epoxy resins irradiated in air (due to inhibition
of cure by oxygen) for both electronic-beam and c-ray initiation.
Palmas et al. (2001) highlighted the c-irradiation curing and thermal ageing of
EPDM synthetic elastomers studied by both 13C high-resolution and 1H wideline solidstate NMR.
6.8.4
Electron-beam-irradiation processing
Process diagram and description of process
Electron accelerators are utilized in industrial processes such as crosslinking of polymers,
curing rubbers, curing paints and adhesives, sterilization of medical devices and pasteurization of foodstuffs.
The interaction of high-energy electrons and materials depends on the kinetic energy
of the electron and the composition of the irradiated materials, so effective operation is
conditioned by this interaction and knowledge of the effective penetration depth of the
electrons into the material. This penetration depth is the depth at which the absorbed dose
has fallen to an arbitrary level (ideally similar to that of the unirradiated surface). An
electron processing facility consists of
Tubing
Fluid lters
Protector caps
Hypodermic syringes
Syringe plungers
Tubing, drip chambers
Hypodermic syringes
Fluoroplastics
Polyamides (nylon)
PE
PP
Polystyrene
PVC
SANb
Acrylics
ABS
Cellulosics
Application examples
Closure-piercing devices,
roller clamps
Luer connectors, injection
sites
Ventilation lters
Polymer
Injection moulding
Injection moulding,
extrusion
Injection moulding
Injection moulding
Extrusion
Injection moulding
Extrusion
Various
Injection moulding
Injection moulding
Processing
Formulationspecic
Good
Formulationspecic
Good
Formulationspecic
Good
Good
Good
Good
Good
Stability up to 25 kGy
To 5000 kGy
To 10000 kGy
To 500 kGy
To 1000 kGy
To 200 kGy
To 1000 kGy
To 1000 kGy
Rigid, clear
Rigid
Discolours
Comments
419
an accelerator (which is rated for the materials thickness and density, and takes into
account optimizing of size, cost, dose distribution and efciency)
a product conveyor (for continuous, batch or bulk liquid/powder processing)
a radiation shield (designed with knowledge of the total radiation ux)
a ventilation system (for removal of ozone)
a safety system (for high voltage, pressures and radiation)
a control system
The advantages of electron-beam processing include that it can be carried out at room
temperature and on the nal product part, the fact that there is no need for catalysts or solvents
and exible control of radiation dose and application. Limitations are the high capital cost and
the limited penetration depth (in comparison with higher-energy c-radiation).
Typical systems
An example system is PVC and polyethylene wire and cabling irradiated to improve stresscracking resistance, abrasion resistance, high-temperature properties and ame retardance,
via controlled electron-beam crosslinking (Loan, 1977). Additionally electron-beam
crosslinking is utilized to impart memory into a polymer system, such as crosslinked PE
materials for heat-shrinkable lms and pipe applications (Baird, 1977). The control of
electron-beam processing has advanced the quality of cell size and shape of PE foams via
control of crosslink distribution (Paterson, 1984).
Process modelling
Cleland et al. (2003) reviewed the application of electron beams with various materials;
however, little process modelling has been developed due to the low industrial usage of
electron-beam processing for reactive polymers. However, many researchers are examining
cure models for promising materials.
Behr et al. (2006) showed the benets of electron-beam curing of glass-bre-reinforcedcomposite veneer specimens for dental applications. Under electron-beam cure the
reconstructions became stiffer and resisted higher load.
Chen et al. (2006) highlight a novel calorimetric method to examine cure during
electron-beam processing of epoxy resins via an energy-balance model and use of a novel
experimental calorimeter.
Das et al. (2005) examined the electron-beam processing of hydrogenated acrylonitrile
butadiene rubber sheets. The gel content and dynamic storage modulus increased with the
radiation dose.
Klosterman et al. (2003) examined the electron-beam curing of a dual-curing (the
monomer contains both acrylate and acetylene reactive groups) liquid-crystal (LC) resin. It
420
Industrial technologies
was shown that electron-beam curing in the LC phase allows tighter molecular packing and
more efcient polymerization, resulting in higher crosslink densities.
6.9
Novel processing
6.9.1
Stereolithography
Process diagram and description
Stereolithography builds plastic parts layer by layer by passing a laser beam (guided by a
three-dimensional image) on the surface of a pool of liquid photopolymer. This photopolymer quickly solidies where the laser beam strikes the surface of the liquid. Once one
layer is completely traced, the part is lowered a small distance into the pool, allowing new
liquid to cover the completed layer, and then a second layer is traced on top of the rst. The
layers bond and eventually form a complete, three-dimensional object after many such
layers have been formed. A schematic diagram of this process is shown in Figure 6.26.
From a material viewpoint, the process may be described as follows (Flach and
Chartoff, 1995a): the photosensitive resin is a mixture of a base resin (typically a
multifunctional acrylate oligomer), a crosslinker (monomeric multifunctional acrylate), a
diluent (monomeric acrylate), a photoinitiator and various additives (surfactants, stabilizers, etc.). The photoinitiator absorbs the UV light and undergoes a reaction leading to
the formation of free radicals. These free radicals initiate the polymerization reaction as
described in Section 1.2.3 and Table 1.5. As polymerization proceeds, the viscosity of the
mixture in the direct vicinity of the laser increases until gelation, then vitrication occurs
(as described in Section 1.2.4 for networks from addition polymerization) and the material
can be considered solid. The chemical reaction is quite fast, with a high degree of conversion to a crosslinked network being reached within 20 ms of exposure.
421
SCANNER
SYSTEM (D)
MOVABLE
TABLE (A)
SEALED
CHAMBER
PHOTOPOLYMER (C)
LASER
OBJECT BEING
FABRICATED
VAT (B)
Typical systems
Photopolymer systems are photocurable resins incorporating reactive liquid monomers,
photoinitiators, chemical modiers and llers. Typically stereolithography utilizes UV
radiation, so UV-curable systems are used. Free-radical-photopolymerizable acrylate systems were originally used; however, newer cationic epoxy-resin and vinyl ether systems
(based on iodinium- or sulfonium-salt cationic initiators) are now being utilized.
An interesting extension of this system is the Soliform system, which focusses on
developing rapid-manufacturing moulds via stereolithography for injection moulds, vacuum
422
Industrial technologies
moulds and cast moulds, via the use of acrylateurethane resins of inherently higher exural
modulus (Chua et al., 2003).
Process modelling
Very little chemorheological or process modelling has been completed for stereolithography, and most operation and optimization has been done on a trial-and-error
basis. However, there are a few key papers that aid the modelling of this process.
Flach and Chartoff (1995a, 1995b) developed a numerical model for laser-induced
photopolymerization that simulates important aspects of stereolithography. The model
considers irradiation, chemical reaction and heat transfer in a small zone of material
exposed to a stationary UV laser source, and generates outputs including spatial
and temporal variations in the conversion, depletion of photoinitiator and local variations
of temperature in and around the region irradiated by the laser light. Here the kinetic
model for conversion of monomer is given by
dM=dt kp MfIa =kt 0:5 ;
6:47
where [M] is the monomer concentration, kp is the propagation rate constant, kt is the
termination rate constant, Ia is the absorbed light and f is the quantum yield for initiation.
This is equivalent to the standard free-radical polymerization of Equation (1.51), in
Section 1.2.3. In this model there is assumed to be no material ow, and the temperature rise
is due to heat of polymerization alone. Results provided insights into effects of laser dwell
time, depth penetration of cure and the overall uniformity of polymer formed during the
stereolithography process. Bartolo (2006) more recently presented an extension to this work in
a phenomenological numerical model for the curing processes of thermosetting resins for
stereolithography. The cure model was developed for light- and/or thermally initiated systems
and incorporates inputs such as initiator concentration, temperature and light intensity in order
to predict output variables such as the diffusion effects after vitrication, the phenomenon of
unimolecular termination and shrinkage effects. The model was experimentally veried.
Tang et al. (2004) presented a cure model that captures transient, thermal and chemical
effects that are ignored in typical threshold-based models. This new model incorporates as
inputs photoinitiation rates, reaction rates, diffusion and temperature distributions, and is
able to determine the spatial and temporal distributions of monomer and polymer concentrations, molar masses, crosslink densities and degree of cure.
423
Advantages of this process include instantaneous curing of the whole of the cross-section,
minimization of shrinkage and internal stresses (due to the full cure of every layer) the
absence of any requirement for nal curing and the fact that there is no need for a support
structure. Since wax supports the structure. The disadvantages are that wax-cleaning
processes are required and the capital cost is high.
424
Industrial technologies
Process modelling
Very little chemorheological modelling has been completed for SGC processing. Gu et al.
(2001) examined the effects of the SGC processing parameters on dimensional accuracy
and mechanical properties. Four key outputs were dened, namely x- and y-dimensional
accuracy, impact strength and the elastic modulus of prototypes. Six input processing
parameters (layer thickness, rst exposure time, second exposure time, tray speed, cool-wax
temperature and cooling time) that had possible effects on the four outputs were identied.
Because of the interactions among the input parameters and the limited number of
experiments, neural networks were used to establish the relationships between the key
inputs and outputs. Key results were the following.
Layer thickness and initial exposure time were key parameters; they were highly
interactive and increasing them increased the modulus. The second exposure time and
cooling time also affect the modulus. Decreasing the cooling temperature increased
the modulus.
Layer thickness, initial exposure time and tray speed affected the, x-direction
dimensional accuracy (the other parameters have negligible effect on the
x-dimensional accuracy).
Tray speed and initial exposure time have the largest effect on the y-direction
dimensional accuracy (the other parameters have negligible effect on the
y-dimensional accuracy).
Tray speed, cooling temperature and cooling time have the greatest effect on the impact
strength (the other parameters have negligible effect on the impact strength).
6.9.2
Microlithography
Process diagram and description
In general there are two lithographic processes used in computer-circuit fabrication
photolithography and radiation (X-ray, laser, electron-beam and deep-UV (248- and
193-nm wavelength)) lithography. The principal difference is the radiation source and
wavelength, which in turn dene the feature size which can be achieved.
Circuit or mask fabrication has the following procedures.
1.. Resist deposition: deposition of the resist (material) lm from dilute-solution spinning
and baking onto a substrate.
2.. Pattern exposure: a pattern is generated by a mask or controlled radiation-beam steering,
where the exposed material is degraded, chemically modied or crosslinked.
3.. Pattern development: the pattern is developed by solvent interaction to form a positive
(irradiated material is removed) or negative (non-irradiated material is removed) and
then baked.
425
4.. Transfer of pattern to substrate: exposure of the patterned substrate to an etchant or gas
plasma allows features to be transferred to the substrate.
5.. Removal of resist: the resist is removed by oxygen ashing without degradation by
product formation.
Key properties of the material (resist) to be used in microlithography are that
the resist must be able to change form in response to the radiation
the resist must adhere well to the substrate and not swell or blister when in contact with
the solvent
the resist must withstand plasma treatment and ion bombardment
the resist must produce a uniform coating
In a chip-fabrication process, the substrate may undergo two or three deposition processes,
20 patterning processes, 10 etching procedures and further ion implantation and metallization procedures.
Typical systems
Typical resists include cyclized polyisoprene with a photosensitive crosslinking agent
(ex bisazide) used in many negative photoresists, novolac resins with diazoquinone
sensitizers and imidazole catalysts for positive photoresists, poly(oxystyrenes)
with photosensitizers for UV resists, polysilanes for UV and X-ray resists, and
polymethacrylates and methacrylatestyrenes for electron-beam resists (Clegg and
Collyer, 1991). Also note the more recent use of novolac/diazonaphthoquinone
photoresists for mid-UV resists for DRAM memory chips and chemically amplied
photoacid-catalysed hydroxystyrene and acrylic resists for deep-UV lithography
(Choudhury, 1997).
Process modelling
Cole et al. (2001) reviewed physical models for most stages of the microlithography
process. For the curing stage, they state that there has been no recent development of cure
models since the work of Dill (1975). Dill (1975) presented models that incorporate both
light absorption and photopolymer cure kinetics modelling, as shown by the following
equations:
a AMz; t B;
dM=dt Iz; tMz; tC;
6:48
6:49
426
Industrial technologies
where a is the absorption coefcient, A, B and C are constants, [M] is the concentration of
photosensitive material at position z and time t, and I is the energy at position z and time t.
One can see clearly that the absorption and cure are coupled.
Some work has been done to model microlithography processing (including broader
material-property considerations). Baek et al. (2004) highlight the importance of the
presence of a uniform liquid layer between the last objective lens and the photoresist, in
optical projection lithography. They also conclude that the refractive index of the liquid
must be known to and maintained within a few parts per million for optimal processing. Hu
and Tsai Ky (2006) highlight thermal modelling of microlithography to model the warpage
of wafers during processing, using temperature measurements and rst-principles thermal
modelling via a simple heat balance.
6.10
Real-time monitoring
The industrial application of the sensor technologies described earlier (Chapter 3) is the
basis for real-time monitoring of processing of reactive polymers. The critical parameters to
be monitored may be seen from an examination of the chemorheological process-control
equations described in the previous sections:
temperature (particularly the reaction exotherm)
viscosity (particularly gelation if a crosslinking system is concerned)
conversion (i.e. the extent of chemical reaction)
pressure (e.g. back pressure due to reaction by-products; premature gel formation etc.)
The changes in these parameters within the liquid polymer can be compared with the
values predicted by the particular chemorheological equation to determine whether the
process is operating within specications. The particular challenge in process operations
compared with laboratory and model studies is the decision regarding where to place the
sensors and how to use the information as a decision-making tool.
One of the rst consolidated approaches to autoclave curing of thermosetting and thermoplastic composite materials was that of Ciriscioli and Springer (1990). They showed that the
evolution of an expert system with sensors (for viscosity, pressure and temperature as well
as dimensional change) is the logical extension of the analytical model approach discussed
in this chapter. The analytical model, being based on heat- and mass-transfer equations is, of
course, an advance on the purely empirical approach based on experience of product quality
without any real understanding of the underlying chemical and physical changes occurring
during material processing. The expert system requires a set of rules that are developed
from the analytical model since they are able to predict the changes to reaction time for
each change of one of the processing parameters. In its most sophisticated embodiment, a
system with a chemorheological model and appropriate sensors can adjust the process cycle
to accommodate changes in the reactivity of starting materials as well as changes in
temperature and pressure that may be routinely monitored (Ciriscioli and Springer, 1990).
427
rheological properties, such as viscosity, and chemical properties, such as extent of reaction,
requires specialized sensors that must be able to withstand the electrically noisy and
aggressive environment of the production line.
428
Industrial technologies
Figure 6.28. A reactive extruder, showing the metal-at-die interface housing the real-time sensor
probes for NIR, Raman and ultrasound (US) sensing of conversion and component
concentration (Fischer et al., 2006). Photograph provided by professor D. Fischer, Leibniz Institute
of Polymer Research, Dresden.
might not be known for a particular commercial resin system, the system may be
calibrated using independent rheological measurements so that real-time viscosity
monitoring is possible. The continuous development of the technique has resulted in a
planar wafer-thin sensor that is able both to monitor and to control processing of highperformance composites of epoxy resins and polyimides (Kranbuehl et al., 1997). Of
particular interest is the comparison of the differences between using a standard cure cycle
for a high-temperature polyimide, PMR-15 (as provided by the manufacturer), and the use
of an intelligent system with criteria for changes in the cure temperature set by the
achievement of target dielectric properties (such as the change in slope of the complex
permittivity (e00 ) with cure time) as the decision point for temperature change. This
resulted in either extension or reduction in duration of various parts of the cure cycle due
to subtle differences in composition among batches of prepreg (Kranbuehl et al., 1997).
This example also illustrates the fact that dielectric sensors are providing not just viscosity
information but rather data on chemical changes as well as physical processes such as
sensor wet-out as the resin ows. As in all systems that are analogues rather than absolute
methods, the system must be trained in order to determine the optimum parameter for
control of the temperaturetime process cycle in the autoclave.
The use of these sensors is not restricted to high-performance composites, but can be
extended to on-line monitoring of resin-transfer moulding (McIlhagger et al., 2000) and
glasspolyester prepreg composite cure (Kim and Lee, 2002).
429
430
Industrial technologies
Table 6.4. Commercially available temperature probes that are based on optical phenomena and fibre-optics
System
Sensor construction
Luxtrona
Metricorb
Ipitek
Takaoka
FISO
Those based on the uorescence lifetime, rather than intensity (e.g. the Ipitek system), allow
one to avoid problems of light loss and other factors that could affect calibration, as discussed
above. Assessment of tilted Bragg gratings and long-period gratings on optical bres has shown
them to be a probe of cure of the resin as well as being both temperature- and strain-sensitive
(Buggy et al., 2007). The complexity of the response of these and bre-optics based Fabry
Perot interferometers to strain, temperature and refractive index makes it necessary to employ
combinations of sensors if measurements of all of these properties are required separately.
References
431
References
Abadie, M., Chia, N. & Boey, F. (2002) J. Appl. Polym. Sci., 86, 15871591.
Alessi, S., Calderaro, E., Fuochi, P., Lavalle, A., Corda, U., Dispenza, C. & Spadaro, G. (2005)
Nucl. Instrum. Methods Phys. Res. B, 236, 5560.
Baek, S., Cole, D., Rothschild, M., Switkes, M., Yeung, M. & Barouch, E. (2004)
J. Microlithography Microfabrication Microsystems, 3, 5260.
Baird, W. (1977) Radiat. Phys. Chem., 9, 225.
Barone, M. & Caulk, D. (1985) Polym. Composites, 6, 105.
Bartolo, P. (2006) CIRP Annals Manufacturing Technol., 55, 221225.
Behr, M., Rosentritt, M., Dummler, F. & Handel, G. (2006) J. Oral Rehabilitation, 33, 447451.
Berzin, F., Vergnes, B., Canevarolo, S., Machado, A. & Covas, J. (2006) J. Appl. Polym, Sci.,
99, 20822090.
Blest, D., Duffy, B., McKee, S. & Zulkie, A. (1999) Composites A, 30, 12891309.
Bucknall, C. & Partridge, I. (1986) Polym. Eng. Sci., 26.
Buggy, S. J., Chehura, E., James, S. W. & Tatam, R. P. (2007) J. Opt. A: Pure Appl. Opt.,
9, S60S65.
Bunyawanichakul, P., Castanie, B. & Barrau, J. (2005) Appl. Composite Mater., 12, 177191.
Campanelli, J., Gurer, C., Rose, T. & Varner, J. (2004) Polym. Eng. Sci., 44, 12471257.
Carlone, P., Palazzo, G. & Pasquino, R. (2006) Math. Computer Modelling, 44, 701709.
Casalini, R., Corezzi, S., Livi, A., Levita, G. & Rolla, P. A. (1997) J. Appl. Polym. Sci., 65, 1725.
Castro, J. (1992) Polym. Eng. Sci., 32, 715.
Castro, J. & Macosko, C. (1982) AIChE J., 28, 251260.
Chachad, Y., Roux, J. & Vaughan, J. (1995) J. Reinforced Plastics and Composites, 14, 495512.
Chachad, Y., Roux, J., Vaughan, J. & Arafa, E. (1996) Composites, 27, 201210.
Chen, J., Johnston, A., Petrescue, L. & Hojjati, M. (2006) Radiation Phys. Chem., 75, 336349.
Chen, L., Hu, G. & Lindt, J. (1996) Int. Polym. Processing XI, 329.
Cherimisinoff, N. (1986) V9-Unit operations in rubber processing, in Encyclopedia of Fluid
Mechanics, Houston, TX: Gulf Publishing Co.
Cho, J. & Hong, J. (2005) J. Appl. Polym. Sci., 97, 13451351.
Choudhury, R. (1997) Handbook of Microlithography, Micromachining and Microfabrication,
Washington: SPIE.
Choulak, S., Couenne, F., Le Gorrec, Y. et al. (2004) Indust. Eng. Chem. Res., 43, 73737382.
Chua, C., Leong, F. & Lim, C. (2003) Rapid Prototyping Principles and Applications,
Singapore: World Scientic.
Ciriscioli, P. R. & Springer, G. S. (1990) Smart Autoclave Cure of Composites, Lancaster,
PA: Technomic.
Clark, D. & Sutton, W. (1996) Ann. Rev. Mater. Sci., 26, 299331.
Clegg, D. & Collyer, A. (1991) Irradiation Effects on polymers, London: Elsevier.
Cleland, M., Parks, L. & Cheng, S. (2003) Nucl. Instrum. Methods Phys. Res. B, 208, 6673.
Cole, D., Barouch, E. & Conrad, E. (2001) Proc. IEEE, 89, 11941213.
Das, P., Ganguly, A. & Banerji, M. (2005) J. Appl. Polym. Sci., 97, 648651.
Decker, C. (1989) Rubber processing, in Cheremisinoff, N. E. (Ed.) Handbook of Polymer Science
and Technology, New York: Marcel Dekker.
Decker, C.(1994) Acta Polymica, 45, 333347.
Decker, C. & Moussa, K. (1990) J. Coatings Technol., 62, 5561.
Degraff, R., Rohde, M. & Janssen, L. (1997) Chem. Eng. Sci., 52, 4345.
Deng, J. & Isayev, A. (1991) Rubber Chem. Technolo., 64, 296324.
Dhavalikar, R. & Xanthos, M. (2003) J. Appl. Polym. Sci., 87, 643652.
432
Industrial technologies
References
433
434
Industrial technologies
Turng, L. & Wang, V. (1993) J. Reinforced Plastics and Composites, 12, 506.
Valliappan, M., Roux, J., Vaughan, J. & Arafat, E. (1996) Composites B, 27, 19.
Venderbosch, R., Meijer, H. E. H. & Lemstra, P. J. (1994) Polymer, 35, 4349.
Venderbosch, R., Meijer, H. & Lemstra, P. (1995) Polymer, 36, 1167.
Vergnes, B. & Berzin, F. (2004) Plastics Rubber Composites, 33, 409415.
Vergnes, B. Della Valle, G. & Delamare, L. (1998) Polym. Eng. Sci., 38, 17811792.
Wan, M. & Isayev, A. (1996) Rubber Chem. Technol., 69, 294312.
Wei, J., Hawle, M. & Demeuse, M. (1995) Polym. Eng. Sci., 35, 46166.
Zagal, A., Vivaldo-Lima, E. & Manero, O. (2005) Indust. Eng. Chem. Res., 44 98059817.
Zhang, B., Wang, D., Du, S. & Song, Y. (1999) Smart Mater. Structures, 8, 515518.
Zhao, H., Turner, I., Yarlagadda, P. & Berg, K. (2001) Int. J. Adv. Manufacturing Technol.,
17, 916927.
Zhu, L., Narh, K. & Hyun, K. (2005) Adv. Polym. Technol., 24, 183193.
A
ABS
AFM
AIBN
ATBN
ATR
ATRP
BHT
C
c*
c
c
Cp
CTBN
CTPEHA
CTT
DB
DCP
DDS
DETDA
Df
DGEBA
DGEBF
DMA
DMA
DMC
DMTA
DP
DSC
DTA
DVB
Ea
EGDMA
EPDM
EPR
ESR
Ev
436
Glossary
EVA
Ex
f
f
FSD
FT-NIR
G
G0
G00
gel Tg
GPC
H
HBP
HDPE
HIPS
HPLC
ICL
IPN
iso-Tg-TTT diagram
k
kB
K
LAOS
LCST
LDPE
LLDPE
LS
M
M0 (g/mol)
MA
MALDI-MS
MCDEA
Mw
Mn
Mv
Mz
MIR
MPDA
MY721
N
n0
NIR
NMR
13
C NMR
OIT
p
p
poly(ethylene-co-vinyl acetate)
activation energy of reaction x (e.g. propagation (p))
free-radical escape efciency
functionality
Fourier self-deconvolution
Fourier-transform near-infrared spectroscopy
Gibbs free energy
storage (solid-like) modulus
loss (viscous-like) modulus
the temperature at which gelation and vitrication coincide
gel permeation chromatography (a.k.a. SEC)
enthalpy
hyperbranched polymer
high-density poly(ethylene)
high-impact polystyrene
high-performance liquid chromatography
the intensity of emission of chemiluminescence
interpenetrating network
TTT diagram with lines of constant Tg
kinetic rate coefcient
Boltzmann constant
equilibrium constant
large-amplitude oscillatory shear
lower critical solution temperature
low-density poly(ethylene)
linear low-density poly(ethylene)
light scattering
the ratio of viscosities of the droplet to the viscosity of the solution
molar mass of the monomeric repeat unit
maleic anhydride
matrix-assisted laser desorption ionisation mass spectrometry
4,40 -methylene bis[3-chloro 2,6-diethylaniline]
weight-average molecular weight
number-average molecular weight
viscosity-average molecular weight
z-average molecular weight
mid-infrared spectroscopy
meta-phenylene diamine
Ciba TGDDM resin
normal stress
refractive index
near-infrared spectroscopy
nuclear magnetic resonance spectroscopy
carbon-13 nuclear magnetic resonance spectroscopy
oxidation induction time
degree of polymerization/extent of reaction
fraction of bonds
Glossary
PALS
PAN
PB
PBAN
PBT
PC
pc
PCA
PCR
PDI Mw/Mn
PE
PEI
PEK
PES
PET
PHB
PLA
PLS
PMMA
POOH
PP
PPE
PRESS
PS
PU
PVAc
PVC
Rrms
R
RAFT
REX
RTM
Rg
RIM
Ro
rp
S
SAN
SANS
SAXS
SEBS
SEC
SEM
SLA
SLS
SMCR
SNR
437
positron-annihilation spectroscopy
poly(acrylonitrile)
poly(butadiene)
poly(butyl acrylonitrile)
poly(butylene terephthalate)
poly(carbonate)
percolation threshold
principal-component analysis
principal-component regression
polydispersity index
poly(ethylene)
poly(ether imide)
poly(ether ketone)
poly(ether sulfone)
poly(ethylene terephthalate)
poly(hydroxy butyrate)
poly(lactic acid)
partial least squares
poly(methyl methacrylate)
polymer hydroperoxide
poly(propylene)
poly(2,6-dimethyl-1,4-phenylene ether)
predicted residual sum of squares
poly(styrene)
poly(urethane)
poly(vinyl acetate)
poly(vinyl chloride)
root mean separation of polymer ends
gas constant
reversible additionfragmentation chain-transfer polymerization
reactive extrusion
resin transfer moulding
radius of gyration
reaction injection moulding
actual chain end to end distance
rate of polymerization
entropy
poly(styrene-co-acrylonitrile)
small-angle neutron scattering
small-angle X-ray scattering
poly(styrene-co-ethylene-b-poly(butene-co-styrene))
size-exclusion chromatography (a.k.a. GPC)
scanning electron microscopy
stereolithography
selective laser sintering
self-modelling curve resolution
signal-to-noise ratio
438
Glossary
T
tan d
TBA
TETA
Tc
Tc
Tc
TEM
Tg
Tg0
Tg1
TGA
TGAP
TDAP
TGDDM
tgel
Tm
TMA
TMA
TMAB
TPU
Trxn
TTT
UCST
UHMWPE
Ult Tg
Vf
Vs
Vs
WAXD
WLF
XRD
temperature
loss tangent
torsional braid analyser
triethylene tetramine
crystallization temperature
polymerization ceiling temperature
isothermal curing temperature
transmission electron microscopy
glass-transition temperature
glass-transition temperature of the initial uncured system
maximum Tg of the cured system
thermo-gravimetric analysis
triglycidyl p-amino phenol
2,4,6-tris(dimethylaminomethyl)phenol
tetraglycidyl diaminodiphenyl methane
gelation time
melting temperature
thermal mechanical analyser
thermo-mechanical analysis
trimethylene glycol di-p-aminobenzoate
thermoplastic poly(urethane)
reaction temperature
timetemperaturetransformation diagram
upper critical solution temperature
ultra-high-molecular-weight poly(ethylene)
the ultimate glass-transition temperature of the fully cured material
polymer free volume
specic volume
wall-slip velocity
wide-angle X-ray diffraction
Williams, Landel and Ferry equation
X-ray diffraction
d
d2
j
q,
sCL
sF
UCL
a
a
ac
agel
a
e00
solubility parameter
cohesive energy density
compressibility
density
the lifetime of decay in chemiluminescence
uorescence lifetime
chemiluminescence quantum yield
the FloryHuggins interaction parameter
coefcient of thermal expansion
cure conversion
critical conversion
cure conversion at gelation
chain expansion co-efcient
dielectric loss
Glossary
e0
f
fm
c
c0
c0e
[g]
g*
ge
gr
gr1
gro
g
gmin
gc(T, t) gc(T, a)
gsr (c 0 , T)
k
h condition
r
rr
rY
x
m
439
dielectric permittivity
volume fraction of particles
maximum packing volume fraction
shear strain
steady shear rate
rate of elongation
intrinsic viscosity
dynamic or complex viscosity in oscillatory ow
elongation viscosity
g/gs the reduced viscosity
the high-shear-rate viscosity
the low-shear-rate viscosity
viscosity or chemoviscosity
minimum viscosity in thermoset processing
Temperature, time and conversion dependent viscosity during cure
shear rate and temperature dependency of chemoviscosity
relaxation time
temperature at which chain expansion coefcient 1
stress
the reduced shear stress
yield stress
the dynamic shear rate (or frequency)
frequency of the fundamental vibration
Index
441
Index
functionality 177
fundamental chemorheological
behaviour 321
gamma-irradiation processing 416
gel-permeation chromatography (GPC) 309
gel Tg 232
gelation 180
extent of reaction (pc) 100
gelation tests 345, 347
glass transition, Tg 17, 18, 202
grafting, high-temperature 957
high-impact poly(styrene) (HIPS) 113
hydroperoxide decomposer 154
hyperbranched polymer
degree of branching 45
living polymerization 989
step growth 43, 45, 47
inhibition, free-radical polymerization 67
interaction forces, polymer chain segments 108
interfacial adhesion 121
internal batch mixers 407
interpenetrating networks (IPNs) 126
interphase 122
ionic conductivity 290
isothermal dynamic frequency sweep 338
isothermal dynamic relaxation test 346
isothermal dynamic time test 342, 345
isothermal multiwave test 346
isothermal steady-shear rate sweep 338
isothermal steady time test 342, 346
isothermal strain sweep 338
James and Guth model 170
kinetic network model 190
kinetics, free-radical polymerization 625
KreigerDoherty equation 171
Kuhn length 3
linear viscoelastic behaviour 322
living polymerization 80, 91
loading vector 273
loss modulus 296
loss tangent 296
lower critical solution temperature (LCST) 106
luminescence spectroscopy, degradation
reactions 2545
mechanical shear 1246, 128
mechanoradical 94, 128
melamine formaldehyde resin 51
melt processing, radical formation 12831
microlithography 424
microwave processing 413
minimum processing viscosity 344
442
Index
mid-infrared (MIR)
absorption and emission analysis, remote
spectroscopy 269
analysis
addition polymerization 2234
end groups for molar mass 2345
network polymerization 22431
oxidation reactions 231
miscibility 1068
mixing extruders 408
mixing mills 408
modified CoxMerz rule 326
molar-mass distribution 8, 9, 28
multiplicative scatter correction (MSC) 277
multivariate calibration 275
multivariate curve resolution 272
network
addition polymerization 99
step growth 47
network-formation models 187
network polymers 1767
Newtonian 301
near-infrared (NIR) absorption and emission
analysis, remote spectroscopy 2678
NIR analysis
addition and condensation polymerization 2367
network polymerization 2378
nitroxide-mediated polymerization (NMP) 81, 92
NMR spectroscopy 212
non-isothermal dynamic sweep tests 344
normal stress 294
normal-stress difference 294
nucleation 13, 1516
phase separation 111
open-mould processes 391
optimal heating rate 344
oxidation induction time (OIT) 197
oxidation
free-radical 13942
kinetics 1424
partial least squares (PLS) 279
peptizer 157
percolation 187
percolation threshold 188
phase-separated systems, morphology 113
phase separation 11112, 11520, 124, 181
phenolic resin, synthesis 48
physical gelation 177
poly(amide)
oxidation 1478
polymerization 32, 33, 77
poly(carbonate), polymerization 31
poly(dimethyl siloxane) ring-opening
polymerization 78
poly(ethylene), oxidation 1456
443
Index
rheometers 305
rheo-NMR 312
rheo-optic 311
rheopectic 302
ring-opening polymerization 77
rotational conformations 57
Rouse model 173
rubber 223
rubber calendaring 410
rubber extrusion 409
rubber mixing 407
rubber moulding 410
scanning electron microscopy (SEM) 310
scavenging, free-radical 150
scission, entangled chains 12931
score vector 273
sealing 377
self-compatibilization 124
self-condensation 31
shear rate 293
shear rheology 293
shear strain 295
shear stress 293
shear-thinning behaviour 294
sheet moulding compound 395
size-exclusion chromatography 9
small-angle neutron scattering (SANS) 307
small-angle X-ray scattering (SAXS) 305
solid dynamic viscosity 296
solid ground curing 422
solubility parameter (d) 109
spatial distribution function 169
spatially dependent network model 190
specific volume 12
spherulite 15
spinodal decomposition, blends 111
statistical network models 187
steady-shear temperature ramps 343
step strain 298
stepwise polymerization 25
stereolithography 420
stereopolymerization 75
SternVolmer quenching equation 247
storage modulus 296
stress build-up 299
stress decay 299
suspension 171
tacticity 78
telechelic polymer 92
Tg0 181
Tg1 182
thermoplastic polymers 176
thermorheologically-simple fluids 298
thermoset injection molding 403
thermoset polymers 176
theta-temperature 4
thixotropic 302
time-independent fluids 302
timetemperature superposition (TTS) 298, 206
timetemperaturetransformation (TTT) diagrams
181
TMA 283
torsional braid analysis 2823
toughened thermoplastics 120
transfer moulding 397
transient shear 171, 298
transition-metal ions 1579
Trommsdorff effect 66
Trouton ratio 301
typical fluid viscosities 302
upper critical solution temperature (UCST) 106
urea formaldehyde resin 51
UV processing 415
UV-visible
absorption and emission analysis, remote
spectroscopy 2637
analysis, polymerization 247
UV-visible and fluorescence analysis, theory 2447
vane rheometer 323
vibrational spectroscopy, theory 21315
viscosity 170, 294
viscous dynamic viscosity 296
vitrification 180
vulcanization 407
wall-slip techniques 325
wide-angle X-ray diffraction (WAXD) 305
X-ray diffraction 305
yield stress 301, 323
Zimm model 173