Beruflich Dokumente
Kultur Dokumente
With Contributions by
M. T. Aronhime, K. Du~ek, J. K. Gillham,
E N. Kelley, J. D. LeMay, E Lohse,
H. Zweifel
2152/3020-543210
Editors
Editors Publisher
Preface
Time-Temperature Transformation ( T T r )
Cure Diagram of Thermosetting Polymeric Systems
M. T. Aronhime, J. K. Gillham . . . . . . . . . . . . . 83
Karel Du~ek
Institute o f M a c r o m o l e c u l a r Chemistry, Czechoslovak A c a d e m y o f Sciences,
16206 Prague 6, Czechoslovakia
The build-up of branched and crosslinked structures from polyepoxides and curing agents is investigated
theoretically and experimentally. The basis of crosslinking theories and their application to curing
of epoxy resins are reviewed and analyzed. The network build-up is dependent on the functionality of the
monomers, reactivity of functional groups and reaction paths as a function of conversion of the reactive
groups. The knowledge of the reaction mechanism is a necessary input information of the branching
theory and it also determines which of the available methods (e.g. statistical or kinetic theory) is to be
used. The curing of polyepoxides with polyamines, polycarboxylic polyacids and cyclic anhydrides is
treated in more detail. The theoretical treatment of polyetherification (polymerization) of epoxy groups
is outlined. Results obtained on simple polyamine-diepoxide systems agree well with the theory, acid
curing still requires refinement of the theoretical treatment and especially much more experimental
studies. The problems of the theoretical treatment of the network build-up in important epoxy resin-
curing agent formulations, to which the branching theory has not yet been applied, are briefly discussed
The problems of homogeneity or inhomogeneity of cured epoxy resins and of the diffusion control are also
analyzed.
1 Introduction . . . . . . . . . . -. . . . . . . . . . . . . . . . . . 5
AdvancesinPolymerScience78
(C)Springer-VedagBerlinHeidelberg1986
2 K. Dugek
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Network Formation in Curing of Epoxy Resins 3
= kt
fsse first-shell substitution effect
gf generating function
pgf probability generating function
wfgf weight-fraction generating function
A amine
C carboxyl
E epoxide
AH2 primary amine
HAE(OH) adduct of one epoxy and one amine group
A(E(OH)) 2 adduct of two epoxy groups and one amine group
DDA dodecylamine
DDM 4,4'-diaminodiphenylmethane
DDS 4,4'-diaminodiphenylsulfone
DGA N,N-diglycidylaniline
DGEBA diglycidyl ether of Bisphenol A
DGER diglycidyl ether of resorcine
EANC elastically active network chain
HMD hexamethylenediamine
PGE phenyl glycidyl ether
T G D D M N,N,N',N'-tetraglycidyl-4,4'-diaminodiphenylmethane
I Introduction
Epoxy resins are typical thermosets. Epoxy resins-curing agents systems exhibit during
cure the same features as other polymeric systems capable of branching and network
formation. These features include: extensive branching, passage through the gel
point, formation of a giant macromolecule with closed circuits -- gel, gradual
transformation of the soluble part, sol, composed of molecules of finite size into gel,
and eventually (in some cases) formation of a dense network. In this respect, the
epoxy resins undergoing cure do not qualitatively differ from other crosslinked
polymers, for instance from vulcanized rubbers or vinyl-divinyl copolymers. In
contrast to rubber vulcanizates, the crosslinking density of typical cured epoxies is
much higher. Moreover, the typical application temperature of cured epoxies is
below their glass transition temperature Tg. However, some epoxy systems have Tg
below room temperature and the crosslinking density of some is close to that of
rubbers.
Although the processing and final physical properties of epoxy-curing agent
systems depend primarily on their chemical composition and degree of cure, the
corresponding relations are often empirical or semiempirical and are not well
understood. The tie between the cure chemistry and structure and properties of the
cured resins consists in the theoretical and experimental study of network formation
as a function of the depth of cure.
6 K. Dugek
In the literature, a view is widely accepted that cured epoxy resins are not homogeneous
in general and inhomogeneously crosslinked in particular. This assumption is based
mainly on the observations of electron microscopy and partly on thermal behaviour
near Tg (cf., e.g., Refs. 1-5)). The aim of this Section is rather modest; it neither
addresses the problem of homogeneity (random fluctuation of density) of the glassy
state, nor it attempts to offer a firm statement saying that in all epoxy systems curing
occurs homogeneously. It wants only to demonstrate that there is a strong experi-
mental evidence that inhomogeneous crosslinking is not an inherent feature of
curing of polyepoxides. On the other hand, the postulate of homogeneity of the system
undergoing cure (in the sense used in homogeneous chemical kinetics) is equally,
or even more, important for the treatment of network build-up.
In an earlier analysis 6), it was shown that the main argument for the existence of
inhomogeneous crosslinking -- the appearance of nodular structure particularly
when etched fracture surfaces are examined -- is not at all characteristic of cured
epoxies. Similar nodular structures could also be seen in other amorphous polymers,
Network Formation in Curing of Epoxy Resins 7
kl
~ A + B ~ -'--~ ~(AB) *--~ k ~ "~'AB~ (1)
k- 1
(where E is the activation energy), in the glass transition region the dependence of k
on T is determined by segmental mobility. Near Tg, it can be approximated by the
free-volume theory and the following proportionality is expected to hold
T--T
lnk~ T--T +c* (3)
g
which says that k should approach zero when T, -- T -- c g, the constant c g is about
50 K. Since Eq. (3), which is an analogy of the W L F equation for the time-temperature
superposition, is no longer valid deeply in the glassy region, the Tg - - T difference
necessary for the arrest o f the reaction is not just 50 K. Experimental results show
that the practical limit for the curing process corresponds to T g - T equal to
25-35 K (cf., e.g., Refs. ~4,27)). Therefore, during curing of some epoxy-curing agent
I I I I
I00
i
GD
5O
100 I I i i
1.0 [ I I I I
0.8 ,/'/ o _3
0
0
0.6
O.4
0,2
,/
4 8 12 ~6 20 24
At .....
Fig. 2. Time (t) dependence of conversion of epoxy groups in the reaction of 4,4'-diamino-3,3'-di-
methyldicyclohexylmethane with DGEBA (rA = 2). Initial course is superimposed with the multi-
plicative factor A. Reaction temperature (°C): 1 100, A = 1; 2 64, A = 0.133; 3 40, A = 0.0308;
..... curve calculated for purely chemical control 27)
Experimentally, the glass transition has also manifested itself by a sharp increase
in relative rigidity (measured by dynamic-mechanical methods) and a simultaneous
drastic decrease in the rate constant o f the autocatalytic epoxy-amine reaction 26).
The mobility or rigidity o f the system is a function of reaction conversion ct; in the
pre-gel region it can be characterized by dynamic viscosity which is proportional
to M n o f the reacting system. Beyond the gel point, still in the rubbery region but
not close to the gel point, the dynamic modulus, G, is at low frequencies proportional
to a m (m ~ 1) 2s).
The dependence o f the rate constant near Tg on ~ at constant T can be obtained
from Eq. (3) by substituting for T 8 from the relation between T s and a like that
depicted in Fig. 1. The relative increase in T, can be expressed as ( T . - T J /
(Tg~ - - T,o) = ~q, where q ~ 1. Taking the first term of the development'of logVk
(Eq. (3)) in power series o f ( T - - T ) , one can see that for isothermal cure k0~exp (--A0~q)
should be expected, where A = (Tg - - Tso)/Cg. The value of A is o f the order o f 2-3.
12 K. Dugek
While the reaction rate for the same system measured at different temperatures
exhibited a marked effect of the overall diffusion control due to glass transition 25),
it was not clear whether this kind of diffusion control could affect the network
structure. If the mobility of segments belonging to smaller molecules were decreased
by the diffusion control to a smaller extent than that of larger molecules, the apparent
reactivity of groups on larger molecules would be lower compared to the reactivity
of groups on smaller molecules. If so, the gel point conversion should be shifted to
higher values. Experiments have shown, however, that this is not the case 27~, which
means that the segmental mobility decreases to the same extent irrespective of the size
of molecules of which the segments are a part. This is an important finding which
should be, however, confirmed by more experiments. If it proves to be generally
valid, the branching theories can be applied equally well for reactions occuring in the
melt, rubbery state and glassy state.
Branching theories are theoretical tools that can describe the structural changes in the
system as a function of time or conversion of functional groups. The initial compo-
sition of the system, functionality of the monomers, reactivity of the groups in terms
of the rate constants or their ratios and the reaction mechanisms, which determine the
sequence of formation of bonds of various types, are the input information. In this
section, the basic postulates of the theories will be outlined together with the procedures
of calculation of main structural parameters. For the sake of simplicity, the procedures
of calculating several selected parameters will be demonstrated here using a single-
component system which is, however, not typical for curing of epoxies. An extension
to multicomponent systems, such as to those composed of diepoxide and diamine, will
follows in Section 4 1)
The existing theories of network build-up can be divided into two major categories:
(I) graph-like models not directly associated with the dimensionality of the space,
(2) simulation of network build-up in n-dimensional space.
Group (1) in based on the tree-like model with uncorrelated circuit closing
in the gel, while the theories of group (2) more or less rigorously simulate spatial
correlations manifested particularly by cyclization.
The generation of the assembly of trees and of the gel according to group (1)
theories can be performed in two ways:
(1) by statistical methods from units at every stage of the reaction,
(2) by kinetic methods which consider every molecule including the gel as compo-
nents and describe the time (conversion) dependence of their fractions by differential
1 The present state and applicability of the network formation theories was analyzed recently in
Ref. 29)
Network Formation in Curing of Epoxy Resins 13
\ \\
I \ ~ \'
IF
g=4 \
b
Fig. 3. Transformation of the molecular forest of trees (a) into a forest of rooted trees (b) for a tri-
functional monomer. • node representinga monomer unit, © terminal node representingan unreacted
functionality
14 K. Dugek
o f this model, the branching theories of Macosko and Miller 36,37) or Durand
et al. 38,39) can be mentioned, which use only somewhat different mathematical
languages.
In the tree-like model, the building (monomer) units and molecules are represented
by graphs: the node o f the graph represents the building unit and the number of edges
issuing from the node is equal to the number of functional groups capable o f forming
bonds. The collection o f branched molecules existing in the reacting system at a given
conversion is represented by a collection o f molecular trees composed of units. This
collection is transformed into a collection of roted trees by choosing every node
(monomer unit) for the root with the same probability and placing it on generation
zero. The covalently bound units appear in the first, second, etc. generations with
respect to the unit in the root. An example is given in Fig. 3 for f = 3. Two important
consequences follow from this transformation: (a) the distribution of units in the
root represents the distribution of units in the system. (2) An i-mer is rooted
i-times so that it appears in the collection of rooted trees - - the rooted forest - -
i-times. The transformation into the rooted trees is performed in order to be able to
generate the trees using simple probabilistic considerations.
The m o n o m e r units (building blocks from which the molecules are built-up)
differ in the number o f reacted functional groups, i.e., in the number o f bonds
in which they take part, or, in other words, in the number of bonds they issue - - the
term which will be used later on. For a single-component system, this distribution is
sufficient (for multicomponent systems the types o f bonds are to be specified, too)
for the build-up o f trees, if the so-called first-shell substitution effect (fsse) is operative.
Fsse means that the reactivity of a group in a unit is independent o f the state o f the
groups in the neighbouring units.
The distribution of monomer units according to the number of bonds they issue is
conveniently expressed through a probability generating function which is a simple
tool used in generation of the trees. Thus, the probability generating function (pgf)
for the number of bonds issuing from a f-functional monomer in the root Fo(z) is
defined as
f
Fo(Z) = ~ piz i (4)
In Fo(Z), Pi is the probability o f finding a m o n o m e r unit in the root which issues i bonds.
This probability is equal to the fraction o f units with i reacted functional groups;
z is an auxiliary (dummy) variable o f the generating function through which the
operations with the pgf are performed. It is important to note that p~ is just the coeffi-
cient at z i 2). By operations, the differentiations or rarely integrations are meant.
In the derived statistical averages, z is put equal to 1 or 0. Thus
F0(z = 1 ) = F0(1) = ~ P i = 1
i
(SF0(z)/Sz)~=, = F'o(1 ) = ~ ip i
i
i.e. Fo(1 ) is the average number of bonds issuing from a monomer unit.
2 In earlier papers, the letter 0 was used for the pgf variable. For typographical convenience, it has
been replaced here by z
Network Formation in Curing of Epoxy Resins 15
While F0(z) represents the distribution of units in the root, it does not apply to
units in generations g > O. The monomer with no reacted group (fraction P0) cannot
appear in generation g > 0 because such unit must be bound at least to the unit in the
preceding generation. The pgf for units on generations g > O, F(z), is in the case of fase
~Fo(z)/(~Fo(z)) ~i iplzi-'
F(z)- ~-z [\--~z-Jz=l- Zip, (5)
i
Thus, the coefficient ipi/E ipi is the probability of finding in generation g > 0 a unit
with i reacted functional groups which issues i-1 bonds to the next generation. This
probability is proportional to i because of the way of construction of the rooted
trees (a unit occurs as first neighbour i times). This can be explained also by a simple
probabilistic argument that a unit with i reacted groups can be combined with another
reacted group in i ways.
The pgf's F 0 and F are sufficient for generation of trees and for derivation of
statistical averages, and they both are determined by a single distribution pi. This
distribution is obtained from the kinetic differential equations as will be shown on the
concrete example of curing of epoxy resins in Section 4.
The distribution of molecules (trees) according to their degree of polymerization
is obtained by so-called cascade substitution, in which the variable z is substituted
by a generating function. Thus, the weight fraction distribution function W(z) is given
by
where w i is the weight fraction of the i-mer. It is easy to see that, while zFo(z) = E pizi + 1
counts units in parts of the trees up to the generation 1, zF0(zF(z)) does it up to gene-
ration 2, etc. The infinite substitution counts units in trees of all sizes, i.e. it gives
the weight fraction distribution. Because of the infinite number of substitutions,
Eq. (6) can be replaced by
u = zF(u) (8)
v:,(1)
vw = w'(1) = 1 + (10)
1 -- V'(1)
16 K. Du~ek
W(z) can easily be transformed into a gf for the number, weight, z, z + 1.... fractions
yielding the respective averages, but the number average can simply be obtained from
stoichiometric arguments -- in this case, from the conversion of functional groups.
The calculation of molecular weight averages of multicomponent systems will be
shown for concrete systems in Section 4.
Indefinite continuation o f the structure is possible if, on average, one bond
issues from a unit on generation g > 0 to the next generation. Since the distribution
for the number of such bonds is given by F(z) and F'(1) is the corresponding average,
the gel point condition is given by
v = F(v) (12)
Figure 4 explains the meaning o f this definition equation: Let us assume that a
unit in generation g > 0 is attached to the preceding unit by a bond with finite conti-
nuation. This happens with the probability v. If so, none of the bonds issuing from
the unit under consideration to generation g + 1 may have an infinite continuation.
Thus the probability that a bond exists must be weighted by the probability v, i.e.
F(z) ~ V(vz) ~ F(v).
The quantity v is then used for calculating the sol fraction, degree of polymeri-
zation distribution and averages o f the sol, the number or concentration of elastically
Network Formation in Curing of Epoxy Resins 17
active network chains, effective functionality of crosslinks or the cycle rank, the
length o f network chains and dangling chains 4o~, and other structural characteristics.
The sol is composed of units issuing bonds with no infinite continuation. Thus,
z in Fo(z ) is to be weighted by v, so that the sol fraction w is given by
in which the probability that a bond is formed is weighted by the probability that it
has a continuation to infinity. The coefficient t i has thus the meaning of the fraction
of units issuing i infinite paths. In terms of these coefficients, Ne is given by
N e = (1/2) ~ it i (15)
i=3
value for z = 0, T'(0) = t~ and the second derivative for z = 0, T"(0) = 2t2, so that
A few words only about multicomponent systems (cf. Ref. 29~) are in order. One can
formulate the pgf Fo(z ) for each component, say F0i for the component i. These
components are a function of several variables z~k, where the subscript ik means that
the bond extends from unit type i unit type k. For units in generation g > 0, we have a
set of Fji(z), where z is the vector of Z~k = (Z~I, Z~Z.... ) and the meaning of the
subscript is the following: Fj~ describes the distribution o f units of type i rooted on the
18 K. Dugek
unit of type j ' the coefficient at z yik is equal to the probability that such unit of type i
extends y bonds i ~ k to unit k in the next generation. Thus,
which is the condition of the gel point. Here, Fi~ = (~Fij(Z)/~Zjk)z= 1 and 8,~ is the
Kronecker delta equal to 1 if ij = jk and to zero otherwise.
Because we have a set of F's and an equal number of z's, also the extinction pro-
bability is described by a set of v's defined by
where v is the vector of k components Vjk. Once the components v 0 are obtained, the
passage to w s, N and other quantities is straightforward using the same reasoning
adopted in the derivation of Eqs. (13-16); each zij is weighted or replaced by vir
Thus, the whole problem is reduced to a proper formulation of F0i's and Fij's.
dcx. 1
k dt lCx'l(I,~xlcx'l + CgWg)
where c is the concentration of unreacted groups in the gel. The first term of Eq. (21)
describes the transformation of the x, 1 molecule to other molecules or gel, the
second term the formation of this molecule by combination of two smaller ones.
By multiplying each equation for c I of the set (21) by variables z~z 1 of the gf
p f
g(zp, zf)
x 1
g(Zp, zf) - ~ c~,lZpZt (22)
x,l
and summing all equations, one obtains a single partial differential equation for the
generating function
where x = kt and % is the number of functional groups per monomer unit. The
solution of this partial differential equation yields
where ~ is given by
and a is the conversion of functional groups. Beyond the gel point (wg > 0),
Eq. (24) gives the distribution in the sol fraction 44)
(b) Initiated living polymerization 45) This process is important in curing of epoxy
resins and, for instance, the polymerization of epoxy groups belongs to this category.
It can be described by the following reaction scheme
kl
M+I "-'~ PI
kp
M + PI---~ P 2
kp
M + Pk----* Pk+t
20 K. Du~ek
di/dt = --k~mi
d q / d t = klmi - - kpmc~
(25)
The set (25) can be converted into a single differential equation (27) for the generating
function g(z)
dg(z)
- rag(z)(z -- 1) + ×mzi (27)
d~
x = kJkp (28)
where i0 is the initiator concentration at t = 0. Equation (29) for g(z) can be trans-
formed into a dependence of g(z) on m. The gf g(z) is in this case the number fraction
gf, and P , Pw, etc. can be obtained by successive differentiations: P = g'(1),
Pw = g"(1)/g'(l) + 1.... ).
The kinetic theory can also be used for polyfunctional systems with unequal
reactivities of groups and substitution effects, but an explicit solution of the partial
differential equation corresponding to Eq. (23) derived for the equireactive system
is not possible. One can use, however, the method of moments for derivation of
certain averages as was explained in 41,43).
The kinetically controlled processes determined by the above sets of differential
equation can be simulated by Monte Carlo methods 46). The species characterized
by x and a vector I for the number of groups differing in reactivity or vectors x and 1
are stored in the computer memory and the random numbers select the reaction
partners. It is necessary to examine the dependence of the results on the number of
m o n o m e r units used in simulation. Recently, the application of this approach was
reported also for epoxy-amine curing 4v~
Network Formation in Curing of Epoxy Resins 2t
The kinetic method can be extended to include cyclization reactions but no suitable
procedures have been developed as yet 48) In recent years, much progress has been
reached in the application of coagulation equations to various cluster formation
and aggregation processes 49)
The advantage of the kinetic theory over the statistical branching theory rests
in its adherence to the kinetically controlled chemical process while the statistical
theory working with units does not take into consideration the connections between
units developed in time (stochastic correlations). The greater mathematical com-
plexity and impossibility to get information on the internal structure of the molecules
and gel are the disadvantages of the kinetic theory.
Therefore, it would be of advantage to use the simpler and more versatile statistical
methods if the neglect of stochastic correlations were not serious. This problem is ana-
lyzed in Section 3.4.
It has been shown that for random polyfunctional step polyaddition both the sta-
tistical and kinetic theories give identical results 44). If substitution effect is operative,
the results are not identical and the results of the statistical theory are only an
approximation due to the neglect of stochastic correlations. The magnitude of these
deviations for this particular case of step polyaddition does not seem to be too
serious for moderate substitution effects 46). The situation appears to be much more
serious in the case of initiated reactions. It is known that the degree-of-polymeri-
zation distribution in linear living polymerization for × >> I approaches the Poisson
distribution, while the build-up of the chain from units generates a distribution which
is close to the most probable one 4s). For a monomer having two such polymerizable
groups of independent reactivity, the calculated critical conversions at the gel point
differ by about 30 %. In this case, the effect of stochastic correlations cannot be
neglected. The magnitude of the difference between results obtained by these two
methods depends on the particular reaction mechanism and at present it cannot be
estimated a priori.
The method of moments 41.43), possibly new numerical methods for the solution
of the partial differential equations for g(z) and computer simulation of the kinetic
process 46) are the possible ways of solving the problem.
A considerable simplification can be achieved, if some of the monomers have
groups of independent reactivity 29.50). Because stochastic correlations do not exist
between groups of independent reactivity, one can severe the connections and reform
them again without any loss in information content. The procedure is as follows 29).
(a) connections between groups of independent reactivity are cut and points of
cut labelled,
(b) the structures are generated using the kinetic method from the fragments obtain-
ed after the cut as well as other monomers in the system,
(c) pairs of points of cut with the same label are combined using the statistical
method.
22 K. l)u~ck
The advantage of this procedure consists in the possibility that the structures
generated in (b) are still finite. This method was applied to the treatment of non-
linear initiated polymerization of a monomer carrying two potymerizable groups 45)
and also to gelation of diamine-diepoxide systems 51). This application will be ex-
plained in more detail in Section 4.
The formation of cycles, or elastically inactive cycles (loops) beyond the gel point,
always accompanies branching and its intensity depends significantly on the reaction
mechanism as well as on chain flexibility and other factors 19,52). Incorporation of
conformationally controlled cyclization into analytical theories is a difficult problem,
because cyclization introduces long range spatial correlations. To preserve the
simplicity of the treatment of the tree-like model, the so-called spanning tree approxi-
mation has been introduced which is only a perturbation of the tree-like model 53.54)
In this approximation, the trees are embedded in the three-dimensional space
and the probability that an unreacted group can react with another unreacted group
in the same tree (molecule) is calculated relative to the probability of its intermolecular
reaction. The probability of ring closure is assumed to be determined by conformatio-
nal statistics of the sequence of bonds connecting these two groups. The groups parti-
cipating in ring closure are considered as unreacted but not capable of a further reac-
tion and bond formation. This approximation works well, if cyclization is weak.
It has been extended beyond the gel point to estimate the fraction of bonds closing
elastically inactive cycles in contrast to the circuits in the gel in which the chains are
elastically active 55,56) It is out of scope of this review to go into details of these and
other approaches to cyclization.
The epoxy curing reactions are usually step reactions or slow polyaddition reactions
which exhibit a lower tendency to cyclization ~9). However, cylization can play a
non-negligible role, if there are special dispositions to ring formation in the monomer
or if the sequence connecting the functional groups is flexible enough and relatively
small cycles can be formed, e.g., in systems containing short aliphatic diepoxide and
diamines, or in monomers having glycidyl groups close to each other like in diglycidyl
esters and ethers with glycidyl groups in the ortho position 5v). Also, N,N-diglycidyl-
aniline (DGA) and its tetrafunctional analogue N,N,N',N'-tetraglycidyl-4,4'-di-
aminodiphenylmethane (TGDDM) exhibit a larger tendency to intramolecular
reactions 58) which is demonstrated by the dependence of the critical molar ratio
at the gel point on dilution 59). Cyclization seems to occur also in the cationic poly-
merization of epoxy compounds 20,21~
Diglycidyl ether of Bisphenol A (DGEBA) is a stiff monomer and in the reaction
with aromatic or even aliphatic diamines it exhibits a very low tendency to cyclization,
so that the ring-free theory can be applied with success. This conclusion was derived
from the fact that the critical conversion at the gel point was independent of dilution
for DGEBA-diamine systems and that the critical conversions correspond to the ring-
free model 16"18). The same conclusion applies to DGEBA-dicarboxylic acids
systems 60).
Network Formation in Curing of Epoxy Resins 23
The interference of cyclization should be, however, considered for any new system
and examined experimentally. The spanning-tree approximation with properly
selected chain flexibility parameters may be suitable. One can also use the simulation
in three-dimensional space mentioned in Section 3.6.
In principle, it is possible to consider ring formation also in the kinetic generation.
However, the differential equations become too complex because of topological
complexity of graphs with cycles unless the cycles are small.
In these models, monomer units are placed on a lattice and bonds are introduced
between them either at random or according to given rules. This type of simulation
is known as percolation and the application of percolation theory to branching
polymer systems was reviewed recently by Stauffer 61). The rings are generated and
their size distribution depends on the dimensionality of space and also on the type
of lattice. The main drawback of the application of this theory to derivation of for-
mation-structure relationships is the fact that this approach does not allow for
conformational rearrangements which occur between formation of two bonds
successive in time. An off-lattice simulation 62,63~ removes the dependence on the
lattice type. For curing of diepoxides with diamines, a variant of the lattice simulation
has been developed 23.24, 57.6~) and applied particularly to deep stages of curing
where the rigidity of the lattice can be justified by the rigidity of the densely cross-
linked network. In this simulation, the diamine units are placed on the lattice sites
and the diepoxide molecules freely migrate. A random number selects the amine
functionality for reaction with an epoxy group. A weighting is applied to the reaction
of a primary or secondary amino group depending on their relative reactivity. When
diepoxide is bound by one epoxy group, the other group can find its reaction partner
within the sphere of action given by the possible conformational rearrangements.
Reactions with only the nearest neighbour amine functionalities or loop formation
with the amine functionality on the same lattice site are considered and the weighting
again applies to these inter- and intramolecular alternatives. Unfortunately, the details
of the algorithms for calculating the structural parameters have not been described in
detail.
Due to the lattice rigidity, some of the unreacted groups cannot find their reaction
partners and remain unreacted; this is called the topological limit of the reaction. A few
of these groups can also form monocycles (cf. also Section 2.VII.4 of Ref. 14)). This
topological limit was confirmed by some experiments, some other authors claim,
however, that they can reach 100% reaction. An accurate determination of the few
unreacted groups in the rigid structure is, however, not an easy task.
24 K. Du~ek
Preget stage
Molecular weight (MW) Gel permeation chromatography
distribution (GPC)
Molecular weight averages VPO, Light scattering and other Ageing of resin-
suitable methods curing agent premixes
Molecular weight vs. compo- GPC + Liquid
sitional distribution chromatography Processing
Radius of gyration Light scattering (LS)
Scattering functions Static and dynamic light scattering Chemorheology
Get point
Critical conversion Solubility
Viscosity
Postgel stage
Sol fraction Extraction
Characteristics of the sol as above Processing
The quantities k I and k 2 are apparent rate constants which depend on the extent
of reaction due to autocatalysis by the formed OH groups as well as on retardation
due to hydrogen-bonded association complexes ~4). If one writes the bimolecular rate
equations as
d[AH2]
- --kl[AH2] [El
dt
d[HAE(OH)]
dt -- k,[AH2] [E] - - k2[HAE(OH)] [El
d[A(E(OH)) 2]
- k2[HAE(OH)I [El
dt
k 1 = k ° + k ~ [ O H ] + ~, (32)
kz = k° + k2[OH] + BE (33)
where t~1 and la2 are conversion-dependent terms expressing the effect of complex
formation. The kinetic results and also the distribution of primary, secondary
and tertiary amino groups ts) approximately conform to the relation
O 0
k2/k , ~ ~/k~ ~ ~2/~, ~ ~ / k , = e (34)
The simplest building blocks are represented by primary, secondary and tertiary
amino groups as well as unreacted and reacted epoxy groups. If the relation (34)
is obeyed, the calculation of fractions of these building blocks is simple is)
mation. It determines whether extensive branching already occurs in the early stages
of curing, or whether chain extension predominates.
The same and independent reactivity of epoxy groups of diglycidylether of Bisphe-
nol A
O CH3 O
DGEBA
seems to be relatively well established 16,18,23, 71-83) It need not be so, however, in
polyepoxides that have different epoxy groups (e.g. the glycidyl group, styrene
oxide, or 1,2-epoxycyclohexane, cf. Chapter 3 of Ref. 65)). There are indications
that in diglycidylaniline
CH2C.HCH2
N 0
o
DGA
the reactivity of the epoxy groups is dependent due to steric interactions which applies
also to N,N,N',N'-tetraglycidyl-4,4'-diaminodiphenyl methane (TGDDM), some-
times also called T G M D A = tetraglycidylmethane dianiline
/CH2~HCH2NH- /,CH2C~-CH2NH-
-N \ OH -o -N \ / O
CH2CH/CH2 CH 2C,x,H
O CH2OH
In the reaction with amine, the proximity of glycidyl groups also makes the ring
formation more probable s9)
-N
,/CH2•HCH2NHR
OH
2
~ -N
/
CH2CHCH 2
I
OH
\
N-R 2
\CH2CH/C H ~ CH2~HCH2
/
O OH
28 K. Dugek
1 I
0.6
i I
As has been discussed in Ref. 14), an analysis of data of bulk kinetics is not easy
due to autocatalysis and autoinhibition and an addition of excess proton donor
(alcohol) may affect the ratio O. In a certain region of Q, the critical conversion is
not too sensitive to Q (cf. Refs. 15-1~)). The best way of determining Q seems to be the
simultaneous determination of concentrations of primary, secondary, and tertiary
amino groups and examination of their ratios; the method (2) offers one of these
possibilities.
3.0
i 2.8
~2.6
Fig. 6. Theoretical dependence of the molar ratio
necessary for gelation (critical excess of amine
2.4 I I __ f groups), (rA)c, in dependence on k2/k1 = 0 is)
~2 0.3 OA 0.5
Q -
A list of k2/k 1 values is available in Section 2.IV.4 of Ref. 14) but, as has been
pointed out, the reliability of data cannot be guaranteed. Nevertheless, one can
conclude that for aliphatic amines Q = kz/k 1 -- 0.3-0.5 (no or weakly negative
substitution effect) and for aromatic amines with the exception of phenylenediamines
Q = 0.t7-0.25.
With some curing agents, the substitution effect can be highly negative as, for
example, in 2,5-dimethylhexane-2,5-diamine due to steric effects 89). A relatively low
reactivity of secondary amino groups in diaminodiphenylsulfone (DDS) observed
in the reaction with the tetraepoxide T G D D M 90,91) could be ascribed to a negative
substitution effect induced by conjugation with the sulfone group. However, model
experiments with tolylglycidyl ether revealed that the substitution effect is com-
parable with that observed with D D M 92,93). Therefore, the low reactivity of the
secondary amino groups in DDS in the D D S - T G D D M system is not due to a
generally low reactivity of this group (highly negative substitution effect) but due to
steric hindrances resulting from the addition of T G D D M . This assumption is sup-
ported by the results of the study of stoichiometric amount of DDS and diglycidyl
ether of butane diol in which a complete reaction of secondary amino groups could
be reached 94)
The last example shows that the value of the substitution effect characterized by the
ratio Q in not a universal constant and may depend on the structure of the reference
epoxy compound. Phenylglycidyl or p-tolylgtycidyl ethers are apparently good
reference compounds for curing of DGEBA, but they may be not that good as models
for DGA, T G D D M and some other polyepoxides.
30 K. Du~ek
dCp/dt = --klcavc~
dCa,/dt = klCapCe--k2CasC e (35)
dCat/dt = k2CasC e
where c is the concentration of epoxy groups. Assuming the validity of relation (34),
i.e., the constancy of the ratio k2/k 1 = Q, one gets from the set (35) two equations in
which t is replaced by c v as a new independent variable
ap = Cap/(C p + C s q- C t ) , etc.;
ap -I- a s + a t = t
Network Formation in Curing of Epoxy Resins 31
Monoomine
Op 0s a t
Diamine
kl = k'-
2
OO=CIp a3=2asa t
M
a=2%a,
Diepoxide
2
eo=(1-c~E)2 el=2c(E(1-~E) e2= ~E
Fig. 7. Fraction of amine, diamine and diepoxide units with different number of reacted functionalities
1
a - l ~ ( a ~ - - a-p ) _ ~ (37)
Q
at - 1 (ap - - a~/£) + 1 (38)
g
a + 2a t = 2ctA (39)
The fraction o f unreacted and reacted epoxy groups is just equal to 1 - - ~E and ~E,
respectively, where 0[E is the conversion o f epoxy groups.
32 K. Du~ek
The probability generating function (pgf) for amine and epoxy units fA(Z) and
fE(z), respectively, just describes this distribution
where the subscript E or A at the variable z means that the bond extends to the
epoxide and amine unit, respectively. The pgf's for the number of bonds issuing
from diamine and diepoxide units in the root is obtained by squaring fA(z) and fE(Z)
because of independence of reactivity of groups
4-
FoA(ZE) = f~(ZE) = (ap + a~z E + atZE2)2 = ~ a~z~i (42)
i=O
From Eqs. (42-45), one can derive all structural characteristics of the system under-
going cure.
(a) Molecular weight averages:
The number average molecular weight is obtained from simple reasoning that the
number of molecules is given by the number of units minus the number of bonds
connecting them. Since the derivatives of FoA(ZE) and FoE(ZA) at ZA = ZE = 1, FEA
and F~E, respectively,
represent the average number of bonds issuing from the respective m o n o m e r units,
the average number of bonds per unit is given by (nF~A + nEF~E)/2 and the number
average molecular weight M is given by
where M A and M E are molecular weights o f the diamine and diepoxide, respectively,
and r A is the initial molar ratio o f amine and epoxide functionalities, r A = 4nA/2n E.
The weight-fraction gf W(z) is obtained by cascade substitution analogous to
Eqs. (6-8) for a single-component system. Depending on the type o f root, the
components o f the wfgf WA(Z) and WE(Z) read
MA
WA(Z ) ~- Z A FOA(HE) (50)
WE(z) = ZE
MEFoE(UA) (51)
with
MA
u A = z E FA(UE) (52)
u E = z~EFE(UA) (53)
The molecular weights o f the units M A and M E appear in the exponent because, in
contrast to Pw, the weight fractions of units must be weighted by M A or M E in order
to get M w.
The wfgf for the whole system is as follows
and
so that
A E
D = 1 -- FEF A ____rA(2~ZA + at ) (59)
Equation (57) can be used for monomers having any number of functional groups
of independent reactivity, if the squares in FoA and FoE (Eqs. (42, 43)) are replaced
by the functionalities fA and f~, and the derived equations modified accordingly.
(b) The gel point:
According to Eq. (19), the determinant D = 0 at the gel point; therefore
F AE F A
E = (2%, + at/atA) % = rA(2a2A + at) = 1 (60)
F r o m Eq. (60), one gets the critical conversion at the gel point by numerical solution
after substitution for at from Eq. (38) and using Eq. (39). The theoretical dependence
for rA = I is given in Fig. 5. For zero substitution effect (0 = 1/2), Eq. (60) yields
the ideal Flory-Stockmayer relation % % = 3-a.
Of interest is the critical value of the molar ratio rA (Eq. (49)), if amino groups are
taken in excess and the system is reacted to % = 1. The critical value of rA is given
by the equation
By substitution for v A in Eq. (62), one gets a cubic equation in VE. The trivial root
v E = 1 can be eliminated and the extinction probability is equal to the root in the
interval (0, 1).
(d) Sol fraction:
The sol is composed of units in which none o f the issuing bonds has continuation
to infinity. This condition is obeyed by multiplying the probability o f finding a unit
issuing i bonds by v i. Each of these units contributes to the weight o f the sol by its
own weight. Thus, the sol fraction w s is given by
4
Ta(z ) = (ap + as0 + atO2)2 = ~ tiz i (65)
i=O
where
0=v E + (1--VE) Z
The probability that a bond exists is thus weighted by the probability that it has an
infinite continuation. Using the above definition of E A N C ' s and the pgf (65), the
number of E A N C ' s per m o n o m e r unit, Ne, is given by
which means that the number of E A N C ' s is contributed by active branch points
issuing 3 and 4 paths to infinity. Expanding the pgfTa(z ) (Eq. (65)) into a power series
of z, one gets the coefficients at z 3 and z4, t 3 and t4, so that
The concentration of EANC's in the whole (unextracted) system re, (mol/vol), and
in the extracted gel, v g, are equal to
v = dNJ19I (68)
v , = d,Ne/1VIwg (69)
where d and dg are densities, wg = 1 - - ws is the gel fraction and IVl is the number
average molecular weight of the monomers, 1VI = nAMA + nEME-
(f) Other structural parameters:
It is possible to calculate a number of other structural parameters, for instance
those listed in Table 1. In Ref. ~7), relations were derived for the average functionality
of active branch points, fo, a quantity which is important in the rubber elasticity
theory for conversion of N into the (effective) cycle rank. In terms of pgfTa(Z), f~ is
defined by
where v0 is the initial volume fraction of the monomers; 1 - - v0 is the fraction of the
diluent possibly present in the system. T e is related to the fully reacted stoichiometric
system in the absence of diluent for which T e = 1. For details, see Ref. 17~;unfortunat-
ely, by misprint the squares of expressions (44) and (49) are missing. An extension
of the approach 4o) for calculating the molecular weight averages of EANC's and
dangling chains to epoxy-amine systems is in progress.
' " I I I
0.5 --
0.4
0.3
0 •
5 Q•
0.2
0 0
0.1
Fig. 8. Dependence of the fraction of tertiary amine groups in the systems dodecytamine (DDA)-
phenyl glycidyl ether (PGE), DDA-DGEBA, and HMD-DGEBA on conversion of amine functio-
nalities. GPC determination: DDA-PGE ~) ; determination of ap and a~ by titration: HMD-DGEBA
epoxide in excess O. amine in excess C) ; DDA-DGEBA amine in excess (13. The curve was calcu-
lated for O = 0.41 ~3)
38 K. Dugek
is only a function of conversion (cf. Eq. (48)) in any correct theory and for any reaction
mechanism. It would be affected only by cyclization. The conclusion one can draw
from these result is that cyclization is unimportant. More information is to be expected
from higher averages of molecular weight. However, only M w data for D G E R - D D S
systems obtained by the method of sedimentation equilibrium are available for
epoxy-amine systems 2z), and they agree within experimental error with the predic-
tions of the cascade theory. Much more data are needed, however.
The position of the gel point on the conversion scale is of primary interest in a
branching process. The critical conversion for the diepoxide-diamine systems with
groups of independent reactivity depends on the initial molar ratio of amino to
epoxy-amine systems 2z), and they agree within experimental error with the predic-
the amino group characterized by Q. Figure 9 shows the theoretical predictions for
various ~ and the experimental points for the system DGEBA-hexamethylene-
diamine (HMDA). The results fit well the curve for Q = 0.5 but they are not at
variance with curves for Q = 0.34-0.40 obtained from model experiments or critical
molar ratios 13,16). It can be seen that the influence of the substitution effect becomes
the weaker the larger is the excess of epoxy groups.
0.8
1 0.6
The influence of the substitution effect on critical conversion has been shown in
Fig. 5 for the stoichiometric mixture of diepoxide and diamine. The critical con-
version changes from 0.5 (Q ~ oo) to 0.618 (Q ~ 0), the ideal value being 0.577 for
Q = 1/2.
The critical excess for amino groups is more sensitive to the substitution effect
within the amino group than the gel point conversion of a stoichiometric system and
it is thus more suitable for characterization of O (Fig. 6). In this way, the value of O
was found to be 0.33-0.40 for the atiphatic amino group and 0.18-0.24 for the
aromatic group in D D M (Refs. 16 and 18 and unpublished measurements). These
values are close to these obtained in model reactions of compounds of low functio-
nality. The determination of the critical molar ratio necessary for gelation 15), i.e. the
Network Formation in Curing of Epoxy Resins 39
critical value of r A for 100~ reaction of epoxy groups (cf. Eq. (61)), is of interest
because of its experimental simplicity-checking of solubility on chemically non-reactive
systems.
While the critical molar ratio at excess amino group has been used for characteri-
zation of the substitution effect within the amino group, a critical excess of epoxy
groups can be used for characterization of a possible substitution effect in the
diepoxide or for verificatibn that such an effect is absent. However, for an excess
of epoxy groups (after all amino groups have reacted) there is a danger of polyetheri-
fication reactions which would increase the critical value of l/rA from the ideal value
of 3 for independent reactivity of epoxy groups and no etherification. Experiments
have shown that for aliphatic amines polyetherification does interfere (it sets in after
all amine hydrogens have been exhausted)13), but this is apparently not so with
aromatic amines; the basicity of the formed tertiary amino alcohol is so weak that
it hardly catalyzes polyetherification 9s). On the other hand, if the tertiary nitrogen
atom is already present in one of the monomers like in N,N-diethyl-l,3-diamino-
propane 99) or in D G A and T G D D M 90,91), etherification can interfere. One remark
important for the application of this method is to be added: The assumption of 100
reaction of the minority groups (usually epoxy groups) should be checked. The final
conversion may be somewhat less in some cases (say 98 ~ ) and this incompleteness
can affect the calculated value of Q somewhat. Then, instead of Eq. (61), Eq. (60) is to
be employed in which the limiting value is used for mE.
In the postgel stage, the sol or gel fractions are parameters suitable for comparing
the branching theory with experiments. The determination of the sol fraction by
extraction is not free from problems. At low degrees of crosslinking, e.g., close to
the gel point, the molecular weight of the molecules in sol is high and their diffusion
slow; moreover, because of a very high degree of swelling of the gel, the thermo-
dynamic driving force for diffusion is weak. An incomplete extraction is possible and,
therefore, the extraction time has to be increased. However, at high extraction times
the network in the gel can undergo stow degradation. For samples, analyzed in the
course of curing, a danger arises from still unreacted functional groups which can
further react during extraction. Then, the extent of reaction appears to be higher
and also some of the sol molecules can react with the gel. To prevent further reaction,
0.6 - O
T
0.4
Fig. 10. Dependence of the gel fraction of cured
DGEBA-HMD systems on the molar ratio ra.
The curve was calculated for the conversion of
epoxy groups % = 0.98 and Qt~ = 1.0 for
0.2 I I I O = 0.35 161
2.3 2.4 2.5
40 K. Du~k
acrylonitrile was added to the solvent used for extraction which readily reacted with
primary and secondary amines 16~. To use chemically non-reactive systems using
excess amine, which is below the critical excess, is another way of solving the
problem ~6,is)
The agreement between the calculated and determined sol or gel fraction is
generally good: Fig. 10 shows the data for D G E B A - H M D A 16) and Fig. 11 for the
three-component system DGEBA-phenylglycidyl ether-DDM; Fig. 10 also shows
that the knowledge of the final conversion is important and that any deviation from
completeness of the reaction can have a strong effect on w~ and the concentration
of elastically active network chains as well a feature characteristic for step poly-
additions with a relatively high critical conversion 52) Also for the D G E R - D D S
system with an excess of epoxy groups 24) _ polyetherification apparently does not
interfere - - the experimental data fit well the theoretical curve (Fig. 12). The only
exception are the results of Bogdanova et al. 79) on the same system, where the sol
fraction was monitored as a function of conversion in a stoichiometric mixture. A
delay of gelation, appearance of a small sol fraction over a wider range of conversion
and eventually an abrupt transition to 100% gel were the main features for which
we have no explanation.
I t ,4 q
0.2
l
1.2 1.6 2.0
rA . - - - - - ~
Fig, 11. Weight fraction of the sol in DGEBA-PGE-DDM systems as a function of the molar ratio r A.
The fractions of epoxy groups of monoepoxide (PGE), s, and final conversion of epoxy groups ctE
determined by ir spectrometry are indicated. The curves were calculated for Q = 0,19
0.3 0.5
0.4
0.2
0.3
0.1
0.1
2 3 4
IE)oI(A)o = 21r
Fig. 12. Weight fraction of unreacted diglycidylether of resorcine (DGER), WDC~R,and sol fraction,
ws, in the system DGER-DDS as a function of initial molar ratio epoxide/amine24): O WAGER,
• W. The curves were calculated theoretically
However, the correlation with the theory is complicated by the fact that one does
not test only the branching theory but also the rubber-elasticity theory. Several
variants are offered for explaining the dependence of the stress-strain data on the
network structure (cf. e.g. Refs 97,100-102)). It is assumed that the equilibrium force
or the equilibrium modulus, Ge, is composed of a chemical contribution, G~, and a
contribution coming from interchain constraints. In the theory by Flory 100,101~,
the constraints affect the stress strain-behaviour only through their effect on the
fluctuation ofcrosslinks, while some other theories consider an interchain contribution
due to chain incrossability using several models. For a phantom-network (volume-
less chain freely interpenetrating one through the other), the equilibrium shear
modulus Ge = Go,
G , ¢ = A~(~z~) v RT (72)
Ap h _ f -- 2
f (73)
where the second term corresponds to entanglements existing in the system before
crosstinking (its concentration is equal to e which is usually close to the plateau
modulus) but entrapped by chains between active junctions. The entrappment pro-
bability is expressed by the trapping factor given by Eq. (71) which varies from 0
at the gel point to 1 for a perfect network.
For bulk polymerized samples containing sol
for extracted swollen samples. Here, Ve and Te are related to the whole non-extracted
system and vcg and Teg only to the gel; v2 is the volume fraction of the polymer in the
swollen gel.
The results for both non-extracted dry and extracted swollen DGEBA-phenyl-
glycidyl ether-DDM networks are shown in Fig. 13a together with the theoretical
curve obtained from v (Eqs. (68, 69)) with the aid of Eqs. (74, 76, 77). The reduced
modulus G d is equal to Go/wgRT for dry non-extracted or to Ge/V~/3wgZ/3for ex-
tracted swollen samples. The data follow well the shape of the dependence pre-
dicted by the theory over two orders of magnitude of the modulus (concentration of
EANC's). They fit relatively well the curve for A = 1. According to the Flory
theory, this would mean that the chains are constrained even in the swollen state.
On the other hand, one can express the data using Eq. (75). Then the difference between
G and the contribution by the phantom network should be proportional to T .
e
This difference indeed fits the T e dependence calculated for different series of samples
(varying r Aand fraction ofmonoepoxide) with a single constant e = 8 x 10 -4 mol/cm 3
(Fig. 13b). Thus, the trapped entanglement concept seems to work well, but the
independence of G d of swelling and the fact that the Mooney-Rivlin C z term is close
to zero ~8) is not usual for typical crosslinked elastomers.
The shape of the dependence of G on r A for excess amine and excess epoxide in
epoxy-amine elastomeric networks predicted by the theory has been well described
by experiments both for rA > 1 and rA < 1 (Refs. 103-1o5~), but the results have not
been analyzed quantitatively.
Network Formation in Curing of Epoxy Resins 43
{ I I I
-3- a
o A=t~-I~1 ° I'-
|/ ..... I I I I
o-4 v
_o A. ~ / o
-5
I I
-0.4 -0.3 -0.2 -0:
tog wg =
Fig. 13. Dependence of the reduced modulus G, (mol/cma) and of the trapped entanglement contribu-
tion A (mol/cm3) on rA for the cured DGEBA-PGE-DDM systems ~7,1s~. O measurements in the
dry state, • samples swollen in dimethyiformamide; a , theoretical dependence for the
value of front factor A indicated, b theoretical dependence using e = 8 x 10-4 mol/em3
In conclusion, it can be said that the theory can well describe the development of
the gel structure. The correlation between the equilibrium modulus and sol fraction
is very good 18) so that the sol fraction can alternatively be used for determination
of the concentration of EANC's if an accurate and precise determination of con-
version meets with difficulties. It is to be recalled here that the Gaussian rubber
elasticity theory does not apply to highly crosslinked networks of usual stoichiometric
systems. When a good theory is available, the calculated value of v , taking possibly
into account the topological limit of the reaction 23.24~, will be needed.
step. However, the actual mechanism of initiation and propagation may not be
simple 14)
Two theoretical approaches are mentioned in Section 4.1 (Refs. 66.84)), but the first
one is based on incorrect statistical treatment. A comparison with experiments is not
available, so that the seriousness of deviations cannot be estimated. Also, a cascade
theory has been developed to cover polyetherification released by the reaction of
diepoxides with dicarboxytic acids 71) but an analysis of experimental data is also
lacking. An extensive pregel study of polymerization of diepoxides released by di-
phenols is analyzed by W. Burchard in another chapter of this volume ~8).
It has been shown in Section 3.3 that the initiated polyetherification should be
described by the kinetic theory and that a simplification is possible, if groups of
independent reactivity in polyfunctional monomers participate in the curing reaction.
The results of the treatment 51) of an ideal postetherification following the epoxy-
amine addition are summarized below.
Let us consider the reaction between diepoxide with independent reactivity of
epoxy groups and a diamine with independent reactivity of amino groups. Because
of independence of reactivity, the connection between the groups can be cut and the
points of cut labelled a and e, respectively
Ap-Ap --~ 2 a-Ap, E-E ~ 2 e-E
in the next step the sequence (clusters)
a-A~(E) x and a-At(E)y
e e
are generated using the kinetic method. In the third step, the labelled points of cut
--a + a-- and --e + e-- are recombined and the original connections restored
using the statistical (cascade) method. The problem is treated sl) as for any ratio of
rate constants for polyaddition and polyetherification.
For the usual systems, however, etherification is much slower than addition, which
means that etherification occurs with a measurable rate only after all amino groups
have been transformed into tertiary amino groups. The reaction then proceeds
according to the scheme:
Ap
2kA1
AtE2 AsE2
lk~ (78)
AtE3 ~ AsE3
Network Formation in Curing of Epoxy Resins 45
Assuming kAp kA2 ~ kE, and a bimolecular reaction mechanism, one can express
the distribution o f clusters AtE~ using a number fraction gf defined by
where (a,ex) is a fraction of AtE x normalized such that Z ate x = 1. In Eq. (80),
x
the dependence o f time can be transformed into a dependence on the concentration
o f unreacted epoxy groups [E0], or etherification conversion %~:rn defined by
where
[E]o - - 2[A]0
R E --
[AI~
and [A]o and [Elo are the initial concentrations of amino and epoxy groups,
respectively. The variable z E in Eqs. (79) or (82) is related to the number o f E units
in the cluster or to the number o f e-functionalities issuing from the cluster.
In the last step, the recombination is performed by using the cascade method.
The final structures are generated from clusters and unreacted epoxy groups. The
pgf's for units in the root can be formulated as follows
F E(Z) = 1
FAc(Z) = g(zr)
w h e r e Fxy means that the unit y is rooted in the preceding generation by the bond
of type x. For example, the gel point condition is given by
or
Once the pgPs for units in the root and higher generations have been formulated, it
is not difficult to calculate other structural parameters like the molecular weight
averages, sol fraction or concentration of EANC's.
For comparison, the same problem was treated by generation from units taking
into account the probabilities of finding A - - E and E - - E bonds and making difference
between E units rooted on an E or A unit ~1). In Fig. 14, the dependence of QtETn
I I
0.'~ - -
T
y 0.05
Fig. 14. Dependence of the critical value
of the conversion of excess epoxy groups
available for polyetherification, ~tEXn, in
diepoxide-diamine systems on the initial
molar ratio of epoxy to amino groups sl);
1 dependence calculated using the kinetic
theory, 2 dependence calculated using the
kinetic theory.
1 2
log ( E ) o / ( A ) o -
on the initial molar ratio of epoxy to amino groups [EP]/[NH2] = 2/r A necessary
for gelation is plotted according to both approaches. The excess of epoxy groups
plays a dual role: (1) it makes the mixture more off-stoichiometric and the network
looser, and (2) the more epoxy groups polymerize, the higher is the functionality of
clusters and the more probable is gelation. The area below the curves corresponds
to a non-gelled system.
Figure 14 shows that the difference in the result obtained by the combination
of the kinetic and statistical method and by the purely statistical method is rather
large. In real systems, the polyetherification reaction may be complicated by termi-
nation reactions e.g. by chain transfer, so that the mechanism m a y deviate from the
Network Formation in Curing of Epoxy Resins 47
pure living polymerization mechanism 14). Therefore, before comparing the experi-
ments with theory, these possible complications have to be taken into account in the
theory. I f the chain transfer reaction is independent of chain length, such amendment
will not represent any serious problem. From the experimental point of view,
a detailed examination of polyetherification released by mono- and bifunctional
initiators and yielding linear chains is desirable. 1
CHCH2OOCR2
+ R~CxHCH~ /R~ I
OH
~
~
+R2COOH
condensation
po,ye~ed-/~ "x~sterilicotion
licorice/
~ . /CH2OOCR2
RI CHCH2OOCR2 disproporti onotion R1CH
I trensesterificotion \OOCR2
(OCH2CHR~)× +
I H20
OH
/CH2OOCR2
R1CH
\OOCR 2
+
RI~HCH20H
OH
For catalysis by bases, addition esterification is the first and the fastes reaction. It can
be followed by slower polyetherification reaction, if epoxy groups are in excess, or
by condensation esterification (even slower), if carboxyl groups are in excess. This
mechanism was verified on a model system caproic acid - - phenylglycidyl ether
i=0 i=2
i=l i=2
i=3 i=4
where o~ and ctc are molar conversion of epoxy and carboxyl groups, respectively,
and fE and fc are functionalities of polyepoxide and polyacid. Relations for the
molecular weight averages, critical conversion, sol fraction and concentration of
EANC's can be derived in analogy with the treatment explained in Section 4.2.3.
Network Formation in Curing of Epoxy Resins 49
In the literature, two additional reactions following addition esterification have been
treated using the cascade theory: the addition esterification followed by polyetheri-
fication with epoxide groups in excess (a reaction used for crosslinking of carboxyl
terminated polydienes) and addition esterification followed by transesterification.
Transesterification often interferes wherever hydroxyester groups are formed, for
example, in synthesis of linear oligomeric polyesters from diepoxide and acids. As
has been explained before, polyetherification is an initiated reaction and, therefore,
the statistical treatment offerend in Refs. 71) should be revised. Below we show
the treatment of transesterification for a system composed of a diepoxide and a
dicarboxylic acid.
The depth of transesterification is expressed by the transesterification conversion
% defined as the fraction of hydroxyester groups transformed into diester and diol
groups. In Equations (87-90), the coefficient ctE at z c is the probability that an
epoxy group reacted. However, only a part, 1 - - %, is still in the form of hydroxy-
ester groups and the fraction % has been transformed into equal amounts of diester
and diol groups, issuing, respectively, 2 and 0 bonds. Therefore, zc is replaced by
One can see that for % > 0, gelation is possible. It should be stressed that the
number of bonds between units does not change as a result of transesterification,
they only become redistributed. Gelation is possible, because some of the diepoxy
units acquire functionality higher than 2.
The extinction probabilities are again given by relations vc = Fc(vE) and v E = FE(Vc)
which yield, after elimination of the trivial roots VE = VC = I, the expressions
I I % % - - (1 - - %/2 + ~ I 2 )
VC~ (96)
%(1 - - a T / 2 )
and
I I I I
0.95
1.0 Q2
0.3
t0.2
T
¢ 0.1
0.1
0 0
0 0.2 OA 0.6 0.8
(X;T -.
Fig. 15. Calculated dependence of the sol fraction, ws, and the number of elastically active network
chains per monomer unit, N , on the extent of transesterification, %. Stoichiomctric mixture of
dicarboxylic acid (M = 188) and diepoxide (M = 340). The extent of addition esterification
ac = ~E (1.0, 0.99, 0.95) is indicated 74~
Network Formation in Curing of Epoxy Resins 51
1.0 I I I f f''- 2
E-
E
.20
-6
0.5 1 E
/
0 I ~f- 0
0 40 80 120 180
t (h)
Fig. 16. Time dependence of the gel fraction, we, and concentration of elastically active network
chains, v, in the stoichiometric mixture of azelaic acid and DGEBA 74~
0.05
0 ~ I ) I0
Fig. 17. Number of EANC elastically active network chains, Ne, calculated from the equilibrium mo-
dulus as a function of the gel fraction, we, in the stoichiometric mixture of azelaic acid and DGEBA.
The curves are calculated theoreticallyfor the extent of addition esterification ac = ~E indicated v4,
52 K. Du~ek
R1CHCH2NR3
X/
Oe
I I lost ®
R ~CHCH2NR 3 + CO CO ~ R ICHCH2NR3
\/
Oe %/ i
OCO CO0°
I I
Propogcztion :
--COO e + R~C.HCH2 k~ RICHCH2OCO -
l I
Oe
- - ] kA
--C--~ + ~O CO ~ IllCOCO COO O
\o / [--J
It has been proved that the tertiary amine is irreversibly bound to the epoxide and
that the tertiary nitrogen atom is transformed into a quarternary one. In the absence
of anhydride, the equilibrium is strongly shifted to the initial components. As soon
as the anhydride is added, the concentration of the quaternary nitrogen atom
starts to increase. Further chain growth occurs by anionic mechanism. The alkoxy
anions have not been detected in the NMR spectra which can be explained by a
faster reaction of alkoxide anion with anhydride than of the carboxylate anion with
epoxide (kA > kE). Thus, the epoxide-anhydride reaction is an initiated reaction.
However, the reaction may be complicated by (a) the presence of acid in the anhy-
dride, (b) possible regeneration of the tertiary amine and reinitiation. Acid is always
present in commercial anhydrides and it is difficult to remove it completely. If
acid is present, initiation can occur by interaction of the carboxyl with tertiary amine
yielding the carboxylate anion. Tertiary amine is not chemically bound and the acid
acts as an initiator. The problem of a possible regeneration and reinitiation is not yet
clear and experiments with strictly proton-donor free cyclic anhydride and diepoxide
are desirable. However, any significant regeneration and reinitiation should yield
modified end groups and make the molecular weight distribution to change from the
Poisson type distribution to the most probable one. This has not been observed so, 1lm.
Network Formation in Curing of Epoxy Resins 53
IN E--E E
E\A /E \A A I A/
t -Ar \/ \ /
a "E~ b E E
Fig. 18. Schematic tree-like representation of the structures monoepoxide-cyclic anhydride (a), and
diepoxide-cyclic anhydride (b)
where ~l, ~A and % are molar conversions of the initiator, cyclic anhydride and
epoxy groups, respectively. The quantity al = 1 - - ~ 2 is the probability that a
reacted A group is the living end, Pt = 1 - - PA is the probability that the neighbour
of the epoxy group (in the left direction) is a bound initiator and ~1 = 1 - - e2 is the
probability that the reacted epoxy group is a living end. These probabilities are
obtained from stoichiometric considerations and from reaction kinetics (ratios of
rate constants k , kA, kE). The pgf's (I 02-104) are a function of five variables zn, ZIE,Zig,
Z,E, Z~a, which are related to bonds extending to the respective neighbour in the
respective direction. Five pgf's for units in generations g > 0 correspond to five
variables z. They are obtained by differentiation of the components F o , FoA and FOE.
Further handling of these pgf's is routine. For details see Ref. 72. A dependence of the
critical conversion on the initiator/monomer ratio follows from this treatment,
whereas for the catalyzed step polyaddition the gel point conversion does not
depend on the concentration of the catalysts. Also, the molecular weights of the linear
polyesters obtained from monoepoxide and cyclic anhydride depend on the concen-
tration of tertiary amine 6o~.
0.6 I I
• •
o
0.4 o
T
Ixt
o . / - / q)
©
Fig. 19. Critical conversion of epoxy groups
at the gel point, aE, for the stoichiometric system
DGEBA-hexahydrophthalic anhydride as a func-
0.2 tion of the relative concentration of the tertiary
amine catalyst (initiator), q = [I]o/[E]o using
various analytical method and reaction condition
(for details cf. Ref. ~z))~ _. . . . . . theoretical de-
pendence taking into account the presence o f 2 %
1 I acid in the anhydride
0 0,05 0.10
q - ~,.
It has been stressed throughout this review that the progress in the application of the
branching theory to network build-up is dependent on the elucidation of the mecha-
Network Formation in Curing of Epoxy Resins 55
nism and kinetic features of curing. From the theoretical point of view, the develop-
ment of statistical and kinetic methods will certainly cover the network build-up
in the majority of systems.
Ionic curing e.g. with BFa-amine complexes, where propagation occurs by ionic
mechanism 116,t17), seems to be tractable theoretically on the basis of combination
of the cascade and kinetic methods. Progress has been made also in the elucidation
of the mechanism of curing with imidazoles ~1s-~20), but it has been shown to
depend on the structure of imidazole Hg~ Thus, the initiation step for 2-ethyl-4-
methytimidazole involves formation of an addition product
CH3 CH3
R~H~H2
A f,
\/ + !
0
C2Hs C2Hs
/ N = N - N=(CH z(~HCH2OPh)2
CH 2 O OH
\CH /
I
CH2OPh
which is a basic catalyst and also a bifunctional initiator for polymerization of epoxy
groups. Also, the elucidation of the mechanism of action of N,N-dimethylurea
accelerators (e.g. Monuron) should be mentioned 12z). The dimethyturea derivatives
are split to yield dimethylamine which reacts with the epoxide and induces polymeri-
zation of epoxy groups. It can be regarded as a monofunctional initiator.
It should be stressed, however, that at present practically no experimental data
on network formation (critical conversions, molecular weights, sol fractions, etc.)
in these systems are available.
A detailed study of the model reaction of tert.amine catalyzed polyetherification
published by Bergerand Lohse 123~has shown that a considerable fraction of products
of reaction of p-cresyl glycidyl ether and benzyldimethylamine are nitrogen-free oligo-
mers
56 K. Du~k
C H 2 ~ C - - O - - - ( - - - C H 2 ~ CH--O - - } n - - H
1 I
CH 2 CH2
1 I
o o
CH3 CH3
At cure temperatures higher than 150 °C, also the isopropanol derivative
@o OH
5 Conclusions
The network formation theory has proved itself capable to treat network formation
in curing of epoxy resins in terms of the dependence of various structural parameters
on reaction conversion. However, the progress in the application of the theory is
dependent on the state of knowledge about the chemistry and chemical kinetics of
curing. At present, only the network formation in simple polyepoxy-polyamine
systems seems to be satisfactorily understood and the applicability of the theory
confirmed by experiments. It is not yet so e.g. for systems involving N,N-diglycidyl-
aniline and its derivatives where the interaction between the glycidyl groups seems to
play a role. However, a progress in this respect is expected to be reached soon. The elu-
cidation of network formation involving initiated polyetherification of epoxy groups
seems to be one of the major issues for the near future.
In conclusion, the importance of understanding the relationships between network
formation (curing) and network structure should be stressed:
(a) It enables the selection of monomers and curing conditions to control the
network structure and processing.
(b) It helps to elucidate the mechanism of curing reactions, which must be
consistent with the network build-up.
(c) It represents a necessary bridge for passing from the initial composition
of the epoxy-curing agent system and depth of the during reaction to the complex
of physical and physico-chemical properties of the epoxy-curing agent compositions
during and after cure. In this respect, much more theoretical work on correlation
of branched and crosslinked structure with properties is needed.
Network Formation in Curing of Epoxy Resins 57
Note
Since the time o f submission of this chapter, several papers on network formation in
curing o f epoxy resins have been published. N o substantial b r e a k t h r o u g h in this field
has occured, but the a u t h o r feels that some amending facts m a y m a k e the picture more
complete.
O f the studies addressing the (in)homogeneity o f cured epoxy resins, the small-
angle neutron scattering o f D G E B A cured with deuterated m-phenylenediamine is
o f interest. 124~In the range of real space 30--400 A, the constant excess S A N S intensity
could be attributed to a uniform distribution o f the curing agent.
In contrary, the dynamic-mechanical behaviour o f D G E B A - t r i e t h y l e n e t e t r a m i n e
(TETA) networks has been interpreted as reflecting an inhomogeneity in crosslink-
ing. 12s~It was found that the modulus and Tg pass through a m a x i m u m as a function
o f the concentration o f the curing agent. The m a x i m u m values were found for 14 wt.-
o f T E T A . However, the lowering o f the crosslinking density a n d T~ in off-stoichio-
metric systems relative to the stoichiometric one has been predicted theoretically
(see Section D.2) and confirmed experimentally lo3-lo5~. The concentration o f T E T A
at which the modulus and T~ exhibit m a x i m a (14 wt.-%) is just the stoichiometric
concentration o f this curing agent. Thus, these results do not offer any evidence o f
inhomogeneous crosslinking.
6 References
1. Kenyon, A. S., Nielsen, L. E.: J. Maeromol. Sci. A 3, 275 (1969).
2. Racich, J. L., Koutsky, J. A. : J. Appl. Polym. Sci. 20, 2111 (1976)
3. Kreibich, U. T., Schmid, R.: J. Polym. Sci., Polym. Symp. 53, 177 (1975)
4. Errath, E. H., Spurr, R. A. : J. Polym. Sei. 35, 391 (1959)
5. Errath, E. H., Robinson, M. J. : J. Polym. Sci. C3, 65 (1963)
6. Du~ek, K., et al.: Polymer 19, 931 (1978)
7. Oberlin, A., et al. : J. Polym. Sci., Polym. Phys. Ed. 20, 579 (1982).
8. Bantle, S., et al.: Polymer 23, 1889 (1982)
9. Wu, W., Bauer, B. J.: Polymer Commun. 26, 39 (1985).
10. Jarry, J. P., Patterson, G. D.: Macromolecules 14, 1281 (1981)
I 1. Stevens, G. C., Champion, L. V., Liddell, P. : J. Polym. Sci., Polym. Phys. Ed. 20, 327 (1982)
12. Bogdanova, L. M., et al.: Polym. Bull. 4, 119 (1981)
13. Du~ek, K., Bleha, M., Lufi~ik,S. : J. Polym. Sci., Polym. Chem. Ed. 15, 2393 (1977)
14. Rozenberg, B. A.: Adv. Polym. Sci. 75, 113 (1985)
15. Du~ek, K., Ilavsk~,, M., Lufi~tk, S. : J. Polym. Sci., Polym. Syrup. 53, 29 (1975)
16. Lufi~tk,S., Du~ek, K.: J. Polym. Sci., Polym. Syrup. 53, 45 (1975).
17. Du~ek, K., Ilavsk~, M. : J. Polym. Sci., Polym. Phys. Ed. 21, 1323 (1983)
18. Ilavsk~, M., Bogdanova, L., Du~ek~ K. : J. Polym. Sci., Polym. Phys. Ed. 22, 265 (1984)
19. Du~ek, K.: Network formation in chain erosslinking (co)polymerisation. In: Developments in
polymerisation. 3. Haward, R. N. (ed)., London: Applied Science Publishers 1982
20. Goethals, E. J. : Adv. Polym. Sci. 23, 103 (1977)
21. Meijer, E. W., et al.: Polymer Commun. 26, 34 (1985)
22. Bailey, R. T., North, A. M., Pethrick, R. A.: Molecular motion in high polymers. Oxford:
Clarendon Press 1981
23. Topotkaraev, V. A., et al.: Vysokomol. Soedin. A21, 1515 (1979)
24. Topolkaraev, V. A., et al. : Vysokomol. Soedin. A21, 1655 (1979)
25. Raspopova, E. N , et al. : Vysokomol. Soedin. B16, 434 (1974)
58 K. Du~ek
Editor: R. Du~ek
Received August 6, 1985
Photocrosslinking of Epoxy Resins
F. Lohse a n d H. Zweifel
C I B A - G E I G Y A G , Central Research Laboratories,
CH-4002 Basle, Switzerland
In this survey, the current status of knowledge with regards to the structures and reactivities
of photoinitiators for epoxies is presented. Especially aryldiazonium, diphenyliodonium, triphenyl-
sulfonium salts and a new class of organometallic cationic photoinitiators are discussed. DSC experi-
ments show that these polymerization reactions have to be considered as dual-step processes.
After irradiation, a thermal activation for complete crosslinking is necessary.
In general, photolysis either leads to protonic acids or Lewis acids, which initiate a cationic
polymerization of epoxies. However, the formation and initiation step of the active species of
several initiators are not yet fully clarified.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3 Photoinitiators . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.1 Aryldiazonium Salts . . . . . . . . . . . . . . . . . . . . . . . 65
3.2 " O n i u m " Salts a n d Related C o m p o u n d s . . . . . . . . . . . . . . 66
3.30rganometallic Compounds . . . . . . . . . . . . . . . . . . . . 69
3.4 Miscellaneous Structures . . . . . . . . . . . . . . . . . . . . . 76
6 Application Characteristics . . . . . . . . . . . . . . . . . . . . . . 77
7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
1 Introduction
Epoxide resins can be crosslinked by polyaddition ~-4~ of active hydrogen-
containing coumpounds, e.g. carboxylic acids, anhydrides (via intermediate ester-
acid steps), amines, phenols, etc. or by polymerization via ionic mechanisms 5,6)
These reactions are generally started by application of heat.
Aside from the technical importance of the photopolymerization of acrylates and
unsaturated polyesters v - n ) as well as the commercially utilized photodimerization
of cinnamates and chalcone derivatives 7-10~, photocrosslinking of epoxies has
become a field of increasing interest v-~4~. In contrast to free radical polymerization
of C: C unsaturated compounds, cationic polymerization of epoxies is not inhibited
by oxygen. The presence of any strong nucleophiles such as amines does, however,
inhibit polymerization. Compared with "conventional" means of effecting cross-
linking by polyreactions (mixing, heating), the use of irradiation brings many
advantages:
- High reaction rates and hence short access times
-
solvents]
-Selective crosslinking possibilities upon imagewise exposure leading to a relief
-
/o\ R /o\
where only the epoxy groups are available for crosslinking;
2) Epoxy resins of the general formula
a) / O \ R-L
or
b) / O x k R-L-R /O~
C H~H C H,)O-"(.
t ~'~C '--(,\ ,)--OCH2C~HCH2 + Initiator
I I
CH, H H CH 2 ,H3
C
c.,
ri t - . . . . . . . . CH
. .3 . . t-----~
i I ~H2
~HCH)O OCHl H
"-I" . . . . . . . t
~CN,~M H, cN~
CH C H 2 ~ C ~
c,.7 o -x=/~. 3
Fig. 1. Schematic network structure formed by polymerization of bisphenol-A diglycidyl ether with
repeating network unit
The schematic network structure shows that the physical properties of these polymers
depend upon the backbone structure of the epoxy resin and upon the achieved
crosslink density. Since it is very difficult to follow polymerization mechanisms and
kinetics in such systems, it is also difficult to determine the exact degree of poly-
merization and the average size of the network mashes. By comparison of the glass
transition temperatures Tg of crosslinked epoxy resins based on Bisphenol-A di-
glycidylether reacted via thermal cationic or anionic polymerization 16) with
analogous resins obtained by photoinitiated cationic polymerization, it can be
deduced that the average crosstink densities are similar in either case, values for n
being found in the range of 3 to 5 17,~8)
In recent years, several classes of cationic photoinitiators have been found and
described in the literature. A survey is given in Table 1.
64 F. Lohse and H. Zweifel
However, a large number o f these initiators are too slow to be o f practical value.
F r o m a commercial point o f view, the most significant catalysts are aryldiazonium,
triphenylsutfonium, diphenyliodonium and iron arene salts which possess anions
o f low nucleophilicity. The most suitable epoxy resins for cationic polymerizations
are those described in Table 2, e.g. technical Bisphenol-A diglycidylether o f formula I
(n ,-, 0.15) and oligomers (n ,-~ 2.14; 5.1; 11.8), cycloaliphatic epoxies based on
cyclohexene oxide derivatives shown in formulas I I - I V , hexahydrophthalic diglycidyl
ester V or multifunctional novolacs VI. Nitrogen-containing epoxies, such as hydan-
toin derivatives, triglycidyl isocyanurate or glycidylized amines give no satisfactory
results.
The cationic polymerization mechanisms by which these initiators (Table 1) work
were examined only in few cases. Such investigations were based on the poly-
merization o f monoepoxides and on the analysis o f the intermediate and final
reaction products. However, the results can clarify crosslinking o f technical epoxy
resins only to a certain extent. It has to be taken into account that these resins
are sold only in a commercial grade, they all contain small amounts o f by-products,
catalysts etc. which can influence and alter the mechanisms as established with
low-molecular epoxy c o m p o u n d s 17"18). Nevertheless, these c o m m o n l y available
epoxies are useful as technical working materials.
Photocrosslinking of Epoxy Resins 65
Table 2. Most suitable epoxy resins for cationic polymerization
O
I!_ --CH
II
Go.. III
°
IV
/°k
--O-"CHz--CH - - C H z
V
O--CHz-'CH ~ C H z
\o /
V1
L R An R
R = H or CH3
3 Photoinitiators
3.1 AryldiazoniumSalts
The first efficient catalysts for the photopolymerization of epoxides to be found
were aromatic diazonium salts with anions of low nucleophilicity 19-24). Upon
66 F. Lohse and H. Zweifel
irradiation, these salts liberate the corresponding Lewis acid (Fig. 2), which rapidly
polymerizes the epoxides.
MX n -~ Y + Nz + MXn_, S
Similarly, arytdiazonium salts containing BF2 19,23), PF6, FeCI4-, AsF~-, SbF6,
SbCIr- 20,23.24) as anions generate upon photolysis BF3, PFs, FeC13, AsFs, SbF5,
or SbC15, respectively; other salts which liberate either trifluoromethane sulfonic
acid 25) or perchloric acid 26) upon photolysis are also known. The efficiency of
aryldiazonium salts as photoinitiators depends upon the structure of the cationic
and anionic moieties of these salts 24LThe spectral sensitivity can be varied throughout
the UV and the blue region of the spectrum by modifying the structure of the
aryl rest of the aryldiazonium compound 23). Photoinitiators of this class require
generally a thermal post-treatment step after irradiation to achieve satisfactory cure
of the epoxy resin.
However, several inherent drawbacks limit the utility of aryldiazonium salts as
photoinitiators in a number of practical applications for epoxy curing. Nitrogen
evolution during photolysis of the initiator causes bubbles and pinholes in coatings.
Other problems arise from the poor thermal stability of aryldiazonium compounds
and from their inherent sensitivity to moisture. The addition of stabilizing additives
such as nitriles z7), amides 28), sutfoxides 29) and poly(vinylpyrrolidone) 3o) has proven
effective in extending the solution stability of aryldiazonium salt/epoxy mixtures.
Major
ArzICXG ~ (Arll~®) 'j ~ Arl~- 4- kr. + X O
Aria.+R-H ~ Arl*H + R -
Minor
(Ar2leX~)'+ R-H --~ (ArRH)~ + ArI + X e
OH r o¢ 0 0 "]
R R2.~
4.- HX
/k
R R
Fig. 5. Mechanism of ylid and Bronsted acid formation by photolysis of dialkylphenacyl sulfonium
salt and dialkyl-4-hydroxyphenylsulfonium salt (X- = PF6, BF£, SbF6 etc.)
-1-
<,
Fig. 6. DSC diagram of the polymerization of Bisphenol-A diglycidylether with 2.5 ~ (w/w) Ph3S÷PFg
Irradiation was carried out at --88 °C; rate of heating was 20 °C/rain
It clearly shows that the polymerization reaction proceeds at a low rate at room
temperature. Although it is possible to obtain tack-free films by irradiation, the
sample should be heated subsequently in order to complete the polymerization of the
available epoxy groups within a reasonable period of time.
Photocrosslinking of Epoxy Resins 69
lI
2oo
Z
3.30rganometallic Compounds
Organometallic compounds which can act as photoinitiators for the epoxy poly-
merization have been described in the past. Strohmeier showed that manganese
decacarbonyl is a photoinitiator for epichlorohydrin 77). Various other organometal
carbonyl compounds are listed in the literature 7s-so) in addition to dicarbonyl
chelates of group IIIa, IVa and Va elements s~). Stark and coworkers ~5) described
a system which is based on the photoinduced curing of epoxy resins with anhydrides.
Kaeriyama has mentioned the use of ferrocene a2) and zirconocene s3) dichloride
as photoinitiators for epichlorohydrin and phenyl glycidyl ether. Organohalogen
compounds in conjunction with organometal derivatives or organometal aluminium
complexes 85, s6) also have been reported as cationic photoinitiators.
70 F. Lohse and H. Zweifel
Recently, Meier and Zweifel have described that iron arene salts 73, 87-90, having
anions with low nucleophilicity are highly efficient photoinitiators for the epoxy
polymerization. A similar result in this area also has been reported by Palazzotto and
Hendrickson 80). The iron arene salt photoinitiators seem to be the most promising
compounds of all the organometal initiators listed above.
Iron arene salts are generally prepared from ferrocene according to the method
reported by Nesmeyanov 9~). Various types of such complexes described by the
following general formula are listed in the literature 92).
I
Fe Xe
x = a 2, SU=
Upon irradiation, iron arene complexes loose the uncharged tridentate arene ligand
and thereby yield a Lewis acid, as shown in Fig. 8.
J
Fe PF6E)
hv
,~=t
tP, I
Fe •
Q
thermally stable LEWIS ACID
light sensitive
2 * 8 / 0N h~ = Fe (pFee)2 . FeCi ~
Considering these results, it appears likely that some ligand-exchanged iron complex
having three coordinated epoxide functions is formed in technical epoxy resins during
the photolytic step. Ring opening and polymerization could thus start in the
ligand sphere of the iron cation leading to cyclic polyethers (Fig. 10). But in the
polymerization reactions with bi- or multifunctional epoxides, ferrocene formation
has not been detected.
Fe X~ hv ~ X
r
/o\. c/°',
A_ B
F,,, °
C
m
x
t I"~ Polymerization
From the DSC experiment, it is evident that the iron cation is less reactive in
epoxide polymerization then Bronsted acids obtained from, e.g. sulfonium salts
(see Fig. 11). Therefore, a heat treatment after the illumination step is necessary. The
narrow enthalpy peak observed could refer to a very uniform pathway.
80-
60-
Fig. 12. Viscosity versus reaction
~o- time of a sample of Bisphenol-A
diglycidyl ether with 2.5~o (w/w)
20- (T16-benzene)(rlS-cyclopentadienyl)-
iron(II)-hexafluorophosphate after
irradiation at 20 °C
0 ~, ............ ~,
Time (h)
300
~ \\\\
$ Tgoo
O.
E
(I)
]00
C)
fication will start. Now the reaction is diffusion-controlled and is halted only when
crosslinking is complete (Tg = 115 °C). To achieve complete cure and thereby
develop ultimate material properties of the cured resin, vitrification during cure should
be avoided. All these data show that for complete crosslinking of epoxies initiated
by photochemical cationic species, a thermal post-cure treatment is necessary.
Figures 14 and 15 show the relations between the amount of iron arene initiator,
the reaction enthalpy (AH) and the glass transition temperature Tg of the polymerized;
Bisphenol-A diglycidylether (cf. Table 2, structure I, x = 0.15) and the oligomer
product based on the former compound (cf. Table 2, structure I, x = 11.8). The
maximum polymerization heat per mole of epoxide is observed with an initiator
concentration of 1.5-2.5 ~ (w/w). At this concentration, Tg of the crosslinked resin
is about 115 °C for the polymerized low-molecular-weight expoxide and about 80 °C
for the polymerized high-molecular-weight epoxide resin.
The relation between the epoxy content of various epoxy resins; based on Bis-
phenol-A diglycidylether and AH and the glass transition temperatures of linear and
cured resins is given in Table 3.
The values obtained in these experiments confirm that the polymerization heat
AH is a function of the amount of epoxy groups in the oligomeric products based on
Bisphenol-A diglycidylether. This correlation shows further that AH is about 50 to
60 kJ per mole epoxy group. The Tg of the cured resins decreases, as expected, with
an increase of the chain length of the advanced epoxy resins 97)
80.
0) II
-6 60- x = 0.15
E x=11.8
~o-
I/,0.
~ 100,
i__~
x=1t.8
80.
Fig. 15. Relation between the
60. II initiator concentration and T~
l ~ ~ ~ (DSC)
Initiator concentration [ w / w ] ( * / , )
74 F. Lohse and H. Zweifel
Table 3. Relation between the epoxy content, heat of polymerization AH and T~. Initiator concentra-
tion: 2,5 % (w/w)
+
,o, F/=,,
H2C ~ C H 2 - - CH2-- O'[--~\
/~-- C'-~t
~"',¢~ .
"~Y-O -- C H2-- C -- CH2- O "1~"
] ~ ~.y=~ ,o,
"~Y'-C "-~\ /,)'- O-- CH2--CH-- CH2
I ~ I X~/ I I ~ I "X~_..~
LCH3 OH JX CH3
b
? o
PI31,,k+k=n3
I A f--yc-o-c.~-c.-c.~
Fe PF60 "1" C ~ 0
O- CH2-Co/CHz Y ~~J / -',n ",,,.~C-o O - CH2-CH--CH
2_O.
. . . . . . . 0-CH2-CH-CH2 ~
-/ \/
Fig. 16. DSC diagram of the polymerizations of different epoxy derivatives with (rl6-benzene)(q s-
cyclopentadienyl)iron(II)-hexafluorophosphate, 2 . 5 ~ (w/w); after irradiation at --88 °C, rate of
heating: 20 °C/min
Photocrosslinking of Epoxy Resins 75
The reaction rate of the iron-catalyzed polymerization depends upon the nucleo-
philicity of the anion employed as well as upon the structure of the oxirane group.
The polymerization rate decreases in the order SbF6 > AsF6 > PF6 > BF£ 8o~
From the DSC experiment, it is evident that the cycloaliphatic epoxides are more
reactive than the glycidyl ethers, as shown in Fig. 16. The presence of oxygen-
containing functional groups other than the epoxide moieties decreases the reactivity
further as, e.g. in the case of glycidyl esters 9o~.
The oxidation state of the comptexed iron in these photoinitiators is 2 + before and
after exposure to actinic radiation. Oxidation of the complexed iron to the 3 ÷ state
is possible after the irradiation step. Irradiation in the presence of an oxidant
yields a stronger Lewis acid with increased reactivity for the epoxide polymerization.
6 5b 16o 1~
Temperature {oC )
Fig. 17. DSC diagram of epoxide polymerizationswith 2.5 ~o (w/w) photoinitiator in the presence
and in absenceof the oxidant cumenehydroperoxide
,ooo . . . . . . ,,
~ Fe PFs
i
\~x I i,""~
"~.
Fig. 18. UV/vis spectra of iron arene com-
0 ~'".b
plexes (CH2CIz)
300 ~00 560 600
A. (nm) -----,.-
Various bifunctional resins are based on acrylic epoxide monomers. Such systems
can photopolymerize by the radical and/or cationic mechanism. With iron arene
photoinitiators in the presence of an oxidant, radical as well as cationic photo-
polymerization of these monomers is possible 88). "Onium"-type photoinitiators
form radical species upon photolysis, as shown in Figs. 3 and 4. The local
radical concentration is, however, too low to permit the polymerization of such
systems lo5).
Recently, Ledwith described combined systems lo6,1o7), composed of a radical
initiator and a cationic photoinitiator, which are very suitable for hybrid systems.
It is particularly convenient to employ common photochemical sources of free
radicals tbr this purpose. Since many of them possess aromatic carbonyl groups
(e.g. benzoin and acetophenone derivatives), these groups provide an extension of
the absorption to longer wavelengths and promote the initiation of cationic
polymerization. Figure 19 shows the proposed formation mechanism of free-radical
and cationic active species by electron transfer:
Photocrosslinking of Epoxy Resins 77
For applications, such hybrid systems are limited to either diaryliodonium salts
or aryldiazonium photoinitiators and suitable radical initiators. Triphenytsulfonium
salts, however, are not active as cocatalysts in the presence of free-radical
initiators.
"o' L ',.="
Fig. 20. Photocrosslinkableepoxy resin with light-sensitivechalcone groups
6 Application Characteristics
Cationic photoinitiators are of basic interest for coating applications, printing
plates for silk screen printing inks, photoresists, name plates etc.
78 F. Lohse and H. Zweifel
7 References
1. Wegler, R., Schmitz-Josten, R., in: Methoden der Organischen Chemie, (Houben-Weyl, ed.)
Bd. XIV/2, pp 462, G. Thieme Verlag, Stuttgart 1963
2. Lee, H., Neville, K. : Handbook of Epoxy Resins, McGraw-Hilt, New York 1967
3. May, A., Tanaka, Y.: Epoxy Resins, Chemistry and Technology, Marcel Dekker Inc.,
New York 1973
4. Lee, H., Neville, K. in: Encyclopedia of Polymer Science and Technology, Vol. 6, p 209,
Interscience Publ., New York t967
5. Ivin, K., Saegusa, T.: Ring-Opening Polymerization, Vol. 1, p 185, 234, Elsevier Applied
Science Publ., London 1984
6. Penczek, S., Kubisa, P., Matyjaszewski, K. : Advances in Polymer Science, Vol. 37, Springer
Verlag, Heidelberg 1980
7. Green, G. E., Stark, B. P.: Chemistry in Britain 17, 228 (1981)
8. Green, G. E., Stark, B. P., Zahir, S. A.: J. Macromol. Sci., Revs. Macromol. Chem. C21,
187 (1982)
9. Delzenne, G. A.: Adv. Photochem. 11, 1 (t979)
10. Finter, J., Haniotis, Z., Lohse, F , Zweifel, H.: Angew. Makromol. Chem. 133, 147
(1985)
11. Crivello, J. V., in: UV Curing, Science and Technology, (S. J. Pappas, ed.) Stamford, Techn.
Marketing Corp., p. 23, 1978
12. Crivello, J. V., in: Developments in Polymer Photochemistry 2, (N. S. Allen, ed.) p. 1,
Applied Science Publ., London 1981
13. Smets, G., Aerts, A., van Erum, J.: Polymer J. 12, 539 (1980)
14. Perkins, W. C.: J. Rad. Curing 16 (1981)
15. Brown, D. L. S., Connor, J. A., Dobinson, D., Stark, B. P. : Angew. Makromol. Chem. 50, 9
(1976)
Photocrosslinking of Epoxy Resins 79
16. Fischer, M., Lohse, F., Schmid, R. : Makromot. Chem. 181, 1251 (1980)
17. Berger, J., Lohse, F.: J. Appl. Polym. Sci. 30, 531 (1985)
18. Berger, J., Lohse, F.: Eur. Polym. J. 21,435 (1985)
19. Licari, J. J., Crepeau, P. C. : US Patent 3205157 (1965)
20. Schlesinger, S. I. : US Patent 3708296 (1973)
21. Fischer, E. : US Patent 3236784 (1966)
22. Pinot de Moira, P., Murphy, J. P.: US Patent 3930856 (1976)
23. Schlesinger, S. I.: Photogr. Sei. Eng., 18, 387 (1974)
24. Schlesinger, S. I.: Polym. Eng. and Sci., 14, 513 (1974)
25. DiPippo, C. A.: US Patent 4482489 (1980)
26. Cripps, H. N. : US Patent 3347676 (1967)
27. Watt, W. R.: US Patent 3721616 (1973)
28. Watt, W. R.: US Patent 3721617 (1973)
29. Feinberg, J. H. : US Patent 3711391 (1973)
30. Feinberg, J. H.: US Patent 3816281 (1974)
31. Smith, G. H.: DOS 2639395 (1977)
32. Smith, G. H. : DOS 2520489 (1976)
33. Nemcek, J. R.: GB Patent 1539192 (1975)
34. Nemeek, J. R. : DOS 2602574 (1976)
35. Crivello, J. V. : US Patent 3981897 (1976)
36. Crivetlo, J. V.: US Patent 4026705 (1977)
37. Crivello, J. V., Lam, J. H. W.: Maeromolecules 10, 1307 (1977)
38. Crivello, J. V. : Chemtech, 10, 624 (1980)
39. Crivello, J. V., Lam, J. H. W.: J. Polym. Sci., Polym. Chem. Edn. 17, 3845 (1979)
40. Irving, E.: Eur. Patent 104143 (1985)
41. Irving, E.: Eur. Patent 106797 (1985)
42. Crivello, J. V. : US Patent 4058401 (1977)
43. Crivello, J. V.: US Patent 4136102 (t979)
44. Crivello, J. V. : US Patent 4108747 (1978)
45. Crivello, J~ V. : U S Patent 4138255 11979)
46. Smith, G. H.: US Patent 4069054 (1978)
47. Chang, K.-T. : US Patent 4197174 (1980)
48. Watt, W. R. : US Patent 420t 640 (1980)
49. Chang, K.-T. : US Patent 4247473 (1981)
50. Ellwood, M.: Eur. Patent 142384 (1984)
51. Crivello, J. V., Lam, J. H. W.: J. Polym. Sci., Polym. Chem. Edn. 17, 977 (1979)
52. Crivello, J. V., Lam, J. H. W.: J. Polym. Sci., Polym. Chem. Edn. 18, 2677 (1980)
53. Crivello, J. V., Lam, J. H. W. : J. Polym. Sci., Polym. Chem. Edn. 18, 2697 (1980)
54. Watt, W. R., Hoffmann, H. T., Pobiner, H., Schkolnick, L. J., Yang, L. S.: J. Polym.
Sci., Polym. Chem. Edn. 22, 1789 (1984)
55. Green, G. E., Irving, E. : Eur. Patent 022081 (1980)
56. Green, G. E., Irving, E. : Eur. Patent 035969 (1981)
57. Green, G. E., Irving, E.: Eur. Patent 044274 (1981)
58. Green, G. E., Irving, E., Stark, B. P.: Eur. Patent 054509 (1981)
59. Crivello, J. V., Lain, J. H. W. : J. Polym. Sci., Polym. Chem. Edn. 17, 2877 (1979)
60. Crivello, J. V., Lam, J. H. W.: J. Polym. Sci., Polym. Chem. Edn. 18, 1021 (1980)
61. Crivello, J. V., Lam, J. H. W.: J. Polym. Sci., Polym. Letters Edn. 16, 563 (1978)
62. Ketley, A. D., Tsao, J. H. : Polymer Preprints, Amer. Chem. Soc. 19, (2), 656 (1978)
63. Ketley, A. D., Tsao, J. H.: J. Radiat. Curing 6, (2), 22 (1979)
64. Crivelto, J. V., Lain, J. H. W.: J. Polym. Sci., Polym. Chem. Edn. 17, 1047 (1979)
65. Crivelio, J. V. : US Patent 4069055 (1978)
66. CriveUo, J. V.: US Patent 4219654 (1980)
67. Crivello, J. V.: US Patent 4234732 (1980)
68. Crivello, J. V.: US Patent 4250311 (1981)
69. Crivello, J. V., Lam, J. H. W. : J. Polym. Sci., Polym. Chem. Edn. 16, 2441 (1978)
70. Crivello, J. V., Lam, J. H. W.: J. Polym. Sci., Polym. Chem. Edn. 17, 1059 (t979)
71. Pappas, S. P., Jilek, J. H.: Photogr. Sci. Eng. 23, 140 (1979)
80 F. Lohse and H. Zweifel
72. Pappas, S. P., Pappas, B. C., Gatchair, L. R.: J. Polym. Sci., Potym. Chem. Edn. 22,
69 (1984)
73. Meier, K., Zweifel, H. : Rad'Cure Europe, Basle, 1985, Technical Paper FC 85-417.
74. Aim, R. R., Carlson, R. C. : Rad'Cure Conference, Chicago, 1982, Technical Paper SME.
75. Berger, J., Lohse, F. : J. Polym. Sci., Polym. Letter Edn. 23, 227 (1985)
76. Yasunobu, O. et al. : Eur. Patent 136679 (1984)
77. Strohmeier, V. W., Barbeau, C. : Makromol. Chem. 81, 86 (1965)
78. Anderson, W. S. : US Patent 3709861 (1973)
79. Irving, E., Johnson, B., Meier, K. : Eur. Patent 094914 (1984)
80. Palazzotto, M. C., Hendrickson, W. A.: Eur. Patent 109851 (1983)
81. Celia, J. A.: US Patent 4086091 (1978)
82. Kaeriyama, K.: J. Polym. Sci., Polym. Chem. Edn. 14, 1547 (1976)
83. Kaeriyama, K. : Makromol. Chem., 153, 229 (1972)
84. Roteman, J. : US Patent 3895954 (t974)
85. Hayase, S., Onishi, Y., Suzuki, S. and Wada, M. : Macromolecules 18, 1799 (1985)
86. Tokyo Shibaura Denki KK: Jap. Patent 59043018 (1982)
87. Meier, K., Bfihler, N., Zweifel, H., Berner, G., Lohse, F.: Eur. Patent 094915 (1984)
88. Meier, K., Eugster, G., Schwarzenbach, F., Zweifel, H. : Eur. Patent 126712 (1984~
89. Lohse, F., Meier, K., Zweifel, H.: Proceedings of the llth International Conference in
Organic Coatings Science and Technology, Athenes, 1985, p. 175
90. Meier, K., Zweifel, H. : J. Rad. Curing, in press (1986)
91. Nesmeyanov, N. A., Vol'kenau, N. A., Bolesova, I. N.: Dokl. Akad. Nauk SSSR 149, 615
(1963)
92. Schumann, H. : Chemiker-Zeitung 108, 345 (1984)
93. Gill, T. P., Mann, K. R.: Inorg. Chem., 19, 3007 (1980)
94. Nesmeyanov, A. N., Vol'kenau, N. A., Shilovtseva, L. S. : Dokl. Akad. Nank SSSR 190,
857 (1970)
95. Meier, K., Rhis, G.: Angew. Chemic 97, 879 (1985)
96. Gillham, J. K., in: Thermal Characterization of Polymeric Materials, (Turi, E. A., ed.) p 438,
565, Academic Press, New York 1981
97. Batzer, H., Lohse, F., Schmid, R. : Angew. Makromol. Chem. 29/30, 349 (1973)
98. Meier, K., Zweifel, H.: Eur. Patent 152377 (1985)
99. Gandini, A., Cheradame, H. : Adv. in Polymer Sei., Vol. 34/35, p. 230, Springer Verlag,
Heidelberg, 1980
100. Monsey, E. G., Bown, D. E.: US Patent 3782952 (1974)
101. Schlesinger, S. I. : US Patent 3782952 (t974)
102. Kinstle, J. F., Tufts, T. A. : Polymer Preprints, Amer. Chem. Soc. 20 (2), 661 (1979)
103. Minnesota Mining Manufacturing Comp.: Brit. Patent 912022 (1962)
104. Hitachi Chemical Co. Ltd. : Brit. Patent 1418169 (1975)
105. Kiihl, G. : SME Technical Papers, Rad'Cure Europe, Basle, 1985, Conference Preprints.
106. Ledwith, A.: Polymer 19, 1217 (1978)
107. Ledwith, A. : Makromol. Chem. Suppl. 3, 348 (1979)
108. Green, G. E., Stark, B. P., Waterhouse, J. S.: Brit. Patent 1527416 (1978)
109. Green, G. E., Waterhouse, J. S.: Brit. Patent 1521933 (1978)
110. Dobinson, B., Green, G. E., Hinton, I. G., Hope, P., Martin, R. J., Stark, B. P., Waterhouse,
J. S., Young, E. W. : Makromot. Chem., 181, 1 (1980)
111. Panda, S. P.: J. Appl. Polym. Sci. 18, 2317 (1974)
112. Panda, S. P.: J. Polym. ScL A-l, 13, 1757 (t975)
113. Panda, S./?.: J. Polym. Sci. A-l, 13, 259 (1975)
114. Panda, S. P.: Indian J. Technol. 9, 387 (1971)
115. Panda, S. P.: Chem. and Ind. 1974, 706
116. Zahir, S. A.: J. Appl. Polym. Sci. 23, 1355 (1979)
117. Crivello, J. V.: ACS Symp. Ser. 242, (Th. Davidson, ed.) p. 1 (1984)
118. Meier, K., Zweifel, H. : Polym. Preprints, Amer. Chem. Soe. 26(2), 347 (1985)
119. Rembold, K.-H. : Proceedings of the 'First Printed Circuit World Convention' L 1.11.1 (London,
June 1978)
Photocrosslinking of Epoxy Resins 81
120. Stark, B. P., Rembold, K.-H. : Proceedings of'E.I.P.C. Seminar: Permanent Polymer Coatings',
Basle, 1979; published by European Institute of Printed Circuits, Ziirich
121. Losert, E. H.: Proceedings of the 'Second Printed Circuit World Convention' L 332
(Munich, June 1981).
Editor: R. Du~ek
Received November 12, 1985
Time-Temperature-TransformationffTT) Cure Diagram
of Thermosetting Polymeric Systems
Marc T. Aronhime
Casali Institute of Applied Chemistry, The Hebrew University of Jerusalem,
91904 Jerusalem, Israel
John K. Gillham
Polymer Materials Program, Department of Chemical Engineering,
Princeton University, Princeton, New Jersey 08544, USA
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
1 Introduction
The transformation of low molecular weight liquid into high molecular weight
amorphous solid polymer by chemical reaction is the fundamental process used in
the coatings, adhesives and thermoset industries. As the chemical reaction proceeds,
the molecular weight and glass transition temperature (Tg) increase, and if the
reaction is carried out isothermally below the glass transition temperature of the
fully reacted system (Tg~), the polymer T~ will eventually reach the reaction
temperature (To,re). During isothermal reaction below Tg~, two phenomena of critical
importance in thermosetting processing can occur: gelation and vitrification. For
linear systems, only vitrification will occur. Gelation generally occurs first, and is
characterized by the incipient formation of material of infinite molecular weight.
Prior to gelation, the system is soluble and fusible, but after gelation both soluble
(sol fraction) and insoluble (gel fraction) materials are present. As the reaction proceeds
beyond gelation, the amount of gel increases at the expense of the sol. As gelation
is approached, the viscosity increases dramatically, and the weight average molecular
weight goes to infinity, but the number average molecular weight is small.
Vitrification is the transformation from liquid or rubbery material to glassy
material. At vitrification, the material solidifies and the chemical reactions can be
quenched; Tg can therefore equal or exceed T~,r~. Other parameters of importance
in the thermosetting process include the changes in viscosity and conversion with
time, and thermal degradation at high temperatures.
An isothermal time-temperature-transformation (TTT) cure diagram results if, for
a series of isothermal cures, the temperature of cure (T .... ) is plotted vs. the times
to getation and vitrification 1-~). The T I T diagram provides a framework for under-
standing the cure process of thermosetting (and linearly polymerizing) materials.
Relationships between cure, structure and properties can also be understood by
studying the TTT diagram. The TTT diagram can be extended to include phase
separation (in the case of rubber-modified thermosets, for example), viscosity, thermal
degradation, and extent of conversion.
Much of the behavior of thermosetting materials can be clarified in terms of the
TTT cure diagram through the influence of gelation, vitrification and devitrification
on properties. For example, gelation retards macroscopic flow, and limits the
growth of a dispersed phase (as in rubber-modified systems); vitrification retards
chemical conversion; and devitrification, due to thermal degradation marks, the limit
in time for the material to support a substantial load.
The TTT diagram is a familiar concept in materials science for studying phase
changes 5~. The TTT diagram is a nonequilibrium diagram, since the transformations
occur as functions of time. TTT diagrams have played an important role in the
control of the properties of metals by permitting thermal history paths to be chosen
so that a desired microstructure can be obtained. The diagrams are specific to a parti-
~cular material composition and considerable insight into the design of alloys can be
achieved once the effects of additives on the TTT diagram have been explored. Such
a diagram for thermosetting systems would permit time-temperature paths of cure
to be chosen so that gelation, vitrification, and phase separation occur in a controlled
manner and consequently give rise to predictable properties of the thermosetting
matrix 2).
Time-Temperature-Transformation(TTT) Cure Diagram of Thermosetting Polymeric Systems 85
The purpose of this review is to summarize the basic features and utility of the
T T T cure diagram, discuss the experimental procedures for obtaining a diagram,
present experimentally-obtained diagrams for model systems, and describe recent
models that have attempted to calculate the time to vitrification on isothermal
polymerization. This review will concentrate on T T T diagrams of epoxy systems.
",.%
",,, o rubber ttr/~,.~. ~
Chor
-- ,.
Sol/Gel '~,~ /e -~ - '
rubber "%. qtioo~-~.
L.
~, - ~ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ ' , ~
E
getTg
Tg o
log time
Fig. 1. Schematictime-temperature-transformation(TTT) isothermal cure diagram for a thermosetting
system, showing three critical temperatures (T~o, ~dTg, T~o) and distinct regions of matter (liquid,
sol/gel rubber, gel rubber, sol/gel glass, gel glass, sol glass, char). The full cure line (Tg = Two~)
separates the sol/gel glass from the gel glass region, and the sol/gel rubber from the gel elastomer
region, respectively. Degradation events are devitrification and char formation. The successive
isoviscous contours shown in the liquid region, which were calculated in the absence of phase
separation, differ by a factor of ten. Phase separation occurs in the liquid region prior to gelation
86 M.T. Aronhimeand J. K. Gillham
gelling, in principle quenching the chemical reactions, and thus precluding gelation.
At vitrification below g~Tg, the system is of low molecular weight, and on subsequent
heating the material will flow and is readily processable.
Tg~o is the maximum glass transition temperature of the system. Between g~Tg and
Tg~, the material is initially in the liquid region, and is soluble and of low
molecular weight. On reaction, gelation occurs, and the sol/gel rubber region is entered.
Finite molecular weight sol and infinite molecular weight gel generally form a miscible
binary mixture in this region. Eventually, Tg will rise to T¢.... and vitrification will
be said to occur. Vitrification greatly decreases the molecular and submolecular
mobilities, and thus in principle the chemical reactions are quenched. In the
absence of full cure, the vitrified region formed between ~,~Tg and Tgoo contains
both sol and gel components.
However, recent work investigating the cure behavior after vitrification has shown
that reactions do occur in the glassy state on prolonged isothermal cure beyond
vitrification 4) The "full cure" line in Fig. 1 is the time required, for any given
T ..... for Tg to equal Tg00. Thus, the glass region between g~Tg and Tg~ is
subdivided into two regions: 1) the sol/gel glass region, below the full cure line where
the cure is not yet complete; and 2) the gel glass region, above the full cure line
where in principle the system is fully reacted as determined by Tg measure-
ments.
The determination of full cure is important because meaningful structure-property
relationships between different systems can only be made by considering fully
cured materials. If the cure state of a material is unknown, then its properties can not
be compared with those of another material, whose cure state is also unknown, for the
purpose of relating the observed properties to the chemical structures. For systems
where Tg~ can be attained in the absence of degradation, curing above Tg~ is the
most direct method for achieving full cure. For high temperature systems, where
curing above Tg~o would lead to thermal degradation, the full cure line (Fig. 1)
presents an alternative guide for achieving complete cure; the system can be cured
below Tg~o,thus avoiding degradation.
In a manner similar to the extension of the full cure line into the glassy region,
note that the gelation line is extended beyond the vitrification curve (Fig. 1). This
implies that gelation can occur below g~Tg once vitrification has occurred; this
could affect the storage temperature of partially-reacted materials.
On isothermal reaction above Tsoo, the material will gel but not vitrify in the
absence of degradation. An elastomer (gel rubber) is formed above Tg® after
prolonged isothermal cure, as indicated by the full cure line.
The calculation of the S-shaped time to vitrification curve forms the basis of
Section 4 of this review. The local maximum just above Tgo and the local minimum
just below Tg~ arise from kinetic factors. The minimum in the time to vitrification is
important in molding applications when the mold can not be opened until the system
has solidified. However, as is discussed later, the time to vitrification curve need not
be S-shaped, but can assume a sigmoidal shape, depending upon the values of the
kinetic parameters used in the model.
The isoviscosity contours of Fig. 1 increase by one order of magnitude from one
curve to the one below it. At high temperature the initial viscosity is naturally lower
than the viscosity at low temperature, but the viscosity increases much more rapidly
Time-Temperature-Transformation (TTT) Cure Diagram of Thermosetting PolymericSystems 87
at high than at low temperatures. The fluid state is limited by gelation above gelTg
and by vitrification between T, and ~Tg 6) By visual extrapolation of the isoviscosity
curves, the vitrification curve below ge~Tgappears to be an isoviscous contour: this
is a consequence of the method of computing the isoviscosity curves 6~. The change in
viscosity with time is one of the most important parameters in thermoset processing.
An isoviscosity level is often used as a practical measure of gelation. In the glassy state,
the isomodulus contours would be expected to approximately parallel the vitrification
curve.
Isoconversion curves, if shown, would approximately parallel the gelation line (as
well as the full cure line) because gelation is considered to be an isoconversion state 7)
The extent of conversion after vitrification changes very slowly, but does not cease.
Reference to a complete T I T diagram enables a time-temperature path of cure to be
selected which will follow a desired viscosity-conversion path.
In the case of a rubber-modified thermoset, the cloud point marks the visual onset
of phase separation of the rubber-rich domains from an initially homogeneous solu-
tion of reactants. The cloud point can be observed by a dramatic decrease in the in-
tensity of transmitted light, or a dramatic increase in the intensity of scattered light.
Whereas most of the phase separation will occur prior to gelation, the end of phase
separation can occur after gelation 8). Phase separation is influenced by competing
thermodynamic and kinetic (transport) factors. At low temperatures, nucleation of a
dispersed phase is favored thermodynamically but transport to nucleating sites is
reduced due to the high viscosities and low temperatures. At high temperatures, the
nucleation rate is low, but the low viscosity and high temperatures facilitate the trans-
port. Thus a maximum in the amount of precipitated rubber is anticipated at an
intermediate temperature which could give rise to a minimum in the T .... vs. time to
phase separation curve 2).
At high temperatures thermal degradation becomes important, and may prevent
full cure from being achieved a~. Two degradation events have been noted in relation
to the TTT diagram: devitrification followed by elastomer formation; and vitrifica-
tion followed by char formation. The devitrification event corresponds to a decrease
in Tg from above to below the isothermal cure temperature; the time to this event
may be considered to be the lifetime of the material since it marks the limit in time
for the material to support a substantial load. The second event is an elastomer-to-
glass transformation, accompanied by an increase in Tg and rigidity, and is presumably
due to the onset of char formation 3).
The T T r diagram (Fig. 1) is a convenient tool for summarizing the changes in
material parameters that occur during the thermosetting process. The necessary time-
temperature paths of cure to achieve desired viscosity-conversion levels, the onset and
end of phase separation, and the proper cure paths to avoid thermal degradation can
be gleaned from examination of a complete T T T cure diagram.
the cure process is followed from the initial liquid state, through gelation and
the sol/gel rubber region, and into the vitrified region. The cured material is then
subjected to a dynamic temperature scan at a fixed scanning rate in order to
determine the Tg and other transition temperatures after cure.
A convenient method for determining transition times and transition temperatures
of polymeric materials is dynamic mechanical analysis. One type of instrument which
is particularly suitable for polymeric solids is the freely oscillating torsion pendulum
(TP). Advantages of the TP include its simplicity, sensitivity, relatively low frequency
( ~ 1 Hz) which permits direct correlation of transition temperatures with static non-
mechanical methods (e.g., ditatometry and calorimetry), and its high resolution of
transitions 2~ A major disadvantage of the conventional TP is that test temperatures
are limited by the inability of materials to support their own weight near load-limiting
transition temperatures.
A variation of the TP is torsional braid analysis (FBA)2,9~, in which a small
sample of the material is supported on a multifilamented glass braid. The TBA
technique has the distinct advantage of being able to take measurements throughout
the cure process, since properties can be recorded above the load-limiting temperatures
of the reactants. Monitoring the entire cure process is essential if a ~ cure
diagram is to be constructed. The elastic and loss moduli of the composite
specimen (resin plus glass substrate) are calculated from the damped oscillations of
each wave and changes are interpreted in terms of the properties of the supported
material.
A schematic diagram of the torsion pendulum is shown in Fig. 2. The pendulum
is intermittently set into motion to generate a series of damped oscillations while
the material behavior of the specimen changes with temperature and/or time.
The time scale of the change in the material is much greater than the time scale of
one damped wave. The nondriven, free oscillations are initiated by an angular step-
displacement of the fixed end of the specimen. The natural frequency range of the
vibrations is 0.05-5 Hz. Conversion of the damped oscillations to electrical analogue
signals is accomplished using a transducer.
A key factor in the instrumentation was the development of a nondrag, optical
transducer that produces an electrical response that is a linear function of the angular
deformation of the pendulum. A polarizing disk is employed as the inertial member
of the pendulum, and a stationary second polarizer is positioned in front of a photo-
detector whose response is a linear function of intensity. The intensity of light
transmission through two polarizers is a cosine squared function of their angular
displacement; over a useful range symmetrical between the crossed and parallel
positions of the pair of polarizers, the transmission function approaches linearity.
As the properties of the specimen change, twisting of the specimen may cause the
inertial mass polarizer to drift out of the linear range. An automated control sequence
was designed to compensate for this behavior 2~.
The composite specimen is supported in the cylindrical vertical shaft in a copper
block around which are band heaters, and cooling coils for liquid nitrogen. The
instrument generally operates over a temperature range of --t90 to 400 °C with
a temperature spread of < 1 °C over a two inch specimen, f f h e instrument has been
modified to give greater temperature ranges2~.) A temperature programmer/
controller permits isothermal, linearly increasing and linearly decreasing temperature
Time-Temperature-Transformation (TTT) Cure Diagram of Thermosetting PolymericSystems 89
ALIGNMENT
AND
ATMOSPHERE INITIATION
PORT MECHAN SM
I
SUPPORTING ROD
TEMPERATURE
CONTROLLED
ENCLOSURE
ROD
LIGHT SOURCE
)OW
_.Jj
II
ATMOSPHERE POLARIZER
PORT II " - - - - VACUUM Fig. 2. Automatedtorsion pendulum:
II PHOTOTUSE schematic. An analog electrical signal
I AMPL'P'ERI results fromusinga light beam passing
through a pair of polarizers, one of
l ~ CONTROLLER which oscillates with the pendulum.
AND DATA The pendulum is aligned for linear
ANALYZER response and initiated by a computer
that also processesthe damped waves
I XYY
PLOTTER
to provide the elastic modulus and
mechanical damping data, which are
plotted vs. temperature or time
modes of operation to be made. The atmosphere is tightly controlled, and for cure
studies an inert atmosphere is generally desired; dry helium is used, rather than
nitrogen, because helium has better heat transfer properties at low temperatures.
The substrate used is a heat-cleaned (425 °C/3 h in air) glass braid, two inches long,
containing about 3600 filaments. A braid is used because it has a balanced twist.
The large surface area of the braid permits pickup of relatively large amounts of
fluid and minimizes flow due to gravity. In most TBA experiments, the reactants are
dissolved in a suitable solvent [generally methylethyl ketone (MEK) for epoxy
systems] in the ratio of 1 g solids/1 ml solvent. The solvent is removed from the
solution-impregnated braid in situ by heating the composite specimen above the boil-
ing point of the solvent.
The TBA/TP instrumental system is available from Plastics Analysis Instruments,
Inc., Princeton, New Jersey, USA.
90 M.T. Aronhime and J. K. Gillham
Two mechanical functions of the specimen, rigidity and damping, are obtained from
the frequency and decay constants which characterize each wave. TBA experiments
provide plots of relative rigidity (1/P2, where P is the period of the oscillation) and
logarithmic decrement (A = In[0i/0i+d, where 0i is the amplitude of the ith
oscillation of the freely damped wave) vs. either time or temperature. The relative
rigidity is proportional to G', the elastic shear modulus; the logarithmic decrement is
proportional to G"/G', where G" is the out-of-phase shear modulus. Due to the
composite nature of the TBA specimen, quantitative values of G' and G" are not
obtained. However, by using homogeneous specimens (e.g., cured films) the TBA
instrument can be used as a conventional TP to obtain quantitative values of G'
and G". In the TBA mode the instrument is used principally to measure transition
times and transition temperatures.
The equation of motion for a torsion pendulum, in complex format, is m):
ItJ+ q0 + K G * 0 = 0 (1)
where co is the angular frequency = 2rc/P and 13is the decay constant. Differentiating
Eq. (2), substituting the results into Eq. (1), and writing G* = G' + iG" yields:
where tn is the time required to reach the 0,th oscillation. Dividing Eq. (4a) by
Eq. (4b) yields:
where 8 is the phase angle between the cyclic stress and the cyclic strain in the
dynamic mechanical experiment. Thus the TBA parameters 1/p2 and A can be related
to the material parameters G ' and G".
The derivation resulting in Eqs. (6a) and (6b) is based on the assumption that
G* is independent of frequency it} If the derivation had been performed with the
assumption that the dynamic viscosity is frequency-independent, then the minus sign
in Eq. (6a) would be replaced by a plus sign 11} Neither assumption is accurate over
a wide range of frequency for high polymers 11j. The plus sign analogue of Eq. (6a)
is generally used in this laboratory. The difference between the two approaches is
small except in the regions of melting, crystallization and the glass transition, where A
can be greater than unity.
In order to construct a TTT cure diagram, a series of isothermal cures are per-
formed at different temperatures, and the relative rigidity and logarithmic decrement
are recorded vs. time. From each damped wave I/P 2 and A are extracted; the
numerical methods used have been described 12~
In a typical TBA experiment, the glass braid is dipped' into a solution containing
the reactants, the impregnated braid is clamped to the pendulum, and the pendulum
assembly is lowered into the TBA chamber at the cure temperature. Tcure is
generally above the boiling point of the solvent. The computer program which controls
the alignment, intermittent oscillation of the pendulum, and data collection and
analysis, is started, and plots of 1/1)2 and A vs. time are produced 12)
The results of a typical isothermal cure are shown in Fig. 3. In general, three
events can be discerned in the A vs. time plot: a shoulder and two distinct peaks.
The first peak is associated with gelation and the second peak with vitrification 2~.
The time to gelation has been compared with gel fraction experiments 1-3); good
agreement is generally observed above 8e~Tg.The pre-gel shoulder has been attributed
92 M. T. Aronhime and J. K. Gillham
10~ 102
100 F 101
.~
"O
.~
Ot Gelation j " Vitrification E
~, 10-1 p r eg el ~,.,7.." - "/'.. 100
,m
'10
: :: "......../ '-... O)
O
\
r,,, i0.2 t0-1
10 -3 I = t 10-2
0 I 2 3 4
log t i m e (min}
Fig. 3. TBA spectrum during isothermal (125 °C, 75 hr) cure showing changes in the relative rigidity
and logarithmic decrement vs. time. Note the location of the pregel shoulder, gelation peak and
vitrification peak. The system studied was a difunctional epoxy resin, DER337 [a diglycidyl ether
of bisphenol A (DGEBA), Dow Chemical Co.], cured with a stoichiometric amount of a tetra-
functional aromatic amine, TMAB (trimethylene glycol di-p aminobenzoate, Polacure 740M,
Polaroid Corp.)
200
150
-- 100
-50 I I |
0 1 2 3
log time Imin)
Fig. 4. TTT cure diagram: temperature of cure vs. time to the isoviscous event, gelation and
vitrification, including TBA and gel fraction data: I-1, isoviscous (TBA); O, gelation (TBA);
A, vitrification ( T B A ) ; . , gelation (gel fraction). The system studied was a difunctional epoxy
resin, Epon 828 (DGEBA, Shell Chemical Co.), cured with a tetrafunctional aliphatic amine,
PACM-20. [bis(p-aminocyelohexyl)methane, DuPont]
Time-Temperature-Transformation(TIT) Cure Diagram of Thermosetting Polymeric Systems 93
to an isoviscous event that is due to the interaction of the liquid and the braid 2).
Note that the peaks in the A curves are located approximately midway through the
transitions in the relative rigidity plots. As the cure reaction proceeds, the relative
rigidity (or modulus) increases and eventually appears to level off in the glassy state
or in the elastomeric state; distinct loss peaks are associated with the transitions from
liquid to sol glass, liquid to sol/gel rubber, and sol/gel rubber to glass.
An example of a T I T cure diagram is shown in Fig. 4, where T¢,~¢ is plotted
vs. the times to gelation and vitrification, as determined by TBA. Note particularly
the good agreement, above genTs, between the gel fraction experiments and the
TBA results for gelation. (Below genTs, a "liquid-to-rubber" transition is observed in
the TBA experiments; the origin of this peak may be due to an interaction between
the braid and the polymer) 2) The vitrification curve is S-shaped, in agreement with the
schematic diagram shown in Fig• 1.
Fig. 4 is a T T T diagram for the reaction of a difunctional epoxy with a tetra-
functional aliphatic amine, which is the only system for which the complete TTT
diagram (from temperatures less than T~o to greater than T ~ ) is available. The
Tso of this system is --19 °C; the amorphous reactants are liquids above this
temperature and so the experiments can be performed in the absence of a solvent.
Most of the systems studied in this laboratory are very viscous liquids at room
temperature and for convenience solvents are used. Only the upper portion of the
TTT diagram, above g,lT~, is usually obtained.
After extended isothermal cure, the system is cooled from T¢,~, to --170 °C,
heated from --170 °C to a specified post-cure temperature, and then immediately
cooled to --170°C, all rates of change of temperature being 1.5 K/min. The
temperature cycling is continued as long as is desired. The initial scan from --170 °C
10 ~ , 10 2
io-2 :"
10-1
"%" s
10-3 I I I I I I I ,I , 10 -2
-200 -150 -100 -50 0 50 100 150 200 250
Temperature (*C)
Fig. 5. TBA spectrum after isothermal cure (90 °C, 10a min) showing changes in the relative
rigidity and logarithmic decrement vs. temperature. Temperature cycle: 90--*--170--*240--+- t70 °C,
1.5 °C/min. Note the increase in T~ and T~c with post-cure, but the decrease in the room temperature
rigidity (modulus) with post-cure. The system studied was DER337/TMAB (see Fig. 3 caption)
94 M . T . Aronhime and J. K. Giltham
From plots such as Fig. 5, Tg after cure is determined, and can be plotted vs.
T .... (Fig. 6). T, is observed to exceed T .... primarily because isothermal TBA cures
are carried out well beyond the time to the second loss peak in the logarithmic
decrement curve (see Fig. 3). If the cure were stopped at the second maximum, then
Tg would equal T .... 2~. However, the system would still be reactive because the
second loss peak occurs approximately midway through the increase in modulus
which accompanies the change from sol/gel rubber to the glass plateau. Similar
considerations apply to the conversion of liquid to ungelled glass below ge~Tg. It is
convenient in an operational sense to associate the second maximum with vitrifica-
tion. However, if vitrification is defined to be the point at which chemical reactions
are quenched, then a more appropriate determination of the time to vitrification
would be when the relative rigidity curve has leveled off with time. In Section 4,
200
o
150
"•T'o o
gO0
o o~i ~ 1 ' ~
•
too
50
50
~ . "
I
I00
~ T g = Teu,e
150
I I
200
Tcure ( * C )
Fig. 6. Tg after prolonged isothermal cure vs. T .... : Ira, Ts; O , T~o~. Tsoo and Tg = T .... lines are
also included. T h e system studied was D E R 3 3 7 / T M A B (see Fig. 3 caption)
Time-Temperature-Transformation (Y/T) Cure Diagram of Thermosetting Polymeric Systems 95
180
141 146
160
1/+0
120
100 I
80 100 120 140 160
Tc~,~ (*C)
Fig. 7. T B vs. temperature of cure for different times of cure. The solid lines are the best straight
line fits to experimental data (symbols); the dashed lines are the extrapolations to Ts0o to obtain
the temperature of cure to reach full cure for the given time. The Tgo~ line is also included. The
system studied was a difunctional epoxy resin, DER331 (DGEBA, Dow Chemical Co.), cured with
TMAB (see Fig. 3 caption). (Peng, X., Gillham, J. K., Ref. 4~)
200
...."~'"'T'g"'"~'"'"'-"""
"' "~" I ...................f .......[""'"'"'""'~ui["'c'ure
.~ 150
0
E
1oo
\\
50 t f ,, , i
0 1 2 3 4
log t i m e (rain)
Fig. 8. TTT cure diagram: temperature of cure vs. the times to gelation, vitrification, and full cure.
The full cure line is obtained from the extrapolated data of Fig. 7. The system studied was
DER331/TMAB (see Fig. 7 caption). (Peng, X., Gillham, J. K., Ref. 4))
96 M° T. Aronhime and J. K. Gillham
250 *C °
250"C
2 0 0 °C
°o •°•
n
200"C
150 *C
fv'"
.F
o_
"0
/ (J
"~""-'--- 150 *C
OJ
._> • °~o/"
,,,,,
"O
/:;
' 12S*C O
0
(z
~ ' ~ " 125 *C
80 °C
- 8o-c
J
"I ....... " - r " d i I 1 I
o I 2 3 4 0 1 2 3 4
a, log t i m e I m i n ) b log time (rain}
Fig. 9a and h. TBA spectra for a series of isothermal cures showing changes in (a) the relative
rigidity and (h) the logarithmic decrement vs. time. Gelation and vitrification are evident in the
80, 125 and 150 °C scans, but only vitrification is observed in the 200 and 250 °C scans. The
system studied was a trifunctionat epoxy resin, XD7342 [triglycidyl ether of tris(hydroxyphenyl)-
methane, Dow Chemical Co.], cured with a tetrafunctional aromatic amine, DDS (diaminodiphenyl
sulfone, Aldrich Chemical Co.)
Time-Temperature-Transformation (TIT) Cure Diagram of Thermosetting PolymericSystems 97
350 e C
• ° .°°. , . . , . . . ~
" ''"" .. ...... ~ 3 3 5 *C
¢.i
(3- 325"C
.... "" " , . ~ 3 2 5 " C
Ot 310"C 310*C
t- .... ,,
¢11
.>_
13
300 *C
n~
300 *C t
290 *C
/"'-x /'x
°
2 go *C
...,...o~j ~.,,s"
I I I
0 1 2 3 4 0 1 2 3
a, log time (rain) b log time (rain)
Fig. 10a and b. TBA spectra for a series of isothermal cures showing changes in (a) the relative
rigidity and (b) the logarithmic decrement vs. time. Gelation and vitrification are not evident in
any of the scans. In the 290-325 °C scans, devitrification and revitrification are observed. In the 335
and 350 °C scans, only the revitrification event (char formation) is observed. The system studied was
XD7342/DDS (see Fig. 9 caption)
98 M.T. Aronhime and J. K. Gillham
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
340
290
Devitrification
2/,0
~ 190 . Vitrification
E lt~0
t,--
9O Gelation
gel Tg "'~'~C'-
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
-112 I I I ,,,I
0 1 2 3 4
tog time (rain)
Fig, 11. TTT cure diagram: temperature of cure vs. the times to gelation, vitrification and
degradation, including TBA and gel fraction data: II, gelation (TBA); ©, vitrification; O, devitrifi-
cation; ~, char formation; A, gelation (gel fraction). Tso,an estimate of ,e~T~,and the hypothetical
value of T.~ are included. The system studied was XD7342/DDS (see Fig. 9 caption)
vitrification, devitrification, and revitrification are events that can occur during the
isothermal polymerization of high temperature epoxy systems. The T I T cure diagram
is a convenient means of summarizing the time-temperature paths of cure that can
lead to degradation.
After isothermal cure, temperature scans are conducted in order to measure the
T s after cure and Tg~. However, due to thermal degradation, postcures can lead to
lower glass transition temperatures than those obtained after cure. Thus, the
determination of Tg~o for high Tg systems is a difficult problem. One approach is to
establish a relationship between T ~ and theoretical crosslink density for systems of
lower Tso~ and similar chemical structure, and extrapolate to the system with higher
crosslink density, thereby obtaining an estimate of T ~ 3)
For epoxy systems with moderately high values of Tg~ (220 °C), thermal degrada-
tion is still a problem, but full cure in the absence of degradation can be achieved 3}
The T T T diagram for such a system is shown in Fig. 12, where the cure temperature
is plotted vs. the times to gelation, vitrification and thermal degradation. Devitrifica-
tion occurs below Tg®, as in Fig. 12, but other degradation events, including char
formation, occur well above Tg~ in the time scale of the experiments.
A comparison of Figs. 11 and 12 also serves to highlight the effect of functionality
on cure and properties. The system of Fig. 11 is a trifunctional epoxy cured with
a tetrafunctional aromatic amine, whereas the system of Fig. 12 is a difunctional
epoxy cured with the same amine. As expected, the more highly functional system
has the higher Tg~o and shorter times to gelation; the times to vitrification are also
shorter. The difference in these transformation times arises from two factors:
Time-Temperature-Transformation(TTT) Cure Diagram of Thermosetting Polymeric Systems 99
340
290
Tgoo
- - 24O
7
190 o.i,r.,oo.,o°
2
A ~.. ~,~ Vitrification
~. t4o
E geiTg Gelatio "':'~:.:'.,
90
4O
Fig. 12. TTT cure diagram: temperature of cure vs. the times to gelation, vitrification and degradation,
including TBA and gel fraction data: I , gelation (TBA); O, vitrification; Q, devitrification; Vl, &,
degradation events; ~, char formation; A, gelation (gel fraction). Tgo, an estimate of 8=~T~,and
the experimental value of Tg® are included. The system studied was a difunctional epoxy resin,
DER332 (DGEBA, Dow Chemical Co.), cured with DDS (see Fig. 9 caption)
1) the extent of conversion at gelation is less for the trifunctional system (Flory's
theory 7)), and the extent of conversion at vitrification is also lower 3); and 2) the
concentration of reactants in the trifunctional system is higher than for the difunctional
system.
High temperature epoxy resins are brittle materials, and one method of improving
their fracture properties is to incorporate reactive liquid rubbers in the formulations
8.15). In situ phase separation occurs during cure; the cured rubber-modified epoxy
resins consist of finely dispersed rubber-rich domains (~0.1-5 pm) bonded to the
epoxy matrix. T I T diagrams can be used to compare different rubber-modified
systems.
Fig. 13 is a T / T cure diagram of three systems: a neat epoxy resin and the same
epoxy modified with two reactive rubbers at the same concentration level. The times
to the cloud point, getation and vitrification are shown for each system. The cloud
point is the point of incipient phase separation, as detected by light transmission.
The modified system with the longer times to the cloud point and gelation, and the
greater depression of Tg=, contains the more compatible of the two rubbers. The
difference in compatibility could then be used to account for differences in the
volume fractions of the phase separated rubber-riCh domains and in the mechanical
properties of the neat and the two rubber-modified systems.
100 M.T. Aronhime and J, K. Giltham
200
/Neot s y s t e m "fg~
............... a ................... e o ........................ ./.,
............................................
Q 8
~150 A 00 ~a • 0
QO0 m • 0
,=
a
~DO ~Lll 0
.Q.
E ~' O 0 ~ 0
a
~" 100
O O
50 I .... J I
0 1 2 3
log t i m e (rain)
Fig. 13. TIT cure diagram: temperature of cure vs. the times to phase separation (cloud point),
gelation and vitrification for a neat and two rubber-modified systems. T~ooof the neat system is also
included. The systems studied were DER331/TMAB: Q, gelation; [], vitrification; modified with
15 parts rubber per hundred parts epoxy: 1) prereacted carboxyl-terminated butadiene-acrytonitrile
(CTBN) copolymercontaining 17~oacrylonitrile (K-293, Spencer Kellog Co.): A, phase separation;
O, gelation; i , vitrification, and 2) polytetramethylene oxide terminated with aromatic amine
(ODA2000, Polaroid Corp.): ~, phase separation; ©, gelation; ~, vitrification. (DER331/TMAB7
K-293 data from Ref. 8~)
Although different aspects of the isothermal TTT cure diagram have been presented
in this review from an experimental point of view, this section wilt present some
recent work that has attempted to model the cure process. Only the gelation and
vitrification processes are examined, and the complicating effects of thermal degrada-
Time-Temperature-Transformation (TTT) Cure Diagram of Thermosetting Polymeric Systems 101
~ 200"C
A
:t °", 200 eC
A
t;
' 170"C
3.
.'.c
;?:" 150"C
~:: 150 *C
"r-
o~
_o =
o
n~ ;~ 120 *C I~ :J" • 120"C
•~)~.........
..~ - - . -. ,.°-'°
: ~. IO0 *C
:" "./" 10o'c
~..':~........
• °
I I I 1 1 I I 1
-200 -100 0 100 200 300 -200 -100 0 I00 200 300
a Temperature (*C} b Temperature (*C }
Fig. 14a mad b. TBA spectra after isothermal cure showing changes in (a) the relative rigidity and (b)
the logarithmic decrement vs. temperature: T¢,,,--*-- 170 ~ 240--+-- 170 °C, 1.5 °C/min. Note presence
of rubber Tj peak at about --50 °C. The system studied was DER331/TMAB/K-293 (see Fig. 13
caption)
where ex/eM ---- ratio of lattice energies for crosslinked and uncrosslinked polymers,
Fx/FM = corresponding ratio of segmental mobilities.
In DiBenedetto's original equation, Tgo represented the glass transition temperature
of a polymer of the same chemical composition as the crosslinked polymer except
without the crosslinks, and Xc was used instead of Pvit, where X¢ is the mole
fraction of m o n o m e r units which are crosslinked in the polymer. Thus, the original
equation was applicable to the crosslinking of long linear polymers.
In order to use Eq. (12) values Ofex/eM and Fx/FM must be determined. Adabbo and
Williams 17~ assumed ex/~M = 1, and they found Fx/Fr~ = 0.733 was an acceptable
value for fitting Pvit vs. Tg data for several epoxy systems. Enns and Gillham 6) fitted
Time-Temperature-Transformation (TTT) Cure Diagram of Thermosetting Polymeric Systems 103
Eq. (I2) to experimental Pvlt vs. T s data, for one particular system, with a non-
linear least squares routine and found ex/EM = 0.34 and FJFM = 0.19.
With values o f ex/e~ and F J F u , it is a simple matter to calculate Pvit at any
value o f Tg ( = T .... ), and then determine the time to vitrification from an assumed
kinetic rate law. Using first order kinetics, which seemed to fit the extent of conversion
vs. time data, the temperature o f cure vs. the times to gelation and vitrification are
shown in Fig. 15. The model fits the data welt at low temperatures but appears to
2.0
1.8
0 0 ~Y"
1.5 ~
g
0 o
"" .
• II
1.2
1.o f t T
o 1 2 3 4
log time
Fig. 15. "ITI" cure diagram: T g ~ s o vs. times to gelation and vitrification. Theoretical (solid lines):
First-order kinetics using the following parameters: Tso = --19 °C; Ts® = 166 °C; ~/eu = 0.34;
Fz/F u = 0.19; E= = 12.6 kcal/mole; A = 4.5 x l06 rain -t ; Ps~t = 0.75; ==iTs = 49 °C. Experimental:
Q, pregel (TBA); II, gelation (TBA); C), vitrification (TBA); f-l, diffusion control (infrared spectro-
scopy);/x, gelation (gel fraction). The system studied was Epon 828/PACM-20 (see Fig. 4 caption)
i) T .... = T,
ii) T, vs. molecular weight or crosslink density
iii) molecular weight or crosslink density vs. extent of reaction
iv) extent of reaction vs. time.
Molecular weight is used for linear systems, and for thermosetting systems that
have not crosslinked (i.e., below ,~T,). There are four cases of importance-
linear systems for step growth and chain reaction mechanisms, and nonlinear systems
for step growth and chain reaction mechanisms -- but only examples of the first three
are discussed here.
For linear systems, an equation relating T, and the number average molecular weight
(Mn) is 21-23~:
where Tg~, Tgg = glass transition temperatures of sol and gel, respectively; ws,
wg = weight fractions of sol and gel, respectively; and I = an interaction parameter.
Tg~ is considered to be given by Eq. (13), where M. is now the molecular weight of the
sol fraction only. Tg, is given by t9):
per volume of polymer. A linear relationship between Tg~ and [X] is one of several
proposed s, 21,26)
In general, for uncrosslinked step growth systems, for the reaction of an f-functional
reactant Af with an h-functional reactant ~ , Mn is related to the extent of reaction
p by 27~:
where MA: M~r, arc mc~lecular weights of reactants Ar and t~,. respectively; Af, Bh
are moles of components At and Bh, respectively; and PA is the extent of reaction of
Af. For the linear step-growth polymerization of A2 with B2, in stoichiometric ratio,
Mn = (MA + MB)/[2(1 -- p)]. For the nonlinear step-growth polymerization of A4
with 2112, Mn = (MA + 2MB)/(3 -- 4p).
For the case of linear, chain growth polymerization, the experimental and computed
number average molecular weights are usually given for the polymeric portion of the
reacting mixture, rather than for the entire reactor contents. Since the Tg in bulk
polymerization is affected by residual monomer as well as polymer, a relationship
between total M~ (i.e., monomer with polymer) and p is needed. The contribution of
initiator is neglected, and the concentration of growing chains is negligible. Since Mn
is defined as the total weight of material divided by the total number of moles (mono-
mer plus polymer), and the number of moles of polymer is given by:
--dc/dt = kc n (19)
For the linear chain reaction case, a free radical kinetic mechanism (e.g., polymeriza-
106 M.T. Aronhime and J. K. Gillham
tion o f styrene) was used as an example. In this the rate of polymerization is given
by 31.32).
where ~]o is the initial initiator concentration and [M]o is the initial monomer
concentration.
For the linear free radical case, one additional equation is needed, relating <Xn>
to the other variables 32):
2°°f
150
P
g ~°°I
iI 5... Glass
- 50, 1 I, I I I
0 1 2 3 4 5
log time to vitrify (rain)
Fig. 16. Reaction temperature vs. time to vitrify for nonlinear step-growth polymerization
(A# + 2B2): nth-order kinetics for n = 1 to 3 in increments of 0.5 using the following parameters:
T=o = --19 °C; piT= = 50 °C; 1"=® = t66 °C; E, = 12.6 kcal/mote; A = 4.51 x l0 s m i n - l ; M^
= 210 gm/mole; MB = 382 gm/mole. Data (squares) are from the study of Epon 828/PACM-20 6~.
(See Fig. 4 caption for description of materials.) [Aronhime, M. T., Gillham, J. K.: J. Coat. Tech.
56 (718), 35 (1984)]
1.0
0.8
0.2,
- 5q
:
..!.
,./
0
.
50
.
Liquid
100 150
Temperature of cure ('C)
J
200
Fig. 17. Extent of reaction at vitrification
vs. reaction temperature for nonlinear
step-growth polymerization (A4 + 2B2).
All kinetic orders have the same p at
vitrification. For model parameters and
system, s¢¢ Fig. 16 caption. [Aronhime,
M. T., Gfllham, J. K.: J. Coat. Tech.
56 (718), 35 (1984)1
108 M. T. Aronhime and J. K. Gillham
400
350
300
i
25O
u n=l
200
.o
®
(3.
150
E
Liq Uid QSS
100
50
0 I I I 1 ,, I I
-1 0 1 2 3 t, 5
tog time to vitrify (min)
Fig. 18. Reaction temperature vs. time to vitrify for nonlinear step-growth polymerization (A4 + 4/3B3) :
nth-order kinetics for n = 1 to 3 in increments o f 0.5 using the following parameters: Tgo = 2 8 °C;
8,~Tg = 42 °C; Tg~o = 352 °C; E a = 13.3 kcal/mole; A = 2.49 × 105 rain -1 ; M^ = 448.4 gm/mole;
MB = 486 gin/mole. Data (squares) are from the study o f XD7342/DDS 3). (See Fig. 9 caption for
description o f materials.) [Aronhime, M. T., Gillham, J. K. : J. Coat. Tech. 56 (718), 35 (1984)]
100
80
r-
(2_ n = l ~ / ~ ~
Lqi u]30taIs.
L.
60
Tgo, Tg~o, and k, the temperature o f reaction vs. time to vitrification is S-shaped
(Fig. 19). However, a change in the activation energy o f the reaction, to a value less
than some critical value (E a < Ea, c,tt), where k = A exp ( - - E j R T ) , results in the
sigmoidally-shaped vitrification curves included in Fig. 20. E a was selected to be
less than E~, erit for n = 2; this value o f E, was used for all values o f n. In the
expression for k, A is the pre-exponential factor, R is the gas constant, and T is the
absolute temperature. Thus, in principle the time to vitrification curve need not be
S-shaped.
F o r the linear free radical case, the time to vitrification is affected by the initial
initiator concentration (Fig. 21). The vitrification curves are again S-shaped. For this
case, the same values o f kp, kt, and ka were used throughout the course o f the
reaction, although it is well known that the termination reaction becomes diffusion
controlled at fairly low degrees o f conversion 33~
Due to the nature o f free radical polymerization, i.e., the reacting system is
essentially a binary mixture composed o f m o n o m e r and high polymer, another appro-
ach was used to calculate P,it, and therefore t , , is, 19). This alternate method is based
on the free volume theory, which predicts a relationship between T s and the volume
fractions o f polymer and m o n o m e r in a binary system 3,, 35).
100 '
6o
4o
2o
0
-6 -5 -4 -3 -2 -1 0
log time to vitrify (min)
Fig. 20. Reaction temperature vs. time to vitrify for linear step-growth polymerization: nth-order
kinetics, for n = 1 to 3 in increments of 0.5 using the following parameters: Tso = 0 °C; T~0o= 100 °C;
Ea = 6 kcal/mole; A = 4.51 x 106 rain -1 ; MA = 200 gm/mole; Ms = 400 gm/mole. [Aronhime,
M. T, Gillham, J. K. : J. Coat. Tech. 56 (718), 35 (1984)]. [In this case, E~ is less than Ea. crit(see text)]
1 10 M.T. Aronhime and J. K. Gitlham
~
100'
50 I] 0 = 0.01
¢-
O Fig. 21. Reaction temperature vs. time
u to vitrify for linear free-radical polymer-
o 0 ization (styrene) for f = 0.5 and [1]o
[I]0=0.20 : = 0.01, 0.10 and 0.20mole/l using the
following parameters: Tgo = --100 °C;
-50 Liquid T,~ = 100 °C; M, (monomer)= 104gm/
0 mole; kp = (1.62× 10l° 1 mole -1 hr-*)
o. x exb (--6.21 kcal mole-l/RT); kt
E -100 = (2.088 x 1011 lmole -1 hr -1)
x exp (--1.91 kcal mole-l/RT);
ka = (2.725 × 1017 h r - l ) e x p (--29.71 kcal
-150 -' ) : ) : ' : ; : ' : ' : -
x mole-1/RT). [Aronhime, M. T., Gill-
-2 0 2 4 6 8 10 12 ham, J. K. : J. Coat. Tech. 56 (718), 35
log time to vitrify (h) (1984)]
w h e r e (~ = density. T h u s :
1.0
J
.,,/:-//
Tg =f (Free volume) . ~ 1 "
O.8
¢-
,0 . . S ( Mn }
0.6
The values of Pvi, from Eq. (26) are compared with the values calculated from
Eq. (21) (Fig. 22), and show good agreement over the entire temperature range.
In this section, two different approaches to calculating the time to vitrification
on isothermal polymerization have been examined. The first approach used an
existing relationship between Tg and Pvit, and the time was calculated from an
assumed rate law. The second method derived the values of Pvit from basic
equations in polymer science and then used an assumed rate law to calculate the
time.
Neither model is entirely satisfactory in fitting the experimental data. The
complexity of epoxy curing reactions contributes to the discrepancies. Many different
mechanisms have been proposed 6,36,37~ The diffusion controlled nature of the
reactions as vitrification is approached is another complicating factor. Both models
do predict the S-shaped vitrification curve, and the second model extends the concept
of the TTT diagram to linear systems.
5 Conclusions
A time-temperature-transformation (TTT) isothermal cure diagram has been de-
veloped to provide an intellectual framework for understanding and comparing the
cure of thermosetting systems. The times to gelation, vitrification, thermal de-
gradation, and phase separation can be conveniently summarized on the TTT diagram.
The TTT diagram can also be extended to linear systems, except that these systems do
not undergo gelation.
In order to obtain a T I T cure diagram, the cure process must be monitored
from the liquid region, through the sol/gel rubber region, and into the glass region.
The torsional braid analyzer (TBA) is an instrument capable of following the entire
cure process. The TBA, unlike the conventional torsion pendulum from which it
was derived, uses supported specimens, and thus can monitor properties above load
limiting transition temperatures.
During isothermal cure, maxima in the logarithmic decrement are associated with
gelation and vitrification. The times to gelation as measured by TBA correlate for
the most part with times as measured in gel fraction experiments. The loss peak
associated with vitrification occurs when T s = Teure and the modulus is midway
between the liquid, or sol/gel rubber, and glass plateaus. The chemical reactions are
quenched not when T 8 = Teure, but when the modulus levels off. The T I T diagram is
constructed by plotting the cure temperature vs. the times to gelation and vitrifica-
tion.
For high temperature and rubber-modified epoxy resins, thermal degradation events
and the cloud point curve are included on the diagrams, respectively. Two
degradation events have been assigned: devitrification, or a glass-to-rubber event;
and revitrification, which is associated with char formation. The cloud points and
depressions of Ts~ for different rubber-modified epoxies can be compared and related
to volume fractions of the second phase and to the mechanical properties of the cured
materials.
Two models for calculating the time to vitrification on isothermal polymerization
have been discussed. One model is based on an existing relationship between T~
112 M.T. Aronhime and J. K. Gillham
and the extent of conversion at T s. Data for the extent of conversion at vitrification
must be available to use this approach. The second model calculates Pvlt from several
relationships between Tg and molecular weight, and molecular weight and extent of
reaction. The second approach can be extended naturally to linear systems. Both
models use an assumed kinetic mechanism to calculate the time to vitrification, and
an S-shaped time to vitrification curve is predicted.
Acknowledgements: The research covered by this review has been supported by the
Army Research Office and the Office of Naval Research.
6 References
34. Horie, K., Mita, I., Kambe, H.: J. Polym. Sci. A-l, 6, 2663 (1968)
35. Sundberg, D. C., James, D. R.: J. Polym. Sci.: Polym. Chem. Ed. 16, 523 (1978)
36. Sourour, S., Kamal, M. R. : Thermoehirnica Acta 14, 41 (1976)
37. Barton, J. M.: Polymer 21, 603 (1980)
Editor: K. Du~ek
Received August 19, 1985
Structure and Ultimate Properties of Epoxy Resins
J. D. LeMay?
Lawrence Livermore National Laboratory, University of California,
Livermore, CA 94550, U S A
F. N. Kelley
Institute of Polymer Science, The University o f Akron, Akron, Ohio 44325, U S A
Common epoxy thermosets are glassy at ambient temperatures and are characterized by a densely
crosslinked microstructure. Under normal use conditions they generally fail by brittle fracture
mechanisms. The influence of network microstructure on glassy fracture is largely undetermined in
spite of a sizeable literature. This can be attributed to a lack of studies on structurally characterized
networks and the often complicated rnicrostructure of typical epoxy systems. To address these
problems we examine structure-fracture relationships in simple epoxy systems whose structural
variables are systematically controlled. Densely crasslinked networks may be characterized by
equilibrium modulus measurements above Ts. Application of rubber elasticity theory yields very
reasonable average network chain molecular weights (Me); surprising in view of the expected
non-Gaussian character of short epoxy network chains. Rubbery fracture energy increases with M¢ when
compared at equivalent temperatures above Tg. In fact, the dependence approximates a M~/2 ordering,
suggesting that the influence of a threshold fracture energy persists well into nonthreshoM testing
conditions. Often, glassy fracture is characterized by brittle, unstable crack propagation leading
to initiation and arrest fracture energies. The initiation values increase with temperature and
generally increase with Mc. In comparison, the arrest values are independent of temperature and display
a strong proportionality to M~/2. A theory presuming material devitrification at a sharp crack tip is
consistent with this observation.
I Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Advancesin PolymerScience78
© Springer-VerlagBerlinHeidelberg1986
116 J.D. LeMay and F. N. Kelley
I Introduction
Thermosets are polymeric materials which when heated form permanent network
structures via the formation of intermolecular crosslinks. Whether the final product
has a glass transition temperature, Tg, above or below room temperature, and
therefore normally exists as an elastomer or a glass, it is, strictly speaking, a thermo-
set. In practice, however, thermosets are identified as highly crosslinked polymers
that are glassy and brittle at room temperature. These materials typically ex-
hibit high moduti, near linear elastic stress-strain behavior, and poor resistance to
fracture.
Among the variety of polymer thermosets, epoxies enjoy the most widespread
use and are certainly the most studied. They are usually synthesized from oligomeric,
end-functional epoxy resins and multifunctional curing agents or "hardeners".
In addition, catalysts sometimes are employed to speed the crosslinking reactions or
allow them to take place at lower temperatures.
Practical uses of epoxies include load bearing applications such as structural ad-
hesives and composite matrices. In these applications, their most detrimental feature
is a characteristic low resistance to brittle fracture. The desire to improve this
property has motivated studies on thermoset fracture behavior for the last two
decades. Of particular interest is the relationship between the molecular structure
and the failure properties of thermosetting epoxies, the subject of this chapter.
The microstructure of epoxy thermosets can be complex, and both molecular
and physical microstructures are presumed. Unfortunately, the intractable nature
of these materials makes direct structural characterization extremely difficult.
The most accessible technique for direct structural characterization is evaluation of
epoxy rubber-like properties above Tg. Sometimes, indirect characterization of epoxy
structure is possible due to the fact that the chemistry of several epoxy systems is
well behaved (e.g., epoxy-amine chemistry). This permits epoxy network structure
to be modeled accurately as a function of the extent of the crosstinking reaction(s).
This approach has been developed extensively by Du~ek and coworkers for amine-
linked epoxies 1-4)
Epoxy networks may be expected to differ from typical elastomer networks
as a consequence of their much higher crosslink density. However, the same
microstructural features which influence the properties of elastomers also exist
in epoxy networks. These include the number average molecular weight and
distribution of network chains, the extent of chain branching, the concentration
of trapped entanglements, and the soluble fraction (i.e., molecular species not
attached to the network). These parameters are typically difficult to isolate and con-
trol in epoxy systems. Recently, however, the development of accurate network for-
mation theories, and the use of unique systems, have resulted in the synthesis of epo-
xies with specifically controlled microstructures 4-s~. Structure-property studies
on these materials are just starting to provide meaningful quantitative information,
and some of these will be discussed in this chapter.
The most popular method by which epoxy thermoset structure is altered for
structure-property investigations is the intentional variation of the curative/epoxy
resin functional group ratio (A/E). Unfortunately, it is impossible to alter indepen-
dently only one structural feature at a time using this technique. For example, the
Structure and Ultimate Properties of Epoxy Resins 117
4000
1001/DDS J
/
• IO02/DDS
3500 o 1004/DDS
3O0O
O
E
2500
2000
Fig. 1. Effectsof reactant ratio on DGEBA/DDS
1500
>.-.-C/ epoxy network M~. (Me calculated from equili-
brium rubbery moduli at T = Ts + 45 K).
/x Network made from Epon 1001 oligomer;
1000 I • Epon 1002; O Epon 1004. (After LeMay s))
0.6 0.8 1,0 1.2 1.4 1.6
A/E
network chain length can be increased only by increasing also the amount of
chain branching and the soluble fraction. Figure 1 shows how the average molecular
weight between crosslinks, Me, is affected by changes in A/E for three different mole-
cular weight end-functional epoxy resins cured with a diamine. At stoichiometry
(A/E = 1), all resin chain ends are ideally joined, by amine crosslinks, and the
resultant network chain length is short. In off-stoichiometric compositions, however,
chain branches and fully detached chains are unavoidable, and these effectivelyincrease
the apparent average network chain length. Thus, epoxy network M c increases as
compositions shift from stoichiometry. The observation that the increase is usually
more pronounced for epoxy excess formulations (A/E < 1) (see Fig. 1) suggests
that these networks" defects have a more profound impact on network structure.
In addition to the molecular microstructural parameters cited above (Me,
soluble fraction, etc.), supermolecular structures commonly called "nodules" are
purported to exist in some epoxies. It has been proposed that nodules result
from inhomogeneous crosslinking, and are sites of higher than average crosslink
density. How and why such inhomogenieties may occur is subject to speculation,
and sometimes is attributed to the following 9-1 ~):
(1) crosslinking reactions initiating in regions "rich" in hardener, perhaps resulting
from incomplete mixing;
(2) localized intramolecular linking and cyclization reactions; and
(3) thermodynamically driven phase separation.
A nodular epoxy network is thought to be a two-phase system in which regions of
relatively high crosslink density are dispersed in a less crosslinked interconnecting
matrix. If this is true, then the nodules should exhibit properties different from
those of the matrix, e.g., a higher T~, and a different specific volume. It is also
reasonable to expect that the size and concentration of nodules should be sensitive
to variations in the reactant ratio and cure conditions.
Whether nodules actually exist, however, is subject to considerable controversy.
The reason for this is that the physical evidence supporting their existence is
118 J.D. LeMay and F. N. Kelley
not yet convincing. For example, results from small-angle X-ray scattering, light
scattering, neutron scattering, birefringence and nuclear magnetic resonance meth-
ods 11-14~ are inconclusive. Differential scanning calorimetry (DSC) experiments
attempting to observe Tg "splitting" or nodule/matrix segregation on aging have met
with limited success 15); however, the results are partly explainable in terms of the
non-equilibrium nature of the glassy state.
Nodules in the 10-100 nm diameter range were first inferred from the inspection
of transmission electron micrographs of fracture surface replicas which had a granular
appearance 16,17). In some cases, it was found that the nodular appearance could
be enhanced by ~eplica staining or surface etching 16~. However, it has been noted
that replicas of thermoplastic fracture surfaces also display a granular appearance. H~,
suggesting that nodules may simply be an artifact of replica preparation techniques.
Some workers report a clear relationship between composition and cure schedule
to nodule size and concentration 10~, while others find no such dependence ~7~
The lack of strong experimental evidence for nodules has not prevented attempts
at correlating mechanical properties to nodule size and concentration. This is
particularly true for fracture measurements where it is proposed that cracks pre-
ferentially propagate through the less crosslinked matrix between nodules 10,1s~
If this is true, then the size, concentration, and distribution of nodules may be
expected to influence fracture. For example, Gledhill and Kinloch 18) have suggested
that epoxies may fracture unstably in the presence of large nodules, but stably in
the presence of smaller nodules. Experimental results obtained by Mijovic and
Koutsky 10) on a series of epoxy resins exhibiting a nodular morphology suggest
that fracture energies pass through a maximum at a particular nodule diameter.
In a study attempting to address the influence of nodules on fracture, Kelley and
Trainor 19) created a nodular network consisting of a composite of highly crosslinked
polystyrene beads dispersed in a mixture of linear and crosslinked polystyrene. The
beads were prepared via emulsion techniques and were synthesized with near
monodisperse sizes in the 100 nm range, similar to the nodule sizes observed in
epoxies. The bead and matrix crosslink densities were capable of being varied at will,
as was the degree of interconnectivity between the beads and the matrix. In
contrast to reports on purported nodular epoxies, Kelley and Trainor found that
the fracture energies of the polystyrene composites were independent of the nodule
size. Their work did suggest, however, that fracture energies may be dependent on
the volume fraction of nodules and the degree of interconnectivitly between the
nodules and the lesser crosslinked phase.
where Q is the density, R the gas constant, T the absolute temperature, A the "front
factor" and (at2) the dilation factor (the ratio of the average chain mean-square
end-to-end distance in the normal unstrained, isotropic state and the reference state).
In simple statistical theory, both A and (~2) are unity. (It should be noted that in
the past the collective term A (at2) has also been called the front factor.) The
front factor in Eq. (2) is needed to describe so-called phanom networks in
120 J.D. LeMay and F. N. Kelley
which the deformation of network junctions is not affine 21, 23, 24). For such networks,
the front factor is related to the junction functionality, f, by 23)
A = ( f - 2)/f (3)
Thus, it is predicted that the modulus of a phantom network is smaller than that
otherwise equivalent affine network.
The contribution of chain entanglements is given by z6,17)
M c < 1,000 g/mole, data recently reported by the authors 6-s~ reveal that Eq. (1)
accurately predicts the rubbery moduli of such networks when A and (0t2)
are taken as unity, and Goe is assumed negligible. (This work is discussed in more
detail below.) For less densely crosslinked epoxies (Mo ranging from 1,000-25,000),
work done by Dusek and coworkers 2's'2s~ leads to two possible conclusions:
(1) the elasticity can be described by Eq. (2) with A and (0d) equal to unity
(i.e., an affine network), and assuming no entanglements, or (2) by assuming the
phantom network model with contributions from trapped entanglements.
Because subsequent discussions detail the fracture behavior of a series of
highly crosslinked epoxies studied by the authors 6-s~, the experimental elasticities
of these networks are now presented. The networks belong to an homologous
series of amine-linked epoxy networks whose chain lengths were varied by the
stoichiometric end-linking of five different molecular weight resins. These resins
are members of Shell's Epon family of DGEBA resins, and range in molecular
weight from 380 to 2600 g/mole. The amine curing agent was 4,4'-diaminodiphenyl
sulfone (DDS) and no catalysts or crosslinking aids were employed. Processing
and curing were performed at 200 °C because of the high viscosity of the higher
molecular weight resins, and the poor solubility of DDS in the resins at much lower
temperatures. Complete cures were achieved after about 15 hours at 200 °C, as
evidenced by the attainment of stable glass transition temperatures and rubbery
Resin M, a EEW b ff
g/mole g/mole
moduli. A list of the properties of the epoxy resins and cured networks is given
in Table 1 and 2.
The approach of using stoichiometric end-linking of oligomers to synthesize
epoxy networks of different M c is not novel, and has been used in several
33,
published investigations 38). tt does, however, represent a departure from the more
common technique of using off-stoichiometric mixtures of liquid, low molecular
weight epoxy resins and amines to create networks with a range of Me. In theory,
the homologous oligomer approach should generate networks which do not contain
the important network defects existing in non-stoichiometric networks of correspond-
ing average network chain lengths (e.g., chain branching and soluble fractions).
In reality, however, commercial oligomeric epoxy resins are not ideally d/functional
(see Table 1). The DGEBA/DDS networks were, therefore, not defect free and this
is demonstrated by the non-zero network soluble fractions listed in Table 2.
The DGEBA/DDS network M¢ (Table 3) were calculated using Eq. (2) (with
Q, A and (~2) equal to unity) and near equilibrium rubbery moduli determined
from both static tensile tests and low frequency (0.16 hz) dynamic mechanical
E Mc E' M¢
GPa g/mole GPa g/mole
12
/ 828 / 1001F
O/( #/e ,1002F
%8 I/Mc /
x
/~ I/~/'t/ o-"IOOAF
lEA
/ js f, .1oo Fig. 2. Rubbery tensile stress versus strain
curves for DGEBA/DDS epoxy networks at
T = T,. + 45 K. Ordinate (true stress) nor-
malized by 3 QRT. O Extension data; • Re-
0 0.05 0.10
covery data. After LeMay8))
Strain
Structure and Ultimate Properties of Epoxy Resins 123
o
o
o
=Y2
possible to estimate theoretical M r from the amine and epoxy resin concentrations.
The Equation for this is
M r = QN/zc (5)
The threshold fracture energy is defined as the amount of energy necessary to break
the molecules crossing a unit area of the crack plane in the absence of chemical
degradation and irreversible energy dissipations. Lake and Lindley first reported
threshold fracture energies for several hydrocarbon elastomers 40). Using cut-growth
fatigue experiments they observed a rate and temperature independent, minimum
fracture energy below which crack propagation could not be observed. These
threshold energies were characteristic of the polymer, the level of crosslinking, and
the environment. Since the pioneering work of Lake and Lindley, threshold fracture
energies have been reported for a variety of elastomers, under different testing
modes and conditions 41-44). For conventional elastomers, 2Jo has been found
to be on the order of 30--100 J/m 2. While these values are considerably lower than
the fracture energies obtained under normal use conditions, they are still about
two orders of magnitude larger than typical covalent bond strengths (around
0.5 j/m2). Lake and Thomas .5) attributed this apparent discrepancy to the structure
of network chains, and subsequently developed a theory which successfully accounted
for the magnitude of 2j0. They pointed out that applied stresses must be
transmitted to the network chains through the crosslinks, and therefore that the
backbone bonds in each chain crossing the fracture plane must be stressed
to near rupture before the chains are severed. Therefore, 2Jo reflects not
only the number of chains crossing the fracture plane, but also the average
length of those chains. Considering a network of chains containing an average
126 J.D. LeMay and F. N. Kelley
where LN/2 is the number of chains crossing a unit area of the fracture plane
and nE is the energy necessary to rupture the chain. N is the number of chains
per unit volume and L is the average, unstrained chain displacement length. In
the case of heteroatomic backbones, E is taken to be the value of the weakest
bond. According to Lake and Thomas, the factor of 1/2 arises because, of the
total number of chains located in the volume defined by the distance L above
and below the crack plane, only 1/2 of them, on the average, will actually
cross the plane.
For a network of uniform length chains, Lake and Thomas substituted for L
with an Equation predicted from rubber elasticity theory. They also derived an
alternate expression for L for a network of random Gaussian chains. The two
expressions differ only by a small numerical constant. Making either substitution,
and rearranging terms, it can be shown that
2 J o = k2M~/2 (8)
where k2 is a collection of constants which reflect the size, mass, flexibility and
strength of an average main-chain bond. k2 is given by
k 2 = k3Moa/2ql/21EQN (9)
function is not strongly affected by the level of crosslinking. If this is the case,
then 2 J can be normalized with respect to 2~o and qb can be studied for
network structure and test condition dependencies. Since 2 J 0 is dependent of
rate and temperature, d~ can be determined as functions of these variables by
plotting 2 J versus temperature or rate.
where log (aT) is the shift factor, To the reference temperature, T the test
temperature, and C1 and C2 empirical constants. When To = Tg, C1 and C2 are
practically "universal" for most polymers, taking on typical values of 17.4 and
51.6K, respectively. In practice, fracture energies are collected as a function
of some testing variable (e.g., extension rate r, or crack velocity, a) at a number
of different temperatures. These data are shifted by Eq. (12) or numerical curve
fitting techniques, and plotted (log--log) against the reduced testing variable, e.g.,
fiaT) or ~i(ax). The shape and magnitude of the resulting "master curve" have
proved to be sensitive to molecular structure features such as crosslink density 53),
entanglements 43) and dangling chains 32)
systems, and both used the amine/epoxy reactant ratio, A/E, to effect changes in M r,
which was determined via equilibrium modulus measurements. The A/E and corre-
sponding M~ for the networks in both studies are summarized in Table 5.
Fracture energy master curves were determind as a function o f nearly equivalent
ranges o f reduced crack velocity (King and Andrews), and extension rate (Swetlin).
In both cases, Tg was used as the reference temperature. King and Andrews' master
curves were obtained using the W L F Equation and the universal constants, while
Swetlin's master curves were determined via numerical "best-fit" shifting.
Swetlin found that all three o f his networks exhibited master curves which
shared the same shape, and which shifted vertically with increasing M c (Fig. 4).
In fact, when normalized by Mlc/2, the three master curves apparently collapse
onto a single curve, as shown in Fig. 5. This strongly suggests that the vertical
shift of the master curves is due to the predicted M¢ dependence of the threshold
3 7s0
,o0 -
.yF.&'Y-
I
TO = "I"6
0 ........ I I i I I I I
-20 -16 -12 -8 -4
log R,,a'r( m / s )
Fig. 4. Rubbery tear energy master curves for Epon 878/diaminodiphenylmethane networks of
different reactant ratios: O A/E = 0.65; • A/E = 1.00; A A/E = 1.60. Reference temperature
is Tg. Curves constructed using a best fit algorithm. (After Swetlin an)
A/E= Mc=
ff- o 0.65 1500 z,
1.60 750
"-3
v • 1.00 300 . , , . ~ '~
i -
+2
0 t I I I .......... I i
Fig. 5. Same as Fig. 4 except tear energies normalized by M~/2. (After Swetlin 31~)
130 J.D. LeMay and F. N. Kelley
fracture, and that the loss functions of the networks are independent of M c. The fact
that the curves have the same shape also suggests that ~b is relatively insensitive
to other structure differences. This is significant in that structural variations, such
as soluble fractions, were found in the networks.
In contrast, King and Andrews observed different master curve shapes among
their networks and attributed this to structural variations other than M r. Specifically,
stoichiometric and amine excess networks behaved differently than epoxy excess
networks. The former exhibited a shape and Mc ordering similar to Swetlin's
networks, while the latter had a different shape and did not show a strong
M c dependence (with the exception of data at low reduced rates in the threshold
region). King and Andrews attributed this to long-chain branching in the epoxy
excess networks, and suggested that the branches internally plasticized the net-
works.
The authors ~,s) measured the rubbery fracture energies of the homologous series
of DGEBA/DDS networks described in Table 2 as a function of temperature
in the range of Tg + 20 to Tg + 100 K, at a single slow rate. The results are
plotted in Fig. 6 as log (2J) versus reduced test temperature, T - Tg. This choice
of the abscissa permits the network response to be compared under equivalent
temperature states. At the higher test temperatures, the tear energies level
off to near constant, apparent threshold values. This is reasonable since the
attainment of near threshold conditions is facilitated at high temperatures and low
rates. These apparent 2o¢o range from 5-25 J/m 2 and are comparable to values
obtained for epoxy thermosets by King and Andrews 32) and Swetlin 31). The curve
shapes are identical and are shifted along the ordinate axis according to increasing
M r. The similarity of the curve shapes suggests that the loss functions of the
DGEBA/DDS networks are insensitive to the level of crosslinking. If this is
E
~2
CN
C~
0
•,•!,= \~
• 828
o 1001F
• t002F
,~ 100L, F
v 1007 F
t I ~ ...... 1
0 40 60 80 100
T = T~ (°C)
Fig. 6. Stoichiometric DGEBA/DDS network rubbery tear energies versus reduced test temperature,
T-TS. Tear energies determined using single edge notch specimens of crosshead rate of 0.05 cm/min.
• Epon 828/DDS; O Epon 1001F/DDS; A Epon 1002F/DDS; A Epon 1004/DDS; x7 Epon
1007F/DDS. (After LeMay 8))
Structure and Ultimate Properties of Epoxy Resins 131
indeed the case, then the tear energies may be expected to display the predicted M c
dependence of the threshold tear energy. To investigate this possibility, the curves
in Fig. 6 were normalized by M 1/2 and M 1/3 per the Lake and Thomas 4s) and
King and Andrews 32) theories, respectively, and the results are illustrated in
Figs. 7 and 8. (The curves are shifted vertically one decade to keep the ordinate
values positive).
It was found that both normalizations yielded tear energy master curves over
all the test temperatures investigated for all but the most highly crosslinked
828/DDS network. The fact that master curves can be generated over the entire
range o f test temperatures shows the important role that M c plays in the rubbery
fracture of these highly crosslinked epoxies.
¢I ~ • 828
+ 2 . ~,\ o 1001F
A 1002F
\+, lOO F
\ ~A v 1007 F
0 " ' i ,
0 20 40 60 80 100
T - - T B (°C)
Fig. 7. Same as Fig. 6, except tear energies normalized by M 1/2. Curves shifted up one decade on ordi-
nate axis to keep values positive, (After LeMaya~)
~+ 3 • 828
, ~ ~ o 1001F
,L 1 0 0 2 F
'~ I 0 0 4 F
v 1007 F
~2
o
I I I I
0 20 40 60 80 100
T--T s (°C)
Considering the data scatter at the threshold end of the master curves, it is
not possible to distinguish one method of normalization over the other. In fact,
for low M r networks it can be argued that the small differences between the two
theories will not be detectable for characteristically scattered measurements such
as tearing. Over the entire Tg shifted temperature range, however, it is obvious that
the M~/2 normalization yields less scattered data and a better defined master
curve.
An explanation for the anomalous behavior of the most highly crosslinked
DGEBA/DDS network may be that it was chemically unstable at the high test
temperatures (250-260 °C) required to reach the rubbery range for this network.
Also, the threshold fracture theories may simply fail to describe the structure-
fracture relationship of this very highly crosslinked network.
In summary, investigations of amine-linked epoxy networks reveal that the
cohesive rubbery fracture energy is dependent on the network chain length. This
relationship, however, may be complicated by network defects, like branches, as
suggested by the data of King and Andrews 3z) It is interesting that the MI¢/z
dependence of the tear energy persists even at temperatures far removed from
those of that special condition called "threshold". Evidently, the threshold
tear energy theory, while developed under highly restricting assumptions, has
applicability for some expoxies even at conditions well removed from those
required by the assumptions.
specimens, including: single edge notch (SEN), three point bend (TPB), double
cantilever beam (DCB), tapered double cantilever beam (TDCB), width tapered
double cantilever beam (WTDCB), and double torsion (DT) specimens.
Crack growth is unstable and catastrophic in the single edge notch and three
point bend specimens, but stable in the cantilever beam and double torsion
specimens, (In this context, crack growth is stable when its propagation can be
halted by simply removing the load on the specimen). Probably the most popular
specimens used to evaluate the fracture behavior of epoxies are the TDCB and DT
specimens, because they not only fracture stably, but yield fracture energies that are
independent of the crack length (so-called linear compliance specimens). Of the two,
the DT specimen is the easiest and least expensive to prepare, and it is therefore
not surprising that the bulk of reported thermoset fracture studies involve this test
specimen.
LEFM specimens yield a stress intensity factor, K, which is a continuous function
of the applied stress field and the crark length. At fracture, the stress intensity factor
takes on a critical value, symbolized by K1~,called the fracture toughness. The Roman
numeral subscript identifies the fracture mode 64), e.g., " I " is tensile opening mode.
In this chapter, mode I is assumed, as is the critical condition; therefore, the unscrip-
ted symbol K will subsequently represent the fracture toughness.
If a material exhibits linear-elastic stress-strain behavior prior to rupture (an
ideal behavior approximated by many therrqosets), then a simple relationship exists
between the material's fracture toughness and its fracture surface energy, J (or G),
i.e.,
a Displacement b Displacement
Fig. 9a and b. Stable and unstable fracture behavior as observedon Instron recorder traces when testing
constant compliance fracture specimens, a) trace, stable crack growth; b) trace, unstable, "stick-
slip" crack growth
crosshead rate. These traces are representative of those obtained for either the DT
or TDCB specimen. In the case of continuous and stable crack growth (Fig. 9(a)),
the load reaches some constant value which is sustained until the test specimen
totally fails. The unstable fracture trace (Fig. 9(b)) has a sawtooth appearance, ex-
hibiting characteristic maximum and minimum loads which alternate until the speci-
men is completely ruptured. It has been observed that the maximum loads are
associated with crack initiation, and the minimum loads with crack arrest 65~. There-
fore, it is common to associate initiation and arrest fracture toughness values,
K i and K~, and fracture energies 2 J i and 2 J , (or G i and G,), with the maximum
and minimum loads, respectively. To differentiate stable fracture from unstable frac-
ture in subsequent discussions, the symbols Ks and 2 i s are introduced.
Klc Klc
a R.~e b TernperoturSa
Fig. 10a and b. Fracture toughness versus rate (a) and temperature (b) showing typical thermoset
fracture behavior. I = initiation; A = arrest; E = stable crack growth
of rates. Comparison between these two studies is complicated by the fact that
two different epoxy resins and amines were used, and that the former work
utilized a stoichiometric network while the latter employed an epoxy excess one.
Like the testing variables just described, material variables can influence the
fracture behavior of epoxy thermosets. Material variables discussed herein include
the types of epoxy resins and amine curatives.
Low molecular weight, end-functional, epoxy resins based on the condensation
product of epichlorohydrin and Bisphenol A are represented most extensively in
the literature. As a class of materials, these resins are often referred to as
DGEBA resins (diglycidyl ether of bisphenol A) and they have the structure
illustrated in Fig. l l(a). Commercial DGEBA resins include Shell's Epon 828
and Dow Chemical's DER 332. These oligomers have number average molecular
weights of about 400 g/mole, are liquids at room temperature, and are popular be-
cause of their ease of handling, availability, and consistent epoxide content.
Higher molecular weight homologs of these resins also are available; however,
they are room temperature solids, difficult to process, and not as chemically
consistent as the liquid resins. While DGEBA resins are multifunctional (containing
epoxide and hydroxyl functional groups), it is generally the epoxide chemistry that
is used to synthesize networks.
Several other epoxy resins have been used in literature fracture studies. Chang
et al. 70-73) used a diglycidyl ether of butane diol resin (DGEB) to prepare
relatively low Tg networks (Fig. t 1(b)). A tetrafunctional, room temperature liquid
DGEBA
0 CH3 OH CH3 0
c.2-c.-c.,t0 I-LW_2-°-c"2-c"- -LW2-I 0-c.,-c.-cH2
CH3 CH3
&
DGEB
/ON /0 X
CH2-CH-CH2-O- (CH2)4-0- CH2--CH--CH2
b
TGDDM
/o. /o\
CH2--CH-CH2~ ~ ~ /CH2-CH-CH2
CH2--CH--Ell2/N ~ CH2~ N,...CH2-CH--CH2
%/ %/
(3
Fig. 11a c. Epoxy resins used to study epoxy network fracture. (a) DGEBA (diglycidylether of
Bisphenol A); (b) DGEB (diglycidylether of butane diol; (e) TGDDM (teragtycidylof diaminodi-
phenyt methane)
Structure and Ultimate Properties of Epoxy Resins 137
The unstable fracture of epoxies has been shown by Mai and Atkins 8o~to be accom-
panied by a negative change of the strain energy release rate, G, with crack
velocity, ~i. This is in contrast to the positive dG/d~ which they find characterizes
stable fracture. Whether a negative dG/d~i is the cause or the consequence of
unstable fracture is, however, subject to debate 80, 81)
The isothermal/adiabatic transition argument was originally applied to poly(me-
thyl methacrylate) in which crack propagation was observed to become unstable
above a certain fast crack velocity s2~. This instability was attributed to the
transition of the crack tip deformation from an isothermal process to an adia-
batic one, resulting in substantial softening at the crack tip. The applicability of this
process to epoxies 83~ is questionable based on the observation that it is slow, not
fast, rates at which the epoxy stable to unstable transition occurs.
An influence of molecular relaxations on observed fracture behavior has
been suggested in several studies on thermoplastics such as poly(methyl methacry-
late) 84), polycarbonate ss), and polystyrene 86). For example, maxima in fracture
energy versus crack velocity plots have been associated with changes in the loss
tangent with rate. Also, activation energies obtained from fracture data have been
correlated with the activation energy of the 13-relaxation. An argument may be
presented that such relationships are fortuitous since sub-Tg relaxations involve
sub-molecular (small strain) deformations while molecules at the crack tip undergo
large strain deformations (e.g., crazes). To the authors' knowledge, such cor-
relations have not been observed with epoxies. However, a molecular relaxation
argument might be given for the stable to unstable transition observed in the
low temperature fracture of the amine-linked epoxies of Scott et al. 69) It is
well known that DGEBA epoxies exhibit a [3-relaxation in the range of --50 °
to --100 °C, 7t, 87, 88) and a lower temperature 7-transition as well 7t, 89).
The role of test specimen geometry in unstable fracture has been described
in a number of reports 84, 90, 91~ For some materials, different geometries can generate
both stable and unstable crack propagation at the same test conditions 80,92,93)
However, epoxies have been extensively tested in a wide variety of test geometries
and they consistently fail by a stick-slip mechanism. It appears that unstable fracture
is an inherent characteristic of epoxies, and cannot be attributed solely to
fracture specimen geometry.
Initial examination of epoxy fracture surfaces resulting from stick-slip fracture
suggests that considerable plastic deformation accompanies this type of failure.
The surface is not smooth, but typically consists of a distribution of rough and
hackled regions 57,76) In fact, these regions often can be associated with the
peaks and valleys on the load-displacement curve. For epoxies, however, such
gross plastic failure would be surprising, especially in mode I (tensile) fracture
where these materials typically display quite brittle behavior. Careful examination
of epoxy fracture surfaces by Phillips et al. 76~ suggested that the two surfaces
actually fit together and, therefore, that the roughness was a result of the
growing crack simply jumping in and out of the fracture plane. This concept was
supported further by surface profile measurements performed by Yamini and
and Young 57) who showed that opposing fracture surfaces tended to inter-
lock, even on levels down to 10 lam. Along with Phillips et al. 76), they concluded
that the surface roughness was a result of crack tip deviation rather than extensive
140 J.D. LeMay and F. N. Kelley
plastic deformation at the crack tip. However, Yamini and Young pointed out
plastic deformations still may have taken place on a scale smaller than 10 Ixm.
It appears that crack tip deviation is a consequence rather than the cause
of unstable fracture. The current most plausible explanation for unstable fracture
was first proposed by Gledhill et al. 66), who invoked the concept of crack
tip blunting due to plastic deformation at the crack tip. The role of crack
blunting in stick-slip fracture was described by Phillips et al. 76) who directly observed
a sharp crack in a double torsion specimen under continuous loading. Apparently,
the sharp crack progressively became more blunted until a certain critical load
was attained. At this toad a new, sharp crack appeared at the blunted crack tip and
propagated rapidly for a short time before arresting. This new crack then be-
haved as the original crack under continued loading.
The instability of the new crack can be attributed to the fact that the stress
field ahead of the original blunted crack was many times greater than that
required to propagate a sharp crack. Once the new crack experiences the excessive
stress field, it propagates unstably until the excess strain energy is dissipated.
The deviation of this crack from the fracture plane may result from the crack
being unaware of the constraints to its growth during rapid, unstable propagation.
Crack tip blunting is attributed to localized yielding at the crack tip. Localized
yielding may result from shear deformation, or normal stress deformation. Unlike
shear deformation, which occurs at constant colume, normal stress deformation
involves a volume dilatation and is considered to be responsible for the formation
of crazes in thermoplastics. Since crazes are not observed in highly crosslinked
epoxies, it is generally assumed that plastic deformation at the crack tip takes
place via a shear yielding process.
Localized yielding at the crack tip successfully accounts for the initiation behavior
of unstable crack growth in epoxy thermosets. The sensitivity of the yield stress to
temperature and rate is reflected by the initiation fracture toughness, i.e., K, which
decreases with rate and increases with temperature. In fact, it has been shown that
the yield stress correlates uniquely with the overall fracture behavior of a variety of
epoxies 57,ss). Epoxies with tow yield stresses undergo significant crack blunting
and fail unstably, while high yield stress epoxies exhibit little crack blunting and
fail by stable crack growth. As for the arrest behavior of unstable crack growth,
little can be said as no significant dependencies of K a on composition, processing
or materials have been reported.
The authors studied the glassy fracture behavior of the homologous series of
D G E B A / D D S networks listed in Table 2. The fracture specimen employed was
the double torsion test piece. Fracture data were collected over the temperature range
T g - - 120 to Tg - - 20 K, and all testing was performed at a single slow crosshead
rate of 0.05 cm/min. This test rate was chosen because it minimized hysteretic
effects and made all the networks fracture unstably over most of the temperatures
investigated.
Structureand UltimatePropertiesof EpoxyResins 141
--+ ~
16l
1.2
1007F
_.
6 T ,t
/
t
f
I
I
I
,,,,£
t I I I }lg
I
I
J
I
1"6I
1,2 t
1004 F " I
• • v i
I
0.8 t~
[ I t I I I
[~ 1.6
no
~ 1.2 1 0 0 ~ . . . . ~ ,~
ff o..~0_~_ 0_.• o.
O.g
t I ~ J 1 I
1"2f
1001F
0.8 2" ~O-O'-w • 0---
I I I I I I
0 ~0 80 120
0.8 [
828 ~e
,L r
0.6
I I
80 120 160 200
T (°C}
Fig. 12. StoichiometricDGEBA/DDS network fracture toughness versus temperature. Fracture
toughness from double torsion specimenat crosshead rate of 0.05 cm/min. Network Tg's shown
by dashed lines. O Initiation; • Arrest; ~ Stable crack growth. (After LeMays))
suspected that a plane strain to plane stress transition is responsible for the
anomalous crack initiation behavior of the 1007F/DDS network.
To investigate the influence of network structure on the fracture behavior
of the networks, the fracture toughnesses in Fig. 12 were converted to fracture
energies via. Eq. (13). The glassy moduli needed for these calculations were
determined as a function of temperature using dynamic mechanical testing. The
temperature dependence of the storage moduli is shown in Fig. 13, for data
collected at the slow cyclic frequency of 0.16 hz. When the five DGEBA/DDS net-
works are compared at the same reduced temperatures ( T g - T), the two highest
crosslinked networks exhibit lower moduli than the others which exhibit apparently
equivalent values. The actual magnitude of the glassy moduli depend on the
test frequency. While noting that ideally K and E should be determined at the
same rate for the sake of accuracy in calculating 2 J or G, it is reasonable
to expect that the relative ordering of fracture energies will not be altered by such dif-
ferences.
The calculated glassy fracture energies of the DGEBA/DDS networks are
shown in Fig. 14 as a function of temperature, shifted with respect to each
network Tg. Also displayed are the rubbery fracture data previously shown in
Fig. 6. Due to data overlap and scatter, the individual data points for the
glassy data are not shown (the single points shown serve only to identify the
curves). Initiation of unstable fracture is indicated by the dashed curves while
the solid curves label the arrest and stable fracture energies. The glassy fracture
energies exhibit the same general dependence on temperature observed for the
fracture toughnesses in Fig. 12. It should be noted, however, that a direct
correspondence between K and 2 J should not be presumed, since the fracture energy
depends on both K and E, which may exhibit different dependencies on test
conditions and network structure.
Q_
c~2
• 828
o 1001F
• 1002F
rad 100/, F
R=I.0 s-
1007F
1
2O 40 60 80 100
Tg--T (°C)
Fig. 13. Stoichiometric DGEBA/DDS network glassy storage modulus versus reduced test tempera-
ture, Tg--T. Measured in rectangular torsion mode at 0.16 hz ( t .0 rad/s) frequency. • Epon 828/DDS;
O Epon 1001F/DDS; A Epon 1002F/DDS; ~ Epon 1004F/DDS; O Epon 1007/DDS. (After
LeMay s~)
Structure and Ultimate Properties of Epoxy Resins 143
O,
if-.
E
--9
v
~2 --- initiation
c,/
... a r r e s t
O
• 828
o 1001 F
A 1002 F
1 1004 F
v 1007 F
., I I
Fig. 14. Stoichiometric DGEBA/DDS network fracture energy versus reduced test temperature,
Ts --T. Both rubbery and glassy fracture behavior are illustrated. Individual data points for
glassy behavior are not plotted because of overlap. Rubbery fracture energies are from Fig. 6, and
glassy fracture energies are from Fig. 12 via Eq. (13) • Epon 828/DDS; O Epon 100tF/DDS;
& Epon 1002F/DDS; A Epon 1004F/DDS; O Epon 1007F/DDS. (After LeMay a))
There are several interesting aspects of the glassy fracture behavior displayed
in Fig. 14. First is the relatively large magnitude of glassy fracture energies.
These suggest that the considerable energy dissipation is accompanying the fracture
of the networks. Second, the glassy fracture energies span a comparatively narrow
range of values compared to the rubbery data over equivalent Tg-shifted temperature
ranges. Third is the apparent ordering o f the glassy fracture energies with Me. The
initiation data, for the most part, increase with increasing Me, but a systematic
relationship is certainly not apparent. This may be due, in part, to the data
scatter, but it is likely that other factors play as yet unidentified roles in
the initiation o f unstable cracks. The arrest fracture energies, on the other hand,
increase systematically with increasing M r. The average arrest energies of each o f
the D G E B A / D D S networks are listed in Table 6. The 4- values are standard
Network 2J~
(A/E = 1) J/m2
828/DDS 158 + 6
1001F/DDS 245 + 6
1002F/DDS 299 5: 6
1004F/DDS 366 5:14
1007F/DDS 533 4- 17
144 J, D, LeMay and F, N. Kelley
2.8i
ff-
.EE
2.6
2.4
03
0
2,2
21s 310
log Mc
Fig. 15. Stoichiometric DGEBA/DDS network arrest fracture energies versus M c. A line of 1/2 slope
fits the data points. (After LeMay s~)
deviations o f the mean value, averaged over all d a t a collected for each network
(usually 50-100 independent points).
A n interesting quantitative relationship is revealed when the arrest fracture energies
are plotted against M r as shown in Fig. 15. It is found that an arbitrary line
o f 1/2-slope provides an excellent fit to the d a t a (in comparison, linear regression
fitting yields a slope o f about 0.55). Since both axes are plotted in logarithmic scales,
a simple power law relationship is indicated, i.e.,
where the proportionality constant, ks, has an approximate value of 6.1 (J/m z)
(g/mole) - m .
To the authors' knowledge, there have been no reports in the literature
quantifying an M r dependence o f the glassy fracture energy o f thermosets. 1
In fact, a n u m b e r o f studies indicate that such a simple dependence does not
exist 31'33'73) F o r example, epoxy networks with nearly equivalent M r have
been observed to display widely different fracture behaviors 73). Apparently micro-
structures other than network chain length also influence glassy fracture, and may
complicate o r obscure simple dependencies such as observed for the D G E B A / D D S
networks.
The effects o f microstructure may be more complex in the initiation o f
c r a c k growth in epoxies. While M r is apparently the only structural variable
to which crack arrest is sensitive in the D G E B A / D D S networks, crack initiation
1 The only related work in this area is a theory put forth by Kramer 94) for the fracture of low mole-
cular weight, thermoplastics where the molecular weight, M, is less than the entanglement molecular
weight. The theory predicts that the fracture energy is proportional to the root-mean-square end-to-
end distance of the chains between entanglements, from which a proportionality to M ~/z is obvious.
Kramer showed that the theory is very successful in predicting the fracture energies of low molecular
weight polystyrene. This theory is not applicable to highly crosslinked polymer glasses, however,
since these materials do not fail by the craze mechanism that formed the basis of Kramer's theory.
Structure and Ultimate Properties of Epoxy Resins 145
Thus, even when dissipative mechanisms were prevalent, the M~/2 dependence
was obeyed. For the observed glassy fracture behavior to be explained in similar
terms, it is necessary that the network chains in the yielded zone at the highly
strained crack tip be characterized by this same behavior. That is, while the
chains may dissipate considerable energy during extension, this loss of energy must
be independent o f the network chain length. Furthermore, the basic Lake and
T h o m a s concept must apply to the network chains at the crack tip, i.e., all
the bonds in the network chains crossing the fracture plane must be stretched
to their fully extended length before chain rupture occurs.
The idea that cracks must grow through the crack tip plastic zone and therefore
are subject to the properties of the material in that zone has been suggested by
others ls'31'67~. However, a network structure dependence such as that found
by the authors has not been previously reported. The similarities in the dependence
o f both glassy and rubbery fracture energies on epoxy network Me suggests the
following: (1) The process o f fracture in highly crosslinked polymer networks, whether
above or below Tg, depends on similar chain rupture mechanisms; (2) Network de-
fects, like chain branching and soluble fractions, are important in the process o f
glassy fracture, especially crack growth initiation, but specific roles have yet to be
determined; and (3) In terms of general behavior, crack arrest in unstable
crack propagation below Tg is similar to threshold fracture above Tg.
5 References
I. Du~ek, K., Ilavsk~,, M., Lurifik, S. : J. Polym. Sci., Poly. Symp. Ed. 53, 29 (1975)
2. Du~ek, K, Ilavsk~,, M.: Colloid and Polym. Sci. 258, 605 (1980)
3. Du.~ek, K.: Rubber Chem. Technol. 55, 1 (i982)
4. Dugek, K., Ilavsk~,, M.: J. Polym. Sci., Phys. Ed. 21, 1323 (1983)
5. Ilavsk~, M., Bogdanova, L., Dugek, K.: J. Polym. Sci., Potym. Phys. Ed. 22, 265 (1984)
6. LeMay, J., Swetlin, B., Kelley, F.: Org. Coatings and Appl. Polym. Sci. Proc. 48, 715
(1983)
7. LeMay, J., Swetlin, B., Kelley, F. In: "Characterization of highly crosslinked polymers."
Labana, S., Dickie, R. (eds.), ACS Symp. Series 243, American Chemical Society, Washington,
D.C. (1984)
8. LeMay, J.: Ph. D. Dissertation, Univ. of Akron (1985)
9. Labana, S. S., Newman, S., Chompff, A. J. In: "'Polymer networks: structural and mechanical
properties." Chompff, A. J., Newman, S. (eds.), p. 453. New York: Plenum Press 1971
10. Mijovic, J., Koutsky, J. A.: Polymer 20, 1095 (1979)
11. Du~ek, K., Ple~til, J., Lednick~,, F., Lufmk, S.: Polymer 19, 393 (t978)
12. Uhlmann, D. R.: Farad. Dis. 68, 87 (1976)
13. Stevens, G. C., Champion, J. V., Liddel, P., Dandridge, A.: Chem. Phys. Lett. 71, 104
(1980)
14. Lind, A. C.: ACS Polym. Preprints 22, 333 (1982)
15. Kreiblich, U. T., Schmid, R. J. Polym. Sci.: Symp. No. 53, 175 (1975)
16. Morgan, R. J., O'Neal, J. E.: J. Mat. Sci. 12 (1966)
17. Racich, J. L., Koutsky, J. A.: J. Appl. Polym. Sci. 20, 2111 (1976)
18. Gledhill, R. A, Kinloch, A. J.: Polym. Eng. Sci. 19, 82 (1979)
19. Kelley, F. N., Trainor, D. R.: Polym. Bull. 7, 369 (1982)
Structure and Ultimate Properties of Epoxy Resins 147
20. Treloar, I. R. G. : "The physics of rubber elasticity", p. 67. Oxford: Clarendon Press 1975
21. Flory, P. J.: Proc. Roy. Soc. Lond. A351, 351 (1976)
22. Flory, P. J. : J. Chem. Phys. 66, 5270 (1977)
23. Graessley, W. W.: Macromolecules 8, 186 (1975)
24. Pearson, D. S., Graessley, W. W. : Macromolecules 11, 528 (1978)
25. Pearson, D. S., Graessley, W. W. : Macromolecules 13, 1001 (1980)
26. Langley, N. R. : Macromolecules 1, 348 (1968)
27. Langley, N. R., Polmanteer, K. E. : J. Polym. Sci., Polym. Phys. Ed. 12, 1203 (1974)
28. Du~ek, K., Ilavsk~, M.: J. Polym. Sci., Polym. Phys. Ed. 21, 1323 (1983)
29. Treloar, L. R. G.: "'The physics of rubber elasticity," p. 105. Oxford: Clarendon Press 1975
30. Bell, J.: J. Polym. Sci., Part A-2, 8, 417 (1970)
31. Swetlin, B. J. : Ph.D. Dissertation, Univ. of Akron (1984)
32. King, N. E., Andrews, E. H.: J. Mat. Sci. 13, 1291 (1978)
33. Manson, J. A., Sperling, L. H., Kim, S. L. In: "Influence of Crosslinking on the mechanical
properties of high T~ polymers," AFML TR-77-109, A. F. Materials Laboratory, WPAFB,
Ohio, 1977
34. Katz, D., Tobolsky, A. V.: Polymer 4, 417 (1963)
35. Kaelble, D. H.: J. Appl. Polym. Sci. 9, 1213 (1965)
36. Muryama, T., Bell, J. P.: J, Polym. Sei., Part A-2, 8, 437 (1970)
37. Lufi~k, S., Du]ek, K.: J. Polym. Sci., Symp. No. 53, p. 45 (1975)
38. Takahama, T., Geil, P. H.: J. Polym. Sci., Polym. Letters 20, 453 (1982)
39. Andrews, E. H.: J. Mat. Sci. 9, 887 (1974)
40. Lake, G, J., Lindley, P. B.: J. Appl. Polym. Sci. 9, 1233 (1965)
41. Mueller, H. K., Knauss, W. G.: Trans. Soc. Rheol. 15, 217 (1971)
42. Mueller, H. K., Knauss, W. G. : Trans. A.S.M.E. 38 (E2), 483 (1971)
43. Ahagon, A., Gent, A. N.: J. Polym. Sci., Phys. Ed. 13, 1903 (1975)
44. Gent, A. N., Tobias, R. H.: J. Polym. Sci., Phys. Ed. 20, 2051 (1982)
45. Lake, G. J., Thomas, A. G.: Proc. Roy. Soc. Ser. A, 300, 108 (1967)
46. Treloar, L. R. G.: "The physics of rubber elasticity", Chap. 9. Oxford: Clarendon Press 1975
47. Morgan, R. J., Treloar, L. R. G.: J. Polym. Sci., Part A-2, 10, 51 (1972)
48. Greensmith, H. W., Mullins, L., Thomas, A. G.: Trans. Soc. Rheol. 4, 179 (1960)
49. Greensmith, H. W., Thomas, A. G.: J. Polym. Sci. 21, 175 (1956)
50. Mullins, L.: Trans. Inst. Rubb. Ind. 35, 213 (1959)
51. Kadir, A., Thomas, A. G.: Rubber Chem. Technol. 54, 15 (1981)
52. Bennett, S. J., Anderson, G. P., Williams, M. L.: J. Appl. Polym. Sci. 14, 735 (1970)
53. Su, L. : Ph.D. Dissertation, Univ. of Akron (1983)
54. Williams, M. L., Landel, R. F., Ferry, J. D.: J. Amer. Chem. Soc. 77, 3701 (1955)
55. Broutman, L. J., McGarry, F. J.: J. Appl. Polym. Sci. 9, 609 (1965)
56. Mostovoy, S., Ripling, E. J.: J. Appl. Polym. Sci. 10, 1351 (1966)
57. Yamini, S., Young, R. J.: J. Mat. Sci. 15, 1823 (1980)
58. Kinloch, A. J., Williams, J. G. : J. Mat. Sci. 15, 987 (1980)
59. Buchman, A., Katz, D.: Polym. Eng. Sci. 19, 923 (1979)
60. Morgan, R. J.: J. Appl. Polym. Sci. 23, 2711 (1979)
61. Ophir, Z. H., Emerson, J. A., Wilkes, G. L.: J. Appl. Phys. 49, 5032 (1978)
62. Kong, E., Wilkes, G. L., McGrath, J. E., Banthia, A. K., Mohajer, Y., Tant, M. R.:
Polym. Eng. Sci. 21, 943 (1981)
63. Kong, E., In: "Characterization of highly crosslinked polymers". Labana, S. S., Dickie, R. A.
(eds.), ACS Symposium Series 243, 125 (1983)
64. Irwin, G. R. In: "Encyclopedia of physics", vol. 6. Berlin: Springer Press (1958)
65. Yamini, S., Young, R. J. : J. Mat. Sci. 14, 1609 (1979)
66. Gledhill, R. A., Kinioch, A. J., Yamini, S., Young, R. J.: Polymer 19, 574 (1978)
67. Young, R. J., Beaumont, P. W. R. : J. Mat. Sci. 11, 779 (1976)
68. Yamini, S., Young, R. J. : Polymer 18, 1075 (1977)
69. Scott, J. M., Wells, G. M., Phillips, D. C.: J. Mat. Sci. 15, 1436 (1980)
70. Chang, T. D., Carr, S. H., Brittain, J. O.: Polym. Eng. Sci. 22, 1205 (1982)
71. Chang, T. D., Carr, S. H., Brittain, J. O.: Polym. Eng. Sci. 22, 1213 (1982)
72. Chang, T. D., Carr, S. H., Brittain, J. O.: Polym. Eng. Sci. 22, 1221 (1982)
148 J . D . LeMay and F. N. Kelley
73. Chang, T. D., Brittain, J. O. : Polym. Eng. Sci. 22, 1228 (1982)
74. Morgan, R. J., Mones, E. T.: Composites Tech. Rev. 1(4), 17 (1979)
75. Morgan, R. J., Happe, J. A., Mones, E. T. : 28th National SAMPE Symp., Anaheim, CA (1983)
76. Phillips, D. C., Scott, J. M., Jones, M.: J. Mat. Sci. 13, 311 (1978)
77. Bell, J. P.: J. Appl. Polym. Sci. 14, 190t (1970)
78. Yamini, S., Young, R. J.: J. Mat. Sci. 15, 1814 (1980)
79. Kim, S. L., Skibo, M. D., Manson, J. A., Hertzberg, R. W., Janiszewski, J. : Polym. Eng. Sci.
18, 1093 (1978)
80. Mai, Y. W., Atkins, A. G.: J. Mat. Sci. 10, 2000 (1975)
8t. Young, R. J. In: "'Developments in polymer fracture-l', p. 183. Andrews, E. H. (ed.). London:
Applied Science 1979
82. Marshall, G. P., Coutts, L. H., Williams, J. G.: J. Mat. Sci. 9, 1409 (1974)
83. Mai, Y. W., Leete, N. B.: J. Mat. Sci. 14, 2264 (1979)
84. Atkins, A. G., Lee, C. S., Caddell, R. M.: J. Mat. Sci. 10, 1381 (1975)
85. Parvin, M., Williams, J. G.: J. Mat. Sci. 10, 1883 (1975)
86. Mai, Y. W., Atkins, A. G.: J. Mat. Sci. 11, 677 (1976)
87. Williams, J. G. : J. Appl. Polym. Sci. 23, 3433 (1979)
88. Takahama, T., Geil, P. H.: J. Polym. Sci., Polym. Phys. Ed. 20, 1979 (1982)
89. Pogany, G. A.: Polymer 22, 66 (1970)
90. Gurney, C., Hunt, J.: Proc. Roy. Soc. A299, 508 (1967)
91. Mai, Y. W.: J. Mat. Sci. 11, 570(1976)
92. Mai, Y. W., Atkins, A. G, Caddell, R. M. : Int. J. Fract. 11, 939 (1975)
93. Mai, Y. W.: Int. J. Fract. 10, 292 (1974)
94. Kramer, E. J.: J. Mat. Sci. 14, 1381 (1978)
95. Kramer, E. J., Hart, F. J.: Polymer, 25, 1667 (1984)
96. Donald, A. M., Kramer, E. J.: Polymer 23, 1183 (1982)
97. Donald, A. M., Kramer, E. J. : J. Polym. Sci., Phys. 20, 899 (1982)
Editor: K. Du~ek
Received September 2, 1985
Author Index Volumes 1-78
B6hm, L. L., Chmeti?, M., L6hr, G., Schmitt, B. J. and Schulz, G. V.: Zustande und Reaktionen
des Carbanions bei der anionischen Polymerisation des Styrols. Vol. 9, pp. 1-45.
Bovey, E. A. and Tiers, G. V. D. : The High Resolution Nuclear Magnetic Resonance Spectroscopy
of Polymers. Vol. 3, pp. 139-195.
Braun, J.-M. and Guillet, J. E.: Study of Polymers by Inverse Gas Chromatography. Vol. 21,
pp. 107-145.
Breitenbach, J. W., Olaj, O. F. und Sommer, F. : Polymerisationsanregung durch Elektrolyse. Vol. 9,
pp. 47-227.
Bresler, S. E. and Kazbekov, E. N. : Macroradical Reactivity Studied by Electron Spin Resonance.
Vol. 3, pp. 688-711.
BucknaH, C. B. : Fracture and Failure of Multiphase Polymers and Polymer Composites. Vol. 27,
pp. 121-148.
Burchard, HI. : Static and Dynamic Light Scattering from Branched Polymers and Biopolymers.
Vol. 48, pp. 1-124.
Bywater, S. : Polymerization Initiated by Lithium and Its Compounds. Vol. 4, pp. 66-110.
Bywater, S. : Preparation and Properties of Star-branched Polymers. Vol. 30, pp. 89-116.
Candau, S., Bastide, J. and Delsanti, M. : Structural. Elastic and Dynamic Properties of Swollen
Polymer Networks. Vol. 44, pp. 27-72.
Carrick, W. L.." The Mechanism of Olefin Polymerization by Ziegler-Natta Catalysts. Vol. 12,
pp. 65-86.
Casale, A. and Porter, R. S.: Mechanical Synthesis of Block and Graft Copolymers. Vol~ 17,
pp. 1-71.
Cerf, R.: La dynamique des solutions de macromolecules dans un champ de vitesses. Vol. 1,
pp. 382-450.
Cesca, S., Priola, A. and Bruzzone, M. : Synthesis and Modification of Polymers Containing a
System of Conjugated Double Bonds. Vol. 32, pp. 1-67.
Chiellini, E., Solaro, R., Galli, G. and Ledwith, A. : Optically Active Synthetic Polymers Containing
Pendant Carbazolyl Groups. Vol. 62, pp. 143-170.
Cicchetti, O.." Mechanisms of Oxidative Photodegradation and of UV Stabilization of Polyolefins.
Vol. 7, pp. 70~112.
Clark, D. 7".." ESCA Applied to Polymers. Vol. 24, pp. 125 188.
Coleman, Jr., L. E. and Meinhardt, N. A. : Polymerization Reactions of Vinyl Ketones. Vol. 1,
pp. 159-179.
Comper, W. D. and Preston, B. N.." Rapid Polymer Transport in Concentrated Solutions. Vol. 55,
pp. 105 152.
Corner, 7". : Free Radical Polymerization -- The Synthesis of Graft Copolymers. Vol. 62, pp. 95-142.
Crescenzi, V. : Some Recent Studies of Polyelectrolyte Solutions. Vol. 5, pp. 358-386.
Crivello, J. V. : Cationic Polymerization -- Iodonium and Sulfonium Salt Photoinitiators, Vol. 62,
pp. 1-48.
Davydov, B. E. and Krentsel, B. A.: Progress in the Chemistra of Polyconjugated Systems. Vol. 25,
pp. 1-46.
Dettenmaier, M. : Intrinsic Crazes in l~olycarbonate Phenomenology and Molecular Interpretation
of a New Phenomenon. Vol. 52/53, pp. 57-104.
Dobb, M. G. and Mclntyre, J. E.." Properties and Applications of Liquid-Crystalline Main-Chain
Polymers. Vol. 60/61, pp. 6148.
D61l, W. : Optical Interference Measurements and Fracture Mechanics Analysis of Crack Tip Craze
Zones. Vol. 52/53, pp. 105-168.
Doi, Y. see Keii, T.: Vol. 73/74, pp. 201-248.
Dole, M. : Calorimetric Studies of States and Transitions in Solid High Polymers. Vol. 2, pp. 221-274.
Donner, J. B., Vidal, A. : Carbon Black-Surface Properties and Interactions with Elastomers. Vol. 76,
pp. 103 128.
Dorn, K., Hupfer, B., and Ringsdorf, H.: Polymeric Monolayers and Liposomes as Models for
Biomembranes How to Bridge the Gap Between Polymer Science and Membrane Biology?
Vol. 64, pp. 1 54.
Aulhor Index Volumes 1-78 151
Eastham, A. M. : Some Aspects of the Polymerization of Cyclic Ethers. Vol. 2, pp. 18-50.
Ehrlieh, P. and Mortimer, G. A. : Fundamentals of the Free-Radical Polymerization of Ethylene.
Vol. 7, pp. 386-448.
Eisenber 9, A.: Ionic Forces in Polymers. Vol. 5, pp. 59-112.
Eiss, N. S. Jr. see Yorkgitis, E. M. Vol. 72, pp. 79-110.
Elias, H.-G., Baress, R. und Watterson, J. G.: Mittelwerte des Molekulargewichts und anderer Ei-
genschaften. Vol. 11, pp. 111-204.
Elsner, G., Rickel, Ch. and Zachmann, H. G. : Synchrotron Radiation Physics. Vol. 67, pp. 1 58.
Elyashevich, G. K. : Thermodynamics and Kinetics of Orientational Crystallization of Flexible-
Chain Polymers. Vol. 43, pp. 207-246.
Enkelmann, V. : Structural Aspects of the Topochemical Polymerization of Diacetylenes. Vol. 63.
pp. 91-136.
Entelis, S. G., Evreinov, V. V., Gorshkov, A. V. : Functionally and Molecular Weight Distribution of
Telchelic Polymers. Vol. 76, pp. 12%175.
Evreinov, V. V. see Entelis, S. G. Vol. 76, pp. 129-175.
Ferruti, P. and Barbueei, R. : Linear Amino Polymers: Synthesis, Protonation and Complex Forma-
tion. Vol. 58, pp. 55-92.
Finkelmann, H. and Rehaoe, G. : Liquid Crystal Side-Chain Polymers. Vol. 60/61, pp. 99 172.
Fischer, H.: Freie Radikale w~ihrend der Polymerisation, nachgewiesen und identifiziert durch
Elektronenspinresonanz. Vol. 5, pp. 463-530.
Flory, P. J.." Molecular Theory of Liquid Crystals. Vol. 59, pp. 1-36.
Ford, W. T. and Tomoi, M. : Polymer-Supported Phase Transfer Catalysts Reaction Mechanisms.
Vol. 55, pp. 49-104.
Fradet, A. and Mar~chal, E.: Kinetics and Mechanisms of Polyesterifications. I. Reactions of Diols
with Diacids. Vol. 43, pp. 51-144.
Franz, G. : Polysaccharides in Pharmacy. Vol. 76, pp. 1-30.
Friedrich, K. : Crazes and Shear Bands in Semi-Crystalline Thermoplastics. Vol. 52/53, pp. 225-274.
Fujita, H. : Diffusion in Polymer-Diluent Systems. Vol. 3, pp. 1~17.
Funke, W. : f2ber die Strukturaufkl~irung vernetzter MakromolekiJle, insbesondere vernetzter Poly-
esterharze, mit chemischen Methoden. Vol. 4, pp. 15%235.
Garbraikh, L. S. and Rioovin, Z. A. : Chemical Transformation of Cellulose. Vol. 14, pp. 87-130.
Galli, G. see Chiellini, E. Vol. 62, pp. 143-170.
Gallot, B. R. M. : Preparation and Study of Block Copolymers with Ordered Structures, Vol. 29,
pp. 85-156.
Gandini, A. : The Behaviour of Furan Derivatives in Polymerization Reactions. Vol. 25, pp. 47-96.
Gandini, A. and Cheradame, H.: Cationic Polymerization. Initiation with Alkenyl Monomers.
Vol. 34/35, pp. 1-289.
Geckeler, K.. Pillai, V. N. R., and Mutter, M. : Applications of Soluble Polymeric Supports. Vol. 39,
pp. 65-94.
Gerrens, H. : Kinetik der Emulsionspolymerisation. Vol. 1, pp. 234-328.
Ghiagino, K. P., Roberts, A. J. and Phillips, D. : Time-Resolved Fluorescence Techniques in Polymer
and Biopolymer Studies. Vol. 40, pp. 69-167.
152 Author Index Volumes 1-78
Hagihara, N., Sonogashira, K. and Takahashi, S. : Linear Polymers Containing Transition Metals in
the Main Chain. Vol. 41, pp. 149 179.
Hasegawa, M. : Four-Center Photopolymerization in the Crystalline State. Vol. 42, pp. 149.
Hatano, M.: Induced Circular Dichroism in Biopolymer-Dye System. Vol. 77, pp. 1-121.
Hay, A. S. : Aromatic Polyethers. Vol. 4, pp. 496-527.
Hayakawa, R. and Wadu, Y. : Piezoelectricity and Related Properties of Polymer Films. Vol. 11,
pp. 1-55.
Heidemann, E. and Rath, W.. Synthesis and Investigation of Collagen Model Peptides. Vol. 43,
pp. 145-205.
Heitz, W. : Polymeric Reagents. Polymer Design, Scope, and Limitations. Vol. 23, pp. 1-23.
Helfferich, F.: Ionenaustausch. Vol. 1, pp. 329-381.
Hendra, P. J. : Laser-Raman Spectra of Polymers. Vol. 6, pp. 151-169.
Hendrix, J. : Position Sensitive "X-ray Detectors". Vol. 67, pp. 59-98.
Henrici-Oliv~, G. und Olive, S. : Ketteniibertragung bei der radikalischen Polymerisation. Vol. 2,
pp. 496-577.
Henrici-Oliv~, G. und OlivO, S. : Koordinative Polymerisation an 16slichen (Jbergangsmetall-Kataly-
satoren. Vol. 6, pp. 421~172.
Henrici-Olivk, G. and OlivO, S. : Oligomerization of Ethylene with Soluble Transition-Metal Catalysts.
Vol. 15, pp. 1-30.
Henrici-OlivO, G. and Olive, S. : Molecular Interactions and Macroscopic Properties of Polyacrylo-
nitrile and Model Substances. Vol. 32, pp. 123-152.
Henriei-OlivO, G. and Olive, S. : The Chemistry of Carbon Fiber Formation from Polyacrylonitrile.
Vol. 51, pp. 1-60.
Herrnans, Jr. J., Lohr, D. and Ferro, D. : Tretament of the Folding and Unfolding of Protein Molecules
in Solution According to a Lattic Model. Vol. 9, pp. 229-283.
Higashimura, T. and Sawamoto, M. : Living Polymerization and Selective Dimerization: Two Extremes
of the Polymer Synthesis by Cationic Polymerization. Vol. 62, pp. 49-94.
Hoffmann, A. S. : Ionizing Radiation and Gas Plasma (or Glow) Discharge Treatments for Prepara-
tion of Novel Polymeric Biomaterials. Vol. 57, pp. 141-157.
Holzmiiller, W. : Molecular Mobility, Deformation and Relaxation Processes in Polymers. Vol. 26,
pp. 1-62.
Hutehison, J. and Ledwith, A. : Photoinitiation of Vinyl Polymerization by Aromatic Carbonyl
Compounds. Vol. 14, pp. 49-86.
Iizuka, E. : Properties of Liquid Crystals of Polypeptides: with Stress on the Electromagnetic Orien-
tation. Vol. 20, pp. 79-107.
Ikada, Y. : Characterization of Graft Copolymers. Vol. 29, pp. 47-84.
Ikada, Y. : Blood-Compatible Polymers. Vol. 57, pp. 103-140.
Imanishi, Y. : Synthese, Conformation, and Reactions of Cyclic Peptides. Vol. 20, pp. 1-77.
Inaoaki, H. : Polymer Separation and Characterization by Thin-Layer Chromatography. Vol. 24,
pp. 189-237.
Inoue, S. : Asymmetric Reactions of Synthetic Polypeptides. Vol. 21, pp. 77-106.
lse, N. : Polymerizations under an Electric Field. Vol. 6, pp. 347-376.
Author Index Volumes 1-78 153
Ise, N. : The Mean Activity Coefficient of Polyelectrolytes in Aqueous Solutions and Its Related
Properties. Vol. 7, pp. 536-593.
lsihara, A. : Intramolecular Statistics of a Flexible Chain Molecule. Vol. 7, pp. 449~,76.
Isihara, A. : Irreversible Processes in Solutions of Chain Polymers. Vol. 5, pp. 531 567.
Isihara, A. and Guth, E. : Theory of Dilute Macromolecular Solutions. Vol. 5, pp. 233-260.
lwatsuki, S. : Polymerization of Quinodimethane Compounds. Vol. 58, pp. 93-120.
Kamachi, M. : Influence of Solvent on Free Radical Polymerization of Vinyl Compounds. Vol. 38,
pp. 55-87.
Kaneko, M. and Yamada, A.: Solar Energy Conversion by Functional Polymers. Vol. 55, pp. 148.
Kawabata, S. and Kawai, H. : Strain Energy Density Functions of Rubber Vulcanizates from Biaxial
Extension. Vol. 24, pp. 89-124.
Keii, 7"., Doi, Y. : Synthesis of "Living" Polyolefins with Soluble Ziegler-Natta Catalysts and Applica-
tion to Block Copolymerization. Vol. 73/74, pp. 201-248.
Kelley, F. N. see LeMay, J. D.: Vol. 78, pp. 115-148.
Kennedy, J. P. and Chou, T.: Poly(isobutylene-co-[3-Pinene): A New Sulfur Vulcanizable, Ozone
Resistant Elastomer by Cationic Isomerization Copolymerization. Vol. 21, pp. 1-39.
Kennedy, J. P. and Delvaux, J. M. : Synthesis, Characterization and Morphology of Poly(butadiene-
g-Styrene). Vol. 38, pp. 141-163.
Kennedy, J. P. and Gillham, J. K. : Cationic Polymerization of Olefins with Alkylaluminium Initiators.
Vol. 10, pp. 1-33.
Kennedy, J. P. and Johnston, J. E. : The Cationic Isomerization Polymerization of 3-Methyl-l-butene
and 4-Methyl-l-pentene. Vol. 19, pp. 57-95.
Kennedy, J. P. and Langer, Jr. A. W.: Recent Advances in Cationic Polymerization. Vol. 3,
pp. 508-580.
Kennedy, J. P. and Otsu, T. : Polymerization with lsomerization of Monomer Preceding Propagation.
Vol. 7, pp. 369-385.
Kenned),, J. P. and Rengachary, S. : Correlation Between Cationic Model and Polymerization Reactions
of Olefins. Vol. 14, pp. 148.
Kennedy, J. P. and Trivedi, P. D. : Cationic Olefin Polymerization Using Alkyl Halide -- Alkyl-
aluminium Initiator Systems. I. Reactivity Studies. II. Molecular Weight Studies. Vol. 28,
pp. 83-151.
Kennedy, J. P., Chang, V. S. C. and Guyot, A. : Carbocationic Synthesis and Characterization of
Polyolefins with Si-H and Si~21 Head Groups. Vol. 43, pp. 1-50.
Khoklov, A. R. and Grosberg, A. Yu. : Statistical Theory of Polymeric Lyotropic Liquid Crystals.
Vol. 41, pp. 53-97.
Kinloch, A. J. : Mechanics and Mechanisms of Fracture of Thermosetting Epoxy Polymers. Vol. 72,
pp. 45~i8.
Kissin, Yu. 11.: Structures of Copolymers of High Olefins. Vol. 15, pp. 91-155.
Kitagawa, T. and Miyazawa, T.: Neutron Scattering and Normal Vibrations of Polymers. Vol. 9,
pp. 335414.
Kitamaru, R. and Horii, F. : NMR Approach to the Phase Structure of Linear Polyethylene. Vol. 26,
pp. 139-180.
Knappe, W.: W~irmeleitung in Polymeren. Vol. 7, pp. 477-535.
Koenik, J. L. see Mertzel, E. Vol. 75, pp. 73-112.
Koenig, J. L. : Fourier Transforms Infrared Spectroscopy of Polymers, Vol. 54, pp. 8~154.
Kola~ik, J. : Secondary Relaxations in Glassy Polymers: Hydrophilic Polymethacrylates and Poly-
acrylates: Vol. 46, pp. 119-161.
Koningsveld, R. : Preparative and Analytical Aspects of Polymer Fractionation. Vol. 7.
154 Author Index Volumes 1 78
Kovacs, A. J.: Transition vitreuse dans les polymers amorphes. Etude ph6nom6nologique. Vol. 3,
pp. 394~507.
Krdssig, H. A. : Graft Co-Polymerization of Cellulose and Its Derivatives, Vol. 4, pp. 111-156.
Kramer, E. J. : Microscopic and Molecular Fundamentals of Crazing. Vol. 52/53, pp. 1-56.
Kraus, G.: Reinforcement of Elastomers by Carbon Black. Vol. 8, pp. 155-237.
Kreutz, W. and Welte, W. : A General Theory for the Evaluation of X-Ray Diagrams of Biomembranes
and Other Lamellar Systems. Vol. 30, pp. 161-225.
Krimm, S. : Infrared Spectra of High Polymers. Vol. 2, pp. 51-72.
Kuhn, W., Ramel, A., Walters, D. H., Ebner, G. and Kuhn, H. J.. The Production of Mechanical
Energy t¥om Different Forms of Chemical Energy with Homogeneous and Cross-Striated High
Polymer Systems. Vol. 1, pp. 540-592.
Kunitake, T. and Okahata, Y.: Catalytic Hydrolysis by Synthetic Polymers. Vol. 20, pp. 159-221.
Kurata, M. and Stockmayer, W. H. : Intrinsic Viscosities and Unperturbed Dimensions of Long
Chain Molecules. Vol. 3, pp. 196 312.
Ledwith, A. and Sherrington, D. C.: Stable Organic Cation Salts: Ion Pair Equilibria and Use in
Cationic Polymerization. Vol. 19, pp. 1-56.
Ledwith, A. see Chiellini, E. Vol. 62, pp. 143-170.
Lee, C.-D. S. and Daly, W. H.: Mercaptan-Containing Polymers. Vol. 15, pp. 61-90.
LeMay, J. D., Kelley, F. N . . Structure and Ultimate Properties of Epoxy Resins. Vol. 78, pp. 115-148.
Lindberg, J. J. and Hortling, B. : Cross Polarization - Magic Angle Spinning NMR Studies of Carbo-
hydrates and Aromatic Polymers. Vol. 66, pp. 1-22.
Lipatov, Y. S.: Relaxation and Viscoelastic Properties of Heterogeneous Polymeric Compositions.
Vol. 22, pp. 1-59.
Lipatov, Y. S.: The Iso-Free-Volume State and Glass Transitions in Amorphous Polymers: New
Development of the Theory. Vol. 26, pp. 63-104.
Lohse, F., Zweifel, H. : Photocrosslinking of Epoxy Resins. Vol. 78, pp. 61-81.
Lustoh, J. and Va~g, F. : Anionic Copolymerization of Cyclic Ethers with Cyclic Anhydrides. Vol. 56,
pp. 91-133.
Madec, J.-P. and Mar(chal, E.: Kinetics and Mechanisms of Polyesterifications. II. Reactions of
Diacids with Diepoxides. Vol. 71, pp. 153-228.
Mano, E. B. and Coutinho, F. M. B.." Grafting on Polyamides. Vol. 19, pp. 97-116.
Mar~chal, E. see Madec, J.-P. Vol. 71, pp. 153-228.
Mark, J. E. : The Use of Model Polymer Networks to Elucidate Molecular Aspects of Rubberlike
Elasticity. Vol. 44, pp. 1-26.
Mark, J. E. see Queslel, J. P. Vol. 71, pp. 229 248.
Maser, F., Bode, K., Pillai, V. N. R. and Mutter, M. : Conformational Studies on Model Peptides.
Their Contribution to Synthetic, Structural and Functional Innovations on Proteins. Vol. 65,
pp. 177-214.
McGrath, J. E. see Yorkgitis, E. M. Vol. 72, pp. 79-110.
Mclntyre, J, E. see Dobb, M. G. Vol. 60/61, pp. 61-98.
Meerwall ~;., E., D. : Self-Diffusion in Polymer Systems. Measured with Field-Gradient Spin Echo
NMR Methods, Vol. 54, pp. 1-29.
Mengoli, G. : Feasibility of Polymer Film Coating Through Electroinitiated Polymerization in Aqueous
Medium. Vol. 33, pp. 1-31.
Mertzel, E., Koenik, J. L. : Application of FT-IR and NMR to Epoxy Resins. Vol. 75, pp. 73-112.
Meyerhoff, G.: Die viscosimetrische Molekulargewichtsbestimmung yon Polymeren. Vol. 3,
pp. 59-105.
• Millich, F. : Rigid Rods and the Characterization of Polyisocyanides. Vol. 19, pp. 117-141.
M6ller, M.." Cross Polarization -- Magic Angle Sample Spinning NMR Studies. With Respect to
the Rotational Isomeric States of Saturated Chain Molecules. Vol. 66, pp. 59-80.
Morawetz, H. : Specific Ion Binding by Polyelectrolytes. Vol. 1, pp. 1--34.
Morgan, R. J.: Structure-Property Relations of Epoxies Used as Composite Matrices. Vol. 72,
pp. 1~14.
Author Index Volumes 1 78 155
Morin, B. P., Breusova, L P. and Rogovin, Z. A. : Structural and Chemical Modifications of Cellulose
by Graft Copolymerization~ Vol. 42, pp. 139-166.
Mulvaney, J. E., Oversberger, C. C. and Schiller, A. M. : Anionic Polymerization. Vol. 3, pp. 106-138.
Nakase, Y., Karijama, L and Odajima, A. : Analysis of the Fine Structure of Poly(Oxymethylene)
Prepared by Radiation-Induced Polymerization in the Solid State. Vol. 65, pp. 79-134.
Neuse, E. : Aromatic Polybenzimidazoles. Syntheses, Properties, and Applications. Vol. 47, pp. 1-42.
Nicolais, L. see Apicella, A. Vol. 72, pp. 69-78.
Nuyken, 0., Weidner, R. : Graft and Block Copolymers vis Polymeric Azo Initiators. Vol. 73/74,
pp. 145-200.
Ober, Ch. K., Jin, J.-L and Lenz, R. W. : Liquid Crystal Polymers with Flexible Spacers in the Main
Chain. Vol. 59, pp. 103-146.
Okubo, T. and Ise, N. : Synthetic Polyelectrolytes as Models of Nucleic Acids and Esterases. Vol. 25,
pp. 135-181.
Osaki, K. : Viscoelastic Properties of Dilute Polymer Solutions. Vol. 12, pp. 1-64.
Oster, G. and Nishijima, Y. : Fluorescence Methods in Polymer Science. Vol. 3, pp. 313-331.
Otsu, T. see Sato, T. Vol. 71, pp. 41-78.
Overberoer, C. G. and Moore, J. A. : Ladder Polymers. Vol. 7, pp. 113-150.
Queslel, J. P. and Mark, J. E. : Molecular Interpretation of the Moduli of Elastomeric Polymer Net-
works of Know Structure. Vol. 65, pp. 135-176.
Queslel, J. P. and Mark, J. E. : Swelling Equilibrium Studies of Elastomeric Network Structures.
Vol. 71, pp. 229-248.
Rempp, P., Herz, J., and Borchard, W.: Model Networks. Vol. 26, pp. 107-137.
Richards, R. W. : Small Angle Neutron Scattering from Block Copolymers. Vol. 71, pp. 1-40.
Rigbi, Z. : Reinforcement of Rubber by Carbon Black. Vol. 36, pp. 21-68.
Rogovin, Z. A. and Gabrielyan, G. A. : Chemical Modifications of Fibre Forming Polymers and
Copolymers of Acrylonitrile. Vol. 25, pp. 97-134.
Roha, M. : Ionic Factors in Steric Control. Vol. 4, pp. 353-392.
Roha, M.: The Chemistry of Coordinate Polymerization of Dienes. Vol. 1, pp. 512-539.
Rostami, S. see Walsh, D. J. Vol. 70, pp. 119-170.
Rozengerk, v. A. : Kinetics, Thermodynamics and Mechanism of Reactions of Epoxy Oligomers with
Amines. Vol. 75, pp. 113-166.
Spiess, H. W. : Deutron NMR -- A new Toolfor Studying Chain Mobility and Orientation in
Polymers. Vol. 66, pp. 23-58.
Sprung, M. M. : Recent Progress in Silicone Chemistry. I. Hydrolysis of Reactive Silane Intermediates,
Vol. 2, pp. 442-464.
Stahl, E. and Briiderle, V.: Polymer Analysis by Thermofractography. Vol. 30, pp. 1-88.
Stannett, V. T., Koros, W. J., Paul, D. R., Lonsdale, H. K., and Baker, R. W.: Recent Advances in
Membrane Science and Technology. Vol. 32, pp. 69-121.
Staverman,/1. I. : Properties of Phantom Networks and Real Networks. Vol. 44, pp. 73-102.
Stauffer, D., Coniglio, A. and Adam, M . : Gelation and Critical Phenomena. Vol. 44, pp. 103-158.
Stille, J. K. : Diels-Alder Polymerization. Vol. 3, pp. 48-58.
Stolka, M. and Pai, D. : Polymers with Photoconductive Properties. Vol. 29, pp. 1-45.
Stuhrmann, H.: Resonance Scattering in Macromolecular Structure Research. Vol. 67, pp. t23-164.
Subramanian, R. V. : Electroinitiated Polymerization on Electrodes. Vol. 33, pp. 35-58.
Sumitoma, H. and Hashimoto, K. : Polyamides as Barrier Materials. Vol. 64, pp. 55-84.
Sumitomo, H. and Okada, M . : Ring-Opening Polymerization of Bicyclic Acetals, Oxalactone, and
Oxalactam. Vol. 28, pp. 47-82.
Szeg6, L. : Modified Polyethylene Terephthalate Fibers. Vol. 31, pp. 89-131.
Szwarc, M . : Termination of Anionic Polymerization. Vol. 2, pp. 275-306.
Szwarc, M. : The Kinetics and Mechanism of N-carboxy~-amino-acid Anhydride (NCA) Polymeri-
zation to Poly-amino Acids. Vol. 4, pp. 1-65.
Szwarc, M. : Thermodynamics of Polymerization with Special Emphasis on Living Polymers. Vol. 4,
pp. 457-495.
Szwarc, M. : Living Polymers and Mechanisms of Anionic Polymerization. Vol. 49, pp. 1-175.
Takahashi,/1. and Kawaouehi, M . : The Structure of Macromolecules Adsorbed on Interfaces. Vol. 46,
pp. 1-65.
Takemoto, K. and Inaki, Y. : Synthetic Nucleic Acid Analogs. Preparation and Interactions. Vol. 41,
pp. 1-51.
Tani, H. : Stereospecific Polymerization of Aldehydes and Epoxides. Vol. 11, pp. 57-110.
Tate, B. E.: Polymerization of Itaconic Acid and Derivatives. Vol. 5, pp. 214-232.
Tazuke, S. : Photosensitized Charge Transfer Polymerization. Vol. 6, pp. 321-346.
Teramoto, A. and Fujita, H. : Conformation-dependent Properties of Synthetic Polypeptides in the
Helix-Coil Transition Region. Vol. 18, pp. 65-149.
Theocaris, P. S.: The Mesophase and its Influence on the Mechanical Behavior of Composites. Vot. 66,
pp. 149-188.
Thomas, W. M. : Mechanismus of Acrylonitrile Polymerization. Vol. 2, pp. 401-441.
Tieke, B. : Polymerization of Butadiene and Butadiyne (Diacetylene) Derivatives in Layer Structures.
Vol. 71, pp. 79-152.
Tobolsky, A. V. and DuPr6, D. B. : Macromolecular Relaxation in the Damped Torsional Oscillator
and Statistical Segment Models. Vol. 6, pp. 103-127.
Tosi, C. and Ciampelli, F. : Applications of Infrared Spectroscopy to Ethylene-Propylene Copolymers.
Vol. 12, pp. 87-130.
Tosi, C. : Sequence Distribution in Copolymers: Numerical Tables. Vol. 5, pp. 451-462.
Tran, C. see Yorkgitis, E. M. Vol. 72, pp. 79-110.
Tsuehida, E. and Nishide, 1t. : Polymer-Metal Complexes and Their Catalytic Activity. Vol. 24,
pp. 1-87.
Tsuji, K.: ESR Study of Photodegradation of Polymers. Vol. 12, pp. 131-190.
Tsvetkov, V. and Andreeva, L. : Flow and Electric Birefringence in Rigid-Chain Polymer Solutions.
Vol. 39, pp. 95-207.
Tuzar, Z., Kratochvil, P., and Bohdaneek~, M.: Dilute Solution Properties of Aliphatie Polyamides.
Vol. 30, pp. 117-159.
Uematsu, L and Uematsu, Y. : Polypeptide Liquid Crystals. Vol. 59, pp. 37-74.
158 Author Index Volumes 1-78
Valvassori, A. and Sartori, G.." Present Status of the Multicomponent Copolymerization Theory.
Vol. 5, pp. 28-58.
Vidal, A. see Donnet, J. B. Vol. 76, pp. 103-128.
Viovy, J. L. and Monnerie, L. : Fluorescence Anisotropy Technique Using Synchrotron Radiation
as a Powerful Means for Studying the Orientation Correlation Functions of Polymer Chains.
Vol. 67, pp. 99-122.
Voi#t-Martin, I. : Use of Transmission Electron Microscopy to Obtain Quantitative Information
About Polymers. Vol. 67, pp. 195-218.
Voorn, M. J. : Phase Separation in Polymer Solutions. Vol. 1, pp. 192-233.
Walsh, D. J., Rostami, S. : The Miscibility of High Polymers: The Role of Specific Interactions.
Vol. 70, pp. 119-170.
Ward, L M. : Determination of Molecular Orientation by Spectroscopic Techniques. Vol. 66, pp.
81-116.
Ward, L M. : The Preparation, Structure and Properties of Ultra-High Modulus Flexible Polymers.
Vol. 70, pp. 1-70.
Weidner, R. see Nuyken, 0 . : Vol. 73/74, pp. 145-200.
Werber, F. X. : Polymerization of Olefins on Supported Catalysts. Vol. 1, pp. 180-191.
Wichterle, 0., Sebenda, J., and Krdlidek, J. : The Anionic Polymerization of Caprolactam. Vol. 2,
pp. 578-595.
Wilkes, G. L. : The Measurement of Molecular Orientation in Polymeric Solids. Vol. 8, pp. 91-136.
Wilkes, G. L. see Yorkgitis, E. M. Vol. 72, pp. 79-110.
Williams, G.." Molecular Aspects of Multiple Dielectric Relaxation Processes in Solid Polymers.
Vol. 33, pp. 59-92.
Williams, J. G.. Applications of Linear Fracture Mechanics. Vol. 27, pp. 67-120.
Wrhrle, D. : Polymere aus Nitrilen. Vol. 10, pp. 35-107.
Wrhrle, D. : Polymer Square Planar Metal Chelates for Science and Industry. Synthesis, Properties
and Applications. Vol. 50, pp. 45-134.
Wolf, B. A. : Zur Thermodynamik der enthalpisch und der entropisch bedingten Entrnischung von
Polymerlrsungen. Vol. 10, pp. 109-171.
Woodward, A. E. and Sauer, J. A. : The Dynamic Mechanical Properties of High Polymers at Low
Temperatures. Vol. 1, pp. 114-158.
Wunderlich, B. : Crystallization During Polymerization. Vol. 5, pp. 568~519.
Wunderlich, B. and Baur, H. : Heat Capacities of Linear High Polymers. Vol. 7, pp. 151-368.
Wunderlich, B. and Grebowicz, J. : Thermotropic Mesophases and Mesophase Transitions of Linear,
Flexible Macromolecules. Vol. 60/61, pp. 1~60.
Wrasidlo, W. : Thermal Analysis of Polymers. Vol. 13, pp. 1-99.
Yamashita, Y. : Random and Black Copolymers by Ring-Opening Polymerization. Vol. 28, pp. 1-46.
Yamazaki, N. : Electrolytically Initiated Polymerization. Vol. 6, pp. 377-400.
Yamazaki, N. and Higashi, F. : New Condensation Polymerizations by Means of Phosphorus Com-
pounds. Vol. 38, pp. 1-25.
Yokoyarna, Y. and Hall, H. K . . Ring-Opening Polymerization of Atom-Bridged and Bond-Bridged
Bicyclic Ethers, Acetals and Orthoesters. Vol. 42, pp. 107--138.
Yorkgitis, E. M., Eiss, N. S. Jr., Tran, C. Wilkes, G. L. and McGrath, J. E.: Siloxane-Modified Epoxy
Resins. Vol. 72, pp. 79-110.
Yoshida, H. and Hayashi, K. : Initiation Process of Radiation-induced Ionic Polymerization as
Studied by Electron Spin Resonance. Vol. 6, pp. 401-420.
Youn 9, R. N., Quirk, R. P. and Fetters, L. J. : Anionic Polymerizations of Non-Polar Monomers
Involving Lithium. Vol. 56, pp. 1-90.
Yuki, H. and Hatada, K. : Stereospecific Polymerization of Alpha-Substituted Acrylic Acid Esters.
Vol. 31, pp. 1-45.
- - theory 13 D G E R - D D S 40
Cationic photoinitiators 63 f. Dialkylphenacylsulfonium salts 67
Chain growth 105 Diamine-diepoxide curing, statistical treatment
Chemical reactivity 9 3O
Cloud point 87, 99 ---diepoxide-monoepoxide systems 24
Clusters 45 Diamines 22
- - o f chemically different units 8 4,4'-Diamino-3,3'-dimethyldicyclohexylmethane
- - - - dissimilar units, size distribution 25
- - 10
Coagulation equations 21 Diaminodiphenylmethane (DDM) 28, 29
Condensation esterification 47 Diaminodiphenylsulfone (DDS) 29
Conversion at gelation 102 Dicyandiamide 55
- vitrification 102ft., 106, ! 11
- a t Dicarboxylic acids 22
Crack arrest 134, 146 Diepoxide-diamine systems 24, 38
- - blunting mechanisms 132 Differential equations 20
growth 134
- - - - scanning calorimetry (DSC) 7
initation 134
- - Diffusion 9
propagation 133
- - - - control 103
- - - - , unstable 145 - - - , glass transition 12
-
- - polymerization 104
Equation of motion for torsion pendulum 90 Living polymerization, initiated 19
Equilibrium rubbery moduli 122 Logarithmic decrement 90 f.
shear modulus 41
- -
- -