78 Advances in Polymer Science

Epoxy Resins and

Composites III
Editor: K. Du~ek

With Contributions by
M. T. Aronhime, K. Du~ek, J. K. Gillham,
E N. Kelley, J. D. LeMay, E Lohse,
H. Zweifel

With 77 Figures and 9 Tables

Springer-Verlag Berlin Heidelberg New York

L o n d o n Paris Tokyo
ISBN-3-540-15936-3 Springer-Verlag Berlin Heidelberg New York
ISBN-0-387-15936-3 Springer-Verlag New York Heidelberg Berlin

Library of Congress Cat,dog Card Number 61-642

This work is subject to copyright. All rights are reserved, whether the whole or part of
the material is concerned, specificallythose of translation, reprinting, re-use of illustrations,
broadcasting, reproduction by photocopying machine or similar means, and storage in
data banks. Under § 54 of the German Copyright Law where copies are made for other
than private use, a fee is payable to "Verwertungsgesellschaft Wort'. Munich.
© Springer-Verlag Berlin Heidelberg 1986
Printed in G D R
The use of registered names, trademarks, etc. in this publication does not imply, even
in the absence of a specific statement, that such names are ex¢mpt from the relevant
protective laws and regulations and therefore free for general use.
Typesetting and Offsetprinting: Th. Miintzer, GDR:
Bookbinding: Liideritz & Bauer, Berlin

2152/3020-543210
Editors

Prof. Henri Benoit, CNRS, Centre de Recherches sur les Macromolecules, 6,

rue Boussingault, 67083 Strasbourg Cedex, France
Prof. Hans-Joachim Cantow, Institut f/Jr Makromolekulare Chemic der Uni-
versit~it, Stefan-Meier-Str. 31, 7800 Freiburg i. Br., FRG
Prof. Gino DalrAsta, Via Pusiano 30, 20137 Milano, Italy
Prof. Karel Dugek, Institute of Macromolecular Chemistry, Czechoslovak
Academy of Sciences, 16206 Prague 616, (~SSR
Prof. John D. Ferry, Department of Chemistry, The University of Wisconsin,
Madison, Wisconsin 53706, U.S.A.
Prof. Hiroshi Fujita, Department of Macromolecular Science, Osaka Univer-
sity, Toyonaka, Osaka, Japan
Prof. Manfred Gordon, Department of Pure Mathematics and Mathematical
Statistics, University of Cambridge CB2 ISB, England
Prof. Gisela Henrici-Oliv~, Chemical Department, University of California,
San Diego, La Jolla, CA 92037, U.S.A.
Prof. Dr. habil. G/inter Heublein, Sektion Chemic, Friedrich-Schiller-Universi-
t~it, HumboldtstraBe 10, 69 Jena, DDR
Prof. Dr. Hartwig H6cker, Deutsches Wollforschungs-Institut e.V. an der
Technischen Hochschule Aachen Veltmanplatz 8, D-5100 Aachen
Prof. Hans-Henning Kausch, Laboratoire de Polym6res, Ecole Polytechnique
F6d6rale de Lausanne, 32, ch. de Bellerive, 1007 Lausanne, Switzerland
Prof. Joseph P. Kennedy, Institute of Polymer Science, The University of
Akron, Akron, Ohio 44325, U.S.A.
Prof. Anthony Ledwith, Department of Inorganic, Physical and Industrial
Chemistry, University of Liverpool, Liverpool L69 3BX, England
Prof. Seizo Okamura, No. 24, Minamigoshi-Machi Okazaki, Sakyo-Ku.
Kyoto 606, Japan
Professor Salvador Oliv6, Chemical Department, University of California,
San Diego, La Jolla, CA 92037, U.S.A.
Prof. Charles G. Overberger, Department of Chemistry. The University of
Michigan, Ann Arbor, Michigan 48 104, U.S.A.
Prof. Helmut Ringsdorf, Institut f~irOrganische Chemic, Johannes-Gutenberg-
Universit~it,J.-J.-Becher Weg 18-20, 6500 Mainz, FRG
Prof. Takeo Saegusa, Department of Synthetic Chemistry, Faculty of
Engineering, Kyoto University, Kyoto, Japan
Prof. John L. Schrag, University of Wisconsin, Department of Chemistry,
1101 University Avenue, Madison, Wisconsin 53706, U.S.A.
Prof. Giinter Victor Schulz, Institut f~ir Physikalische Chemie der Universit~it,
6500 Mainz, FRG
Prof. William P. Slichter, Chemical Physics Research Department, Bell Tele-
phone Laboratories, Murray Hill, New Jersey 07971, U.S.A.
Prof. John K. Stille, Department of Chemistry. Colorado State University, Fort
Collins, Colorado 80523, U.S.A.
Editorial

With the publication of Vol. 51 the editors and the publisher

would like to take this opportunity to thank authors and readers
for their collaboration and their efforts to meet the scientific
requirements of this series. We appreciate the concern of our
authors for the progress of "Advances in Polymer Science" and
we also welcome the advice and critical comments of our readers.
With the publication of Vol. 51 we would also like to refer to a
editorial policy: this series publishes invited, critical review articles
of new developments in all areas of polymer science in English
(authors may naturally also include workes of their own). The
responsible editor, that means the editor who has invited the
author, discusses the scope of the review with the author on the
basis of a tentative outline which the author is asked to provide.
The author and editor are responsible for the scientific quality of
the contribution.
Manuscripts must be submitted in content, language and form
satisfactory to Springer-Verlag. Figures and formulas should be
reproducible. To meet the convenience of our readers, the
publisher will include a "volume index" which characterizes the
content of the volume.
The editors and the publisher will make all efforts to publish
the manuscripts as rapidly as possible. Contributions from diverse
areas of polymer science must occasionally be united in one volume.
In such cases a "volume index" cannot meet all expectations, but
will nevertheless provide more information than a mere volume
number.
Starting with Vol. 51, each volume will contain a subject index.

Editors Publisher
Preface

This volume 80 of ADVANCES IN POLYMER SCIENCE

contains the fourth part of a series of critical reviews on selected
topics concerning epoxy resins and composites. The last decade
has been marked by an intense development of applications of
epoxy resins in traditional and newly developing areas such as
coatings, adhesives, civil engineering or electronics and high-
performance composites. The growing interest in applications
and requirements of high quality and performance has provoked
a new wave in fundamental research in the area of resin synthesis,
curing systems, properties of cured products and methods of
their characterization.
The collection of reviews to be published in ADVANCES IN
POLYMER SCIENCE is devoted just to these fundamental
problems. The epoxy resin-curing agent formulations are typical
thermosetting systems of a rather high degree of complexity.
Therefore, some of the formation-structure-properties relationships
are still of empirical or semiempirical nature. The main objective
of this series of articles is to demonstrate the progress in research
towards the understanding of these relationships in terms of
current theories of macromolecular systems.
Because of the complexity of the problems discussed, the
theoretical approaches and interpretation of results presented by
various authors and schools may be somewhat different. It may
be hoped, however, that a confrontation of ideas may positively
contribute to the knowledge about this important class of poly-
meric materials.

In view of the wide range of this volume, it was not possible to

publish all contributions in successive volumes of ADVANCES
IN POLYMER SCIENCE. Part I of the articles is published in
Vol. 72; Part II appeared in Vol. 75 and Part III in Vol. 78.

The reader may appreciate receiving a list of all contributions

to EPOXY RESINS AND COMPOSITES I-IV appearing in
ADVANCES IN POLYMER SCIENCE:

M. T. Aronhime and J. K. Gillham:

The Time-Temperature-Transformation (TTT) Cure Diagram of
Thermosetting Polymeric Systems.
A. Apicella and L. Nicolais (University of Naples, Naples, Italy)
Effect of Water on the Properties of Epoxy Matrix and Composites
(Part I, Vol, 72).
J. M. Barton (Royal Aircraft Establishment, Farnborough, UK):
The Application of Differential Scanning Calorimetry (DSC) to
the Study of Epoxy Resins Curing Reactions (Part I, Vol. 72).
L. T. Drzal (Michigan State University, East Lansing, MI, USA)
The Interphase in Epoxy Composites (Part II, Vol. 75).
K. Dugek (Institute of Macromolecular Chemistry, Czechoslovak
Academy of Sciences, Prague, Czechoslovakia).
Network Formation in Curing of Epoxy Resins (Part III, Vol. 78).
T. Kamon and H. Furukawa (The Kyoto Municipal Research
Institute of Industry, Kyoto, Japan).
Curing Mechanism and Mechanical Properties of Cured Epoxy
Resins (Part IV, Vol. 80).
J. L. Kardos and M. P. Dudukovie (Washington University,
St. Louis. MO, USA).
Void Growth and Transport During Processing of Thermosetting
Matrix Composites (Part IV, Vol. 80).
A. J. Kinloch (Imperial College, London, UK).
Mechanics and Mechanisms of Fracture of Thermosetting Epoxy
Polymers (Part I, Vol. 72).
E. S. W. Kong (Hewlett-Packard Laboratories, Palo Alto, CA,
USA).
Physical Aging in Epoxy Matrices and Composites (Part IV,
Vol. 80).
J. D. LeMay and F. N. Kelley (University of Akron, Akron, OH,
USA).
Structure and Ultimate Properties of Epoxy Resins (Part III,
Vol. 78).
F. Lohse, and H. Zweifel (Ciba-Geigy, Basle, Switzerland).
Photocrosstinking of Epoxy Resins (Part III, Vol. 78).
E. Mertzel and J. L. Koenig (Case Western Reserve University,
Cleveland, OH, USA).
Application of FT-IR and NMR to Epoxy Resins (Part II, Vol. 75).
R. J. Morgan (Lawrence Livermore National Laboratory, Liver-
more, CA, USA).
Structure-Properties Relations of Epoxies Used as Composite
Matrices (Part I, Vol. 72).
E. F. Oleinik (Institute of Chemical Physics, Academy of Sciences
of USSR, Moscow, USSR).
Structure and Properties of Epoxy-Aromatic Amine Networks in
the Glassy State (Part IV, Vol. 80).
B. A. Rozenberg (Institute of Chemical Physics, Academy of
Sciences of USSR, Moscow, USSR).
Kinetics, Thermodynamics and Mechanism of Reactions of Epoxy
Oligomers with Amines (Part II, Vol. 75).
S. D. Senturia and N. F. Sheppard (Massachusetts Institute of
Technology, Cambridge, MA, USA).
Dielectric Analysis of Epoxy Cure (Part IV, Vol. 80).
R. G. Schmidt and J. P. Bell (Universi'ty of Connecticut, Storrs,
CT, USA).
Epoxy Adhesion to Metals (Part I1. Vol. 75).
E. M. Yorkgitis, N. S. Eiss, Jr., C. Tran, G. L. Wilkes and J. E.
Me Grath (Virginia Polytechnic Institute, Blacksburg, VA, USA).
Siloxane Modified Epoxy Resins (Part I, Vol. 72).

The editor wishes to express his gratitude to all contributors for

their cooperation.

Prague, January 1986 Karel Du~ek

Editor
Table of Contents

Network Formation in Curing of Epoxy Resins

K. Du~ek . . . . . . . . . . . . . . . . . . . . . .

Photocrosslinking of Epoxy Resins

F. Lohse, H. Zweifel . . . . . . . . . . . . . . . . . 61

Time-Temperature Transformation ( T T r )
Cure Diagram of Thermosetting Polymeric Systems
M. T. Aronhime, J. K. Gillham . . . . . . . . . . . . . 83

Structure and Ultimate Properties of Epoxy Resins

J. D. LeMay, F. N. Kelley . . . . . . . . . . . . . . . 115

Author Index Volumes 1-78 . . . . . . . . . . . . . . 149

Subject Index . . . . . . . . . . . . . . . . . . . . 161

Network Formation in Curing of Epoxy Resins

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.

List of Symbols and Abbreviations . . . . . . . . . . . . . . . . . . . . 3

1 Introduction . . . . . . . . . . -. . . . . . . . . . . . . . . . . . 5

2 0n)Homogeneity of Cured Epoxy Resins and Control of the Curing Reaction . . 6

2.1 H o m o g e n e i t y or In_homogeneity o f C u r e d Epoxies . . . . . . . . . . 6
2.2 C o n t r o l o f Curing by Chemical Reactivity or Diffusion . . . . . . . . 9

3 The Branching Theories . . . . . . . . . . . . . . . . . . . . . . . 12

3.1 Classification o f Branching Theories . . . . . . . . . . . . . . . . 12
3.2 Statistical Methods . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 The Kinetic M e t h o d . . . . •. . . . . . . . . . . . . . . . . . . 18
3.4 C o m p a r i s o n o f the Statistical and Kinetic Theories - -
Their C o m b i n a t i o n . . . . . . . . . . . . . . . . . . . . . . . 21
3.5 Statistical and Kinetic Theories and A p p r o x i m a t i o n o f Cyclization. 22
3.6 Simulation o f N e t w o r k Build-up in n-Dimensional Space . . . . . . . 23

4 Application of Branching Theories to Curing Reactions Theory

and Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4,1 Review o f Studies . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Addition o f Polyepoxides and Polyamines . . . . . . . . . . . . . 25
4.2.1 Reaction Mechanism and Kinetics . . . . . . . . . . . . . . 25

AdvancesinPolymerScience78
(C)Springer-VedagBerlinHeidelberg1986
2 K. Dugek

4.2.2 Reactivity and Substitution Effect . . . . . . . . . . . . . . 26

4.2.3 Statistical Treatment of Diamine-diepoxide Curing . . . . . . . 30
4,2.4 Extension to Multicomponent Polyepoxy-polyamine Systems . . . 36
4.2.5 Comparison with Experiments . . . . . . . . . . . . . . . . 37
4.3 Treatment of Polyetherification . . . . . . . . . . . . . . . . . . 43
4.4 Acid Curing . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4.1 Curing with Polycarboxylic Acids . . . . . . . . . . . . . . 47
4.4.2 Curing with Cyclic Anhydrides . . . . . . . . . . . . . . . . 52
4.5 Possible Application to Other Curing Systems . . . . . . . . . . . . 54

5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Network Formation in Curing of Epoxy Resins 3

List of Symbols and Abbreviations

a
p fraction of primary amine units
a fraction of secondary amine units
at fraction of tertiary amine units
ai
fraction of diamine units with i reacted hydrogens (Fig. 7)
Cx concentration of molecules composed of x monomer units
Cx, 1 concentration of molecules composed of x monomer units and bearing 1
unreacted functional groups
Cg constant in the Williams-Landell-Ferry equation
% concentration of unreacted functional groups in the gel
f,f,, functionality of the monomer (X)
number-average functionality of an active branch point, Eq. (73)
fx(Z) probability generating function
g(z) generating function for the number fraction distribution
i concentration of the initiator, io initial value
k, k i rate constants
k1 rate constant for the reaction of a primary amine group with epoxide
group
ks rate constant for the reaction of a secondary amine group with an epoxide
group
kI rate constant of initiation
kp rate constant o f propagation
k? = kl/2
1 number of unreacted functional groups in a molecule
m x weight (mass) fraction of monomer X
m molar concentration of the monomer
nx molar (number) fraction of component X
Pi probability of finding a unit with i reacted functional groups
rA initial molar ratio of amine hydrogens to epoxy groups = 2[A]o/[E] o
t time
ti coefficient of the pgf T(z)
U, U i variable in the cascade substitution, Eqs. (52-53)
V, Vi extinction probability
V2 volume fraction of the monomer (polymer) in mixture with a diluent
Wi weight fraction of i-mer
Wg weight fraction of gel
Ws weight fraction of sol

X nmnber of monomer units in a molecule

Z~ Zx, Zxy auxiliary variable in a probability generating function
A front factor in the rubber elasticity theory
Am front factor for a phantom network
[A] concentration of amine groups
[A]o initial concentration of amine groups
C 1, C~ Mooney-Rivlin constants
E activation energy
4 K. Dugek

[E0] concentration of unreacted epoxy groups

[E]o initial concentration of epoxy groups
F(z)
Fx(z) probability generating function for the number of bonds issuing from a
Fxy(Z) unit
F(N) value of F(z) for z = N
F'(N) value of OF(z)/~z for z = N
F"(N) value of~2F(z)/~z2 f/Jr z = N
F, x value of OFox(Z)/Szv for z = 1
Fv value of ~Fx(Z)/~z v for z = 1
X
Fx Z value of OFxy(Z)/~Zyxfor z = 1
G~ equilibrium shear modulus; G~, c chemical, G .... t trapped entanglement
contributions
M molecular weight
M number-average molecular weight
Mw weight-average molecular weight
Mz z-average molecular weight
r~ number average molecular weight of monomers
Mx molecular weight of component X
N~ number of elastically active network chains (EANC) per monomer unit
P degree of polymerization
Pn number-average degree of polymerization
Pw weight-average degree of polymerization
R gas constant
RE = ([E]o -- 2[A]o)/[A]o
T temperature in K
Tg glass transition temperature
T go glass transition temperature of the system before curing
T~ glass transition temperature at full cure
T , Teg trapping factor in the trapped entanglement theory related to the whole
system and to the gel, respectively
T(z), Tx(z) probability generation function for the number of bonds with infinite
continuation issuing from a unit
W(z) weight-fraction generating function
conversion of functional groups
% conversion of functionalities (functional groups) of type X
0~I ~ 5 2 probabilities defined by Eq. (103)
(~ETH etherification conversion of excess epoxy groups, Eq. (81)
dilation factor in the rubber elasticity theory
8k~ Kronecker delta equal to 1 for ij = kl and zero otherwise
ij
E proportionality constant in G e, ent
~;1 ~ f;2 probabilities, Eq. (104)
× = k,/kp
Ye~ Yeg
concentration of elastically active network chains in the whole system
and in the gel, respectively
= k 2 / k 1, ration of rate constant for the reaction of secondary and
primary amine group with epoxide group
Network Formation in Curing of Epoxy Resins 5

= 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

This contribution reviews the application of branching theories to curing of epoxy

resins and their experimental verification. Because the theory plays a crucial role, the
methods of the branching theory will be briefly explained. The application of the
theory requires the knowledge of the reaction mechanism and chemical kinetics as
input information and this information is obtained assuming a uniform distribution
of reacting groups throughout the volume or, in other words, the mass action law
employing the average concentrations of reactive groups is assumed to be valid.
Moreover, the application of the mass action law implies that the network build-up
is controlled by chemical kinetics and not by specific diffusion. Therefore, the
problems of homogeneity of the reacting epoxy systems and the possible diffusion
control of the reaction will be briefly examined first. The application of the branching
theories will be dealt with in more detail for amine-epoxy systems in order to explain
the derivation of the main relations, and also to some extent for curing with poly-
carboxylic acids and cyclic anhydrides. The theoretical treatment ofpolyetherification
(polymerization) of epoxy groups, which can accompany the important curing reac-
tions or is trigerred by special initiators, will only be outlined.
Two points should already be stressed in the Introduction: (1) The quantitative
description of the network build-up in curing of epoxies is only in its initial stage and
is limited to relatively simple system, Its role is, however, important for making
predictions useful for the choice of the curing system and processing conditions.
(2) The knowledge of the reaction mechanism and chemical kinetics is a necessary
condition for the application of any branching theory. In this respect, the reader
is referred to reviews in this volume by Barton, Fedtke, Lohse, Morgan, Rozenberg
and some others.

2 (In)Homogeneity of Cured Epoxy Resins and Control of the

Curing Reaction

2.1 Homogeneity or Inhomogeneity of Cured Epoxies

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

both crosslinked and uncrosslinked, such as polystyrene or poly(methyl methacrylate).

Also within the series of samples of Bisphenol A diglycidylether cured with diamines
and cyclic anhydrides, there was no correlation between the nodular structure and
degree of crosslinking. Some of the off-stochiometric epoxy-amine systems with
very low crosslinking density exhibited a more clearly expressed structure than some
highly crosslinked systems. No correlation between noduli size and crosslinking
density has been found. An analysis of structures seen by transmission electron
microscopy has revealed that these structures can be interprete ' as artefacts 7). There-
fore, electron microscopy does not offer any evidence of inhomogeneous cross-
linking. The anomalies in glass transition observed by differential scanning calori-
metry (DSC) and interpreted as due to the existence of regions of higher and lower
crosslinking densities 3) can be explained by what is called the physical ageing
of glasses.
The small-angle X-ray scattering studies 6) of a number of cured epoxy resins did
not detect any fluctuation in electron density characteristic for inhomogeneous
systems. The level of scattering at low angles was again comparable with that
obtained for non-crosslinked polymer glasses and, moreover, swelling of the cured
resin in a solvent of different electron density, performed with the aim to increase
the contrast between the possibly existing regions of higher and lower crosslinking
density, did not caused an increase but rather a decrease in scattering intensity
at Bragg angles corresponding to the size of the noduli observed by electron
microscopy. Thus, this result is contrary to what would be expected if inhomogeneous
crosslinking were real. However, an increased scattering was observed at Bragg
distance smaller than 1 nm and this scattering became stronger after swelling. It is
assumed that this scattering was due to local ordering on the scale of dimensions o f
monomer units.
Small-angle neutron scattering (SANS) of pre-gel branched products of diglycidyl
ether of Bisphenol A (DGEBA) polymers 8) correspond to the behaviour typical
of randomly branched polycondensates without any evidence of local inhomogeneities.
In another study by SANS applied to DGEBA cured with poly(oxypropylene)di-
and triamines (Jeffamines), some fluctuations of the order < 1 nm were again observed,
i.e., corresponding to the size o f monomer units of their dimers which were inter-
preted as correlations of packing of intranetwork structure. No evidence of regions
of higher and lower crosslinking density can be traced in Brillouin scattering
studies 10) of reaction products of a branched epoxy prepolymer crosslinked with
poly(oxypropylene) diamine.
Static light-scattering studies it), however, have revealed some inhomogeneities
20-70 nm in size existing already in the original epoxy resins derived from DGEBA.
Their size was higher in the higher molecular weight resin and exceeded the size
of noduli observed by electron microscopy. These aggregates dissociate on increasing
the temperature and are assumed to be due to intermolecular hydrogen bonding.
It is to be noted that commercial non-purified resins were used in these studies.
Another light scattering study concerned the changes occuring during the curing
process 12~
The possible inhomogeneous course of network formation should be reflected first
of all in the reaction kinetics itself or, what is more relevant here, in the distribution
of groups in various reaction states at conversions higher than zero. Specifically, for a
8 K. Dugek

diepoxide-diamine system without substitution effect between epoxy and amino

groups it means that the fractions of primary, secondary and tertiary amino groups
should be the same as in a model monoepoxide-monoamine system at the same con-
version. Such a behaviour was observed in reality 13~up to and slightly beyond the gel
point where the analysis of the reacting system was still feasible. Considering the
strictly alternating epoxy-amine reaction (unless polyetherification interferes if
epoxide groups are in excess), there is no other way for bond (crosslink) formation
than the reaction between the amino and epoxy group. If regions existed in the
stoichiometric system richer in one or the other component, the crosslinking density
of both regions would be lower than that expected for a stoichiometric system because
both regions would be off-stoichiometric. It should be added that in hundreds of
papers devoted to kinetic studies of curing, the possible inhomogeneity of the reacting
system was not taken into account. One should, however, consider with care the high
conversion stages of the stoichiometric systems where the reaction kinetics may
deviate from that of a model system of low functionality due to steric hindrances and
other factor discussed in the chapter by B. A. Rozenberg 14~
Another proof against inhomogeneous cure in simple epoxy-amine and other
systems has been supplied by gel point measurements. The critical conversion at the
gel point (cf. Sect. 4) is a sensitive function of any inhomogeneity. For epoxy-amine
systems, the gel point conversion has been found to agree well with the prediction
of the theory assuming uniform distribution of reactive groups throughout the
volume 15 18~. The deviation does not exceed 1%. In contrast, for free-radical
copolymerization of bis-unsaturated monomers the observed gel point conversions
are higher than the calculated ones by a factor of up to 102-103. The high propagation
rate and extensive cyclization determine the inhomogeneous character of network
formation which proceeds via microgel-like particles ~9~.
It has been suggested that the autocatalytic character of the uncatalyzed epoxy-
amine reaction can be a reason for the formation of regions differing in conversion ~4~
It could be so, if the mobility of the reacting groups were low compared with the
reaction rate, i.e. specific diffusion control (see below) were operative. However, in
contrast to chain reactions the step addition is slow and diffusion control is not
operative unless stages of high crosslinking densities are reached. Autoacceleration
is operative only in the beginning of the reaction when the segmental mobility is
sufficiently high and the reaction rate is controlled by chemical kinetics.
One can conclude that there is no experimental or theoretical reason to expect a
general tendency" to inhomogeneous course of curing of epoxy resins and formation of
inhomogeneously crosslinked products. This conclusion has been obtained mainly
by analysis of simple epoxy-amine systems in which crosslinking occurs by a single
alternating reaction. This conclusion is not necessarily valid for other systems. The
following factors can assist inhomogeneous cure:
(1) Thermodynamic instability or segregation developed during cure in systems
containing partly compatible components,
(2) simultaneously or consecutively occuring reactions producing sequences or
clusters of chemically different units, which by itself can be a reason for inhomo-
geneous crosslinking or can additionally induce physical segregation,
(3) extensive cyclization, which is not typical for curing involving alternate reactions
and stiff monomers, but can become important in the homopolymerization of epoxy
Network Formation in Curing of Epoxy Resins 9

groups (polyetherification), provided the connection between epoxy groups in poly-

epoxide is not too stiff. For example, a number of cyclic structure was found to
result from the homopolymerization 20) and this kind of cyclization was assumed to be
the reason for the formation of inhomogeneities zl).
Therefore, it is worthwhile to examine the homogeneity of any new epoxy system
using physical methods and methods of chemical kinetics.

2.2 Control of Curing by Chemical Reactivity or Diffusion

Control of the curing rate by chemical reactivity of functional groups is another

assumption required by chemical kinetics based on the mass action law and by the
corresponding branching theory. In this case, the reaction rate is controlled neither
by diffusion of reacting species (specific diffusion control) nor by segmental mobility
which sets in when the system enters the glass transition region (overall diffusion
c o n t r o l ) 22).
It can be shown that the specific diffusion control is not operative for the con-
ventional mechanisms of curing of epoxy resins. The situation becomes somewhat
complicated when high crosslinking densities are reached.
In terms of the activated complex theory of reaction rate, the reaction between
groups A and B can be written down as

kl
~ A + B ~ -'--~ ~(AB) *--~ k ~ "~'AB~ (1)
k- 1

where ~(AB)*~- means the activated complex ~ A B ~ the product -- the AB

bond.
If ki, k_ 1 >> k2, the overall reaction rate is controlled by the equilibrium concen-
tration of the activated complex and is independent of the diffusion constants of the
species carrying the A and B groups. This behaviour is typical of relatively slow
step reactions including addition reactions of epoxy groups and the rate has been
found to be independent of the molecular weight (Flory principle). Specific diffusion
control has been found to be operative for fast reactions like fluorescence quenching
or recombination of macroradical in free-radical polymerization (the Trommsdorff
or gel effect), when k 2 becomes comparable to or greater than k 1 and k_ 1. Also,
the specific diffusion control is expected to be observable rather in systems dilute
in polymer, where the molecules participate in the reaction as coils with little
overlap, in contrast to bulk systems, where the interpenetration of all species is
almost perfect. A kind of diffusion control is likely in the densely crosslinked net-
works even above their T , but it is questionable to what extent it can be specific.
There are no experimental indications of such a specificity. The impossibility for a
few remaining groups to find their reaction partners in highly crosslinked networks
due to the stiffness of the network structure (topological limit of the reaction 14,23- 26))
can be regarded as manifestation of such diffusion control. Therefore, the lowering
of the curing rate due to steric hindrances and slow segmental diffusion may be rather
regarded as a kind of overall diffusion control although it is not associated with
glass transition.
10 K. Du~ek

2.2.2 Overall diffusion control due to glass transition

Curing o f epoxy resins is a typical example where overall diffusion control can
become operative. During curing, the glass transition temperature of the system
increases and may reach or exceed the reaction temperature. This phenomenon is
dealt with in several reviews of this volume, particularly in those by J. K. Gillham
and E. F. Oleinik.
While well above Tg, the dependence o f the rate constant k on temperature
is governed by the Arrhenius proportionality

--In k o~E/RT (2)

(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

Fig. 1. Dependence of the glass transition

temperature Tg and of the difference 0
50 between the reaction temperature and Tg,
0 = T -- Tg, on conversion of epoxy groups
in the system DGEBA-4,4'-diamino-3,3'-
dimethyldicyclohexylmethane (rA = 2) 27)
Reaction temperature T (°C): 1 (I) 100,
0
2 • 64, 3 0 40 (raised to 64)
G2 0.4 C6 C8
Network Formation in Curing of Epoxy Resins 11

systems the reaction is initially controlled by chemical reactivity of groups but,

because Tg increases with conversion, the reaction rate becomes eventually controlled
by overall diffusion.
Figures 1 and 2 show an increase in Tg during cure which brings about a decrease
in 0 = T - - T in the isothermal r6gime. Assuming the Arrhenius dependence o f the
rate constants on T, one can get a good superposition of the kinetic ctirves in the
region well above T 8 and a considerable retardation o f the reaction at reaction
temperatures near or below T (Fig. 2).

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.

3 The Branching Theories

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)

3.1 Classification of Branching Theories

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

equations of chemical kinetics. It is to be remarked that these kinetic equations are a

degenerated case of Smoluchowski coagulation equations.
The fact that various details of the structure can be described is the advantage of
the statistical methods. This is particularly important for the gel. Such insight into
the molecules and gel is not possible using the kinetic method in which each finite
molecule and also the gel appear as species differing in the number and type of un-
reacted groups regardless their internal structure. The disregard of the spatial corre-
lations (cyclization) can be serious in some cases. There exist approximate procedures
in which the trees are embedded in a n-dimensional (n is usually 3) space, the number
of ring-forming bonds is calculated but the formalism of the generation of trees is
retained.

3.2 Statistical Methods

The statistical methods are based on the Flory-Stockmayer model of branching trees
with uncorrelated circuit closing in the gel. This model was generalized by Gordon
and coworkers and resulted in what is now called the theory of branching processes,
or cascade theory because it uses cascade substitution for generation of trees. The
generalization has included treatment of multicomponent systems, unequal reactivity,
substitution effect and approximate treatment ofcyclization 31-357 The use of mathe-
matically simple probability generating functions makes the procedures for derivation
of various statistical averages highly economical and routine. O f the other variants

\ \\
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

W(z) = Y. w l z i = Fo(zF(zF(zF ...))) (6)

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

W(z) = zFo(u) (7)

and a recurrent equation

u = zF(u) (8)

The weight-average degree of polymerization, Pw, is obtained by differentiation of

W(z)
Pw = ~ iwl = W'(I) (9)
i

From Eqs. (7) and (8), one obtains

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

F'(1) = ~, i ( i - 1)Pl/)-'. ipl = 1 (tl)

i i

Equation (10) shows that P~ diverges at the gel point.

Beyond the gel point, some units belong to finite molecules in the sol and some
are a part o f the (infinite) gel. That means that units in the sol must issue none and
units in the gel at least one bond that has (via a sequence of bonds) a continuation
to infinity. The key quantity which determines the probability that a bond has no
continuation to infinity is called extinction probability v. The extinction probability
is defined as the root of the equation

v = F(v) (12)

in the interval (1, 0).

V Fig. 4. Schematic explanation of the calculation of extinction

probability v

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

w~ = Fo(v) = ~ plY i (13)

i

The number o f elastically active chains, N e, determining the equilibrium rubber

elasticity, is derived from the following consideration. A chain in the gel is elastically
active, if the branch points at each of its ends issue at least three paths to infinity. Such
elastically active network chain (EANC) can have many long side branches but none
of them may have an infinite continuation. The number of EANC's, N e, is thus
calculated from the number of E A N C ends, i.e., branch points issuing three or more
bonds with infinite continuation. The distribution of units according to the number
of bonds with infinite continuation is described by a pgf T(z)

T(z) = F0(v + (1 - v) z) = y, pi[v + (1 - v) Z] i = E tizi (14)

i i

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

because a bond with infinite continuation contributes by 1/2 to the number of

EANC's. The sum in Eq. (15) can conveniently be evaluated using the values of
derivatives of T(z). Thus, the first derivative, T'(1), is equal to ~] iq, while its
i=1

value for z = 0, T'(0) = t~ and the second derivative for z = 0, T"(0) = 2t2, so that

N e = (1/2) ~ ' ( 1 ) - - T'(0) - - T"(0)] (16)

which holds for any form o f F0(z).

I f the reactive groups are of equal and independent reactivity, the probability of
finding a reacted group is just equal to the molar conversion of these groups ~. Then,
the pgf Fo(z) for a f-functional m o n o m e r is just equal to the convolution Fo(z)
= (1 - - ~ + ~z) f and the coefficient p, in Eq. (14) is equal to the coefficient o f binomial

expansion (fi) (1 - ~)f-' ~ ' •

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,

F3i(z) = N(OFoi(Z)/~zlj) (17)

where N is a normalizer to have Fji(1) = 1. This is a physically correct formulation

in contrast to formulation of a multicomponent Galton-Watson process defined
in 31) according to which we have only sets of z k and F i defined by

F~(z) = (~k ~Foi(Z)/~Zk) N (18)

However, for an alternating bicomponent system (like in diepoxide-diamine addi-

tion) both formulations give the same result. These problems are exposed in more
detail in Ref. 29)
In the cascade substitution, zij is substituted by Fij or zljFij and, for an inf'mite
continuation of the structure to be possible,

det 18i} - - Fill = 0 (19)

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

vii = Fij(v ) (20)

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.

3.3 The Kinetic Method

The network build-up is described by a (infinite) set of kinetic differential equations

for the concentration o f each i-mer. This approach has been developed mainly by
K u c h a n o v et al. (cf., e.g., Refs. ,1-,s)) and is demonstrated here on two examples:
(a) Random irreversible step polyaddition of a f-functional monomer: the molecules
are characterized by the number of units x and number of unreacted groups 1 (for
trees 1 = x ( f - 2) + 2). The formation of molecules composed of x units and
bearing 1 unreacted functional groups, Ax, ~, proceeds according to the scheme

Ak,j+ 1 + A x _ k , l _ j + l --4 Ax,~

Assuming bimolecular kinetics, the time (t) dependence of the concentration Cx,l
reads 44)
Network Formation in Curing of Epoxy Resins 19

dcx. 1
k dt lCx'l(I,~xlcx'l + CgWg)

+ ± ~ (j + 1)(1- j + 1) Ck,~+1 C~_k,l_j+1 (21)

j=0 k=l

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

Og/~x = cfzf(Sg/~zf) + (1/2) (Og/Ozf)2 (23)

where x = kt and % is the number of functional groups per monomer unit. The
solution of this partial differential equation yields

g(Zp, zf) = zp~f--(1/2) fZp~f-l[~--r(1 --~)] (24)

where ~ is given by

~--Zp~ f-1 = zf(1 -- ~)

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

and the following differential equations

di/dt = --k~mi
d q / d t = klmi - - kpmc~
(25)

dCk/dt = kpmck_l -- kpmck

The set (25) can be converted into a single differential equation (27) for the generating
function g(z)

g(z) = i~ ciz i (26)

dg(z)
- rag(z)(z -- 1) + ×mzi (27)
d~

In these equations, i, m, c i are molar concentrations of the initiator I, m o n o m e r M,

and polymer o f degree of polymerization i, respectively; ~ = kpt and

x = kJkp (28)

The solution of Eq. (27) yields

×Zio (i/io)-~z- 1)/~, _ (i/io) (29)

g(z)- z-- 1 + ×

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.

3.4 Comparison of the Statistical and Kinetic Theories - -

Their Combination

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.

3.5 Statistical and Kinetic Methods and Approximation of Cyclization

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.

3.6 Simulation of Network Build-up in n-Dimensional Space

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

4 Application of Branching Theories to Cuiing Reactions:

Theory and Experiment

4.1 Review of Studies

Although the major interest in experimental and theoretical studies of network

formation has been devoted to elastomer networks, the epoxy resins keep apparently
first place among typical thermosets. Almost exclusively, the statistical theory based
on the tree-like model has been used. The problem of curing was first attacked by
Japanese authors (Yamabe and Fukui, Kakurai and Noguchi, Tanaka and Kakiuchi)
who used the combinatorial approach of Flory and Stockmayer. Their work has
been reviewed in Chapter IV of May's and Tanaka's monograph 65). Their experi-
mental studies included molecular weights and gel points. However, their conclusions
were somewhat invalidated by the fact that the assumed reaction schemes were too
simplified or even incorrect. It is to be stressed, however, that Yamabe and Fukui 66)
were the first who took into account the initiated mechanism of polymerization of
epoxy groups (polyetherification). They used, however, the statistical treatment
which is incorrect as was shown in Section 3.3.
The importance of deviations from the structure of an ideal network due to
stoichiometric imbalance or incomplete reaction was recognized in amine-epoxy
curing by Bell 67,68) who developed semi-empirical corrections. Their applicability
was, however, limited to rather small deviations from the perfect state. The degree
of approximation has never been tested against the complete theory.
A complex approach to curing of epoxies was enabled by introducing the modern
branching theories, particularly the theory of branching processes (cascade theory).
A brief account of these studies includes the theoretical and experimental treatment
of diamine-diepoxide and diamine-diepoxide-monoepoxide systems (molecular
weights, critical conversions, sol fractions, concentration of elastically active network
chains, entanglement trapping factor and equilibrium elasticity in the rubbery state)
performed by the Prague group 13,15-18,2~,59,69,70) Also, the problem of acid curing
by polycarboxylic acids and cyclic anhydrides has been attacked by the same group 60,
71-74). Recently, the combination of the kinetic and statistical theory was suggested
for a treatment of polyetherification accompanying the amine-epoxy addition sl)
The work of Burchard's group in Freiburg has been concentrated on the pregel
and critical regions in the polyetherification of epoxides released by diphenols
(molecular weights, radii of gyration, diffusion coefficients obtained by static and
dynamic light scattering) 68-70~. These and new results have been reviewed in a special
chapter of this volume 71). The studies performed at the Institute of Chemical Physics
of the Soviet Academy of Sciences in Moscow and headed by Irzhak, Topolkaraev
and Rozenberg have been directed to aromatic diamine-diepoxide systems (molecular
weights, composition, sol fractions, topological limit of the reaction); statistical
(cascade) method or the computer simulation mentioned in Section 3.6 have been
used 14,23-25,57,64.,79- 81)
Using the Macosko-Miller version of the tree-like model, Charlesworth a2.a3)
analyzed his experimental results obtained for the diamine-diepoxide reaction
(molecular weights, pregel composition, sol fraction). Bokare and Ghandi 84~ derived
Network Formation in Curing of Epoxy Resins 25

relations to cover polyetherification of epoxy group as a consecutive reaction to

epoxy-amine addition.
A selection of structural parameters which can be calculated using the branching
theory, experimental methods of their determination and their connection with
properties of materials are given in Table 1.

Table 1. Calculated structural parameters

Structural parameter Experimental method Utilization

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

Concentration of elastically active :Equilibrium elasticity and Mechanical, optical

network chains (EANC) visco-elasticity and ultimate
Length and distribution of EANC properties
and dangling chains Dynamic LS Chemical and thermal
Entanglement trapping factor stability
Size distribution of clusters of Scattering methods
chemically dissimilar units Viscoelasticity Chemical and physical
ageing

4.2 Addition of Polyepoxides and Polyamines

4.2.1 Reaction mechanism and kinetics

The reaction mechanism and kinetics o f the addition o f a p r i m a r y or secondary
amino group on the epoxy group is analyzed in detail in the review by Rozenberg
in this volume 14) Here, the p r o b l e m is considered merely from the point o f view
o f the distribution o f building-blocks. The reaction can be represented by the scheme
26 K. Dugek

R1CH_CH2 + H2NR 2 k_~lR1CH_CHzNHR2

\/ I
O OH
E AH 2 HAE(OH)
(30)

R1CH_CH2NHR2 + R1CH_CH2 k__~2R I C H _ C H 2

\/ \/ I \
O O OH NR 2
/
HAE(OH) E RlffH • C H 2
|
OH
A(E(OH))2

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

the apparent rate constants can formally be expressed as

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)

4.2.2 Reactivity and substitution effect

The intrinsic or induced (substitution effect) difference in reactivity of epoxy groups
in polyepoxides and amino groups in polyamines can greatly affect the network for-
Network Formation in Curing of Epoxy Resins 27

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

CH2CHCH2\ ~ /'---k tCH2C.H.CH2

\o/ 'o
CH
\~
CHCH2/ ~ "~ \CH2CHCH2
\I
O TGDDM 0

Moreover, the position of the pair of glycidyl groups in D G A and T G D D M enhances

the probability of an internal etherification under formation of a morpholine
ring sa, as)

/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

Moreover, the analysis by LC and N M R is complicated by the existence of well

distinguishable stereoisomers in the amine-DGA adducts 7o). These complications
make the determination of the substitution effect difficult. At present, no studies are
available concerning the possible substitution effect in another polyfunctional
epoxide - - tris(hydroxyphenylmethane)s6), but the reactivity of epoxy groups can
be expected to be independent.
For polyamines, two substitution effects are possible. One concerns the amino
groups in diamines and the other the activity of hydrogens in primary amine
compared to that in the formed secondary amine (generally, k 1 # k 2 in Eqs. (30,
31)). In diamines like diaminodiphenylmethane (DDM), diaminodiphenylsulfone
(DDS) or hexamethylenediamine (HMD), the reactivity of amino groups is practic-
ally independent. It need not be so, however, in case of a short distance between
amino groups like in diethylenetriamine, or when the reactivity is affected by conju-
gation like in isomeric phenylenediamines, where the para isomer is expected to
exhibit the strongest substitution effect. Again, if the polyamines contain amino
groups of intrinsically different reactivity, network formation can be affected.
The substitution effect within the amino is caused by steric and electronic effects.
In the ideal case of equal reactivity of amine hydrogens, the ratio ~ = k2/k 1 = 1/2
because the primary amine has two hydrogens and the secondary amine only one.
Sometimes, the factor 2 appears in front of the 1.h.s. of Eq. (30) because primary
amine is bifunctional 15-18). The rate constant defined in this way k~ = kl/2 and for
the ideal case one has k2/k f = t.
In principle, three methods are available for determination of this ratio: (1) reaction
kinetics preferably on model monoamine-monoepoxide systems by monitoring the
time change in the concentration of epoxy or amino groups 14,s7,88), (2) chromato-
graphic determination of reactants and products of the reaction of a monoamine or
diamine with monoepoxide for excess amine over the stoichiometric ratio 13), (3) criti-
cal conversion at the gel point or preferably determination of the so-called critical
molar ratio necessary for gel formation at 100 9/0 reaction of epoxide 15-~8,59). The
theoretical dependence of the critical conversion ~c in a stoichiometric mixture of
diamine and diepoxide and of the critical molar ratio (rA)c is shown in Figs. 5 and 6.

1 I

0.6

i I

Fig. 5. Theoretical dependence of the critical

tl I conversion of epoxy groups or amine hydrogens
in the stoichiometric mixture of a diepoxide
and a diamine as a function of the ratio of rate
0.5 I constantsk2/kl = Q 15~
-2 -1 0
tog ( k 2 / k ~) ,"
Network Formation in Curing of Epoxy Resins 29

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

The determination of the substitution effect is of advantage because of a simplified

kinetic treatment based on the assumption of additivity of activation energies. In
this case, the rate constant of the reaction of components M and N, kMN, is pro-
portional to the product o f rate constants k M and k N (kMN~kMkN) for the reaction
M and N, respectively, with reference compounds. Thus, if the additivity of activation
energies is valid, one can get the ratios of rate constant required by the
branching theory just from once predetermined reactivities (k M, k N, etc.). If this
condition is not met, a kinetic study of the particular system is required (cf. Ref. sg)).

4.2.3 Statistical treatment of diamine-diepoxide curing

The application of the branching theory to amine curing is demonstrated in this
section using the example of a diamine and diepoxide with independent reactivities
of amino groups and epoxy, groups, respectively. This assumption is valid e.g. for
DGEBA and D D M . Cyclization can be neglected because the critical conversion
at the gel point was found to be independent of dilution up to 60% solvent 18,59)
Because of the independence of reactivity of groups in diamine and diepoxide and a
step polyaddition mechanism, it is legitimate to use the statistical approach. Only
one type of bond can be formed.
The building blocks are represented by diamine units with 0, 1, 2, 3 and 4 reacted
hydrogens (diamine is tetrafunctional) and by diepoxide units with 0, 1 and 2
reacted epoxy groups. Originally 15), the distribution of fractions of these units was
obtained by solving the complete set of differential equations (Eqs. (50-56) of
Ref. ~5)), but this is unnecessary because of independent reactivities. It is sufficient
to generate the distribution of monoamine and monoepoxide units, with respect to
the number of bonds they issue from reaction kinetics and to obtain the distribution
of diamine and diepoxide units, respectively, by convolution 13,17) (cf. Fig. 7).
According to Eqs. (30, 31), one can express the molar concentrations of primary,
secondary, and tertiary amino groups, c p, cas, and c~, respectively, by the following
kinetic equations

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

das/da p = --1 + QaJap (36)

dat/da p = - - a s a p

where ap, a s, and a t are molar fractions of the respective units.

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

O,n ureocted ~ 1,1 reocted functionotities

C~Econversion of epoxy groups

Fig. 7. Fraction of amine, diamine and diepoxide units with different number of reacted functionalities

The solution o f Eq. (36) yields

1
a - l ~ ( a ~ - - a-p ) _ ~ (37)

Q
at - 1 (ap - - a~/£) + 1 (38)
g

The dependence o f ap, a a n d a t o n conversion o f amine hydrogens, ~A, is obtained

by combination of Eqs. (37) and (38) with the balance equation

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

fA(ZE) = ap + asz E + a z~ (40)

fE(ZA) = 1 -- otE + 0~EZA (41)

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

FoE(ZA) = f~(ZA) = (1 -- ~E + 0~EZA)2 = E eiZA

i (43)
i=O

(cf. Fig, 7). The diepoxide (fractions %, e l, %) and diamine (fractions %, a 1, a 2, a 3, a 4)

units are placed in the root with frequency given by their molar fractions, n E and n A,
respectively, i.e. nee i and nAa v
In analogy with Eq. (5), the p g f s for the number of bonds issuing from units in
generation g > 0 are obtained by differentiation and renormalization

FA(ZE) = ('a + aszE + atz2) (a s + 2atz E) (44)

a s + 2a,

FE(ZA) = 1 ........0tE + ~EZA (45)

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,

F~A ~- 2 ( a + 2a,) = 4~ A (46)

F~E = 20tE (47)

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

nAMA + nEME nAMA + nEME (48)

M = 1 --- (nAFEA + nBFoAE)/2 = 1 - - 2nEg E
Network Formation in Curing of Epoxy Rcsins 33

because it holds that

4nA0~ A = 2 n ~ E o r ctE = rAOtA (49)

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

W(z) = mAWA(z) + mEWE(z) (54)

and

(0W(z) 0W(z) (55)

Mw = \ \ + ~Z--~/ZA=ZE=,

M w is thus obtained by differentiation o f Eqs. (50-54) and by substitution o f the

result into Eq. (55). The derivatives o f u A and u E are obtained from Eqs. (52, 53). The
result can also be expressed in a matrix form

(1 - v~ -V~ yl (egA F~E~

Mw = (MA, ME) E
-F A ,-v~; \F~A V~/ + ('0 ~)](m:)
(56)

where the - - 1 means the inversion of the matrix;

\ ~Zy /ZA=ZE= 1 \ ~Zy ./Zn=ZE= I

34 K, Du~ek

For FOA , FOE , FA, F E given by Eqs. (42-45)

F(~ A = 2(a s + 2at) = 4~ A F E = [(a s + 2at)2 + 2at]/(a ` + 2at)

FoAE = 20t E = 20t A q- at/0tA
FOAA= FEE = F ] = FE =0 =

so that

M w = mAM A + mEM E + 1/D[MA(mAFEA + mEF~F) -k

A E
+ ME(mAFEA + mEF~)EFA)] (57)

where D is the determinant

t l-F~ -F~ (58)

D = -F~, 1 - F~

In this case (cf. also Eq. (49)),

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

1/rA + rAat = 1 (61)

where at is an implicit function of rA. The determination of the critical value of

rA, (rA)c, is experimentally very simple (determination of solubility) and the value
of the ratio of rate constants 0 can be determined using Eq. (61) in combination
with Eqs. (37-39). The calculated dependence is given in Fig. 6.
(c) Extinction probabilities:
The extinction probabilities necessary for obtaining the structural characteristics
beyond the gel point are defined by Eq. (20). For the diamine-diepoxide system
we have
Network Formation in Curing of Epoxy Resins 35

v.A = FA(VE) = (av + asVE + a t e ) ( a s + 2a,v E) (62)

( a + 2a,)

VE = FE(VA) : 1 -- ot E --~ 0[EV A (63)

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

Ws = mA(a p + a~ve + atv~)2 + mE(1 - - 0~E + 0~EVA)2 (64)

(e) Concentration of elastically active network chains (EANC)

An elastically active network chain is active in the equilibrium elastic response
of the network to deformation. F r o m the topological point of view, an E A N C is a
chain between two active branch points. An active branch point is a unit from which
at least three paths issue to infinity. In the case under consideration, only some of
the chemically tetrafunctional diamine units can become active branch points. I f
the polyepoxide were more than bifunctional, it would also contribute to the number
o f E A N C ' s . In analogy with Eq. (14), the pgf for the number of bonds with infinite
continuation issuing from a diamine unit Ta(z ) 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

No = (1/2) nA(3t 3 + 4t4) (66)

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

N = (1/2) nA{6asat(l - - VE)3 + 4at2[4VE(1 - - VE)3 + (1 - - VE)4]} (67)

36 K. Dugek

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

f = (3t 3 + 4t4)/(t 3 + t4) (70)

Also the trapping factor, T , in Langley-Graessley's theory 95-97) of trapped

entanglement is accessible using the pgf T(z). Thus,

Te = [hA(t2 + t~ + t4) + ne~(1 - - VA)2]2 (71)

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.

4.2.4 Extension to Multicomponent Polyepoxy-polyamine Systems

The approach explained in the preceding section can be extended to systems con-
taining more than one polyepoxide or polyamine. When selecting the building blocks
for the network build-up, one should recall the conclusions of Section 3.4 and
allow for the possible stochastic correlations due to substitution effect. However,
even in case the reactivity of groups in polyamines and polyepoxides is independent,
the reactivity of groups in the epoxy and amine components can be intrinsically
different. If so, the distribution of amine units given by ap, a s and a t for each compo-
nent (i.e. aip, ais and ait ) and epoxy units is not sufficient. It is necessary to build up the
molecules from units differing in the type of neighbouring epoxy groups and this
distribution must be obtained from chemical kinetics. This problem has been exposed
in more detail for the diepoxide-monoepoxide-diamine system 17~: the blocks are
now Ap, As(E1), As(E2), At(E1Et), At(E1E 2) and At(E2E2) , where E 1 and E 2 are
epoxy groups of type 1 and 2 and the contents of bracket determines to which epoxy
units the amine units are bound. The fractions of the building blocks are determined
by the concentrations of the respective triads. Their concentration is obtained by
Network Formation in Curing of Epoxy Resins 37

solving the corresponding kinetic differential equations. If the reactivities o f E 1 and

Ez are equal, the probability that a bond extends from any amine unit to reacted
either E 1 or E 2 is just equal to their molar fractions.

4.2.5 Comparison with Experiments

A comparison of the predictions o f the statistical theory with experiments on
diepoxy-diamine systems can be found in papers by Du~ek et al. 13.15-18), Topoi-
karaev et al. 23,24), Bogdanova et al. 79) and Charlesworth 82,83)
Whether the distribution o f units in the system undergoing cure is the same as in a
model system yielding only low molecular products is one o f the most important
issues in application o f the theory. The results obtained on D G E B A and aliphatic
amines 13) and diglycidylether of resorcine ( D G E R ) 22) give a positive answer. Figure 8
shows the fraction of tertiary amino group determined in monoamine-monoepoxide,
monoamine<liepoxide and diamine-diepoxide systems as a function o f conversion.
Practically no difference can be detected. This means that the distribution o f groups
in the reacting system is only a function o f conversion and is not affected by the
macromolecular character o f the reaction in curing.
In the literature, there is a number o f data on the number-average molecular
weight 16.s2), and they roughly correspond to the predictions. However, the agreement
cannot be considered as a support for the validity o f the branching theory, since M ,

' " I I I

0.5 --

0.4

0.3
0 •
5 Q•
0.2

0 0
0.1

0.2 0.4 0.6

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

Fig. 9. Dependence of the critical conversion

of epoxy groups on the molar ratio of amine
to epoxy functionalities r A in the system
DGEBA-HMD. The curves were calculated
for the value of O indicated 16~
a5 t.0 1.5 29
rA

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

0.4 ~ . ~ 0.33 0 (30"2 S=0

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

Weakly crosslinked epoxy-amine networks above their Tg exhibit rubbery be-

haviour like vulcanized rubbers and the theory of rubber elasticity can be applied
to their mechanical behaviour. The equilibrium stress-strain data can be correlated
with the concentration of elastically active network chains (EANC) and other
statistical characteristics of the gel. This correlation is important not only for veri-
fication of the theory but also for application of crosslinked epoxies above their T v
Network Formation in Curing of Epoxy Resins 41

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)

where R is the gas constant, T is temperature, v, is the concentration of EANC's;

( ~ ) is the dilation factor equal to the ratio of mean square end-to-end distance
of network chains in the isotropic and reference states. A ~ is the front factor of
a phantom network equal to

Ap h _ f -- 2
f (73)

where f is the number-average functionality of active junctions given by Eq. (70).

For a real network, G~, ~ can be expressed by the equation

Ge, c = A ( ~ ) veRT (74)

42 K. Du~ek

in which A is dependent on deformation v , and possibly other network parameters.

The theory of Flory predicts, however, that in the limiting case of fully suppressed
fluctuation of crosslinks and small-strain modulus A can reach the value 1. On the
other hand, A ~ A ~ and G¢¢ ~ C1 of the Mooney-Rivlin equation for the defor-
mation ~ ~ oc. The other models (e.g. the tube model) are not bound by the limit
of A ~ 1.
The Langley-Graessley concept of trapped entanglements 95-97), which is very
simple, has proved itself applicable to a number of crosslinked rubbers. In this case,

G = (Ap, v + e T ) (~]) RT = Ge, c + G . . . . t (75)

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

v e ( ~ ) = ve = v~gw, and T(~) = T e = T gw, (76)

if the stress-strain measurement is performed on dry non-extracted sample and

v(C~oz) = VcgV~/aw2/ag and T(~) = T~gvzt/2wg2/3 (77)

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.

4.3 Treatment of Polyetherifieation

In spite of the great importance of polyaddition (polyetherificationl) of epoxy

groups in many formulations, a rigorous branching theory is still to be developed.
Polyetherification is often released by a primary reaction of epoxide with amines,
carboxyl groups, dicyandiamide, or is initiated by BF3-amine complexes, imidazoles,
etc. The polyetherification reaction is to be regarded as an initiated reaction and the
initiator is either added at the beginning of the reaction of formed in the first reaction

1 Polyetherification means formation of an ether bond or sequence of ether bonds by (poly)addition

of epoxy groups which is released by proton donors (like the OH group formed in the amine-epoxy
addition), ionic catalysts or other initiators
44 K. Du~k

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

g(ZE) = Z [A,EJ z~/N = • (atex) z~ (79)

x=2 x=2

By multiplying by z~ each o f the kinetic equations corresponding to the scheme (78)

for [AtEx] one gets

dg(ZE) -- g(ZE) (ZE - - 1) [Eo] (80)

2k E dt

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

[E]0 - - 2[A]0 - - [E0]

O~ETH :
(81)
[E]o -- 2[A]o

with the solution

g(ZE) = z~ exp [(zE -- 1) RE0~ETn] (82)

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

Foc(Z) = ZCAg(ZE) ; FoE(Z) = ZE (83)

Foc is the gf for clusters which always contain one amino group and this fact is ex-
pressed by the variable ZCA; the subscript means the direction o f bonds extending
from the amino group in a cluster to another amino group. FOE now concerns only
unreacted epoxy groups which issue only one e-functionality. The variable z E refers
to e - - e bonds but either an E-unit in the cluster or an unreacted epoxy group can be
the partner o f an E unit. The probabilities of finding such bonds are equal to QtE
and 1 - - ~E, respectively, and are related to the pgf variables Z¢c and ZeE

ZE = 0~EZeC -[- ( 1 --~)Z E (85)

Accordingly, we have three pgf's for the number of bonds issuing from units on
generation g > 0
46 K. Du~ek

Fec(Z) = ZAcg(ZE)(2 + ZERE0~ETH)/ZE (86)

2 + RE0~ETH

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

1 - - [(2 + RE~ETH) (3 + 2RE0~ETH) + RE~ETrl]/(RE 4- 2) = 0 (87a)

or

R E = 2(RE~ETH) 2 4- 8RE0~ETrt + 4 (87b)

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

4.4 Acid Curing

Curing of epoxy resins by polycarboxylic acids and cyclic anhydrides is also important
in applications, but it is much less understood due to more complex reaction mecha-
nism. Also, the statistical treatment is less developed and partly requires a revision.
In this section, the statistics of curing of epoxy resins with polycarboxylic acids
and cyclic anhydrides is discussed.
4.4.1 Curing with Polycarboxylic Acids
The reaction of an epoxy group with a carboxyl group can be described by the following
simplified scheme lo6)
RI CNH#H2 + R2COOH
0
Iaddiiion
es~erificotion

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

1 The polyetherification of epoxide groups catalyzed by tertiary amines is characterized by formation

of polymers of low degree of polymerization, a part of which is nitrogen-free (cf. Berger and Lohse,
Section 4.5).
48 K. Du~ek

catalyzed by tertiary amines using a number of analytical methods 106). Important

is the disproportionation reaction which can proceed even in a stoichiometric system
after all epoxy and carboxyl groups have been exhausted. Therefore, a poly(hydroxy-
ester) is not a chemically inactive substance.
While in a poly(hydroxyester) the diepoxide unit has functionality 2, after trans-
esterification it can be bound to neighbouring units by the number of bonds i
ranging from 0 to 4:

HOCtt2\ /CH2OH ~OCH2\ /CH20~

Ho/CHRCH\oH Ho/CHRCHNoH

i=0 i=2

HOCH2\ /CH20~" ~OCH2\ /CH2OH

Ho/CHRCHNoH ~o/CHRCHNoH

i=l i=2

~-OCH2\ /CH20~ ~OCHz\ /CH20~

Ho/CHRCH\o ~ ~o/CHRCH\o ~

i=3 i=4

Thus, transesterification causes branching as well as chain scission 73,74).

Let us consider the branching statistics for the case of a polycarboxylic acid and
polyepoxide with groups of independent reactivity. The use of the statistical method
is justified because neither the substitution effect nor the initiated chain growth
are operative. If only addition esterification occurs (and indeed this reaction can be
selectively accelerated by special catalysts 106)), the statistics is analogous to the
polyamine-polyepoxide addition and even simpler because the substitution effect
is here absent. The pgf's for the number of bonds issuing from polyacid (C) and
polyepoxide (E) units read

Foc(ZE) = (1 ~c + ~cZE)fc (87)

FoE(Zc) = (1 - - 0~E + 0~EZc) fE (88)

and for units in generation g > 0

.fc-1
Fc(ZE) = (1 - - ~ c + ~cZE) (89)
"fE - 1
FE(Zc) = (1 --~E + ~EZc) (90)

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

zc --+ (1 - - %) zc + %(1/2 + z~/2)

Thus, for a diepoxide and a diacid

FoE(Zc) = {1 - - % + %[(1 - - %) Zc + (%/2) (1 + ZCZ)]} (91)

Foc(ZE) = (1 - - % + %ZE)2 (92)

FE(Zc) = {1 - - 0tE + ~E[(1 - - % ) Zc +

+ (~r/2) (1 + g)l} (1 -- o~ r + ~rZc) (93)

Fc(zt) = 1 - - % + acZE (94)

The gel point conversion is determined by the relation

~F~ = %(% + %) = 1 (95)

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 )

VE __ VC--(1 --~C) (97)

~c
50 K. Dugek

The sol fraction w s is then given by

ws = mcFoc(VE) + mEFoE(Vc) = mc(1 - - ~c + 0~cVE)2 +

+ mE{1 - - OtE + ORE[(1 - - Otr) v c + (Z(T/2) (1 + v2)]} 2 (98)

where m c a n d m E are weight fractions o f diacid and d i e p o x i d e units, respectively.

The n u m b e r o f elastically active n e t w o r k chains E A N C , N e, is c o n t r i b u t e d only
by d i e p o x i d e units. A c c o r d i n g to the r e a s o n i n g given in Section 4.2.3, the distri-
b u t i o n o f d i e p o x i d e units with respect to the n u m b e r of b o n d s with infinite conti-
n u a t i o n is given by the p g f

TE(z) = FoE[V c + (1 -- Vc) z] = y ' tlz i (99)

i

and

N e = (1/2)nE(3t 3 + 4t4) (100)

where n E is the m o l a r fraction of diepoxide, Eq. (100) eventually yields

N = (1/2) nE~0tx(1 - - VC)3 [3 - - 2~X(1 - - VC)] (101)

F i g u r e 15 shows that the sol fraction is p r e d i c t e d to increase with increasing ~r but

to reach a steady value or even to pass t h r o u g h a m i n i m u m while N e continues to
increase. This b e h a v i o u r is similar to the effect o f b r a n c h i n g induced by degra-
d a t i o n 1o7); also here, every cut ( f o r m a t i o n o f diol unit) p r o d u c e s a new b r a n c h
p o i n t (diester unit).

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

The theoretical predictions were tested using the system DGEBA-1,7-heptane-

dicarboxylic (azelaic) acid a n d a tertiary amine as catalyst. In a stoichiometric system,
gelation was observed at etr = 0tc = 0.96-0.98 which, according to Eq. (95), would
c o r r e s p o n d to 0Vr = 0.08-0.04; this value is in agreement with the ratio o f rate
constants for addition esterification and transesterification o f the order o f 102-10 a
(Ref. 60). Figure 16 shows the time dependence o f the sol fraction and equilibrium
modulus o f these networks and it can be seen that the expected trend, particularly for
the sol fraction, is obeyed. Since the d a t a on ~r were not available, c o m p a r i s o n
could be made using the relation between Ne a n d wg. This plot is independent o f Oh-.
The agreement is satisfactory (Fig. 17).

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~

0.1 ........... 1 .....

0.05

0 ~ I ) I0

0.6 0.7 0.8

Wg

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

4.4.2 Curing with Cyclic Anhydrides

In spite of a great number of studies devoted to curing with cyclic anhydrides, there
existed a number of contraversial views concerning the reaction mechanism. Recent
studies have revealed 1o8-11o) that in the presence of tertiary amines the reaction
can proceed also in the absence of proton donors. The tertiary amine reacts first
with epoxide most probably via a zwitterion 11o). The simplified mechanism is as fol-
lows 1io):
@
Initiation : NR3
|
R I CHCH2 + NR3 ~ R ~ C H - - C Hi 2
V X/

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

Network formation in epoxide-cyclic anhydride curing was studied by Tanaka

et al. ,1-114) and Dugek et al. 72). Tanaka et al. considered the diepoxide-cyclic
anhydride curing as a random step polyaddition of a tetrafunctional (diepoxide) and
bifunctionat (cyclic anhydride) monomers 112). They did not take into account any
initiation mechanism. It has been shown elsewhere 72) that even if the reaction
were a non-initiated stepwise reaction, the expected critical conversion of epoxy
group 0.58 could not concern epoxy groups which are bifunctional. The correct critical
conversion of epoxy groups would be 0.82. Du~ek et al. 72) considered the initiated
mechanism and used the cascade method for derivation of the condition for gelation.
From Section 3.3 and 3.4 it follows, however, that this abproach is also not rigorous
and that one should rather use the kinetic method. The solution is unfortunately
not so simple as in the case of polyetherification (Sect. 4.3) and the differential
equation for the number fraction generating function has to be solved numerically.
Nevertheless, the statistical treatment correctly observes the main feature of
initiated network build-up - - the fact that the number of chain ends is equal twice
the number of reacted molecules of the initiator. The reacted initiator molecules
themselves represent one type of ends and the other type is represented by living
ends (anions). It is instructive to briefly outline the procedure 72), because it may be
found to be a good approximation.
If a chain has different ends and the same type of ends cannot be on both ends of
one chain (e.g. two initiator molecules or two living ends), this fact must be
respected in the cascade generation. Therefore, the bond directions are specified with
respect to the type of chain end (cf. also Ref. llS~). For example, for linear chains
formed from cyclic anhydride and monoepoxide (Fig. 18) one can formally place
the bound molecule of the initiator in the left branch and denote this direction
by 1 and the direction to the living end in the right branch by r.

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)

In case of cyclic anhydride-diepoxide curing, the building units are represented

by initiator (I), cyclic anhydride (A) and diepoxide (E) units. The pgf's for units
in the root then read

Fol(Z) = (1 - - ~l + ~IZ~E) (102)

FoA(Z) = 1 ---~A + ~AZIE((gl "~ (~2ZrE) (103)

FoE(Z) : [1 --0~E + ~E(PlZI! -k pAZIA)(gl + ;g2ZrA)]2 (104)

54 K. Dugek

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 - ~,.

Figure 19 shows that such a dependence was indeed found experimentally. In

comparing the theory with experiment, it has been assumed that k~ ~ k E and
k A ~ k E and in the calculations it has been taken into account that the hexahydro-
phthalic anhydride contained t.9 % acid which has been assumed also to act as an
initiator.
Much more experimental and theoretical studies are needed, however, before the
network build-up is well understood.

4.5 Possible Application to Other Curing Systems

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

followed by polymerization of epoxy groups; the reaction of 1-methylimidazole is

more complicated, however.
Recently, an important contribution to the curing mechanism involving dicyan-
diamide was presented ~21).The main product of the model reaction of dicyandiamide
with phenylglycidyl ether is 2-amino-oxazolidine formed from one molecule of
dicyandiamide and there molecules of phenylglycidyl ether

/ 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

is formed. The degree of polymerization of reaction products was found to be depend-

ent on conversion but, in the range of concentration of 2--4 % benzyldimethylamine,
not on the concentration of the ter.amine. The latter finding is in contradiction with
the pure initiation mechanism in which the initiating species remain chemically bound.

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

26. Pachomova, L. K., et al. : Vysokomol. Soedin. B20, 554 (1978)

27. Lufihk, S., Vladyka,K., Du~ek, K.: Polymer 19, 931 (1978)
28. Salamantina, O. B., et al. : Vysokomol. Soedin. A23, 2360 (1981)
29. Du~ek, K.: Brit. Polym. J. 17, 185 (1985)
30. Gordon, M.: Proc. Roy. Soc. London A268, 240 (1962)
31. Gordon, M , Malcolm, G. N.: Proc. Roy. Soc. London A295, 29 (1966)
32. Gordon, M., Ross-Murphy, S. B. : Pure Appl. Chem. 43, I (1975)
33. Dobson, G. R., Gordon, M.: J. Chem. Phys. 43, 705 (1975)
34. Dugek, K.: Makromol. Chem., Suppl. 2, 35 (1979)
35. Burchard, W.: Adv. Polym. Sci. 48, 1 (1982)
36. Macosko, C. W., Miller, D. R. : Macromolecules 9, 199 (1976)
37. Miller, D. R., Macosko, C. W. : Macromolecules 9, 206 (1976)
38. Durand, D., Bruneau, C.-M.: Macromolecules 12, 1216 (1979)
39. Durand, D., Bruneau, C.-M. : Polymer 24, 587, 592 (1983)
40. Du~ek, K. : Macromolecules 17, 716 (1984)
41. Kuchanov, S. I. : Methods of kinetic calculations in polymer chemistry (in Russian). Moscow,
Khimiya 1978
42. Kuchanov, S. I., Povolotskaya, E. S.: Vysokomol. Soedin. A24, 2179 (1982)
43. Kuchanov, S. I., Povolotskaya, E. S. : Vysokomol. Soedin. A24, 2190 (1982)
44. Du~ek, K. : Polym. Bull. 1, 523 (1979)
45. Dugek, K., ~omv~rsky, J. : Polym. Bull. 13, 313 (1985)
46. Mike~, J., Du~ek, K. : Macromotecules 15, 93 (1982)
47. Li6geois, J.-M. : private communication.
48. Du~ek, K. : in Physics of finely divided matter. Boccara, N. and Daoud, M. (eds.) Berlin, Heidel-
berg, New York, Tokyo: Springer 1985, p. 107
49. Kinetics of aggregation and gelation. Family, F., Landau, D. P. (eds.), Amsterdam: North-
Holland 1984.
50. Irzhak, V. I., Tai, M. L. : Dokl. Akad. Nauk SSSR 259, 856 (1981)
51. Dugek, K.: Polym. Bull. 13, 321 (1985)
52. Dugek, K.: Rubber Chem. Technol. 55, 1 (1982)
53. Gordon, M., Scantlebury, G. R. : J. Polym. Sci. C16, 3933 (t968)
54. Gordon, M., Scantlebury, G. R. : Trans. Faraday Soc. 60, 604 (1964)
55. Dugek, K., Gordon, M., Ross-Murphy, S. B. : Macromolecules 11,236 (1978)
56. Dugek, K., Vojta, V.: Brit. Polym. J. 9, I64 (1977)
57. Chepel, L. M. et al. : Vysokomol. Soedin. A24, 1646 (1982)
58. Morgan, R.: Adv. Polym. Sci. 72, 1 (1985)
59. Mat6jka, L., Du~ek, K., Dobbin, I.: Polym. Bull. 14, 309 (1985)
60. Mat6jka, L., Pokorn~,, S., Du~ek, K. : Makromol. Chem. 186, 2025 (1985)
61. Stauffer, D., Coniglio, A., Adam, M. : Adv. Polym. Sci. 44, 103 (1982)
62. Leung, Y.-K., Eichinger, B. E. : J. Chem. Phys. 80, 3877 (1984)
63. Leung, Y.-K., Eichinger, B. E.: J. Chem. Phys. 80, 3885 (t984)
64. Chepel, L. M. et al.: Vysokomol. Soedin. A26, 362 (1984)
65. May, C. A., Tanaka, Y.: Epoxy resins. Chemistry and technology. New York: M. Dekker
1973
66. Yamabe, T., Fukui, K. : Bull. Chem. Soc. Japan 42, 2112 (1969)
67. Bell, J. P.: J. Polym. Sci. A-2, 417 (1970)
68. Lin, C. J., Bell, J. P.: J. Appl. Polym. Sci. 16, 172I (1972)
69. Du~ek, K., Ilavsk~, M. : Colloid Polym. Sci. 258, 605 (1980)
70. Dosko6ilov~, D., et al.: Polym. Bull. 14, 123 (1985)
71. Dugek, K., et at.: Internat. Rubber Conf. Kiev, Proc. A1, 18 (t978)
72. Du~ek, K., Lufi~tk, S., Mat6jka, L. : Pol3;m. Bull. 7, 145 (1982)
73. Mat6jka, L , Du~ek, K. : Preprints Div. Potym. Mat. Sci. Eng. 49, 388 (1983)
74. Dugek, K., Mat~jka, L. : ACS Syrup. Rubber Toughened Thermosets, Adv. Chem. Ser. 208,
15 (1983)
75. Burchard, W., et al. : Pure Appl. Chem. 53, 1519 (1981)
76. Burchard, W., Bantle, S., Zahir, S. A. : Makromol. Chem. 182, 143 (t981)
77. Zahir, S. A , Bantle, S. : Preprints ACS Div. Org. Coatings Plastics Chem. 46, 651 (1982)
Network Formation in Curing of Epoxy Resins 59

78. Burchard, W. : Adv. Polym. Sci., will be published at a later date

79. Bogdanova, L. M., et ai. : Vysokomol. Soedin. A18. ! 100 (1976)
80. Irzhak, V. I., Rozenberg, B. A., Enikolopyan, N. S. : Network polymers (in Russian), Moscow:
Nauka 1979
81. Oleinik, E. F. :Adv. Polym. Sci., this volume
82. Charlesworth, J. M.: J. Polym. Sci., Polym. Phys. Ed. 17, 1577 (1979)
83. Charlesworth, J. M.: J. Polym. Sci., Polym. Phys. Ed. 17, 1571 (1979)
84. Bokare, V. M., Ghandi, K. S. : J. Polym. Sci., Polym. Chem. Ed. 18, 857 (1980)
85. Hagnauer, G. L., Pearce, P. J. : Preprints ACS Div. Org. Coatings Appl. Polym. Sci. 46, 580
(1982)
86. Hawthorne, K. L., Henson, F. C. : Preprints ACS Div. Org. Coatings Appl. Polym. Sci. 46,
493 (1982)
87. Charlesworth, J. M.: J. Polym. Sci., Polym. Chem. Ed. 18, 621 (1982)
88. Horie, K., et al.: J. Polym. Sci. A-i, 8, 1357 (1980)
89. Buckley, L , Roytance, D. : Polym. Eng. SCi. 22, 166 (1982)
90. Mones, E. T., Morgan, J. R. : Polym. Preprints 22(2), 248 (1981)
91. Gupta, A., et al.: J. Appi. Polym. Sci. 28, 1011 (1983)
92. Eichler, J., D o l ~ , I.: Collect. Czech. Chem. Commun. 38, 2602 (1973)
93. DobbS, I., Eichler, J., Klaban, J. : Collect. Czech. Chem. Commun. 40, 2989 (1975)
94. Chang, T. D., Carr, S. H., Brittain, J. O.: Polym. Eng. Sci. 22, 1213 (1982)
95. Langley, N. R. : Macromolecules I, 348 (1968)
96. Langley, N. R., Polmanteer, K. E, : J. Polym. Sci., Polym. Phys. Ed. 12, 1023 (1974)
97. Pearson, D. S., Graessley, W. W.: Macromolecules 13, t00t (1980)
98. Byrne, C. A., Schneider, N. S., Hagnauer, G. L.: Proc. IUPAC Macro 82, Amherst 1982,
p. 686
99. Whitting, D. A., Kline, D. E. : J. Appl. Polym. Sci. 18, 1043 (1974)
100. Erman, B., Wagner, W., Flory, P. J. : Macromolecules 13, 1554 (1980)
101. Mark, J. E.: Makromol. Chem., Suppl. 2, 180, 87 (1979)
102. Valles, E. M., Macosko, C. W. : Macromolecules 12, 673 (1979)
103. Morgan, R. J., Kong, F. M., Walkup, C. M. : Polymer 25, 375 (1984)
104. Le May, J. D., Swetlin, B. J., Kelley, F. N. : ACS Symp. Structure and fracture of highly cross-
linked networks. ACS Adv. Chem. Ser.
105. Le May, J. D., Kelley, F. N. : Adv. Polym. Sci., this volume.
t06. Mat6jka, L., Pokorn2~, S, Du~ek, K.: Polym. Bull. 7, 123 (1982)
107. Demjanenko, M., Du~ek, K. : Macromolecules 13, 571 (1980)
108. Antoon, M. K., Koenig, J. L. : J. Polym. Sci., Polym. Chem. Ed. 19, 549 (1981)
109. Lustoft, J. : Adv. Polym. Sci. 56, 91 (1984)
110. Mat~jka, e. al. : J. Polym. Sci., Polym. Chem. Ed. 21, 2873 (1983)
111. Tanaka, Y., Kakiuchi, H.: J. Appl. Polym. Sci. 7, 1951 (1963)
112. Tanaka, Y., Kakiuchi, H.: J. Polym. Sci., Pt. A, 3, 3279 (1965)
113. Tanaka, Y., Kakiuchi, H.: J. Macromol. Sci.-Chem. AI, 307 (1966)
114. Tanaka, Y., Huang, C. M.: Makromol. Chem. 120, 1 (1968)
115. Burchard, W., Ullisch, B., Wolf, Ch. : Faraday Disc. Chem. Soc. 57, 56 (1974)
116. Fischer, M., Lohse, F., Schmid, R. : Makromol. Chem. 181, 1251 (1980)
117. Lohse, F., Schmid, R. : Proc. 5th Internat. Conf. Org. Coatings Sci. Technol., Athens 1979, p. 31
118. Ricciardi, F., Romanchik, W. A., Joulli6, M. M.: J. Polym. Sci., Polym. Chem. Ed. 21, 1475
(1983)
t 19. Berger, J., Lohse, F. : J. Appl. Polym. Sci. 30, 531 (1985)
120. Berger, J., Lohse, F.: Polym. Bull. 12, 535 (1984)
121. Zahir, S. A. : Proc. 6th Internat. Conf. Org. Coatings Sci. Technol., Athens 1981, p. 83
122. Byrne, C. A., Hagnauer, G. L., Schneider, N. S. : Polym. Composites 4, 206 (1983)
123. Berger, J., Lohse, F.: Eur. Polym. J. 21, 435 (1985)
124. Bai, S. J.: Polymer 26, 1053 (1985)
125. Spathis, G., Kontou, E., Theocaris, P. S.: J. Polym. Sci., Polym. Chem. Ed. 23, 1439 (1985)

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

2 Network Structure, Classes of Initiators and Epoxy Resins . . . . . . . . 63

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

4 Hybrid Systems for Cationic and Radical Photopolymerization . . . . . . . 76

5 Epoxy Resins of Dual Functionality . . . . . . . . . . . . . . . . . . 77

6 Application Characteristics . . . . . . . . . . . . . . . . . . . . . . 77

7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Advances in Polymer Science 78

© S~rina¢r-Verla2 Berlin Heidelher~ 1986
62 F. Lohse and H. Zweifel

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
-

-- Low energy requirements

-- Low operating temperatures
- Ecological and related advantages [Many photopolymer systems do not require
-

solvents]
-Selective crosslinking possibilities upon imagewise exposure leading to a relief
-

image after development

Such systems are known as negative-working photoresists and can be utilized
in producing printing plates, printed circuits and name plates, to name
a few.
This review is intended to provide a survey of the chemistry in these areas,
to discuss some applications for photosensitive epoxy systems and to show recent
advances. A number of previous reviews are available 7 -9,11 -14} which have described
various photocrosslinkable epoxides.
Since epoxy groups can be attached on differently structured backbones R and
combined with other photosensitive groups L, taylor-made photosensitive resin
systems can be prepared. From a formal point of view, photocrosslinkable epoxy
resins may come in three types 7,8~:
I) Pure epoxy resins of the general formula

/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~

containing an epoxy and a photosensitive functional group L:

Photocrosslinking of Epoxy Resins 63

In the case of a) photocrosslinking is achieved by two different mechanisms:

cationic as well as free-radical-induced photopolymerization
(hybrid systems, see Sect. 4),
in the case of b) L is crosslinked by light, while the epoxy groups are crosslinked
by heat (epoxy resins of dual functionality, see Sect. 5).

2 Network Structure, Classes of Initiators and Epoxy Resins

Photoinitiated epoxy crosslinking is generally based (with one exception is) on

cationic ring-opening polymerization of the oxirane group, yielding polyether
structures.

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

Table 1. Cationic photoinitiators for the epoxide polymerization

Initiator type Refs.

A. Aryldiazonium salts 19-25.27-30)

B. 'Onium' salts and related compounds
Diaryliodonium salts 31 39,*
-

Diaryliodosyt salts ,to.41~

Triarylsulfonium salts and related compounds ~2-54~
T riarylsulfoxonium salts and related compounds 55-ss~
Dialkylphenacylsulfonium salts and related compounds 59.6o)
Diarylchloronium and Diarylbromonium salts 61)
Thiopyrylium salts 62~63)
Triarylselenonium salts 64)
'Onium' salts of group Va elements 65-68.43)
C. Organometal compounds
Organometal carbonyl compounds 77-so)
Dicarbonyl chelates of group Ilia, IVa and Va elements 81)
Ferrocene-Titaniumtetrachloride complex s2)
Zirconocene diiaalides s3)
Iron arene salts 73.87-90)
Aluminium complexes as, 86)
D. Various photoinitiators
Organohalogen compounds in conjugation with organometal derivatives sag
Fluorinated alkanesulfonic acid salts 26)
Chromates of alkali metal, alkaline earth metal and ammonium 1oo)
Phototropic o-Nitrobenzene compounds 101
Iodocyclohexene 102)
Unsaturated nitrosamines ~oa)
o-Nitrobenzyl esters and carbamates lo4)

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

[ f f " ~ c..JT~k ]~ YH,/7"-~

CH~CI"ICH~-K t '~)-C - ( / ~'~-OCH=CHCH:~O
"l~/ '~/~'-C-( / "~-OCH2CHCH=
L
. . . . , ~,. j,=--, c..,=-, o
X=0.15; 2.1z,; 5.1 or 11.8

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

Fig. 2. Principle o f Lewis acid formation by photolysis of aryldiazonium salts, M X . = PF 5, BF 3,

SbF s etc.

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.

3.2 "Onium" Salts and Related Compounds

Many different photoinitiators based on "onium"-type compounds with anions
of low nucleophilicity also have been described in the literature as effective
catalysts for the polymerization of epoxides: Thus, diaryliodonium salts 3a-39),
diaryliodosyl salts 4°'41), triarylsulfonium salts and related compounds 4z-54), tri-
phenylsulfoxonium salts 55-58), dialkytphenacylsulfonium salts 59) and dialkyl-4-
hydroxyphenylsulfonium salts 6o) seem to be most suitable as photoinitiators for
epoxy curing. Some of the principles of the reaction mechanism involving these
initiators are discussed in detail in the following Sections. Various other "onium"
photoinitiators such as diarylchloronium and diarylbromonium salts 6x),thiopyrylium
salts 62,63),triarylselenonium salts ~) and "onium" salts of group Va elements 43,65 -68)
have been mentioned, but they have not found technical acceptance as yet.
In recent years, the photochemistry and polymerization behaviour of iodonium 1~.
~z, 37, 38, 39) and triphenylsulfonium salts 11,11,51-54) was investigated. Crivello and
coworkers postulate homolytic cleavage of one of the aryl bonds induced by
a photochemical reaction as the first reaction step (Figs. 3 and 4) followed by
hydrogen abstraction from a suitable donor and loss of a proton yielding the
Bronsted acid HX:
Photocrosslinking of Epoxy Resins 67

Major
ArzICXG ~ (Arll~®) 'j ~ Arl~- 4- kr. + X O

Aria.+R-H ~ Arl*H + R -

ArleH ~ Arl + H~Xe

Minor
(Ar2leX~)'+ R-H --~ (ArRH)~ + ArI + X e

(ArRH)®.---~ ArR + H~Xo

Fig. 3. Mechanismof Brensted acid formation by photolysisofa diphenyliodoniumsalt (X- = PF6,

BF~-, SbF~ etc.)

Ar3SCX - -,h~" ( Ar2S~ At-) X e ~ Ar2S~- + At- + X ~

ArzS~- + R-H ~ Ar2S~-H + R-
Ar, S~H "--~ Ar~S + H*x e

Fig. 4. Mechanismof Bmnstedacid formation by photolysisofa triphenylsulfoniumsalt (X- = PF~-,

BF;, SbF6 etc.)

The powerful Bronsted acid HX produced by the photolysis of an "onium" salt

protonates the oxirane group in an initial step, and subsequently ring-opening
polymerization occurs. Several ways of chain termination are possible: the reaction
of the growing cationic chain end with nucleophilic or basic impurities or correspond-
ing reactive sites of polymers. Generally, only "onium" salts with anions BF4-,
PF6, AsF6 or SbF6 can be employed, the polymerization rate increases according to
the sequence given above. Cycloaliphatic epoxies show higher reactivities than
glycidyl ethers and glycidyl esters t 1)
"Onium" salt photoinitiators have strong absorption bands in the deep UV
region, but their sensitivity can be extended to longer wave lengths 69-72). Triaryl-
sulfonium salt photoinitiators with extended conjugation systems and improved
spectral sensitivity have been described 51,53,54), the photoinduced Bronsted acid
formation being similar to triphenylsulfonium salt initiators.
The photolysis of dialkylphenacylsulfonium salts 59) and dialkyl-4-hydroxyphenyl-
sulfonium salts 60) is different from that of triphenylsulfonium salts. The latter
compounds undergo irreversible photoinduced carbon-sulfur bond cleavage; the
former compounds, however, react by reversible photodissociation and form reso-
nance-stabilized ylids as shown in Fig. 5. Because of the slow thermally induced
reverse reaction, only small equilibrium concentrations of the ylid and acid are
present during irradiation and the concentration will rapidly decrease when photolysis
has been terminated. Therefore, in contrast to triarylsulfonium salt initiation, no
'dark' reaction will continue after the irradiation step.
68 F. Lohse and H. Zweifel

_%.. ,o, ...7

O ~/R
Ar-C-CH~-S,~ R
X~ hv
., -C=CH-~\R < ~ Ar-C-CH =SNR ~ -I-HX

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.)

Thermodynamic data of the epoxy polymerization with triphenylsulfonium salt

photoinitiators can be obtained by differential scanning calorimetric measurements
(DSC). Figure 6 shows the DSC diagram of the polymerization of Bisphenol-A di-
glycidylether with triphenylsulfonium hexafluorophosphate, irradiation being carried
out at --88 °C 73,

-1-
<,

s'o ~5o ~,~o 260

Temperature (*C)

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

The measurements of the pendulum hardness confirm further the necessity of

a final thermal activation step. While irradiated, samples may exhibit up to 180 s
pendulum hardness upon standing for 24 hours at 20 °C; the same irradiated samples
show a pendulum hardness of 220 s if heated at 110 °C for 3 min. immediately after
irradiation (Method of determination according to K6nig.)

lI
2oo
Z

~150- Fig. 7. Pendulum hardness beha-

viour of a coating composed of
Bisphenol-A diglycidyl ether with
2.5 % (w/w) Ph3S+PF6.
I: irradiation carried out with the
~100- use of an IR filter; storage at
20 °C over period of time
II: irradiation similar to I, sub-
sequent thermal cure for 3 min
50 at 110 °C
0 2'0 4'06'0 120 2/,0 /,80 960 Time(mi~)

It can be concluded that cationic photopolymerization of glycidyl ethers with

"onium salt" initiators should be considered as a dual-step process: 1) liberation of
the active initiator species by irradiation and 2) heat treatment to complete the poly-
merization reaction 74)
Unfortunately, the initiating species, the Bronsted acid HX, is not consumed
during polymerization; it remains in the hardened layer and can lead to undesired
ether cleavage reactions 75) causing degradation of the network or it can be a source
of corrosion problems in coatings on metal substrates. Recent work with triphenyl-
sulfonium and diphenyliodonium salts having phosphotungstate, phosphomolybdate
and the related silicon counterions has been reported to exhibit improved corrosion
inhibition behaviour in epoxy coatings 76)

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

Fig. 8. F o r m a t i o n o f a Lewis acid by irradiation o f an iron arene salt

Photolysis of the arene complexes in the presence of monodentate ligands, e.g.

carbon monoxide, leads to new complexes of the type CpFe(L)~- 93) whereas in pure
aprotic solvents, ferrocene and iron salts are formed 94). Investigation of the photo-
lytic reaction of an iron arene complex with excess ethylene oxide in methylene
chloride solution (Meier and Rhis 95)) showed that a crystalline crown ether
complex (structure shown in Fig. 9) was obtained in high yield. Only traces of
dioxane could be detected.
• q2°

2 * 8 / 0N h~ = Fe (pFee)2 . FeCi ~

Fig. 9. Formation of [Fe(1,4,7,10-tetraoxacyctododecane)2](PF6)2 by irradiation of ethylene oxide

in the presence of CpFe(Toluene)(PF6) Cp = Cyclopentadien
Photocrosslinking of Epoxy Resins 71

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

Fig. 10. General principle of epoxide polymerization with

iron arene complexes

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.

SIS Fig. 11. DSC diagram of the

polymerization of Bisphenol-
A diglycidyl ether with (q6.
benzene)(q5-cyclopentadi-
enyl)iron(II)-hexafluorophos-
phate, 2.5~ (w/w), irradia-
tion carried out at --88 °C,
rate of heating 20 °C/min
6 ..... 50 160 150 2(~0
Temperature (%)
72 F. Lohse and H. Zweifel

The viscosity of the irradiated Bisphenol-A diglycidylether/initiator reaction mixture

increases only slightly with storage time. Even after prolonged storage of the exposed
sample, no tackfree coating could be obtained (see Fig. 12).

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)

Figure 13 shows the time-temperature-transition diagram which was obtained

according to the method described by Gillham 96) The measurements were carried
out by T M A analysis, the time to vitrification is plotted as a function of isothermal
cure temperature. At temperatures below the glass transition of the irradiated
epoxy re~in 'initintnr mixture (Bi~phenol-A di~lycidylether with iron arene salt, 2.5
w/w, Tgo = - - 9 °C), the reaction occurs in the glassy state and is, therefore, extremely
slow. Above Tgo, the reaction mixture becomes liquid, but the reaction rate of the
curing reaction is still very slow. By rising the temperature above 80 °C, the reac-
tion rate increases sharply and the liquid resin reacts until the continuously rising
glass transition temperature approaches the cure temperature, upon which vitri-

300
~ \\\\

200 Rubber region

o)

$ Tgoo
O.
E
(I)
]00

C)

Fig. 13. Time-temperature-transition dia-

0 ,~,,xx~ \ Liquid region x ~~ gram of Bisphenol-A diglycidyl ether poly-
merized with 2.5~o (w/w) (rl6-benzene)-
Tgo (rl5-cyclopentadienyl)iron(II)-hexafluoro-
....... . ................... '///////3 phosphate
t lO 100
Time {h)
Photocrosslinking of Epoxy Resins 73

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-

20- Fig. 14. Relation between the

initiator concentration and AH
(DSC)
Initiator concentration [ w / w ] ( * / , )

I/,0.

120. --II x =0.15

~ 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

Epoxy content Repeating unit Heat of polymerization T~ (°C) resin

mole epoxide x AH (DSC) (DSC)
per kg resin
J/g resin k J/mole epoxide linear cured

5.2 0,15 308.4 59.3 --9 114.5

2.34 2.14 160.9 68.7 25 100.5
1.11 5.1 53.6 48.3 60 82.3
0.59 11.8 38 64.3 76 78,5

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 ~

...... "-,.-"-" "~.- ...'~":;'~-

:~.'~,",- "i"---" )" . . . . .

-/ \/

2 57"C 68"C 98 C 1090C 151"C

55KJ 33KJ 32KJ 72KJ 56KJ 46KJ /MOLE EPOXY

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

Figure 17 shows the enthalpy recording of DSC experiments with Bisphenol-A

diglycidylether polymerized with (q6-naphthalene)(rlS-cyclopentadienyl)iron-hexa-
fluorophosphate after irradiation at low temperature in the presence and absence of
equimolar amounts of the oxidant cumene hydroperoxide.
In the presence of the oxidant, the polymerization reaction occurs already at
about 50 °C. With such compositions, tack-free coatings can be obtained solely by
irradiation without further heating ss~,
Iron arene photoinitiators have excellent light absorption properties in the ultra-
violet and visible parts of the spectrum. As shown in Fig. 18, the absorption can be
varied over a wide range by structural changes in the ligands. Iron arene salts can
be sensitized, for example with anthracene derivatives 98~:
During irradiation of sensitized epoxide matrices, the absorption spectrum is
altered and colourless coatings are obtained. Iron arene complexes are thermally
stable up to 300 °C.
76 F. Lohse and H. Zweifel

,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) -----,.-

3.4 Miscellaneous Structures

Cationic polymerization of epoxides by irradiation of charge-transfer complexes
has been mentioned in the literature 99) Fluorinated alkanesulfonic acid salts 26)
chromates and dichromates of alkali metals, alkaline earth metals and ammonium 100),
phototropic o-nitrobenzyl esters 101j, iodocyclohexene 102) unsaturated nitrosamines
and carbamates 10a,104~ have been reported to act as cationic photoinitiators.
However, none of these compounds has found technical acceptance yet.

4 Hybrid Systems for Cationic and Radical Photopolymerization

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

PhCOCHPh "'%, Ph(~O 4-¢HPh

~e OMe

Ph~H 4" Ar21~F~ " PhCH~PF'6

e 4- Arl 4" Ar-
OMe 6Me
¢ Fig. 19. Generation of aryl radicals
PhC,O 4- Ar21ePF~ ~ PhCO PF~ 4- Arl 4- At- by electron transfer processes

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.

5 Epoxy Resins of Dual Functionality

The combination of epoxy groups with a second functional group opens new
possibilities for various technical applications. In such a case, one group can be
used for thermally induced and the other for photoinitiated reactions.
Epoxy resins containing nitrogen-heterocyclic rings bearing photosensitive,
unsaturated substituents have been described 108) These materials can be prepared
by chain-extension ('advancement') of low-molecular-weight epoxides with un-
saturated derivatives of hydantoin. Another possibility exists in introducing the
light-sensitive groups at suitable positions of an epoxy resin molecule ~o9,110). Such
systems contain both light-sensitive units and groups capable of being activated
by heat and can first be partially reacted or crosslinked by exposure to light.
Subsequently, they can then be further crosslinked in a second step by heating, e.g.
in the presence of suitable hardeners. Chalcone-group-containing epoxy resins are
of particular interest 7, s) Such resins can be obtained by condensation ofp-hydroxy-
benzaldehyd with p-hydroxyacetophenone m-HS). By reaction of the resulting
bisphenol with epichlorohydrin and further 'advancement' of the low-molecular-
weight diepoxide, a diglycidyl ether shown in Fig. 20 is obtained.
The light-induced cyclodimerization of the chalcone groups leads to insoluble resins ~"
s~ The residual epoxide groups can subsequently take part in seperate, thermally
induced crosslinking processes ~16)

"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

F o r m u l a t i o n s o f epoxy coatings can be varied over a wide range by combining

different epoxy resins and cationic photoinitiators discussed in this paper. The
reactivity o f a formulation depends u p o n the structure of the epoxies and the
photoinitiators which can be further influenced by additives such as fillers,
pigments, colorants, stabilizers etc. Complete crosslinking o f epoxide/initiator systems
can be achieved by an irradiation and a heat treatment step, both o f a very short
d u r a t i o n 9o~
D u e to the chemical structure and the UV absorption ranges o f the iron arene
salts 73) (the optical density in the UV and visible part o f the spectrum decreases
during irradiation), even thick films up to several 100 micrometers can be cross-
linked.
Besides the most i m p o r t a n t area o f surface coatings, the use of photopolymers
as photoresists in the manufacture o f printed circuits is well established. Photo-
imaging with aryldiazonium salt photoinitiators a n d multifunctional cresol-novolac
epoxides was first described by Schlesinger 23~. Crivello has mentioned several
new photoresists based on the photopolymerization o f epoxides with " o n i u m "
initiators ~1~). Meier and Zweifel 1~a) have shown that iron arene salts in combination
with multifunctional cresol-novolac epoxides yield photoresists with high resolution
and contrast. D u a l functional epoxides (cf. Sect. 5) containing chalcone groups as
light-sensitive units have been described as suitable photoresists 119.120,~21) especially
as solder masks.

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

An isothermal time-temperature-transformation (TTT) cure diagram for thermosetting systems has

been developed to aid in the understanding o f the cure process and the properties after cure. The diagram
summarizes in a convenient manner the temperature o f cure ( T,u,e) vs. the times to gelation, vitrification,
phase separation (in the case o f rubber-modified systems), full cure, and thermal degradation. A more
complete diagram could atso include contours o f constant conversion, viscosity, and mothdus. The principal
experimental technique which has been used to obtain a T T T diagram, torsional braid analysis ( TBA ), is
described. Two recent models describing the cure process are revieved; from these the time to vitrifica-
tion vs. To,re is computed from the chemical kinetics and the conversion at vitrification. One model uses
a relationship between T 8 and the extent o f conversion at vitrification in conjunction with experimental
data for the extent o f conversion at vitrification, whereas the other predicts the extent o f conversion at
Tg from relationships in the literature. Both models predict an S-shaped time to vitrification curve, as
has been obtained experimentally. The second model has been apptied also to linearlypolymeriz ing systems.

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

2 Generalized Schematic T I T Cure Diagram . . . . . . . . . . . . . . . 85

3 Experimental Aspects of the T T T Cure Diagram . . . . . . . . . . . . . 86

3.1 Torsional Braid Analysis (TBA) . . . . . . . . . . . . . . . . . . 86
3.2 Relationship of TBA Parameters to Material Parameters . . . . . . . 90
3.3 Isothermal and Temperature Scans . . . . . . . . . . . . . . . . 91
3.4 High Tg Epoxy Resins . . . . . . . . . . . . . . . . . . . . . . 97
3.5 Rubber-Modified Epoxy Resins . . . . . . . . . . . . . . . . . . 99

4 Modeling the T I T Cure Diagram . . . . . . . . . . . . . . . . . . . 100

5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

Advances in Polymer Science 78

(C3 Snrin~er-Verla~ Berlin ]-leide|Ea~ro I OR~
84 M.T. Aronhimeand J. K. Gillham

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.

2 Generalized Schematic TTT Cure Diagram

A schematic T I T cure diagram, illustrating the main features of the thermosetting
process, is shown in Fig. 1. The diagram indicates distinct regions of matter
encountered in the thermosetting process; these include liquid, sol/gel rubber, gel
rubber, sol/gel glass, gel glass, sol glass, and char. Three critical temperatures are
also shown: Tso, ~=ETg,and Tg=. A "'full cure" line is indicated on the diagram which
separates the sol/gel rubber region from the gel rubber region, and the sol/gel
glass region from the gel glass region. Isoviscosity contours, the onset of phase
separation, and thermal degradation events are also included. Finally, note that
the temperature of cure vs. the time to vitrification curve is S shaped.
Tgo is the glass transition temperature of the uncured reactants. Below this tempera-
ture, in principle the system has no reactivity, g=]Tgis the temperature at which gela-
tion and vitrification coincide. Between Tgo and ~=~Tgthe system will vitrify before

",.%
",,, o rubber ttr/~,.~. ~
Chor
-- ,.
Sol/Gel '~,~ /e -~ - '
rubber "%. qtioo~-~.

.................. "" Oevi t r t f .

o

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.

3 Experimental Aspects of the TTT Cure Diagram

3.1 Torsional Braid Analysis ~BA)
The isothermal TI'I" cure diagram summarizes the transitions that occur on iso-
thermal polymerization, such as gelation, vitrification and devitrification. Typically,
88 M.T. Aronhime and J. K. Gillham

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

DRIVE GEAR TRAIN

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

3.2 Relationship of TBA Parameters to Material Parameters

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 I = moment of inertia of the oscillating system

rl = air-damping coefficient, which is related to the moving parts of the
system
K = factor which depends on the dimensions of the sample
G* = complex shear modulus
0 = angular displacement
The solution of Eq. (1) is:

0 = 0o exp (--130 exp (imt) (2)

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:

1(132-m 2)-[311 + K G ' = 0 (3a)

vim -- 2113m + KG" = 0 (3b)

which are obtained by grouping real and imaginary terms.

Now:

O. = Oo exp [t~(im -- 13)] (4a)

0.+1 ='0o exp [(t. + P) (ira -- 13)] (4b)

where tn is the time required to reach the 0,th oscillation. Dividing Eq. (4a) by
Eq. (4b) yields:

0,10n+, = exp [P(13 -- ira] = exp (P13) exp (~imP)

= exp (P13) exp (--i 2n) = exp (P13) (5)
Time-Temperature-Transformation(TIT) Cure Diagram of ThermosettingPolymericSystems 91

Thus, the logarithmic decrement= A = In (0./0.+ 1) = ~P.

Finally, from Eqs. (3 a) and (3 b):

KG' = (4~2I/p2) (1 --A2/4rc2 + APrl/4~2I) (6a)

KG" = (4gI/P 2) (A - - rlP/2I) (6b)

If 1 > A ~> qP/I (the air-damping coefficient is usually negligible):

KG' ~ 4rtzI/P2 (7a)

KG" ~ 4nlA/P 2 (7b)

In a TBA experiment K is unknown, so G' oc 1/p2, and

A ~ KG"/(4gI/P 2) = gG"/G' = rc tan 6 (8)

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.

3.3 Isothermal and Temperature Scans

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

..................... ;,*:.~.,~.¢o ,oo~.

10 0 101 .~
C
"I0
0
°m
%.,, ~"t ~ * ""
E
~- 10-1 1o0 ~,

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

to the post-cure temperature yields the Tg after cure, which is observed to be

greater than To,re if the material had vitrified on isothermal cure. The subsequent
scan from the post-cure temperature to --170 °C yields a transition temperature,
here designated T ~ , for that particular To,re- Tgoo is obtained by averaging the
values of T'g~o for the different temperatures of cure. The post-cure temperature
depends upon the system being investigated. For a variety of tetrafunctional
aliphatic and aromatic amines, the post-cure temperature was 240-250 °C for a di-
functional epoxy and 280-300 °C for a trifunctional epoxy.
A typical temperature scan of a system after prolonged isothermal cure is shown in
Fig. 5. In comparing post-cure behavior with that after cure at Tc.... note the increase
in Tg as well a~ in the temperature of the secondary transition (Ts~c). Also note that
the relative rigidity (modulus) of the post-cured material is lower at room temperature
(RT) than that of the partially-cured specimen. This behavior is anomalous, because
post-cure would be expected to increase the crosslinking, and hence the stiffness of the
material. The lower modulus manifests itself in a lower density and greater water
absorption at RT for the more highly cured material than for the partially-cured
one 13, 14).

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

vitrification is defined to occur when Tg = T . . . . . Two other reasons for T, being

greater than T . . . . are: 1) the temperature scan after isothermal cure can p r o m o t e
further reaction as Tg is being measured; a n d 2) reactions can proceed in the glassy
state to an extent determined by the reaction mechanism 4).
In Fig. 6, Tg is shown to be only a function o f T . . . . ° However, even for cures carried
out well beyond the point where Tg = T ..... Tg is actually a function o f both the
temperature and time o f cure 4), as shown in Fig. 7. In Fig. 7, the approximately
parallel lines represent a series o f isochrones. I f each isochrone is extrapolated to
Ts~, the isochrone will intersect the horizontal Tg® line at different values o f T~,,,.
F o r each o f these values o f T¢,,, (136, 141, I42 °C, etc., Fig. 7), the time for the
isochrone is the curing time required to produce full cure (defined to occur when
T z equals Ts~ ). F o r example, a T .... o f 152 °C, for 720 min, will result in a material
with a Tg = Ts~. Plotting T¢,,¢ obtained from the extrapolation vs. the isochronal
cure time on the T T T diagram results in the full cure line (Fig. 8). The full cure
line summarizes the time required, at any given T ..... to obtain a material with
Tg = Tg~.

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

3.4. High Tg Epoxy Resins

For high temperature systems, thermal degradation is an important consideration,
and the competition between cure and degradation can prevent full cure, or Ts~, from
being reached 3~. For one particular system, with a theoretical T ~ of 352 °C, iso-
thermal experiments for prolonged times at elevated temperatures can detect
degradation events such as devitrification and subsequent revitrification (char forma-
tion) 3~. Isothermal plots of relative rigidity and logarithmic decrement, from 80 to
350 °C, for a high T s epoxy system, are shown in Figs. 9 and 10. For the lower
temperatures (Fig. 9), gelation and vitrification peaks are evident in the logarithmic
decrement, and the modulus increases as the cure proceeds. At the higher tem-
peratures (Fig. 10) gelation and vitrification occur too rapidly to be measured;
however, additional events occur, as evidenced by the maxima in the loss peaks and
the associated changes in the modulus. The first maximum (Fig. 10) is associated
with a devitrification event (decrease in modulus), whereas the second peak is
associated with a revitrification event (increase in modulus). The locus of these
events can be placed on the TTI" diagram for this system (Fig. 11). Thus, gelation,

350 e C

• . ...... "" " ' " ~ 350"C

335 *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

-10 ! I ..................... I .......... !

0 1 2 3 /~
log t i m e {min)

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.

3.5 Rubber-Modified Epoxy Resins

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~)

Isothermal TBA scans of rubber-modified systems show no evidence of phase

separation, and they only differ from the neat systems in terms of the transition
times. In the subsequent temperature scans, however, the presence of the rubber-rich
domains is evident by the distinct loss peaks at the Tg of the rubber (,,,--50 °C)
(Fig. 14). The relative rigidity plots are qualitatively similar to those seen previously
(e.g., Fig. 5), including the inversion of the moduli at RT.
In this Section, an experimental approach for constructing isothermal TTT cure
diagrams has been described, T r T diagrams of representative epoxy systems including
high Tg and rubber-modified epoxy resins have been discussed, and perturbations to
the TTT cure diagram due to thermal degradation and rubber modification have
been illustrated.

4 Modeling the TTT Cure Diagram

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)

tion, rubber modification, and viscosity (diffusion control) are ignored. M o r e

sophisticated models could incorporate the refinements.
D u r i n g isothermal polymerization below Tg~, the molecular weight and Tg increase,
and eventually Tg will equal To,re. The m a i n p u r p o s e o f this section is to discuss the
calculation o f the time to vitrification, where vitrification is defined to occur when
T~ equals T . . . . . The concepts o f vitrification and the T T T cure diagram are extended to
linear systems for both step growth and chain reaction mechanisms, although m o s t
o f the discussion will focus on the nonlinear step growth case, o f which the cure o f
epoxy resins is a n example.
The calculation o f the time to gelation is straightforward if gelation is assumed
to be an isoconversion state 7~, and if the kinetics o f the reaction are known. The rate
o f reaction in general is:

- - d c / d t = kf(c, cl, c2 .... ) (9)

where c is the concentration o f the reactant under consideration, k is the temperature-

dependent rate constant, and cl, c2, ... are the concentrations o f other reactants in
the system, f(c, cl, c2 .... ) is a function o f the reaction mechanism and the extent o f
conversion. F o r the simple case o f two reactants in stoichiometric ratio, as is
102 M.T. Aronhime and J. K. Gillham

considered in this section, f(c, Cl, c 2. . . . ) can be reduced to f(c). Substituting

c = Co(1 - - p) into Eq. (9), and integrating yields:

kt = S Co dp/tlco(1 - - p)] (1o)

F o r example, if the reaction is first order, f [ c o ( l - p)] = c o ( l - p), and kt

- In (1 - - p), where p is the extent of reaction.
The conversion at gelation is generally assumed to follow from Flory's theory
(7):

Pg~l = 1 / i f - - 1) 1/2 (11)

where f is the functionality of the multifunctional unit in a nonlinear reaction.

Equation (11) is valid for the stoichiometric reaction of a multifunctional reactant
Af with a difunctional reactant B2. For the typical case of a difunctional material
cured with a tetrafunctional material, f = 4 and Pge~ = 0.577. Experimental values of
p~e~ are usually observed to be greater than the predicted values because of non-
idealities relative to the theory, such as intramolecutar ring formation and unequal
reactivities of the same functional groups.
Equations (9) and (10) assume that the reactions are not diffusion controlled and
only one temperature-independent reaction mechanism is operable. Epoxy thermo-
setting reactions are actually complex, and complicated kinetic expressions and
competing reaction mechanisms have been proposed 16)
Whereas the calculation of the time to gelation is relatively simple, the calculation
of the time to vitrification (t~t) is not so elementary. The critical point is to obtain
a relationship between T~ and the extent of conversion at T~ (Pvit). Once the
conversion at Tg is known, then the time to vitrification can be calculated from the
kinetics of the reaction. Two approaches have been examined: one calculates tvlt
based on a relationship between Tg and Pvit in conjunction with experimental values of
Pvit 6, 17); the other approach formulates the Tg vs. Pvlt relationship from equations in
the literature relating Tg to molecular weight and molecular weight to extent of
reaction :8.19.~
The first method of calculating tvi t is based on an equation from DiBenedetto, as
presented in Nielsen 20~.

(Tg - - T~)fFgo = (ex/eM - - Fx/FM) pvl,/[1 - - (1 - - Fx/FM) Pvit] (12)

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)

fail at high temperatures, where the time to vitrification is very short. In a T B A

experiment the system generally requires several minutes to equilibrate thermally,
which could account for the lack o f agreement at high temperature.
Data for the time to the onset o f diffusion control, as determined by infrared
spectroscopy, are also included in Fig. 15. The time to diffusion control was
selected as the point at which the extent of conversion vs. time data, plotted for
first order kinetics, deviated from linearity. The onset o f diffusion control corresponds
with vitrification, as determined by TBA, for the system o f Fig. 15 6~.
In the above model, data for the extent of reaction at vitrification are needed. These
data were obtained using infrared spectroscopy. The extent of conversion o f the epoxy
group was monitored as a function o f time at a series o f temperatures. A corresponding
set o f T B A experiments was performed, and the time to vitrification data were
104 M.T. Aronhimeand J. K. GiUham

superimposed on the extent of conversion data, at the different temperatures, to

obtain the conversion at vitrification.
The second approach attempts to predict P, it, and then calculates tvltfrom those pre-
dictions 18,~9~. In addition, this second model can be extended easily to linear
polymerizations for different reaction mechanisms.
Several relationships are needed to calculate tvlt. These are:

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~:

1/Tg = lfI',~ + K/M, (13)

where K is a constant. This Equation is applicable over a wide range of values of

molecular weight.
For nonlinear systems, two regimes are distinguishable: (1) vitrification that occurs
below genT,; and (2) vitrification that occurs above genTr Below ,~lTg the systems will
vitrify without gelling, so the material is not crosslinked. Above ,e~Tg, both
crosslinked and uncrosslinked material are present at vitrification.
Eq. (13) was used to relate Tg and M, below gelTg, even though the material is
branched at vitrification. (An estimate of T,~ for uncrosslinked material was
obtained by using Eq. (13) at Tgo and g~T,.) A more appropriate expression, which
attempts to account for the effect of branching on T,, has been proposed 24~
Above g~Tg, T, must be related to the molecular weight of the sol fraction and the
crosslink density of the gel fraction. If the system is considered to be a miscible binary
mixture of sol and gel fractions, then 25~:

Tg = wsTgs + wgTg, + Iw~wg (14)

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):

T,, = ,~Tg + Kx[X] (15)

where Kx is a constant and [X] the concentration of-crosslinks in moles of crosslinks

Time-Temperature-Transformation(TTT) Cure Diagram of ThermosettingPolymericSystem~ |05

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~:

Mn = (MAfAf + M~Bh)/(Af + Bh -- pAfAf) (16)

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:

moles polymer = (moles of monomer in polymer) _ p (17)

(moles of monomer/polymer) <X~)

where <Xn> is the cumulative number average degree of polymerization of the

polymer, then:

M, = Mo/[(1 - - p ) + p/(X,)] (18)

where Mo is the monomer molecular weight.

For the nonlinear step growth case above ,elT,, the crosslink density must be
related to p. A relevant model, based on calculating the probabilities of finite chains
being formed, has been published 2s) For the reaction of A4 + 2[32 (e.g., tetra-
functional amine + difunctional epoxy), A4 is considered to be an effective cross-
linking site if three or more of its arms lead out to the infinite network. The
probability of finding an effective crosslink is related to one minus the probability of
a randomly chosen A4 leading to the start of a finite chain, which in turn is related to
the extent of reaction. Application of this procedure to the system of Fig. 15 has been
presented in detail 19). The more complicated reaction of a tetrafunctional amine with
a trifunctional epoxy was also considered 19).
The last step in this second model is to relate the extent of reaction at vitrification
to time. For the step growth mechanism, for both linear and nonlinear systems,
nth order kinetics were assumed 29.30).

--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).

Rp = --d[M]/dt = kp[M] (flq[I]/kt) 1/2 (20)

where [M] = m o n o m e r concentration at time t

kp = propagation rate constant
f = initiator efficiency
kd = initiator decomposition rate constant
[I] = initiator concentration at time t
lq = termination rate constant
If first order decomposition o f the initiator is assumed, and from [M] = [M]o (1 - - p),
integrating Eq. (20) over time yields:

- - In (1 - - p) = 2kp(f[I]o/kakt) 1/2 (1 - - exp (--kdt/2)) (21)

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):

( X , > = ([M]o - - [Ml)/(f{[I]0 - - [I]})

= [M]o p/(t~]o {1 - - exp (--kdt)}) (22)

In Eq. (22), termination by combination was used.

The time to vitrification, as a function o f reaction temperature, can now be solved
for each of the three cases considered. The only case for which experimental data are
available for t,, is the nonlinear step growth case. Combining Eqs. (13)-(16), (19),
and those relating the crosslink density to p, results in the plot o f Toure vs. t,, shown
in Fig. 16. The system used was the same one used in Fig. 15. Different values of
the reaction order (n) were used in Fig. 16. The value of k obtained for n = 1 was
used for all values of n. The fit is not entirely satisfactory, but the lack o f an accurate
kinetic model mitigates against a good fit. The calculated time to vitrification curve
is S-shaped, as is seen experimentally.
The model predictions o f the extents o f reaction at vitrification vs. reaction
temperature are compared to the experimental values in Fig. 17. The model predictions
are too low, and the inherent simplifications in the model could account for some o f
the discrepancy. These simplifications include 2s): all functional groups o f the same
type are equally reactive; all groups react independently o f one another; and no
intramolecular reactions occur in finite species.
The model was also applied to the reaction o f a tetrafunctional amine with
a trifunctional epoxy, denoted A , + 4/3Ba, and was compared with available data
(Fig. 18). A n approximate value of k was obtained from the times to gelation. This
model appears to provide a reasonable framework within which the vitrification
process for nonlinear systems can be discussed.
Time to vitrification data for the other two cases are not available. For a hypothe-
tical linear step growth reaction o f A2 + B2, with reasonable values o f MA, Ms,
Time-Temperature-Transformation (TTT) Cure Diagram of Thermosetting Polymeric Systems 107

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

3 ~o Fig. 19. Reaction temperature vs. time

o to vitrify for linear step-growth poly-
CU
merization: nth-order kinetics, for
n = 1 to 3 in increments o f 0.5 using the
20
following parameters: Tso = 0 °C; T=o~
= 100 °C; E= = 12.6 kcal/mole;
A = 4.51 x 106 min -1 ; M A = 200 gm/
mole; MB = 400 gin/mole. [Aronhime,
M. T., Gillham, J. K. : J. Coat. Tech.
l o g t i m e to v i t r i f y (rain} 56 (718), 35 (1984)]
Time-Temperature-Transformation (TI'I') Cure Diagram of Thermosetting Polymeric Systems 109

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).

Tg = (cz~,¢pTgp + ~x,.(1 - - COp)Tgm)/(cxpq~p + ~Xm(l - - %,)) (23)

where a = volume coefficient o f expansion of liquid minus volume coefficient o f

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)]

e x p a n s i o n o f glass, ~p = volume fraction, and the subscripts p and m refer to

p o l y m e r and m o n o m e r , respectively. F r o m Eq. (23):

q~p = [ ~ ( T , . - - Ts)I/[%(T s - - T,p) + ~ ( T , m - - Ts)I (24)

F r o m a m a s s balance on the p o l y m e r and m o n o m e r :

~ = (p/Qp)/[(l - - p)/Q~, + p/Qp] (25)

w h e r e (~ = density. T h u s :

p = l/[(l/q~p- l)e.,/ep + 11 (26)

1.0
J

.,,/:-//
Tg =f (Free volume) . ~ 1 "

O.8
¢-
,0 . . S ( Mn }
0.6

Fig. 22. Extent of reaction at vitrification

"~
¢1
0.4 vs. reaction temperature for linear free-
radical polymerization (styrene) for
UJ f = 0.5 and [1]~ = 0.10 mole/l. The solid
O.2 line is for the results from the Ts-mote-
cular weight model [Eq. (21)]; the dashed
line is for the results from the free volume
0 I l I theory [Eq. (26)]. [Aronhime, M. T.,
-100 -50 0 50 100 Gillham, J. K,: J. Coat. Tech. 56 (718),
Temperature of reaction (*C) 35 (1984)]
Time-Temperature-Transformation(TTT) Cure Diagram of ThermosettingPolymericSystems 111

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

1. Enns, J. B., Gillham, J. K. : ACS Adv. Chem. Ser. 203, 27 (1983)

2. Gillham, J. K.: in Developments in Polymer Characterisation-3, J. V. Dawkins, Ed., p. 159,
Applied Science, London 1982
3. Chan, L. C., Na6, H. N., Gillham, J. K.: J. Appl. Polym. Sci. 29, 3307 (1984)
4. Peng, X., Gillham, J. K.: J. Appl. Polym. Sci., J. Appl. Polym. Sci. 30, 4685 (t985)
5. Chadwick, G. A. : Metallography of Phase Transitions, Crane, Russak & Co., Inc., N.Y. t972
6. Enns, J. B., Gillham, J. K.: J. Appl. Polym. Sci. 28, 2567 (1983)
7. Flory, P. J. : Principles of Polymer Chemistry, Cornell University, Ithaca, N.Y. 1953
8. Chan, L. C., Gillham, J. K., Kinloch, A. J., Shaw, S. J. : ACS Adv. Chem. Set. 208, 235
(1984)
9. Gillham, J. K.: AIChE J. 20 (6), 1066 (1974)
10. Read, B. E., Dean, G. D.: The Determination of Dynamic Properties of Polymers and
Composites, pp. 54-56, John Wiley & Sons, N.Y. 1978
11. N i e l s e n , L. E. : Mechanical Properties of Polymers, pp. 194-5, Reinhold Publishing Corp., N.Y.
1962
12. Enns, J. B., Gillham, J. K.: ACS Symp. Set. 197, 329 (1982)
13. Enns, J. B., Gillham, J. K.: J. Appl. Polym. Sci. 28, 2831 (1983)
14. Aronhime, M. T., Peng, X., GiUham, J. K., Small, R. D.: J. Appl. Polym. Sci,, in press
15. Chan, L. C., Gillham, J. K., Kinloch, A. J., Shaw, S. J.: ACS Adv. Chem. Ser. 208, 261 (1984)
16. Osinski, J. S., Manzione, L. T. : ACS Symp. Set. 221,263 (1983)
17. Adabbo, H. E., Williams, R. J. J.: J. Appl. Polym. Sci.: 27, 1327 (1982)
18. Aronhime, M. T., Gillham, J. K.: J. Appl. Pole,an. Sci. 29, 2017 (1984)
19. Aronhime, M. T., Gillham, J. K.: J. Coat. Techn. 56 (718), 35 (1984)
20. Nielsen, L. E.: J. Macromol. Sci.-Revs. MacromoL Chem. C3 (1), 69 (1969)
21. Fox, T. G., Loshaek, S. : J. Polym. Sci. 15, 371 (1955)
22. Ueberreiter, K., Kanig, G.: J. Colloid Sci. 7, 569 (1952)
23. Couchman, P. R. : Polym. Eng. Sci. 21 (7), 377 (1981)
24. Kow, C., Morton, M., Fetters, L. J., Hadjichristidis, N.: Rubb. Chem. Tech. 55, 245 (1982)
25. Fried, J. R. : in Developments in Polymer Characterisation-4, J. V. Dawkins, Ed., p. 39,
Applied Science, London 1983
26. Horie, K., Hiura, H., Sawada, M., Mita, I.: J. Polym. Sci. A-1 8, 1357 (1970)
27. Macosko, C. W., Miller, D. R. : Macromolecules 9 (2), 199 (1976)
28. Miller, D. R., Macosko, C. W.: Macromolecules 9 (2), 206 (1976)
29. Prime, R. B. : in Thermal Characterization of Polymeric Materials, Tuff, E. A., Ed., pp. 441,
480-482, Academic Press, N.Y. 1981
30. Prime, R. B. : Polym. Eng. Sci., 13 (5), 365 (1973)
31. Odian, G . : Principles of Polymerization, 2nd Ed., John Wiley & Sons, N.Y. 1981
32. Rosen, S. L. : Fundamental Principles of Polymeric Materials, 2nd Ed., John Wiley & Sons,
N.Y. 1982
33. Kwant, P. W.: J. Polym. Sci.: Polym. Chem. Ed. 17, 3397 (1979)
Time-Temperature-Transformation (TIT) Cure Diagram of Thermosetting Polymeric Systems 113

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

2 Molecular Structure Characterization . . . . . . . . . . . . . . . . . . 118

2.1 Rubber-like Elasticity Theory . . . . . . . . . . . . . . . . . . . 118
2.2 Rubber-like Elasticity Theory and Highly Crosslinked Epoxies . . . . . 120

3 Rubbery Cohesive Fracture . . . . . . . . . . . . . . . . . . . . . . 125

3. I Generalized Fracture Theory . . . . . . . . . . . . . . . . . . . 125
3. I.I Threshold Fracture . . . . . . . . . . . . . . . . . . . . . 125
3.1.2 T h e Loss Function . . . . . . . . . . . . . . . . . . . . . 127
3.2 Rate and Temperature Effects . . . . . . . . . . . . . . . . . . . 128
3.3 Effect of Molecular Structure on the Rubbery Tear o f Highly Crosslinked
Epoxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

4 Glassy Cohesive Fracture . . . . . . . . . . . . . . . . . . . . . . . 132

4.1 Fracture Testing . . . . . . . . . . . . . . . . . . . . . . . . . . 132
4.2 Effects o f Testing, Material and Processing Variables . . . . . . . . . 134
4.3 Origin of Unstable Crack Growth . . . . . . . . . '. . . . . . . . 138
4.4 Effect o f Structure on the Glassy Fracture o f Highly
Crosslinked Epoxies . . . . . . . . . . . . . . . . . . . . . . . 140

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.

2 Molecular Structure Characterization

2.1 Rubber-like Elasticity Theory

The average length (or molecular weight) of network chains in a crosslinked polymer
can be experimentally determined from the equilibrium rubbery modulus. This
relationship is a direct result of the statistical theory of rubber-like elasticity zo)
In the last decade or so, modern theories of rubber-like elasticity 21-2v) have further
refined this relationship but have not altered its basic foundation. In essence, it is
Structure and Ultimate Properties of Epoxy Resins 119

accepted that the mechanical response of an ideal etastomeric network is simply

the sum response of the elastically active chains of which it is composed, and
that the dimensions of these chains can be accurately described by the Gaussian
statistics of random coils. Modern refinements typically take a harder look at the
fundamental assumptions of the basic statistical theory, for example, the assumption
that network deformation is "affine" (i.e., that network chain junctions deform in
proportion to the bulk elastomer). Experimental verification of such refinements,
however, is extremely difficult because of the practical difficulties involved with
synthesizing suitable "ideal" networks which yield completely unambiguous results.
For the epoxy networks above Tg, the applicability of even the simplest form
of statistical network theory must be critically examined. The primary concern is,
of course, whether epoxy network chains have sufficient length to be accurately
described by Gaussian statistics. For common commercial epoxy thermosets it is
unlikely that this criterion is met, as these networks have apparent Mc in the
200-400 g/mole range. More suitable epoxies, for which statistical theory may be
applicable (i.e., M~'s of 103-105 g/mole), can, however, be synthesized.
Thus, the level of sophistication which one may consider for the application
of rubber-like elasticity theory to epoxy networks may depend on the application.
For highly crosslinked systems (M c < 1,000), a quantitative dependence of the
rubbery modulus on network chain length has recently been demonstrated s),
but the relevance of higher order refinements in elasticity theory is questionable.
Less densely crosslinked epoxies, however, are potentially suitable for testing modern
elasticity theories because they form via near quantitative stepwise reactions.
Detailed investigations of such networks have been reported by Dusek and
coworkers in recent studies 2, 5,28~
The modern theory of rubber-like elasticity theory suggests that there are
two types of elastically active network chains which contribute to the overall
equilibrium rubbery modulus, G~ : (1) chains attached to the network by chemical
crosslinks, G,~ and (2) chains attached by physical crosslinks or entangelements,
Gee-"That is,

G~ = G.c + Ge~ (1)

In moderately to highly crosslinked systems, the entanglement term is generally

of smaller magnitude.
For dry (unswollen) Gaussian networks, the chemical contibution to the modulus
is given by

Ge~ ----A ( ~ 2) QRT/M~ (2)

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)

Gee = kiTe (4)

where kl is a proportionality constant and Te is the trapping factor. The

constant kl is often estimated using the plateau modulus of the corresponding linear
polymer; however, the validity of this approach is a matter of debate 2s). The
trapping factor can be estimated from network formation theories.

2.2 Rubber-like Elasticity Theory and Highly Crosslinked Epoxies

Treloar suggests that non-Gaussian behavior is expected when network chains

are extended so that the end-to-end distance, (r2) 1/2, exceeds 30-50 percent of the
fully extended length 207. This phenomenon occurs at high strain~ for loosely
crosslinked polymers and is largely responsible for the sharp upturn in stress-
strain curves at high strains. Highly crosslinked polymers,, such as epoxies,
attain this critical extension even at low strains. In fact, very short network
chains apparently achieve the Gaussian limit even in the unstrained state.
Consider a typical DGEBA (diglycidyl ether of bisphenol-A) epoxy network chain
with a molecular weight of about 360 g/mole, containing approximately 9 flexible
main chain bonds. If this molecule could be described by" Gaussian statistics,
then its relative dimensions compared to the corresponding free chain, would
be given by ( r Z ) t / 2 / n l = nl/Zl/nl = n - l f 2 , where n is the number of freely rotating
bonds, and 1 is the bond length. Assuming that n is given approximately by the
number of flexible main chain bonds, the DGEBA molecule is apparently extended
to about 33 percent of its fully extended length, even before crosslinking.
A number of workers have treated non-Gaussian networks theoretically in
terms of this finite extensibitity problem. The surprising conclusion is that the
effect on simple statistical theory is not as severe as might be expected. Even
for chains as short as 5 statistical random links at strains of up to 0.25,
the equilibrium rubbery modulus is increased by no more than 20-30 percent
(typical epoxy elastomers rupture at much lower strains). Indeed, literature reports
of highly crosslinked epoxy M¢ calculated from equilibrium rubbery moduli 3o-38)
are consistently reasonable, apparently confirming this mild finite extensibility
effect.
Apparently Eqs. (1) and (2) do a reasonable job in describing the elasticity of
epoxy networks, even at high crosslink densities. For well-defined epoxies of
Structure and Ultimate Properties of Epoxy Resins 121

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

Table 1. Epoxy Restin Properties

Resin M, a EEW b ff
g/mole g/mole

828 380 190 2.0

100IF 940 493 1.9
1002F 1200 671 1.8
1004F 1450 860 1.7
1007F 2600 1880 1.4

a Via vapor phase osmometry.

b Epoxy equivalent weight. Via direct titration
(ASTM D-1652).
Functionality = Mn/EEW

Table 2. Stoichiometric DGEBA/DDS Network Properties

Network Tg Density Solubles a

(23 °C)
~C g/cna 3 ~o

828/DDS 212 1.232 0.3

1001F/DDS 132 1.204 1.3
1002F/DDS 121 1.200 1.9
1004F/DDS 113 1.196 2.8
1007F/DDS 105 1.187 9.0

Weight percent extractable solids.

122 J.D. LeMay and F. N. Kelley

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

Table 3. Equilibrium Rubbery Moduti I (T = T~ + 45 K) and M~of DGEBA/DDS

Networks

Network Tensile Dynamic Mechanical

E Mc E' M¢
GPa g/mole GPa g/mole

828/DDS 36.0 360 44.6 296

1001FtDDS 13.6 824 15,t 742
1002F/DDS 10.1 1090 11.8 930
1004F/DDS 7.19 1510 9.16 1180
t007F/DDS 3.16 3330 4.15 2598

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

Fig. 3. Stoichiometric DGEBA/DDS network

Mc versus prepolymer resin molecular weight,
M,. (Mc calculated from equilibrium rubbery mo-
duli at T = Tg + 45 K). O M, from equilibrium
tensile experiments; • Mr from 0.16 hz dynamic
mechanical storage modulus measurements (After
I , ,I,, LeMay s))
0 1 2 :3
M ./1000

storage modulus measurements. All measurements were taken at temperatures near

45 °C above the network T s's. Representative network true stress versus strain curves
from the tensile experiments are shown in Fig. 2. The ordinate axis, true stress,
is normalized by 30RT to account for the different test temperatures employed. The
resultant curves are thereby directly comparable for structural differences, since
the instantaneous slopes are proportional to 1/M c after Eq. (2). The curves of all
five networks are linear and reversible up to strains of around 10 percent. The
reversibility suggests that the measurements were performed under near-equilibrium
conditions and that the networks were stable at the high test temperatures
employed.
Figure 3 shows the relationship between the network M0 and the Epon resin
prepolymer molecular weights (M,) (the dashed line shows exact equivalence).
The Mc apparently increases in proportion to the resin molecular weights with
the exception of some deviation at the highest values. The deviation is undoubtedly
a consequence of the decreasing functionality of the higher molecular weight
resins. This results in soluble fractions in the networks formed from them, and
certainly some network chain branching as well. For example, the 9.0 percent soluble
fraction in the 1007 F/DDS network lowers the effective modulus and increases
M c by about 30 percent over the expected value. The M~ calculated from the
dynamic mechanical moduli are consistently lower than those obtained from the
tensile moduli, and probably reflect the fact that the dynamic moduli were not
collected under true equilibrium conditions. Despite the differences, however, both
sets of M c are remarkably consistent and of similar magnitude as the prepolymer
molecular weights.
For comparison to the experimental D G E B A / D D S network Me in Table 3, it is
124 J.D. LeMay and F. N. Kelley

Table 4. Comparison of Theoretical and Experimental M,

Network Theoretical Experimental

a Ratio
g/mole g/mole

828/DDS 504 360 1.40

1001F/DDS 1110 824 1.35
1002F/DDS 1470 1090 1.35
1004F/DDS 1840 1510 1.22
1007F/DDS 3890 3330 1.17

Tensile values from Table 3.

possible to estimate theoretical M r from the amine and epoxy resin concentrations.
The Equation for this is

M r = QN/zc (5)

where N is Avogadro's number, z is the number of elastically active chains per

crosslink, and c is the number of crosslinks per unit volume. This Equation is
easily derived from Eq. (2) using unity for A and (Qt2).
Since the DGEBA/DDS networks are tetrafunctional and of stoichiometric
composition, the theoretical value ofz is 2. Furthermore, the crosslink concentration, c,
is simply the DDS molecule concentration. Performing the necessary calculations
yields the theoretical M r listed in Table 4. Compared to the experimental Mr,
the theoretical values are very consistent. If it is assumed that the DGEBA/DDS
networks are not phantom-like (i.e., A -----I), then the ratio of the theoretical and
experimental values may serve as an estimate of the dilation factor, ( ~ ) . These
ratios are listed in Table 4, and show that (~2) is approximately unity for all the
networks. If the experimental Me had been calculated using the actual network
densities (instead of Q = 1 g/cm), the ratios would be even closer to unity,
being reduced by approximately 20 percent.
In summary, the apparent success of Eq. (2) is predicting the M r of the
short chain DGEBA/DDS networks suggests that these materials obey the
assumptions of statistical rubber elasticity theory, as do many more loosely
crosslinked etastomers. Yet, as previously noted, there are reasons why short
chain epoxies may be expected not to behave like ideal elastomers. In particular,
there are concerns about the non-Gaussian character and finite extensibility of short
network chains. Experimental results show, however, that in spite of these concerns,
simple rubber elasticity theory yields remarkably reasonable values. Evidently,
the storage of elastic strain energy in these networks is based largely in the
configurational changes of the network chains, even though the assumption that the
networks consist of randomly coiled Gaussian chains may not be justified.
From a structure-property viewpoint, perhaps the most useful outcome is that the
rubbery modulus of highly crosslinked epoxies is sensitive to small changes in
crosslinking, and therefore can be used as a practical means by which to correlate
crosslink density and physical properties.
Structure and Ultimate Properties of Epoxy Resins 125

3 Rubbery Cohesive Fracture

Above the glass transition temperature, thermosets are weak elastomers (because
of their densely crosslinked structure) and are of no known practical use.
Apparently, only King and Andrews 32k Swetlin 31) and LeMay s) have investigated
the cohesive fracture or tear of thermosets above Tg, all using amine-linked
epoxies. These studies have demonstrated that the rubbery fracture of epoxy
thermosets is quite similar to that of more conventional crosslinked elastomers.

3.1 Generalized Fracture Theory

Andrews 39) has suggested that the overall fracture energy, 2 f , can be separated
into reversible (equilibrium) and irreversible (non-equilibrium) components. This
is represented by the Equation

2 J = 2Jo~(%, ai, T) (6)

where the reversible component is represented by 2J0, the threshold or intrinsic

fracture energy, and the irreversible component by qb, the loss function. Under
equilibrium conditions, where energy dissipations disappear, the loss function equals
unity and 2 J = 2Jo. The toss function is expected to be a function of the
overall strain, %, the crack velocity,/t, and the temperature, T.

3.1.1 Threshold Fracture

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

of n backbone bonds of dissociation energy E, Lake and Thomas predicted

that

2 J o = (LN/2) (nE) (7)

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)

where k3 is a numerical constant in the range of 1/3 to 2/3 (depending on the

L Equation used), M0 the average bond molecular weight, q the number of bonds
per equivalent statistical freely-jointed link, I the bond length, Q the elastomer
density and N Avogadro's number. For C - - C elastomers k2 typically falls in the
range of 0.3--1.0 (J/m 2) (g/mole) -1/2 44)
All of the parameters in Eq. (9) can be reasonably estimated or experimentally
determined. In particular, q can be estimated via experimental measurement of the
stress-optical coefficient 46). For typical C - - C backbone elastomers (e.g., cis- and
trans-polyisoprene and polybutadienes), q is in the range of 5-10 main bonds 47).
Although not explicitly stated, it is assumed in Eq. (8) that the number of statistical
random links n per network chain is large. Common hydrocarbon elastomers
exhibit M c in the range of 25000 g/mole, and, therefore, contain around 1400 main
chain bonds (assuming a reasonable M0 of 18 g/mole), or about 200 random links
(using q = 7). Obviously, the assumption of large n is valid for these hydrocarbon
elastomers. Consider, however, the crosslinking of any of the C - - C vinyl elastomers
mentioned above to an M c of 2000. These network chains would contain around
l l 0 main chain bonds, or about 15 statistical links. Although an M¢ of 2000
corresponds to a relatively high level of crosslinking, Gent and Tobias ~) have
demonstrated that Eq. (8) predicts 2 J o values which compare favorably with experi-
ment for even these highly crosslinked elastomers. Apparently, a C - - C backbone
network chain with 10--20 statistical links is still reasonably flexible.
Structure and Ultimate Properties of Epoxy Resins 127

For densely crosslinked networks such as epoxies, n could conceivably be

less than 5, and the validity of Eq. (8) is therefore questionable. King and
Andrews 32) were apparently the first investigators to address this point. The
lack of q values for epoxy polymers motivated them to derive an alternate
expression for 2Jo that did not incorporate q. The main assumption of their
theory was that the distance between nearest crosslinks, L', in short-chain networks
was equivalent to the mean displacement length, L, of the network chains.
They then showed that the number of chains crossing a unit area of crack
plane was given by (1/2) N z/3 and therefore expressed 2Jo as

2Jo = (1/2) NZ/3nE (10)

Since N is proportional to M~-1 and n is proportional to M r this Equation predicts

that

2Jo = k a M c 1/3 (1t)

where k4 is a proportionality constant. This result predicts a slightly weaker

dependence of 2Jo on M c than does Eq. (8).
Eqs. (8) and (11) are important because they predict a direct relationship
between a failure property and network structure. If the Equations hold, then
the following experimental results would be expected: (1) 2J0 for a series of
networks of the same chemical composition will reflect only differences in
crosslinking, and (2) 2Jo for chemically different networks of the same M¢ will
reflect the differences in their backbone composition. Experiments such as these
have been performed on a variety of elastomers 41-44), e.g., polyurethanes,
polybutadienes, polyisoprenes, polysiloxanes, and polyphosphazenes, and the results
have consistently supported the predictions.
Very little experimental fracture work has been done in the rubbery state
for thermosets. After attaining threshold conditions for only the least crosslinked
sample of a series of amine-linked epoxies (Me in the range of 900-5000 g/mole),
King and Andrews 32) resorted to Eq. (11) to estimate threshold fracture energies.
For an M r = 5000 network, which did yield an experimental 2Jo of 3.7 J/m 2,
they calculated a value of about 4.1 J/m 2, in reasonable agreement with their
theory. The calculated 2Jo for their other networks ranged from 3 to 5 J/m 2.
Interestingly, King and Andrews compared the theshold fracture energy
(2J0 = 3.05 J/m 2) for their M r = 2000 amine-linked epoxy to that of a poly-
butadiene (PB) network of similar M r from which a value of 37 J/m 2 had been
determined by Ahagon and Gent 43). They noted that the epoxy threshold fracture
energy was about an order of magnitude lower than that of PB and attributed
this to: (1) the epoxy having a weaker backbone bond than the C - - C bond of PB,
and (2) the considerably bulkier epoxy backbone having significantly fewer network
main chain bonds than the equivalent M r, but less bulky PB.

3.1.2 The Loss Function

Substituting Eq. (8) or (11) into Eq. (6) yields an expression which suggests
that the gross fracture energy, 2 J , may display an M r dependence if the loss
128 J.D. LeMay and F. N. Kelley

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.

3.2 Rate and Temperature Effects

The cohesive fracture of conventional, non-strain crystallizing, unfilled elastomers
is sensitive to rate and temperature 32,41,48-53), exhibiting increased values of 2 J
with increasing rate and decreasing temperature. The basic viscoelastic nature of the
fracture of these materials is evidenced by the fact that it can be described over
wide ranges of temperature and rate by time-temperature superposition as described
by the WLF Equation 54)

log (aT) = --CI(T - - To)/(C 2 + T - - To) (12)

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)

3.3 Effect of Molecular Structure on the Rubbery Tear

of Highly Crosslinked Epoxies
The work of King and Andrews 32) and Swetlin 31) has shown that the rubbery
fracture energies of epoxy thermosets are time-temperature superposable and
sensitive to network structure. These studies incorporated different amine/DGEBA

Table5. Networks used to Study the Rubbery Fracture of Epoxies

Reference A/E T~ M
°C g/mole

King and Andrews 32~ 0.50 38 5150

0.63 5t 2000
1.00 86 900
1.50 59 t420
Swetlin 31~ 0.65 75 1500
1.00 162 300
1.60 115 750
Structure and Ultimate Properties of Epoxy Resins 129

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

-20 -16 -12 - 8 -4

log R.ar (m/s)

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)

Fig. 8. Sameas Fig. 6, excepttearenergiesnormalizedbyM~ 13. Curvesshiffed up onedecade on ordi-

nate axisto keep values positive.(Afier LeMaya))
t 32 J . D . LeN ty and F. N. Kelley

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.

4 Glassy Cohesive Fracture

The fracture behavior of epoxy thermosets has been of growing interest since
the mid-1960's when investigations by Broutman and McGarry 55) and Mostovoy
and Ripling 56) were published. Literature references seem to have peaked in the late
1970's and early 1980's when studies on crack blunting mechanisms 57,5s), speculations
of a nodular morphology in epoxy networks lo,17.18) and the effects of physical
(sub-Tg) aging 59-63) became of interest. Collectively, these investigations have
detailed the dependence of crack growth in epoxies on variations in compounding,
cure, and test conditions. Unfortunately, very few of these studies have been able
to correlate observed fracture behavior systematically with structural features on the
molecular level.
Although epoxies dominate the thermoset fracture literature, work has been
reported on other systems, e.g., polyester resins, phenol-formaldehyde compounds,
peroxide cured polystyrene, and highly crosslinked polyurethanes. In general,
these materials exhibit fracture behaviors similar to epoxies, and suggest that
thermosets, as a class of materials, display characteristic crack growth properties.

4.1 Fracture Testing

The fact that thermosets are typically brittle and generally exhibit linear elastic
stress-strain behavior suggests that linear elastic fracture mechanics (LEFM) and
test methods may be applicable. In fact, these approaches have proven very
popular, as is evidenced by the successful use of a number of LEFM-based fracture
Structure and Ultimate Properties of Epoxy Resins 133

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.,

2o¢ = G = K2/E * (13)

where E* is Young's modulus given by

E* = E (plane stress) (14a)

E* -- (1 -- v~)E (plane strain) (14b)

where E is the experimental modulus measured at the same test conditions as K,

and v is Poisson's ratio. Often G is called the strain energy release rate, but
its units are those of surface fracture energy.
One of the most curious aspects of crack growth in most epoxies is the
apparently unstable manner by which propagation occurs, even over wide ranges of
temperature and test rate. This behavior is commonly referred to as "stick-slip", and
is characterized by the crack growing in a series of discrete, unstable jumps. Even
some of the earliest works on epoxy fracture 5s,56~ report this mode of crack
growth. The suspected origins of stick-slip fracture behavior in epoxies is discussed
in a subsequent section. Unlike epoxies, thermoplastic polymers, such as poly(methyl
methacrylate) and polystyrene, are characterized by stable, continuous crack growth.
This mode of fracture sometimes can be observed in epoxies, in particular, when
they are tested at fast rates and/or low temperatures.
Both stable and stick-slip modes of crack propagation are illustrated in Fig. 9,
which shows typical load-displacement traces from a testing machine at a constant
134 J.D. LeMay and F. N. Kelley

Load~ Load i Initiation

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.

4.2 Effects of Testing, Material and Processing Variables

Under typical test conditions, thermoset fracture behavior is characteristically sen-

sitive to variations in testing rate 66,67) and temperature 65,66,68) Interesting and
complicated behavior also has been observed at very low temperatures (--200 °C) 66)
There are environmental factors which also may affect the fracture of thermosets,
for example, w a t e r 65'69) and perhaps gases near their liquefaction tempera-
ture 69).
Figure 10 generalizes the typical rate and temperature effects reported in the litera-
ture. The effect of rate is shown by the upper diagram a. At low rates, unstable crack
growth dominates, and the difference between K i and K, is large, which indicates that
the crack propagates by long jumps. Compared to the initiation toughnesses, which
are quite sensitive to the rate, the arrest values are comparatively rate insensitive. At
higher rates, the jump distance decreases and the magnitude of Ki approaches that of
K a. In fact, at sufficiently high rates, crack growth apparently becomes stable and a
rate controlled transition from unstable to stable fracture is observed. The magnitude
of K~ at this transition is generally about the same value of K a.
The effects of temperature (Fig. 10(b))"are very similar to those of the
rate. Crack growth at high temperatures is characteristically unstable, while tending
Structure and Ultimate Properties of Epoxy Resins 135

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

toward stable growth at lower temperatures; arrest behavior is temperature

insensitive; and a temperature controlled unstable/stable crack propagation transition
is observed.
The observation that an increase in temperature or a decrease in rate both
result in the same fracture response points toward a viscoelastic influence on
thermoset fracture behavior, especially crack initiation. This characteristic behavior
of epoxies has been explained qualitatively by consideration of the temperature
and strain rate effects on the plasticity of the material at the crack tip ls, ss:66).
In effect, test conditions which promote the formation of a so-called crack tip
plastic zone, or blunt the crack by a ductile process, promote unstable crack
propagation. This aspect of unstable fracture is subsequently discussed in more
detail.
The work of Scott et al. 69~ suggests that at low temperatures, the fracture
of epoxies may not be characterized as simply as indicated by Fig. 10(b). While
measuring the fracture energy at low temperatures for a series of networks
made from different epoxy and amine compositions, a typical unstable to stable
crack growth transition was observed at temperatures around 0 °C. Stable fracture
then persisted as the temperature dropped to --100 to --150 °C where a new
transition to unstable crack growth was observed. The origin of this lower temperature
transition was not determined; however, energy dissipations due to low temperature
molecular relaxations, or interactions of the crack tip with the cooling gas (N2)
near its liquefaction temperature, were speculated as causes.
Careful examination of the data of Scott et al. shows that over the wide
temperature range of - - t 0 0 ° to 50 °C the arrest and stable fracture energies
can be represented by a single averaged value within a scatter of about + 25
percent. At colder temperatures, however, the arrest energies drop rapidly below
this value.
The effects of moisture on epoxy fracture are not conclusive. Scott et al. 69~
reported that an amine cured epoxy, normally displaying stick-slip fracture at
room temperature and low rates, exhibited stable behavior when immersed in
distilled water. Also, they found that the rate necessary to promote the un-
stable to stable crack growth transition at room temperature was increased
by two orders of magnitude in the presence of the water. Yamini and Young 65),
on the other hand, found that testing in water tended to suppress stable behavior
and promote stick-slip fracture in an amine cured epoxy over a wider range
136 J.D. LeMay and F. N. Kelley

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

resin called N,N,N',N'-tetraglycidyl-4,4'-diaminodiphenyl methane (TGDDM) is

a commonly used resin in high performance fiber reinforced composites (Fig. 11 (c)).
This resin yields very high Tg networks ( > 200 °C) when cured with DDS, and
may produce complicated network structures due to the close proximity of the
epoxide pairs (e.g., forming intramolecular rings 74, 7s)).
From the limited fracture data available for similarly cured networks generated
from these various resin types, little can be concluded as to the role of the
resin in fracture. Certainly, the resin backbone contributes to the Tg of the
network, and comparisons should take this into account because, as previously
discussed, the initiation of crack growth is very sensitive to temperature. If
the epoxy resin structure results in complicated network forming reactions, as is
possible for T G D M / D D S networks, the structure of the final network will be
affected and may likely influence fracture.
While epoxies can be cured with a multitude of agents, the general ease
and simplicity of the reaction with polyfunctional amines has made them most
popular. Aliphatic amines generally yield networks with lower Tg's than aromatic
amines, but if this is taken into account, no significant differences in general
fracture behavior have been observed. Phillips et al. 76~ showed, however, that
the amine type can influence the fracture behavior of DGEBA networks. Their
study involved stoichiometric networks cured with the n = 1, 2 and 3 homologs
of two series of aliphatic amines: (1) difunctional amines of the type
NH2(CH2CH2)nNH2, and (2) polyfunctional amines of the type NH2CH2(CH2 •
• NHCHE)nCH2NH 2. While the amine molecular weight and type did not sig-
nificantly affect arrest fracture energies for either series, or initiation fracture
for the second series, the initiation fracture energies of the first series increased
with the amine molecular weight. It was proposed that the crosslink density
of the first series decreased as the number of ethylene groups separating the
amines increased, while the crosslink density of the second series did not change
because the molecular distance between amine groups was constant. The initiation
fracture energies were therefore considered to be reflecting the crosslink density
differences generated by the two amine types.
Some principal epoxy processing variables include: (1) the reactant ratio, (2) the
cure schedule, and (3) the postcure thermal treatment. These processing parameters
are important because they influence the final microstructure of the thermoset.
The reactant ratio determines the number of functional groups available for the
crosslinking reactions, and therefore strongly controls the network structure. Cure
schedules (time at temperature sequences) have a direct influence on the reaction
kinetics through which the network structure is developed. Postcures may act to
increase the extent of cure, introduce new crosslinking chemistry, or induce
oxidative and degradative mechanisms, all of which can effectively alter network
structure.
The effect of the reactant ratio, A/E, on the physical properties and fracture be-
havior of epoxy systems has been the subject of many studies lo, 31,32,65,66,77, 78)
and the results have been inconclusive. This is due largely t o t h e fact that network
structure changes dramatically with changes in A/E, especially in epoxy excess
(A/E < 1) and amine excess (A/E > 1) compounds. Comparison of different systems,
therefore, must take into consideration whether the networks involved are amine
138 J.D. LeMay and F. N. Kelley

excess, epoxy excess, or stoichiometric compositions. In addition, some epoxy

systems are capable of side reactions which may be enhanced in off-stoichiometric
mixtures. For example, the epoxy ring opening reaction by hydroxyl groups
(etherification) can lead to chain extension and/or increased crosslinking. Even in
systems in which the amine-epoxy reaction is preferred, etherification may play a role
in the later stages of cure in epoxy excess formulas.
Even after careful accounting of the composition reactant ratios, it is difficult
to arrive at specific conclusions about epoxy structure and ultimare property rela-
tionships from the literature. For example, the tensile strength has been found
to display both a maximum 56) and a minimum 79) at stoichiometry, while another
study suggests that there is virtually no effect 77) Also the elongation at break
has been found to be a maximum at A/E = 1 in one study 79) while displaying no
A/E dependence in others s6,77~. As noted above, it is found that a minimum
M c is obtained at stoichiometry, while off-stoichiometry ratios yield higher values.
Taking this approach, A/E has been used in a number of studies as a means to study
the effect of crosslinking on fracture. Again, the results are inconclusive: K and G
have been found to pass through a maximum at some A/E value 79) (not stoichio-
metry), while increasing with A/E through stoichiometry in another study 31).
Controlled variation of network structure has often been attempted through
the use of cure and postcure schedules to control of the extent of cross-
linking 65, 70-73, 78). A typical methodology is to subject a partially cured network
to different time-at-temperature sequences to yield networks crosslinked to dif-
ferent extents of complete cure. For example, Chang et al. 70-73) used this
technique to investigate the M r dependence of the physical and fracture properties
of an amine-linked epoxy network. They utilized a stoichiometric mixture of the
flexible epoxy and the amine DDS to obtain networks with low Tg'S, but high
reaction temperatures. They anticipated that the resultant undercured networks
would be structurally stable, even near Tg, and exhibit widely different M r.
What they actually prepared were a series of networks with apparent Me in the
narrow range of 300--400 g/mole which displayed inconclusive and complicated
fracture behavior. This was most likely a consequence of the fact that their
networks were not as structurally simple as they had anticipated, and were
actually capable of undergoing significant structural changes over a narrow range
of extent of cure. While stoichiometric reactant ratios were used, the resultant
undercured networks were undoubtedly structurally complicated, likely sharing
characteristics of off-stoichiometric systems. Thus, the problem accompanying
the use of cure schedules to control structure is that while network structure can be
altered, the changes cannot be easily characterized.

4.3 Origin of Unstable Crack Growth

The origin of unstable crack growth in thermosets is a subject of speculation. Some

suggested causes for this behavior include: (1) dG/d~i becoming negative, (2) crack
tip thermal softening due to an isothermal to adiabatic transition, (3) molecular
relaxation processes, (4) test specimen geometry effects, and (5) crack tip blunting
by localized plastic deformation.
Structure and Ultimate Properties of Epoxy Resins 139

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.

4.4 Effect of Structure on the Glassy Fracture

of Highly Crosslinked Epoxies

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))

The temperature dependence of the fracture toughness for the DGEBA/DDS

networks is shown in Fig. 12. Crack growth initiation data are symbolized by
the open circles, arrest data by the closed circles, and stable data by the
half-filled circles. With the exception of the 1007F/DDS network, the initiation
fracture toughnesses increase with the test temperature. In contrast, the arrest
values exhibit no noticeable temperature sensitivity, and when observed, the stable
fracture toughnesses are of the same magnitude as the arrest values. It is
142 J.D. LeMay and F. N. Kelley

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

Rubbery state Glassy state

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

-121 -80 -40 0 40 80 120

T, - - T (°C)

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

Table 6. DGEBA/DDS Network

Arrest Fracture Energies

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.,

2J, -~- v~ 5 ~,~1/2

lvJt e (15)

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

behavior is less easily characterized. It is possible that initation is affected

by microstructures such as the soluble fraction, and chain branching, as well
as the crosslink density. More research must be done, however, before any
specific conclusions can be drawn on the structure-fracture relationship in initiation.
The unique behavior of arrest fracture in the DGEBA/DDS networks, however,
suggests that this energy may represent a unique and characteristic property of the
material. This is supported in a new fracture theory by Kramer and Hart 957,
which predicts that K a is a limiting fracture toughness, independent of both testing
rate and temperature.
There are several reasons to expect that long-range structural variations, such
as M c, should not influence the glassy fracture of the DGEBA/DDS networks.
For instance, other glassy properties, such as the modulus and yield stress,
do not exhibit discernible Mc dependencies 7,8~ This is hardly surprising, how-
ever, since only short-range molecular features are expected to influence these
properties. In order for long-range structural features, such as M~, to influence
a glassy property, it is necessary that molecular chains be so short that they
are in the realm of short-range motions, or that network chains experience
increased mobility. In the highly stressed region ahead of a crack tip, it is
generally accepted that the yield criterion is met, and therefore that network
chains are given increased mobility. The chains in the yielded zone are certainly more
mobile than those in the bulk glass, yet undoubtedly much less mobile than they
would be above Tg, due to a much higher internal viscosity. Thus, it would
still require a considerable strain on the yielded material to involve molecular seg-
ments as large as entire network chains. It may be possible that this condition is met
in the highly strained region at the crack tip.
At sub-T~ temperatures, long-range structure is apparently reflected in the
large strain, post-yield behavior of glassy polymers, where the phenomenon of strain
hardening is observed. While it is generally accepted that this response is due to glassy
polymer chains experiencing sumcient mobility to be stretched between their cross-
links or entanglements, and thereby become oriented with the applied stress,
it also is noted that these chains are far from rubber-like. Specimens yielded
to such strains do not recover their dimensions when unloaded, even though the
entropic driving force certainly exists for the recovery of the random coil
conformations of the polymer chains. Obviously, the internal viscosity is simply
too high.
Further evidence that long-range structure influences the behavior of highly strai-
ned and yielded glassy polymers was shown by Donald and K r a m e r 96'97)
Their microscopic examination of plastic zones formed at flaws during the
tensile deformation of thin polymer sheets demonstrated convincingly that the
material in the plastic zone is composed of extended molecular chains, as they
found a semi-quantitative relationship between the relative dimensions of the
zone and the length of fully extended chains trapped between entanglements.
The observed MI~/2dependence of arrest glassy fracture energies is curious because
this is the same dependence predicted by Lake and Thomas 45) for the threshold
tearing of etastomers. It was previously shown that this dependence is exhi-
bited by the rubber tear of the DGEBA/DDS epoxies over a wide range of
temperatures, even though they were far removed from the threshold region.
146 J.D. LeMay and F. N. Kelley

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.

Acknowledgement: The authors wish to acknowledge the financial Support of the

United States Air Force Office o f Scientific Research.

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

Allegra, G. and Bassi, I. IV.: Isomorphism in Synthetic Macromolecular Systems. Vol. 6,

pp. 549-574.
Andrews, E. H. : Molecular Fracture in Polymers. Vol. 27, pp. 1~i6.
Anufrieva, E. V. and Gotlib, Yu. Ya. : Investigation of Polymers in Solution by Polarized Lumines-
cence. Vol. 40, pp. 1~58.
Apicella, A. and Nicolais, L. : Effect of Water on the Properties of Epoxy Matrix and Composite.
Vol. 72, pp. 6%78.
Apicella, A., Nicolais, L. and de Cataldis, C. : Characterization of the Morphological Fine Structure
of Commercial Thermosetting Resins Through Hygrothermal Experiments. Vol. 66, pp. 189-208.
Aroon, A. S., Cohen, R. E., Gebizlioglu, O. S. and Schwier, C. : Crazing in Block Copolymers and
Blends. Vol. 52/53, pp. 275-334
Aronhime, M. T., Gillham, J. K. : Time-Temperature Transformation (TTT) Cure Diagram of Thermo-
setting Polymeric Systems. Vol. 78, pp. 83-113.
Arridge, R. C. and Barham, P. J. : Polymer Elasticity. Discrete and Continuum Models. Vol. 46,
pp. 67-117.
Aseeva, R. M., Zaikov, G. E.: Flammability of Polymeric Materials. Vol. 70, pp. 171-230.
A y r e y , G. : The Use of Isotopes in Polymer Analysis. Vol. 6, pp. 128-148.

Bdssler, H. : Photopolymerization of Diacetylenes. Vol. 63, pp. 1-48.

Baldwin, R. L. : Sedimentation of High Polymers. Vol. 1, pp. 451-511.
Balta-Calleja, F. J. : Microhardness Relating to Crystalline Polymers. Vol. 66, pp. 117-148.
Barton, J. M. : The Application of Differential Scanning Calorimetry (DSC) to the Study of Epoxy
Resins Curing Reactions. Vol. 72, pp. 111-154.
Basedow, A. M. and Ebert, K.: Ultrasonic Degradation of Polymers in Solution. Vol. 22,
pp. 83-148.
Batz, H.-G. : Polymeric Drugs. Vol. 23, pp. 25-53.
Bell, J. P. see Schmidt, R. G. : Vol. 75, pp. 33-72.
Bekturov, E. A. and Bimendina, L. A. : Interpolymer Complexes. Vol. 41, pp. 99-147.
Berosma, F. and Kruissink, Ch. A.: Ion-Exchange Membranes. Vol. 2, pp. 307-362.
Berlin, AI. AI., Volfson, S. A., and Enikolopian, N. S.: Kinetics of Polymerization Processes. Vol. 38,
pp. 89-140.
Berry, G. C. and Fox, T. G.: The Viscosity of Polymers and Their Concentrated Solutions. Vol. 5,
pp. 261-357.
Bevington, J. C. : Isotopic Methods in Polymer Chemistry. Vol. 2, pp. 1-17.
Bhuiyan, A. L. : Some Problems Encountered with Degradation Mechanisms of Addition Polymers.
Vol. 47, pp. 1~55.
Bird, R. B., Warner, Jr., H. R., and Evans, D. C. : Kinetic Theory and Rheology of Dumbbell
Suspensions with Brownian Motion. Vol. 8, pp. 1-90.
Biswas, M. and Malty, C. : Molecular Sieves as Polymerization Catalysts. Vol. 31, pp. 47-88.
Biswas, M., Packirisamy, S. : Synthetic Ion-Exchange Resins. Vol. 70, pp. 71-118.
Block, H.: The Nature and Application of Electrical Phenomena in Polymers. Vol. 33, pp. 93-167.
Bodor, G. : X-ray Line Shape Analysis. A. Means for the Characterization of Crystalline Polymers.
Vol. 67, pp. 165-194.
150 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

Dreyfuss P. and Dreyfuss, M. P. : Polytetrahydrofuran. Vol. 4, pp. 528-590.

Drobnik, J. and Ryp6(ek, F.: Soluble Synthetic Polymers in Biological Systems. Vol. 57, pp. 1-50.
Dr6seher, M.: Solid State Extrusion of Semicrystalline Copolymers. Vol. 47, pp. 120-138.
Drzal, L. T. : The Interphase in Epoxy Composites. Vol. 75, pp. 1-32.
Dugek, K. : Network Formation in Curing of Epoxy Resins. Vol. 78, pp. 1--60.
Dugek, K. and Prins, W.." Structure and Elasticity of Non-Crystalline Polymer Networks. Vol. 6,
pp. 1-102.
Duncan, R. and Kope~ek, J.." Sotuble Synthetic Polymers as Potential Drug Carriers. Vol. 57,
pp. 51-101.

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

Gillham, J. K. see Aronhime, M. T. : Vol. 78, pp. 83-113.

Godovsky, Y. K. : Thermomechanics of Polymers. Vol. 76, pp. 31-102.
Goethals, E. J. : The Formation of Cyclic Oligomers in the Cationic Polymerization of Heterocycles.
Vol. 23, pp. 103-130.
Gorshkov, A. V.: see Entelis, S. G. Vol. 76, pp. 129-175.
Graessley, W. W. : The Etanglement Concept in Polymer Rheology. Vol. 16, pp. 1-179.
Graesslev, W. W. : Entagled Linear, Branched and Network Polymer Systems. Molecular Theories.
Vol. 47, pp. 67-117.
Grebowicz, J. see Wunderlich, B. Vol. 60/61, pp. 1-60.
Greschner, G. S. : Phase Distribution Chromatography. Possibilities and Limitations. Vol. 73/74,
pp. 1-62.

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.

Janeschitz-Kriegl, H. : Flow Birefrigence of Elastico-Viscous Polymer Systems. Vol. 6, pp. 170-318.

Jenkins, R. and Porter, R. S. : Upertubed Dimensions of Stereoregular Polymers. Vol. 36, pp. 1-20.
Jenngins, B. R. : Electro-Optic Methods for Characterizing Macromolecules in Dilute Solution.
Vol. 22, pp. 61-81.
Johnston, D. S.: Macrozwitterion Polymerization. Vol. 42, pp. 51-106.

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.

Packirisamy, S. see Biswas, M. Vol. 70, pp. 71-118.

Papkov, S. P.: Liquid Crystalline Order in Solutions of Rigid-Chain Polymers. Vol. 59, pp. 75-102.
Patat, F., Killrnann, E. und Schiebener, C. : Die Absorption yon Makromolek~len aus Lrsung. Vol. 3,
pp. 332--393.
Patterson, G. D. : Photon Correlation Spectroscopy of Bulk Polymers. Vol. 48, pp. 125-159.
Penczek, S., Kubisa, P. and Matyjaszewski, K. : Cationic Ring-Opening Polymerization of Heterocyclic
Monomers. Vol. 37, pp. 1-149.
Penczek, S., Kubisa, P. and Matyjaszewski, K.: Cationic Ring-Opening Polymerization; 2. Synthetic
Applications. Vol. 68/69, pp. 1-298.
Peticolas, W. L. : Inelastic Laser Light Scattering from Biological and Synthetic Polymers. Vol. 9,
pp. 285-333.
Petropoulos, J. H. : Membranes with Non-Homogeneous Sorption Properties. Vol. 64, pp. 85~134.
Pino, P. : Optically Active Addition Polymers. Vol. 4, pp. 393-456.
Pitha, J. : Physiological Activities of Synthetic Analogs of Polynucleotides. Vol. 50, pp. 1-16.
Plat~, N. A. and Noak, O. 1I. : A Theoretical Consideration of the Kinetics and Statistics of Reactions
of Functional Groups of Macromolecules. Vol. 31, pp. 133-173.
Plat~, N. A. see Shibaev, V. P. Vol. 60/61, pp. 173-252.
Plesch, P. 11.: The Propagation Rate-Constants in Cationic Polymerisations. Vol. 8, pp. 137-154.
Porod, G. : Anwendung und Ergebnisse der Rrntgenkleinwinkelstreuung in festen Hochpolymeren.
Vol. 2, pp. 363-400.
Posplgil, J. : Transformations of Phenolic Antioxidants and the Role of Their Products in the Long-
Term Properties of Polyolefins. Vol. 36, pp. 69-133.
Postelnek, W., Colemann, L. E. and Lovelace, A. M. : Fluorine-Containing Polymers. I. Fluorinated
Vinyl Polymers with Functional Groups, Condensation Polymers, and Styrene Polymers. Vol. 1,
pp. 75-113.

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.

Rehage, G. see Finkelmann, H. Vol. 60/61, pp. 99-172.

Rempp, P. F. and Franta, E. : Macromonomers: Synthesis, Characterization and Applications. Vol. 58,
pp. 1 54.
t 56 Author Index Volumes 1-78

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.

SafJord, G. J. and Naumann, A. W. : Low Frequency Motions in Polymers as Measured by Neutron

Inelastic Scattering. Vol. 5, pp. 1-27.
Sato, T. and Otsu, T. : Formation of Living Propagating Radicals in Microspheres and Their Use
in the Synthesis of Block Copolymers. Vol. 71, pp. 41-78.
Sauer, J. A. and Chen, C. C. : Crazing and Fatigue Behavior in One and Two Phase Glassy
Polymers. Vol. 52/53, pp. 169-224.
Sawamoto, M. see Higashimura, T. Vol. 62, pp. 49-94.
Schmidt, R. G., Bell, J. P. : Epoxy Adhesion to Metals. Vol. 75, pp. 33 72.
Sehuereh, C.." The Chemical Synthesis and Properties of Polysaccharides of Biomedical Interest.
Vol. 10, pp. 173-194.
Schulz, R. C. und Kaiser, E.." Synthese und Eigenschaften yon optisch aktiven Polymeren. Vol. 4,
pp. 236-315.
Seanor, D. A. : Charge Transfer in Polymers. Vol. 4, pp. 317-352.
Semerak, S. N. and Frank, C. W. : Photophysics of Excimer Formation in Aryl Vinyl Polymers,
Vol. 54, pp. 31-85.
Seidl, J., Malinskj,, J., Dugek, K. und Heitz, W. : Makroporrse Styrol-Divinylbenzol-Copolymere
und ihre Verwendung in der Chromatographic und zur Darstellung von Ionenaustauschern.
Vol. 5, pp. 113-213.
Semjonow, V. : Schmelzviskosit/iten hochpolymerer Stoffe. Vol. 5, pp. 387-450.
Semlyen, J. A. : Ring-Chain Equilibria and the Conformations of Polymer Chains. Vol. 21, pp. 41-75.
Sen, A. : The Copolymerization of Carbon Monoxide with Olefins. Vol. 73/74, pp. 125-144.
Sharkey, W. H. : Polymerizations Through the Carbon-Sulphur Double Bond. Vol. 17, pp. 73-103.
Shibaev, V. P. and Platd, N. A.: Thermotropic Liquid-Crystalline Polymers with Mesogenic Side
Groups. Vol. 60/61, pp. 173-252.
Shimidzu, T. : Cooperative Actions in the Nucleophile-Containing Polymers. Vol. 23, pp. 55-102.
Shutov, F. A. : Foamed Polymers Based on Reactive Oligomers, Vol. 39, pp. 1-64.
Shutov, F. A. : Foamed Polymers. Cellular Structure and Properties. Vol. 51, pp. 155-218.
Shutov, F. A. : Syntactic Polymer Foams. Vol. 73/74, pp. 63-124.
Siesler, H. IV. : Rheo-Optical Fourier-Transform Infrared Spectroscopy: Vibrational Spectra and
Mechanical Properties of Polymers. Vol. 65, pp. 1-78.
Silvestri, G., Gambino, S., and Filardo, G.. Electrochemical Production of Initiators for Polymeri-
zation Processes. Vol. 38, pp. 27-54.
Sixl, H. : Spectroscopy of the Intermediate States of the Solid State Polymerization Reaction in
Diacetylene Crystals. Vol. 63, pp. 49-90.
Sliehter, W. P. : The Study of High Polymers by Nuclear Magnetic Resonance. Vol. 1, pp. 35-74.
Small, P. A.: Long-Chain Branching in Polymers. Vol. 18.
Smets, G.: Block and Graft Copolymers. Vol. 2, pp. 173-220.
Smets, G. : Photochromic Phenomena in the Solid Phase. Vol. 50, pp. 17-44.
Sohma, J. and Sakaguchi, M. : ESR Studies on Polymer Radicals Produced by Mechanical Destruction
and Their Reactivity. Vol. 20, pp. 109-158.
Solaro, R. see Chiellini, E. Vol. 62, pp. 143-170.
Sotobayashi, H. und Springer, J.." Oligomere in verdiinnten Lrsungen. Vol. 6, pp. 473-548.
Sperati, C. A. and Starkweather, Jr., H. W. : Fluorine-Containing Polymers. II. Polytetrafluoroethy-
lene. Vol. 2, pp. 465~195.
Author Index Volumes 1-78 157

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.

Zachmann, H. G.: Das Kristallisations- und Schmelzverhalten hochpolymerer Stoffe. Vol. 3,

pp. 581-687.
Author Index Volumes 1-78 159

Zaikov, G. E. see Aseeva, R. M. Vol. 70, pp. 171-230.

Zakharov, V. ,4., Bukatov, G. D., and Yermakov, Y. I.: On the Mechanism of Olifin Polymerization
by Ziegler-Natta Catalysts. Vol. 51, pp. 61-100.
Zambelli, A. and Tosi, C. : Stereochemistry of Propylene Polymerization. Vol. 15, pp. 31-60.
Zucchini, U. and Cecchin, G.: Control of Molecular-Weight Distribution in Polyolefins Synthesized
with Ziegler-Natta Catalytic Systems. Vol. 51, pp. 101-154.
Zweifel, H. see Lohse, F.: Vol. 78, pp. 61-81.
Subject Index

Acid curing 47f. - - - deviation 139

-

Active branch points, average functionality 36 - - - - plastic zone 146

Addition esterification 47, 48 Critical conversion 25, 38, 54
Aliphatic amines 37 Crosslink concentration 124
Amine curing, branching theory 30 - - density 104f., 137
- excess networks 130
- - - - , average ll7
-

Amines, polyfunctional 137 Crosslinked networks, dense 127

tertiary 48, 51 52
- - Crosslinking 75
Amino groups 29 - - , inhomogeneous 6, 117
Anhydrides, cyclic 33, 47, 52 - - schedules 138
Arrest 14l Cure diagram, TTT 85, 93, 96ff., 103
Arrhenius dependence o f rate constants 11 - - , full 86, 96
Aryldiazonium salts 65 Cured epoxy resins, (in)homogeneity 6f.
Autocatalysis 26, 29 Curing 1 ft.
Autoinhibition 29 - - , acid 47f.
- agents 5
-

BFa-amine complexes 55 Cyclic anhydrides 47, 52, 53

Branching processes, theories 13 Cyclization 8, 22
theories 12f.
- -

, amine curing 30 D D M (Diaminodiphenylmethane) 28, 29

Bronsted acid formation 66 Degree o f polymerization 15
D G A (N,N-Diglycidylaniline) 22, 29, 39
Caproic acid 47 D G E B A 22, 37
Cascade method 45 - - - H M D A 38, 40
substitution 15, 18
- - - resins 136
-

- - 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
-

tip blunting 138, 140

- - -- --, overall 9f.
 162 Subject Index

, specific 9 - - energies 142, 146

- -

N,N-Diglycidylaniline (DGA) 22, 29, 39 moduli 142

- -

Diglycidylether o f B i s p h e n o l A (DGEBA) 22, 37 Gyration, radius of 25

- of resorcine ( D G E R ) 37 1,7-Heptanedicarboxylic (azelaic) acid 51
N,N-Dimethylurea 55 Hexamethylenediamine ( H M D A ) 28, 38
Disproportionation 48 High Tg epoxy resins 97
Double torsion (DT) specimens 133 Hybrid systems 76
Dual functionality 77
D y n a m i c mechanical analysis 88, 90 Ideal networks 119
lmidazoles 55
Elastically active chains 17, 119 Infrared spectroscopy 103
- - network chains, concentrations 25, 35
- -
Initiated reaction 43
, length and distribution 25 Initiation 141
, n u m b e r of 50, 51 Initiator 20
Electron microscopy 6 Iodonium salts 66
Energy dissipations 125 Iron aren salt photoinitiators 70-75
Entanglements, trapped 121
-- trapping factor 25 Kinetics 25
Epoxy-amine reaction, alternating 8 - equations 30, 45
-

- - excess networks 130 -- method 18

functionality 98 -- theory 20, 21, 44
resin 93- -

Lewis acid formation 65

functional group ratio 116 - -

Linear elastic fracture mechanics (LEFM) 132

structure, characterization 116
- -

- - polymerization 104
Equation of motion for torsion pendulum 90 Living polymerization, initiated 19
Equilibrium rubbery moduli 122 Logarithmic decrement 90 f.
shear modulus 41
- -

Long-chain branching 130

- stress/strain 40
-

Loss function 125, 127

Esterifications 47 Loss modulus (G") 89f.
Etherification, internal 27
Extensibility, finite 120, 124 Master curve 128
Extinction probabilities 16, 34, 49 M a x i m u m glass transition temperature, Tg~ 86
M c dependence 129
Fracture energy master curves 129 Method of m o m e n t s 21
- - , stable/unstable 134 Microstructures 144
-- surfaces 139 - - , controlled 116
- - -- replicas 118 Moisture, effects 135
-toughness 133, 141
-
Molecular relaxations 135
Free radical kinetic mechanism 106 - weight 54 -

-- volume theory 109 -- - averages 25, 32

-

Front factor 41, 119f. -- - distribution 25

-

Full cure 86, 96 • number-average 37

Functionality 123 Monoepoxide 53
Gelation 84ff., 97 M o n o m e r s 14
- - , critical molar ratio 38 M o n t e Carlo methods 20
- time 101
-
Multicomponent systems 17
Gel formation, critical molar ratio 28 Network build-up, simulation 23
- - fractions 39 formation 5
- -

Gel point 8, 34, 38 - - , short-chain 124, 127

- -- condition 46
- - - soluble fractions 122
- - - - conversion 49 structures 63
- -

Glass transition 11 Nodules 117

- - - temperature 5
-
- - morphology 132
-- - - after cure 94, 96, 98 Non-Gaussian behavior 120
- - - - - of uncured reactants, Tgo 85
-
Nonlinear polymerization 104
Glassy cohesive fracture 132 Non-stoichiometric networks 122
- fracture 140, 143
-
N u m b e r average degree of polymerization 105f.
 Subject Index 163

functionality 41 Sol fraction 17, 25, 35, 39, 50

molecular weight 104f. Soluble fractions 123, 130
Spanning tree approximation 22
Off-stoichiometric compositions 117 Stable fracture 134
-----mixtures 138 Static light-scattering 7
Onium salts and related compounds 66 Statistical links 126
-- --, photochemistry 66 Stick-slip fracture 133, 135
Organometallic compounds as photoinitiators Strain hardening 145
69 Stress intensity factor 133
Oxidation state 74 ---strain measurement 42
Substitution effect 21, 28, 30
Percolation 23
Phantom networks 119 TBA 89f.
Phase separation 87, 99 --, isothermal scan 91
Phenylglycidyl ether 47 Tear energy master curves 131
- - - - - D D M 40 Temperature scan 93 f.
Photocrosslinking 61-77 N,N,N',N'-Tetraglycidyl-4,4'-diaminodiphenyl-
Photoinitiators 63 f., 69 methane (TGDDM) 22, 29, 39
Physical aging 132 Tgoo 94
Plasticity at the crack tip 135 T 8 splitting 118
Plastic zone 135 T G D D M 22, 29, 39
Poisson distribution 21 Thermal degradation 86 f., 97 f.
Polyaddition, irreversible step 18 Thermodynamic instability 8
Polyamines, addition to polyepoxides 25f. Thermosetting materials 84
Polycarboxylic acids 47 -- process 85
Polyepoxides and polyamines 25 f. Threshold fracture energy 125, 127
Polyepoxy-polyamine systems, multicomponent Time-temperature superposition 128
36 - - - - - transition diagram 72
Polyetherification 24, 43 Topological limit of the reaction 24
Poly(hydroxyester) 48 Torsion pendulum (TP) 88
Postcures 137 Torsional braid analysis (TBA) 88
Power law relationship 144 Transesterification 48
Probability generating function 14 -- conversion 49
Transition, isothermal to adiabatic 138 f.
Trapped entanglement 36, 42
Radius of gyration 25
Trapping factor 36
Rate of reaction 101
Triads 36
Reactant ratio 137 f.
Triphenylsulfonium salts 66
Reactivity, independent 21
TTT cure diagram 85, 93, 96ff., 103
Relative rigidity 90, 94
13-Relaxation 139
Unstable fracture 134
Rubber-like elasticity theory 118
UV absorption 75
---modified epoxy resins 99
Rubbery cohesive fracture 125
Vitrification 84ff., 94, 97
fracture energies 143, 146
time 101 f., 104, 106, 109
- -

- -

Volume dilatation 140

Scattering functions 25
Segmental mobility 10 WLF equation 128
Shear modulus 90, 91
Small-angle neutron scattering (SANS) 7 Yield criterion 145
- - - - - X-ray scattering (SAXS) 7 - - stress 145